Economics Letters 236 (2024) 111589
Contents lists available at ScienceDirect
Economics Letters
journal homepage: www.elsevier.com/locate/ecolet
An impossibility theorem concerning positive involvement in voting
Wesley H. Holliday ∗
University of California, Berkeley, United States of America
ARTICLE
INFO
Dataset link: https://github.com/wesholliday/
pos-inv
Keywords:
Social choice theory
Impossibility theorem
Positive involvement
Condorcet winner
Condorcet loser
Resolvability
ABSTRACT
In social choice theory with ordinal preferences, a voting method satisfies the axiom of positive involvement if
adding to a preference profile a voter who ranks an alternative uniquely first cannot cause that alternative to
go from winning to losing. In this note, we prove a new impossibility theorem concerning this axiom: there
is no ordinal voting method satisfying positive involvement that also satisfies the Condorcet winner and loser
criteria, resolvability, and a common invariance property for Condorcet methods, namely that the choice of
winners depends only on the ordering of majority margins by size.
1. Introduction
A basic assumption of democratic voting is that increased voter
support for a candidate should not harm that candidate’s chances
of winning. In social choice theory with ordinal preferences, one
formalization of this idea is given by the axiom of positive involvement (Saari 1995): adding to a preference profile a voter who ranks
an alternative uniquely first cannot cause that alternative to go from
winning to losing.1 Perhaps surprisingly, many well-known voting
methods—especially Condorcet-consistent voting methods—violate this
axiom (see Pérez 1995). In this note, we prove a new impossibility
theorem in this vein: there is no ordinal voting method satisfying
positive involvement that also satisfies the Condorcet winner and
loser criteria (Condorcet 1785), resolvability (Tideman 1986), and a
common invariance property for Condorcet methods, namely that the
choice of winners depends only on the ordering of majority margins by
size. Methods satisfying this ordinal margin invariance include Minimax
(Simpson 1969, Kramer 1977), Ranked Pairs (Tideman 1987), Beat
Path (Schulze 2011), Weighted Covering (Dutta and Laslier 1999,
Pérez-Fernández and De Baets 2018), Split Cycle (Holliday and Pacuit
2023a), and Stable Voting (Holliday and Pacuit 2023b). Thus, our
theorem helps explain why none of these voting methods satisfies all
the stated axioms, as well as motivating the search for new methods
satisfying all the axioms except for ordinal margin invariance.
2. Preliminaries
Fix infinite sets and of alternatives and voters, respectively. A
profile 𝐏 is a function from some nonempty finite 𝑉 (𝐏) ⊆ , called the
set of voters in 𝐏, to the set of strict weak orders on some nonempty
finite 𝑋(𝐏) ⊆ , called the set of alternatives in 𝐏.2 For 𝑖 ∈ 𝑉 (𝐏) and
𝑥, 𝑦 ∈ 𝑋(𝐏), take (𝑥, 𝑦) ∈ 𝐏(𝑖) to mean that 𝑖 strictly prefers 𝑥 to 𝑦. 𝐏 is
linear if for each 𝑖 ∈ 𝑉 (𝐏), 𝐏(𝑖) is a linear order.3 Given 𝑥, 𝑦 ∈ 𝑋(𝐏), the
margin of 𝑥 over 𝑦 is defined by
Margin𝐏 (𝑥, 𝑦) = Support𝐏 (𝑥, 𝑦) − Support𝐏 (𝑦, 𝑥), where
Support𝐏 (𝑎, 𝑏) = |{𝑖 ∈ 𝑉 (𝐏) ∣ (𝑎, 𝑏) ∈ 𝐏(𝑖)}|.
A voting method is a function 𝐹 assigning to each profile 𝐏 a nonempty
𝐹 (𝐏) ⊆ 𝑋(𝐏).4 If |𝐹 (𝐏)| > 1, we assume some further (possibly random)
tiebreaking process ultimately narrows 𝐹 (𝐏) down to one alternative.
∗ Correspondence to: 314 Philosophy Hall #2390, University of California, Berkeley, CA 94720-2390, United States of America.
E-mail address: wesholliday@berkeley.edu.
