Improving the security, transparency and efficiency of California's 1% manual tally procedures

EVT 2008 - 2008 USENIX/ACCURATE Electronic Voting Technology Workshop

Volumn , Issue , 2008, Pages

  Hall, Joseph Lorenzo ^a  

^a UNIVERSITY OF CALIFORNIA   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EFFICIENCY; TRANSPARENCY; VOTING MACHINES;

CALIFORNIA; VOTING TECHNOLOGY;

ELECTRONIC VOTING;

EID: 77955424593     PISSN: None     EISSN: None     Source Type: Conference Proceeding    
DOI: None     Document Type: Conference Paper

References (67)

1. Due to space limitations, the detailed procedures are presented in a separate document: Joseph Lorenzo Hall, The 1% Manual Audit in California. UC Berkeley School of Information, April 2008 hURL: http://josephhall.org/procedures/ca_tally_procedures-2008.pdfi
2. Ronald L. Rivest and John Wack, On the notion of “software independence” in voting systems. July 2006 hURL: http://vote.nist.gov/SI- in- voting.pdfi.
3. Top-To-Bottom Review of California's Voting Systems. California Secretary of State, March 2007 hURL: http://www.sos.ca.gov/elections/elections_vsr. htmi;
4. Patrick McDaniel et al., EVEREST: Evaluation and Validation of Election-Related Equipment, Standards and Testing (Academic Final Report). December 2007 hURL: http://www.sos.state.oh.us/SOS/upload/everest/14-AcademicFinalEVERESTReport.pdfi;
5. Software Reviews and Security Analyses of Florida Voting Systems. Florida State University's Security and Assurance in Information Technology Laboratory, February 2008 hURL: http://www.sait.fsu.edu/research/evoting/index. shtmli.
6. Lawrence Norden et al., Post-Election Audits: Restoring Trust in Elections. Brennan Center for Justice at The New York University School of Law and The Samuelson Law, Technology and Public Policy Clinic at the University of California, Berkeley School of Law (Boalt Hall), 2007 hURL: http://www.brennancenter.org/dynamic/subpages/download_file_50227.pdfi.
7. The Verified Voting Foundation keeps an up-to-date list of state paper record laws on its front page (see: http://www.verifiedvoting.org/). Many researchers believe that paper is currently the only feasible form of such an audit record, while others advocate for “end-to-end” verification systems or take issue with the lack of verifiability of most paper-based systems for people with disabilities.
8. See: Ben Adida and C. Andrew Neff, Ballot Casting Assurance. USENIX/ACCURATE Electronic Voting Technology 2006 (EVT'06) Workshop, 2006 hURL: http://www.usenix.org/events/ evt06/tech/full_papers/adida/adida.pdfi;
9. Daniel Tokaji, The Paperless Chase: Electronic Voting and Democratic Values. Fordham Law Review, 57 2005 hURL: http://ssrn.com/abstract=594444i
10. Philip B. Stark, Conservative Statistical Post-Election Audits (in press). The Annals of Applied Statistics, 2008 hURL: http://www.stat.berkeley.edu/~stark/Preprints/conservativeElectionAudits07. pdfi;
11. Javed A. Aslam, Raluca A. Popa and Ronald L. Rivest, On Auditing Elections When Precincts Have Different Sizes. USENIX/ACCURATE Electronic Voting Technology Workshop 2008 July 2008.
12. California's 1% manual tally was introduced in 1965. See: California Statutes 1965, c. 2040, p. 4659, Sec. 1.
13. Saltman provides interesting cases of historical punchcard mishaps. See: Roy G. Saltman, Effective Use of Computing Technology in Vote-Tallying. National Bureau of Standards, March 1975 hURL: http://csrc.nist.gov/publications/nistpubs/NBS_SP_500-30.pdfi
14. Note that the current manual tally statute, CA Elec. Code Sec. 15360, was consolidated from a number of historical statutes.
15. See: Former Sec. 15645, enacted by CA Stats. 1994, c. 920, Sec. 2, (derived from: former Sec. 15281 (added by CA Stats. 1976, c. 246, Sec. 3), former Sec. 15417 (added by CA Stats. 1965, c. 2040, p. 4659, Sec. 1), and former Sec. 17190 (added by CA Stats. 1978, c. 847, Sec. 5, amended by CA Stats. 1986, c. 1277, Sec. 14.).
16. CA Elec. Code Sec. 336.5.
17. CA Elec. Code Sec. 19253(b)(2).
18. California Top-To-Bottom Review (as in n. 3);
19. David Jefferson et al., Evaluation of Audit Sampling Models and Options for Strengthening California's Manual Count. California Secretary of State, July 2007 hURL: http://www.sos.ca.gov/elections/peas/final_peaswg_report.pdfi;
20. Post-Election Manual Tally Requirements. California Secretary of State, October 2007 hURL: http://www.sos.ca.gov/elections/voting_systems/ttbr/post_election_req.pdfi.
21. Of course, voters must actually check that the contents of the paper record match their intent and what is displayed on the voting system screen. Recent evidence suggests that few voters do this.
22. See: Sarah P. Everett, The Usability of Electronic Voting Machines and How Votes Can Be Changed Without Detection. Rice University PhD Thesis, May 2007 hURL: http://chil.rice.edu/alumni/ petersos/EverettDissertation.pdfi. This concern does not apply to technologies where voters directly-mark the paper record, such as optical scanners.
