scieee Science in your language
[en] (orig)
Games and Economic Behavior 144 (2024) 203–224
Available online 1 February 2024
0899-8256/Β© 2024 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
Contents lists available at ScienceDirect
Games and Economic Behavior
journal homepage: www.elsevier.com/locate/geb
Impartial selection with additive guarantees via iterated deletion
Javier Cembrano a, Felix Fischer b, David Hannonb, Max Klimma,βˆ—
aInstitut fΓΌr Mathematik, Technische UniversitΓ€t Berlin, Germany
bSchool of Mathematical Sciences, Queen Mary University of London, UK
A R T I C L E I N F O A B S T R A C T
Keywords:
Voting
Impartial selection
Additive approximation
Impartial selection is the selection of an individual from a group based on nominations by other
members of the group, in such a way that individuals cannot influence their own chance of
selection. For this problem, we give a deterministic mechanism with an additive performance
guarantee of 𝑂(𝑛(1+πœ…)βˆ•2)in a setting with 𝑛individuals where each individual casts 𝑂(π‘›πœ…)
nominations, where πœ…βˆˆ[0, 1]. This bound is 𝑂(βˆšπ‘›)for πœ…=0and 𝑂(𝑛)for πœ…=1. The latter is
trivial, as even a mechanism that never selects provides an additive guarantee of 𝑛 βˆ’1. We show,
however, that it is also best possible: for every deterministic impartial mechanism there exists a
situation in which some individual is nominated by every other individual and the mechanism
either does not select or selects an individual not nominated by anyone.
1. Introduction
Many important decisions are made by groups of people who select one member of the group based on nominations from within
the group. In fact, such a procedure is applied in a wide range of settings such as the papal conclaves (Mackenzie, 2020), the national
captain’s votes for the best FIFA Women’s and Men’s Players Awards, or the voting for spokespersons of committees or student
representatives. A common feature of these situations is that there is at least a partial overlap between the set of voters and the
set of candidates. This overlap may be a source of incentive issues as candidates who have a reasonable chance of winning may be
motivated not to reveal their true opinion on who should win in order to increase their own chance of winning. Incentive issues of
this kind were first studied in a systematic way by Holzman and Moulin (2013)and Alon et al. (2011), who formalized the problem
in terms of a directed graph in which vertices correspond to voters and a directed edge from one voter to another indicates that the
former nominates the latter. A (deterministic) selection mechanism then takes such a nomination graph as input and returns either
one of its vertices or no vertex. In order to allow voters to express their true opinions about other voters without having to worry
about their own chance of selection, an important property of a selection mechanism is its impartiality: a mechanism is impartial if,
for all nomination graphs, a change of the outgoing edges of some vertex 𝑣does not change whether 𝑣is selected or not. It is easy
to see that a mechanism that selects a vertex with maximum indegree and breaks ties in some consistent way is not impartial: if ties
are broken in favor of greater index, for example, a vertex with maximum indegree that currently nominates another vertex with
maximum indegree but greater index has an incentive to instead nominate a different vertex.
Holzman and Moulin have shown that impartial mechanisms are in fact much more limited even in a setting where each voter
casts exactly one vote, i.e., where each vertex has outdegree one: for every impartial mechanism that selects a vertex in every
* Corresponding author at: Max Klimm, TU Berlin, Institut für Mathematik, MA 5-2, Straße des 17. Juni 136, 10623 Berlin, Germany.
E-mail addresses: [emailΒ protected] (J. Cembrano),
felix.fi[email protected] (F. Fischer),
[emailΒ protected] (D. Hannon),
https://doi.org/10.1016/j.geb.2024.01.008
Received 27 December 2022
Games and Economic Behavior 144 (2024) 203–224
204
J. Cembrano, F. Fischer, D. Hannon et al.
nomination graph, there exists a graph where the mechanism selects a vertex with indegree zero or a graph where it fails to select
a vertex with indegree 𝑛 βˆ’1, where 𝑛is the number of voters. This shows in particular that the best multiplicative approximation
guarantee for impartial mechanisms, i.e., the worst case over all nomination graphs of the ratio between the maximum indegree
and the indegree of the selected vertex is at least 𝑛 βˆ’1. On the other hand, a multiplicative guarantee of 𝑛 βˆ’1 is easy to obtain
by always following the outgoing edge of a fixed vertex. As multiplicative guarantees do not allow for a meaningful distinction
among deterministic impartial mechanisms, Caragiannis et al. (2022, 2023) proposed to instead consider an additive guarantee,
i.e., the worst case over all nomination graphs of the difference between the maximum indegree and the indegree of the selected
vertex. An adaptation provided by Mackenzie (2020)of a mechanism due to Holzman and Moulin achieves an additive guarantee
of βŒˆπ‘›βˆ•2βŒ‰.1Caragiannis et al. (2022)further proposed a randomized mechanism with an additive guarantee of 𝑂(βˆšπ‘›). It remained
open, however, whether there exists a deterministic mechanism with a sublinear additive guarantee.
The setting studied by Holzman and Moulin, where each vertex has outdegree one, is commonly referred to as the plurality setting.
The impossibility results of Holzman and Moulin regarding multiplicative guarantees carry over to the more general approval setting,
where outdegrees can be arbitrary, but here even less is known about possible additive guarantees. While Caragiannis et al. (2022)
have shown that deterministic impartial mechanisms cannot provide a better additive guarantee than 3, no mechanism is known that
improves on the trivial guarantee of 𝑛 βˆ’1achieved by selecting a fixed vertex. Caragiannis et al. also gave a randomized mechanism
for the approval setting with an additive guarantee of Θ(𝑛2βˆ•3 ln1βˆ•3 𝑛).
1.1. Our contribution
We develop a new deterministic mechanism for impartial selection that is parameterized by a pair of thresholds on the indegrees
of vertices in the graph. The mechanism seeks to select a vertex with large indegree, and to achieve impartiality it iteratively deletes
outgoing edges from vertices in decreasing order of their indegrees, until only the outgoing edges of vertices with indegrees below
the lower threshold remain. It then selects a vertex with maximum remaining indegree if that indegree is above the higher threshold,
and otherwise does not select. Any ties are broken according to a fixed ordering of the vertices. We give a sufficient condition for
choices of thresholds that guarantee impartiality. The iterative nature of the deletions requires a fairly intricate analysis but is key to
achieving impartiality. The additive guarantee is then obtained for a good choice of thresholds, and the worst case is the one where
the mechanism does not select.
For instances with 𝑛vertices and maximum outdegree at most 𝑂(π‘›πœ…), where πœ…βˆˆ[0, 1], the mechanism provides an additive
guarantee of 𝑂(𝑛
1+πœ…
2). This is the first sublinear bound for a deterministic mechanism and any πœ…βˆˆ[0, 1], and is sublinear for all
πœ…βˆˆ[0, 1). For settings with constant maximum outdegree, which includes the setting of Holzman and Moulin where all outdegrees
are equal to one, our bound matches the best known bound of 𝑂(βˆšπ‘›)for randomized mechanisms and outdegree one, due to
Caragiannis et al. (2022).
When the maximum outdegree is unbounded, the bound becomes 𝑂(𝑛). This is of course trivial, as even a mechanism that never
selects a vertex provides an additive guarantee of 𝑛 βˆ’1. For a setting without abstentions, i.e., with minimum outdegree one, the
guarantee can be improved slightly to 𝑛 βˆ’2by following the outgoing edge of a fixed vertex. We show that both of these bounds
are best possible by giving matching lower bounds. This improves on the only lower bound known prior to our work, again due
to Caragiannis et al., which is equal to 3and applies to the setting with abstentions and mechanisms that select a vertex for every
graph.
Just like the lower bounds regarding multiplicative guarantees for plurality, our lower bounds for approval are obtained through
an axiomatic impossibility result. Holzman and Moulin have shown that in the case of plurality, impartiality is incompatible with
positive and negative unanimity. Here, positive unanimity requires that a vertex with the maximum possible indegree of 𝑛 βˆ’1must
be selected, and negative unanimity that a vertex with indegree zero cannot be selected.2We show that in the case of approval
this impossibility can be strengthened even further: call a selection mechanism weakly unanimous if it selects a vertex with positive
indegree whenever there exists a vertex with the maximum possible indegree of 𝑛 βˆ’1; then weak unanimity and impartiality are
incompatible, even if we are not always required to select.
This result is obtained by analyzing the behavior of impartial mechanisms on a restricted class of graphs with a high degree of
symmetry among vertices. Like Holzman and Moulin, we can assume that isomorphic vertices are selected with equal probabilities
by a randomized relaxation of a mechanism. A suitable class of graphs for our purposes are those generated by partial orders on the
set of ordered partitions. Our result is then obtained by combining counting results for ordered partitions and an argument similar
to Farkas’ Lemma.
1.2. Related work
Impartiality as a formal property of social and economic mechanisms was first considered by De Clippel et al. (2008)for the
distribution of a divisible commodity among a set of individuals according to the individuals’ subjective claims. Holzman and Moulin
(2013)and Alon et al. (2011)studied impartial selection in two different settings, plurality and approval, and established strong
1The mechanism does not always select a vertex, but we will see that for our purposes there is little difference between mechanisms that do and do not always
select.
2This result is formulated for mechanisms that always select a vertex, but not selecting a vertex clearly does not improve the situation.
Games and Economic Behavior 144 (2024) 203–224
205
J. Cembrano, F. Fischer, D. Hannon et al.
impossibility results regarding the ability of deterministic mechanisms to approximate the maximum indegree in a multiplicative
sense. Alon et al. (2011)also proposed randomized mechanisms for the selection of one or more vertices. Fischer and Klimm (2015)
then obtained a randomized mechanism with the best possible multiplicative guarantee of 2for the selection of a single vertex in
the approval setting, and Bjelde et al. (2017)gave improved deterministic and randomized mechanisms for the selection of more
than one vertex. Niemeyer and Preusser (2023)studied strategyproof mechanisms for the selection of a single individual under more
general preferences.
Starting from the observation that impossibility results for randomized mechanisms in particular are obtained from graphs with
very small indegrees, Bousquet et al. (2014) developed a randomized mechanism that is optimal in the large-indegree limit, i.e.,
that chooses a vertex with indegree arbitrarily close to the maximum indegree as the latter goes to infinity. Caragiannis et al.
(2022, 2023)used the same observation as motivation to study mechanisms with additive rather than multiplicative guarantees.
They developed new mechanisms that achieve such guarantees, established a relatively small but nontrivial lower bound of 3for
deterministic mechanisms in the approval setting, and gave improved deterministic mechanisms for a setting with prior information.
The axiomatic study of Holzman and Moulin has been refined and extended in a number of ways, for example with a focus on
symmetric mechanisms (Mackenzie, 2015) and to the selection of more than one vertex (Tamura and Ohseto, 2014). Mackenzie
(2020)provided a detailed axiomatic analysis of mechanisms used in the papal conclave and showed that impartial mechanisms
satisfying natural symmetry and anonymity axioms are exactly those that select a vertex whose indegree exceeds a supermajority
threshold. Various selection mechanisms have also been proposed that are tailored to applications like peer review and exploit the
particular preference and information structures of those applications (Kurokawa et al., 2015; Xu et al., 2019; Aziz et al., 2019;
Mattei et al., 2021). Impartial mechanisms have finally been considered for other objectives, specifically for the maximization of
progeny (Babichenko et al., 2020; Zhang et al., 2021) and for rank aggregation (Kahng et al., 2018). For an overview of mechanisms
for peer selection and evaluation, the reader is referred to the survey of Olckers and Walsh (2022).
The proof of our impossibility result uses a class of graphs constructed from ordered partitions of the set of vertices. The class has
been studied previously (e.g., Insko et al., 2017; Diagana and MaΓ―ga, 2017), and some of its known properties including its lattice
structure and the number of graphs isomorphic to each graph within the class are relevant to us.
2. Preliminaries
For 𝑛 βˆˆβ„•, let ξˆ³π‘›={(𝑁,𝐸)βˆΆπ‘={1,2,…,𝑛},𝐸 βŠ†(𝑁×𝑁)β§΅β‹ƒπ‘£βˆˆπ‘{(𝑣, 𝑣)}}be the set of directed graphs with 𝑛vertices and no
loops. Let  =β‹ƒπ‘›βˆˆβ„•ξˆ³π‘›. For 𝐺=(𝑁, 𝐸) ∈and 𝑣 βˆˆπ‘, let 𝑁+(𝑣, 𝐺) ={𝑒 βˆˆπ‘βˆΆ(𝑣, 𝑒) ∈𝐸}be the out-neighborhood and π‘βˆ’(𝑣, 𝐺) =
{𝑒 βˆˆπ‘βˆΆ(𝑒, 𝑣) ∈𝐸}the in-neighborhood of 𝑣in 𝐺. Let 𝛿+(𝑣, 𝐺) =|𝑁+(𝑣, 𝐺)|and π›Ώβˆ’(𝑣, 𝐺) =|π‘βˆ’(𝑣, 𝐺)|denote the outdegree and
indegree of 𝑣in 𝐺,
π›Ώβˆ’
𝑆(𝑣, 𝐺)=|{π‘’βˆˆπ‘†βˆΆ(𝑒, 𝑣)∈𝐸}|
the indegree of 𝑣from a particular subset π‘†βŠ†π‘of the vertices, and Ξ”(𝐺) =max
π‘£βˆˆπ‘π›Ώβˆ’(𝑣, 𝐺)the maximum indegree of any vertex
in 𝐺. When the graph is clear from the context, we will sometimes drop 𝐺from the notation and write 𝑁+(𝑣), π‘βˆ’(𝑣), 𝛿+(𝑣), π›Ώβˆ’(𝑣),
π›Ώβˆ’
𝑆(𝑣), and Ξ”. For 𝑛, π‘˜ βˆˆβ„•, let +
𝑛={(𝑁, 𝐸) βˆˆξˆ³π‘›βˆΆπ›Ώ+(𝑣) β‰₯1for every 𝑣 βˆˆπ‘}be the set of graphs in ξˆ³π‘›where all outdegrees are
strictly positive,
ξˆ³π‘›(π‘˜)={(𝑁,𝐸)βˆˆξˆ³π‘›βˆΆπ›Ώ+(𝑣)β‰€π‘˜for every π‘£βˆˆπ‘}
the set of graphs in ξˆ³π‘›where outdegrees are at most π‘˜, and +
𝑛(π‘˜) =+
π‘›βˆ©ξˆ³π‘›(π‘˜)for the set of graphs satisfying both conditions. Let
+=β‹ƒπ‘›βˆˆβ„•ξˆ³+
𝑛, (π‘˜) =β‹ƒπ‘›βˆˆβ„•ξˆ³π‘›(π‘˜), and +(π‘˜) =β‹ƒπ‘›βˆˆβ„•ξˆ³+
𝑛(π‘˜).
A (deterministic) selection mechanism is then given by a family of functions π‘“βˆΆξˆ³π‘›β†’2𝑁that maps each graph to a subset of its
vertices, where we require throughout that |𝑓(𝐺)| ≀1for all 𝐺∈. By slightly abusing notation, we will use 𝑓to refer to both the
mechanism and to individual functions from the family. Mechanism 𝑓is impartial on ξˆ³β€²βŠ†ξˆ³if on this set of graphs the outgoing
edges of a vertex have no influence on its selection, i.e., if for every pair of graphs 𝐺=(𝑁, 𝐸)and 𝐺′=(𝑁, 𝐸′)in ξˆ³β€²and every
𝑣 βˆˆπ‘, 𝑓(𝐺) ∩{𝑣} =𝑓(𝐺′) ∩{𝑣}whenever 𝐸⧡({𝑣} ×𝑁) =𝐸′⧡({𝑣} ×𝑁). Mechanism 𝑓is 𝛼-additive on ξˆ³β€²βŠ†ξˆ³, for 𝛼β‰₯0, if for
every graph in ξˆ³β€²the indegree of the choice of 𝑓differs from the maximum indegree by at most 𝛼, i.e., if
sup
πΊβˆˆξˆ³β€²{Ξ”(𝐺)βˆ’π›Ώβˆ’(𝑓(𝐺),𝐺)}≀𝛼,
where for π‘†βŠ†π‘we let π›Ώβˆ’(𝑆, 𝐺) =βˆ‘π‘£βˆˆπ‘†π›Ώβˆ’(𝑣, 𝐺)denote the sum of the indegrees of the vertices in 𝑆.
3. Iterated deletion of nominations
When outdegrees are at most one, the following simple mechanism is βŒŠπ‘›βˆ•2βŒ‹-additive: if there is a vertex with indegree at least
βŒŠπ‘›βˆ•2βŒ‹ +1, select it; otherwise, do not select. This mechanism is a slight modification of a mechanism Holzman and Moulin (2013)
called majority with default and corresponds to the supermajority rule of Mackenzie (2020)with threshold βŒŠπ‘›βˆ•2βŒ‹ +1. As there can be
at most one vertex with degree βŒŠπ‘›βˆ•2βŒ‹ +1or more and a vertex cannot influence its own indegree, the mechanism is clearly impartial.
We will borrow from this mechanism the idea of imposing a threshold on the minimum indegree a vertex needs to be selected, but
will seek to lower the threshold in order to achieve a better additive guarantee and also to relax the constraint on the maximum
Games and Economic Behavior 144 (2024) 203–224
206
J. Cembrano, F. Fischer, D. Hannon et al.
Algorithm 1: Twin threshold mechanism with thresholds 𝑇and 𝑑.
Input: Digraph 𝐺=(𝑁, 𝐸) βˆˆξˆ³π‘›.
Output: Set π‘†βŠ†π‘of selected vertices with |𝑆| ≀1.
Initialize 𝑖 ←0and 𝑑←Δ;
π·π‘–β†βˆ…; // vertices with deleted outgoing edges in iteration 𝑖or before
ξ˜•
𝛿𝑖(𝑣) β†π›Ώβˆ’(𝑣)for all 𝑣 βˆˆπ‘; // indegree of 𝑣omitting edges deleted up to 𝑖
while 𝑑β‰₯𝑑do
if {𝑒 βˆˆπ‘β§΅π·π‘–βˆΆξ˜•
𝛿𝑖(𝑒) =𝑑} =βˆ…then
update π‘‘β†π‘‘βˆ’1and continue
end
𝑣 ←max{𝑒 βˆˆπ‘β§΅π·π‘–βˆΆξ˜•
𝛿𝑖(𝑒) =𝑑};
update
ξ˜•
𝛿𝑖+1(𝑒) β†ξ˜•
𝛿𝑖(𝑒) βˆ’1for all 𝑒 βˆˆπ‘+(𝑣); // delete outgoing edges of 𝑣
update
ξ˜•
𝛿𝑖+1(𝑒) β†ξ˜•
𝛿𝑖(𝑒)for all 𝑒 βˆˆπ‘β§΅π‘+(𝑣);
update 𝐷𝑖+1 ←𝐷𝑖βˆͺ{𝑣}and 𝑖 ←𝑖 +1;
end
𝐼←𝑖;
ξ˜•
Δ←maxπ‘£βˆˆπ‘ξ˜•
𝛿𝐼(𝑣);
if ξ˜•
Ξ”β‰₯𝑇then
return 𝑆←{max{𝑣 βˆˆπ‘βˆΆξ˜•
𝛿𝐼(𝑣) =ξ˜•
Ξ”}}
end
return π‘†β†βˆ…
outdegree. Of course, lower thresholds and larger outdegrees both mean that more and more vertices become eligible for selection,
and we will no longer get impartiality for free.
