WSB University in Wroclaw
Research Journal
ISSN 1643-7772 I eISSN 2392-1153
Vol. 16 I No. 3
Zeszyty Naukowe Wyższej Szkoły
Bankowej we Wrocławiu
ISSN 1643-7772 I eISSN 2392-1153
R. 16 I Nr 3
On Axiomatization of Plurality Decisions with Veto
Author: Jacek Mercik
Abstract
The article presents an analysis of the axioms associated with the plurality
method of aggregation of individual preferences, both when it is necessary
to select one of many alternatives and when it is necessary to approve a single alternative. Also, we investigate the impact of the introduction of a new
attribute, being the right of veto (absolute and relative), on the axioms given.
In the conclusion, the emphasis is that the commonly used method of aggregation, i.e. the plurality method is not, in this sense, the best method, .
Keywords: aggregation of preferences, axioms, power indices.
JEL: C71, D71, D72.
History: Otrzymano 2016-03-21, poprawiono 2016-03-29, zaakceptowano
2016-06-01
Introduction
or qualified). This rule implies that if
an alternative should be chosen from
a set of alternatives, then the alternative for which the majority of decision-makers opted is chosen. Every
specific situation requires that the
notion of “majority” be defined and
a method for settlement provided
when a majority cannot be reached,
yet, a decision must be taken. An additional and frequently deployed decision scheme is to endow some of decision-makers with the so called right
of veto. This right means that one or
more decision-makers can defer the
choice of a particular alternative absolutely (that is forever) or relatively,
where the decision to use veto can
be overruled. Obviously, adding veto
Making group decisions1 occurs in
multiple decision-making situations.
It occurs when decisions are made by
people (e.g. voters, parliamentarians
or supervisory board members), as
well as when they are made by software-embedded machines (e.g. multiple agents systems, image analysis
techniques or decision-support systems). The decision-making rule that
is applied most frequently in those
situations is the majority rule (simple
1
In various fields such decisions are variously named. E.g. in social choice theory
one refers to social preference or social
decision [Lisowski, 2008].
Jacek Mercik
Wyższa Szkoła Bankowa we Wrocławiu
jacek.mercik@wsb.wroclaw.pl
WSB University in Wroclaw Research Journal I ISSN 1643-7772 I eISSN 2392-1153 I Vol. 16 I No. 3
Plurality aggregation with veto occurs
when at least one decision-maker has
the right to veto the choice made.
In the paper, we examine decisions
where the decision-makers have
to choose one alternative from a set
of alternatives 𝑊={𝑤1,𝑤2,…,𝑤𝑟},
𝑟=1,2,…,𝑛. Analyzing group decisions, we will investigate, due to the
amount of alternatives, two classes of
decisions, K1 and K2. The first class K1
is made up of the decisions in which
an 𝑁 number decision-makers participate (|𝑁|=𝑛≥2), having to choose
one alternative from the set of alternatives 𝑊, |𝑊|=𝑚≥2. The second
class, K2, is made up of the decisions
in which an 𝑁 number decision-makers take part (|𝑁|=𝑛≥2) who have
to accept a single alternative (|𝑊|=1).
The investigation of group aggregation
methods should be carried out in the
following ways:
1. By characterizing individual aggregation methods, describing the
conditions in which they provide
a clear-cut result, etc.
2. By specifying a set of criteria for
which no aggregation method
exists capable of providing clearcut results and simultaneously fulfilling all those criteria, and
3. By examining aggregation methods
in terms of their satisfying or failing to satisfy a particular criterion
(axiom).
With respect to the decision made
in class K1, we assume that every
changes the importance of some or
all of decision-makers2. In the paper,
we intend to analyze a set of axioms
describing such decision-making situations and outline how adding the right
to veto affects the selected axioms.
The structure of the paper is as follows: the first chapter presents the
idea behind group decision with veto.
Depending on the amount of alternatives from which the choice is to be
made, subsequent chapters analyze
the set of postulates (axioms) associated with the plurality aggregation.
The second chapter includes a discussion on axioms associated with the
plurality choice of one out of more
than two alternatives. The components of simple plurality voting games
theory (voting theory) are introduced
in the third chapter. Moreover, a minimum set of plurality decision axioms
together with the relevant theorems
on non-existence or existence of aggregation functions are also discussed.
The paper ends with a conclusion and
proposals for further research.
Initial Presumptions of Group
Decisions with Veto
For the analysis of group aggregation methods, the so called individual profiles provide a starting point.
They state that every decision-maker is capable of ordering (partially or
completely) a set of alternatives using
their own function of preferences.
The plurality method of aggregation
is to be understood as a way of aggregating individual profiles, where every
decision-maker indicates their best
alternative, and the alternative that
is going to be chosen is the one which
gets the highest amount of votes3 .
2 The example of real absolute right of
veto can be found, among others, in the
work of Mercik (2009), while the relative
veto in Ramsey, Mercik (2015).