1
Of course, such a voter expresses more than just support for their favorite alternative if they also rank other alternatives. A weaker formalization is therefore
bullet-vote positive involvement : adding a voter who ranks an alternative uniquely first and all other alternatives in a tie below their favorite cannot cause their
favorite to go from winning to losing.
2
The proof of Theorem 4 easily adapts to a setting where the set of alternatives cannot vary between profiles, assuming at least four alternatives (for more,
add indefensible ones below the four in the proof). But we need the set of voters to be variable.
3
Theorem 4 holds even if we restrict the domain of voting methods to linear profiles, but we allow non-linear profiles in our setup for the sake of the
distinction between the two parts of Lemma 2.
4
Thus, we build the axiom of universal domain into the definition of our voting methods.
https://doi.org/10.1016/j.econlet.2024.111589
Received 10 January 2024; Received in revised form 4 February 2024; Accepted 5 February 2024
Available online 6 February 2024
0165-1765/© 2024 The Author. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Economics Letters 236 (2024) 111589
W.H. Holliday
Margin𝐏 (𝑧, 𝑦) < Margin𝐏 (𝑦, 𝑥). Let 𝑘 = 𝑚𝑎𝑥{Margin𝐏 (𝑧, 𝑦) ∣ 𝑧 ∈ 𝑋(𝐏)},
so 𝑘 < Margin𝐏 (𝑦, 𝑥). Since 𝐏 is linear, all margins have the same
parity, so 𝑘 + 1 < Margin𝐏 (𝑦, 𝑥). Let 𝐏′ be obtained from 𝐏 by adding
𝑘 + 1 voters who rank 𝑥 uniquely first and 𝑦 uniquely second, followed
by any linear order of the remaining alternatives. Then by positive
involvement, 𝑥 ∈ 𝐹 (𝐏′ ). But in 𝐏′ , 𝑦 is the Condorcet winner, so by
the Condorcet winner criterion, 𝐹 (𝐏′ ) = {𝑦}, contradicting 𝑥 ∈ 𝐹 (𝐏′ ).
Thus, 𝑥 ∈ 𝐷(𝐏).
For part 2 and an arbitrary 𝐏, the argument is similar only the
margins may have different parities, so we cannot infer from 𝑘 <
Margin𝐏 (𝑦, 𝑥) that 𝑘 + 1 < Margin𝐏 (𝑦, 𝑥). In this case, we add only
𝑘 voters who rank 𝑥 uniquely first and 𝑦 uniquely second, followed
by any linear order of the remaining alternatives. Then in 𝐏′ , 𝑦 is
a weak Condorcet winner, and Margin𝐏′ (𝑦, 𝑥) > 0. It follows by the
weak Condorcet winner criterion that 𝑥 ∉ 𝐹 (𝐏′ ), contradicting positive
involvement. □
A voting method 𝐺 refines 𝐹 (for some class of profiles) if for any profile
𝐏 (in the class), 𝐺(𝐏) ⊆ 𝐹 (𝐏).
A voting method 𝐹 satisfies positive involvement (see Saari 1995,
Pérez 2001, Holliday and Pacuit 2021) if for any profiles 𝐏, 𝐏′ , if
𝑥 ∈ 𝐹 (𝐏) and 𝐏′ is obtained from 𝐏 by adding one voter who ranks
𝑥 uniquely in first place, then 𝑥 ∈ 𝐹 (𝐏′ ).5 Thus, you cannot cause 𝑥 to
lose by ranking 𝑥 first. According to Pérez (2001, p. 605), this ‘‘may
be seen as the minimum to require concerning the coherence in the
winning set when new voters are added’’.
A voting method 𝐹 satisfies the Condorcet winner criterion (Condorcet 1785) if for every profile 𝐏 with a Condorcet winner, 𝐹 (𝐏)
contains only the Condorcet winner. Recall that 𝑥 is a Condorcet winner (resp. weak Condorcet winner) if 𝑥 beats (resp. does not lose to)
any other alternative head-to-head, i.e., for every 𝑦 ∈ 𝑋(𝐏) ⧵ {𝑥},
Margin𝐏 (𝑥, 𝑦) > 0 (resp. Margin𝐏 (𝑥, 𝑦) ≥ 0). 𝐹 satisfies the weak
Condorcet winner criterion if for every profile 𝐏 with a weak Condorcet
winner, 𝐹 (𝐏) contains only weak Condorcet winners.
