Back to Board of Elections and Ethics main page
Government and People
Restoring Confidence in the District’s Elections
Preliminary Report and Recommendations of the Council Board of Elections and Ethics Investigation Special Committee
October 8, 2008
Mary M. Cheh, Councilmember, Ward 3, Chairperson
Responding to voting irregularities during the primary election held on September 9, 2008, the Council of the District of Columbia adopted a resolution on September 16, 2008, establishing the Council Board of Elections and Ethics Investigation Special Committee (“Committee”). The charge to the Committee is contained in the resolutions, attached here as Attachments A and B.
On October 3, 2008, the Committee held the first of several planned hearings related to the District’s elections processes and procedures. This report provides preliminary findings of the Special Committee based on both its work and investigation thus far and the testimony received at the Committee’s October 3 hearing.1 The Committee is releasing this preliminary report just days after the hearing so that its recommendations may be considered and implemented in time for the general election scheduled for November 4, 2008.
The Committee thanks the Council Chairman, Vincent Gray, for his leadership in establishing this Committee. The Committee also wishes to acknowledge the pro bono assistance of Jenner & Block LLP, which has already provided substantial guidance to this Committee.
On September 9, 2008, the District of Columbia conducted its primary election. Only 13% of registered voters participated in the primary election which, overall, ran relatively smoothly.
Less than two hours after the polls closed, at approximately 9:37 p.m. the District of Columbia Board of Elections and Ethics (“Board”) released its second, unofficial election-results summary report. That report was based on 134 of 143 precincts reporting, contained obviously inaccurate numbers, and proved to be fundamentally flawed.
This second unofficial report was based on all precincts in Ward 2 reporting, and showed Jack Evans and Cary Silverman, both candidates for the Ward 2 Council seat, as receiving 4,379 votes and 3,097 votes, respectively. Remarkably, the report also showed 1,554 write-in votes, despite the absence of a write-in campaign. By 10 p.m., the Board found itself “embroiled in chaos.”2
Soon afterward, realizing the error, the Board essentially shut down. Rather than provide information to the public and try to explain and defend the situation, the Board permitted an already tense situation to grow worse. Candidates and supporters, recognizing the obvious inconsistencies in the reported results, descended on the Board’s offices and demanded answers. Security did not stop these individuals from flooding into the Board’s offices, though apparently no one was able to get beyond the reception area.
While the situation deteriorated, the Board began reviewing the election results to
locate the source of the anomaly. That anomaly was traced to Precinct 141,
which — notwithstanding a registered voting population of 2,388 —
reported a turnout of 4,759 voters. More specifically, the Board traced the error back to a memory cartridge used
for a “Precinct Count Optical Scan” voting machine in Precinct 141.3 According to the
At 11:10 p.m. and at 11:40 p.m., the Board released its third and fourth unofficial election results summary reports. In those reports, the 1,542 write-in votes in Precinct 141 for the Democratic Party Ward 2 Member of the Council dwindled down to two. Similarly, the vote totals for Jack Evans and Cary Silverman fell from 4,379 votes and 3,097 votes, to 188 votes and 114 votes, respectively. Despite these extreme changes, the Board confidently told the public that the fourth election results summary report, unlike the second report, “accurately reflected the Election results with respect to ballots cast on that day.”5
On Thursday, September 18, 2008, the Board conducted a post-election audit of four precincts.6 The Board reported that the audited precincts were selected randomly, but the selection was conducted outside of the public view. None of the Board members were present at the audit.
Although the Board was under no formal obligation to hold the audit, it had previously agreed to do so. Despite the Board’s willingness to hold a post-election manual audit, according to witness testimony heard by the Committee, the Board provided virtually no notice to the public in advance of the audit and provided little information concerning the conduct of the audit.
The audit was conducted by four teams of three individuals and took two days to complete.7 Several days later, the Board posted the results of the audit on its website. According to the report, most of the election-night results were confirmed by the audit; yet, some of the results could not be reconciled. Other discrepancies were attributed to ballots from two precincts where “voters had brought in pencils with erasers and the marks were evident on the ballots” and where voters “changed their original choice [and] wrote the word ERROR.”