As a first step, it is instructive to again consider the outdegree-one case but to use a lower threshold of 𝑑 =βŒŠπ‘›βˆ•3βŒ‹ +1. This choice
of threshold implies that there can be at most two vertices with indegrees equal to or higher than the threshold. Analogously to the
supermajority rule, the general idea is that only vertices with indegree at least 𝑑are eligible for selection. In cases where there is only
one such vertex impartiality would follow by a similar argument as for the supermajority rule, so the critical case is when there are
two vertices 𝑒and 𝑣with indegree at least 𝑑. If 𝑒is selected, π›Ώβˆ’(𝑒) =𝑑, and (𝑣, 𝑒) ∈𝐸, then 𝑣could remove the edge (𝑣, 𝑒), cause 𝑒to
drop below the threshold, and leave 𝑣as the only vertex eligible for selection. To prevent impartiality from being compromised in
this case, it makes sense to introduce another threshold 𝑇=𝑑 +1 =βŒŠπ‘›βˆ•3βŒ‹ +2in the understanding that only vertices with indegree
at least 𝑇are eligible for selection, but the outgoing edges of all vertices with indegree at least 𝑑 =βŒŠπ‘›βˆ•3βŒ‹ +1will automatically be
ignored. It turns out that in order to actually achieve impartiality, the outgoing edges of the vertices with indegree at least 𝑑must
be removed in decreasing order of indegrees, breaking ties according to some fixed ordering of the vertices. In the end we select
the vertex with maximum indegree among those with indegree at least 𝑇, breaking ties as before, and call the resulting mechanism
two contenders (TC). It is not too difficult to convince ourselves that the mechanism is indeed impartial.3Indeed, if π›Ώβˆ’(𝑒) β‰₯𝑇and
π›Ώβˆ’(𝑣) β‰₯𝑇, neither 𝑒nor 𝑣can influence whether the respective other vertex ends up with an indegree of at least 𝑑 =π‘‡βˆ’1; the
outgoing edges of both 𝑒and 𝑣will be removed, and neither 𝑒nor 𝑣can influence which of the two vertices is selected. If π›Ώβˆ’(𝑒) β‰₯𝑇
and π›Ώβˆ’(𝑣) =𝑑, only 𝑒is eligible for selection; the outgoing edge of 𝑒, which has the highest indegree, is removed first, so 𝑒cannot
influence whether 𝑣will have indegree at least 𝑑and thus whether a potential edge (𝑣, 𝑒)is removed or not. All other cases are either
analogous or lead to no vertex being selected, and impartiality follows.
It is natural to ask whether the threshold can be lowered further while maintaining impartiality, and whether guarantees can be
obtained in a similar fashion for graphs with larger outdegrees. The answer to the first question is not obvious, as the number of
graphs that need to be considered to establish impartiality grows very quickly in the number of vertices eligible for selection. The
obvious generalization of the mechanism with threshold βŒŠπ‘›βˆ•2βŒ‹ +1to a setting with outdegrees at most π‘˜, of selecting the unique
vertex with indegree at least βŒŠπ‘˜π‘›βˆ•2βŒ‹ +1if it exists and not selecting otherwise, is impartial and βŒŠπ‘˜π‘›βˆ•2βŒ‹-additive, but this guarantee
is trivial when π‘˜ β‰₯2.
Our main result answers both questions in the affirmative. It applies to settings with outdegree at most π‘˜ =𝑂(π‘›πœ…)for πœ…βˆˆ[0, 1]
and provides a nontrivial guarantee when πœ…βˆˆ[0, 1). When π‘˜is constant the guarantee is 𝑂(βˆšπ‘›), which matches the best guarantee
known for randomized mechanisms and outdegree one (Caragiannis et al., 2022).
Theorem 1. For every 𝑛 βˆˆβ„•, πœ…βˆˆ[0, 1], and π‘˜ =𝑂(π‘›πœ…), there exists an impartial and 𝑂(𝑛
1+πœ…
2)-additive mechanism on ξˆ³π‘›(π‘˜). Specifically,
for every 𝑛 βˆˆβ„•, there exists an impartial and
√7.25𝑛-additive mechanism on ξˆ³π‘›(1).
The result is achieved by a mechanism of a family we call twin threshold mechanisms and describe formally in Algorithm 1. Each
of these mechanisms iteratively deletes the outgoing edges from vertices with indegree above a first threshold 𝑑from the highest to
the lowest indegree and, in the end, selects the vertex with maximum remaining indegree as long as that indegree is above a second,
higher threshold 𝑇. The parameters 𝑑and 𝑇characterize a specific mechanism of this family and will be chosen in order to achieve
impartiality and obtain the desired bounds. Throughout the mechanism ties are broken as before, in favor of greater index.
3This will follow as a special case from a more general result in Lemma 3, but we find it insightful to give the informal argument here.
Games and Economic Behavior 144 (2024) 203–224
207
J. Cembrano, F. Fischer, D. Hannon et al.
𝑛23451020501005001000
SR 112251025 50250500
TC 23335818 35168335
TTβˆ—11225816 23 56 81
(𝑑, 𝑇 )(2,2) (2,2) (3,3) (3,3) (4,5) (7,8) (7,11) (13,18) (28,41) (36,56)
LB 112233 3 3 3 3
Fig. 1. Values of 𝛼𝑛, for certain values of 𝑛, such that the supermajority rule with threshold βŒŠπ‘›βˆ•2βŒ‹ +1(SR), the two contenders mechanism (TC), and an optimal twin
threshold mechanism (TTβˆ—) are 𝛼𝑛-additive on ξˆ³π‘›(1), together with a lower bound (LB) on 𝛼𝑛for any impartial 𝛼𝑛-additive mechanism. Each bound for TTβˆ—in the
table is accompanied by values of the thresholds 𝑑and 𝑇that achieve the bound, which may not be the unique such values.
For any choice of the maximum outdegree π‘˜and the threshold parameters 𝑇and 𝑑, the twin threshold mechanism with thresholds
𝑇and 𝑑achieves its worst additive performance guarantee in cases where it does not select, and this guarantee can be obtained in
a straightforward way by bounding the maximum indegree. The proof of impartiality, on the other hand, uses a relatively subtle
argument to show that for certain values of 𝑇and 𝑑, a vertex above the higher threshold 𝑇cannot influence whether another vertex
ends up above or below the lower threshold 𝑑when edges have been deleted. Vertices above 𝑇then have no influence on the set of
edges taken into account for selection, and since these are the only vertices that can potentially be selected impartiality follows.
For the outdegree-one case and specific values of 𝑛, we can optimize over 𝑑and 𝑇in order to obtain the best guarantee; bounds for
an optimal twin threshold mechanism and selected values of 𝑛are shown in Fig. 1, alongside bounds for the supermajority rule with
threshold βŒŠπ‘›βˆ•2βŒ‹ +1, corresponding to a twin threshold mechanism with 𝑇=𝑑 =βŒŠπ‘›βˆ•2βŒ‹ +1, the two contenders mechanism, corre-
sponding to a twin threshold mechanism with 𝑇=𝑑 +1 =βŒŠπ‘›βˆ•3βŒ‹ +2, and lower bounds for any impartial mechanism. It is interesting
to note at this point that the supermajority rule with threshold βŒŠπ‘›βˆ•2βŒ‹ +1possesses natural symmetry properties (Mackenzie, 2020),
and that the optimal twin threshold mechanism does not possess these properties due to its use of tie-breaking and its distinction
among edges depending on their origin. The improved additive guarantee provided by the optimal twin threshold mechanism and
shown in the figure thus comes at the expense of a certain amount of fairness among vertices.
A twin threshold mechanism maintains a set 𝐷𝑖of vertices whose outgoing edges have been deleted up to iteration 𝑖, and indegrees
ξ˜•
𝛿𝑖(𝑣)where all edges deleted up to iteration 𝑖are not counted. After the iterations during which outgoing edges are deleted, we add
an artificial final iteration during which no edges are deleted; this iteration counter is denoted by 𝐼. In order to prove Theorem 1,
we will compare runs of the mechanism with fixed thresholds for different graphs, and denote by
ξ˜•
𝛿𝑖(𝑣, 𝐺), 𝐷𝑖(𝐺), and 𝐼(𝐺)the
respective values of
ξ˜•
𝛿𝑖(𝑣), 𝐷𝑖, and 𝐼when the mechanism is invoked with input graph 𝐺. We use πœ’to denote the indicator function
for logical propositions, i.e., πœ’(𝑝) =1when proposition 𝑝is true and πœ’(𝑝) =0otherwise. For a graph 𝐺and a vertex 𝑣in 𝐺whose
outgoing edges are deleted by the mechanism, we use π‘–βˆ—(𝑣, 𝐺)to denote the iteration in which this deletion takes place, such that
π·π‘–βˆ—(𝑣,𝐺)+1(𝐺) β§΅π·π‘–βˆ—(𝑣,𝐺)(𝐺) ={𝑣}. We use the convention that π‘–βˆ—(𝑣, 𝐺) =𝐼(𝐺)if 𝑣 βˆ‰π·πΌ(𝐺)(𝐺)to extend the function to all vertices. For
a graph 𝐺and vertex 𝑣, we write π›Ώβˆ—(𝑣, 𝐺)= ξ˜•
π›Ώπ‘–βˆ—(𝑣,𝐺)(𝑣, 𝐺)for the indegree of 𝑣, not taking into account any incoming edges deleted
previously, at the last moment before the outgoing edges of 𝑣are deleted. When the graph 𝐺is clear from the context, we again drop
𝐺from the notation. It is clear from the definition of the mechanism that for any graph 𝐺=(𝑁, 𝐸)and vertex 𝑣 ∈𝐷𝐼(𝐺)(𝐺),
π›Ώβˆ’(𝑣, 𝐺)βˆ’π›Ώβˆ—(𝑣, 𝐺)=|{π‘’βˆˆπ‘βˆ’(𝑣, 𝐺)βˆΆπ‘–βˆ—(𝑒, 𝐺)<𝑖
βˆ—(𝑣, 𝐺)}|.(1)
When comparing tuples of the form (π›Ώβˆ—(𝑣), 𝑣), we use βͺ―and β‰Ίto denote lexicographic order and strict lexicographic order. These
comparisons are relevant to our analysis because they determine the order in which edges are deleted. Specifically, for any graph
𝐺=(𝑁, 𝐸)and vertices 𝑒, 𝑣 ∈𝐷𝐼(𝐺)(𝐺),
π‘–βˆ—(𝑒, 𝐺)<𝑖
βˆ—(𝑣, 𝐺)if and only if (π›Ώβˆ—(𝑒, 𝐺),𝑒)≻(π›Ώβˆ—(𝑣, 𝐺),𝑣).(2)
Fig. 2shows an example of the edge deletion process over four iterations of Algorithm 1. Observe that for π‘—βˆˆ{1, 2}, (π›Ώβˆ’(𝑣), 𝑣) ≻
(π›Ώβˆ’(𝑒𝑗), 𝑒𝑗)but (π›Ώβˆ—(𝑣), 𝑣) β‰Ί(π›Ώβˆ—(𝑒𝑗), 𝑒𝑗). This is caused by a drop in the indegree of 𝑣over the course of the algorithm, and this drop
occurs before the algorithm considers possible outgoing edges of 𝑣for deletion. For our analysis, it will be important to bound how
much the indegree of a vertex 𝑣can drop before 𝑣loses its outgoing edges. The following lemma, which we prove in Appendix A,
characterizes this quantity in terms of the indegrees of the in-neighbors of 𝑣.
Games and Economic Behavior 144 (2024) 203–224
208
J. Cembrano, F. Fischer, D. Hannon et al.
Fig. 2. Illustration of the edge deletion process in Algorithm 1. Here and in other figures, we arrange the vertices above 𝑑vertically according to their indegree, with
larger indegrees above, and horizontally according to their index, with greater indices to the left. Vertices below 𝑑, as well as edges with one or both endpoints below
𝑑, are omitted for simplicity.
Fig. 3. Illustration of Lemma 1for π‘Ÿ =3. If the indegree of 𝑣drops as shown by the dashed arrow, there must be a vertex with an edge to 𝑣in 𝐴, another vertex with
an edge to 𝑣in 𝐴 βˆͺ𝐡, and a third vertex with an edge to 𝑣in 𝐴 βˆͺ𝐡βˆͺ𝐢. Note that this exact condition is satisfied for the example of Fig. 2.
Lemma 1. Let 𝐺=(𝑁, 𝐸) βˆˆξˆ³π‘›, 𝑣 βˆˆπ‘and (𝑑, 𝑧) βˆˆβ„•2such that
(π›Ώβˆ’(𝑣),𝑣)≻(𝑑,𝑧)βͺ°(π›Ώβˆ—(𝑣),𝑣).
Let π‘Ÿ =π›Ώβˆ’(𝑣) βˆ’π‘‘+πœ’(𝑣 >𝑧). Then there exist vertices 𝑒0, … , π‘’π‘Ÿβˆ’1 such that for each π‘—βˆˆ{0, 1, … , π‘Ÿ βˆ’1}, (𝑒𝑗, 𝑣) ∈𝐸and (π›Ώβˆ—(𝑒𝑗), 𝑒𝑗) ≻
(π›Ώβˆ’(𝑣) βˆ’π‘—, 𝑣). Moreover, if (𝑑, 𝑧) =(π›Ώβˆ—(𝑣), 𝑣), then for every vertex 𝑒 βˆˆπ‘βˆ’(𝑣) β§΅{𝑒0, … , π‘’π‘Ÿβˆ’1}, (π›Ώβˆ—(𝑒), 𝑒) β‰Ί(π›Ώβˆ—(𝑣), 𝑣).
If we take (𝑑, 𝑧) =(π›Ώβˆ—(𝑣), 𝑣), the lemma implies that for any vertex 𝑣 ∈𝐷𝐼,
π›Ώβˆ’(𝑣)βˆ’π›Ώβˆ—(𝑣)=|{π‘’βˆˆπ‘βˆ’(𝑣)∢(π›Ώβˆ—(𝑒),𝑒)≻(π›Ώβˆ—(𝑣)+1,𝑣)}|.(3)
In other words, if the indegree of a vertex 𝑣drops from π›Ώβˆ’(𝑣)to π›Ώβˆ—(𝑣) =π›Ώβˆ’(𝑣) βˆ’π‘Ÿbefore the outgoing edges of 𝑣are deleted, there
must be π‘Ÿvertices with edges to 𝑣that satisfy the following property: at least one of them, 𝑒0, must have indegree high enough for its
outgoing edges to be deleted before those of 𝑣, i.e., (π›Ώβˆ—(𝑒0), 𝑒0) ≻(π›Ώβˆ’(𝑣), 𝑣); another vertex, 𝑒1, must have indegree high enough for
its outgoing edges to be deleted before those of 𝑣after its indegree is reduced by one, i.e., (π›Ώβˆ—(𝑒1), 𝑒1) ≻(π›Ώβˆ’(𝑣) βˆ’1, 𝑣); and so forth, up
to π‘’π‘Ÿβˆ’1 with (π›Ώβˆ—(π‘’π‘Ÿβˆ’1), π‘’π‘Ÿβˆ’1) ≻(π›Ώβˆ’(𝑣) βˆ’(π‘Ÿ βˆ’1), 𝑣) =(π›Ώβˆ—(𝑣) +1, 𝑣). The other vertices 𝑒with edges to 𝑣must satisfy (π›Ώβˆ—(𝑒), 𝑒) β‰Ί(π›Ώβˆ—(𝑣), 𝑣).
This is illustrated in Fig. 3for the case where the indegree of 𝑣drops by π‘Ÿ =3. When (𝑑, 𝑧) ≻(π›Ώβˆ—(𝑣), 𝑣)it is enough to carry out the
analysis for the first π‘Ÿ =π›Ώβˆ’(𝑣) βˆ’π‘‘+πœ’(𝑣 >𝑧)vertices 𝑒0, … , π‘’π‘Ÿβˆ’1.
To establish impartiality of the twin threshold mechanism with certain thresholds, we need to compare runs of the mechanism
on graphs that differ in the outgoing edges of a single vertex. Intuitively, a change in the outgoing edges will make a difference to
the outcome of the mechanism only if it affects the position of some other vertex relative to the lower threshold 𝑑at the time that
vertex is considered: If at that time the vertex is above the threshold its outgoing edges are deleted, otherwise the edges remain
and are used in the decision of which vertex to select. We are thus interested in pairs of graphs 𝐺1and 𝐺2that differ only in the
outgoing edges of a vertex ξ˜‘π‘£, and which contain a second vertex 𝑣 β‰ ξ˜‘π‘£ such that π›Ώβˆ—(𝑣, 𝐺1) >𝛿
βˆ—(𝑣, 𝐺2). Using Lemma 1, we can derive
conditions in terms of the indegrees of the in-neighbors of 𝑣under which this can happen. Moreover, we can show that one of two
additional conditions must be satisfied: either (i) ξ˜‘π‘£ has an edge to 𝑣in 𝐺1, or (ii) there exists a vertex 𝑒0with an edge to 𝑣in both
𝐺1and 𝐺2such that π›Ώβˆ—(𝑒0, 𝐺1) <𝛿
βˆ—(𝑒0, 𝐺2). We obtain the following lemma.
Lemma 2. Let 𝐺1=(𝑁, 𝐸1), 𝐺2=(𝑁, 𝐸2) βˆˆξˆ³π‘›, 𝑣, ξ˜‘π‘£ βˆˆπ‘with 𝑣 β‰ ξ˜‘π‘£ be such that 𝐸1β§΅({ ξ˜‘π‘£} ×𝑁) =𝐸2β§΅({ ξ˜‘π‘£} ×𝑁), π›Ώβˆ—(𝑣, 𝐺1) >𝛿
βˆ—(𝑣, 𝐺2),
and π›Ώβˆ—(𝑣, 𝐺1) β‰₯𝑑. Consider (𝑑, 𝑧) βˆˆβ„•2such that (π›Ώβˆ—(𝑣, 𝐺1), 𝑣) ≻(𝑑, 𝑧) βͺ°(π›Ώβˆ—(𝑣, 𝐺2), 𝑣), and let π‘Ÿ =π›Ώβˆ—(𝑣, 𝐺1) βˆ’π‘‘+πœ’(𝑣 >𝑧). Then, there
exist vertices 𝑒0, … , π‘’π‘Ÿβˆ’1 such that, for every π‘—βˆˆ{1, … , π‘Ÿ βˆ’1},
(𝑒𝑗,𝑣)∈𝐸1∩𝐸2,(π›Ώβˆ—(𝑒𝑗,𝐺
2),𝑒
𝑗)≻(π›Ώβˆ—(𝑣, 𝐺1)βˆ’π‘—,𝑣),(π›Ώβˆ—(𝑒𝑗,𝐺
1),𝑒
𝑗)β‰Ί(π›Ώβˆ—(𝑣, 𝐺1),𝑣),
and one of the following holds:
(i) (𝑒0, 𝑣) ∈𝐸1⧡𝐸2and if (π›Ώβˆ—(𝑣, 𝐺1), 𝑣) β‰Ί(π›Ώβˆ’(ξ˜‘π‘£, 𝐺1), ξ˜‘π‘£), taking ξ˜‘π‘Ÿ =π›Ώβˆ’(ξ˜‘π‘£, 𝐺1) βˆ’π›Ώβˆ—(𝑣, 𝐺1) +πœ’(ξ˜‘π‘£>𝑣)we have that there are vertices
ξ˜‘π‘’0, … , ξ˜‘π‘’ξ˜‘π‘Ÿβˆ’1, none of them equal to ξ˜‘π‘£, such that (π›Ώβˆ—(ξ˜‘π‘’π‘—, 𝐺1), ξ˜‘π‘’π‘—) ≻(π›Ώβˆ’(ξ˜‘π‘£, 𝐺1) βˆ’π‘—, ξ˜‘π‘£)for every π‘—βˆˆ{0, … , ξ˜‘π‘Ÿ βˆ’1}; or
(ii) (π›Ώβˆ—(𝑒0, 𝐺2), 𝑒0) ≻⋯ ≻(π›Ώβˆ—(π‘’π‘Ÿβˆ’1, 𝐺2), π‘’π‘Ÿβˆ’1), (𝑒0, 𝑣) ∈𝐸1∩𝐸2, and (π›Ώβˆ—(𝑒0, 𝐺2), 𝑒0) ≻(π›Ώβˆ—(𝑣, 𝐺1), 𝑣) ≻(π›Ώβˆ—(𝑒0, 𝐺1), 𝑒0).
