Brief Descriptions of Voting Methods, Part 2
Bruce Anderson
landerso at ida.org
Sun Jun 2 04:40:38 PDT 1996
A. Elemental Direct and Sequential Deletion Methods (continued):
/Copeland/2a/: is sequential deletion by Copeland's score (i.e., by # of
pairwise wins minus # of pairwise losses, or by any of the equivalent scores
listed in Part 1). The /2a/ means (in my notation) to delete all of the
candidates tied with the lowest Copeland score at the same time (provided this
does not delete every remaining candidate), and to continue until only 1
candidate is left (or until all of the remaining candidates have the same
Copeland score).
Copeland-mod: selects the candidate(s) with no pairwise losses if any such
candidate(s) exist(s); otherwise, it selects the candidate(s) with the best
pairwise won-lost percentage ignoring pairwise ties (as if such ties never
occurred). Copeland-mod reduces to Copeland when there are no pairwise ties.
Dodgson: selects the candidate(s) that minimize the number of (contiguous) pair
reversals in voters rankings needed to make that candidate not lose any pairwise
matchup.
Fishburn: selects the candidate(s) that is (are) not a Fishburn losers. A
candidate, say i, is a Fishburn loser if there is some other candidate, say j,
such that every candidate that pairwise beats j also pairwise beats i, and there
is at least one candidate that pairwise beats i but does not pairwise beat j.
Hare: is sequential deletion by least first-place rankings, i.e., by the
minimum of b(i,1) -- see Part 1 for the definition of b. Hare is also called
Majority Preference Voting, Single Transferable Voting (with one winner), and
some other IMHO less desirable names. In my notation, Hare is equivalent to
/Plurality/1a/.
Kemeny: Consider all n! permutations of the n candidates. Give each
permutation 1 point for every voter and every distinct pair of candidates such
that both the voter and the permutation agree on the relative ranking of that
pair of candidates. Give each permutation 1/2 a point for every voter and every
distinct pair of candidates such that the voter ranks the two candidates in that
pair has been tied with each other. Kemeny selects the candidate(s) that is
(are) ranked first on the (any) permutation that receives the maximum number of
these points.
Max-Tourneys: Consider all of the meaningfully different ways of seeding a
(balanced, non-adaptive) single elimination tournament among the n candidates.
(I have proven that the number of such seedlings is much less than n!.)
Max-Tourneys selects the candidate(s) that win(s) the most such tournaments
(counting pairwise ties in the logical manner).
Nanson: is sequential deletion by Borda score. Using my notation, I define
Nanson as being equivalent to /Borda/1a/ (i.e., delete at most one candidate at
a time), but others define Nanson as being equivalent to /Borda/2a/.
Plurality: selects the candidate(s) with largest number of first-place
rankings, i.e., by the maximum of b(i,1).
Plurality-w/runoff: selects the candidate that is the majority winner, if such
a candidate exists; otherwise, it selects the pairwise non-loser(s) between the
two candidates that receive more first-place rankings (i.e., higher b(i,1)) than
any other candidate, if exactly two such candidates exist; otherwise, it selects
any candidate such that no more than one other candidate receives more
first-place rankings than it does and that is a pairwise non-loser in some
matchup between it and some other such "first or second place" candidate.
Schwartz: selects the candidate(s) that is (are) not a Schwartz losers. A
candidate, say i, is a Schwartz loser if there are one or more other candidates,
say j1,...,jm, such that j1 pairwise beats ... who pairwise beats jn who
pairwise beats i, but there are not one or more other candidates, say k1,...,kn,
where k1 is j1, such that i pairwise beats kn who pairwise beats ... who
pairwise beats k1 who is j1.
Sister: selects the candidate(s) with the most pairwise wins, with ties in
most wins being broken by least pairwise losses -- I am indebted to Mike for
this simplification of my original, more complex definition. Sister reduces
to Copeland when there are no pairwise ties.
Smith: A candidate is a Smith winner if it belongs to the smallest non-empty
set of candidates such that each candidate in that set pairwise beats each
candidate not in that set.
Young: selects the candidate(s) that minimize the number of voters whose
ballots need to be ignored in order to make that candidate not lose any pairwise
matchup.
Beats-all: selects the candidate that pairwise beats each other candidate, if
such a candidate exists; otherwise, it selects all of the candidates under
consideration.
Plurality-ext: selects the candidate(s) with largest number of first-place
rankings, i.e., by the maximum of b(i,1), with ties broken by the largest number
of second-place rankings (i.e., by b(i,2)), and so on.
To be continued.
Bruce
More information about the Election-Methods
mailing list