Elections: The Nation's Evolving Election System as Reflected in 
the November 2004 General Election (06-JUN-06, GAO-06-450).	 
                                                                 
The 2004 general election was the first presidential election	 
that tested substantial changes states made to their election	 
systems since the 2000 election, including some changes required 
by the Help America Vote Act of 2002 (HAVA). HAVA required some  
major changes in the nation's elections processes, not all which 
had to be implemented by the November 2004 election. HAVA	 
addressed issues of people, processes, and technology, all of	 
which must be effectively integrated to ensure effective election
operations. GAO initiated a review under the authority of the	 
Comptroller General to examine an array of election issues of	 
broad interest to Congress. For each major stage of the election 
process, this report discusses (1) changes to election systems	 
since the 2000 election, including steps taken to implement HAVA,
and (2) challenges encountered in the 2004 election. For this	 
report, GAO sent a survey to the 50 states and the District of	 
Columbia (all responded) and mailed a questionnaire to a	 
nationwide sample of 788 local election jurisdictions about	 
election administration activities (80 percent responded). To	 
obtain more detailed information about experiences for the 2004  
election, GAO also visited 28 local jurisdictions in 14 states,  
chosen to represent a range of election system characteristics.  
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-06-450 					        
    ACCNO:   A55161						        
  TITLE:     Elections: The Nation's Evolving Election System as      
Reflected in the November 2004 General Election 		 
     DATE:   06/06/2006 
  SUBJECT:   Absentee voting					 
	     Elections						 
	     Intergovernmental relations			 
	     Jurisdictional authority				 
	     Mechanization					 
	     Surveys						 
	     Voting						 
	     Voter registration 				 
	     Voting systems					 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Product.                                                 **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO-06-450

                 United States Government Accountability Office

GAO

                                   June 2006

ELECTIONS

The Nation's Evolving Election System as Reflected in the November 2004 General
                                    Election

                                       a

ELECTIONS

The Nation's Evolving Election System as Reflected in the November 2004 General
Election

  What GAO Found

In passing HAVA, Congress provided a means for states and local
jurisdictions to improve upon several aspects of the election system, but
it is too soon to determine the full effect of those changes. For example,
41 states obtained waivers permitted under HAVA until January 1, 2006, to
implement a requirement for statewide voter registration lists. States
also had discretion in how they implemented HAVA requirements, such as the
identification requirements for first-time mail registrants. Some local
election jurisdictions described different identification procedures for
first-time mail registrants who registered through voter registration
drives. Although states differed regarding where voters who cast
provisional ballots for federal office must cast those ballots in order
for their votes to be counted, provisional voting has helped to facilitate
voter participation. HAVA also created the Election Assistance Commission,
which has issued best practice guides and voluntary voting system
standards and distributed federal funds to states for improving election
administration, including purchasing new voting equipment. The results of
our survey of local election jurisdictions indicate that larger
jurisdictions may be replacing older equipment with technology-based
voting methods to a greater extent than small jurisdictions, which
continue to use paper ballots extensively and are the majority of
jurisdictions. As the elections technology environment evolves, voting
system performance management, security, and testing will continue to be
important to ensuring the integrity of the overall elections process.

GAO found that states made changes-either as a result of HAVA or on their
own-to address some of the challenges identified in the November 2000
election. GAO also found that some challenges continued-such as problems
receiving voter registration applications from motor vehicle agencies,
addressing voter error issues with absentee voting, recruiting and
training a sufficient number of poll workers, and continuing to ensure
accurate vote counting. At the same time, new challenges arose in the
November 2004 election, such as fraudulent, incomplete, or inaccurate
applications received through voter registration drives; larger than
expected early voter turnout, resulting in long lines; and counting large
numbers of absentee ballots and determining the eligibility of provisional
voters in time to meet final vote certification deadlines.

The Election Process Involves the Integration of People, Process, and
Technology People

Absentee Conducting Voting with Vote

Voter Process registration and elections provisional counting

early voting and

ballots

certification

    Technology

Source: GAO.

United States Government Accountability Office

Contents

Transmittal Letter

3

Executive Summary

Purpose 3 Background 4 Results in Brief 5 Principal Findings 12 Concluding
Observations 31

33

Chapter 1

Election Authority 34 Election System Elements 40 Scope and Methodology 45

47

Chapter 2

Overview Changes Required by HAVA Subsequent to the 2000 General  48
Election 50 Processing Registration Applications Received from MVAs and
Other NVRA Entities Remained a Challenge 77 Jurisdictions Encountered New
Challenges Processing the Large Volume of Voter Registration Applications
for the November  2004 General Election 81 Concluding Observations 96

99

Chapter 3

Overview

Absentee Voting 100

         102 Voting Uniformed and Overseas Citizens Absentee Voting 127 Early
                                       Voting 133 Concluding Observations 145

148

Chapter 4

Overview HAVA Made Changes Intended to Improve Election  148
Administration 149 Some States Reported Changing Identification
Requirements for  All Prospective Voters 166

Recruiting a Sufficient Number of Skilled, Reliable Poll  Workers
Continued to Be a Challenge for Large and Medium Jurisdictions 170

Election Officials in Some Jurisdictions We Visited Reported on Challenges
Training Poll Workers 182 Poll Worker Performance Problems in Some Large
and Medium Jurisdictions 186 Obtaining Enough Polling Places That Met
Standards Continued  to Be a Challenge for Some Jurisdictions 190
Election Officials in Some Jurisdictions We Visited Reported That
Designing Clear Ballots Continued to Be a Challenge 199

Preparing Ballots or Voting Equipment Not a Problem Reported by Most
Jurisdictions, but These Activities Can Be a Challenge for Some
Jurisdictions 202

Election Officials in Some Jurisdictions We Visited Reported Experiencing
Long Lines on Election Day, while Officials in  Other Jurisdictions Did
Not 206

Overloaded Phones on Election Day Posed Problems for Some  Large and
Medium Jurisdictions 207

Election Officials in Some Jurisdictions We Visited Reported  That
Third-Party Observers Were a Challenge on Election  Day 2004 209

Concluding Observations 211

213

Chapter 5

214

Provisional Voting Overview

Events in the November 2000 Election Preceded the HAVA Requirement for
Provisional Ballots 215 State and Local Jurisdictions Varied in Their
Implementation for Providing Provisional Ballots for the November 2004
Election 218 State and Local Jurisdictions Reported Variation in Several
Areas Affecting whether Provisional Ballots Are Counted 234 A Number of
Factors Contribute to Difficulties in Estimating and Comparing Provisional
Ballots Cast and Counted 242 Concluding Observations 248

249

Chapter 6

Overview In the 2004 General Election the Vote-Counting Process Remained
249 Complex and Marked by Local Diversity 250

  Page ii GAO-06-450 Elections

For the 2004 General Election, More States Had Requirements or

Guidance for Determining Voter Intent 259 Recounts and Contested Elections
Remained Uncommon for the

November 2004 General Election 264 Several States Had Further Developed
Their Specifications for an

Election Recount since the 2000 General Election 267 Vote Count Audits
Were Not Prevalent in the 2004 General Election,

but Some States Reported Taking Action to Require Them 270 Local
Jurisdictions Reported Some Mistakes and Technical

Challenges in Counting Votes for the 2004 General Election 273 In the 2004
General Election, Some Election Jurisdictions 

Reported Difficulty Meeting Deadlines for Certifying the 

Final Vote Count 276 The Recount in Washington State Revealed the
Intricacies of the

Election Process, but Also Yielded Many State Reforms 277 Concluding
Observations 283

286

Chapter 7

Overview Voting Methods Vary among Jurisdictions and Are Being  287
Technologies Influenced by States' Choices and HAVA 291

Voting Equipment Performance Was Not Consistently Measured during the 2004
General Election, but Estimated Levels of Performance Satisfaction Were
High 311

Attention to Voting System Security Management Continues  to Vary amid
Published Concerns and Federal Improvement Efforts 328

Certain Types of Tests and Evaluations Were Widely Performed on Voting
Systems, while Others Were Less Common 350

Information on Jurisdictions' Election Technology Integration Is Limited,
but Current and Emerging Technologies Offer Greater Opportunities 363

Concluding Observations 375

Appendixes

       Appendix I: Related GAO Products 378 Appendix II: 2005 Survey of State
           Election Officials 380 Appendix III: 2005 Survey of Local Election
Jurisdictions 420 Appendix IV: Summary of Local Election Jurisdictions GAO
                                                                  Visited 468

      Appendix V: Objectives, Scope, and Methodology                      485 
Appendix VI:   Voter Registration Deadlines for States and the             
                  District  of Columbia                                  497
Appendix VII:  Characteristics of Early Voting Jurisdictions Visited   499 
Appendix VIII: Selected State Statutory Requirements for Absentee         
                  Voting                                                  501
Appendix IX:   State Provisions for Accessibility of Polling Places        
                  and Alternative Voting Methods                          505
      Appendix X: States and the District of Columbia Reported                
                  Requirements for Local Jurisdictions to Use Federal     
                  Standards for Voting Systems                            508
Appendix XI:   GAO Contacts and Acknowledgments                        510 

Glossary

Tables Table 1: Estimated Percentage of Jurisdictions That Removed     
                   Names from Voter Registration List for 2004 General    
                   Election for Various Reasons                            59 
          Table 2: Estimated Percentage of Local Jurisdictions That       
                   Engaged                                                
                   in Various Quality Assurance Activities when Entering  
                   Completed Voter-Registration Applications               88 
          Table 3: Options Allowed by States for Requesting and Returning 
                   Absentee Ballot Applications, November 2004 General    
                   Election                                               106 
          Table 4: Estimated Percentages of All Local Jurisdictions and   
                   by                                                     
                   Jurisdiction Size That Encountered Voter Error         
                   Problems                                               
                   in Processing Absentee Ballot Applications, November   
                   2004 General Election                                  120 
          Table 5: Estimated Percentages of All Local Jurisdictions and   
                   by                                                     
                   Jurisdiction Size Using Various Contact Methods When   
                   Absentee Ballot Applications Could Not Be Processed,   
                   November 2004 General Election                         121 
          Table 6: Estimated Percentages of All Local Jurisdictions and   
                   by                                                     
                   Jurisdiction Size That Encountered Problems in         
                   Processing Submitted Absentee Ballots, November 2004   
                   General Election                                       122 

Table 7:    Estimated Percentages of All Local Jurisdictions and by  
              Jurisdiction Size Using Various Contact Methods when
Absentee Ballots Could Not Be Processed, November 2004
General Election                                                       123 
Table 8:    Estimated Percentages of Local Jurisdictions That Used   
                 Specific Procedures to Help Ensure Voters Did Not Vote 
              More than Once, November 2004 General Election              125 
Table 9:    Estimated Percentages of Local Jurisdictions That Used   
               Specific Procedures to Help Ensure Absentee Ballots Were 
                Actually Completed by the Person Requesting the Ballot, 
November 2004 General Election                                         126 
Table 10: Estimated Percentages of Local Jurisdictions, by           
               Jurisdiction Size, That Encountered Specific Voter Error 
                Problems on Absentee Applications That Used the Federal 
              Post Card Application, November 2004 General             
Election                                                               128 
Table 11: States Reporting Differing Mail-in Absentee Ballot         
Deadlines from Inside and Outside the United States,                 
November 2004 General Election                                         129 
Table 12: Early Voting Period in 24 Early Voting States and the      
District of Columbia, November 2004 General Election                   137 
Table 13: Estimated Percentages of Local Jurisdictions That Used     
                Various Staff Mixes at Early Voting Locations, November 
2004 General Election                                                  141 
Table 14: Estimated Percentages of Local Jurisdictions That Used     
Various Methods to Train Early Voting Staff, November                
2004 General Election                                                  142 
Table 15: Estimated Percentages of All Local Jurisdictions and by    
            Jurisdiction Size That Provided Alternative Voting          
Methods and Accommodations for the November 2004                     
General Election                                                       159 
Table 16: Estimated Percentages of All Local Jurisdictions and by    
            Jurisdiction Size That Took Steps to Provide Voter          
              Education Prior to the November 2004 General             
Election                                                               163 
Table 17: Estimated Percentages of All Local Jurisdictions and by    
           Jurisdiction Size That Used Sources to Recruit Poll          
              Workers for the November 2004 General Election              174 
Table 18: Estimated Percentages of Local Jurisdictions Nationwide    
             That Provided Provisional, Challenge, or Other Ballots for 
the November 2004 Election under Various Scenarios                     228 

Table 19: Estimated Percentages of Local Jurisdictions Using         
Various Methods to Inform Voters of the Outcome of Their             
Provisional Ballot, November 2004                                      240 
Table 20: Reported Provisional Ballots Cast and Counted in Selected  
Jurisdictions                                                          246 
Table 21: Estimated Percentages of Jurisdictions Having Written      
                 Ballot Security Policies and Procedures, November 2004 
General Election                                                       252 
Table 22: Estimated Percentages of Local Jurisdictions Using         
              Particular Voting Methods by Ballot Type for the November 
2004 Election                                                          254 
Table 23: States Reported Changes to Recount Provisions since 2000   
for the 2004 General Election                                          268 
Table 24: Voting System Security Tasks and Responsibilities for the  
                2004 General Election Reported by Election Officials in 
Jurisdictions Visited by GAO                                           333 
Table 25: Security Controls Reportedly Used in the 2004 General      
Election Reported by Election Officials in Jurisdictions             
Visited by GAO                                                         337 
Table 26: Voting System Security Topics Addressed in Security Plans  
                Submitted by Jurisdictions Visited by GAO                 343 
Table 27: Types of Testing and Evaluation for Voting Systems, with   
                 Common Time Frames and Responsibilities                  351 
Table 28: Population in Each State Group                               488 
Table 29: Local Election Survey Sample Allocation                      491 
Table 30: Local Election Jurisdiction Survey Response Rates,         
Overall and by Sample Strata                                           493 
Table 31: Local Election Jurisdictions Visited                         496 
Table 32: States Reporting on November 2004 General Election         
Registration Deadlines                                                 497 
Table 33: Selected Information on Jurisdictions We Visited That Had  
Early Voting                                                           499 
Table 34: Selected State Requirements for Domestic Absentee          
             Voting, November 2000 and 2004 General Elections             501 
Table 35: States' Mail-in Absentee Ballot Application and Mail-in    
Absentee Ballot Deadlines for Inside the United States,              
November 2004 General Election                                         503 
Table 36: State Provisions Concerning Accessibility of Polling      
Places                                                                 506 
Table 37: State Provisions for Alternative Voting Methods and        
Accommodations                                                         506 

                                    Figures

Table 38: State Provisions Concerning Accessibility of Polling Places
and Accommodations for Individuals with Disabilities for
the November 2004 General Election                                     507 
Table 39: States and the District of Columbia Reported               
                Requirements for Use of Voluntary Federal Standards for 
New Voting Systems for the November 2006 General                     
             Election                                                     508 
Figure 1: Estimated Percentage of Jurisdictions Having Problems      
                 Entering the Number of Voter Registration Applications 
               Received for the 2004 General Election                      14 
Figure 2:   Estimated Percentage of Local Jurisdictions That         
                 Encountered Lateness with Absentee Ballot Applications 
              and Absentee Ballots, November 2004 General              
               Election                                                    17 
Figure 3:   Estimated Percentages of All Local Jurisdictions and by  
              Jurisdiction Size That Encountered Difficulties Obtaining 
                   a Sufficient Number of Poll Workers for the November 
               2004 General Election                                       21 
Figure 4:   State-Reported Locations Where a Provisional Vote Had    
              to Be Cast in Order for It to Be Counted for the November 
               2004 General Election                                       24 
Figure 5:   Number of States Reporting Vote Count Audit              
               Requirements in Place for the 2004 General Election         26 
Figure 6:   Estimated Percentages of Jurisdictions Using a           
                 Predominant Voting Method in the 2000 and 2004 General 
               Elections                                                   29 
Figure 7:   The Election Process Involves the Integration of People, 
               Process, and Technology                                     40 
Figure 8:   Many States Report Challenges Matching Voter             
               Registration Information with State Records                 57 
Figure 9:   States Reporting That They Had or Would Have as of       
                 January 2006, Capability for Real-Time Access to Voter 
             Registration List to Enter, Update, and Query             
Information                                                             64 
Figure 10: States' Reported Capabilities to Electronically Match     
                Voter Registration Information with State Death Records 
and Felony Conviction Records                                           65 
Figure 11: Virginia Voter Registration Form Indicating whether the   
Voter Had Registered Elsewhere                                          70 
Figure 12: Colorado Voter Registration Form with Instructions for    
First-Time Mail Registrants                                             75 

Figure 13: Example of the Voter Registration Application Process        78 
Figure 14: Estimated Percentage of Jurisdictions Having Problems     
                 Entering the Number of Voter Registration Applications 
Received for 2004 General Election                                      82 
Figure 15: Total Weekly Voter Registration Applications              
                Documented by a Large Jurisdiction in 2004                 83 
Figure 16: Estimated Percentage of Local Jurisdictions with          
              Insufficient Election Workers to Process Voter            
Registration Applications                                               86 
Figure 17: Estimated Percentage of Local Jurisdictions That Had      
                  Names on Voter Registration Applications Appearing to 
                    Be Fraudulent and Volume of Fraudulent Applications 
                 Identified for the 2004 General Election                  91 
Figure 18: Mail-in Absentee Voting Process                             103 
Figure 19: Example of a Standard Absentee Ballot Application Form    
                  for the November 2004 General Election                  105 
Figure 20: Two Examples of Absentee Ballot Envelopes with the        
               Inclusion of Affidavit Information-One with              
               Certification on Inner Envelope and One with             
                  Certification on Outer Envelope-for the November 2004 
General Election                                                       108 
Figure 21: Estimated Percentage of Local Jurisdictions That          
                 Encountered Problems Processing Absentee               
Applications because the Applicant Did Not Meet the                  
                    Excuse Required by State Law, November 2004 General 
Election                                                               111 
Figure 22: Estimated Percentage of Local Jurisdictions That          
                 Encountered Lateness with Absentee Ballot Applications 
              and Absentee Ballots, November 2004 General              
Election                                                               115 
Figure 23: Example of Envelope Illustrating Official Postal Marking  
                   for Absentee Ballot Materials, November 2004 General 
Election                                                               118 
Figure 24: The Federal Post Card Application Used before and after   
the November 2004 General Election                                     131 
Figure 25: Twenty-four States and the District of Columbia Reported  
               Providing Early Voting as an Option in the November 2004 
General Election                                                       135 
Figure 26: Estimated Percentage of Local Jurisdictions That Offered  
                   Early Voting for Various Time Periods, November 2004 
General Election                                                       138 

Figure 27: King County, Washington, Poster Used to Inform            
Prospective Voters of New Identification                            
Requirements                                                           153 
Figure 28: Examples of Voter Information Signs Posted at Polling     
              Places for the November 2004 General Election               161 
Figure 29: Efforts to Inform Voters Prior to the November 2004       
General Election                                                       165 
Figure 30: States' Reported Identification Requirements for          
             Prospective Voters for the November 2004 General           
Election                                                               168 
Figure 31: Estimated Percentages of All Local Jurisdictions and by   
              Jurisdiction Size That Encountered Difficulties Obtaining 
                   a Sufficient Number of Poll Workers for the November 
2004 General Election                                                  172 
Figure 32: Estimated Percentages of All Local Jurisdictions and by   
             Jurisdiction Size That Encountered Difficulties            
              Recruiting Enough Poll Workers with Specific             
           Political Party Affiliation for the November 2004           
General Election                                                       176 
Figure 33: Estimated Percentages of All Local Jurisdictions and by   
             Jurisdiction Size That Encountered Difficulties            
                    Recruiting Poll Workers with Information Technology 
            Skills or Computer Literacy for the November 2004           
General Election                                                       178 
Figure 34: Materials Used to Recruit Poll Workers in Jurisdictions   
We Visited                                                             180 
Figure 35: Reported State Requirements for Poll Worker Training for  
the November 2004 General Election                                     183 
Figure 36: Examples of Instructions and Checklists Provided to Poll  
              Workers for the November 2004 General Election              189 
Figure 37: Estimated Percentages of All Local Jurisdictions and by   
              Jurisdiction Size That Encountered Difficulties Obtaining 
                 a Sufficient Number of Polling Places for the November 
2004 General Election                                                  191 
Figure 38: Town Hall in Madbury, New Hampshire                         192 
Figure 39: Estimated Percentages of All Local Jurisdictions and by   
              Jurisdiction Size That Encountered Difficulties Obtaining 
                   Enough Polling Places That Were Accessible to Voters 
with Disabilities for November 2004 General Election                   194 
Figure 40: El Paso County, Colorado, Optical Scan Ballot               200 

Figure 41: DRE and Optical Scan Voting Equipment That Large          
Jurisdictions Had to Prepare and Deliver Prior to Election           
Day                                                                    205 
Figure 42: Provisional Ballot Envelope                                 222 
Figure 43: Middlesex County, New Jersey, Challenged Voter            
Affidavit                                                              227 
Figure 44: Affidavit Containing Provisional Ballot Instructions for  
Poll Workers in Champaign, Illinois                                    233 
Figure 45: State-Reported Locations Where a Provisional Vote Had     
              to Be Cast in Order for It to Be Counted for the November 
2004 General Election                                                  235 
Figure 46: Examples of Written Information Provided to Voters to     
Inquire whether Their Provisional Vote Was Counted                     241 
Figure 47: Sealed Voting Equipment Bag                                 251 
Figure 48: Punch Card Tabulator Connected to Computer                  256 
Figure 49: Punch Card Voting Booth                                     262 
Figure 50: New Hampshire Guidance for Determining Voter Intent       
on Paper Ballots                                                       263 
Figure 51: Number of States Reporting Federal and Statewide          
            Recounts for the 2004 Primary or General Elections            265 
Figure 52: Number of States Reporting Vote Count Audit               
Requirements in Place for the 2004 General Election                    271 
Figure 53: Examples of Voting Equipment                                287 
Figure 54: Estimated Percentages of Jurisdictions Using              
             Predominant Voting Methods in the 2000 and 2004            
General Elections                                                      293 
Figure 55: Estimated Percentages of Jurisdictions Using              
                  Predominant Voting Methods in 2004, by Jurisdiction  
Size                                                                   295 
Figure 56: Estimated Percentages of Jurisdictions with Plans to      
                  Acquire Particular Voting Systems before the November 
2006 General Election                                                  297 
Figure 57: State-Reported Status of Lever and Punch Card Voting      
Equipment in 2000, 2004, and 2006                                      299 
Figure 58: Estimated Percentages of Local Jurisdictions That Used    
                Single and Multiple Voting Methods in the November 2004 
General Election                                                       302 
Figure 59: Precinct Count Optical Scan Voting Equipment                303 
Figure 60: State-Reported Involvement in the Use of Specific Voting  
              Methods in the 2004 and 2006 General Elections              305 

Figure 61: Estimated Percentages of Local Jurisdictions Identifying
Factors That Influenced Buying or Leasing Voting Equipment since the 2000
General Election 307

Figure 62: Federal Payments to States in Fiscal Year 2003 under HAVA to
Replace Lever and Punch Card Voting Equipment 308

Figure 63: Example of DRE Instructions and Equipment 310

Figure 64: Estimated Percentages of Jurisdictions That Collected Voting
System Performance Information for the 2000 and 2004 General Elections 316

Figure 65: Estimated Percentages of Jurisdictions That Collected
Information on Voting Accuracy for the 2004 General Election, by
Jurisdiction Size 318

Figure 66: Estimated Percentages of Jurisdictions That Collected
Information on Voting Equipment Reliability for the 2004 General Election,
by Jurisdiction Size 319

Figure 67: Estimated Percentages of Jurisdictions That Collected
Information on Voting Equipment Failures for the 2004 General Election, by
Predominant Voting Method 321

Figure 68: Estimated Percentages of Jurisdictions That Collected
Information on Voting System Efficiency for the 2004 General Election, by
Jurisdiction Size 322

Figure 69: Estimated Percentages of Jurisdictions Satisfied with Voting
System Performance for the 2004 General Election, by Jurisdiction Size 325

Figure 70: Estimated Percentages of Jurisdictions That Identified Security
Responsibilities for the 2004 General Election, by Jurisdiction Size 332

Figure 71: Estimated Percentages of Jursidictions That Used Security
Controls in the 2004 General Election, by Jurisdiction Size 335

Figure 72: Estimated Percentages of Local Jurisdictions That Documented
Security Policies or Procedures for Their Voting Systems in the 2004
General Election, by Jurisdiction Size 339

Figure 73: Estimated Percentages of Jurisdictions That Established
Policies or Procedures for Electronic Transmission of Voted Ballots in the
2004 General Election, by Predominant Voting Method 341

Figure 74: State-Reported Standards Required for Voting  Systems 346
Abbreviations:

Figure 75: State-Reported Voting System Certification Requirements  
for the 2004 General Election                                          356 
Figure 76: Estimated Percentages of Jurisdictions Conducting        
                Postelection Voting System Audits for the 2004 General 
                 Election, by Predominant Voting Method                   362 
Figure 77: Examples of Portable Memory Cards and Cartridges Used    
with Voting Equipment                                                  365 
Figure 78: Examples of Ballot Tabulation Equipment and              
Environments                                                           366 
Figure 79: Example of an Assembly for Producing a Voter-Verifiable  
               Paper Audit Trail for DRE Voting Equipment                 368 
Figure 80: Estimated Percentages of Jurisdictions That Used         
                   Networking or Remote Access for Voting Equipment in 
the 2004 General Election                                              370 
Figure 81: Estimated Percentages of Jurisdictions Using Automated   
           Election Support for the 2004 General Election, by          
Jurisdiction Size                                                      372 
Figure 82: Estimated Percentages of Local Jurisdictions with Plans  
to Acquire Voting Equipment                                            374 

          ADA          Americans with Disabilities Act                        
          DRE          direct recording electronic                            
          EAC          Election Assistance Commission                         
          FBI          Federal Bureau of Investigation                        
          FEC          Federal Election Commission                            
          FPCA         Federal Post Card Application                          
          FVAP         Federal Voting Assistance Program                      
          HAVA         Help America Vote Act                                  
          MCD          minor civil division                                   
          MVA          motor vehicle agency                                   
          NASED        National Association of State Election Directors       
          NIST         National Institute for Standards and Technology        
          NVRA         National Voter Registration Act                        
          PKI          public key infrastructure                              
          SSA          Social Security Administration                         
          UOCAVA       Uniformed and Overseas Citizens Absentee Voting Act    

                       VVPT voter-verifiable paper trail

This is a work of the U.S. government and is not subject to copyright
protection in the United States. It may be reproduced and distributed in
its entirety without further permission from GAO. However, because this
work may contain copyrighted images or other material, permission from the
copyright holder may be necessary if you wish to reproduce this material
separately.

Comptroller General.A of the United States

United States Government Accountability Office Washington, D.C. 20548

June 6, 2006

Congressional Committees

The November 2004 presidential election was not as close as the 2000
presidential election, but it still raised concerns about our election
processes. Following the 2004 general election, a number of members of
Congress asked GAO to review aspects of that election. In response to
these requests, GAO initiated a review under the authority of the
Comptroller General to examine an array of election issues of broad
interest to Congress.

This report focuses on the changing election processes in the United
States and the November 2004 general election. Specifically, for each
major stage of the election process-voter registration, absentee and early
voting, preparing for and conducting elections, provisional voting, and
counting the votes-plus voting methods, this report discusses (1) changes
to election systems since the 2000 election, including steps taken to
implement the Help America Vote Act, and (2) challenges encountered by
election officials in the November 2004 election.

Copies of this report are being sent to the congressional leadership and
the Chairman and Ranking Minority Member of the House Committee on
Science. Copies will also be sent to state election officials and the
election official for the District of Columbia and local elections
jurisdictions that participated in our research and will be made available
to other interested parties upon request. As a courtesy, we are providing
other members of Congress a copy of the report's highlights page and
executive summary.

If you or your offices have any questions about matters discussed in this
report, please contact me at (202) 512-5500; Norman J. Rabkin, Managing
Director, Homeland Security and Justice, at (202) 512- 8777; or William O.
Jenkins, Jr., Director, at (202) 512-8757. They can also be reached by
e-mail at [email protected] and [email protected], respectively. Contacts
and key contributors are listed in appendix XI.

David M. Walker Comptroller General of the United States

List of Congressional Committees

The Honorable Trent Lott Chairman  The Honorable Christopher J. Dodd
Ranking Minority Member  Committee on Rules and Administration United
States Senate

The Honorable Vernon J. Ehlers, III Chairman The Honorable Juanita
Millender-McDonald Ranking Minority Member Committee on House
Administration House of Representatives

The Honorable Arlen Specter Chairman The Honorable Patrick J. Leahy
Ranking Minority Member Committee on the Judiciary United States Senate

The Honorable F. James Sensenbrenner, Jr. Chairman The Honorable John
Conyers, Jr. Ranking Minority Member Committee on the Judiciary House
of Representatives

The Honorable Susan M. Collins Chairman The Honorable Joseph I.
Lieberman Ranking Minority Member Committee on Homeland Security and
Governmental Affairs United States Senate

The Honorable Tom Davis Chairman The Honorable Henry A. Waxman Ranking
Minority Member Committee on Government Reform House of Representatives

Executive Summary

Purpose

Faith in the fairness and accuracy of the U.S. election system is at the
foundation of our democracy. All eligible persons, but only eligible
persons, should be able to cast their votes and have their validly cast
votes counted accurately. Reports of problems encountered in the close
2000 presidential election with respect to voter registration lists,
absentee ballots, ballot counting, and antiquated voting systems raised
concerns about the fairness and accuracy of certain aspects of the U.S.
election system. Subsequently, the Help America Vote Act of 2002 (HAVA)
was enacted,1 and major election reforms are now being implemented. The
November 2004 general election highlighted some of the same challenges as
in 2000 as well as some new challenges in areas such as electronic voting
technology and implementation of some HAVA requirements. The issues that
arose in both elections highlighted the importance of the interaction of
people, processes, and technology in ensuring effective election
operations and maintaining public confidence that our election system
works.

The November 2004 general election was the first federal election that
tested changes states have made to their systems of election
administration since the 2000 general election and the first presidential
election since the enactment of HAVA. HAVA includes a number of provisions
related to the administration of federal elections affecting voter
registration, absentee voting, voting systems, and other election
administration activities.

The November 2004 general election was not as close or contentious as the
2000 general election, but media reports, interest groups, and members of
Congress raised concerns about various aspects of the elections process.
Following the November 2004 election, a number of members of Congress
asked us to review aspects of that election. In response to these
requests, we initiated a review under the authority of the Comptroller
General to examine an array of election issues of broad interest to
Congress using an approach similar to that we used to examine election
issues following the November 2000 election. During the design of the
review, GAO kept key committees of jurisdiction and interested parties
informed of its work.

This report focuses on the changing election processes in the United
States and the November 2004 election. It discusses (1) changes to
election systems since the 2000 election, including steps taken to
implement HAVA, and (2) challenges encountered by election officials in
the November 2004

1Help America Vote Act of 2002, Pub. L. No. 107-252, 116 Stat. 1666
(2002).

                                   Background

election for each major stage of the election process-voter registration,
absentee and early voting, Election Day preparation and activities,
provisional voting, and vote counting-and for voting technology.

In the United States, election authority is shared by federal, state, and
local entities. In addition to HAVA, federal laws have been enacted in
several major areas of the voting process such as the National Voter
Registration Act of 1993 (NVRA), which was designed to expand the
opportunities for citizens to register to vote in federal elections by
allowing registration by mail and at state motor vehicle agencies (MVA)
and other public agencies, and the Uniformed and Overseas Citizens
Absentee Voting Act of 1986 (UOCAVA), which facilitated absentee voting by
these populations.

The U.S. election system is highly decentralized, with primary
responsibility for managing, planning, and conducting elections residing
at the local jurisdiction level-generally, the county level in most
states, but some states have delegated election responsibility to
subcounty governmental units. Subcounty election jurisdictions in 9 states
account for about 75 percent of about 10,500 local election jurisdictions
in the United States, but about 12 percent of the 2000 U.S. Census
population. Local election jurisdictions vary widely in size and
complexity, ranging from small New England townships to Los Angeles
County, whose number of registered voters exceeds that of many states. Our
election system is based upon a complex interaction of people (voters,
election officials, and poll workers), processes (controls), and
technology that must work effectively together to achieve a successful
election. Every stage of the election process-registration, absentee and
early voting, preparing for and conducting Election Day activities,
provisional voting, and vote counting- is affected by the interface of
people, processes, and technology.

Following the November 2000 general election, GAO issued a series of
reports addressing a range of issues that emerged during that election.2
These reports also identified challenges that election officials reported
they faced in major stages of the election process. We have also issued
reports since the November 2004 general election on voter registration
issues and security and reliability of electronic voting. As appropriate,
information from our October 2001 comprehensive report on election
processes

2See appendix I for a list of GAO reports on elections since 1983.

                                Results in Brief

                               Executive Summary

nationwide was used as a basis for determining changes since 2000.3 Our
more recent reports were used to supplement this report on challenges
election officials faced in the November 2004 election. Our methodology
for this report included a Web-based survey of all 50 states and the
District of Columbia (all 51 responded) and a mail questionnaire sent to a
representative probability sample of 788 local election jurisdictions
nationwide, stratified by population (632, or 80 percent, responded). We
also conducted site visits to a nonprobability sample of 28 local election
jurisdictions in 14 states, selected to reflect variation in such factors
as geographic location, whether early voting was offered, whether recounts
for federal or statewide offices occurred, and voting technology used.4
Some of the 28 jurisdictions visited were among those we had visited for
our 2001 election report. In stratifying our nationwide mail survey of
local election jurisdictions, we grouped election jurisdictions by their
2000 U.S. Census population-small (less than 10,000), medium (10,000 to
100,000), and large (more than 100,000). These categories are also used in
this report to describe jurisdictions we visited. The results of our state
and local surveys are presented in two supplemental products that can be
found on our Web site at www.gao.gov.5

The most prevalent changes to state and local elections systems since the
2000 presidential election were changes required under HAVA, which, among
other things, established the U.S. Election Assistance Commission (EAC)
with wide-ranging duties that include providing information and assistance
to states and local jurisdictions with regard to election administration.
EAC is led by four Commissioners who are to be appointed by the President
and confirmed by the Senate. The Commissioners who, under HAVA, were to be
appointed by February 26, 2003, were appointed by the President in October
2003 and confirmed by the Senate in December 2003. Since beginning
operations in January 2004, EAC has achieved many of its objectives;
however, EAC has reported that its delayed start-up affected its ability
to conduct some HAVA-mandated activities within the

3GAO, Elections: Perspectives on Activities and Challenges across the
Nation, GAO-02-3, (Washington, D.C.: October 2001).

4Appendix V provides more detailed information about our scope and
methodology.

5GAO, Elections: 2005 Survey of State Election Officials, GAO-06-451SP
(Washington, D.C.: June 6, 2006); and GAO, Elections: 2005 Survey of Local
Election Jurisdictions,  GAO-06-452SP (Washington, D.C.: June 6, 2006).

  Page 5 GAO-06-450 Elections

time frames specified in the act. In turn, according to its fiscal year
2004 annual report, the delayed EAC start-up affected states' procurement
of new voting systems and the ability of some states and local
jurisdictions to meet related HAVA requirements by statutory deadlines.

In addition, HAVA included specific changes to certain aspects of state
administration of federal elections. Some key changes included
requirements for states to implement statewide voter registration lists, a
requirement that certain first-time mail registrants provide
identification with their registration application or when they vote for
the first time at the polls, and a requirement that most states permit,
under certain circumstances, the casting of provisional ballots-those cast
by voters at the polls whose eligibility to vote is unclear and to be
determined later-in elections for federal office. HAVA also provided for
funding to encourage states to replace their punch card and lever voting
equipment, and set out voting system standards that state voting systems
used in federal elections must meet. While HAVA defined some parameters
for these requirements, the act leaves the states discretion in choosing
the methods of implementing them. It is too early to determine the full
effect that HAVA's requirements may have on the elections process because
those requirements are in different stages of implementation. States had
to implement HAVA's requirements for provisional voting and identification
for first-time voters who register by mail prior to the November 2004
election. However, 41 states obtained waivers, allowed under HAVA, to
delay the implementation of their statewide voter registration systems
from January 1, 2004, to January 1, 2006. Moreover, states are in
different stages of replacing their older voting equipment, such as punch
card and lever machines, with newer technology.

On the basis of our surveys of states and local jurisdictions and visits
to selected jurisdictions, we found that states varied in their progress
in implementing their statewide voter registration lists and how they have
implemented their voting systems. Except for the 9 states that did not
obtain a waiver from HAVA's requirements for establishing a statewide
voter registration lists, all other states subject to the statewide list
requirement were not required to perform list maintenance activities as
defined in HAVA until the extended waiver deadline of January 2006. By the
November 2004 general election, states were in various stages of
implementing provisions of HAVA related to their statewide voter
registration lists and performing voter list verification and maintenance,
and had different capabilities and procedures at the state and local level
for performing required list maintenance functions. Thus, states are still
working to fully implement HAVA's voter registration requirements. As
states gain more experience with their statewide voter registration and
data-matching systems and processes, it is likely their systems and
processes will evolve. Given the continuing challenge of maintaining
accurate voter registration lists in a highly mobile society, this is to
be expected.

We also found that implementation of the identification provision for
certain first-time mail registrants varied. One noteworthy variation is in
the definition of mail registration: Some local jurisdictions we visited
said that applications received through voter registration drives would be
treated as mail registrations and thus would be subject to the HAVA
identification requirements. Other local jurisdictions we visited said
applications from registration drives were not to be treated as mail
registrations and therefore were not subject to the HAVA identification
requirements. As to the other two provisions, the results of GAO's survey
of state and local officials and jurisdictions we visited showed that
states varied in their implementation of HAVA's requirement for
provisional voting. One variation of particular note during the November
2004 election was the difference in state requirements regarding the
location where voters must cast their provisional ballots in order for
them to be counted. For example, in some jurisdictions, once the voter's
eligibility to vote had been verified, the provisional ballot was counted
if it was cast within the voter's county of residence, while in other
jurisdictions the ballot was counted only if the voter had cast it in the
assigned precinct. Notwithstanding these variations for implementing
provisional voting, it is clear that provisional voting has helped to
facilitate voter participation of those encountering eligibilityrelated
issues when attempting to vote.

Many states have taken advantage of federal funding to replace their punch
card and lever voting equipment with other voting methods. The results of
our survey of local election jurisdictions indicate that large
jurisdictions are replacing older voting equipment with more
technology-based voting methods to a greater extent than small
jurisdictions, which continue to use paper ballots extensively and
constitute the majority of jurisdictions across the United States. On the
basis of states' reported plans and local jurisdictions' estimated plans
for acquiring voting systems for future elections, the election technology
environment can be characterized as varied and evolving. Accordingly,
voting system performance management, security, and testing will continue
to be important to ensuring the integrity of the overall election process.

In addition to reporting the required HAVA changes, some states reported
having taken other actions since the 2000 general election to reform
election administration; for example, 6 states reported they had
eliminated the need for an excuse to vote absentee, and 9 states reported
establishing procedures to conduct an automatic recount (audit), in full
or in part, of the vote tabulation to help ensure accuracy of the vote
prior to certification.

Election officials identified challenges faced in the November 2004
general election. Some of these challenges were also identified as
challenges in GAO's October 2001 comprehensive report on the election
processes, while others were raised with us for the first time.

Continuing Challenges in 2004 o  Voter registration. According to our      
                                    nationwide survey of local election       
                                    jurisdictions and visits to selected      
                                    jurisdictions, many local                 
                                    jurisdictions reported that they          
                                    continued to encounter challenges with    
                                    the voter registration lists that they    
                                    had experienced in the 2000 general       
                                    election, such as difficulties related to 
                                    voter registration applications           
                                    with inaccurate and incomplete voter      
                                    registration information, multiple        
                                    registrations by the same person, or      
                                    ineligible voters appearing on the        
                                    list. Election jurisdictions also         
                                    continued to face challenges obtaining    
                                    voter registration applications from      
                                    motor vehicle agencies and other          
                                    NVRA entities.                            
                                    Absentee voting. The results of our       
                                 o  nationwide survey indicate that           
                                    election jurisdictions continued to       
                                    experience absentee voting                
                                    challenges that included receiving late   
                                    absentee voter applications and           
                                    ballots, managing large workloads with    
                                    inadequate resources,                     
                                    addressing voter error issues such as     
                                    unsigned or otherwise incomplete          
                                    absentee applications, and preventing     
                                    potential fraud. Although election        
                                    officials in jurisdictions we visited     
                                    provided examples of procedures           
                                    used to help protect against fraud such   
                                    as comparing signatures on                
                                    absentee applications to signatures on    
                                    registration applications, election       
                                    officials still suspected instances of    
                                    fraud. In 1 jurisdiction we visited,      
                                    election officials reported they referred 
                                    to the district attorney for              
                                    investigation matters pertaining to 44    
                                    individuals who allegedly voted           
                                    absentee ballots with invalid signatures. 
                                    Election Day activities. According to our 
                                 o  nationwide survey of local                
                                    election jurisdictions and visits to      
                                    selected jurisdictions, many local        
                                    jurisdictions reported that they          
                                    encountered many of the same              
                                    challenges preparing for and conducting   
                                    Election Day activities in the            

November 2004 general election as they did in November 2000, including
recruiting and training an adequate supply of skilled poll workers,
locating a sufficient number of polling places that met requirements,
designing ballots that were clear to voters when there were many
candidates or issues to include, having long lines at polling places, and
handling the large volume of telephone calls received from voters and poll
workers on Election Day. On the basis of our nationwide survey, we
estimate that large jurisdictions and, to some extent, medium
jurisdictions encountered these challenges more than small jurisdictions.

     o Vote counting. On the basis of interviews with election officials,
       many of the problems in managing people, processes, and technology for
       vote counting that had confronted election officials across the
       country in the November 2000 general election continued to be a
       challenge for them in the 2004 general election. Voting equipment
       problems, poll worker errors, and voter errors were reported as making
       it difficult to tabulate the votes quickly and accurately.
     o Voting technology. According to our local jurisdiction survey and
       visits, voting system performance measures have not been
       systematically embraced, reliable performance data have not been
       collected, and security and testing activities have not been
       consistently required and performed for all voting systems. As a
       result, effective management of voting technology remains a challenge
       for many states and local jurisdictions because election officials may
       continue to use a patchwork of operational indicators and anecdotal
       experiences, rather than requirements-based information on voting
       system performance, to support decisions regarding voting system
       investments and operations.

New Challenges in 2004  o  Registrations from registration drives.
Election officials in some local election jurisdictions we visited
reported that efforts of various groups to "get out the vote" by
registering new voters through voter registration drives created new
challenges not identified to us in the 2000 general election.
Specifically, at some local jurisdictions we visited, election officials
told us they faced a challenge processing large volumes of voter
registration applications just prior to the deadlines for registration.
The conditions that election officials reported experiencing in processing
the volume of voter registration applications, such as long hours and lack
of time to fully train temporary workers, can result in data entry errors
that would have the impact of not properly registering eligible voters and
not identifying ineligible voters. Moreover, while not reported as a
prevalent problem, applications received from voter registration drives
was a challenge reported by election officials, who said that some of
these applications had incomplete or invalid addresses, fictitious names,
or questionable signatures. On the basis of our nationwide survey of local
election jurisdictions, we estimate that  5 percent of local
jurisdictions had voter registration applications that appeared to have
fraudulent names.

     o Early voting. Election officials reported encountering new challenges
       managing early voting. Some local jurisdictions we visited reported
       that they experienced long lines at early voting locations resulting
       from larger than expected early voter turnout. In some jurisdictions
       we visited, election officials said that factors such as inadequate
       planning on their part, limitations on types of facilities that could
       be used for early voting locations, and funding constraints on hiring
       more staff or acquiring more voting locations affected their
       management of large early voter turnout.
     o New UOCAVA provision. A new challenge could develop for election
       officials as a result of a HAVA amendment to UOCAVA. In an effort to
       help make registration and voting easier for absent uniformed service
       voters and certain other civilian voters residing outside of the
       United States, this 2002 amendment extended the period of time that
       can be covered by a single application from the year during which the
       application was received to a time period covering up to the next two
       subsequent general elections for federal office, or 4 years. However,
       election officials in 4 jurisdictions we visited told us that a
       possible unintended consequence of this amendment could be that when
       uniformed services personnel are reassigned to other duty posts,
       absentee ballots may not be sent to the correct address for subsequent
       general elections. Even with a 2005 revision to the Federal Post Card
       Application form where voters can indicate that they want ballots for
       one federal election only, election officials were concerned that many
       absentee ballots would be returned as undeliverable.
          * Third-party polling place activities. Election officials in some
            of the jurisdictions we visited in states where the presidential
            race was considered close (often referred to as battleground
            states) reported encountering challenges with disruptive
            third-party (e.g., poll watchers, observers, and electioneers)
            activities at polling places on Election Day. In some instances,
            these third parties simply increased the number of
          * people that poll workers were to manage at a polling location; in
            others, election officials told us third-party observers'
            behavior negatively affected poll workers and voters.
     o Provisional voting. The implementation of provisional voting
       requirements as specified under HAVA highlighted another instance
       where states varied in their election systems, with somewhat distinct
       approaches for providing and counting provisional ballots. That is,
       states reported various differences in their counting processes such
       as the prescribed location (e.g., county or precinct) in which a voter
       must cast a provisional ballot in order for it to be counted. Another
       way states varied included circumstances, apart from those specified
       in HAVA, where a provisional ballot would be offered, such as when
       voters claimed they did not receive an absentee ballot. States also
       varied in the design of provisional ballots and how they tracked them.
     o Vote counting deadlines. A new phenomenon emerged as a challenge to
       election officials with respect to counting the votes: Some
       jurisdictions reported difficulty completing the extra steps required
       to verify and count provisional votes within the time allowed for
       tallying the final vote count.
     o Voting systems. States and local jurisdictions face a broad challenge
       in ensuring consistent accuracy, integrity, and security among their
       voting systems in light of their adoption of various versions of
       federal voluntary voting system standards containing somewhat
       different-and, in some cases, outdated-performance thresholds for
       voting equipment. Adoption of the 2005 Voluntary Voting System
       Guidelines by EAC on December 13, 2005, provided updated criteria that
       states and local jurisdictions can choose to apply when evaluating and
       certifying their voting equipment. Organizations involved with voting
       system certification-including federal, state, and local governments;
       testing authorities; and vendors-may need the capacity to assume the
       workloads associated with the adoption of current standards, including
       upgrading, testing, and certifying newly acquired voting systems to
       meet the standards, particularly if the standards are to be applied to
       the 2006 general election. Furthermore, as states and jurisdictions
       move to a more integrated suite of election systems, proactive and
       systematic efforts in areas such as standards will be essential to
       addressing emerging technical, security, and reliability interactions
       among systems and managing risks in this dynamic election environment.

                               Principal Findings

Voter Registration In managing the voter registration process and
maintaining voter registration lists, state and local election officials
must balance two goals- minimizing the burden on eligible persons of
registering to vote, and ensuring that voter lists are accurate, that is,
limited to those eligible to vote and that eligible registered voters are
not inadvertently removed from the registration lists. During 2004 and
2005, many states were in the process of implementing their HAVA-required
statewide voter registration lists and associated requirements for
maintaining such a list. Thus, the potential benefits to be gained from
HAVA's requirement for a statewide voter registration list were not
evident in many states at the time of the November 2004 general election.
Maintenance requirements in HAVA intended to help states and local
election jurisdictions have access to more accurate voter registration
list information, such as identifying duplicate registrations and matching
the voter information against other state agency databases or records,
were not yet fully implemented by many states. Many local jurisdictions
were not yet seeing the benefits of being able to verify voter
registration application information with state motor vehicle agency
databases to identify eligible voters, or to match voter registration
lists with a state agency's records to identify felons who may be
ineligible to vote.

Local jurisdictions also encountered instances where voters claimed to be
registered to vote and their names were not on the voter registration
list. When this occurs, under HAVA's provisional voting requirements,
states must permit voters to cast provisional ballots if the voters assert
that they are registered in the jurisdiction where they desire to vote and
are eligible to vote in a federal election. The results of our nationwide
survey of local election jurisdictions indicate that many local
jurisdictions encountered problems determining whether a provisional
ballot was eligible to be counted where voters claimed to have registered
at a motor vehicle agency or at another NVRA entity but there was
insufficient evidence that the voter had submitted a registration
application at one of those offices.

While registering to vote appears to be a simple step in the election
system generally, applying to register and being registered are not
synonymous, and election officials face challenges in processing the voter
registration applications they receive. Local election jurisdictions
continued to encounter challenges with the voter registration lists for
the November 2004 election such as difficulties related to receiving
inaccurate and incomplete voter registration information, multiple
registrations, and ineligible voters appearing on the lists. The surge of
last-minute registrations in many jurisdictions prior to the November 2004
election illustrated the challenge of balancing ease of registration with
assurances that only eligible voters are on the registration rolls. In
some cases, election officials reported that hundreds or thousands of
applications were submitted just before the registration deadline and
close to Election Day. According to our nationwide survey and visits to
selected jurisdictions, entering voter registrations in a timely manner
presented a challenge for some election officials in marshaling the needed
resources, including in some cases hiring and training temporary
employees, to review the applications, obtain missing or incomplete
information from applicants, determine that the registrants were eligible
to vote in the jurisdiction, and ensuring that the names of eligible
voters were added to the voter registration rolls prior to Election Day.
As shown in figure 1, we estimate that 19 percent of jurisdictions
nationwide received applications just prior to the registration deadline
that posed problems in entering them prior to Election Day, with large
jurisdictions experiencing problems more than medium and small
jurisdictions.6

6Our nationwide survey of local election jurisdictions was designed to
have maximum sampling errors of +/- 5 percentage points for the complete
sample.

  Page 13 GAO-06-450 Elections

Figure 1: Estimated Percentage of Jurisdictions Having Problems Entering
the Number of Voter Registration Applications Received for the 2004
General Election

Percentage of jurisdictions

50 49

40

30

21

19

20

16

10

0 Experienced problems entering applications

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000)

Overall Source: GAO 2005 survey of local election jurisdictions.

Note: Large jurisdictions are statistically different from both medium and
small jurisdictions.

During our site visits, 1 large jurisdiction we visited reported that on a
daily basis it was 30,000 to 40,000 applications behind in data entry. As
a result, election officials reported that they hired 80 full-time
temporary workers who worked two full-time shifts to enter all eligible
applications into the voter registration list used at the polls on
Election Day. Election officials in another large jurisdiction told us
that they unexpectedly received about 10,000 last-minute registration
applications.

According to our nationwide survey of local election jurisdictions and
election officials in jurisdictions we visited, many local election
jurisdictions had processes to help manage receipt of voter registration
applications such as training for MVA and other NVRA entities' staff and
local election office's staff for data entry and tracking of registration
application forms. However, some local jurisdictions did not report having
such management processes. We estimate that 76 percent of all
jurisdictions provided training to data entry staff about the processing
and inputting of registration applications, and we estimate that over half
of all jurisdictions tracked incoming registration applications to
ascertain the total number received, the number entered into registration
lists, and the number not processed because of omission or application
error, and to identify ineligible voters based on age or residence. In
addition, some local jurisdictions we visited reported that they
implemented processes such as tracking the number of applications
distributed and the source from which applications are received, and
providing receipts to voter registrants to help alleviate problems
encountered with properly registering voters.

In addition to challenges encountered processing the large volume of
registration applications received through voter registration drives, on
the basis of our nationwide survey of local election jurisdictions, we
estimate that 5 percent of local jurisdictions had voter registration
applications that appeared to have fraudulent names. Election officials in
some jurisdictions we visited reported receiving voter registration
applications that had irregularities. For example, election officials in 1
jurisdiction reported receiving applications that were unreadable, had
questionable signatures, were incomplete, or had invalid addresses.
Election officials in another jurisdiction also reported receiving
applications with fictitious names and fake signatures. Generally,
election officials reported that the number of applications that were
irregular were few in number, especially in relation to the total number
of applications received.

Absentee and Early Voting Some states have increased the opportunities for
citizens to vote absentee or early. For the November 2004 general
election, 3 additional states reported that they no longer required voters
to provide excuses such as being ill, having a disability, or being away
from the precinct on Election Day to vote absentee. Three states reported
expanding their provisions for permanent absentee status (usually reserved
for the elderly or individuals with disabilities), allowing voters to
receive absentee ballots for a statespecified time period, such as 4
years. One state reported eliminating its requirement that mail-in
absentee voters provide an affidavit from a notary or witness for their
signature along with the completed absentee ballot. Furthermore, HAVA
amended UOCAVA to, among other things, extend the period of time that can
be covered by a single absentee ballot application by absent uniformed
service voters and certain other civilian voters residing outside of the
United States from the year during which the application was received to a
time period covering up to the two next regularly scheduled general
elections for federal office.

Absentee voting. Voting prior to Election Day can make voting easier for
voters but can also create challenges for election officials. On the basis
of our nationwide survey of local election jurisdictions, more than half
of all jurisdictions encountered problems receiving absentee ballot
applications and absentee ballots from voters too late to process-an
estimated  55 percent of jurisdictions received applications too late and
an estimated 77 percent received ballots too late, as shown in figure 2.

Figure 2: Estimated Percentage of Local Jurisdictions That Encountered
Lateness with Absentee Ballot Applications and Absentee Ballots, November
2004 General Election

Percentage of jurisdictions

100 97

Applications Ballots received received too latea too lateb

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000)

Overall Source: GAO 2005 survey of local election jurisdictions.

aAll size categories are statistically different from one another.

bSmall jurisdictions are statistically different from medium and large
jurisdictions.

Although the extent of the problem in terms of the number of applications
and ballots that could not be processed is unknown, the estimated number
of jurisdictions encountering the problem would seem to be of some concern
to state and local election officials. Absentee application deadlines
close to Election Day provide citizens increased time to apply to vote
absentee. But such deadlines can create difficulties for election
officials, providing a short period of time to ensure that eligible voters
receive absentee ballots in time to vote, including having time to notify
voters and have the voters correct errors on their ballot applications,
such as failing to sign them. The impact of absentee ballot application
deadlines on voters' ability to complete and return the absentee
application and ballot in time for their votes to be counted is another
example of the difficulties of balancing voter access and ease of voting
with appropriate election administration processes and controls.

Election officials identified problems related to voter errors on absentee
ballot applications and ballots. On the basis of our nationwide survey of
election jurisdictions, we estimate that almost half of the jurisdictions
encountered problems with missing or illegible signatures on absentee
ballot applications, and our survey results also indicate that local
jurisdictions encountered problems with, among other things, missing or
inadequate voting residence addresses on absentee applications and missing
or incomplete witness information for a voter's signature or information.
In jurisdictions that we visited, some election officials told us of steps
they took, when time permitted, to address voter errors. States reported
having information on their Web sites that included information on the
basic requirements for requesting and casting an absentee ballot. In
addition, some absentee voting applications and ballots provided to us by
elections jurisdictions we visited included instructions for voting
absentee.

Mail-in absentee ballots are considered by some to be particularly
susceptible to fraud. Election fraud could include such activities as
completion of a ballot by someone other than the registered voter or an
attempt by a voter to cast more than one ballot in an election. On the
basis of our nationwide survey, we estimated that a majority of
jurisdictions used procedures in the November 2004 election designed to
help ensure that absentee voters did not vote more than once and that
absentee ballots were actually completed by the person requesting the
ballot. However, some mail-in absentee voter fraud concerns remained. In
particular, election officials expressed concern regarding absentee voters
being unduly influenced or intimidated while voting by third parties who
went to voters' homes and offered to assist them in voting ballots.
Election officials also expressed concerns about the influence of third
parties on voters for early voting when voters waiting in line were
approached by candidates and poll watchers.

Uniformed military and overseas absentee voters. Election officials in a
few jurisdictions we visited told us of a possible unintended consequence
that may create a challenge with respect to provisions in UOCAVA as
amended by HAVA, whereby the Federal Post Card Application can possibly
cover as many as two subsequent general elections for federal office.
Election officials in some jurisdictions we visited said that when
uniformed service personnel are reassigned to other duty posts, ballots
might not be sent to the correct address for subsequent general elections.
Election officials in some of these jurisdictions said they were taking
steps to help ensure that absentee ballots would be sent to the correct
address for the 2006 general election, such as requesting e-mail addresses
as a means to obtain information to update mailing addresses or conducting
mass mailings to these voters to confirm mailing addresses. However,
election officials were concerned that if these efforts are not
successful, a number of the ballots mailed to addresses provided on the
Federal Post Card Application for the November 2006 election would be
returned as undeliverable. This was a concern for these election officials
because the jurisdictions would have to absorb the expense of mailing
ballots that would be undeliverable. Furthermore, a potential effect may
be that some uniformed services voters, who applied to vote absentee using
the Federal Post Card Application, may not receive their ballots for
subsequent general elections. As noted in our April 2006 report on
election assistance provided to uniformed service personnel, one of the
top two reasons for disqualifying absentee ballots for these voters was
that they were undeliverable.7

Early voting. With respect to early voting, election officials in some
jurisdictions we visited identified obtaining adequate staffing for
conducting early voting as a challenge, especially when given the
unanticipated large early voter turnout. In 11 of the 14 early voting
jurisdictions visited, election officials emphasized the importance of
staffing early voting locations with experienced staff such as election
office staff or experienced and seasoned poll workers. According to our
nationwide survey of local jurisdictions, we estimate that 30 percent of
jurisdictions used permanent staff to work early voting polling locations.
Our nationwide survey also showed that jurisdictions used other types of
staff and combinations of staff such as permanent and part-time staff.
Depending on the number of early voting locations to be staffed, using
experienced staff may not always be feasible, and using other staff may
affect the speed with which voters can be processed and may contribute to
long lines. As states fully implement their statewide voter registration
lists, processing voters at early voting locations may become easier as
the voter registration systems evolve and systems become user-friendly so
that all types of staff can be more effective in processing voters.

7GAO, Elections: Absentee Voting Assistance to Military and Overseas
Citizens Increased for the 2004 General Election, but Challenges Remain,
GAO-06-521 (Washington, D.C.: Apr. 7, 2006).

  Page 19 GAO-06-450 Elections

                              Conducting Elections

States and local jurisdictions have reported making changes since the
November 2000 general election as a result of HAVA requirements and other
state actions to improve the administration of elections in the United
States. HAVA established requirements with respect to elections for
federal office for, among other things, certain voters who register by
mail to provide identification prior to voting; mandated that voting
equipment accessible to individuals with disabilities be located at each
polling place; and required that voter information be posted at polling
places on Election Day. Since the November 2000 general election, some
states have also reported making changes to their identification
requirements for all voters.

Many of the challenges that election officials reported encountering in
preparing for and conducting the November 2004 general election were not
new challenges. Recruiting and training an adequate supply of poll
workers, finding accessible polling places, and managing communications on
Election Day were challenges that we identified in our October 2001 report
on the November 2000 general election.8 As shown in figure 3, on the basis
of our nationwide survey, we estimate that more large jurisdictions
encountered difficulties than medium and small jurisdictions when it came
to obtaining a sufficient number of poll workers.

8GAO-02-3.

Figure 3: Estimated Percentages of All Local Jurisdictions and by
Jurisdiction Size That Encountered Difficulties Obtaining a Sufficient
Number of Poll Workers for the November 2004 General Election

Percentage of jurisdictions

                                       57

50

47e

46 44

40

35 32

30

30

24

22 20

20

17 14

10

5

4 11

0 Difficult or very Neither difficult Easy or very Not applicabled

c

difficulta nor easyb easy

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000)

Overall Source: GAO 2005 survey of local election jurisdictions.

aAll size categories are statistically different from one another.

bThe difference between small and medium jurisdictions is statistically
significant.

cThe differences between both large and medium jurisdictions and small
jurisdictions are statistically significant.

dJurisdictions could indicate not applicable for a variety of reasons,
including that poll workers are not recruited, but elected or appointed;
that elections are conducted by mail ballot, and as a result there is not
a need for poll workers to staff polling places on Election Day; or that
the election officials themselves serve as poll workers.

eThe 95 percent confidence interval for this percentage is +/-8 percentage
points.

Administering an election in any jurisdiction is a complicated endeavor
that involves effectively coordinating the people, processes, and
technologies associated with numerous activities. However, we found in our
survey of local jurisdictions and site visits to 28 localities that more
large and, to some extent, medium jurisdictions reported that they
encountered

                               Provisional Voting

challenges in preparing for and conducting the November 2004 general
election than small jurisdictions did. This may be because the complexity
of administering an election and the potential for challenges increase
with the number of people and places involved and the scope of activities
and processes that must be conducted, such as the need to provide ballots
and voter assistance in languages other than English. The results of our
local election jurisdiction survey indicate that more large and medium
jurisdictions than small jurisdictions took steps-such as through voter
education or providing instructions at polling places for poll workers-
designed to minimize potential problems. Many of the election officials in
large jurisdictions we visited told us that being well prepared, having
established policies and procedures in place, and having high-quality
election staff were factors that contributed to a smooth Election Day. One
problem that election officials in some jurisdictions reported
encountering on Election Day was actions by poll watchers and other third
parties that election officials considered disruptive. This presents
another issue that election officials may need to include in their
Election Day preparations and training.

Concerns were raised with respect to the November 2000 election that some
eligible voters were not allowed to vote because of questions regarding
the voters' eligibility. HAVA required that by January 1, 2004, states
permit the casting of provisional ballots in elections for federal office
by voters who assert that they are eligible to vote and registered in that
jurisdiction, but are not found on the voter registration list.9 Such
states are also required under HAVA to provide provisional ballots in
federal elections under other circumstances, such as for certain voters
who registered by mail and do not have required identification. While HAVA
requires that states permit an individual under certain circumstances to
cast a provisional ballot in a federal election, the act left the specific
choices on the methods of implementation to the discretion of the states.
Under HAVA, election officials receiving provisional voter information are
to determine whether such individuals are eligible to vote under state
law. If an individual is determined to be eligible, HAVA specifies that
such individual's provisional ballot be counted as a vote in that election
in accordance with state law.

9Six states are exempt from this requirement, in general, either because
they permit voter registration on Election Day (Idaho, Minnesota, New
Hampshire, Wisconsin, and Wyoming) or because they do not require voter
registration (North Dakota).

  Page 22 GAO-06-450 Elections

On the basis of our survey of state election officials, our nationwide
survey of election jurisdictions, and our visits to jurisdictions, states
and local jurisdictions varied in a number of ways in how they implemented
HAVA's provisional voting requirements in the November 2004 election. For
example, in addition to those specified in HAVA, the circumstances
reported by states and local jurisdictions when a provisional ballot would
be offered varied, with some jurisdictions allowing voters claiming they
did not receive an absentee ballot to vote provisionally. The results of
our survey of state elections officials showed that states also varied as
to the location where voters must cast their provisional ballots in order
for such ballots to be eligible to be counted, as shown in figure 4.

Figure 4: State-Reported Locations Where a Provisional Vote Had to Be Cast
in Order for It to Be Counted for the November 2004 General Election

Source: GAO survey of state election officials.

Note: Six states are not subject to HAVA's provisional voting
requirements, but 2 of these 6 (Wisconsin and Wyoming) authorize some
measure of provisional voting. Both of these states are included with the
32 states that reported requiring that provisional voters must cast their
votes in the specific precincts in which they are registered in order for
their votes to be eligible to be counted.

                               Counting the Votes

On the basis of our interviews with local election officials, local
election procedures and unique circumstances add to the differences among
jurisdictions. For example, in some jurisdictions we visited, election
officials described various factors that affected the counting of
provisional ballots, such as the time allowed for provisional voters to
provide missing identification. Specifically, in 1 jurisdiction, voters
had to provide the required identification before the polls closed for the
ballot to be counted, while in other jurisdictions the ballot would be
counted if the voter provided the required identification within a
specified number of days after Election Day.

These variations in provisional voting implementation highlight how
individual state rules, procedures, and practices may have affected the
number of provisional ballots cast and counted in the November 2004
election. These differences and limited data availability make it
difficult to determine with certainty how many provisional ballots were
cast and counted nationally in the November 2004 election. However, the
data that are available indicate that the HAVA requirement for provisional
voting has helped to better facilitate voter participation of those
encountering eligibility-related issues when attempting to vote.

Although the methods used to secure and count ballots vary across the 50
states and the District of Columbia, the goal of vote counting is the same
across the nation: to accurately process those ballots requiring
verification and accurately count every valid ballot. As with the
elections process overall, conducting an accurate vote count is not a
simple process. It requires many steps, an unerring attention to detail,
and the seamless integration of people, processes, and technology.

In 2004, vote counting remained an intricate multistep process
characterized by a great variety of local procedures depending on a local
jurisdiction's technology, size, and preferences. The multistep process
can involve such activities as the initial vote count, a vote count audit
to verify the accuracy of the count, certification of the vote count, and
recounts of the votes when an election is close.10 There were some notable
developments related to conducting recounts that may be mandatory

10As used in this report, a vote count audit is an automatic recount, in
full or in part, of the vote tabulation, irrespective of the margin of
victory, in order to ensure accuracy before certification.

  Page 25 GAO-06-450 Elections

(because of a close margin of victory) or requested. Some states reported
that they added rules for mandatory recounts. Others reported that they
changed their guidance for who may request a recount. Regarding vote count
audits, while 29 states and the District of Columbia reported they did not
require audits of vote counts, 21 states reported having provisions that
required or allowed audits of vote counts, as shown in figure 5.
Furthermore, 9 states reported having taken some legislative or executive
steps toward doing so since November 2004.

Figure 5: Number of States Reporting Vo t e Count Audit Requirements in
Place fo r the 2004 General Election

Number of states 35

30

30 25 20 15

11

10

8

5

2

0

                     d auditsfor d auditsfor jurisdictions

Requirewed localall local jurisdictionssome local jurisdictions to conduct
                         auditsDid not requirew auditsa

Require

                                    or allo

                                      Allo

Audits

Source: GAO 2005 survey of state election officials.

aIncludes District of Columbia.

Providing eligible voters multiple means and times within a jurisdiction
for casting their ballots-early, absentee, provisional, and Election Day
voting-enhances eligible voters' opportunity to vote. At the same time,

  Page 26 GAO-06-450 Elections

multiple voting methods and types of ballots can make the vote-counting
process more complicated. In addition, short deadlines for certifying the
final vote-as little as 2 days in 1 state-provide little time for election
officials to review, verify, and count provisional and absentee ballots.
Larger jurisdictions generally face more challenges than smaller
jurisdictions because of the sheer volume of votes cast by all ballot
types- absentee, provisional, and regular ballots. Provisional ballots
were new for many jurisdictions in November 2004 and created some
challenges in tracking, verifying, and counting. On the basis of their
experience in November 2004, some election officials in jurisdictions we
visited said that they are implementing new procedures for provisional
voting, such as printing provisional ballots in a color different from
other types of ballots or using paper ballots rather than direct recording
electronic (DRE) machines for provisional voters. These procedures are
intended to help election officials track provisional ballots to ensure
that they are all accounted for and included in the vote count.

Two jurisdictions we visited in Washington told us that they are moving to
all-mail elections, which was authorized on a countywide basis by a new
state law. Although replacing in-person voting with all-mail voting
eliminates some challenges that can affect accurate vote counting-e.g.,
poll worker training on voting equipment operations and provisional voting
or the chance of malfunctioning voting equipment at the polls-it magnifies
the importance of other aspects of the process, such as accurately
matching voter signatures and having clear guidance for determining voter
intent from improperly or unclearly marked ballots.

The recount in the close gubernatorial election in Washington revealed the
interdependence of every stage of the elections process in ensuring an
accurate vote count. In the initial statewide count, a mere 261 votes
separated the two top candidates, and an initial recount reduced that
margin of victory to just 42 votes out of more than 2.7 million cast, and
the final recount resulted in a 129-vote margin of victory for the
candidate who came in second in the first two vote counts. The experiences
of election jurisdictions that had to conduct the recounts illustrated how
small errors in the election administration process can affect the vote
count. For example, in at least 11 counties provisional ballots were found
by a Washington state superior court to have been counted without
verifying voter signatures or before verification of voter registration
status had been completed. Furthermore, 573 absentee ballots were
erroneously disqualified in one county, and 22 absentee and provisional
ballots were discovered in the base units of optical scan machines in
another county

Voting Methods and Technologies

after the election had been certified. Were any state's election processes
subjected to the very close scrutiny that characterized the recount in
Washington state, it is likely that imperfections would be revealed. Votes
are cast and elections are conducted by people who are not and cannot be
100 percent error free in all their tasks all the time. Thus, the
consistently error-free vote count may be elusive, particularly in very
large jurisdictions with hundreds of thousands of ballots cast in person,
absentee, or provisionally. However, diligent efforts to achieve a
consistently error-free count can help to ensure that any errors are
reduced to the minimum humanly possible.

The technology of the voting environment can be characterized as varied
and evolving, according to our 2005 state survey results and local
jurisdiction survey estimates. Figure 6 shows the estimated percentages of
all jurisdictions' use of a predominant voting method in the 2000 and 2004
general elections. Two key patterns emerged in the use of voting methods
between the 2000 and 2004 general elections. First, we estimate that the
percentage of large jurisdictions using DREs doubled from 15 percent in
the 2000 general election to 30 percent in 2004. The predominant voting
method most often used for large jurisdictions changed from precinct count
optical scan in 2000 to both DRE and precinct count optical scan in 2004.
In contrast, we estimate that the predominant voting methods most often
used remained the same for small and medium jurisdictions (paper ballots
and precinct count optical scan, respectively) from 2000 to 2004.
Furthermore, on the basis of our local jurisdiction survey, we estimate
that at least one-fifth of jurisdictions plan to acquire DRE or optical
scan equipment before the 2006 general election. Second, in response to
our state survey, 9 states reported that they eliminated the lever machine
and punch card voting methods for the 2004 general election. In addition,
 18 other states plan to eliminate lever or punch card voting methods for
the 2006 general election. This greater state involvement in
jurisdictions' choice of voting methods, the availability of federal
funding to replace lever and punch card voting equipment, and certain HAVA
requirements-among other factors-are likely influences on the adoption of
DRE and optical scan voting methods.

Figure 6: Estimated Percentages of Jurisdictions Using a Predominant
Voting Method in the 2000 and 2004 General Elections

Percentage of jurisdictions

                                      scan

                                   scan hine

Central count

                          Precinct countCentral count

                           Precinct countaper ballot

DRE

                                       d

                                       d

                                     h car

                                     h car

                                    ver mac

optical

                                    optical

                                      punc

                                      punc

                                       P

                                       Le

Predominant voting method

2000

2004 Source: GAO 2005 survey of local election jurisdictions.

HAVA recognized the importance of voting system performance by specifying
requirements for error rates in voting systems and providing for updates
to the federal voting system standards, including the performance
components of those standards. According to our local jurisdiction survey,
most local jurisdictions adopted performance standards for accuracy,
reliability, or efficiency for the 2004 general election-usually standards
selected by their respective states. It is important that system
performance be measured during an election, when the system is being used
and operated according to defined procedures by voters and election
workers, to provide a basis for determining where performance needs,
requirements, and expectations are not being met so that timely corrective
action can be taken. As was the case for the 2000 general election,
jurisdictions collected various types of voting system performance
measures for the 2004 general election, although some types of measures
were collected by fewer jurisdictions than others-in part because they
were not well suited to particular voting methods. From our local
jurisdiction survey, we estimate that the vast majority of all
jurisdictions were very satisfied or satisfied with their systems'
performance during the 2004 general election, even though performance data
may not have been collected to an extent that would provide firm support
for these views. The moderate collection levels of data on operational
voting systems' performance may present a challenge to state and local
election officials in their efforts to make informed decisions on both
near-term and long-term voting system changes and investments.

Having secure voting systems is essential to maintaining public confidence
in the election process, and according to our local jurisdiction survey
estimated results, accomplishing this was a shared responsibility among
states, local jurisdictions, vendors, law enforcement officials, and
others for most jurisdictions. According to our state survey, estimates
from our local jurisdiction survey, and visits to jurisdictions, there
were differences across states and jurisdictions in areas such as the
adoption of system security standards and reported implementation of
system security controls, which was generally consistent with what we
reported in our October 2001 report on election processes. In addition, 27
states reported in our state survey that they are requiring jurisdictions
to apply voluntary federal standards to voting systems used for the first
time in the November 2006 general election that are outdated, unspecified,
or entail multiple versions. In the area of testing, most states reported
that they required national or state certification of their voting
systems, but the systems covered by those requirements and the criteria
used for certification also varied by state and by voting method.
Readiness (logic and accuracy) testing continued to be commonly performed
by an estimated 92 percent of local jurisdictions that used automated
voting systems for the 2004 general election, but the local election
officials we talked with described a variety of testing approaches.11 We
estimate that two other forms of testing- parallel testing and
postelection auditing of voting equipment-were much less prevalent than
readiness testing and were conducted by 2 percent and 43 percent of
jurisdictions that used automated voting, respectively.12

11For the questions in our local survey related to types of testing,
jurisdictions that used only hand-counted paper ballots on Election Day
were excluded.

12We estimate that 91 percent of jurisdictions considered parallel testing
to be not applicable.

  Page 30 GAO-06-450 Elections

Appropriately defined and implemented standards for system functions and
testing processes are essential to ensuring the accuracy, integrity, and
reliability of voting systems across all phases of the elections process.
States and local jurisdictions face the challenge of regularly updating
and consistently applying appropriate standards and other directives for
security management and testing to address vulnerabilities and risks in
their specific election environments.

The number of jurisdictions that had integrated particular aspects of
voting system components and technologies was limited for the 2004 general
election according to estimates from our local jurisdiction survey and
visits to local jurisdictions for the selected areas of integration we
examined, such as electronic programming or setup and electronic
management. Furthermore, relatively few local jurisdictions we visited
reported having plans for integrating or further integrating their
election-related systems and components for the 2006 general election, and
in the instances where jurisdictions reported plans, the scope and nature
of the plans varied. Nevertheless, there is real potential for greater
integration among voting systems, election systems, and components as
states and jurisdictions act on plans to acquire optical scan and DRE
equipment that lends itself to integration. It is unclear if and when this
migration to more technologybased voting methods will produce more
integrated election system environments. However, suitable standards and
guidance for these interconnected components and systems-some of which
remain to be developed-could facilitate the development, testing,
operational management, and maintenance of components and systems, thereby
maximizing the benefits of current and emerging election technologies and
achieving states' and local jurisdictions' goals for performance and
security. The challenge inherent in such a dynamic environment is to
update system standards so that emerging technical, security, and
reliability interactions are systematically addressed.

The administration of election systems will never be error free or
perfect.

Concluding

Each stage of the election process poses a major challenge for election

officials. Effective management of the election system requires a variety
of resources that must be prepared, mobilized, and deployed at regular
intervals. These resources include the people who conduct the election and
participate in it, the processes that govern what the people do and how
the election is conducted, and the technology that facilitates the efforts
of the people as they work through the election processes. Although
responsibility for election administration falls largely on local
governmental units, state and federal governments have a role to play in
helping to minimize the types of errors that can occur. Thus, as
technology evolves and circumstances warrant, state and federal
governments might consider what, if any, actions on their part could help
to improve election processes.

GAO found that states have made changes-either as a result of HAVA or on
their own-that addressed some of the challenges identified in the 2000
general election. GAO also found that some challenges continue and new
challenges occurred in the 2004 general election. In passing HAVA,
Congress provided a means for states and local jurisdictions to improve
upon several aspects of the voting administration system. It is too soon
to determine the full effect of those changes, especially the requirement
for statewide voter registration lists for federal elections and new
voting systems, both of which are at different stages of implementation
across the states.

Chapter 1

Introduction

The basic goal of the elections system in the United States is
straightforward: All eligible persons, but only eligible persons, should
be able to cast their votes and, if such votes have been properly cast by
the voters, have those votes counted accurately. Faith in the fairness and
accuracy of the U.S. election system is at the foundation of our
democracy. Reports of problems encountered in the close 2000 presidential
election with respect to voter registration lists, absentee ballots,
ballot counting, and antiquated voting equipment raised concerns about the
fairness and accuracy of certain aspects of the U.S. election system.
After the events surrounding the November 2000 general election, the Help
America Vote Act of 2002 (HAVA) was enacted and major election reforms are
now being implemented.1 The November 2004 general election highlighted
some of the same challenges as 2000 as well as some new challenges in
areas such as electronic voting technology and implementation of some HAVA
requirements. The issues that arose in both elections highlighted the
importance of the effective interaction of people, processes, and
technology in ensuring effective election operations and maintaining
public confidence that our election system works.

Since 2001, GAO has issued a series of reports covering aspects of the
election process primarily with respect to federal elections.2 This report
focuses on the changing of such election processes in the United States
and the November 2004 general election. Specifically, primarily with
respect to federal elections, our objectives were to examine each major
stage of the election process to (1) identify changes to election systems
since the 2000 election, including steps taken to implement HAVA, and (2)
describe the issues and challenges encountered by election officials in
the November 2004 election.

1Pub. L. No. 107-252, 116 Stat. 1666 (2002).

2See appendix I for a list of these reports. One of the most comprehensive
is GAO, Elections: Perspectives on Activities and Challenges across the
Nation, GAO-02-3 (Washington, D.C.: Oct. 15, 2001).

  Page 33 GAO-06-450 Elections

                               Election Authority

                             Chapter 1 Introduction

Election authority is shared by federal, state, and local officials in the
United States. Congressional authority to affect the administration of
elections derives from various constitutional sources, depending upon the
type of election.3 Congress has passed legislation in several major areas
of the voting process. For example, the National Voter Registration Act of
1993 (NVRA),4 expanded the opportunities for citizens to register to vote
for federal elections by, among other things, requiring most states to
accept registration applications for federal elections by mail and at
state motor vehicle agencies (MVA) and at certain other state agencies.
The act also requires that in the administration of elections for federal
office, states are to take certain steps to accurately maintain voter
registration lists, and it limits the circumstances for removing names
from voter lists. The Uniformed and Overseas Citizens Absentee Voting Act
of 1986 (UOCAVA) requires states to, among other things, permit uniformed
services voters absent from the place of residence where they are
otherwise qualified to vote, their dependents, and U.S. citizens residing
outside the country to register and vote absentee in elections for federal
office.5

The Help America Vote Act was enacted into law on October 29, 2002. As
discussed below, the act includes a number of provisions related to voter
registration, provisional voting, absentee voting, voting equipment, and
other election administration provisions, and authorizes the appropriation
of funds to be used toward implementing the law's requirements. HAVA also
provides that the choices on the methods of implementation of such
requirements, for example, a computerized statewide voter registration
list, provisional voting, voter information requirements at the polling
place, identification requirements, and voting system standards (for
ballot verification, manual audit capacity, accessibility, and error
rates), are left to the discretion of the states. HAVA further specifies
that such requirements are minimum requirements and should not be
construed to prevent states from establishing election technology and
administration requirements that are stricter than HAVA requirements as
long as they are not inconsistent with certain other specified provisions.

3GAO, Elections: The Scope of Congressional Authority in Election
Administration,  GAO-01-470 (Washington, D.C.: Mar. 13, 2001). 442 U.S.C.
S:S: 1973gg to 1973gg-10. 542 U.S.C. S:S: 1973ff to 1973ff-6.

  Page 34 GAO-06-450 Elections

HAVA, in general, applies to all 50 states and the District of Columbia.
Areas covered by the law include

     o Computerized statewide voter registration list: HAVA requires most
       states to implement a single, uniform, centralized, computerized
       statewide voter registration list to serve as the official voter
       registration list for the conduct of all elections for federal office
       in each such state.6 Under HAVA, the computerized statewide voter
       registration list was to have been implemented by 2004. However,  40
       states and the District of Columbia received waivers to extend the
       deadline until January 1, 2006. States are required to perform regular
       maintenance of the voter list by comparing it to state records on
       felons and deaths, and to match voter registration applicant
       information on the voter list with information in the state motor
       vehicle agency's records and Social Security Administration records,
       as appropriate.
     o Absentee ballots: HAVA contains various amendments to UOCAVA regarding
       absentee voting for absent uniformed service voters and certain other
       civilian voters residing outside of the United States.  The
       amendments, among other things, (1) required that the secretaries of
       each military department, to the maximum extent practicable, provide
       notice to military personnel of absentee ballot deadlines, (2)
       extended the time that can be covered by a single absentee ballot
       application from UOCAVA voters, and (3) prohibited states from
       refusing to accept or process, with respect to federal elections, a
       voter registration application or an absentee ballot application by an
       absent uniformed services voter on the ground that the application was
       submitted before the first date that the state otherwise accepts or
       processes applications for that year from  nonuniformed service
       absentee voters.

6Under HAVA, states with no voter registration requirements for voters
with respect to federal elections on and after the date of HAVA's
enactment (e.g., North Dakota) are not subject to the act's computerized
statewide voter registration list requirement.

  Page 35 GAO-06-450 Elections

     o Provisional ballots: HAVA requires most states to implement
       provisional voting for elections for federal office.7 Under HAVA, in
       an election for federal office, states are to provide a provisional
       ballot to an individual asserting (1) to be registered in the
       jurisdiction for which he or she desires to vote and (2) eligible to
       vote in a federal election but (3) whose name does not appear on the
       official list of eligible voters for the polling place. Provisional
       ballots are also to be provided in elections for federal office to
       individuals who an election official asserts to be ineligible to vote,
       and for court-ordered voting in a federal election after the polls
       have closed. These various types of individuals, under HAVA, are to be
       permitted to cast the provisional ballot upon the execution of written
       affirmation at the polling place that they are registered voters in
       the jurisdiction and that they are eligible to vote in that election.
       If election officials determine that the individual is eligible under
       state law to vote, the individual's provisional ballot is to be
       counted as a vote in accordance with state law. HAVA also requires
       that a free access system be established to inform voters if their
       votes were counted, and if not, the reason why.
     o Polling places: HAVA provisions targeted, among other things,
       improving information at polling places and Election Day procedures.
       To improve the knowledge of voters regarding voting rights and
       procedures, HAVA requires election officials8 to post voting
       information at each polling place on the days of elections for federal
       office, including, for example, a sample ballot, polling place hours,
       how to vote, instructions for first-time voters who registered by
       mail, and general information on federal and state voting rights laws
       and laws prohibiting fraud and misrepresentation. The act also
       authorized the appropriation of funds for payments to states for
       educating voters concerning voting procedures, voting rights, and
       voting technology. Under HAVA, voting systems used in elections for
       federal office are required to meet specified accessibility
       requirements for individuals with disabilities. With respect to

7Under HAVA, states that had either (1) no voter registration requirements
for voters with respect to federal elections (North Dakota) or (2) polling
place registration on Election Day with respect to federal elections
(Idaho, Minnesota, New Hampshire, Wisconsin, and Wyoming) in effect on and
after August 1, 1994, are not subject to HAVA's provisional voting
requirements.

8Jurisdictions call their poll workers by different titles, including
clerks, wardens, election judges, inspectors, captains, and precinct
officers and often have a chief poll worker for each polling place.

improving accessibility, HAVA also authorized the appropriation of funds
for payments to states to be used for improved accessibility of polling
places for, among others, individuals with disabilities and those with
limited English proficiency. HAVA also requires that such voting systems
provide individuals with disabilities with the same opportunity for access
and participation (including privacy and independence) as for other
voters. In connection with this requirement, HAVA provides for the use of
at least one direct recording electronic (DRE) device or other voting
system equipped for individuals with disabilities at each polling place.9

     o Identification requirements: Under HAVA, states are to require that
       certain voters who register by mail to provide specified types of
       identification when voting at the polls or send a copy of the
       identification with their mailed applications.10 Acceptable
       identification includes a current and valid photo identification or
       current utility bill, bank statement, government check, paycheck, or
       other government document that shows the name and address of the
       voter. Under HAVA, voters at the polls who have not met the
       identification requirement may cast a vote under HAVA's provisional
       voting section. Similarly, mail-in ballots from persons who have not
       provided the required identification also are to be counted as HAVA
       provisional ballots.
     o Election administration: HAVA also established an agency with
       wide-ranging duties to help improve state and local administration of
       federal elections. The Election Assistance Commission is to be
       involved with, among other things, providing voluntary guidance to
       states implementing certain HAVA provisions, serving as a national
       clearinghouse and resource for information with respect to the

9There are two types of DREs, push-button, and touch screen. For
push-button machines, voters press a button next to the candidate's name
or ballot issue, which then lights up to indicate the selection.
Similarly, voters using touch screen DREs make their selections by
touching the screen next to the candidate or issue, which is then
highlighted. When voters are finished on a push-button or touch screen
DRE, they cast their votes by pressing a final "vote" button on the
machine or screen.

10Voters subject to this provision are those who have registered to vote
in a jurisdiction by mail and have not previously voted in an election for
federal office in the state or those who have not previously voted in such
an election in the jurisdiction and the jurisdiction is located in a state
that does not have a statewide computerized voter registration list, as
required by HAVA.

administration of federal elections, conducting studies, administering
programs that provide federal funds for states to make improvements to
some aspects of election administration, and helping to develop testing
for voting systems, and standards for election equipment. EAC is led by
four Commissioners, who are to be appointed by the President and confirmed
by the Senate. The Commissioners, who, under HAVA, were to be appointed by
February 26, 2003, were appointed by the President in October 2003 and
confirmed by the Senate in December 2003. Since beginning operations in
January 2004, EAC has achieved many of its objectives. Among other things,
EAC has held hearings on the security of voting technologies and the
national poll worker shortage; established a clearinghouse for information
on election administration by issuing two best practices reports;
distributed payments to states for election improvements, including
payments for voter education and voting equipment replacement; drafted
changes to existing federal voluntary standards for voting systems; and
established a program to accredit the national independent certified
laboratories that test electronic voting systems against the federal
voluntary standards. However, EAC has reported that its delayed start-up
affected its ability to conduct some HAVA-mandated activities within the
time frames specified in the act. In turn, according to its fiscal year
2004 annual report, the delayed EAC start-up affected states' procurement
of new voting equipment and the ability of some states and local
jurisdictions to meet related HAVA requirements by statutory deadlines.

o  Voting systems: One of the primary HAVA provisions relates to
encouraging states to replace punch card voting systems and lever voting
systems and authorizing appropriations for payments to support states in
making federally mandated improvements to their voting systems. A voting
system includes the people, processes, and technology associated with any
voting method. It encompasses the hardware and software used to define the
ballot, conduct the vote, and transmit and tally results, and system
maintenance and testing functions. With respect to standards for voting
systems used in elections for federal office, HAVA requirements for such
systems include providing voters with the ability to verify their votes
before casting their ballots, producing permanent paper records for manual
auditing of voting systems, and compliance of voting system ballot
counting error rates with those set out in specified federal voting system
standards. HAVA also directs that updates to the federal voluntary voting
system standards for these requirements be in place

        by January 1, 2004,11 and provides for additional updates to the

           voluntary standards as approved by the Election Assistance

         Commission. Mechanisms are also specified that can be used by

        states and localities in acquiring and operating voting systems,

       including accreditation of laboratories to independently test and

evaluate voting systems and federal certification for voting systems

that undergo independent testing.

