Schulze method

The Schulze method (/ˈʃʊltsə/) is an electoral system developed in 1997 by Markus Schulze that selects a single winner using votes that express preferences. The method can also be used to create a sorted list of winners. The Schulze method is also known as Schwartz Sequential dropping (SSD), cloneproof Schwartz sequential dropping (CSSD), the beatpath method, beatpath winner, path voting, and path winner.

The Schulze method is a Condorcet method, which means that if there is a candidate who is preferred by a majority over every other candidate in pairwise comparisons, then this candidate will be the winner when the Schulze method is applied.

The output of the Schulze method (defined below) gives an ordering of candidates. Therefore, if several positions are available, the method can be used for this purpose without modification, by letting the k top-ranked candidates win the k available seats. Furthermore, for proportional representation elections, a single transferable vote variant has been proposed.

The Schulze method is used by several organizations including Debian, Ubuntu, Gentoo, Software in the Public Interest, Free Software Foundation Europe, Pirate Party political parties and many others.

Description of the method

Ballot

The input for the Schulze method is the same as for other ranked single-winner electoral systems: each voter must furnish an ordered preference list on candidates where ties are allowed (a strict weak order).[1]

One typical way for voters to specify their preferences on a ballot (see right) is as follows. Each ballot lists all the candidates, and each voter ranks this list in order of preference using numbers: the voter places a '1' beside the most preferred candidate(s), a '2' beside the second-most preferred, and so forth. Each voter may optionally:

  • give the same preference to more than one candidate. This indicates that this voter is indifferent between these candidates.
  • use non-consecutive numbers to express preferences. This has no impact on the result of the elections, since only the order in which the candidates are ranked by the voter matters, and not the absolute numbers of the preferences.
  • keep candidates unranked. When a voter doesn't rank all candidates, then this is interpreted as if this voter (i) strictly prefers all ranked to all unranked candidates, and (ii) is indifferent among all unranked candidates.

Computation

Let be the number of voters who prefer candidate to candidate .

A path from candidate to candidate is a sequence of candidates with the following properties:

  1. and .
  2. For all .

In other words, in a pairwise comparison each candidate in the path will beat the following candidate.

The strength of a path from candidate to candidate is the smallest number of voters in the sequence of comparisons:

For all .

For a pair of candidates and that are connected by at least one path, the strength of the strongest path is the maximum strength of the path(s) connecting them. If there is no path from candidate to candidate at all, then .

Candidate is better than candidate if and only if .

Candidate is a potential winner if and only if for every other candidate .

It can be proven that and together imply .[1]:§4.1 Therefore, it is guaranteed (1) that the above definition of "better" really defines a transitive relation and (2) that there is always at least one candidate with for every other candidate .

Example

In the following example 45 voters rank 5 candidates.

The pairwise preferences have to be computed first. For example, when comparing A and B pairwise, there are 5+5+3+7=20 voters who prefer A to B, and 8+2+7+8=25 voters who prefer B to A. So and . The full set of pairwise preferences is:

Directed graph labeled with pairwise preferences d[*, *]
Matrix of pairwise preferences
20263022
25163318
19291724
15122814
23272131

The cells for d[X, Y] have a light green background if d[X, Y] > d[Y, X], otherwise the background is light red. There is no undisputed winner by only looking at the pairwise differences here.

Now the strongest paths have to be identified. To help visualize the strongest paths, the set of pairwise preferences is depicted in the diagram on the right in the form of a directed graph. An arrow from the node representing a candidate X to the one representing a candidate Y is labelled with d[X, Y]. To avoid cluttering the diagram, an arrow has only been drawn from X to Y when d[X, Y] > d[Y, X] (i.e. the table cells with light green background), omitting the one in the opposite direction (the table cells with light red background).

One example of computing the strongest path strength is p[B, D] = 33: the strongest path from B to D is the direct path (B, D) which has strength 33. But when computing p[A, C], the strongest path from A to C is not the direct path (A, C) of strength 26, rather the strongest path is the indirect path (A, D, C) which has strength min(30, 28) = 28. The strength of a path is the strength of its weakest link.

