The following testimony was presented by Verified Voting President Pamela Smith to the New Hampshire House Election Law Committee on January 21, 2014.
No voting system is perfect. Nearly all elections in New Hampshire, as in most of the nation, are counted using electronic vote counting systems. Such systems have produced result-changing errors through problems with hardware, software and procedures. Error can also occur when compiling results. Even serious error can go undetected if results are not audited effectively.
In a municipal election in Palm Beach County, Florida in 2012 a “synchronization” problem with the election management software allotted votes to both the wrong candidate and the wrong contest; this was uncovered during a post-election audit. The results were officially changed after a public hand count of the votes.[1. Jaikumar Vijayan, E-voting System Awards Election to Wrong Candidates in Florida Village, COMPUTERWORLD, (Apr. 4, 2012).] Particularly noteworthy about that example is the fact that Florida has one of the nation’s weakest audit provisions; even so, it enabled the discovery of this critical error. In another state, a software malfunction caused thousands of votes to be added to the total. A manual audit revealed the mistake and officials were able to correct the results and avoid a costly run-off election.[1. Emilie Rusch, Scanner Glitch Blamed for Election Miscounts, RAPID CITY JOURNAL., June 3, 2009.] In a Republican primary in Iowa, a manual check of the physical ballots revealed a programming error that was attributing votes to the wrong candidates. Thanks to the manual audit, the correct person was seated in office.[1. Tim Rohwer, Faulty Voting Machines Delay Results; Counting Under Way, THE DAILY NONPAREIL ONLINE (June 7, 2006).]
These are examples of issues uncovered by and resolved because of audits. With any such example of malfunction or error that can cause the wrong outcome to be reported, regular audits help find and correct for those errors. This is because the voter-verified paper ballot is the record of the voter’s intent, and by reviewing some subset of those ballots and comparing the physical ballots with the machine reported totals, the will of the voters can be accurately construed without relying exclusively on the ability of sometimes fallible equipment to interpret that intent. The purpose of the paper ballot is two-fold: to enable voters to check that their votes were marked correctly, and to serve as a tool to election officials to demonstrate that all the votes were counted accurately.
In states that lack a voter-verified paper ballot, anomalous-seeming outcomes can occur that cannot be explained, and which also cannot be audited nor recounted. In states that have paper ballots but fail to use the voter-verified paper ballot to check the system’s correct functioning, that benefit is greatly diminished.
The fact that New Hampshire voters enjoy the greater reliability of a paper-ballot voting system is partially undermined by the lack of a regular post-election check on the accurate functioning of the system. In a recent report, we rated New Hampshire in the top six states (with a rating of “good”) for election preparedness based on several factors. Had New Hampshire an audit provision in place, it would have been the only state to receive a score of “excellent” in our report.[1. Report: Counting Votes 2012: A State by State Look At Election Preparedness]
Post election audits can:
• reveal when recounts are necessary to verify election outcomes; and, by extension, help reduce the need for recounts;
• find error, whether accidental or intentional;
• deter fraud;
• provide for continuous improvement in the conduct of elections; and
• promote public confidence in election outcomes.
Pre-election testing is often cited as a tool that potentially could serve instead of audits as a check on the system. Pre election testing, also known as “logic and accuracy” testing, or L&A, has enormous value for several purposes, not least of which is proofing the ballot (checking for correct ballot layout, that all candidate names are included and spelled correctly, that all contests are included, etc.), as well as checking that the system tabulates as expected. The process generally involves voting a series of known, pre-marked test ballots and comparing the tabulation with the expected outcome. Importantly, however, pre-election testing seeks to predict how a system WILL perform come Election Day. Post-election audits instead examine details about how a system DID perform on Election Day. Audits look at actual ballots and vote counts, not hypotheticals.
Problems can arise – or be introduced – between pre-election testing and the actual vote counting on Election Day, so in addition to routine pre-election testing, routine post-election audits are essential. In our recent report[1. Report: Changes Ahead: A Look at Voting System Testing and Certification, 2013.] about voting system testing and certification, we note that some jurisdictions not only do pre- and post-election L&A testing, but also post-election audits. A county official said “we can tell not only how the system performed in our center prior to the election, but also after it was jostled about on its way to the polling place.”
Audits are not the same as recounts. While audits involve what you might call a manual “recount” of some of the ballots, they differ from recounts in important ways. Audits are nonpartisan; recounts may not be. Recounts can provide redress of a perceived or actual problem, while audits check to ensure the voting system is working, whether there’s an obvious problem or not. Recounts are usually candidate or party-specific, geared toward “who won in this instance” while audits are geared toward making sure the tally is correct, without regard for party or candidate. Recounts may be triggered by a close margin, but without a request or legal requirement for a recount, no checking of the vote totals would occur in the absence of an audit. Audits do not have to be requested.
Most critical business or government processes have some form of quality assurance process, for good reason. These practices can help prevent future problems. Audits are one of the least costly elements in the conduct of elections; estimates range from as low as $0.03 per vote to $0.10 per vote. Audits are arguably one of the most important components in election administration, however. Just as banks regularly conduct audits to make sure all the dollars and decimal points are going where they belong, elections are important enough to us to make sure they work correctly and that the public can have confidence in the results. Post election audits are a tool to help do just that.
Transparent: Audits should be observable by the public, and readily understood by those observers. Transparency through all phases of the audit is desirable. Even in the event no one observed the audit, making the audit observable still demonstrates transparency is a key element of the protocol. More importantly perhaps, failure to make the audit observable may hint at something suspect, if it is being conducted behind closed doors. Another way to put this: “Even if correct, a non-transparent audit would have trouble quelling skepticism and could thereby fail to provide confidence in an election outcome.”[1. Cordero, Dill et al, The Role of Dice in Election Audits, June 2006.]
Verifiably Random: Audits should incorporate verifiably random selection. Some jurisdictions have used a method of selecting using ten-sided dice.[1. See also www.josephhall.org/dicebins.php for ways to make the use of ten-sided dice for audit selection more efficient.] Others have used a “lottery ball” system, and still others have written all precinct numbers onto pieces of paper and then drawn those from a hat. These can all work as long as members of the public have the opportunity to satisfy themselves that in fact no precincts have been left out of the drawing process.
Inclusive: Audits should include all ballot types, that is: polling place ballots, absentee or vote by mail ballots, military and overseas ballots, provisional that have been adjudicated, etc.
Discrepancies: Audits should operate from a protocol which incorporates an advance plan for how to investigate and handle discrepancies, and that plan should be disclosed to the public. We further recommend the establishment of a mechanism for expansion if unexplained discrepancies are found during the audit. We recommend publishing the audit results also, including discrepancies noted and how those were addressed and resolved.
Timing: Ideally, we recommend completing the audit prior to certification, or at least in a manner timed so that any potentially outcome-changing errors can be resolved. Timing of the selection for the audit must occur after all votes are counted and initial results have been made public.
Flexibility: Substantial research is ongoing, on how to improve the efficiency and effectiveness of post-election audits in ways that can minimize the resources necessary to carry them out. We strongly recommend that while initiating an audit process, New Hampshire be open to adapting its protocol to emerging methods –risk limiting, small-batch and ballot-level audits, e.g. – that can provide as great or greater confidence in the accuracy of the outcome with less effort.
Audits are critical to our democratic process. Properly conducted audits offer multiple benefits, not least of which is protecting election officials from unfounded criticism, ensuring accuracy and increasing transparency, reducing disputes and promoting public confidence. Importantly, audits create a feedback loop that enables us to assess what is working and what isn’t.