After the election, a substantial amount of speculation began to gather around how the cartridge in question generated the erroneous results. To get answers, the Board turned to the vendor, Sequoia Voting Systems, which provides the District of Columbia with voting machines, software licenses, ballots, training, and other election-related products.
Initially, Sequoia stated that it found “no anomalies or irregularities in either the data or the internal event logs that can be identified as having caused or contributed to the issue experienced election night.”8
After the Board demanded a fuller explanation, Sequoia prepared a report explaining its theory for the irregularities.9 While it was unable to replicate the error, Sequoia claimed that the problems reported on election night were “not a problem with Sequoia’s software or hardware [but rather] a matter of human interaction with a mechanical device . . . and a cartridge with data, combined with a process issue in the [Board] not thoroughly reviewing their reports before releasing to the media.”10 According to Sequoia, “[e]ndemic hardware and software failures [had] been ruled out as the cause.”11
Rather, Sequoia identified four possibilities for failure: (1) a transient malfunction of the memory pack reader (“MPR”), which reads data from the memory cartridge and transmits the data to the vote tally database; (2) improper seating of the cartridge in the MPR; (3) accidental manual ejection of the cartridge prior to completing the cartridge reading process; or (4) electrostatic discharge during the reading or ejection of the cartridge. Moreover, the report indicated that “[n]one of [those] instances would be recorded in the voting system event logs, making it impossible to provide a more definitive answer.”
Thus, though it admitted that it could not identify the cause of the malfunction with certainty, Sequoia opined that the error was the result of a “rush to release results.”
Sequoia recommended that, in order to avoid such “process errors,” staff should report to work later in the day to avoid fatigue, and take “the necessary time to review, cross reference to previous results summaries and make sure everything is in proper order” before releasing the election results to the public.
After the primary election, the Board initiated an internal review of the reported irregularities. In that review, the Board found that, “in its zeal to meet the expectations of the speedy release of unofficial Election Night results, [it] erred by deviating from its standard practice and failing to properly review the erroneous data contained in Report No. 2 prior to its release.”12
The Board also found that because it processed the cartridges at a rate of 10.3 cartridges per minute, and because the Board “unwarrantedly relied upon” the indication that data had been successfully uploaded, it failed to notice the erroneous election results.13 The Sequoia system did not inform the Board personnel tabulating the election results that any data or cartridge errors had occurred.
As a result of its findings, the Board recommended, inter alia, that it: (1) “[d]ecrease the speed at which [cartridges] are tabulated so as to reduce and/or eradicate the risk of the incidence of static buildup and/or electronic impulses during the tabulation process”; “release results in smaller increments so as to accommodate the review and audit procedure associated with unofficial results summary reports”; and “[d]evelop a checklist for proofing tasks before releasing unofficial summary reports.”14
The following findings are based upon the evidence and testimony collected by the Committee before and during the hearing. They are not exhaustive and may be supplemented pending further investigation.
1. SEQUOIA’S EXPLANATION FOR THE ANOMALY ON SEPTEMBER 9, 2008, IS INADEQUATE
Sequoia attributed the mistaken election results contained in unofficial report #2 to human error in interacting with equipment (i.e. improper seating of the cartridge, premature removal of the cartridge, and/or an electrostatic discharge). Witnesses, including experts, cast considerable doubt on these explanations.
Principally, based on Sequoia’s inability to replicate these results under the same conditions, the technical implausibility of the explanation15, and expertise of experienced elections specialists16 suggest that Sequoia was too quick to exonerate itself and the equipment used in the tabulation process.
Until an independent technical investigation is conducted, the precise cause is still not known. To date, the evidence appears to indicate that there was a problem both in equipment (the server) and in the software.17
2. THE EQUIPMENT PROVIDED BY SEQUOIA VOTING SYSTEMS — BOTH HARDWARE AND SOFTWARE — FAILED TO GIVE ANY WARNING THAT THE DATA READ FROM THE CARTRIDGE WAS ERRONEOUS.
Despite the Board’s acknowledgment that it failed “to properly review the erroneous data contained in Report No. 2 prior to its release,” testimony received by the Committee confirms that the Board is not entirely to blame.