A key tool for reasoning about voting methods satisfying both
positive involvement and the Condorcet winner criterion is given by
the following voting method, appearing implicitly in Lemma 3 of Pérez
1995 and studied explicitly in Kasper et al. 2019.
Definition 1.
Given the appeal of positive involvement and the Condorcet winner
criterion, it is of significant interest to explore refinements of the
defensible set (cf. Kasper et al. 2019, Section 4). The Minimax and Split
Cycle methods satisfy these axioms (Pérez 2001, Holliday and Pacuit
2023a), so they refine the defensible set.9 However, refinements of the
defensible set may violate positive involvement. Examples include some
refinements of Split Cycle such as Beat Path, Ranked Pairs, and Stable
Voting (see Holliday and Pacuit 2023a).
Given a profile 𝐏, define the defensible set of 𝐏 as
𝐷(𝐏) = {𝑥 ∈ 𝑋(𝐏) ∣ for all 𝑦 ∈ 𝑋(𝐏), there exists 𝑧 ∈ 𝑋(𝐏) ∶
3. Impossibility
Margin𝐏 (𝑧, 𝑦) ≥ Margin𝐏 (𝑦, 𝑥)}.
Let us now try adding further axioms to our wish list. A Condorcet
loser in a profile 𝐏 is an 𝑥 ∈ 𝑋(𝐏) that loses head-to-head to every other
alternative, i.e., for every 𝑦 ∈ 𝑋(𝐏) ⧵ {𝑥}, Margin𝐏 (𝑦, 𝑥) > 0. 𝐹 satisfies
the Condorcet loser criterion if for every profile 𝐏, 𝐹 (𝐏) does not contain
a Condorcet loser.
The resolvability axiom can be formulated in two ways, either of
which works below. The first formulation (from Tideman 1986), singlevoter resolvability, states that for any profile 𝐏, if 𝐹 (𝐏) contains multiple
alternatives, then there is a 𝐏′ obtained from 𝐏 by adding only one
new voter such that 𝐹 (𝐏′ ) contains only one alternative; thus, every
tie can be broken by adding just one voter. The second formulation
(see Schulze 2011, Section 4.2.1), asymptotic resolvability, states that
for any positive integer 𝑚, in the limit as the number of voters goes
to infinity, the proportion of linear profiles 𝐏 for 𝑚 alternatives with
|𝐹 (𝐏)| > 1 goes to zero; thus, ties become vanishingly rare.
The final axiom is ordinal margin invariance. Informally, 𝐹 satisfies
this axiom if its selection of alternatives depends only on the ordering of
majority margins by size, not on the absolute margins or other features
of the profile. Formally, given a profile 𝐏, define M(𝐏) = (𝑀, ≻), the
ordinal margin graph of 𝐏, where 𝑀 is a directed graph whose set of
vertices is 𝑋(𝐏) with an edge from 𝑥 to 𝑦 when Margin𝐏 (𝑥, 𝑦) > 0, and
≻ is a strict weak order of the edges of 𝑀 such that (𝑎, 𝑏) ≻ (𝑐, 𝑑) if
Margin𝐏 (𝑎, 𝑏) > Margin𝐏 (𝑐, 𝑑). Then 𝐹 satisfies ordinal margin invariance if for any 𝐏, 𝐏′ , if M(𝐏) = M(𝐏′ ), then 𝐹 (𝐏) = 𝐹 (𝐏′ ). As a corollary
of Debord’s Theorem (Debord 1987, cf. Fischer et al. 2016, Theorem
4.1), any pair M = (𝑀, ≻) of an asymmetric directed graph and a strict
weak order of its edges is the ordinal margin graph of some profile.
Hence if 𝐹 satisfies ordinal margin invariance, we may also regard 𝐹
as a function that takes in M = (𝑀, ≻) and returns a nonempty subset
𝐹 (M) of its vertices.
