The EAC’s hearing on Vote Counting and Recounting was quite interesting. The first panel had representatives from Washington and Virginia discussing their recent recounts, as well as a local election official from Sacramento California discussing recounting VVPAT ballots. The Washington and Virginia cases illustrated the great difference among states on election practice and procedures. One of the most interesting differences was in the scope of the recounts – Washington conducted a statewide recount but Virginia only recounted a very small number of precincts were there was clear evidence of problems. Also of interest was the difference in how the two states handle ballots; in Washington, every absentee ballot is hand checked to discern voter intent before it is tabulated. If a voter makes an error, for example, by circling the names instead of bubbling the circle next to the name, the ballot is remade. In Virginia, the election officials do not do anything similar.
The most amusing discussion, however, was a discussion of the time it took to conduct a recount of the Avante voting system in Sacramento, using a VVPAT. A plethora of problems arose during the recounting process and, in the end, it took 127.5 hours to count 114 ballots! This works out to 67 minutes a ballot. Not surprisingly, the county decided to purchase an optical scan system.
The second panel was Doug Chapin and me. My testimony is below.
Good morning. My name is Thad Hall and I am an assistant professor of political science at the University of Utah and principal investigator of the Election Assistance Commission’s (EAC) Vote Count and Vote Recount Project. There are three other primary investigators on this project: Michael Alvarez of Caltech, Kim Brace of Election Data Services, and Doug Chapin of electionline.org. In my testimony this morning, I will describe our effort to study vote counting and recounting, provide an overview of our findings, and discuss the best practice component of our work. My discussion will largely focus on the Count component of the project; Doug Chapin will be discussing the recount and challenge analysis we have done to date.
Our view of the vote counting process has been rather comprehensive in scope and this view was largely influenced by our experiences observing elections and in talking with election officials. My own views were greatly influenced by a trip I took to Austin Texas, where I met with Dana DeBouvoir, the Travis County election administrator. Her election administration practices are predicated on the idea that an election should be viewed as maintaining the chain of custody of all election materials—from pre-election logic and accuracy machine tests to the counting of absentee and precinct-cast ballots. This goal is accomplished by (1) having a product—some piece of paper or report—for every aspect of the election process, (2) having a witness for every product (i.e., a signature from a witness for the production of every product), and (3) securing every product effectively. Should a challenge arise regarding any aspect of the election process, Dana can produce (1) a product showing what was done, (2) witnesses regarding how it was done, and (3) show that the item in question was stored securely.
We have viewed the vote counting process within this broad scope, starting with the printing of ballots and the securing of voting machines before an election and ending with the auditing of the election. Our study focuses on this process for all channels of voting—precinct, absentee, and in-person early voting. We developed a specific data collection instrument that was reviewed by several election officials, election lawyers, and the EAC staff. The instrument was completed at the University of Utah by reviewing state laws and state regulations relating to each activity in the instrument. The draft of each state’s process is being sent to each state for their review to ensure that the final document is complete and reflects any changes in election law made since 2005.
At the outset, I would like to make four observations that arise from our data collection.
- First, the transparency and accessibility of state election laws varies greatly. In some states, it is possible for a citizen to go to the website of the state election official, click a link, and access the state’s election code and regulations. In other states, the election code is not online and several states could not produce easily a copy of their election regulations. Transparency is a hallmark of free and fair elections, and the public should be able to easily know what the rules of the game are for elections.
- Second, there is wide variation in the terms used to describe any given practice in elections; a common set of terms or a Rosetta stone document would be very helpful for everyone in comparing election practices across states.
- Third, some states have clearly updated their election codes in light of changes in voting procedures and technologies. Other states retain election processes and procedures from 60 or 80 years ago but apply them to new technologies. Georgia is an example of a state that radically updated their election code in response to the development of new voting technologies. The processes in statute and regulation related to securing, operating, and counting ballots tracks the technologies used today. By contrast, some other states use lever machine laws from 60 to 80 years ago to govern the use of direct recording equipment (DRE) voting systems.
- Fourth, there is great variation among states in their use of regulations to govern elections and provide greater guidance—guidance with the force of law—to local election officials regarding how to implement specific activities. For example, Maryland has detailed regulatory guidance regarding how every make and model of voting technology used in the state is to be handled throughout the election process.