23. Note that this election law (CA Elec. Code Sec. 15360(c)) now specifies that election officials must use a “random number generator or other method” from regulations adopted by the Secretary of State. This means that there is no prohibition on computer-generated pseudorandom numbers, as we would prefer.
24. Arel Cordero, David Wagner and David Dill, The Role of Dice in Election Audits-Extended Abstract. IAVoSS Workshop on Trustworthy Elections 2006 (WOTE 2006), June 2006 hURL: http://www.cs.berkeley.edu/~daw/papers/dice-wote06.pdfi.
25. But compare with: Joseph A. Calandrino, J. Alex Halderman and Edward W. Felten, In Defense of Pseudorandom Sample Selection. USENIX/ACCURATE Electronic Voting Technology Workshop 2008 July 2008.
26. CA Elec. Code Sec. 15372.
27. We met twice with Yolo county to discuss how what we were learning would apply to Yolo, a much smaller county. This author did not have the opportunity to observe the manual tally in Yolo County, although we benefited from the observations of collaborators.
28. David Wagner, Thoughts on the Nov 16, 2006 1% Manual Tally in Yolo County. UC Berkeley Department of Computer Science, November 2006 hURL: http://www.yoloelections.org/news/snews/reactions.pdfi.
29. Hall, The 1% Manual Audit in California (as in n. 1).
30. Others involved in this work include David Dill (Stanford Computer Science), David Wagner (UC Berkeley Computer Science), Arel Cordero (UC Berkeley Computer Science), Aaron Burstein (UC Berkeley Law) and Kim Alexander (California Voter Foundation).
31. For example, Alameda County worked with Arel Cordero, David Wagner, David Dill and members of their public advisory committee to design a random selection process using a tumbler with numbered ping-pong balls. We provided a research memo describing how imperfections in their random selection process resulted in non-uniform random selection. We describe this case in Section 4.1.
32. Joseph Lorenzo Hall, The 1% Manual Audit in California: Proposed Procedures and Rationale. November 2006 hURL: http://www.josephhall.org/papers/manual_audit_procedures-final.pdfi;
33. Rebekah Gordon, Elections Office gets tips from experts. San Mateo County Times, November 2006 hURL: http://www.shapethefuture.org/press/2006/insidebayareacom113006.aspi.
34. By “ballot types” we mean ballots cast using a distinct type of voting technology such as in-precinct optical scan ballots, DRE ballots, paper-based provisional ballots, etc.
35. The Secretary of State has published manual tally reports that show many counties performing random selection very soon after election day. County Manual Tally Reports. April 2008 hURL: http://www.sos.ca.gov/elections/manual_count_reports.htmi In one case, a county seems to have even performed the selection before election day (February 5, 2008).
36. Summary Information-Post-Election Manual Tally. County of Fresno County Clerk / Registrar of Voters, March 2008 hURL: http://www.sos.ca.gov/elections/county_manual_count_reports/Fresno/Fresno.pdfi
37. This allows beginning the manual tally for ballot types where counting is completed before all ballot types are counted. In February 2008, we participated with Philip Stark (UC Berkeley Statistics) and Elaine Ginnold (Marin County Registrar of Voters) in the first statistical confidence-based audit of an election. In this pilot, each ballot type was sampled in separate strata so that the tally could begin for ballots where counting was completed.
38. When counting optical scan ballots-where the ballot marking can be less definitive-we observed that it helped to relax the blind counting requirements after two discrepant tallies. This allowed the talliers to try and determine which ballot had been read (or not read) by the machine but not seen (or seen) as a valid vote by the talliers.
39. Joseph Lorenzo Hall, Research Memorandum: On Improving the Uniformity of Randomness with Alameda County's Random Selection Process. UC Berkeley School of Information, March 2008 hURL: http://josephhall.org/papers/alarand_memo.pdfi. Note: we did not discover this independently; two public observers, Meg Holmberg and Tim Erickson, brought this to our attention.
40. Lawrence Norden and Eric Lazarus, The Machinery of Democracy: Protecting Elections in an Electronic World: Brennan Center Task Force on Voting System Security. Brennan Center for Justice at NYU School of Law, 2006 hURL: http://www.brennancenter.org/dynamic/ subpages/download_file_39288.pdfi.
41. Kim Alexander (California Voting Foundation), Philip Stark (UC Berkeley Statistics) and myself assisted the Office of the California Secretary of State in developing the reporting instrument: Post-Election Manual Tally Log: Suggested Instructions for Post-Election Manual Tally Requirements (CCROV-08048). California Secretary of State, 2008 hURL: http://josephhall.org/ccrov/CCROV-08048.pdfi