As a voting method, 𝐷 is a special case, based on majority margins,
of Heitzig’s (2002) family of methods that are weakly immune to binary
arguments.6 The idea can also be found in the argumentation theory
of Dung (1995).7 Our choice of the term ‘defensible’ is based on the
following intuition. Although 𝑥 may lose head-to-head to 𝑦, possibly
prompting the supporters of 𝑦 over 𝑥 to call for the overthrow of 𝑥 in
favor of 𝑦, we can defend the choice of 𝑥 on the following grounds,
when applicable: there is another 𝑧 that beats 𝑦 head-to-head by a
margin at least as large as that by which 𝑦 beat 𝑥; thus, the reason
invoked to overthrow 𝑥 in favor of 𝑦 would immediately provide at
least as strong a reason to overthrow 𝑦 in favor of 𝑧, undercutting the
original idea to overthrow 𝑥 in favor of 𝑦.
Part 1 of the following lemma is equivalent to Lemma 3 of Pérez
1995 (cf. Lemma 1 of Pérez 2001),8 which is in turn an adaptation of a
similar result of Moulin 1988, but we include a proof to keep this note
self-contained.
Lemma 2.
1. Any voting method satisfying positive involvement and the Condorcet
winner criterion refines the defensible set for linear profiles.
2. Any voting method satisfying positive involvement and the weak
Condorcet winner criterion refines the defensible set for all profiles.
Proof. For part 1, let 𝐹 be such a method, 𝐏 a linear profile, and
𝑥 ∈ 𝐹 (𝐏). Toward a contradiction, suppose 𝑥 ∉ 𝐷(𝐏). Hence there
is a 𝑦 ∈ 𝑋(𝐏) with Margin𝐏 (𝑦, 𝑥) > 0 and for every 𝑧 ∈ 𝑋(𝐏),
5
The requirement that the new voter ranks 𝑥 uniquely first is crucial; a
stronger version of the axiom that applies whenever the new voter does not
strictly prefer any 𝑦 to 𝑥 is inconsistent with the Condorcet winner criterion
(Pérez 2001, Duddy 2014).
6
In particular, 𝐷 is the coarsest voting method satisfying all of Heitzig’s
axioms wIm𝑃𝛼 for 𝛼 ∈ (0, 1].
7
In Dung’s (1995, Definition 6.1) terms, the alternatives in 𝐷(𝐏) are
acceptable with respect to 𝑋(𝐏) given each attack relation 𝑎𝑡𝑡𝑎𝑐𝑘𝑠𝑛 , for each
positive integer 𝑛, defined by 𝑎𝑡𝑡𝑎𝑐𝑘𝑠𝑛 (𝑥, 𝑦) if Margin𝐏 (𝑥, 𝑦) ≥ 𝑛.
8
For linear profiles, where Support𝐏 (𝑥, 𝑦) + Support𝐏 (𝑦, 𝑥) = |𝑉 (𝐏)|, one
can show that Margin𝐏 (𝑧, 𝑦) ≥ Margin𝐏 (𝑦, 𝑥) is equivalent to Support(𝑥, 𝑦) ≥
Support(𝑦, 𝑧), which is the relevant condition in Lemma 3 of Pérez 1995.
9
This is also clear directly. Minimax selects the alternatives whose worst
head-to-head loss is smallest among any alternative’s worst head-to-head loss.
Thus, if 𝑥 is a Minimax winner that loses head-to-head to 𝑦, there must be an
alternative 𝑧 to which 𝑦 loses by at least as large a margin. As noted by Young
(1977), the Minimax winner is the ‘‘most stable against overthrow’’ (p. 350). If
𝑥 is a Split Cycle winner that loses head-to-head to 𝑦, then there is a sequence
of alternatives 𝑦1 , … , 𝑦𝑛 with 𝑦 = 𝑦1 and 𝑥 = 𝑦𝑛 such that each loses to the next
by a margin at least as large as that by which 𝑥 lost to 𝑦; hence the reason for
overthrowing 𝑥 in favor of 𝑦 would similarly justify a sequence of revolutions
leading right back to 𝑥.
