Ms. Poucher has stated that it “would be fantastic” if the EAC or other federal agency would establish asearchable database that would allow election officials to see what kind of problems other jurisdictionsaround the country had with the same systems she was using, whether the vendor had provided themwith patches or other fixes, and what procedures they were using to prevent similar problems in thefuture. 134 She noted that if she discovered a problem during a small, off-year election, it might drawlittle notice in her own county, but an EAC database would provide her with an opportunity to in<strong>for</strong>mother election officials throughout the country using the same system of the malfunction, and preventmore damaging problems down the road. “I think this kind of database would get integrity back in thesystem,” she added. 135Additional case studies further suggest that vendors are too often slow to acknowledge problemswith their systems and frequently do not cooperate as fully or as transparently as public officials (andmembers of the public) would like when problems are confirmed. At the same time, because the lastfew years have seen a number of voting system vendors go out of business or get bought out by rivals,election officials may not have any vendor to turn to <strong>for</strong> explanation when a problem occurs (see CaseStudy 11 below and The Suboptimal Structure of the Voting System Market at pp. 25 - 26). This is whywe believe it is critical that vendors be required to report problems to the clearinghouse within a certaintime period when certain events occur. 13611. Fairfax County, Virginia, March 2009• • •In March of 2009, during the post-election canvass <strong>for</strong> a closely contested special election to fill avacancy <strong>for</strong> the County Board of Supervisors, election officials and observers noticed that the combinedtotals <strong>for</strong> two AVC WinVote DRE voting machines in a precinct showed a total of 359 votes cast, with377 votes recorded <strong>for</strong> the Republican, 328 <strong>for</strong> the Democrat, and eighteen <strong>for</strong> other candidates, <strong>for</strong>a total of 723 votes – or 364 more votes recorded than cast. 137 Officials in Virginia were lucky to catchthis problem. As the Brennan Center and others have shown, several jurisdictions in the United Stateshave inadequate ballot accounting and reconciliation practices. 138 In such cases, it is possible to missthis kind of error.The post-election canvass showed that while one machine recorded 723 votes, only 707 voters hadsigned-in to that precinct on Election Day, and 348 votes had been cast and correctly counted onanother machine in the same precinct. Election officials decided to print the “ballot images” (or thesoftware’s digital representation of the ballots cast by voters), and add those up by hand. When they didthat, the number of votes on the two machines combined equaled the number of voters who checkedin at the polls, and <strong>for</strong> each machine the number of voters equaled the sum of the number of votes <strong>for</strong>each candidate. It was only by checking the number of votes against the poll books and by using analternative method of vote counting that officials were able to determine the result with some certainty.It is important to note, however, that while the ballot images correlated with the number of voters whoused the machine, the WinVote does not produce a paper trail; the ballot images are created by thevoting system itself, not by voters.While Fairfax County officials were satisfied that ballot accounting redundancies allowed them todetermine accurate election results, they were never able to determine the cause of the problem withcertainty. 139 They surmised that the problematic machine did not reset properly after pre-electiontesting, but they are not sure why this occurred in 2009 after years of using the same machines without20 | Brennan Center <strong>for</strong> Justice
incident. 140 Because the manufacturer of the system was no longer in business, the county had to trackdown a programmer familiar with the system to corroborate its theory. 141 County Election ManagerJudy Flaig acknowledges that conducting the type of investigation necessary to determine the causeof the problem to even this limited extent would not have been possible <strong>for</strong> every county, and that asmuch as jurisdictions try to communicate with manufacturers and one another, some in<strong>for</strong>mation fallsthrough the cracks.“Most jurisdictions don’t have the resources to do this kind of work,” Flaig said,“something like [a database of voting machine problems] would really be helpful.” 142Rokey Suleman, Fairfax County Registrar at the time of the incident, says he still does not knowwhat caused the problem. Although he no longer works in Fairfax County, he believes other officialsusing these machines “should certainly want to know what happened, so that they can put the properprocedures in place to ensure it doesn’t happen again.” 143Hinds County, Mississippi also uses the AVS WinVotes. Suleman noted that “unless they read a tinyVienna, Virginia newspaper where this story was reported, there’s no way they would know about thisproblem.” 