Games and Economic Behavior 144 (2024) 203–224
209
J. Cembrano, F. Fischer, D. Hannon et al.
We prove Lemma 2in Appendix B, but provide some intuition for its correctness here. Assume that the indegree of a vertex 𝑣
drops from π›Ώβˆ’(𝑣)to π›Ώβˆ—(𝑣) =π›Ώβˆ’(𝑣) βˆ’π‘Ÿbefore its outgoing edges are deleted. Then, by Lemma 1, there must be π‘Ÿin-neighbors of 𝑣
whose indegrees are at least π›Ώβˆ—(𝑣) +1when their outgoing edges are deleted. Thus, when 𝑣, 𝐺1, and 𝐺2are as in the statement of
Lemma 2, and defining π‘Ÿ1=π›Ώβˆ’(𝑣, 𝐺1) βˆ’π›Ώβˆ—(𝑣, 𝐺1)and π‘Ÿ2=π›Ώβˆ’(𝑣, 𝐺2) βˆ’π›Ώβˆ—(𝑣, 𝐺2), then π‘Ÿ1in-neighbors of 𝑣must have indegree at least
π›Ώβˆ—(𝑣, 𝐺1) +1in 𝐺1upon deletion of their outgoing edges, and π‘Ÿ2in-neighbors of 𝑣must have indegree at least π›Ώβˆ—(𝑣, 𝐺2) +1in 𝐺2
upon deletion of their outgoing edges. There are then two possible reasons for the difference between π›Ώβˆ—(𝑣, 𝐺1)and π›Ώβˆ—(𝑣, 𝐺2). The
first, which can only occur when π‘Ÿ1=π‘Ÿ2, is given in Condition (i) of the lemma: ξ˜‘π‘£ has an edge to 𝑣in 𝐺1but not in 𝐺2, while all the
other indegrees remain the same. However, for this difference to have an impact, the outgoing edges of ξ˜‘π‘£ must be deleted after those
of 𝑣, and Lemma 1implies the existence of in-neighbors of ξ˜‘π‘£ with indegrees as shown. The other reason, which can happen when
π‘Ÿ1=π‘Ÿ2and necessarily happens otherwise, is that some in-neighbor 𝑒0of 𝑣in both 𝐺1and 𝐺2loses its outgoing edges after 𝑣when
the input to the mechanism is 𝐺1, but before 𝑣when the input is 𝐺2. This must happen due to a change in the indegree of 𝑒0at the
time its outgoing edges are deleted, i.e.,
(π›Ώβˆ—(𝑒0,𝐺
2),𝑒
0)≻(π›Ώβˆ—(𝑣, 𝐺1),𝑣)≻(π›Ώβˆ—(𝑒0,𝐺
1),𝑒
0).
This is captured in Condition (ii). In both cases, Lemma 1implies the existence of π›Ώβˆ—(𝑣, 𝐺1) βˆ’π›Ώβˆ—(𝑣, 𝐺2) βˆ’1further in-neighbors of 𝑣
in both graphs, denoted as 𝑒1, … , π‘’π‘Ÿβˆ’1, which lose their outgoing edges after 𝑣in 𝐺1but before 𝑣in 𝐺2. When (𝑑, 𝑧) ≻(π›Ώβˆ—(𝑣, 𝐺2), 𝑣),
it again suffices to carry out the analysis for a smaller subset of the vertices with edges to 𝑣.
Lemma 2implies that whenever 𝐺1and 𝐺2differ only in the outgoing edges of a single vertex ξ˜‘π‘£, and 𝑣is a different vertex with
π›Ώβˆ—(𝑣, 𝐺1) >𝛿
βˆ—(𝑣, 𝐺2), then either (i) ξ˜‘π‘£ has an edge to 𝑣in 𝐺1, or (ii) there exists a vertex 𝑒0with π›Ώβˆ—(𝑒0, 𝐺1) <𝛿
βˆ—(𝑒0, 𝐺2). The fact that
this relationship is the opposite of that for 𝑣naturally suggests an iterative analysis, where the roles of 𝐺1and 𝐺2are exchanged in
each iteration as long as Condition (ii) holds. Such an analysis leads to the following lemma, which establishes a sufficient condition
for impartiality in terms of 𝑇,𝑑, and π‘˜. Impartiality for a particular choice of 𝑇and 𝑑that guarantees the bound of Theorem 1can
then be obtained in a straightforward way.
Lemma 3. For every 𝑛, π‘˜ βˆˆβ„•with π‘˜ ≀𝑛 βˆ’1, the twin threshold mechanism with thresholds 𝑇and 𝑑such that
1
2(𝑇2+𝑇+3π‘‘βˆ’π‘‘2)βˆ’min{π‘‡βˆ’π‘‘, π‘˜}>π‘˜π‘›
is impartial on ξˆ³π‘›(π‘˜).
Proof. Let 𝑓be the selection mechanism given by the twin threshold mechanism with thresholds 𝑇and 𝑑. We suppose that 𝑓
is not impartial and we want to see that the inequality in the statement of the lemma is thus violated. Specifically, let 𝑛 βˆˆβ„•,
𝐺=(𝑁, 𝐸), 𝐺′=(𝑁, 𝐸′) βˆˆξˆ³π‘›, and ξ˜‘π‘£ βˆˆπ‘be such that 𝐸⧡({ ξ˜‘π‘£} ×𝑁) =𝐸′⧡({ ξ˜‘π‘£} ×𝑁)and ξ˜‘π‘£ βˆˆπ‘“(𝐺) □𝑓(𝐺′), i.e., ξ˜‘π‘£ is selected only
for one of these graphs. In particular, π›Ώβˆ’(ξ˜‘π‘£, 𝐺) =π›Ώβˆ’(ξ˜‘π‘£, 𝐺′) β‰₯𝑇. For ξ˜‘π‘£ to be selected only for one of the graphs, there has to be a
vertex 𝑣 βˆˆπ‘β§΅{ξ˜‘π‘£}whose outgoing edges are deleted when the mechanism runs with input 𝐺but not when it runs with input 𝐺′, or
vice versa. Suppose w.l.o.g. that the former holds, so denoting 𝑣0=𝑣we have that π›Ώβˆ—(𝑣0, 𝐺) >𝑑 βˆ’1 β‰₯π›Ώβˆ—(𝑣0, 𝐺′). From Lemma 2with
𝑣 =𝑣0, 𝐺1=𝐺, 𝐺2=𝐺′, and (𝑑, 𝑧) =(𝑑 βˆ’1, 𝑣0), we have that there are π‘Ÿ0=π›Ώβˆ—(𝑣0, 𝐺) βˆ’(𝑑 βˆ’1)vertices 𝑒0
0, … , 𝑒0
π‘Ÿ0βˆ’1 for which (𝑒0
𝑗, 𝑣0) ∈
πΈβˆ©πΈβ€², (π›Ώβˆ—(𝑒0
𝑗, 𝐺′), 𝑒0
𝑗) ≻(π›Ώβˆ—(𝑣0, 𝐺) βˆ’π‘—, 𝑣0), (π›Ώβˆ—(𝑒0
𝑗, 𝐺), 𝑒0
𝑗) β‰Ί(π›Ώβˆ—(𝑣0, 𝐺), 𝑣0)for every π‘—βˆˆ{1, … , π‘Ÿ0βˆ’1}, and one of the conditions in
the lemma holds. If Condition (i) holds, we denote π‘š =0. Otherwise, we have that (π›Ώβˆ—(𝑒0
0, 𝐺′), 𝑒0
0) ≻(π›Ώβˆ—(𝑣0, 𝐺), 𝑣0) ≻(π›Ώβˆ—(𝑒0
0, 𝐺), 𝑒0
0),
thus we can define 𝑣1=𝑒0
0and apply Lemma 2with 𝑣 =𝑣1, 𝐺1=𝐺′, 𝐺2=𝐺, and (𝑑, 𝑧) =(π›Ώβˆ—(𝑣0, 𝐺), 𝑣0).
The argument can be repeated until Condition (i) holds at some iteration, which we denote π‘š. This necessarily happens because
𝑛is finite, and we denote as πΊβˆ—βˆˆ{𝐺, 𝐺′}the graph such that (ξ˜‘π‘£, π‘£π‘š)is an edge of πΊβˆ—. In particular, πΊβˆ—=𝐺if π‘šis even and
πΊβˆ—=𝐺′if π‘šis odd. For every iteration π“βˆˆ{1, … , π‘š βˆ’1}the following holds: There is a vertex 𝑣𝓁=π‘’π“βˆ’1
0and a strictly positive
value π‘Ÿπ“=π›Ώβˆ—(𝑣𝓁, 𝐺) βˆ’π›Ώβˆ—(π‘£π“βˆ’1, 𝐺′) +πœ’(𝑣𝓁>𝑣
π“βˆ’1)(if 𝓁is even) or π‘Ÿπ“=π›Ώβˆ—(𝑣𝓁, 𝐺′) βˆ’π›Ώβˆ—(π‘£π“βˆ’1, 𝐺) +πœ’(𝑣𝓁>𝑣
π“βˆ’1)(if 𝓁is odd) such that
there are vertices 𝑒𝓁
0, … , 𝑒𝓁
π‘Ÿπ“βˆ’1 such that for each π‘—βˆˆ{1, … , π‘Ÿπ“βˆ’1},
(𝑒𝓁
𝑗,𝑣
𝓁)βˆˆπΈβˆ©πΈβ€²,(π›Ώβˆ—(𝑒𝓁
𝑗,𝐺′),𝑒
𝓁
𝑗)≻(π›Ώβˆ—(𝑣𝓁,𝐺)βˆ’π‘—,𝑣𝓁),(π›Ώβˆ—(𝑒𝓁
𝑗,𝐺),𝑒
𝓁
𝑗)β‰Ί(π›Ώβˆ—(𝑣𝓁,𝐺),𝑣
𝓁)(4)
if 𝓁is even, and
(𝑒𝓁
𝑗,𝑣
𝓁)βˆˆπΈβˆ©πΈβ€²,(π›Ώβˆ—(𝑒𝓁
𝑗,𝐺),𝑒
𝓁
𝑗)≻(π›Ώβˆ—(𝑣𝓁,𝐺′)βˆ’π‘—,𝑣𝓁),(π›Ώβˆ—(𝑒𝓁
𝑗,𝐺′),𝑒
𝓁
𝑗)β‰Ί(π›Ώβˆ—(𝑣𝓁,𝐺′),𝑣
𝓁)(5)
if 𝓁is odd. Furthermore, we claim that for every 𝓁, π“β€²βˆˆ{0, … , π‘š βˆ’1} and every π‘—βˆˆ{0, … , π‘Ÿπ“βˆ’1}, π‘—β€²βˆˆ{0, … , π‘Ÿπ“β€²βˆ’1} with
(𝓁, 𝑗) β‰ (𝓁′, 𝑗′)it holds 𝑒𝓁
𝑗≠𝑒𝓁′
𝑗′. This is straightforward if 𝓁=𝓁′, so we only need to show it when 𝓁≠𝓁′. In order to see this, we
actually show the following properties, that directly imply the previous one:
(π›Ώβˆ—(𝑒𝓁
𝑗,𝐺′),𝑒
𝓁
𝑗)≻(π›Ώβˆ—(𝑒𝓁′
𝑗′,𝐺′),𝑒
𝓁′
𝑗′)for every π“βˆˆ{2,…,π‘š}even,π“β€²βˆˆ{0,…,π“βˆ’1},
π‘—βˆˆ{0,…,π‘Ÿ
π“βˆ’1},π‘—β€²βˆˆ{0,…,π‘Ÿ
π“β€²βˆ’1}.(6)
(π›Ώβˆ—(𝑒𝓁
𝑗,𝐺),𝑒
𝓁
𝑗)≻(π›Ώβˆ—(𝑒𝓁′
𝑗′,𝐺),𝑒
𝓁′
𝑗′)for every π“βˆˆ{1,…,π‘š}odd,π“β€²βˆˆ{0,…,π“βˆ’1},
π‘—βˆˆ{0,…,π‘Ÿ
π“βˆ’1},π‘—β€²βˆˆ{0,…,π‘Ÿ
π“β€²βˆ’1}.(7)
Games and Economic Behavior 144 (2024) 203–224
210
J. Cembrano, F. Fischer, D. Hannon et al.
We prove this by induction over 𝓁, distinguishing whether this is an even or odd value. Let first π“βˆˆ{2, … , π‘š}be an even value and
note that
(π›Ώβˆ—(𝑒𝓁
𝑗,𝐺′),𝑒
𝓁
𝑗)≻(π›Ώβˆ—(𝑣𝓁,𝐺)βˆ’π‘—,𝑣𝓁)≻(π›Ώβˆ—(π‘£π“βˆ’1,𝐺′),𝑣
π“βˆ’1)for every π‘—βˆˆ{0,…,π‘Ÿ
π“βˆ’1},
where the first inequality follows from (4)and the second one from the definition of π‘Ÿπ“. By (5)and the fact that π“βˆ’1 β‰ π‘š, we have
that (π›Ώβˆ—(π‘£π“βˆ’1, 𝐺′), π‘£π“βˆ’1) ≻(π›Ώβˆ—(π‘’π“βˆ’1
𝑗′, 𝐺′), π‘’π“βˆ’1
𝑗′)for each π‘—β€²βˆˆ{0, … , π‘Ÿπ“βˆ’1 βˆ’1}, and from Condition (ii) of Lemma 2we also have that
(π›Ώβˆ—(π‘£π“βˆ’1, 𝐺′), π‘£π“βˆ’1) =(π›Ώβˆ—(π‘’π“βˆ’2
0, 𝐺′), π‘’π“βˆ’2
0) ≻(π›Ώβˆ—(π‘’π“βˆ’2
𝑗′, 𝐺′), π‘’π“βˆ’2
𝑗′)for each π‘—β€²βˆˆ{1, … , π‘Ÿπ“βˆ’2 βˆ’1}. Therefore, we conclude that for every
even 𝓁we have both
(π›Ώβˆ—(𝑒𝓁
𝑗,𝐺′),𝑒
𝓁
𝑗)≻(π›Ώβˆ—(π‘’π“βˆ’1
𝑗′,𝐺′),𝑒
π“βˆ’1
𝑗′)for every π‘—β€²βˆˆ{0,…,π‘Ÿ
π“βˆ’1 βˆ’1},
and
(π›Ώβˆ—(𝑒𝓁
𝑗,𝐺′),𝑒
𝓁
𝑗)≻(π›Ώβˆ—(π‘’π“βˆ’2
𝑗′,𝐺′),𝑒
π“βˆ’2
𝑗′)for every π‘—β€²βˆˆ{0,…,π‘Ÿ
π“βˆ’2 βˆ’1}.
This proves the claim for the base case 𝓁=2, and also implies that if it holds for every even π“β‰€ξ˜•
𝓁where
ξ˜•
𝓁β‰₯2is even, then it holds
for
ξ˜•
𝓁+2as well. For odd values of 𝓁, the claim follows from an analogous argument where the roles of 𝐺and 𝐺′are exchanged,
with the only difference that for the base case 𝓁=1, 𝓁′=π“βˆ’1is the only possibility.
For the last step 𝓁=π‘š, if π›Ώβˆ—(π‘£π‘š, πΊβˆ—) <𝛿
βˆ’(ξ˜‘π‘£, 𝐺), taking ξ˜‘π‘Ÿ =π›Ώβˆ’(ξ˜‘π‘£, 𝐺) βˆ’π›Ώβˆ—(π‘£π‘š, πΊβˆ—) +πœ’(ξ˜‘π‘£>𝑣
π‘š)we have that Lemma 2guarantees
the existence of vertices ξ˜‘π‘’0, … , ξ˜‘π‘’ξ˜‘π‘Ÿβˆ’1, none of them equal to ξ˜‘π‘£, such that for each π‘—βˆˆ{0, … , ξ˜‘π‘Ÿ βˆ’1},
(ξ˜‘π‘’π‘—,ξ˜‘π‘£)βˆˆπΈβˆ©πΈβ€²,(π›Ώβˆ—(ξ˜‘π‘’π‘—,πΊβˆ—),ξ˜‘π‘’π‘—)≻(π›Ώβˆ’(ξ˜‘π‘£, 𝐺)βˆ’π‘—, ξ˜‘π‘£).(8)
Furthermore, we claim for every π“βˆˆ{0, … , π‘š βˆ’1}, π‘—βˆˆ{0, … , π‘Ÿπ“βˆ’1}, and π‘—β€²βˆˆ{0, … , ξ˜‘π‘Ÿ βˆ’1}that 𝑒𝓁
π‘—β‰ ξ˜‘π‘’π‘—β€². In order to see this, we
observe that (8)and the definition of ξ˜‘π‘Ÿ imply that for every π‘—βˆˆ{0, … , ξ˜‘π‘Ÿ βˆ’1}and every π‘—β€²βˆˆ{1, … , π‘Ÿπ‘šβˆ’1},
(π›Ώβˆ—(ξ˜‘π‘’π‘—,πΊβˆ—),ξ˜‘π‘’π‘—)≻(π›Ώβˆ—(π‘£π‘š,πΊβˆ—),𝑣
π‘š)=(π›Ώβˆ—(π‘’π‘šβˆ’1
0,πΊβˆ—),𝑒
π‘šβˆ’1
0)≻(π›Ώβˆ—(π‘’π‘šβˆ’1
𝑗′,πΊβˆ—),𝑒
π‘šβˆ’1
𝑗′),
where the last inequality follows from Condition (ii) of Lemma 2. Therefore, by (7), if π‘šis even, then for every π“βˆˆ{0, … , π‘š βˆ’1},
π‘—βˆˆ{0, … , π‘Ÿπ“βˆ’1}, and π‘—β€²βˆˆ{0, … , ξ˜‘π‘Ÿ βˆ’1},
(π›Ώβˆ—(ξ˜‘π‘’π‘—β€²,𝐺),ξ˜‘π‘’π‘—β€²)=(π›Ώβˆ—(ξ˜‘π‘’π‘—β€²,πΊβˆ—),ξ˜‘π‘’π‘—β€²)≻(π›Ώβˆ—(𝑒𝓁
𝑗,𝐺),𝑒
𝓁
𝑗).
Similarly, if π‘šis odd, (6)implies for every π“βˆˆ{0, … , π‘š βˆ’1}, π‘—βˆˆ{0, … , π‘Ÿπ“βˆ’1}, and π‘—β€²βˆˆ{0, … , ξ˜‘π‘Ÿ βˆ’1}that
(π›Ώβˆ—(ξ˜‘π‘’π‘—β€²,𝐺′),ξ˜‘π‘’π‘—β€²)=(π›Ώβˆ—(ξ˜‘π‘’π‘—β€²,πΊβˆ—),ξ˜‘π‘’π‘—β€²)≻(π›Ώβˆ—(𝑒𝓁
𝑗,𝐺′),𝑒
𝓁
𝑗).
This concludes the proof of the claim.