Experts testified that, on at least two occasions, Sequoia’s software should have generated a warning to the user that the data was compromised. First, if there was some mishandling of the cartridge, the hardware should have indicated that the cartridge was compromised. Second, and perhaps more importantly, according to basic principles of defensive programming, Sequoia’s software should have had an “elementary sanity check,” which would have flagged the erroneous data before it was released to the public.
Therefore, although it is true that the Board did not review the results before releasing them, and is appropriately subject to criticism on that account, the Board should not have been placed in the position of having only its on-the-scene scan of the results to determine whether data was compromised or not. Instead, the equipment provided to the Board by Sequoia should have given warnings of any potentially corrupted data. Without warnings, too much reliance is placed upon the individual judgments of Board members to determine whether Sequoia’s system is working properly.
3. THE BOARD’S POST-ELECTION MANUAL AUDIT FAILED TO PROVIDE CONFIDENCE IN THE ACCURACY OF THE VOTING MACHINES.
The Committee acknowledges that the Board was under no legal obligation to conduct a post-election audit after the September 2008 primary election and commends the Board for this proactive action. Nevertheless, the Committee finds that the audit performed by the Board suffered from a number of procedural defects.
For example, the audit fundamentally suffered from a lack of transparency insofar as there was little advance public notice of the audit, and no procedures for conducting the audit were identified in advance of the election. At times during the audit, individuals who were not part of the auditing teams handled ballots.
Moreover, some of the explanations offered by the Board for discrepancies between election night data and the audit results are puzzling. For example, in Precinct 21, there were several discrepancies related to Statehood Green candidate vote totals. The only explanation offered by the Board for these errors was that “[b]allot could not be read by the machine.” This explanation is insufficient.
Nor was the audit comprehensive. The audit apparently tabulated only paper records from the randomly selected precincts and made no attempt to tabulate other paper records — including absentee and special ballots. Rather than audit all paper records, the Board chose to review only ballots cast in each of the randomly selected precincts. Significantly, the audit did not include the precinct at the eye of the storm, namely Precinct 141.
Finally, the Committee heard testimony that the ballots were re-counted repeatedly, and that the audit process appeared to have been designed with the sole intent to reconcile the machine count to the manual count rather than to evaluate the accuracy of the machine counts. This should not be the primary purpose of the audit.
4. THE BOARD FAILED TO ADEQUATELY COMMUNICATE WITH THE PUBLIC DURING A CRISIS ON ELECTION NIGHT, AND HAS CONTINUED TO FRUSTRATE PUBLIC INTERACTION.
One of the greatest failings of the Board has been its inability to assure the public that it is taking steps to avoid mistakes in the future. On election night, when the Board released the erroneous reports to the public, the Board’s communication processes essentially shut down. Instead of offering explanations and providing assurance that the Board was responding appropriately to the crisis, the Board invited additional scrutiny through its silence. After the election, Board representatives similarly refused to comment publicly.
More troublingly, the Committee has heard on numerous separate occasions that the Board’s past dealings have been opaque to public criticism, and that election night merely exposed a pre-existing bunker mentality. Though the Board has been cooperative thus far with the present investigation, the Committee finds that the Board’s communication strategy failed in the face of a crisis.
At this stage in the investigation, it is clear that a substantial number of problems exist and that the election process in the District is in need of comprehensive reform. In the upcoming months, the Committee will continue to investigate the events of September 9, 2008. In the Committee’s opinion, the mistakes of the September primary election brought to the fore larger, systemic deficiencies in the operation of the District’s elections.
The Committee anticipates looking into a wide range of issues, including, but not limited to: aging hardware and software; the adequacy of resources and training (including increased recruiting); proper maintenance of voter rolls; proactive negotiation of contract terms with the voting systems vendor; the energy, experience, and leadership of the Board and its top staff; and the Board’s broader engagement with the public in elections issues.
Nevertheless, given the short time frame before the general election, the Committee makes the following recommendations that it believes can be implemented immediately, without change in law, and with little additional expense or administrative burden.