I would like to provide a brief overview of our findings to date. Because the data we have collected is currently being reviewed by the states, I am going to avoid focusing on exact numbers but instead will report trends and issues we have identified from our data analysis.
States vary in the type of pre-election processes that they have. Almost all states have procedures for testing tabulation equipment before the election. These tests vary in the level of detail but generally involve casting a pre-audited set of votes on a machine and comparing the tabulation results with the known results. Beyond the tabulation testing process, however, there is little consistency in pre-election processes. For example:
- Only half of all states require tabulation equipment to be sealed after the test and few states explicitly require precinct count optical scan tabulators to be sealed after the pre-election tabulation test is completed. Often, machine sealing rules date back to the days of lever machines, creating requirements that are not appropriate for today’s machines.
- Most states have no laws or rules governing the production and securing of ballots upon delivery from a printer, or after being printed in-house.
- Often, the rules governing the security of voting machines and vote tabulation machines prior to the opening of the polls date back to the time of lever machines and relate to the delivery of keys in a sealed envelope.
Voting Through Early and Absentee Channels
Absentee voting is one area where almost all states would be expected to have detailed rules governing the process. However, there are again variations in the way in which absentee ballots are treated across states. For example:
- Just over half of all states explicitly state how an absentee ballot is to be secured upon receipt from the voter. (e.g., “when received, the county election board shall cause envelopes containing absentee ballots to be placed in a ballot box, locked with three locks, in a secure place.”)
- Some states transport absentee ballots from the central election office back out to precincts for counting, but other states count such ballots centrally.
- More than half of states allow absentee ballots to be counted before the election day ends, by more than 20 states do not.
- Almost every state provides a detailed set of guidance regarding what is a valid absentee ballot for inclusion in the tabulation process. States do vary greatly on the exact time by which ballots have to be received to be counted, however.
In-person early voting is another area where there is wide variation among states. Specifically, there are varying rules regarding the clarity of how ballots and voting machines are to be handled at the close of each day of voting.
Voting in Precincts
Most states have very clear rules about how ballots and voting equipment are to be treated in precincts. However, when it comes to actually counting ballots, there are variations. One of the places where such variation is seen most starkly is in the area of in-precinct vote reconciliation. In at least 17 states, there are no provisions for reconciling the ballots counted in the precinct against the number of voters who cast ballots. This means that there is no effort made to determine that the number of voters who cast ballots and the number of ballots counted are related. Even in states with such rules, the rules are not equitable across voting system platforms. This problem is not addressed in most states in the canvass process either; more than 40 states have no requirement that the canvass process balance the number of voters voting and the number of ballots cast.
For example, when paper ballots are used (including optical scan ballots), several states have rules that the ballots are counted and if the count exceeds the number of voters who are listed as having cast ballots, all ballots are put back into the ballot box and then ballots are randomly removed until there are no excess ballots. (The excess ballots are then carefully stored and accounted for). Obviously, such a process cannot be done the same way with electronic voting systems and although many states have explicit rules regarding noting discrepancies between the number of ballots cast on an electronic system and the number of voters who cast ballots, no state explicitly listed a process for conducting a similar random reduction in electronic ballots when excess ballots are discovered.
The problem of election accounting also exists in other aspects of the vote count process. For example, at least 12 states have no rules governing the handling of a spoiled ballot when it is returned by the voter to the precinct election official. In these states, voters may be allowed to spoil a ballot—often, a voter is allowed to spoil up to three ballots—but there is no law or rule for accounting for these ballots after they are spoiled, or even a process for formally voiding the ballot.
The key issue on vote counting, obviously, is what constitutes a vote on various voting platforms. Here, we see two types of states. First, there are states with very clear statutory or regulatory language contained in sections entitled “vote definition.” For example, the language might state that a vote on a DRE is:
- the statement of results of votes cast produced by operating its mechanism may be considered the results of the votes cast; or
- the results on the tape from each DRE unit and the memory card.
Many of these states also have long sections in their codes where they have descriptions and illustrations of what constitutes a vote on a paper ballot. This might include having descriptions of every possible combination of markings on an optical scan ballot.
Second, there are states that define a vote much less clearly. For example, several states define a vote as “the standard for all of these voting systems is voter intent.”