In the following, we will establish a lower bound on the sum of the indegrees of the vertices in graph 𝐺. We start by observing
that, for any subset of vertices π‘†βŠ†π‘β§΅{𝑣0}, defining πΊπ‘£βˆˆ{𝐺, 𝐺′}arbitrarily for each 𝑣 βˆˆπ‘†we have
βˆ‘
π‘£βˆˆπ‘β§΅{𝑣0}
π›Ώβˆ’(𝑣, 𝐺)β‰₯max{βˆ‘
π‘£βˆˆπ‘†
(π›Ώβˆ’(𝑣, 𝐺𝑣)βˆ’1)+1,βˆ‘
π‘£βˆˆπ‘†
π›Ώβˆ’(𝑣, 𝐺𝑣)βˆ’π‘˜}.(9)
To see this fix 𝑆and {𝐺𝑣}π‘£βˆˆπ‘†as described, and note that since 𝐺and 𝐺′only differ in the outgoing edges of ξ˜‘π‘£, it holds that
βˆ‘
π‘£βˆˆπ‘†
π›Ώβˆ’(𝑣, 𝐺𝑣)βˆ’ βˆ‘
π‘£βˆˆπ‘β§΅{𝑣0}
π›Ώβˆ’(𝑣, 𝐺)β‰€βˆ‘
π‘£βˆˆπ‘†
π›Ώβˆ’(𝑣, 𝐺𝑣)βˆ’βˆ‘
π‘£βˆˆπ‘†
π›Ώβˆ’(𝑣, 𝐺)β‰€π‘˜.
Moreover, we know that π›Ώβˆ’(ξ˜‘π‘£, 𝐺) =π›Ώβˆ’(ξ˜‘π‘£, πΊξ˜‘π‘£) β‰₯𝑇β‰₯1, and for every 𝑣 βˆˆπ‘that π›Ώβˆ’(𝑣, 𝐺) β‰₯π›Ώβˆ’(𝑣, 𝐺𝑣) βˆ’1. If ξ˜‘π‘£ βˆˆπ‘†, this yields
βˆ‘
π‘£βˆˆπ‘β§΅{𝑣0}
π›Ώβˆ’(𝑣, 𝐺)β‰₯π›Ώβˆ’(ξ˜‘π‘£, 𝐺 ξ˜‘π‘£)+ βˆ‘
π‘£βˆˆπ‘†β§΅{ξ˜‘π‘£}
(π›Ώβˆ’(𝑣, 𝐺𝑣)βˆ’1)=βˆ‘
π‘£βˆˆπ‘†
(π›Ώβˆ’(𝑣, 𝐺𝑣)βˆ’1)+1.
On the other hand, if ξ˜‘π‘£ βˆ‰π‘†, we have
βˆ‘
π‘£βˆˆπ‘β§΅{𝑣0}
π›Ώβˆ’(𝑣, 𝐺)β‰₯π›Ώβˆ’(ξ˜‘π‘£, 𝐺)+βˆ‘
π‘£βˆˆπ‘†
(π›Ώβˆ’(𝑣, 𝐺𝑣)βˆ’1)β‰₯βˆ‘
π‘£βˆˆπ‘†
(π›Ώβˆ’(𝑣, 𝐺𝑣)βˆ’1)+1.
This concludes the proof of (9).
We now claim that if π›Ώβˆ—(π‘£π‘š, πΊβˆ—) β‰₯π›Ώβˆ’(ξ˜‘π‘£, 𝐺), then
βˆ‘
π‘£βˆˆπ‘β§΅{𝑣0}
π›Ώβˆ’(𝑣, 𝐺)+π›Ώβˆ’(𝑣0,𝐺)
Games and Economic Behavior 144 (2024) 203–224
211
J. Cembrano, F. Fischer, D. Hannon et al.
β‰₯max⎧
βŽͺ
⎨
βŽͺ
βŽ©βˆ‘
𝓁<π‘š even
π‘Ÿπ“βˆ’1
βˆ‘
𝑗=0
(π›Ώβˆ’(𝑒𝓁
𝑗,𝐺′)βˆ’1)+ βˆ‘
𝓁<π‘š odd
π‘Ÿπ“βˆ’1
βˆ‘
𝑗=0
(π›Ώβˆ’(𝑒𝓁
𝑗,𝐺)βˆ’1)+1,
βˆ‘
𝓁<π‘š even
π‘Ÿπ“βˆ’1
βˆ‘
𝑗=0
π›Ώβˆ’(𝑒𝓁
𝑗,𝐺′)+ βˆ‘
𝓁<π‘š odd
π‘Ÿπ“βˆ’1
βˆ‘
𝑗=0
π›Ώβˆ’(𝑒𝓁
𝑗,𝐺)βˆ’π‘˜βŽ«
βŽͺ
⎬
βŽͺ
⎭
+π›Ώβˆ’(𝑣0,𝐺)
β‰₯max⎧
βŽͺ
⎨
βŽͺ
βŽ©βˆ‘
𝓁<π‘š even
π‘Ÿπ“βˆ’1
βˆ‘
𝑗=0
(π›Ώβˆ—(𝑣𝓁,𝐺)βˆ’π‘—βˆ’1)+ βˆ‘
𝓁<π‘š odd
π‘Ÿπ“βˆ’1
βˆ‘
𝑗=0
(π›Ώβˆ—(𝑣𝓁,𝐺′)βˆ’π‘—βˆ’1)+1,
βˆ‘
𝓁<π‘š even
π‘Ÿπ“βˆ’1
βˆ‘
𝑗=0
(π›Ώβˆ—(𝑣𝓁,𝐺)βˆ’π‘—)+ βˆ‘
𝓁<π‘š odd
π‘Ÿπ“βˆ’1
βˆ‘
𝑗=0
(π›Ώβˆ—(𝑣𝓁,𝐺′)βˆ’π‘—)βˆ’π‘˜βŽ«
βŽͺ
⎬
βŽͺ
⎭
+π›Ώβˆ’(𝑣0,𝐺)
β‰₯max{π›Ώβˆ—(π‘£π‘š,πΊβˆ—)βˆ’π‘‘
βˆ‘
𝑗=0
(π›Ώβˆ—(π‘£π‘š,πΊβˆ—)βˆ’π‘—βˆ’1)+1,
π›Ώβˆ—(π‘£π‘š,πΊβˆ—)βˆ’π‘‘
βˆ‘
𝑗=0
(π›Ώβˆ—(π‘£π‘š,πΊβˆ—)βˆ’π‘—)βˆ’π‘˜}+𝑑
=
π›Ώβˆ—(π‘£π‘š,πΊβˆ—)
βˆ‘
𝑗=𝑑
π‘—βˆ’min{π›Ώβˆ—(π‘£π‘š,πΊβˆ—)βˆ’π‘‘, π‘˜}+𝑑
β‰₯
𝑇
βˆ‘
𝑗=𝑑
π‘—βˆ’min{π‘‡βˆ’π‘‘, π‘˜}+𝑑=1
2(𝑇2+𝑇+3π‘‘βˆ’π‘‘2)βˆ’min{π‘‡βˆ’π‘‘, π‘˜}.
Indeed, the first inequality follows from (9) applied to 𝑆={𝑒𝓁
π‘—βˆΆπ“βˆˆ{0, … , π‘š βˆ’1}, π‘—βˆˆ{0, … , π‘Ÿπ“βˆ’1}}, and the facts that 𝑣0≠𝑒𝓁
𝑗
for any π“βˆˆ{0, … , π‘š βˆ’1} and π‘—βˆˆ{0, … , π‘Ÿπ“βˆ’1} and that 𝑒𝓁
𝑗≠𝑒𝓁′
𝑗′whenever (𝓁, 𝑗) β‰ (𝓁′, 𝑗′). The second inequality uses that
π›Ώβˆ—(𝑣,

𝐺) β‰€π›Ώβˆ’(𝑣,

𝐺)for every

𝐺=( 
𝑁,

𝐸) ∈and 𝑣 ∈
𝑁, as well as (4) and (5). The third inequality uses that π›Ώβˆ’(𝑣0, 𝐺) β‰₯𝑑and that,
by Lemma 2, π›Ώβˆ—(𝑣𝓁, 𝐺) β‰₯π›Ώβˆ—(𝑒𝓁
0, 𝐺) =π›Ώβˆ—(𝑣𝓁+1, 𝐺′) βˆ’π‘Ÿπ“+1 if 𝓁is even, and π›Ώβˆ—(𝑣𝓁, 𝐺′) β‰₯π›Ώβˆ—(𝑒𝓁
0, 𝐺′) =π›Ώβˆ—(𝑣𝓁+1, 𝐺) βˆ’π‘Ÿπ“+1 if 𝓁is odd. The
first equality follows from rearranging terms, the equality from simple computations, and the final inequality by observing that the
expression is increasing in π›Ώβˆ—(π‘£π‘š, πΊβˆ—)and this value is bounded from below by π›Ώβˆ’(ξ˜‘π‘£, 𝐺) β‰₯𝑇.
On the other hand, if π›Ώβˆ—(π‘£π‘š, πΊβˆ—) <𝛿
βˆ’(ξ˜‘π‘£, 𝐺), then the final inequality does not hold. We can, however, use (8)to conclude that
βˆ‘
π‘£βˆˆπ‘β§΅{𝑣0}
π›Ώβˆ’(𝑣, 𝐺)+π›Ώβˆ’(𝑣0,𝐺)
β‰₯max⎧
βŽͺ
⎨
βŽͺ
βŽ©βˆ‘
𝓁<π‘š even
π‘Ÿπ“βˆ’1
βˆ‘
𝑗=0
(π›Ώβˆ’(𝑒𝓁
𝑗,𝐺′)βˆ’1)+ βˆ‘
𝓁<π‘š odd
π‘Ÿπ“βˆ’1
βˆ‘
𝑗=0
(π›Ώβˆ’(𝑒𝓁
𝑗,𝐺)βˆ’1)+
ξ˜‘π‘Ÿβˆ’1
βˆ‘
𝑗=0
(π›Ώβˆ’(ξ˜‘π‘’π‘—,πΊβˆ—)βˆ’1)+1,
βˆ‘
𝓁<π‘š even
π‘Ÿπ“βˆ’1
βˆ‘
𝑗=0
π›Ώβˆ’(𝑒𝓁
𝑗,𝐺′)+ βˆ‘
𝓁<π‘š odd
π‘Ÿπ“βˆ’1
βˆ‘
𝑗=0
π›Ώβˆ’(𝑒𝓁
𝑗,𝐺)+
ξ˜‘π‘Ÿβˆ’1
βˆ‘
𝑗=0
π›Ώβˆ’(ξ˜‘π‘’π‘—,πΊβˆ—)βˆ’π‘˜βŽ«
βŽͺ
⎬
βŽͺ
⎭
+π›Ώβˆ’(𝑣0,𝐺)
β‰₯max{π›Ώβˆ—(π‘£π‘š,πΊβˆ—)βˆ’π‘‘
βˆ‘
𝑗=0
(π›Ώβˆ—(π‘£π‘š,πΊβˆ—)βˆ’π‘—βˆ’1)+
ξ˜‘π‘Ÿβˆ’1
βˆ‘
𝑗=0
(π›Ώβˆ’(ξ˜‘π‘£, 𝐺)βˆ’π‘—βˆ’1)+1,
π›Ώβˆ—(π‘£π‘š,πΊβˆ—)βˆ’π‘‘
βˆ‘
𝑗=0
(π›Ώβˆ—(π‘£π‘š,πΊβˆ—)βˆ’π‘—)+
ξ˜‘π‘Ÿβˆ’1
βˆ‘
𝑗=0
(π›Ώβˆ’(ξ˜‘π‘£, 𝐺)βˆ’π‘—)βˆ’π‘˜}+π›Ώβˆ’(𝑣0,𝐺)
β‰₯max{π›Ώβˆ’(ξ˜‘π‘£,𝐺)
βˆ‘
𝑗=𝑑
(π‘—βˆ’1)+1,
π›Ώβˆ’(ξ˜‘π‘£,𝐺)
βˆ‘
𝑗=𝑑
π‘—βˆ’π‘˜}+𝑑
=
π›Ώβˆ’(ξ˜‘π‘£,𝐺)
βˆ‘
𝑗=𝑑
π‘—βˆ’min{π›Ώβˆ’(ξ˜‘π‘£, 𝐺)βˆ’π‘‘, π‘˜}+𝑑
β‰₯
𝑇
βˆ‘
𝑗=𝑑
π‘—βˆ’min{π‘‡βˆ’π‘‘, π‘˜}+𝑑=1
2(𝑇2+𝑇+3π‘‘βˆ’π‘‘2)βˆ’min{π‘‡βˆ’π‘‘, π‘˜}.
Games and Economic Behavior 144 (2024) 203–224
212
J. Cembrano, F. Fischer, D. Hannon et al.
Fig. 4. By changing its outgoing edge, ξ˜‘π‘£ is able to affect whether it is selected by the mechanism or not. Lemma 3gives a condition over 𝑇, 𝑑, and π‘˜such that this
cannot happen. In this example, the sum of the indegrees of the vertices of 𝐺1that are shown in the figure is 5(𝑑 +1), so we need that 5(𝑑 +1) β‰€π‘˜π‘›. On the other
hand, when 𝑇=𝑑 +2Lemma 3states that impartiality is guaranteed as long as 4𝑑 +3 >π‘˜π‘› +min{π‘˜, 2}. The graphs of this example clearly violate this condition.
Indeed, the first inequality follows from (9) applied to 𝑆={𝑒𝓁
π‘—βˆΆπ“βˆˆ{0, … , π‘š βˆ’1}, π‘—βˆˆ{0, … , π‘Ÿπ“βˆ’1}} βˆͺ{ξ˜‘π‘’π‘—βˆΆπ‘—βˆˆ{0, … , ξ˜‘π‘Ÿ βˆ’1}},
and the facts that 𝑣0≠𝑒𝓁
𝑗for any π“βˆˆ{0, … , π‘š βˆ’1}and π‘—βˆˆ{0, … , π‘Ÿπ“βˆ’1}, that 𝑒𝓁
𝑗≠𝑒𝓁′
𝑗′whenever (𝓁, 𝑗) β‰ (𝓁′, 𝑗′), and that 𝑒𝓁
π‘—β‰ ξ˜‘π‘’π‘—β€²
for every π“βˆˆ{0, … , π‘š βˆ’1}, π‘—βˆˆ{0, … , π‘Ÿπ“βˆ’1}, and π‘—β€²βˆˆ{0, … , ξ˜‘π‘Ÿ βˆ’1}. The second inequality follows analogously to the previous
sequence of inequalities but also uses (8)to bound π›Ώβˆ’(ξ˜‘π‘’π‘—, πΊβˆ—). The other inequalities follow analogously to before, with the exception
that for the last inequality we now use that the expression is increasing in π›Ώβˆ’(ξ˜‘π‘£, 𝐺), which is again bounded from below by 𝑇.
Since the sum of the indegrees in a graph is at most the maximum number π‘˜π‘› of edges, we conclude in any of the previous cases
that
1
2(𝑇2+𝑇+3π‘‘βˆ’π‘‘2)βˆ’min{π‘‡βˆ’π‘‘, π‘˜}β‰€π‘˜π‘›. β–‘
Fig. 4illustrates this result by showing a situation where the outgoing edge of a vertex ξ˜‘π‘£ with π›Ώβˆ’(ξ˜‘π‘£, 𝐺1) =π›Ώβˆ’(ξ˜‘π‘£, 𝐺2) =𝑇determines
whether π›Ώβˆ—(𝑣) β‰₯𝑑for another vertex 𝑣, and thus whether ξ˜‘π‘£ itself is selected or not. Note that in the example there exist vertices 𝑀𝑗
such that π›Ώβˆ’(𝑀𝑗, 𝐺) β‰₯𝑑 +𝑗for every π‘—βˆˆ{0, … , π‘‡βˆ’π‘‘}and some 𝐺∈{𝐺1, 𝐺2}.
It is worth noting that Lemma 3implies the impartiality of both the supermajority rule with threshold βŒŠπ‘›βˆ•2βŒ‹ +1and the two
contenders mechanism. When taking a single threshold 𝑇=𝑑, the sufficient condition for impartiality becomes 𝑇>π‘˜π‘›βˆ•2, which
is clearly satisfied when π‘˜ =1and 𝑇=βŒŠπ‘›βˆ•2βŒ‹ +1as in the supermajority rule with threshold βŒŠπ‘›βˆ•2βŒ‹ +1. When taking thresholds
that differ by 1, 𝑇=𝑑 +1, the sufficient condition for impartiality becomes 𝑇>π‘˜π‘›βˆ•3 +1, which is clearly satisfied when π‘˜ =1
and 𝑇=βŒŠπ‘›βˆ•3βŒ‹ +2as in the two contenders mechanism. We will now prove Theorem 1by using this same condition to guarantee
impartiality, but choosing the thresholds in such a way that the additive deviation is as small as possible.
Proof of Theorem 1.Let 𝑛 βˆˆβ„•+, πΊβˆˆξˆ³π‘›(π‘˜)and πœ…, 𝑐>0such that π‘˜ β‰€π‘π‘›πœ…. Let 𝑓be the selection mechanism given by the twin
threshold mechanism with thresholds 𝑇=⌊√8𝑐𝑛
1+πœ…
2βŒ‹ +1and 𝑑 =βŒˆβˆšπ‘π‘›
1+πœ…
2βŒ‰. Since π‘˜(𝑛 +1) ≀𝑐𝑛1+πœ…+π‘π‘›πœ…β‰€2𝑐𝑛1+πœ…, we have from
Lemma 3that a sufficient condition for impartiality is that
1
2(𝑇2+𝑇+3π‘‘βˆ’π‘‘2)>2𝑐𝑛1+πœ….
Replacing 𝑇and 𝑑yields
1
2(𝑇2+𝑇+3π‘‘βˆ’π‘‘2)>1
2(8𝑐𝑛1+πœ…+√8𝑐𝑛
1+πœ…
2+3
βˆšπ‘π‘›
1+πœ…
2βˆ’βŒˆβˆšπ‘π‘›
1+πœ…
2βŒ‰2)>2𝑐𝑛1+πœ…,
where we used that π‘π‘›πœ…>1and thus βŒˆβˆšπ‘π‘›
1+πœ…
2βŒ‰2<4𝑐𝑛1+πœ…. We conclude that the mechanism is impartial for these values of 𝑇and 𝑑.
In order to obtain the additive bound, first consider a graph 𝐺=(𝑁, 𝐸)such that 𝑓returns the empty set when run with this
graph as input. Let π‘£βˆ—be such that π›Ώβˆ’(π‘£βˆ—) =Ξ”(𝐺)and note that necessarily
ξ˜•
𝛿𝐼(π‘£βˆ—) β‰€π‘‡βˆ’1. Since there are at most βŒŠπ‘˜π‘›βˆ•π‘‘βŒ‹vertices
with indegree 𝑑or higher, a maximum of βŒŠπ‘˜π‘›βˆ•π‘‘βŒ‹ βˆ’1 in-neighbors of π‘£βˆ—has their outgoing edges deleted during the algorithm.
Therefore, we conclude that Ξ”(𝐺) ≀𝑇+βŒŠπ‘˜π‘›βˆ•π‘‘βŒ‹ βˆ’2.