1. THE DISTRICT SHOULD ESTABLISH A TEAM OF INDEPENDENT, VOTING-TECHNOLOGY EXPERTS TO CONDUCT A TARGETED FORENSIC EVALUATION OF THE VOTING EQUIPMENT AND SOFTWARE PROVIDED BY SEQUOIA VOTING SYSTEMS.
Sequoia initially ruled out any “[e]ndemic hardware and software failures” as the cause of the tally issue. The Committee concludes, based on the testimony during the October 3 hearing, that Sequoia’s explanation is unreliable and undoubtedly self serving.
According to expert testimony, there are ample reasons to doubt Sequoia’s claims.18 For example, the summary report, which is generated by Sequoia’s software, had internal mathematical inconsistencies within it independent of the problem of phantom votes. Specifically, during the hearing, Sequoia admitted that data entry anomalies alone could not account for the fact that, in the erroneous report, the total number of voters cast throughout the District was fewer than the number of votes cast in certain individual races.19 Thus, there is reason to believe that the problems are due, at least in part, to deficiencies in the software design.
2. THE BOARD SHOULD ADOPT PROCEDURES AND PRACTICES, TO BE FOLLOWED IMMEDIATELY UPON THE CLOSING OF POLLS, TO ASSURE THE ACCURATE REPORTING OF UNOFFICIAL ELECTION RESULTS.
In retrospect, the District is fortunate that the erroneous results reported on September 9th were so obviously flawed. With all precincts reporting in Ward 2, had the number of votes been less preposterous, there is a good chance the error might have gone unnoticed.20 Irrespective of the explanation, it is beyond doubt that the unofficial election night results were simply incorrect and, had they gone unnoticed, there could have been significant consequences both to specific races and even more serious erosion of public confidence in the election system.
3. THE BOARD SHOULD DESIGN AND IMPLEMENT MANUAL POST-ELECTION AUDIT PROCEDURES.
It is absolutely critical that the Board immediately adopt and routinely follow a clearly defined post-election audit procedure. The implementation of a postelection audit procedure that is transparent, comprehensive, and targeted will go a long way toward confirming the accuracy of the voting machines and promoting the confidence in the reliability of the results.
There are several principles that should be strictly adhered to by the Board related to its audit.
Scope of the audit. After the September election, the Board audited four precincts, which represents less than 3% of the total number of precincts. The Board should routinely audit at least four randomly selected precincts after each election. Because of the heightened concerns related to the reliability of the voting machines, the Board should audit eight precincts (more than 5% of the total number of precincts) after the November election.
In addition to ballots cast in any given precinct, the post-election audit should include absentee and special ballots, which are centrally counted by the Board. The Committee recommends that the Board audit 3% of all absentee and special ballots cast during the November elections before certifying the election results.
The Committee also recommends that the Board expand its audit to target additional races, precincts, and machines. First, the Board should implement a policy of automatically auditing races where the margin of victory is 1% or less than the total number votes cast in that election. Second, the Board should audit those precincts where the difference between the number of voters reported by the machines and the number of ballots cast reported by the precinct captains is greater than the margin of victory. Finally, the Board should identify additional criteria that will guide the exercise of its discretion in auditing additional races, precincts, and machines beyond those already identified, and make its policies publicly available before the election.
Transparency. Because a post-election audit process is aimed at measuring the accuracy of the voting machines and promoting the confidence in the reliability of the results, the process should be public and clearly explained. The Board should post key dates on its website and coherently explain the procedure it intends to follow in the conduct of its audit.
In that same vein, the Board should provide greater transparency in the reconciliation process. According to the Board’s report, “[i]n precinct 21, the Statehood Green ballots could not be read by the machine because of the ballot header.” Yet the Board appears to have made no effort to determine the cause of that error, nor did it investigate whether that error may have affected other precincts. The Board should undertake greater efforts to meaningfully explain and investigate any discrepancies that are discovered through the audit process.
Independence of counters. In addition, when conducting the audit, the Board should not inform those counting the ballots of the voting-machine-generated results. The purpose of the audit is measure — not to rubberstamp — the integrity and accuracy of the results.