2
Economics Letters 236 (2024) 111589
W.H. Holliday
Table 2
Satisfaction (✓) or violation (−) of the axioms by selected voting methods.
Table 1
Among the 1920 linearly edge-ordered tournaments for four alternatives up to isomorphism, the number with multiple winners for a given method and the average
(resp. maximum) size of the set of winners.
Smith set
Uncovered set
Copeland
Defensible set
Defensible set ∩ Smith set
Split Cycle
Minimax
# with > 1 winner
avg. size of set
max. size of set
960
960
960
598
583
104
0
2.375
2
1.625
1.359375
1.34375
1.05416
1
4
3
3
3
3
2
1
Positive involvement
Condorcet winner
Condorcet loser
Resolvability
Ordinal margin invariance
Beat Path
Ranked Pairs
Stable Voting
Black’s
Borda
Minimax
Split Cycle
–
✓
✓
✓
✓
−
✓
✓
✓
–
✓
–
✓
✓
−
✓
✓
–
✓
✓
✓
✓
✓
–
✓
satisfying the remaining axioms. This is shown by Beat Path/Ranked
Pairs/Stable Voting, Minimax,11 and Split Cycle in Table 2, which also
includes the Borda method (Borda 1781) and Black’s method12 (Black
1958) for comparison. However, no method satisfies all of the axioms.13
While the normative appeal of positive involvement and the Condorcet criteria, as well as the practical relevance of resolvability, is
clear, it is less obvious whether ordinal margin invariance should
be a desideratum. One point in favor of methods satisfying ordinal
margin invariance is that a small amount of noise in the collection of
voter preferences is unlikely to change the ordinal margin graph and
therefore unlikely to change the selection of winners, rendering such
rules quite robust to noise (cf. Procaccia et al. 2006).
For the following, a linearly edge-ordered tournament is a pair (𝑀, ≻)
where 𝑀 is a tournament, i.e., an asymmetric directed graph in which
any two distinct vertices are related by an edge in some direction, and
≻ is a linear order of the tournament’s edges.
Theorem 4. There is no voting method satisfying positive involvement, the
Condorcet winner criterion, the Condorcet loser criterion, resolvability, and
ordinal margin invariance.
Proof. Assume there is such an 𝐹 . To derive a contradiction, we use
the ordinal margin graphs in Fig. 1. The numbers indicate the ordering
≻ from the smallest margin (1) to the largest (6). The defensible set
for each graph is shaded in gray. Since 𝐹 satisfies ordinal margin
invariance and resolvability, 𝐹 returns a singleton for each graph by
Lemma 3. Given M and M′ and an alternative 𝑥, we write M ⇒𝑥 M′ in
Fig. 1 if there are profiles 𝐏, 𝐏′ such that M is the ordinal margin graph
of 𝐏, M′ is the ordinal margin graph of 𝐏′ , and 𝐏′ is obtained from 𝐏
by adding voters all of whom rank 𝑥 uniquely first. The construction of
such profiles is an integer linear programming problem, whose solution
yields the profiles in Figs. 2–5.14
The defensible set for M1 is {𝑎, 𝑑}. Then since 𝐹 satisfies positive
involvement and the Condorcet winner criterion, 𝐹 (M1 ) ⊆ {𝑎, 𝑑} by
Lemma 2, so 𝐹 (M1 ) = {𝑎} or 𝐹 (M1 ) = {𝑑} by resolvability.
Suppose 𝐹 (M1 ) = {𝑑}. Then by positive involvement and resolvability, 𝐹 (M2 ) = {𝑑}. On the other hand, 𝐹 (M3 ) ⊆ {𝑏, 𝑑} by Lemma 2, but
𝑑 ∉ 𝐹 (M3 ) by the Condorcet loser criterion, so 𝐹 (M3 ) = {𝑏}. Then by
positive involvement, 𝑏 ∈ 𝐹 (M2 ), contradicting 𝐹 (M2 ) = {𝑑}.
Thus, 𝐹 (M1 ) = {𝑎}. Then by positive involvement and resolvability,
𝐹 (M4 ) = {𝑎}. But 𝐹 (M5 ) = {𝑑} by Lemma 2, so by positive involvement,
𝑑 ∈ 𝐹 (M4 ), contradicting 𝐹 (M4 ) = {𝑎}. □
Lemma 3.