Games and Economic Behavior 144 (2024) 203–224
213
J. Cembrano, F. Fischer, D. Hannon et al.
Consider now 𝐺=(𝑁, 𝐸)such that the mechanism returns a set {𝑣}, and let π‘£βˆ—be such that π›Ώβˆ’(π‘£βˆ—) =Ξ”(𝐺). Once again, a
maximum of βŒŠπ‘˜π‘›βˆ•π‘‘βŒ‹ βˆ’1in-neighbors of π‘£βˆ—has their outgoing edges deleted during the algorithm. Using the fact that
ξ˜•
𝛿𝐼(π‘£βˆ—) β‰€ξ˜•
𝛿𝐼(𝑣)
since 𝑣is selected, we conclude that
Ξ”(𝐺)βˆ’π›Ώβˆ’(𝑣)≀(π›Ώβˆ’(𝑣)+βŒŠπ‘˜π‘›
π‘‘βŒ‹βˆ’1
)βˆ’π›Ώβˆ’(𝑣)=βŒŠπ‘˜π‘›
π‘‘βŒ‹βˆ’1.
Since the value obtained in the former case is greater than or equal to the one obtained in the latter for any values of 𝑇and 𝑑, we
have that 𝑓is 𝛼-additive for 𝛼=𝑇+βŒŠπ‘˜π‘›βˆ•π‘‘βŒ‹ βˆ’2. Therefore, for the specified values of 𝑇and 𝑑and given the upper bound on π‘˜, 𝑓is
𝛼-additive for
𝛼=⌊√8𝑐𝑛
1+πœ…
2βŒ‹+1+⎒⎒⎒⎒⎣
𝑐𝑛1+πœ…
βŒˆβˆšπ‘π‘›
1+πœ…
2βŒ‰βŽ₯βŽ₯βŽ₯βŽ₯⎦
βˆ’2≀(√8+1)
βˆšπ‘π‘›
1+πœ…
2=𝑂(𝑛
1+πœ…
2).
We conclude that 𝑓is 𝑂(𝑛
1+πœ…
2)-additive.
When π‘˜ =1, a more detailed analysis yields the bound of
√7.25𝑛. First observe that in this case, by Lemma 3, impartiality holds
whenever
1
2(𝑇2+𝑇+3π‘‘βˆ’π‘‘2)βˆ’(𝑛+1)>0
and thus when
𝑇2+π‘‡βˆ’(𝑑2βˆ’3𝑑+2𝑛+2)>0.
The left-hand side is equal to zero if and only if
𝑇=βˆ’1 Β± √4(𝑑2βˆ’3𝑑+2𝑛+2)+1
2=Β±
βˆšπ‘‘2βˆ’3𝑑+2𝑛+9
4βˆ’1
2,
and since 𝑇has to be non-negative impartiality holds if
𝑇>
βˆšπ‘‘2βˆ’3𝑑+2𝑛+9
4βˆ’1
2.
This is trivially satisfied if we take
𝑑=⌈1.23βˆšπ‘›βŒ‰,𝑇=⌊√⌈1.23βˆšπ‘›βŒ‰2βˆ’3⌈1.23βˆšπ‘›βŒ‰+2𝑛+9
4+1
2βŒ‹.
As before, the twin threshold mechanism with thresholds 𝑇and 𝑑is 𝛼-additive for any 𝛼β‰₯𝛼(𝑛) ∢= 𝑇+βŒŠπ‘›βˆ•π‘‘βŒ‹ βˆ’2. In order to obtain
an upper bound on 𝛼(𝑛), we start by bounding 𝑇from above:
π‘‡β‰€βˆš(1.23βˆšπ‘›+1)
2βˆ’3(1.23βˆšπ‘›+1)+2𝑛+9
4+1
2
=√3.5129π‘›βˆ’1.23βˆšπ‘›+1
4+1
2,
where the first inequality holds because π‘Ž2βˆ’3π‘Žis increasing for π‘Ž β‰₯3βˆ•2. Then,
𝛼(𝑛)=𝑇+βŒŠπ‘›βˆ•π‘‘βŒ‹βˆ’2 β‰€βˆš3.5129π‘›βˆ’1.23βˆšπ‘›+1
4+1
2+βŒŠπ‘›
⌈1.23βˆšπ‘›βŒ‰βŒ‹βˆ’2
β‰€βˆš3.5129π‘›βˆ’1.23βˆšπ‘›+1
4+βˆšπ‘›
1.23 βˆ’3
2=∢ β„Ž(𝑛).
Let 𝑔(𝑛) ∢= √7.25π‘›βˆ’β„Ž(𝑛). Then, by the previous inequality,
𝑔(1) = √7.25 βˆ’ √3.5129 βˆ’ 1.23 + 1
4βˆ’1
1.23 +3
2
=√7.25 βˆ’ √2.5329 βˆ’ 1
1.23 +3
2β‰ˆ1.79 >0.
Moreover,
𝑔′(𝑛)=√7.25
2βˆšπ‘›
βˆ’7.0258βˆšπ‘›βˆ’1.23
2βˆšπ‘›βˆš14.0516π‘›βˆ’4.92βˆšπ‘›+1
βˆ’1
2.46βˆšπ‘›
Games and Economic Behavior 144 (2024) 203–224
214
J. Cembrano, F. Fischer, D. Hannon et al.
=
(2.46√7.25 βˆ’ 2)√14.0516π‘›βˆ’4.92βˆšπ‘›+1βˆ’17.283468βˆšπ‘›+3.0258
4.92βˆšπ‘›βˆš14.0516π‘›βˆ’4.92βˆšπ‘›+1
,
β‰₯
(2.46√7.25 βˆ’ 2)√14.05π‘›βˆ’4.92βˆšπ‘›+1βˆ’17.3βˆšπ‘›+3
4.92βˆšπ‘›βˆš14.0516π‘›βˆ’4.92βˆšπ‘›+1
,
=
45(2.46√7.25 βˆ’ 2)√14.05π‘›βˆ’4.92βˆšπ‘›+ 1 βˆ’ 778.5βˆšπ‘›+ 135
221.4βˆšπ‘›βˆš14.0516π‘›βˆ’4.92βˆšπ‘›+1
,
which is non-negative when 𝑛 β‰₯1.
To see this, note that the denominator of the last expression is positive and that 45(2.46√7.25 βˆ’ 2) >208, so that 𝑔′(𝑛) β‰₯0if
208√14.05π‘›βˆ’4.92βˆšπ‘›+1β‰₯778.5βˆšπ‘›βˆ’ 135,
i.e., if
1796.95π‘›βˆ’ 2663.88βˆšπ‘›+ 25039 β‰₯0.
This inequality indeed holds for every 𝑛 β‰₯1. This is immediate for 𝑛 ∈{1, 2, 3}. For 𝑛 β‰₯4we have that 𝑛 β‰₯2βˆšπ‘›, and thus the
expression on the left is greater than or equal to 930.02βˆšπ‘›+ 25039, which is clearly non-negative. We conclude that 𝑔(𝑛) β‰₯0for
every 𝑛 βˆˆβ„•and thus 𝛼(𝑛) β‰€β„Ž(𝑛) β‰€βˆš7.25𝑛for every 𝑛 β‰₯1, so 𝑓is
√7.25𝑛-additive. β–‘
Recalling Fig. 1, we note that for small values of 𝑛, the guarantee given by the optimal twin threshold mechanism is much smaller
than √7.25𝑛, and that the bound becomes significantly better than those given by the other mechanisms as 𝑛grows. Closing the
gap between the lower bound of Theorem 4and the upper bound given by the optimal twin threshold mechanism is left as our main
open question.
4. A tight impossibility result for approval
So far, we have developed a new mechanism for impartial selection and have established an additive performance guarantee for
the mechanism relative to the maximum outdegree in the graph. We will now take a closer look at the case where the maximum
outdegree is unbounded, i.e., at the approval setting.
When applied to the approval setting, Theorem 1provides a performance guarantee of 𝑂(𝑛). As the maximum indegree in a graph
with 𝑛vertices is 𝑛 βˆ’1, this bound is trivially achieved by any impartial mechanism including the mechanism that never selects.
Caragiannis et al. (2022)have used a careful case analysis to show that deterministic impartial mechanisms cannot be better than
3-additive. We show that the trivial upper bound of 𝑛 βˆ’1is in fact tight for all 𝑛, which means that the mechanism that never selects
provides the best possible additive performance guarantee among all deterministic impartial mechanisms. Our result is in fact more
general and again holds relative to the maximum outdegree π‘˜.
Theorem 2. Let 𝑛 βˆˆβ„•and π‘˜ ≀𝑛 βˆ’1. Let 𝑓be an impartial deterministic selection mechanism such that 𝑓is 𝛼-additive on ξˆ³π‘›(π‘˜). Then
𝛼β‰₯π‘˜. In particular, if 𝑓is 𝛼-additive on ξˆ³π‘›, then 𝛼β‰₯𝑛 βˆ’1.
In the practically relevant case where individuals are not allowed to abstain and the minimum outdegree is therefore at least 1,
a small improvement can be obtained by selecting a vertex with an incoming edge from a fixed vertex, and again breaking ties by a
fixed ordering of the vertices. The selected vertex then has indegree at least 1, which for the approval setting implies (𝑛 βˆ’2)-additivity.
This guarantee is again best possible.
Theorem 3. Let 𝑛 βˆˆβ„•and π‘˜ ≀𝑛 βˆ’1. Let 𝑓be an impartial deterministic selection mechanism such that 𝑓is 𝛼-additive on +
𝑛(π‘˜). Then
𝛼β‰₯π‘˜ βˆ’1. In particular, if 𝑓is 𝛼-additive on +
𝑛, then 𝛼β‰₯𝑛 βˆ’2.
To prove both theorems we study the performance of impartial, but not necessarily deterministic, selection mechanisms on a
particular class of graphs which in the case of 𝑛vertices we denote by ξˆ³π‘‡
𝑛. For each 𝑛 βˆˆβ„•, a graph 𝐺=(𝑁, 𝐸) βˆˆξˆ³π‘›belongs to ξˆ³π‘‡
𝑛
if and only if there exists an π‘Ÿ-partition of 𝑁for some π‘Ÿ β‰₯1, which we denote by (𝑆1, … , π‘†π‘Ÿ), such that (i) 𝑒 <𝑣for every 𝑒 βˆˆπ‘†π‘–,
and 𝑣 βˆˆπ‘†π‘—with 𝑖 <𝑗, and (ii) 𝐸={(𝑒, 𝑣) βˆˆπ‘†π‘–Γ—π‘†π‘—βˆΆπ‘–, π‘—βˆˆ{1, … , π‘Ÿ}, 𝑖 ≀𝑗, 𝑒 ≠𝑣}. In other words, ξˆ³π‘‡
𝑛contains all graphs obtained by
taking an ordered partition of a set of 𝑛unlabeled vertices and adding edges from each vertex to all other vertices in the same part
and in greater parts. We will not be interested in isomorphic graphs within the class and thus only consider partitions of the vertices
in increasing order. A graph in ξˆ³π‘‡
𝑛is thus characterized by the partition (𝑆1, … , π‘†π‘Ÿ), or by the tuple (𝑠1, … , π‘ π‘Ÿ)where 𝑠𝑖=|𝑆𝑖|for
Games and Economic Behavior 144 (2024) 203–224
215
J. Cembrano, F. Fischer, D. Hannon et al.
𝑠(𝐺)
πœ†πΊ
(1,1)
2
(2)
1
(1,1,1)
6
(1,2)
3
(2,1)
3
(3)
1
Fig. 5. Graphs in ξˆ³π‘‡
2and ξˆ³π‘‡
3.
each 𝑖 ∈{1, … , π‘Ÿ}. For a given graph πΊβˆˆξˆ³π‘‡
𝑛, we denote the former by 𝑆(𝐺), the latter by 𝑠(𝐺), and the length of 𝑠(𝐺)by π‘Ÿ(𝐺).
Finally, for πΊβˆˆξˆ³π‘‡
𝑛, let πœ†πΊbe the number of graphs with 𝑛vertices isomorphic to 𝐺, i.e.,
πœ†πΊ=𝑛!
βˆπ‘Ÿ(𝐺)
𝑖=1 (𝑠(𝐺))𝑖!
.
Fig. 5shows the graphs in ξˆ³π‘‡
2and ξˆ³π‘‡
3, along with their tuple representation 𝑠(𝐺)and associated values πœ†πΊ. The sums, for 𝑛 βˆˆβ„•, of
the values πœ†πΊfor all graphs πΊβˆˆξˆ³π‘‡
𝑛are known as Fubini numbers and count the number of weak orders on an 𝑛-element set. The
following lemma establishes a property of Fubini numbers that is readily appreciated for the cases shown in Fig. 5but in fact holds
for all 𝑛. The property was known previously (Diagana and MaΓ―ga, 2017), but we provide an alternative proof in Appendix Cfor the
sake of completeness.
Lemma 4. For every 𝑛 βˆˆβ„•, 𝑛 β‰₯1,
βˆ‘πΊβˆˆξˆ³π‘‡
π‘›πœ†πΊis an odd number.
For every pair of graphs 𝐺, πΊβ€²βˆˆξˆ³π‘‡
𝑛and π‘—βˆˆ{2, … , π‘Ÿ(𝐺)}, we say that there is a 𝑗-transition from 𝐺to 𝐺′if π‘Ÿ(𝐺) =π‘Ÿ(𝐺′) +1,
(𝑠(𝐺))𝑗=1, and
(𝑠(𝐺′))𝑖=⎧
βŽͺ
⎨
βŽͺ
⎩
(𝑠(𝐺))𝑖if π‘–β‰€π‘—βˆ’2,
(𝑠(𝐺))𝑖+1 if 𝑖=π‘—βˆ’1,
(𝑠(𝐺))𝑖+1 if 𝑖β‰₯𝑗.
Intuitively, a 𝑗-transition can be obtained by changing the outgoing edges of the single vertex in the set (𝑆(𝐺))𝑗, including not only
edges to vertices in β‹ƒπ‘Ÿ(𝐺)
𝑖=𝑗(𝑆(𝐺))𝑖but also to vertices in (𝑆(𝐺))π‘—βˆ’1. When the value of 𝑗is not relevant in a particular context, we
simply say that there is a transition from 𝐺to 𝐺′if there exists some π‘—βˆˆ{2, … , π‘Ÿ(𝐺)} such that there is a 𝑗-transition from 𝐺to
𝐺′. Observe that for every pair of graphs 𝐺, πΊβ€²βˆˆξˆ³π‘‡
𝑛there is at most one π‘—βˆˆ{2, … , π‘Ÿ(𝐺)} such that there is a 𝑗-transition from 𝐺to
𝐺′, and that if there is a transition from 𝐺to 𝐺′, there cannot be a transition from 𝐺′to 𝐺. This kind of relation between ordered
partitions has been exploited by Insko et al. (2017)for studying an expansion of the determinant, giving rise to a partial order on ξˆ³π‘‡
𝑛.
In our context, it turns out to be relevant because whenever there is a 𝑗-transition from 𝐺to 𝐺′, any impartial mechanism either
selects the vertex in (𝑆(𝐺))𝑗both in 𝐺and 𝐺′, or in none of them.
In the case of plurality, impartiality was shown by Holzman and Moulin (2013)to be incompatible with two further axioms:
positive unanimity, which requires for all 𝐺=(𝑁, 𝐸) ∈(1) that 𝑣 βˆˆπ‘“(𝐺)if π›Ώβˆ’(𝑣) =|𝑁| βˆ’1; and negative unanimity, which requires
for all 𝐺=(𝑁, 𝐸) ∈(1) that 𝑣 βˆ‰π‘“(𝐺)if π›Ώβˆ’(𝑣) =0. This result holds even on a restricted class of graphs, consisting of a single
cycle and additional vertices with edges onto that cycle, and has immediate and very strong implications on the best multiplicative
approximation guarantee that an impartial mechanism can achieve. For additive performance guarantees, however, the incompat-
ibility of impartiality with the other two axioms implies only a lower bound of 2. We will see in the following that on the class
ξˆ³π‘‡
𝑛, impartiality is incompatible with a single axiom that weakens positive and negative unanimity. Strong lower bounds regarding
additive performance guarantees for approval then follow immediately.
The class ξˆ³π‘‡
𝑛is very different, and has to be very different, from the class of graphs used by Holzman and Moulin, and will
ultimately require a new analysis. We can, however, follow the approach of Holzman and Moulin to consider randomized mechanisms
rather than deterministic ones, which allows us without loss of generality to restrict attention to mechanisms that treat vertices
symmetrically. A randomized selection mechanism for is given by a family of functions π‘“βˆΆξˆ³π‘›β†’[0, 1]𝑛that maps each graph to
a probability distribution on the set of its vertices, such that βˆ‘π‘›
𝑖=1 (𝑓(𝐺))𝑖≀1for every graph πΊβˆˆξˆ³π‘›. Analogously to the case
of deterministic mechanisms, we say that a randomized selection mechanism 𝑓is impartial on ξˆ³β€²βŠ†ξˆ³if for every pair of graphs
𝐺=(𝑁, 𝐸)and 𝐺′=(𝑁, 𝐸′)in ξˆ³β€²and every 𝑣 βˆˆπ‘, (𝑓(𝐺))𝑣=(𝑓(𝐺′))𝑣whenever 𝐸⧡({𝑣} ×𝑁) =𝐸′⧡({𝑣} ×𝑁). We say that a
randomized selection mechanism 𝑓satisfies weak unanimity on ξˆ³π‘›if for every 𝐺=(𝑁, 𝐸) βˆˆξˆ³π‘›such that π›Ώβˆ’(𝑣) =𝑛 βˆ’1 for some
𝑣 βˆˆπ‘,
βˆ‘
π‘’βˆˆπ‘βˆΆπ›Ώβˆ’(𝑒)β‰₯1
(𝑓(𝐺))𝑒β‰₯1.
Games and Economic Behavior 144 (2024) 203–224
216
J. Cembrano, F. Fischer, D. Hannon et al.
2
3
2
2
4
3
2
3
2
3
2
2
(1,1) (2)
(1,2)
(2,1)
(1,1,1)
(3)
(1,1,1,1)
(1,3)
(2,2)
(3,1)
(1,1,2)
(1,2,1)
(2,1,1)
(4)
Fig. 6. Directed versions of 2, 3, and 4. For 𝑛 ∈{2, 3, 4}, each graph πΊβˆˆξˆ³π‘‡
𝑛is represented by the tuple 𝑠(𝐺)and each edge (𝐺, 𝐺′)labeled with π‘—βˆˆβ„•represents
a 𝑗-transition from 𝐺to 𝐺′.
In other words, weak unanimity requires that a vertex with positive indegree is chosen with probability 1whenever there exists a
vertex with indegree 𝑛 βˆ’1. We finally say that a randomized mechanism 𝑓is symmetric if it is invariant with respect to renaming of
the vertices, i.e., if for every 𝐺=(𝑁, 𝐸) ∈, every 𝑣 βˆˆπ‘and every permutation πœ‹=(πœ‹1, … , πœ‹|𝑁|)of 𝑁, (𝑓(πΊπœ‹))πœ‹π‘£=(𝑓(𝐺))𝑣, where
πΊπœ‹=(𝑁, πΈπœ‹)with πΈπœ‹={(πœ‹π‘’, πœ‹π‘£) ∢(𝑒, 𝑣) ∈𝐸}. For a given randomized mechanism 𝑓, we denote by 𝑓sthe mechanism obtained by
applying a random permutation πœ‹to the vertices of the input graph, invoking 𝑓, and permuting the result by the inverse of πœ‹. Thus,
for all 𝑛 βˆˆβ„•, πΊβˆˆξˆ³π‘›, and 𝑣 βˆˆπ‘,
(𝑓s(𝐺))𝑣=1
𝑛!βˆ‘
πœ‹βˆˆξˆΏπ‘›
(𝑓(πΊπœ‹))πœ‹π‘£,
where ξˆΏπ‘›is the set of all permutations πœ‹=(πœ‹1, … , πœ‹π‘›)of a set of 𝑛elements.