4. THE BOARD SHOULD TRAIN POLL-WORKERS TO ENCOURAGE VOTERS TO USE PRECINCT COUNT OPTICAL-SCAN VOTING MACHINES DURING THE GENERAL ELECTION.
Though the phantom votes arose in connection with reading a cartridge used by an optical-scan voting machine, the Committee recommends that the Board provide specific training to precinct captains to encourage voters to use optical-scan voting machines instead of the touch-screen voting machines. There are two reasons for this recommendation.
First, in light of the unresolved concerns with the Sequoia software and hardware, the Committee believes that the District of Columbia should make every effort to provide a “voter verified paper trail” — in this case, the ballots themselves — during the general election. By implementing meaningful ballot accounting and manual post-election audit procedures, there can be greater confidence in the outcome of the election.
Second, the Committee notes that the operation of optical-scan machines is substantially faster than that of touch-screen machines because multiple voters can fill out paper ballots simultaneously, while only one voter can use a touch-screen voting machine at a time. During the hearing, the Committee also learned that the Board intends to place an addition optical-scan machine in each precinct. The active encouragement of optical-scan machines will mitigate, at least in part, some of the strain caused by the very high turnout expected in November.
5. THE BOARD SHOULD DESIGN AND IMPLEMENT A PROTOCOL TO ENSURE TIMELY RELEASES OF UNOFFICIAL ELECTION RESULTS WITHOUT SACRIFICING ACCURACY.
In the aftermath of the primary election, there may be a tendency to proceed overly cautiously in the release of results. It would be a mistake, however, to unnecessarily slow the release of results. Though accuracy must be ensured, by unnecessarily or unreasonably delaying the release of results, the public’s confidence in the Board’s ability to efficiently process vote tabulation could be further eroded.
The best way forward is to establish a clear procedure for the processing of elections results — including the aforementioned “sanity checks” and ballot accounting — and adhering to that procedure. Though it is obviously difficult to predict the rate at which precincts will report, the Board would do well to release the results in smaller batches. Beyond the advantage of having timely release of results, issuing results with fewer precincts reporting allows for a more thorough check for errors.
6. THE BOARD SHOULD ALSO IDENTIFY STRATEGIES TO COMMUNICATE WITH THE PUBLIC IN THE EVENT THAT FURTHER ANOMALIES ARISE.
As one witness observed: though mistakes happen, the Board must be prepared to respond appropriately. Elections rarely run as anticipated, and the Board should be working diligently to plan for contingencies that may arise on November 4. The Board must develop a clear plan to act swiftly when crises develop. It must also make it a top priority to establish a plan to effectively communicate — in real time — what steps it is taking to respond to evolving situations. There is a pressing need to adopt a timely and comprehensive public communications strategy, particularly given the events of the most recent election. The Board should act immediately to implement such a strategy.
The District of Columbia may ultimately never know the precise cause of the erroneous election results released on September 9, 2008. Yet, expert testimony has made clear that the explanations given by Sequoia, resting essentially on human error or static discharge, are implausible and self serving. Indeed, there was no consensus, even among Sequoia’s representatives, that the reported anomaly could be traced to user error. Thus, it is unwise to assume that future problems can be avoided simply by being more careful. The better course is for the District to independently undertake a targeted forensic audit of the equipment. The District should also review Sequoia’s contractual obligations to ensure that they were, and will be, met. And while that investigation continues, we must look forward. On November 4, 2008, thousands of District residents will cast ballots. The Board’s first priority should be to make efforts to restore confidence in its ability to accomplish the task of accurately counting those votes. Through the immediate implementation of sound ballot accounting practices, post-election audits, and clearly defined processes, the Board can make substantial progress toward that goal. In addition, voters should be encouraged to use paper ballots and optical-scan voting machines on Election Day. This may help to speed the movement of voters through the polls, but, as importantly, will provide a paper trail of votes. Finally, the Board should build in redundancies in the vote tabulation process, and should establish clear processes for the release of election results.
