An Election Collection (June 2009: now includes over 100 elections!)

Tideman data collection: 87 ranked-ballot elections

The data in these files were gathered by two different investigators. Elections 1-32 and 48-99 were gathered in 1987 and 1988 respectively, by Nicolaus Tideman, with support from NSF grant SES86-18328. Elections 33-35 were gathered by David Hill whose address is Laverton, Berry Lane, Chorleywood, Rickmansworth, Herts, WD3 5EY, England. Numbers 36-47 were not assigned.

The data are records of ballots from elections of British organizations (mostly trade unions using PR-STV or IRV voting) in which the voters ranked the candidates. The data were gathered under a stipulation that the organizations involved would remain anonymous.

The data are in a format developed by David Hill. The first line contains the number of candidates and the number to be elected. (Many but not all elections were multiwinner.) In subsequent lines that represent ballot papers, the first number is always 1. (The format was designed for a counting program that treats the first number as the number of instances of the ordering of the candidates on the line.) Next on these lines is a sequence of numbers representing a voter's reported ranking: The number of the candidate ranked first, the number of the candidate ranked second, and so on. The end of the reported ranking is signaled by a zero. A zero at the beginning of the ranking is a signal that the list of ballot papers has ended.

Next come the names of the candidates, each in parentheses, as required by the counting program, and finally the name of the election.

Invalid votes and a small number of references to write-in candidates were deleted from the data.

Any questions or comments regarding the data should be addressed to:

     Nicolaus Tideman       
     Economics Department   
     Virginia Tech          
     Blacksburg VA 24061    
     (540) 231-7592         
     e-mail: ntideman AT             

Comments on Tideman data

The least number of voters in any Tideman election is 9, in A33.HIL, which has 18 candidates. The greatest number of voters in any Tideman election is 3422, in A9.HIL, which had 12 candidates and 11 winners (although this could be thought of as a 1-loser election!); this also is the greatest number of winners in any Tideman election. The greatest number of candidates in any Tideman election is 26, in A13.HIL (with 849 voters). The least number of candidates in any Tideman election is 3, in several elections including A77.HIL. The average number of voters in the 87 Tideman elections is 410.

2010 UK Labour Party leader election data

1998 skating data

Sandra J. Loosemore collected all the judge's rankings of the skaters in the 48 skating subcontests of the 4 top skating contests during 1998. Note: actually each judge rated each skate with a real number from 0 to 6 (such as "5.9") but those real numbers were ignored by the voting system for allocating gold, silver and bronze medals – which converted them all to rank-orders and from then onward the numbers played no role. Unfortunately Loosemore did not provide the original real numbers. But she did provide the rank-orders. Read down the Kth column for each event to get the Kth judge's rank ordering, where "1" is the rank of the best skater according to that judge. It appears there were always 9 judges (voters). For example, in the Europeans, the first judge ranked TELENKOV 27th in the men's short program.


Additional Elections (beyond those collected by Tideman)

Anybody who wants to contribute additional new election data (preferably in the same format or very near...) please contact me:
   warren.wds at

WDS has added the annual Debian Leader Elections 2001-2006. These all were single-winner elections conducted using Schulze-beatpaths Condorcet voting method. The number of voters ranged from 311 to 505 and the number of candidates from 4 to 8. The last candidate was always "None of The Above" (which can be viewed as an "approval threshold"; so far, NOTA has never won a Debian election). In these elections some votes expressed equal rankings, e.g. ranking two candidates both 1 or both 2 (this was allowed by the rules). Unfortunately Tideman's HIL format does not allow equal rankings, so the votes have been converted to strict rankings by favoring the lower-indexed candidate to break any and all such ties.

Tideman data files

A10.HIL, A11.HIL, A12.HIL, A13.HIL, A14.HIL, A15.HIL, A16.HIL, A17.HIL, A18.HIL, A19.HIL,
A20.HIL, A21.HIL, A22.HIL, A23.HIL, A24.HIL, A25.HIL, A26.HIL, A27.HIL, A28.HIL, A29.HIL,
A30.HIL, A31.HIL, A32.HIL, A33.HIL, A34.HIL, A35.HIL,
A48.HIL, A49.HIL,
A50.HIL, A51.HIL, A52.HIL, A53.HIL, A54.HIL, A55.HIL, A56.HIL, A57.HIL, A58.HIL, A59.HIL,
A60.HIL, A61.HIL, A62.HIL, A63.HIL, A64.HIL, A65.HIL, A66.HIL, A67.HIL, A68.HIL, A69.HIL,
A70.HIL, A71.HIL, A72.HIL, A73.HIL, A74.HIL, A75.HIL, A76.HIL, A77.HIL, A78.HIL, A79.HIL,
A80.HIL, A81.HIL, A82.HIL, A83.HIL, A84.HIL, A85.HIL, A86.HIL, A87.HIL, A88.HIL, A89.HIL,
A90.HIL, A91.HIL, A92.HIL, A93.HIL, A94.HIL, A95.HIL, A96.HIL, A97.HIL, A98.HIL, A99.HIL

Debian Leader Election data files

In Tideman's format:

D01.HIL, D02.HIL, D03.HIL, D04.HIL, D05.HIL, D06.HIL

In the format originally provided by Debian (which is better since it allows equal rankings in ballots):

leader2001_tally.txt,   leader2002_tally.txt,
leader2003_tally.txt,   leader2004_tally.txt,
leader2005_tally.txt,   leader2006_tally.txt,

Same thing but cleaned up to make them all the same, machine-readable, format:

DB2001.DEB,   DB2002.DEB,   DB2003.DEB,   DB2004.DEB,   DB2005.DEB,   DB2006.DEB,

Multiwinner elections carried out using reweighted range voting

A real-world high stakes election with 9 winners and 39 voters that was carried out with RRV (you will need to view it with wide window).

More election data

The "Haiku" icon election is another dataset. Raw data for it.

The San Francisco 2007 Mayoral race (about 150000 voters; 12 on-ballot candidates plus 6 write-ins; conducted using top-3-only bastardized IRV, won by Gavin Newsome; also G.N. won under bastardized Borda count and with every Condorcet method). Format with one unique-ballot-type per line, each line comes with count of how many times that ballot was cast. UNDERVOTES and OVERVOTES are two pseudo-candidates used by San Francisco to denote somebody leaving that slot in the 3-slot-rank-order blank or ranking two or more candidates in that slot, respectively (I presume).

The Dublin North STV race (13 candidates, 43943 voters), Dublin West (10 candidates, 29988 voters) and County Meath (14 candidates, 64081 voters) multiwinner STV races (all same format, all year 2002 and anonymized). The ballots from these three races were unusually made publicly electronically available by Ireland as an experiment (which apparently was never repeated) and I converted them to the present format. In the County Meath 14-candidate race, voters ranked an average of 6.15 candidates; candidate "B" was the IRV and Condorcet beats-all winner; but "Q" was the Borda Count winner.

The Burlington VT 2009 Mayoral race was conducted using IRV. There were 5 on-ballot candidates (plus write-ins) and about 9000 voters. It exhibited numerous pathologies. Data supplied by Juho Laatu based on processing data from Burlington web site.

The Gay Asian Pacific Association Foundation 2013 Pride Run Beneficiary Election, 86 voters, 14 candidates, $8000-$10000 at stake. Also similar 2012 beneficiary election. Both supplied by Dominic Paris.

Greg Dennis has collected a database of STV elections he calls Open STV.

More elections to be added (hopefully soon)...

Return to main page