The following lemma, which we prove in Appendix D, establishes that 𝑓sis symmetric for every randomized mechanism 𝑓and
inherits impartiality and weak unanimity from 𝑓. The lemma is a straightforward variant of a result of Holzman and Moulin and will
allow us to restrict attention to symmetric randomized mechanisms.
Lemma 5. Let 𝑓be a randomized selection mechanism that is impartial and weakly unanimous on ξˆ³π‘›. Then, 𝑓sis symmetric, impartial, and
weakly unanimous on ξˆ³π‘›.
We are now ready to state our axiomatic impossibility result, which can be seen as a stronger version of that of Holzman and
Moulin for the case of unbounded outdegree. Both lower bounds follow easily from this result.
Lemma 6. For every 𝑛 βˆˆβ„•, 𝑛 β‰₯2, there exists no randomized selection mechanism 𝑓satisfying impartiality and weak unanimity on ξˆ³π‘‡
𝑛.
Proof. Let 𝑛 βˆˆβ„•, 𝑛 β‰₯2and suppose that there exists a randomized selection mechanism 𝑓satisfying impartiality and weak una-
nimity on ξˆ³π‘›. Since we can assume symmetry due to Lemma 5, for each graph πΊβˆˆξˆ³π‘‡
𝑛we have that for every 𝑖 ∈{1, … , π‘Ÿ(𝐺)} and
every 𝑒, 𝑣 ∈(𝑆(𝐺))𝑖, (𝑓(𝐺))𝑒=(𝑓(𝐺))𝑣, and thus we denote this value simply as (𝑓(𝐺))𝑖. We consider for the proof an undirected
graph ξˆ΄π‘›=(ξˆ³π‘‡
𝑛, ), such that for every pair of graphs 𝐺, πΊβ€²βˆˆξˆ³π‘‡
𝑛we have that {𝐺, 𝐺′} ∈if and only if there is a transition from
𝐺to 𝐺′or from 𝐺′to 𝐺. By definition of a transition, for each {𝐺, 𝐺′} ∈we have |π‘Ÿ(𝐺) βˆ’π‘Ÿ(𝐺′)| =1. Therefore, ξˆ΄π‘›is bipartite
with partition (𝐿𝑛, 𝑅𝑛)where 𝐿𝑛={πΊβˆˆξˆ³π‘‡
π‘›βˆΆπ‘Ÿ(𝐺)is even}and 𝑅𝑛={πΊβˆˆξˆ³π‘‡
π‘›βˆΆπ‘Ÿ(𝐺)is odd}. Fig. 6depicts the graphs 2, 3, and
4. For each graph πΊβˆˆξˆ³π‘‡
𝑛, we define 𝑖(𝐺) =2if (𝑠(𝐺))1=1and 𝑖(𝐺) =1otherwise.
Let 𝐺=(𝑁, 𝐸)and 𝐺′=(𝑁, 𝐸′)be two graphs in ξˆ³π‘‡
𝑛such that there is a 𝑗-transition from 𝐺to 𝐺′for some π‘—βˆˆ{2, … , π‘Ÿ(𝐺)}.
Denoting by 𝑣the unique vertex in (𝑆(𝐺))𝑗, which is also in 𝑆(𝐺′)π‘—βˆ’1, we have that 𝐸⧡({𝑣} ×𝑁) =𝐸′⧡({𝑣} ×𝑁), and since 𝑓is
impartial, (𝑓(𝐺))𝑗=(𝑓(𝐺′))π‘—βˆ’1. Therefore,
πœ†πΊβ€²β‹…(𝑠(𝐺′))π‘—βˆ’1 β‹…(𝑓(𝐺′))π‘—βˆ’1 =𝑛!
βˆπ‘Ÿ(𝐺′)
𝑖=1 (𝑠(𝐺′))𝑖!
(𝑠(𝐺′))π‘—βˆ’1(𝑓(𝐺′))π‘—βˆ’1
=𝑛!
((𝑠(𝐺′))π‘—βˆ’1 βˆ’1)!βˆπ‘–βˆˆ{1,…,π‘Ÿ(𝐺′)}β§΅{π‘—βˆ’1} (𝑠(𝐺′))𝑖!(𝑓(𝐺′))π‘—βˆ’1
=𝑛!
((𝑠(𝐺))π‘—βˆ’1)! βˆπ‘–βˆˆ{1,…,π‘Ÿ(𝐺)}β§΅{π‘—βˆ’1,𝑗}(𝑠(𝐺))𝑖!(𝑓(𝐺))𝑗
=𝑛!
βˆπ‘–βˆˆ{1,…,π‘Ÿ(𝐺)} (𝑠(𝐺))𝑖!(𝑠(𝐺))𝑗⋅(𝑓(𝐺))𝑗
Games and Economic Behavior 144 (2024) 203–224
217
J. Cembrano, F. Fischer, D. Hannon et al.
Algorithm 2: Selection mechanism 𝑓based on π‘“π‘˜.
Input: Digraph 𝐺=(𝑁, 𝐸) βˆˆξˆ³π‘˜+1, mechanism π‘“π‘˜and integer 𝑛 β‰₯π‘˜ +1.
Output: Set 𝑆of selected vertices with |𝑆| ≀1.
Let 𝐻=(𝑁βˆͺ𝑁′, 𝐸), where 𝑁′=β‹ƒπ‘›βˆ’π‘˜βˆ’1
𝑗=1 {𝑒𝑗};
Return π‘“π‘˜(𝐻)
=πœ†πΊβ‹…(𝑠(𝐺))𝑗⋅(𝑓(𝐺))𝑗.
The first two and last two equalities are obtained by replacing known expressions and simple calculations. The third equality comes
from the fact that (𝑠(𝐺′))𝑖=(𝑠(𝐺))𝑖for every 𝑖 β‰€π‘—βˆ’2, (𝑠(𝐺′))π‘—βˆ’1 =(𝑠(𝐺))π‘—βˆ’1 +1, and (𝑠(𝐺′))𝑖=(𝑠(𝐺))𝑖+1 for every 𝑖 β‰₯𝑗. Moreover,
observe that for each πΊβˆˆξˆ³π‘‡
𝑛and for each π‘—βˆˆ{𝑖(𝐺), … , π‘Ÿ(𝐺)}, there exists exactly one πΊβ€²βˆˆξˆ³π‘‡
𝑛such that there is a 𝑗-transition from
𝐺to 𝐺′(if (𝑠(𝐺))𝑗=1) or a (𝑗+1)-transition from 𝐺′to 𝐺(if (𝑠(𝐺))𝑗β‰₯2). Therefore,
βˆ‘
πΊβˆˆπΏπ‘›
π‘Ÿ(𝐺)
βˆ‘
𝑖=𝑖(𝐺)
πœ†πΊβ‹…(𝑠(𝐺))𝑖⋅(𝑓(𝐺))𝑖=βˆ‘
πΊβˆˆπ‘…π‘›
π‘Ÿ(𝐺)
βˆ‘
𝑖=𝑖(𝐺)
πœ†πΊβ‹…(𝑠(𝐺))𝑖⋅(𝑓(𝐺))𝑖.(10)
We now derive two important sets of inequalities. From the fact that 𝑓is a selection mechanism, for each πΊβˆˆξˆ³π‘‡
𝑛we have that
π‘Ÿ(𝐺)
βˆ‘
𝑖=𝑖(𝐺)
(𝑠(𝐺))𝑖⋅(𝑓(𝐺))𝑖≀1,(11)
where replacing 1 by 𝑖(𝐺)on the left-hand side is possible since it can only make the sum smaller and thus keeps the inequality. On
the other hand, from the fact that 𝑓satisfies weak unanimity, for each πΊβˆˆξˆ³π‘‡
𝑛we have that
βˆ’
π‘Ÿ(𝐺)
βˆ‘
𝑖=𝑖(𝐺)
(𝑠(𝐺))𝑖⋅(𝑓(𝐺))π‘–β‰€βˆ’1,(12)
where we can omit the term for 𝑖 =1on the left-hand side whenever (𝑠(𝐺))1=1, because the vertex in (𝑆(𝐺))1has indegree 0 in
such case.
In order to cancel out the left-hand sides of the previous inequalities, we assign a sign to each part of the bipartition of ξˆ΄π‘›.
Let sign(𝐿𝑛), sign(𝑅𝑛) ∈{βˆ’1, 1} with sign(𝐿𝑛) β‹…sign(𝑅𝑛) =βˆ’1, and let sign(𝐺) =sign(𝐿𝑛)for each πΊβˆˆπΏπ‘›and sign(𝐺) =sign(𝑅𝑛)
for each πΊβˆˆπ‘…π‘›. Summing up the inequalities (11)multiplied by πœ†πΊfor every πΊβˆˆξˆ³π‘‡
𝑛with sign(𝐺) =1and the inequalities (12)
multiplied by πœ†πΊfor every πΊβˆˆξˆ³π‘‡
𝑛with sign(𝐺) =βˆ’1, we obtain
βˆ‘
πΊβˆˆξˆ³π‘‡
𝑛
π‘Ÿ(𝐺)
βˆ‘
𝑖=𝑖(𝐺)
sign(𝐺)β‹…πœ†πΊβ‹…(𝑠(𝐺))𝑖⋅(𝑓(𝐺))π‘–β‰€βˆ‘
πΊβˆˆξˆ³π‘‡
𝑛
sign(𝐺)β‹…πœ†πΊ.
By expression (10)the left-hand side is equal to 0. However, we know from Lemma 4that
βˆ‘πΊβˆˆξˆ³π‘‡
π‘›πœ†πΊis odd, so the right-hand side
cannot be equal to 0. For one of the two possible choices of sign(𝐿𝑛)and sign(𝑅𝑛)the right-hand side is negative, and we obtain
a contradiction. We conclude that a randomized selection mechanism 𝑓satisfying impartiality and weak unanimity on ξˆ³π‘›cannot
exist. β–‘
As an illustration, the counterexamples constructed for 𝑛 =3and 𝑛 =4are shown in Fig. 7.
We are now ready to prove Theorems 2and 3. In order to be able to apply Lemma 6to deterministic mechanisms, we need a
simple definition. For a given deterministic selection mechanism π‘“βˆΆξˆ³π‘›β†’2𝑁, let 𝑓rand βˆΆξˆ³π‘›β†’[0, 1]𝑛be the randomized selection
mechanism such that (𝑓rand(𝐺))𝑣=1if 𝑣 βˆˆπ‘“(𝐺)and (𝑓rand(𝐺))𝑣=0otherwise. It is then easy to see that whenever 𝑓is impartial,
𝑓rand is impartial as well.
Proof of Theorem 2.The result is straightforward when 𝑛 =1. Let 𝑛 β‰₯2and π‘˜ ≀𝑛 βˆ’1, and suppose that there is an impartial
deterministic selection mechanism π‘“π‘˜with
Ξ”(𝐺)βˆ’π›Ώβˆ’(π‘“π‘˜(𝐺),𝐺)β‰€π‘˜βˆ’1
for every πΊβˆˆξˆ³π‘›(π‘˜). We define the deterministic selection mechanism 𝑓based on π‘“π‘˜as specified in Algorithm 2. This mechanism
receives a graph 𝐺=(𝑁, 𝐸)in ξˆ³π‘˜+1 and adds new vertices, if necessary to complete 𝑛vertices. These vertices are isolated, in the
sense that the set of edges in this new graph 𝐻remains the same. For every input graph 𝐺, the graph constructed belongs to ξˆ³π‘›(π‘˜),
so the mechanism finally applies π‘“π‘˜. We claim that 𝑓rand is impartial and weakly unanimous on ξˆ³π‘‡
π‘˜+1. To see that 𝑓rand is weakly
unanimous, observe that
π›Ώβˆ’(𝑣, 𝐻)=π›Ώβˆ’(𝑣, 𝐺)for every π‘£βˆˆπ‘, and π›Ώβˆ’(𝑣, 𝐻)=0for every π‘£βˆ‰π‘.
Games and Economic Behavior 144 (2024) 203–224
218
J. Cembrano, F. Fischer, D. Hannon et al.
Fig. 7. Counterexamples for the existence of a randomized selection mechanism satisfying impartiality and weak unanimity for 𝑛 =3and for 𝑛 =4. The graphs are
depicted with the impartial probabilities assigned to each vertex with indegree at least one. The inequalities included below each of them are obtained from imposing
either that the probabilities sum up to at most 1β€”since at most one vertex is to be selectedβ€”or sum up to at least 1β€”due to weak unanimity. The values in parentheses
show that the inequalities written for each set of graphs constitute an infeasible system because if one multiplies the corresponding inequality by them and sums the
resulting inequalities, the contradiction 0 β‰€βˆ’1 is achieved.
Algorithm 3: Selection mechanism 𝑓based on 𝑓+
π‘˜.
Input: Digraph 𝐺=(𝑁, 𝐸) βˆˆξˆ³π‘˜, mechanism 𝑓+
π‘˜and integer 𝑛 β‰₯π‘˜ +1.
Output: Set 𝑆of selected vertices with |𝑆| ≀1.
Let 𝐻=(𝑁βˆͺ𝑁′, 𝐹), where 𝑁′=β‹ƒπ‘›βˆ’π‘˜
𝑗=1 {𝑒𝑗}and 𝐹=𝐸βˆͺ(𝑁′×𝑁) βˆͺ({𝑣 βˆˆπ‘βˆΆπ›Ώ+(𝑣, 𝐺) =0} Γ—{𝑒1});
Return 𝑓+
π‘˜(𝐻)
For each πΊβˆˆξˆ³π‘‡
π‘˜+1, we have that Ξ”(𝐺) =π‘˜and thus Ξ”(𝐻) =π‘˜. From the fact that π‘“π‘˜is (π‘˜ βˆ’1)-additive on ξˆ³π‘›(π‘˜), we conclude that
𝑓returns a vertex π‘£βˆ—of 𝐻with π›Ώβˆ’(π‘£βˆ—, 𝐻) β‰₯π‘˜ βˆ’(π‘˜ βˆ’1) =1. This implies, in the first place, that π‘£βˆ—βˆˆπ‘, thus 𝑓is indeed a selection
mechanism on ξˆ³π‘‡
π‘˜+1. Furthermore, we have π›Ώβˆ’(π‘£βˆ—, 𝐺) β‰₯1, thus
βˆ‘
π‘£βˆˆπΊβˆΆπ›Ώβˆ’(𝑣)β‰₯1
(𝑓rand(𝐺))𝑣=1,
i.e., 𝑓rand is weakly unanimous on ξˆ³π‘‡
π‘˜+1. Impartiality of 𝑓rand is straightforward since π‘“π‘˜is impartial and the set of edges is not
modified in the mechanism. This contradicts Lemma 6, so we conclude that mechanism π‘“π‘˜cannot exist. β–‘
Proof of Theorem 3.The result is straightforward when 𝑛 ∈{1, 2}. Let 𝑛 β‰₯3and π‘˜ ≀𝑛 βˆ’1, and suppose that there is an impartial
deterministic selection mechanism 𝑓+
π‘˜with
Ξ”(𝐺)βˆ’π›Ώβˆ’(𝑓+
π‘˜(𝐺),𝐺)β‰€π‘˜βˆ’2
for every 𝐺∈+
𝑛(π‘˜). We define the deterministic selection mechanism 𝑓based on 𝑓+
π‘˜as specified in Algorithm 3. This mechanism
requires a graph 𝐺in ξˆ³π‘˜and adds new vertices 𝑁′={𝑒1, … , π‘’π‘›βˆ’π‘˜}, as well as edges from each of these vertices to every vertex of 𝐺,
and edges from every vertex of 𝐺with outdegree zero to 𝑒1. We first claim that for every input graph 𝐺, the graph 𝐻constructed
in the mechanism belongs to +
𝑛(π‘˜). Indeed, since πΊβˆˆξˆ³π‘˜every vertex 𝑣 βˆˆπ‘satisfies 𝛿+(𝑣, 𝐺) β‰€π‘˜ βˆ’1. Moreover, an outgoing edge
to 𝑒1is added for every 𝑣with 𝛿+(𝑣, 𝐺) =0, thus 1 β‰€π›Ώβˆ’(𝑣, 𝐻) β‰€π‘˜ βˆ’1for each 𝑣 βˆˆπ‘. On the other hand, each node in 𝑁′has
outdegree π‘˜. Therefore, the mechanism is well defined, in the sense that in its last step it applies 𝑓+
π‘˜to a graph in +
𝑛(π‘˜). We claim
that 𝑓rand is impartial and weakly unanimous on ξˆ³π‘‡
π‘˜, which is a clear contradiction to Lemma 6and thus implies that mechanism 𝑓+
π‘˜
cannot exist. We prove the claim in what follows.
Games and Economic Behavior 144 (2024) 203–224
219
J. Cembrano, F. Fischer, D. Hannon et al.
For each 𝐺=(𝑁, 𝐸) βˆˆξˆ³π‘‡
π‘˜, the set {𝑣 βˆˆπ‘βˆΆπ›Ώ+(𝑣) =0}is equal to (𝑆(𝐺))π‘Ÿ(𝐺)if (𝑠(𝐺))π‘Ÿ(𝐺)=1, or to the empty set, otherwise.
Therefore, for 𝑒1, … , π‘’π‘›βˆ’π‘˜and 𝐻as defined in the mechanism we have that π›Ώβˆ’(𝑒1, 𝐻) ≀1, π›Ώβˆ’(𝑒𝑗, 𝐻) =0for every π‘—βˆˆ{2, … , 𝑛 βˆ’π‘˜},
and π›Ώβˆ’(𝑣, 𝐻) =π›Ώβˆ’(𝑣, 𝐺) +𝑛 βˆ’π‘˜for every 𝑣 βˆˆπ‘. Since Ξ”(𝐺) =π‘˜ βˆ’1from the definition of the set ξˆ³π‘‡
π‘˜, we have that Ξ”(𝐻) =𝑛 βˆ’1.
Using that 𝑓+
π‘˜is (π‘˜ βˆ’2)-additive on +
𝑛(π‘˜), we conclude that 𝑓returns a vertex π‘£βˆ—βˆˆπ‘βˆͺ𝑁′with π›Ώβˆ’(π‘£βˆ—, 𝐻) β‰₯𝑛 βˆ’1 βˆ’(π‘˜ βˆ’2) =
𝑛 βˆ’π‘˜ +1 β‰₯2. This implies, in the first place, that π‘£βˆ—βˆˆπ‘, thus 𝑓is indeed a selection mechanism on ξˆ³π‘‡
π‘˜. Furthermore, we have
π›Ώβˆ’(π‘£βˆ—, 𝐺) =π›Ώβˆ’(π‘£βˆ—, 𝐻) βˆ’(𝑛 βˆ’π‘˜) β‰₯1, thus
βˆ‘
π‘£βˆˆπΊβˆΆπ›Ώβˆ’(𝑣)β‰₯1
(𝑓rand(𝐺))𝑣=1,
i.e., 𝑓rand is weakly unanimous on ξˆ³π‘‡
π‘˜.
To see that 𝑓is impartial on ξˆ³π‘‡
π‘˜, let 𝐺=(𝑁, 𝐸), 𝐺′=(𝑁, 𝐸′) βˆˆξˆ³π‘‡
π‘˜and ξ˜‘π‘£ βˆˆπ‘be such that 𝐸⧡({ ξ˜‘π‘£} ×𝑁) =𝐸′⧡({ ξ˜‘π‘£} ×𝑁).