3 In practice the majority rule is often
modified, usually by adding an additional
182
requirement with respect to the number
of votes which needs to be fulfilled so
that a particular choice could be accepted. Amongst the classical plurality methods we can distinguish, e.g. plurality with
run off method, where 50% of votes is
reached by (if it is necessary) eliminating
sequentially alternatives which receive
the lowest amount of support; Hare method, where individual profiles do not get
changed while applying plurality with run
off method, or Coombs’ method, which is
in fact the reverse of Hare method.
Jacek Mercik | On Axiomatization of Plurality Decisions with Veto
decision-maker is capable of ordering
a set of alternatives (in the best case
by individual preference relation:
either weak or strong). In the worst
case the ordering unfolds by comparing pairs with the admissible partial
order and equivalent alternatives4.
With respect to class K2, we assume
that every decision-maker can make
a decision whether or not to accept a particular alternative5. All decision-makers present their views
on a particular alternative (e.g. by
voting) and if it receives the required
amount of support (votes), then it is
accepted by the group. For this purpose we introduce the following notation derived from simple games.
Let 𝑁 represent a finite set of decision-makers, 𝒒 the required number
of votes (quota) to make a decision
and 𝑤𝑗 denotes the weight of decision-maker 𝑗, 𝑗∈𝑁. Let us consider
a special class of cooperative games
called weighted majority games.
Weighted majority game G is defined
by 𝒒 and a sequence of non-negative weights 𝜔𝑖,𝑖∈𝑁. We can think of
𝜔𝑖 as of number of votes of an 𝑖-th
decision-maker, and of 𝒒 as number
of votes necessary for an alternative
to be collectively accepted, that is, necessary for the group of decision-makers to become a winning group (coalition). In the interest of simplicity, we
assume that both 𝒒 and 𝜔𝑖 are positive
integers. Any subset of decision-makers is named a coalition.
The decision acceptance is thus
equivalent to the formation of
a winning coalition of decision-makers. Simple game (𝑁, 𝑣), where
𝑁 is a set of decision-makers
4 Interesting findings on ordering equivalent objects with partial orders can be
found, e.g. in works of Bury, Wagner
(2008).
5 In simple game theory, this corresponds
to yes or no voting
(players) and 𝑣 is a characteristic
function of the game6, is only then
a proper game when the following
condition is fulfilled: for all coalitions
𝑇 ⊏ 𝑁, if 𝑣(𝑇)=1, then 𝑣(𝑁\𝑇)=0.
This means that decisions are made
by the majority of at least 𝒒 votes
((0,5<𝑞≤1). Let us consider only
simple games (SG) where a decision-maker (player) may vote yes-no,
or yes-no-abstain. The last one is an
analogue of a weak preference where
there is a greater number of alternatives. Alternatively, we can describe
the decision-making body as a triplet
(𝑁,𝑞,𝑣)=(𝑁,𝑞,𝜔1,𝜔2,…,𝜔𝑛).
If a particular member of the decision-making body can transform the
winning coalition into the non-winning one by utilizing veto, then this
type of veto is called veto of the first
degree. If a particular member of
the decision-making body can transform only some of the winning coalitions, that is, not all of them, into the
non-winning ones, even without being
their member, then this type of veto is
called veto of the second degree (Mercik, 2011).
Let us note that the behaviour of
the decision-makers in the winning coalition can provide a basis
for the evaluation of their impact
on the final outcome measured by
the so called power index. From
a formal point of view, the representation 𝜑:𝑆𝐺→𝑅𝑛 is called power index. For each 𝑖∈𝑁 and 𝑣∈𝑆𝐺, the
𝑖-th coordinate 𝜑(𝑣)𝜖𝑅𝑛,𝜑(𝑣)(𝑖) is interpreted as the voting power of player 𝑖 in game 𝑣. In general, power indexes are a priori in their nature.
Potential situations are analysed in
which individual decision-makers
(players) change their mind (e.g. from
voting yes to voting no) and how this
change affects the position of a given
6
183
This means that if Σ𝑘∈𝑇⊑𝑁𝜔𝑘≥𝑞, then
𝑣(𝑇)=1
WSB University in Wroclaw Research Journal I ISSN 1643-7772 I eISSN 2392-1153 I Vol. 16 I No. 3
to all sets of postulates concerning
the behaviour of decision-makers.
Possible assertions that there exists
or does not exist an optimal group aggregation method (including the plurality method) always depend on the
selected set of axioms. This also applies to potential measurement of the
power of the decision-maker in the
plurality method.
player. If a particular member of
a winning coalition transforms the
winning coalition into a non-winning
one by changing the vote, then he/
she is in the position called swing position (which leads to Penrose-Banzhaf
power index (1965), with the coalition
being “sensitive” to the behaviour of
a particular decision-maker, which, in
turn, leads to Johnston power index
(1978). If, however, the decision-maker
changes his/her mind and joins a particular coalition and thus transforms it
into a winning coalition, then he/she is
a pivotal player, which leads to Shapley-Shubik power index (1954)7.