The time frames for implementing various HAVA requirements ranged from as
early as 45 days after enactment (a deadline for establishing a grant
program for payment to the states for improved election administration) to
as late as January 1, 2006, for various voting system standards.12 Several
key deadlines were set for January 1, 2004, including implementation of
HAVA's provisional voting requirements and the establishment of a
statewide voter registration list (or to request a waiver from the
deadline until January 1, 2006). States receiving funds to replace punch
card voting systems or lever voting systems could also request a waiver
until January 1, 2006; otherwise such systems were to be replaced in time
for the November 2004 general elections. The deadline for states and
jurisdictions to comply with specific requirements for voting systems,
such as producing a paper record for audit purposes, was January 1, 2006.

HAVA vests enforcement authority with the Attorney General to bring a
civil action against any state or jurisdiction as may be necessary to
carry out specified uniform and nondiscriminatory election technology and
administration requirements under HAVA. These requirements pertain to HAVA
voting system standards, provisional voting and voting information
requirements, the computerized statewide voter registration list
requirements, and requirements for persons who register to vote by mail.
The enforcement of federal statutes pertaining to elections and voting
has, with certain exceptions, been delegated by the Attorney General to
the Civil Rights Division.

11The Election Assistance Commission approved the Voluntary Voting System
Guidelines in December, 2005.

12These HAVA voting system standards pertain to, among other things, voter
ballot verification prior to casting a vote, permanent paper records with
a manual audit capacity, federal standards for error rates, alternative
language accessibility, and accessibility for individuals with
disabilities. In addition, HAVA requires that voting systems purchased
with specified HAVA funds on or after January 1, 2007, must meet HAVA
voting system standards for disability access.

Election System Elements

The U.S. election system is highly decentralized and based upon a complex
interaction of people (election officials and voters), processes, and
technology. Each of the 50 states and the District of Columbia has its own
election system with a somewhat distinct approach. Within each of these 
51 systems, the guidelines and procedures established for local election
jurisdictions can be very general or specific. Each election system
generally incorporates elements that are designed to allow eligible
citizens to vote and ensures that votes are accurately counted. While
election systems vary from one local jurisdiction to another, most
election systems have the elements identified in figure 7.

Figure 7: The Election Process Involves the Integration of People, Process, and
                                   Technology

                                  Source: GAO.

Delegation of Election Responsibility

Typically, states have decentralized elections so that the details of
administering elections are determined at the local jurisdiction. States
can be divided into two groups according to how they delegate election
responsibilities to local jurisdictions. The first group include 41 states
where election responsibilities are delegated to counties, with a few of
these states delegating election responsibilities to some cities, and 1
state that delegates these responsibilities to election regions. We
included the District of Columbia along with this group. The second group
is composed of 9 states that delegate election responsibilities to
subcounty governmental units, known by the U.S. Census Bureau as minor
civil divisions (MCD). However, in 1 of these states, Minnesota, election
functions are split between county-level governments and MCDs. For
example, registration is handled exclusively by county officials, and
functions, such as polling place matters, are handled by MCDs. Overall,
about 10,500 local government jurisdictions are responsible for conducting
elections nationwide, with the first group of states containing about
onefourth of the local election jurisdictions and about three-fourths of
the local election jurisdictions located in the states delegating
responsibilities to MCDs. Although more election jurisdictions are in the
9 states, most of the population (88 percent of the U.S. population based
on the Census of 2000) lives in the states delegating responsibilities
primarily to counties.

                               Voter Registration

While voter registration is not a federal requirement, the District of
Columbia and all states, except North Dakota, generally require citizens
to register before voting.13 The deadline for registering, and what is
required to register, varies; at a minimum, state eligibility provisions
typically require a person to be a U.S. citizen, at least 18 years of age,
and a resident of the state, with some states requiring a minimum
residency period. Citizens apply to register to vote in various ways, such
as at motor vehicle agencies, during voter registration drives, by mail,
or at local voter registrar offices. Election officials process
registration applications and compile and maintain the list of registered
voters to be used throughout the administration of an election. Prior to
HAVA, voter registration lists were not necessarily centralized at the
state level, and separate lists were often

13Historically, Wisconsin has not required voters in all jurisdictions to
register to vote. Only municipalities with populations exceeding certain
specified thresholds were required to register voters. Changes made in
2003 to Wisconsin's election laws will require voter registration in every
municipality regardless of population size. This registration requirement
first applies to the 2006 spring primary election.

Page 41 GAO-06-450 Elections

                           Absentee and Early Voting

managed by local election officials. HAVA requires voter registration
information for federal elections to be maintained as a statewide
computerized list and matched with certain state data, and that voter
registration application information be matched with certain state data
and, in some cases, with federal data, to help ensure that the voter list
is accurate.

All states and the District of Columbia have provisions allowing voters to
cast their ballot before Election Day by voting absentee with variations
on who may vote absentee, whether the voter needs an excuse, and the time
frames for applying and submitting absentee ballots. In addition, some
states also allow early voting, in which the voter goes to a specific
location to vote in person prior to Election Day. As with absentee voting,
the specific circumstances for early voting-such as the dates, times, and
locations-are based on the state and local requirements. In general, early
voting allows voters from any precinct in the jurisdiction to cast their
vote before Election Day either at one specific location or at one of
several locations. The early voting locations are staffed by poll workers
who have a registration list for the jurisdiction and ballots specific to
each precinct. The voter is provided with and casts a ballot for his or
her assigned precinct.

Conducting Elections Election officials perform a broad range of
activities in preparation for and on Election Day itself. Prior to an
election, officials recruit and train poll workers to have the skills
needed to perform their Election Day duties, such as opening and closing
the polls, operating polling place equipment, and explaining and
implementing provisional voting procedures for certain voters such as
those who are not on the registration list. Where needed and required,
election officials must also recruit poll workers who speak languages
other than English. Polling places have to be identified as meeting basic
standards for accessibility and having an infrastructure to support voting
machines as well as voter and poll worker needs. Ballots are designed and
produced to meet state requirements, voter language needs, and identify
all races, candidates, and issues on which voters in each precinct in
their jurisdiction will vote. Election officials seek to educate voters on
topics such as what the ballot looks like, how to use a voting machine,
and where their particular polling place is located. Finally, election
officials seek to ensure that voting equipment, ballots, and supplies are
delivered to polling places.

                               Provisional Voting

On Election Day, poll workers set up and open the polling places. This can
include tasks such as setting up the voting machines or voting booths,
readying supplies, testing equipment, posting required signs and voter
education information, and completing paperwork such as confirming that
the ballot is correct for the precinct. Before a voter receives a ballot
or is directed to a voting machine, poll workers typically are to verify
his or her eligibility. The assistance provided to voters who are in the
wrong precinct depends on the practices for that particular location.

One of the most significant post-2000 election reforms found in HAVA,
according to the Election Assistance Commission, is that states are
required to permit individuals, under certain circumstances, to cast a
provisional ballot in federal elections.14 More specifically, states are
to provide a provisional ballot to an individual asserting to be (1)
registered in the jurisdiction for which he or she desires to vote and (2)
eligible to vote in a federal election, but (3) whose name does not appear
on the official list of eligible voters for the polling place. In
addition, provisional ballots are to be provided in elections for federal
office to individuals who an election official asserts to be ineligible to
vote, and for court-ordered voting in a federal election after the polls
have closed. Although many states had some form of provisional balloting
prior to the passage of HAVA, 44 of the  50 states and the District of
Columbia were required to provide provisional ballots for the 2004 general
election. Under HAVA, 6 states were exempt from HAVA's provisional voting
requirements because they either permitted voters to register on Election
Day or did not require voter registration.15  If individuals are
determined to be eligible voters, their provisional ballots are to be
counted as votes in accordance with state law, along with other types of
ballots, and included in the total election results.

14The United States Election Assistance Commission, 2004 Election Day
Survey; How We Voted: People, Ballots, and Polling Places. September 2005.

15Under HAVA, states that had either (1) no voter registration
requirements for voters with respect to federal elections (e.g., North
Dakota) or (2) polling place registration on Election Day with respect to
federal elections (as in Idaho, Minnesota, New Hampshire, Wisconsin, and
Wyoming) in effect on and after August 1, 1994, are not subject to HAVA's
provisional voting requirements.

Vote Counting and Recounting

Following the close of the polls, election officials and poll workers
complete a number of basic steps to get the votes counted and determine
the outcome of the election. Equipment and ballots are to be secured, and
votes are to be tallied or transferred to a central location for counting.
The processes used to count or to recount election votes vary with the
type of voting equipment used in a jurisdiction, state statutes, and local
jurisdiction policies. Votes from Election Day, absentee ballots, early
votes (where applicable), and provisional ballots are to be counted and
consolidated for each race to determine the outcome. While preliminary
results are available usually by the evening of Election Day, the
certified results are generally not available until days later. Some
states establish a deadline for certification of results, while other
states do not.

Voting Methods and Technologies

Voting methods are tools for accommodating the millions of voters in our
nation's approximately 10,000 local election jurisdictions. Since the
1980s, ballots in the United States have been cast and counted using five
methods: paper ballots, lever machines, punch cards, optical scan, and
DREs. Four of the five methods by which votes are cast and counted involve
technology; only the paper ballot system does not use technology. The
three newer methods-punch card, optical scan, and DRE-depend on computers
to tally votes. Punch card and optical scan methods rely on paper ballots
that are marked by the voter, while many DREs use computers to present the
ballot to the voter. Voting systems utilize technology in different ways
to implement these basic voting methods. For instance, some punch card
systems include the names of candidates and issues on the printed punch
card, while others use a booklet of candidates and issues that must be
physically aligned with the punch card. The way systems are designed,
developed, tested, installed, and operated can lead to a variety of
situations where misunderstanding, confusion, error, or deliberate actions
by voters or election workers can, in turn, affect the equipment's
performance in terms of accuracy, ease of use, security, reliability, and
efficiency. In fact, some recent election controversies have been
specifically associated with particular voting methods and systems.
Nevertheless, all voting methods and systems can benefit from established
information technology management practices that effectively integrate the
people, processes, technologies.

Scope and Methodology

For this report, we conducted a Web-based survey of election officials in
all 50 states and the District of Columbia, surveyed by mail a nationally
representative stratified random probability sample of 788 local election
jurisdictions, and conducted on-site interviews with election officials in
 28 local jurisdictions in 14 states. Copies of the survey instruments
are in appendixes II and III. In addition, the results of our state and
local surveys are presented in two supplemental GAO products that can be
found on our Web site at www.gao.gov.16 Appendix IV provides a summary of
jurisdictions we visited. In reporting the state survey data, actual
numbers of states are provided. When reporting local jurisdiction survey
data, we provide estimates for jurisdictions nationwide. Unless otherwise
noted, the maximum sampling error, with 95 percent confidence, for
estimates of all jurisdictions from our local jurisdiction survey is plus
or minus  5 percentage points (rounded).17 We also provide some national
estimates by jurisdiction population size, and the sampling errors for
these estimates are slightly higher. For these estimates, large
jurisdictions are defined as those with a population over 100,000, medium
jurisdictions have a population of over 10,000 to 100,000, and small
jurisdictions have a population of 10,000 or less. Unless otherwise noted,
all estimates from our local jurisdiction survey are within our planned
confidence intervals.

Jurisdictions in which we conducted on-site interviews were chosen based
on a wide variety of characteristics, including voting methods used,
geographic characteristics, and aspects of election administration, such
as whether early voting was offered. We did not select jurisdictions we
visited on the basis of size, but as appropriate, we identify the size of
a jurisdiction we visited using the same groupings we used for our
nationwide mail survey.

16GAO, Elections: 2005 Survey of State Election Officials, GAO-06-451SP
(Washington, D.C.: June 6, 2006); and GAO, Elections: 2005 Survey of Local
Election Jurisdictions,  GAO-06-452SP (Washington, D.C.: June 6, 2006).

17Measures of sampling error are defined by two elements: the width of the
confidence intervals around the estimate (sometimes called the precision
of the estimate) and the confidence level at which the intervals are
computed. Because we followed a probability procedure based on random
selections, our sample is only one of a large number of samples that we
might have drawn. As each sample could have provided different estimates,
we express our confidence in the precision of our particular sample
results as a 95-percent confidence interval (e.g., plus or minus 5
percentage points). This is the interval that would contain the actual
population value for 95 percent of the samples we could have drawn. As a
result, we are 95 percent confident that each of the confidence intervals
based on the mail survey includes the true values in the sample
population.

We also reviewed extensive prior GAO work and other national studies and
reports, and attended an annual election official conference. A
comprehensive description of our methodology for this report is contained
in appendix V.

We conducted our work between March 2005 and February 2006 in Washington,
D.C.; Dallas; Los Angeles; and 28 local election jurisdictions in 14
states, in accordance with generally accepted government auditing
standards.

Chapter 2

Voter Registration

In general, the goal of a voter registration system is to ensure that
eligible citizens who complete all the steps required of them to register
to vote in their jurisdictions are able to have their registrations
processed accurately and in a timely fashion, so they may be included on
the rolls in time for Election Day. The November 2000 general election
resulted in widespread concerns about voter registration in the United
States. Headlines and reports questioned the mechanics and effectiveness
of voter registration by highlighting accounts of individuals who thought
they were registered being turned away from polling places on Election
Day, the fraudulent use of the names of dead people to cast additional
votes, and jurisdictions incorrectly removing the names of eligible voters
from voter registration lists. With the passage of HAVA,1 with respect to
federal elections, most states were required to establish statewide
computerized voter registration lists and perform certain list maintenance
activities as a means to improve upon the accuracy of voter registration
lists.2 List maintenance is performed by election officials and consists
of updating registrants' information and deleting duplicate registrations
and the names of registrants who are no longer eligible to vote.

The voter registration process includes the integration of people,
processes, and technology involved in registering eligible voters and in
compiling and maintaining accurate and complete voter registration lists.
In managing the voter registration process and maintaining voter
registration lists, state and local election officials must balance two
goals- minimizing the burden on eligible persons registering to vote, and
ensuring that voter lists are accurate, that is, limited to those eligible
to vote and that eligible registered voters are not inadvertently removed
from the voter registration lists. This has been a challenging task, and
remains so, as we and others have noted. While registering to vote appears
to be a simple step in the election system generally, applying to register
and being registered are not synonymous, and election officials face
challenges in processing the voter registration applications they receive.
This chapter describes various HAVA and state changes related to the voter
registration processes that have occurred since the 2000 general election.
It also examines

1Pub. L. No. 107-252,116 Stat. 1666 (2002).

2HAVA section 303(a) (codified at 42 U.S.C.S: 15483(a)). North Dakota did
not have voter registration requirements for persons with respect to
federal elections as of the date of HAVA's enactment and, under HAVA, is
therefore not subject to the act's requirement to create and maintain a
computerized statewide voter registration list.

Page 47 GAO-06-450 Elections

continuing and new registration challenges encountered by local
jurisdictions for the 2004 general election.

With respect to voter registration, a significant change since the 2000

Overview

general election is the HAVA requirement for states to each establish a
single, uniform, statewide, computerized voter registration list for
conducting elections for federal office. The HAVA requirements for states
to develop statewide lists and verify voter information against state and
federal agency records presented a significant shift in voter list
management in many states. While the initial deadline to implement HAVA's
statewide list requirement was January 1, 2004, more than 40 states took
advantage of a waiver allowing an extra 2 years to complete the task, or
until January 1, 2006. The statewide registration lists for federal
elections are intended to implement a system capable of maintaining voter
registration lists that are more accurate by requiring states to (1) match
voter registration application information against other state and federal
agency databases or records to help ensure that only eligible voters are
added to such lists, (2) identify certain types of ineligible voters whose
names should be removed from the lists, and (3) identify individual voter
names that appear more than once on the list to be removed from the lists.
While HAVA defined some parameters for the required statewide voter
registration lists and required matching voter information with certain
state and federal records, the act leaves the choices on the methods of
implementing such statewide list requirement to the discretion of the
states. On the basis of our survey of state election officials, states
varied in the progress made in implementing their statewide voter
registrations lists, how they have implemented these systems, and the
capabilities of their systems to match information with other state and
federal agency records as well as many other features of the state
systems.

In addition to requiring states to develop statewide voter registration
lists, HAVA provides that states must require that mail registrants who
have not previously voted in a federal election in the state are to
provide certain specified types of identification with their mail
application, and if they do not provide such identification with their
application, these first-time mail registrants are to provide the
identification at the polls. Furthermore, if such a voter does not have
the requisite identification at the polls, HAVA requires that the voter be
provided a provisional ballot with the status of his or her ballot to be
determined by the appropriate state or local official. As with the
statewide voter registration list requirement, HAVA leaves the choices on
the methods of implementing the provisional voting requirement to the
discretion of the states. On the basis of interviews of officials in 28
local election jurisdictions, implementation of the requirement for
first-time voters who registered by mail varied. One noteworthy variation
is in the definition of mail registration, where some local jurisdictions
we visited told us that applications received through voter registration
drives would be treated as mail registrations subject to HAVA
identification requirements and other local jurisdictions we visited told
us applications from registration drives were not treated as mail
registrations and therefore were not treated as subject to HAVA
identification requirements.

As noted above, during 2004 and 2005 many states were in the process of
implementing their HAVA-required statewide voter registration lists and
associated requirements for maintaining the lists. Thus, the potential
benefits to be gained from HAVA's requirement for the statewide voter
registration lists were not evident in many states at the time of the
November 2004 general election. Maintenance requirements in HAVA intended
to help states and local election jurisdictions to have access to more
accurate voter registration list information, such as identifying
duplicate registrations and matching the voter information against other
state agency databases or records, were not yet fully implemented by many
states. Many local jurisdictions were not yet seeing the benefits of being
able to verify voter registration application information with state motor
vehicle agency databases to identify eligible voters, or to match voter
registration lists with state vital statistics agency records to identify
deceased persons, and to appropriate state agency's records to identify
felons who may be ineligible to vote. Thus, on the basis of our nationwide
survey and local election jurisdictions we visited, many local
jurisdictions continued to encounter challenges with the voter
registration lists that they had experienced in the 2000 general election,
such as difficulties related to receiving inaccurate and incomplete voter
registration information, multiple registrations, or ineligible voters
appearing on the list. In addition, election officials in some
jurisdictions we visited told us they continued to face challenges
obtaining voter registration applications from motor vehicle agencies and
other NVRA entities.3

3In 1993, the National Voter Registration Act (42 U.S.C. S:S: 1973gg to
1973gg-10), among other things, expanded the number of locations and
opportunities of citizens to apply to register to vote in federal
elections at MVAs and other public organizations, such as public
assistance agencies and armed forces recruiting centers.

  Page 49 GAO-06-450 Elections

Changes Required by HAVA Subsequent to the 2000 General Election

In addition, for some local election jurisdictions we visited, election
officials told us that efforts on the part of various groups to get out
the vote by registering new voters through voter registration drives
created new challenges not identified to us as a problem in the 2000
general election. Specifically, at some local jurisdictions we visited,
election officials told us they faced a challenge processing large volumes
of voter registration applications just prior to the deadlines for
registration, which included challenges in some large jurisdictions to
resolve issues of incomplete or inaccurate (and potentially fraudulent)
applications submitted by entities conducting voter registration drives.

HAVA requires states to, among other things, (a) implement a single,
uniform, computerized statewide voter registration list for conducting
elections for federal office;4 (b) perform regular maintenance by
comparing the voter list against state records on felons and deaths; (c)
verify information on voter registration applications with information in
state motor vehicle agency databases or with a Social Security
Administration database, as appropriate. In addition, HAVA imposes new
identification requirements for certain mail registrants-such as,
individuals who register by mail and have not previously voted in a
federal election within the state.5

4This HAVA-required statewide voter registration list is, among other
things, to be defined, maintained, and administered at the state level. In
addition, this list is to serve as a secure, centralized, and interactive
database that is coordinated with other state agency databases and grants
state and local election official immediate electronic access to
information contained in the list.

5These HAVA identification requirements also apply to individuals who
register to vote in a jurisdiction by mail or have not previously voted in
a jurisdiction when the jurisdiction is located in a state that does not
have a HAVA compliant statewide voter registration list.

HAVA Requirements for Statewide Voter Registration List

Historically, to ensure that only qualified persons vote, states and local
jurisdictions have used various means to establish and compile voter
registration lists.6 Prior to HAVA, we noted in our October 2001
comprehensive report on election processes nationwide that in compiling
these lists, election officials used different methods to verify the
information on registration forms, check for duplicate registrations, and
update registration records, and we noted that states' capabilities for
compiling these lists varied. At the time, some states had statewide voter
lists, but others did not and were not required to do so. Moreover, most
jurisdictions we visited at the time maintained their own local,
computerized voter lists.7 Under HAVA, this has changed. HAVA requires the
chief election official in the state to implement a "single, uniform,
official, centralized, interactive, computerized statewide voter
registration list" that must contain the name and registration information
of every legally registered voter in the state. Under HAVA, states were
required to be in compliance with the statewide voter registration list
requirement by January 2004 unless they obtained a waiver until January
2006. Forty-one states and the District of Columbia obtained a waiver and
thus, for the 2004 general election, were not required to have their
statewide voter registration lists in place.8

With respect to the HAVA required statewide voter registration list,
states are to, among other things:

     o Make the information in such lists electronically accessible to any
       election officials in the state.
     o Ensure that such voter lists contain registration information on every
       legally registered voter in the state, with a unique identifier
       assigned to each legally registered voter.

6Throughout this chapter, the use of the term "voter registration list"
refers specifically to the names of registered voters compiled by state
and local officials in accordance with HAVA provisions. The term "voter
registration system" refers more broadly to computer systems (stand-alone
or Internet- or network-based) where voter registration lists reside in
accordance with appropriate security and privacy measures.

7GAO-02-3.

8Nine states (Alaska, Arizona, Georgia, Hawaii, Kentucky, Minnesota, South
Carolina, South Dakota, and West Virginia) did not seek a waiver.

     o Verify voter identity; most states are required to match voter
       information obtained on the voter registration application for the
       applicant's drivers' license number or the last four digits of the
       voter's Social Security number, when available, to state MVAs or the
       Social Security Administration databases. In connection with this
       requirement to verify voter registration application information,
       states must require that individuals applying to register to vote
       provide a current and valid driver's license number, or the last four
       digits of their Social Security number; if neither has been issued to
       the individual, then the state is to assign a unique identifier to the
       applicant. The state MVA must enter into an agreement with the Social
       Security Administration (SSA), as applicable, to verify the applicant
       information when the last four digits of the Social Security number
       are provided, rather than a driver's license number or state ID
       number.9
     o Perform list maintenance on the statewide voter registration lists by
       coordinating them on a regular basis with state records on felony
       status and deaths, in order to identify and remove names of ineligible
       voters.10 List maintenance is also to be conducted to eliminate
       duplicate names.
     o Implement safeguards ensuring that eligible voters are not
       inadvertently removed from statewide lists.

9Seven states (Georgia, Hawaii, Kentucky, New Mexico, South Carolina,
Tennessee, and Virginia) required, at the time of our review, full Social
Security numbers on applications for voter registration. HAVA provides
that for states using full Social Security numbers on applications in
accordance with section 7 of the Privacy Act of 1974, the HAVA voter
registration verification requirements are optional. Georgia's right to
require full Social Security numbers under this Privacy Act provision,
however, has been the subject of recent litigation. In January 2005 a
federal district court found that Georgia did not qualify to use full
Social Security numbers under this Privacy Act provision (Schwier v. Cox,
Civil No. 1:00-CV-2820, (N.D. Ga. January 31, 2005)). The district court
determined, in part, that Georgia must correspondingly revise its voter
registration forms and instructions and expressly inform applicants that
they are not required to provide their Social Security numbers. On
February 16, 2006, the U.S. Court of Appeals for the Eleventh Circuit
affirmed the district court's January 2005 ruling (Schwier v. Cox, 439 F.
3d 1285 (2006)).

10In 1993, the National Voter Registration Act, among other things,
limited the circumstances under which states could remove the names of
registrants from registration lists for federal elections and required
states to take certain steps to accurately maintain such voter
registration lists by removing the names of certain types of ineligible
persons.

Progress Establishing Statewide Voter Registration Lists Varied

o  Include technological security measures as part of the statewide list
to prevent unauthorized access to such lists.

Except for the 9 states that did not obtain a waiver from HAVA's
requirements for establishing a statewide voter registration list, all
other states subject to the statewide list requirement were not required
to perform list maintenance activities as defined in HAVA until the
extended waiver deadline of January 2006. By the November 2004 general
election, states were in various stages of implementing provisions of HAVA
related to their statewide voter registration lists and performing voter
list verification and maintenance, and had different capabilities and
procedures at the state and local levels for performing required list
maintenance functions. Many states reported that their statewide voter
registration systems implementing the statewide list requirement include
or will include additional election management features not required under
HAVA.

Voter registration system development was an ongoing process in 2004 and
2005. For the November 2004 general election, the use of technology to
compile voter registration information remained an issue. Developing and
implementing statewide computerized voter lists has been an ongoing
process for many states, and state and local election officials reported
encountering difficulties along the way. Our state survey and site visits
suggest that states and jurisdictions were still coming to terms, as of
the last half of calendar year 2005, with how their systems should be
updated and whether states or jurisdictions should control the flow of
information into statewide registration systems.

As mentioned in chapter 1, HAVA vests the Attorney General with the
responsibility of enforcing certain HAVA requirements with respect to the
states. In January 2006, the Justice Department asked all states, the
District of Columbia, and other covered territories to provide a detailed
statement of their compliance with voting systems standards and
implementation of a single, uniform, official, centralized, interactive
computerized statewide voter registration list. If the states, the
District of Columbia, or covered territories were not implementing HAVA's
requirements for the computerized statewide voter registration lists as of
January 2006, the Justice Department reported that it then asked them to
identify steps they planned to take to achieve full implementation of the
HAVA-compliant statewide voter registration list and the date on which
each step would be accomplished. According to Justice Department
officials, they are reviewing the information provided by the states, the
District of Columbia, and such territories to make determinations of what,
if any, enforcement action might be needed. The Department of Justice
reports that it entered into a memorandum of agreement with California in
November 2005 after that state realized it would not be able to fully meet
HAVA's requirements by the January 1, 2006, deadline. On March 1, 2006,
the Department of Justice also filed suit in a federal district court
against the state of New York alleging the state not to be in compliance
with, among other things, HAVA's requirement for a computerized statewide
voter registration list and seeking a judicial determination of
noncompliance and a court order requiring the state to develop a plan for
how it will come into compliance.11

During our site visits in 2005, we asked local election officials about
the status of their statewide registration systems. Election officials in
some local jurisdictions we visited cited difficulties related to
implementing their statewide voter registration systems involving, among
other things, internal politics and technology-related challenges. For
example, election officials in a large jurisdiction reported that a
disagreement between the State Board of Elections and local election
officials over the type of system to implement delayed the project for a
year. State election officials wanted a system requiring all voter
registrations to be entered at the state level but maintained locally. The
local election officials expressed the view that such a system would
result in a lack of control over data entry at the local level at the
front end, while imposing accountability on them on the back end (data
maintenance). During our interview in August 2005, these election
officials told us that a statewide registration system had not been
implemented yet. In some jurisdictions, the difficulties cited by election
officials may have reflected the fact that they were establishing
statewide voter registration systems for the first time. For example, in 1
large jurisdiction that was establishing a HAVA voter registration list
from scratch, local election officials noted that at the time of our
interview in August, the system was behind schedule, lacked the ability to
identify duplicates, had no quality control, and was not planned to
function as a real-time system.

11United States v. New York State Board of Elections, Civil No. 06-CV-0263
(N.D. NY, March 1, 2006).

Page 54 GAO-06-450 Elections

State Capabilities for Matching  In our survey of states and the District  
Voter Registration Lists with    of Columbia, and our survey of local      
State and Federal Records, as    election jurisdictions nationwide, among  
                                    other things, we inquired about the       
                                    status of their capabilities for meeting  
                                    HAVA provisions for (1) verifying         
Required by HAVA, Were Under Way voter registration application            
or Not Yet Achieved              information against MVA and SSA databases 
                                    and (2) maintaining the statewide voter   
                                    lists by comparing information on the     
                                    statewide voter registration list against 
                                    state death records and felon             
                                    information, and discussed the issues     
                                    during our local site visits. Our work    
                                    focused on how states had matched or      
                                    planned to match voter registration       
                                    lists against other state records, as     
                                    required by HAVA. However, it is          
                                    important to note that the success of     
                                    such matching in ensuring accurate        
                                    voter registration lists is dependent     
                                    upon the accuracy and reliability of the  
                                    data in the databases used for matching.  
                                    If that state's MVA databases, felon      
                                    records, death records, or other records  
                                    used for matching are inaccurate,         
                                    they can result in voter registration     
                                    list errors.                              
                                    Matching to MVA Databases                 
                                    When a driver's license or driver's       
                                    license number is presented as            
                                    identification when registering to vote   
                                    in an election for federal office,        
                                    HAVA requires that states match the voter 
                                    registration application information      
                                    presented with that in the MVA records.12 
                                    In our survey of state                    
                                    election officials, we asked states       
                                    whether their voter registration systems  
                                    would have the capability to perform      
                                    electronic matching of such voter         
                                    registration information with state motor 
                                    vehicle agency records for the            
                                    purposes of verifying the accuracy of     
                                    information on the registration           
                                    application. Twenty-seven states reported 
                                    they will have or currently had           
                                    the capability to match on a real-time    
                                    basis, 15 states and the District of      
                                    Columbia reported they will have or       
                                    currently had capability to match in      
                                    batches, and 4 states reported they would 
                                    not have the capability to                
                                    perform electronic matching. The          
                                    remaining 4 states included 2 states that 
                                    reported that they are not subject to     
                                    HAVA's registration information           
                                    verification requirement because they     
                                    collect the full Social Security numbers  
                                    on voter registration applications;13 1   
                                    state, North Dakota, which                

12HAVA section 303(a)(5) (codified at 42 U.S.C.S: 15483(a)(5)).

13The authority of one of these states (Georgia) to require full Social
Security numbers has been the subject of recent litigation. In February
2006 a federal appeals court affirmed a January 2005 federal district
court ruling that Georgia did not have the authority, under the Privacy
Act of 1974, to require voter applicants to disclose their Social Security
numbers on voter registration forms. Schwier v. Cox, 439 F. 3d 1285 (11th
Cir. 2006).

Page 55 GAO-06-450 Elections

does not require voter registration, did not respond, and 1 state reported
that it was uncertain of its capability to perform electronic matching.

  Matching with SSA Records

With respect to matching voter information with SSA data when a Social
Security number is presented instead of a driver's license, in our state
survey, 7 states had and 26 states and the District of Columbia reported
that they would have the capability, by January 1, 2006, to electronically
match voter registration information with SSA (through the MVA); 10 states
reported they planned to have this capability in place but not by January
2006; and 6 states had not yet determined whether they could do so. Many
states reported concerns with whether SSA would be able to return
responses to verify requests in a timely manner. Specifically, 30 states
and the District of Columbia reported some level of concern about the
issue. When asked whether they thought local jurisdictions would be able
to resolve nonmatches resulting from SSA verification checks, opinions
were divided, with a number of states (21) expressing some degree of
concern about this, while a nearly equal number (22 states and the
District of Columbia) did not.

In our June 2005 report on maintaining accurate voter registration lists,
we found that in one state (Iowa) that had verified its voter registration
list with SSA before the 2004 general election, there was no unique match
for 2,586 names, according to the SSA records.14 As we stated in our
report, Iowa officials said that the biggest problem they faced was that
SSA did not specify what specific voter information did not match (i.e.,
was the mismatch in name, date of birth, or final four-digit Social
Security number). Without that information, they were not able to
efficiently resolve the nonmatching problems. In that same report, we also
noted that an SSA official said that the system established to perform the
HAVA matching on the fourdigit Social Security number is not able to
provide that detail. In addition, we found that use of SSA's database to
identify deceased registrants, which is linked with the system established
to perform the HAVA verification of voter registration application
information, had matching and timeliness issues.

14GAO, Elections: Additional Data Could Help State and Local Elections
Officials Maintain Accurate Voter Registration Lists, GAO-05-478
(Washington, D.C.: June 2005).

Page 56 GAO-06-450 Elections

  Overall Matching Challenges

As shown in figure 8, many states reported that they faced significant
challenges when trying to match voter registration information with state
records. For example, in our survey, 29 states and the District of
Columbia reported that records with incomplete data posed a challenge; 19
states and the District of Columbia reported that obtaining records not
maintained electronically was a challenge; and 23 states reported that
verifying information against incompatible electronic record systems was
also a challenge.

Figure 8: Many States Report Challenges Matching Voter Registration
Information with State Records

Number of states

30

                 hingMatcds withrecorincomplete dataaonicallya

                       gainst incompatibleormationsystems

ds not

maintained electrObtaining recor

                                  erifying inf

                                  onic record

V

                                       a

                                     electr

Source: GAO 2005 survey of state election officials.

aIncludes the District of Columbia.

During our site visits to local jurisdictions, we obtained additional
views on how well, in general, states were believed to perform various
data-

Page 57 GAO-06-450 Elections

Removing Voters Names from the Registration List

matching functions.15 We asked local election officials to describe their
state system's ability to match voter registration information with MVA
and SSA records and the system's ability to verify information on
eligibility status for felons, noncitizens, and others with other state
databases or records. One jurisdiction in Illinois reported it was not
sure how or if its voter registration system would be able to match data
with MVA and SSA databases or to verify eligibility status for felons and
by age. An official in a jurisdiction in Florida said that Florida's
system could not verify information on the eligibility status of felons,
noncitizens, the mentally incompetent, or the underaged-though plans were
under way to obtain information from the Clerk of Courts Information
System to perform some of these tasks.

HAVA's list maintenance provisions require states to match the statewide
voter registration list information against certain state records to
identify ineligible voters and duplicate names.16 If a voter is ineligible
under state requirements and is to be removed from the statewide voter
registration list, states are generally required to remove such names in
accordance with NVRA provisions relating to the removal of voter names
from registration lists for federal elections. Under NVRA, in the
administration of voter registration for federal elections, states may not
remove the names of people who are registered to vote for nonvoting and
names may be removed only for certain specified reasons: at the request of
the registrant; by reason of criminal conviction, as provided by state
law; by reason of mental incapacity, as provided by state law; or pursuant
to a general program that makes a reasonable effort to remove the names of
ineligible voters from the official lists by reason of the death of the
voter or on the ground that the voter has changed address to a location
outside the election jurisdiction on the basis of change of address
information from the

U.S. Postal Service (but only if either (1) the voter confirms in writing
a change of address to a place outside the election jurisdiction or (2)
the voter has failed to respond to a confirmation mailing and the voter
has not voted or appeared to vote in any election between the time of such
notice and the passage of two federal general elections).17

15HAVA's list maintenance provisions require states to compare the
statewide registration list with state records on felons and deaths to
identify ineligible voters as well as to identify duplicate registrations.

16HAVA section 303(a)(2) (codified at 42 U.S.C.S: 15483).

1742 U.S.C. S: 1973gg-6.

Page 58 GAO-06-450 Elections

  Reasons Names Removed from Registration Lists

In our survey of local election jurisdictions nationwide, we asked about
the reasons names were removed from voter registration lists. On the basis
of our survey of local election jurisdictions, the following table shows
various reasons that jurisdictions removed names from voter registration
lists for the 2004 general election and our estimates of how frequently
names were removed for that reason. For example, the most frequent reason
was the death of the voter (76 percent). Names were removed with about
equal frequency because the voter requested that his or her name be
removed (54 percent) or the registrant's name appeared to be a duplicate
(52 percent). The least frequent reason was for mental incompetency (10
percent). In many jurisdictions, names were not removed but rather placed
on an inactive list for a period of time. In our survey of local
jurisdictions, nearly half, or an estimated 46 percent, took this step.

Table 1: Estimated Percentage of Jurisdictions That Removed Names from
Voter Registration List for 2004 General Election for Various Reasons

                          Page 59 GAO-06-450 Elections

                                                            Percentage of all
Name removed because:                                        jurisdictions 
Information received from state/county vital statistics                 76 
offices                                                      
identified registrants as deceased                           
Registrants requested that their names be removed from the              54 
voter registration list (e.g., moved out of jurisdiction or  
other                                                        
reason)                                                      
Registrant's name appeared to be a duplicate                            52 
Change of address information received from U.S. Postal                 45 
Service                                                      
showed that the registrants had moved outside of the         
jurisdiction                                                 
where registered                                             
Registrant failed to respond to a notice from the registrar             38 
and had                                                      
not voted or had not appeared to vote in the most recent two 
federal elections                                            
Felony records received from federal/state/local                        38 
governmental                                                 
entities identified registrant as ineligible to vote or      
register to vote                                             
because of a felony conviction                               
Newspaper obituaries identified registrant as deceased                  31 
Names removed from the voter registration list for other                11 
reasons                                                      
Information received from federal/state/local courts                    10 
indicating that                                              
registrant had been judged to be mentally incompetent        
Source: GAO 2005 survey of local election jurisdictions.     

In our June 2005 report on maintaining accurate voter registration
lists,18 on the basis of interviews of election officials in 14
jurisdictions and 7 state election offices, we reported that in larger
jurisdictions, the task of identifying and removing registrants who died
can be substantial. For example, in the city of Los Angeles, in 1 week in
2005 alone, almost 300 persons died.

The issue of felons voting unlawfully-that is, voting when their felony
status renders them ineligible to voter under state law-was a high-profile
issue in some jurisdictions. According to an election official in a
Washington jurisdiction we visited, this issue was identified during the
November 2004 general election. This official also told us that the
Secretary of State is working to establish a database that will indicate
felony status and cancel the registration of felons. This election
official noted that the jurisdiction rarely receives information from
federal courts on felony convictions. Under federal law, U.S. Attorneys
are to give written notice of felony convictions in federal district
courts to the chief state election official of the offender's state of
residence.19 In our June 2005 report on maintaining accurate voter
registration lists, we found that U.S. Attorneys had not consistently
provided this information, and while the law did not establish a
standardized time frame or format for forwarding the federal felony
conviction information, election officials in 7 states we visited reported
that the felony information received from U.S. Attorneys was not always
timely and was sometimes difficult to interpret.20 We recommended that the
U.S. Attorneys provide information in a more standardized manner.

  Removing Duplicate Names

Under HAVA, duplicate names on the statewide voter registration list are
also to be identified and removed. In our state survey, 49 states and the
District of Columbia reported that their voter registration systems will
include a function for checking duplicate voter registration records.21 On
the basis of our nationwide survey of local jurisdictions, we estimate
that

18GAO-05-478. 1942 U.S.C. S: 1973gg-6(g). 20GAO-05-478. 21One state
responded that it did not know whether its system would include checking
for

duplicates.

Page 60 GAO-06-450 Elections

72 percent of local jurisdictions employed a system of edit checks
(automated controls to identify registration problems) to identify
duplicates. Our prior work has also found that states were, for the most
part, able to handle duplicate registrations-though obtaining timely,
accurate data to facilitate the identification of duplicate registrations
has been viewed as a challenge among some state election officials.
Specifically, in our February 2006 report on certain states' (9 states
that did not seek a waiver until January 1, 2006 and were to implement a
computerized statewide voter registration list by January 1, 2004)
experiences with implementing HAVA's statewide voter registration lists,
we found that 8 of the 9 states we reviewed screened voter applications to
identify duplicate registrations, and most did so in real time.22 We also
reported that 8 of these 9 states checked voter registration lists for
duplicate registrations on an annual, monthly, or other periodic basis.
And 4 of the 9 states reported that implementing the HAVA requirements led
to some or great improvement in the accuracy of their voter lists by
reducing duplicate registrations or improving the quality of voter
information before it was entered into the statewide voter list.

Checking for duplicates remained a challenge for some in 2004 and 2005,
however. In our June 2005 report on maintaining accurate voter
registration lists, we noted that officials in 7 of the 21 local election
jurisdictions we spoke with during 2004 and 2005 had some concern about
the accuracy and timeliness of data they received to identify duplicate
registrants and verify that registrants resided within the jurisdiction.23
 They noted that the matching and validation of names are complex and
made more so when considering aliases and name changes, as are matches
such as "Margie L. Smith" with "Margaret Smith." Officials from several
states who reported, at the time of our review, that their state had not
implemented a statewide voter registration system noted that there was no
way to identify duplicates outside their jurisdiction.

22These are the 9 states that implemented the HAVA requirement to have
statewide voter registration lists in place by the earlier deadline of
January 1, 2004.

23GAO-05-478.

Page 61 GAO-06-450 Elections

Most States Reported Having Established Centralized Voter Registration
Systems, and Half Reported They Can Enter Voter Information on a Real-Time
Basis

While HAVA requires that both state and local election officials have
immediate electronic access to information in the statewide voter
registration list, HAVA grants states discretion as to the method used to
ensure that this capability is established.24 According to EAC, state and
local election officials may determine whether to establish (a) a top-down
system, whereby the statewide voter registration list resides on a state
database hosted on a single, central platform (e.g., a mainframe or client
servers), which state and local election officials may query directly; 

(b) a bottom-up system, whereby the statewide voter list is stored on a
state-level database that can be downloaded to jurisdictions and updated
by the state only when the jurisdictions send new registration information
back to the state;25 or (c) take another approach. According to the EAC
voluntary guidance on HAVA's statewide voter registration system, the 
top-down approach most closely matches HAVA requirements-but other
configurations may be used as long as they meet the HAVA requirement  for
a single, uniform list that allows election officials to have  immediate
access.

Our 2005 survey of state election officials sought information on how
states were implementing statewide computerized voter registration
systems. We asked, among other things, whether states were using a
top-down or a bottom-up approach. In response, 40 states and the District
of Columbia reported that they have a database maintained by the state,
with information supplied by local jurisdictions (top-down system); 4
states reported that local jurisdictions retain their own lists and
transmit information to a statewide list (a bottom-up system); and 5
states reported they use a hybrid of these two options. We also asked
whether state election officials would have immediate, real-time access to
their state lists for the purposes of entering new voter registration
information, updating existing information, and querying voter
registration records. About half the states and the District of Columbia
reported they had or would have all these capabilities. Specifically, 24
states and the District of Columbia reported they had or would have as of
January 2006, real-time access for entering new voter registration
information, while 23 states reported they did not plan to do so and 2
states did not respond. In addition, 26 states and

24HAVA section 305 (codified at 42 U.S.C.S: 15485).

25According to EAC, a bottom-up system remains static until the state
electronically provides the next updated version; registration information
held solely in a local database is not part of the official registration
list until it is electronically transmitted to the state and added to the
list.

Page 62 GAO-06-450 Elections

the District of Columbia reported that they had or would have as of
January 2006, real-time access for updating existing voter registration
information, while 21 states reported they did not plan to do so and 2
states did not respond. And 47 states and the District of Columbia
reported they had or would have as of January 2006 real-time access for
querying all state voter registration records, while 1 state reported it
would not do so and 1 state did not respond. For each of these questions,
one state reported it too would have these capabilities, but not by the
January 1, 2006, HAVA deadline.

We also sought state election officials' views on whether election
officials in local jurisdictions would have immediate, real-time access to
voter list information for the same three purposes stated above: entering
new information, updating existing information, and querying records. In
our state survey, most states and the District of Columbia reported that
local jurisdictions had these capabilities. Specifically, 46 states and
the District of Columbia reported that local jurisdictions had or would
have as of January 2006, real-time access for entering new voter
registration information, and 3 other states reported that they planned to
do so as well, but not by January 1, 2006. Also, 46 states and the
District of Columbia reported that local jurisdictions had or would have
as of January 2006,  real-time access for updating existing voter
registration information, and  3 other states planned to do so as well,
but not by the deadline. Finally,  47 states and the District of Columbia
reported local jurisdictions had or would have as of January 2006 the
capability to query records for their jurisdictions in real time, and 2
states planned to do so, but not by January 2006. Figure 9 compares the
capability of state and local jurisdiction election officials to access
the voter registration lists to perform certain tasks.

Most States Reported They Will Match Statewide Voter Registration List
Information Electronically against State Databases

Figure 9: States Reporting That They Had or Would Have as of January 2006,
Capability for Real-Time Access to Voter Registration List to Enter,
Update, and Query Information

Number of states and District of Columbia

50 48 48

47 47

40

30

27 25

20

10

0 Access for Access for Access for entering updating querying information
information information

States and District of Columbia Local jurisdictions

Source: GAO 2005 survey of state election officials.

While HAVA's list maintenance provisions require states to coordinate
statewide voter registration list information with certain other state
records within their state in order to identify and remove ineligible
names, the act does not specifically provide that such coordination must
be done electronically. However, to determine whether state systems had or
would have the capability to perform electronic data matching, our survey
asked states about existing or planned electronic capability. As shown in
figure 10, more than half the states reported that they had, or planned to
have, the ability to match voter registration information electronically
with state records on felony convictions and deceased registrants.
Specifically,  25 states reported they had and 15 states reported that
they would have the capability to electronically match against state death
records as of January 2006, and 6 states and the District of Columbia
planned to have the capability, but not by January 2006. Three states
reported that they did not plan to have this capability. With respect to
identifying ineligible felons,  16 states reported they had, and 15
reported they would have the capability to electronically match against
felony conviction records as of January 2006, while 9 states planned to do
so but would not have done so by January 2006. In addition, 7 states and
the District of Columbia did not plan to have this capability, and 2
states had not determined whether to have the capability.

Figure 10: States' Reported Capabilities to Electronically Match Voter
Registration In f ormation with State Death Record s and Felon y Con
viction Record s

Number of states 50

40

40

31

30

20

10

98

7

3

2

0

ould have

                                 No capabiltiya

2006

or capability

                                     2006a

                                   ould have

                               No capabiltiy 2006

                               or capability 2006

                                   Don't know

Jan.

                                      Jan.

                                   after Jan.

after Jan.

Had or wcapability by

                             Had or wcapabiltiy by

Planned f

                                   Planned f

Matching state death records Matching state felony conviction records
Source: GAO 2005 survey of state election officials. aIncludes District of
Columbia.

On the topic of states' efforts to meet HAVA's data-matching requirements
electronically-as opposed to transmitting paper records-EAC

Security of Voter Information in the Statewide Voter Registration Lists

recommends that voter registration information be transmitted
electronically, particularly between states and their MVAs. EAC further
recommends that to the extent allowed by state law and available
technologies, the electronic transfer between statewide voter registration
lists and coordinating verification databases should be accomplished
through direct, secure, interactive, and integrated connections. While EAC
provided guidance to states for their statewide systems, under HAVA, the
states are to define the parameters for implementing interactive and
integrated systems.

HAVA requires election officials to provide adequate technological
database security for statewide voter registration lists that is designed
to prevent unauthorized access.26 EAC provided states with voluntary
guidance, issued in July 2005, to help clarify HAVA's provisions for
computerized statewide voter registration lists. Among other things, the
EAC guidance noted that such computer security must be designed to prevent
unauthorized users from altering the list or accessing private or
otherwise protected information contained on the list. Access may be
controlled through a variety of tools, including network- or system-level
utilities and database applications (such as passwords and "masked" data
elements). Special care must be taken to ensure that voter registration
databases are protected when linked to outside systems for the purposes of
coordination. Any major compromise of the voter registration system could
lead to considerable election fraud.27

We sought information on what documented standards or guidance for
computer and procedural controls would be in place to prevent unauthorized
access to the lists. In our state survey, 45 states and the District of
Columbia reported having such standards or guidance, 3 plan to do so, and
1 reported that it did not know. We also asked states what actions they
had taken or planned to take to deal with privacy and intrusion issues. We
asked, for instance, what, if anything, had been done to install or
activate mechanisms to detect or track unauthorized actions affecting the
state's computerized voter registration system. A majority of states

26HAVA section 303(a)(3) (codified at 42 U.S.C.S: 15483(a)(3)).

27Election fraud includes conduct that corrupts the electoral process for
(1) registering voters; (2) obtaining, marking, or tabulating ballots; or
(3) canvassing and certifying election results. Types of fraudulent
conduct may include, among other things, voting by ineligibles, voting
more than once, voter impersonation, intentional disruption of polling
places either physically or by corrupting tabulating software, or
destroying ballots or voter registrations.

Page 66 GAO-06-450 Elections

reported actions had been taken or were to be taken at some point.
Specifically, 26 states reported taking action as of August 1, 2005, while
 12 states and the District of Columbia reported they would do so by
January 1, 2006. An additional 4 states reported that actions were
planned, but at no particular point in time. In a related question, we
asked what actions had been taken or were planned to install or activate
mechanisms to protect voter privacy. Again, a majority of states reported
actions had been taken or were to be taken at some point. Specifically, 32
states reported taking action as of August 1, 2005, while 13 states and
the District of Columbia reported they would do so by January 1, 2006. Two
other states reported actions would be taken at a later point in time.

During our site visits, we asked local election officials what standards
or procedures were used for the November 2004 general election to help
ensure that the registration list was secure and that the privacy of
individuals was protected. Election officials in most jurisdictions
reported that voter information (such as name and address) is public
information if it is to be used for political purposes-though some do not
release Social Security numbers, and others limit access to this
information by requiring a fee. Some local election officials noted that
security standards for this information were not set by the state but
rather at the county or local level, though many look to the state for
future guidance on standards. The type of security in place to restrict
access to voter registration records varied by jurisdiction; among the
procedures commonly used were password protection (so that only certain
election officials could log onto the voter registration system to access
the information); storage of voter registration records in locked
facilities; use of "best practice" protocols such as system firewalls; and
in some cases, registration information is maintained on a computer system
that is separate from the jurisdiction's central system. Along these
lines, 1 jurisdiction noted that it planned to implement a public key
infrastructure (PKI). A PKI is a system of computers, software, policies,
and people that can be used to facilitate the protection of sensitive
information and communications. The official noted it is a felony in that
jurisdiction to use a PKI authorization without authorization from the
State Board of Elections. Election officials in another jurisdiction we
visited told us that all voter registration system users must log on using
unique user IDs and passwords, which are maintained by the county
registrar. The system tracks all data entries and changes, which user made
them, and when they were made. In a few jurisdictions, election officials
said they grant additional privacy to the records of voters involved in
domestic disputes or other law enforcement matters.

Sharing Registration and Eligibility Information among States Is Limited

When asked whether they had any plan to develop or change existing
security standards or procedures, local election officials in 16 of the 
28 jurisdictions we visited told us there were no plans to alter current
practices, though some noted they were not sure. Among those indicating
that security procedures were being enhanced, election officials in 1
large jurisdiction said they planned to enclose their computer systems
server in a secure case with restricted access. Another official in a
large jurisdiction in another state said that because of a change in state
law in 2004, a hard copy of voter records was no longer available for
public inspection.

As mentioned earlier, the HAVA computerized statewide voter registration
list provisions require states to perform list maintenance to identify
duplicate registrations, deceased registrants, and registrants who may be
ineligible to vote under state law based upon a felony conviction.
However, we note that requirements for matching voter registration lists
with certain state records leaves some potential gaps for incomplete and
inaccurate voter registration lists because election officials may not
have information regarding registered voters who die out of state or who
are in prison in another state and ineligible because of a criminal
conviction. To determine whether states went beyond HAVA requirements to
share voter registration data with other states to identify registrants
who died in another state, were incarcerated in another state, or
registered in another state, we asked on our survey of state election
officials whether they had taken action to electronically exchange voter
registration information with at least 1 other state and whether they were
sharing registration information routinely with other states. In our state
survey, 31 states and the District of Columbia reported that they did not
plan to electronically exchange voter registration information with
another state. However, 35 states and the District of Columbia reported
they share information with states when a new registrant indicates he or
she previously resided in another state. Other types of information
sharing across state lines were less common. For instance, 6 states
reported sharing voter registration information with neighboring states,
and 1 state reported that it shared information with states where an
individual is known to reside part of the year. In our state survey, 14
states reported they do not currently share voter registration information
with other states.

We analyzed state and federal voter registration applications to determine
whether these applications provided space for applicants to indicate they
were registered in other states or in other jurisdictions within the same
state to identify duplicate registrations.28 We obtained state application
forms during site visits with local election jurisdictions, from state Web
sites or, if not available from there, we obtained the application from
the state. Registration forms were those on the Web site or obtained from
the states as of January 2006. Applications for the 46 states and the
District of Columbia and both federal applications had a place on their
registration application where registration applicants could indicate
prior registration in another state on their forms. Three states
(Kentucky, Texas, and Wyoming) did not include a place on their
registration forms to identify prior registration information in another
state.29 Forty-five states and the District of Columbia included a space
for registration applicants to indicate prior registration in another
jurisdiction within their state on their forms, or in the case of the
District of Columbia applicants were to indicate the address of their last
registration. Four states (Alaska, Hawaii, Kentucky, and Wyoming) did not
provide space to indicate prior registration within their state. Figure 11
is an example of a state registration form that provided a space for the
voter registration applicant to indicate that he or she had registered in
another state.

28The National Voter Registration Act required the Federal Election
Commission (FEC) to develop a mail voter registration application form for
elections for federal office. NVRA also requires states to accept and use
this federal mail voter registration application form for the registration
of voters in elections for federal office. Pursuant to HAVA, various FEC
functions, including those relating to the federal mail voter registration
form application were transferred to the Election Assistance Commission.

29The total number of states does not add to 50 because North Dakota does
not require voters to register to vote and therefore does not have voter
registration forms.

  Figure 11: Virginia Voter Registration Form Indicating whether the Voter Had
                              Registered Elsewhere

Source: State Board of Elections Web site for Virginia.

On the basis of our survey of local election jurisdictions, we estimated
that 12 percent of local jurisdictions administered their own registration
application form in addition to the state registration application. Of the
 12 percent who had their own form, we estimate that 70 percent had space
on their voter registration applications so that an applicant can
indicate  

Many States Have or Plan to Have Additional Election Management Features
in Voter Registration Systems

whether he or she was previously registered in another state.30 However,
we estimate that about a third did not capture this information on their
forms.

Although HAVA's voter registration-related provisions focus primarily on
state election management activities for developing, verifying, and
maintaining voter lists, we sought information on what other types of
registration system upgrades, if any, states planned, and we asked at the
sites we visited what additional system capabilities, if any, had been
implemented or planned. In our state survey, 15 states reported taking
action to upgrade the processing speed or records capacity of their
systems as of August 2005; 6 states reported that such actions would be
taken by January 2006; and 12 states and the District of Columbia reported
they would take such action at a later time.

In other recent work, we have also looked at selected states' efforts to
enhance their statewide voter list systems. In our February 2006 report on
certain states' experiences with implementing HAVA's statewide voter
registration lists, we found that 7 of 9 states that reported implementing
HAVA provisions for a computerized, statewide voter registration system by
January 1, 2004, also reported that they have upgraded or enhanced their
systems, or planned to so do, to include additional election management
capabilities.31 For example, Arizona reported plans to upgrade its current
system to reflect reciprocity agreements with other states, so that
election officials can be alerted when a voter moves from state to state,
and will allow election officials to retrieve data on such issues as voter
petitions, provisional ballots, poll worker training, and polling
locations. Other states reported adding or planning similar enhancements.
Kentucky reported another type of enhancement: It has used its statewide
computerized voter registration system to establish voter information
centers on the state's Web site, to assist applicants and staff in the
voter registration process.

During our site visits, we asked local election officials to comment on
the election management functions their voter registration systems might
perform. While some local election officials noted they were not certain

30The 95 percent confidence interval for this percentage is +11 or -13
percentage points.

31GAO, Election Reform: Nine States' Experiences Implementing Federal
Requirements to Establish Computerized Voter Registration Lists,
GAO-06-247 (Washington, D.C.: February 2006).

Page 71 GAO-06-450 Elections

whether their new statewide voter registration systems would include the
same array of features as the local county versions, other local election
officials in some jurisdictions responded that they expect their statewide
systems to be able to perform some or all of the following functions:

     o maintain records confirming mailings to new registrants,
     o generate letters informing rejected applicants of reasons for
       rejection,
     o generate forms or mailing labels,
     o note status or date of absentee applications and ballots sent and
       received,
     o identify polling places for use on Election Day, and
     o identify poll workers.

In some jurisdictions, other capabilities were mentioned; 2 large
jurisdictions noted, for instance, that bar coding would be used to
identify registrants, and 2 other large jurisdictions indicated that their
systems would track and maintain candidate petition information.

Not all jurisdictions expressed equal confidence in the extra
(non-HAVA-related) capabilities of their systems. Election officials in a
couple of large jurisdictions, for instance, told us they were not certain
their statewide voter system would have features comparable to those
already in place, and that their vendor or state was taking a
one-size-fits-all approach for all jurisdictions regardless of size,
rather than taking specific local needs into account. In some
jurisdictions, election officials stated that their statewide systems were
still too new to know whether these additional functions would be
operational, and some said they were not yet familiar with all the
system's capabilities.

Implementation of Identification Requirements for First-Time Mail
Registrants Varied

HAVA imposed new identification requirements for certain mail
registrants-such as, individuals who register by mail and have not
previously voted in a federal election within the state.32 These
individuals (first-time mail registrants) must provide certain specified
types of identification either by submitting copies of such identification
during the mail registration process or by presenting such identification
when voting in person for the first time following their mail
registration.33 Moreover, first-time mail registrants are to be informed
on the application that appropriate identifying information must be
submitted with the mailed form in order to avoid additional identification
requirements upon voting for the first time. An individual who asserts
that he or she has registered by mail and desires to vote in person but
who does not meet the identification requirements may cast a provisional
ballot under HAVA's provisional language. However, according to election
officials in some jurisdictions we visited, casting a provisional ballot
requires that these voters are to provide identification to election
officials by a specified time (e.g., by close of polls on Election Day or
within a certain number of days following Election Day) to have their
ballot count. On the basis of our local survey, we estimate that 32
percent of local jurisdictions encountered a problem in counting
provisional ballots because voters did not provide identification as
specified by HAVA for mail-in registrants and were voting for the first
time in the precinct or jurisdiction.34 Our discussion of provisional
voting processes appears in chapter 5.

32HAVA section 303(b) (codified at 42 U.S.C.S: 15483(b)).

33Acceptable forms of identification to meet this HAVA requirement
include, for example, a current and valid photo identification, a copy of
a current utility bill, bank statement, government check, paycheck, or
other government document that shows the name and address of the voter.

34The 95 percent confidence interval for this percentage is +/- 6
percentage points.

Page 73 GAO-06-450 Elections

HAVA, in general, provides states with discretion as to the methods of
implementing HAVA's identification requirements for first-time mail
registrants, such as ensuring that voters comply with the requirements
and, subject to certain limitations, allows states to establish
requirements that are stricter than those required under HAVA.35 According
to our state survey, 7 states reported that such HAVA requirements were
already covered by existing state legislation or some type of state
executive action (such as orders, directives, regulations, or policies);
44 states and the District of Columbia reported that they enacted new
legislation or took some type of state executive action (such as orders,
directives, regulations, or polices) to address the identification
requirements in HAVA for first-time mail registrants.36

We analyzed state and federal (NVRA) voter registration application forms
to determine whether the applications provided instructions on
identification requirements for individuals registering in a jurisdiction
for the first time. We obtained some state application forms during site
visits with local election jurisdictions, and others from state Web sites
or, if not available from there, we obtained the application from the
state. Registration forms were those on the Web site or obtained from the
states as of January 2006. Our analysis showed that 39 states and the
District of Columbia had information on their application forms and 10
states did not provide this information on their forms.37 The NVRA voter
registration form included this information. Figure 12 is an example of a
voter registration form that included instructions for first-time mail
registrants.

35HAVA sections 304 and 305 (codified at 42 U.S.C.S:S: 15484, and 15485,
respectively). For example, Arizona submitted an inquiry to Department of
Justice's Civil Rights Division asking whether it was permissible under
HAVA for a state to mandate that a potential voter show identification at
the polls prior to receiving a provisional ballot. The Civil Rights
Division responded, in part, in September 2005, that while HAVA requires
states to allow voters who meet certain specified conditions the
opportunity to cast a provisional ballot, states are free to prescribe
their own rules for deciding whether to count those ballots. (September 1,
2005, letter to the State of Arizona from the Civil Rights Division).

36Three states (California, Texas, and Washington) reporting that existing
legislation or executive action addressed HAVA first-time voter
identification requirements also reported enacting additional legislation
or taking executive action to address such HAVA requirements.

37The 10 states are Alabama, Hawaii, Iowa, Kentucky, New Hampshire,
Pennsylvania, South Dakota, Tennessee, Utah, and Wyoming. Three of these
states, Hawaii, Kentucky, and Tennessee, require full Social Security
numbers on applications for voter registration.

  Figure 12: Colorado Voter Registration Form with Instructions for First-Time
                                Mail Registrants

               Source: Secretary of State Web site for Colorado.

During our site visits, we asked local election officials whether they
considered registering by mail to only include when someone mails in a
single application or to also include mailed-in applications from voter
registration drives. Five local jurisdictions told us that applications
received by mail as a result of voter registration drives are not treated
as mail-in applications and therefore are not treated as subject to mail
registration identification requirements under HAVA; 3 jurisdictions told
us that applications submitted by voter registration drives were treated
as mail-in applications subject to HAVA's mail registration identification
requirements. Election officials in 1 of these jurisdictions told us that
under their state law (Pennsylvania) all voters who are voting for the
first time in a district must show a valid form of identification,
regardless of how they registered to vote.

Also, during our site visits we asked local election officials how they
processed voter registration applications from first-time mail registrants
for the 2004 general election. Election officials reported taking
different approaches, many involving mailed communications from election
officials sent back to the applicant, particularly if required information
was missing. For example, at least 2 large jurisdictions reported that
first-time voters who did not mail in identification with their
applications were sent letters instructing them to do so. Similarly,
officials in 2 jurisdictions in another state said letters were sent to
applicants whose applications were incomplete, advising them of the need
to provide photo ID-and informed applicants that if they failed to do so,
they may have to use a provisional ballot on Election Day, which would be
subject to the voter subsequently providing identification. In other
jurisdictions, though local election officials reported taking steps to
process incomplete applications from first-time voters, they did not
necessarily give the applicant a chance to correct the application prior
to Election Day. For example, in a medium jurisdiction we visited,
first-time voter applicants who did not submit proper identification were
to have been given provisional ballots. However, the election official
told us her office did not inform them about this in advance for the 2004
general election.

In addition to contacting applicants to inform them of the need to provide
identification discussed above, 1 jurisdiction we visited told us that it
periodically provided a list of applicants who provided driver's license
numbers but did not provide identification at the time of registration to
the state MVA as another means to verify the registrant's identity. In
this case, the MVA compared the county clerk office's registration list
against its list of licensed drivers to see if the name, date of birth,
and driver's license number matched, and returned the results to election
officials. If all these data elements matched, the election official
certified the records and these prospective voters were not required to
show identification at the polling place.

If a registrant did not provide identification prior to Election Day,
local election officials at all 28 sites we visited reported having a
system for recording first-time voters who failed to provide
identification and transferring that information to a polling site by
annotating the poll book. One large jurisdiction, for example shaded the
voter line in the poll book, while another printed the words "ID required"
next to the voter's name. With respect to voters who presented themselves
at a polling place and did not have identification, election officials at
some local jurisdictions we visited described different ways that the
voter's provisional ballot could become verified. For example, a
jurisdiction in Georgia said that if a voter did not provide
identification at the polls, it allowed the voter to vote a provisional
ballot and the voter had until 2 days after the election to provide
identification. Another jurisdiction in Kansas told us that the voter had
until the day that votes were canvassed to provide identification. Other
jurisdictions told us that voters would have until the close of the polls
on Election Day to provide identification to election officials. A local
jurisdiction in Washington told us that if the voter did not have
identification on Election Day, the voter would vote a provisional ballot
and election officials would subsequently have the voter's signature
matched against the registration application to verify the voters
identity.

Citizens generally have numerous opportunities to apply to register to
vote. Figure 13 shows several of these opportunities-such as applying at a
local election office, at a motor vehicle agency, or through a voter
registration drive-and the processes used to submit an application.

Processing Registration Applications Received from MVAs and Other NVRA
Entities Remained a Challenge

        Figure 13: Example of the Voter Registration Application Process

Source: GAO.

Problems with applications submitted to MVAs have been identified as a
challenge since 1999. Our October 2001 report on election processes found
that 46 percent of local jurisdictions nationwide had problems processing
applications submitted at MVAs and other public registration sites  
designated pursuant to NVRA requirements.38 In its reports to Congress on
the impact of NVRA on federal elections in 1999 through 2002, the Federal
Election Commission (FEC) found that several states reported problems with
election officials receiving applications from MVA offices in a timely
manner, resulting in, the FEC stated, "the effective disenfranchisement"
of citizens who had applied to vote but were not processed by Election
Day.39 FEC recommended in both reports that states develop ongoing
training programs for personnel in NVRA agencies, such as MVAs.

HAVA includes requirements providing that voters who contend that they
registered (at MVAs or through other means) in the jurisdiction in which
they desire to vote, but whose names are not on the voter registration
list for that polling place, be allowed to cast a provisional ballot.40
HAVA also requires that voters who an election official asserts is not
eligible to vote also be permitted to cast a provisional ballot. Election
officials would determine the voter's eligibility under state law and
whether the vote should count as part of the vote counting process. From
our local jurisdiction survey, we estimate that for the 2004 general
election,  61 percent of local jurisdictions had a problem in counting
provisional ballots because of insufficient evidence that individuals had
submitted voter registration applications at MVAs.41 In addition, we
estimate that  29 percent of local jurisdictions had a problem in
counting provisional ballots because of insufficient evidence that
individuals had submitted voter registration applications at NVRA agencies
other than MVAs.42 Also, our September 2005 report on managing voter
registration reported that  4 of 12 jurisdictions we surveyed reported
that election office staff experienced challenges,43 either to a great
extent or some extent, receiving

38GAO-02-3.

39Federal Election Commission, The Impact of the National Voter
Registration Act on Federal Elections 1999-2000 and 2001-2002. These
reports surveyed 44 states, and the District of Columbia, that were
subject to NVRA provisions for the administration of elections for federal
office.

40HAVA section 302 (codified at 42 U.S.C.S: 15482).

41The 95 percent confidence interval for this percentage is +/- 7
percentage points.

42The 95 percent confidence interval for this percentage is +/- 6
percentage points.

43GAO, Elections: Views of Selected Local Election Officials on Managing
Voter Registration and Ensuring Eligible Citizens Can Vote, GAO-05-997
(Washington, D.C.: September 2005).

voter registration applications from motor vehicle agencies.44 They
reported taking steps to address the problem by hiring additional staff to
handle the volume of applications received and by contacting applicants to
obtain correct information.

There is evidence that, at least in 1 jurisdiction, election officials
took steps since the 2000 general election to address the MVA voter
registration issue, though problems persisted for the November 2004
general election. When we revisited the same small jurisdiction in 2005
that we had visited in 2001, election officials reported they were still
experiencing problems receiving registration forms from the MVA, for all
those who registered to vote there-but noted that the process had
improved. For example, they said elections staff now have access to the
MVA database directly, so they can verify whether someone who claimed to
have registered at the MVA actually did so.

In our local jurisdictions survey, we estimate that few jurisdictions
provided training to MVA or other NVRA agencies. Specifically, for the
2004 general election, we estimate that 12 percent of local jurisdictions
provided training or guidance to MVA offices and an estimated 3 percent
provided training to other NVRA entities regarding procedures for
distributing and collecting voter registration applications. Large
jurisdictions are statistically different from small or medium
jurisdictions, and medium jurisdictions are statistically different from
small jurisdictions. Specifically, we estimate that 34 percent of large
jurisdictions provided training to MVA offices, an estimated 18 percent of
medium jurisdictions did so, and an estimated 9 percent of small
jurisdictions did this. In addition, large jurisdictions are statistically
different from both medium and small jurisdictions in providing training
to other NVRA entities.

In our October 2001 comprehensive report on election processes nationwide,
we identified measures such as improving the training of MVA staff as a
means of addressing challenges related to applications received from
MVAs.45 After the November 2004 general election, the National Task Force
on Election Reform-composed almost exclusively of officials who 

44We surveyed 14 jurisdictions altogether, but Wisconsin, represented by 2
jurisdictions on this question, was not, at the time of our survey,
subject to NVRA, and therefore the MVA question did not apply.

45GAO-02-3.

Page 80 GAO-06-450 Elections

Jurisdictions Encountered New Challenges Processing the Large Volume of
Voter Registration Applications for the November 2004 General Election

served in voter registration and administration of elections capacities46-
reported that while the NVRA expanded the number of locations and
opportunities where citizens can apply to register to vote, supporting the
voter registration application process is a secondary duty for entities
that do so under this law. The task force report noted that it is a
challenge for these entities to provide this service in a consistent
manner and to transfer the registrations collected accurately and
efficiently to voter registration offices.

In our October 2001 report on election processes, some election officials
noted that while extending voter registration deadlines gave voters
additional chances to register, it shortened the time for processing
applications.47 And a few election officials raised concerns about short
time frames for processing applications in relation to the possibility of
voter fraud if there was insufficient time to verify an applicant's
eligibility. For the 2004 general election, the time frame for processing
applications had the potential to pose an even greater challenge given the
increase in the number of voter registration applications that elections
officials reported receiving for the November 2004 general election. The
conditions that election officials experienced in processing the volume of
voter registration applications, such as long hours and lack of time to
fully train temporary workers, could have resulted in data entry errors
that would have had the impact of not properly registering eligible voters
and not identifying ineligible voters.

During our site visits to local jurisdictions, election officials told us
that for the 2004 general election, entering applications in a timely
manner was possible-but challenges did arise, and election officials
described actions taken to help ensure that voters were properly
registered. Furthermore, on the basis of our survey of local election
jurisdictions, we estimate that 81 percent of local jurisdictions were
able to process applications received just prior to the registration
deadline-though we estimate 19 percent of the jurisdictions received
applications just prior to the registration deadline

46The task force, composed of local election officials, was convened by
the nonprofit Election Center, also known as the National Association of
Election Officials, following the 2000 federal election to study and
address questions about the election. The task force was reconvened in
2005 to examine the 2004 election.

47GAO-02-3.

Page 81 GAO-06-450 Elections

that posed problems in entering them prior to Election Day. As shown in
figure 14, we estimate that large jurisdictions experienced problems in
entering the number of voter registration applications more than small and
medium jurisdictions. Large jurisdictions are statistically different from
both medium and small jurisdictions. This may be attributable to larger
jurisdictions having larger populations with more registration activity,
among other things.

Figure 14: Estimated Percentage of Jurisdictions Having Problems Entering
the Number of Voter Registration Applications Received for 2004 General
Election

Percentage of jurisdictions

50 49

40

30

21

19 16

20

10

0 Experienced problems entering applications

Small (<10,000) Medium (10,000-100,000) Large (>100,000) Overall Source:
GAO 2005 survey of local election jurisdictions.

Note: Large jurisdictions are statistically different from both medium and
small jurisdictions.

All jurisdictions we visited reported that they were able to enter all
eligible applications into the voter registration lists. Nevertheless,
most reported it was a challenge to process the large volume of
applications received. For example, 1 large jurisdiction we visited
reported that on a daily basis it was 30,000 to 40,000 applications behind
in data entry. As a result, election officials reported that they hired 80
full-time temporary workers who worked two full-time shifts to enter all
eligible applications into the voter registration list used at the polls
on Election Day. Election officials in another large jurisdiction told us
that they unexpectedly received about 10,000 last-minute registrants.
Another large jurisdiction reported it was "swamped" with registration
applications right before the registration deadline and was not prepared
for the volume of applications submitted. Several jurisdictions required
permanent employees to work extended hours or on weekends. To manage
registration workloads, other jurisdictions reported hiring temporary
workers and recruiting county employees to handle processing workloads.
Figure 15 shows the reported spike in voter registration applications
received prior to Election Day in  1 large jurisdiction. Some
applications were received after the final week allowed for voter
registration and could not be registered for the 2004 general election but
were registered for future elections.

Figure 15: Total Weekly Voter Registration Applications Documented by a
Large Jurisdiction in 2004

Number of applications (in thousands) 30

25

20

15

10

5

0 15 10152025303540455052 2004 (by week)

Source: GAO (analysis), Clark County, Nevada (data).

In our state survey, a few states reported that since the 2000 general
election they increased the time that voters in their states have to
register. Although setting registration deadlines close to Election Day
itself provides citizens increased time to apply to register, reducing the
number of days from the registration deadline to Election Day can make it
difficult for election officials to ensure that all eligible voters are
included on the voter registration list. Specifically, in our state
survey, 3 states (Maryland, Nevada, and Vermont) reported changing their
registration deadlines for the November 2004 general election. For the
2000 general election, Maryland's registration deadline had been 25 days
before the election, but for the 2004 general election, the deadline for
registration was 21 days before the election, extending the time that
voters could register by 4 days. Nevada's 2000 registration deadline (9
p.m. on the fifth Saturday preceding any primary or general election)
remained the same for mail-in registrations. However, for the 2004 general
election, the state extended inperson registration by 10 days. Vermont's
voter registration deadline changed from the second Saturday before the
election to the second Monday before the election, allowing voters 2 more
days to register. Appendix VI provides information on state laws
pertaining to registration deadlines.

On the basis of our local jurisdiction survey, entering all voter
registration applications for the time between the registration deadline
and the November 2004 general election posed problems for large
jurisdictions more than it did for small and medium jurisdictions.
Specifically, we estimate that 41 percent of large jurisdictions
experienced problems,  18 percent of medium jurisdictions, and 13 percent
of small jurisdictions. Large jurisdictions are significantly different
from both medium and small jurisdictions. Inasmuch as large jurisdictions
have more potential registrants, it is reasonable to expect that they
would experience more difficulty entering all voter registration
applications by Election Day than smaller ones would.

For the 2004 general election, while many states reported having
registration deadlines that were 20 to 30 days prior to Election Day, a
few states reported having registration deadlines that were 10 days or
less prior to Election Day, and some states reported having same-day
registration. Four states (Alabama, Maine, New Hampshire, and Vermont)
reported having registration deadlines that were 10 days or less prior to
Election Day. Idaho, Maine, Minnesota, New Hampshire, Wisconsin, and
Wyoming reported having Election Day registration at the polling place.

According to Our Nationwide Survey and Jurisdictions Visited, Some
Jurisdictions Lacked Sufficient Staff to Process Applications

Having sufficient staff to process the increased number of voter
registration applications was an issue for large local election
jurisdictions. On the basis of our nationwide survey, most local
jurisdictions (an estimated 89 percent) had a sufficient number of
election workers (whether full-time, part-time, or temporary) who were
able to enter registration applications in a timely manner. However, we
estimate that  11 percent had an insufficient workforce for this task.
Large jurisdictions experienced problems with insufficient election
workers to enter voter registrations applications more than small and
medium jurisdictions did, as shown in figure 16. The difference between
large jurisdictions and both medium and small jurisdictions is
statistically significant. This difference could be attributable to larger
jurisdictions having a greater need for additional staff.

Figure 16: Estimated Percentage of Local Jurisdictions with Insufficient
Election Workers to Process Voter Registration Applications

Percentage of jurisdictions

25

25

20

16

15

11

10

8

5

0

            Experienced Insufficient workers to process applications

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000)

Overall Source: GAO 2005 survey of local election jurisdictions.

Note: The difference between large jurisdictions and both medium and small
jurisdictions is statistically significant.

Several jurisdictions we visited reported that there was a price to pay
for the large volume of registration applications received, such as the
need to hire temporary workers or extend the hours of permanent employees
in order to process voter registration applications for the November 2004
general election. Election officials in several jurisdictions we visited
commented on the financial impact of the temporary workers hired, overtime
hours, and the purchase of needed equipment, such as computers.

In our September 2005 report on managing voter registration, we noted that
all but 1 of the 14 jurisdictions we surveyed faced challenges receiving
and processing voter registration applications during the 2004 general
election

Larger Jurisdictions More Likely than Small or Medium Ones to Provide
Training to Staff and Track Receipt of Voter Applications

and took various steps to address them.48 For example, election officials
in 7 of the 14 jurisdictions reported challenges checking voter
registration applications for completeness, or for accuracy, or for
duplicates. At that time, as in our more recent site visits, jurisdictions
reported hiring extra staff, among other things, to address these
challenges.

Providing training to data entry staff and tracking applications provide
ways for election officials to manage the flow of applications for
processing that can help ensure that voter registration applications are
appropriately entered into the voter registration list. As part of our
inquiry into the methods jurisdictions used to enter completed
registration application data into voter lists, our questionnaire to local
election jurisdictions asked how they went about accomplishing this task.
On the basis of our survey, we estimate that 76 percent of all local
jurisdictions provided training to data entry staff about the processing
and inputting of registration applications. Seventy-five percent of small
jurisdictions provided this training, 73 percent of medium jurisdictions
did so, and  94 percent of larger jurisdictions did so, too. Large
jurisdictions are statistically different from both medium and small
jurisdictions. Another activity that election officials undertook when
entering completed registration applications included tracking incoming
registrations. The results of our survey show that over half of local
jurisdictions tracked incoming registration applications to ascertain the
total number received, the number entered into registration lists, and the
number not processed because of omission or application error, and to
identify ineligible voters based on age or residence. Again, large
jurisdictions are statistically different from both medium and small
jurisdictions. Table 2 provides information on the different activities
that local election jurisdictions undertake when entering completed
registration applications into the official voter registration list.

48GAO-05-997.

Table 2: Estimated Percentage of Local Jurisdictions That Engaged in
Various Quality Assurance Activities when Entering Completed
Voter-Registration Applications

Jurisdictions                 Percentage of    Percentage of Percentage of 
engaged in the                small                                  large 
following       Percentage of jurisdictions medium           jurisdictions 
activities                all               jurisdictions    
when entering                                                
completed       jurisdictions     (<10,000) (10,000-100,000)    (>100,000) 
registration                                                 
applications                                                 
Provided                                                                   
training to                76            75               73            94
data entry                                                   
staff                                                        
about the                                                    
processing and                                               
inputting of                                                 
registration                                                 
applicationsa                                                
Employed a                                                                 
system of edit             72            66               82            95
checks in                                                    
the voter                                                    
registration                                                 
database to                                                  
identify                                                     
duplicatesb                                                  
Employed a                                                                 
system of edit             67            62               74            89
checks in                                                    
the voter                                                    
registration                                                 
database to                                                  
identify                                                     
ineligibles                                                  
based on agea                                                
Employed a                                                                 
system of edit             64            61               68            84
checks in                                                    
the voter                                                    
registration                                                 
database to                                                  
identify                                                     
ineligibles                                                  
based on                                                     
residencea                                                   
Verified input                                                             
of data by the             60            57               64            77
same or a                                                    
different                                                    
individual to                                                
confirm                                                      
initial                                                      
input                                                        
accuracyc                                                    
Tracked                                                                    
incoming                   59            55               64            74
registration                                                 
applications                                                 
for total                                                    
number                                                       
received,                                                    
number entered                                               
into                                                         
registration                                                 
list,                                                        
and number not                                               
processed                                                    
because                                                      
of an omission                                               
or error on                                                  
applicationc                                                 

Processing Applications Obtained through Voter Registration Drives Posed
Additional Challenges for Some Jurisdictions

Source: GAO 2005 survey of local election jurisdictions. a Large
jurisdictions are statistically different from both medium and small
jurisdictions. b All size categories are statistically different from one
another. c Large jurisdictions are statistically different from small
jurisdictions.

Nongovernmental organizations in many states sponsored voter registration
drives for the November 2004 general election in an effort to increase the
number of citizens eligible to vote. Voter registration drives pose a
dilemma for some election officials. On one hand, voter registration
drives provide another means by which persons can apply to register to
vote. On the other hand, they pose challenges in assessing the validity of
submitted registrations and in processing large numbers of registrations
submitted close to the registration deadline. For the November 2004
general election, election officials in some jurisdictions we visited told
us they encountered challenges validating and processing the large number
of voter registration applications obtained through voter registration
drives that employed either paid staff (where workers are paid for each
voter registration application completed and submitted to election
authorities prior to Election Day)49 or used volunteers. For example,
Wisconsin's state legislative audit bureau conducted an evaluation of the
2004 general election in its state.50 It found, among other things, that
many registration deputies appointed for the November 2004 general
election worked for special interest groups or political parties
interested in increasing voter turnout.51 The evaluation states that
investigators found that registration deputies had submitted 65 falsified
names for the 2004 general elections and that district attorneys in two
counties charged four individuals with submitting fraudulent registration
forms. According to the evaluation report, these registration deputies
were reportedly paid by their employer on a per registrant basis, which
may have encouraged them to submit fraudulent registration forms to
increase their compensation.

Such questions about the integrity of the voter registration process were
of particular concern in battleground states such as Florida, Ohio, and
Pennsylvania, where margins of victory were slim and accurate tallies of
eligible votes were therefore of consequence.52 In our state survey
several states reported that their state election provisions do not
address the issue of voter registration drives that involve payment per
application, while relatively fewer states reported prohibiting them
outright. Specifically,  19 states and the District of Columbia reported
that state laws or executive actions are silent about these drives (that
is, it is left up to each local jurisdiction to decide). However, 1 of
these 19 states further reported that while its state law does not address
voter registration drives that involve

49With respect to federal voter registration forms, the National Voter
Registration Act of 1993 provides that the chief election official of a
state shall make such forms available through governmental and private
entities, with particular emphasis on making them available for organized
voter registration programs. 42 U.S.C. S: 1973gg-4(b).

50Legislative Audit Bureau, State of Wisconsin. An Evaluation: Voter
Registration, Report 05-12 (Madison, Wisconsin: September 2005).

51In some states, election officials appoint registration deputies who
assist with voter registration. Some registration deputies are municipal
officials and are permitted to work for interest groups or political
parties.

52EAC, using various news media sources, identified 17 states deemed to be
most competitive in the 2004 presidential contest. These "battleground or
highly contested" states included Arkansas, Arizona, Colorado, Florida,
Iowa, Michigan, Minnesota, Missouri, Nevada, New Hampshire, New Mexico,
Ohio, Oregon, Pennsylvania, Washington, West Virginia, and Wisconsin. We
visited 8 of these 17 states during our site visits; alternatively,  16
of the 28 jurisdictions we visited were located in these battleground
states.

payment per application, the conduct of such drives is not left up to each
local jurisdiction-the local jurisdictions have no authority in regulating
such matters. Sixteen states reported that voter registration drives are
allowed either by state law or by executive action, 13 states reported
that they are prohibited by state law, and 2 states did not respond.

In addition, our nationwide survey of local election jurisdictions
inquired about their awareness and handling of registration drives, and
any actions taken to deter fraudulent applications from being submitted by
persons or groups participating in paid registration drives, and we
discussed this matter during our site visits to selected jurisdictions as
well. In our nationwide survey, we estimate that 91 percent of all local
jurisdictions were not aware of such drives, while 9 percent were aware.
About a third (an estimated 32 percent) of the large jurisdictions-those
with populations greater than 100,000-were aware of such drives. We also
queried local election jurisdictions whether any names on voter
registration applications appeared to be fraudulent. On the basis of our
local survey, nearly all jurisdictions-an estimated 95 percent-did not
have any names that appeared to be fraudulent. Although only 5 percent of
local election jurisdictions had voter registration applications that
appeared to have fraudulent names, an estimated 70 percent identified
receiving 10 fraudulent applications or fewer, an estimated 14 percent
identified receiving 10 or more fraudulent applications, and an estimated
 16 percent did not know the volume of fraudulent applications received.
The distribution of the volume of fraudulent applications received is of a
smaller subset of our total sample and therefore has larger confidence
intervals than other estimates. Figure 17 shows the extent to which local
jurisdictions identified experiencing fraudulent voter registration
applications.

Figure 17: Estimated Percentage of Local Jurisdictions That Had Names on
Voter Registration Applications Appearing to Be Fraudulent and Volume of
Fraudulent Applications Identified for the 2004 General Election

Did not have applications with names that appeared fraudulent Had
applications with names that appeared fraudulent

Source: GAO 2005 survey of local election jurisdictions.

Note: The distribution of the volume of fraudulent applications received
is of a smaller subset of our total sample and therefore has larger
confidence intervals than other estimates. For the category of  10 or
fewer fraudulent applications the confidence interval is +/- 17 percent,
for the category of 10 or more the confidence level is +/- 16 percent, and
for the category of don't know the confidence level is +/- 12 percent.
Furthermore, not all respondents who had names on voter registration
applications that appeared to be fraudulent answered our question about
the volume of these applications. Therefore, our estimate of the subset
for the volume of applications that were received is based on 4 percent of
jurisdictions that answered that question.

In addition, our prior work raised concerns about the quality of voter
registration applications obtained through voter registration drives. In
our September 2005 report on managing voter registration, we reported that

Election Officials' Views on Irregular Voter Registration Applications
from Registration Drives

among 12 of 14 local jurisdictions we surveyed, processing applications
received from voter registration drives sponsored by nongovernmental
organizations posed a challenge to election officials because applications
were incomplete or inaccurate.53

During our site visits, we sought local officials' views on a host of
issues related to the integrity of the voter registration process,
including how or whether voter registration drive applications were
tracked, how many registration applications were submitted by volunteer or
paid registration drives in calendar year 2004 leading up to the November
election, and how their jurisdictions dealt with irregular applications.
(We defined irregular applications as those using fictitious names,
unusual dates of birth, nonexistent addresses, or fake signatures or party
affiliations.) We also asked election officials whether they had the
ability to determine if individuals were using false or fictitious names.
Many local jurisdictions that we visited told us that they did not have
specific procedures to ensure that voter applications obtained through
voter registration drives were collected or tracked. This was because, in
some cases, the application forms could simply be downloaded from the
Internet. One large jurisdiction that did not track applications coming
from various sources told us it planned to begin doing so, using a
drop-down menu in its statewide voter registration system that will allow
staff to record the information.

Overall, at local jurisdictions that we visited where applications from
voter registration drives were tracked or at least estimated, the number
and proportion of applications submitted through voter registration drives
relative to total registrations-and the number and proportion considered
irregular-varied widely. For example, in 1 large jurisdiction, election
officials reported that approximately 30,000 registrations received in
2004-about 90 percent-were submitted by registration drives. Of these, the
election officials estimated that only about 50 applications were
irregular-that is, they were unreadable, had questionable signatures, were
incomplete, or had invalid addresses. The election official from this
jurisdiction noted that it appeared some of the applications had been
filled out by individuals who took addresses from the phone book and
changed them slightly. In another large jurisdiction in a battleground
state, local election officials estimated that 70,000 registration
applications were submitted by volunteer or paid registration drives, and
here too

53GAO-05-997.

irregularities were noted-such as fictitious names and fake signatures-
but election officials stated that these irregular applications
represented a "low" percentage of the total. In other large jurisdictions,
fewer voter registration applications were received; 1 jurisdiction, for
example, in another battleground state, reported receiving 2,500 such
applications and estimated that about 20 percent of them were irregular.
Two medium jurisdictions we visited reported receiving a few hundred voter
registration applications or fewer, and both reported that there were no
irregularities. One small jurisdiction did not report any voter
registration drives taking place.

When we asked local election officials during our site visits whether they
had the ability to determine whether a person actually tried to vote using
a false or fictitious name, responses were mixed: Election officials in 3
large jurisdictions we visited told us they did not have the ability to
make this determination. An election official in another large
jurisdiction stated that "there is no way to know if someone falsely
registered has voted." Others, however, reported that they were able to
determine whether false identities had been used. For example, in 1 large
jurisdiction, election judges check voter IDs and signatures at the polls
to prevent the use of fictitious identities. One large jurisdiction
verifies voter registration information against Social Security and
driver's license information and checked voter history internally;
election officials in this jurisdiction reported that they believe anyone
who attempted to use a false or fictitious name in the November 2004
general election would have been caught. And in another jurisdiction,
election officials told us that if an individual attempted to vote using a
fictitious name that was not in the poll book, that individual would be
issued a provisional ballot-which would not be verified if it was
determined that the name was indeed fictitious. Election officials in some
jurisdictions we visited said there was no way to know whether the poll
book already contained fictitious names.