Denoting 𝐹and 𝐹′the edges of the graphs defined in the mechanism 𝑓when run with input 𝐺and 𝐺′, respectively, it is enough to
show that 𝐹⧡({ ξ˜‘π‘£} Γ—(𝑁βˆͺ𝑁′)) =𝐹′⧡({ ξ˜‘π‘£} Γ—(𝑁βˆͺ𝑁′)), because this would imply
𝑓(𝐺)∩{ξ˜‘π‘£}=𝑓+
π‘˜(𝑁βˆͺ𝑁′,𝐹)∩{ξ˜‘π‘£}=𝑓+
π‘˜(𝑁βˆͺ𝑁′,𝐹′)∩{ξ˜‘π‘£}=𝑓(𝐺′)∩{ξ˜‘π‘£}
where the second equality holds since 𝑓+
π‘˜is impartial on +
𝑛(π‘˜)by hypothesis. Indeed,
𝐹⧡({ ξ˜‘π‘£}Γ—(𝑁βˆͺ𝑁′)) = (𝐸βˆͺ(𝑁′×𝑁)
βˆͺ({π‘£βˆˆπ‘βˆΆπ›Ώ+(𝑣, 𝐺)=0}Γ—{𝑒1})) β§΅({ ξ˜‘π‘£}Γ—(𝑁βˆͺ𝑁′))
=𝐸⧡({ ξ˜‘π‘£}×𝑁)βˆͺ(𝑁′×𝑁)
βˆͺ({π‘£βˆˆπ‘β§΅{ξ˜‘π‘£}βˆΆπ›Ώ+(𝑣, 𝐺)=0}Γ—{𝑒1})
=𝐸′⧡({ ξ˜‘π‘£}×𝑁)βˆͺ(𝑁′×𝑁)
βˆͺ({π‘£βˆˆπ‘β§΅{ξ˜‘π‘£}βˆΆπ›Ώ+(𝑣, 𝐺′)=0}Γ—{𝑒1})
=(𝐸′βˆͺ(𝑁′×𝑁)
βˆͺ({π‘£βˆˆπ‘βˆΆπ›Ώ+(𝑣, 𝐺′)=0}Γ—{𝑒1})) β§΅({ ξ˜‘π‘£}Γ—(𝑁βˆͺ𝑁′))
=𝐹′⧡({ ξ˜‘π‘£}Γ—(𝑁βˆͺ𝑁′)),
where the third equality uses the fact that the outgoing edges of every vertex in (𝑁βˆͺ𝑁′) β§΅{ξ˜‘π‘£}are the same in 𝐺and 𝐺′. This
implies that 𝑓rand is impartial as well, concluding the proof of the claim and the proof of the theorem. β–‘
Theorems 2and 3provide tight bounds for the approval setting but have very weak implications for plurality. We end with small
but nontrivial lower bounds for the latter.
Theorem 4. Let 𝑛 βˆˆβ„•and let 𝑓be an impartial deterministic selection mechanism such that 𝑓is 𝛼-additive on ξˆ³π‘›(1). Then, 𝛼β‰₯𝛼𝑛, where
𝛼𝑛=⎧
βŽͺ
⎨
βŽͺ
⎩
1if π‘›βˆˆ{2,3},
2if 4≀𝑛≀9,
3if 𝑛β‰₯10.
Proof. Denote 𝑆1={2, 3}, 𝑆2={4, 5, … , 9}, and 𝑆3={10, 11, …} for simplicity. The graphs in Fig. 8, Fig. 9, and Fig. 10 show the
probabilities that impartial randomized selection mechanisms on ξˆ³π‘›(1) assign to each vertex for 𝑛 βˆˆπ‘†1, 𝑆2, and 𝑆3, respectively.
Vertices not shown in the figures have no incident edges. The inequalities below each graph πΊβˆˆξˆ³π‘›(1) are obtained either from (i)
imposing that the corresponding mechanism on this set selects with probability 1a vertex with indegree at least Ξ”(𝐺), Ξ”(𝐺) βˆ’1, or
Ξ”(𝐺) βˆ’2for 𝑛in 𝑆1, 𝑆2, or 𝑆3, respectively, or (ii) the fact that the probabilities assigned to the vertices of a single graph add up to at
most one. The values in parentheses show that the inequalities written for each set of graphs constitute an infeasible system because
if one multiplies the corresponding inequality by them and sums the resulting inequalities, the contradiction 0 β‰€βˆ’1 is achieved.
Defining 𝛼𝑛as in the statement of the theorem, this shows that for 𝑛 βˆˆβ„•and for every randomized mechanism β„ŽβˆΆξˆ³π‘›β†β†β†’ [0, 1]𝑛, there
exists 𝐺=(𝑁, 𝐸) βˆˆξˆ³π‘›(1) such that
βˆ‘
π‘£βˆˆπ‘βˆΆπ›Ώβˆ’(𝑣)β‰₯Ξ”βˆ’(π›Όπ‘›βˆ’1)
(β„Ž(𝐺))𝑣<1.
Suppose now for contradiction that there exists 𝑛 βˆˆβ„•and π›Όβ‰€π›Όπ‘›βˆ’1 such that there is an impartial deterministic selection
mechanism 𝑓which is 𝛼-additive on ξˆ³π‘›(1). It is clear that 𝑓rand is also impartial and, moreover, for every graph 𝐺=(𝑁, 𝐸) βˆˆξˆ³π‘›(1)
with Ξ”(𝐺) β‰₯𝛼𝑛,
Games and Economic Behavior 144 (2024) 203–224
220
J. Cembrano, F. Fischer, D. Hannon et al.
Fig. 8. Proof of impossibility of the existence of an impartial randomized selection mechanism selecting vertices with indegree Ξ”(𝐺)for each graph πΊβˆˆξˆ³π‘›(1) and
𝑛 ∈{2, 3}. Scaling and summing the inequalities leads to 0 β‰€βˆ’1, a contradiction.
Fig. 9. Proof of impossibility of the existence of an impartial randomized selection mechanism selecting vertices with indegree at least Ξ”(𝐺) βˆ’1 for each graph
πΊβˆˆξˆ³π‘›(1) and 4 ≀𝑛 ≀9. Scaling and summing the inequalities leads to 0 β‰€βˆ’1, a contradiction.
βˆ‘
π‘£βˆˆπ‘βˆΆπ›Ώβˆ’(𝑣)β‰₯Ξ”βˆ’(π›Όπ‘›βˆ’1)
(𝑓rand(𝐺))𝑣=1,
since 𝑓(𝐺) =π‘£βˆ—with π›Ώβˆ’(π‘£βˆ—) β‰₯Ξ” βˆ’(π›Όπ‘›βˆ’1). But this contradicts the non-existence of such impartial randomized selection mechanisms.
We conclude that for every 𝑛 βˆˆβ„•, if 𝑓is an impartial deterministic selection mechanism which is 𝛼-additive on ξˆ³π‘›(1), then 𝛼β‰₯
𝛼𝑛.β–‘
It is worth noting that these bounds apply to the setting without abstentions as well, with the only difference that any deterministic
selection mechanism is impartial and 0-additive on +
2(1), thus the lower bound of 1 does not hold for 𝑛 =2.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.
Data availability
No data was used for the research described in the article.
Acknowledgments
We are grateful to the anonymous referees for their careful reading of the manuscript and their detailed comments. Funded by the
Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under project number 431465007 and by the Engineering
and Physical Sciences Research Council under grant EP/T015187/1.
Appendix A. Proof of Lemma 1
Let 𝐺and 𝑣be as defined in the statement of the lemma. We prove the existence of vertices as claimed by induction over 𝑗. For
the base case, suppose that for every 𝑒 βˆˆπ‘βˆ’(𝑣)it holds (π›Ώβˆ—(𝑒), 𝑒) β‰Ί(π›Ώβˆ’(𝑣), 𝑣). From the definition of the mechanism we thus have
that π‘–βˆ—(𝑒) >𝑖
βˆ—(𝑣)for every 𝑒 βˆˆπ‘βˆ’(𝑣)and therefore π›Ώβˆ—(𝑣) =π›Ώβˆ’(𝑣), which contradicts the hypothesis of the lemma.
Now, let π‘—β€²βˆˆ{0, … , π‘Ÿ βˆ’2}and assume that for every π‘—βˆˆ{0, … , 𝑗′}there is a vertex 𝑒𝑗such that (𝑒𝑗, 𝑣) ∈𝐸and (π›Ώβˆ—(𝑒𝑗), 𝑒𝑗) ≻
(π›Ώβˆ’(𝑣) βˆ’π‘—, 𝑣). Suppose that for every vertex 𝑒 βˆˆπ‘βˆ’(𝑣) β§΅{𝑒1, … , 𝑒𝑗′}it holds (π›Ώβˆ—(𝑒), 𝑒) β‰Ί(π›Ώβˆ’(𝑣) βˆ’(𝑗′+1), 𝑣). The expression π›Ώβˆ’(𝑣) βˆ’
(𝑗′+1)is exactly the indegree of 𝑣after deleting the incoming edges from 𝑒0, … , 𝑒𝑗′, thus from the definition of the mechanism we
have that π‘–βˆ—(𝑒) >𝑖
βˆ—(𝑣)for every 𝑒 βˆˆπ‘βˆ’(𝑣) β§΅{𝑒0, … , 𝑒𝑗′}. This yields
π›Ώβˆ—(𝑣)β‰₯π›Ώβˆ’(𝑣)βˆ’(𝑗′+1)β‰₯π›Ώβˆ’(𝑣)βˆ’(π‘Ÿβˆ’1),
implying π‘Ÿ β‰₯π›Ώβˆ’(𝑣) βˆ’π›Ώβˆ—(𝑣) +1. In the case 𝑣 >𝑧, this is equivalent to π‘‘β‰€π›Ώβˆ—(𝑣), but these two inequalities contradict (𝑑, 𝑧) βͺ°(π›Ώβˆ—(𝑣), 𝑣).
If 𝑣 <𝑧, on the other hand, it is equivalent to π‘‘β‰€π›Ώβˆ—(𝑣) βˆ’1, which is again a contradiction. This concludes the existence of vertices
𝑒0, … , π‘’π‘Ÿβˆ’1 as in the statement of the lemma.
Games and Economic Behavior 144 (2024) 203–224
221
J. Cembrano, F. Fischer, D. Hannon et al.
Fig. 10. Proof of impossibility of the existence of an impartial randomized selection mechanism selecting vertices with indegree at least Ξ”(𝐺) βˆ’2 for each graph
πΊβˆˆξˆ³π‘›(1) and 𝑛 β‰₯10. Scaling and summing the inequalities leads to 0 β‰€βˆ’1, a contradiction.
To prove the last claim, we fix (𝑑, 𝑧) =(π›Ώβˆ—(𝑣), 𝑣). We know that taking π‘Ÿ =π›Ώβˆ’(𝑣) βˆ’π›Ώβˆ—(𝑣)there are vertices such that for each
π‘—βˆˆ{0, 1, … , π‘Ÿ βˆ’1}, (𝑒𝑗, 𝑣) ∈𝐸and (π›Ώβˆ—(𝑒𝑗), 𝑒𝑗) ≻(π›Ώβˆ’(𝑣) βˆ’π‘—, 𝑣). Suppose that there is a vertex 𝑒 βˆˆπ‘βˆ’(𝑣) β§΅{𝑒0, … , π‘’π‘Ÿβˆ’1}with (π›Ώβˆ—(𝑒), 𝑒) ≻
(π›Ώβˆ—(𝑣), 𝑣). It is clear that
|{π‘’βˆˆπ‘βˆ’(𝑣)∢(π›Ώβˆ—(𝑒),𝑒)≻(π›Ώβˆ—(𝑣),𝑣)}|β‰₯π‘Ÿ+1,
thus by expressions (1) and (2)we conclude that π›Ώβˆ’(𝑣) βˆ’π›Ώβˆ—(𝑣) β‰₯π‘Ÿ +1, which is a contradiction.
Appendix B. Proof of Lemma 2
Since 𝐸1β§΅({𝑒} ×𝑁) =𝐸2β§΅({𝑒} ×𝑁), it is clear that
π›Ώβˆ’(𝑣, 𝐺2)β‰₯π›Ώβˆ’(𝑣, 𝐺1)βˆ’1=𝑑+π‘Ÿ+(π›Ώβˆ’(𝑣, 𝐺1)βˆ’π›Ώβˆ—(𝑣, 𝐺1)) βˆ’ 1 βˆ’ πœ’(𝑣>𝑧).(B.1)
In the following, we denote π‘Ÿβ€²=π‘Ÿ +(π›Ώβˆ’(𝑣, 𝐺1) βˆ’π›Ώβˆ—(𝑣, 𝐺1)) for ease of notation. If π›Ώβˆ’(𝑣, 𝐺2) =π›Ώβˆ’(𝑣, 𝐺1), then we have the equation
π‘Ÿβ€²=π›Ώβˆ’(𝑣, 𝐺2) βˆ’π‘‘+πœ’(𝑣 >𝑧). In addition,
Games and Economic Behavior 144 (2024) 203–224
222
J. Cembrano, F. Fischer, D. Hannon et al.
Fig. B.11. Illustration of Lemma 2for the case π›Ώβˆ’(𝑣, 𝐺2) =π›Ώβˆ’(𝑣, 𝐺1) +1, shown on the left, and for the case π›Ώβˆ’(𝑣, 𝐺2) =π›Ώβˆ’(𝑣, 𝐺1), shown on the right. In contrast to
Fig. 2only the initial iteration 𝑖 =0is shown, and the overall drop in the indegree of 𝑣is illustrated by a dashed arrow. Observe that (π›Ώβˆ—(𝑒0, 𝐺2), 𝑒0) ≻(π›Ώβˆ—(𝑣, 𝐺1), 𝑣) ≻
(π›Ώβˆ—(𝑒0, 𝐺1), 𝑒0)and (π›Ώβˆ—(𝑒1, 𝐺2), 𝑒1) ≻(π›Ώβˆ—(𝑣, 𝐺1) βˆ’1, 𝑣), (π›Ώβˆ—(𝑒1, 𝐺1), 𝑒1) β‰Ί(π›Ώβˆ—(𝑣, 𝐺1), 𝑣).
(π›Ώβˆ’(𝑣, 𝐺2),𝑣)=(π›Ώβˆ’(𝑣, 𝐺1),𝑣)≻(𝑑,𝑧)βͺ°(π›Ώβˆ—(𝑣, 𝐺2),𝑣),
thus from Lemma 1there are vertices 𝑒′
0, … , 𝑒′
π‘Ÿβ€²βˆ’1 such that for every π‘—βˆˆ{0, … , π‘Ÿβ€²βˆ’1} we have that (𝑒′
𝑗, 𝑣) ∈𝐸1∩𝐸2β€”since
π›Ώβˆ’(𝑣, 𝐺2) =π›Ώβˆ’(𝑣, 𝐺1)we have 𝐸1βˆ©π‘βˆ’(𝑣) =𝐸2βˆ©π‘βˆ’(𝑣)β€”and (π›Ώβˆ—(𝑒′
𝑗, 𝐺2), 𝑒′
𝑗) ≻(π›Ώβˆ’(𝑣, 𝐺2) βˆ’π‘—, 𝑣). From expression (3), we know that
|{π‘’βˆˆ{𝑒′
0,…,𝑒
β€²
π‘Ÿβ€²βˆ’1}∢(π›Ώβˆ—(𝑒, 𝐺1),𝑒)≻(π›Ώβˆ—(𝑣, 𝐺1),𝑣)}|β‰€π›Ώβˆ’(𝑣, 𝐺1)βˆ’π›Ώβˆ—(𝑣, 𝐺1),
thus it is possible to take {𝑒0, … , π‘’π‘Ÿβˆ’1} βŠ†{𝑒′
0, … , 𝑒′
π‘Ÿβ€²βˆ’1}with the property that the inequality (π›Ώβˆ—(𝑒𝑗, 𝐺1), 𝑒𝑗) β‰Ί(π›Ώβˆ—(𝑣, 𝐺1), 𝑣)also holds
for π‘—βˆˆ{0, … , π‘Ÿ βˆ’1}. Moreover,
(π›Ώβˆ—(𝑒𝑗,𝐺
2),𝑒
𝑗)≻(π›Ώβˆ’(𝑣, 𝐺2)βˆ’π‘—βˆ’((π›Ώβˆ’(𝑣, 𝐺1)βˆ’π›Ώβˆ—(𝑣, 𝐺1))),𝑣)=(π›Ώβˆ—(𝑣, 𝐺1)βˆ’π‘—,𝑣).(B.2)
Relabeling these vertices so that (π›Ώβˆ—(𝑒0, 𝐺2), 𝑒0) ≻⋯ ≻(π›Ώβˆ—(π‘’π‘Ÿβˆ’1, 𝐺2), π‘’π‘Ÿβˆ’1)it is clear that the previous inequalities still hold, and in
particular
(π›Ώβˆ—(𝑒0,𝐺
2),𝑒
0)≻(π›Ώβˆ—(𝑣, 𝐺1),𝑣)≻(π›Ώβˆ—(𝑒0,𝐺
1),𝑒
0).
We conclude that Condition (ii) holds.
Similarly, if π›Ώβˆ’(𝑣, 𝐺2) =π›Ώβˆ’(𝑣, 𝐺1) +1, then π‘Ÿβ€²+1 =π›Ώβˆ’(𝑣, 𝐺2) βˆ’π‘‘+πœ’(𝑣 >𝑧). As before, (π›Ώβˆ’(𝑣, 𝐺2), 𝑣) ≻(𝑑, 𝑧) βͺ°(π›Ώβˆ—(𝑣, 𝐺2), 𝑣), thus
from Lemma 1there are vertices 𝑒′
0, … , 𝑒′
π‘Ÿβ€²such that for every π‘—βˆˆ{0, … , π‘Ÿβ€²}we have that (𝑒′
𝑗, 𝑣) ∈𝐸2and
(π›Ώβˆ—(𝑒′
𝑗,𝐺
2),𝑒
β€²
𝑗)≻(π›Ώβˆ’(𝑣, 𝐺2)βˆ’π‘—,𝑣).
But |(𝐸2βˆ©π‘βˆ’(𝑣)) β§΅(𝐸1βˆ©π‘βˆ’(𝑣))| =1, thus at least π‘Ÿβ€²of these vertices 𝑒′
𝑗are such that (𝑒′
𝑗, 𝑣) ∈𝐸1∩𝐸2. As before, from expression (3)
we conclude that it is possible to take {𝑒0, … , π‘’π‘Ÿβˆ’1} βŠ†{𝑒′
0, … , 𝑒′
π‘Ÿβ€²}such that (𝑒𝑗, 𝑣) ∈𝐸1∩𝐸2and (π›Ώβˆ—(𝑒𝑗, 𝐺1), 𝑒𝑗) β‰Ί(π›Ώβˆ—(𝑣, 𝐺1), 𝑣)for
every π‘—βˆˆ{0, … , π‘Ÿ βˆ’1}. Since expression (B.2)still holds, relabeling the vertices such that (π›Ώβˆ—(𝑒0, 𝐺2), 𝑒0) ≻⋯ ≻(π›Ώβˆ—(π‘’π‘Ÿβˆ’1, 𝐺2), π‘’π‘Ÿβˆ’1)
we conclude the lemma with Condition (ii) as well. An example for each of the cases addressed so far is included in Fig. B.11.
In what follows, we suppose that the first inequality in Condition (B.1)is an equality, so (ξ˜‘π‘£, 𝑣) ∈𝐸1⧡𝐸2and (𝐸2βˆ©π‘βˆ’(𝑣)) βŠ‚
(𝐸1βˆ©π‘βˆ’(𝑣)). In this case, π‘Ÿβ€²βˆ’1 =π›Ώβˆ’(𝑣, 𝐺2) βˆ’π‘‘+πœ’(𝑣 >𝑧). Suppose first that (π›Ώβˆ’(𝑣, 𝐺1) βˆ’1, 𝑣) ≻(𝑑, 𝑧), thus
(π›Ώβˆ’(𝑣, 𝐺2),𝑣)=(π›Ώβˆ’(𝑣, 𝐺1)βˆ’1,𝑣)≻(𝑑,𝑧)βͺ°(π›Ώβˆ—(𝑣, 𝐺2),𝑣).