If 𝐹 satisfies ordinal margin invariance and single-voter
(resp. asymptotic) resolvability, then 𝐹 selects a unique winner in any profile
whose ordinal margin graph is a linearly edged-ordered tournament.
Proof. For single-voter resolvability, suppose 𝐹 selects multiple winners in a profile 𝐏 whose ordinal margin graph is a linearly edgeordered tournament. Consider 3𝐏, the profile obtained from 𝐏 by
replacing each voter with three copies of that voter. Since 3𝐏 has the
same ordinal margin graph as 𝐏, 𝐹 selects the same winners in 3𝐏 as
in 𝐏. But by adding a single voter to 3𝐏, it is impossible to obtain a 𝐏′
whose ordinal margin graph differs from that of 3𝐏, so it is impossible
to obtain a profile with a unique winner. Hence 𝐹 does not satisfy
single-voter resolvability.
For asymptotic resolvability, Harrison-Trainor (2022, Theorem 11.2)
shows that for every positive integer 𝑚 and linearly edge-ordered tournament 𝑇 for 𝑚 alternatives, in the limit as the number of voters goes
to infinity, the proportion of linear profiles for 𝑚 alternatives whose
ordinal margin graph is 𝑇 is nonzero. Thus, if 𝐹 picks multiple winners
in such a tournament, 𝐹 does not satisfy asymptotic resolvability. □
Thus, in the search for the ‘‘holy grail’’ of a voting method satisfying
positive involvement, the Condorcet winner and loser criteria, and resolvability, we must drop the restriction of ordinal margin invariance.15
Whether such a method exists or another impossibility theorem awaits
us is an important open question.
Thanks to Lemma 3, either version of resolvability works below, so
we simply speak of ‘resolvability’. The defensible set does not satisfy
either version, since it returns multiple winners for some linearly edgeordered tournaments. Table 1 shows the extent of its irresoluteness for
linearly edge-ordered tournaments for four alternatives in comparison
to several other voting methods.10
If we drop any of the axioms introduced so far besides the Condorcet
winner criterion and ordinal margin invariance, then there is a method
11
It is plausible that with more combinatorial work, our proof strategy for
Theorem 4 could yield a characterization of Minimax for four alternatives
using the axioms other than the Condorcet loser criterion (cf. Holliday and
Pacuit 2023c).
12
Black’s method selects the Condorcet winner only, if one exists, and
otherwise selects all Borda winners.
13
In Theorem 4, positive involvement can be replaced by negative involvement (see Pérez 2001) by the proof of Proposition 3.19 in Ding et al. 2023,
since ordinal margin invariance implies the neutral reversal axiom in that
proposition.
14
Although we minimize the number of voters in these profiles subject to
the relevant constraints, this does not answer an interesting question: what
is the minimal number of voters needed for the impossibility theorem itself
(cf. Brandt et al. 2017)?
15
Natural weakenings of ordinal margin invariance to consider are the
C1.5 (De Donder et al. 2000) and C2 (Fishburn 1977) invariance conditions.
10
The Smith set (Smith 1973) is the smallest nonempty set of alternatives
such that every alternative in the set beats ever alternative outside the set
head-to-head. The uncovered set (Fishburn 1977, Miller 1980) is the set of all
alternatives 𝑥 for which there is no 𝑦 such that 𝑦 beats 𝑥 and beats every 𝑧
that 𝑥 beats; this definition is equivalent to others (see Duggan 2013) provided
there are no zero margins between distinct alternatives, which is the case
for Table 1. In this case, the Copeland winners (Copeland 1951) are the
alternatives that beat the most other alternatives. A notebook with code to
generate Table 1 and Figs. 2–5 is available at github.com/wesholliday/pos-inv.
3
Economics Letters 236 (2024) 111589
W.H. Holliday
Fig. 1. Ordinal margin graphs for the proof of Theorem 4.
Fig. 2. Profiles realizing M1 and M2 and the transition between them in Fig. 1.
Fig. 4. Profiles realizing M1 and M4 and the transition between them in Fig. 1.