When asked what steps, if any, local jurisdictions we visited took to
notify law enforcement or other legal authorities on irregular
registration applications received, most reported taking some actions. For
example,  1 large jurisdiction we visited reported providing irregular
registration applications to the Federal Bureau of Investigation (FBI) and
the district attorney's office and to the Secretary of State's office for
investigation. Both the FBI and the district attorney declined to pursue
the matter on the ground that they were understaffed, the jurisdiction
reported. The Secretary of State's office concluded that while the
registration applications were fraudulent or fictitious, a purposeful
fraud was not

Some Jurisdictions Have Controls to Manage Registration Drives

committed and that the people completing the fake applications were not
trying to alter an election, but to obtain money by working for the
registration drives. Four other jurisdictions that we visited said they
contacted appropriate state or federal authorities, such as state law
enforcement, a State's Attorney, a state election enforcement agency, or
the FBI, but election officials did not know whether any action had been
taken.

In addition, in our June 2005 report on maintaining voter registration
lists, we reported that election officials in seven locations we visited
referred reported instances of voter registration fraud allegations to
appropriate agencies, such as the district attorney and the U.S. Attorney
for investigation.54 Also, EAC issued voluntary guidance in July 2005 to
help states implement HAVA. EAC's guidance suggested that when the voter
registration verification process indicates the possible commission of an
election crime, such as the submission of false registration information,
such matters should be forwarded to local, state, and federal law
enforcement authorities for investigation.

When we asked local jurisdictions that we visited whether they had
procedures in place for registration groups to follow when submitting
applications, election officials in most jurisdictions reported that some
type of system was in place to control registration drives. For example, 
1 large jurisdiction reported that it had a program to train volunteer
field registrars to register citizens on behalf of the county registrar;
these field registrars were to comply with all registration rules and laws
and must themselves be registered voters, and noncandidates, have proof of
identify, complete a 2-hour training course, and pass a brief examination
before taking an oath. In addition, this same jurisdiction required that
any group requesting more than 50 voter registration forms was required to
provide a plan to the state elections department for when, where, and how
it would distribute the forms-all of which were numbered so that election
offices could track them. Some jurisdictions reported, however, that no
procedures were in place that registration groups had to follow. One large
jurisdiction, for instance, reported that anyone can run a voter
registration drive simply by downloading the voter registration form from
the election office Web site.

On the topic of what actions, if any, local jurisdictions had taken to
deter paid registration drives from submitting fraudulent registration

54GAO-05-478.

Actions Taken to Help Prevent Fraudulent Registrations and Ensure
Submission of Registrations to Election Offices

applications, from our nationwide survey, we estimate that roughly half of
the estimated 9 percent of local jurisdictions that were aware that paid
registration drives were occurring provided training or guidance on how to
accurately complete an application, and an estimated 41 percent of these
jurisdictions notified the persons or groups engaged in paid registration
drives that they had submitted incomplete, inaccurate, or fraudulent
applications.55 In addition, on the basis of our survey, 41 percent of
local jurisdictions that were aware of the drives helped prevent
submission of incomplete, inaccurate, or fraudulent applications by
working with persons and groups engaged in paid registration drives.56

In a couple of jurisdictions, election officials told us they took other
steps, such as meeting with registration drive organizers and contacting
the registrant identified on the application, to help prevent fraudulent
registrations. A jurisdiction in Colorado reported that numerous
complaints had been received from voters who claimed to have completed
registrations through a drive but for whom the county had no record of
application. The jurisdiction reported that Colorado's legislature passed
a bill pertaining to voter registration drives. Subsequently, Colorado
enacted legislation effective in June 2005 that, among other things,
requires voter registration organizers to file a statement of intent with
the Secretary of State, fulfill training requirements pursuant to rules
promulgated by the Secretary of State, and, in general, submit or mail
registration applications within 5 business days. In addition, the 2005
state legislation provides that voter registration organizers may not
compensate persons circulating voter registration application forms based
on the number of applications distributed or collected. The Secretary of
State issued rules in November 2005 implementing such requirements,
including rules that require registration drive organizers to file a
statement of intent with the Secretary of State and require persons
circulating such application forms to ensure that the tear-off receipt on
the application is completed and given to the applicant. Election
officials in 17 jurisdictions we visited told us that they had procedures
in place for managing voter registration drives to some extent. For
example, in 1 medium jurisdiction, election officials stated that groups
or persons seeking to run registration drives must be trained and
deputized by the registrar's office.

55The 95 percent confidence interval is +/- 13 percentage points. 56The 95
percent confidence interval is +/- 14 percentage points.

Page 95 GAO-06-450 Elections

Concluding Observations

In 43 of the 50 states and the District of Columbia, successfully
registering to vote prior to Election Day is a prerequisite for casting a
ballot and having that ballot counted. States are still working to fully
implement HAVA's voter registration requirements. As states gain more
experience with their statewide voter registration and data matching
systems and processes, it is likely their systems and processes will
evolve. Given the continuing challenge of maintaining accurate voter
registration lists in a highly mobile society, this is to be expected.

For election officials, the voter registration process presents a
continuing challenge in balancing ease of registration for eligible voters
with sufficient internal controls to help ensure that only eligible voters
are added to and remain on the voter registration rolls. To maintain
accurate voter registration lists, election officials must use and rely
upon data from a number of sources, such as state death and criminal
records and applications from MVAs.

HAVA's requirements for creating and maintaining statewide voter
registration lists and its identification requirements for first-time
voters who register by mail were designed to help improve the accuracy of
voter registration lists and reduce the potential for voter fraud.
Specifically, HAVA's requirements for creating and maintaining a statewide
voter registration list was designed to improve voter registration list
accuracy by identifying duplicate registrations within the state and
identifying those ineligible to vote because of death, criminal status, or
other reasons. HAVA requires states to match the names and other
identifying information on their statewide voter registration lists
against death and felony records in the state. States may voluntarily
match their voter registration lists with the voter registration lists,
death, felony, or other records in other states. In the absence of
voluntary cross-state matching, it is possible to fully implement HAVA's
statewide voter registration provisions and still have ineligible persons
on the state's voter registration rolls on Election Day, such as those who
died out of state or were convicted in federal courts or other states.57
Nor would implementing HAVA's statewide matching requirements identify
persons who are registered to vote in more than one state. Although some
states report sharing registration and eligibility information among
states, the practice was generally limited to neighboring states or
dependent upon a registrant indicating that he or she previously resided
in another state.

57GAO-05-478.

HAVA includes a provision that requires certain first-time voters who
register by mail to provide identification as proof of their identity and
eligibility to vote in the jurisdiction. Which voters must present
identification either with their mail application or when they vote for
the first time depends upon how states and local jurisdictions define
"mail registrations" subject to HAVA's identification requirement. In our
site visits, we found that some local jurisdictions considered
registration applications submitted by registration drives to be mail
registrations subject to HAVA's identification requirement for first-time
voters, while other jurisdictions did not consider such registrations to
be mail registrations subject to the identification requirement. This
distinction has importance on Election Day for first-time voters who
registered through registration drives. In those jurisdictions that
considered mail registrations to include registration drive applications,
first-time voters who registered through registration drives would be
required to show an acceptable form of identification at the polls on
election day. If they did not do so, they are to be permitted to cast a
provisional ballot, but the ballot would only be counted upon a state
determination that the voter is eligible to vote under state law. In
contrast, in those jurisdictions that did not consider mail applications
to include those submitted through registration drives, firsttime voters
would not be treated as subject to the HAVA identification requirement and
could generally cast a regular ballot that would be counted with all other
regular ballots.

Election jurisdictions continue to face challenges in obtaining voter
registration applications from NVRA entities, including MVAs. Some local
jurisdictions have established processes to manage receipt of voter
registration applications from these entities, such as training for staffs
of these agencies. To the extent that NVRA entities do not track and
forward to the appropriate election jurisdiction the voter applications
that they have received, voters may be required to cast provisional
ballots instead of regular ones because their names do not appear on the
voter registration lists. In addition, the provisional ballot will not be
counted if the voter's valid registration cannot be verified. Our survey
of local election jurisdictions found that many local jurisdictions
encountered problems counting provisional ballots in cases where voters
claimed to have registered at an MVA or some other NVRA entity but there
was insufficient evidence that the voter had submitted a registration
application at the MVA or NVRA entity.

A surge of last-minute registrations in many jurisdictions prior to the
November 2004 election illustrated the challenge of balancing ease of
registration with assurance that only eligible voters are on the
registration rolls. Some election jurisdictions reported registration
drive groups submitted hundreds or thousands of applications just before
the registration deadline. When the registration deadline is close to
Election Day, processing these applications presents a tremendous
challenge in checking applications for completeness, having time to
contact applicants to obtain missing information, verifying applicants'
eligibility to vote, and adding the name of eligible voters to the
registration list. Some jurisdictions reported hiring and training
temporary employees to process the applications. The enormous workload and
time constraints associated with processing large numbers of last-minute
applications can increase the chances that errors will be made in
determining voter eligibility, and the names of some eligible voters may
not be added to the list in time for Election Day.

Chapter 3

Absentee and Early Voting

A growing number of citizens seem to be casting their ballots before
Election Day using absentee and early voting options that are offered by
states and local jurisdictions.

However, circumstances under which these voters vote and the manner in
which they cast their ballots before Election Day differ because there are
51 unique election codes.1 Because of the wide diversity in absentee and
early voting requirements, administration, and procedures, citizens face
different opportunities for obtaining and successfully casting ballots
before Election Day.

To collect information about absentee and early voting options, in our
state and local surveys we asked questions about each of these voting
options separately. We defined absentee voting as casting a ballot,
generally by mail, in advance of Election Day (although ballots could be
returned through Election Day and dropped off in person). We defined early
voting as generally in-person voting in advance of Election Day at
specific polling locations, separate from absentee voting. However, there
is some measure of overlap between absentee voting and early voting
reported by the states, especially where states have reported in-person
absentee voting to be, in effect, early voting. This may be due, in part,
to the fact that the relational statutory framework for early voting and
absentee voting varies among the states-with some states, for example,
providing early voting within the context of the state's absentee voting
provisions, while others, for example, provide for absentee voting within
the context of the state's early voting provisions. Similarly, local
jurisdictions that completed our survey may also have had some measure of
overlap in relation to their practices for absentee and early voting.
During our interviews with local election officials in jurisdictions that
offered early voting, we were able to obtain more detailed information
about absentee and early voting procedures and practices for those
jurisdictions.

On the basis of our site visits to jurisdictions that had early voting,
absentee and early voting were similar in some ways and distinct in
others. Election officials described to us that when voters cast absentee
ballots, they typically followed a specific process including applying for
and receiving the ballot and returning their marked ballots before
Election Day or, in some cases, returning the ballot up until the close of
polls on Election Day.2 According to the description that election
officials gave us, early voting

1This total includes the 50 states and the District of Columbia.

                                    Overview

                      Chapter 3 Absentee and Early Voting

was distinct from in-person absentee voting in that in-person absentee
voters usually applied for and received a ballot, and cast it at the
registrar's office, while early voters reported to a voting location where
early voting staff verified their eligibility to vote, usually by
accessing the jurisdiction's voter registration list. Also, early voting
usually did not require citizens to provide an excuse, as some states
required for absentee voting, and it was usually allowed for a shorter
period of time than absentee voting. For example, in the 14 jurisdictions
we visited in 7 states that reported having early voting, the time frame
allowed for absentee voting was almost always at least twice as long as
that for early voting (e.g., Colorado allowed 30 days for absentee voting
and 15 days for early voting). Early voting was similar to Election Day
voting in that the voting methods were usually the same. However,
according to election officials in jurisdictions we visited that had early
voting, voters were not limited to voting in their precinct because all
early voting locations had access to a complete list of registered voters
for the jurisdiction (not just precinct specific) and had appropriate
ballots that included federal, state, and precinct-specific races.
Appendix VII provides a description of selected characteristics of the
early voting jurisdictions we visited.

In this chapter, we will discuss changes since 2000 and challenges related
to (1) absentee voting in general, (2) overseas military and civilian
absentee voting, and (3) early voting.

Some states have increased the opportunities for citizens to vote absentee
or early. For the November 2004 general election, 21 states reported that
they no longer required voters to provide excuses such as being ill,
having a disability, or being away from the precinct on Election Day to
vote absentee-an increase of 3 states from the November 2000 general
election. Three states reported expanding their provision for permanent
absentee status (usually reserved for the elderly or those with
disabilities), allowing voters to receive absentee ballots for a
state-specified time period, such as 4 years. One state reported
eliminating its requirement that mail-in absentee voters provide an
attestation from a notary or witness for their signature along with the
completed absentee ballot. Eliminating the need for a notary or witness
removes a potential barrier to an absentee ballot being counted. According
to election officials in 2 jurisdictions in 1 state

2Some states require voters to provide an excuse to cast an absentee
ballot. This subject is discussed in more detail later in this chapter.

Page 100 GAO-06-450 Elections

we visited that required a notary or witness signature, an absentee ballot
may not be counted if voters neglect to have their ballots witnessed or
notarized. Furthermore, HAVA amended the Uniformed and Overseas Citizens
Absentee Voting Act (UOCAVA) to, among other things, extend the period of
time that can be covered by a single absentee ballot application by absent
uniformed service voters and certain other civilian voters residing
outside of the United States from the year during which the application
was received to a time period covering up to the two next regularly
scheduled general elections for federal office.

Election officials reported facing some of the same challenges in the
November 2004 general election that they had identified to us for the
November 2000 general election, and they also reported some new
challenges. Continuing absentee voting challenges included (1) receiving
late absentee voter applications and ballots; (2) managing general
workload, resources, and other administrative constraints; (3) addressing
voter error issues such as unsigned or otherwise incomplete absentee
applications and ballot materials; and (4) preventing potential fraud.

Election officials also told us that they encountered new challenges in
the November 2004 general election. Some election officials said that the
increased early voter turnout during this election resulted in long lines.
In some local jurisdictions we visited, election officials said that
factors such as inadequate planning on their part, limitations on types of
facilities that could be used for early voting locations, and funding
constraints on hiring more staff or acquiring more voting locations
affected their management of large early voter turnout. In addition, some
election officials reported that they encountered a challenge handling
disruptive third parties as they attempted to approach early voters who
were in line to vote. Another challenge could develop as a result of a
2002 HAVA amendment to UOCAVA. In an effort to help make registration and
voting easier for absent uniformed service voters and certain other
civilian voters residing outside of the United States, this 2002
amendment, as noted above, extended the period of time that can be covered
by a single application from the year during which the application was
received to a time period covering up to the next two subsequent general
elections for federal office. Election officials in 4 jurisdictions we
visited told us that a possible unintended consequence of this amendment
could be that when uniformed services personnel are reassigned to other
duty posts, absentee ballots may not be sent to the correct address for
subsequent general elections. Even with a 2005 revision to the ballot
request form whereby voters can indicate that they want ballots for one
federal election only, election officials in 3 of these jurisdictions were
concerned many absentee ballots would be returned as undeliverable.

Absentee voting allows citizens the opportunity to vote when they are

Absentee Voting

unable to vote at their precinct on Election Day. Although availability,
eligibility requirements, administration, and procedures vary across the 
50 states and the District of Columbia, absentee voting generally follows
a basic process. As figure 18 shows, this process included four basic
steps for the November 2004 general election.

aIn general, for a mail-in absentee ballot application to be approved, it
must meet certain state or local requirements. For example, in some
states, the voter must be registered, and in some jurisdictions, the
signature on the absentee application must match the one on the voter
registration application, among other things.

Jurisdictions we visited typically provided absentee ballot applications
that registered voters used to request absentee ballots in a standard
state or jurisdiction form, as shown in figure 19.

Chapter 3 Absentee and Early Voting

Figure 19: Example of a Standard Absentee Ballot Application Form for the
November 2004 General Election

                       Source: El Paso County, Colorado.

According to our state survey, state election officials reported that
registered voters could visit or write their local election office, or in
some cases visit a state or local election Web site, to obtain an
application or learn what information was required to request an absentee
ballot. State election officials reported registered voters could return a
completed absentee ballot application via the U.S. mail or in many other
different ways as allowed by state absentee ballot provisions. Also, some
election officials in jurisdictions we visited told us that voters could
complete any part of the absentee voting process in person at their local
elections office. Table 3 shows the various options allowed by states for
requesting and returning absentee ballot applications. However, it is
important to note that particular local jurisdictions might not have
offered all of the options described below.

Table 3: Options Allowed by States for Requesting and Returning Absentee
Ballot Applications, November 2004 General Election

Options                                                  Number of statesa 
Allowed for requesting absentee ballot applications 
In person                                                               51 
U.S. mail                                                               51 
Telephone                                                               41 
E-mail                                                                  42 
Via facsimile                                                           48 
Via state Web site                                                      30 
Download from Web site                                                  41 
Allowed for returning absentee ballot applications  
In person                                                               51 
U.S. mail                                                               51 
E-mail                                                                  11 
Via facsimile                                                           34 
Via state Web site                                                       4 

Source: GAO 2005 survey of state election officials.

aStates include the 50 states and the District of Columbia.

According to our state survey results, states reported that applicants
could find out the status of their absentee ballot application after it
was submitted and offered at least one of several ways, including
telephoning a state or local jurisdiction office, telephoning a hotline or
toll-free number,

Page 106 GAO-06-450 Elections

or e-mailing a state or local jurisdiction office. For example, in 49
states and in the District of Columbia, applicants could telephone a state
or local jurisdiction office, and in 47 states and in the District of
Columbia, applicants could e-mail a state or local jurisdiction office to
find out their absentee ballot applications' status. Thirty-nine states
and the District of Columbia notified the applicant if the application was
rejected.

While absentee ballots are generally provided to the voter through the
mail, unless voting in person, on the basis of our survey of a
representative sample of local jurisdictions nationwide, some
jurisdictions provided absentee ballots using fax and e-mail.
Specifically, for the November 2004 general election, we estimate that 17
percent of local jurisdictions provided absentee ballots by fax, and 4
percent of local jurisdictions provided absentee ballots by e-mail. On the
basis of our discussions with election officials in jurisdictions we
visited, absentee ballots are generally returned through the mail.

Election officials in most jurisdictions we visited said that voters used
a combination of envelopes for returning completed absentee ballots so
that voters' indentities would be distinct from the ballots they were
casting. For example, a voter would place the completed ballot in a
secrecy (inner) envelope, which would then be placed in an outer envelope.
The secrecy envelope would be to ensure that the voted ballot was not
linked to the voter, while the voter's affidavit information, such as a
name, address, and signature, needed to certify that the voter was
eligible to vote, would be marked on the outer envelope. Election
officials in some jurisdictions provided examples of the envelopes used to
return absentee ballots. One of these examples had a separate affidavit
envelope, which was to be placed in a pre-addressed return envelope and
mailed to the local elections jurisdiction. Other examples allowed the
voter to include the affidavit information on the back of the
pre-addressed return envelope. Once the local elections jurisdiction
certified that the absentee ballots could be counted using the affidavit
information, election officials in jurisdictions we visited told us that
they removed the secrecy envelope (with the voted ballot sealed inside)
and set it aside for counting. Figure 20 shows examples of absentee ballot
return envelopes and the inclusion of affidavit information.

Figure 20: Two Examples of Absentee Ballot Envelopes with the Inclusion of
Affidavit Information-One with Certification on Inner Envelope and One
with Certification on Outer Envelope-for the November 2004 General
Election

         Sources: Champaign County, Illinois; Clark County, Washington.

In our survey of state election officials, we asked whether absentee
voters were able to find out the status of their submitted absentee
ballots in various ways. According to our state survey, 44 states and the
District of Columbia reported that absentee voters were able to telephone
a state or local jurisdiction office, 32 states and the District of
Columbia reported that absentee voters were able to e-mail a state or
local jurisdiction office, 16 states reported that absentee voters could
telephone a hotline or tollfree number, and 5 states reported that
absentee voters' ballot status was available via a Web site. Furthermore,
16 states reported that either state or local jurisdictions would notify
the voter if the absentee ballot was not counted. However, 6 states
reported that they do not allow voters to check the status of their
absentee ballots at all. For example, Vermont reported that state law does
not allow voters to find out whether or not the absentee ballot was
counted. Kentucky reported that it does not track whether or not an
individual voter's ballot was counted because linking a voted ballot back
to a specific voter violates that voter's right to a secret ballot.

A Few States Reported A few states reported changes to their requirements
with respect to absentee voting by (1) no longer requiring a reason or
excuse for voting

Changes to Absentee Voting

Requirements since 2000 absentee; (2) eliminating the need for a mail-in
absentee voter to have a notary or witness for the voter's signature to
accompany the ballot; and 

(3) not limiting permanent absentee voting status to individuals with
disabilities or the elderly.

Excuse Requirement According to our state survey regarding the November
2004 general election, all 50 states and the District of Columbia had some
provisions allowing registered voters to vote before Election Day, but not
every registered voter was eligible to do so. Twenty-one states reported
allowing voters to vote absentee for the November 2004 general election
without first having to provide a reason or excuse. The other 29 states
and the District of Columbia reported requiring voters to meet one of
several criteria, or "excuses," to be eligible to vote before Election
Day, such as having a disability, being elderly, or being absent from the
jurisdiction on Election Day. The following are examples of excuses that
some states required:

     o absent from the state or county on Election Day;
     o a member of the uniformed services or a dependent;
     o a permanent or total disability;
          * ill or having a temporary disability;
          * o  over a certain age, such as 65;
     o at a school, college, or university;
     o employed on Election Day in a job for which the nature or hours
       prevent the individual from voting at his or her precinct, such as an
       election worker; and
     o involved in emergency circumstances, such as the death of a family
       member.

In our survey of local jurisdictions, we asked about problems encountered
when processing absentee ballot applications. As shown in figure 21, we
estimate that 9 percent of local jurisdictions received absentee
applications that did not meet the excuse required by law, in states where
excuses were required. The issue of applicants not meeting the required
excuse is more of a problem for large jurisdictions than small or medium
jurisdictions.

Figure 21: Estimated Percentage of Local Jurisdictions That Encountered
Problems Processing Absentee Applications because the Applicant Did Not
Meet the Excuse Required by State Law, November 2004 General Election

Percentage of jurisdictions

33

30

25

20

15

13

9

10

6

5

0 Problems processing absentee applications when excuse was required

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000)

Overall Source: GAO 2005 survey of local election jurisdictions.

Note: The difference between large jurisdictions and small and medium
jurisdictions is statistically significant.

According to our state survey, the number of states that allowed absentee
voting without an excuse increased from 18 in 2000 to 21 in 2004. Since
November 2004, 2 more states reported that they have eliminated their
excuse requirement. Specifically, during visits to local jurisdictions in
New Jersey, election officials told us that state law had changed since
the November 2004 general election. According to these officials,
no-excuse absentee voting was adopted by the New Jersey legislature and
became

Notary or Witness Signature Requirement

Permanent Absentee Voting

effective in July 2005. Ohio also amended its absentee voter provisions,
effective January 2006, to provide for no-excuse absentee voting.

Election officials in 2 jurisdictions in 1 state we visited told us that
if voters returned a completed (voted) ballot without having the signature
notarized or affirmed by a witness, the vote would be disqualified and not
counted. For the November 2004 general election, according to our state
survey,  12 states reported requiring that mail-in absentee ballots
contain attestation by a notary or witness for a voter's signature to
accompany the absentee ballot. From the November 2000 election to the
November 2004 election, Florida was the only state that reported in our
state survey that it had dropped the requirement that mail-in absentee
ballots contain attestation by a notary or witness for a voter's
signature.

Permanent absentee voting, which typically was available to individuals
with disabilities or the elderly, was another way some states sought to
help enfranchise certain categories of voters. Permanent absentee status,
where offered, generally allowed the voter to apply for mail-in absentee
ballots once (rather than for each separate election) over a specified
time period. State requirements dictated when and how often a voter must
apply for permanent absentee status. For example, for the November 2004
general election, in a New Jersey jurisdiction that we visited, election
officials told us that state law required those eligible for permanent
absentee status to apply at the beginning of the calendar year to receive
absentee ballots for that year. According to the absentee ballot
application provided by this jurisdiction, a voter's permanent absentee
status remains in effect throughout that year unless the voter notifies
the election office otherwise. An election official in a Pennsylvania
jurisdiction we visited said that his state allowed permanent absentee
voters to apply once every 4 years. In this state, permanent absentee
voters were to receive absentee ballots for all elections during the
4-year period, according to the election official. In  2 Washington
jurisdictions we visited, election officials told us that any voter could
qualify for permanent absentee status for all future elections (e.g., no
time period specified). In one of these Washington jurisdictions, election
officials provided a copy of the permanent absentee application
instructing voters that their permanent absentee status would be
terminated upon the (1) voter's written request, (2) cancelation of the
voter's registration record, (3) death or disqualification, or (4) return
of an ongoing absentee ballot as undeliverable.

Our state survey results showed that since the November 2000 general
election, 3 states (California, Rhode Island, and Utah) reported state

Receiving Late Absentee Ballot Applications and Ballots for the November
2004 General Election Continued to Be a Challenge

changes that expanded, in some manner, the use of permanent absentee
voting. For example, California, reported changes for the November 2004
election that allowed any voter to apply for and receive permanent
absentee status. For the November 2000 general election, California
previously reported that only certain categories of voters with
disabilities (e.g., blind voters) were eligible for permanent absentee
status. Overall, the results of our state survey showed that at the time
of the November 2004 general election, 17 states reported having some
provision for permanent absentee status, 32 states and the District of
Columbia reported that they did not provide for permanent absentee status,
and Oregon reported conducting its election entirely by mail-making
permanent absentee status unnecessary in this state.

Appendix VIII provides information on states' requirements for no-excuse
absentee voting and witness or notary signature provisions for the
November 2000 and 2004 general elections and shows where changes occurred.
States did not report any changes to their permanent absentee requirements
since the November 2000 general election.

The results from our state survey show that deadlines for voters to both
apply for absentee ballots and return them to local jurisdictions to be
counted differed among states. According to our state survey for the
November 2004 general election, 47 states and the District of Columbia
reported that they had absentee ballot application deadlines that ranged
from Election Day (5 states: Connecticut, Maine, New Jersey, Ohio, and
South Dakota) to 21 days before Election Day (Rhode Island). Three states
(Florida, New Hampshire, and Oregon) reported having no absentee ballot
application deadline, although ballots in these states had to be returned
by the close of polls on Election Day.3

With respect to state deadlines for returning absentee ballots, many
states reported having more than one deadline to correspond with differing
methods of returning such ballots to election officials. In our state
survey, 44 states reported having provisions requiring that absentee
ballots be returned by or on Election Day; 7 states reported having
provisions requiring that absentee ballots be returned a certain number of
days before Election Day; and 8 states and the District of Columbia
reported having

3Oregon conducts its entire election by mail.

provisions allowing mailed absentee ballots to be returned a certain
number of days after Election Day, if such ballots were postmarked by a
specified date. For example, for the 2004 November general election,
Alaska reported two deadlines: (1) mail-in absentee ballots were to be
received by close of business on the 10th day after the election when
postmarked on or before Election Day, and (2) in-person absentee ballots
were to be delivered by 8:00 p.m. on Election Day.

Also, according to our state survey, Nebraska reported that for absentee
ballots returned by mail, the deadline changed from no later than 2 days
after Election Day for the November 2000 general election to the close of
polls on Election Day for the November 2004 general election. According to
our state survey, these deadlines may be different for absent uniformed
service voters and certain other civilian voters residing outside the
United States, a subject that will be discussed later in this chapter.

In our October 2001 comprehensive report on election processes, we
reported that election officials for the 2000 general election identified
receiving applications and ballots after state statutory deadlines as a
challenge.4 According to our nationwide survey, local jurisdictions
encountered similar problems with processing absentee ballot applications
and absentee ballots for the November 2004 general election. More
specifically, on the basis of our survey, we estimate that 55 percent of
local jurisdictions received absentee ballot applications too late to
process. We also estimate 77 percent of local jurisdictions encountered
problems in processing absentee ballots because ballots were received too
late. Furthermore, we asked jurisdictions about which problems were
encountered most frequently. An estimated 25 percent of local
jurisdictions encountered the ballot lateness problem most frequently.
Figure 22 shows that medium and large jurisdictions encountered lateness
with absentee ballots more than small jurisdictions did.

4GAO-02-3.

Figure 22: Estimated Percentage of Local Jurisdictions That Encountered
Lateness with Absentee Ballot Applications and Absentee Ballots, November
2004 General Election

Percentage of jurisdictions

100 97

Applications Ballots received received too latea too lateb

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000)

Overall Source: GAO 2005 survey of local election jurisdictions.

aAll size categories are statistically different from one another.

bSmall jurisdictions are statistically different from medium and large
jurisdictions.

Appendix VIII summarizes states' deadlines for receiving domestic mail-in
absentee ballot applications and absentee ballots.

Election officials in the local jurisdictions we visited told us that they
tried to approve applications and mail absentee ballots to voters as
quickly as possible, assuming that the ballots had been finalized and
printed. In  8 jurisdictions we visited in 5 states (Colorado, Kansas,
New Mexico, Pennsylvania, and Washington), election officials said that
their states mandated that local election jurisdictions process absentee
ballot applications within a specified time period, such as within 24, 48,
or  72 hours of receipt of the application. In 2 Pennsylvania
jurisdictions we visited, election officials stated that they established
a local policy encouraging election staff to process absentee ballot
applications faster (such as on the day of receipt) than the time period
specified in state law (which was 48 hours). In 1 Illinois and 1 Nevada
jurisdiction we visited, election officials said that while a 24- or
48-hour turnaround time for absentee ballot applications was not mandated
in state law, local office policy was to process them as quickly as
possible-such as within 24 hours of receipt of the application.

During our site visits, election officials in 9 jurisdictions stated that
they received large numbers of mail-in absentee ballot applications just
prior to the deadlines prescribed by state law. Most of these election
officials said they were able to meet their state-mandated or office
policy applicationprocessing time, although they had to work long hours
and hire additional staff to process the absentee ballot applications by
the deadline. In  1 Florida jurisdiction we visited, local election
officials said that even though they had no absentee ballot application
deadline, they processed applications using "long hours and extra people"
and tried to send out absentee ballots within 24 hours of receiving a
complete application.

In jurisdictions we visited in Pennsylvania and Colorado, election
officials said that sometimes the 24- or 48-hour turnaround was impossible
to meet because the state did not finalize the ballots for printing until
the days immediately preceding Election Day for the November 2004
election. For example, an election official in the Pennsylvania
jurisdiction we visited told us that determining whether or not an
independent presidential candidate's name was to be included on the
November 2004 general election ballot proved to be a challenge. In this
jurisdiction, the validity of petition signatures supporting the
independent candidate's request to be included on the ballot was
challenged in state court about 10 weeks before the election. As a result,
according to the election official, election officials were required to
participate in a court-mandated process of verifying the signatures.
According to the election official, it took about 10 days in court to
resolve the situation, which delayed the printing of the ballots.

In 6 jurisdictions we visited, election officials told us that slowness in
the delivery of the mail added to the processing time crunch during the
week before Election Day-a problem that is out of election officials'
control and may contribute to the local election officials' receipt of
absentee voting materials after state-mandated deadlines. Although
envelopes can use an "official election mail" designation, election
officials in these 6 jurisdictions we visited said that the U.S. Postal
Service did not always process absentee voting materials in a timely
manner. For example, in one New Mexico jurisdiction we visited, election
officials stated that they experienced serious problems with the U.S.
Postal Service delivering absentee ballot applications. These officials
felt that the post office ignored the envelopes' official election mail
designation and did not process and deliver them quickly. Election
officials in this jurisdiction said that their telephone system crashed
numerous times leading up to Election Day in November 2004, given the
heavy volume of incoming calls from voters checking on the status of their
absentee ballot applications. In one Pennsylvania jurisdiction that we
visited, election officials said that postal concerns were raised when
some college students' absentee ballot applications were received after
Election Day. These officials could not definitely say at what point these
applications might have been delayed and explained that the mail delivery
delay could have been attributable to either the U.S. Postal Service or
the university's mailing center. Figure 23 illustrates the use of special
postal markings for absentee ballot materials.

Figure 23: Example of Envelope Illustrating Official Postal Marking for
Absentee Ballot Materials, November 2004 General Election

Source: Muscogee County, Georgia.

While election officials in 6 jurisdictions we visited told us about
challenges with mail delivery, election officials in 7 jurisdictions we
visited told us that they did not have problems with mail delivery or
coordinating with the U.S.

Postal Service. In an Illinois jurisdiction we visited, election officials
told us that prior to the election, staff from his office met with the
postmaster to establish a good working relationship. Election officials in
a New Hampshire and Ohio jurisdiction we visited stated that the post
office was very helpful. In a Nevada jurisdiction we visited election
officials said that they received excellent service from the postal
service.

When an absentee application was received after the state-mandated
deadline, election officials in 13 jurisdictions we visited told us that
they often sent these applicants a letter explaining that their
application was received too late. In 5 of these same jurisdictions,
election officials said they also provided an alternative to absentee
voting such as early voting, voting on Election Day, or in-person absentee
voting, where the voter could visit the election office and complete the
absentee voting process in person.

Voter Errors in the In our October 2001 report on election processes, we
reported that election officials for the 2000 general election identified
voters' failure to provide

November 2004 Election

Continued to Be a Challenge critical information, with respect to
signatures and addresses, as

challenges to successfully processing mail-in absentee applications and
with Processing Absentee verifying ballots for counting.5 According to our
nationwide survey for the Ballot Applications and November 2004 election,
local jurisdictions encountered similar voter Ballots errors that could
affect the jurisdictions' ability to establish voter eligibility

or approve the ballot for counting when processing absentee ballot
applications and absentee ballots.

Absentee Ballot Applications In our nationwide survey, we asked local
jurisdictions what problems they encountered in processing absentee ballot
applications. We estimate that 48 percent of them identified problems
receiving absentee ballot applications that contained a missing or
illegible voter signature. Furthermore, we asked about which problems were
encountered most frequently. An estimated 20 percent of local
jurisdictions encountered the problem of receiving absentee ballot
applications that contained a missing or illegible voter signature most
frequently. Table 4 shows our estimates of the types of voter errors local
jurisdictions encountered with absentee ballot applications submitted for
the November 2004 general election.

5GAO-02-3.

Table 4: Estimated Percentages of All Local Jurisdictions and by
Jurisdiction Size That Encountered Voter Error Problems in Processing
Absentee Ballot Applications, November 2004 General Election

                                                  Small     Medium      Large 
Problem                 All jurisdictions (< 10,000) (10,000 to (>100,000) 
                                                        100,000)   
Missing or illegible                   48        44c         55         73 
signaturea                                                      
Missing or inadequate                                                      
voting residence                       35     30             42         70
addressa                                                        
Applied to wrong                       33     32             32        58d 
jurisdictiona                                                   
Missing or inadequate                  32     25             43         74 
voting mailing addressb                                         

Source: GAO 2005 survey of local election jurisdictions.

aLarge jurisdictions are statistically different from small and medium
jurisdictions. bAll size jurisdictions are statistically different from
one another. cThe 95 percent confidence interval for this percentage is
+/- 6 percentage points. dThe 95 percent confidence interval for this
percentage is +/- 8 percentage points.

On the basis of our nationwide survey, large jurisdictions had more of a
problem than small or medium jurisdictions concerning missing or illegible
signatures. Specifically, we estimate that 73 percent of large
jurisdictions encountered this problem, while we estimate 44 percent and
55 percent of small and medium jurisdictions respectively encountered it.
Large jurisdictions are statistically different from medium and small
jurisdictions.

When elections officials were unable to process absentee ballot
applications, our nationwide survey showed that some local jurisdictions
contacted applicants to inform them of the status of their application
using the methods listed in table 5. Specifically, on the basis of our
survey of local jurisdictions, we estimate that 72 percent of all
jurisdictions telephoned applicants when their absentee applications could
not be processed. We found no significant difference based on the size of
the jurisdiction with regard to this contact method. However, we estimate
that 84 percent of medium jurisdictions and 90 percent of large
jurisdictions contacted absentee applicants by U.S. mail. In contrast, 63
percent of small jurisdictions contacted absentee applicants with problem
applications via

U.S. mail. Small jurisdictions are statistically different from medium and
large jurisdictions. We also estimate that 10 percent of local
jurisdictions did not inform any applicants about the status of their
application.

Table 5: Estimated Percentages of All Local Jurisdictions and by
Jurisdiction Size Using Various Contact Methods When Absentee Ballot
Applications Could Not Be Processed, November 2004 General Election

 Methods used to inform applicants of application Small Medium Large status All
                            jurisdictions (100,000)

                      Telephoned the applicant 72 72 73 77

                         Contact by maila 69 63d 84 90

                         Contact by e-mailb 20 18 21 46

                               Otherc 15 18 9 11

Source: GAO 2005 survey of local election jurisdictions.

aSmall jurisdictions are statistically different from medium and large
jurisdictions. bLarge jurisdictions are statistically different from small
and medium jurisdictions. c"Other" included contact by facsimile or
contacting voters' relatives, among other things. dThe 95 percent
confidence interval for this percentage is +/- 6 percentage points.

In an Illinois jurisdiction that we visited, elections officials told us
that they would do everything possible in an attempt to obtain complete
absentee applications from voters. If the absentee ballot application was
incomplete, election office staff said they contacted the voter and
attempted to resolve the problem in the best way practical, according to
the election officials. For example, if the application was missing the
voter's signature and there was enough time, the staff mailed the
application back to the voter for signature. If time was limited, the
staff called the voter and asked him or her to visit the election office
to sign the application. An election official in a Pennsylvania
jurisdiction we visited told us that if applicants forgot to include one
part of an address, such as a ZIP code, but election staff could match the
rest of the address and voters' identifying information with their
registration information, the application was approved. Election officials
in another Pennsylvania jurisdiction and a Nevada jurisdiction told us
that the voter registration system automatically generated letters to
voters when the application could not be processed for any reason.

Absentee Ballots In our nationwide survey, we asked local jurisdictions
what problems they encountered in processing submitted absentee ballots.
We estimate that  61 percent of all jurisdictions reported that absentee
ballots were received without the voter's signature on the envelope. We
estimate 54 percent of small jurisdictions, 76 percent of medium
jurisdictions, and 90 percent of large jurisdictions encountered this
problem. Jurisdictions of all sizes are statistically different from one
another. Table 6 shows our estimates of the types of problems election
officials encountered on absentee ballots. We estimate that 81 percent of
local jurisdictions encountered at least one of the problems listed.

Table 6: Estimated Percentages of All Local Jurisdictions and by
Jurisdiction Size That Encountered Problems in Processing Submitted
Absentee Ballots, November 2004 General Election

                                         All      Small     Medium      Large 
Problems encountered        jurisdictions (< 10,000) (10,000 to (>100,000) 
                                                         100,000)  
Envelope not signeda                   61         54         76         90 
Missing or incomplete                  36         38         33         36 
witness signature or                                            
information                                                     
Improper or missing notary             18         19         16        12c 
signature                                                       
Signature on the envelope              13          9         18        48c 
did not match the                                               
application or digitized                                        
signature on fileb                                              
Voter identification marks              9          9          7        23c 
on envelope or ballotb                                          
Voter identification number                                                
not included                            3          2          5         7c

Source: GAO 2005 survey of local election jurisdictions.

a Jurisdictions of all sizes are statistically different from one another.
bLarge jurisdictions are statistically different from small and medium
jurisdictions. cThe 95 percent confidence interval for this percentage is
+/- 8 percentage points.

If the ballot was not able to be verified, election officials in some
jurisdictions we visited told us that they attempted to contact the voter,
time permitting, so that the affidavit envelope could be corrected and
approved for counting. In 10 jurisdictions we visited, election officials
said that they reviewed the affidavit envelope information to approve the
ballots as they received them rather than waiting until Election Day. On
the basis of our nationwide survey, we estimate that 40 percent of local
jurisdictions contacted the voter by mail in an attempt to address a
problem with the affidavit envelope, and 39 percent contacted the voter
via telephone. Table 7 shows our estimates of the contact methods used by
local jurisdictions when absentee ballots had problems that could prevent
them from being approved for counting if not corrected.

Table 7: Estimated Percentages of All Local Jurisdictions and by
Jurisdiction Size Using Various Contact Methods when Absentee Ballots
Could Not Be Processed, November 2004 General Election

Methods used to inform                      Small        Medium      Large 
voters of ballot                                                
status                  All jurisdictions (< 10,000) (10,000 to (>100,000) 
                                                         100,000)  
Contacted the voter by                 40         31         61         66 
maila                                                           
Telephoned the voter                   39        38c         42         41 
Did not inform voters                  27         29         24         25 
Did not receive any                                                        
ballots that could not                 21         25         12          3
be                                                              
processed                                                       
Contacted the voter by                  8          8          7         15 
e-mail                                                          
Otherb                                 11         11          7         16 

Source: GAO 2005 survey of local election jurisdictions.

aSmall jurisdictions are statistically different than medium and large
jurisdictions.

b"Other" included contacting the voter by facsimile, contacting a family
member of the voter, or providing a hotline number for voters to check
their ballot status, among other things.

cThe 95 percent confidence interval for this percentage is +/- 6
percentage points.

Differences in whether voters were contacted by mail when there were
problems with their absentee ballots were based on the size of the local
elections jurisdiction. Specifically, we estimate that 31 percent of
small,  61 percent of medium, and 66 percent of large jurisdictions
contacted voters by mail. Small jurisdictions are statistically different
from medium and large jurisdictions.

While election officials in 10 jurisdictions we visited told us that they
qualified absentee ballots prior to Election Day-allowing them time to
follow up with voters, in 6 local jurisdictions we visited, election
officials told us that they qualified or approved absentee ballots for
counting on Election Day. According to election officials in these
jurisdictions, contacting the voter for corrected or complete ballot
information was not a viable option because there was not enough time.
These election officials stated that absentee ballots with incomplete or
inaccurate information on the affidavit envelope would not be qualified or
counted.

Some election officials in jurisdictions we visited told us that voters
can visit local election offices and complete all or part of the absentee
process in person. Some election officials told us that when voters vote
in-person absentee, officials are well situated to help ensure that the
application and ballot are complete and accurate before accepting them.
For example, in one Connecticut jurisdiction we visited, election
officials told us that they

Some Election Jurisdictions Continued to Have Concerns about Fraud and Had
Procedures to Address the Potential for Fraud

did not have incomplete absentee ballot applications from voters who
visited the office in person because they reviewed the application and
required the person to correct any errors before leaving.

In our October 2001 report on election processes, we reported that
election officials for the 2000 general election had concerns with mail-in
absentee voting fraud, particularly regarding absentee voters being unduly
influenced or intimidated while voting.6 However, we also reported that
election officials identified that they had established procedures to
address certain potential for fraud, such as someone other than the
registered voter completing the ballot or voters casting more than one
ballot in the same election.

Once the voters received and voted absentee ballots in accordance with any
state or local requirements (such as providing a signature or other
information on the affidavit envelope), such ballots were to be returned
to specified election officials. In general, local election officials or
poll workers were to review the information on the affidavit envelope and
subsequently verified or disqualified the ballot for counting based on
compliance with these administrative requirements, according to election
officials in some local jurisdictions we visited.

In our state survey, we asked states whether they specified how local
jurisdictions were to determine eligibility of absentee ballots. According
to our survey, 44 states and the District of Columbia reported that at the
time of our survey, they specified how to determine absentee ballot
eligibility, while 6 states reported that they did not. Colorado, for
example, specified that the poll worker is to compare the signature of the
voter on a selfaffirmation envelope with a signature on file with the
county clerk and recorder. Wisconsin specified, among other things, that
inspectors ascertain whether a certification has been properly executed,
if the applicant is a qualified elector of the ward or election district,
and that the voter has not already voted in the election.

Our survey of local elections jurisdictions asked election officials if
they used any of the procedures described in table 8 to ensure that the
absentee voter did not vote more than once for the November 2004 general
election.

6GAO-02-3.

These procedures could have been conducted either manually by elections
officials or through system edit checks. On the basis of our survey of
local jurisdictions, we estimate that 69 percent of jurisdictions checked
the Election Day poll book to determine whether the voter had been sent an
absentee ballot, and 68 percent of jurisdictions checked the Election Day
poll book to determine whether the voter had completed an absentee ballot.

Table 8: Estimated Percentages of Local Jurisdictions That Used Specific
Procedures to Help Ensure Voters Did Not Vote More than Once, November
2004 General Election

                                         All      Small     Medium      Large 
Procedure                   jurisdictions (< 10,000) (10,000 to (>100,000) 
                                                        100,000)   
Election Day poll book                                                     
checked to determine                   69         65         77         82
whether the                                                     
voter had been sent an                                          
absentee ballot                                                 
Election Day poll book                                                     
checked to determine                   68         66         74         56
whether the                                                     
voter had completed an                                          
absentee ballot                                                 
Election Day poll book                                                     
checked against the                    64         68         54         48
absentee                                                        
ballots prior to counting                                       
thema                                                           
Applications or voter                                                      
registration system checked            58        52c         70         85
to                                                              
determine whether the voter                                     
had already applied for an                                      
absentee ballotb                                                

Source: GAO 2005 survey of local election jurisdictions.

aSmall jurisdictions are significantly different from medium and large
jurisdictions. bAll size categories are statistically different from one
another. cThe 95 percent confidence interval for this percentage is +/- 6
percentage points.

On our survey of local jurisdictions, we also asked if any of the
procedures listed in table 9 were in place to ensure that the absentee
ballots were actually completed by the person requesting the ballot. On
the basis of our survey of local jurisdictions, we estimate that 70
percent of jurisdictions compared the absentee ballot signature with the
absentee application signature.

Table 9: Estimated Percentages of Local Jurisdictions That Used Specific
Procedures to Help Ensure Absentee Ballots Were Actually Completed by the
Person Requesting the Ballot, November 2004 General Election

                                                  Small     Medium      Large 
Procedure                  All            (< 10,000) (10,000 to (>100,000) 
                              jurisdictions             100,000)   
Absentee ballot signature                                                  
was compared with the                  70         72         69        40c
absentee                                                        
application signaturea                                          
Absentee ballot signature                                                  
had to be witnessed (in                54         64         28         25
only those                                                      
states that required it)b                                       
Absentee ballot signature                                                  
was compared with the                  39         36         47        62c
voter                                                           
registration signature                                          
Absentee ballot signature                                                  
had to be notarized (in                 8          7         14         6c
only those                                                      
states that required it)                                        

Source: GAO 2005 survey of local election jurisdictions.

a Large jurisdictions are statistically different from small and medium
jurisdictions. b Small jurisdictions are statistically different from
medium and large jurisdictions. cThe 95 percent confidence interval for
these percentages is +/- 8 percentage points.

With respect to comparing the absentee ballot application signature with
the absentee ballot signature, there were differences based on the size of
the jurisdiction. On the basis of our survey of local jurisdictions, we
estimate that 72 percent of small, 69 percent of medium, and 40 percent of
large jurisdictions compared these signatures. Large jurisdictions are
significantly different from small and medium jurisdictions. One reason
that large jurisdictions may differ is that they have a large volume of
absentee ballots to process and it may be too resource intensive to
compare signatures, among other things.

During our site visits, elections officials provided examples of the
procedures they used to ensure against fraud. For example in 20 local
jurisdictions that we visited, election officials said that when the
ballot signature was compared with the absentee application signature,
voter registration signature, or some other signature on file, the
signatures had to match for the ballot to be approved and counted. In
addition to matching signatures, election officials in 2 Illinois
jurisdictions and 1 New Jersey jurisdiction we visited told us that during
the Election Day absentee ballot qualification process, poll workers were
instructed to check the poll book to determine if the voter had cast an
Election Day ballot. In 1 of these Illinois jurisdictions, if poll workers
found both an Election Day and absentee ballot were cast, they were
instructed to void the absentee ballot so that it would not be counted. In
addition to matching signatures, election officials in a Nevada
jurisdiction we visited said that they used an

Uniformed and Overseas Citizens Absentee Voting

electronic poll book to manage absentee, early, and Election Day voting to
ensure that voters cast only one ballot. Once a ballot was cast in this
jurisdiction, the electronic poll book was annotated and the voter was not
allowed to cast another ballot.

Although election officials in the 20 jurisdictions mentioned above told
us that they had procedures in place designed to help prevent fraud during
the absentee voting process, election officials told us that they still
suspected instances of fraud. For example, in a Colorado jurisdiction we
visited, election officials told us that they referred 44 individuals who
allegedly voted absentee ballots with invalid signatures to the district
attorney for investigation. In a New Mexico jurisdiction that we visited,
election officials told us that organized third parties went door to door
and encouraged voters to apply for absentee ballots. Once these voters
received their ballots, according to election officials, the third parties
obtained the voters' names (in New Mexico this is public information,
according to such officials), and went to the voters' homes and offered to
assist them in voting the ballots. These election officials said that they
were concerned that the latter part of this activity might be intimidating
to voters and could result in voter fraud.

In general, the Uniformed and Overseas Citizens Absentee Voting Act
requires, among other things, that states permit absent uniformed services
members and U.S. citizen voters residing outside the country to register
and vote absentee in elections for federal office.7 In addition, states
also generally offer some measure of absentee voting for registered voters
in their states not covered under UOCAVA. The basic process for absentee
voting under UOCAVA is generally similar to that described in figure 18
for absentee voters not covered under UOCAVA in that UOCAVA voters also
must establish their eligibility to vote on their absentee ballot
application, and the ballot must be received by the voter's local
jurisdiction to verify it for counting. Election officials in some
jurisdictions we visited told us that they allow UOCAVA voters to submit a
voted ballot via facsimile-a method that might not be allowed for absentee
voters not covered under UOCAVA because of concerns about maintaining
ballot secrecy. In  6 jurisdictions we visited, election officials told
us that they require voters

742 U.S.C. S:S: 1973ff to 1973ff-6.

under UOCAVA to submit a form acknowledging that ballot secrecy could be
compromised when ballots are faxed.

One mechanism used to simplify the process for persons covered by UOCAVA
to apply for an absentee ballot is the Federal Post Card Application
(FPCA), which states are to use to allow such absentee voters to
simultaneously register to vote and request an absentee ballot. On our
survey of local jurisdictions, we asked if any problems were encountered
in processing absentee applications when the applicant used the FPCA. We
estimate that 39 percent of local jurisdictions received the FPCA too late
to process-a problem also encountered with other state-provided absentee
ballot applications. Table 10 shows our estimates of problems local
jurisdictions encountered when processing Federal Post Card Applications.
In addition, we asked about which problems were encountered most
frequently when the FPCA was used, and an estimated 19 percent of local
jurisdictions encountered the problem of receiving the FPCA too late to
process more frequently than other problems.

Table 10: Estimated Percentages of Local Jurisdictions, by Jurisdiction
Size, That Encountered Specific Voter Error Problems on Absentee
Applications That Used the Federal Post Card Application, November 2004
General Election

                                         All      Small     Medium      Large 
Problems                    jurisdictions (< 10,000) (10,000 to (>100,000) 
                                                        100,000)   
Application received too               39         32         53         70 
latea                                                           
Missing or inadequate                  31         24         41         75 
voting residence addressa                                       
Applied to wrong                       29         25         36         60 
jurisdictionb                                                   
Missing or inadequate                  24         17         36         67 
voting mailing addressa                                         
Missing or illegible                   21         16         27         57 
signatureb                                                      
Application not witnessed,             13         13         13         14 
attested, or notarized                                          
Did not meet excuse                     3          2          4         13 
required by state law                                           

Source: GAO analysis of local election jurisdiction survey data.

aAll size categories are statistically different from one another. bLarge
jurisdictions are statistically different from small and medium
jurisdictions.

Also, uniformed services voters and U.S. citizen voters residing outside
of the country are allowed to use the Federal Write-In Absentee Ballot to
vote for federal offices in general elections. This ballot may be used
when such voters submit a timely application for an absentee ballot (i.e.,
the application must have been received by the state before the state
deadline or at least 30 days prior to the general election, whichever is
later) but do not receive a state absentee ballot. Some states' absentee
ballot application forms included serving in a uniformed service or
residing outside the country as excuses for voting absentee. According to
our state survey,  4 states (Minnesota, Missouri, Oklahoma, and Rhode
Island) reported that they require attestation by a notary or witness for
a voter's signature on voted mail-in absentee ballots but do not require
uniformed service voters and U.S. citizen voters outside the country to
provide this on their voted ballots.

For the 2004 November general election, according to our state survey,  9
states reported having absentee ballot deadlines for voters outside the
United States that were more lenient than the ballot deadlines for voters
inside the United States. Table 11 lists these 9 states and the difference
between the mail-in ballot deadline from inside the United States and the
mail-in absentee ballot deadline from outside the United States.

Table 11: States Reporting Differing Mail-in Absentee Ballot Deadlines
from Inside and Outside the United States, November 2004 General Election

Mail-in absentee ballot deadline from
State         inside the United States    Mail-in absentee ballot deadline 
                                             from outside the United States   
Alaska         10 days after Election Day 15 days after Election Day and   
                              and postmarked postmarked by Election Day       
                 by Election Day             
Arkansas      Election Day                10 days after Election Day       
                                             No later than 10 days after      
Florida       Election Day                Election Day if postmarked or    
                                             signed and                       
                                             dated by Election Day (federal   
                                             races only)                      
Louisiana     1 day before Election Day   Election Day                     
Maryland      1 day after Election Day if 10 days after Election Day and   
                 postmarked                  postmarked before Election Day   
                 before Election Day         
Massachusetts Election Day                10 days after Election Day and   
                                             postmarked by Election Day       
Ohio          Election Day                10 days after Election Day       
Pennsylvania  4 days before Election Day  Deadline extended per court      
                                             order for November 2004 general  
                                             election for not only absentee   
                                             ballots from outside the United  
                                             States                           
                                             but also for those voters        
                                             covered by UOCAVA, including     
                                             domestic                         
                                             uniformed service members, who   
                                             are nonetheless absent from the  
                                             place of residence where they    
                                             are otherwise qualified to vote  
Texas         Election Day                5 days after Election Day        

Source: GAO 2005 survey state election officials.

HAVA amended the UOCAVA to, among other things, extend the period of time
that can be covered by a single absentee ballot application-the Federal
Post Card Application-by absent uniformed service voters and

Page 129 GAO-06-450 Elections

citizen voters residing outside the United States from the year during
which the application was received to a time period covering up to the two
next regularly scheduled general elections for federal office. To
illustrate, if uniformed service voters or civilian voters residing
outside the United States submitted a completed FPCA in July 2004, they
would have been allowed to automatically receive ballots for the next two
federal general elections, including those held in 2004 and 2006. (See
fig. 24 for an example of the FPCA used in 2004.)

In 4 local jurisdictions we visited, election officials told us that the
amendment described above may present a challenge for successfully
delivering absentee ballots to the uniformed services members because they
tend to move frequently. For example, in a North Carolina jurisdiction
that we visited, election officials stated that addresses on file for such
voters at the time of the November 2004 general election may be no longer
correct and that mail sent to these voters could be returned as
undeliverable. Also, in 1 jurisdiction in Georgia that we visited,
election officials told us that they were concerned that many of the
absentee ballots sent in subsequent general elections would be returned as
undeliverable. In an Illinois jurisdiction we visited, elections officials
expressed concerns about paying the postage for mail that may be
undeliverable will be a challenge in future years. Also, we noted in our
March 2006 report on election assistance provided to uniformed service
personnel, that one of the top two reasons for disqualifying absentee
ballots for UOCAVA voters was that the ballots were undeliverable.8

The Federal Post Card Application was revised in October 2005, after the
November 2004 general election, and now allows overseas military and
civilians to designate the time period for which they want to receive
absentee ballots. (See figure 24 for the revised FPCA.) Those who do not
wish to receive ballots for two regularly scheduled general elections can
designate that they want an absentee ballot for the next federal election
only and then complete the form and request a ballot for each subsequent
federal election separately. The FPCA used at the time of the November
2004 election did not allow overseas military and civilian voters to make
this designation.

8See GAO, Elections: Absentee Voting Assistance to Military and Overseas
Citizens Increased for the 2004 General Election, but Challenges Remain,
GAO-06-521 (Washington, D.C.: April 7, 2006), for more information.

Page 130 GAO-06-450 Elections

Figure 24: The Federal Post Card Application Used before and after the November
                             2004 General Election

 Source: U.S. Department of Defense Federal Voting Assistance Program Web site.

Even with the revised FPCA, some applications might not have this box
checked, and jurisdictions could continue to have absentee ballots
returned as undeliverable. In an attempt to mitigate these problems,
election officials in 3 local jurisdictions we visited told us that they
planned several activities in an attempt to maintain and update the
addresses of uniformed services voters and civilian voters residing
outside the country. In a Washington jurisdiction we visited, election
officials told us that they began requesting e-mail addresses from such
voters so that any problems with these applications or ballots could be
corrected more efficiently. In previous elections, when e-mail addresses
were not available, elections officials in this jurisdiction told us that
many absentee applications and ballots sent to uniformed services members
and civilian voters residing outside the United States were often returned
as undeliverable. In a Georgia jurisdiction that we visited, election
officials said that they planned to create a subsystem within their voter
registration system. This subsystem will, according to the election
officials, allow staff in the election office to produce a form letter for
each uniformed services voter that will verify the voter's current
address. The election officials also told us letters will be mailed in
January asking the voter to contact the jurisdiction to confirm that he or
she continues to reside at the address on the letter. If the jurisdiction
does not receive confirmation from the uniformed services voter, the
election officials told us that they will contact the Federal Voting
Assistance Program (FVAP) for assistance in locating the voter.9 In an
Illinois jurisdiction we visited, election officials stated that they plan
to canvass all uniformed services members and civilians residing outside
the United States who are registered in the state in 2006. Election
officials in this jurisdiction told us that they had approximately 7,400
such registered voters who completed the FPCA and that the jurisdiction
planned to canvass these voters to confirm that they continued to reside
at the address on the FPCA. This jurisdiction expects that as many as half
of these canvass cards will be returned as undeliverable. Once the cards
are returned, state law allows those voters whose canvass cards are
returned to be deleted from the voter registration list, according to the
election officials.

9FVAP provides overseas miliary and civilian voters a broad range of
nonpartisan information and assistance to facilitate their participation
in the electoral process.

Page 132 GAO-06-450 Elections

                                  Early Voting

Early voting is another way to provide registered voters with the
opportunity to cast ballots prior to Election Day. However, conducting
early voting is generally more complicated for election officials than
conducting Election Day voting. In the jurisdictions we visited in 7
states with early voting, election officials described early voting as
generally inperson voting at one or more designated polling locations
usually different from polling locations used at the precinct level on
Election Day. The voting may or may not be at the election registrar's
office. Early voting is distinct from in-person absentee voting in that
in-person absentee voters usually apply for an absentee ballot at the
registrar's office and vote at the registrar's office at that time. Also,
early voting usually does not require an excuse to vote, which some states
require for absentee voting, and in the jurisdictions we visited in 7
states with early voting, it was usually offered for a shorter period of
time than absentee voting.10 The time frame allowed for absentee voting
was almost always at least twice as long as for early voting. For example,
election officials in the Colorado jurisdictions we visited said that they
allow 30 days for absentee voting and 15 days for early voting. In the
jurisdictions we visited in 7 states with early voting, election officials
said early voting is similar to Election Day voting in that the voter
generally votes using the same voting method as on Election Day. However
they added that it differs from Election Day voting in that voters can
vote at any early voting polling location because all early voting
locations have access to a list of all registered voters for the
jurisdiction (not just precinct specific) and can provide voters with
appropriate ballots that include federal, state, and precinct-specific
races.

10The seven states are Colorado, Florida, Georgia, Kansas, Nevada, New
Mexico, and North Carolina.

Page 133 GAO-06-450 Elections

Proponents argue that early voting is convenient for voters and saves
jurisdictions money by reducing the number of polling places and poll
workers needed on Election Day, and also provides the voter with more
opportunity to vote. Opponents counter that those who vote early do so
with less information than Election Day voters, and there is no proof that
early voting increases voter turnout. Statistics on voter turnout for
early voting can be difficult to come by, partly because some states and
localities combine early and absentee voting numbers.11 Nevertheless,
early voting in certain jurisdictions appears to be popular with voters
and on the rise. In a New Mexico jurisdiction, election officials told us
that early voting accounted for about 34 percent of the ballots cast in
that jurisdiction. In North Carolina and Colorado elections jurisdictions
we visited, election officials said that early voters cast about 35 and 38
percent of the jurisdictions' total votes in the November 2004 election,
respectively. In a Nevada jurisdiction we visited, election officials told
us that the percentage of voters who voted early steadily increased over
time. The officials said that in 1996, about 17 percent of voters voted
early; in 2000, 43 percent voted early; and in the November 2004 general
election, about 50 percent (271,500) of their voters voted early.

Our prior work on the 2000 general election did not identify states that
offered early voting as we have defined it. Rather, we reported on
absentee and early voting together. Thus, we are unable to identify the
change in the number of states that offered early voting for the November
2000 general election and the November 2004 general election. We describe
the availability of early voting throughout the nation and the challenges
and issues that election officials encountered in the November 2004
general election as they conducted early voting in selected jurisdictions.
Many early polling locations in Florida and elsewhere received media
publicity about voters standing in long lines and waiting for long periods
of time to vote early. In half of the local election jurisdictions we
visited, election officials described encountering challenges that
included long lines, and some identified challenges dealing with
disruptive third-party activities at the polls.

11The statutory framework for early voting and absentee voting varies
among the states- with some states, for example, providing early voting
within the context of the state's absentee voting provisions, while others
provide for absentee voting in the context of the state's early voting
provisions.

Page 134 GAO-06-450 Elections

Early Voting Appears to Be Gaining in Popularity

For the November 2004 general election, in our state survey, 24 states and
the District of Columbia reported offering early voting. In addition,  2
states-Illinois and Maine-reported, in our state survey, that they had
enacted legislation or taken executive action since November 2004 to
provide for early voting in their states. Another 7 states reported that
with respect to early voting, they (1) had legislation pending, (2)
considered legislation in legislative session that was not enacted, or (3)
had an executive action that was pending or was considered. Figure 25
shows where early voting was provided for the November 2004 general
election.

Figure 25: Twenty-four States and the District of Columbia Reported
Providing Early Voting as an Option in the November 2004 General Election

         Sources: GAO survey of state election officials, MapArt (map).

The Number of Days and Hours to Conduct Early Voting

On the basis of our survey of local jurisdictions, we estimate 23 percent
of jurisdictions were in states that offered early voting. Furthermore, we
estimate that 16 percent of small jurisdictions, 40 percent of medium
jurisdictions, and 52 percent of large jurisdictions were in states that
offered early voting. Small jurisdictions are statistically different from
both medium and large jurisdictions.

The number of days that early voting was available in these 24 states and
the District of Columbia varied. In some cases, early voting was allowed
no sooner than a day or a few days prior to Election Day, while in other
cases voters had nearly a month or longer to cast an early ballot. Table
12 shows the range of days for early voting among the states and the
District of Columbia that reported providing early voting for the November
2004 election.

Table 12: Early Voting Period in 24 Early Voting States and the District
of Columbia, November 2004 General Election

                         Page 137 GAO-06-450 Elections

States that reported providing
early voting              Statutory early voting perioda                   
Alaska                    16 days                                          
Arizona                   At least 30 days                                 
Arkansas                  15 days                                          
California                30 days                                          
Colorado                  15 days                                          
Florida                   15 days                                          
Georgia                   5 days                                           
Hawaii                    10 days                                          
Idaho                     Beginning of early voting period is unspecified- 
                               early voting period ends on the day before the 
                             election                                         
Iowa                      Up to 40 days                                    
Kansas                    Up to 20 days                                    
Louisiana                 12 to 6 days                                     
Montana                   45 days                                          
Nebraska                  Beginning of early voting period is unspecified- 
                             early voting period ends by close of polls on    
                             Election Day                                     
Nevada                    14 days                                          
New Mexico                17 days                                          
North Carolina            Up to 19 days                                    
North Dakota              15 days                                          
Oklahoma                  3 days                                           
Tennessee                 Up to 20 days                                    
Texas                     14 days                                          
Utah                      Beginning of early voting period is unspecified- 
                               early voting period ends on the day before the 
                             election                                         
Vermont                   Unspecified                                      
West Virginia             18 days                                          
District of Columbia      15 days                                          

Source: GAO analysis.

aFor the purposes of this table, the phrase "early voting period" refers
to the period of time during which in-person early ballots may be cast in
elections for federal office. Different time periods may be applicable
under state laws with respect to early ballot applications and nonfederal
elections.

On the basis of our survey of local jurisdictions, we estimate that  75
percent of the jurisdictions that offered early voting offered it for 2 or
more weeks prior to Election Day. Figure 26 shows the estimated percentage
of local jurisdictions that offered early voting for various time periods.

Figure 26: Estimated Percentage of Local Jurisdictions That Offered Early
Voting for Various Time Periods, November 2004 General Election

Offered early voting 1-13 days prior to election day

Offered early voting 28 days prior to election days

Offered early voting 14-27 days prior to election day

Source: GAO 2005 survey of local election jurisdictions.

Note: The 95 percent interval for these percentages is +/- 10 percentage
points or less. Figures do not total 100 because of rounding.

Among the local jurisdictions that we visited in the 7 states that
provided early voting, we found that the shortest time frame allowed for
early voting was in Georgia, which had 5 days, and the longest time frame
allowed for early voting was in New Mexico, with 28 days. Furthermore, in
the local jurisdictions we visited in the 7 states that provided early
voting, election officials supplied information on early voting hours that
ranged from weekday business hours to those that included weekends and
evenings. For more details on the characteristics of early voting sites we
visited, see appendix VII.

During the course of our work, a limited review of state statutes showed,
for example, that Nevada statute requires early voting polling places be
open Monday through Friday, 8 a.m. to 6 p.m., during the first week of
early voting and possibly to 8 p.m. during the second week, dependent upon
the county clerk's discretion. In addition, under the Nevada provision,
polling places must be open on any Saturdays within the early voting
period from 10 a.m. to 6 p.m., and may be open on Sundays within the early
voting period dependent upon the county clerk's discretion. Under these
provisions, the early voting period is to begin the third Saturday prior
to an election and end the Friday before Election Day. Similarly, Oklahoma
statute provides that voters be able to cast early ballots from 8 a.m. to
 6 p.m. on the Friday and Monday immediately before Election Day, and
from 8 a.m. to 1 p.m. on the Saturday immediately before Election Day.
Some states' statutes are less prescriptive, such as those of Florida,
where the statute specifies that early voting should be provided for at
least  8 hours per weekday during the early voting period, and at least 8
hours in the aggregate for each weekend during the early voting period,
without specifying the specific hours such voting is to be offered. Other
states, such as Kansas, however, do not specify in statute the hours for
voting early.12 Kansas statute, in general, leaves it to county election
officials to establish the times for voting early. Officials at some local
jurisdictions we visited said that their hours of operations were set
based on the hours of the election office or by the hours of the facility
that was hosting early voting such as a shopping mall or a library.

According to our survey of local jurisdictions, an estimated 34 percent13
of local jurisdictions that provided early voting for the November 2004
general election offered early voting during regular business hours (e.g.,
from  8 a.m. until 4 p.m.) on weekdays, and 16 percent14 offered early
voting during regular business hours on weekdays and during other hours.
Other hours included weekday evenings (after 4 p.m. or 5 p.m. until 7 p.m.
or  9 p.m.) and Saturdays (all day) and Sundays (any hours) for about 2
percent of the jurisdictions, respectively.

12Kansas election officials reported that early voting in Kansas is called
advance voting. Any registered voter may choose to vote an advance ballot
by mail or in person with in-person voting in the county election office
up to 20 days before any election, according to the election officials.

13The 95 percent confidence interval for this percentage is +/- 9
percentage points.

14The 95 percent confidence interval for this percentage is +7 or -5
percentage points.

Page 139 GAO-06-450 Elections

Determining Number and Types of Early Voting Locations

As with early voting time frames, some states reported having requirements
for local election jurisdictions regarding the number of early voting
locations. In our state survey, 17 of the 25 entities (including 24 states
and the District of Columbia) that reported offering early voting for the
November 2004 general election also reported having requirements for local
jurisdictions regarding the number or distribution of early voting
locations. Kansas election standards, for example, provide for one such
voting location per county unless a county's population exceeds 250,000,
in which case the election officer may designate additional sites as
needed to accommodate voters.

Election officials in 1 jurisdiction we visited said that state statute
determined the number of locations, while election officials in 13 other
jurisdictions told us they decided the number of locations. For example,
New Mexico's early voting statutory provisions specifically require that
certain counties with more than 200,000 registered voters establish not
fewer than 12 voting locations each. During our site visits, we asked
jurisdictions how they determined the number of early voting locations. In
a Nevada jurisdiction that we visited, election officials said that the
number of locations was determined by the availability of resources such
as fiscal and manpower needs. In a Colorado jurisdiction we visited, an
election official said he would like to have had more early voting
locations but could not because the jurisdiction did not have the funds to
pay for additional costs associated with additional sites, such as the
cost for computer connections needed for electronic voter registration
list capability. In a North Carolina jurisdiction we visited, election
officials said that they had only one early voting location because they
did not have election staff that would be needed to manage another site.

Conducting Early Voting In many ways, early voting is conducted in a
manner substantially similar to Election Day voting in that polling
locations are obtained, workers are recruited to staff the sites for each
day polling locations are to be open, and voting machines and supplies are
delivered to the polling locations. However, as described by election
officials in jurisdictions we visited that had early voting, early voting
differs from Election Day voting in that staff are generally required to
perform their voting day-related duties for more than 1 day, and staff
generally do not use poll books to identify eligible voters and check them
in. Instead, as described by some of these jurisdictions, early voting
staff usually access the jurisdiction's voter registration list to
identify eligible voters and to indicate the voter voted early to preclude
voting on Election Day or by absentee ballot. Also, election officials
told us that, generally, staff must possess some computer skills and need
to be trained in using the jurisdiction's voter registration system.
Furthermore, staff must be aware that ballots are specific to the voter's
precinct.

In our nationwide survey of local election jurisdictions, we asked about
the type of staff who worked at early voting polling places. According to
our survey for the November 2004 general election, local election
jurisdictions relied on permanent election jurisdiction staff most often
to work at early voting polling locations. As table 13 shows, we estimate
30 percent of local jurisdictions offered early voting using only
permanent election jurisdiction staff to work at the early voting polling
places; we estimate that 14 percent of local jurisdictions used poll
workers exclusively; and we estimate  14 percent used other staff (e.g.,
county or city employees).

Table 13: Estimated Percentages of Local Jurisdictions That Used Various
Staff Mixes at Early Voting Locations, November 2004 General Election

Staff                                                              Percent 
Permanent staff                                                        30a 
Poll workers                                                          14 b 
Other (respondents wrote in other categories)                          14c 
Permanent and temporary part-time staff                                 7d 
Permanent staff and poll workers                                         4 
Permanent, temporary full-time, and temporary part-time staff and        4 
poll workers                                                       
Permanent and temporary full-time staff                                  4 
Permanent and temporary part-time staff and poll workers                 4 
Temporary part-time staff and poll workers                               3 
Temporary full-time staff and poll workers                               2 
Permanent, temporary full-time, and temporary part-time staff            2 
Temporary full-time and temporary part-time staff and poll workers       1 

Source: GAO 2005 survey of local election jurisdictions. aThe 95 percent
confidence interval for this percentage is +/- 9 percentage points. bThe
95 percent confidence interval for this percentage is +7 or -6 percentage
points. cThe 95 percent confidence interval for this percentage is +8 or
-6 percentage points. dThe 95 percent confidence interval for this
percentage is +6 or -4 percentage points.

Election officials at 11 jurisdictions we visited emphasized the
importance of staffing early voting locations with experienced staff such
as election office staff or experienced and seasoned poll workers. Even
with experienced staff working early voting locations, election officials
at local jurisdictions we visited mentioned that staff were required to
take training and were provided tools to help them perform their duties.

In our nationwide survey, we asked local jurisdictions that provided early
voting about the ways that staff were trained for early voting. As shown
in table 14, the majority of jurisdictions used methods, such as providing
a checklist of procedures, written guidance for self-study or reference,
and quick reference materials for troubleshooting, to train early voting
staff.

Table 14: Estimated Percentages of Local Jurisdictions That Used Various
Methods to Train Early Voting Staff, November 2004 General Election

Percentage of local Ways that early voting staff were trained
jurisdictionsa

Provided a checklist of procedures

Provided written guidance for self-study or reference

Provided quick reference materials for troubleshooting

Provided classroom training

Viewed training video

Source: GAO 2005 survey of local election jurisdictions.

aThe 95 percent confidence interval for all percentages is +/- 11
percentage points or less.

Local jurisdictions could do more than one of the above ways to train
early voting staff. On the basis of our local survey, we estimate that 14
percent of local jurisdictions used classroom training, written guidance
for self-study or reference, a checklist of procedures, and quick
reference materials for troubleshooting to train early voting staff.15

When asked about what worked particularly well during early voting,
election officials in 1 jurisdiction we visited in Colorado said that that
they provided 8 hours of training and had on-site supervision that they
thought contributed to a successful early voting experience. The election
officials also said they used a feature in their electronic poll book
system to track

15 The 95 percent confidence interval for this percentage is +8 or -6
percentage points.

the number of ballots used at each site to determine whether sites had
adequate inventories of ballots. The program for the poll book system had
an alarm that went off if any site was running low on ballots, according
to these election officials. Two other jurisdictions we visited in Kansas
and Florida noted the importance of having experienced staff for early
voting, with the election officials in 1 Kansas jurisdiction saying that
designating a group of workers to work on early voting helped the process
run effectively and the election officials in 1 Florida jurisdiction
saying that having the supervisor of elections office staff on site to
support early voting helped make the process work well.

Some Local Jurisdictions We When asked about challenges with early voting
faced during the November Visited Encountered Long lines 2004 general
election, in half of the local jurisdictions we visited that Resulting
from Larger than offered early voting election officials identified long
lines as a major Expected Early Voter Turnout challenge. Election
officials at 5 local jurisdictions we visited said that they

had not anticipated the large number of voters who had turned out to vote

early. Officials attributed challenges handling the large number of voters

and resulting long lines to problems with technology, people, and

processes. Election officials at local jurisdictions we visited made the

following comments:

     o Election officials in one Florida jurisdiction we visited said that
       their jurisdiction faced more early voters than anticipated and this
       fact, coupled with slowness in determining voter eligibility, resulted
       in long lines. They said that on the first day of early voting, staff
       was unable to access the voter registration list because laptops were
       not functioning properly. To address the problem, a worker at the
       early voting location paired with another worker, who called the
       supervisor of elections office to obtain voter registration
       information and provide information on the voter seeking to vote
       early.
          * An election official in another Florida jurisdiction said that
            while state law provides for early voting in the main office of
            the supervisor of elections, other locations may be used only
            under certain conditions. For example, in order for a branch
            office to be used, it must be a full-service facility of the
            supervisor and must have been designated as such at least 1 year
            prior to the election. In addition, a city hall or public library
            may be designated as an additional early voting location, but
            only if the sites are located so as to provide all voters in the
            county an equal opportunity to cast a ballot, insofar as is
            practicable. The official thought more flexibility was needed to
            allow him to either have more early voting locations or use other
            types of
          * facilities, such as a local community center, that could
            accommodate more voters.
     o An election official in a Nevada jurisdiction we visited said that the
       jurisdiction's process flow was inadequate to handle the large turnout
       for early voting. The election official said that the jurisdiction had
       not planned sufficiently to manage the large turnout for early voting
       and did not have enough staff to process voters. The election official
       said that in the future, he will hire temporary workers and will
       assign one person to be in charge of each process (e.g., checking in
       voters, activating the DRE machine, etc.) In addition, the election
       official said that, in hindsight, he made a questionable decision to
       close all but two early voting locations for the last day of early
       voting. The closing of all but two locations on the last day of early
       voting coincided with a state holiday so children were out of school.
       The decision to close all but two locations caused 3 to 3 1/2 hours of
       wait time, with parents waiting in line with children. The election
       official said he has set a goal for the future that no wait time
       should be longer than half an hour.16

To address challenges related to heavy early voter turnout, election
officials in 1 Nevada jurisdiction said they have gradually added new
early voting locations each year to keep up with the increasing number of
people who vote early. In a New Mexico jurisdiction we visited, election
officials said that they used a smaller ratio of voters to machines than
required by state statute. According to these election officials, the
state required at least one machine for every 600 voters, and during early
voting, the election officials said they used one machine for every 400
voters registered in the jurisdiction.

In 1 Colorado jurisdiction we visited, election officials said that they
addressed the challenge of long lines by having greeters inform voters
about the line and make sure the voters had required identification with
them. They said they provided equipment demonstrations and passed out
sample ballots so people in line could consider their choices, if they had

16The election official said that he closed all but two locations because
he wanted to have enough time before Election Day to make sure that he got
all of the data off the DRE machines (which were new) and have them ready
to be transported to Election Day polling places.

Page 144 GAO-06-450 Elections

Some Jurisdictions We Visited Encountered Challenges Dealing with
Disruptive Third-Party Activities

not already done so. They also said they offered people in line the option
of absentee ballot applications.

In 3 jurisdictions we visited, election officials stated that they
encountered challenges dealing with disruptive third-party activities at
early voting sites. In particular, concerns were raised about various
groups attempting to campaign or influence voters while the early voters
waited in line. State restrictions on various activities in or around
polling places on Election Day include prohibitions relating to, for
example, the circulation of petitions within a certain distance of a
polling place, the distribution of campaign literature within a certain
distance of the polls, the conduction of an exit or public opinion poll
within a certain distance of the polls, and disorderly conduct or violence
or threats of violence that impede or interfere with an election. Election
officials in 1 jurisdiction we visited stated that campaign activities too
close to people waiting in line were a concern to the extent that police
were called in to monitor the situation at one early voting location.
Election officials in a Florida jurisdiction we visited said that they
were concerned about solicitors, both candidates and poll watchers,
approaching people waiting in line to vote early and offering them water
or assistance in voting.17 While Florida's statutory provisions in place
for the November 2004 general election contained restrictions of various
activities in or around polling places on Election Day, such provisions
did not explicitly address early voting sites. Amendments to these
provisions, effective January 2006, among other things, explicitly applied
certain restrictions of activities in or around polling places to early
voting areas. With respect to poll watchers, these amendments also
prohibit their interaction with voters to go along with the pre-existing
prohibition on obstructing the orderly conduct of any election by poll
watchers.

Making voting easier prior to Election Day has advantages for voters and

Concluding

election officials, but also presents challenges for elections officials.
Many

17Florida statutory provisions in place for the November 2004 general
election provided, in part, that each political party and each candidate
may have one watcher in each polling room at any one time during an
election. Such statutory provisions further provided, among other things,
that no watcher shall be permitted to come closer to the official's table
or the voting booths than is reasonably necessary to properly perform his
or her functions, but each shall be allowed within the polling room to
watch and observe the conduct of electors and officials.

states and local jurisdictions appear to be moving in the direction of
enabling voters to vote before Election Day by eliminating restrictions on
who can vote absentee and providing for early voting. Many states allowed
voters to use e-mail and facsimiles to request an absentee ballot
application and, in some cases, to return applications. To the extent that
large numbers of voters do vote absentee or early, it can reduce lines at
the polling place on Election Day and, where permitted by state law, ease
the time pressures of vote counting by allowing election officials to
count absentee and early votes prior to Election Day.

However, there are also challenges for election officials. An estimated 
55 percent of jurisdictions received absentee ballot applications too late
to process, and an estimated 77 percent received ballots too late.
Although we do not know the extent of these problems in terms of the
number of applications and ballots that could not be processed, the
estimated number of jurisdictions encountering these problems may be of
some concern to state and local election officials. Absentee application
deadlines close to Election Day provide citizens increased time to apply
to vote absentee. However, the short time period between when applications
are received and Election Day may make it difficult for election officials
to ensure that eligible voters receive absentee ballots in time to vote
and return them before the deadline for receipt at election offices.

Voter errors on their absentee applications and ballots also create
processing problems for election officials. These include missing or
illegible signatures, missing or inadequate voting residence addresses,
and missing or incomplete witness information for a voter's signature or
other information. In addition, mail-in absentee ballots are considered by
some to be particularly susceptible to fraud. This could include such
activities as casting more than one ballot in the same election or someone
other than the registered voter completing the ballot. Despite efforts to
guard against such activities, election officials in some of the
jurisdictions we visited expressed some concerns, particularly regarding
absentee voters being unduly influenced or intimated by third parties who
went to voters' homes and offered to assist them in voting their ballots.
Some election officials expressed similar concerns about the influence of
third parties on early voters waiting in line who were approached by
candidates and poll watchers. After this happened in Florida in November
2004, the state amended its election provisions to prohibit such activity
with respect to early voters.

Getting absentee ballots to uniformed service personnel and overseas
citizens is a continuing challenge. UOCAVA permitted such voters to
request an absentee ballot for the upcoming election, and HAVA extended
the covered period to include up to two subsequent general elections for
federal office. Because the duty station of uniformed service personnel
may change during the period covered by the absentee ballot requests,
election officials in jurisdictions we visited were concerned that they
have some means of knowing the current mailing address. Some jurisdictions
are taking action to ensure that they have the correct address for sending
absentee ballots for the November 2006 election, such as requesting e-mail
addresses that can be used to obtain the most current address information
prior to mailing the absentee ballot. To the extent there are problems
identifying the correct address, uniformed service personnel and overseas
civilians may either not receive an absentee ballot or receive it too late
to return it by the deadline required for it to be counted.

Chapter 4

Conducting Elections

Overview

Whether voters are able to successfully vote on Election Day depends a
great deal on the planning and preparation that occur prior to the
election. Election officials carry out numerous activities-including
recruiting and training poll workers; selecting and setting up polling
places; designing and producing ballots; educating voters; and allocating
voting equipment, ballots, and other supplies to polling places-to help
ensure that all eligible voters are able to cast a ballot on Election Day
with minimal problems.1 In our October 2001 comprehensive report on
election processes nationwide we described these activities as well as
problems encountered in administering the November 2000 general election.2
Since then, federal and state actions have been taken to help address many
of the challenges encountered in conducting the November 2000 general
election. However, reports after the November 2004 general election
highlighted instances of unprepared poll workers, confusion about
identification requirements, long lines at the polls, and shortages of
voting equipment and ballots that voters reportedly encountered on
Election Day. This chapter describes changes and challenges-both
continuing and new-that election officials encountered in preparing for
and conducting the November 2004 general election.

States and local jurisdictions have reported making changes since the
November 2000 general election as a result of HAVA provisions and other
state actions to improve the administration of elections in the United
States. In addition to establishing a commission-the U.S. Election
Assistance Commission-with wide-ranging duties that include providing
information and assistance to states and local jurisdictions-HAVA also
established requirements with respect to elections for federal office for,
among other things, certain voters who register by mail to provide
identification prior to voting; mandated that voting systems accessible to
individuals with disabilities be located at each polling place; and
required voter information to be posted at polling places on Election Day.
HAVA also authorized the appropriation of federal funds for payments to
states to implement these provisions and make other improvements to
election administration. Since the November 2000 general election, some
states

1Jurisdictions call their poll workers by different titles, including
clerks, wardens, election judges, inspectors, captains, and precinct
officers and often have a chief poll worker for each polling place.

2See GAO-02-3.

Page 148 GAO-06-450 Elections

HAVA Made Changes Intended to Improve Election Administration

have also reported making changes to their identification requirements for
all voters.

Election officials reported encountering many of the same challenges
preparing for and conducting the November 2004 general election as they
did in 2000, including recruiting and training an adequate supply of
skilled poll workers, locating a sufficient number of polling places that
met requirements, designing ballots that were clear to voters when there
were many candidates or issues (e.g., propositions, questions, or
referenda), having long lines at polling places, and handling the large
volume of telephone calls received from voters and poll workers on
Election Day. Election officials in some of the jurisdictions we visited
also reported encountering new challenges not identified to us in the 2000
general election with third-party (e.g., poll watchers, observers, or
electioneers) activities at polling places on Election Day. On the basis
of our survey of a representative sample of local election jurisdictions
nationwide and our visits to 28 local jurisdictions, the extent to which
jurisdictions encountered many of these continuing challenges varied by
the size of election jurisdiction. Large and medium jurisdictions-those
jurisdictions with over 10,000 people-generally encountered more
challenges than small jurisdictions. In most results from our nationwide
survey where there are statistical differences between the size categories
of jurisdictions, large jurisdictions are statistically different from
small jurisdictions.

HAVA established EAC to provide voluntary guidance and assistance with
election administration, for example, by providing information on election
practices to states and local jurisdictions and administering programs
that provide federal funds for states to make improvements to some aspects
of election administration.3 HAVA also added a new requirement for states
to in turn require certain first-time voters who register by mail who have
not previously voted in a federal election in the state to provide
identification prior to voting,4 and jurisdictions reported taking steps
to implement this requirement and inform voters about it. In addition,
HAVA includes provisions to facilitate voting for individuals with
disabilities, such as requirements for accessible voting systems in
elections for federal 

3HAVA section 201 et seq. (codified at 42 U.S.C. S: 15321 et seq.). 4HAVA
section 303(b) (codified at 42 U.S.C. S: 15483(b)).

Page 149 GAO-06-450 Elections

HAVA Established EAC to Provide Guidance and Assistance with Election
Administration

office.5 HAVA established voter information requirements at polling places
on the day of election for federal office6 and authorized the
appropriation of funding for payments to states to expand voter education
efforts.

HAVA established EAC, in part, to assist in the administration of federal
elections by serving as a national clearinghouse for information and
providing guidance and outreach to states and local officials.7 In our
October 2001 report on election processes, we estimated that on the basis
of our survey of local election jurisdictions in 2001, 40 percent of local
election jurisdictions nationwide were supportive of federal development
of voluntary or mandatory standards for election administration similar to
the voluntary standards available for election equipment. We also reported
in 2001 that some election officials believed that greater sharing of
information on best practices and systematic collection of information
could help improve election administration across and within states.8 To
assist election officials, since its establishment, EAC has produced two
clearinghouse reports, one of which covers election administration.9 EAC
released a Best Practices Toolkit on Election Administration on August 9,
2004, to offer guidance to election officials before the November 2004
general election. The document is a compilation of practices used by
election officials that covers topics such as voter outreach, poll
workers, polling places, and election operations. Of note, this
compilation provided election officials with a checklist for HAVA
implementation that covers identification for new voters, provisional
voting, complaint procedures, and access for individuals with
disabilities. EAC has made this guidance available to states and local
jurisdictions via its Web site and engaged in

5HAVA section 301(a)(3) (codified at 42 U.S.C. S: 15481(a)(3)).

6HAVA section 302(b) (codified at 42 U.S.C. S: 15482(b)).

7Prior to the establishment of EAC, the Federal Election Commission's
Office of Election Administration served as a national clearinghouse for
information regarding the administration of federal elections.

8See GAO-02-3.

9The second report covers voting by absent uniformed service voters and
citizens residing outside of the United States. On September 21, 2004, EAC
released Best Practices for Facilitating Voting by U.S. Citizens Covered
by UOCAVA, which was compiled in consultation with the U.S. Department of
Defense Federal Voting Assistance Program. This report describes methods
to enhance absentee voter registration by UOCAVA voters and ensure the
timely receipt of their ballots.