Beyond these recommendations, the Committee will continue to seek out ways to ensure that elections in the District of Columbia are accessible, accurate, transparent, and efficient. It is basic that everyone’s right to vote should be preserved and that everyone’s vote should count. It is also essential that voters have confidence that their votes have been counted and that the elections are professionally managed. The Committee looks forward to finding ways to restore the District’s place as a model jurisdiction in managing elections.
A. R17-0769, Council Board of Elections and Ethics Investigation Special Committee
Emergency Declaration Resolution of 2008
1 At that hearing, the Committee received extremely helpful testimony from a number of witnesses, including: Sequoia Voting Systems; the Board of Elections and Ethics; and others with particularized knowledge of the District’s elections, including Dorothy Brizill, Executive Director, DC Watch. The Committee gives special thanks to voting-technology experts Professor Doug Jones from the University of Iowa and Jeremy J. Epstein, Principal Consultant, Cigital Inc., and to Lawrence Norden, Counsel and Project Director, Voting Technology Project, Brennan Center for Justice.
2 Marcia Davis, Voting Irregularities Interrupt Count, D.C. Wire, Washington Post (Sept. 9, 2008).
3 A precinct count optical scan voting machine is an electronic voting machine that uses optical sensors to detect how a voter marks a paper ballot and tabulates the number of votes at the precinct location. The District of Columbia also uses a touchscreen Direct-Recording Electronic (DRE) voting machine, which records votes directly without the use of a paper ballot, in part for the visually impaired and others with disabilities.
4 District of Columbia Board of Elections & Ethics, Internal Review Committee’s Investigative Report into Election Night Results Summary Reporting Irregularities During the September 9, 2008 District of Columbia Congressional & Council Primary Election 8 (Oct. 1, 2008), available at http://www.dcboee.org/pdf_files/nr_172.pdf [hereinafter “Board Report”]. Before certifying the election results, the Board conducted a hand recount of the ballots in Precinct 141, which confirmed that the final election results were accurate within that Precinct.
5 Board Report 9.
6 The precincts selected for review were not released to the public before the audit process began.
7 According to the District of Columbia Board of Elections & Ethics, Final Report for the Congressional and City Council Primary Post Election Audit (Sept. 24, 2008), available at http://www.dcboee.org/pdf_files/nr_169.pdf [hereinafter “Audit Report”], there was one person read the votes, one person who observed, and a third who tallied.
8 Sequoia Voting System, Sequoia Voting Systems’ Analysis of Washington, D.C. 09/09/08 Tabulation Issue (Sept. 9, 2008).
9 Sequoia Voting Systems, Report to the District of Columbia Board of Elections & Ethics (Sept. 22, 2008) [hereinafter “Sequoia Report”].
10 Id. at 3.
12 Board Report 9.
14 Id. at 10.
15 Expert witnesses testified that, had these conditions caused the data to be corrupted, the database should have rejected the data rather than accept inaccurate voting results. By way of analogy, if a DVD of Terminator II were seated improperly in a DVD player, it would not likely play “Gone with the Wind.”
16 For example, Jeremy Epstein cited to a situation in Ohio where a different vendor, Premier Election Systems, “blamed election officials for incorrect results, where one precinct’s votes were lost. Later, the vendor changed their explanation, and blamed the anti-virus vendor for the error. Eventually, the vendor admitted that it was a bug in their software.”
17 During the hearing, Sequoia’s representatives indicated that after they had generated their report, they identified what they believed to be a problem in the database server. They could not, though, offer any explanation for the cause of that problem. E.g.,
18 E.g., Statement of Jeremy J. Epstein, Principal Consultant, Cigital Inc. (Oct. 3, 2008) (“Sequoia’s claim that incorrect seating of the memory card causing incorrect results is hard to believe from a technical perspective.”).
19 For example, in the race for Democratic At-Large Member of the Council, there were almost 5,000 more votes (including votes for the candidates, overvotes, and undervotes) than the total number of Democratic voters.
20 Witnesses did express concern that past anomalies may have occurred but avoided detection. The Committee will continue to investigate ways to increase public confidence in the reliability and accuracy of election results in the District.
Back to top of page
Send mail with questions or comments to firstname.lastname@example.org
Web site copyright ©DCWatch (ISSN 1546-4296)