For each pair of candidates X and Y, the following table shows the strongest path from candidate X to candidate Y in red, with the weakest link underlined.

Strongest paths
To
From
ABCDE
A N/A
A-(30)-D-(28)-C-(29)-B
A-(30)-D-(28)-C
A-(30)-D
A-(30)-D-(28)-C-(24)-E
A
B
B-(25)-A
N/A
B-(33)-D-(28)-C
B-(33)-D
B-(33)-D-(28)-C-(24)-E
B
C
C-(29)-B-(25)-A
C-(29)-B
N/A
C-(29)-B-(33)-D
C-(24)-E
C
D
D-(28)-C-(29)-B-(25)-A
D-(28)-C-(29)-B
D-(28)-C
N/A
D-(28)-C-(24)-E
D
E
E-(31)-D-(28)-C-(29)-B-(25)-A
E-(31)-D-(28)-C-(29)-B
E-(31)-D-(28)-C
E-(31)-D
N/A E
ABCDE
From
To
Strengths of the strongest paths
28283024
25283324
25292924
25282824
25282831

Now the output of the Schulze method can be determined. For example, when comparing A and B, since , for the Schulze method candidate A is better than candidate B. Another example is that , so candidate E is better than candidate D. Continuing in this way, the result is that the Schulze ranking is , and E wins. In other words, E wins since for every other candidate X.

Implementation

The only difficult step in implementing the Schulze method is computing the strongest path strengths. However, this is a well-known problem in graph theory sometimes called the widest path problem. One simple way to compute the strengths therefore is a variant of the Floyd–Warshall algorithm. The following pseudocode illustrates the algorithm.

 1 # Input: d[i,j], the number of voters who prefer candidate i to candidate j.
 2 # Output: p[i,j], the strength of the strongest path from candidate i to candidate j.
 3 
 4 for i from 1 to C
 5    for j from 1 to C
 6       if (i ≠ j) then
 7          if (d[i,j] > d[j,i]) then
 8             p[i,j] := d[i,j]
 9          else
10             p[i,j] := 0
11 
12 for i from 1 to C
13    for j from 1 to C
14       if (i ≠ j) then
15          for k from 1 to C
16             if (i ≠ k and j ≠ k) then
17                p[j,k] := max ( p[j,k], min ( p[j,i], p[i,k] ) )

This algorithm is efficient, and has running time O(C3) where C is the number of candidates.

Ties and alternative implementations

When allowing users to have ties in their preferences, the outcome of the Schulze method naturally depends on how these ties are interpreted in defining d[*,*]. Two natural choices are that d[A, B] represents either the number of voters who strictly prefer A to B (A>B), or the margin of (voters with A>B) minus (voters with B>A). But no matter how the ds are defined, the Schulze ranking has no cycles, and assuming the ds are unique it has no ties.[1]

Although ties in the Schulze ranking are unlikely,[2] they are possible. Schulze's original paper[1] proposed breaking ties in accordance with a voter selected at random, and iterating as needed.

An alternative way to describe the winner of the Schulze method is the following procedure:

  1. draw a complete directed graph with all candidates, and all possible edges between candidates
  2. iteratively [a] delete all candidates not in the Schwartz set (i.e. any candidate x which cannot reach all others who reach x) and [b] delete the graph edge with the smallest value (if by margins, smallest margin; if by votes, fewest votes).
  3. the winner is the last non-deleted candidate.

There is another alternative way to demonstrate the winner of the Schulze method. This method is equivalent to the others described here, but the presentation is optimized for the significance of steps being visually apparent as you go through it, not for computation.

  1. Make the results table, such as used above in the Example and called the 'matrix of pairwise preferences'. If using margins rather than raw vote totals, subtract it from its transpose. Then every positive number is a pairwise win for the candidate on that row (and marked green), ties are 0, and losses are negative (marked red). Order the candidates by how long they last in elimination.
  2. If there's a candidate with no red on their line, they win.
  3. Otherwise, draw a square box around the Schwartz set, in the upper left corner. You can describe it as the minimal 'winner's circle' of candidates who do not lose to anyone outside the circle. Note that to the right of the box there is no red, which means it's a winner's circle, and note that within the box there is no reordering possible that will produce a smaller winner's circle (if the Schwartz set has no more than three candidates, this not hard).
  4. Cut away every part of the table that isn't in the box.
  5. Otherwise, something needs to be compromised on - every candidate lost some race, and the loss we tolerate the best is the . So, take the red cell with the highest number (if going by margins, the least negative), make it green, or anyway not the same color as a loss, and go back step 2.