HAVA Added a New Requirement for Certain Voters, and Jurisdictions
Reported Taking Steps to Inform Voters

public hearings and outreach efforts to inform the election community
about the resource tool.

EAC also administers programs that provide federal funds for states under
HAVA to make improvements to aspects of election administration, such as
implementing certain programs to encourage youth to become involved in
elections; training election officials and poll workers; and establishing
tollfree telephone hotlines that voters may use to, among other things,
obtain general election information. The results of our state survey of
election officials show that as of August 1, 2005, most states reported
spending or obligating HAVA funding for a variety of activities related to
improving election administration. For example, 45 states and the District
of Columbia reported spending or obligating HAVA funding for training
election officials, and 32 states and the District of Columbia reported
spending or obligating funding to establish toll-free telephone hotlines.

As discussed in chapter 2, under HAVA, states are to require certain
firsttime voters who registered to vote by mail to provide identification
prior to voting.10 Voters who are subject to this provision are those
individuals who registered to vote in a jurisdiction by mail and have not
previously voted in a federal election in the state, or those who have not
voted in a federal election in a jurisdiction which is located in a state
that has not yet established a computerized voter registration list, as
required by HAVA. When voting in person, these individuals must (if not
already provided with their mailed application) present a current and
valid photo identification, or a copy of a current utility bill, bank
statement, government check, paycheck, or other government document that
shows the name and address of the voter. Under HAVA, voters at the polls
who have not met this identification requirement may cast a vote under
HAVA's provisional voting provisions. Additional information on
provisional voting processes and challenges is presented in chapter 5.

Election officials in 21 of the 28 jurisdictions we visited reported
encountering no problems implementing the HAVA first-time voter ID
requirement, and officials in some of these jurisdictions provided reasons

10HAVA requirements to present ID at the time of voting for mail
registrants are, in general, not applicable to certain types of voters
such as, for example, persons who have previously submitted required ID
with their mail registration, and those entitled to vote by absentee
ballot under UOCAVA.

Page 151 GAO-06-450 Elections

why there were no problems.11 For example, election officials in  2
jurisdictions in Colorado told us that they did not encounter
implementation problems because all voters, under state requirements, were
required to show identification. Election officials in some other
jurisdictions we visited reported that they took steps to inform voters of
the new HAVA ID requirement for such voters registering by mail. For
example, election officials in a jurisdiction in Ohio reported that they
contacted about 300 prospective voters twice, either by phone or by
letter, prior to the election to inform them that that they needed to show
identification. Figure 27 illustrates a poster used in a jurisdiction we
visited to inform prospective voters about the new identification
requirements.

11Election officials in 6 other jurisdictions we visited reported
encountering some problems, such as poll worker confusion or
administrative burdens associated with mailing postcards to all voters who
needed to show identification at polling places. Election officials in  1
jurisdiction we visited told us that this HAVA requirement was not
implemented until 2005.

Page 152 GAO-06-450 Elections

Figure 27: King County, Washington, Poster Used to Inform Prospective
Voters of New Identification Requirements

Source: GAO (photo).

HAVA Includes Provisions to Facilitate Voting for Individuals with
Disabilities

HAVA Requirements for Accessible Voting Systems

HAVA contains provisions to help facilitate voting for individuals with
disabilities,12 including requirements for the accessibility of voting
systems used in elections for federal office, effective January 1, 2006.
HAVA also authorized the appropriation of funding for payments to states
to improve the accessibility of polling places. In October 2001, we issued
a report that examined state and local provisions and practices for voting
accessibility, both at polling places and with respect to alternative
voting methods and accommodations.13 We reported in 2001 that all states
and the District of Columbia had laws or other provisions concerning
voting access for individuals with disabilities, but the extent and manner
in which these provisions addressed accessibility varied from state to
state. In addition, in our 2001 report we noted that various features of
the polling places we visited had the potential to prove challenging for
voters with certain types of disabilities. On the basis of our
observations on Election Day 2000, we also estimated that most polling
places in the contiguous United States had one or more physical features,
such as a lack of accessible parking or barriers en route to the voting
room, that had the potential to pose challenges for voters with
disabilities.14 Results from our 2005 surveys show that at the time of the
November 2004 general election, many states and local jurisdictions had
taken steps to meet HAVA's requirement for accessible voting systems, as
well as making other changes to help improve the accessibility of voting
for individuals with disabilities.

HAVA requires that, effective January 1, 2006, each voting system used in
a federal election must meet certain accessibility requirements. These
voting systems are required to provide individuals with disabilities with
the same opportunity for access and participation (including independence
and

12Other federal laws affecting voting for individuals with disabilities
include the Americans with Disabilities Act of 1990 (Pub. L. No. 101-336,
104 Stat. 327 (1990)), which, in general, provides civil rights protection
to individuals with disabilities similar to those provided to individuals
on the basis of race, color, sex, national origin, age, and religion, and
the Voting Accessibility for the Elderly and Handicapped Act (Pub. L. No.
98-435, 98 Stat. 1678 (1984)), which, in general, requires registration
facilities and polling places for federal elections to be physically
accessible to the elderly and those with disabilities.

13For more information, see GAO, Voters with Disabilities: Access to
Polling Places and Alternative Voting Methods, GAO-02-107 (Washington,
D.C.: Oct. 15, 2001).

14In our October 2001 report we reported that these potential impediments
would primarily affect individuals with mobility impairments. We also
reported that polling places generally provide accommodations, such as
curbside voting, voting stations designed for people with disabilities,
and voter assistance inside the voting room.

privacy) as for other voters. These HAVA requirements specify that such
accessibility include nonvisual accessibility for voters who are blind or
visually impaired. HAVA provides for the use of at least one DRE or other
voting system equipped for voters with disabilities at each polling place.
The results of our state survey show that as of August 1, 2005, 41 states
and the District of Columbia reported having laws (or executive action) in
place to provide each polling location with at least one DRE voting system
or other voting system equipped for individuals with disabilities by
January 1, 2006. Of the remaining 9 states, 5 reported having plans to
promulgate laws or executive action to provide each polling location with
at least one DRE voting system or other voting system equipped for
individuals with disabilities, and 4 reported that they did not plan to
provide such equipment or were uncertain about their plans.15

Some local election jurisdictions provided accessible voting machines at
polling places for the November 2004 general election. On the basis of our
survey of a representative sample of local election jurisdictions
nationwide, we estimate that 29 percent of all jurisdictions provided
accessible voting machines at each polling place in the November 2004
general election.16 Further, more large and medium jurisdictions provided
accessible voting machines than small jurisdictions. We estimate that  39
percent of large jurisdictions, 38 percent of medium jurisdictions, and 
25 percent of small jurisdictions provided accessible voting machines at
each polling place. The differences between both large and medium
jurisdictions and small jurisdictions are statistically significant.
Election officials from some small jurisdictions who provided written
comments on our survey questionnaire expressed concerns about how this
requirement would be implemented in their jurisdictions and whether
electronic voting machines were the best alternative. For example, one
respondent wrote: "We [live] in a small town ... and use paper ballots and
that has worked very well in the past and I believe will work very well in
the future. Voting machines should be decided on for much larger areas
with a lot more than our 367 population with 150 voters." Another wrote:
"We are a small rural

15The 5 states that reported having plans were Kansas, Louisiana, Ohio,
Oklahoma, and Wyoming. The 4 states that reported having no plans or were
uncertain about their plans were Delaware, Massachusetts, Missouri, and
Tennessee.

16Unless otherwise noted, the maximum sampling error for estimates of all
jurisdictions is plus or minus 5 percentage points (rounded); +/- 7
percentage points (rounded) for large population size jurisdictions; +/- 7
percentage points (rounded) for medium population size jurisdictions; and
+/- 5 percentage points (rounded) for small population size jurisdictions.

township with about 160 voters. Our 2004 election went well; as usual, we
had no problems. We use paper ballots. We have some concerns with the
implementation of HAVA. We are being forced to use expensive voting
machines that will require expensive programming for every election. We
are concerned about these costs.... If our limited budget can't afford
those expensive machines and programming, we may need to combine our
township polling place with another township-maybe several townships. The
additional driving to a different polling place miles away will discourage
voters from voting-particularly our elderly residents. So these efforts
(HAVA) to help voters will actually hinder voters."

In an effort to address these issues, Vermont, which has about 250 small
and medium election jurisdictions that use paper and optical scan ballots,
took an alternative approach to meeting the HAVA requirement, according to
an election official. Instead of providing one DRE machine for each of its
280 polling places, Vermont plans to implement a secure vote-by-phone
system that allows voters to mark a paper ballot, in private, using a
regular telephone at the polling place. According to the Vermont's
Secretary of State's Office, a poll worker uses a designated phone at the
polling place to call a computer system located at a secure location and
access the appropriate ballot for the voter. The computer will only permit
access to the system from phone numbers that have been entered into the
system prior to the election, and only after the proper poll worker and
ballot access numbers have been entered. The phone system reads the ballot
to the voter, and after the voter makes selections using the telephone key
pad, the system prints out a paper ballot that is automatically scanned by
the system and played back to the voter for verification. The voter may
then decide to cast the ballot or discard it and revote. The system does
not use the Internet or other data network, and it produces a
voter-verified paper ballot for every vote cast. In addition, according to
an election official, voters are able to dial into a toll-free telephone
number for at least 15 days prior to an election to listen to, preview,
and practice with the actual ballot they will vote on Election Day. This
is a way of providing a sample ballot to voters, as well as providing an
opportunity for voters to become familiar with using the telephone system.

Provisions for Polling Place Accessibility and Other Accommodations

For our October 2001 report on voters with disabilities, our analysis
included a review of state statutes, regulations, and written policies
pertaining to voting accessibility for all 50 states and the District of
Columbia, as well as policies and guidelines for a statistical sample of 
100 counties.17 As part of our 2005 surveys, we asked states to report on
provisions concerning accessibility and local jurisdictions whether they
provided accommodations or alternative voting methods for individuals with
disabilities in the November 2004 general election. While the
methodologies in the 2001 report and this report differ, results of our 
2005 surveys show that states and local jurisdictions have taken actions
to help improve voting for individuals with disabilities by, for example,
using HAVA funds, taking steps to help ensure accessibility of polling
places, and providing alternative voting methods or accommodations.

Most states reported that they had spent or obligated HAVA funding to
improve the accessibility of polling places, including providing physical
or nonvisual access. The results of our state survey of election officials
show that as of August 1, 2005, 46 states and the District of Columbia
reported spending or obligating HAVA funding for this purpose. For
instance, election officials in a local jurisdiction we visited in
Colorado told us they had used HAVA funds to improve the accessibility of
polling places by obtaining input from the disability community, surveying
the accessibility of their polling places, and reviewing the DRE audio
ballot with representatives of the blind community.

States and local jurisdictions reported taking a variety of actions
designed to help ensure that polling places are accessible for voters with
disabilities, including specifying guidelines or requirements, inspecting
polling places to assess accessibility, and reporting by local
jurisdictions on polling place accessibility to the state. In our October
2001 report on voters with disabilities, we noted that state involvement
in ensuring polling places are accessible and the amount of assistance
provided to local jurisdictions could vary widely. For example, in 2001 we
reported that 29 states had provisions requiring inspections of polling
places, and 20 states had provisions requiring reporting by local
jurisdictions.18 According to our 2005 state survey, 43 states and the
District of Columbia reported requiring or allowing inspections of polling
places, and 39 states and the District of

17See GAO-02-107. 18See GAO-02-107.

Page 157 GAO-06-450 Elections

Columbia reported that they required or allowed reporting by local
jurisdictions. From our local jurisdiction survey, we estimate that  83
percent of jurisdictions nationwide used state provisions to determine the
accessibility requirements for polling places. During our site visits to
local jurisdictions, we asked election officials to describe the steps
they took to ensure that polling places were accessible. Election
officials in many of the jurisdictions we visited told us that either
local or state officials inspected each polling location in their
jurisdiction using a checklist based on state or federal guidelines. For
example, election officials in the 4 jurisdictions we visited in Georgia
and New Hampshire told us that state inspectors conducted a survey of all
polling locations. Election officials in the 2 jurisdictions we visited in
Florida told us that they inspected all polling places using a survey
developed by the state. Appendix IX presents additional information about
state provisions for alternative voting methods and accommodations for the
November 2000 and 2004 general elections.

In addition to making efforts to ensure that polling places are
accessible, some local jurisdictions provided alternative voting methods
pursuant to state provisions (such as absentee voting) or accommodations
at polling places (such as audio or visual aids) that could facilitate
voting for individuals with disabilities. Table 15 presents results from
our survey of local election jurisdictions about the estimated percentages
of jurisdictions that provided alternative voting methods or
accommodations to voters for the November 2004 general election.

Table 15: Estimated Percentages of All Local Jurisdictions and by
Jurisdiction Size That Provided Alternative Voting Methods and
Accommodations for the November 2004 General Election

                                                                   Percentage
                                  Percentage of      of medium Percentage of  
                                                               large          
               Percentage of all small           jurisdictions  jurisdictions 
                                 jurisdictions                 
Provisiona      jurisdictions      (<10,000)     (10,000 to     (>100,000) 
                                                      100,000) 
Early voting b             27             20             41            49c 
Absentee voting                                             
(no excuse or                                               
an allowable                                                
excuse)                    86             84             91             89 
Permanent                                                   
absentee voting                                             
(for instance,                                              
absentee voting                                             
status for a                                                
time period or                                              
number of                  58            52d             74             68 
elections)b                                                 
Curbside voting            55            55d             54             60 
Audio or visual                                             
aids to assist                                              
voters with                                                 
disabilities                             38d                               
(magnifying                42                            50            57c
lens)e                                                      
Braille ballots                                                            
or voting                  15             15             16             10
methods                                                     
Large-font                                                                 
ballots or                 13             11             19             21
instructions                                                

HAVA Helped to Expand State and Local Jurisdictions' Voter Education
Efforts

Source: GAO 2005 survey of local election jurisdictions.

aSome provisions, such as early voting, might not be provided by some
jurisdictions because state provisions do not authorize them. Some
provisions may not be applicable for the type of voting system a
jurisdiction uses.

bThe differences between both large and medium jurisdictions and small
jurisdictions are statistically significant. cThe 95 percent confidence
interval for these percentages is +/- 8 percentage points. dThe 95 percent
confidence interval for these percentages is +/- 6 percentage points. eThe
difference between large and small jurisdictions is statistically
significant.

Election officials' efforts to educate citizens can help minimize problems
that could affect citizens' ability to successfully vote on Election Day.
Informing the public about key aspects of elections includes communicating
how to register, what opportunities exist to vote prior to Election Day,
where to vote on Election Day, and how to cast a ballot. This information
can be distributed through a number of different media, including signs or
posters, television, radio, publications, in-person demonstrations, and
the Internet. In our October 2001 report on election processes, we stated
that lack of funds was the primary challenge cited by election officials
in expanding voter education efforts. From our 2001 survey of local
election jurisdictions, we estimated that over a third of jurisdictions
nationwide believed that the federal government should 

HAVA Required Voter Information at Polling Places and Provided for Funding
to States

provide monetary assistance for voter education programs.19 Since the
November 2000 election, changes in voter education efforts include HAVA
requiring certain information to be posted at polling places and
authorizing the payment of federal funds to states to use for educating
voters, and states and local jurisdictions reported expansion of voter
education efforts.

To help improve voters' knowledge about voting rights and procedures, HAVA
required election officials to post voting information at each polling
place on the day of each election for federal office and authorized the
payment of funding to states for such purposes. This required voting
information includes a sample ballot, polling place hours, instructions on
how to vote, first-time mail-in instructions, and general information on
federal and state voting rights laws and laws prohibiting fraud and
misrepresentation. Results of our state survey of election officials show
that as of August 1, 2005, 40 states and the District of Columbia reported
spending or obligating HAVA funding for voting information, such as sample
ballots and voter instructions, to be posted at polling places. Election
officials in all 28 jurisdictions we visited told us they posted a variety
of voter information signs at polling places on Election Day 2004. Figure
28 illustrates examples of some of these signs.

19See GAO-02-3.

Figure 28: Examples of Voter Information Signs Posted at Polling Places for the
                         November 2004 General Election

Source: Clark County, Nevada, and Chicago, Illinois, posters (GAO photos).

HAVA also authorized the payment of funding for voter education programs
in general, and according to our state survey, as of August 1, 2005, 44
states and the District of Columbia reported spending or obligating HAVA
funding for these programs. For example, according to its HAVA plan,
Florida required local election officials to provide descriptions of
proposed voter education efforts, such as using print, radio, or
television to advertise to voters, in order to receive state HAVA funds in
fiscal years 2003 and 2004. Election officials in 2 jurisdictions we
visited in Florida provided us information about voter education campaigns
that they implemented. Election officials in 1 of these jurisdictions
reported designing election

Variety of Local Election Jurisdictions' Actions to Educate Prospective
Voters

advertisements to be shown on movie theater screens in the beginning of
the summer season; election officials in the other jurisdiction told us
they implemented a "Get Out the Vote" television advertising campaign with
a cable company intended to reach hundreds of thousands of households
during the weeks prior to the November 2004 general election.

More local election jurisdictions appear to have taken steps to educate
prospective voters prior to Election Day in 2004 than in 2000, and on the
basis of our 2005 survey of local jurisdictions, more large and medium
jurisdictions took these steps than small jurisdictions. In our October
2001 report on election processes, we noted that local election
jurisdictions provided a range of information to prospective voters
through multiple media. For example, on the basis of our 2001 survey of
local jurisdictions, we reported that between 18 and 20 percent of local
jurisdictions nationwide indicated they placed public service ads on local
media, performed community outreach programs, or put some voter
information on the Internet. On the basis of our 2005 survey, we estimate
that more jurisdictions provided these measures. For instance, we estimate
that  49 percent of all jurisdictions placed public service ads on local
media, and 43 percent of all jurisdictions listed polling places on the
Internet. However, increases in the overall estimates from the 2001 and
2005 surveys are, in part, likely due to differences in the sample designs
of the two surveys and how local election jurisdictions that were minor
civil divisions (i.e., subcounty units of government) were selected.
Because of these sample design differences, comparing only election
jurisdictions that are counties provides a stronger basis for making
direct comparisons between the two surveys' results.20 These county
comparisons show increases as well. For instance, for the November 2000
election, we estimate that 21 percent of county election jurisdictions
placed public service ads on local media, while for the November 2004
election, we estimate that 61 percent of county election jurisdictions
placed such ads.21

20For this reason, some estimates from the 2001 survey are slightly
different than the overall sample estimates provided in our prior report.
For these comparisons, the 95 percent confidence interval is +/- 5 percent
or less for the 2001 survey estimates and +/- 8 percent or less for the
2005 survey estimates. See appendix V for further details about the
sampling differences between the 2001 and 2005 local election jurisdiction
surveys.

21In addition, some of the increase may be because of changes in how the
question was worded. However, as noted above, HAVA authorized the payment
of funding to states for voter education programs, and according to our
survey of state election officials, most states reported obligating or
spending HAVA funds for voter education.

In our 2005 survey, we also looked at whether there were differences
between the size categories of jurisdictions, and generally, more large
jurisdictions provided voter education prior to Election Day than medium
and small jurisdictions. For instance, we estimate that 88 percent of
large jurisdictions, 46 percent of medium jurisdictions, and 38 percent of
small jurisdictions listed polling place locations on Internet Web sites.
Table 16 presents estimated percentages of jurisdictions that provided
various voter education steps prior to the November 2004 general election.

Table 16: Estimated Percentages of All Local Jurisdictions and by
Jurisdiction Size That Took Steps to Provide Voter Education Prior to the
November 2004 General Election

                                                                   Percentage
                                   Percentage of of medium     
                                           small jurisdictions  Percentage of 
                                                                        large 
                 Percentage of all jurisdictions    (10,000 to  jurisdictions 
Step              jurisdictions     (<10,000)      100,000)     (>100,000) 
Provided sample                                             
ballots, either                                             
by mail or by                                               
printing in                  87            86            91             86 
newspaper                                                   
Placed polling                                              
place locations                                             
on Internet                                                 
Web sitea                    43           38b            46             88 
Placed public                                               
service ads on                                              
local media,                                                
such as TV,                  49           43b            62             63 
radio, or                                                   
newspapersc                                                 
Conducted                                                   
outreach with                                               
local                                                       
organizations,                                              
such as political                                           
parties or                                                  
charitable or                                               
social groupsd               26            16            43             76 
Mailed                                                      
vote-casting                                                
instructions on                                             
using the                                                   
jurisdiction's                                              
voting system to                                            
registered                                                  
votersa                      12            11             9             28 
Demonstrations of                                           
voting equipment                                            
(at                                                         
county fairs,                28            21            41             61 
registrar's                                                 
office, public                                              
events)d                                                    
Provided                                                    
information to                                              
voters about                                                
their                                                       
specific polling             83            82            83             92 
place locatione                                             

Source: GAO 2005 survey of local election jurisdictions.

aThe differences between large jurisdictions and both medium and small
jurisdictions are statistically significant

bThe 95 percent confidence interval for these percentages is +/- 6
percentage points. cThe differences between both large and medium
jurisdictions and small jurisdictions are statistically significant.

dAll size categories are statistically different from one another. eThe
difference between large and small jurisdictions is statistically
significant.

Large jurisdictions may have provided voter education through multiple
media in order to reach a broader audience of prospective voters. For
instance, Web sites were used to provide information to voters by nearly
all large jurisdictions. On the basis of our 2005 survey of local
jurisdictions, we estimate that 93 percent of large jurisdictions, 60
percent of medium jurisdictions, and 39 percent of small jurisdictions had
a Web site. The differences between all size categories are statistically
significant. During our site visits, election officials in large
jurisdictions described a variety of voter education mechanisms used to
reach a number of prospective voters. For example, election officials in a
large Nevada jurisdiction we visited told us that their office partnered
with power, water, and cable companies to provide voter registration
information in subscribers' billing statements. Election officials in
other jurisdictions we visited reported using a variety of other media to
encourage participation or provide information to a broad audience of
prospective voters. For example, figure 29 illustrates a billboard,
cab-top sign, and milk carton used in local jurisdictions we visited.

Figure 29: Efforts to Inform Voters Prior to the November 2004 General Election

  Source (top): Broward County, Florida, (bottom left to right): Leon County,
            Florida; Clark County, Nevada, milk carton (GAO photo).

Some States Reported Changing Identification Requirements for All
Prospective Voters

Whether or not all voters should be required to show identification prior
to voting is an issue that has received attention in the media and reports
since the November 2000 general election. Recent state initiatives, such
as those in Georgia, that in general require voters to provide photo
identification, exemplify the challenge that exists throughout the
election process in maintaining balance between ensuring access to all
prospective voters and ensuring that only eligible citizens are permitted
to cast a ballot on Election Day. Results of our state and local
jurisdiction surveys show that while providing identification could be one
of several methods used to verify identity, it was not required by the
majority of states, nor was it the only way used to verify voters'
identities in the majority of local jurisdictions for the November 2004
election. Voter identification requirements vary in flexibility, in the
number and type of acceptable identification allowed, and in the
alternatives available for verifying identity if a voter does not have an
acceptable form of identification.

Results of our state survey of election officials show that for the
November 2004 general election 28 states reported that they did not
require all prospective voters to provide identification prior to voting
in person.22 Twenty-one states reported that they required all voters to
provide identification prior to voting on Election Day 2004. However, 14
of these states reported allowing prospective voters without the required
identification an alternative. In 9 of these 14 states the alternative
involved voting a regular ballot in conjunction with, for example, the
voter providing some type of affirmation as to his or her identity. For
example, Connecticut, in general, allowed voters who were unable to
provide required identification to swear on a form provided by the
Secretary of State's Office that they are the elector whose name appears
on the official registration list. Kentucky allowed an election officer to
confirm the identity of a prospective voter by personal acquaintance or by
certain types of documents if the prospective voter did not have the
required identification. The other 5 states reporting that they offered an
alternative did so through the use of a provisional ballot if a
prospective voter did not have the required identification. For the
November 2004 election, 5 of the 21 states that reported having
identification requirements also had statutory provisions requiring, in
general, that such identification include a

22Oregon conducts all-mail voting on Election Day; identification
requirements for in-person voting are not applicable. The District of
Columbia also reported on our survey that this requirement was not
applicable.

Page 166 GAO-06-450 Elections

photograph of the prospective voter.23 For the other 16 states that
reported requiring identification, there was a range of acceptable forms
of identification, including photo identification, such as a driver's
license, and other documentation, such as a copy of a government check or
current utility bill with a voter's name and address. Figure 30 presents
information on the identification requirements for prospective voters for
the November 2004 general election for all 50 states and the District of
Columbia.

23These states were Florida, Louisiana, Montana, South Carolina, and South
Dakota.

                Source: GAO survey of state election officials.

                         Page 168 GAO-06-450 Elections

In our nationwide survey, we asked local jurisdictions about how they
checked voters' identities, such as by asking voters to state their name
and address, verifying voters' signatures, or asking voters to provide a
form of identification or documentation. On the basis of this survey, we
estimate that 65 percent of all local jurisdictions checked voters'
identification as one way to verify their identities on Election Day.
However, in an estimated 9 percent of all jurisdictions, providing
identification was the only way voters could verify their identities.

Since the November 2004 general election, several states have reported
that they have considered establishing identification requirements for all
prospective voters, and some reported that they have implemented
requirements. Results of our state survey show that at the time of our
survey, 9 states reported having either considered legislation (or
executive action) or legislation (or executive action) was pending to
require voters to show identification prior to voting on Election Day. 24
Four states, at the time of our survey, reported having taken action since
November 2004 to require that voters show identification for in-person
Election Day voting. For example, changes in Arizona law and procedure
emanating from a November 2004 ballot initiative were finalized in 2005
after receiving approval from the Department of Justice. These Arizona
changes require voters to present, prior to voting, one form of
identification with the voter's name, address, and photo, or two different
forms of identification that have the name and address of the voter.
Indiana enacted legislation in 2005 requiring, in general, that voters
provide a federal- or state-of-Indiana-issued identification document with
the voter's name and photo prior to voting, whereas 2005 legislation in
New Mexico and Washington imposed identification requirements but allowed
prospective voters to provide one of several forms of photo or nonphoto
forms of identification. In all four states, if voters are not able to
provide a required form of identification, they are allowed to cast a
provisional, rather than a regular, ballot. Finally, a state that had
identification requirements in place for the November 2004 general
election may have taken additional actions to amend such requirements.
Georgia, for instance, required voters in the November 2004 general
election to provide 1 of 17 types of photo or nonphoto identification. In
2005 Georgia enacted legislation that, in general, amended

24These 9 states are Maryland, Massachusetts, Michigan, New Hampshire, New
Jersey, Ohio, Oklahoma, Pennsylvania, and Wisconsin. Since our survey,
more recent actions by states include, for example, Ohio's enactment of
legislation, effective in May 2006, requiring voters to provide
identification prior to voting.

Page 169 GAO-06-450 Elections

Recruiting a Sufficient Number of Skilled, Reliable Poll Workers Continued
to Be a Challenge for Large and Medium Jurisdictions

and reduced the various forms of acceptable identification and made the
presentation of a form of photo identification, such as a driver's
license, a requirement to vote.25

Having enough qualified poll workers to set up, open, and work at the
polls on Election Day is a crucial step in ensuring that voters are able
to successfully vote on Election Day. The number of poll workers needed
varies across jurisdictions, and election officials recruit poll workers
in a variety of ways using different sources and strategies. Some poll
workers are elected, some are appointed by political parties, and some are
volunteers. Election officials in jurisdictions we visited reported
considering several different factors-such as state requirements,
registered voters per precinct, historical turnout, or poll worker
functions at polling places-to determine the total number of poll workers
needed. On the basis of our survey of local jurisdictions, we estimate
that recruiting enough poll workers for the November 2004 general election
was not difficult for the majority of jurisdictions. However, large and
medium jurisdictions encountered difficulties to a greater extent than
small jurisdictions. To meet their need, election officials recruited poll
workers from numerous sources, including in some cases, high schools and
local government agencies, to help ensure that they were able to obtain
enough poll workers for Election Day. Poll workers with specialized
characteristics or skills were also difficult for some large and medium
jurisdictions to find. Election officials in some jurisdictions we visited
reported that finding qualified poll workers could be complicated by
having a limited pool of volunteers willing to work long hours for low
pay. Poll worker reliability continued to be a challenge for some
jurisdictions-especially large jurisdictions-that depend on poll workers
to arrive at polling places on time on Election Day.

25In October 2005 a federal district court order (Common Cause/GA v.
Billups, 406 F. Supp. 2d 1326 (N.D. Ga. 2005)) granted a preliminary
injunction enjoining and restricting the state from enforcing or applying
certain provisions of the 2005 Georgia amendments. Plaintiffs in the case
allege, in general, that such amendments unduly burden the right to vote.
On appeal, the U.S. Court of Appeals for the Eleventh Circuit remanded the
case to the district court on February 10, 2006, for further proceeding in
light of subsequent Georgia amendments to its voter identification laws
enacted in January 2006.

Recruiting Enough Poll Workers Was Not Difficult for the Majority of
Jurisdictions, but Many Large and Medium Jurisdictions Encountered
Difficulties

We estimate that recruiting enough poll workers for the November 2004
general election was not difficult for the majority of jurisdictions, and
may have been less of a challenge for the November 2004 election than it
was for the November 2000 election. For example, on the basis of our 2001
survey of local jurisdictions, we estimate 51 percent of county election
jurisdictions found it somewhat or very difficult to find a sufficient
number of poll workers for the November 2000 election.26 In contrast, from
our 2005 survey, we estimate that 36 percent of county election
jurisdictions had difficulties obtaining enough poll workers for the
November 2004 election. In our 2005 survey, there are differences between
size categories of election jurisdictions in the difficulties encountered
obtaining a sufficient number of poll workers, with more large and medium
jurisdictions encountering difficulties than small jurisdictions. As shown
in figure 31, we estimate that 47 percent of large jurisdictions, 32
percent of medium jurisdictions, and 14 percent of small jurisdictions
found it difficult or very difficult to obtain a sufficient number of poll
workers.

26These estimates include only county election jurisdiction subgroup
comparisons between the 2001 and 2005 surveys. See appendix V for further
details about the sampling differences between these surveys.

Page 171 GAO-06-450 Elections

Figure 31: Estimated Percentages of All Local Jurisdictions and by
Jurisdiction Size That Encountered Difficulties Obtaining a Sufficient
Number of Poll Workers for the November 2004 General Election

Percentage of jurisdictions

                                       57

50

47e

46 44

40

35 32

30

30

24

22 20

20

17 14

10

5

4 11

0 Difficult or very Neither difficult Easy or very Not applicabled

c

difficulta nor easyb easy

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000)

Overall Source: GAO 2005 survey of local election jurisdictions.

aAll size categories are statistically different from one another.

bThe difference between small and medium jurisdictions is statistically
significant.

cThe differences between both large and medium jurisdictions and small
jurisdictions are statistically significant.

dJurisdictions could indicate not applicable for a variety of reasons,
including that poll workers are not recruited, but elected or appointed;
that elections are conducted by mail ballot, and as a result there is not
a need for poll workers to staff polling places on Election Day; or that
the election officials themselves serve as poll workers.

eThe 95 percent confidence interval for this percentage is +/- 8
percentage points.

Election officials in large and medium jurisdictions, with typically more
polling places to staff, are generally responsible for obtaining more poll
workers than officials in small jurisdictions. For example, election
officials in a large jurisdiction we visited in Illinois told us that
recruiting enough poll workers for Election Day was always a challenge and
November 2004 was no different. They said that state law specifies a
minimum of 5 poll workers per precinct, and there were 2,709 precincts in
their jurisdiction for the November 2004 general election, requiring at
least 13,545 poll workers. In contrast, election officials in a small
jurisdiction we visited in New Hampshire told us that they never had
difficulties finding poll workers because they were able to use a pool of
volunteers to staff the 9 poll worker positions at their one polling
place.

While election officials in 10 of the 27 large and medium jurisdictions we
visited told us they had difficulties recruiting the needed number of poll
workers, election officials in the other 17 jurisdictions did not report
difficulties. These officials provided a variety of reasons why they did
not encounter difficulties, including having a set number of appointed or
elected poll workers for each precinct, having a general public interest
in being involved in a presidential election, and using a variety of
strategies and sources to recruit poll workers. For example, election
officials in a large jurisdiction in New Mexico told us that their lack of
problems with recruitment was due to the fact that they had a full-time
poll worker coordinator who began the search for poll workers very early
and, as a result, was able to fill all of the positions needed (about
2,400) for the November 2004 election. Election officials in other large
jurisdictions reported that they were able to obtain enough poll workers
by relying on multiple sources. For example, election officials in a large
jurisdiction in Kansas told us that they made an exhaustive effort to
recruit about  1,800 poll workers for the November 2004 general election
that included soliciting from an existing list of poll workers, working
with organizations, using a high school student program to obtain about
300 student poll workers, recruiting from a community college, using
county employees, and coordinating with the political parties. On our
nationwide survey we asked local jurisdictions about the sources they used
to recruit poll workers for the November 2004 general election, and table
17 presents estimates from this survey on a variety of sources that
jurisdictions used.

Table 17: Estimated Percentages of All Local Jurisdictions and by
Jurisdiction Size That Used Sources to Recruit Poll Workers for the
November 2004 General Election

                            Percentage of small  Percentage of  Percentage of 
                                                                        large 
                Percentage of all jurisdictions medium          jurisdictions 
                                                jurisdictions  
Source           jurisdictions     (<10,000) (10,000 to         (>100,000) 
                                                100,000)       
Lists or rosters                                            
of poll workers                                             
from                                                        
past electionsa             89            86             96             98 
Referrals from              62           48b             92             97 
poll workersa                                               
Political                   48           36b             72             82 
partiesa                                                    
High schoolsc               21            13             36            59d 
Collegesc                   11             4             18            64d 
Public                                                      
announcements or                                            
information                                                 
posted on                   18                              
jurisdiction's                                              
Web sitec                                  9             31             77 
City/county                                                                
government                  18            12             25            65d
officesc                                                    
Local school                10             5             17            43d 
districtsc                                                  
Civic, cultural,                                            
or religious                                                
organizationsc              10             3             21            63d 
Private firmse               2             0              2            24f 

Source: GAO 2005 survey of local election jurisdictions.

aThe differences between both large and medium jurisdictions and small
jurisdictions are statistically significant.

bThe 95 percent confidence interval for these percentages is +/- 6
percentage points.

cAll size categories are statistically different from one another.

dThe 95 percent confidence interval for these percentages is +/- 8
percentage points.

eThe differences between large jurisdictions and both medium and small
jurisdictions are statistically significant.

fThe 95 percent confidence interval for this percentage is +8 and -7
percentage points.

In our October 2001 report on election processes, we identified several
recruiting strategies that election officials reported helped in their
efforts to obtain enough poll workers.27 On the basis of our local
jurisdictions survey, student poll workers and county or city employees
were used as sources for poll workers by many medium and large
jurisdictions in the November 2004 general election, as shown in table 17.
These two sources were also cited as having worked well by election
officials in several of the jurisdictions we visited. For example,
election officials in a jurisdiction in

27See GAO-02-3.

According to Our Nationwide Survey, Poll Workers with Specialized Skills
Were Difficult to Find for Some Jurisdictions

Colorado told us that their high school student poll worker programs
helped them to obtain a sufficient number of skilled poll workers and
reported that 200 of their about 600 poll workers were high school
students. Election officials in other jurisdictions we visited reported
that high school students often helped them in obtaining enough poll
workers with specialized skills or characteristics, such as needed
language skills. According to our state survey, 38 states and the District
of Columbia reported allowing poll workers to be under the age of 18.

Local government offices were another source of poll workers for the
November 2004 general election. As shown in table 17, we estimate that 
65 percent of large jurisdictions, 25 percent of medium jurisdictions, and
 12 percent of small jurisdictions recruited poll workers from city or
county government offices. For example, election officials in a large
jurisdiction in Nevada told us that the chief poll worker at most of the
jurisdiction's  329 polling places is a county employee, and described
benefits of recruiting local government employees as poll workers,
including their experience in dealing with the public.

The specific skills and requirements needed for poll workers varies by
jurisdiction, and in some cases by precinct, but can include political
party affiliation, specific technical or computer skills, or proficiency
in languages other than English. On the basis of our survey of local
jurisdictions, we estimate that most jurisdictions nationwide did not
encounter difficulties recruiting poll workers with these specific skills
and requirements. However, the results show that the ease of obtaining
poll workers with these skills varied by the size of the election
jurisdiction, with large and medium jurisdictions generally experiencing
more difficulties than small jurisdictions.

Some states require political balance between poll workers at polling
places. For example, New York election law, which requires that each
election district must be staffed with four election inspectors (i.e.,
chief poll workers) and a variable number of poll workers (depending upon
specified conditions), requires that appointments to such positions for
each election district be equally divided between the major political
parties. Election officials in some jurisdictions we visited told us that
even though not required, they tried to maintain a balance in poll
workers' political party affiliation. Recruiting enough poll workers with
specific political party affiliations continued to be a challenge for
some, in particular large and medium jurisdictions. From our local
jurisdiction survey, we estimate that 49 percent of large jurisdictions,
41 percent of medium jurisdictions, and 22 percent of small jurisdictions
had difficulties recruiting enough Democratic or Republican poll workers,
as shown in figure 32.

Figure 32: Estimated Percentages of All Local Jurisdictions and by
Jurisdiction Size That Encountered Difficulties Recruiting Enough Poll
Workers with Specific Political Party Affiliation for the November 2004
General Election

Percentage of jurisdictions

50 49d

41

40

35 35

30

30

29 26 23

28

22

21

20

17

12 10 16

10

8

0 Difficult or very Neither difficult Easy or very Not applicablea,c

a

difficulta nor easyb easy

Small (<10,000) Medium (10,000-100,000) Large (>100,000) Overall Source:
GAO 2005 survey of local election jurisdictions.

aThe differences between both large and medium jurisdictions and small
jurisdictions are statistically significant. bThe difference between small
jurisdictions and medium jurisdictions is statistically significant. cNot
all local jurisdictions are required to obtain poll workers with specific
party affiliations. dThe 95 percent confidence interval for this
percentage is +/- 8 percentage points.

Election officials in 11 of the 28 jurisdictions we visited reported
experiencing some difficulties finding enough poll workers with needed
party affiliations. For example, election officials in a jurisdiction in
Connecticut told us that because their jurisdiction was predominantly one
political party it was difficult to find minority party poll workers.
Election officials in these 11 jurisdictions told us that they recruited
independents, unaffiliated persons, or student poll workers to fill
minority party poll worker positions.

Recruiting poll workers with necessary information technology skills or
computer literacy was also a challenge for some large and medium
jurisdictions, according to our survey of local jurisdictions. We estimate
that 34 percent of large jurisdictions and 28 percent of medium
jurisdictions found it difficult or very difficult to obtain poll workers
with these skills, whereas, we estimate that 5 percent of small
jurisdictions had difficulties, as shown in figure 33.

Figure 33: Estimated Percentages of All Local Jurisdictions and by
Jurisdiction Size That Encountered Difficulties Recruiting Poll Workers
with Information Technology Skills or Computer Literacy for the November
2004 General Election

Percentage of jurisdictions

60

50

42d

Difficult or very Neither difficult Easy or very Not applicableb,c

a

difficulta nor easy easy

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000)

Overall Source: GAO 2005 survey of local election jurisdictions.

aThe differences between both large and medium jurisdictions and small
jurisdictions are statistically significant.

bThe difference between small and large jurisdictions is statistically
significant.

cJurisidictions may not need to recruit poll workers with technical or
computer skills because they, among other reasons, use paper or mechanical
voting systems, use paper poll books, or specifically train poll workers
for any skills needed in these areas.

dThe 95 percent confidence interval for this percentage is +/- 6
percentage points.

Election officials in 23 of the 28 jurisdictions we visited told us that
computer or technically skilled poll workers were not needed in their
jurisdictions for the November 2004 general election. However, election
officials in some of these jurisdictions reported that they foresaw a need
for poll workers with these skills with the implementation of electronic
poll books or new voting technology. Among the reasons cited for not
needing technically skilled poll workers were the use of paper ballots or
lever machines, the ease of use of DRE voting equipment, and that any
needed skills were taught. In addition, election officials in many
jurisdictions we visited told us that they recruited and trained
technicians or troubleshooters to maintain, repair, and in some cases set
up voting equipment prior to Election Day.

Some jurisdictions may be required under the language minority provisions
of the Voting Rights Act to, in general, provide voting assistance and
materials in specified minority languages in addition to English. We asked
on our survey of local jurisdictions whether jurisdictions encountered
difficulties recruiting poll workers who were fluent in the languages
covered under the Voting Rights Act for their jurisdiction and estimate
that for the majority (61 percent) of all jurisdictions, this requirement
was not applicable. We estimate that 15 percent of all jurisdictions
indicated that recruiting poll workers fluent in languages other than
English was difficult or very difficult.28 Jurisdictions of all size
categories may encounter difficulties recruiting poll workers with needed
language skills for different reasons. For instance, small jurisdictions
may find it difficult to recruit enough poll workers fluent in other
languages because of a limited pool of potential recruits, whereas large
jurisdictions may be required to provide voters with assistance in
multiple languages other than English. Los Angeles County, for example,
was required to provide voters assistance in six languages other than
English for the November 2004 election. Election officials in some of the
large jurisdictions we visited reported encountering difficulties
obtaining poll workers with needed language skills, but these officials
also told us about their efforts to recruit poll workers with language
skills. For example, election officials in a large jurisdiction in
Illinois reported that they recently established an outreach department to
assist in the recruitment of poll workers with specialized language
skills. The jurisdiction has hired outreach coordinators for the Hispanic,
Polish, and Chinese communities to assist with recruiting. Figure 34
illustrates materials used by election officials in some jurisdictions we
visited to recruit poll workers with a variety of skills for the November
2004 general election.

28The differences between size categories were not statistically
significant. We estimate that 23 percent of large jurisdictions, 20
percent of medium jurisdictions, and 13 percent of small jurisdictions had
difficulties recruiting poll workers with needed language skills.

Page 179 GAO-06-450 Elections

 Figure 34: Materials Used to Recruit Poll Workers in Jurisdictions We Visited

Election Officials in Some Jurisdictions We Visited Described Factors That
Affected Recruiting Poll Workers

  Source (left to right): Cuyahoga County, Ohio; King County, Washington (GAO
                    photo); Guilford County, North Carolina.

In our October 2001 report on election processes, we identified long
hours, low pay, and an aging volunteer workforce as factors that
complicated election officials' efforts to recruit enough poll workers.29
Election officials in some, but not all, of the jurisdictions we visited
in 2005 told us that one or more of these factors complicated their
efforts to find enough quality poll workers for the November 2004 general
election. For example, election officials in a large jurisdiction in
Nevada told us that it was difficult to find people who wanted to work,
considering that most families are two-income

29See GAO-02-3.

Poll Worker Reliability Continued to Be a Challenge, Especially in Large
Jurisdictions

households and Election Day is a long-14 hours-grueling day. Election
officials in a large jurisdiction in Washington told us that they never
have enough poll workers, noting that the pay is minimal, the hours are
long, and the majority of the poll worker population is elderly. Election
officials in several of these jurisdictions we visited reported concerns
about finding poll workers in light of a limited pool of volunteers. For
example, election officials in a large jurisdiction in Colorado told us
the average age of poll workers was over 70 years old and expressed
concerns about obtaining poll workers who could physically work a 12-hour
day. Alternatively, election officials in a large jurisdiction in Florida
told us that the younger generation does not have the same commitment to
civic duty that the older poll worker generation had and recruiting enough
qualified poll workers may be a challenge in the future. These officials
noted that about threequarters of their poll workers are return
participants. An election official in a large jurisdiction in
Pennsylvania, where the median age of poll workers is about 75 years old,
suggested that serving as a poll worker should be treated similarly as
serving on jury duty-it should be everyone's civic duty to serve as a poll
worker.

In our October 2001 report on election processes, we noted that poll
worker reliability was a challenge for election officials, who depended on
poll workers to arrive on time, open, and set up polling places.30 Poll
worker absenteeism was a challenge for large and, to some extent, medium
jurisdictions in the November 2004 general election. On the basis of our
nationwide survey of local jurisdictions, we estimate that 61 percent of
large jurisdictions, 20 percent of medium jurisdictions, and 2 percent of
small jurisdictions encountered problems with poll workers failing to show
up on Election Day. The differences between all size categories are
statistically significant. One way that election officials in several
large jurisdictions we visited minimized the impact of poll worker
absenteeism was to recruit backup poll workers to ensure that polling
places were set up and adequately staffed, even if some poll workers
failed to show up. For example, election officials in a large jurisdiction
we visited in Illinois reported that approximately 1 to 2 percent of about
13,000 poll workers did not show up on Election Day. However, these
officials reported that they had recruited stand-by judges who were to
report to the elections office on Election Day in case an already
scheduled judge did not show up. Election

30See GAO-02-3.

Election Officials in Some Jurisdictions We Visited Reported on Challenges
Training Poll Workers

officials in a few other jurisdictions we visited told us that they called
poll workers before Election Day to help ensure they showed up. For
instance, election officials in a large jurisdiction in Pennsylvania told
us that they called all of the chief poll workers-about 1,300
people-during the week prior to the election. Election officials in a
large jurisdiction we visited in Connecticut went a step further,
reporting that in addition to placing wakeup calls to all of the chief
poll workers, they offered rides to poll workers to help ensure they
showed up on time.

Voters' experiences on Election Day are largely informed by their
interactions with poll workers, who are responsible for conducting many
Election Day activities, such as setting up polling places, checking in
voters and verifying their eligibility to vote, providing assistance to
voters, and closing the polling places. Although these workers are usually
employed only for 1 day, the success of election administration partly
depends on their ability to perform their jobs well. Depending on the
applicable state requirements and the size of the jurisdiction, the steps
that election officials take to adequately prepare all of their poll
workers can vary, but may include training, testing, or certification.
Ensuring that poll workers were adequately trained for Election Day was a
challenge reported by some election officials in large and medium
jurisdictions we visited, but these officials also reported a variety of
steps they took to help prepare poll workers for Election Day.

State Requirements for Training Poll Workers Varied

Most states and the District of Columbia reported having training
requirements for poll workers for the November 2004 general election, but
the frequency and content of training varied. Some states also reported
providing guidance related to the training of poll workers. According to
our state survey, for the November 2004 general election, 18 states
reported having had poll worker training requirements and providing
guidance;  20 states and the District of Columbia reported having had
training requirements; 9 states reported providing guidance; 1 state
reported that it did not require training nor provide guidance; and
Oregon, which conducted all-mail voting on Election Day 2004, indicated
this requirement was not applicable.31 Figure 35 shows reported state
requirements for

31The total does not add to 50 because 1 state did not respond to this question
                                 on our survey.

                         Page 182 GAO-06-450 Elections

training for the chief poll worker at a precinct or polling place and for
poll workers.

Figure 35: Reported State Requirements fo r P o ll Worker Training for the
November 2004 General Election

Number of states 25

21 21

20

20

18

15

10

9

7

5

1

1

0

Had both requirementsand guidance

Pro

      Had requirements onlyavided guidance onlyNo requirementsor guidance

                             Had both requirements

Pro

                 Had requirements onlyaand guidance dance only

                           No requirementsor guidance

                                   vided gui

Chief poll worker at a precinct or polling placeb

Poll workersc Source: GAO 2005 survey of state election officials.

aIncludes the District of Columbia.

bThe total does not add to 50 because 2 states indicated that this
requirement was not applicable- 1 of which was Oregon, which conducted
all-mail voting on Election Day 2004.

cThe total does not add to 50 because 1 state did not respond to this
question on our survey and Oregon indicated this requirement was not
applicable.

About half of the states with training requirements reported requiring
that poll workers be trained prior to every election or every general
election. According to our survey, of the 38 states and the District of
Columbia that reported having training requirements for poll workers, 22
states and the District of Columbia reported requiring poll workers to be
trained prior to every election or every general election. For example,
Florida provisions in place for the November 2004 general election
required that poll workers have a minimum of 3 hours of training prior to
each election and demonstrate a working knowledge of the laws and
procedures relating to voter registration, voting system operation,
balloting, and polling place procedures, and problem-solving and conflict
resolution skills. These Florida provisions also require, among other
things, that local election officials are to contract with a "recognized
disability-related organization" to develop and assist with training for
disability sensitivity programs, which must include actual demonstrations
of obstacles confronted by persons with disabilities during the voting
process, including obtaining access to the polling place and using the
voting system. Ten states reported requiring that poll workers be trained
on a scheduled basis (e.g., yearly or every 2 years). For example, under
provisions in place for the November 2004 general election, New Jersey
required that all district board members attend training sessions for each
election at least once every 2 years. The other 6 states reported that
training was required at least once, but not prior to every general
election; that the frequency of training was not specified; or that they
did not know.

For the November 2004 general election, fewer states reported requiring
testing or certification than training for poll workers. According to our
state survey, 12 states reported having requirements for testing or
certification for poll workers, and 16 states reported having these
requirements for the chief poll worker at a precinct or polling place.
Election officials in 6 of the 28 jurisdictions we visited reported that
poll workers were certified or tested after training. Election officials
in 6 other jurisdictions told us that they used informal tests or quizzes
or informally monitored poll workers performance in training. For
instance, election officials in a jurisdiction in Kansas told us that they
gave poll workers a nongraded quiz at the end of training. In Nevada,
where state election officials indicated in our state survey that there
are no requirements for poll worker training or testing, election
officials in the 2 jurisdictions we visited told us that they required
poll workers to attend training. Election officials in 1 of these
jurisdictions required all poll workers to attend a training class each
year and to pass a hands-on performance test in which they demonstrate
their ability to perform their assigned function, such as checking in
voters or programming the DRE voting equipment.

Election Officials Reported on Training Conducted and Challenges
Encountered

Training provided to poll workers varies greatly among local election
jurisdictions. Therefore, we asked questions about training challenges as
part of our site visits only where we were able to gain an understanding
of the types of training and specific conditions faced by local
jurisdictions. Election officials in a small jurisdiction we visited in
New Hampshire reported that they did not conduct training for the November
2004 general election because poll workers only receive training if they
have not previously worked in the polling place, and all nine poll workers
had worked in the polling place before. Election officials in the 27 other
jurisdictions we visited described the training that they provided poll
workers for the November 2004 general election. According to these
officials, poll worker training generally occurred in the weeks or month
before the election and ranged from 1 hour to 2 days, depending on the
type of poll worker being trained. Election officials in most of these
jurisdictions reported that training was mandatory. However, the frequency
varied, with election officials in the majority of jurisdictions reporting
that they required training prior to every election. Election officials in
a few jurisdictions reported that poll workers received training at least
once or on a scheduled basis, such as once every 2 years. Election
officials in many jurisdictions told us that poll workers were paid to
attend training, and payments could range from $5 to $50.

While election officials in nearly all of these jurisdictions reported
that training was conducted by these officials and their staffs, the
manner in which the training was conducted varied. For example, election
officials in a large jurisdiction in Nevada told us that poll workers were
trained in a workshop fashion in which 15 to 20 poll workers were provided
hands-on training for their specific function, such as operating voting
machines or processing voters. In a large jurisdiction in Kansas, election
officials told us that they conduct the training for between 70 and 100
poll workers using a formal presentation as well as the documents poll
workers use on Election Day and the voting equipment. Election officials
in a large jurisdiction in Washington told us that poll worker training
consisted of a PowerPoint presentation conducted in a train-the-trainer
style where election officials trained the chief poll workers, who then
trained the poll workers.

Election officials in 9 of the 27 large and medium jurisdictions we
visited reported encountering some challenges with training poll workers,
but generally reported that they overcame them. Some of the challenges
reported by these officials included keeping poll workers informed about
new or changing requirements, conveying a vast amount of information about
election processes to a large number of people in a limited time, and

Poll Worker Performance Problems in Some Large and Medium Jurisdictions

ensuring that poll workers understand their tasks and responsibilities.
For instance, election officials in a large jurisdiction in Ohio told us
that it was challenging keeping up with state changes and incorporating
such changes into poll worker training. Election officials in a large
jurisdiction in Connecticut told us that effectively training poll workers
on a variety of new changes (such as those required by HAVA) could be
challenging because the procedures can be difficult to understand,
especially for tenured poll workers who have been working at the polls for
many years. Election officials in a large jurisdiction in Kansas noted
that addressing the need to have a systematic way to evaluate poll worker
performance at polling places was a challenge. These officials said that
they currently rely on the fact that the poll worker showed up, general
observations of the poll workers' performance, and feedback cards
completed by voters exiting the polls. Election officials in the
jurisdictions we visited reported taking steps to address these
challenges, such as providing poll workers training manuals or booklets
for reference on Election Day, training poll workers to perform one
function, and conducting training in a workshop fashion with smaller class
sizes.

Election officials and poll workers perform many tasks throughout the day
to ensure that elections run smoothly and that voters move efficiently
through the polling place. These activities can include checking in
voters, providing instructions for voting machine operation, or assisting
voters at the polls. We asked on our survey of local jurisdictions whether
for the November 2004 general election jurisdictions encountered poll
workers failing to follow procedures for a variety of activities,
including, among others, procedures for voter identification requirements,
providing correct instructions to voters, and voting machine operation.
Overall, according to this survey, most local election jurisdictions
nationwide did not encounter problems with poll worker performance. For
example, we estimate that

     o 90 percent of all jurisdictions did not encounter poll workers failing
       to follow procedures related to voter identification requirements,
     o 92 percent of all jurisdictions did not encounter poll workers failing
       to provide correct instructions to voters, and
     o 94 percent of all jurisdictions did not encounter poll workers failing
       to follow procedures for voting machine operation.

Most Jurisdictions Provided Guidance at Polling Places to Help Poll
Workers

However, we estimate that poll worker performance problems encountered
varied by size category of jurisdiction, with more large jurisdictions
encountering problems than medium and small jurisdictions. For example, we
estimate that 37 percent of large jurisdictions, 19 percent of medium
jurisdictions, and 3 percent of small jurisdictions encountered problems
with poll workers failing to follow procedures related to voter
identification requirements. In terms of providing correct instructions to
voters, we estimate that 31 percent of large jurisdictions, 12 percent of
medium jurisdictions, and 1 percent of small jurisdictions encountered
problems with poll worker performance in this area. For both results, the
differences between all size categories are statistically significant.

Large jurisdictions could have encountered problems for a variety of
reasons, including having more poll workers to train and oversee or having
fewer options for recruiting skilled poll workers. While jurisdictions may
have reported on our survey that they encountered problems with a
particular aspect of poll workers' performance, written comments provided
on the questionnaire indicated that these problems may not have been
widespread or may have been easily remedied after they occurred. For
example, one survey respondent wrote: "Errors were few and far between,
but with 4,500 poll workers, it is very difficult to answer that [our
jurisdiction did not encounter any problems with poll workers'
performance.]" Election officials in 12 of the 28 jurisdictions we visited
reported that they encountered some problems with poll workers'
performance, but that generally the majority of poll workers performed
well. For example, an election official in a large jurisdiction in
Pennsylvania we visited told us that while the jurisdiction did not
encounter serious problems with performance, in the official's opinion, it
would be disingenuous to report that there were no problems with the 6,500
poll workers working the polls on Election Day.

In an effort to minimize poll worker confusion or performance problems,
many jurisdictions provided written guidelines or instructions for poll
workers to use at the polling place. On our nationwide survey we asked
local jurisdictions whether or not for the November 2004 general election
they had written guidelines or instructions at the polling place for poll
workers covering a variety of topics, such as voting equipment operation;
procedures related to verifying voters' eligibility to vote; and assisting
voters with special needs, such as voters with disabilities or who spoke a
language other than English. We estimate that 94 percent of all
jurisdictions had at least one set of written guidelines at polling places
for poll workers.

Further, more large and medium jurisdictions provided instructions to poll
workers than small jurisdictions. For example, we estimate that 99 percent
of large jurisdictions, 96 percent of medium jurisdictions, and 80 percent
of small jurisdictions provided written instructions for poll workers to
use at polling places if a voter's name was not on the poll list. In
addition, we estimate that 96 percent of large jurisdictions, 92 percent
of medium jurisdictions, and 71 percent of small jurisdictions provided
written guidelines to use at the polls for identification requirements for
first-time voters who registered by mail and did not provide
identification with their registration. For both of these results, small
jurisdictions are statistically different from both medium and large
jurisdictions.

During our site visits, election officials in 26 of the 28 jurisdictions
we visited reported that they provided written instructions or checklists
for poll workers to have at polling places. Election officials in the 2
smallest population size jurisdictions we visited reported that they did
not provide written instructions for poll workers. As the officials in a
small jurisdiction in New Hampshire said, they are at the polling place to
resolve issues personally as they arise. Figure 36 illustrates examples of
some checklists that election officials in jurisdictions we visited
provided to us.

Figure 36: Examples of Instructions and Checklists Provided to Poll Workers for
                       the November 2004 General Election

Source (left to right): Clark County, Washington; Leon County, Florida.

Written instructions and checklists may help poll workers, but problems on
Election Day can still be encountered with some issues, in particular
issues related to voter registration. We asked on our survey of local
jurisdictions whether for the November 2004 general election jurisdictions
maintained a written record to keep track of issues or problems that
occurred on Election Day. We estimate that 55 percent of all jurisdictions
nationwide maintained a written record to keep track of issues. Of those
that did

Obtaining Enough Polling Places That Met Standards Continued to Be a
Challenge for Some Jurisdictions

maintain a record and provided written comments on our survey, the issues
most frequently cited by election officials were problems with voter
registration (e.g., not being registered, being registered at another
polling location, or being in the wrong polling location).

Election officials are responsible for selecting and securing a sufficient
number of polling places that meet basic requirements and standards.
Polling place locations vary across jurisdictions but can include public
and private facilities, such as schools, government buildings, fire
departments, community centers, libraries, churches, and residential
facilities. To meet the needs of the voting population, polling places
should be easily accessible to all voters, including voters with
disabilities. Polling places also need to have a basic infrastructure,
including electricity, heating and cooling units, and communication lines,
to support some voting machines and be comfortable for voters and poll
workers. In our October 2001 report on election processes, we stated that
obtaining polling places for the November 2000 election was not a major
challenge for most jurisdictions.32 On the basis of our 2005 survey of
local jurisdictions, obtaining a sufficient number of polling places was
not difficult for the majority of jurisdictions. However, finding polling
places that met these standards was generally more difficult for large and
medium jurisdictions than for small jurisdictions. Election officials in
many jurisdictions reported combining precincts in one polling place, with
minimal challenges, for the November 2004 general election.

Finding a Sufficient Number of Polling Places Was Difficult for Some, but
Not Most, Jurisdictions

For the November 2004 election, obtaining a sufficient number of polling
places was not difficult for the majority of jurisdictions. On the basis
of our survey of local jurisdictions, we estimate that 3 percent of all
jurisdictions found it difficult or very difficult to obtain a sufficient
number of polling places for the November 2004 general election. However,
the difficulty encountered in finding enough polling places varied by the
size category of jurisdiction. We estimate that 14 percent of large
jurisdictions, 8 percent of medium jurisdictions, and 1 percent of small
jurisdictions had difficulties obtaining enough polling places, as
presented in figure 37.

32See GAO-02-3.

Figure 37: Estimated Percentages of All Local Jurisdictions and by
Jurisdiction Size That Encountered Difficulties Obtaining a Sufficient
Number of Polling Places for the November 2004 General Election

Percentage of jurisdictions

                                       75

68

70

60

56

50

45a

40

38

33

30

20

20

14 13 11 9

10 8 3

32

1

0 Difficult or very Neither difficult Easy or very Not applicableb,d

                                       bc

difficultb nor easyeasy

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000)

Overall Source: GAO 2005 survey of local election jurisdictions.

aThe 95 percent confidence interval for this percentage is +/- 8
percentage points.

bThe differences between both large and medium jurisdictions and small
jurisdictions are statistically significant.

cAll size categories are statistically different from each other.

dJurisdictions may indicate not applicable if they do not use polling
places on Election Day because of all mail balloting, or because only one
polling location is used on Election Day.

Small jurisdictions may not experience difficulties obtaining polling
places for a variety of reasons, among them because they do not have to
find as many locations to support an election as large jurisdictions do.
For example, election officials in a small jurisdiction we visited in New
Hampshire told us that because of the small voting population (about
1,200), they only needed to use one polling place-the town hall-for the
November 2004 general election, as shown in figure 38.

                 Figure 38: Town Hall in Madbury, New Hampshire

Source: GAO.

In contrast, large jurisdictions could be responsible for selecting
hundreds of polling places for Election Day. Election officials from a
large jurisdiction we visited in Illinois reported that they used over
1,800 polling places for the November 2004 election and hired staff to
find polling places that met standards for their jurisdiction. Although
election officials in some large and medium jurisdictions told us that
they needed to find numerous polling places, officials in only 1 large
jurisdiction we visited in Kansas told us that they encountered
difficulties finding suitable polling places, in part because of low
payments provided to use polling place facilities. Election officials in
this jurisdiction reported that in 2003 they implemented a campaign to
"recruit" polling places and sent letters to schools and other possible
locations in addition to conducting site visits and inspections. These
election officials reported that after their efforts, they added about 70
polling places for use on Election Day 2004.

Finding Accessible Polling Places and Polling Places with Parking and
Phone Lines Was Difficult for Some Jurisdictions

Selecting accessible polling places includes assessing parking areas,
routes of travel, exterior walkways, and entrances, as well as interior
voting areas. In our October 2001 report on voters with disabilities, we
identified a variety of challenges faced by election officials in
improving the accessibility of voting-including the limited availability
of accessible buildings and the lack of authority to modify buildings to
make them more accessible.33 Finding accessible polling places continued
to be a challenge for some jurisdictions for the November 2004 general
election. On the basis of our local jurisdiction survey, we estimate that
36 percent of large jurisdictions, 25 percent of medium jurisdictions, and
5 percent of small jurisdictions found it difficult or very difficult to
find enough accessible polling places, as shown in figure 39.

33See GAO-02-107.

Figure 39: Estimated Percentages of All Local Jurisdictions and by
Jurisdiction Size That Encountered Difficulties Obtaining Enough Polling
Places That Were Accessible to Voters with Disabilities for November 2004
General Election

Percentage of jurisdictions 80

70 69

60

60

50

40

40

36 37 32

30

25

25

20

20

15 12

11

8 5

10

32

0 Difficult or very Neither difficult Easy or very Not applicablea,c

a

difficulta nor easyeasyb

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000)

Overall Source: GAO 2005 survey of local election jurisdictions.

aThe differences between both large and medium jurisdictions and small
jurisdictions are statistically significant.

bAll size categories are statistically different from one another.

cJurisdictions may indicate not applicable if they do not use polling
places on Election Day because of all mail balloting, or because only one
polling location is used on Election Day.

Election officials in some jurisdictions we visited told us that they
encountered challenges finding accessible polling places. For example,
election officials in 2 large jurisdictions we visited reported that it
was challenging to find polling places that were accessible because many
of the public buildings in their jurisdiction were older facilities and
were not compliant with the Americans with Disabilities Act (ADA).
However, election officials reported taking steps to help ensure that
polling places were accessible. For example, election officials in a large
jurisdiction in Georgia reported that they hired a private company to
conduct surveys of the polling locations and determine whether they were
accessible and what, if any, changes needed to be made to make the
facilities compliant. Some election officials described making minor or
temporary modifications to polling places to ensure that they were
accessible, for example, by adding ramps, using doorstops for heavier
doors, or clearly identifying accessible entrances.

In addition to being accessible for all voters, polling places should have
sufficient parking for voters and phone lines to provide for communication
on Election Day. From our local jurisdiction survey, more large and medium
jurisdictions encountered difficulties in finding polling places with
these characteristics than small jurisdictions. On the basis of this
survey, we estimate that 38 percent of large jurisdictions, 18 percent of
medium jurisdictions, and 4 percent of small jurisdictions had
difficulties obtaining polling places with adequate parking. The
differences between all size categories are statistically significant. In
terms of finding polling places with adequate phone lines, we estimate
that 35 percent of large jurisdictions, 33 percent of medium
jurisdictions, and 9 percent of small jurisdictions had difficulties
obtaining polling places with adequate phone lines. Providing cell phones
to poll workers was one way for some jurisdictions to help ensure
communication between polling places and the election office on Election
Day. Also on the basis of our survey, we estimate that cell phones
provided by the jurisdiction were the primary means of communication for
29 percent (plus or minus 9 percent) of large jurisdictions, 15 percent
(+9 percent, -6 percent) of medium jurisdictions, and 3 percent of small
jurisdictions.34 For both of these results, the differences between both
large and medium jurisdictions and small jurisdictions are statistically
significant. Election officials in some large jurisdictions we visited
included cell phones as part of the supplies provided to each polling
place. For example, officials in a large jurisdiction we visited in Nevada
told us they paid poll workers $5 to use their own cell phones.

34The 95 percent confidence interval for both the large and medium
percentages is greater than the +/-7 percent sampling error for their
respective samples.

Page 195 GAO-06-450 Elections

Combining Precincts at a Polling Location Continues to Be a Strategy to
Address Challenges with Obtaining Polling Places

We identified several strategies in our October 2001 report on election
processes that election officials said helped in their efforts to obtain
enough polling places, including locating more than one precinct at a
single polling place.35 Results of our 2005 state and local surveys and
site visits show that combining precincts at a polling location continued
to be a strategy used by local jurisdictions, predominantly large and
medium jurisdictions, to find adequate polling locations for voters in all
precincts. According to our state survey, nearly all states (47) reported
that they allowed precincts to be colocated in a polling place for the
November 2004 general election. Ten states reported allowing colocation
only under specified conditions, for instance, if no suitable polling
place existed for a precinct, and 37 states reported allowing colocation
but did not specify conditions. 36 On the basis of our survey of local
jurisdictions, we estimate 33 percent of all jurisdictions had multiple
precincts located in the same polling place. However, more large and
medium jurisdictions combined precincts than small jurisdictions. We
estimate that 78 percent of large jurisdictions, 63 percent of medium
jurisdictions, and 19 percent of small jurisdictions had multiple
precincts located in the same polling location. The differences between
all size categories are statistically significant. During our site visits,
election officials in 22 of the 28 jurisdictions we visited told us that
they combined precincts in the same polling location for the November 2004
general election. Included in the 6 jurisdictions that did not report
combining precincts in a single polling place were the 1 small and 2
medium jurisdictions we visited. Further, in many of the large
jurisdictions we visited, election officials told us that most of their
polling places had more than one precinct. For example, election officials
in a large jurisdiction in Ohio told us that there was an average of three
precincts per polling location, but that there could be up to nine
precincts in one polling place.

Although combining precincts may help solve the issue of obtaining a
sufficient number of voting places that meet requirements, other
challenges may surface, including voter confusion in not finding the
correct precinct at a location, poll worker confusion about eligibility if
a voter is not in the correct precinct poll book at a polling place, and
the possibility

35See GAO-02-3.

36North Carolina and the District of Columbia reported that they did not
allow precincts to be colocated in a polling place. Oregon, which uses
all-mail voting on Election Day, and Kentucky reported on our survey that
this was not applicable.

Page 196 GAO-06-450 Elections

of voters voting on the wrong voting machine for their precinct. However,
on the basis of our local survey, few challenges were encountered in
polling places where precincts were combined for the November 2004 general
election. We estimate that of the 33 percent of jurisdictions with
multiple precincts at a polling location, 85 percent (+6 percent, -5
percent) did not experience challenges in terms of voters locating their
correct precinct.37 Election officials in jurisdictions we visited
described steps they took to help ensure that voters were able to easily
find their correct precinct, including posting signage to direct voters to
the correct precinct, using specially designated poll workers as greeters
to direct voters as they entered the polling location, setting up separate
tables or voting areas for each precinct, and locating the precincts in
distinct areas of the building, for example, in the gym and cafeteria of a
school building.

Election officials in a few jurisdictions we visited told us that they
consolidated functions, such as the check-in table or voting equipment,
for precincts located in the same polling location in order to avoid voter
confusion or problems with voting. For example, election officials in a
jurisdiction in Kansas reported that they used one registration table with
a consolidated poll book for all precincts at a polling location. As a
result, voters only needed to locate one table. Election officials in a
jurisdiction in Nevada reported that once voters checked in, they were
able to vote on any voting machine in the polling location because the
machines were programmed with ballots from each of the precincts located
at the polling place, and poll workers activated the particular ballot
style for a particular voter.

Beyond consolidating some functions at a polling place, in 2004 Colorado
authorized the use of "vote centers," which are polling places at which
any registered voter in the local election jurisdiction may vote,
regardless of the precinct in which the voter resides. Each vote center is
to use a secure electronic connection to a computerized registration list
maintained by the local election office to allow all voting information
processed at any vote center to be immediately accessible to computers at
all other vote centers in the jurisdiction. Larimer County, with 143
precincts and about  200,000 registered voters, reported using 31 vote
centers for the November 2004 general election. Election officials in
Larimer County described several benefits of vote centers, including voter
convenience; cost

37The 95 percent confidence interval for this percentage is greater that
the +/-5 percent sampling error for the complete sample.

Page 197 GAO-06-450 Elections

effectiveness; minimal voter wait time on Election Day; and overall easier
management, including requiring fewer poll workers. Election officials
told us that voters liked the convenience of being able to vote anywhere
in the jurisdiction, regardless of the precinct they live in. Vote centers
can also be cost-effective, according to election officials, for
jurisdictions faced with replacing voting equipment to comply with HAVA
accessibility requirements for voting systems used in federal elections.
Using vote centers also reduces the number of polling places a
jurisdiction needs, which can be cost-effective with respect to finding
enough accessible polling places. Election officials also told us that on
Election Day they were able to avoid having long lines at most vote
centers by issuing media announcements to voters throughout the day
specifying which vote centers were busy and which were not, and by using
their electronic poll book technology to process voters quickly and to
monitor ballots and supplies. Officials told us that on average there was
a 15-minute wait time for voters. Finally, officials told us that from the
perspective of election officials, vote centers facilitated aspects of
election administration because there were fewer locations (about 30
instead of about 140) and fewer poll workers overall to recruit and train.

While other jurisdictions in Colorado have used vote centers since the
November 2004 election or are planning to pilot vote centers in elections
in 2006, election officials in a second jurisdiction we visited in
Colorado explained why their jurisdiction opted to not use vote centers.38
Officials told us that their jurisdiction assessed the feasibility of
implementing vote centers and concluded that despite several advantages
offered by vote centers, the cost of implementation was prohibitive. For
example, election officials identified costs including the connectivity
for the electronic poll books, so that voters can be credited with voting
in real time; potential rental costs for facilities, such as hotels, to
house vote centers; and the expense of purchasing additional voting
equipment. Because a voter in a jurisdiction using vote centers can vote
at any vote center, each vote center needs to be stocked with all
applicable ballot styles for an election or have DRE voting machines
capable of being programmed with all applicable ballot styles, according
to election officials. For the November 2004 general election, these
officials told us that they used optical scan for absentee and Election
Day voting and DREs for early voting. To avoid the cost and

38Other jurisdictions outside of Colorado are considering vote centers.
For example, the Indiana Secretary of State formed a bipartisan delegation
of legislators and local election officials to conduct a study of vote
centers and their applicability to Indiana.

Page 198 GAO-06-450 Elections

Election Officials in Some Jurisdictions We Visited Reported That
Designing Clear Ballots Continued to Be a Challenge

confusion of having to print and keep track of ballot styles for their 
378 precincts-compared to Larimer County's 143 precincts-election
officials said that they would need to purchase additional DRE voting
machines if they were to implement vote centers.

Election officials are responsible for designing ballots that meet various
state requirements, possibly federal requirements under the minority
language provisions of the Voting Rights Act relating to offering voting
materials in specified minority languages in addition to English, and the
requirements of the particular voting equipment, and these ballots must be
easy for voters to understand. Ballot design generally involves both state
and local participation. Most states (46 states and the District of
Columbia) were involved in ballot design for the November 2004 general
election. For instance, according to our state survey, 17 states and the
District of Columbia reported designing ballots for local jurisdictions, 3
states reported requiring approval of the ballot design, and 26 states
reported having requirements for local jurisdictions regarding ballot
design  (e.g., layout, candidate order, or paper stock).39 Specifically,
election officials must determine all races, candidates, and issues that
voters in each precinct in a jurisdiction will vote on and construct
layouts for these races and issues for the particular types of ballots
used with their election equipment. Figure 40 illustrates an optical scan
ballot used in El Paso County, Colorado, for the November 2004 general
election.

39Three states (Kentucky, Nevada, and Washington) reported that they were
not involved in ballot design; local jurisdictions designed ballots. One
state did not respond to this question on our survey.

Page 199 GAO-06-450 Elections

            Figure 40: El Paso County, Colorado, Optical Scan Ballot

                       Source: El Paso County, Colorado.

In our October 2001 report on election processes, we noted that despite
the controversy over the "butterfly ballot" and other ballot problems in
the aftermath of Florida's 2000 general election, very few jurisdictions
nationwide thought that confusing ballot design was a major problem.40
Ballot design problems were not highlighted by voters as a problem in the
November 2004 election; therefore, we did not inquire about the extent of
ballot design problems in our local survey of jurisdictions. However, we
asked about ballot design processes and problems during our visits to
local election jurisdictions. Election officials in all of the
jurisdictions we visited reported that they did not encounter voter
problems with confusing ballot designs for the November 2004 general
election. However, election officials in 7 jurisdictions we visited told
us that designing easily understood ballots that meet the particular
constraints of the voting equipment can be challenging when there are a
large number of races or issues to include on the ballot. For example,
election officials in a jurisdiction we visited in Colorado that used
optical scan ballots told us that fitting all of the races and questions
on the ballot is always challenging, but they managed to do so by limiting
the number of words on ballot questions and using small fonts. These
officials noted that they provided magnifying glasses at polling places to
assist voters. Election officials in a jurisdiction we visited in Florida
reported that they had to use oversized optical scan ballots to
accommodate the number of constitutional amendments that had to be
included on the ballot.

Some ballot design options taken to help ensure clarity for voters could
lead to problems later. For example, election officials in a jurisdiction
in Kansas reported that they used a two-sided ballot design requiring that
the optical scan counting equipment read the ballot front and back, which
presented a problem. Chapter 6 discusses challenges with counting ballots.
The requirements of the voting equipment may also limit options election
officials can take related to ballot design. For example, election
officials in a jurisdiction in Illinois that used punch cards reported
that lengthy ballots could have been a problem in the November 2004
election, but they decided to change the type of punch card ballot used.
These officials told us that increasing the number of punch positions
allowed for more space on the ballot and prevented challenges related to
length of ballot. However, with punch card ballots, the greater the number
of choices on a punch card, the greater the potential for voter error in
punching the preferred choice, as voters must align the ballot carefully.

40See GAO-02-3.

Preparing Ballots or Voting Equipment Not a Problem Reported by Most
Jurisdictions, but These Activities Can Be a Challenge for Some
Jurisdictions

Election officials in jurisdictions we visited that designed their ballots
described steps they took to ensure that ballots were clear to voters,
including using templates from the state or election management systems,
proofreading both before and after printing, and public viewing or testing
of ballots. For example, election officials in a jurisdiction in Colorado
told us that prior to printing they send proofs of the ballot designs to
candidates for their review. After printing, election officials said that
staff members and representatives of the political parties test the ballot
designs to ensure that there are no problems with how the ballots are
processed through the counting equipment. Election officials in another
jurisdiction in Colorado reported conducting a mock election with county
employees to review the ballot and test a ballot from each package of
printed ballots. Election officials in a jurisdiction in Ohio told us that
they displayed the ballots for the general public to view and test.

The activities and plans that election officials undertake related to
preparing ballots or voting equipment can have a direct impact on a
voter's Election Day experience. For example, reports about the November
2004 election highlighted shortages of ballots and voting machines at some
polling places. While election officials may not be able to prepare for
every contingency that could affect a voter's wait time or experience at
the polls, ensuring that there is a sufficient number of ballots or voting
machines can minimize potential problems. On the basis of our survey of
local jurisdictions, we estimate that few jurisdictions had problems with
ballot or voting equipment shortages for the November 2004 general
election. We estimate that 4 percent of all jurisdictions experienced
problems with Election Day ballot shortages, and an estimated 4 percent of
all jurisdictions did not have enough voting equipment on Election Day.
However, there were statistical differences between large and small
jurisdictions in having enough voting equipment. We estimate that  12
percent of large jurisdictions, 4 percent of medium jurisdictions, and  3
percent of small jurisdictions did not have enough voting equipment.
Election officials in 23 of the 28 jurisdictions we visited reported that
they encountered no challenges with preparing and delivering ballots,
voting equipment, and supplies for the November 2004 general election.
However, these activities could present logistical challenges for
jurisdictions if there are unexpected delays, or for jurisdictions that
are required to prepare ballots in multiple languages or prepare and
deliver numerous voting machines to a large number of polling places.

To ensure that there is an adequate supply of machine-readable paper
ballots on Election Day, election officials may conduct numerous
activities, such as designing, reviewing, proofreading, printing, and
testing ballots. Uncertainties about ballot content, such as whether or
not certain candidates or issues will be included on the ballot, could
affect these activities by delaying printing or leading to a last-minute
rush to ensure that ballots are printed in time for the election. While
election officials in most of the jurisdictions we visited did not report
encountering these uncertainties, election officials did in 4
jurisdictions. For example, election officials in a jurisdiction in
Colorado reported that ballot printing was delayed by three statewide
lawsuits regarding the content of the ballot. These officials reported
that they prepared two ballot designs-one with a particular candidate's
name and one without-so that they would be prepared to send the ballots to
an external printer regardless of the lawsuits' outcome.

Some jurisdictions are required to provide ballots in languages other than
English. Producing ballots in multiple languages can add to the complexity
of preparing ballots because election officials must take steps to ensure
proper translation and printing for each required language. On the basis
of our local jurisdictions survey, we estimate that 6 percent of
jurisdictions nationwide provided ballots in other languages. We estimate
that significantly more large jurisdictions provided ballots in languages
other than English than medium and small jurisdictions. We estimate that 
26 percent of large jurisdictions (compared to 10 percent of medium
jurisdictions and 3 percent of small jurisdictions) provided ballots in
languages other than English.41

Once voting equipment, ballots, and supplies have been prepared, ensuring
that they are transported to polling places can be a logistical challenge
for jurisdictions with thousands of voting machines and hundreds of
polling places. Election officials in 18 of the 28 jurisdictions we
visited told us that they contracted with moving companies to deliver
voting equipment to polling places prior to Election Day. For example,
election officials in a jurisdiction in Pennsylvania told us that they
contract with a moving company that transports about 1,000 DREs to about
400 polling places in

41On the basis of our nationwide survey, we estimate that all of the
jurisdictions that provided ballots in a language other than English
provided ballots in Spanish. Our survey results also showed that some
large jurisdictions provided ballots in other languages, such as Chinese,
Vietnamese, Korean, and Tagalog, for the November 2004 general election.

Page 203 GAO-06-450 Elections

the week prior to Election Day. Election officials in a jurisdiction in
Nevada told us that to ensure that voting machines were delivered to the
correct polling places, they bar-coded each DRE and also assigned a bar
code to each polling place. Upon delivery, contract movers used scanners
to read the bar codes on each DRE and the bar code for the specific
polling place. Prior to Election Day, these officials said that teams of
election staff technicians then went to each polling place to set up the
DREs and verify the scanned bar codes. After setting up the DREs, the
rooms in which they were located were secured until Election Day. In
contrast, in a jurisdiction we visited in New Hampshire, two election
workers delivered 12 optical scan counters to the 12 polling places at
4:00 a.m. on Election Day. Figure 41 shows stored voting equipment-with
accompanying delivery instructions for each DRE for 1 location-in 3 large
jurisdictions we visited that needed to be prepared and delivered to
polling places prior to  Election Day.

Figure 41: DRE and Optical Scan Voting Equipment That Large Jurisdictions Had to
                   Prepare and Deliver Prior to Election Day

                                  Source: GAO.

Election Officials in Some Jurisdictions We Visited Reported Experiencing
Long Lines on Election Day, while Officials in Other Jurisdictions Did Not

Long voter wait times are a problem that election officials try to avoid.
However, voters waiting in line at the polls was an issue identified in
reports reviewing the November 2004 general election. These reports
identified a variety of factors, including confusion about a voter's
registration status, ballot or voting equipment shortages, or
malfunctioning voting equipment that led to long voter wait times. We
asked election officials during our site visits whether or not any polling
places in their jurisdictions had long lines during the November 2004
general election and to describe factors they thought contributed to or
helped to reduce long lines.

Election officials in 17 of the 28 jurisdictions we visited reported
having long lines at one or more polling places in their jurisdiction at
some point on Election Day. However, there was variation in the reported
voter wait times, times of day, and numbers of polling places with lines.
For instance, election officials described voter wait times that ranged
from 15 minutes to 1  1/2 hours. Some election officials reported that the
longer lines occurred in the morning; others told us that they kept
polling places open past the official closing time to accommodate voters
who were in line when the polls closed. Election officials in over half
these 17 jurisdictions attributed long lines to higher than expected voter
turnout, both in general and at peak voting times. Some of these
jurisdictions were located in states where the presidential race was
considered close (often referred to as "battleground states").42 For
example, the election official in a jurisdiction in Nevada attributed long
lines to using a new voting system in addition to being a battleground
state and encountering high voter turnout. This official estimated that
there were between 30,000 and 35,000 more voters for the November 2004
general election than in previous elections. Election officials in 2
jurisdictions we visited in Ohio told us that higher than expected voter
turnout in some precincts led to long lines. For example, election
officials in 1 of these jurisdictions reported that at a polling place
where two precincts were located there was higher than expected turnout
because of a school board race. According to these officials, at this
polling place there was a single line for voters from both precincts to
check in at

42EAC, using various news media sources, identified 17 states deemed to be
most competitive in the 2004 presidential contest. These "battleground or
highly contested" states included Arkansas, Arizona, Colorado, Florida,
Iowa, Michigan, Minnesota, Missouri, Nevada, New Hampshire, New Mexico,
Ohio, Oregon, Pennsylvania, Washington, West Virginia, and Wisconsin. We
visited 8 of these 17 states during our site visits; alternatively,  16
of the 28 jurisdictions we visited were located in these battleground
states.

Page 206 GAO-06-450 Elections

Overloaded Phones on Election Day Posed Problems for Some Large and Medium
Jurisdictions

the registration table, and this line backed up. Election officials in
another jurisdiction in Ohio told us that some precincts had long lines,
and one precinct in particular had a waiting time of up to 1 hour. These
officials said that one precinct closed 30 to 45 minutes after closing
time for the voters that were in line at 7:30 p.m.

Election officials in 11 of 28 jurisdictions we visited told us that none
of the polling places in their jurisdictions had long lines, and some
described factors that helped to reduce or prevent lines. High voter
turnout prior to Election Day-either during early voting or through
absentee voting-was one factor they identified. For example, election
officials in 2 jurisdictions we visited-a second jurisdiction in Nevada
and 1 in New Mexico-told us that about 60 percent of those who cast
ballots voted early or absentee. Election officials in a jurisdiction we
visited in Washington (which reported that it did not require or allow
early voting) told us that they attributed their lack of long lines on
Election Day to the fact that two-thirds of voters in their jurisdiction
vote by absentee ballot. Election officials in a jurisdiction in Florida
reported that in planning for the November 2004 general election, they
decided to encourage early and absentee voting as alternatives to Election
Day voting in anticipation that there would be heavy turnout for the
general election. Their voter education campaign, which included buying
airtime on radios and in movie theaters, stressed early voting options. In
the end, about 40 percent of voters cast early ballots, which, according
to election officials, made crowds easier to manage on Election Day.

On Election Day, poll workers may need to communicate with election
officials at the central office for a variety of reasons-to inquire about
a person's eligibility to vote if his or her name does not appear in the
poll book, to report voting equipment problems, or to report other issues
that could occur at a polling place on Election Day. On the basis of our
nationwide survey of local jurisdictions, for the November 2004 general
election, we estimate that for 48 percent of all jurisdictions, the
primary means of communication between polling places and the central
office was telephones installed at polling places. Cell phones were also
used as a primary means of communication in some jurisdictions. For
example, on the basis of our local survey results, we estimate that for 25
percent of all jurisdictions, personal cell phones were the primary means
of communication.

Having inadequate communication lines on Election Day was a problem for
election officials in the November 2000 election, as we noted in our
October 2001 report on election processes.43 On the basis of our 2005
survey of local jurisdictions, communication problems between polling
places and the election office on Election Day were a challenge for some
jurisdictions in the November 2004 election, and these problems varied by
the size category of jurisdiction, with more large jurisdictions
encountering major problems than medium and small jurisdictions. We
estimate that  36 percent of large jurisdictions, 63 percent of medium
jurisdictions, and  89 percent of small jurisdictions encountered no
major problems with the communication system used at polling places. Small
jurisdictions may not have experienced communication problems on Election
Day for a variety of reasons, among them because a single polling place is
located in the same building as the central election office, allowing the
election officials to be physically present to resolve any questions or
issues. Election officials in small jurisdictions provided comments on our
nationwide survey of local jurisdictions about the primary communication
system used in their jurisdictions on Election Day, including "personal
contact-the clerk's office is across the hall from the polling place,"
"[we] yelled across the room," or "we are the central office and the
polling place." In addition, the election official in the small
jurisdiction we visited in New Hampshire told us that the town clerk was
on site at the one polling place.

Election Day communication problems encountered by some large and medium
jurisdictions included overloaded phones because of the volume of calls.
On the basis of our local jurisdictions survey, we estimate that  49
percent (plus or minus 8 percent) of large jurisdictions, 14 percent of
medium jurisdictions, and 1 percent of small jurisdictions experienced
overloaded phone systems.44 The differences between all size categories
are statistically significant. Election officials in many large
jurisdictions we visited reported receiving numerous phone calls on
Election Day, both from polling places and from the public. In addition to
poll workers calling from polling places, election officials at the
central office may receive phone calls from citizens asking about the
location of their polling place or whether or not they are registered to
vote. For example, a large jurisdiction we visited in Nevada reported
receiving over 35,000 calls on Election Day

43See GAO-02-3.

44The 95 percent confidence interval for the large percentage is greater
than the +/- 7 percent sampling error for the sample.

Page 208 GAO-06-450 Elections

Election Officials in Some Jurisdictions We Visited Reported That
Third-Party Observers Were a Challenge on Election Day 2004

2004, about three times the number reportedly received in 2000. Election
officials reported that most calls received were from people wanting to
know whether or not they were registered or where their polling place was,
despite providing polling place locations on their Web site, printing the
locations in the newspaper, and mailing a sample ballot listing polling
place locations to every registered voter in the jurisdiction. Election
officials in  2 other large jurisdictions in Florida and Kansas reported
that the volume of calls received was extremely high and that most
inquiries concerned voter eligibility. In 1 of these 2 jurisdictions,
election officials told us that many poll workers could not get through to
the elections office to verify voter registration information, which may
have increased the number of provisional ballots issued during the
election.