Fig. 3. Profiles realizing M2 and M3 and the transition between them in Fig. 1.
Fig. 5. Profiles realizing M4 and M5 and the transition between them in Fig. 1.
4
Economics Letters 236 (2024) 111589
W.H. Holliday
Data availability
Heitzig, Jobst, 2002. Social choice under incomplete, cyclic preferences:
Majority/minority-based rules, and composition-consistency. arXiv:math/0201285
[math.CO].
Holliday, Wesley H., Pacuit, Eric, 2021. Measuring violations of positive involvement
in voting. In: Halpern, J.Y., Perea, A. (Eds.), Theoretical Aspects of Rationality and
Knowledge 2021. TARK 2021, In: Electronic Proceedings in Theoretical Computer
Science, vol. 335, pp. 189–209. http://dx.doi.org/10.4204/EPTCS.335.17.
Holliday, Wesley H., Pacuit, Eric, 2023a. Split Cycle: A new Condorcet-consistent voting
method independent of clones and immune to spoilers. Public Choice 197, 1–62.
http://dx.doi.org/10.1007/s11127-023-01042-3.
Holliday, Wesley H., Pacuit, Eric, 2023b. Stable Voting. Const. Political Econ. 34,
421–433. http://dx.doi.org/10.1007/s10602-022-09383-9.
Holliday, Wesley H., Pacuit, Eric, 2023c. An extension of May’s theorem to three
alternatives: Axiomatizing minimax voting. arXiv:2312.14256 [econ.TH].
Kasper, Laura, Peters, Hans, Vermeulen, Dries, 2019. Condorcet consistency and the
strong no show paradoxes. Math. Social Sci. 99, 36–42. http://dx.doi.org/10.1016/
j.mathsocsci.2019.03.002.
Kramer, Gerald H., 1977. A dynamical model of political equilibrium. J. Econom.
Theory 16 (2), 310–334. http://dx.doi.org/10.1016/0022-0531(77)90011-4.
Miller, Nicholas R., 1980. A new solution set for tournaments and majority voting:
Further graph-theoretical approaches to the theory of voting. Am. J. Political Sci.
24 (1), 68–96. http://dx.doi.org/10.2307/2110925.
Moulin, Hervé, 1988. Condorcet’s principle implies the no show paradox. J. Econom.
Theory 45 (1), 53–64. http://dx.doi.org/10.1016/0022-0531(88)90253-0.
Pérez, Joaquín, 1995. Incidence of no-show paradoxes in Condorcet choice functions.
Investig. Econ. XIX (1), 139–154.
Pérez, Joaquín, 2001. The strong no show paradoxes are a common flaw in Condorcet
voting correspondences. Soc. Choice Welf. 18 (3), 601–616. http://dx.doi.org/10.
1007/s003550000079.
Pérez-Fernández, Raúl, De Baets, Bernard, 2018. The supercovering relation, the
pairwise winner, and more missing links between Borda and Condorcet. Soc. Choice
Welf. 50, 329–352. http://dx.doi.org/10.1007/s00355-017-1086-0.
Procaccia, Ariel D., Rosenschein, Jeffrey S., Kaminka, Gal A., 2006. On the robustness
of preference aggregation in noisy environments. In: Endriss, Ulle, Lang, Jérôme
(Eds.), Proceedings of the 1st International Workshop on Computational Social
Choice. COMSOC-2006, ILLC, University of Amsterdam.
Saari, Donald G., 1995. Basic Geometry of Voting. Springer, Berlin, http://dx.doi.org/
10.1007/978-3-642-57748-2.
Schulze, Markus, 2011. A new monotonic, clone-independent, reversal symmetric, and
Condorcet-consistent single-winner election method. Soc. Choice Welf. 36, 267–303.
http://dx.doi.org/10.1007/s00355-010-0475-4.
Simpson, Paul B., 1969. On defining areas of voter choice: Professor Tullock on stable
voting. Q. J. Econ. 83 (3), 478–490. http://dx.doi.org/10.2307/1880533.
Smith, John H., 1973. Aggregation of preferences with variable electorate.