Here is a margins table made from the above Example. Note the change of order used for demonstration purposes.

Initial Results Table
EACBD
E 1-3917
A -17-515
C 3-713-11
B -95-1321
D -17-1511-21

The first drop doesn't help shrink the Schwartz set.

First Drop
EACBD
E 1-3917
A -17-515
C 3-713-11
B -95-1321
D -17-1511-21

So we get straight to the second drop, and that shows us the winner, E, with its clear row.

Second Drop, final
EACBD
E 1-3917
A -17-515
C 3-713-11
B -95-1321
D -17-1511-21

This method can also be used to calculate a result, if you make the table in such a way that you can conveniently and reliably rearrange the order of the candidates on both row and column (use the same order on both at all times).

Satisfied and failed criteria

Satisfied criteria

The Schulze method satisfies the following criteria:

Failed criteria

Since the Schulze method satisfies the Condorcet criterion, it automatically fails the following criteria:

Likewise, since the Schulze method is not a dictatorship and agrees with unanimous votes, Arrow's Theorem implies it fails the criterion

The Schulze method also fails

Comparison table

The following table compares the Schulze method with other preferential single-winner election methods:

Comparison of Preferential Electoral Systems
Monotonic Condorcet Majority Condorcet loser Majority loser Mutual majority Smith ISDA LIIA Clone independence Reversal symmetry Participation, Consistency Later-noharm Later-nohelp Polynomial time Resolvability
Schulze YesYesYesYesYesYesYesYesNoYesYesNoNoNoYesYes
Ranked pairs YesYesYesYesYesYesYesYesYesYesYesNoNoNoYesYes
Alternative smith NoYesYesYesYesYesYesYesNoYesNoNoNoNoYesYes
Alternative schwartz NoYesYesYesYesYesYesYesNoYesNoNoNoNoYesYes
Kemeny-Young YesYesYesYesYesYesYesYesYesNoYesNoNoNoNoYes
Copeland YesYesYesYesYesYesYesYesNoNoYesNoNoNoYesNo
Nanson NoYesYesYesYesYesYesNoNoNoYesNoNoNoYesYes
Instant-runoff voting NoNoYesYesYesYesNoNoNoYesNoNoYesYesYesYes
Borda YesNoNoYesYesNoNoNoNoNoYesYesNoYesYesYes
Baldwin NoYesYesYesYesYesYesNoNoNoNoNoNoNoYesYes
Bucklin YesNoYesNoYesYesNoNoNoNoNoNoNoYesYesYes
Plurality YesNoYesNoNoNoNoNoNoNoNoYesYesYesYesYes
Contingent voting NoNoYesYesYesNoNoNoNoNoNoNoYesYesYesYes
Coombs[4] NoNoYesYesYesYesNoNoNoNoNoNoNoNoYesYes
MiniMax YesYesYesNoNoNoNoNoNoNoNoNoNoNoYesYes
Anti-plurality[4] YesNoNoNoYesNoNoNoNoNoNoYesNoNoYesYes
Sri Lankan contingent voting NoNoYesNoNoNoNoNoNoNoNoNoYesYesYesYes
Supplementary voting NoNoYesNoNoNoNoNoNoNoNoNoYesYesYesYes
Dodgson[4] NoYesYesNoNoNoNoNoNoNoNoNoNoNoNoYes

The main difference between the Schulze method and the ranked pairs method can be seen in this example:

Suppose the MinMax score of a set X of candidates is the strength of the strongest pairwise win of a candidate A ∉ X against a candidate B ∈ X. Then the Schulze method, but not Ranked Pairs, guarantees that the winner is always a candidate of the set with minimum MinMax score.[1]:§4.8 So, in some sense, the Schulze method minimizes the largest majority that has to be reversed when determining the winner.