Election officials in many of the large jurisdictions we visited reported
taking steps to manage, or even reduce, the volume of calls from both
polling places and the public. These actions included setting up call
centers or phone banks, installing additional phone lines in their
offices, or hiring temporary workers. For example, election officials in a
large jurisdiction in Pennsylvania reported that after experiencing
problems being able to handle the volume of calls on Election Day 2000,
they implemented a call center at their office with 30 phone lines for the
November 2004 election. While these election officials reported receiving
"a lot" of calls for the 2004 general election, they said they were able
to successfully handle the volume because of the new phone lines. Election
officials in a large jurisdiction in Illinois reported that a feature, new
for the November 2004 election, on the jurisdiction's Web site that
allowed voters to determine their polling place online helped to reduce
the number of phone calls received from people asking about polling
location.

After the November 2004 general election, some reports highlighted
allegations of voter intimidation by third parties (e.g., poll watchers,
observers, or electioneers) at polling places. To gain a better
understanding of the extent to which this alleged behavior occurred and
because the range of behaviors and circumstances in which they could have
occurred was difficult to capture on a structured survey, we asked
election officials during our site visits about challenges they faced
conducting voting on Election Day-specifically, we asked them about any
problems they encountered with voter intimidation. Election officials in
19 of the  28 jurisdictions we visited did not report experiencing
problems with third parties on Election Day. However, election officials
in 9 jurisdictions we visited in battleground states reported challenges
with disruptive third-party activities.45 In some instances these third
parties simply increased the number of people that poll workers were to
manage at a polling location; in others, election officials told us
third-party observers provided misinformation to voters or even used
intimidation tactics.

Election officials in a jurisdiction in Nevada told us that poll watchers
were the biggest challenge on Election Day. Poll watchers, according to
election officials, had been bused in from another state to observe the
election because Nevada was a battleground state, which led to having 14
poll watchers at some locations. These officials noted that while most
poll watchers simply observed, the poll watchers did increase the number
of people at polling places, creating more for poll workers to manage.
Election officials in other jurisdictions reported that third-party
behavior negatively affected poll workers and voters. For example,
election officials in a jurisdiction in Pennsylvania reported that one of
the biggest challenges on Election Day was managing poll workers' stress
levels in an especially contentious election where poll watchers and
observers yelled at them throughout the day. Election officials in another
jurisdiction in Nevada told us that outside observers' behavior was
disruptive and noted that the observers were contentious, violated
electioneering limits at the polling place, and questioned every action
that poll workers made. Election officials in a jurisdiction in Colorado
reported that at one polling location on a college campus, poll watchers
and representatives of a national organization were encouraging students
to go to the polling place at one time to create a disruption. Students
were also being encouraged to get back in line after they had voted, which
caused long lines for other voters. Election officials said that they
ended up calling security officers to help manage the situation.

In other instances, election officials reported that observers provided
misinformation to voters or even used intimidation tactics. Election
officials in a jurisdiction in Florida reported that third-party
organizations caused confusion at polling places by misinforming voters
and staging demonstrations. In a jurisdiction we visited in Colorado,
election officials told us that poll watchers caused problems at some
polling places by providing misinformation to voters, such as informing
them that their provisional ballots would not be counted. In a
jurisdiction in New Mexico, election officials said that one polling place
had to remain open until 

45As previously noted, during our site visits we visited 8 of the 17
states defined by EAC as battleground states.

Page 210 GAO-06-450 Elections

Concluding Observations

10:30 p.m. because voters were encouraged by local political advocates to
go to that polling place to vote even though the polling location for
their precinct had been changed. As a result, according to these
officials, hundreds of provisional ballots were cast at the polling place,
which made for long waiting times. Election officials in another
jurisdiction in New Mexico reported that outside candidate advocates and
observers from political parties tried intimidation tactics and treated
people at the polls "terribly." For example, these election officials told
us that some advocates were observed taking photographs of the license
plates of Hispanic voters as they arrived at polling places.

We did not ask a specific question about third-party activities at polling
places on our survey of local jurisdictions because of the complexities in
capturing the range of alleged behaviors on a structured survey. However,
we asked whether local election jurisdictions maintained a written record
of issues that occurred on Election Day and, if so, what issue or problem
occurred most frequently on Election Day. Several election officials from
jurisdictions in battleground states that provided comments on our
nationwide survey wrote that electioneering or poll watchers did. For
example, election officials from Florida, Colorado, and Iowa wrote "voters
complained about being harassed by demonstrators while waiting in line to
vote," "poll watchers acting aggressively," and "poll watchers (who were
attorneys, mostly) were interfering with the process, intimidating
precinct officials, and giving erroneous advice to voters who showed up at
the wrong polling place."

Administering an election in any jurisdiction is a complicated endeavor
that involves effectively coordinating the people, processes, and
technologies associated with numerous activities. Many of the challenges
that election officials reported encountering in preparing for and
conducting the November 2004 election were not new. Recruiting and
training an adequate supply of poll workers, finding accessible polling
places, and managing communications on Election Day were challenges we
identified in our October 2001 report on the November 2000 election.

Data from our local elections jurisdiction survey and site visits to  28
locations, indicate that more large, and to some extent medium,
jurisdictions encountered challenges in preparing for and conducting the
November 2004 general election than did small jurisdictions. This is not
surprising. Larger, diverse jurisdictions may face challenges smaller
jurisdictions do not, such as recruiting poll workers with non-English
language skills. Larger jurisdictions are also likely to need to rely to a
greater degree on technology to manage their elections administration
process, and this brings its own set of challenges. The complexity of
administering an election and the potential for challenges increase with
the number of people and places involved, the ethnic diversity and
language skills of the voting population, and the scope of activities and
processes that must be conducted. Many of the election officials in large
jurisdictions we visited told us that being well prepared, having
established policies and procedures in place, and having qualified
election staff were factors that contributed to a smooth Election Day. One
problem that occurred on Election Day in some jurisdictions that election
officials reported encountering was the actions of poll watchers and other
third parties that election officials considered disruptive. This presents
another issue that election officials may need to include in their
Election Day preparations and training.

Chapter 5

Provisional Voting

A goal of the election process is to ensure that every eligible voter is
able to cast a vote and have that vote counted. In the November 2000
general election, reports of some voters showing up at the polls and not
being able to vote raised concerns about eligible voters' names not
appearing on the voter registration list at the polling place or poll
workers not otherwise being able to determine voters' eligibility. While
many jurisdictions reported in 2001 having at least one procedure in place
to help resolve eligibility questions for voters whose names did not
appear on a polling place registration list, only 20 states plus the
District of Columbia reported using some form of provisional ballot for
the 2000 general election.1

One of the major changes since the 2000 general election has been the
implementation of a HAVA provision requiring, in general, that states
permit individuals, under certain circumstances, to cast provisional
ballots in elections for federal office. In general, under HAVA, voters
who claim to be eligible to vote and registered in the jurisdiction they
desire to vote in but whose names do not appear on the polling place
registration list are to be allowed to cast provisional ballots in a
federal election. These ballots are called provisional because they are
counted only if an election official determines that the voter is eligible
under state law to vote. In terms of ballot access, provisional ballots
benefit voters by allowing an individual to cast a vote, in general, when
there is some question as to the individual's eligibility such as when the
individual's name is not on the registration list or the individual's
eligibility has been questioned by an election official. In terms of
ballot integrity, provisional ballots benefit election officials by
allowing them to determine voter eligibility prior to counting such
ballots (i.e., verifying provisional ballots).

In this chapter, we describe (1) events that preceded HAVA's provisional
voting requirements, (2) how states and local jurisdictions implemented
the requirement to provide provisional ballots, (3) how states and local
election jurisdictions qualified provisional ballots for counting, and (4)
the difficulties of estimating and comparing the number of provisional
ballots that were cast and counted.

1GAO-02-3.

                                    Overview

                          Chapter 5 Provisional Voting

Concerns were raised with respect to the November 2000 election that some
eligible voters were not allowed to vote because of questions regarding
the voters' eligibility. HAVA required that by January 1, 2004, most
states permit the casting of provisional ballots in elections for federal
office by voters who affirm in writing that they believe they are eligible
to vote and registered in that jurisdiction, but are not found on the
voter registration list.2 Such states are also required under HAVA to
provide provisional ballots in federal elections under other circumstances
such as for certain voters who registered by mail and do not have required
identification, and where an election official asserts that an individual
is ineligible to vote. Provisional votes cast under HAVA's provisional
voting requirements are to be counted in accordance with state law if
election officials determine that the voter is eligible to vote under
state law. Under HAVA, 6 states are exempt from the act's provisional
voting requirements because they either permitted the voter to register on
Election Day or did not require voter registration.3

On the basis of reports from state election officials and in local
election jurisdictions we surveyed and visited, states and local
jurisdictions varied in a number of ways regarding how they implemented
HAVA's provisional voting requirements in the November 2004 election.
Among other things, we found variation in the

     o additional circumstances, apart from those circumstances specified in
       HAVA, where a provisional ballot would be offered, such as when voters
       claimed they did not receive an absentee ballot;
     o design of ballots themselves and how they were tracked; and
     o voting method used for casting provisional ballots, such as optical
       scan ballots or DRE.

2HAVA section 302 (codified at 42 U.S.C. S: 15482).

3Under HAVA, states that had either (1) no voter registration requirements
for voters with respect to federal elections (North Dakota) or (2) polling
place registration on Election Day with respect to federal elections
(Idaho, Minnesota, New Hampshire, Wisconsin, and Wyoming) in effect on and
after August 1, 1994, are not subject to HAVA's provisional voting
requirements.

Page 214 GAO-06-450 Elections

Events in the November 2000 Election Preceded the HAVA Requirement for
Provisional Ballots

With respect to the counting of provisional votes, states reported various
differences in their counting processes such as the prescribed location
from which a voter must cast a provisional ballot in order for it to be
counted. Also, with respect to the counting of provisional ballots,
according to our estimates from our survey of local election jurisdictions
nationwide, a voter not meeting residency requirements was the most
frequently cited problem, followed by insufficient evidence that the voter
was registered. In jurisdictions we visited, election officials also
varied in how they handled a lack of information from the voter that was
needed to verify a provisional ballot.

National figures on provisional ballots for the November 2004 election are
difficult to estimate because of a lack of data on provisional ballots
cast and counted, and variation in how states implemented provisional
voting. Nevertheless, we estimate that between 1.1 million and 1.7 million
provisional ballots were cast in the November 2004 election. The variation
in how provisional voting was implemented makes it difficult to compare
the use and counting of provisional ballots among jurisdictions. A number
of factors can affect the number of provisional ballots cast and counted.
For example, one such factor could be an instance in which the polling
location hours were extended and votes cast during the extended hours were
cast provisionally.

Following the November 2000 election, in our October 2001 comprehensive
report on election processes nationwide, we noted that the biggest
problems on Election Day involved resolving questions about voter
eligibility.4 Typically, a voter's eligibility is established before a
voter receives a ballot, most often by a poll worker examining a poll book
or registration list for the person's name. If the name appears on the
list and other identification requirements are met, the voter is given a
regular ballot and is allowed to vote. We also noted in our report that in
the November 2000 election, a large number of voters with eligibility
issues created frustration for voters, long lines, and problems
communicating between the polls and election headquarters as workers tried
to resolve eligibility issues.

4GAO-02-3.

For the 2000 general election, when the voter's name did not appear on the
registration list, we reported in October 2001 that jurisdictions had
different procedures for dealing with the question of the voter's
eligibility. More specifically, we reported that 20 states plus the
District of Columbia used some form of provisional ballot when a voter's
name was not on the voter list, with verification of registration
conducted after the election.5 As we reported, provisional balloting
measures went by different names among the states, including provisional
ballot, challenged ballot, special ballot, emergency paper ballot, and
escrow ballot. Further, in 5 states in the 2000 general election, we
reported that voters could complete an affidavit when voting with no
further verification of their registration information being required by
state law prior to the ballot being counted.

The U.S. Census Bureau estimated that of the 19 million registered voters
who did not vote in 2000, 6.9 percent did not vote because of uncertainty
regarding their registration.6 In our October 2001 report, we noted that
headlines and reports questioned the effectiveness of voter registration
by highlighting accounts of individuals who thought they were registered
being turned away from polling places on Election Day and jurisdictions
incorrectly removing the names of eligible voters from voter registration
lists. Our report also found that almost half of the jurisdictions
nationwide in 2000 reported having problems with registration applications
submitted at motor vehicle agency offices that election officials believed
could result in individuals showing up at the polls to vote and
discovering that they were not registered.

Numerous recommendations were made for federal regulations to require that
all states provide provisional voting. For example, the Federal Election
Commission in June 2001 recommended that all states devise procedures for
voters to cast provisional ballots at the polls under certain conditions,
as did the National Commission of Federal Election Reform in August 2001
and the National Task Force on Election Reform in July 2001, among others.

5See GAO-02-3, table 22, for a list and description of measures, including
provisional voting requirements, in the 50 states and the District of
Columbia to address instances of voter names not appearing on registration
lists for the 2000 general election.

6U.S. Census Bureau, Voting and Registration in the Election of November
2000, U.S. Department of Commerce, February 2002.

Page 216 GAO-06-450 Elections

Under HAVA, in an election for federal office, most states are to permit
individuals to cast a provisional ballot under certain circumstances.7 The
statutory deadline for implementing HAVA's provisional voting requirement
was January 1, 2004. For federal elections, states are, in general,
required to allow the casting of a provisional ballot by an individual

     o asserting to be registered in the jurisdiction in which the individual
       desires to vote and eligible to vote but whose name does not appear on
       the official list of eligible voters for the polling place, or
     o whom an election official asserts to be ineligible to vote, or
     o who registered to vote by mail but does not have (and has not
       previously provided) the required registration identification when
       trying to vote in person or by mail, or
     o casting a vote pursuant to a court order or other type of order
       extending poll closing times.

HAVA requires that an individual be permitted to cast a provisional ballot
upon the execution of a written affirmation before an election official at
the polling place. The written affirmation must state that the individual
is registered to vote in that jurisdiction and eligible to vote in that
election. HAVA specifies that either the provisional ballot or the written
affirmation information be transmitted to an appropriate election official
for a determination as to whether the individual is eligible to vote under
state law. Under HAVA, if an individual is determined to be eligible, the
provisional ballot is to be counted as a vote in accordance with state
law. Election officials, under HAVA, are to give the individual written
information on how to ascertain whether the vote was counted and, if the
vote was not counted, the reason the vote was not counted. HAVA directs
that state or local election officials establish a free access system,
such as a toll-free number, for provisional voters to ascertain such
information.

7Although many states had some form of provisional balloting prior to the
passage of HAVA, 44 of the 50 states and the District of Columbia were
required to permit certain individuals to cast provisional ballots for the
2004 general election. Under HAVA, 6 states were exempt from HAVA's
provisional voting requirements. States that had either no voter
registration requirements for voters with respect to federal elections
(North Dakota) or polling place registration on Election Day with respect
to federal elections (Idaho, Minnesota, New Hampshire, Wisconsin, and
Wyoming) in effect on and after August 1, 1994, are not subject to HAVA's
provisional voting requirements.

State and Local Jurisdictions Varied in Their Implementation for Providing
Provisional Ballots for the November 2004 Election

While HAVA established conditions under which an individual must be
allowed to cast a provisional ballot, states are not prohibited from
offering provisional ballots for other reasons, or from using ballots with
other names (e.g., a challenged ballot) to serve provisional vote
purposes. HAVA explicitly provides that the specific choices on the
methods of complying with certain act requirements, including the
provisional voting requirements, are left to the discretion of the state.8
In addition, HAVA provides that a state may establish election technology
and administration requirements that are stricter than HAVA requirements,
so long as they are not inconsistent with other specified federal
requirements.9

On the basis of reports from state election officials and in local
election jurisdictions we surveyed and visited, states and local
jurisdictions provided for provisional voting in a variety of ways for the
November 2004 election. These differences contributed to the variation in
the number of provisional votes cast among jurisdictions.

The results of our state survey of election officials show that states
reported using new or existing legislative or executive actions (which
included executive orders, directives, regulations, or policies) to
implement HAVA's provisional voting requirements. Specifically, our state
survey showed

o  27 states reported enacting new legislation or taking executive action
to meet HAVA's provisional voting requirements;

8HAVA section 305 (codified at 42 U.S.C. S: 15485).

9HAVA section 304 (codified at 42 U.S.C. S: 15484). For example, Arizona
submitted an inquiry to Department of Justice's Civil Rights Division
asking whether it was permissible under HAVA for a state to mandate that a
potential voter show identification at the polls prior to receiving a
provisional ballot. The Civil Rights Division responded, in part, on
September 1, 2005, that while HAVA requires states to allow voters who
meet certain specified conditions the opportunity to cast a provisional
ballot, states are free to prescribe their own rules for deciding whether
to count those ballots. (September 1, 2005, letter to the state of Arizona
from the Department of Justice's Civil Rights Division).

     o 11 states and the District of Columbia reported using the state's
       existing legislative or executive action to meet the requirements;10
     o 7 states said HAVA provisional requirements were met by a combination
       of new legislation or executive action and existing actions;11
     o 5 states (Idaho, Minnesota, New Hampshire, North Dakota, and
       Wisconsin), in response to the question of how their state established
       the provisional voting requirements set forth in HAVA, answered that
       they were exempt from such requirements; these 5 states are exempt
       from HAVA provisional requirements, in general, because they have
       same-day voter registration or no voter registration.12

Connecticut officials responded, for example, that the state enacted
legislation after HAVA to establish HAVA provisional voting requirements.
Connecticut state laws were enacted in June 2003 related to the
application for a provisional ballot, casting of the ballot, and
determination of eligibility for counting of provisional ballots, among
other things. In contrast, Alaska election officials reported that
existing legislation met HAVA's provisional voting requirements. According
to Alaska's 2005 updated HAVA plan, the state had an existing provisional
voting process known as Questioned Voting. This process, established in
the early 1980s, required only minimal changes to meet HAVA provisional
voting requirements. Alaska requires use of a questioned ballot for any
voter who votes at a polling location where his or her name does not
appear on the precinct register, or if the voter does not have
identification and is not personally known by the election official. In
our state survey, New Jersey reported meeting HAVA provisional voting
requirements with a combination of existing and new legislation. In one
New Jersey jurisdiction we visited, election officials stated that state
provisional voting procedures were first established in

10Alaska, Colorado, Florida, Georgia, Kansas, Maine, Missouri, New York,
North Carolina, South Carolina, and West Virginia reported using existing
legislative or executive action to meet HAVA requirements.

11Arizona, California, Iowa, Michigan, New Jersey, Ohio, and Washington
reported meeting HAVA requirements with a combination of new and existing
state actions.

12Wyoming is also exempt from HAVA's provisional voting requirements.
States are exempt because they had either (1) no voter registration
requirements for voters with respect to federal elections (North Dakota)
or (2) polling place registration on Election Day with respect to federal
elections (Idaho, Minnesota, New Hampshire, Wisconsin, and Wyoming), in
effect on and after August 1, 1994. Despite the exemption, Wyoming
reported enacting legislation to authorize some measure of provisional
voting to be consistent with HAVA.

Paper Ballots and DRE Were Voting Methods Used to Cast Provisional Ballots

1999. According to these officials, the state amended its provisional
ballot election law after HAVA to allow use for voting by court (or other)
order after the polls have closed, and by first-time mail registrants who
do not provide identification.

Election officials in 25 of the 26 jurisdictions we visited that provide
for provisional voting told us that they used some form of paper ballot
for Election Day provisional voting for the November 2004 election.13 For
example, election officials in the Illinois jurisdictions we visited said
that the regular punch card ballot was used by provisional voters, and
then placed in provisional ballot envelopes. In the New Jersey
jurisdictions we visited officials said that provisional votes were cast
on paper ballots that could be counted with optical scan machines (if
voters were determined to be eligible). Election officials in Connecticut
jurisdictions said that they used hand-counted paper ballots for
provisional voters.

According to election officials in 1 Ohio jurisdiction and 1 Nevada
jurisdiction, DRE was used for Election Day provisional voters.14
According to election officials or documents they provided in the 2
jurisdictions we visited that used DRE for provisional voting on Election
Day, the processes used for casting provisional votes were as follows:

o  In the Ohio jurisdiction, election officials said voters first
completed an affidavit statement with a preprinted code number, and signed
a special section of the poll book. The poll worker then inserted a unit
into the DRE that contained the ballot for the precinct. The poll worker
then pressed the provisional ballot selection on the DRE and entered the
code number for the individual voter associated with the voter's affidavit
statement. The individual then voted.

13Two additional jurisdictions we visited were in New Hampshire, which is
not subject to HAVA provisional voting requirements.

14In each of these jurisdictions paper ballots were used for some
provisional voting-for those casting a provisional vote Election Day in
the election office in Ohio and those casting a provisional ballot during
early voting in the Nevada jurisdiction. Voters in Nevada, including early
voters, who were first-time voters who registered by mail and did not
provide identification when they registered and could not provide required
identification when at the polls, were allowed to cast a provisional
ballot. Election officials in the other Nevada jurisdiction also told us
they used DRE for provisional voting but not on Election Day. This
jurisdiction used DREs for provisional early voting and optical scan paper
ballots on Election Day.

The Provisional Ballot Design and Tracking Procedures Varied among States

o  In one Nevada jurisdiction, DREs were used for Election Day provisional
voting, but optical scan ballots were used for provisional voters
participating in early voting. According to the poll worker's manual
provided by election officials, Election Day provisional voters completed
an affirmation with identifying information and the reason they were
casting a provisional ballot. As described to us by election officials at
this jurisdiction, the poll worker then added precinct information, and
both signed the affirmation. The poll worker then activated the DRE with a
card. To indicate that the ballot was provisional, the poll worker pressed
"0" and the machine provided a provisional voter identification number
that the poll worker copied onto the voter affirmation and provisional
voter receipt. The voter then voted.

According to election officials in the jurisdictions we visited, the
design of provisional ballots varied for the November 2004 election. The
provisional ballot differences included variation in terms of the races
included, ballot and envelope color, the envelopes they were placed in,
and the information included on the provisional ballot envelopes. For
example, in the Nevada jurisdictions, the provisional ballot only included
races for federal offices, while in the Kansas jurisdictions, officials
said that the provisional ballot was the same as a regular ballot. In 1
Georgia jurisdiction, election officials stated that they were using an
absentee ballot for provisional voters but were inserting it into a
salmon-colored envelope, whereas in an Illinois jurisdiction we visited,
"Provisional" was printed in pink letters across the punch card ballot
used in that jurisdiction so that these ballots were distinguishable from
other ballots.

The provisional ballot envelopes also varied in terms of what information
was provided in the jurisdictions we visited, according to example
envelopes provided to us (or described) by election officials. The outside
of the provisional ballot envelopes in most of the jurisdictions we
visited served as the voter's written affirmation that is required by
HAVA. For example, in a jurisdiction in Illinois, the ballot envelope
included instruction to voters on how to cast a provisional ballot; in a
Florida jurisdiction (as well as in Illinois) the provisional envelope
includes information on the reason why the provisional ballot was cast. In
New Mexico and Colorado jurisdictions we visited, the envelope included a
tearoff tab with information on how voters could find out whether their
vote counted, and if not, why it was not counted. In addition, election
officials in some jurisdictions we visited described provisional ballots
being placed in envelopes, sometimes with a second security envelope
covering the ballot inside. Figure 42 shows an example of a provisional
ballot envelope.

                     Figure 42: Provisional Ballot Envelope

Source: Broward County, Florida.

Officials in jurisdictions we visited described a variety of methods used
for tracking provisional ballots in the November 2004 election. Methods
included having individual ballots numbered, maintaining an inventory or
log, accounting for provisional ballots at the beginning and end of
Election

Additional Circumstances for Using Provisional Ballots Varied for the
November 2004 Election

Day, and using specially colored ballots or envelopes for holding
provisional ballots. The following are examples of how election officials
in four jurisdictions we visited said they tracked provisional ballots for
the November 2004 election:

     o In a Pennsylvania jurisdiction, election officials tracked provisional
       ballots cast at the polling place on a form provided by the election
       officials. Provisional ballots were marked with a sticker indicating
       that they were provisional. The sticker also had an identification
       number for tracking the ballot, and the voter was provided a receipt
       with the identification number to use when calling for information on
       the status of their ballot. All provisional ballots were placed inside
       of green envelopes.
     o In a New Mexico jurisdiction, an election official said that ballots
       were numbered sequentially, so that the poll workers could track the
       numbers. The precinct judges certified the numbers of the ballots they
       received, used, delivered, and destroyed.
     o In a New Jersey jurisdiction, the municipal clerk issued a specific
       number of provisional ballots (25) to each precinct, with a "Custody
       Receipt" form that identified who was in possession of the orange bag
       with the provisional ballots and an accounting of all ballots
       originally issued. A ballot that had been voted was enclosed in a gray
       envelope and then put back in the orange bag.
     o In a Kansas jurisdiction, separate poll books, separate envelopes for
       provisional ballots, and separate pouches for envelopes containing
       provisional ballots (all blue in color) facilitated tracking the
       ballots as separate items from regular Election Day ballots. No
       tracking of the actual ballot occurred (before it was voted) because
       the same optical scan paper ballot was used for regular Election Day
       voters.

Apart from permitting voters to cast provisional ballots under the
circumstances specified in HAVA, some jurisdictions we surveyed or spoke
with had additional reasons for providing provisional ballots to voters in
the November 2004 election and other types of ballots that could be used
for different circumstances. In addition, election officials in
jurisdictions we visited told us about different approaches for offering
provisional ballots.

Provisional Ballots Were In the local election jurisdictions we visited,
election officials described Provided for a Variety of Reasons various
circumstances, in addition to those required by HAVA, in which a in
Jurisdictions We Visited provisional ballot was provided to a prospective
voter in the November

2004 election. The additional circumstances under which provisional

ballots were provided are established by state officials. For example,

     o In one Colorado jurisdiction we visited, election officials stated
       that provisional ballots were available to voters who did not have the
       identification required of all voters in the state and also available
       if a person was listed as a felon in the poll book. Further, election
       officials told us that the Colorado Secretary of State issued guidance
       just prior to the 2004 general election that allowed individuals-
       claiming to have registered at a voter registration drive but for whom
       the jurisdiction had no record-to vote provisionally.
     o Election officials in jurisdictions we visited in Colorado, Florida,
       Kansas, Ohio, and Washington said that voters claiming they had not
       received their absentee ballots were provided with provisional
       ballots. In other jurisdictions, such as the 2 we visited in
       Connecticut, voters were allowed to vote regularly if their absentee
       ballot did not arrive.
     o Kansas election officials reported that they allowed voters to cast
       provisional ballots if the voter did not trust the voting machines and
       wanted a paper ballot, or if the voter had a different last name than
       the listed one because of marriage or divorce.

The extent to which voters are provided with provisional ballots varied
depending on whether states required identification of all voters or only
certain voters, according to our state survey. Some states reported that
they require all voters to provide identification; some reported that they
require only provisional voters to produce identification, while others
reported that they do not require identification from voters other than
firsttime voters who registered by mail, as required by HAVA.15 Chapter 4
on conducting elections discusses state requirements for voter
identification for all voters. According to our state survey, 6
states-Arizona, Massachusetts, Michigan, New Mexico, Utah, and
Wisconsin-reported

15HAVA mandates all states to require identification from first-time
voters who registered by mail unless a copy of the identification is
included with their mail-in voter registration application.

Page 224 GAO-06-450 Elections

Alternatives to Provisional Voting Available in Some Jurisdictions We
Visited

requiring identification from only provisional voters in the November 2004
election, but Michigan and Utah reported allowing an alternative to
identification for provisional voters who did not have required
identification. In Michigan, for example, a voter receiving a provisional
ballot who was unable to meet the identification requirement was
permitted, according to election officials responding to our state survey,
to fax, mail, or hand-deliver an acceptable form of photo identification
to the clerk anytime during the 6 days following the election.

Some jurisdictions we visited reported that Election Day voting options
other than provisional ballots were available. For example, election
officials in jurisdictions we visited in Ohio said that provisional
ballots were the only special ballots available for that election. In
contrast, in a New Mexico jurisdiction we visited, election officials said
the state offered an in-lieu-of ballot for voters who requested an
absentee ballot, and claimed it did not arrive. These election officials
said the in-lieu-of ballot was the same as a provisional ballot, but it
was placed in a different sleeve for later determination of whether an
absentee ballot had been cast or not. At a Connecticut jurisdiction we
visited, election officials described the state's presidential ballot,
available at the clerk's office on Election Day for the November 2004
election. A presidential ballot, according to election officials and
documents they provided, allowed voting for president and vice-president
by former Connecticut residents who had moved to another state within 30
days of the election and for that reason could not vote in their new state
of residence.16

Election officials in some jurisdictions we visited, such as 1
jurisdiction in Florida and 2 jurisdictions in New Jersey, said their
procedures allowed challenged voters to sign a statement, such as an
affidavit declaring their eligibility, and to vote on a regular ballot
that would be counted with other ballots on Election Day. According to
poll worker guidance provided by election officials in the Florida
jurisdiction, a written challenge must be submitted under oath and given
to the voter; then the voter has the right to submit an oath affirming his
or her eligibility. The polling place clerk and inspectors must resolve
the challenge by majority vote, providing a regular ballot if the decision
is in the prospective voter's favor. The guidance states that a challenged
voter who refuses to sign the oath must be offered a provisional ballot.
In both jurisdictions we visited in New Jersey, voters

16To be eligible for a presidential ballot, a person must be a citizen at
least 18 years old and have not forfeited electoral privileges because of
a disfranchising crime.

Page 225 GAO-06-450 Elections

who were challenged were not issued a provisional ballot, according to
documents provided by election officials. As stated in the poll worker
manual for one of the jurisdictions for the 2004 general election, a voter
who was challenged completed a challenged voter affidavit, as shown in
figure 43. The manual stated that the location's four poll workers take a
vote to decide whether the voter would be allowed to vote. On the basis of
the decision, the challenged voter cast a regular ballot or was not
allowed to vote, according to the manual (in case of a tie, the voter was
allowed to vote).

      Figure 43: Middlesex County, New Jersey, Challenged Voter Affidavit

                     Source: Middlesex County, New Jersey.

In our survey of local election jurisdictions nationwide, we asked for
information on the use of provisional ballots, challenged ballots, or
other types of ballots under various scenarios for the November 2004
election.17 Table 18 shows the extent to which we estimate that local
jurisdictions provided provisional ballots as compared to providing other
types of ballots.

Table 18: Estimated Percentages of Local Jurisdictions Nationwide That
Provided Provisional, Challenge, or Other Ballots for the November 2004
Election under Various Scenarios

                                        Percentage of jurisdictions providing
                                    Provisional Challenge Other
Scenario                             ballots  ballots  ballots         Not 
                                                                   applicable 
Individuals who registered by mail               49    6     7          39 
(without providing                                             
identification),voted for the first                            
time, and did not bring proper                                 
identification with them to the                                
polling place                                                  
Individuals who claimed they were at          48    7       13          33 
the correct polling place but                                  
were not on the voter registration                             
list at the polling place                                      
Individuals who were challenged by            23    23       3          49 
an election official as                                        
ineligible to vote                                             
Individuals who voted after the                   6    1     2          91 
polling place closing times when                               
the polling place was kept open late                           
because of a federal court,                                    
state court, or other order                                    
extending the polling place hours                              
Individuals who said they had                    18   1     19          60 
requested an absentee ballot but                               
that the ballot never arrived                                  

Jurisdictions Varied in Their Approach to Providing Provisional Ballots

Source: GAO 2005 survey of local election jurisdictions.

Note: Jurisdictions could indicate using more than one type of ballot for
a particular scenario; therefore, the row of percentages may not add to
100 percent.

Apart from permitting voters to cast provisional ballots under the
circumstances specified in HAVA, election officials in jurisdictions we
visited described differing approaches under which provisional ballots
were utilized for the November 2004 election.

Election officials in most of the 28 jurisdictions we visited said that in
the November 2004 election they would not refuse an individual a
provisional

17Jurisdictions in Idaho, Minnesota, New Hampshire, and North Dakota were
not asked questions about provisional voting because they are not subject
to HAVA's provisional voting requirements and do not provide for
provisional voting. Maine was inadvertently not asked to answer these
questions.

Page 228 GAO-06-450 Elections

ballot. In a Colorado jurisdiction, election officials said that election
judges were instructed to direct all voters meeting the criteria for
voting provisionally (e.g., claiming to be registered and eligible, but
with some eligibility question) to the provisional voting table. In 1
Nevada jurisdiction, the election official said that anyone could receive
a provisional ballot. He said that they had Las Vegas tourists who wanted
to vote a provisional ballot, even though they were informed that it would
not be counted. Election officials in 1 Washington jurisdiction said
voters knew that they could cast a ballot regardless of circumstances, and
election officials in the other Washington jurisdiction said that
provisional ballots served as a conflict avoidance tool at the polls.
Election officials in both New Mexico jurisdictions said that if a voter
was not on the registration list, he or she was immediately given a
provisional ballot. According to the New Mexico election officials,
precinct officials were not to direct a voter to the correct precinct;
instead, under the provisional voting rule, they were to offer a
provisional ballot to the voter.

Election officials in some other jurisdictions we visited told us that
poll workers may have taken certain steps before providing a voter with a
provisional ballot. In 1 Illinois jurisdiction, an election official said
that if a potential voter was not listed, the poll workers first tried to
determine if the voter was registered in another jurisdiction. If that was
the case, the poll workers then directed the voter to that jurisdiction,
but they did not refuse to provide a provisional ballot if a voter
requested one. In 1 Ohio jurisdiction, election officials told us that if
a voter was registered in Ohio, everything was done to get the voter to
the correct precinct. In a New Jersey jurisdiction we visited, election
officials explained that poll workers take several steps when the voter's
name was not listed in the poll book. Poll workers were instructed,
according to the poll worker's manual, to check the poll book for
misspellings or for the name being out of alphabetical sequence, and to
check the county street guide to see if the voter was in the wrong
location. Election officials in this jurisdiction also told us that voters
who were in the wrong location were directed to the correct location. They
added that voters who did not wish to vote provisionally were told to go
before a superior court judge to plead their cases.

In 5 jurisdictions we visited, election officials said there were
instances where election officials would refuse to provide a provisional
ballot on Election Day. In 1 Ohio jurisdiction, election officials said
that a provisional ballot was provided if the potential voter appeared at
the polling place. However, if the person came to the election office on
Election Day and no

Newness of Providing Provisional Ballots Presented Some Challenges in
November 2004

record of voter registration was found by the Registrar, then the voter
was not allowed to vote provisionally. A potential voter stating that he
or she was not registered or not a resident was a reason not to offer the
individual a provisional ballot, according to election officials in 1
jurisdiction in Nevada, and 1 in New Jersey, and both jurisdictions in
North Carolina. Officials in 1 Georgia jurisdiction we visited said that
an individual might not be offered a provisional ballot if he or she was
on the voter registration list and therefore eligible to vote a regular
ballot.

Whether a provisional ballot was provided or not might have been based, in
part, on the size of the jurisdiction and the familiarity of the poll
workers with the voters. Several election officials in small local
jurisdictions included in our nationwide survey made this point in written
comments. For example, comments included the following:

     o "This is a small township. We don't have the problems big cities have.
       People know who lives in the township. They know their neighbors."
     o "Most voters are personally known, including their addresses."
     o "We were told that the state voter list was the bible for the day. But
       we had one lady who should have been provisional but we all knew where
       she lived so we let her vote. It was the choir lady's niece. Her
       signature was on file."

In larger jurisdictions, poll workers might be less likely to know the
voters in the precinct and may have made greater use of provisional
ballots than in smaller jurisdictions.

Some jurisdictions we visited reported that knowing how many provisional
ballots to have available for the November 2004 election was a challenge.
However, on the basis of our survey of local jurisdictions, we estimate
that for the November 2004 election, only 1 percent of jurisdictions had a
shortage of provisional ballots. The difficulty with anticipating the need
for provisional ballots, according to an Illinois jurisdiction election
official, was that officials had no historical experience to rely upon in
deciding how many to make available at each site. In this jurisdiction,
provisional ballots were used for the first time in the November 2004
election, according to the election official. Similarly, in a Pennsylvania
jurisdiction we visited, election officials stated that they had no basis
to plan for the number needed, and that they had to rush to produce (e.g.,
placing a provisional ballot sticker over an absentee ballot) additional
provisional ballots at the last minute because some precincts needed more
than were initially allocated. Election officials in one Nevada
jurisdiction we visited said some polling places were overstocked while
others were understocked, requiring them to shuttle the ballots between
polling places. In a Colorado jurisdiction we visited, election officials
said that last-minute changes by state officials created a need for more
provisional ballots because this change allowed individuals who registered
during a voter registration drive but who were not on the voter list to
vote provisionally.

On the basis of our local survey, poll workers failing to follow
procedures for conducting provisional voting surfaced as an issue in some
jurisdictions in the November 2004 election. We estimate that 12 percent
of jurisdictions nationwide encountered poll worker performance problems
related to their failure to follow procedures with provisional voting. The
newness of the provisional procedures or last-minute changes in the
guidance were challenges that confused poll workers, according to election
officials in jurisdictions we visited. Specifically,

     o In a Georgia jurisdiction, election officials told us there was a
       question regarding whether several college students were eligible to
       vote provisionally, and state election officials were called for
       clarification (the students were allowed to vote provisionally).
     o In a Connecticut jurisdiction, election officials said poll workers
       were confused about the process, issuing provisional ballots in some
       cases before checking with the Registrar to try to locate the
       prospective voters in the statewide database.
     o In both Nevada jurisdictions, election officials we visited identified
       poll worker training needs; for example, in 1 of the Nevada
       jurisdictions election officials said provisional ballot materials
       were not adequately tracked and returned.
     o In an Ohio jurisdiction, election officials identified poll worker
       handling of provisional ballots as an area for improvement based on
       finding valid provisional ballots returned in envelopes for soiled and
       defaced ballots. In addition, they said about half of the provisional
       voters did not sign the poll book, as they were supposed to have done
       under this jurisdiction's requirements. Furthermore, voters were to
       place their provisional ballots in a colored provisional sleeve for
       determination of eligibility before the vote was submitted, but the
       election official estimated that

about 10 percent of the provisional ballots were placed directly in the

ballot box instead.

Some election officials in jurisdictions we visited described actions they
took to implement provisional voting that worked well for the November
2004 election. Several identified training given to poll workers that
prepared them for provisional voting, or had staff dedicated to handling
provisional votes, or poll workers with prior provisional voting
experience. For example, election officials in 1 Colorado jurisdiction
said that they had election judges whose sole responsibility was
conducting provisional voting. According to these election officials, the
election judges (i.e., poll workers) were well trained and sat at a
separate table to handle provisional voting. One jurisdiction we visited
in Illinois had specific instructions on the voter affidavit for election
workers to follow. Figure 44 provides an example of the affidavit.

Figure 44: Affidavit Containing Provisional Ballot Instructions for Poll
Workers in Champaign, Illinois

Source: Champaign County, Illinois.

State and Local Jurisdictions Reported Variation in Several Areas
Affecting whether Provisional Ballots Are Counted

HAVA specifies that voters casting ballots under HAVA's provisional
balloting requirements must, in general, execute a written affirmation
stating that they are registered in the jurisdiction in which they desire
to vote and that they are eligible to vote in that election. Polling place
officials, under HAVA, are to transmit either the ballot or the written
affirmation information to an appropriate election official for
verification to ascertain if the individual is eligible to vote under
state law. In the November 2004 election, state requirements regarding the
location from which voters had to cast their provisional ballot in order
for it to be counted (e.g., in the specific precinct in which the voter is
registered or anywhere within the county-city, parish, township-in which
the voter was registered) was one key difference among states. States also
varied in how missing voter information was handled and how voters were
informed whether their vote was counted or not. On the basis of our
national survey of local jurisdictions, the most frequent problem
encountered by local jurisdictions in counting provisional ballots was
that voters did not meet residency eligibility requirements for the
precinct or jurisdiction.

Location Where Voters Must Cast Their Ballots in Order to Be Counted
Varied among States

HAVA requires states to provide provisional balloting where, among other
things, individuals assert that they are registered in the jurisdiction in
which they desire to vote. The term "jurisdiction" in HAVA's provisional
voting requirements is not specifically defined. As a result, states
establish, under their own election codes, the applicable jurisdiction
where voters must cast their provisional ballot from in order for such
ballot to be eligible to be counted. For example, in some states this
location is the specific precinct in which the voter is registered, and in
other states, the voter may be anywhere within the county (city, parish,
township) in which the voter resides and is registered. Our survey of
state election officials asked where a provisional voter needed to cast a
vote in order for it to be counted for the November 2004 election. Figure
45 shows where states reported that provisional voters needed to cast
their votes in order for such votes to be eligible to be counted.

Figure 45: State-Reported Locations Where a Provisional Vote Had to Be
Cast in Order for It to Be Counted for the November 2004 General Election

Source: GAO survey of state election officials.

Note: Six states are not subject to HAVA's provisional voting
requirements, but 2 of these 6 (Wisconsin and Wyoming) authorize some
measure of provisional voting. Both of these states are included with the
32 states that reported requiring that provisional voters must cast their
votes in the specific precincts in which they are registered in order for
their votes to be eligible to be counted.

Variation in state requirements as to the location where a provisional
ballot must have been cast in order to be counted was also evident in the
jurisdictions we visited. For example, voters in Kansas could, according
to election officials, vote provisionally in precincts other than where
they were registered (but within the same county) and if otherwise
eligible to vote have their vote partially counted (e.g., for county,
state, or federal offices or issues). Nevada election officials said they
count provisional votes cast anywhere in the county where the voter was
registered and otherwise eligible, but all provisional ballots only
included federal races. Election officials in both Washington
jurisdictions we visited reported that a voter in the November 2004
election was allowed to cast a provisional ballot anywhere in the state of
Washington, and the ballot would be forwarded to the correct county (if
the ballot was cast in a county other than the one in which the voter was
registered) and counted if the voter was eligible. Election officials in 1
Washington jurisdiction we visited said that county election workers
mailed the provisional ballots for non-Washington residents to the
Secretary of State of the state where the voter claimed to be registered,
but these officials were not knowledgeable of what became of the ballots.

Election officials in several states have faced court challenges to their
state requirements regarding the location where a provisional ballot must
have been cast in order to be counted. The litigation has primarily arisen
in states requiring that a provisional voter had to cast a vote in the
specific precinct in which he or she was registered, in order for that
vote to be counted. In this context, the courts have generally held that
HAVA does not require a state to count provisional votes cast in the wrong
precinct as legal votes when they would otherwise be considered invalid
under state law.18

In our state survey, we also asked state election officials if they
anticipated that their state would change, by November 2006, where a
provisional voter must cast a vote for it to be counted. Forty states
reported that they did not anticipate such rules would change. Election
officials in 4 states reported they anticipated a change by November 2006.
Three out of the 4 states (Arkansas, Nevada, and New Jersey) reporting
that they anticipated a change for 2006 had reported for the November 2004
general election that a

18For example, the United States Court of Appeals for the Sixth Circuit,
in ruling on the counting of provisional votes under Ohio law, held that
ballots cast in a precinct where the voter does not reside and which would
be invalid under state law for that reason are not required by HAVA to be
counted as valid votes. Sandusky County Democratic Party v. Blackwell, 387
F.3d 565, 568 (6th Cir. 2004).

Page 236 GAO-06-450 Elections

Residency Requirements and Evidence of Registration Were the Most Frequent
Problems Nationwide Affecting whether Provisional Ballots Were Counted

provisional voter could have cast a vote anywhere within the county (city,
parish, township) in which the voter resides and have such vote counted.
The fourth state, Colorado, had reported for the November 2004 general
election that provisional voters had to cast their votes in the specific
precincts in which they were registered in order for their votes to be
counted. Georgia, Maryland, and the District of Columbia said they did not
know whether rules specifying where a provisional voter must cast a ballot
in order to be counted could be anticipated to change, and the remaining 
4 states responded that they will not have provisional voting. These 4
states are not subject to provisional voting requirements.

In our survey of local election jurisdictions nationwide, we asked about
problems that local jurisdictions encountered during the November 2004
election in counting provisional ballots. On the basis of our survey, in
jurisdictions where provisional ballots were cast we estimate that the
most frequent problems concerned voters not meeting residency requirements
or lacking evidence that the voter was registered. Specifically, we
estimate

     o 66 percent (plus or minus 7 percent) of jurisdictions had a problem
       with voters not meeting residency eligibility requirements for the
       precinct or jurisdiction,19
     o 61 percent (plus or minus 7 percent) received insufficient evidence
       that individuals had submitted voter registration applications at
       motor vehicle agency offices,
     o 61 percent (plus or minus 7 percent) had instances of insufficient
       evidence that individuals had registered or tried to register directly
       with the election office,
     o 34 percent (plus or minus 7 percent) had registration applications
       received by the registrar very close to or after the registration
       deadline,
     o 32 percent (plus or minus 7 percent) had voters not providing
       identification as specified by HAVA for registrants who registered by
       mail and were voting for the first time in the precinct or
       jurisdiction,

19The 95 percent confidence interval for this percentage is greater than
the +/- 5 percent sampling error for the complete sample.

Page 237 GAO-06-450 Elections

Jurisdictions Visited Identified Variations in How Missing Information Was
Handled when Provisional Ballots Were Counted

     o 29 percent (plus or minus 6 percent) received insufficient evidence
       that individuals had submitted voter registration applications at
       National Voter Registration Act agencies other than motor vehicle
       agency offices,
     o 28 percent (plus or minus 6 percent) had provisional ballot envelopes
       or ballots that were incomplete or illegible, and
     o 20 percent of jurisdictions had problems with voters who did not sign
       a sworn statement that they met the qualifications to be eligible to
       vote in the precinct or jurisdiction.

Written comments made by local election officials in our nationwide survey
identified some additional problems encountered with counting provisional
ballots. Examples included uncertainty whether a convicted felon's voting
rights, lost as a result of such conviction, had been restored; a voter's
registration records that had been sealed by a court; and the state
changing the rules several times right up to Election Day, creating
confusion, according to election officials.

In addition to variation in where states required provisional ballots to
be cast in order to be counted for the November 2004 election, local
jurisdictions we visited reported a variation in how to handle a lack of
identification or a missing signature. For example, election officials in
one New Mexico jurisdiction we visited said that first-time voters that
did not provide the required identification had until the close of the
polls on Election Day to bring their identification to the county clerk's
office. In contrast, according to election officials in a New Jersey and a
Georgia jurisdiction, provisional voters were allowed up to 2 days to
produce identification for their vote to be counted, and in a Nevada
jurisdiction, voters had until 5:00 p.m. the Friday after the election.

With respect to mail registrants who were permitted to cast provisional
ballots because they did not provide required identification when voting
for the first time, election officials in 1 Illinois jurisdiction we
visited reported a lack of clarity as to what subsequent
identification-related verification was needed prior to counting
provisional ballots. According to the Illinois election officials, the
state's guidance resulted in a situation where one Illinois jurisdiction
required the voter to provide to the county clerk's office identification
with an address that matched the address in the voter registration list
within 48 hours after the election in order to be counted,

Telephone Was Used Most Often to Provide Voters with the Outcome of Their
Provisional Vote

while another jurisdiction did not require the two such addresses match.
The Illinois officials stated that this issue has been clarified.

Jurisdictions we visited also varied in how they handled a missing voter
signature. For example, in 1 Colorado jurisdiction, election officials
said that they mailed letters to voters who failed to sign their
provisional ballot envelopes and allowed the voters up to 10 days after
the election to come in and sign so that their votes would be counted.
This was not a procedure described in all jurisdictions we visited. In 1
jurisdiction in New Mexico, ballots would not be counted for voters who
did not sign the provisional ballot affidavit or roster. In 1 Georgia
jurisdiction we visited, voters had to complete a new voter registration
form or their provisional ballots were not counted.

HAVA requires that provisional voters be provided with written information
about how to find out whether their vote was counted (and if not, why)
using a free access system established by state or local election
officials. On the basis of our local jurisdiction survey, we estimate that
the majority of local jurisdictions that had provisional ballots cast used
the telephone (often toll-free) as the free access system for voters in
the November 2004 election to obtain information on whether their
provisional ballot was counted, and if not counted, why not. Table 19
shows the estimated percentage of jurisdictions that used various methods.
Some jurisdictions used more than one method.

Table 19: Estimated Percentages of Local Jurisdictions Using Various
Methods to Inform Voters of the Outcome of Their Provisional Ballot,
November 2004

                         Percentage of jurisdictions using Method used method

                   Local election office telephone number 84a

                   State election office telephone number 50b

     Letter informing voters of the outcome of their provisional ballot 51b

                         Internet Web site address 35b

     E-mail informing voters of the outcome of their provisional ballot 3c

Source: GAO 2005 survey of local election jurisdictions. The estimates are
based on a subgroup of jurisdictions where provisional ballots were cast.

aThe 95 percent confidence interval for this percentage is +/- 6
percentage points.

bThe 95 percent confidence interval for this percentage is +/- 8
percentage points.

cThe 95 percent confidence interval for this percentage is +6,-3
percentage points.

Note: Jurisdictions could indicate using more than one method, therefore
the percentages add to more than 100 percent.

Election officials from jurisdictions we visited described a number of
ways that provisional voters were provided information about how to learn
the outcome of their votes for the November 2004 election, such as ballot
receipts, a copy of the voter's affidavit, a form letter, or a tear-off
portion of the provisional ballot envelope. In a New Jersey jurisdiction
we visited, provisional voters were given a toll-free number at which to
leave their name and address, and then the results were mailed to them,
according to election officials. The jurisdiction election officials noted
that this process worked well. Figure 46 provides examples of the
information voters were provided to inquire whether their vote was
counted.

Figure 46: Examples of Written Information Provided to Voters to Inquire whether
                       Their Provisional Vote Was Counted

Source (top to bottom): Allegheny County, Pennsylvania; Santa Fe County,
New Mexico; El Paso County, Colorado.

A Number of Factors Contribute to Difficulties in Estimating and Comparing
Provisional Ballots Cast and Counted

In our local jurisdiction survey, we asked how soon after Election Day
information on the outcome of a provisional ballot was made available to
voters. According to written comments, feedback was reported by some
election officials to be available to voters after the November 2004
election as early as the next day, or within 7 days after the election,
although some allowed 1 month, or until the election was certified.

Election officials in some of the jurisdictions we visited reported that
few voters called to find out if their provisional votes were counted. For
example, in a Colorado jurisdiction, officials reported approximately 
100 calls out of over 6,100 ballots cast; a Kansas jurisdiction election
official estimated receiving calls from 6 provisional voters out of over 
3,600 that voted; a New Jersey jurisdiction reported receiving 69
inquiries from voters out of over 6,300 cast; and in 3 other jurisdictions
we visited, election officials reported no one called to find out if his
or her vote was counted.

Estimating the number of provisional ballots initially cast and those that
were counted in the November 2004 election is difficult because complete
information is not available, and because of differences in how state and
local jurisdictions have implemented HAVA provisional voting requirements
affecting how and whether such ballots are provided and counted. Those
same factors limit the value of comparing provisional ballots cast and
counted among jurisdictions. Although estimation is difficult, our survey
allowed us to estimate provisional ballots cast, but with strong caveats.

While HAVA required that most states permit individuals to cast
provisional ballots under certain circumstances, not all jurisdictions
reported having provisional ballots cast in their jurisdiction in the
November 2004 election. On the basis of our survey of local jurisdictions,
we estimate that provisional votes were cast in 33 percent of
jurisdictions and none were cast in 67 percent of jurisdictions.20 Our
estimates varied by size of jurisdiction regarding whether provisional
votes were cast or not. We estimate that in 99 percent of large
jurisdictions, 84 percent of medium jurisdictions, and 12 percent of small
jurisdictions provisional votes were

20Jurisdictions in Idaho, Maine, Minnesota, New Hampshire, and North
Dakota were not included in this estimate.

Page 242 GAO-06-450 Elections

Data Were Lacking for National Estimate of Provisional Votes Cast and
Counted

cast in the November 2004 election. The differences between all sizes of
jurisdictions were statistically significant.

The difference between different sizes of jurisdictions' use of
provisional ballots may be explained in part by comments from election
officials in local jurisdictions surveyed and from officials in
jurisdictions we visited. For example, officials in several smaller
jurisdictions included in our nationwide survey who reported that
provisional ballots were not cast in their jurisdiction had indicated in
written comments that election workers are likely to have personal
knowledge of a voter's eligibility. As one election official from a
Wisconsin jurisdiction wrote, provisional ballots were available, but use
of the ballots was not necessary. Similarly, in a small jurisdiction we
visited in New Hampshire, election officials told us that given the town's
small population of roughly 1,600 residents, 99 percent of the time
someone in the room knew the individual and could vouch for his or her
identity. In this circumstance, according to election officials, no
verification was necessary at the poll to ensure the voter's
identification.

The number of provisional ballots cast and counted nationally is difficult
to estimate with precision because of the limited data available and data
quality concerns. Estimates that are available, however, do serve as an
indication that the HAVA provisional voting requirements have allowed
potentially eligible voters who otherwise might have been turned away to
participate. We requested November 2004 data on provisional ballots cast
and counted in our survey of local election jurisdictions nationwide, but
because of missing information and other methodological concerns, our
estimate is provided only with strong caveats. We estimate that a total of
between 1.1 million and 1.7 million provisional ballots were cast. Our
range reflects the fact that an estimated 20 percent of the jurisdictions
in our survey did not provide data on how many provisional ballots were
cast. We could not estimate the number of provisional ballots that were
counted with any level of certainty, because of a very high level of
missing data-an estimated 40 percent of the jurisdictions did not provide
data on the number of provisional ballots counted.

In addition, some jurisdictions in our survey providing the number of
provisional ballots cast may have actually provided the number of
provisional votes counted. It is possible this may have occurred because
jurisdictions would more likely have a record of the number of provisional
votes determined to be qualified and counted than they would have the
number of provisional votes originally submitted at polling places
(cast).21 For example, in 1 jurisdiction we visited, provisional ballot
numbers were provided only on the number of provisional votes that were
counted. If some responses to our survey of local jurisdictions actually
provided the number of votes counted rather than the number of votes cast,
then our estimate of provisional votes cast may be an underestimate.

HAVA specifies that information be made available to individuals through a
free access system (such as a toll-free telephone number or an Internet
Web site) regarding whether their provisional votes were counted and, if a
vote was not counted, the reason it was not counted.22 The specifics of
implementing such a system, such as the methods by which such information
is to be identified, collected, and maintained, however, under HAVA, are
left to the discretion of state and local election officials. The National
Task Force on Election Reform recommended that states develop a uniform
method for reporting provisional ballots at the state and national levels,
and also that states collect data on the number of provisional ballots
cast on Election Day.23 Some states might require the information on
ballots cast and counted be sent for statewide figures. Election officials
in a Connecticut jurisdiction we visited, for example, said that the
Registrar completed a provisional ballot report for the Secretary of State
in accordance with state guidance.

21Lack of agreement for terms was a data quality issue pointed out by the
Election Assistance Commission for the commission's Election Day survey,
with some responses providing the same number for Election Day ballots
cast and ballots counted. The U.S. Election Assistance Commission. 2004
Election Day Survey: How We Voted: People, Ballots and Polling Places.
September 2005.

22HAVA section 302(a) (codified at 42 U.S.C. S: 15482(a)).

23National Task Force on Election Reform, Election 2004, Review and
Recommendations by the Nation's Elections Administrators. May 2005.

Other national estimates of the number of provisional votes cast and
counted in the November 2004 election have been affected by data quality
issues. The Election Assistance Commission, using data from its survey of
election administrators,24 estimated that 1.9 million voters cast
provisional ballots at the polls in November 2004, and that 1.2 million of
those votes cast were counted.25 As with our estimates, EAC cautioned that
the coverage, or response rate, for its estimates was limited. The
response rate for provisional ballots cast and counted was 46 percent and
38 percent, respectively. The report authors stated that data quality
issues, such as missing data or data error entries (such as in 15
jurisdictions in the EAC report where the number of provisional ballots
counted was greater than the jurisdiction reported as cast) were
identified and corrected where possible.

On the basis of data collected at different times from different sources
in different states, electionline.org estimated that over 1.6 million
provisional ballots were cast, and nearly 1.1 million of them were
counted. However, readers are cautioned here as well about the limitations
of the available data. For example, figures are not definitive because of
the variation in requirements and procedures among (or even within)
states, and estimates are based on incomplete information.26 The authors
stated that they provided provisional voting estimates with the intent of
moving the discussion of provisional voting forward.

Number of November 2004 Information provided by some of the jurisdictions
we visited illustrates the variation in the reported number of provisional
ballots cast and counted

Provisional Votes Cast and

Counted in Local during the November 2004 election, as shown in table 20.
Jurisdictions We Visited

24EAC surveyed all 50 states (which, in turn, worked with local election
officials), the District of Columbia, Guam, Puerto Rico, American Samoa,
and the U.S. Virgin Islands, in the fall of 2004 and with follow-up
requests in 2005.

25U.S. Election Assistance Commission, 2004 Election Day Survey. September
2005.

26Electionline.org, Briefing: Solution or Problem? Provisional Ballots in
2004. March 2005. Electionline.org is produced by the Election Reform
Information Project.

Page 245 GAO-06-450 Elections

      Table 20: Reported Provisional Ballots Cast and Counted in Selected
                                 Jurisdictions

                                                                   Percentage
                                             Total             of
                                       provisional provisional          Total 
                                           ballots ballots cast       ballots 
                         Provisional  counted of      that were    counted in 
Jurisdiction         ballots cast    those cast      counted  jurisdiction 
El Paso, CO                 6,158         4,779           78       242,888 
Larimer, CO                 2,636         1,798           68       147,112 
New Haven, CT                  75            67           89 Not available 
Dougherty, GA                 130            12            9        33,809 
Muscogee, GA                  210           107           51        64,336 
Champaign, IL                 292            64           22        84,153 
Chicago, IL                22,611        13,838           61     1,063,860 
Johnson, KS                10,942         7,375           67       259,599 
Wyandotte, KS               3,664         2,780           76        53,630 
Clark, NV                   4,352         1,543           35       546,858 
Washoe, NV                  1,465           880           60       159,511 
Bernalillo, NM             12,367         6,233           50       263,054 
Santa Fe, NM                  580           230           40        67,782 
Carteret, NC                1,099           888           81        31,770 
Guilford, NC                2,260         1,291           57       201,854 
Cuyahoga, OH               25,309        16,750           66       687,255 
Mahoning, OH                2,786         2,350           84       134,290 
Clark, WA                   5,214         4,215           81       172,277 
King, WA                   31,712        28,010           88       898,238 

Source: GAO summary of information provided by jurisdictions we visited.

When looking at provisional ballots cast and counted for a particular
jurisdiction, the variability in the implementation of provisional voting
by states and jurisdictions makes interpretation and comparison among
jurisdictions difficult. As mentioned earlier, the number of provisional
votes cast and counted may vary based on a number of factors. In general,
states and jurisdictions vary in why and how provisional ballots are
provided to potential voters, as well as the state and local procedures
for how provisional ballots are counted. A partial list of these factors
includes the following:

     o State provisions varied regarding the additional circumstances (apart
       from the minimum requirements specified in HAVA) under which a
       provisional ballot may be offered.
     o Some states offered other voting options in addition to provisional
       ballots to voters with eligibility issues (such as signing an
       affidavit, then voting normally or casting a challenged ballot).
     o The manner and extent to which the provisional ballot options
       available to voters are actually utilized varied in connection with
       the size and approach of the jurisdictions. For example, smaller
       jurisdictions were, according to election officials, less likely to
       use the provisional ballot option than larger jurisdictions because
       they were more knowledgeable of voters in their jurisdictions and
       therefore better positioned to address eligibility issues than larger
       jurisdictions, and some jurisdictions reported taking additional steps
       to send the voter to the correct precinct before offering a
       provisional ballot, whereas other jurisdictions might not do so.
     o States established the location where voters must cast their
       provisional ballots from in order for such ballots to be eligible to
       be counted. For example, in some states this location is the specific
       precinct in which the voter is registered, and in other states, the
       voter may be anywhere within the county (city, parish, township) in
       which the voter resides and is registered.
     o States or local jurisdictions established other conditions (e.g., the
       time limit for providing required identification) that varied in
       determining whether a provisional vote was to be counted.
     o There were other factors, such as instances in which the polling
       location was kept open late because of a federal court, state court,
       or other order extending the polling hours.

Notwithstanding the variations we have identified in provisional voting
processes and challenges identified by some election officials in
jurisdictions we visited, several election officials reported that they
thought the provisional voting process worked well for the November 2004
election, in that people who would normally not have been able to cast a
ballot were allowed to do so, and some of those ballots were counted.

Concluding Observations

While many jurisdictions reported that for the November 2000 election
having at least one procedure in place to help resolve eligibility
questions for voters whose name did not appear on a polling place
registration list, only 20 states plus the District of Columbia reported
using some form of provisional voting in the November 2000 election. In
those states in which it was not available, voters whose names did not
appear on polling place registration lists, but stated they had properly
registered to vote, were often not permitted to cast a regular ballot.
Provisional voting is an important means of enhancing voter access to the
polls.

HAVA required all states that required registration prior to Election Day
to provide for provisional balloting by the November 2004 election, but
left to states the specific choices on how they would implement that
requirement. In exercising this discretion, states have created varied
provisional voting rules and practices. Under HAVA, provisional ballots
are to be counted as a vote under state law if the person casting the
ballot is determined to be eligible to vote under state law. These
statutory provisions and determinations of eligibility and what
constitutes a properly voted ballot vary by state and thus affect the
state rules and procedures used to determine whether provisional ballots
are counted. At least 1 state, for example, allows voters to cast a
provisional ballot for statewide offices anywhere in the state, with the
ballot returned for eligibility verification and counting to the
jurisdiction in which the voter said he or she was registered. Other
states required that voters cast provisional ballots in their assigned
precinct for the ballots to be counted. The actual impact of these varying
practices on provisional balloting and vote counting is unknown.
Comparable data across states are not available to determine whether or
how these variations affect the number of voters who are permitted to cast
provisional ballots or the percentage of provisional ballots that are
actually counted. Thus, it is difficult to assess the potential impact of
a state changing its existing rules and practices. However, based on the
data that are available, it is clear that provisional voting has helped to
facilitate voter participation of those encountering eligibility-related
issues when attempting to vote.

Chapter 6

Counting the Votes

Once the polls close on Election Day, the process of determining and
certifying the final results begins. Vote counting is a complex, multistep
process with many variations across the nation. The exact process depends
upon a number of variables. Among them are state requirements that define
standards for determining voter intent for ballots that are not clearly
marked, deadlines for certifying the final count, and specifications for
conducting recounts when required. The types of ballots to be counted
affect vote tabulations because absentee and provisional ballots typically
undergo some type of verification before counting, while early and regular
Election Day ballots typically do not require this processing. The types
of technology used for vote casting and counting-hand-counted paper
ballots and machine-counted ballots (punch card, optical scan, and those
cast electronically)-also add variance to how votes are handled. The
counting process requires attention to detail, and problems in any one
election stage can affect the final vote count. Moreover, its
orchestration requires the effective interaction of people, processes, and
technology. This chapter discusses the continuity and key changes since
the 2000 general election and challenges-new and ongoing-encountered by
election officials in the 2004 general election with respect to counting
votes.

In the 2004 general election, vote counting remained an intricate,
multistep

Overview

process characterized by a great variety of local procedures depending on
a local jurisdiction's technology, size, and preferences. As with the 2000
general election, the proportion of jurisdictions nationwide reporting
recounts or contested elections remained small in the 2004 general
election. There were some notable developments related to vote counting. A
significant change was the fact that by the 2004 general election more
states had developed guidance for determining voter intent on unclear
ballots. Eighteen states that reported not having guidance in the 2000
general election reported in our survey they had such guidance in place
for the 2004 general election. In addition, 9 states reported changes
relating to the process of conducting recounts. Some added requirements
for mandatory recounts. Others changed their conditions and guidance for
conducting recounts. The results of our state survey showed that while 29
states and the District of Columbia did not require audits of vote counts,
9 states reported having taken some legislative or executive steps toward
doing so.

Many of the problems in managing people, processes, and technology that
had confronted election officials across the country in the November 2000
general election continued to challenge them in the 2004 general election.

In the 2004 General Election the Vote-Counting Process Remained Complex
and Marked by Local Diversity

Equipment problems, poll worker errors, and voter errors made it difficult
to tabulate the votes quickly and accurately, according to some election
officials. A new phenomenon emerged as a challenge to election officials,
as well: Some jurisdictions reported difficulty completing the extra steps
required to verify and count provisional votes within the time allowed for
tallying the final vote count. Finally, while recounts and contested
elections remained rare in the 2004 general election, those that did
occur, particularly in Washington state, revealed the intricacies and
vulnerabilities of the election process.

The basic elements of the vote-counting process we described in our
October 2001 comprehensive report on election processes nationwide
remained in practice for the general election of 2004.1 Of necessity, it
was a complex, multistep process, with many variations, depending on a
jurisdiction's technology, size, and preferences. As with other elections,
vote counting in the 2004 general election involved certain common steps:
closing and securing the polls and voting equipment; securing the ballots;
reconciling the number of ballots at the polls (e.g., the number available
at the polls compared to the number cast, spoiled, and remaining);
transporting ballots and equipment from the polling places to a central
location where they were secured; in some cases electronically
transmitting results from polling place voting equipment to a central
tally location; verifying provisional and absentee ballots for counting;
determining whether and how to count ballots that may be improperly or
unclearly marked; conducting any necessary recounts; and certifying the
final count.

Local Jurisdictions We Surveyed and Visited Secured Machines and Ballots
in a Variety of Ways

Preliminary to counting, a key step was to secure the voting machines and
ballots so that no additional votes could be cast. Procedures for securing
equipment varied with the equipment that was in use. However, on the basis
of our survey of a representative sample of local election jurisdictions
nationwide, we estimate that 91 percent of all jurisdictions used hardware
locks and seals as one of their predominant security measures. In our site
visits, local election officials also described securing DRE tapes and
cartridges under lock and key before and after they were delivered to
boards of elections or other authorities. For example, election officials

1GAO-02-3.

from 1 jurisdiction we visited described securing memory cards in optical
scan counting machines by attaching a plastic band with a serial number.
The band would have to be severed in order for the memory card to be
removed, according to election officials. One such band is shown in Figure
47 securing a voting equipment bag.

                     Figure 47: Sealed Voting Equipment Bag

Source: GAO.

Election officials in 2 jurisdictions we visited also described a variety
of measures they took to ensure that ballots were not lost or miscounted.
In  1 Washington jurisdiction, officials said they secured punch card
ballots at polling places for counting elsewhere by transporting ballots
twice: once earlier on Election Day and the other time after the polls had
closed. These officials also said that ballots were bundled into groups of
50, separated by type (Election Day, provisional, and absentee ballots),
and put into transport carrier safe boxes. Two poll workers, one from each
political party, accompanied the ballots when they were transported to the
elections office for counting. Similarly, in a Colorado jurisdiction we
visited, election officials said that at the close of Election Day they
sealed optical scan ballots from the polling place and the optical scan
counter to prevent tampering. Then, two election judges transferred the
ballots and optical scan counter to the counting center.

While ballot-securing methods varied, the results of our local
jurisdiction survey showed that most jurisdictions had written policies
and procedures in place in the November 2004 general election to secure
ballots (including paper and electronically stored ballots). As shown in
table 21, on the basis of our survey we estimate that two-thirds of local
jurisdictions had written procedures for transporting ballots, and about
three-quarters had written procedures in place for secure ballot storage
rooms.

Table 21: Estimated Percentages of Jurisdictions Having Written Ballot
Security Policies and Procedures, November 2004 General Election

                                                                Percentage of
                                                           jurisdictions that
                                                                  had written
                                                                 policies and
Policies and procedures                                         procedures 
Transporting unvoted ballots to polling places                          66 
Transporting voted ballots or e-memory to locations for                 66 
countinga                                                       
Providing secure rooms or building where unvoted ballots are            71 
stored                                                          
Providing security for rooms or building where voted ballots            76 
are                                                             
stored                                                          
Securing the electronic transmission of voted ballots for               18 
countingb                                                       

Source: GAO 2005 survey of local election jurisdictions.

aTransporting ballots or e-memory to locations was not applicable for an
estimated 20 percent of jurisdictions.

bSecuring electronic transmissions was not applicable for an estimated 57
percent of jurisdictions.

In addition, reconciling ballots with the number of voters was a common
step in securing ballots before they were counted. According to our state
survey, 47 states and the District of Columbia reported that they required
jurisdictions to count or keep track of ballots that were unused, spoiled,
rejected, or issued but not returned. Two states, Montana and Maine,
reported not requiring jurisdictions to count or keep track of such
ballots. New York reported in our survey that because it does not have
paper ballots, such tracking was not necessary. New York reported that it
did not use paper, optical scan, or punch card ballots. During our visits
to election jurisdictions, we asked officials how they reconcile ballot
and voter numbers. The election officials reported conducting cross-checks
in a

Counting, Canvassing, and Certifying the Vote Was an Intricate and Varied
Operation

number of ways, but generally followed a process of reconciling any
discrepancies between the total numbers of ballots on hand at the
beginning of the day, the number of voters who signed in at the polling
place, and the number of ballots cast.

Once the ballots were reconciled in the November 2004 election, local
jurisdictions tabulated and canvassed (or reviewed) the vote. Both
counting and canvassing the count were an ongoing process in the effort to
ensure an accurate tally. After initial tabulations of votes on election
night, which were typically released to the public, canvassing was
typically the process of reviewing all votes by precinct, resolving
problem votes, and counting all types of votes (including absentee and
provisional votes) for each candidate and issue on the ballot and
producing an official total for each. The official total was usually
certified by an election official. This process varied among jurisdictions
in terms of how and where it was done and who was responsible.

The counting process involved several different types of ballots, cast
under different circumstances:

     o General election votes are cast at polling places on Election Day by
       voters who appeared in the registration lists for that precinct and
       voted a regular ballot.
     o Provisional votes are cast by those, for example, whose registration
       (and qualification to vote) could not be established at the time of
       voting at the polls on Election Day.
     o Absentee votes are generally votes received and cast by mail before
       Election Day.
     o Early votes are generally cast in person before Election Day.

According to our local survey, for the November 2004 general election,
local jurisdictions nationwide used different voting methods for different
ballot types. As shown in table 22, we estimate the largest percentages of
jurisdictions used optical scan and paper hand-counted ballots for
Election Day. Also, optical scan and punch card vote-counting methods were
used at precincts or at central locations. Jurisdictions could check more
than one voting method.

Table 22: Estimated Percentages of Local Jurisdictions Using Particular
Voting Methods by Ballot Type for the November 2004 Election

                   Percentage of  Percentage of  Percentage of  Percentage of 
             jurisdictions using jurisdictions  jurisdictions   jurisdictions 
                                     using      using                   using 
              method for general method for         method for     method for 
                                 absentee                      
Voting method        Election         voting  provisional     early voting 
                      Day voting                    voting     
DRE                         7              4              1              4 
Central count              24             20              9              5 
optical scan                                                
Precinct count             32             24            10a              5 
optical scan                                                
Lever machine               7              1              1              1 
Central count               3              3              2              1 
punch card                                                  
ballot                                                      
Precinct count              2              1             2a              1 
punch card                                                  
ballot                                                      
Paper                                                                      
(hand-counted)             35             36             18              8
ballot                                                      

Source: GAO 2005 survey of local election jurisdictions. a

A small number of jurisdictions indicated on our questionnaire that they
use precinct count equipment for provisional ballots. This could be true
when a jurisdiction does not physically count provisional ballots at a
precinct, but when officials programmed optical scan or punch card
tabulating machines to accept provisional ballots only from specified
precincts.

In our local jurisdiction survey, we also asked what predominant voting
method was used to process the largest number of ballots in the 2004
general election. We estimate that hand-counted paper ballots were the
predominant tabulation method for 30 percent of all jurisdictions,
although these were almost all small jurisdictions. Specifically, we
estimate that  41 percent of small jurisdictions, 3 percent of medium
jurisdictions, and no large jurisdictions hand-counted paper ballots.2
Small jurisdictions were statistically different from large jurisdictions.

As in the November 2000 general election, the counting process for the
November 2004 election took place at precincts or at centralized
locations, such as election headquarters at town halls and even
warehouses. In jurisdictions we visited, we learned about some of the
substantial variations in the sequence, procedures, and precautions taken
to conduct the count. We found in our site visits that vote counting
ranged from a very simple process in a small jurisdiction to more complex
processes in larger jurisdictions. For example, a small New Hampshire
jurisdiction, with just

2The estimate for large jurisdictions has a 95 percent confidence interval
of 0 to  +2 percentage points.

Page 254 GAO-06-450 Elections

over 1,000 registered voters, had one polling place and one precinct open
on Election Day, according to election officials. They told us the paper
ballots were not transferred to any location for counting and were
handcounted by 25 election workers. These officials also said that five
teams of five individuals each reviewed votes cast on each paper ballot
and used paper and pencil to record and tally vote totals. The final
election outcomes were written on a standard form and submitted to the New
Hampshire Secretary of State's office, according to election officials.

In contrast, election officials in a large Washington jurisdiction
described a more complex process for their centralized vote count of punch
card ballots. As described by these officials, their process enabled them
to begin reporting results on Election Day evening by precinct and to
provide updates of the count every 30 minutes. Once Election Day ballots
were transferred to the election office by poll workers, the ballots were
counted to determine total numbers, according to election officials. They
also told us that after the ballots were separated by precinct, up to 20
inspection boards, composed of two Republicans and two Democrats each,
inspected the ballots one precinct at a time. In the inspection process,
the officials said that the ballots were further separated into
categories-those that were machine-readable and those that required
further examination, such as ballots with write-in candidates or with a
chad hanging by two or more corners. Once all questions were resolved
(including any that would require review by a canvassing board), they told
us ballots in batches of  500 each were placed in trays by precinct and
brought to the ballot tabulation area. According to these officials, the
jurisdiction used a punch card tabulator, which was connected to a
computer and had a processing speed of 600 ballots per minute (see fig.
48). Once all ballots were counted, jurisdiction election officials told
us they generated an unofficial report with results for all races and
voting propositions. This initial tally was posted on the county Web site
and released to the press, candidates, and public, according to election
officials.

             Figure 48: Punch Card Tabulator Connected to Computer

Source: GAO.

Six of the jurisdictions we visited told us that they counted Election Day
votes at the local precinct, where poll workers would tabulate results and
resolve any ballot issues that could be handled locally. For example, in a
large Kansas jurisdiction, election officials said that voters were able
to place their ballots in an optical scanner at the polling place that
read the ballot and rejected it if there were any problems. According to
officials there, the machines could return to the voter any ballot that,
for example, had too few or too many votes for a specific office and
provide a screen message for what to correct before resubmitting the
ballot. After the polls closed, the optical scan machines with their
memory cards-which had been programmed for the specific precinct-were
transferred to election headquarters, according to election officials. The
officials also said the optical scan machines were linked electronically
to one computer and data from the memory cards were uploaded so that votes
from all precincts could be tallied.

Additional Steps Were Required to Manage Absentee, Provisional, and Early
Votes

Absentee, provisional, and early votes each required some additional steps
to manage in order to include them in the vote count.

Absentee votes: According to our state survey, all states reported having
some provision for absentee voting in the 2004 general election. As we
discussed in chapter 3, on absentee voting, absentee ballots must
typically undergo some type of verification prior to counting. At 1
Colorado jurisdiction we visited, officials said that they began verifying
and counting absentee ballots 10 days before Election Day. At 1
jurisdiction in Washington election officials said that they qualified the
absentee ballots as they were received at the election office, but did not
count the votes until

3:00 p.m. on Election Day. Additionally, at a jurisdiction in Illinois,
election officials said that they distributed most absentee ballots to
their respective precincts to be counted along with the Election Day
ballots. In each of these jurisdictions, however, according to election
officials, the absentee ballot results were not released until after the
Election Day polls were closed. Also, on the basis of our local
jurisdiction survey, we estimate that 99 percent of election jurisdictions
included the counts of qualified absentee ballots in the final certified
count, regardless of their effect on the outcome.

Provisional votes: Provisional voting, which was required by HAVA in all
but 6 states during the 2004 general election, generally required several
steps. At all of the local jurisdictions we visited that used provisional
ballots, election officials said that the ballots were transferred to an
election office or central count location, where the eligibility of the
voter was verified before they were counted. We estimate, on the basis of
our local jurisdiction survey, that 83 percent of jurisdictions that
provided provisional ballots during the 2004 general election transferred
the provisional ballots to a central location for counting. Those
jurisdictions that did not engage in transfers may have been jurisdictions
with only one precinct, in which case, the votes were tallied on-site. At
all of the jurisdictions we visited that used provisional ballots election
officials said they included provisional ballots determined to be verified
in certified vote counts regardless of their effect on the outcome of the
election.

Early votes: According to our state survey, for the November 2004
election, 24 states and the District of Columbia reported they allowed
early voting, and from our local jurisdiction survey, we estimate that
about 23 percent of local jurisdictions allowed early voting in the
election. In early voting jurisdictions we visited, a variety of
reconciliation and counting processes were used, according to election
officials. At one jurisdiction we visited,

                 Election Certification Periods Varied by State

election officials told us that early voting DRE votes were reconciled
daily. According to these officials, at the end of the early voting
period, election department staff shut down the DRE machines and removed
the memory cards (which stored cast votes). The officials said that the
memory cards were sealed and returned to the election department office
for counting, in a manner similar to Election Day DRE votes. In another
jurisdiction we visited that used optical scan machines for early voting,
officials told us that ballots were inserted by voters into the machines
at the polls-the same procedure used on Election Day. At the end of each
early voting day, according to the officials, the ballots from that day
were physically transferred to the clerk's office and the optical scan
results were submitted by modem to the jurisdiction's headquarters.

Election returns posted on election night are unofficial and are not
considered final until canvassing-the process described earlier of
reviewing all votes by precinct, resolving problem votes, and counting all
types of votes-is complete and the count is certified. Certification is
when the vote count is finalized, generally by state and local officials.
Our state survey showed that for the 2004 general election, states
reported varied practices for when counts were certified and by whom,
similar to the general election of 2000.

Our state survey showed that most states reported setting certification
deadlines, but the certification periods varied from state to state. Four
states (Alaska, Nebraska, New Hampshire, and Rhode Island) and the
District of Columbia reported not specifying a deadline following Election
Day for certification of election results, while all other states reported
specifying such a deadline. For example, certification on the second day
after Election Day was reported by Delaware, while not later than 40 days
was reported by Michigan. Some states reported caveats and varying levels
of specificity in the certification deadlines. Maine reported allowing 3
days for local election official certification and 20 days for state-level
certification. Missouri's reported deadline was by the fourth Tuesday
following the election. North Dakota reported a deadline of not less than
 3 days, but not more than 6. Similarly, the requirement reported for
Texas was 15 to 30 days after the election.

For the 2004 General Election, More States Had Requirements or Guidance
for Determining Voter Intent

An important facet of the canvassing process is the consideration that may
or may not be given to ballots that have not been marked properly. An
improper mark, for example, could be a circle around a candidate's name
instead of a checked box on a ballot that is to be scanned optically. For
those states providing for the determination of voter intent, the
importance of having explicit and consistent criteria for treating unclear
ballots became evident in the 2000 general election when different
interpretations for punch card ballots in Florida made the close
presidential race extremely contentious. While subsequent federal reforms
have not specified standards for treating unclear ballots, HAVA requires
that each state adopt uniform standards, by January 2006, that define what
constitutes a vote and what will be counted as a vote for each category of
voting system used in the state.3

In our state survey, 39 states and the District of Columbia reported that
for the November 2004 general election they had requirements or guidance
for determining voter intent that focused primarily on improper ballot
marks. Forty-five states and the District of Columbia reported they had
requirements or guidance for determining how or whether to count a
machine-unreadable ballot-one that cannot be processed by machine because
it is damaged.

Eighteen states that had reported not having provisions in place for the
2000 general election reported to us in our 2005 state survey that they
had voter intent guidance for the November 2004 general election. Georgia,
for example, had developed requirements for four methods: DRE machines,
lever-type machines, optical scan, and hand-counted paper ballots. Some of
Georgia's requirements were for certain ballots rejected by optical scan
machines. These requirements provide for some measure of subjective
determination of a voter's intent by election officials in certain
specified instances. In such an instance, a vote shall be counted, under
these Georgia provisions, if in the opinion of the vote review panel, the
voter has clearly and without question indicated a choice for which the
voter desired to vote. In addition, under specified circumstances, these
Georgia provisions also provide for a similar type of voter intent
determination with respect to hand-counted paper ballots.

3HAVA section 301(a)(6) (codified at 42 U.S.C. S: 15481(a)(6)).

As described below, we found in our site visits that under state or local
guidance, local jurisdictions we visited had gone to varied lengths in the
2004 general election to salvage ballots that were improperly marked or
that were machine unreadable. These efforts varied by the type of voting
equipment used in the jurisdiction.

Optical scan ballots: In some jurisdictions, election officials told us
that optical scan machines located at polling places could notify the
voter of an unreadable or incorrectly marked ballot at the moment it was
submitted. However, where the ballots were transferred to a central
location for counting this would not be the case. In one jurisdiction in
Colorado where optical scanning was done centrally for absentee ballots,
election officials told us they were required to interpret voter intent or
replace an unreadable ballot. According to election officials, the
jurisdiction had instructions, which they stated were based on state
statutes, specifying that bipartisan election judges would be the
responsible parties for determining voter intent. Their deliberations,
however, would be observed by others, according to the instructions. If a
decision was reached on voter intent, a replacement ballot could be
created and run through the optical scanner, according to the officials.

Officials in a Kansas jurisdiction we visited said that state election
standards were very specific for interpreting an incorrectly marked
optical scan ballot. They would count a vote if an oval shape is marked,
near but not inside the oval, and not closer to another candidate's name.
A completed oval would also be counted if another oval for the same race
was scribbled or crossed out. If the ballot could be interpreted locally,
officials said election workers duplicated the vote on a new ballot for
the optical scanner to read. According to election officials, if the
intent was not clear, the ballot would be sent to the Board of Canvassers
for further examination. State guidance also included standards for
hand-counted paper ballots.

In Florida, guidance in place for the November 2004 general election was
even more specific than that provided in Colorado or Kansas. The guidance
specified, for example, that, with respect to manual recounts, a vote may
be counted if "there is an `X', a check mark, a plus sign, an asterisk or
a star, any portion of which is contained in a single oval or within the
blank space between the head and tail of a single arrow and which does not
enter into another oval or the space between the head and tail of another
arrow." It also allowed for a vote to be counted under additional
specified circumstances including if "there is a diagonal, horizontal, or
vertical line, any portion of which intersects two points on the oval and
which does not intersect another oval at any two points," provided that
the horizontal line does not strike through the name of the candidate.

Punch cards: While federal election reforms included provisions promoting
replacement of punch card ballots, on the basis of our local jurisdiction
survey, some jurisdictions continued to use them in the 2004 election. As
was the case for other types of ballots, levels of guidance for
interpreting voter intent varied by state. Illinois reported that it had
no requirements or guidance for determining voter intent, according to our
state survey. Election officials in 2 Illinois jurisdictions using punch
card ballots told us in our site visit that election workers did not
attempt to ascertain the intent of voters on punch card ballots that were
improperly punched. If the ballot could not be counted by a punch
card-counting machine because of an improper punch or mark, the votes were
not to be counted.

In contrast to Illinois, Washington reported that it had guidelines or
requirements regarding voter intent and allowed for remaking an unreadable
or damaged punch card. In a Washington jurisdiction we visited that used
punch card ballots in the 2004 general election, election officials said
that state law guided their jurisdiction's written instructions for
determining voter intent. Election officials said voters were given very
specific instructions for how to change their vote before casting their
vote, if necessary, on a punch card ballot while at the polls. These
officials also said ballots could be either enhanced or duplicated if it
was clear that a voter had followed these instructions. Also, according to
the officials, a problem ballot could be enhanced or duplicated by
officials if voter intent could easily be determined. If voter intent was
at all unclear, the ballot was to be sent to the canvassing board for
review. According to officials, canvassing board meetings were open to the
public and state guidelines were to be used to interpret voter intent.
Figure 49 shows a punch card voting booth.

                       Figure 49: Punch Card Voting Booth

Source: GAO.

Hand-counted paper ballots: While we estimate, on the basis of our local
jurisdiction survey, that no large jurisdictions and only 3 percent of
medium jurisdictions used paper ballots in the November 2004 general
election for their predominant voting method, 41 percent of small
jurisdictions did. This voting method presented yet another variation in
the process of determining voter intent. For example, in one small
jurisdiction we visited in New Hampshire, election officials we spoke with
said a senior election official was on hand during ballot counting. They
said if a ballot was unclear, the senior official would be involved to
discuss it. If it was still unresolved, state guidance called for an
unclear ballot to "be counted in accordance with a majority vote of the
election officials present." The guidance, which we examined, also
provided examples of what marks on a paper ballot to accept, as shown in
figure 50.

Figure 50: New Hampshire Guidance for Determining Voter Intent on Paper Ballots

Source: New Hampshire Election Procedure Manual: 2004-2005; provided by election
                      officials in Madbury, New Hampshire.

Recounts and Contested Elections Remained Uncommon for the November 2004
General Election

As with the 2000 general election, recounts and contested elections were
an uncommon event in the 2004 general election. On the basis of our local
survey, we estimate that 92 percent of election jurisdictions nationwide
did not conduct a recount for federal or statewide office. Also on the
basis of our survey, recounts were more prevalent in large than in small
election jurisdictions. Specifically, we estimate that 4 percent of small
jurisdictions, 16 percent of medium, and 24 percent of large jurisdictions
conducted recounts for federal or statewide offices. Both large and medium
jurisdictions were statistically different from small jurisdictions.
Similarly, in our state survey, 37 states and the District of Columbia
reported they had no recounts for federal or statewide offices during the
primary or general elections of 2004, as shown in figure 51.

Figure 51: Number of States Reporting Federal and Statewide Recounts for
the 2004 Primary or General Elections

Number of states 40 38

35 30 25 20 15 10

6

4

5

2

1

0

al andwide offices

r

oy fOnlOnl

f

ederal officesstate

                                       r

o y f

wide offices

                               ederawide offices

                                       a

                                       wl

                                   Don't kno

eder

                                      or f

                                     nts f

Both f

state

                                       e

                                  No recoustat

                                       or

Recounts

Source: GAO 2005 survey of state election officials.

aIncludes District of Columbia.

Recounts are, in general, conducted because a candidate, voter, or group
of voters has requested it or because the margin of victory was within a
certain specified margin such that state provisions required or allowed
for a recount. Election officials in local jurisdictions we visited in
several states where recounts were conducted described to us the
procedures they used for their 2004 general election recounts. In a New
Hampshire jurisdiction, where a recount was conducted of the presidential
race of 2004, officials said the recount was requested by a presidential
candidate to test the accuracy of the optical scan vote-counting
equipment. The officials provided the following description of the
recount: Five wards in the jurisdiction had been selected for a sample
recount. It was conducted by the New Hampshire Secretary of State's
office, not by the local election jurisdiction. The jurisdiction's only
role in the recount was to provide the Secretary of State with the optical
scan ballots from the applicable wards. After the Secretary of State
recounted a portion of the optical scan ballots and found no significant
discrepancies between the initial vote tally and the partial recount, a
full recount was not conducted statewide, according to these officials.