I would note here that most states tend to have very clear rules for what it takes to be an eligible write-in candidate and for counting write-in ballots. The same holds true for those states allowing straight-party balloting.
One of the hallmarks of free elections is that the election is transparent to the public. We measured transparency in three aspects of the election. First, we examined if there were provisions for observing pre-election machine testing. Second, we examined who can observe voting in precincts. Third, we considered who can observe vote tabulation. In localities where votes are tabulated in-precinct, these last two items obviously are conflated.
At each point, there is a strong bifurcation between states that allow the public to observe each of these components of the voting process and states that only allow representatives of candidates or political parties to observe such activities. In some states, observers are explicitly not allowed to observe voting, machine preparations, or vote counts. Instead, a very limited number of registered and credentialed challengers are allowed in polling places, but they are not there to observe the conduct but to challenge the credentials of voters to cast ballots. In these states with restrictive laws, there is often not even an exception for the media to observe the election. In other states, the law explicitly requires the public to be allowed into polling places to observe the voting process, as well as vote counting and pre-election processes.
One of the areas where there is a small but very important effort to improve vote counting is in the post-election auditing. Most vote tallying today is done by some electronic means, either through the use of DREs or by scanning ballots through an electronic tabulator. To ensure that there are not problems with the system, a small number of states have adopted procedures for auditing the results of the election by comparing a manual tabulation with the electronic tabulation of the same ballots.
These manual recount procedures allow jurisdictions to have some increased level of confidence that the tabulation process is accurate. In addition to the manual counting of paper ballots and comparing the totals to the machine tabulation totals, paper audit trails that many DREs produce allow for the audit receipts from a sample of machines to be compared with the electronic totals produced by the same machines.
The post-election auditing procedures differ from the canvass of results at the local level. In many cases, the canvass can also audit election results that seem to be in error, but many states do not allow ballots to be re-tabulated after the initial tabulation is conducted.
Best Practices in Counting
The best practice component of our project is being guided by two streams of work. First, the Government Accountability Office has published a series of works on best practices and the methodology that is most effective in identifying best practices. Second, there are also international principles for best practices for vote counting that have been elaborated. The Administration and Cost of Elections (ACE) Project (www.aceproject.org) – a collaboration of IFES, the International Institute for Democracy and Electoral Assistance (IDEA), and the United Nations Department of Economic and Social Affairs (UNDESA) – has identified eight (8) guiding principles for vote counting. The Project states:
to establish and maintain public confidence in the electoral process, vote counting systems and procedures should incorporate the fundamental principles of vote counting in a democratic election.
These “guiding principles” are: transparency, security, professionalism, accuracy, secrecy, timeliness, accountability, and equity.
Using data from a related international election project, the Election Process Information Center (EPIC) Project, we can see that many aspects of these goals for counting and recounting are also norms for activities in Organization for Economic Cooperation and Development (OECD) nations. Key findings from in this overview include:
o Most nations require that the number of ballots counted equal the number of voters listed in the poll books.
o Most nations have explicit requirements for transporting sealed and secured ballot containers.
o Computers and electronic tabulation are used in more than half of OECD countries in the initial tabulation process.
o Twenty-five (25) of the twenty-six (26) OECD nations studied allow for recounts; most commonly, nations allow for recounts by request (allowed for in 13 OECD nations).
o Five (5) OECD nations have rules requiring that all ballots cast to be recounted in all elections to ensure an accurate initial count at the polling place or count center.
We are currently completing the best practice component of our work. Our goal is to identify a limited number of highly transferable practices that address the issues identified in the international guiding principles for vote counting and recounting.
To highlight a couple of strong practices that currently exist, Oregon has a very flexible yet effective mechanism for handling ballot security. They require all jurisdictions to submit a security plan to the state that encompasses all aspects of their election process, from ballot printing to ballot tabulation. Likewise, Washington State has developed a new but comprehensive procedure for post-election accounting of ballots that allows local governments to ensure that every ballot is accounted for. New technologies and practices are also being used in Washington to facilitate this accounting process.
Over the next two weeks, we will be finalizing our work on this project and look forward to getting feedback from you and your stakeholder organizations regarding the work products we will be presenting you.