Excel ate my elections! Jason Kitcat, Sussex University Ian Brown, Oxford University
UK elections May 2007
Scottish Parliamentary elections used e-counting
10 e-voting/e-counting/telephone vote trials in English local council elections
England e-voting trials System Election Software AG ( Dominion Voting, OPT2VOTE) Warwick D. Tata ( Everyone Counts) Swindon B. Software AG ( Dominion Voting, OPT2VOTE) Stratford D. ES&S ( Scytl, IntelliVote, Reliant) South Bucks D. OPT2VOTE Shrewsbury B. OPT2VOTE Sheffield C. ES&S ( Scytl, IntelliVote, Reliant) Rushmoor B. OPT2VOTE Dover D. Indra Breckland B. Indra Bedford B.
2006 Act allows accredited observers to attend polls and counts
Does not cover remote servers, testing lab results, source code etc.
FoI rights v. helpful
To evaluate the integrity of technologies and processes used in the electoral pilot schemes.
To examine whether the pilot schemes might increase the risk of electoral fraud or error.
To observe whether the pilot schemes risk the secrecy of the ballot.
To collect the views of voters, candidates and officials on the schemes piloted.
Pilot programme announced 17 Oct 2006
Closing date 17 Nov 2006
Final approval 29 Jan 2007
Polling day 3 May 2007
“ T he current systems to combat electoral abuse in Great Britain are unsatisfactory already, so to proceed with these pilot schemes, appears ill-timed and betrays confusion over priorities. Unfortunately it appears to come down to the obsession with modernisation as a means of increasing participation at elections.” - Sir Alistair Graham, chairman, Committee on Standards in Public Life, February 2007
Failing laptops in Swindon
Incorrect tally and error messages in Rushmoor
Remote voters with problems in South Bucks couldn’t use polling station
Scottish voters confused by combination of AMS parliamentary ballots (Gould 2007 p.52)
In Glasgow, 3-5% regional votes spoilt, 10% constituency
“ [T]he responses of 100 voters (for each election) to various ballot paper designs hardly seems adequate given the widespread change in the system.” (Gould 2007 p.48)
75% misvotes included only one x; 15% 2 in one column (Gould 2007 p.50)
Spoilt Scottish ballots
Ballot papers cast: > two million votes cast for each election (regional: 2,102,623; constituency: 2,101,638; and local government: 2,099,945).
Rejected ballot papers - Scottish Parliament: 146,099 ballot papers (regional: 60,455 or 2.88%; constituency: 85,644 or 4.075%) rejected >> rejected ballot paper rate in 2003 - 0.65% of the regional ballot papers and 0.66% of the constituency ballot papers.
Rejected ballot papers - local government: 38,352 rejected ballot papers or 1.83% >> 0.64% in 2003/4
Source: Ron Gould (2007) Scottish Elections Review
Problems with counting code
Serious problem occurred during Edinburgh count
Investigation by DRS technical staff found corrupted SQL Server database indices
Patch applied to all counting systems next morning (Gould 2007 p.102)
Lack of checking
Pilot regulations: “46(4) Re-count … the returning officer may treat a request for a re-count as unreasonable unless he has reason to believe that the system has not scanned the votes correctly”
One manual recount in Dereham-Humbletoft ward found 56.1% more votes
Lack of reliable audit trail: “ one local authority was not able to provide any information, two count centres produced the electronic images but not the decisions taken by Returning Officers, and two further count centres were able to produce the decisions taken by Returning Officers but not the electronic images.” (Gould 2007 p.106)
Excel ate my votes!
Provisional count of regional votes in Highlands and Islands unexpectedly found 0 SNP wins
Recheck found that votes had been missed due to an overwide spreadsheet
New result gave control of Parliament to SNP
The golf club attack
Voter attacked Edinburgh ballot boxes with golf club, ripped up 100 ballot papers
Officials taped up ballots for counting
Source: BBC News
Loss of confidence in counting
“ A lthough material had been produced explaining how the count operated and briefings were offered to candidates on the subject, it is clear that many political party agents, candidates, and other observers did not fully understand the counting process.” (Gould 2007 p.106)
“ [C]ombined elections are not only a disservice to the local councils and candidates but also to the electorate as well.” (Gould 2007 p.115)
Independent election observers are a key mechanism for increasing public trust
What should they be observing in e-elections?
Plenty of time is required to plan, build consensus, procure, design and test systems - and fallback plans
Real-world UK use of e-voting and e-counting systems is a world away from the fully verifiable systems being designed by cryptographers