In North Carolina, races for two statewide offices (the Agricultural
Commissioner and the Superintendent of Public Instruction) were subject to
recounts because, under state law, the close margin of victory allowed the
losing candidates to request a recount, according to election officials.
In 1 North Carolina jurisdiction we visited, which used DRE machines,
local election officials described the recount process as follows: The
recount was conducted in a different manner from the initial count. For
the initial count, votes were electronically transferred from each DRE
machine to vote storage devices at the polls that stored the vote totals
by precinct. The precinct totals were then downloaded from the vote
storage devices onto a computer located at the jurisdiction's election
headquarters, and vote tabulation software summed vote totals from each
precinct for each election contest in the jurisdiction. During the
recount, rather than relying on aggregated votes totaled by precinct for a
vote count, officials tabulated individual DRE ballots. To complete this
process, the jurisdiction's tabulation software recognized individual
ballot images from the DRE machines rather than aggregated votes per
precinct. The individual ballot images were downloaded onto the computer
in election headquarters, and votes for the races in question were
retabulated (by voter, rather than by precinct as in the initial count).
The outcomes of both the Agricultural Commissioner and the Superintendent
of Public Instruction races were unaffected by the recount results.

Generally, contested elections are court actions initiated by a candidate
or voter alleging, for example, that some type of misconduct or fraud on
the part of another candidate, election officials, or voters, occurred in
a particular election. The results of our local survey indicate that
contested elections were rare during the period from 2001 to the 2004
general election. In our local survey, we asked local jurisdictions
whether they held any primary or general elections for federal or
statewide offices during this period that were contested, and if so,
whether the outcomes for these elections changed. On the basis of our
nationwide survey, we estimate that 5 percent of local election
jurisdictions held a federal or statewide election that was contested
during this period. The contested elections in which the winner did change
involved races for offices such as state judge or governor, or for the
U.S. House of Representatives.

Several States Had Further Developed Their Specifications for an Election
Recount since the 2000 General Election

Perhaps the most heavily contested election in November 2004, which
received a great deal of press coverage, was the Washington state
governor's race. A close margin of victory and a candidate request
prompted two recounts, and after the state certification of a winner in
the election, the second place candidate's campaign and seven voters filed
a petition in a state Superior Court contesting such certification,
alleging that errors, omissions, mistakes, neglect, and other wrongful
acts had occurred in conducting the election. The Chelan County Superior
Court dismissed the election contest petition, finding that the
petitioners failed to prove that grounds for nullification of the election
existed.4 The Superior Court held, in general, that while there was some
evidence of irregularities, the petitioners failed to adequately prove
that the outcome of the election was changed as a result. The recount
itself, however, revealed the substantial complexities involved in
accomplishing an error-free count. We discuss this case more closely later
in this chapter.

State provisions for recount processes vary, and not all states have
provided for or required them in the past. For the November 2004 general
election, however, several states reported that they had introduced or
further developed their specifications for election recounts since the
2000 general election.

In our October 2001 report on election processes, we reported that  47
states and the District of Columbia had provisions for recounts, though
most did not have mandatory recount provisions.5 To better understand
recount reform efforts to help ensure vote count accuracy since the  2000
election, we asked states in our 2005 survey about changes to their
mandatory recount provisions in place for the November 2004 general
election. Nineteen states reported requiring a mandatory recount
predominantly in cases of a tie or close margin of victory, whereas in
2001, 17 states indicated they required mandatory recounts. Thus, 2 more
states reported requiring mandatory recounts for the 2004 general election
than for the 2000 general election. In addition, 3 other states reported
amending their existing provisions for mandatory recounts, while 3 said
they had changed their requirements or guidance for who may request a
recount as

4Borders v. King County, No. 05-2-00027-3 (Chelan County Super. Ct., June
6, 2005) (transcript of oral decision).

5GAO-02-3.

Page 267 GAO-06-450 Elections

shown in table 23. Three states-Hawaii, Mississippi, and Tennessee-
reported not having any formal provision for conducting recounts-both for
the 2000 or 2004 general elections.

Table 23: States Reported Changes to Recount Provisions since 2000 for the
2004 General Election

               States reporting changes Type of change since 2000

        Alabama, Pennsylvania, Texas Mandatory recount requirement added

Arizona, Minnesota, Washington Mandatory recount requirements amended

Florida, Maine, Rhode Island Requirement or guidance for who may request a
recount changed

Source: GAO 2005 survey of state election officials

Alabama, Pennsylvania, and Texas were the states that reported adding
mandatory recount provisions for the 2004 election. Alabama law, in place
for the 2004 general election, requires a recount when the election
returns for any public office indicate that a candidate or ballot measure
is defeated by not more than one-half of 1 percent of the votes cast for
the office or the ballot measure-unless the defeated candidate submits a
written waiver. In Pennsylvania, a recount is mandatory if an election is
decided by one-half of 1 percent or less-unless the defeated candidate
requests in writing that a recount and recanvass not be made. Texas
reported that a recount was required only if two or more candidates tie in
an election.

For the 2004 general election, Arizona, Minnesota, and Washington reported
adding more specifications to the vote margins that trigger recounts in
their states than were in effect during the 2000 general election. Arizona
added triggers for different types of races. For the 2000 general
election, Arizona reported requiring a mandatory recount when the margin
of votes between the two candidates receiving the most votes was not more
than 0.1 percent of votes cast for both candidates, or 200 votes for
statewide offices and 50 votes for the state legislature. For the 2004
general election, Arizona reported in our state survey that it had amended
its mandatory recount requirements so that the thresholds triggered by the
number of votes only applied when the total number of votes cast was
25,000 or fewer.

Washington's mandatory recount provisions in place for the November 2004
general election had changed since the November 2000 general election. The
requirement in 2000 for a mandatory recount by machine was a margin of 0.5
percent or less of total votes cast for the top two candidates. If the
margin was less than 150 votes and less than 0.25 percent of total votes
cast for the top two candidates, a manual recount was required. The
amended requirement, in place for the November 2004 general election,
specified that a recount by machine was required when the margin is both
fewer than 2,000 votes and less than 0.5 percent of total votes cast for
the top two candidates. If the margin was fewer than 150 votes and less
than

0.25 percent of total votes cast for the top two candidates, there was to
be a manual recount.

Since the November 2000 election, Minnesota amended its mandatory recount
triggers to include a specific percentage margin of victory in certain
circumstances, rather than only a specified difference in the absolute
number of votes between the top two candidates. While a margin of 100
votes or fewer in an election had previously triggered a recount for the
2000 general election, Minnesota election officials reported in our state
survey that for the 2004 general election their state required a recount
if the margin was determined to be either less than one-half of 1 percent
of the total number of votes counted or, was 10 votes or less when no more
than 400 votes are cast.

According to our state survey, state requirements or guidance for who may
request a recount, in place for the November 2004 general election,
changed in Florida, Maine, and Rhode Island since 2000. While any Florida
candidate or candidate's political party in 2000 could request a recount,
this was no longer true for the November 2004 general election. For the
2004 general election, Florida election officials reported that no
candidate or political party could request a recount, and that the only
authorized recounts were mandatory recounts to be conducted when the
margin of victory was 0.5 percent or less of the total votes cast. Rhode
Island, which reported that for the November 2000 general election it had
allowed recount requests by any candidate who trailed the winning
candidate by less than 5 percent, reported that for the November 2004
general election, it required a smaller margin before a losing candidate
could request a recount. For example, for races with between 20,001 and
100,000 votes, Rhode Island reported that it required a margin of 1
percent or less (or  500 votes) before a trailing candidate could request
a recount, and for races with more than 100,000 votes the required margin
was one-half of 1 percent (or 1,500 votes) before a trailing candidate
could request a recount. Maine, on the other hand, reported that its
recount provisions in place for the November 2004 general election were
clarified to provide that an apparent

Vote Count Audits Were Not Prevalent in the 2004 General Election, but
Some States Reported Taking Action to Require Them

losing candidate, rather than only the second-place candidate, could
request a recount.

Twenty-nine states and the District of Columbia reported that for the 2004
general election, they did not have provisions requiring or allowing local
jurisdictions to conduct a vote count audit of election results. However,
in our state survey, 9 states reported taking action since November 2004 
(e.g., enacted legislation or took executive action) to require audits of
vote counts. As used in this report, a vote count audit is an automatic
recount, in full or in part, of the vote tabulation, irrespective of the
margin of victory, in order to ensure accuracy before certification.

On the basis of our state survey, as shown in figure 52, 8 states reported
that for the 2004 general election they had a vote count audit requirement
for all local jurisdictions, and 2 states reported requiring vote count
audits for some local jurisdictions. Election officials from 29 states and
the District of Columbia reported that for the 2004 general election they
did not require or allow local jurisdictions to conduct vote count audits.
Eleven states reported that they allowed them. We estimate, on the basis
of our local survey, that 15 percent of all local jurisdictions were
required by their states to conduct such audits as part of the
certification process for the 2004 general election. Larger and medium
jurisdictions were more likely to have been required to do so than smaller
jurisdictions. Nine percent of small jurisdictions, 27 percent of medium,
and 38 percent of large jurisdictions conducted a required vote count
audit of the 2004 general election. Both large and medium jurisdictions
were statistically different from small jurisdictions.

Figure 5 2 : Number of States Reporting Vote Count Audit Requirements in
Place fo r the 2004 General Election

Number of states 35

30

30 25 20 15

11

10

8

5

2

0

Required audits for all local jurisome local jurisdictionsRequired audits fowed
                              local jurisdictionsr

sdictions

                   to conduct auditsDid not requirew auditsa

                                    or allo

                                      Allo

    Audits

Source: GAO 2005 survey of state election officials.

aIncludes District of Columbia.

Nine states reported in our state survey that they had enacted legislation
or taken some executive action to require audits since November 2004.6 For
example, in Washington, beginning January 1, 2006, prior to election
certifications, county officials must audit the results of votes cast on
DRE machines. The audit must be conducted by randomly selecting up to  4
percent of the DRE voting machines or one machine, whichever is

6These 9 states were Alabama, Alaska, Colorado, Connecticut, Minnesota,
New Mexico, North Carolina, Washington, and Wisconsin.

Page 271 GAO-06-450 Elections

greater, and for each device, comparing the results recorded
electronically with the results recorded on paper.

During our visits to local election jurisdictions, election officials in 5
jurisdictions described conducting vote count audits as a part of the
election certification process for the November 2004 general election. For
instance, 2 large jurisdictions in Nevada reported that the state requires
each jurisdiction to randomly audit election results when DRE machines
were used. According to officials in 1 of these Nevada jurisdictions, they
were required to select 1 percent of DRE machines, or 20 machines,
whichever amount is greater, and to perform a manual audit of the
machine-tabulated vote totals. The officials said that they used a
computer program to randomly select which of the jurisdiction's 740 DRE
machines to audit. To conduct a paper-based audit, they told us that for
each randomly selected machine, election workers printed the DRE result
tapes from the voter-verified paper trail printer, manually counted the
vote data on the tapes, and compared the manual count results to the
original electronic results.

In one large Illinois jurisdiction we visited, election officials told us
they were required by the state to automatically audit (by retabulating
votes) results of punch card ballots in 5 percent of their precincts,
which were randomly selected. According to the officials, the State Board
of Elections sent the jurisdiction officials a letter specifying which
randomly selected precincts had to retabulate their votes.

Election officials in a Pennsylvania jurisdiction we visited said that
state law required random audits when electronic voting machines were
used. According to these officials, they were required to audit 2 percent
of DRE vote totals following an election. They told us, however, that in
practice they actually audit all DRE machine vote totals to ensure an
accurate vote count. They stated that vote data stored on DRE backup
memory cards is printed and compared to vote data stored on DRE cartridges
used in original vote counts. They said they operated on the assumption
that because the internal memory cards serve as a backup system, there
should be no difference in the totals.

Local Jurisdictions Reported Some Mistakes and Technical Challenges in
Counting Votes for the 2004 General Election

As in the general election of 2000, the 2004 general election saw failures
to properly employ voting equipment. At several of the jurisdictions we
visited, officials recounted mistakes in using the DRE systems, for
example, that echoed other recent findings (in our September 2005 report
on the security and reliability of electronic voting), noting inadequate
understanding of the equipment on the part of those using it.7

In our September 2005 report on electronic voting, we noted that instances
of fewer votes counted than cast in one Pennsylvania county in the 2004
general election had resulted from incorrectly programmed DRE machines.
Similarly, in our 2005 site visits to election jurisdictions for this
report, officials with whom we spoke recounted difficulties that had
resulted from mistakes in programming the electronic equipment. In 1
Florida jurisdiction, for example, officials reported that the storage
capacity of an optical scan accumulator (used to combine vote data from
DREs and optical scanners) had been inadequately programmed to capture all
of the votes cast. Officials there were able to discover and rectify the
problem so that all votes were counted. In a Nevada jurisdiction,
officials said that on Election Day, there were 198 provisional ballots
(out of 4,532 cast) that were incorrectly programmed on the DRE machines
at several polling locations, resulting in the provisional votes being
counted without the voter first being qualified. According to these
officials, poll workers forgot to add the "0" to the beginning of the
precinct number. The officials noted that 2004 was the first time that the
jurisdiction had used provisional voting and that in the future they
planned to use paper provisional ballots to avoid any confusion.

In a North Carolina jurisdiction we visited, election officials told us
about how a misunderstanding of the voting equipment resulted in the loss
of votes. Specifically, election officials were unclear about the vote
storage capacity of a DRE machine used in early voting and failed to
notice the machine's warning that its file was full. The software
installed on this machine was an older version of the program and only
recognized up to 3,500 votes, according to election officials. Election
administrators believed that it could recognize up to 10,500 votes. They
discovered the error at the close of Election Day when reconciling the
number of votes

7GAO, Elections: Federal Efforts to Improve Security and Reliability of
Electronic Voting Systems Are Under Way, but Key Activities Need to Be
Completed, GAO-05-956 (Washington, D.C.: Sept. 21, 2005).

Page 273 GAO-06-450 Elections

cast on the DRE machine used in early voting with the number of voters
credited with early voting at the polls. Furthermore they said it was not
until they subsequently conducted a simulation of votes cast that they
discovered the cause of the problem. They also discovered that while the
machine's software flashed warnings on its screen when the voter file
became full, election workers had not seen it because of the screen's
positioning. Also, according to the officials, they had been operating
under the assumption that the machine would have automatically stopped
accepting votes once the limit had been reached. Instead, the machine had
continued to accept votes cast, overwriting earlier votes in order to
accommodate the new ones. The officials said they determined that  4,235
votes were lost.

Not all equipment failures resulted in lost votes, but some did create
technical challenges. Officials in a Colorado jurisdiction stated that
memory cards for optical scan machines at early voting sites sometimes
failed, which meant that all affected optical scan ballots were rescanned
using a new card once poll workers realized that the original card was
malfunctioning. Also, in our September 2005 report on the security and
reliability of electronic voting mentioned earlier, we noted that a
Florida county experienced several problems with its DRE system, including
instances where each touch screen took up to 1 hour to activate and had to
be activated separately and sequentially, causing delays at the polling
place. In addition, we reported that election monitors discovered that the
system contained a flaw that allowed one DRE machine's ballots to be added
to the canvass totals multiple times without being detected. In another
instance, our report notes that a malfunction in a DRE system in Ohio
caused the system to record approximately 3,900 votes too many for one
presidential candidate in the 2004 general election. We also reported that
a statedesignated voting system examiner in a Pennsylvania jurisdiction
noted that the county DRE system had technical problems, such as failure
to accurately capture write-in votes, frozen computer screens, and
difficulties sensing voters' touches.

During our 2005 site visits, officials from 3 jurisdictions also described
several cases of jamming problems with optical scan and punch card ballot
tabulators. For example, election officials in a Kansas jurisdiction we
visited told us that an extensive two-sided optical scan ballot frequently
jammed voting machines because of its length. These officials told us that
they used a two-sided ballot design which required that the optical scan
counting equipment read the ballot front and back, which presented a
problem. According to the officials, the ballot was not scored properly to
feed easily through the equipment and paper jams occurred frequently.
Election officials said the ballots had to be hand-sorted into 13 groups
before scanning, which took time. Similarly, officials in a New Jersey
jurisdiction told us that their optical scan machines had frequently
jammed when reading provisional and absentee ballots. According to the
officials, the ballots had two or three folds, which in combination with
the high volume of ballots being read, jammed the machine regularly. To
repair the jams, officials told us they would straighten ballots and run
them through again, or, if needed, would remake the ballot. Also,
officials in an Illinois jurisdiction we visited said punch cards had also
jammed in their tabulator. Officials there said that this had been likely
due to the punch cards swelling in humid weather, and this problem had
caused the scanner to misread ballots on several occasions. In all of
these instances, the problems were corrected.

While we heard in our site visits about some human error at the polls, in
our survey of local jurisdictions we found that human error was a problem
for a small portion of election jurisdictions in terms of at least one key
function. Specifically, we estimate that 6 percent of local jurisdictions
nationwide experienced poll worker errors in tracking and accounting for
ballots. To the extent that these errors occurred, they were more common
in large jurisdictions. We estimate 1 percent of small jurisdictions,  14
percent of medium jurisdictions, and 34 percent of large jurisdictions had
these errors. The differences between all size categories are
statistically significant. In 10 of the jurisdictions we visited, election
officials cited poll worker or voter errors as the cause of discrepancies
in the number of ballots and voters. In 1 Ohio jurisdiction, for example,
election officials said the discrepancy in the number of ballots and votes
was caused by the fact that poll workers did not track some voters who
left the polling place without voting. In a Florida jurisdiction,
according to election officials, some voters left the polling place
without signing a poll book (which was used to reconcile voter numbers).
Another cause for discrepancies in the number of ballots and voters cited
by election officials in a Washington jurisdiction was that poll workers
erroneously counted some provisional ballots as regular Election Day
ballots, which led to the appearance of more regular Election Day ballots
cast than voters credited with voting in that manner.

Finally, from election officials in 2 jurisdictions we visited, we learned
of voter errors in using voting technology. In one Kansas jurisdiction,
officials reported that some voters did not know how to scroll down the
electronic screen to see all of the information. Also, we were told by
election officials

In the 2004 General Election, Some Election Jurisdictions Reported
Difficulty Meeting Deadlines for Certifying the Final Vote Count

in a New Jersey jurisdiction that poll workers had noticed that some
voters had failed to press a button to finalize their votes. According to
these officials, the poll workers watched for such a mistake, and in at
least one instance, reached under the curtain to register a vote while
both a Democrat and a Republican poll worker observed the maneuver.

According to state survey responses, 7 states (Arkansas, California,
Georgia, Oklahoma, Pennsylvania, South Carolina, and Virginia) encountered
a challenge during the 2004 general election related to timely completion
of the certification process. For example, Georgia election officials
reported difficulty in certifying election results in a timely manner that
would allow a runoff election to commence within 3 weeks of Election Day.
California officials responded that achieving an appropriate balance
between vote count accuracy and the speed of vote tabulation was a
challenge statewide. Arkansas officials said that the Secretary of State's
office had to contact local election jurisdictions numerous times to
receive certified election results in a timely manner.

In some local jurisdictions we visited, we also heard about difficulty
meeting certification deadlines, particularly with regard to provisional
ballots. In 7 local election jurisdictions we visited, election officials
cited concerns with the timing requirements of election certifications.
Specifically, the task of verifying voter information with respect to
provisional ballots and counting provisional ballots made achieving
certification deadlines difficult. For example, officials in 1 Colorado
jurisdiction said that verifying and counting provisional ballots within
the state-mandated 12-day period required that the county hire additional
workers. A Florida jurisdiction reported a similar challenge, but in this
instance, these officials stated that the county canvassing board was
required to consider each provisional ballot individually, which added to
the challenge to meet the short state certification deadline. One large
jurisdiction in Illinois also reported that its 14-day certification
deadline was difficult to achieve because of the large number of
provisional ballots that had to be verified and counted. In a Washington
jurisdiction, officials stated that verifying and counting all ballots
(including provisional ballots) within state-mandated periods had been a
challenge in 2004. In 2005, the Washington state legislature extended the
mandated certification deadline from 15 to 21 days following any general
election.

The Recount in Washington State Revealed the Intricacies of the Election
Process, but Also Yielded Many State Reforms

While the 2004 recount in Washington was one of few statewide recounts
conducted across the country, the types of issues that surfaced during the
recount about Washington's election system identified problems in all
three key elements of elections-people, process, and technology. The close
gubernatorial race and the recount subjected these elements to close
scrutiny, revealing the vulnerability and interdependence of the various
stages of the elections process and the unerring attention to detail that
is required to run an error-free election. It was, in fact, the closest
gubernatorial race in United States history. In the initial statewide
count, a mere 261 votes separated the top two candidates-about 0.001
percent of the total votes cast. An initial recount reduced that margin of
victory to just 42 votes out of more than 2.7 million cast, and the final
recount resulted in a 129-vote margin of victory for the candidate who
came in second in the first two vote counts.

In part because it is the largest election jurisdiction (in number of
voters) in Washington state, King County was the subject of some of the
greatest scrutiny. However, problems were identified by courts in other
jurisdictions in the state as well. As a result of this scrutiny, as
discussed below, Washington state, and King County itself, has
subsequently instituted many reforms.

Breakdowns in Interaction of People, Process, and Technology Can Impair
the Vote Count

We reviewed a variety of reports and studies on this extraordinary
election, including state task force studies, an internal county review, a
management audit sponsored by the Election Center,8 and the findings of a
state Superior Court that resulted from a lawsuit challenging the results
of the final recount. The principal problems we identified in these
materials ranged from poll worker errors to challenges in using equipment.
Described here, they illustrate how breakdowns in the interface of people,
process, and technology may, at any stage of an election, impair an
accurate vote count.

8The Election Center is a nonprofit organization composed of government
employees whose profession it is to serve in voter registration and
elections administration (e.g., voter registrars, elections supervisors,
elections directors). The Election Center in October 2005 completed an
audit of King County's elections operations following the November 2004
general election. The audit was initiated in response to a mandate from
King County's governing council requesting that such an audit be
completed.

Page 277 GAO-06-450 Elections

Provisional Ballots Counted In at least 11 counties provisional ballots
were found by a Washington state

before Being Validated Superior Court to have been counted without
verifying voter signatures or before verification of voter registration
status was completed.9 For example, in Pierce County, Washington, 77
provisional ballots were found by the Superior Court to have been
improperly cast. Provisional ballots were to have included on the ballot
envelope the voter's name and residence. Because the provisional voter's
identity or residence was not marked on the provisional ballot envelope
for these 77 ballots, voter registration status could not be verified. In
King County, the court found that 348 provisional ballots were improperly
cast without verifying voter eligibility. The Election Center management
audit found this had occurred because the provisional voters had been
allowed to put their ballots, which had not been verified, directly into
the optical scan machines at the voting precincts. The Superior Court
found that of these 348 provisional ballots, 252 were ultimately
determined to have been cast by registered voters. According to the audit,
this error resulted from poll worker confusion about who was accountable
for the provisional voting process at the polls. No one poll worker was
assigned responsibility for tracking provisional ballots.

Illegal Votes Cast by Felons The Superior Court also found that more than
1,400 votes had been cast illegally by felons during the November 2004
general election in counties across Washington. Under Washington state
law, in general, persons convicted of a federal or state felony are not
eligible to vote unless their right to vote has been restored. According
to the King County audit, some felons were registered to vote in King
County. The audit stated that election registration officials had very
limited information available to them regarding such felons that would
have allowed them to periodically purge the rolls. Moreover, according to
the audit report, when a former felon who wished to register signed an
affidavit to attest to the fact that his or her voting rights had been
restored, election officials had no expedient way to verify the claim,
particularly for former felons convicted in a different county. In
addition, the audit report noted that election officials did not
necessarily have the authority to refuse to accept a registration form. In
our June 2005 report on maintaining accurate voter registration lists, we
found   

9Borders v. King County, No. 05-2-00027-3 (Chelan County Super. Ct., June
6, 2005) (transcript of oral decision).

Page 278 GAO-06-450 Elections

                         Page 279 GAO-06-450 Elections

        that similar challenges in identifying and removing felons from voter
                                                                        rolls
                    were reported in other states as well.10
                                    The Superior Court found that more votes  
More Votes Counted than Cast     were counted than the number              
                                    of voters credited with voting.           
                                    Specifically, a judge cited evidence of  
                                    190 excess votes counted in Clark County, 
                                    77 excess votes counted in                
                                    Spokane County, 20 excess votes counted   
                                    in Island County, and 14 excess           
                                    votes counted in Kittitas County. In a    
                                    King County internal report, election     
                                    officials reported that the discrepancy   
                                    between voters credited with voting       
                                    and ballots cast was about 0.2 percent,   
                                    or over 1,000 votes. The Election         
                                    Center management audit concluded that    
                                    the discrepancy may have been             
                                    due, in part, to the use of an electronic 
                                    wand held by temporary employees          
                                    to scan the entry codes in the poll book  
                                    when registrants came to vote. The        
                                    audit noted space limitations and         
                                    difficulty hearing the wand's beep when   
                                    it                                        
                                    processed a bar code may have prevented   
                                    an accurate count of voters.              
                                    During our site visit with King County    
                                    officials, they told us that separate     
                                    from the wanding issue, poll worker       
                                    training deficiencies may have            
                                    contributed to discrepancies in the       
                                    number of votes credited and cast when    
                                    voter information was not entered         
                                    properly into poll books.                 
                                    According to the Superior Court's         
Uncounted Ballots Discovered     findings, in several counties uncounted   
                                    ballots were discovered after the         
after Results Had Been Certified certification of the initial election     
                                    results.                                  
                                    The Superior Court found that there were  
                                    64 uncounted absentee ballots             
                                    found in Pierce County and 8 in Spokane   
                                    County. According to the Election         
                                    Center audit, in King County, the         
                                    uncounted ballots were both absentee and  
                                    provisional ballots, and 22 absentee and  
                                    provisional ballots were                  
                                    discovered in the base units of optical   
                                    scan machines after the election was      
                                    certified. The audit concluded that poll  
                                    workers had failed to adhere to           
                                    their procedures for checking these units 
                                    when reconciling ballots after the        
                                    polls closed, and recommended             
                                    strengthening both procedures and         
                                    training.                                 
                                    In King County, during the second         
Absentee Ballots Erroneously     recount, the King County Canvassing       
Disqualified                     Board discovered that election workers    
                                    had disqualified 573 absentee             
                                    ballots during initial canvassing when    
                                    they could not find the voters'           
                                    signatures in the county's new            
                                    computerized voter registration list for  
                                    verification. In addition, the election   
                                    workers had not checked elsewhere         
                                    for these signatures, such as on the      
                                    voters' paper registration forms. In the  

10GAO-05-478.

recount, the King County Canvassing Board decided to recanvass these
ballots to determine whether their disqualification had been appropriate
or whether these ballots should have been counted.11 According to the King
County audit, the voter registration list had been very recently updated,
and for this reason, not all voter signatures had been scanned and
electronically stored in time for the general election so that election
workers would have been able to find them.

Verifying absentee ballots was another issue highlighted during the
recount. According to press accounts, differences existed in how local
jurisdictions in the state verified the signatures of absentee and
provisional voters. The Seattle Times reported conducting a survey in
which it found that signatures went through as many as four levels of
review in one county and only one level in another. Also, the newspaper
reported that some counties would look for as many as six different
identifying traits of a signature, while others "eyeballed the
handwriting." Recommendations by the Governor's Election Reform Task Force
identified the verification of voter signatures as one of several areas
needing more procedural consistency among the counties.

The Washington Recount Washington enacted into law a series of election
reform measures in 2005 designed to clarify, standardize, and strengthen
election requirements and

Fueled Several Reforms at

the State and Local Levels procedures. Several of the statewide reforms
specifically address problems described above, but others are broader
measures designed to improve election administration. Examples of these
measures are listed below.

o  Unique provisional and absentee ballots: All provisional and absentee
ballots are required to be visually distinguishable from one another and
must be either printed on colored paper or imprinted with a bar code for
the purpose of identifying the ballot as a provisional or absentee ballot.
The bar code must not identify the voter. Provisional and absentee ballots
must be incapable of being tabulated by polling place counting devices.

11The King County Canvassing Board's authority to recanvass these ballots
was litigated in late 2004. The Washington State Supreme Court eventually
ruled that the county canvassing board could, in its discretion, recanvass
the 573 uncounted ballots (eventually totaling 735 after another 162
ballots were discovered) in certain circumstances, thereby allowing the
recanvassing of the ballots and inclusion in the total tally of votes.
Wash. State Republican Party v. King County, 103 P.3d 725 (2004); See also
In Re Election Contest filed by Coday, 2006 Wash. LEXIS 185 (Wash. Mar. 9,
2006.).

     o Standardized guidelines for signature verification processes: The
       Secretary of State is to establish guidelines for signature
       verification relating to, for example, signatures on absentee and
       provisional ballot envelopes. All election personnel assigned to
       verify signatures are required to receive training on the established
       guidelines. State law also provides that while signatures on certain
       mail-in ballot envelopes (such as absentee ballots) must be compared
       with the voter's signature in the county registration files, variation
       between the signature on a return envelope and the signature of that
       voter in the registration files due to the substitution of initials or
       the use of common nicknames (e.g., Joseph Smith versus Joe Smith) is
       permitted so long as the surname and handwriting are clearly the same.
     o Triennial review of county election processes and reports listing
       corrective actions: Instead of being performed periodically, 
       state-conducted reviews of county election-related policies,
       procedures, and practices are to be performed at least once every 3
       years. If staffing or budget levels do not permit a 3-year review
       cycle, such reviews must be done as often as possible. The county
       auditor or the county canvassing board must respond to the review
       report in writing, listing steps to be taken to correct any problems.
       Before the next primary or general election, the Secretary of State's
       office must visit the county and verify that the corrective action has
       been taken.
     o Election law manuals for use in all vote-counting centers: The
       Secretary of State must prepare a manual explaining all election laws
       and rules in easy-to-understand, plain language for use during the
       vote counting, canvassing, and recounting process. The manuals must be
       available for use in all vote-counting centers throughout the state.
          * Option to conduct voting entirely by mail: Another change
            introduced by the state, which may avoid errors at the polls, has
            been to give county officials the option to conduct elections
            entirely by mail. The new measure authorizes the use of all-mail
            voting in counties upon the express approval by a county's
            legislative authority and provides that such approval must apply
            to all primary, special, and general elections conducted by the
            county. For example, King County has announced plans to conduct
            elections entirely by mail in 2007. The King County Independent
            Task Force on Elections found in 2005 that the King County
            election process basically involved simultaneously conducting two
            dissimilar elections. The task force stated that increasingly, a
            majority of King County voters (565,011, or slightly more than 
          * 62 percent in 2004) used the permanent absentee or vote-by-mail
            process. Despite this fact, the task force reported that the
            county also conducted a traditional election involving about
            330,000 voters assigned to over 2,500 precincts and 540
            individual polling places, and the use of hundreds of temporary
            election workers who must be trained and who work at the polling
            places for more than 13 hours on election days. Furthermore, the
            task force stated that both election processes contain
            independent, complex, and often conflicting requirements that
            have clearly caused significant problems for King County election
            officials. Having one means of voting for all citizens is
            perceived to be both more efficient and cost-effective than the
            previous process, according to the task force.
          * Paper records for electronic voting devices and precertification
            audits of electronic voting results: All electronic voting
            devices must, beginning January 1, 2006, produce an individual
            paper record of each vote, at the time of voting, that may be
            accepted or rejected by the voter before finalizing his or her
            vote. This audit is to be conducted by randomly selecting a
            specified percentage of electronic voting devices and, for each
            device, comparing the results recorded electronically with the
            paper records. The audit process must be open to observation by
            political party representatives if such representatives have been
            appointed and are present at the time of the audit.
          * Separate from changes made at the Washington state level, King
            County, as reported in the Election Center audit, also
            implemented or was in the process of implementing changes to
            improve election administration that specifically address issues
            that arose during the 2004 general election.12 Examples of such
            reported changes are below:
     o Controls to manage provisional ballots: Provisional ballots will be
       colorcoded for easy recognition and will have timing marks that
       prevent the counter at the polling place from accepting them.
       Therefore, the voter has no option but to return his or her
       provisional ballot to a poll worker, who will place it in a
       provisional envelope. One additional poll worker is to be assigned to
       each polling place to exclusively manage provisional ballots for all
       voters at that polling place.

12Election Center, King County Elections Operations (Houston, Texas,
2005).

Concluding Observations

     o Controls to prevent misplaced ballots: Poll workers are required to
       record the serial number located at the bottom of the optical scan
       bins on the ballot reconciliation transmittal form. The serial number
       is not visible if any ballots remain in the bin. Increased poll worker
       training, attaching a flashlight to the inside of each bin, and
       continued adherence to existing procedures for troubleshooters to
       examine each bin before certification are also intended to help ensure
       that all ballots are properly handled and counted in future elections.
     o Additional procedures for tracking absentee ballots and registration
       signatures: King County performed a database search of the entire
       voter file prior to the fall 2005 elections, in order to identify
       missing or unreadable signatures. On the basis of the search results,
       elections personnel contacted voters and made significant progress in
       updating the files. In addition, procedures at the absentee ballot
       operation center have been enhanced. New logs were created for
       tracking absentee ballots that required additional research because
       they were not easily verified. Also, in any instance where a voter
       registration signature is not on file, or is illegible, a search for
       the original record, as well as a call and a letter to the voter, is
       required.
     o Improvements to procedures for reconciling ballots and voters: For the
       2005 primary and general elections, the use of electronic hand wands
       to scan poll books, when reconciling ballot and voter numbers, was to
       be done at a county center where more space would be available. New
       checklists were developed that required staff to balance the number of
       signatures recorded with the wand against the number of ballots
       counted by the computer. Also, the hand-wand process was to occur at
       the beginning rather than at the end of the canvass to allow more time
       for any necessary research into potential discrepancies.

Although the methods used to secure and count ballots vary across the  50
states and the District of Columbia, the goal of vote counting is the same
across the nation: to accurately count all ballots cast by eligible
voters. As with the elections process overall, conducting an accurate vote
count is not a simple process. It requires many steps, an unerring
attention to detail, and the seamless integration of people, processes,
and technology.

Providing eligible voters multiple means and times within a jurisdiction
for casting their ballots-early, absentee, provisional, and Election Day
voting-enhances eligible voters' opportunity to vote. At the same time,
multiple voting methods and types of ballots can make the vote-counting
process more complicated. In addition, short deadlines for certifying the
final vote-as little as 2 days in 1 state-provide little time for election
officials to review, verify, and count provisional and absentee ballots.
Larger jurisdictions generally face more challenges than smaller
jurisdictions because of the sheer volume of votes cast by all ballot
types- absentee, provisional, and regular ballots. Provisional ballots
were new for many jurisdictions in November 2004 and created some
challenges in tracking, verifying, and counting. On the basis of their
experience in November 2004, some jurisdictions are implementing new
procedures for provisional voting, such as printing provisional ballots in
a color different from other types of ballots or using paper ballots
rather than DRE machines for provisional voters.

Two jurisdictions we visited in Washington have announced plans to move to
all-mail elections, which was authorized on a county-wide basis by recent
state law. Although replacing in-person voting with all-mail voting
eliminates some challenges-e.g., poll worker training on voting equipment
operations and provisional voting or the chance of malfunctioning voting
equipment at the polls-in some circumstances it could magnify the
importance of other aspects of state election processes, such as verifying
votes, accurately matching voter signatures and having guidance for
determining voter intent from improperly or unclearly marked ballots. For
those jurisdictions allowing or requiring the determination of a voter's
intent from an improperly or unclearly marked ballot, the importance of
having explicit and consistent criteria for treating such ballots became
evident in the 2000 general election when different interpretations for
such ballots in Florida made the close presidential race extremely
contentious. Eighteen states that reported they did not have voter intent
guidance in place for the November 2000 general election reported to us in
our state survey that they did have voter intent requirements or guidance
in place for the November 2004 general election. While federal election
provisions do not address the state counting issue of ascertaining voter
intent, HAVA did require states to adopt, by January 2006, uniform and
nondiscriminatory standards defining what constitutes a vote and what will
be counted as a vote for each type of voting system used by the state.

The recount in the close gubernatorial election in Washington revealed the
interdependence of every stage of the elections process in ensuring an
accurate vote count. That experience also illustrated how small errors in
election operations can affect the vote counting process. Were any state's
election processes subjected to the very close scrutiny that characterized
the recount in Washington state, it is likely that imperfections would be
revealed. Votes are cast and elections are conducted by people who are not
and cannot be 100 percent error free in all their tasks all the time.
Thus, the consistently error-free vote count may be elusive, particularly
in very large jurisdictions with hundreds of thousands of ballots cast in
person, absentee, or provisionally. However, diligent efforts to achieve
consistent error-free vote counts can help to ensure that any errors are
reduced to the minimum humanly possible.

Chapter 7

                        Voting Methods and Technologies

Voting methods can be thought of as tools for accommodating the millions
of voters in our nation's more than 10,000 local elections jurisdictions.
These tools are as simple as a pencil, paper, and a box, or as
sophisticated as programmable computer-based touch screens.1 Regardless of
method, however, the proper operation and functioning of each depends on
its effective interplay with the people who participate in elections (both
voters and election workers) and the processes (governed by policies,
procedures, and so forth) that govern the interaction of people with one
another and with the voting method.

This chapter focuses on voting methods-the technology variable in the
people, process, and technology election equation. It describes the use of
voting methods in the 2004 general election, compares this technology
environment with that of the 2000 general election, and examines plans for
voting technologies in the 2006 election, particularly in light of the
roles being played by states and HAVA. It also examines efforts to measure
and understand how well voting equipment performed in the 2004 election 
(see fig. 53 for equipment examples), including the state of performance
standards and local jurisdictions' overall satisfaction with their
respective voting methods. Additionally, this chapter discusses the state
of practice relative to voting system security, testing, and integration,
and presents key challenges facing all levels of governments as voting
systems, related election systems, and supporting technologies continue to
evolve.

1As described in the glossary of this report, the five types of voting
methods are paper ballot, lever machine, punch card, optical scan, and
direct recording electronic (DRE). All except paper ballot are considered
automated methods. The punch card, optical scan, and DRE methods utilize
computers to automate voting or tabulation. Of these three, punch card and
optical scan equipment typically automate vote counting, while DREs
automate both vote casting and counting. Optical scan and DRE are
considered the more technology-based voting methods.

                    Figure 53: Examples of Voting Equipment

Source: GAO.

DRE (top left), precinct count optical scan (top center), central count
optical scan (top right), lever machine (bottom left), precinct count
punch card (bottom center), and central count punch card (bottom right).

The technology of the voting environment can be characterized as varied

Overview

and evolving, according to our 2005 state survey results and local
jurisdiction survey estimates. We estimate on the basis of our local
jurisdiction survey that the predominant voting methods most often used
for the 2004 general election by large jurisdictions were DRE and precinct
count optical scan, while medium jurisdictions most often used precinct
count optical scan and small jurisdictions most often used paper ballot.
In addition, the predominant voting method most often used for large
jurisdictions changed from precinct count optical scan in 2000 to both DRE
and precinct count optical scan in 2004, while the predominant voting
methods remained the same for the other jurisdiction sizes. Also in the
2004 general election, an estimated one-fifth of jurisdictions used
multiple voting methods to support voting activities. Most states
generally exercised influence over the voting methods used by their
respective elections jurisdictions through a range of approaches such as
requiring the use of one specific voting method, helping with local
acquisition efforts, or eliminating voting methods, according to our 2005
state survey. Ten states and the District of Columbia reported that they
required the use of one specific method for the 2004 general election, and
4 additional states planned to require a specific method for the 2006
general election. Sixteen states and the District of Columbia reported
that they were involved to some extent in local jurisdiction efforts to
acquire voting systems, components, and services. States also reported
that they were eliminating lever and punch card equipment between the 2000
and 2006 general elections. Specifically, for the November 2000 general
election, 37 states reported that they used lever or punch card voting
equipment; by the November 2006 general election, only 4 states had plans
to use lever and punch card equipment. HAVA has influenced state and local
decisions regarding particular voting methods by providing funds to states
to replace punch card and lever voting equipment with other voting
methods. This greater state involvement in jurisdictions' choice of voting
methods, combined with federal funding to replace lever and punch card
voting equipment and certain HAVA requirements-among other factors-is
likely to influence the adoption of DRE and optical scan voting methods.

Federal and state standards provide an important baseline for the
performance of voting systems and were widely adopted for the 2004 general
election. However, according to our local jurisdiction survey, voting
equipment performance was not consistently measured during the 2004
general election and varied by jurisdiction size and voting method, in
part because some types of measures were not well suited to particular
voting methods. For example, small jurisdictions were generally less
likely to collect accuracy measures such as accuracy of voting equipment
(estimated at 31 percent for small jurisdictions) than large and medium
jurisdictions (66 percent and 54 percent, respectively), and this may be
because the predominant voting method most used by small jurisdictions was
paper ballot. On the other hand, on the basis of our local jurisdiction
survey, we estimate that the vast majority of all jurisdictions were very
satisfied or satisfied with their systems' performance during the 2004
general election. For instance, we estimate that 78 percent of
jurisdictions were very satisfied or satisfied with the accuracy of their
voting system performance. The estimated high satisfaction levels
demonstrated across different voting system performance areas and
jurisdiction sizes contrast with our lower estimates of the performance
measures that were collected for the 2004 general election. Although the
reasons for moderate collection levels for performance measures are
unclear, jurisdictions that may not have collected performance data or may
have considered such information not applicable to their situation may
lack sufficient insight into their system operations to adequately support
their satisfaction in the variety of performance areas we surveyed. The
moderate collection levels of data on operational voting system
performance may present a challenge to state and local election officials
in their efforts to make informed decisions on both near-term and
long-term voting system changes and investments.

A wide range of recently published concerns for the security of voting
systems and the development of nationwide mechanisms under HAVA to improve
security standards and processes have not yet produced a consistent
approach across all jurisdictions for managing the security of voting
systems. Our 2005 local jurisdiction survey and our visits to local
jurisdictions found that voting system security has been primarily
shouldered by local jurisdictions. However, states, vendors, law
enforcement officials, and others shared in these efforts to varying
degrees for the 2004 general election. Our state survey for the 2004
general election and visits to local jurisdictions indicated that security
mechanisms employed by some states-but not others-included promulgation of
policies and guidance, compliance of voting equipment with security
standards, and monitoring and evaluation of implemented security controls.
According to our local jurisdiction survey estimates and visits to local
jurisdictions, jurisdictions and their support organizations were largely
responsible for implementation of security controls, such as access
restrictions to voting equipment, system backup capabilities, and
securityrelated testing. Estimates from our local jurisdiction survey also
showed, however, that many jurisdictions nationwide had not documented
their security measures, and we found that several of the jurisdictions we
visited reported that they had not implemented recommended measures, such
as security plans, training, and documentation of policies and procedures.
Furthermore, decisions by states to continue using outdated voting system
standards may allow the vulnerabilities of newer technologies to go
unevaluated and impair effective management of the corresponding security
risks. States and local jurisdictions face the challenge of regularly
updating and consistently applying appropriate standards and other
directives to meet the vulnerabilities and risks of their specific
election environments.

Testing and evaluation of voting systems also varied across states and
jurisdictions for the 2004 general election. Our state survey found that
most states required certification testing of their voting systems using a
range of criteria. However, responsibility for purchasing a certified
system typically rested with local jurisdictions. Other results from our
2005 state survey and responses from jurisdictions we visited indicated
that acceptance testing continued to be commonly performed, but there was
wide variation in the responsibilities and practices for this type of
testing, including whether such testing was applied to new systems or
upgrades, the extent of vendor participation, and the coverage of hardware
and software functions. Also on the basis of our local jurisdiction
survey, we estimate that most jurisdictions conducted readiness (logic and
accuracy) testing for the 2004 general election as they did for the 2000
election, but in some jurisdictions we visited, we found they used
different procedures that may have included one or more processes such as
diagnostic tests, mock elections, or suites of test votes. In contrast,
our local survey estimates indicate that parallel testing was employed by
fewer than an estimated 2 percent of jurisdictions. This may be due to, in
part, the lack of directives for conducting such tests.2 Finally,
postelection voting system audit tests were conducted by fewer than half
of jurisdictions for the 2004 general election, according to our local
survey estimates, although many more large and medium jurisdictions
performed these tests than small jurisdictions. As with other types of
testing, the requirements and practices for audit tests were diverse.
Factors associated with the testing of voting systems may further
challenge states and local jurisdictions as they adapt to changes in
voting system capabilities, standards, and national certification for the
2006 general election. Those factors are likely to include increased
certification testing workloads to recertify systems with new
capabilities, ongoing limits to the number of available testing
laboratories until a new laboratory accreditation process becomes fully
operational, and more complex testing because a new version of the federal
voluntary voting system guidelines has been added in 2005 to older federal
standards from 1990 and 2002 that states are already using.

The number of jurisdictions that had integrated particular aspects of
voting system components and technologies was limited for the 2004 general
election, according to estimates from our local jurisdiction survey and
visits to local jurisdictions for the selected areas of integration we
examined, such as electronic programming or setup and electronic

2We estimate that 91 percent of jurisdictions considered parallel testing to be
                                not applicable.

                         Page 290 GAO-06-450 Elections

Voting Methods Vary among Jurisdictions and Are Being Influenced by
States' Choices and HAVA

management. Two-thirds of the jurisdictions we visited told us that they
used electronic programming or setup of voting equipment, and an estimated
7 percent of jurisdictions that used voting methods other than paper
ballots, according to our local survey, connected their voting equipment
via a local network at polling locations. Relatively few local
jurisdictions we visited also reported having plans for integrating or
further integrating their election-related systems and components for the
2006 general election, and in the instances where jurisdictions reported
plans, the scope and nature of the plans varied. For instance, officials
at 5 jurisdictions we visited reported plans to introduce a
voter-verifiable paper trail (VVPT) capability for future elections, and
officials from 1 jurisdiction reported plans to purchase an optical
scanner with the ability to tabulate both DRE and optical scan election
results. Nevertheless, the potential for greater integration in the future
does exist as states and jurisdictions act on plans to acquire the kind of
voting equipment (e.g., optical scan and DRE products) that lends itself
to integration. For example, on the basis of our local jurisdiction
survey, we estimate that at least one-fifth of jurisdictions plan to
acquire DRE or optical scan equipment before the 2006 general election,
and officials from 2 jurisdictions we visited who used DRE equipment told
us that their state planned to purchase electronic poll books for its
precincts to use during the 2006 elections to electronically link its
voter registration system with its voting systems. It is unclear if and
when this migration to more technology-based voting methods will produce
more integrated election system environments. However, suitable standards
and guidance for these interconnected components and systems-some of which
remain to be developed-could facilitate the development, testing,
operational management, and maintenance of components and systems, thereby
maximizing the benefits of current and emerging election technologies and
achieving states' and local jurisdictions' goals for performance and
security. The challenge inherent in such a dynamic environment is to
update system standards so that emerging technical, security, and
reliability interactions are systematically addressed.

The technology of the voting environment can be characterized as varied
and evolving, according to our 2005 state survey results and local
jurisdiction survey estimates. We estimate on the basis of our local
jurisdiction survey that the predominant voting methods most often used
for the 2004 general election by large jurisdictions were DRE and precinct
count optical scan, while medium jurisdictions most often used precinct
count optical scan and small jurisdictions most often used paper ballot.

Page 291 GAO-06-450 Elections

Voting Methods Differ by the Size of Local Jurisdictions and States'
Choices

Two key patterns emerged in the use of voting methods between the 2000 and
2004 general elections. First, we estimate that the percentage of large
jurisdictions using DREs doubled from 15 percent in the 2000 general
election to 30 percent in 2004. The predominant voting method for large
jurisdictions changed from precinct count optical scan in 2000 to both DRE
and precinct count optical scan in 2004. In contrast, we estimate that the
predominant voting methods remained the same for small and medium
jurisdictions (paper ballots and precinct count optical scan,
respectively) from 2000 to 2004. Furthermore, on the basis of our local
jurisdiction survey, we estimate that at least one-fifth of jurisdictions
plan to acquire DRE or optical scan equipment before the 2006 general
election. Second, in response to our state survey, 9 states reported that
they eliminated the lever machine and punch card voting methods for the
2004 general election. In addition, 18 other states plan to eliminate
lever or punch card voting methods for the 2006 general election. This
greater state involvement in jurisdictions' choice of voting methods, the
availablilty of federal funding to replace lever and punch card voting
equipment, and certain HAVA requirements-among other factors-are likely
influences on the adoption of DRE and optical scan voting methods.

Since the November 2000 general election, the DRE voting method has become
more widely used in large jurisdictions, according to our local
jurisdiction 2005 survey. During the same period, states' reported use of
lever machine and punch card voting methods has decreased, according  to
responses to our 2005 state survey. Our state and local jurisdiction
surveys also indicate plans for changes to voting technologies for the
2006 general election.

Overall, the estimated percentages of predominant voting methods used by
local jurisdictions in the 2000 and 2004 general elections did not change
appreciably. In particular, from our local jurisdiction survey, we
estimate that the mix of predominant voting methods used in the November
2000 general election was 5 percent DRE, 21 percent central count optical
scan, 26 percent precinct count optical scan, 5 percent central count
punch card, 2 percent precinct count punch card, 8 percent lever, and 31
percent paper.3 In comparison, we estimate that the mix for the November
2004 general

3We defined the predominant voting method as one that processed the
largest number of ballots regardless of when the vote was cast: on general
Election Day, as a provisional vote, during absentee voting, or during
early voting.

Page 292 GAO-06-450 Elections

election (in the same order) was 7 percent DRE, 21 percent central count
optical scan, 30 percent precinct count optical scan, 2 percent central
count punch card, 2 percent precinct count punch card, 7 percent lever,
and 30 percent paper. Figure 54 compares these percentage changes.

Figure 54: Estimated Percentages of Jurisdictions Using Predominant Voting
Methods in the 2000 and 2004 General Elections

Percentage of jurisdictions

                                      scan

                                      scan

                                  ver machine

Central count

                          Precinct countCentral count

                           Precinct countaper ballot

DRE

                                       d

                                       d

                                     h car

                                     h car

optical

                                    optical

                                      punc

                                      punc

                                       P

                                       Le

Predominant voting method

2000

2004 Source: GAO 2005 survey of local election jurisdictions.

According to our local jurisdiction survey, there may have been a small
shift away from punch card and lever machine voting methods (estimated at
3 percent or 1 percent loss of jurisdictions, respectively) and may have
been an increase in optical scan and DRE voting equipment (estimated at 
5 percent and 2 percent gain of jurisdictions, respectively) for the 2004
general election. However, these differences are not statistically
significant. During the same time frame, we estimate that 16 percent of
jurisdictions acquired new voting equipment through their own purchases or
leases and 15 percent of jurisdictions through purchases or leases by
their state. Thus, the new voting equipment acquired by many jurisdictions
since 2000 did not substantively affect the predominant voting methods
that were already in use.

One notable change did occur, however, in the use of predominant voting
methods in the 2000 and 2004 general elections. The percentage of large
jurisdictions using DREs doubled (estimated at 15 percent in 2000 and  30
percent in 2004, respectively)-an increase that is statistically
significant. This increase in the use of DREs changed the predominant
voting method most often used for large jurisdictions, which was precinct
count optical scan in 2000, to both DRE and precinct count optical scan in
2004. A smaller increase in the use of DREs among medium jurisdictions
(from an estimated 13 percent in 2000 to 20 percent in 2004) is not
statistically significant, and there was virtually no change in DRE use
among small jurisdictions (an estimated 1 percent for both elections). In
contrast, the use of paper ballots as a predominant voting method did not
appreciably change between the 2000 and 2004 general elections (with
overall use at 30 percent in 2000 and 31 percent in 2004, respectively).
Small jurisdictions were the major contributors to this steady use of
paper ballots (estimated at 43 percent in 2000 and 41 percent in 2004,
respectively); medium jurisdictions were minor contributors (3 percent for
each election). (No large jurisdictions used paper ballots as their
predominant voting method for either of these elections.) We also estimate
that use of precinct count optical scan as the predominant voting method
for medium jurisdictions did not change appreciably between the 2000 and
2004 elections (estimated at 35 percent in 2000 and 39 percent in 2004,
respectively). Figure 55 shows the estimated use of predominant voting
methods for small, medium, and large jurisdictions in the 2004 general
election.

Figure 55: Estimated Percentages of Jurisdictions Using Predominant Voting
Methods in 2004, by Jurisdiction Size Percentage of jurisdictions

41

40 35 30 25 20 15 10 5

       DREa           Central count   Precinct count 
                      optical scanb     optical scan 
Predominant voting method        

Central count punch      Precinct count punch Lever machinec Paper ballota 
cardb                    card                                
                  Small (<10,000)                               
                         Medium (10,000-100,000)                
                 Large (>100,000)                               

Source: GAO 2005 survey of local election jurisdictions.

Note: Percentages for predominant voting methods within each jurisdiction
size may not add to 100 because of rounding.

aThe differences between small jurisdictions and both medium and large
jurisdictions are statistically significant.

bThe differences between both small and medium jurisdictions and large
jurisdictions are statistically significant.

cThe difference between small jurisdictions and large jurisdictions is
statistically significant.

The more widespread adoption of DREs by large jurisdictions was consistent
with their greater proportion among jurisdictions that acquired voting
equipment since 2000. According to our local jurisdiction survey, we
estimate that 37 percent of large jurisdictions bought or leased new
voting equipment since 2000, compared with 21 percent of medium
jurisdictions and 12 percent of small jurisdictions, where the differences
between large jurisdictions and both medium and small jurisdictions are
statistically significant.

Furthermore, on the basis of our local jurisdiction survey, we estimate
that at least one-fifth of jurisdictions plan to acquire DRE or optical
scan equipment before the 2006 general election.4 Both large and medium
jurisdictions are more likely to have plans to acquire DREs before the
November 2006 general election (estimated at 34 percent each) than small
jurisdictions (estimated at 13 percent), while small jurisdictions are
more likely to have plans to acquire precinct count optical scan voting
equipment (estimated at 28 percent) than medium or large jurisdictions
(estimated at 17 percent and 15 percent, respectively). In general, fewer
jurisdictions expected to acquire central count optical scan voting
equipment than the other two voting methods, although the differences were
not statistically significant. The percentages of jurisdictions planning
to acquire the newer voting systems before the next general election are
shown in figure 56 by the size of jurisdiction.

4Jurisdictions could identify more than one voting method to be acquired.

Figure 56: Estimated Percentages of Jurisdictions with Plans to Acquire
Particular Voting Systems before the November 2006 General Election

Percentage of jurisdictions

34 34

30

28

25

20

                                       17

15 13 11

15

15

10

5

4

0

DREa Central count Precinct count optical scanb optical scanc

Voting method to be acquired

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000) Source: GAO 2005 survey of local election jurisdictions.

aThe differences between small jurisdictions and both medium and large
jurisdictions are statistically significant. The 95 percent confidence
interval for large jurisdictions planning to acquire DREs is +/- 8
percentage points.

bThe difference between small jurisdictions and medium jurisdictions is
statistically significant.

cThe difference between small jurisdictions and large jurisdictions is
statistically significant.

Another interesting pattern emerged in voting methods between November
2000 and November 2004 at the statewide level. Thirty-seven states
reported that at least 1 jurisdiction used lever machine or punch card
voting equipment for the November 2000 general election. By the time of
the November 2004 general election, the number of states that continued to
employ these voting methods decreased to 28. Specifically, our state
survey results show that 9 states reported that they completed replacement
of all their punch card or lever voting equipment before the November 2004
general election, and 4 other states reported that they completed their
replacements since the 2004 election. Of the remaining 24 states that
reported using the punch card and lever methods in 2000 but had not yet
replaced them at the time of our survey, 18 reported that they planned to
replace all punch card and lever voting equipment by the November 2006
general election, while 3 planned to replace a portion of their equipment
by then. One state reported no replacement plans prior to the November
2006 general election. Figure 57 summarizes the states' progress and plans
for replacing punch card and lever voting equipment.

Figure 57: State-Reported Status of Lever and Punch Card Voting Equipment
in 2000, 2004, and 2006

Sources: GAO 2005 survey of state election officials and follow-up with 2
states for clarification of responses (analysis), MapArt (map).

Our local jurisdiction survey provided insight into jurisdictions' plans
for acquiring technology-based voting methods and the time frames for
executing these plans, which may increase the predominance of these
methods in future elections. Specifically, we estimate that 25 percent of
local jurisdictions are planning to acquire precinct count optical scan
machines by the November 2006 general election, 19 percent expect to
acquire DREs by then, and about 7 percent plan to acquire central count
optical scan equipment before that election.5 In addition, we estimate
that between 4 and 10 percent of local jurisdictions had plans to acquire
additional equipment in each of these voting methods but had not set a
target date for doing so at the time of our survey.6

During visits to election jurisdictions across the country, local election
officials explained some of their motivations behind plans to acquire DRE
or optical scan voting equipment. For example, election officials in  6
jurisdictions cited HAVA as the reason for purchasing new DRE equipment,
particularly HAVA's requirement that each voting place have at least one
voting method that is accessible to persons with disabilities, as we
discussed earlier in chapter 4.7 More specifically, officials in 1 large
jurisdiction in Connecticut said that they would evaluate the use of DREs
to meet HAVA accessibility requirements before deciding whether to
purchase more DREs in time for the November 2006 general election.
Election officials from 5 other jurisdictions stated that they planned to
purchase new voting equipment to provide a VVPT, a requirement levied by 3
of the 14 states we visited (Colorado, Nevada, and New Mexico). Officials
from 5 other jurisdictions said that they expected to acquire new voting
equipment but did not give a reason and, in some cases, did not yet know
what type of equipment they would obtain. Officials in jurisdictions that
did not plan to purchase new voting equipment told us that their existing
equipment was sufficient or that budget constraints prevented the
acquisition of new equipment, among other reasons.

5Some jurisdictions may be planning to acquire equipment for more than one
voting method.

6We also estimate that about one-fourth of jurisdictions did not know
whether they planned to acquire DRE or optical scan voting equipment in
time for the 2006 general election. 7HAVA S: 301(a)(3) (codified at 42
U.S.C. S: 15481(a)(3)).

Page 300 GAO-06-450 Elections

                 More Jurisdictions Used Several Voting Methods

As for the 2000 general election, some jurisdictions used multiple voting
methods to support the 2004 general election, and some of these methods
were more widely used than others for particular types of voting. In our
October 2001 comprehensive report on election processes nationwide, we
reported that 5 percent of jurisdictions used more than one voting
method.8 On the basis of our 2005 local jurisdiction survey, we estimate
that  21 percent of jurisdictions used more than one voting method in the
November 2004 general election, with the most common combination of
methods being central count optical scan with paper ballot (estimated to
be 5 percent of jurisdictions).9 Other common combinations in 2004 were
lever machine with paper ballot (4 percent) and DRE with paper ballot  (3
percent). DRE with central count optical scan was one of numerous other
combinations used by 2 percent or less of local jurisdictions. Figure 58
shows the estimated proportion of jurisdictions with the most prevalent
single and combination voting methods.

8GAO-02-3.

9The percentage of jurisdictions reported for the November 2000 general
election was based on GAO analysis of data from Election Data Services and
states; the current estimate of jurisdictions is based on our 2005 survey
of local election jurisdictions. Some of the differences may be due to
differences in these methods.

Page 301 GAO-06-450 Elections

Figure 58: Estimated Percentages of Local Jurisdictions That Used Single
and Multiple Voting Methods in the November 2004 General Election

Central count optical scan and paper Lever and paper

3%

DRE and paper All other combinations

1%

DRE

Optical scan

3%

Punch card

2%

Lever Paper ballot

Single voting method Multiple voting methods

Source: GAO analysis of GAO's 2005 survey of local election jurisdictions.

Note: Percentages for single and multiple voting methods do not add to 100
because of rounding.

The specific mix of voting methods used can also be viewed with respect to
particular types of voting (e.g., absentee, early, provisional) that were
supported in the 2004 election. In this regard, some voting methods were
applied to a particular type of voting more frequently than others. We
estimate that paper ballot was the most widely used voting method for
absentee voting (36 percent of jurisdictions), provisional voting  (18
percent), and early voting (8 percent). Precinct count optical scan (shown
in fig. 59) was generally the second most widely used voting method for
these types of voting (24 percent of jurisdictions for absentee, 10
percent for provisional, and 5 percent for early voting, respectively),
while central count optical scan was the third most widely used method 
(20 percent of jurisdictions for absentee, 9 percent for provisional, and
 5 percent for early voting, respectively).

            Figure 59: Precinct Count Optical Scan Voting Equipment

Jurisdictions' Voting Methods Were Influenced by the States and HAVA

Source: GAO. Precinct count optical scan voting equipment at the warehouse
(left), polling location (center), and tabulation facility (right).

Most states have generally exercised influence over the voting methods
used by their respective elections jurisdictions through a range of
approaches. In particular, for our state survey, a majority of states (32)
and the District of Columbia said that they restricted the voting methods
employed by local jurisdictions in the 2004 election either by requiring
the use of one specific method (10 states and the District of Columbia) or
providing a list of approved voting methods for the jurisdiction to select
from (22 states). An alternate approach reported by 10 states was to
require local jurisdictions to obtain state approval when selecting a
voting method. The remaining 8 states said that local jurisdictions chose
the voting method they used without any state involvement.

In addition to affecting the choice of voting methods, 16 states and the
District of Columbia reported that they were involved to some extent in
local jurisdiction efforts to acquire voting systems, components, and
services. For example, 1 state reported that it evaluated voting equipment
options and vendors, and then contracted with a single vendor to supply
voting equipment for all jurisdictions in the state. Jurisdictions within
this state then had the option of purchasing additional voting equipment
from this vendor, as needed. The top map of figure 60 shows the role of
each state in the selection of specific voting methods for jurisdictions
in the 2004 general election.

Figure 60: State-Reported Involvement in the Use of Specific Voting
Methods in the 2004 and 2006 General Elections

Sources: GAO 2005 survey of state election officials (analysis), MapArt
(map).

Responses to our state survey indicate that state influence over the
voting methods to be used in the November 2006 general election will
continue to increase. Four additional states planned to require the use of
a single voting method statewide, which will bring the total number of
states doing so to 14, and the District of Columbia will do so as well.
Also, 5 additional states reported that they will require local
jurisdictions to select a voting method or methods from a state-approved
list, bringing this total to 27; 8 states intended to continue to allow
local jurisdictions to select their voting methods with state approval.
Only 1 state was not expecting to be involved in decisions on voting
methods for its jurisdictions for 2006. The bottom map of figure 60 shows
the role of each state in the selection of specific voting methods for
jurisdictions in the 2006 general election.

Consistent with state survey responses indicating their contributions to
local jurisdictions' selection of voting methods and on the basis of our
local jurisdiction survey, one of the most frequent factors that
influenced the  16 percent of local jurisdictions that bought or leased
new voting equipment since the November 2000 general election was state
requirements or certification of the equipment (an estimated 83 percent of
the 16 percent of jurisdictions that bought or leased the new voting
equipment). Other widely influential factors included ease of equipment
use (91 percent), vendor demonstrations (72 percent), and affordability 
(68 percent). In contrast, local requirements and HAVA funding were less
influential factors for local jurisdictions' acquisition of voting
equipment (44 percent and 45 percent of jurisdictions, respectively). (See
fig. 61.)

Figure 61: Estimated Pe r centag es of Local Jurisdictions Identifying F
actors That Influenced Buying or Leasing Voting Equipment since the 2000
General Election

P e rcentag e of jurisdictions 100

91

90

83

80

                                       72

                                       68

70

                                       63

                                     59 59

60

                                       51

                                       48

50

                                     45 44

                                 40 30 20 10 0

oter ease of use

                State requirementsendor demonstrationstification

                       System or vendor requirementswhere

                                     system

                                  successwhere

                    Local requirementsState funding funding

                                    dability

                                 ectiveness or

                                      else

                              or failure elseAffor

                             ormance of prior HAVA

or cer

                                   ception of

                                    ormance

                                  Cost effHAVA

                                      Per

                                      perf

                                       V

                                      perf

                  Factors in buying or leasing voting systems

Source: GAO 2005 survey of local election jurisdictions.

Note: These percentages represent responses from the estimated 16 percent
of our sample of local jurisdictions that responded to our local
jurisdiction survey that they bought or leased new voting equipment for
the 2004 general election or later. The 95 percent confidence interval for
values in all categories is +/- 13 percentage points or less.

HAVA has also influenced state and local decisions regarding particular
voting methods through mechanisms to encourage the adoption of technology.
Among other things, HAVA provided funds to states to replace punch card
and lever voting equipment with other voting methods (Section 102
funds).10 During fiscal year 2003, the General Services Administration

10HAVA S: 102 (codified at 42 U.S.C. S: 15302). HAVA S: 101(b) (codified
at 42 U.S.C. S: 15301(b)) and HAVA S: 102(a)(2) (codified at 42 U.S.C. S:
15302(a)(2)) govern the use of HAVA Title I funds.

Page 307 GAO-06-450 Elections

(GSA) reported distributing about $300 million to 30 states that applied
for these funds. Figure 62 depicts an overview of the funds distributed to
states specifically to replace lever machines and punch card voting
equipment. (Fig. 57 presented an overview of states' progress in replacing
lever and punch card voting equipment.) In responding to our state survey,
24 of the 30 states reported that they had invested at least a portion of
these funds to replace lever or punch card voting equipment as of August
1, 2005.

Figure 62: Federal Payments to States in Fiscal Year 2003 under HAVA to
Replace Lever and Punch Card Voting Equipment

Sources: GAO analysis of EAC data (analysis), MapArt (map).

Note: These figures represent lower limits on payments for replacement of
lever and punch card voting equipment as made under HAVA Section 102.
States may use additional HAVA funding obtained under HAVA Section 101 for
voting equipment replacement or upgrade. However, these uses have not been
separately tracked or reported by EAC or GSA to date.

In addition to the funding that HAVA earmarked for voting equipment
replacement, states could also apply for other HAVA funds that could be
used for multiple purposes, including replacement or upgrade of voting
systems (Section 101 funds).11 In its 2004 annual report, EAC reported
that almost $344 million had been distributed to each of the 50 states and
the District of Columbia under this multiple purpose funding category. In
all,  44 states and the District of Columbia reported in our state survey
that they had spent or obligated funds from one or both of these HAVA
funding sources in order to improve, acquire, lease, modify, or replace
voting systems and related technology. EAC requires states to submit
detailed annual reports on the use of those funds but has not yet compiled
data from the state reports about spending for voting equipment covered in
HAVA Section 101.

Besides authorizing funding for changes to voting methods, HAVA also has
the potential to influence voting methods through new requirements for the
usability and accountability of voting systems. Among other things, HAVA
requires that voting systems used in federal elections provide voters with
ballot verification and correction capabilities by January 1, 2006,
including

     o the opportunity to verify their ballots in a private and independent
       manner before they are cast;
     o the ability to change their ballots or correct any error in a private
       and independent manner before the ballots are cast and counted; and
     o the capability to both notify the voter whenever more than one
       candidate has been selected for a single office and correct the
       ballots.12

HAVA also requires voting equipment to generate a permanent paper record
with manual audit capacity as an official record of the election.13

Our October 2001 report on election processes described how voting methods
varied in their ability to support features such as error 

11HAVA S: 101(b)(1)(F) (codified at 42 U.S.C. S: 15301).

12HAVA S: 301(a)(1) (codified at 42 U.S.C. S: 15481(a)(1)).

13HAVA S: 301(a)(2) (codified at 42 U.S.C. S: 15481(a)(2)). HAVA
requirements for maximum error rates in counting ballots are discussed in
the section on voting system performance, which follows.

Page 309 GAO-06-450 Elections

identification and correction for voters.14 With regard to minimizing
voter error at the polls, our local jurisdiction survey for the 2004
general election found that, for instance, voters were provided the
opportunity to correct a ballot or exchange a spoiled ballot for a new one
in most jurisdictions, and such capabilities were largely available for
all voting methods. Our estimates of the availability of ballot correction
capabilities range from  100 percent15 (for jurisdictions whose
predominant voting method was central count punch cards) to 70 percent
(for jurisdictions predominantly using DREs).16 However, the differences
among these voting methods were not statistically significant. Figure 63
shows one approach that allows voters to verify and correct their ballots
using a particular voting method (DRE).

              Figure 63: Example of DRE Instructions and Equipment

Source: GAO.

DRE voting instructions (left), DRE voting unit (right).

14GAO-02-3.

15The 95 percent confidence interval for this value is between 100 percent
and 89 percent. 16The 95 percent confidence interval for DREs is +13 or
-15 percentage points.

Page 310 GAO-06-450 Elections

Voting Equipment Performance Was Not Consistently Measured during the 2004
General Election, but Estimated Levels of Performance Satisfaction Were
High

With regard to voting equipment that generated a permanent paper record
with a manual audit capability for election audits in the 2004 general
election (including solutions such as VVPT), we estimate that few
jurisdictions that used DREs had this capability.17 Specifically, from our
local jurisdiction survey, a small proportion of jurisdictions that used
DREs for the 2004 election had manual audit capabilities such as VVPT
(estimated at 8 percent of DRE jurisdictions)18 or printing of ballot
images (11 percent of DRE jurisdictions).19 An estimated 52 percent of
jurisdictions using DREs had equipment that produced an internal paper
record that was not voter-verifiable.20 With this limited implementation
of HAVA-related capabilities in the 2004 general election, it appears that
most of the voting system and election process changes to comply with
these specific HAVA usability and accountability requirements will need to
be satisfied by jurisdictions for the 2006 general election.

Voting system performance can be viewed in terms of accuracy, reliability,
and efficiency. Accuracy refers to how frequently the equipment completely
and correctly records and counts votes; reliability refers to a system's
ability to perform as intended, regardless of circumstances; and
efficiency refers to how quickly a given vote can be cast and counted.
Performance in each of these areas depends not only on how well a given
voting system was designed and developed, but also on the procedures
governing its operation and maintenance and the people who use and operate
it. Thus, it is important that system performance be measured during an
election when the system is being used and operated according to defined
procedures by voters and election workers. As we have previously reported
in our October 2001 report on election processes,

17Jurisdictions that used optical scan, punch card, and paper ballot
voting methods were not included in our survey responses on manual audit
capabilities because they already provide a key component for manual
audits-a paper record created by the voter. Jurisdictions that used lever
machines were not included in survey responses because the machines are no
longer manufactured.

18The 95 percent confidence interval for VVPT in DRE jurisdictions is +9
or -6 percentage points.

19The 95 percent confidence interval for printing of ballot images in DRE
jurisdictions is +6 or -5 percentage points.

20The 95 percent confidence interval for internal paper records that were
not voter-verifiable in DRE jurisdictions is +/- 13 percentage points.

HAVA Has Increased the Focus on Voting System Performance

measuring how well voting systems perform during a given election allows
local election officials to better position themselves for ensuring that
elections are conducted effectively and efficiently.21 Such measurement
also provides the basis for knowing where performance needs, requirements,
and expectations are not being met so that timely corrective action can be
taken.

HAVA recognized the importance of voting system performance by specifying
requirements for error rates in voting systems and providing for updates
to the federal voting system standards, including the performance
components of those standards. Moreover, according to our local
jurisdiction survey, most local jurisdictions adopted performance
standards for the 2004 general election-usually standards selected by
their respective states.22 As was the case for the 2000 general election,
jurisdictions collected various types of voting system performance
measures for the 2004 general election, although some types of measures
were collected by fewer jurisdictions than others-in part because they
were not well suited to particular voting methods. Furthermore, from our
local jurisdiction survey, we estimate that the vast majority of all
jurisdictions were very satisfied or satisfied with their systems'
performance during the 2004 general election, even though performance data
may not have been collected to an extent that would provide firm support
for these views.

In our October 2001 report on voting equipment standards, we reported that
the national voluntary voting system standards being used by some states
and local jurisdictions at that time were originally approved in 199023
and were thus out of date.24 Among other things, these standards
identified minimum functional and performance thresholds for voting
systems in terms of accuracy, reliability, and efficiency. In 2002, the
Federal Election Commission updated these standards and, in doing so,
provided new or

21GAO-02-3.

22Jurisdictions that used only hand-counted paper ballots on Election Day
were excluded from this survey question.

23Federal Election Commission, Performance and Test Standards for
Punchcard, Marksense, and Direct Recording Electronic Voting Systems, Jan.
1990.

24GAO, Elections: Status and Use of Federal Voting Equipment Standards,
GAO-02-52 (Washington, D.C.: Oct.15, 2001).

enhanced coverage of certain performance requirements for, among other
things,25

     o voting system components that define, develop, and maintain election
       databases; perform election definition and setup functions; format
       ballots; count votes; consolidate and report results; and maintain
       records to support vote recounts;
     o direct feedback to the voter that indicates when an undervote or
       overvote is detected in DRE and paper-based voting systems that
       encompass punch cards and optical scan;
     o system standards to meet the needs of voters with disabilities,
       including specific standards for DREs; and
     o strengthened election record requirements to address a range of
       election management functions, including such functions such as ballot
       definition and election programming.

HAVA further focused attention on voting system performance by
establishing a performance requirement for systems used in elections for
federal offices and by providing for updates to federal voting system
standards. Specifically, HAVA required that voting systems used in federal
elections comply with error rate standards specified in the 2002 federal
voting system standards.26 Under these standards, the maximum acceptable
error rate during testing is 1 in 500,000 ballot positions.27 In addition,
HAVA directed EAC to revise the voluntary national voting system
standards, and to test, certify, decertify, and recertify voting system
hardware and software with respect to national voting system standards
using accredited testing laboratories.

25Federal Election Commission, Voting System Performance and Test
Standards, 2002.

26HAVA S: 301(a)(5) (codified at 42 U.S.C. S: 15481(a)(5)).

27The law did not specify whether the accuracy standards are to be
measured in a test environment or an operational environment. However, the
standard itself specifies a test environment.

Page 313 GAO-06-450 Elections

Most Jurisdictions Adopted Voting System Performance Standards, but
Collection of Performance Measures Varied by Jurisdiction Size and Voting
Method

On the basis of our local jurisdiction survey, we estimate that the vast
majority of jurisdictions that used some type of automated voting
equipment on Election Day generally established written standards for the
performance of their voting equipment for the November 2004 general
election.28 Of these, most jurisdictions (an estimated 77 percent) had
adopted their state's standards or requirements pertaining to voting
system performance, although a few had adopted performance standards from
a source other than their state (10 percent) or developed their own  (8
percent).29 The apparently high adoption rate for standards among states
and local jurisdictions is important because it indicates broad acceptance
of a basic management tool needed for systematic performance measurement
and evaluation.

Consistent with our results on voting system performance measurement from
our October 2001 report on election processes,30 estimates from our local
jurisdiction survey indicated that jurisdictions used several specific
measures that could be generally grouped into the areas of accuracy,
reliability, and efficiency to assess the performance of their voting
systems for the 2004 general election. However, jurisdictions measured how
well their systems actually performed in the 2004 election to varying
degrees. In the discussion below, we compare jurisdictions' collection of
selected information on voting system performance for the 2000 and 2004
general elections, and then examine jurisdictions' performance monitoring
in each of the three performance areas.

On the basis of on our local jurisdiction surveys for the 2000 and 2004
elections, we estimate that about 50 percent of jurisdictions collected
performance information in both elections using three measures- accuracy,
undervotes, and overvotes. The percentage of jurisdictions that collected
information on a fourth performance measure-average time to vote-was much
smaller (estimated at 10 percent or less). The differences  

28Jurisdictions that used only hand-counted paper ballots on Election Day
were excluded from this survey question.

29Fourteen percent of the jurisdictions that adopted standards for the
2004 election did not know the source of the performance standards that
they had adopted. Responses for these categories do not add to 100 percent
because respondents were allowed to select multiple items for this
question.

30GAO-02-3.

between estimates for the two elections are not statistically
significant.31 Figure 64 shows the percentages of jurisdictions that
collected these performance measures for the 2000 and 2004 general
elections.

31Differences in overall estimates from the 2001 and 2005 GAO surveys of
local election jurisdictions are, in part, likely due to differences in
the sample designs of the two surveys and how local election jurisdictions
that were minor civil divisions were selected. Because of these sample
design differences, comparing only election jurisdictions that are
counties provides a stronger basis for making direct comparisons between
the two surveys' results. The estimates compared in this report are of the
county local election jurisdictions only; for this reason, some estimates
from the 2001 survey are slightly different than the overall sample
estimates provided in our prior report (GAO-02-3). For these comparisons,
the 95 percent confidence interval is +/- 5 percentage points or less for
the 2001 survey estimates and +/- 8 percentage points or less for the 2005
survey estimates.

Figure 64: Estimated Percentages of Jurisdictions That Collected Voting
System Performance Information for the 2000 and 2004 General Elections

Percentage of jurisdictions 70

60 57 55 54

                                       52

50

50

46

40

30

20

10

10

5

0 Accuracy Undervotes Overvotes Average time to vote

Performance measure

2000

2004 Sources: 2000 data: GAO 2001 survey of local election jurisdictions;
2004 data: GAO 2005 survey of local election jurisdictions.

Note: These estimates include only county election jurisdiction subgroup
comparisons between the 2001 and 2005 surveys. See appendix V for further
details about sampling differences between the surveys.

Accuracy In the area of accuracy, we estimate that 42 percent of
jurisdictions overall monitored the accuracy of voting equipment in the
2004 general election.32 Other widely used measures of accuracy in the
2004 general election were spoiled ballots (estimated at 50 percent of
jurisdictions),33 undervotes  (50 percent of jurisdictions), and
overvotes (49 percent of jurisdictions).

32An estimated 38 percent of respondents selected "not applicable" to the
question on accuracy of voting equipment in their survey response.

33An estimated 25 percent of respondents selected "not applicable" to the
question on spoiled/ruined ballots in their survey response.

Page 316 GAO-06-450 Elections

During our visits to local jurisdictions, election officials in several
jurisdictions told us that measuring overvotes was not a relevant
performance indicator for jurisdictions using lever machines and DREs
because neither permits overvoting. Election officials in several local
jurisdictions we visited also told us that undervotes were not a
meaningful metric because most voters focused on a limited range of issues
or candidates and thus frequently chose not to vote on all contests.

Jurisdictions' collection of the accuracy measures we studied for the 2004
general election varied according to jurisdiction size, with small
jurisdictions generally less likely to collect these measures than other
jurisdiction sizes. Both large jurisdictions (an estimated 66 percent) and
medium jurisdictions (54 percent) were significantly more likely than
small jurisdictions (31 percent) to collect data on vote count accuracy.
In addition, large jurisdictions (65 percent) were significantly more
likely than small jurisdictions (47 percent) to collect data on
undervotes. (See fig. 65.) This disparity may be due to the proportion of
smaller jurisdictions that use paper ballots and for whom collection of
these data would be a manual, time-consuming process.

Figure 65: Estimated Percentages of Jurisdictions That Collected
Information on Voting Accuracy for the 2004 General Election, by
Jurisdiction Size

Percentage of jurisdictions 70

                                     66 65

60

58

56 56

55

54

53 49

50

47

46

40

31

30 20 10

0 Accuracya Spoiled ballotsb Overvotesb Undervotesc Accuracy measures
collected

Small (<10,000) Medium (10,000-100,000) Large (>100,000) Source: GAO 2005
survey of local election jurisdictions. aThe differences between small
jurisdictions and both medium and large jurisdictions are statistically
significant. bThe 95 percent confidence interval for small jurisdictions
in these categories is +/- 6 percentage points. cThe difference between
small jurisdictions and large jurisdictions is statistically significant.
The 95 percent confidence interval for small jurisdictions is +/- 6
percentage points.

Reliability In the area of reliability, we estimate that 15 percent of
jurisdictions measured the reliability of their voting equipment in terms
of pieces of equipment that failed, and 11 percent measured equipment
downtime.34 

34An estimated 66 percent of respondents selected the response "not
applicable" for the survey questions on measurement of pieces of equipment
that failed and equipment downtime.

Page 318 GAO-06-450 Elections

As with accuracy, a higher percentage of large and medium jurisdictions
collected such reliability data than small jurisdictions, and in the case
of equipment failures, there were statistically significant differences in
the collection of this information among different sizes of jurisdictions.
 (See fig. 66.) Importantly, an estimated 55 percent of all jurisdictions
kept a written record of issues and problems that occurred on Election
Day, which could be a potential source of reliability data.

Figure 66: Estimated Percentages of Jurisdictions That Collected
Information on Voting Equipment Reliability for the 2004 General Election,
by Jurisdiction Size

Percentage of jurisdictions 70

59

60 50 40

30 30 28

23

20

10 8 5

0 Equipment Equipment failuresa downtime Reliability measures collected

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000) Source: GAO 2005 survey of local election jurisdictions.

aThe differences between all categories of jurisdiction size are
statistically significant.

Collection of reliability data for automated voting equipment was also
related to the predominant voting method used by a jurisdiction, with
jurisdictions that predominantly used DREs more likely to collect
reliability data than those that used optical scan voting methods.  An
estimated 45 percent of jurisdictions whose predominant method was DREs
collected information on the number of pieces of voting equipment that
failed. The next most frequently collected information on machine failures
was for precinct count optical scan systems (an estimated  23 percent of
jurisdictions) and central count optical scan systems  (an estimated 10
percent). The differences in data collection on equipment failures among
jurisdictions that predominantly used DREs and those that used precinct
count optical scan or central count optical scan voting methods are
statistically significant. (See fig. 67.)

Figure 67: Estimated Percentages of Jurisdictions That Collected
Information on Voting Equipment Failures for the 2004 General Election, by
Predominant Voting Method

Percentage of jurisdictions 50

45

45

40 35 30 25

23

20 15

10

10 5 0

                                       a

                                       nt

                                       b

                                     count

                                       c

                                      DREn

                                       n

                                 Central cousca

                                      sca

                                Precinctoptical

                                    optical

Predominant voting methods

Source: GAO 2005 survey of local election jurisdictions.

Note: The differences between DRE and both central count and precinct
count optical scan voting methods are statistically significant.

aThe 95 percent confidence interval for DRE is +/- 13 percentage points.

bThe 95 percent confidence interval for central count optical scan
percentages is +7 or -5 percentage points.

cThe 95 percent confidence interval for precinct count optical scan
percentages is +8 or -7 percentage points.

Efficiency In the area of efficiency, we estimate that 13 percent of
jurisdictions measured their voting system's speed of counting votes, 17
percent measured the time it took for election workers to set up
equipment, and  4 percent measured the average length of time it took for
voters to cast ballots on Election Day. Large jurisdictions (34 percent)
were significantly more likely than were both medium jurisdictions (19
percent) and small jurisdictions (9 percent) to collect information on
counting speed. There were no significant differences for other efficiency
measures by jurisdiction size. (See fig. 68.)

Figure 68: Estimated Percentages of Jurisdictions That Collected
Information on Voting System Efficiency for the 2004 General Election, by
Jurisdiction Size

Percentage of jurisdictions

Equipment Election Day Counting setup time vote time speeda

Efficiency measures collected

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000) Source: GAO 2005 survey of local election jurisdictions.

aThe differences between all categories of jurisdiction size are
statistically significant.

It is worth noting that for several types of performance measures in our
local jurisdiction survey, jurisdiction size was a factor in whether
system performance information was collected. Generally, large
jurisdictions were most likely to record voting system performance and
small jurisdictions were least likely, with medium jurisdictions in
between. Moreover, large jurisdictions were more likely to keep a written
record of issues or problems that occurred on Election Day. Specifically,
on the basis of our local jurisdiction survey, we estimate that 79 percent
of large jurisdictions kept such records, compared with 59 percent of
medium jurisdictions and 52 percent of small jurisdictions.35 The
differences between large jurisdictions and both medium and small
jurisdictions are statistically significant. The responsibilities for
monitoring or reporting voting system performance most often rested with
local jurisdictions. On the basis of our local jurisdiction survey, we
estimate that 83 percent of local jurisdictions had local officials
responsible for performance monitoring or reporting, while states or other
organizations (such as independent consultants or vendors) held such
responsibilities in 11 percent and 13 percent of jurisdictions,
respectively.

Information obtained during our visits to local election jurisdictions was
generally consistent with the above estimates from our local jurisdiction
survey. For example, election officials in the 28 jurisdictions we visited
most frequently cited number of undervotes (14 jurisdictions), overvotes
(10 jurisdictions), and equipment failures (10 jurisdictions) as types of
performance metrics collected. Another collected metric (cited by election
officials in 6 jurisdictions we visited) was equipment speed, measured in
terms of how fast the voting equipment downloaded vote totals or
transmitted totals to its central count location, and the time required to
cast a vote (reported by election officials in 4 jurisdictions, although
officials in 2 of these 4 jurisdictions limited their measurements to
early voting). Another measurement that election officials in some
jurisdictions told us they collected was comments from poll workers and
voters on the efficiency of the equipment. For instance, an election
official in a large jurisdiction in Georgia told us that poll workers
commented that it took  20 minutes to vote using the voting equipment's
audio feature. In addition, election officials in several jurisdictions
that we visited told us that they had established performance management
programs for their voting systems. For example, election officials in 1
jurisdiction reported that they collected data on the time it took to vote
to better allocate its voting equipment to various locations. Officials in
a large jurisdiction in Kansas said they had conducted a survey of voters
concerning their satisfaction with the ease of use of voting equipment
during the 2004 general election and determined that they were very
satisfied.

35The 95 percent confidence interval for small jurisdictions is +/- 6
percentage points.

Local Jurisdictions Were Generally Satisfied with Their Voting Systems in
2004, although Some Problems Were Reported

In our October 2001 report on election processes, we reported that  96
percent of local jurisdictions nationwide were satisfied with the
performance of the voting equipment during the November 2000 general
election. On the basis of our local jurisdiction survey for the 2004
general election, we estimate that election officials were generally
satisfied with their voting system performance. Estimated satisfaction
varied for specific areas of voting system performance, ranging from
relatively high levels for accuracy (78 percent), speed of vote counting
(73 percent), time to set up equipment (63 percent), and number of spoiled
or ruined ballots  (61 percent), to relatively low levels for equipment
failures (37 percent), and downtime (36 percent).36 Some of these measures
may not be applicable to all jurisdictions, such as those using only
hand-counted paper ballots. When jurisdictions that used only hand-counted
paper ballots were excluded from our results, satisfaction levels were
higher in all performance areas-accuracy (86 percent), speed of vote
counting  (83 percent), time to set up equipment (76 percent), number of
spoiled ballots (68 percent), equipment failures (54 percent), and
downtime  (52 percent). However, even with the exclusion of paper ballot
jurisdictions, "not applicable" responses were often selected by
jurisdictions in the areas of equipment failures (41 percent not
applicable) and downtime (43 percent not applicable).

Also on the basis of our local jurisdiction survey, for five of six
satisfaction measures, we estimate that medium and large jurisdictions
were satisfied or very satisfied with their voting systems more frequently
than small jurisdictions and that most of these differences are
statistically significant. These ratings may be related to the widespread
use of paper ballots by small jurisdictions, where this voting method was
predominant in an estimated 41 percent of jurisdictions. Figure 69 shows
the frequency of satisfaction in each of six performance areas for large,
medium, and small jurisdictions.

36Percentages represent the combination of "satisfied" or "very satisfied"
responses for each survey question item.

Page 324 GAO-06-450 Elections

Figure 69: Estimated Percentages of Jurisdictions Satisfied with Voting
System Performance for the 2004 General Election, by Jurisdiction Size

Percentage of jurisdictions

100 96 95 91

87

90

86 86 81

79 78 79

80

                                       70

70

67

70

64

62

60

60

50 47 42

40

30

20

10

0 Equipment Equipment Speed of Accuracy of Equipment Spoiled or failuresa
downtimeb counting votesc counting votesd setup timee ruined ballotsf
Performance area

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000) Source: GAO 2005 survey of local election jurisdictions.