Econometrica 41 (6), 1027–1041. http://dx.doi.org/10.2307/1914033.
Tideman, T.N., 1986. A majority-rule characterization with multiple extensions. Soc.
Choice Welf. 3, 17–30. http://dx.doi.org/10.1007/BF00433521.
Tideman, T.N., 1987. Independence of clones as a criterion for voting rules. Soc. Choice
Welf. 4, 185–206. http://dx.doi.org/10.1007/bf00433944.
Young, H.P., 1977. Extending Condorcet’s rule. J. Econom. Theory 16, 335–353.
http://dx.doi.org/10.1016/0022-0531(77)90012-6.
Associated code is available at https://github.com/wesholliday/
pos-inv.
Acknowledgments
I thank Yifeng Ding, Jobst Heitzig, Milan Mossé, Eric Pacuit, Zoi Terzopoulou, Nic Tideman, Snow Zhang, Bill Zwicker, and an anonymous
referee for helpful comments.
References
Black, Duncan, 1958. The Theory of Committees and Elections. Cambridge University
Press, Cambridge.
Borda, Jean-Charles de Chevalier, 1781. Mémoire sur les Élections au Scrutin. Histoire
de l’Académie Royale des Sciences, Paris.
Brandt, Felix, Geist, Christian, Peters, Dominik, 2017. Optimal bounds for the no-show
paradox via SAT solving. Math. Social Sci. 90, 18–27. http://dx.doi.org/10.1016/
j.mathsocsci.2016.09.003.
Condorcet, M.J.A.N. de C., Marque de, 1785. Essai sur l’application de l’analyse à la
probabilitié des décisions rendues à la pluralité des voix. l’Imprimerie Royale, Paris.
Copeland, A.H., 1951. A ‘Reasonable’ Social Welfare Function. In: Notes from a Seminar
on Applications of Mathematics To the Social Sciences, University of Michigan.
De Donder, Philippe, Le Breton, Michel, Truchon, Michel, 2000. Choosing from a
weighted tournament. Math. Social Sci. 40, 85–109. http://dx.doi.org/10.1016/
S0165-4896(99)00042-6.
Debord, Bernard, 1987. Caractérisation des matrices des préférences nettes et méthodes
d’agrégation associées. Math. Sci. Hum. 97, 5–17.
Ding, Yifeng, Holliday, Wesley H., Pacuit, Eric, 2023. An axiomatic characterization of
Split Cycle. arXiv:2210.12503 [econ.TH].
Duddy, Conal, 2014. Condorcet’s principle and the strong no-show paradoxes. Theory
and Decision 77, 275–285. http://dx.doi.org/10.1007/s11238-013-9401-4.
Duggan, John, 2013. Uncovered sets. Soc. Choice Welf. 41 (3), 489–535. http://dx.doi.
org/10.1007/s00355-012-0696-9.
Dung, Phan Minh, 1995. On the acceptability of arguments and its fundamental role
in nonmonotonic reasoning, logic programming and 𝑛-person games. Artificial
Intelligence 77 (2), 321–357. http://dx.doi.org/10.1016/0004-3702(94)00041-X.
Dutta, Bhaskar, Laslier, Jean-Francois, 1999. Comparison functions and choice
correspondences. Soc. Choice Welf. 16, 513–532. http://dx.doi.org/10.1007/
s003550050158.
Fischer, Felix, Hudry, Olivier, Niedermeier, Rolf, 2016. Weighted tournament solutions.
In: Brandt, Felix, Conitzer, Vincent, Endriss, Ulle, Lang, Jérôme, Procaccia, Ariel D.
(Eds.), Handbook of Computational Social Choice. Cambridge University Press, New
York, pp. 86–102. http://dx.doi.org/10.1017/CBO9781107446984.005.
Fishburn, Peter C., 1977. Condorcet social choice functions. SIAM J. Appl. Math. 33
(3), 469–489. http://dx.doi.org/10.1137/0133030.
Harrison-Trainor, Matthew, 2022. An analysis of random elections with large numbers
of voters. Math. Social Sci. 116, 68–84. http://dx.doi.org/10.1016/j.mathsocsci.
2022.01.002.
5