144 Indeed, the Brennan Center contacted Hinds County Supervisor Robert Graham, and hehad never heard of the problems in Fairfax County.Suleman, who is now the Executive Director of the District of Columbia’s Board of Elections and Ethics,noted that in addition to providing election officials with notice of potential problems with the machinesthey are using, a central database would have other benefits <strong>for</strong> election officials. Mr. Suleman stated thatWashington, D.C. “just went through a procurement process <strong>for</strong> new voting machines,” and that hewould have been greatly helped by a central database that could “serve as a repository to let me know whatissues exist with the machines, rather than having to rely on what the vendors spoon-feed me.”12. District of Columbia, September 2008• • •A District of Columbia Council investigation after the District’s 2008 primary found that vote totalsoriginally produced by a Sequoia tally server on election night were “obviously inaccurate.” 145 TheBoard of Elections traced the problem to a cartridge <strong>for</strong> a Sequoia precinct-count optical scanner in oneprecinct, which reported voter turnout nearly twice that of the registered population of the precinct andshowed 1,554 write-in votes in a race without a write-in campaign. 146 In two reports in the WashingtonPost, election officials indicated that the malfunctioning cartridge caused other problems with thepreliminary vote totals. 147 According to acting Executive Director of the D.C. Board of Elections SylviaGoldsberry-Adams, “one defective cartridge caused vote totals to be duplicated into multiple races onthe summary report issued by our office;” on this summary report, 1,542 appeared as the number ofovervotes in five contests. 148The District Board of Elections asked Sequoia to explain the problems. In an initial response, Sequoiastated that it found “no anomalies or irregularities in either the data or the internal event logs that canbe identified as having caused or contributed to the issue experienced on election night.” 149 In responseto the Board’s request <strong>for</strong> a more detailed explanation, Sequoia issued a report that attributed theproblem to human error and ruled out “[e]ndemic hardware and software failures […] as the cause.” 150As <strong>for</strong> the Board’s request <strong>for</strong> in<strong>for</strong>mation about past occurrences of this error, Sequoia responded,“[s]ince our customers conduct the actual elections – not Sequoia – we do not have any way of keepingtrack of such incidents, nor is it our responsibility to do so.” 151Brennan Center <strong>for</strong> Justice | 21
- Page 5: IV.A BETTER WAY TO TRACK AND ADDRES
- Page 11: 4. Pressure Vendors to Voluntarily
- Page 14 and 15: its newly established Voting System
- Page 16 and 17: iii.failures of the current system:
- Page 18 and 19: 2. Humboldt County, California, Nov
- Page 20 and 21: 4. Pulaski County, Arkansas, May 20
- Page 22 and 23: 7. Florida, June 2004According to t
- Page 24 and 25: Denise Lamb, who currently serves a
- Page 28 and 29: Sequoia identified four possible ca
- Page 30 and 31: Had advocates and researchers in Ne
- Page 32 and 33: elative to voting system vendors. A
- Page 34 and 35: or potential vulnerabilities by any
- Page 36 and 37: Finally, the Department of Justice
- Page 38 and 39: Provide Timely & Organized Access t
- Page 40 and 41: the CPSC may compel the manufacture
- Page 42 and 43: Civil Division of the Department of
- Page 44 and 45: C. Analogous RegimesCivil penalty p
- Page 46 and 47: Of course, adding these kinds of pr
- Page 49 and 50: v. conclusionVoting is the most imp
- Page 51 and 52: System Vulnerabilities: should incl
- Page 53 and 54: February2008September2004November20
- Page 55 and 56: November2006November2006ES&SiVotron
- Page 57 and 58: June 2008SequoiaOptech 400CCaliforn
- Page 59 and 60: March 2004March 2004DieboldAccuVote
- Page 61 and 62: February2008September2008PremierAcc
- Page 63 and 64: event of an overvote displayed a co
- Page 67 and 68: January 2008January 2008PremierAccu
- Page 69 and 70: November2004ES&SM650FloridaSupervis
- Page 71 and 72: October 2008February2008PremierAccu
- Page 73 and 74: February2008February2008October 200
- Page 75 and 76: 2004 AccuVote 2000ESNovember2008Nov
- Page 77 and 78:
November2008November2008Hart InterC
- Page 79 and 80:
February2008May 2006August 2004May2
- Page 81 and 82:
March 2008June 2009November2008Prem
- Page 83 and 84:
February2008November2006SequoiaAVC
- Page 85 and 86:
November2008SequoiaImageCastNew Yor
- Page 87 and 88:
November2004November2004UnilectPatr
- Page 89 and 90:
November2008November2008March 2008E
- Page 91 and 92:
May 2008April 2008November2006May 2
- Page 93 and 94:
2006 iVotronicThe Post & Courier re
- Page 95 and 96:
AccuVote TSR6; HartInterCiviceScan;
- Page 97 and 98:
November2006ES&S:iVotronicTexasHida
- Page 99 and 100:
November2003November2009November200
- Page 101 and 102:
October 2008October 2008May 2008ES&
- Page 103 and 104:
appendix c : dupage county election
- Page 105 and 106:
and resolve problems with certified
- Page 107 and 108:
49. E-mail from Carolyn Crnich, Cle
- Page 109 and 110:
95. Kleinberg, supra note 92.96. Ma
- Page 111 and 112:
143. Telephone Interview with Rokey
- Page 113 and 114:
187. Id. at 22.188. See Thomas, sup
- Page 115 and 116:
1801 (“[T]he Secretary shall init
- Page 117 and 118:
258. See United States Department o
- Page 119 and 120:
294. John Archibald & Brett J. Blac
- Page 121 and 122:
334. Jane Musgrave, Palm Beach Coun
- Page 123 and 124:
374. Glitch Keeps Fulton Voters Wai
- Page 125 and 126:
417. Joe Dejka & Chris Olson, A Lat
- Page 127 and 128:
457. Letter from Dave Byrd, Preside
- Page 129:
498. Lauren Glendenning, Voting Gli
- Page 132:
ennancenterfor justiceAt New York U