Note: Satisfaction combines percentages with responses for "satisfied" or
"very satisfied." Jurisdictions

that used only hand-counted paper ballots on Election Day were excluded
from this figure. aThe differences between both small and medium
jurisdictions and large jurisdictions are statistically significant.

bThe differences between all categories of jurisdiction size are
statistically significant. cThe difference between small jurisdictions and
medium jurisdictions is statistically significant. dThe differences
between small jurisdictions and both medium and large jurisdictions are
statistically

significant.

eThe differences between small jurisdictions and both medium and large
jurisdictions are statistically significant. The 95 percent confidence
interval for small jurisdictions is +/- 7 percentage points. fThe 95
percent confidence interval for small jurisdictions is +/- 7 percentage
points.

The estimated high satisfaction levels demonstrated across different
voting system performance areas and jurisdiction sizes contrast with our
lower estimates of the performance measures that were collected for the
2004 general election. Although the reasons for moderate collection levels
for performance measures are unclear, jurisdictions that may not have
collected performance data or may have considered such information not
applicable to their situation may lack sufficient insight into their
system operations to adequately support their satisfaction in the variety
of performance areas we surveyed.

Local election officials at most of the 28 jurisdictions we visited also
expressed satisfaction with the performance of their voting systems or
method. For example,

     o Election officials in several jurisdictions using optical scan systems
       stated that they were pleased with their equipment because it produced
       a paper trail and permitted fast processing. Officials in  1 large
       jurisdiction in Florida added that their use of the same equipment
       over several elections made it easy for voters to use the equipment in
       both 2000 and 2004.
     o Election officials in several other jurisdictions using DREs told us
       that their equipment was easy to use and provided results that were
       accurate and timely. Officials in 1 large jurisdiction in New Jersey
       reported that, in contrast to paper ballots, DREs do not require poll
       workers to interpret a voter's ballot.
     o Election officials in a large Connecticut jurisdiction using lever
       machines said that voters were happy with the equipment and that it
       had worked well for over 60 years. They emphasized that the simplicity
       and transparency of the equipment's counting mechanisms gave voters
       confidence that their votes would be counted correctly.
     o Election officials in a small New Hampshire jurisdiction using paper
       ballots reported that they had used the same hand-counted paper ballot
       system for decades and it has been very cost-effective for the small
       population of voters in the jurisdiction.

Overall, election officials in few of the 28 jurisdictions that we visited
reported substantive performance issues, such as overvoting, undervoting,
or equipment failure.

Although the estimated level of satisfaction with voting equipment
performance in the 2004 general election was high overall, some
dissatisfaction existed. On the basis of our local jurisdiction survey, we
estimate that between 1 and 4 percent of jurisdictions were dissatisfied
or very dissatisfied with their voting systems in the 2004 general
election for the six performance areas of our survey.

Our local jurisdiction survey provided additional insight into the role of
voting equipment in jurisdictions' dissatisfaction ratings. Of almost 
300 responses to our open-ended question about the issue or problem that
occurred most frequently on Election Day, November 2004, fewer than  20
responses were specifically related to voting equipment. The most frequent
reason for voting system dissatisfaction was voting equipment malfunction.
Ballot errors related to voting equipment were much less frequently
mentioned.

Although such problems were rarely mentioned by election officials during
our visits to local jurisdictions, some did describe a few reasons for
dissatisfaction with voting equipment, including

     o the additional time required to count ballots using DREs versus the
       optical scan equipment previously used,
     o the perceived lower reliability and greater failure rates of DREs over
       the voting equipment used in the past,
     o accuracy problems with DRE computer programs, and
     o difficulty in first-time poll worker operation and voter use of DREs.

Election officials in a few jurisdictions we visited noted situations that
required considerable effort to resolve. For example, as mentioned in our
discussion of vote counting in chapter 6, election officials in a North
Carolina jurisdiction told us that 4,235 ballots were lost by one of the
DREs used for early voting because the software manufacturer had not
installed an upgrade that would have allowed the machine to record up to
10,000 ballots rather than its original limit of 3,500 ballots. The
machine continued to show the number of people who voted on the machine
after 3,500 ballots had been cast, but did not store the results of their
ballots. As a result, the jurisdiction switched to hand-counted paper
ballots for elections after the 2004 general election until its state can
approve a new automated system for use. Given the real and potential
impacts of situations where dissatisfaction was reported, systematic
collection and analysis of performance information may help provide
election officials with objective support for decisions to improve the
operation and upgrade of these systems.

Attention to Voting System Security Management Continues to Vary amid
Published Concerns and Federal Improvement Efforts

Having secure voting systems is essential to maintaining public confidence
in the election process, and accomplishing this is a shared responsibility
among federal, state, and local jurisdiction authorities. Among other
things, voting system security involves ensuring that technical security
controls embedded in voting equipment operate as intended, as well as
ensuring that security policies and procedures governing the testing,
operation, and use of the systems are properly defined and implemented by
state and local election officials.

Our October 2001 report on election processes identified voting system
security challenges facing local jurisdictions, such as consistent
application of controls and adequacy of resources.37 HAVA recognized some
of these challenges by requiring specific system security controls and
providing improved security management guidance. Nevertheless, while we
estimate from our local survey that most jurisdictions have assigned
responsibility for voting system security to individuals and implemented
certain security controls, the nature and extent of their respective
security efforts and activities varied widely. In particular, according to
our state survey, estimates from our local jurisdiction survey, and visits
to jurisdictions, there are differences across jurisdictions in the (1)
adoption of system security standards, with some states requiring
jurisdictions to use outdated standards for voting systems; (2) reported
implementation of system security controls; and (3) testing performed to
ensure that security controls are functioning properly. For instance, we
estimate on the basis of our local jurisdiction survey that at least 19
percent of local jurisdictions nationwide (excluding jurisdictions that
reported using paper ballots) did not conduct security testing for the
systems they used in the November 2004 general election. In addition, 27
states reported in our state survey that they are requiring jurisdictions
to apply federal standards to voting systems used for the first time in
the November 2006 general election that are outdated, unspecified, or
entail multiple versions. This variability in implementation and testing
of controls is generally consistent with what we reported for the 2000
general election. Moreover, our September 2005 report on the security and
reliability of electronic voting highlighted substantial security issues
and concerns for more modern electronic voting systems and reinforced the
importance of effective security management.38

37GAO-02-3. 38GAO-05-956.

Page 328 GAO-06-450 Elections

               HAVA Has Increased Focus on Voting System Security

HAVA recognized the importance of effective voting system security through
two primary mechanisms. First, it required voting systems to produce a
permanent paper record that provides a manual review capability and
constitutes the official record for recounts by January 1, 2006.39 The
paper record can be compared with polling place records and voting system
documentation to ensure that authorized ballots have been completely and
accurately counted. Second, HAVA provided various means to assist states
and localities in acquiring and operating secure voting systems. These
include provisions for EAC to (1) update voting system standards for
voting systems, including standards for security;40 

(2)
           establish processes for accrediting voting system testing
           laboratories and conducting tests of voting systems against the
           standards;41 and 

(3)
           create a process for federal certification of voting systems that
           undergo the testing process.42 In doing so, HAVA created tools and
           resources that states and local jurisdictions can leverage when,
           for example, acquiring systems from vendors, conducting system
           testing, and operating and auditing voting systems.

However, delays in establishing EAC and commission funding challenges
resulted in the first update to the 2002 voluntary voting system
standards, and its provisions for system security, not being approved
until  December 2005.43 Further, commission efforts to establish
processes for accrediting testing laboratories, conducting testing, and
certifying systems are still under way.

39HAVA S: 301(a)(2)(B) (codified at 42 U.S.C. S: 15481(a)(2)(B)).

40HAVA S: 221 (codified at 42 U.S.C. S: 15361).

41HAVA S: 231 (codified at 42 U.S.C. S: 15371).

42HAVA S: 231 (codified at 42 U.S.C. S: 15371).

43EAC was to be appointed by March 2003, but the appointment of EAC
commissioners did not occur until December 13, 2003. The $1.2 million in
funding received for fiscal year 2004 supported limited activities.
Significant startup funding was not received until fiscal year 2005, when
its budget was increased to $13.8 million. The largest portion of EAC's
2005 budget (31 percent) was allocated to improving voting technology,
including $2.8 million for National Institute of Standards and Technology
support to develop the Voluntary Voting System Guidelines.

Efforts to Address System Security Continued to Vary Widely, and
Jurisdictions Were Inconsistent in Following Common Security Practices

As was the case for the November 2000 general election, the nature and
extent of voting system security efforts and activities during the 2004
election varied among jurisdictions. Moreover, these efforts and
activities do not in all cases reflect the use of recommended system
security management practices and current voting system security
standards.

In our October 2001 report on election processes, we reported that
jurisdictions had taken a number of steps to manage the security of their
respective voting systems for the 2000 general election.44 In particular,
we estimated that 89 percent of the local jurisdictions assigned
responsibility for performing security-related functions to one or more
individuals, and implemented some type of controls to protect their
equipment during the election. Examples of implemented security controls
included such physical controls as locks and surveillance, and such
embedded controls as access restrictions and firewalls. However, we also
reported in 2001 that an estimated 40 percent of the jurisdictions had not
assessed the security threats and risks on which their controls were
based,45 and 19 percent had not reviewed the sufficiency of their security
controls. Moreover, the nature of established controls varied by type of
system, and these controls were not uniformly followed across
jurisdictions.46

For the November 2004 general election, jurisdictions addressed system
security to varying degrees and through various means. At the foundation
of these approaches, responsibilities for voting system and network
security were distributed among local officials, the state, and third
parties (e.g., independent consultants and vendors) in varying
proportions. On the basis of our 2005 local jurisdiction survey, we
estimate that 90 percent of all jurisdictions (excluding those that used
only hand-counted paper ballots on Election Day) specifically assigned
responsibility for voting system 

44GAO-02-3.

45Threats and risks to voting systems include loss of electronic voting
data, loss or theft of ballots, and unauthorized access to software.

46Common security controls included (1) identification names and passwords
to control access to voting equipment and software, (2) redundant storage
media for recovery in the event of power or equipment failure, (3)
encryption to ensure privacy of votes and confidentiality of election
results, (4) audit trails to document the integrity of the voting process,
and (5) hardware locks and seals to prevent unauthorized access to voting
equipment components.

security in the 2004 general election.47 We estimate that 67 percent of
these local jurisdictions assigned responsibilities for voting system and
network security to local election officials, 14 percent relied on state
officials to perform these responsibilities, and 24 percent assigned them
to third parties.48 Moreover, this distribution varied somewhat according
to jurisdiction size, with large jurisdictions depending on local
officials the most and medium jurisdictions depending on local officials
the least. Figure 70 shows how voting system and network security
responsibilities were distributed among various parties for each size of
jurisdiction.

47Jurisdictions that used only hand-counted paper ballots on Election Day
were excluded from this survey question.

48Jurisdictions may have identified more than one group with assigned
security responsibilities.

Page 331 GAO-06-450 Elections

Figure 70: Estimated Percentages of Jurisdictions That Identified Security
Responsibilities for the 2004 General Election, by Jurisdiction Size

Percentage of jurisdictions 90

80

80

70

70 60

57

50 40 30

27 23

19

20

14 15

12

10

0 Local election State Othersb officialsa

Responsibility for voting system security

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000) Source: GAO 2005 survey of local election jurisdictions.

Note: Jurisdictions may have identified more than one group with assigned
security responsibilities. Jurisdictions that used only hand-counted paper
ballots on Election Day were excluded from this survey question.

aThe difference between medium jurisdictions and large jurisdictions is
statistically significant. The  95 percent confidence interval for small
jurisdictions is +/- 7 percentage points.

bOthers include third parties, such as independent consultants and
vendors.

On the basis of our visits to local jurisdictions, the types of system
security responsibilities and the groups that performed them further
demonstrate the variation among security approaches and controls applied
to voting systems. Specifically, election officials in these jurisdictions
were typically responsible for implementing security controls, state
officials were usually involved with developing security policy and
guidance and monitoring local jurisdictions' implementation of security,
and third parties performed tasks such as ensuring adequate security of
voting equipment during transport or storage. Table 24 shows examples of
security tasks and the parties that performed them as reported to us by
election officials in the jurisdictions that we visited.

Table 24: Voting System Security Tasks and Responsibilities for the 2004
General Election Reported by Election Officials in Jurisdictions Visited
by GAO

Examples of voting system security                  Performing entity
tasks                                               
identified by local officials       Local officials State  Third parties   
Secure ballot programming                  X              
Sealing of voted ballots                   X              
Secure storage of voting equipment         X                     X         
                                                             (e.g., schools)  
Video surveillance of stored               X              
equipment or ballots                                      
Access control to stored election          X              
materials                                                 
Protection of voting equipment and         X                     X         
materials during                                          
transport                                                    (e.g., law    
                                                                  enforcement 
                                                                   officials) 
Inventory management of voting             X              
equipment and                                             
ballots                                                   
Monitoring vote tallying systems           X              
for unauthorized                                          
connections                                               
Impoundment of election materials          X              
after elections                                           
Monitoring or testing of equipment         X          X          X         
accuracy before,                                          
during, or after elections                                
Security awareness training for            X          X          X         
election personnel                                        
Certification of voting equipment          X          X   
Development of security policies           X          X   
and guidance for                                          
local jurisdiction use                                    
Monitoring implementation of               X          X   
security policies                                         

Source: GAO analysis of documents provided by local jurisdictions we
visited.

Responses to our state survey showed that both states and third parties
participated in security responsibilities related to monitoring and
evaluating security and privacy controls. Although the most frequently
cited party responsible for this area was local officials (identified by
38 states), just less than one-half of the states (22 states and the
District of Columbia) reported that they had some level of responsibility
for security monitoring and evaluation as well. In addition, 22 states
responded that third parties (e.g., independent consultants or vendors)
were involved in monitoring and evaluating controls. Overall, security
monitoring and evaluation was performed by two or more entities in 26 of
the states.

The use of certain security controls was similarly varied. On the basis of
our local jurisdiction survey, we estimate that 59 percent of
jurisdictions used power or battery backup, 67 percent used system access
controls,  91 percent used hardware locks and seals, and 52 percent used
backup electronic storage for votes.49 We further estimate that 95 percent
of jurisdictions used at least one of these controls, with hardware locks
and seals being most consistently used across the automated voting methods
associated with this survey question.50 Furthermore, we estimate that a
lower percentage of small jurisdictions used power or battery backup and
electronic backup storage of votes for their voting equipment than large
or medium jurisdictions, and these differences are statistically
significant in most cases. Figure 71 presents the use of various security
controls by jurisdiction size.

49Jurisdictions that used only hand-counted paper ballots on Election Day
were excluded from this survey question.

50We were unable to reliably estimate percentages for jurisdictions whose
predominant voting methods were central count punch card or precinct count
punch card voting methods for all but one of these security controls. We
estimate that 95 percent of jurisdictions whose predominant voting method
was central count punch card used hardware locks and seals.

Page 334 GAO-06-450 Elections

Figure 71: Estimated Percentages of Jursidictions That Used Security
Controls in the 2004 General Election, by Jurisdiction Size

Percentage of jurisdictions 100

95 91 90

90 80

75

72 70 6867 6666

63

60

53

50

43

40 30 20 10

0 Power or Access controlb Hardware Electronic battery backupa locks and
seals backup storagec Voting system security controls

Small (<10,000) Medium (10,000-100,000) Large (>100,000) Source: GAO 2005
survey of local election jurisdictions. Note: More than one type of
security control may have been identified. aThe difference between small
jurisdictions and medium jurisdictions is statistically significant. The
95 percent confidence interval for small jurisdictions is +/- 8 percentage
points. bThe 95 percent confidence interval for small jurisdictions is +/-
8 percentage points. cThe differences between small jurisdictions and both
medium and large jurisdictions are statistically significant. The 95
percent confidence interval for small jurisdictions is +/- 8 percentage
points.

We estimate that a small percentage of local jurisdictions (10 percent)
provided remote access to their voting systems for one or more categories
of personnel-local election officials, state election officials, vendors,
or other parties.51 Small jurisdictions, in particular, were less likely
to provide remote access to their voting systems (estimated at 7 percent)
than either medium jurisdictions (13 percent) or large jurisdictions (19
percent). The difference between small jurisdictions and large
jurisdictions is statistically significant. For each category of
personnel-local officials, state election officials, vendors, or other
parties-7 to 8 percent of jurisdictions did not know if remote access was
available to their systems, a situation that could increase the risk of
unauthorized access to these systems. Some of the jurisdictions responding
to this survey question described a variety of protections to mitigate the
risk of unauthorized remote access, including locally controlled
passwords, passwords that change for each access, and local control of
communications connections.

Among the jurisdictions that we visited, election officials reported that
various security measures were in use during the 2004 general election to
safeguard voting equipment, ballots, and votes before, during, and after
the election. However, the measures were not uniformly reported by
officials in these jurisdictions, and officials in most jurisdictions
reported that they did not have a security plan to document these measures
or other aspects of their security program. The security controls most
frequently cited by officials for the jurisdictions that we visited were
locked storage of voting equipment and ballots, and monitoring of voting
equipment. Other security measures mentioned during our visits included
testing voting equipment before, during, or after the election to ensure
that the equipment was accurately tallying votes; planning and conducting
training on security issues and procedures for elections personnel; and
video surveillance of stored ballots and voting equipment. Table 25
summarizes the types and frequency of security measures reported by
election officials in the jurisdictions we visited.

51Jurisdictions that used only hand-counted paper ballots on Election Day
were excluded from this survey question.

Page 336 GAO-06-450 Elections

Table 25: Security Controls Reportedly Used in the 2004 General Election
Reported by Election Officials in Jurisdictions Visited by GAO

               Reported security control Number of jurisdictions

Locked/sealed storage of voting equipment and ballotsa

Monitoring of voting equipment a

Encrypted ballots or election results a

Security plansa

Testing of voting equipment a

Control of voting machine memory cards by precinct personnel during
elections a

Video surveillance for voting equipment or ballots

Security training

Source: GAO analysis of interviews from local jurisdictions we visited.

aOne or more jurisdictions we visited indicated this security control was
not applicable because of the voting method used.

Notwithstanding this range of reported security controls that were used in
the 2004 general election by jurisdictions we visited, jurisdictions'
activities and efforts for managing voting system security were not always
in line with recommended system security practices. Our research of
recommended practices shows that effective system security management
involves having, among other things, (1) defined policies governing such
system controls as authorized functions and access, and documented
procedures for secure normal operations and incident management; 

(2)
           documented plans for implementing policies and procedures; 

(3)
           verified implementation of technical and procedural controls
           designed to reduce the risk of disruption, destruction, or
           unauthorized modification of systems and their information; and
           (4) clearly assigned roles and responsibilities for system
           security.

On the basis of our local jurisdiction survey, we estimate that 46 percent
of election jurisdictions nationwide that used some type of automated
voting method had written policies for voting system security and access
in place for the November 2004 general election, while 45 percent had
formal security procedures.52 Written security policies were more
prevalent among large jurisdictions, an estimated 65 percent, compared to
an estimated  52 percent of medium jurisdictions and an estimated 41
percent of small jurisdictions. The difference between large and small
jurisdictions is statistically significant. More large and small
jurisdictions had formal security procedures (an estimated 51 percent and
47 percent, respectively) than medium jurisdictions (an estimated 39
percent), although these differences are not statistically significant.
Figure 72 shows the estimated percentages of jurisdictions with written
security policies and procedures by jurisdiction size.

52Jurisdictions that used only hand-counted paper ballots on Election Day
were excluded from this survey question.

Page 338 GAO-06-450 Elections

Figure 72: Estimated Percentages of Local Jurisdictions That Documented
Security Policies or Procedures for Their Voting Systems in the 2004
General Election, by Jurisdiction Size

Percentage of jurisdictions 70

65

60

52

51

50

47 41

39

40

30

20

10

0 Written policiesa Formal proceduresb Voting system security and access
measures

Small (<10,000)

Medium (10,000-100,000)

Large (>100,000) Source: GAO 2005 survey of local election jurisdictions.

Note: Percentages in each category of jurisdiction size may not add to 100
because of rounding. Jurisdictions that used only hand-counted paper
ballots on Election Day were excluded from this survey question.

aThe difference between small jurisdictions and large jurisdictions is
statistically significant. The 95 percent confidence interval for small
and medium jurisdictions is +/- 8 percentage points.

bThe 95 percent confidence interval for small and large jurisdictions is
+/- 8 percentage points.

In our earlier discussion of local survey responses related to counting
votes in chapter 6, we estimated that many jurisdictions had written
policies and procedures for ballot security in the 2004 general election.
However, we estimate that up to one-fifth of jurisdictions did not have
written policies and procedures uniformly in place, including policies and
procedures for transporting unvoted and voted ballots or electronic
memory, storing unvoted and voted ballots, and electronic transmission of
voted ballots. The disparity in written policies and procedures was
observed for electronic transmission of voted ballots for counting, where
an estimated 18 percent of jurisdictions had such security management
tools,53 compared with between 66 and 76 percent of jurisdictions for each
of the other four types of ballot controls-a difference that is
statistically significant but which may be linked to the percentage of
jurisdictions that used paper ballot and older technologies in the 2004
general election. Yet we also found that an estimated 17 percent of
jurisdictions whose predominant method was DRE had no policies or
procedures for electronic transmission of voted ballots for counting.54 In
addition, the differences in estimates of policies and procedures for
electronic ballot transmission among jurisdictions whose predominant
voting method was punch cards and those whose methods were DRE or optical
scan are statistically significant. Figure 73 shows the variation in
estimates of documented policies and procedures for electronically
transmitting ballots among jurisdictions that used specific voting
methods.

53 An estimated 57 percent of respondents selected "not applicable" for
written policies and procedures for electronic transmission of voted
ballots for counting in their survey response.

54The 95 percent confidence interval for jurisdictions using DREs was +13
or -9 percentage points. An estimated 38 percent of respondents from
jurisdictions whose predominant method was DRE selected "not applicable"
in their survey response to written policies and procedures for electronic
transmission of voted ballots for counting.

Figure 73: Estimated Percentages of Jurisdictions That Established
Policies or Procedures for Electronic Transmission of Voted Ballots in the
2004 General Election, by Predominant Voting Method

Percentage of jurisdictions 45

40 39

35

30

30 25

19

20 15 10

8 5

5

2

                                       1

0

a

nt

                                       b

                                     count

                                       c

                                       nt

                                     count

                                       e

                                       d

                                       d

                                       d

                                       n

DRE

                                       n

                                  balloth car

                                    h carne

Central cou

                                  Central cou

sca

                                      sca

                                       hi

                                    ver mac

                                    Precinct

                                  Precinctpunc

                                      punc

optical

                                  opticalaper

                                       P

                                       Le

Predominant voting methods

Source: GAO 2005 survey of local election jurisdictions.

Note: Estimates of respondents who selected "not applicable" for this
survey question ranged between 38 and 75 percent.

aThe 95 percent confidence interval for DRE is +/- 14 percentage points.

bThe 95 percent confidence interval for central count optical scan is +9
or -8 percentage points.

cThe 95 percent confidence interval for precinct count optical scan is +/-
8 percentage points.

dThe 95 percent confidence interval for lever machine is +11 or -6
percentage points.

eThe 95 percent confidence interval for paper ballot is +6 or -4
percentage points.

Moreover, our visits to local jurisdictions found diverse approaches to
documenting security policies and procedures. Election officials in 8 of
the jurisdictions that we visited told us that they had written
instructions for managing security aspects of their voting equipment and
processes. However, some guidance we reviewed did not cover these topics.
Election officials in some jurisdictions stated that their security
measures were contained in the voting process documentation for the voting
system or were covered in election worker training. For example, the
hardware guide for the voting system used by some jurisdictions described
the verification and authentication functions that were built into the
system to secure vote counts during transmission of the precinct results
to the jurisdiction, including processes for ballot creation and vote
tabulation that also included security procedures. In contrast, several
other jurisdictions that we visited had published detailed security
policies and procedures for their voting systems that included, for
example, network security policies for election tabulation, procedures for
securing and protecting election equipment and software, testing voting
equipment to ensure accurate recording of votes, and disaster recovery
plans, and they provided them to GAO. Officials in several jurisdictions
also described their steps to ensure that election workers had access to,
and were trained in, the contents of the policies and procedures for
securing ballots and voting equipment.

Information system security plans typically identify the responsibilities,
management approach, and key controls to be implemented for an information
system, based on an assessment of identified risks to the information.
Election officials in a few of the jurisdictions that we visited told us
that they had security plans in place for the November 2004 general
election (8 of 28). Officials at 4 of the jurisdictions that we visited
stated that they had security plans or plan components that were approved
at the state level, and officials in 1 large jurisdiction in Nevada
reported having a state statutory requirement for a voting system security
plan. However, jurisdictions that employed advanced security technologies,
such as encryption, in their systems did not always have a plan that would
document how the elections people, process, and technologies would work
together to provide comprehensive protections. Moreover, the contents of
plans we obtained from our visits to local jurisdictions varied widely.
One of the jurisdiction security plans we examined covered most aspects of
the voting process, from ballot preparation through recount, while another
plan focused on the security of its vote-tallying system in a stand-alone
environment. Two security plans covered several security topics including
risk assessment, physical and personnel controls, and incident response.
Table 26 shows the variation in topics covered in the security plans we
reviewed.55

55Election officials in three of the jurisdictions that we visited
reported having security plans but did not provide them for this study.

Page 342 GAO-06-450 Elections

Table 26: Voting System Security Topics Addressed in Security Plans
Submitted by Jurisdictions Visited by GAO

                             Jurisdiction coverage of topic in security plans
Security topic                   1         2         3        4          5 
Risk assessment                            X         X                   X 
Awareness training                         X         X              
Physical controls                          X         X        X     
Personnel controls                         X                        
Access controls                  X         X         X        X     
Security testing or audit                                           
Backup and recovery              X         X         X              
Incident response                          X         X              

Source: GAO analysis of documents provided by local jurisdictions we
visited.

Security testing is an important way to verify that system security
controls have been implemented and are functioning properly. From our
survey of state election officials, 17 states and the District of Columbia
reported that they had conducted security testing of the voting systems
used in the 2004 general election, and 7 other states reported that they
required local jurisdictions to conduct such testing. The remaining 22
states said that they did not conduct or require system security testing.
(Three states reported that security testing was not applicable for their
voting systems.) Moreover, from our local jurisdiction survey, we estimate
that at least 19 percent of local jurisdictions nationwide (excluding
jurisdictions that reported that they used paper ballots) did not conduct
security testing for the systems they used in the November 2004 general
election. Although jurisdiction size was not a factor in whether security
testing was performed, the percentage of jurisdictions performing security
testing was notably higher when the predominant voting method was DRE (63
percent)56 and lower for jurisdictions where the predominant method was
central count optical scan (38 percent)57 or precinct count optical scan
(45 percent).58 However,

56The 95 percent confidence interval for DRE is +14 or -15 percentage
points.

57The 95 percent confidence interval for central count optical scan is +/-
10 percentage points.

58The 95 percent confidence interval for precinct count optical scan is
+/- 9 percentage points.

Page 343 GAO-06-450 Elections

Security Standards Being Used Vary by State and Jurisdiction

the difference in the percentages of jurisdictions performing security
testing on DRE or central count optical scan is not statistically
significant.

Beyond jurisdictions' efforts to verify implementation of voting system
security controls, some states required that their voting systems be
nationally qualified against the federal voluntary voting system
standards, which include a security component. In particular, from our
state survey, most states that used a new voting system for the first time
in the November 2004 general election said that they required the system
to go through qualification testing. For example, all 26 states that used
DREs for the first time in the 2004 general election, as well as the
District of Columbia, required qualification testing and approval by the
National Association of State Election Directors (NASED).59 Similarly, of
the  35 states and the District of Columbia that used optical scan
systems for the first time in the 2004 general election, 31 reported that
they required voting systems to be qualified. Nine of the 10 states that
used new punch card systems for the first time in the 2004 general
election also reported that they required voting systems to be qualified.

States and jurisdictions are applying a variety of security standards to
their voting systems, some of which are no longer current. Specifically,
44 states and the District of Columbia reported on our state survey that
they were requiring local jurisdictions' voting systems being used for the
first time in the November 2006 general election to comply with voluntary
federal voting system standards, which include security standards.
However, they are not all using the same version of the voluntary
standards. This is troublesome because the 2002 standards are more
stringent than the 1990 standards in various areas, including security.
For instance, the 2002 standards establish security requirements and
acceptable levels of performance for the telecommunications components of
voting systems, while the 1990 standards do not include detailed
requirements for this control measure.

According to our analysis of responses states reported in our state
survey, 17 of the 44 states and the District of Columbia reported that
their voting systems must comply solely with the 2002 standards that were
developed

59Prior to the November 2004 general election, NASED conducted the
qualification program to test voting systems against the federal voluntary
voting system standards. In 2005, EAC assumed this responsibility.

Page 344 GAO-06-450 Elections

and approved by the Federal Election Commission and later adopted by EAC.
However, 27 other states are requiring their jurisdictions to apply
federal standards to their new voting systems that are outdated,
unspecified, or entail multiple versions. In the case of 5 of these 27
states where multiple versions of voluntary federal standards will be
applied, one of the versions is the Voluntary Voting System Guidelines,
which was approved by the EAC in December 2005.60 These guidelines promote
security measures that address gaps in prior standards and are applicable
to more modern technologies, such as controls for distributing software
and wireless operations. Nevertheless, these same 5 states reported that
they will also apply older federal standards to systems that are new to
the 2006 election. Furthermore, 2 other states responded that they do not
plan to require their voting systems to comply with any version of the
voluntary federal standards, while 3 additional states reported that they
had not yet made a decision on compliance with voluntary federal standards
for 2006. (One state did not respond.) Figure 74 depicts the number of
states that reported applying voluntary federal voting system standards to
their new voting systems. Appendix X summarizes responses for all states
and the District of Columbia regarding reported requirements for local
jurisdictions' use of federal standards for their voting systems.

60 After the Voluntary Voting System Guidelines become effective in
December 2007, voting systems will no longer be tested by federally
accredited laboratories to prior versions of the federal standards.

Page 345 GAO-06-450 Elections

Figure 7 4 : State-Reported Standar d s Required fo r V oting Systems
Number of states 35 30 29

               2002 FECstandards otingguidelines ederalstandards

                                   standards

                                   guidelines

                            l compliancerequirement

                         Not yetdetermined No response

1990 FECstandards

                                  Draft EAC v

                                  Hybrid of f

                                     ederal

                                   system or

                                     edera

                                     Any f

                                      No f

Applicable standards

Source: GAO 2005 survey of state election officials.

Note: State respondents were allowed to select multiple items for this
survey question.

Simultaneous use of multiple versions of voting system standards is not
new for the 2006 election. Not all NASED-qualified voting systems that may
have operated during the 2004 election were tested against a single
version of security standards. For example, many systems that were
qualified before the 2004 general election had been tested against the
1990 Federal Election Commission standards, rather than the more stringent
2002 standards.

The use of outdated system security standards increases the risk of system
integrity, availability, and confidentiality problems for all voting
methods, but it is of special concern for jurisdictions that use their
systems in a networked environment or transmit election data using
telecommunications capabilities. This is because the use of such
connectivity introduces vulnerabilities and risks that the older versions
of

Recent Studies and Analyses Have Raised Concerns about the Security
Vulnerabilities and Weaknesses of Modern Voting Systems

the standards do not adequately address, as we have previously described
in our September 2005 report on the security and reliability of electronic
voting.61

After the 2000 general election, Congress, the media, and others cited
numerous instances of problems with the election process. As the use of
electronic voting systems expanded and the 2004 general election
approached, the media and others continued to report problems with these
systems that caused some to question whether they were secure and
reliable. To clarify the wide range of concerns and issues raised and
identify recommended practices for addressing them, our September 2005
report on the security and reliability of electronic voting analyzed over
 80 recent and relevant studies related to the security and reliability
of electronic voting systems.62 We focused on systems and components
associated with vote casting and counting, including those that define
electronic ballots, transmit voting results among election locations, and
manage groups of voting machines.

In summary, our September 2005 report stated that while electronic voting
systems hold promise for a more accurate and efficient election process,
numerous organizations and individuals have raised concerns about their
security, citing instances of weak security controls, system design flaws,
inadequate system version control, inadequate security testing, incorrect
system configuration, poor security management, and vague or incomplete
voting system standards, among other issues. For example, we reported that
studies found (1) some electronic voting systems did not encrypt cast
ballots or system records of ballots, and it was possible to alter both
without being detected; (2) it was possible to alter the files that define
how a ballot looks and works so that the votes for one candidate could be
recorded for a different candidate; and (3) vendors installed uncertified
versions of voting system software at the local level. We also reported
that some of these concerns were said to have caused local problems during
national elections-resulting in the loss or miscount of votes.63 We added,
however, that many of the reported concerns were drawn from specific

61GAO-05-956. 62GAO-05-956.

63Several of the problems included in our 2005 report on voting system
security and reliability have been discussed in this report.

Page 347 GAO-06-450 Elections

system makes and models or from a specific jurisdiction's election, and
that there has been a lack of consensus among election officials and other
experts on the pervasiveness of the concerns.

We also reported in September 2005 that federal organizations and
nongovernmental groups have issued recommended practices and guidance for
improving the election process, including electronic voting systems, as
well as general practices for the security of information systems. For
example, in mid-2004, EAC issued a collection of practices recommended by
election experts, including state and local election officials.64 This
guidance includes approaches for making voting processes more secure and
reliable through, for example, risk analysis of the voting process, poll
worker security training, and chain of custody controls for Election Day
operations, along with practices that are specific to ensuring the
security and reliability of different types of electronic voting systems.
As another example, in July 2004, the California Institute of Technology
and the Massachusetts Institute of Technology issued a report containing
recommendations pertaining to testing equipment, retaining records of
ballots, and physically securing voting systems.65 In addition to such
election-specific practices, numerous recommended practices are available
that are relevant to any information system. For instance, we, the
National Institute for Standards and Technology (NIST), and others have
issued guidance that emphasizes the importance of incorporating security
and reliability into the life cycle of information systems through
practices related to security planning and management, risk management,
and procurement.66 We noted that the recommended practices in these
electionspecific and information technology-focused documents provide
valuable guidance that, if implemented effectively, should help improve
the security of voting systems.

64EAC, Best Practices Tool Kit (July 2004),

http://www.eac.gov/bp/docs/BestPracticesToolKit.doc.

65California Institute of Technology/Massachusetts Institute of
Technology, Immediate Steps to Avoid Lost Votes in the 2004 Presidential
Elections: Recommendations for the Election Assistance Commission (July
2004).

66For example, GAO, Federal Information Systems Controls Audit Manual, 
GAO/AIMD-12.19.6 (Washington, D.C.: January 1999); NIST, Generally
Accepted Principles and Practices for Securing Information Technology
Systems, SP 800-14 (September 1996) and Security Considerations in the
Information System Development Life Cycle,  SP 800-64, Revision 1 (June
2004); and International Systems Security Engineering Association, Systems
Security Engineering Capability Maturity Model, ISO/IEC 21827, version 3.0
(June 2003).

Further, our September 2005 report stated that since the passage of HAVA,
the federal government has begun a range of actions that are expected to
improve the security and reliability of electronic voting systems.
Specifically, after beginning operations in January 2004, EAC was leading
efforts to (1) draft changes to the existing federal voluntary standards
for voting systems, including provisions related to security;67 (2)
develop a process for certifying, decertifying, and recertifying voting
systems; 

(3) establish a program to accredit the national independent testing
laboratories that test electronic voting systems against the federal
standards; and (4) develop a software library and clearinghouse for
information on state and local elections and systems. However, we observed
that these actions were unlikely to have a major effect in the 2006
federal election cycle because at the time of our report publication the
changes to the standards had not yet been completed, the system
certification and laboratory accreditation programs were still in
development, and the software library had not been updated or improved
since the 2004 elections. Further, we stated that EAC had not defined
tasks, processes, and time frames for completing these activities, and we
recognized that other organizations had actions under way that were
intended to improve the security of electronic voting systems. These
actions include developing and obtaining international acceptance for
voting system standards, developing voting system software in an open
source environment (i.e., not proprietary to any particular company), and
cataloging and analyzing reported problems with electronic voting systems.

To improve the security and reliability of electronic voting systems, we
made recommendations to EAC for establishing tasks, processes, and time
frames for improving the federal voluntary voting system guidelines,
testing capabilities, and management support available to state and local
election officials. The EAC commissioners agreed with our recommendations
and stated that actions to address each were either under way or intended,
and the NIST director agreed with our conclusions.

67The Federal Election Commission used the general term "voting system
standards" for its 2002 publication Voting Systems Performance and Test
Standards. Consistent with HAVA terminology, EAC refers to its revision of
these standards as Voluntary Voting System Guidelines. For this report, we
refer to the contents of both of these documents as standards.

Page 349 GAO-06-450 Elections

Certain Types of Tests and Evaluations Were Widely Performed on Voting
Systems, while Others Were Less Common

To ensure that voting systems perform as intended during use, the systems
must be effectively tested, both before they are accepted from the
manufacturer and before each occasion that they are used. Further
confidence in election results can be gained by conducting Election Day
and postelection audits of voting systems. For the November 2004 general
election, voting system testing was conducted for almost all voting
systems, but the types and content of the testing performed varied
considerably. Most states and local jurisdictions employed national and
state certification testing and readiness testing to some extent, but the
criteria used in this testing were highly dependent on the state or
jurisdiction. Also, many, but not all, states and jurisdictions conducted
acceptance testing of both newly acquired systems and those undergoing
changes or upgrades. In contrast, relatively few states and jurisdictions
conducted parallel testing during elections or audits of voting systems
following elections.

To assist election officials in testing voting systems for the 2004
general election, most local jurisdictions documented policies and
procedures related to some types of testing, according to estimates based
on our survey of local jurisdictions. However, the testing approaches
embodied in policies and procedures that the local jurisdictions we
visited shared with us varied considerably. Furthermore, in jurisdictions
we visited, few voting system problems were reported as a result of local
testing, and correspondingly few changes were made to the systems or
election processes. The variability in testing approaches among states and
jurisdictions underscores our previously reported concerns from our
September 2005 report about whether actual testing of voting systems is
sufficient to ensure satisfaction of system requirements, including those
associated with accuracy, reliability, and security.68

Voting system test and evaluation can be grouped into various types or
stages: certification testing (national level), certification testing
(state level), acceptance testing, readiness testing, parallel testing,
and postelection voting system audits. Each of these tests has a specific
purpose, and is conducted at the national, state, or local level at a
particular time in the election cycle. Table 27 summarizes these types of
tests.

68GAO-05-956.

 Table 27: Types of Testing and Evaluation for Voting Systems, with Common Time
                          Frames and Responsibilities

Test type        Purpose               When conducted  Responsibility      
Certification    To verify compliance  Prior to (or as Federal authorities 
(national)a      of voting equipment   a condition of) and independent     
                    with federal          system          laboratoriesb       
                    standards             acceptance      
Certification    To validate           Before election State election      
(state)          compliance of voting                  authorities         
                       equipment with                     
                       state-specific                     
                    requirements                          
                    To verify that voting                      State or local 
Acceptance       equipment             Before election            election 
                                                                  authorities 
                    delivered by a vendor                 
                              meets state                 
                    or local requirements                 
Readiness (logic To verify that voting Before election Local election      
and                       equipment is                 authorities         
accuracy)        functioning properly,                 
                    usually by                            
                          confirming that                 
                      predictable outputs                 
                      are produced from                   
                         predefined                       
                    inputsc                               
                       To verify accurate                      State or local 
Parallel                performance of During election            election 
                                                                  authorities 
                         voting equipment                 
                           through random                 
                    selection and                         
                    systematic                            
                    evaluation of                         
                    operational                           
                    equipment                             
                    To review and                              State or local 
Audit            reconcile election    After election             election 
                                                                  authorities 
                       records to confirm                 
                          correct conduct                 
                        of an election or                 
                         uncover evidence                 
                         of problems with                 
                         voting equipment                 
                    or election processes                 

Source: GAO analysis based on GAO-02-3 and GAO-05-956.

aWith the enactment of HAVA in 2002, responsibility for overseeing
national testing of voting systems and certifying those that met federal
standards was assigned to EAC in HAVA S: 231(a)(1) (codified at 42 U.S.C.
S: 15371(a)(1)). EAC assumed this responsibility in August 2005, when it
was transferred from NASED. Under NASED, national testing against federal
standards was called qualification testing.

bRefers to EAC and testing laboratories accredited by them as provided for
in HAVA S: 231 (codified at 42 U.S.C. S: 15371).

cReadiness testing that is conducted to confirm the proper functioning of
election equipment on Election Day just before the polls open is sometimes
called verification testing.

Many states have laws or regulations that mandate specific types of
testing for voting equipment and time frames for conducting those tests.
Documented policies and procedures for testing and evaluation provide an
important means for ensuring that testing is effectively planned and
executed. Effective test and evaluation can greatly reduce the chances of
unexpected or unknown equipment problems and errors. From our local
jurisdiction survey for the 2004 election, we estimate that 85 percent of
local jurisdictions had documented policies and procedures for some type
of voting system testing, 6 percent of jurisdictions did not have policies
and procedures for testing, and 9 percent did not know whether their
jurisdictions had them. Larger jurisdictions were more likely to have
these management tools than smaller ones. An estimated 96 percent of large
jurisdictions had documented testing policies and procedures, compared
with 89 percent of medium and 82 percent of small jurisdictions.69 The
difference between large and small jurisdictions is statistically
significant.

The testing policies and procedures of the local jurisdictions we visited
presented a wide variety of approaches and details for the 2004 general
election. For instance, election officials in 1 large jurisdiction in
Connecticut told us that they did not conduct acceptance testing on their
lever equipment, which had been in use for many years, and did not conduct
either parallel testing or audit testing, stating that these tests were
not applicable to its systems for 2004. However, officials said they did
conduct readiness testing at the polling place prior to the election.
Election officials in a large Ohio jurisdiction that used punch card
voting equipment told us that readiness testing had been conducted by
local officials. However, election officials stated that certification and
acceptance testing were not performed for 2004 because this system had
been used in prior elections. They also said that neither parallel testing
nor audit testing of voting systems was performed. Officials in a large
Colorado jurisdiction we visited that used central count optical scan
equipment told us that they obtained state certification of the newly
purchased equipment, conducted acceptance and readiness testing prior to
the election, and executed another readiness test following the election.
Election officials in a large Georgia jurisdiction that used DRE voting
equipment reported that the state performed both certification and
acceptance testing when the equipment was purchased and conducted a
parallel test of the tabulation system during the election. Further, local
officials reported that they conducted readiness testing prior to the
election, but did not perform postelection audit testing. For the 5 local
jurisdictions that provided us with copies of procedures for readiness
testing, three sets of procedures were developed by the jurisdictions
themselves and two sets were developed by the voting equipment vendors.

69The 95 percent confidence interval for small jurisdictions is +6 or -7
percentage points.

National Certification Testing Was Widely Performed, but Its Standards and
Usage Varied

The enactment of HAVA in 2002 established federal responsibilities for the
certification of voting systems to meet federal standards and provided the
framework for a national testing program. The act charged EAC, supported
by NIST, with instituting a federal program for the development and
adoption of voluntary voting system guidelines against which voting
systems can be evaluated, establishing processes and responsibilities for
accrediting laboratories to test systems, and using the results of testing
by the accredited labs to certify the voting systems. In 2005, EAC
developed guidelines for the certification process and defined the steps
needed for the process to transition from NASED to EAC. States and local
jurisdictions are to decide whether and how to use the testing and
certification results from the federal program in their elections
processes.

Most states continued to require that voting systems be nationally tested
and certified. In our October 2001 report on election processes, we
reported that 38 states required that their voting systems meet federal
standards for the November 2000 general election, which meant that the
systems were tested by NASED. For voting systems being used for the first
time in the 2004 general election, national certification testing was
almost uniformly required. From our prior discussion of state survey
responses in the context of voting system security, 26 of 27 states using
DRE for the first time in this election, as well as the District of
Columbia, required them to be nationally certified, while 9 of the 10
states using punch card equipment for the first time, and 30 of 35 states
and the District of Columbia using optical scan equipment for the first
time, said they had such requirements.

It is unclear whether the proportion of nationally certified systems
changed between the 2000 and 2004 general elections. In our October 2001
report on election processes nationwide, we reported that an estimated 39
percent of jurisdictions used NASED-qualified voting equipment for the
2000 general election.70 However, for the 2004 general election, we
estimate that  68 percent of jurisdictions did not know whether the
respective systems that they used were NASED-qualified. This uncertainty
surrounding the national qualification status of a specific version of
voting system at the local level underscores a concern we recently
reported with respect to electronic voting security and reliability in our
September 2005 report on this topic-that is, even though voting system
software may have been

70GAO-02-3.

States Generally Required Certification of Voting Systems Using a Range of
Criteria

qualified and certified at the national or state levels, software changes
and upgrades performed at the local level may not be qualified and
certified.71

The upcoming 2006 general election can be viewed as a challenging
transition period in the voting system capabilities, standards, and
national certification, with several testing-related factors potentially
increasing the difficulty of this transition. First, HAVA's requirements
for voting system capabilities, such as voter error correction and manual
audit, along with the attendant new guidelines, are likely to require
additional testing at the national level to recertify previously fielded
and certified systems that have been upgraded. Second, this increased
workload is not likely to be met with added national testing capacity,
since the process for accrediting new voting system testing laboratories
is not expected to produce newly accredited labs in time for the 2006
election. Third, the complexity of the testing being performed is likely
to increase because states report that they will collectively apply the
full range of available standards-1990, 2002, and 2005 standards, as well
as various combinations of these-to voting systems first used for the
November 2006 election. As a result, a range of test protocols must be
developed or maintained, and a variety of corresponding tests must be
planned, executed, and analyzed to meet the variety of standards.

Most states continue to certify voting systems to ensure that they meet
minimum state election requirements. In our October 2001 report on
election processes, we reported that 45 states and the District of
Columbia had certification programs for their voting systems, 38 of which
required that the systems be tested before they were certified for the
2000 general election.72 In addition, we reported that an estimated 90
percent of local jurisdictions used state-certified voting equipment for
the November 2000 general election. However, we also reported that state
officials had expressed concerns with voting system changes that did not
undergo recertification. Since then, we have reported that security
experts and election officials have expressed similar concerns.73

71GAO-05-956. 72GAO-02-3. 73GAO-05-956.

Page 354 GAO-06-450 Elections

For the November 2004 general election, 42 states and the District of
Columbia reported on our state survey that they required state
certification of voting systems.74 (See fig. 75.) Seven states required
certification of the voting equipment purchased at the state level for
local jurisdictions in the 2004 election. However, in 35 states and the
District of Columbia, officials reported that responsibility for
purchasing a state-certified system rested with the local jurisdiction.
While state certification requirements often included NASED testing, as
well as approval or confirmation of functionality for particular ballot
conditions, some states also included additional requirements for features
such as quality of construction, transportation safety, and documentation.
Although the remaining 8 states did not require state certification, the
officials we contacted described other mechanisms to address the
compliance of voting equipment with state-specific requirements, such as a
state approval process or acceptance of voting equipment based on federal
certification. Figure 75 shows states' reported certification requirements
for voting systems used in the 2004 general election.

74Initial state responses were obtained from our 2005 state survey; we
later obtained clarifications by phone or electronic mail from 19 states.

Page 355 GAO-06-450 Elections

Figure 75: State-Reported Voting System Certification Requirements for the
2004 General Election

Sources: GAO 2005 survey of state election officials and follow-up with
selected states for clarification of responses (analysis), MapArt (map).

For the 2006 general election, 44 states reported that they will have
requirements for certification of voting systems, 2 more states than for
the 2004 general election. The District of Columbia reported that it will
not require voting system certification for the 2006 general election. Of
the  44, all but 1 expected to conduct the certification themselves; the
1 state reported that it would rely solely on a national independent
testing authority to make its certification decision. Furthermore, of the
43 other states conducting certification themselves, 41 reported that they
would include testing of system functions to obtain certification. In
addition,  18 of the 43 states planned to involve a national testing
laboratory in their certification process.

Acceptance Testing Continued to Be Commonly Performed, but Implemented
Practices Varied Widely

As we reported previously in our October 2001 report on election
processes, either states or local jurisdictions conducted acceptance tests
prior to the 2000 general election.75 However, the testing processes, test
steps, and involvement of vendors in the testing performed varied by
jurisdiction and by type of equipment. Also, we reported in our 2001
report that states and local jurisdictions sometimes relied heavily on
vendors to design and conduct acceptance tests. With respect to vendor
involvement in particular, we reported that vendors were sometimes heavily
relied upon to design and conduct acceptance tests.

For the 2004 election, the extent and variety of acceptance testing was
similar to those for the 2000 election. With regard to state roles and
involvement in acceptance testing of new voting systems, 26 states and the
District of Columbia reported responsibilities at some level of
government. Specifically, 8 states and the District of Columbia reported
on our survey that they had responsibility for performing acceptance
testing, 15 states required local jurisdictions to perform such testing,
and 3 states reported that requirements for acceptance testing existed at
both the state and local levels. Twenty-two states either did not require
such testing or did not believe that such testing was applicable to them.
(Two states did not know their acceptance testing requirements for the
2004 election.) More states required that acceptance testing be performed
for changes and upgrades to existing systems than they did for new
systems-30 states in all and the District of Columbia. Specifically, 15
states and the District of Columbia were responsible for performing
acceptance tests for changes and upgrades, 10 states required local
jurisdictions to perform these tests, and  5 states required acceptance
testing at both the state and local levels.

Election officials at a majority of the local jurisdictions that we
visited told us that they conducted some type of acceptance testing for
newly acquired voting equipment. As with the 2000 general election, these
officials described a variety of approaches to acceptance testing for the
2004 general election. For example, the data used for testing could be
vendorsupplied, developed by election officials, or both, and could
include system initialization, logic and accuracy, and tamper resistance.
Other steps, such as diagnostic tests, physical inspection of hardware,
and software configuration checks, were also mentioned as testing
activities by local election officials. Further, election officials from 3
jurisdictions that we

75GAO-02-3.

Readiness Testing Continued to Be Widely Performed Using Various
Approaches

visited said that vendors were heavily involved in designing and executing
the acceptance tests, while officials from another jurisdiction that we
visited said that vendors contributed to a portion of their testing. In 
2 jurisdictions in Georgia, officials said that acceptance tests were
conducted at a university center for elections systems.

Most jurisdictions conducted readiness testing, also known as logic and
accuracy testing, for both the 2000 and 2004 general elections. In
addition, some states reported that they conducted readiness testing for
the 2004 general election. The content and nature of these tests varied
among jurisdictions.

According to our state survey, 49 states and the District of Columbia
reported that they performed readiness testing of voting systems at the
state level, the local level, or both (1 state did not require readiness
testing). Most states required local jurisdictions to perform readiness
testing (37 states in all). However, 7 states reported that they performed
their own readiness testing of voting equipment for the 2004 general
election in addition to local testing. Five states and the District of
Columbia reported that they had no requirements for local jurisdictions to
perform readiness testing but conducted this testing themselves. State
laws or regulations in effect for the 2004 election typically had specific
requirements for when readiness testing should be conducted and who was
responsible for testing, sometimes including public demonstrations of
voting system operations. For example, one state mandated that local
jurisdictions conduct three readiness tests using all types of election
ballots including audio ballots. One test took place before Election Day
and two occurred on Election Day-before the official counting of ballots
began and after the official counting had been completed. Another state
required the Secretary of State to conduct testing using pre-audited
ballots before Election Day, as well as on Election Day before ballots
were counted.

On the basis of a subgroup of local election jurisdictions from our 2000
election survey, we estimate that 96 percent of jurisdictions nationwide
conducted readiness testing before the 2000 general election.76 For a

76These estimates include only county election jurisdiction subgroup
comparisons between 2001 and 2005 surveys. See the scope and methodology
in appendix V for further details about sampling differences.

Page 358 GAO-06-450 Elections

comparable subgroup of jurisdictions in the 2004 general election, we
estimate that 95 percent of local jurisdictions conducted readiness
testing. The frequency with which readiness testing was conducted in 2004
was largely stable across all jurisdictions of various sizes that did not
solely use hand-counted paper ballots, ranging between an estimated 90
percent (for small jurisdictions) to an estimated 96 percent (for large
jurisdictions). Whenever the sample of jurisdictions permitted statistical
comparison, there were also no significant differences between the
percentages of jurisdictions that said they conducted readiness testing
for various predominant voting methods.

The variety of readiness testing activities performed by jurisdictions for
the 2000 general election was also evident for the 2004 general election.
Election officials in all of the local jurisdictions we visited following
the 2004 election reported that they conducted readiness testing on their
voting equipment using one or more of the approaches we identified for the
2000 election, such as diagnostic tests, integration tests, mock
elections, and sets of test votes. Election officials in many of these
jurisdictions told us that they combined test approaches. For example,
officials in 1 large jurisdiction in Florida told us that they conducted
pre-election testing using complete ballots (not test decks) to determine
the accuracy of the marks and to see if there were any errors in voting
machine programming. They told us that logic and accuracy testing was
performed for each machine using undervoted ballots and overvoted ballots,
and that zero tapes were run for each voting machine before the
election.77 In addition, a diagnostic test was run before the election on
each voting machine. According to the local officials, this was the test
approach described in the manufacturer's preparation checklist. Election
officials in another Florida jurisdiction stated that readiness testing
included integration testing to demonstrate that the voting system is
properly programmed; the election is correctly defined on the system; and
all system inputs, outputs, and communication devices are in working
order. In the case of these jurisdictions, the state requires logic and
accuracy testing and submission of the test parameters to the state.

77Zero tapes record the vote counts stored in a piece of voting equipment
for each of the contests prior to voter inputs, when all counts should be
zero.

Page 359 GAO-06-450 Elections

Parallel Testing Was Not Frequently Performed

Parallel testing was not widely performed by local jurisdictions in the
2004 general election, although 7 states reported on our state survey that
they performed parallel testing of voting systems on Election Day, and
another  6 states reported that this testing was required by local
jurisdictions.78 From our survey of local jurisdictions, we estimate that
2 percent of jurisdictions that did not solely use hand-counted paper
ballots conducted parallel testing for the 2004 general election.79 Large
and medium jurisdictions primarily performed this type of testing (7
percent and  4 percent of jurisdictions, respectively). The percentage of
small jurisdictions performing this type of testing was negligible (0
percent). The differences between both large and medium jurisdictions and
small jurisdictions are statistically significant.

Our visits to local jurisdictions affirmed the limited use of parallel
testing. Specifically, election officials in 2 of the 28 jurisdictions
that we visited told us that they performed parallel testing. Officials in
1 large jurisdiction in Georgia told us that parallel testing was
conducted by the state in conjunction with a university center for voting
systems. In another case, officials in a large jurisdiction in Kansas told
us that parallel testing was required by the local jurisdiction and was
publicly conducted. In both cases, the tests were conducted on voting
equipment for which security concerns had been raised in a voting
equipment test report issued by the state of Maryland prior to the 2004
general election. Local officials who told us that parallel testing was
not performed on their voting systems attributed this to the absence of
parallel testing requirements, a lack of sufficient voting equipment to
perform these tests, or the unnecessary nature of parallel testing because
of the stand-alone operation of their systems.

Postelection Voting System According to our state survey, 22 states and
the District of Columbia reported that they performed postelection voting
system audits for the 2004

Audit Requirements and

Practices Were Diverse general election. Specifically, 4 states and the
District of Columbia reported that they conducted postelection audits of
voting systems themselves, 

78Both EAC's Best Practices Tool Kit and the 2005 Voluntary Voting System
Guidelines recommend development of parallel testing procedures for all
types of automated voting equipment.

79We estimate that 91 percent of jurisdictions considered parallel testing
to be not applicable.

Page 360 GAO-06-450 Elections

16 states required that audits of voting systems be conducted by local
jurisdictions, and 2 states reported that audits of voting systems were
performed at both the state and local levels. State laws or regulations in
effect for the 2004 general election varied in when and how these audits
were to be conducted. In addition, a variety of statutes cited by states
for testing requirements did not mention postelection voting system
audits, and the one that did lacked details on the scope or components of
such audits.

According to our local jurisdiction survey, postelection voting system
audits were conducted by an estimated 43 percent of local jurisdictions
that did not solely use hand-counted paper ballots on Election Day. This
practice was much more prevalent at large and medium jurisdictions  (62
percent and 55 percent, respectively) than small jurisdictions  (34
percent).80 The differences between small jurisdictions and both medium
and large jurisdictions are statistically significant. We further estimate
that these voting system audits were conducted more frequently in
jurisdictions with central count optical scan voting methods (54 percent)
than they were in jurisdictions with precinct count optical scan voting
methods (35 percent).81 Figure 76 shows the estimated use of postelection
audits for jurisdictions with different voting methods in the 2004 general
election.

80The 95 percent confidence interval for large jurisdictions is +/- 8
percentage points. The  95 percent confidence interval for small
jurisdictions is +/- 7 percentage points.

81Percentages for the other voting methods had confidence intervals
greater than  +/- 10 percent or were unreliable.

Page 361 GAO-06-450 Elections

Figure 76: Estimated Percentages of Jurisdictions Conducting Postelection
Voting System Audits for the 2004 General Election, by Predominant Voting
Method

Percentage of jurisdictions 70

60

53 54

50 40

35

30 20 10

DREnt

                                       b

                                     count

a

                                       c

                                       n

                                       n

Central cou

                                      sca

                                      sca

                                    Precinct

optical

                                    optical

Predominant voting method

Source: GAO 2005 survey of local election jurisdictions.

aThe 95 percent confidence interval for DRE is +/- 14 percentage points.
bThe 95 percent confidence interval for central count optical scan is +/-
10 percentage points. cThe 95 percent confidence interval for precinct
count optical scan is +/- 8 percentage points.

Election officials in 14 of 28 local jurisdictions that we visited told us
that they conducted postelection voting system audits. However, the
conditions and scope of voting system audits varied. Some were routine,
while others were conducted only in the event of close races or challenges
to results. Among the 14 jurisdictions, most of the officials we spoke
with said that they focused on reconciling voting machine counts with
known votes, and officials in 2 of these jurisdictions characterized the
voting system audits largely as voting system logic and accuracy tests.
However, officials with a few jurisdictions told us that they also
reviewed voting machine logs, sampled results from random precincts, or
employed independent auditors to repeat and verify vote counting. In 1
large jurisdiction in Nevada, an election official told us that paper
results were compared to the tabulated

Information on Jurisdictions' Election Technology Integration Is Limited,
but Current and Emerging Technologies Offer Greater Opportunities

results of votes counted on 24 machines. In addition, every voting machine
was activated and the same scripts used for pre-election testing were
rerun through the machines. According to the election official, this level
of testing was required by law.

The number of jurisdictions that have integrated particular aspects of
voting system components and technologies was limited for the 2004 general
election for the areas of integration we examined, based on estimates from
our local jurisdiction survey and visits to local jurisdictions. For the
areas of integration we did examine, the scope and nature of this
integration was diverse and included remote programming of electronic
ballots, statewide tabulation of voting results, and end-to-end management
of the election process. Nevertheless, the potential for greater
integration in the future does exist as states and jurisdictions act on
their earlier discussed plans to acquire the kind of voting equipment
(e.g., optical scan and DRE products) that lends itself to integration. It
is unclear if and when this migration to more technology-based voting
methods will produce more integrated election system environments.
However, suitable standards and guidance for these interconnected
components and systems-some of which remain to be developed-could
facilitate the development, testing, operational management, and
maintenance of components and systems, thereby maximizing the benefits of
current and emerging election technologies and achieving states' and local
jurisdictions' goals for performance and security.

Reported Instances of Certain Election Systems Integration Approaches Have
Been Inconsistent

Various voting systems, components, and technologies-some of which have
been available since the 2000 general election-encompass a wide range of
functional capabilities and system interactions. According to our local
jurisdiction survey estimates and visits to election jurisdictions for the
2004 general election, officials reported various types of integration,
but there were few instances. The areas in which integration was reported
can be grouped into four categories: (1) electronic programming or setup
of voting equipment from a centralized facility, (2) electronic
aggregation and tabulation of voting results from multiple voting systems
or locations,

(3) add-on voting features and technologies, and (4) electronic management
of voting equipment and operations.

Electronic Programming or Setup

Electronic programming or setup of voting equipment involves integration
between an administrative system and voting equipment to initialize vote
count totals, load ballot definitions, and authorize voter access. As we
previously reported in our September 2005 report on the security and
reliability of electronic voting, this type of integration has raised
security concerns.82 Election officials in 19 of the 28 jurisdictions that
we visited used portable memory cartridges or cards for electronic
programming or setup of their voting equipment. To accomplish programming
or setup, officials at some of the local jurisdictions that we visited
said that they used a computer to preload voting equipment with ballots or
tabulation logic prior to transporting the equipment to polling locations.
At 1 large New Jersey jurisdiction, officials stated that the
administrative computer used a dedicated connection to the election server
to electronically transmit the data and logic necessary to program and
enable the units for the election. Election officials in some
jurisdictions told us that an administrative system loaded ballot
definitions onto portable electronic devices, such as memory cartridges or
smart cards, which were then physically transported to the locations where
the voting equipment was being prepared for the election-either at a
storage facility or polling location (see fig. 77). The cartridges or
cards were then inserted into individual voting units to prepare or
activate them for the election. Some electronic ballot cards were provided
directly to the voter to activate the voting equipment, then returned to
election workers when the ballot has been cast.

82GAO-05-956.

Figure 77: Examples of Portable Memory Cards and Cartridges Used with
Voting Equipment

                                  Source: GAO.
                             Electronic aggregation or tabulation of cast     
Electronic Aggregation or ballots also requires                            
Tabulation                integration between voting equipment and another 
                                                      computer system that is 
                             responsible for collecting and aggregating the   
                             votes. Figure 78 shows                           
                             examples of computer systems used for vote       
                             tabulation.                                      

      Figure 78: Examples of Ballot Tabulation Equipment and Environments

Source: GAO.

Transfer of votes or election results between the voting equipment and the
central tabulator may employ portable electronic media or
telecommunication lines. Portable electronic media were the means that
officials at 7 of the 28 jurisdictions that we visited said they used to
electronically aggregate election results from multiple voting locations.
For DRE equipment, memory cartridges that stored cast ballots from
individual voting units were transferred to the election office, and the
data they contained were uploaded and tallied by an electronic tabulation
system.

Add-on Features and Technologies

Some jurisdictions also used telecommunications services to transfer
election data from polling locations or election coordination centers to
tabulation facilities, although how these services were used varied.
Officials at 4 jurisdictions that we visited told us that they employed
dial-up connections to transmit local vote tallies for further tabulation.
For instance, election officials in a large jurisdiction in Washington
told us that after the polls were closed and all ballots were scanned and
recorded by the optical scan machines at each polling place, the machines
were taken to storage areas, where the results were transmitted to the
central computer for tabulation using the jurisdiction's phone line.
Officials at a large jurisdiction that we visited in Ohio said that they
had election judges take voting machine memory cartridges from their
polling locations to facilities where laptop computers would read the
cartridges and transmit vote tallies over phone lines to a remote access
server at the elections office. In a large jurisdiction that we visited in
Illinois, election officials told us that they took their portable
precinct ballot counters to 1 of 10 stations throughout the city, where
vote totals from the counters were encrypted and transmitted to a remote
access server via a cellular network.

Add-on features and technologies to ensure the accuracy of votes, provide
easier access to persons with disabilities or special needs, and enhance
security or privacy were also integrated into voting systems by a few
states and jurisdictions for the 2004 general election. Officials at both
large jurisdictions in Nevada that we visited told us that they had
integrated a VVPT capability into their DREs to meet a state requirement
for VVPT. Figure 79 shows one example of a VVPT voting system component.
Overall, we estimate that about 8 percent of jurisdictions operating DRE
voting equipment in the November 2004 general election produced VVPT.83

83The 95 percent confidence interval for this item is +9 or -6 percentage
points.

Figure 79: Example of an Assembly for Producing a Voter-Verifiable Paper
Audit Trail for DRE Voting Equipment

Source: GAO.

Audio features were also added to voting systems for the 2004 election.
Officials at 6 of the jurisdictions that we visited reported that they had
incorporated an audio ballot component into their DRE machines for voters
with sight impairments. Election officials in 3 jurisdictions reported
that they offered audio ballots in languages other than English.

Security and privacy capabilities, such as data encryption and virtual
private networks, were also reportedly integrated into several
jurisdictions' voting system environments for the 2004 general election to
protect electronically transferred election data or to secure remote
system access. Election officials at 6 of the 28 jurisdictions that we
visited said they used encryption to protect ballots during electronic
storage. Officials at both jurisdictions in Georgia explained that their
state-selected DRE equipment used individual access cards for each voter,
uniquely encrypted data on the card (including the voter's cast ballot)
for each polling location, and a separately encrypted electronic key
needed to access the voter's ballot. Officials at 7 jurisdictions said
they applied encryption to the transmission of election results during the
2004 general election. Election officials in  1 large Colorado
jurisdiction stated that they used a virtual private network to ensure the
secrecy of data and authenticity of parties when transmitting election
results from jurisdictions to the state.

                             Electronic Management

Electronic management of voting equipment and operations was another form
of integration employed for the 2004 general election. Electronic
management covers such functions as equipment testing, initializing,
operational monitoring, diagnosis, troubleshooting, shutdown, and
auditing. It also includes election operations that affect voting
equipment, such as voter processing at the polling place and handling of
absentee ballots. We previously reported that some of these capabilities
were available during the 2000 general election in our October 2001 report
on election processes.84 For the 2004 general election, on the basis of
our local jurisdiction survey, we estimate that 7 percent of jurisdictions
that used voting methods other than paper ballots connected their voting
equipment via a local network at their polling locations. The frequency
with which remote access to voting systems was provided for the 2004
general election was similarly low (estimated at 10 percent of
jurisdictions that used voting methods other than paper ballots) but was
again affected by the size of jurisdictions. We estimate that a higher
percentage of large jurisdictions used remote access to voting equipment
(estimated at 19 percent) than medium jurisdictions (13 percent) or small
jurisdictions (7 percent). The difference between large and small
jurisdictions is statistically significant.85 Furthermore, we estimate
that remote access was primarily provided to local election officials (in
6 percent of jurisdictions) and to a lesser extent, state election
officials, voting equipment vendors, and third parties. Figure 80 shows
the estimated percentages of jurisdictions of various sizes that used
networking or various types of remote access. These capabilities pose
voting system security and reliability concerns as reported in our
September 2005 report on the security and reliability of electronic
voting.86 From approximately 20 open-ended text responses to our survey of
local jurisdictions that described steps taken to prevent unauthorized
remote access to voting systems, four safeguards were identified:
employing passwords for remote users, limiting operations to specific
election activities, use of virtual private networks, and system
monitoring.

84GAO-02-3.

85Percentages represent the aggregation of responses for all sources of
remote access. Percentages exclude jurisdictions that use only
hand-counted paper ballots.

86GAO-05-956.

Page 369 GAO-06-450 Elections

Figure 80: Estimated Percentages of Jurisdictions That Used Networking or
Remote Access for Voting Equipment in the 2004 General Election

Percentage of jurisdictions

7

7

6

6 5

4

4

3

3 2 1

0

0

                                      ess

orked atpolling places

                                       a

                                       a

                                       a

                                       a

                                       y

y

                                       y

                                       s

                                     ess b

                               election officials

                                       s

                                    support

                                     ess b

ess b

                                    official

                                     vendor

                                   Remote acc

                                   Remote acc

                                   Remote acc

Remote accNetw

local election

                                      ory

                                       b

                                      ty f

                                     d par

                                     state

                                      thir

Types of connections to voting systems

Source: GAO 2005 survey of local election jurisdictions.

aJurisdictions may have selected one or more types of remote access.

As we previously reported in our September 2001 report on voting
assistance to military and overseas citizens, state and local election
officials used technologies like electronic mail and faxing to better
integrate activities during the 2000 general election and to improve
communications with absentee voters.87 According to our estimates from the
local jurisdiction survey for the 2004 election, jurisdictions continued
to use electronic mail to interact with voters and also relied on Web
sites for a variety of election needs including voter registration status,
the application and processing of absentee ballots, and the status of
provisional ballots. For seven items in our survey where we asked about

87 GAO-01-1026.

jurisdictions' use of e-mail and Web sites for voter services, we estimate
that large jurisdictions generally used these technologies more frequently
than both medium and small jurisdictions, and that differences in six of
these items were statistically significant. Figure 81 shows the extent to
which jurisdictions of different sizes employed e-mail and Web sites for
selected voter services. In addition to using technology to support
individual voters, election officials in 1 large jurisdiction we visited
in New Mexico described their use of telecommunications technology to
support early voting at multiple locations. This jurisdiction connected
its registration database to its early voting locations with dedicated
phone lines, thus making voter registration information electronically
available at each location.

Figure 81: Estimated Percentages of Jurisdictions Using Automated Election
Support for the 2004 General Election, by Jurisdiction Size

    Percentage of jurisdictions

90

80

70

60

50 46

                 Status of voter         Status of   Notification of 
                   registration          absentee           absentee 
                   on Web sitea         application      application 
                                        by e-maila          problems 
                                                           by e-mail 
            Voter notification support               
            Percentage of jurisdictions              
            90       83                              

       Application     Ballot provided by e-mailb Application submitted by
downloaded from Web e-mailc Application requested by e-maila
          siteb        
       Absentee ballot                                     
               support                                     
       Small (<10,000)       Medium (10,000-100,000)         Large (>100,000) 

            Source: GAO 2005 survey of local election jurisdictions.

Prospects for Expanded Use of Integrated Election Systems Are Unclear

Note: The 95 percent confidence interval for all jurisdiction sizes is +/-
8 percentage points or less.

aThe differences between both small and medium jurisdictions and large
jurisdictions are statistically significant. bThe differences between all
categories of jurisdiction size are statistically significant. cThe
difference between small jurisdictions and large jurisdictions is
statistically significant.

Relatively few local jurisdictions we visited reported having plans for
integrating or further integrating their election-related systems and
components for the 2006 general election, and in cases where they had
plans, the scope and nature of the plans varied. At the same time, we
estimate on the basis of our local jurisdiction survey that a relatively
large proportion of jurisdictions expect to acquire DREs and optical scan
systems, which will introduce greater integration opportunities. However,
given the uncertainty surrounding the specific types of systems and
features to be acquired, the extent and timing of greater integration of
voting systems and components, as well as election-related systems,
remains to be seen.

More specifically, officials in several jurisdictions that we visited told
us about plans to integrate relatively modular add-on components to their
systems, while officials with several other jurisdictions described plans
for more complex end-to-end interactions among election systems and
technologies. For example, officials at 5 jurisdictions that we visited
reported plans to introduce a VVPT capability for future elections, and
officials at 2 jurisdictions reported plans to integrate an audio
component to comply with HAVA requirements. In another case, officials in
 2 jurisdictions told us that their state is planning to purchase
electronic poll books for its precincts to use during the 2006 elections
to electronically link its voter registration system with its voting
systems. Officials at another jurisdiction told us that they plan to
obtain a new optical scanner that will be used to tabulate both DRE and
optical scan election results.

The scope and magnitude of election system integration may be influenced,
in part, by the jurisdictions' adoption of the optical scan and DRE voting
methods and the corresponding products that support add-on automated
features, such as languages and accessibility tools, and interactions
among automated components of the election process, such as ballot
generation and tabulation. As we discussed earlier in this chapter,
one-fifth of local jurisdictions are planning to acquire new optical scan
and DRE voting equipment in time for the 2006 general election. For
instance, on the basis of our survey of local jurisdictions, we estimate
that 25 percent of jurisdictions plan to acquire precinct count optical
scan voting equipment by the November 2006 general election. However, some
jurisdictions had not yet finalized their time frame for acquiring voting
equipment at the time of our survey. In addition, their acquisition plans
also include technologies for their election Web sites. Figure 82
estimates the percentages of jurisdictions with acquisition plans for
various technologies and their implementation time frames. While the
advent of more technology-based voting methods provides greater
opportunities for integration, the uncertainty around the timing and
nature of their introduction makes the future extent of this integration
unclear at this point.

Figure 82: Estimated Percentages of Local Jurisdictions with Plans to
Acquire Voting Equipment

Percentage of jurisdictions

30

                                       25

25

20 19

15

10

10

8

7

5

4

4 3

0 DRE Central count Precinct count Technology optical scan optical scan
for election Web site

Information technology to be acquired

By November 2006

Date uncertain Source: GAO 2005 survey of local election jurisdictions.

It is important for voting system standards developers to recognize the
opportunity and potential for greater integration of election systems. EAC
recently adopted a new version of the voluntary voting system guidelines
in

Concluding Observations

December 2005 that will become effective in December 2007. However, this
version does not address some of the capabilities discussed above. For
instance, the guidelines do not address the integration of registration
systems with voting systems. Neither do they address
commercial-off-the-shelf devices (such as card readers, printers, or
personal computers) or software products (such as operating systems or
database management systems) that are used in voting systems without
modification. EAC has acknowledged that more work is needed to further
develop the technical guidelines in areas such as voting accessibility,
usability, and security features. Such efforts have the potential to
assist states and local jurisdictions in maximizing the benefits of
emerging election technologies.

The challenges confronting local jurisdictions in acquiring and operating
voting technologies are not unlike those faced by any technology user-
adoption and consistent application of standards for system capabilities
and performance, reliable measures and objective data to determine whether
the systems are performing as intended, rigorous and disciplined
performance of security and testing activities, and successful management
and integration of the people, process, and technology components of
elections during system acquisition and operation. These challenges are
heightened by other conditions common to both the national elections
community and other information technology environments: the distribution
of responsibilities among various organizations, technology changes,
funding opportunities and constraints, emerging requirements and guidance,
and public attention.

The extent to which states and local jurisdictions adopt and consistently
apply up-to-date voting systems standards will directly affect the
security and performance of voting systems. A substantial proportion of
jurisdictions have yet to adopt the most current federal voting system
standards or related performance measures. Even if this happens, however,
other challenges loom because systems will need to be tested and
recertified by many states (and by federal processes whenever states have
adopted national standards) to meet any newly adopted voting standards and
HAVA requirements for accuracy. Organizations involved with
recertification-including federal, state, and local governments; testing
authorities; and vendors-may need the capacity to assume the workloads
associated with expected increases in the adoption of current standards
and the use of new voting systems so that potential risks to near-term
election processes are minimized.

Reliable measures and objective data are also considered essential
management practices for determining whether the technology being used is
meeting the needs of the jurisdiction's user communities (both the voters
and the officials who administer the elections). Looking back to the
November 2000 and 2004 general elections, we estimate that the vast
majority of jurisdictions were satisfied with the performance of their
respective technologies. However, considering that our local jurisdiction
surveys for the 2000 and 2004 elections indicated limited collection of
voting system performance data, we conclude that estimated levels of
satisfaction with voting equipment found in our local surveys have been
mostly based on a patchwork of operational indicators and, based on site
visits to local jurisdictions, have involved anecdotal experiences of
election officials. Although these impressions should not be discounted,
informed decision making on voting system changes and investment would
benefit from more objective data about how well existing equipment is
meeting specific requirements, such as those governing system accuracy,
reliability, efficiency, and security. No one voting method, or particular
voting system make and model, will meet the needs of every jurisdiction.
The challenge is thus to ensure that decisions about staying with an
existing voting method or investing in new or upgraded voting equipment
are made on the basis of reliable and relevant data about the operational
performance of the existing method against requirements and standards, as
well as the benefits to be derived versus the costs to be incurred with
each choice.

Effective execution of well-planned security and testing activities
provides opportunities to anticipate and address potential problems before
they affect election results. This is important because even a few
instances of election errors or disruptions can have a sizable impact if
election results are close. We estimate that the vast majority of
jurisdictions performed security and testing activities in one form or
another for the 2004 general election. However, the nature and extent of
these activities varied among jurisdictions-to some degree by jurisdiction
size, voting method, or perceived applicability of the activities. These
activities were also largely responsive to-and limited by-formal state and
local directives. When appropriately defined and implemented, such
directives can promote the effective execution of security and testing
practices across all phases of the elections process. As voting
technologies and requirements evolve, states and local jurisdictions face
the challenge of regularly updating and consistently implementing the
directives to meet the needs of their specific election environments.

As we noted for the 2000 general election, managing the three election
components of people, process, and technology as interrelated and
interdependent variables presents an important challenge in the
acquisition or operation of a given voting method. Whether a state or
jurisdiction is acquiring, testing, operating, or maintaining a new voting
system or an existing one, how successfully the system actually performs
throughout the election cycle will depend not only on how well the
technology itself has been designed, but also on how well the people and
processes associated with the system fulfill their roles for each stage.
The technical potential of more extensive integration of voting equipment,
components, and election systems also holds the prospect for even more
interrelationships and interdependencies among the people, processes, and
technologies, with all their attendant risks. In addition to establishing
minimum functional and performance requirements and processes for voting
system aspects of the election process, system standards can also be used
to govern the integration of election systems; address the accuracy,
reliability, privacy, and security of components and interfaces; and
deliver needed support for the people and processes that will use the
integrated election systems. Timely development of integration standards
presents a challenge to the election community to keep pace with the
advancement of election systems and technology.

Appendix I

Related GAO Products

Elections: Absentee Voting Assistance to Military and Overseas Citizens
Increased for 2004 General Election, but Challenges Remain. GAO-06-521.
Washington, D.C.: April 7, 2006.

Election Reform: Nine States' Experiences Implementing Federal
Requirements for Computerized Statewide Voter Registration Lists.
GAO-06-247. Washington, D.C.: February 7, 2006.

Elections: Views of Selected Local Election Officials on Managing Voter
Registration and Ensuring Eligible Citizens Can Vote. GAO-05-997.
Washington, D.C.: September 27, 2005.

Elections: Federal Efforts to Improve Security and Reliability of
Electronic Voting Systems Are Under Way, but Key Activities Need to Be
Completed. GAO-05-956. Washington, D.C.: September 21, 2005.

Elections: Additional Data Could Help State and Local Elections Officials
Maintain Accurate Voter Registration Lists. GAO-05-478. Washington, D.C.:
June 10, 2005.

Department of Justice's Activities to Address Past Election-Related Voting
Irregularities. GAO-04-1041R. Washington, D.C.: September 14, 2004.

Elections: Electronic Voting Offers Opportunities and Presents Challenges.
GAO-04-975T. Washington, D.C.: July 20, 2004.

Elections: A Framework for Evaluating Reform Proposals. GAO-02-90.
Washington, D.C.: October 15, 2001.

Elections: Perspectives on Activities and Challenges across the Nation.
GAO-02-3. Washington, D.C.: October 15, 2001.

Elections: Statistical Analysis of Factors That Affected Uncounted Votes
in the 2000 Presidential Election. GAO-02-122. Washington, D.C.: October
15, 2001.

Elections: Status and Use of Federal Voting Equipment Standards.
GAO-02-52. Washington, D.C.: October 15, 2001.

Voters with Disabilities: Access to Polling Places and Alternative Voting
Methods. GAO-02-107. Washington, D.C.: October 15, 2001.

Elections: Voting Assistance to Military and Overseas Citizens Should Be
Improved. GAO-01-1026. Washington, D.C.: September 28, 2001.

Comparison of Voting Age Population to Registered Voters in the 40 Largest
U.S. Counties. GAO-01-560R. Washington, D.C.: March 23, 2001.

Elections: The Scope of Congressional Authority in Election
Administration. GAO-01-470. Washington, D.C.: March 13, 2001.

Bilingual Voting Assistance: Assistance Provided and Costs. GAO/GGD-97-81.
Washington, D.C.: May 9, 1997.

Puerto Rico: Confusion over Applicability of the Electoral Law to
Referendum Process. HRD-93-84. Washington, D.C.: May 28, 1993.

Voting: Some Procedural Changes and Informational Activities Could
Increase Turnout. PEMD-91-1. Washington, D.C.: November 2, 1990.

Bilingual Voting Assistance: Costs of and Use during the November 1984
General Election. GGD-86-134BR. Washington, D.C.: September 15, 1986.

Justice Can Further Improve Its Monitoring of Changes in State/Local
Voting Laws. GGD-84-9. Washington, D.C.: December 19, 1983.

                                  Appendix II

                    2005 Survey of State Election Officials

                                  Appendix III

                  2005 Survey of Local Election Jurisdictions

Appendix IV

Summary of Local Election Jurisdictions GAO Visited

Local election jurisdictions, such as counties, cities, townships, and
villages, conduct elections-including federal and state contests. We
visited 28 local election jurisdictions to collect information about the
election administration process and their experiences during the November
2004 general election. In addition to information such as population size,
number of precincts, and type of voting system used in the November 2004
general election,1 we include information about whether these
jurisdictions provided bilingual voting materials. Under section 203 of
the Voting Rights Act,2 the Census Bureau Director is to make
determinations3 based upon the most recent Census data as to which
jurisdictions must provide voting materials4 in specified minority
languages as well as in the English language. In our summaries we
identify, pursuant to the Census Bureau Director's determinations under
the Voting Rights Act, whether or not each of the 28 jurisdictions we
visited is currently (e.g., for purposes of this appendix, as of the
November 2004 election) required to provide bilingual voting materials to
any particular group of language minority voters.

Following is a brief summary of these 28 local jurisdictions, as reported
to us by election officials we met with. The summaries, except where
another source is specifically identified, are based upon information
provided to us by local election officials and provide a general
description of the environment in which the local election jurisdictions
operate. The statements in this appendix that are based upon information
from local election officials were not independently verified by us.

1We asked election officials about the number of precincts in their
jurisdiction for the November 2004 general election. Some officials
indicated whether the total number included absentee ballot precincts,
while others did not. In most jurisdictions we visited, the number of
precincts was not the same as the number of polling places operating on
Election Day.

242 U.S.C. S: 1973aa-1a.

3Determinations made by the Census Bureau Director are to be based upon a
prescribed statutory formula. The most recent determinations made as to
which jurisdictions are to be covered were published on July 26, 2002, at
67 Fed. Reg. 48,871 (2002).

4In this context, the term "voting materials" means registration or voting
notices, forms, instructions, assistance, or other materials or
information relating to the electoral process, including ballots.

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

Larimer County, Colorado

The County Clerk and Recorder is responsible for overseeing election
administration in Larimer County, from registration through recounts. At
the time of our visit, Larimer County had 10 full-time employees and 1
parttime employee working on election administration activities. The 2000
Census showed that Larimer County had a population of about 251,000.
During our site visit, Larimer County election officials told us that for
the 2004 election, the county had about 200,000 registered voters.
Pursuant to the Census Bureau Director's determinations under the Voting
Rights Act, Larimer County is not currently required to provide bilingual
voting materials to any particular group of language minority voters. As a
result, the county provides voting materials in English only. For the
November 2004 election, the county had 144 precincts. However, Larimer
County used 31 "vote centers" as polling places. On Election Day, any
registered voter could go to any of the 31 vote centers to cast a ballot.
The county also offered its registered voters the opportunity to vote
early at designated early voting locations prior to Election Day. For the
November 2004 general election, the county used optical scan ballots for
early, absentee, provisional, and Election Day voting. For early and
Election Day voting, the ballots were counted at the early voting sites
and vote centers. Absentee and provisional ballots were counted centrally
using optical scan counters. This county was included in our site visits
to local jurisdictions for our October 2001 comprehensive report on
election processes nationwide.5

El Paso County, Colorado

The County Clerk and Recorder is responsible for overseeing election
administration in El Paso County, from registration through recounts. At
the time of our visit, El Paso County had nine full-time staff working on
election administration activities. The 2000 Census showed that El Paso
County had a population of about 517,000. During our site visit, El Paso
County election officials told us that for the 2004 election, the county
had about 354,000 registered voters. Pursuant to the Census Bureau
Director's determinations under the Voting Rights Act, El Paso County is
not currently required to provide bilingual voting materials to any
particular group of language minority voters. As a result, the county
provides voting materials in English only. For the November 2004 general
election, the county had 378 precincts. The county also offered its
registered voters the opportunity to vote early at designated locations
prior to Election Day. For the

5For more information, see GAO, Elections: Perspectives on Activities and
Challenges across the Nation, GAO-02-3 (Washington, D.C.: Oct. 15, 2001).

Page 469 GAO-06-450 Elections

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

City of New Haven, Connecticut

November 2004 general election, the county used direct recording
electronic (DRE) voting systems for early voting. Optical scan ballots
were used to record absentee, provisional, and Election Day votes. Ballots
cast on Election Day were counted using optical scan counters located in
precincts. Absentee and provisional ballots were counted centrally.

Responsibility for election administration for the City of New Haven is
divided between two entities-the Registrars of Voters and the City Clerk's
Office. The Registrars of Voters have responsibility for voter
registration, Election Day activities, provisional voting, vote counting
and recounts, voting equipment, and voter education. The City Clerk is
responsible for designing and disseminating absentee ballots, printing
paper ballot strips for the lever machines, and providing sample ballots
to polling places. At the time of our visit, there were six full-time
staff in the Registrars of Voters Office and three full-time staff in the
City Clerk's Office working on election administration activities. The
2000 Census showed that the City of New Haven had a population of about
124,000. During our site visit, New Haven election officials told us that
for the 2004 general election, the city had about 58,000 registered
voters. Pursuant to the Census Bureau Director's determinations under the
Voting Rights Act, the City of New Haven is currently required to provide
bilingual voting materials to a particular group of language minority
voters. As a result, the city provides voting materials in English and
Spanish. For the November 2004 general election, the city had 42 precincts
operating on Election Day and used lever machines for Election Day voting.
Paper ballots were used to record absentee and provisional votes and were
hand-counted in a central location. In addition, the city provided
presidential ballots to Connecticut residents that were not registered to
vote, and former Connecticut residents who had moved to another state
within 30 days of the election. These ballots could be used only to vote
for presidential and vice presidential candidates.

City of Hartford, Connecticut

Responsibility for election administration for the City of Hartford is
divided between two entities-the Registrars of Voters and the City Clerk's
Office. The Registrars of Voters have the responsibility for voter
registration, Election Day activities, provisional voting, vote counting
and recounts, voting equipment, and voter education. The City Clerk is
responsible for designing and disseminating absentee ballots, printing
paper ballot strips for the lever machines, and providing sample ballots
to polling places. At

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

Broward County, Florida

the time of our visit, there were nine full-time staff in the Registrars
of Voters and three full-time staff in the City Clerk's Office working on
election administration activities. The 2000 Census showed that the City
of Hartford had a population of about 122,000. According to the Secretary
of State's Web site, at the time of the 2004 general election, the city
had about 58,000 registered voters. Pursuant to the Census Bureau
Director's determinations under the Voting Rights Act, the city of
Hartford is currently required to provide bilingual voting materials to a
particular group of language minority voters. As a result, the city
provides voting materials in English and Spanish. For the November 2004
general election, the city had 23 precincts operating on Election Day and
used lever machines for Election Day voting. Paper ballots were used to
record absentee and provisional votes and were hand-counted in a central
location. In addition, the city provided presidential ballots to
Connecticut residents who were not registered to vote, and former
Connecticut residents who had moved to another state within 30 days of the
election. These ballots could be used only to vote for presidential and
vice presidential candidates.