On the other hand, Ranked Pairs minimizes the largest majority that has to be reversed to determine the order of finish, in the minlexmax sense. [5] In other words, when Ranked Pairs and the Schulze method produce different orders of finish, for the majorities on which the two orders of finish disagree, the Schulze order reverses a larger majority than the Ranked Pairs order.

History

The Schulze method was developed by Markus Schulze in 1997. It was first discussed in public mailing lists in 1997–1998[6] and in 2000.[7] Subsequently, Schulze method users included Software in the Public Interest (2003),[8] Debian (2003),[9] Gentoo (2005),[10] Topcoder (2005),[11] Wikimedia (2008),[12] KDE (2008),[13] the Free Software Foundation Europe (2008),[14] the Pirate Party of Sweden (2009),[15] and the Pirate Party of Germany (2010).[16] In the French Wikipedia, the Schulze method was one of two multi-candidate methods approved by a majority in 2005,[17] and it has been used several times.[18] The newly formed Boise, Idaho chapter of the Democratic Socialists of America in February chose this method for their first special election to be held in March (2018). [19]

In 2011, Schulze published the method in the academic journal Social Choice and Welfare.[1]

Users

sample ballot for Wikimedia's Board of Trustees elections

The Schulze method is not currently used in parliamentary elections. However, it has been used for parliamentary primaries in the Swedish Pirate Party. It is also starting to receive support in other public organizations. Organizations which currently use the Schulze method include:

Notes

  1. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Markus Schulze, A new monotonic, clone-independent, reversal symmetric, and condorcet-consistent single-winner election method, Social Choice and Welfare, volume 36, number 2, page 267–303, 2011. Preliminary version in Voting Matters, 17:9-19, 2003.
  2. Under reasonable probabilistic assumptions when the number of voters is much larger than the number of candidates
  3. 1 2 3 Douglas R. Woodall, Properties of Preferential Election Rules, Voting Matters, issue 3, pages 8-15, December 1994
  4. 1 2 3 Anti-plurality, Coombs and Dodgson are assumed to receive truncated preferences by apportioning possible rankings of unlisted alternatives equally; for example, ballot A > B = C is counted as  A > B > C and  A > C > B. If these methods are assumed not to receive truncated preferences, then Later-no-harm and Later-no-help are not applicable.
  5. Tideman, T. Nicolaus, "Independence of clones as a criterion for voting rules," Social Choice and Welfare vol 4 #3 (1987), pp 185-206.
  6. See:
  7. See:
  8. 1 2 Process for adding new board members, January 2003
  9. 1 2 See:
  10. 1 2 See:
  11. 1 2 2007 TopCoder Collegiate Challenge, September 2007
  12. See:
  13. 1 2 section 3.4.1 of the Rules of Procedures for Online Voting
  14. 1 2 See:
  15. 1 2 See:
  16. 1 2 11 of the 16 regional sections and the federal section of the Pirate Party of Germany are using LiquidFeedback for unbinding internal opinion polls. In 2010/2011, the Pirate Parties of Neukölln (link), Mitte (link), Steglitz-Zehlendorf (link), Lichtenberg (link), and Tempelhof-Schöneberg (link) adopted the Schulze method for its primaries. Furthermore, the Pirate Party of Berlin (in 2011) (link) and the Pirate Party of Regensburg (in 2012) (link) adopted this method for their primaries.
  17. 1 2 Choix dans les votes
  18. fr:Spécial:Pages liées/Méthode Schulze
  19. Chumich, Andrew. "DSA Special Election". Retrieved 2018-02-25.
  20. Article 7.1.3 of its Working Format of the Agora, p. 54, July 2016
  21. Election of the Annodex Association committee for 2007, February 2007
  22. Ajith, Van Atta win ASG election, April 2013
  23. §6 and §7 of its bylaws, May 2014
  24. §9a of the bylaws, October 2013
  25. See:
  26. Project Logo, October 2009
  27. resolution, December 2013
  28. Civics Meeting Minutes, March 2012
  29. Adam Helman, Family Affair Voting Scheme - Schulze Method
  30. See:
  31. "Guidance Document". Eudec.org. 2009-11-15. Retrieved 2010-05-08.
  32. Democratic election of the server admins, July 2010
  33. Campobasso. Comunali, scattano le primarie a 5 Stelle, February 2014
  34. Fondi, il punto sui candidati a sindaco. Certezze, novità e colpi di scena, March 2015
  35. article 25(5) of the bylaws, October 2013
  36. 2° Step Comunarie di Montemurlo, November 2013
  37. article 12 of the bylaws, January 2015
  38. Ridefinizione della lista di San Cesareo con Metodo Schulze, February 2014
  39. article 57 of the statutory rules
  40. Voters Guide, September 2011
  41. See:
  42. §7(3) of the voting rules, November 2015
  43. GnuPG Logo Vote, November 2006
  44. §14 of the bylaws
  45. "User Voting Instructions". Gso.cs.binghamton.edu. Retrieved 2010-05-08.
  46. Haskell Logo Competition, March 2009
  47. "Hillegass-Parker House Bylaws § 5. Elections". Hillegass-Parker House website. Retrieved 4 October 2015.
  48. article VI section 10 of the bylaws, November 2012
  49. A club by any other name ..., April 2009
  50. See:
  51. Knight Foundation awards $5000 to best created-on-the-spot projects, June 2009
  52. Kubuntu Council 2013, May 2013
  53. See:
  54. article 8.3 of the bylaws
  55. The Principles of LiquidFeedback. Berlin: Interaktive Demokratie e. V. 2014. ISBN 978-3-00-044795-2.
  56. "Madisonium Bylaws - Adopted". Google Docs.
  57. "Wahlmodus" (in German). Metalab.at. Retrieved 2010-05-08.
  58. Benjamin Mako Hill, Voting Machinery for the Masses, July 2008
  59. See:
  60. bylaws, September 2014
  61. "2009 Director Elections". noisebridge.net.
  62. "Online Voting Policy". openembedded.org.
  63. See:
  64. Election Process, June 2016
  65. National Congress 2011 Results, November 2011
  66. §6(10) of the bylaws
  67. The Belgian Pirate Party Announces Top Candidates for the European Elections, January 2014
  68. bylaws
  69. Rules adopted on 18 December 2011
  70. Verslag ledenraadpleging 4 januari, January 2015
  71. "23 January 2011 meeting minutes". pirateparty.org.nz.
  72. Piratenversammlung der Piratenpartei Schweiz, September 2010
  73. Committee Elections, April 2012
  74. Squeak Oversight Board Election 2010, March 2010
  75. See:
  76. Election status update, September 2009
  77. §10 III of its bylaws, June 2013
  78. Minutes of the 2010 Annual Sverok Meeting, November 2010
  79. article VI section 6 of the bylaws
  80. Ubuntu IRC Council Position, May 2012
  81. See:
  82. "/v/GAs - Pairwise voting results". vidyagaemawards.com.
  83. See e.g. here (May 2009), here (August 2009), and here (December 2009).
  84. See here and here.
  85. "Девятнадцатые выборы арбитров, второй тур" [Result of Arbitration Committee Elections]. kalan.cc. Archived from the original on 2015-02-22.
  86. See here

  • Official website
  • Condorcet Computations by Johannes Grabmeier
  • Spieltheorie (in German) by Bernhard Nebel
  • Accurate Democracy by Rob Loring
  • Christoph Börgers (2009), Mathematics of Social Choice: Voting, Compensation, and Division, SIAM, ISBN 0-89871-695-0
  • Nicolaus Tideman (2006), Collective Decisions and Voting: The Potential for Public Choice, Burlington: Ashgate, ISBN 0-7546-4717-X
  • preftools by the Public Software Group
  • Arizonans for Condorcet Ranked Voting
  • Condorcet PHP PHP library supporting multiple Condorcet methods, including that of Schulze.
  • Implementation in Java
  • Implementation in Ruby
  • Implementation in Python 2
  • Implementation in Python 3
This article is issued from Wikipedia. The text is licensed under Creative Commons - Attribution - Sharealike. Additional terms may apply for the media files.