In this case, from Lemma 1there are vertices 𝑒′
0, … , 𝑒′
π‘Ÿβ€²βˆ’2 such that for all π‘—βˆˆ{0, … , π‘Ÿβ€²βˆ’2} we have that (𝑒′
𝑗, 𝑣) ∈𝐸1∩𝐸2and
(π›Ώβˆ—(𝑒′
𝑗, 𝐺2), 𝑒′
𝑗) ≻(π›Ώβˆ’(𝑣, 𝐺2) βˆ’π‘—, 𝑣). If (π›Ώβˆ—(ξ˜‘π‘£, 𝐺1), ξ˜‘π‘£) ≻(π›Ώβˆ—(𝑣, 𝐺1), 𝑣), from expression (3)we have that
|{π‘’βˆˆ{𝑒′
0,…,𝑒
β€²
π‘Ÿβ€²βˆ’2}∢(π›Ώβˆ—(𝑒, 𝐺1),𝑒)≻(π›Ώβˆ—(𝑣, 𝐺1),𝑣)}|β‰€π›Ώβˆ’(𝑣, 𝐺1)βˆ’π›Ώβˆ—(𝑣, 𝐺1)βˆ’1.
Therefore, we conclude once again that it is possible to take {𝑒0, … , π‘’π‘Ÿβˆ’1} βŠ†{𝑒′
0, … , 𝑒′
π‘Ÿβ€²βˆ’2}such that (π›Ώβˆ—(𝑒𝑗, 𝐺1), 𝑒𝑗) β‰Ί(π›Ώβˆ—(𝑣, 𝐺1), 𝑣)
for every π‘—βˆˆ{0, … , π‘Ÿ βˆ’1}. We also have that
(π›Ώβˆ—(𝑒𝑗,𝐺
2),𝑒
𝑗)≻(π›Ώβˆ’(𝑣, 𝐺2)βˆ’π‘—βˆ’(π›Ώβˆ’(𝑣, 𝐺1)βˆ’π›Ώβˆ—(𝑣, 𝐺1)βˆ’1),𝑣)=(π›Ώβˆ—(𝑣, 𝐺1)βˆ’π‘—,𝑣),
so after relabeling the vertices with (π›Ώβˆ—(𝑒0, 𝐺2), 𝑒0) ≻⋯ ≻(π›Ώβˆ—(π‘’π‘Ÿβˆ’1, 𝐺2), π‘’π‘Ÿβˆ’1)Condition (ii) follows again. On the other hand, if
(π›Ώβˆ—(ξ˜‘π‘£, 𝐺1), ξ˜‘π‘£) β‰Ί(π›Ώβˆ—(𝑣, 𝐺1), 𝑣), from expression (3)
|{π‘’βˆˆ{𝑒′
0,…,𝑒
β€²
π‘Ÿβ€²βˆ’2}∢(π›Ώβˆ—(𝑒, 𝐺1),𝑒)≻(π›Ώβˆ—(𝑣, 𝐺1),𝑣)}|β‰€π›Ώβˆ’(𝑣, 𝐺1)βˆ’π›Ώβˆ—(𝑣, 𝐺1).
Thus, it is possible to take {𝑒1, … , π‘’π‘Ÿβˆ’1} βŠ†{𝑒′
0, … , 𝑒′
π‘Ÿβ€²βˆ’2}such that both (π›Ώβˆ—(𝑒𝑗, 𝐺2), 𝑒𝑗) ≻(π›Ώβˆ—(𝑣, 𝐺1) βˆ’π‘—, 𝑣)and (π›Ώβˆ—(𝑒𝑗, 𝐺1), 𝑒𝑗) β‰Ί
(π›Ώβˆ—(𝑣, 𝐺1), 𝑣)hold for every π‘—βˆˆ{1, … , π‘Ÿ βˆ’1}. In addition, whenever (π›Ώβˆ—(𝑣, 𝐺1), 𝑣) β‰Ί(π›Ώβˆ’(ξ˜‘π‘£, 𝐺1), ξ˜‘π‘£)we know from Lemma 1that,
taking ξ˜‘π‘Ÿβ€²=π›Ώβˆ’(ξ˜‘π‘£, 𝐺1) βˆ’π›Ώβˆ—(ξ˜‘π‘£, 𝐺1) +πœ’(ξ˜‘π‘£>𝑣), there exist vertices ξ˜‘π‘’0, … , ξ˜‘π‘’ξ˜‘π‘Ÿβ€²βˆ’1, different than ξ˜‘π‘£, such that for every π‘—βˆˆ{0, … , ξ˜‘π‘Ÿβ€²βˆ’1}
it holds (ξ˜‘π‘’π‘—, ξ˜‘π‘£) ∈𝐸and (π›Ώβˆ—(ξ˜‘π‘’π‘—, 𝐺1), ξ˜‘π‘’π‘—) ≻(π›Ώβˆ’(ξ˜‘π‘£, 𝐺1) βˆ’π‘—, ξ˜‘π‘£). Taking the first ξ˜‘π‘Ÿ =ξ˜‘π‘Ÿβ€²βˆ’(π›Ώβˆ—(𝑣, 𝐺1) βˆ’π›Ώβˆ—(ξ˜‘π‘£, 𝐺1)) such vertices and 𝑒0=ξ˜‘π‘£,
Condition (i) follows.
Finally, if π›Ώβˆ’(𝑣, 𝐺2) =π›Ώβˆ’(𝑣, 𝐺1) βˆ’1 and (π›Ώβˆ’(𝑣, 𝐺1) βˆ’1, 𝑣) β‰Ί(𝑑, 𝑧), we obtain from the inequality (𝑑, 𝑧) β‰Ί(π›Ώβˆ—(𝑣, 𝐺1), 𝑣)that
π›Ώβˆ—(𝑣, 𝐺1) =π›Ώβˆ’(𝑣, 𝐺1)and π‘Ÿ =1. Therefore, we must have (π›Ώβˆ—(ξ˜‘π‘£, 𝐺1), ξ˜‘π‘£) β‰Ί(π›Ώβˆ—(𝑣, 𝐺1), 𝑣), and the same analysis above leads to the
existence of vertices ξ˜‘π‘’0, … , ξ˜‘π‘’ξ˜‘π‘Ÿβˆ’1, different than ξ˜‘π‘£, such that for every π‘—βˆˆ{0, … , ξ˜‘π‘Ÿ βˆ’1} it holds (ξ˜‘π‘’π‘—, ξ˜‘π‘£) ∈𝐸and (π›Ώβˆ—(ξ˜‘π‘’π‘—, 𝐺1), ξ˜‘π‘’π‘—) ≻
(π›Ώβˆ’(ξ˜‘π‘£, 𝐺1) βˆ’π‘—, ξ˜‘π‘£). Condition (i) follows in this case as well. An example is shown in Fig. B.12.
Games and Economic Behavior 144 (2024) 203–224
223
J. Cembrano, F. Fischer, D. Hannon et al.
Fig. B.12. Illustration of Lemma 2for the case where π›Ώβˆ’(𝑣, 𝐺2) =π›Ώβˆ’(𝑣, 𝐺1) βˆ’1with (π›Ώβˆ’(𝑣, 𝐺1), 𝑣) β‰Ί(π›Ώβˆ’(ξ˜‘π‘£, 𝐺1), ξ˜‘π‘£). Only the initial iteration 𝑖 =0is shown, and the
overall drop in the indegree of ξ˜‘π‘£ is illustrated by a dashed arrow. Observe that (π›Ώβˆ—(ξ˜‘π‘’π‘—, 𝐺1), ξ˜‘π‘’π‘—) ≻(π›Ώβˆ’(ξ˜‘π‘£, 𝐺1) βˆ’π‘—, ξ˜‘π‘£)for π‘—βˆˆ{0, 1}.
Appendix C. Proof of Lemma 4
Let ξˆ³β€²
𝑛={πΊβˆˆξˆ³π‘‡
π‘›βˆΆ(𝑠(𝐺))𝑖=(𝑠(𝐺))π‘Ÿ(𝐺)+1βˆ’π‘–for every 𝑖 ∈{1, … , π‘Ÿ(𝐺)}} be the set of graphs πΊβˆˆξˆ³π‘‡
𝑛such that the tuple 𝑠(𝐺)is
symmetric. From the definition of ξˆ³π‘‡
𝑛, it is clear that for every graph 𝐺with a non-symmetric tuple 𝑠(𝐺), i.e., for every graph
πΊβˆˆξˆ³π‘‡
π‘›β§΅ξˆ³β€²
𝑛, there exists a unique π»βˆˆξˆ³π‘‡
𝑛⧡(ξˆ³β€²
𝑛βˆͺ{𝐺}) such that π‘Ÿ(𝐺) =π‘Ÿ(𝐻) =∢ π‘Ÿand (𝑠(𝐺))𝑖=(𝑠(𝐻))π‘Ÿ+1βˆ’π‘–for every 𝑖 ∈{1, … , π‘Ÿ}.
Moreover, πœ†πΊ=πœ†π»for such a pair of graphs. This implies that
βˆ‘
πΊβˆˆξˆ³π‘‡
π‘›β§΅ξˆ³β€²
𝑛
πœ†πΊ
is even. In what follows we show how to conclude the lemma using the following claim: for every πΊβˆˆξˆ³β€²
𝑛with π‘Ÿ(𝐺) β‰₯2, πœ†πΊis even.
Observe that
βˆ‘
πΊβˆˆξˆ³π‘‡
𝑛
πœ†πΊ=βˆ‘
πΊβˆˆξˆ³π‘‡
π‘›β§΅ξˆ³β€²
𝑛
πœ†πΊ+βˆ‘
πΊβˆˆξˆ³β€²
π‘›βˆΆπ‘Ÿ(𝐺)β‰₯2
πœ†πΊ+βˆ‘
πΊβˆˆξˆ³β€²
π‘›βˆΆπ‘Ÿ(𝐺)=1
πœ†πΊ.
If the claim is true, we have that the first two sums on the right-hand side are even. The third sum only contains one term, namely
πœ†πΊfor the complete graph 𝐺, for which 𝑠(𝐺) =(𝑛)and thus πœ†πΊ=𝑛!βˆ•π‘›! =1. Therefore,
βˆ‘πΊβˆˆξˆ³π‘‡
π‘›πœ†πΊis the sum of an even term plus 1,
and we conclude the result.
We now prove the claim. Let πΊβˆˆξˆ³β€²
𝑛with π‘Ÿ(𝐺) β‰₯2. The multiplicity of the prime factor 2 in the numerator of πœ†πΊ, due to Legendre’s
formula, is simply
∞
βˆ‘
𝓁=1 βŒŠπ‘›
2π“βŒ‹,
while its multiplicity in the denominator is
π‘Ÿ(𝐺)
βˆ‘
𝑖=1
∞
βˆ‘
𝓁=1 ⌊(𝑠(𝐺))𝑖
2π“βŒ‹.
Therefore, we need to prove that this last term is strictly lower than the former. It is easy to see that for every π“βˆˆβ„•, 𝓁β‰₯1we have
βˆ‘π‘Ÿ(𝐺)
𝑖=1 ⌊(𝑠(𝐺))𝑖
2π“βŒ‹β‰€βŒŠπ‘›
2π“βŒ‹, since without the floor functions we would have the equality, and the fractional parts of the terms (𝑠(𝐺))π‘–βˆ•2𝓁
must add up β€”when summing over 𝑖— to at least the fractional part of π‘›βˆ•2𝓁. We now show that there exists some 𝓁′for which this
inequality is strict, by distinguishing two cases. If (𝑠(𝐺))π‘–β‰€π‘›βˆ•2 for every 𝑖 ∈{1, … , π‘Ÿ(𝐺)}, we let π“β€²βˆˆβ„•such that 2𝓁′≀𝑛 <2𝓁′+1,
and then the following holds:
π‘Ÿ(𝐺)
βˆ‘
𝑖=1 ⌊(𝑠(𝐺))𝑖
2π“β€²βŒ‹β‰€
π‘Ÿ(𝐺)
βˆ‘
𝑖=1 βŒŠπ‘›βˆ•2
2π“β€²βŒ‹=
π‘Ÿ(𝐺)
βˆ‘
𝑖=1 βŒŠπ‘›
2𝓁′+1 βŒ‹=0<1=βŒŠπ‘›
2π“β€²βŒ‹.
On the other hand, if (𝑠(𝐺))𝑖′>π‘›βˆ•2 for some π‘–β€²βˆˆ{1, … , π‘Ÿ(𝐺)}, from the symmetry of 𝑠(𝐺)we have that π‘Ÿ(𝐺)is odd and, moreover,
𝑖′=(π‘Ÿ(𝐺) +1)βˆ•2. We now define π“β€²βˆˆβ„•such that 2𝓁′≀𝑛 βˆ’(𝑠(𝐺))𝑖′<2𝓁′+1, which implies (𝑠(𝐺))π‘–β‰€π‘›βˆ’(𝑠(𝐺))𝑖′
2<2𝓁′for every 𝑖 ≠𝑖′.
Therefore, the following holds:
π‘Ÿ(𝐺)
βˆ‘
𝑖=1 ⌊(𝑠(𝐺))𝑖
2π“β€²βŒ‹=⌊(𝑠(𝐺))𝑖′
2π“β€²βŒ‹β‰€βŒŠπ‘›βˆ’2
𝓁′
2π“β€²βŒ‹=βŒŠπ‘›
2π“β€²βˆ’1
βŒ‹<βŒŠπ‘›
2π“β€²βŒ‹.
In either case, we obtain that
π‘Ÿ(𝐺)
βˆ‘
𝑖=1
∞
βˆ‘
𝓁=1 ⌊(𝑠(𝐺))𝑖
2π“βŒ‹<
∞
βˆ‘
𝓁=1 βŒŠπ‘›
2π“βŒ‹,
and thus πœ†πΊis even, which concludes the proof of the claim and the proof of the lemma.
Games and Economic Behavior 144 (2024) 203–224
224
J. Cembrano, F. Fischer, D. Hannon et al.
Appendix D. Proof of Lemma 5
To see that 𝑓sis impartial, let 𝐺=(𝑁, 𝐸), 𝐺′=(𝑁, 𝐸′) βˆˆξˆ³π‘›and 𝑣 βˆˆπ‘such that 𝐸⧡({𝑣} ×𝑁) =𝐸′⧡({𝑣} ×𝑁). Since 𝑓is
impartial,
(𝑓s(𝐺))𝑣=1
𝑛!βˆ‘
πœ‹βˆˆξˆΏπ‘›
(𝑓(πΊπœ‹))πœ‹π‘£=1
𝑛!βˆ‘
πœ‹βˆˆξˆΏπ‘›
(𝑓(𝐺′
πœ‹))πœ‹π‘£=(𝑓s(𝐺′))𝑣,
and thus 𝑓sis impartial.
To prove that 𝑓ssatisfies weak unanimity, let 𝐺=(𝑁, 𝐸) βˆˆξˆ³π‘›and 𝑣 βˆˆπ‘with π›Ώβˆ’(𝑣) =𝑛 βˆ’1. Since 𝑓satisfies weak unanimity,
βˆ‘
π‘’βˆˆπ‘βˆΆπ›Ώβˆ’(𝑒)β‰₯1
(𝑓s(𝐺))𝑒=βˆ‘
π‘’βˆˆπ‘βˆΆπ›Ώβˆ’(𝑒)β‰₯1
1
𝑛!βˆ‘
πœ‹βˆˆξˆΏπ‘›
(𝑓(πΊπœ‹))πœ‹π‘’=1
𝑛!βˆ‘
πœ‹βˆˆξˆΏπ‘›βˆ‘
π‘’βˆˆπ‘βˆΆπ›Ώβˆ’(𝑒)β‰₯1
(𝑓(πΊπœ‹))πœ‹π‘’
β‰₯1
𝑛!βˆ‘
πœ‹βˆˆξˆΏπ‘›
1=1.
This concludes the proof of the lemma.
References
Alon, N., Fischer, F., Procaccia, A., Tennenholtz, M., 2011. Sum of us: strategyproof selection from the selectors. In: Proceedings of the 13th Conference on Theoretical
Aspects of Rationality and Knowledge, pp. 101–110.
Aziz, H., Lev, O., Mattei, N., Rosenschein, J.S., Walsh, T., 2019. Strategyproof peer selection using randomization, partitioning, and apportionment. Artif. Intell. 275,
295–309.
Babichenko, Y., Dean, O., Tennenholtz, M., 2020. Incentive-compatible selection mechanisms for forests. In: Proceedings of the 21st ACM Conference on Economics
and Computation, pp. 111–131.
Bjelde, A., Fischer, F., Klimm, M., 2017. Impartial selection and the power of up to two choices. ACM Trans. Econ. Comput. 5, 1–20.
Bousquet, N., Norin, S., Vetta, A., 2014. A near-optimal mechanism for impartial selection. In: Proceedings of the 10th International Conference on Web and Internet
Economics. Springer, pp. 133–146.
Caragiannis, I., Christodoulou, G., Protopapas, N., 2022. Impartial selection with additive approximation guarantees. Theory Comput. Syst. 66, 721–742.
Caragiannis, I., Christodoulou, G., Protopapas, N., 2023. Impartial selection with prior information. In: Proceedings of the ACM Web Conference, pp. 3614–3624.
De Clippel, G., Moulin, H., Tideman, N., 2008. Impartial division of a dollar. J. Econ. Theory 139, 176–191.
Diagana, T., MaΓ―ga, H., 2017. Some new identities and congruences for Fubini numbers. J. Number Theory 173, 547–569.
Fischer, F., Klimm, M., 2015. Optimal impartial selection. SIAM J. Comput. 44, 1263–1285.
Holzman, R., Moulin, H., 2013. Impartial nominations for a prize. Econometrica 81, 173–196.
Insko, E., Johnson, K., Sullivan, S., 2017. An ordered partition expansion of the determinant. Integers 17, A36.
Kahng, A., Kotturi, Y., Kulkarni, C., Kurokawa, D., Procaccia, A.D., 2018. Ranking wily people who rank each other. In: Proceedings of the 32nd AAAI Conference on
Artificial Intelligence.
Kurokawa, D., Lev, O., Morgenstern, J., Procaccia, A.D., 2015. Impartial peer review. In: Proceedings of the 24th International Joint Conference on Artificial Intelli-
gence.
Mackenzie, A., 2015. Symmetry and impartial lotteries. Games Econ. Behav. 94, 15–28.
Mackenzie, A., 2020. An axiomatic analysis of the papal conclave. Econ. Theory 69, 713–743.
Mattei, N., Turrini, P., Zhydkov PeerNomination, S., 2021. Relaxing exactness for increased accuracy in peer selection. In: Proceedings of the 29th International Joint
Conferences on Artificial Intelligence, pp. 393–399.
Niemeyer, A., Preusser, J., 2023. Simple allocation with correlated types. Working Paper.
Olckers, M., Walsh, T., 2022. Manipulation and peer mechanisms: a survey. arXiv preprint. arXiv :2210 .01984.
Tamura, S., Ohseto, S., 2014. Impartial nomination correspondences. Soc. Choice Welf. 43, 47–54.
Xu, Y., Zhao, H., Shi, X., Shah, N.B., 2019. On strategyproof conference peer review. In: Proceedings of the 28th International Joint Conference on Artificial Intelligence,
pp. 616–622.
Zhang, X., Zhang, Y., Zhao, D., 2021. Incentive compatible mechanism for influential agent selection. In: Proceedings of the 14th International Symposium on
Algorithmic Game Theory. Springer, pp. 79–93.