The County Supervisor of Elections is responsible for overseeing election
administration in Broward County, from registration through recounts. At
the time of our visit, Broward County had 72 full-time staff working on
election administration activities. The 2000 Census showed that Broward
County had a population of about 1.6 million. During our site visit,
Broward County election officials told us that for the November 2004
general election, the county had about 1 million registered voters.
Pursuant to the Census Bureau Director's determinations under the Voting
Rights Act, Broward County is currently required to provide bilingual
voting materials to a particular group of language minority voters. As a
result, the county provides voting materials in English and Spanish.
County officials told us that they were unable to provide voting materials
in Seminole, a Native American language, because the language is phonetic
and cannot be written. In addition, county officials told us that they
provided some voter education materials in English, Spanish, Portuguese,
and Creole. For the November 2004 general election, the county had 793
precincts operating on Election Day. The county also offered its
registered voters the opportunity to vote early at designated locations
prior to Election Day. For the November 2004 general election, the county
used DRE voting systems on Election Day and for early voting. Paper
ballots were used to record absentee and provisional votes and were
counted centrally using optical scan counters.

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

                              Leon County, Florida

The County Supervisor of Elections is responsible for overseeing election
administration in Leon County, from registration through recounts. At the
time of our visit, Leon County had 16 full-time staff working on election
administration activities. The 2000 Census showed that Leon County had a
population of about 239,000. During our site visit, Leon County election
officials told us that for the November 2004 general election, the county
had about 171,000 registered voters. Pursuant to the Census Bureau
Director's determinations under the Voting Rights Act, Leon County is not
currently required to provide bilingual voting materials to any particular
group of language minority voters. As a result, the county provides voting
materials in English only. Additionally, for the November 2004 general
election, the county provided ballots in Braille for the blind. The county
had 179 precincts operating on Election Day for the November 2004 general
election. The county also offered its registered voters the opportunity to
vote early at the Supervisor of Elections' office building prior to
Election Day. For the November 2004 general election, the county used
paper ballots to record absentee, early, provisional, and Election Day
votes. Election Day and early voting ballots were counted using precinct
optical scan counters, and absentee and provisional ballots were counted
centrally using optical scan counters.

Dougherty County, Georgia

The Supervisor of Elections is responsible for overseeing election
administration in Dougherty County, from registration through recounts. At
the time of our visit, Dougherty County had three full-time staff working
on election administration activities. The 2000 Census showed that
Dougherty County had a population of about 96,000. During our site visit,
Dougherty County election officials told us that for the 2004 general
election, the county had about 46,000 registered voters. Pursuant to the
Census Bureau Director's determinations under the Voting Rights Act,
Dougherty County is not currently required to provide bilingual voting
materials and assistance to any particular group of language minority
voters. As a result, the county provides voting materials in English only.
For the November 2004 general election, the county had 31 precincts
operating on Election Day. The county also offered its registered voters
the opportunity to vote early in the election office prior to Election
Day. For the November 2004 general election, the county used DRE voting
systems for Election Day, early, and in-person absentee voting. Paper
ballots were used to record provisional and mail-in absentee votes and
were counted centrally using optical scan 

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

Muscogee County, Georgia

counters. This county was included in our site visits to local
jurisdictions for our October 2001 report on election processes.6

The Director of Elections, appointed by the City Council, is responsible
for overseeing election administration in Muscogee County, from
registration through recounts. At the time of our visit, Muscogee County
had five fulltime staff working on election administration activities. The
2000 Census showed that Muscogee County had a population of about 186,000.
During our site visit, Muscogee County election officials told us that for
the 2004 general election, the county had about 106,000 registered voters.
Pursuant to the Census Bureau Director's determinations under the Voting
Rights Act, Muscogee County is not currently required to provide bilingual
voting materials to any particular group of language minority voters. As a
result, the county provides voting materials in English only. For the
November 2004 general election, the county had 49 precincts, with 1
absentee voting precinct for purposes of counting ballots, operating on
Election Day. The county also offered its registered voters the
opportunity to vote early at the election office prior to Election Day.
For the November 2004 general election, the county used DRE voting systems
for Election Day and early voting. Paper ballots were used to record
absentee and provisional ballots and were counted centrally using optical
scan counters.

Champaign County, Illinois

The County Clerk is responsible for overseeing election administration in
Champaign County, from registration through recounts. At the time of our
visit, Champaign County had 16 full-time staff working on election
administration activities. The 2000 Census showed that Champaign County
had a population of about 180,000. During our site visit, Champaign County
election officials told us that for the 2004 general election, the county
had about 123,000 registered voters. Pursuant to the Census Bureau
Director's determinations under the Voting Rights Act, Champaign County is
not currently required to provide bilingual voting materials to any
particular group of language minority voters. As a result, the county
provides voting materials in English only. For the November 2004 general
election, the county had 117 precincts operating on Election Day. The
county used punch card ballots to record Election Day, absentee, and
provisional votes. These ballots were counted centrally using punch card
counter machines.

6See GAO-02-3.

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

                           City of Chicago, Illinois

This county was included in our site visits to local jurisdictions for our
October 2001 report on election processes.7

The Executive Director of the Board of Elections is responsible for
overseeing election administration in the City of Chicago, from
registration through recounts. At the time of our visit, the City of
Chicago had 300 fulltime staff on the Board of Elections working on
election administration activities. The 2000 Census showed that the City
of Chicago had a population of about 2.9 million. During our site visit,
Chicago election officials told us that for the 2004 general election, the
city had about 

1.4 million registered voters. Pursuant to the Census Bureau Director's
determinations under the Voting Rights Act, Cook County, in which the City
of Chicago is located, is currently required to provide bilingual voting
materials to particular groups of language minority voters. As a result,
the city provides voting materials in English, Spanish, and Chinese
(including Mandarin and Cantonese). In addition, election officials told
us that they provide voter registration information and other materials in
15 languages, including Croatian, Korean, Russian, and Tagalog. For the
November 2004 general election, the city had 2,709 precincts operating on
Election Day. The city used punch card ballots to record Election Day,
absentee, and provisional votes. Election Day and absentee ballots were
counted in each precinct using punch card counters. Provisional ballots
were counted centrally using punch card counters.

Wyandotte County, Kansas

The County Election Commissioner is responsible for overseeing election
administration in Wyandotte County, from registration through recounts. At
the time of our visit, Wyandotte County had six full-time staff working on
election administration activities. The 2000 Census showed that Wyandotte
County had a population of about 158,000. During our site visit, Wyandotte
County election officials told us that for the November 2004 general
election, the county had about 89,000 registered voters. Pursuant to the
Census Bureau Director's determinations under the Voting Rights Act,
Wyandotte County is not currently required to provide bilingual voting
materials to any particular group of language minority voters. As a
result, the county provides voting materials in English only. For the
November 2004 general election, the county had 159 precincts operating on
Election

7See GAO-02-3.

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

Johnson County, Kansas

Day. The county also offered its registered voters the opportunity to vote
early at the headquarters office prior to Election Day. For the November
2004 general election, the county used paper optical scan ballots to
record Election Day, absentee, early, and provisional votes. Election Day
ballots were counted at the precincts using optical scan counters, and
early ballots were counted at the early voting location using optical scan
counters. Absentee and provisional ballots were counted using optical scan
counters in a central location.

The County Election Commissioner is responsible for overseeing election
administration in Johnson County, from registration through recounts. At
the time of our visit, Johnson County's Election Office had 16 full-time
staff working on election administration activities. The 2000 Census
showed that Johnson County had a population of about 451,000. During our
site visit, Johnson County election officials told us that for the
November 2004 general election, the county had about 349,000 registered
voters. Pursuant to the Census Bureau Director's determinations under the
Voting Rights Act, Johnson County is not currently required to provide
bilingual voting materials to any particular group of language minority
voters. As a result, the county provides voting materials in English only.
For the November 2004 general election, the county had 415 precincts
operating on Election Day. The county also offered its registered voters
the opportunity to vote early at designated early voting locations prior
to Election Day. For the November 2004 general election, the county used
DRE voting systems for Election Day voting. Paper optical scan ballots
were used to record early, absentee, and provisional votes and were
counted in a central location using optical scan counters.

City of Manchester, New Hampshire

The City Clerk is responsible for overseeing election administration for
the City of Manchester, from registration through recounts. At the time of
our visit, the City Clerk's Office had 14 full-time staff working on
election administration activities. The 2000 Census showed that the City
of Manchester had a population of about 107,000. During our site visit,
Manchester's City Clerk told us that for the November 2004 general
election, the city had about 57,000 registered voters. In terms of
population, the City of Manchester is the largest city in New Hampshire
and is located in the largest county. Pursuant to the Census Bureau
Director's determinations under the Voting Rights Act, the City of
Manchester is not currently required to provide bilingual voting materials
to any particular

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

Town of Madbury, New Hampshire

group of language minority voters. As a result, the city provides voting
materials in English only. For the November 2004 general election, the
county had 12 precincts operating on Election Day. The city used paper
optical scan ballots to record Election Day and absentee votes, which were
counted in each precinct using optical scan counters.

The Town Clerk is responsible for overseeing election administration for
the Town of Madbury, from registration through recounts. At the time of
our visit, the town had no full-time staff working on election
administration activities, but had nine part-time workers, including the
Town Clerk, assigned to administer elections. The 2000 Census showed that
the Town of Madbury had a population of about 1,500. During our site
visit, Madbury election officials told us that for the November 2004
general election, the town had about 1,200 registered voters. Pursuant to
the Census Bureau Director's determinations under the Voting Rights Act,
the Town of Madbury is not currently required to provide bilingual voting
materials to any particular group of language minority voters. As a
result, the town provides voting materials in English only. For the
November 2004 general election, the town had one precinct operating on
Election Day. The town used hand-counted paper ballots to record Election
Day and absentee votes.

Middlesex County, New Jersey

Responsibility for election administration for the County of Middlesex is
divided between two entities-the County Clerk's Office and the Board of
Elections. The County Clerk's Office is responsible for sending out
absentee ballots, designing ballots, and certifying elections to the
state. The Board of Elections has responsibility for voter registration,
poll workers, polling places, vote counting and recounting, and voting
equipment. At the time of our visit, the Board of Elections had 21
full-time staff and 4 parttime commissioners working on election
administration activities. The 2000 Census showed that Middlesex County
had a population of about 750,000. During our site visit, election
officials told us that about  427,000 residents of the county are
registered voters. Pursuant to the Census Bureau Director's determinations
under the Voting Rights Act, Middlesex County is currently required to
provide bilingual voting materials to a particular group of language
minority voters. As a result, the county provides voting materials in
English and Spanish. For the November 2004 general election, the county
had 597 precincts operating on Election Day. The county used DRE voting
systems on Election Day. Paper ballots

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

Passaic County, New Jersey

were used to record absentee and provisional votes, and were counted
centrally using optical scan counters. In addition, paper ballots were
used to record emergency votes, and were hand counted in a central
location. This county was included in our site visits to local
jurisdictions for our October 2001 report on election processes.8

Responsibility for election administration for Passaic County is divided
among three entities-the Board of Elections, the County Clerk's Office,
and the Superintendent of Elections Office. The Board of Elections has
responsibility for poll workers, polling places, vote counting, and
recounts. The County Clerk's Office has responsibility for ballot design,
accepting absentee ballot applications and sending out ballots, and
certifying the election to the state. The Superintendent of Elections has
responsibility for voter registration and voting equipment. At the time of
our visit, there were 5 full-time staff in the Board of Elections Office,
3 full-time staff in the County Clerk's Office, and 25 full-time staff in
the Superintendent of Elections Office working on election administration
activities. The 2000 Census showed that Passaic County had a population of
about 490,000. During our site visit, Passaic County election officials
told us that for the 2004 general election, the county had about 255,000
registered voters. Pursuant to the Census Bureau Director's determinations
under the Voting Rights Act, Passaic County is currently required to
provide bilingual voting materials to a particular group of language
minority voters. As a result, the county provides voting materials in
English and Spanish. For the November 2004 general election, the county
had 288 precincts operating on Election Day. The county used DRE voting
systems on Election Day. Paper ballots were used to record absentee and
provisional votes, and were counted centrally using optical scan machines.
In addition, emergency ballots were available for "emergency" situations,
such as DRE malfunction or power loss, and were hand-counted in the
precinct.

8See GAO-02-03.

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

                              Clark County, Nevada

The Registrar of Voters, appointed by the County Manager, is responsible
for overseeing election administration in Clark County, from registration
through recounts. At the time of our visit, Clark County had 38 full-time
staff working on election administration activities. The 2000 Census
showed that Clark County had a population of about 1.4 million. During our
site visit, Clark County election officials told us that for the 2004
general election, the county had about 684,000 registered voters. Pursuant
to the Census Bureau Director's determinations under the Voting Rights
Act, Clark County is currently required to provide bilingual voting
materials to a particular group of language minority voters. As a result,
the county provides voting materials in English and Spanish. For the
November 2004 election, the county had 1,042 precincts operating on
Election Day. The county also offered its registered voters the
opportunity to vote early at designated early voting locations prior to
Election Day. For the November 2004 general election, the county used DRE
voting systems for Election Day, early voting, and provisional voting.
Absentee and provisional votes, other than those cast on Election Day,
were recorded on paper ballots and counted centrally using optical scan
counters.

Washoe County, Nevada

The Registrar of Voters, appointed by the County Manager, is responsible
for overseeing election administration in Washoe County, from registration
through recounts. At the time of our visit, Washoe County had eight
fulltime staff working on election administration activities. The 2000
Census showed that Washoe County had a population of about 339,000. During
our site visit, Washoe County election officials told us that for the 2004
general election, the county had about 234,000 registered voters. Pursuant
to the Census Bureau Director's determinations under the Voting Rights
Act, Washoe County is not currently required under the Voting Rights Act
to provide bilingual voting materials to any particular group of language
minority voters. Nonetheless, the county provided voting materials in
English and Spanish. For the November 2004 general election, the county
had 236 out of 489 precincts operating on Election Day. The county also
offered its registered voters the opportunity to vote early at designated
early voting locations prior to Election Day. For the November 2004
general election, the county used DRE voting systems for Election Day and
early voting. Paper ballots were used to record absentee and provisional
votes, and were counted centrally using optical scan machines.

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

Bernalillo County, New Mexico

The Elections Administrator is responsible for overseeing election
administration in Bernalillo County, from registration through recounts.
At the time of our visit, Bernalillo County had 29 full-time staff working
on election administration activities. The 2000 Census showed that
Bernalillo County had a population of about 557,000. During our site
visit, Bernalillo County election officials told us that for the 2004
general election, the county had about 357,000 registered voters. Pursuant
to the Census Bureau Director's determinations under the Voting Rights
Act, Bernalillo County is currently required to provide bilingual voting
materials to particular groups of language minority voters. As a result,
the county provides voting materials in English and Spanish, and provides
interpreters in three Native American languages, Navajo, Tiwa, and Keres,
which have no written form. In addition, the county has a Native American
coordinator who works with the various pueblos to discuss specific issues.
For the November 2004 general election, the county had 413 precincts
operating on Election Day. The county also offered its registered voters
the opportunity to vote early at designated early voting locations prior
to Election Day. For the November 2004 general election, the county used
DRE voting systems for Election Day and early voting. Absentee ballots
were counted centrally using optical scan machines. Provisional votes were
recorded on paper ballots and hand-counted centrally. This county was
included in our site visits to local jurisdictions for our October 2001
report on election

9

processes.

Santa Fe County, New Mexico

The Bureau of Elections is responsible for overseeing election
administration in Santa Fe County, from registration through recounts. At
the time of our visit, Santa Fe County had eight full-time staff working
on election administration activities. The 2000 Census showed that Santa
Fe County had a population of about 129,000. During our site visit, Santa
Fe County election officials told us that for the 2004 general election,
the county had about 95,000 registered voters. Pursuant to the Census
Bureau Director's determinations under the Voting Rights Act, Santa Fe
County is currently required to provide bilingual voting materials to
particular groups of language minority voters. As a result, voting
materials were provided in English and Spanish. The county also provides
oral translation and audiotapes in Native American languages, some of
which have no written form. For the November 2004 general election, the
county had 87 precincts,

9See GAO-02-3.

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

Carteret County, North Carolina

including one absentee voting precinct for purposes of counting absentee
ballots, operating on Election Day. The county also offered its registered
voters the opportunity to vote early at designated early voting locations
prior to Election Day. For the November 2004 general election, the county
used DRE voting systems for Election Day. In addition, absentee ballots
were counted centrally using optical scan machines. Provisional votes were
recorded on paper ballots and centrally hand-counted. This county was
included in our site visits to local jurisdictions for our October 2001
report on election processes.10

The Board of Elections is responsible for overseeing election
administration in Carteret County, from registration through recounts. At
the time of our visit, Carteret County had two full-time staff working on
election administration activities. The 2000 Census showed that Carteret
County had a population of about 59,000. During our site visit, Carteret
County election officials told us that for the 2004 general election, the
county had about 42,000 registered voters. Pursuant to the Census Bureau
Director's determinations under the Voting Rights Act, Carteret County is
not currently required to provide bilingual voting materials to any
particular group of language minority voters. Nonetheless, the county
provided some voting materials, such as polling place posters, in English
and Spanish. Election ballots and other materials were provided in English
only. For the November 2004 general election, the county had 34 precincts
operating on Election Day. The county also offered its registered voters
the opportunity to vote early at the Board of Elections' office prior to
Election Day. For the November 2004 general election, the county used DRE
voting systems for Election Day and early voting. Optical scan ballots
were used to record absentee and provisional votes, and were centrally
counted using optical scan machines.

10See GAO-02-3.

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

Guilford County, North Carolina

The Board of Elections is responsible for overseeing election
administration in Guilford County, from registration through recounts. At
the time of our visit, Guilford County had 14 full-time staff working on
election administration activities. The 2000 Census showed that Guilford
County had a population of about 421,000. During our site visit, Guilford
County election officials told us that for the 2004 general election, the
county had about 275,000 registered voters. Pursuant to the Census Bureau
Director's determinations under the Voting Rights Act, Guilford County is
not currently required to provide bilingual voting materials to any
particular group of language minority voters. Nonetheless, the county
provided ballots in English and voting instructions in English and
Spanish. For the November 2004 general election, the county had 159
precincts operating on Election Day. The county also offered its
registered voters the opportunity to vote early at designated early voting
locations prior to Election Day. For the November 2004 general election,
the county used DRE voting systems for most Election Day and early voting
ballots. Optical scan ballots were used to record absentee and provisional
votes, and were centrally counted using optical scan machines.

Cuyahoga County, Ohio

The Board of Elections is responsible for overseeing election
administration in Cuyahoga County, from registration through recounts. At
the time of our visit, Cuyahoga County had about 86 full-time staff
working on election administration activities. The 2000 Census showed that
Cuyahoga County had a population of about 1.4 million. During our site
visit, Cuyahoga County election officials told us that for the 2004
general election, the county had about 1 million registered voters.
Pursuant to the Census Bureau Director's determinations under the Voting
Rights Act, Cuyahoga County is not currently required to provide bilingual
voting materials to any particular group of language minority voters. As a
result, the county provides voting materials in English only. For the
November 2004 general election, the county had 1,436 precincts operating
on Election Day. The county used punch card ballots for Election Day,
absentee, and provisional voting that were counted in a central location.
This county was included in our site visits to local jurisdictions for our
October 2001 report on election processes.11

11See GAO-02-3.

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

Mahoning County, Ohio

The Board of Elections is responsible for overseeing election
administration in Mahoning County, from registration through recounts. At
the time of our visit, Mahoning County had 12 full-time staff working on
election administration activities. The 2000 Census showed that Mahoning
County had a population of about 258,000. During our site visit, Mahoning
County election officials told us that for the 2004 general election, the
county had about 195,000 registered voters. Pursuant to the Census Bureau
Director's determinations under the Voting Rights Act, Mahoning County is
not currently required to provide bilingual voting materials to any
particular group of language minority voters. As a result, the county
provides voting materials in English only. For the November 2004 general
election, the county had 312 precincts operating on Election Day. The
county used DRE voting systems for Election Day and Election Day
provisional voting. Optical scan ballots were used to record absentee
votes and were centrally counted using optical scan machines.

Allegheny County, Pennsylvania

The Allegheny County Elections Division is responsible for the day-to-day
activities of and overseeing election administration, from registration
through recounts. At the time of our visit, the county Elections Division
had 45 full-time staff working on election administration activities. The
2000 Census showed that Allegheny County had a population of about 

1.3 million. According to Pennsylvania's Department of State's Web site,
Allegheny County had about 919,000 registered voters for the 2004 general
election. Pursuant to the Census Bureau Director's determinations under
the Voting Rights Act, Allegheny County is not currently required to
provide bilingual voting materials to any particular group of language
minority voters. As a result, the county provides voting materials in
English only. For the November 2004 general election, the county had 1,317
precincts operating on Election Day. For the November 2004 general
election, the county used lever machines for Election Day voting. Paper
ballots were used to record absentee and provisional votes. Absentee
ballots were handcounted at the precincts, and provisional ballots were
hand-counted in a central location.

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

Montgomery County, Pennsylvania

The Voter Services Office is responsible for overseeing election
administration in Montgomery County, from registration through recounts.
At the time of our visit, Montgomery County had 18 full-time staff working
on election administration activities. The 2000 Census showed that
Montgomery County had a population of about 750,000. During our site
visit, Montgomery County election officials told us that for the 2004
general election, the county had about 565,000 registered voters. Pursuant
to the Census Bureau Director's determinations under the Voting Rights
Act, Montgomery County is not currently required to provide bilingual
voting materials to any particular group of language minority voters. As a
result, the county provides voting materials in English only. For the
November 2004 general election, the county had 407 precincts operating on
Election Day. The county used DRE voting systems for Election Day. Paper
ballots were used to record absentee and provisional votes. Absentee
ballots were hand-counted in each precinct and recounted by county staff
to obtain the official count. Provisional ballots were hand-counted in a
central location. This county was included in our site visits to local
jurisdictions for our October 2001 report on election processes.12

Clark County, Washington

The County Auditor is responsible for overseeing election administration
in Clark County, from registration through recounts. At the time of our
visit, Clark County had eight full-time staff working on election
administration activities. The 2000 Census showed that Clark County had a
population of about 345,000. During our site visit, Clark County election
officials told us that for the 2004 general election, the county had about
208,000 registered voters. Pursuant to the Census Bureau Director's
determinations under the Voting Rights Act, Clark County is not currently
required to provide bilingual voting materials to any particular group of
language minority voters. As a result, the county provides voting
materials in English only.  For the November 2004 general election, the
county had 180 precincts operating on Election Day. The county used punch
card ballots to record Election Day, absentee, and provisional votes.
These votes were counted in a central location. This county was included
in our site visits to local jurisdictions for our October 2001 report on
election processes.13

12See GAO-02-3. 13See GAO-02-3.

Page 483 GAO-06-450 Elections

        Appendix IV Summary of Local Election Jurisdictions GAO Visited

                            King County, Washington

The Director of Records, Elections and Licensing Services is responsible
for overseeing election administration in King County, from registration
through recounts. At the time of our visit, King County had 38 full-time
staff working on election administration activities. The 2000 Census
showed that King County had a population of about 1.7 million. In terms of
population, it has the largest population in Washington state. During our
site visit, King County election officials told us that for the 2004
general election, the county had about 1 million active registered voters.
Pursuant to the Census Bureau Director's determinations under the Voting
Rights Act, King County is currently required to provide bilingual voting
to a particular group of language minority voters. As a result, the county
provides voting materials in English and Chinese. For the November 2004
general election, the county had 2,616 precincts operating on Election
Day. The county used paper ballots on Election Day and for absentee and
provisional votes. Election Day ballots were counted at the precincts
using optical scan machines. Absentee and provisional ballots were counted
centrally using optical scan machines.

Appendix V

                       Objectives, Scope, and Methodology

                                  State Survey

Our objectives were to

     o identify changes to election systems since the November 2000 general
       election, including steps taken to implement provisions of the Help
       America Vote Act (HAVA) and
     o describe the issues and challenges encountered by election officials
       in the 2004 general election.

For both objectives, we (1) conducted a Web-based survey of the 50 states
and the District of Columbia on state laws and other actions taken at the
state level related to major stages of election administration-voter
registration, absentee and early voting, conducting elections, and vote
counting; (2) sent a mail questionnaire to election officials in a
representative sample of local election jurisdictions nationwide; and 

(3) visited 28 local election jurisdictions in 14 states to obtain
information about their experiences during the November 2004 general
election. To obtain additional information about local election
jurisdiction experiences, we attended the Election Center's 21st Annual
National Conference in August 2005. The Election Center is a national
association of state and local election officials.

To identify questions to ask in our surveys and during our site visits to
local jurisdictions about changes and issues and challenges, we reviewed
prior GAO reports and relevant studies. The studies included those done by
national or state organizations and state or local governments to assess
the November 2004 general election. We also attended several hearings held
by the Election Assistance Commission regarding election administration
issues and attended the winter 2005 joint meeting of the National
Association of Secretaries of State and National Association of State
Election Directors. Overall, we did our work between March 2005 and
February 2006 in accordance with generally accepted government auditing
standards.

Our state survey sought to update information on selected state statutory
requirements included in our 2001 elections report and inquired about
other changes that included actions taken to implement HAVA. To develop
our survey of state election officials, we reviewed existing studies about
the election process, including previous and ongoing GAO work. Social
science survey specialists designed a draft questionnaire in close
collaboration with GAO subject matter experts, and we conducted pretests

Page 485 GAO-06-450 Elections

Mail Survey of Local Jurisdictions

with representatives of 5 states to help further refine our questions,
develop new questions, clarify any ambiguous portions of the survey, and
identify any potentially biased questions. These pretests were conducted
in person and by telephone with election officials from states with
varying election system characteristics.

Prior to fielding our state survey, we contacted the Secretaries of State,
Chairs of Election Boards, or other responsible state-level officials to
confirm the contact information for the director of elections for each
state. We launched our Web-based survey on August 15, 2005, and received
all responses by November 7, 2005. Log-in information for the Web-based
survey was e-mailed to directors of elections. We sent one follow-up
e-mail message to all nonrespondents after the questionnaire had been
online for 2 weeks. After another 4 weeks, we contacted by telephone or
e-mail all those who had not completed the questionnaire. A representative
of the Council of State Governments also coordinated with us to encourage
a select number of nonresponding states to complete the survey. We
obtained responses from all 50 states and the District of Columbia for a
100 percent response rate. Even so, the number of responses to individual
questions may be fewer than 51, depending upon how many states were
eligible to or chose to respond to a particular question.

Because our state survey was not a sample survey, but rather a census of
all states, including the District of Columbia, there are no sampling
errors. However, the practical difficulties of conducting any survey may
introduce nonsampling errors. For example, differences in how a particular
question is interpreted or the sources of information available to
respondents can introduce unwanted variability into the survey results. We
included steps in both the data collection and data analysis stages for
the purpose of minimizing such nonsampling errors. As indicated above,
social science survey specialists designed a draft questionnaire, and
versions of the questionnaire were pretested with 5 members of the
population. We examined the survey results and performed computer analyses
to identify inconsistencies and other indications of error. A second,
independent analyst checked the accuracy of all computer analyses.

To obtain national information from local election officials on changes to
election systems since 2000 and election administration, we conducted a
mail survey of local election jurisdictions nationwide. Overall, there are
about 10,500 local government jurisdictions responsible for conducting
elections nationwide. States can be divided into two groups according to

Page 486 GAO-06-450 Elections

how they delegate election responsibilities to local jurisdictions. The
first group is composed of 41 states that delegate election
responsibilities primarily to counties, with a few of these states
delegating election responsibilities to some cities, and 1 state that
delegates these responsibilities to election regions. We included the
District of Columbia in this group of states. The first group contains
about one-fourth of the local election jurisdictions nationwide. The
second group is composed of  9 states that delegate election
responsibilities to subcounty governmental units, known by the U.S. Census
Bureau as minor civil divisions (MCD). This group of states contains about
three-fourths of the local election jurisdictions nationwide. The division
of the 50 states and the District of Columbia by how election
responsibilities are organized is as follows (states in bold delegate
election responsibilities to some cities independently from counties).

     o County-level states: Alaska (four election regions), Alabama, Arizona,
       Arkansas, California, Colorado, Delaware, the District of Columbia,
       Florida, Georgia, Hawaii, Idaho, Illinois, Indiana, Iowa, Kansas,
       Kentucky, Louisiana, Maryland, Mississippi, Missouri, Montana,
       Nebraska, Nevada, New Jersey, New Mexico, New York, North Carolina,
       North Dakota, Ohio, Oklahoma, Oregon, Pennsylvania, South Carolina,
       South Dakota, Tennessee, Texas, Utah, Virginia, Washington, West
       Virginia, and Wyoming
     o Minor civil division-level states: Connecticut, Maine, Massachusetts,
       Michigan, Minnesota, New Hampshire, Rhode Island, Vermont, and
       Wisconsin

While only about one-fourth of election jurisdictions nationwide are in
states that delegate election responsibilities primarily to counties,
according to Census 2000, 88 percent of the U.S. population lived in these
states. The U.S. population distribution between the two state groups is
shown in table 28.

2001 Local Election Jurisdiction Mail Survey

                    Table 28: Population in Each State Group

State group                                Population in 2000   Percentage 
County-level states                               247,277,791           88 
Minor civil division-level states                  34,143,449           12 
Total                                             281,421,240          100 

                Source: GAO analysis of U.S. Census Bureau data.

After the 2000 general election, GAO also conducted a mail survey of local
election jurisdictions nationwide. The sample frame for GAO's 2001 survey
consisted of (1) all county election jurisdictions, or their equivalents,
in  39 states that delegate election responsibilities primarily to
counties;1 

(2)
           the largest MCD in each county (based on 1999 Census population
           estimates) in the 9 states that delegate election responsibilities
           to MCDs;

(3)
           the District of Columbia; and (4) Alaska. Survey results were
           generalized to this sample frame, which covered 90 percent of the
           U.S. population.

The sample frame for the 2001 survey was stratified into three groups-
jurisdictions that used electronic voting machines; those that used
optical scan; and those that used any other method, including punch cards,
lever machines, and hand-counted paper ballots. To determine the
population of jurisdictions that used each type of voting method, we used
two databases from Election Data Services2-one for counties nationwide and
one for New England MCDs---supplemented by data we obtained from other
sources. We created separate strata for DRE and optical scan because these
were the two types of methods usually considered at the time as options by
jurisdictions purchasing new voting equipment.

Where possible, the results of some questions in the 2001 survey were
compared with results in the 2005 local election jurisdiction survey.
Differences in overall estimates from the 2001 and 2005 surveys of local

1We did not include any election jurisdictions in Oregon because statewide
voting in Oregon for the November 2000 election was conducted entirely by
mail and absentee balloting.

2Election Data Services is a private company that collects
election-related data from state and local jurisdictions, such as the
number of registered voters and voting methods used in local election
jurisdictions. We used several methods to check the reliability of data in
these databases and found the databases to be sufficiently reliable for
our purposes in this survey.

Page 488 GAO-06-450 Elections

2005 Local Election Jurisdiction Mail Survey

election jurisdictions are, in part, likely due to differences in the
sample designs of the two surveys and how MCDs were selected. Because of
these sample design differences, comparing only election jurisdictions
that are counties provides a stronger basis for making direct comparisons
between the two surveys' results. The estimates compared in this report
are of the county local election jurisdictions only; for this reason, some
estimates from the 2001 survey are slightly different than the overall
sample estimates provided in our prior report. For these comparisons, the
95 percent confidence interval is plus or minus 5 percentage points or
less for the 2001 survey estimates and plus or minus 8 percentage points
or less for the 2005 survey estimates.

Unlike for the 2001 GAO election survey, for the 2005 survey GAO
constructed its own sample frame. The initial list for each state group
was constructed from the 2000 decennial census data. Census population
data were available for all counties and county equivalents (cities that
are independent from counties and boroughs, municipalities, and other
census areas in Alaska); the county equivalents for Alaska were assigned
to their respective election regions. Census population data were also
available for all MCDs. We used jurisdiction population size to define
sample strata because these census data were readily available for all
counties and MCDs nationwide. Ideally, it would have been useful to define
the sample using national data on all registered voters or all eligible
voters. However, we did not use numbers of registered voters to define the
strata because census data on registered voters were not available at the
county and MCD level nationwide. And we did not use numbers of eligible
voters (individuals  18 years and over) to define the strata because
census data allowing us to exclude noncitizens and felons, groups that are
not eligible to vote, from the 18 years and over population, were also not
available at the county and MCD levels nationwide.

The large imbalance between the distribution of population and election
jurisdictions between the two groups of states created unique challenges
in designing a survey sample that, at the same time, is representative of
all election jurisdictions nationwide and covers the entire U.S.
population. Because about three-fourths of all election jurisdictions are
in the MCD states, in large part, a survey of a representative sample of
jurisdictions nationwide is a survey of the MCD jurisdictions and the
election processes, procedures, and practices that result from the laws of
the 9 states in which they are located. And, because of the large number
of MCD jurisdictions in Minnesota, Michigan, and Wisconsin, these 3 states
play a big role in the 



sample.3 Contrariwise, a survey that is representative of the U.S.
population nationwide, and not of local election jurisdictions nationwide,
similar to GAO's 2001 election survey, would place heavy emphasis on the
41 countylevel states in which 88 percent of the U.S. population lives. In
a sample like this, about 88 percent of the sampled jurisdictions would
come from the county-level states, and about 12 percent would come from
the MCD states. So neither the MCD jurisdictions nor their states would
receive as much emphasis in the sample. Therefore, we designed a survey
sample that would allow us to provide survey results that can be
generalized to all local election jurisdictions nationwide, as well as for
local election jurisdictions by population size.

We surveyed a stratified random probability sample of 788 local election
jurisdictions nationwide. The sampling unit was the geographically
distinct local election jurisdiction at the county, city, or MCD level of
local government (or, in Alaska, the election region). The population of
election jurisdictions was divided into the two state groups-county-level
states and MCD-level states, and each state group was then divided into
strata according to jurisdiction population size using Census 2000 data.
Countylevel states were divided into 4 strata, and MCD-level states were
divided into 6 strata. The allocation of units, or jurisdictions, to
strata was done in two stages (see table 29). In the first stage, 533
units were allocated across the 9 strata in proportion to the number of
jurisdictions in the population in each stratum. We used only 9 strata
because 1 stratum (stratum 5) did not have any jurisdictions. Sample
allocation at this stage allowed us to have a random sample of all local
election jurisdictions nationwide. In the second stage, we allocated an
additional 255 sample units to the 5 strata having the largest population,
with all jurisdictions in county-level states having a population greater
than 1 million being selected. Sample allocation at this stage allowed us
also to have a random sample of local jurisdictions nationwide according
to population size-large, medium, and small. To group jurisdictions by
population size, we combined jurisdictions in likesized population strata
in county-level and MCD-level states. We defined large jurisdictions as
those with a population greater than 100,000 (strata  1, 2, and 6),
medium jurisdictions as those with a population of more than 10,000 to
100,000 (strata 3 and 7), and small jurisdictions as those with a
population of 10,000 or less (strata 4, 8, 9, and 10). Thus, our survey
results

3There are more than 6,100 MCD jurisdictions in these 3 states combined.

can be generalized to all local election jurisdictions nationwide, as well
as to jurisdictions by population size category.

               Table 29: Local Election Survey Sample Allocation

                        Total number of Number in  Number in  Total number of 
                       jurisdictions in  stage 1      stage 2   jurisdictions 
Stratum                   population allocation allocation         sampled 
      County/city-greater                                                     
    1 than 1 million                 28          1         27              28
      population                                              
    2 County/city-100,001 to        422         21        136             157 
      1 million population                                    
    3 County/city-10,001 to       1,716         87         65             152 
      100,000 population                                      
    4 County/city-1 to              648         33          0              33 
      10,000 population                                       
    5 MCD-greater than 1              0          0          0               0 
      million population                                      
    6 MCD-100,001 to 1               25          1          7               8 
      million population                                      
    7 MCD-10,001 to 100,000         673         33         20              53 
      population                                              
    8 MCD-1,001 to 10,000         3,053        155          0             155 
      population                                              
    9 MCD-301 to 1,000            2,341        119          0             119 
      population                                              
10 MCD-0 to 300                1,585         83          0              83 
      population                                              
      Total                      10,491        533        255             788 

Source: GAO analysis of U.S. Census Bureau data for the population counts
and GAO for sample allocations.

As indicated above, GAO constructed an initial sample frame for each state
group using 2000 decennial census data. To check the comprehensiveness and
correctness of this list, information about election jurisdictions was
obtained from state election Web sites in nearly all states. In making
these checks, we determined that the census data did not coincide with all
local election jurisdictions in 7 states-Hawaii, Virginia, Maine, New
Hampshire, Vermont, Minnesota, and Wisconsin. All of these states, except
Wisconsin, had counties or MCDs that were not identified as election
jurisdictions, and Maine and Wisconsin had election jurisdictions that
were not census MCDs. Therefore, we adjusted the original sample frame to
reflect the election jurisdictions in these 7 states as best we could
determine. Additionally, during the course of fielding this survey, we
learned that the  5 counties, or boroughs, that constitute New York City
are not separate election jurisdictions, but are a single election
jurisdiction governed by the New York City Board of Elections. We adjusted
our sample frame based on census data again to reflect this circumstance.
Two states, Minnesota and Michigan, did not provide comprehensive listings
of election jurisdictions on their state election Web sites. However, from
the Minnesota Municipal Clerks and Finance Officers' Association and the
Minnesota Association of Townships, we obtained lists of Minnesota MCD
election officials; and from the Michigan Townships Association Web site,
the Michigan Association of Municipal Clerks, and a listing of Michigan
local government Web sites contained in an online document from the
University of Michigan, we obtained Internet links to Michigan MCDs that
enabled us to identify local election officials in this state. From all of
these data sources, GAO constructed a complete list of local election
jurisdictions in the United States. The information from state election
Web sites and other sources was used also to obtain contact information
for the head official in local election jurisdictions selected to be in
our survey sample. For some local election jurisdictions officials whose
contact information was not readily available, we called or e-mailed state
election officials, county or city hall administration offices, or
jurisdiction membership organizations in order to learn appropriate names
and addresses. A contractor was used to call all jurisdictions in the
sample to confirm or correct, as necessary, this contact information.

We had to make adjustments in how we surveyed the election jurisdictions
in two states-Minnesota and Texas. During questionnaire development, we
learned that election functions in Minnesota are split between countylevel
government and MCDs. For example, registration is handled exclusively by
county officials, and county auditors may delegate other functions, such
as absentee voting, to MCDs. There are some functions that MCDs are
responsible for handling, such as polling place matters. Therefore, we
also surveyed election officials in the 67 Minnesota counties that had
MCDs selected to be in our sample-our sample included no MCDs from the
remaining Minnesota counties. We created separate versions of our
questionnaire for Minnesota county and MCD election officials. GAO staff
reviewed and combined the responses from counties with the responses from
their included MCDs to create a single completed questionnaire per
Minnesota MCD in our sample. Also, during the course of fielding this
survey, we learned that some Texas counties split election
responsibilities between the county clerk and the tax assessor-collector.
In the few Texas counties in our sample with these split responsibilities,
we sent a copy of the main questionnaire to each official, and again used
only a single completed questionnaire for the county.

Our survey period was from mid-August until mid-December 2005. A
contractor was used to make up to two follow-up telephone calls to all
nonrespondents to encourage them to return a completed questionnaire. In
all, we received 632 usable questionnaires for an overall response rate of
 80 percent. The response rates by strata are presented in table 30. As
can be seen, the response rates were high in all strata, with the lowest
response rate being 69 percent in stratum 2. And according to our
groupings of jurisdictions by population size, the response rates for
large jurisdictions was 72 percent, for medium jurisdictions it was 82
percent, and for small jurisdictions it was 83 percent. We believe that
these response rates combined with relatively small sampling errors,
presented below, allow us to describe the issues and challenges
encountered by local election jurisdictions, both in total and by
population size categories, in the 2004 general election with an
acceptable level of precision.

  Table 30: Local Election Jurisdiction Survey Response Rates, Overall and by
                                 Sample Strata

                                       Number of Number of jurisdictions
                                   jurisdictions returning usable
Stratum                               sampled questionnaires Response rate 
    1 County/city-greater than 1 million      28             24           86% 
      population                                                
    2 County/city-100,001 to 1 million       157            108           69% 
      population                                                
    3 County/city-10,001 to 100,000          152            121           80% 
      population                                                
    4 County/city-1 to 10,000 population      33             26           79% 
    5 MCD-greater than 1 million               0              0             0 
      population                                                
    6 MCD-100,001 to 1 million                 8              7           88% 
      population                                                
    7 MCD-10,001 to 100,000 population        53             48           91% 
    8 MCD-1,001 to 10,000 population         155            126           81% 
    9 MCD-301 to 1,000 population            119             96           81% 
10 MCD-0 to 300 population                 83             76           92% 
      Total                                  788            632           80% 

Source: GAO

All sample surveys are subject to sampling error-that is, the extent to
which the survey results differ from what would have been obtained if the
whole population had been observed. Measures of sampling error are defined
by two elements, the width of the confidence intervals around the estimate
(sometimes called the precision of the estimate) and the confidence level
at which the intervals are computed. Because we followed a probability
procedure based on random selections, our sample is only one of a large
number of samples that we might have drawn. As each sample could have
provided different estimates, we express our confidence in the precision
of our particular sample's results as a 95 percent confidence interval
(e.g., plus or minus 5 percentage points). This is the interval that would
contain the actual population value for 95 percent of the samples we could
have drawn. As a result, we are 95 percent confident that each of the
confidence intervals based on the mail survey includes the true values in
the sample population. Unless otherwise noted, the maximum sampling error
for estimates of all jurisdictions is plus or minus  5 percentage points,
plus or minus 7 percentage points for large jurisdictions, plus or minus 7
percentage points for medium population size jurisdictions; and plus or
minus 5 percentage points for small population size jurisdictions.

In addition to the reported sampling errors, as indicated earlier, the
practical difficulties of conducting any survey may introduce errors,
commonly referred to as nonsampling errors. For example, differences in
how a particular question is interpreted, the information sources
available to respondents, or the types of sample members who do not
respond can introduce unwanted variability into the survey results. We
took extensive steps in questionnaire development, data collection, and
the editing and analysis of the survey data to minimize nonsampling
errors. As with the questionnaire for our state survey, social science
survey specialists designed the draft questionnaire for local
jurisdictions in close collaboration with GAO subject matter experts. An
early draft was reviewed by an expert in the election field who is also a
long-time local election administrator. We pretested the questionnaire in
person and by telephone with officials in election jurisdictions in 7
states. From this review and these pretests, we made revisions, as
necessary. We also consulted with election officials in several counties
in Minnesota when developing separate questionnaire versions for Minnesota
counties and MCDs. Our questionnaire was sent to all jurisdictions in
booklet form. A copy of the main mail questionnaire is included in
appendix III. All returned questionnaires were reviewed and edited, and we
called respondents to obtain information for blank responses or where
clarification was needed. For example, many jurisdictions returned a
questionnaire that had two or more adjacent pages left blank. We called
these jurisdictions to determine whether the questions on these pages had
not been answered accidentally or intentionally. We then obtained answers
to these questions from those respondents willing to provide answers.
Also, when necessary, we called Minnesota election officials to resolve
conflicts that occurred when we were combining questionnaire data from
Minnesota MCDs and associated counties into a single completed
questionnaire per MCD. All questionnaire data were double key-entered into
an electronic file in batches (that is, the entries were 100 percent
verified), and a random sample of each batch was selected for further
verification for completeness and accuracy. Computer

                Visits to Selected Local Election Jurisdictions

analyses were also performed to identify any inconsistencies in response
patterns or other indications of errors. All computer syntax was
peerreviewed and verified by separate programmers to ensure that the
syntax had been written and executed correctly.

To obtain a more detailed understanding of the stages of the election
process, challenges associated with it in local jurisdictions, and how
local election officials address those challenges, we visited and
interviewed officials in a nonprobability sample of 28 local election
jurisdictions in  14 states nationwide. We obtained and reviewed
available documentation on the requirements, people, processes, and
technology of election administration within each jurisdiction. Although
the information obtained from the visits to these 28 jurisdictions cannot
be generalized to other local election jurisdictions, these jurisdictions
were chosen based on a wide variety of characteristics, including voting
methods used, geographic characteristics, and aspects of election
administration. Regarding election administration, we sought to have a mix
of jurisdictions where the following varied: registration time frames,
absentee voting requirements (i.e., excuse versus no excuse), whether
early voting was offered, whether voters were required to provide
identification to vote, and whether recounts for federal or state offices
occurred. We did not select jurisdictions we visited on the basis of size,
but as appropriate, we identify the size of a jurisdiction we visited
using the same groupings we used for our nationwide mail survey. The
jurisdictions we visited are shown in  table 31.

                 Table 31: Local Election Jurisdictions Visited

State                                                 Jurisdiction visited 
Colorado                                                    El Paso County 
                                                               Larimer County 
Connecticut                                               City of Hartford 
                                                            City of New Haven 
Florida                                                     Broward County 
                                                                  Leon County 
Georgia                                                   Dougherty County 
                                                              Muscogee County 
Illinois                                                  Champaign County 
                                                              City of Chicago 
Kansas                                                      Johnson County 
                                                             Wyandotte County 
New Hampshire                                              Town of Madbury 
                                                           City of Manchester 
New Jersey                                                Middlesex County 
                                                               Passaic County 
Nevada                                                        Clark County 
                                                                Washoe County 
New Mexico                                               Bernalillo County 
                                                              Santa Fe County 
North Carolina                                             Carteret County 
                                                              Guilford County 
Ohio                                                       Cuyahoga County 
                                                              Mahoning County 
Pennsylvania                                              Allegheny County 
                                                            Montgomery County 
Washington                                                    Clark County 
                                                    King County               

Source: GAO

We visited these 28 local election jurisdictions between July and October
2005.

Page 496 GAO-06-450 Elections

Appendix VI

Voter Registration Deadlines for States and the District of Columbia

The following table presents information on voter registration deadlines
in the 50 states and the District of Columbia as reported by state
election officials in our state survey. In our state survey, we provided
states the deadline they had reported to us for submitting a registration
application for the November 2000 general election and asked if the
deadline for submitting a registration application remained the same or
had changed for the November 2004 general election. As shown in bold text,
three states (Maryland, Nevada, and Vermont) reported that their states
had changed their registration deadlines for the November 2004 general
election.

Table 32: States Reporting on November 2004 General Election Registration
                                   Deadlines

State             2004 general election registration deadline (days before 
                                                                    election) 
Alabama     10 days before election                                        
Alaska      30                                                             
Arizona     29                                                             
Arkansas    30                                                             
California  15                                                             
Colorado    29                                                             
Connecticut 14                                                             

Delaware    20 days prior to a general election and 21 days prior to a     
               primary election                                               
District of 30                                                             
Columbia    
Florida     29                                                             

                         Page 497 GAO-06-450 Elections

Georgia       The fifth Monday before a general primary, general election, 
                 presidential preference primary, or special election         
                 unless special election or special primary is held on date   
                 other than uniform election date, in this case deadline      
                 would be fifth day after the date of the call for the        
                 special primary or election                                  
Hawaii        30-9 October, 1 month prior to election                      
                 Must be delivered by 5 p.m., 10 days before a state primary  
Iowa          or general election, 11 days before all others, or           
                 postmarked 15 or more days before an election                
Kansas        15 days prior to any election                                
Kentucky      No state response to this question                           
Louisiana     30                                                           
Maine         Registration by mail or delivered by a person other than the 
                 voter had to be received by the registrar of voters by       
                 the 10th business day before the election. Registration      
                 permitted in person up to and on Election Day                
Maryland      21 days before the election (the deadline was reduced from   
                 25 days before the election)                                 
Massachusetts 20                                                           

Idaho     25 days before an election if mailed, 24 days for in-person      
             Election Day registration at polling precincts                   
Illinois  28 days before a general and primary election                    
Indiana   29                                                               

Appendix VI Voter Registration Deadlines for States and the District of Columbia

                         (Continued From Previous Page)

    State 2004 general election registration deadline (days before election)

                                  Michigan 30

     Minnesota Delivered by 5 p.m. 21 days before an election. Election Day
                       registration at polling precincts

                                 Mississippi 30

Missouri 28-always fourth Wednesday prior to an election. In some years
this is 28 days, in some years it may be more than 28 days

                                   Montana 30

Nebraska For applicants submitted other than in person (mail, MVA, public
assistance agency, etc.) the deadline is the third Friday prior to
election. For in-person registrations at the local election official's
office, the deadline is 6 p.m. the second Friday before election

Nevada        Deadline for registration by mail (9 p.m. on the fifth       
                 Saturday preceding any primary or general election) stayed   
                 the same. For registration in person the deadline was        
                 lengthened 10 days (9 p.m. on the third Tuesday preceding    
                 any primary or general election).                            
New Hampshire Must be received 10 days before an election. Election Day    
                 registration at polling precincts                            
New Jersey    29                                                           
New Mexico    28                                                           
New York      25                                                           

North Carolina Postmarked 25 days before an election or received 25 days   
                  before an election in the elections office or designated    
                  voter registration agency by 5 p.m.                         
North Dakota   No voter registration                                       
Ohio           No state response to this question                          
Oklahoma       24                                                          

Oregon         Postmarked or received 21 days before an election           
Pennsylvania   30                                                          
Rhode Island   30                                                          
South Carolina 30                                                          
South Dakota   Postmarked 30 days before an election or delivered 15 days  
                  before an election                                          

Tennessee     Postmarked or received 30 days before an election            
Texas         30                                                           
Utah          Postmarked 20 days before an election; in person at a county 
                 clerk's office or at a county clerk designated               
                 "satellite" registration location 8 days before an election  
Vermont       Noon on the second Monday before the election (changed from  
                 the second Saturday before the election)                     
Virginia      28 days prior to the general or primary election             
Washington    30 days before an election or delivered in person up to 15   
                 days before an election at a location designated by the      
                 county elections officer                                     
West Virginia 21 days before the election                                  
Wisconsin     Postmarked or accepted by second Wednesday preceding         
                 election, or completed in the local voter registration       
                 office 1 day before the election. Election Day registration  
                 at polling precincts                                         
Wyoming       30-Election Day registration at polling precincts            
                 Source: GAO 2005 state survey.                               

Appendix VII

Characteristics of Early Voting Jurisdictions Visited

GAO's review of the November 2004 general election included visits to  14
early voting election jurisdictions in 7 states, selected as part of a
nonprobability sample of local election jurisdictions.1 The election
jurisdictions varied widely in their provisions of early voting locations,
equipment, workers, and calendar days and hours. As shown in the table
below, for example, election jurisdictions in Florida, Georgia, Kansas,
and North Carolina had one early voting location, while others had
multiple sites such as 3 in Johnson County, Kansas; 5 in Larimer County,
Colorado; 14 in Broward County, Florida; and 63 in Clark County, Nevada.
The broad range of early voting locations involved the use of as few as
one voting machine in Georgia and North Carolina counties we visited to
several hundred voting machines used in counties visited in New Mexico and
Nevada. The cadre of poll workers staffing early voting sites included as
few as eight or nine workers in Georgia, Kansas, and North Carolina
counties; dozens in jurisdictions such as Larimer County, Colorado;
Bernalillo County, New Mexico; and hundreds in Broward County, Florida;
Clark County, Nevada; and Guilford County, North Carolina. The period for
early voting ranged from 5 days in Georgia to 28 days in New Mexico. Early
voting hours varied extensively, but early voting was generally available
during weekday business hours and sometimes on Saturday and Sunday.

Table 33: Selected Information on Jurisdictions We Visited That Had Early Voting

                                           Locations                           Early voting
                                         with voting                           hours  
                              Locations     machines                           Election/main
                             accessible   accessible                     Early office Other
                     Total    to voters    to voters                    voting locations
                 number of         with         with Voting   Poll    calendar 
State    County  locations disabilities disabilities machines workers     days 
                                                                               8              8 
Colorado El Paso         3            3            3       69      14       15 a.m.-5 a.m.-4:30 
                                                                               p.m.        p.m. 
Colorado Larimer         5            5           0a        5      92       15 8      8 a.m.-5  
                                                                               a.m.-5   p.m.    
                                                                               p.m.   
Florida  Broward        14           14           14      183     220       15 8:30             
                                                                               a.m.-           
                                                                               6 p.m.           
                                                                               M-F;  8:30      
                                                                               10     a.m.- 6  
                                                                               a.m.- p.m. M-F; 
                                                                               3 p.m.  10      
                                                                               Sat.;  a.m.- 3  
                                                                                1    p.m.      
                                                                               p.m.- Sat.;  1
                                                                               4 p.m. p.m.- 4
                                                                               Sun.   p.m. Sun.

1Election responsibility for these jurisdictions resided at the county
level.

Appendix VII Characteristics of Early Voting Jurisdictions Visited

Page 500 GAO-06-450 Elections

(Continued From Previous Page)
                                         Locations                              Early voting hours
                                              with                           
                                            voting                           
                            Locations     machines                           
                   Total accessible   accessible                       Early 
                  number    to voters    to voters                    voting 
                      of         with         with   Voting    Poll calendar Election/main
State   County locations disabilities disabilities machines workers     days office             Other
                                                                                            locations
                                                                             8:30       
Florida Leon           1            1           0a        1      15       15 a.m.-     
                                                                             5:30 p.m.  N/Ab
                                                                             M-F;      
                                                                             9 a.m.-  
                                                                             5 p.m. Sat.; 
                                                                             9 a.m.-  
                                                                             5 p.m. Sun.
Georgia  Dougherty     1            1            1        7       8        5 8:30 a.m.-5  N/Ab        
                                                                             p.m.         
Georgia  Muscogee      1            1            1        9      20        5 8 a.m.-5     N/Ab        
                                                                             p.m.         
Kansas   Johnson       3            3            3        6      24       20 8 a.m.-     11 a.m.-   
                                                                             7 p.m. M-F;  7 p.m. M-F; 
                                                                                                    
                                                                             8 a.m.-     9 a.m.-    
                                                                             5 p.m. Sat.  5 p.m. Sat. 
                                                                             (generally)  (generally) 
Kansas   Wyandotte     1            1           0a        4       8       20 8 a.m.-5     N/Ab        
                                                                             p.m.         
Nevada   Clark        63           63           63      390     347       17 Various      Various     
Nevada   Washoe       11           11           11       85      29       17 8 a.m.-6     8 a.m.-6    
                                                                             p.m.         p.m.        
                                                                             (generally)  (generally) 
New      Bernalillo   13           13           13      283      55      28c     8 a.m.-5 12 p.m.-   
                                                                                 p.m. M-F 
Mexico                                                                       (generally)  8 p.m. T-F; 
                                                                                                     
                                                                                            10 a.m.-
                                                                                           5p.m. Sat.
                                                                                          (generally)
New       Santa Fe                                                           12 p.m.- 12 p.m. -
Mexico                 5            5           0a        5      20      17c 8 p.m. T  8 p.m. T-F; 
                                                                             - F;     
                                                                             10 a.m.-  10 a.m. - 
                                                                             6 p.m.     6 p.m. Sat.
                                                                             Sat.      
North     Carteret     1            1            1       13       9       19 8 a.m.-  N/Ab
Carolina                                                                     5 p.m.    
                                                                             M-F;     
                                                                             8 a.m.-1 p.m. Sat.
North     Guilford    13           13           13      114     116       19 8 a.m.-5    10 a.m.-
                                                                             p.m. M-F    
Carolina                                                                     (generally) 6:30 p.m.
                                                                                         M-F;
                                                                                            10 a.m.-
                                                                                      1, 3, or 4 p.m.
                                                                                     Sat. (generally)

Source: GAO interviews of local election jurisdiction officials.

aTo vote early, voters with disabilities completed a paper ballot which an
election worker fed into a precinct count optical scan machine.

bN/A-not applicable because jurisdiction had only 1 location for early
voting.

cNew Mexico statute provides that a voter may vote early up to 28 days
prior to an election at the office of the county clerk or at an alternate
location.

Appendix VIII

Selected State Statutory Requirements for Absentee Voting

This appendix presents selected state statutory requirements for absentee
voting. Table 34 summarizes certain mail-in absentee voting requirements
in the 50 states and the District of Columbia. The statutory changes are
in bold to highlight differences between requirements in the November 2000
and 2004 general elections, including excuse requirements and notary or
witness requirements. Table 35 summarizes the deadlines for returning
mail-in absentee ballot applications and absentee ballots in the 50 states
and the District of Columbia. Some of the states reporting that an excuse
was required to vote absentee also reported providing voters with other
means to vote prior to Election Day without requiring an excuse. According
to our state survey, of the 30 states that reported requiring an excuse to
vote absentee, 8 states reported that they offered voters the option of
early voting without requiring an excuse. Oregon reported, on our state
survey, that it offered all-mail voting.

  Table 34: Selected State Requirements for Domestic Absentee Voting, November
                        2000 and 2004 General Elections

                         Page 501 GAO-06-450 Elections

                            Excuse required        Witness or notary required
State                 2000       2004           2000                  2004 
Alabama               Yes        Yes             Yes              Yes      
Alaska                 No        Yes             Yes              Yes      
Arizona                No        Yes                      No            No 
Arkansas              Yes        Yes                      No            No 
California             No         No                      No            No 
Colorado               No         No                      No            No 
Connecticut           Yes        Yes                      No            No 
Delaware              Yes        Yes                      No            No 
Florida               Yes         No             Yes                    No 
Georgia               Yes        Yes                      No            No 
Hawaii                 No         No                      No            No 
Idaho                  No         No                      No            No 
Illinois              Yes        Yes                      No            No 
Indiana               Yes        Yes                      No            No 
Iowa                  Yes         No                      No            No 
Kansas                 No         No                      No            No 
Kentucky              Yes        Yes                      No            No 
Louisiana             Yes        Yes             Yes              Yes      
Maine                      No     No                      No            No 

Appendix VIII Selected State Statutory Requirements for Absentee Voting

Page 502 GAO-06-450 Elections

(Continued From Previous Page)
                               Excuse required     Witness or notary required
State                     2000      2004        2000                  2004
Maryland                Yes          Yes               No               No
Massachusetts           Yes          Yes               No               No
Michigan                Yes          Yes               No               No
Minnesota               Yes          Yes         Yes            Yes
Mississippi             Yes          Yes         Yes            Yes
Missouri                Yes          Yes         Yes            Yes
Montana                        No        No            No               No
Nebraska                       No        No            No               No
Nevada                         No        No            No               No
New Hampshire           Yes          Yes               No               No
New Jersey              Yes          Yes               No               No
New Mexico                     No        No            No               No
New York                Yes          Yes               No               No
North Carolina          Yes              No      Yes            Yes
North Dakota                   No        No            No               No
Ohio                    Yes          Yes               No               No
Oklahoma                       No        No      Yes            Yes
Oregon                         No    Yes               No               No
Pennsylvania            Yes          Yes               No               No
Rhode Island            Yes          Yes         Yes            Yes
South Carolina          Yes          Yes         Yes            Yes
South Dakota            Yes              No            No               No
Tennessee               Yes          Yes               No               No
Texas                   Yes          Yes               No               No
Utah                    Yes              No            No               No
Vermont                 Yes              No            No               No
Virginia                Yes          Yes         Yes            Yes
Washington                     No        No            No               No
West Virginia           Yes          Yes               No               No
Wisconsin                      No        No      Yes            Yes
Wyoming                        No        No            No               No
District of Columbia    Yes          Yes               No               No
Total                    33   18    30 21         13 38            12   39 

Source: GAO 2005 survey of state election officials and GAO, Elections:
Perspectives on Activitiesand Challenges Across the Nation, GAO-02-3
(Washington, D.C.: October 2001).

Information in bold shows states with changes in excuse or witness or
notary requirement.

Appendix VIII Selected State Statutory Requirements for Absentee Voting

Table 35: States' Mail-in Absentee Ballot Application and Mail-in Absentee
Ballot Deadlines for Inside the United States, November 2004 General
Election

                         Page 503 GAO-06-450 Elections

State         Mail-in absentee ballot        Mail-in absentee ballot       
                 application deadline           deadline                      
Alabama       5 days before Election Day     Election Day and postmarked   
                                                by day prior to Election      
                                                Day                           
Alaska        7 days before Election Day     10 days after Election Day    
                                                and postmarked by             
                                                Election Day                  
Arizona       11 days before Election Day    Election Day by 7 p.m.        
Arkansas      7 days before Election Day     Election Day no later than    
                                                7:30 p.m.                     
California    7 days before Election Day     Election Day by close of      
                                                polls                         
Colorado      11 days before Election Day    Election Day by 7 p.m.        
Connecticut   Election Day                   Election Day by close of      
                                                polls                         
Delaware      1 day before Election Day      1 day before Election Day by  
                                                noon                          
Florida       No deadline                    Election Day by 7 p.m.        
Georgia       4 days before Election Day     Election Day by 7 p.m.        
Hawaii        7 days                         Election Day by close of      
                                                polls                         
Idaho         6 days before Election Day     Election Day by 8 p.m.        
Illinois      5 days before Election Day     Election Day by close of      
                                                polls.                        
                      Varies by county, type of In time for the board to      
Indiana          voter, and delivery method, deliver the ballot to the     
                                            but 
                 earliest deadline is 8 days    precinct on Election Day      
                 before Election Day            
                                                Either (1) Election Day,      
Iowa          4 days before Election Day     received by close of polls,   
                                                or                            
                                                (2) postmarked no later than  
                                                the day before close of       
                                                polls and received not later  
                                                than 12 p.m. on the           
                                                Monday following the election 
Kansas        4 days before Election Day     Election Day by close of      
                                                polls                         
Kentucky      7 days before Election Day     Election Day by close of      
                                                polls                         
Louisiana     4 days before Election Day     1 day before Election Day     
Maine         Election Day                   Election Day by close of      
                                                polls                         
Maryland      7 days before Election Day     1 day after Election Day by 4 
                                                p.m. if postmarked            
                                                before Election Day           
Massachusetts 1 day before Election Day      Election Day by the close of  
                                                polls                         
Michigan      3 days before Election Day     Election Day by close of      
                                                polls                         
Minnesota     1 day before Election Day      Election Day                  
Mississippi   1 day before Election Day      1 day before Election Day by  
                                                5 p.m.                        
Missouri      6 days before Election Day     Election Day by close of      
                                                polls                         
Montana       1 day before Election Day      Election Day by close of      
                                                polls                         
Nebraska      6 days before Election Day     Election Day by close of      
                                                polls                         
Nevada        7 days before Election Day     Election Day by close of      
                                                polls                         
New Hampshire No deadline                    Election Day by 5 p.m.        

Appendix VIII Selected State Statutory Requirements for Absentee Voting

(Continued From Previous Page)
State        Mail-in absentee ballot         Mail-in absentee ballot       
                application deadline            deadline                      
New Jersey   Election Day                    Election Day by the closing   
                                                of the polls                  
New Mexico   7 days before Election Day      Election Day by 7 p.m.        
New York     7 days before Election Day      7 days after Election Day and 
                                                postmarked before             
                                                Election Day                  
North Carolina 7 days before Election Day    1 day before Election Day by  
                                                5 p.m.                        
North Dakota   1 day before Election Day     Postmarked by midnight of the 
                                                          day before Election 
                                                 Day (if no postmark legible, 
                                                    must be received within 2 
                                                days after Election Day)      
Ohio            Election Day                 Election Day by close of the  
                                                polls                         
Oklahoma        6 days before Election Day   Election Day no later than 7  
                                                p.m.                          
Oregon          No deadline                  Election Day not later than 8 
                                                p.m.                          
Pennsylvania    7 days before Election Day   4 days before Election Day by 
                                                5 p.m.                        
Rhode Island    21 days before Election Day  Election Day not later than 9 
                                                p.m.                          
South Carolina  4 days before Election Day   Election Day by closing of    
                                                the polls                     
South Dakota    Election Day                 Election Day                  
Tennessee       5 days before Election Day   Election Day by close of      
                                                polls                         
Texas           7 days before Election Day   Election Day by close of      
                                                polls                         
Utah            20 days before Election Day  6 days after Election Day and 
                                                postmarked before             
                                                Election Day                  
Vermont       1 day before Election Day      Election Day                  
Virginia      5 days before Election Day     Election Day                  
Washington    1 day before Election Day      Postmarked no later than      
                                                Election Day                  
West Virginia 6 days before Election Day      Either (1) received Election 
                                                     Day by close of polls or 
                                                (2) postmarked no later than  
                                                Election Day and              
                                                received before canvassing    
                                                begins (generally the         
                                                fifth day after a general     
                                                election)                     
Wisconsin        4 days before Election Day  Election Day                  
Wyoming          1 day before Election Day   Election Day by 7:00 p.m.     
District of      7 days before Election Day     10 days after Election Day 
Columbia                                          and postmarked not later 
                                                than Election Day             

Source: GAO 2005 survey of state election officials.

Note: All of the reported state mail-in ballot deadlines remained the same
for the November 2000 and 2004 general elections except for Nebraska.
Nebraska reported that the mail-in absentee ballot deadline for the
November 2000 general election was no later than 2 days after Election
Day.

Appendix IX

State Provisions for Accessibility of Polling Places and Alternative Voting
Methods

State Provisions Concerning Accessibility for the 2000 General Election

This appendix presents additional information on the accessibility of
polling places and alternative voting methods.

In October 2001, we issued a report that examined state and local
provisions and practices for voting accessibility, both at polling places
and with respect to alternative voting methods and accommodations. For
that report, our analysis included a review of state statutes,
regulations, and written policies pertaining to voting accessibility for
all 50 states and the District of Columbia. We noted that all states and
the District of Columbia had laws or other provisions concerning voting
access for individuals with disabilities, but the extent and manner in
which these provisions addressed accessibility varied from state to
state.1 Table 36 presents state provisions concerning the accessibility of
polling places, and table 37 presents provisions related to alternative
voting methods, as reported in our October 2001 report.

1For more information, see GAO, Voters with Disabilities: Access to
Polling Places and Alternative Voting Methods, GAO-02-107 (Washington,
D.C.: Oct. 15, 2001).

Page 505 GAO-06-450 Elections Appendix IX State Provisions for
Accessibility of Polling Places and Alternative Voting Methods

Table 36: State Provisions Concerning    
Accessibility of Polling Places          
                                            Number of states with provisions  
                                                  Statute or Number of states 
State provisions                           regulation Policy onlya with no 
                                                                    provision 
Polling place accessibility              

All polling places must/should be accessible                      36 7   8 
State provisions contain one or more polling place accessibility  23 19  9 
standards                                                               
Inspection of polling places to assess accessibility is required  15 14 22 
Reporting by counties to state on polling place accessibility is  10 10 31 
required                                                                
Voting booth areas and equipment                                        
Voting booth areas must/should accommodate wheelchairs            17 16 18 
Voting systems must/should accommodate individuals with           13 11 27 
disabilities                                                            
Aids for visually impaired voters                                       
Braille ballot or methods of voting must/should be provided       3  3  45 
Ballots with large type must/may be provided                      2  2  47 
Magnifying instruments must/may be provided                       7  15 29 

Source: GAO-02-107, p. 17; GAO analysis of statutes, regulations, and
other written provisions in 50 states and the District of Columbia.
Provision categories were identified based on our review of these legal
and policy documents.

aPolicies for a particular provision were identified only if a state did
not have either a statute or a regulation for that provision.

Table 37: State Provisions for Alternative Voting Methods and
Accommodations                          
                                                             Number of states
Methods and accommodations              Permitting No provision Prohibited 
Absentee voting by mail                         51            0          0 
Permanent absentee voting                       17           34          0 
Curbside voting on Election Day                 28           19          4 
Ballot can be taken to voter's                  21           25          5 
residence on or before Election Day                             
Early voting                                    39           12          0 

Source: GAO-02-107, p. 21; GAO analysis of statutes, regulations, and
other written provisions provided by election officials in 50 states and
the District of Columbia.

Appendix IX State Provisions for Accessibility of Polling Places and Alternative
                                 Voting Methods

Table 38 presents information from our 2005 survey of state election

State Provisions For

officials about provisions concerning accessibility and accommodations for
individuals with disabilities for the November 2004 general election.

General Election

Table 38: State Provisions Concerning Accessibility of Polling Places and
Accommodations for Individuals with Disabilities for the November 2004
General Election

                                         Not       Not        Not Required or 
Provision        Required Allowed allowed addressed applicable     allowed 
Polling place                                                              
accessibility          41       6       0         2         1b          47
standardsa                                                     
Inspections of                                                             
polling place          28      16       0         6         1b          44
accessibility                                                  
Reporting by                                                               
local                  32       8       0         9        2 b          40
jurisdictions to                                               
the state                                                      
on polling place                                               
accessibility                                                  
Accommodation of                                                           
wheelchairs in         39       4       0         7         1b          43
voting                                                         
areas                                                          
Curbside voting                                                            
available on           17      13      18         2         1b          30
Election Day                                                   
Ballot can be                                                              
taken to voter's        5      20      18         8          0          25
residence on                                                   
or before                                                      
Election Day                                                   
Notification of                                                            
voters of any          16      11       1        16        6 b          27
inaccessible                                                   
polling placesc                                                
Provision of                                                               
ballot or               1      13       2        33          1          14
methods of                                                     
voting in                                                      
Brailled                                                       
Provision of                                                               
ballots with            5      17       3        26          0          22
large type                                                     
Provision of                                                               
magnifying              8      34       0         7         1b          42
instrumentsd                                                   

Source: GAO 2005 survey of state election officials.

aElection officials in one state responded that they did not know.

bOregon conducts Election Day voting by mail; thus, provisions for polling
place accessibility are not applicable. cElection officials in one state
did not respond to this question. dElection officials in one state did not
respond to this question.

Appendix X

States and the District of Columbia Reported Requirements for Local
Jurisdictions to Use Federal Standards for Voting Systems

The following table summarizes reported state requirements for compliance
with voluntary federal standards for voting equipment that will be used
for the first time in the 2006 general election in the 50 states and the
District of Columbia. In our survey of state election officials, we asked
states to identify from a list of federal standards or guidelines any or
all that they would require local jurisdictions to comply with for new
voting equipment for the November 2006 general election. According to our
state survey, states are requiring local jurisdictions to apply a variety
of federal standards to their voting systems.

Table 39: States and the District of Columbia Reported Requirements for
Use of Voluntary Federal Standards for New Voting Systems for the November
2006 General Election

                              Hybrid of one or   Will use federal    
                              more of 1990 and   standards/guidelines 
      1990 FEC 2002 FEC  Draft EAC  2002 FEC   but version was not  Not yet   
                                     standards                         
State standards standards guidancea   and EAC     specified             determined 
                                      guidance                          

Alabama                                                                  x 
Alaska                              x                      
Arizona               x             x      x               

                                   Arkansas x

California             x       x                                         x 
Colorado                       x                     
Connecticut            x       x                     

Delawareb                                x                       x
Florida                                            x     
Georgia                                            x     
Hawaii                                                           x
Idaho                          x                         
Illinois                       x                                 x
Indiana                                            x     
Iowa                           x                         
Kansas               x         x         x               
Kentucky                       x                                         x 
Louisiana                                          x              

Maine                                                                    x 
Maryland                             x                      
Massachusetts           x            x      x               
Michigan                             x                      
Minnesota                            x      x               

Appendix XStates and the District of Columbia Reported Requirements for
Local Jurisdictions to Use Federal Standards for Voting Systems
(Continued From Previous Page)
Hybrid of one or Will use federal more of 1990 and standards/guidelines
 2002 FEC  Draft EAC  2002 FEC standards  but version was not Not
yet State standards standards guidancea and EAC guidance specified
determined
Mississippi x
Missouri x
Montana x
Nebraska x
Nevada x
New Hampshireb
New Jersey x x
New Mexico x
New York x
North Carolina x
North Dakota x
Ohio x
Oklahomab
Oregon x x
Pennsylvania x

Rhode Island

South Carolinac

South Dakota x
Tennessee x
Texas x x
Utah x
Vermont
Virginia x x
Washington x x
West Virginia x
Wisconsin x
Wyoming x
District of Columbia x
Totals 1129 5 10 6 5

Source: GAO 2005 survey of state election officials.

aAt the time of our 2005 state survey, the proposed EAC guidelines were
open to public comment for a period of 90 days. On December 13, 2005, EAC
adopted the 2005 Voluntary Voting System Guidelines, which will take
effect in December 2007.

bThree states (Delaware, New Hampshire, and Oklahoma) indicated in their
response that they do not require that local jurisdictions comply with
federal standards or guidance.

cSouth Carolina did not respond to any of the response categories.

Appendix XI

                        GAO Contacts and Acknowledgments

For questions regarding this report, please contact Norman Rabkin at .

Contacts

(202) 512-8777 ([email protected]), Bill Jenkins at (202) 512-8757
([email protected]), or Randolph C. Hite at (202) 512-3870
([email protected]).

Additional staff making major contributions to this report were .

Acknowledgments

Linda Watson, Project Manager and Assistant Director; David Alexander;
Pille Anvelt; Orin B. Atwater; Carl Barden; Amy Bernstein; Susan
Bernstein; Fred Berry; Stefanie Bzdusek; Grace Coleman; Willie Commons;
John Dale; Lester Diamond; Neil Doherty; Michele Fejfar; Alice Feldesman;
 Joanne Fiorino; Dan Garcia; Dan Gordon; Rich Griswold; Brandon Haller;
Geoffrey Hamilton; Mike Holland; Valerie Hopkins; Ashfaq Huda;  Mary
Catherine Hult; Cathy Hurley; Sairah Ijaz; Monica Kelly;  Mohammad Khan;
George Kovachick; Ron La Due Lake; Frank Maguire; Matthew R. Michaels; Jan
Montgomery; Paula Moore; Jean Orland; Brandon S. Pettis; Tomas Ramirez;
Emmy Rhine; Amy Sheller; Jay Spaan; Derek Updegraff; Ashanta Williams; and
Ellen Wolfe.

We gratefully acknowledge the substantial time and cooperation of the
election officials whom we visited in the following locations: Allegheny
County, Pennsylvania; Bernalillo County, New Mexico; Broward County,
Florida; Carteret County, North Carolina; Champaign County, Illinois;
Chicago, Illinois; Clark County, Nevada; Clark County, Washington;
Cuyahoga County, Ohio; Dougherty County, Georgia; El Paso County,
Colorado; Guilford County, North Carolina; Hartford, Connecticut;  New
Haven, Connecticut; Johnson County, Kansas; King County, Washington;
Larimer County, Colorado; Leon County, Florida; Madbury, New Hampshire;
Mahoning County, Ohio; Manchester, New Hampshire; Middlesex County, New
Jersey; Montgomery County, Pennsylvania; Muscogee County, Georgia; Passaic
County, New Jersey; Santa Fe County, New Mexico; Washoe County, Nevada;
and Wyandotte County, Kansas.

Glossary

This glossary is provided for reader convenience. It is not intended as a
definitive, comprehensive glossary of election-related terms.

Absentee and Early Voting These are programs that, in general, permit
eligible persons to vote prior to Election Day. Absentee voting is
generally conducted by mail in advance of Election Day and early voting is
generally in-person voting in advance of Election Day at specific polling
locations.

Acceptance Testing Acceptance testing is the examination of voting systems
and their components by the purchasing election authority in a
simulated-use environment to validate performance of delivered units in
accordance with procurement activities.

Audit Testing Audit testing reviews and reconciles election records to
confirm correct conduct of an election or uncover evidence of problems
with voting equipment or election processes.

Ballot This is the official presentation of all of the contests to be
decided in a particular election.

Canvassing This is the process of reviewing votes by precinct, resolving
problem votes, and counting all types of votes (including absentee and
provisional votes) for each candidate and issue on the ballot and
producing an official total for each.

      Central Count Tabulation Ballots are counted at a central location.

Certification Certification is the point in the election process at which
the vote count is finalized and made official. There are generally two
stages of the certification process for statewide elections: First, the
local election jurisdiction certifies the vote count to the state, and
second, the state certifies the final vote count.

Page 511 GAO-06-450 Elections

Certification Testing Certification testing validates the compliance of
the voting equipment with federal standards or state-specific requirements
prior to (or as a condition of) system acceptance. Under NASED, national
testing was called qualification testing.

Computerized Statewide The Help America Vote Act of 2002 requires most
states to implement a single, uniform, centralized, and interactive
computerized statewide voter

Voter Registration List

registration list to serve as the official voter registration list for the
conduct of all elections for federal office in each such state.

Contested Elections These generally involve either an administrative or a
judicial process specified under state provisions whereby certain
individuals (e.g., a voter, a group of voters, or candidates) initiate an
action challenging an election on the basis of certain state-specified
reasons that may include, for example, ineligibility of the person
elected, illegal votes, fraud, or errors by election officials.

Direct Recording Electronic (DRE)

There are two types of DREs, push-button and touch screen. For pushbutton
machines, voters press a button next to the candidate's name or ballot
issue, which then lights up to indicate the selection. Similarly, voters
using touch screen DREs make their selections by touching the screen next
to the candidate or issue, which is then highlighted. When voters are
finished on a push-button or a touch screen DRE, they cast their votes by
pressing a final "vote" button on the machine or screen.

Election Administration This is the people, processes, and tasks
associated with registering voters and preparing for and conducting
elections.

                      Election Assistance Commission (EAC)

HAVA established this commission with wide-ranging duties to help improve
the administration of federal elections. The commission is to be involved
with, among other things, providing voluntary guidance to states on
implementing certain HAVA provisions, serving as a national clearinghouse
of state experiences implementing such guidelines and in the operation of
voting systems in general, conducting studies, and helping to develop
optional testing and standards for election equipment.

Election Day Activities Activities carried out on Election Day include
opening and closing polling places, verifying voter qualifications,
assisting voters in casting their ballots, resolving problems that may
arise during the day, and safeguarding the ballots.

Election Day Preparation Tasks carried out in preparation for Election Day
include recruiting and training poll workers, arranging for polling
places, educating voters, designing ballots, and preparing voting
equipment for use in casting and tabulating votes.

                                 Election Fraud

Election fraud includes conduct that corrupts the electoral process for 

(1)
           registering voters; (2) obtaining, marking, or tabulating ballots;
           or 

(3)
           canvassing and certifying election results. Types of fraudulent
           conduct may include, among other things, voting by ineligibles,
           voting more than once, voter impersonation, intentional disruption
           of polling places either physically or by corrupting tabulating
           software, or destroying ballots or voter registrations.

Election Jurisdictions These are counties, cities, townships, and villages
that have responsibility for election administration.

Election Management This is a system that integrates the functions
associated with readying votecasting and -tallying equipment for a given
election with other election

System

management functions.

Federal Election In 1975, Congress established FEC to administer and
enforce the Federal Election Campaign Act-the statute that governs the
financing of federal

Commission (FEC)

elections. To carry out this role, FEC discloses campaign finance
information; enforces provisions of the law, such as limits and
prohibitions on contributions; and oversees the public funding of
presidential elections.

Federal Voting Equipment Standards

In 1990, FEC issued voluntary federal voting equipment standards that
identified minimum functional and performance requirements for electronic
voting systems and specified test procedures to be used to ensure that
voting equipment met these requirements. FEC later revised the standards
in 2002 to address new or additional functional and technical voting
system capabilities. With the passage of HAVA in 2002, the responsibility
for issuing standards passed to EAC. In December 2005, EAC issued the
Voluntary Voting System Guidelines, which include additions and revisions
for system functional requirements, performance characteristics,
documentation requirements, and test evaluation criteria for the national
certification of voting systems.

Integration Testing Integration testing determines whether vote-casting,
tallying, and election management systems function together for the
election. Integration testing is conducted before the election.

Lever Machines Lever machines are mechanical; the ballot is composed of a
rectangular array of levers, which can be physically arranged either
horizontally or vertically. Voters cast their votes by pulling down those
levers next to the candidates' names or ballot issues of their choice.
After voting, the voter moves a handle that simultaneously opens the
privacy curtain, records the vote, and resets the levers.

Minor Civil Divisions (MCD) These are subcounty governmental units, as
designated by the U.S. Census Bureau.

                                  Optical Scan

An optical scan voting system is composed of computer-readable ballots,
appropriate marking devices, privacy booths, and a computerized tabulation
machine. The ballots can vary in size and list the names of the candidates
and the issues. Voters record their choices using an appropriate writing
instrument to fill in boxes or ovals or to complete an arrow next to the
candidate's name or the issue. Optical scan ballots are counted by being
run through a computerized tabulation machine.

Overvotes These are votes for more choices than are permitted for the
contest.

Paper Ballots Voters generally complete their paper ballots in the privacy
of a voting booth and record their choices by placing marks in boxes
corresponding to the candidates' names and the ballot issues. After making
their choices, voters drop the ballots into sealed ballot boxes. Paper
ballots are manually counted and tabulated.

Parallel Testing Parallel testing verifies the accurate performance of
voting equipment through random selection and systematic evaluation of
operational equipment. Parallel testing is conducted during the election.

     Precinct Count Tabulation Votes are cast and counted at the precinct.

                               Provisional Voting

Provisional voting is generally used by states to address certain voter
eligibility issues encountered at the polling place on Election Day. A
provisional ballot cast by an individual with an eligibility issue would
typically not be counted until the individual's eligibility to vote under
state law has been verified. HAVA specifies that states permit individuals
to cast provisional ballots under certain circumstances. These
circumstances include instances when voters assert they are registered in
the jurisdiction in which they desire to vote and that they are eligible
to vote in an election for federal office but their names do not appear on
the polling place registration list.

Punch Card Punch card voting equipment generally consists of a ballot, a
voterecording device that keeps the ballot in place and allows the voter
to punch holes in it, a privacy booth, and a computerized tabulation
device. The voter inserts a machine-readable card with prescored numbered
boxes representing ballot choices into the vote-recording device and uses
a stylus to punch out the appropriate prescored boxes. The ballot must be
properly aligned in the vote-recording device for the holes in the ballot
card to be punched all the way through. Punch card ballots are counted by
being run through a computerized tabulation machine.

Readiness Testing Also referred to as logic and accuracy testing,
readiness testing checks that the voting equipment is functioning
properly, usually by confirming that predictable outputs are produced from
predefined inputs. Readiness testing is typically conducted in the weeks
leading up to Election Day.

Recount

Some states authorize certain persons (e.g., defeated candidates and
voters) to request an election recount under specified circumstances.
These circumstances can include, for example, when there is a tie vote,
when the margin of victory is within a specified percentage or number of
votes, or when inaccuracies in the vote count are alleged. The scope and
method of such recounts can vary to include, for example, partial recounts
of certain precincts, complete recounts of all ballots, machine recounts,
and hand recounts for the office or issue in question. Some states provide
for mandatory (or automatic) recounts that are, for example, mandated in a
specified percentage or number of precincts irrespective of victory
margins, or triggered by a tie vote, or a margin of victory within a
specified percentage or number of votes.

                                  Registration

For the 2004 election, most states and the District of Columbia required
individuals to apply to register prior to Election Day and be registered
with the appropriate local election officials before such individuals
could vote. Some states had same-day registration and allowed individuals
to register to vote at the polls on Election Day. One state, North Dakota,
did not have a voter registration requirement. Voter registration includes
the processes, people, and technology involved in registering eligible new
voters and in compiling and maintaining voter registration lists.

Spoiled Ballot This is a ballot that has been voted but will not be cast.

System Verification Testing System verification testing is testing before
the election to verify that the voting equipment is operating properly on
Election Day.

Page 516 GAO-06-450 Elections

Undervotes Undervotes are votes for fewer choices than permitted, such as
not voting for President. An undervote may or may not be an error. A voter
might have tried to vote for a candidate but failed to mark the ballot
unambiguously or might have chosen not to vote for any candidate for a
particular office.

U.S. Election SystemsThose federal and state constitutional provisions,
statutes, regulations, people, processes, and technology associated with
the framework of, preparation for, and conduct of elections.

Vote Tabulation Vote tabulation is the counting of the ballots cast at the
polling places on Election Day and those cast in person or by mail prior
to Election Day; determining whether and how to count ballots that cannot
be read by the vote-counting equipment; certifying the final vote counts;
and performing recounts, if required.

                                Voter Education

Voter education is essentially education about elections, and the primary
target is the voter. It includes information about how to register, vote
absentee or early, and use the voting method employed in the state or
jurisdiction, as well as information needed to vote on Election Day. Voter
education is usually identified as a function of the election authority.
It may also be fostered by public interest organizations. Political
parties may provide information about candidates, but this type of
information is not included under our definition of voter education.

                                  Voter Intent

The determination of voter intent can be an issue under certain
circumstances, such as when a voter has improperly or unclearly marked his
ballot (for example, with a punch card ballot, not making a clean punch)
and the state provides for resolving the situation by determining the
"intent of the voter." State direction to election officials in making
such a determination can be set out as a general or a specific standard by
which the election official should judge such a ballot.

Voter-Verifiable Paper Trail This is a human-readable printed record of
all of a voter's selections, presented to the voter to view and check for
accuracy.

(VVPT)

Page 517 GAO-06-450 Elections

     Voting Equipment All devices used to display the ballot, accept voter
          selections, record voter selections, and tabluate the votes.

Voting Method The classes or types of machines used in a voting system.
There are five types of voting methods used in U.S. elections: paper
ballot, lever machine, punch card, optical scan, and DRE.

Voting System A voting system is the people, processes, and technology
associated with any specific method of casting and counting votes, such as
optical scan. Technology includes the mechanical, electromechanical, or
electronic equipment, software, firmware, and documentation required to
program, control, and support voting equipment.

GAO's Mission

The Government Accountability Office, the audit, evaluation and
investigative arm of Congress, exists to support Congress in meeting its
constitutional responsibilities and to help improve the performance and
accountability of the federal government for the American people. GAO
examines the use of public funds; evaluates federal programs and policies;
and provides analyses, recommendations, and other assistance to help
Congress make informed oversight, policy, and funding decisions. GAO's
commitment to good government is reflected in its core values of
accountability, integrity, and reliability.

The fastest and easiest way to obtain copies of GAO documents at no cost
is through GAO's Web site (www.gao.gov). Each weekday, GAO posts GAO
Reports and newly released reports, testimony, and correspondence on its
Web site. To

have GAO e-mail you a list of newly posted products every afternoon, go to
www.gao.gov and select "Subscribe to Updates."

                             Order by Mail or Phone

The first copy of each printed report is free. Additional copies are $2
each. A check or money order should be made out to the Superintendent of
Documents. GAO also accepts VISA and Mastercard. Orders for 100 or more
copies mailed to a single address are discounted 25 percent. Orders should
be sent to:

U.S. Government Accountability Office 441 G Street NW, Room LM
Washington, D.C. 20548

To order by Phone: Voice: (202) 512-6000  TDD: (202) 512-2537  Fax:
(202) 512-6061

Contact:

To Report Fraud, Web site: www.gao.gov/fraudnet/fraudnet.htm.

E-mail: [email protected].

Federal Programs Automated answering system: (800) 424-5454 or (202)
512-7470

Gloria Jarmon, Managing Director, [email protected] (202) 512-4400 U.S.
Government Accountability Office, 441 G Street NW, Room 7125 .Relations
Washington, D.C. 20548

Paul Anderson, Managing Director, [email protected] (202) 512-4800.

Public Affairs

U.S. Government Accountability Office, 441 G Street NW, Room 7149 
Washington, D.C. 20548
*** End of document. ***