42. We have been unable to confirm with this particular county how they count VVPAT records.
43. CA Elec. Code Sec. 19253(b)(1) which says, in part, “The voter verified paper audit trail shall be considered the official paper audit record and shall be used for the required 1-percent manual tally [... ]”.
44. For example, one explanation is that there were very few ballots cast on the VVPAT rolls we observed. In that case, most of the candidates on the ballot would have zero votes. Tallying would entail a brief tally of any candidates that do have votes with a lengthy tally of zero votes for each of the remaining candidates. Without pre-filled tally sheets, the task of tallying votes on VVPAT records becomes more a task of confirming the lack of votes.
45. For example, we observed a tallier in Alameda erasing tally marks after they had to retally a specific candidate. See: http://www.flickr.com/photos/joebeone/2295569830/sizes/l/.
46. Joseph Lorenzo Hall, Transparency and Access to Source Code in Electronic Voting. USENIX/ACCURATE Electronic Voting Technology Workshop 2006, June 2006 hURL: https://www.usenix.org/events/evt06/tech/full_papers/hall/hall.pdfi.
47. CA Elec. Code Sec. 15360(d).
48. One Percent Manual Recount Procedures. San Mateo County Clerk Assessor Recorder Elections, December 2007 hURL: http://josephhall.org/procedures/sanmateo_tally_procedure_122007.pdfi;
49. Procedures for One Percent Manual Tally. Marin County Registrar of Voters, February 2008 hURL: http://josephhall.org/procedures/marin_tally_procedure_022008.pdfi;
50. Process Overview of the 1% Manual Tally. Alameda County Registrar of Voters, February 2008 hURL: http://josephhall.org/procedures/alameda_tally_procedure_022008.pdfi.
51. Election Observer Handbook. San Mateo County Clerk Assessor Recorder Elections, February 2008 hURL: http://josephhall.org/procedures/sanmateo_obsprocs_ 022008.pdfi In addition, the California Secretary of State requires counties to publish an Election Observation Panel Plan
52. (see: http://www.sos.ca.gov/elections/eop.htm) and the California Election Code imposes certain requirements to facilitate public observation in certain cases.
53. See CA Elec. Code Sec. 15104.
54. We call this hardcopy mapping a Master Selection Spreadsheet. Recall that once the 1% selection is complete, additional selection may be necessary as there may be races for which no precinct was selected for auditing. If a jurisdiction has decided to follow best practice, rather than the letter of the law, and select precincts randomly for these additional races, they will need to provide similar mapping spreadsheets for each race in the election (we call these Contest Selection Spreadsheets).
55. See the image linked in n. 43.
56. Cordero, Wagner and Dill (as in n. 17).
57. Joseph Lorenzo Hall, Dice Binning Calculator for Post-Election Audits. March 2008 hURL: http://www.josephhall.org/dicebins.phpi.
58. To see our dicebins.php calculator “in action” in San Mateo, see: http://www.flickr.com/photos/joebeone/2293490290/sizes/l/.
59. We observed one team quickly recounting stacks of ballots to make sure that they arrived at the result they got with the slower tally process. This might have been developed in response to having to retally an entire precinct.
60. Stark (as in n. 6).
61. See: http://www.oasis-open.org/committees/tc_home.php?wg_abbrev=election.
62. We note that, in one exceptional case, one tally team member was a supervisor working in the local elections department. We found the interaction between this tally team member and the other members of the team to be particularly different from other teams. We were uncertain if the other team members were directly supervised by this individual. If so, this obviously highlights an undesired point of friction. Election officials need to be conscious of these kinds of power dynamics and seek to neutralize them. One option would be to only place supervisors on teams of people that they do not directly supervise (such as temporary employees). Another possible solution is to emphasize in training that all tally team members have equal authority for the period of time in which the tally is conducted.
63. For images of these types of tally sheets in Alameda and San Mateo, see the following images, respectively: http://www.flickr.com/photos/joebeone/2266221884/sizes/l/ and http://www.flickr.com/photos/joebeone/2240342264/sizes/l/.
64. This observation was based on San Mateo's VVPAT tally area which used 5 tally teams of 4 people each. San Mateo did photocopy the filled-out tally sheets once each tallier had completed them so that they could be reused when the team started a new VVPAT roll.
65. Roger G. Johnston, Tamper-Indicating Seals. American Scientist, 94 November 2006.
66. San Mateo 1% Manual Tally Procedures (as in n. 37).
67. Hall, The 1% Manual Audit in California (as in n. 1).