Penros-Banzhaf
power
index (1965) for a simple game
is a value equal to: 𝜑:𝑆𝐺→𝑅𝑛,
𝑣→(φ1(v),φ2(v),…,φn(v)), where for
every 𝑖ε𝑁; 𝑐𝑎𝑟𝑑 {𝑁}=𝑛;𝑐𝑎𝑟𝑑{𝑆}=𝑠
Σ
Axiomatization of
Decisions in Class K1
Let us recall that the first class, K1, is
made up of decisions in which an 𝑁
number of decision-makers participate
(|𝑁|=𝑛≥2), having to choose one alternative from the set of alternatives
𝑊, |𝑊|=𝑚≥2.
The fundamental assumption of the
theory of collective decisions is the
rationality of the decision. At the level
of individual decision-making, this
implies, in crude terms, that if a particular decision-maker considers the
choice of a given alternative to be the
best (assuming that there are more alternatives than just one), then he/she
will chose this alternative.
If, however, the decision-maker cannot indicate a single alternative, then
it means that there are at least two
alternatives which he/she considers
to be the best, according to his/her
own criteria. If the choice of a single
alternative is still necessary, then we
agree on some way of settling such
“tie” situations, e.g. by drawing lots or
resorting to some other criterion (e.g.
alphabetical order of the names of the
alternatives). However, there should
be no doubt that such decisions are
rational in the sense that a better alternative (or at least not worse than
the remaining ones) gets always
chosen.
We thus assume, from a formal viewpoint, that, between a pair of alternatives a and b coming from a given
set of alternatives 𝑊, one of three
[𝑣(𝑆∪{𝑖})−𝑣(𝑆)]
𝑆⊆𝑁{𝑖}
Moreover,
Shapley-Shubik
power index (1954) represents
the following value: 𝜑:𝑆𝐺→𝑅𝑛,
𝑣→(φ1(v),φ2(v),…,φn(v)), where
{𝑁}=𝑛;
for every 𝑖ε𝑁, 𝑐𝑎𝑟𝑑
𝑐𝑎𝑟𝑑{𝑆}=𝑠
𝑆⊆𝑁{𝑖}
The above values are obtained directly
from the games defined with the use
of the characteristic function, where
marginal values of the power increase
are calculated for every winning
coalition.
We will demonstrate that deploying
the plurality aggregation method and
including veto results in the lack of an
unequivocal set of axioms pertaining
7
It is worth noting that the aforementioned power indices (and a variety of
other indices) are not equivalent and
their application depends on the context
of a decision-making situation.
184
Jacek Mercik | On Axiomatization of Plurality Decisions with Veto
shows that the group relation is not always transitive even if individual preferences are transitive.
It is expected (e.g. Mercik, 1998, Lissowski, 2008) that the employed
method of the individual preferences
aggregation should be immune to manipulations (individual11 and agenda
manipulations12), be effective computationally (i.e. it should lead to choosing one alternative) and it should satisfy the so called Arrow’s axioms, that is:
(1) it should be defined as a set of all
possible individual voter preferences
(postulate of unrestricted domain),
(2) it should fulfil the requirement of
Pareto optimality (at least in its weak
version)13, should be independent of
irrelevant alternatives14, and (4) there
should be no decision-maker who is
a dictator.
Arrow (1951, the so called impossibility theorem) showed that if there
are at least two decision-makers and
at least three alternatives to be voted
on, then there is no method of group
aggregation which would satisfy
the aforementioned criteria. Similar
to that, Gärdenfors (1976) shows that
there exists no aggregation technique
that could satisfy both Condorcet’s
possible preference relations occurs
(≻):𝑎≺𝑏, 𝑏≺𝑎 or 𝑎∼𝑏 (strong alternative) or 𝑎≼𝑏, 𝑏≼𝑎 or 𝑎∼𝑏 (weak
alternative). We usually assume with
respect to this kind of preference that
it is reflexive, transitive and complete.8
The rationality of group decisions, i.e.
decisions made by more than one decision-maker is no longer so obvious,
since what seems rational at a level of
individual decision-making need not
be so at the level of a group. The aggregation of individual decisions can lead
to the choice of an alternative which
is not the best in terms of the criteria used. Although Condorcet (1785)
already introduced the two main criteria9 which the group decision-making should fulfil, what soon followed is
that the majority of decision-making
methods do not satisfy those criteria,
resulting in what is known in literature
as a voting paradox.
Let us then take a closer look at the
axioms characterizing the rationality of the aggregation of group decisions. It is assumed (just like for the
preferences describing individual decision-making) that the group preference relation is rational if it satisfies
the requirements of the reflexive,
complete and transitive relation. Unfortunately, the Condorcet’s paradox10
8
That this assumption is rather idealistic
has been demonstrated in the research
on human behaviour, where the majority
of created orders do not retain transitivity (e.g. Aumann, 1985 or Gilboa et al.,
2012).
9 Two main criteria of Condorcet are: (1)
if a particular alternative wins in a pairwise comparison with the remaining alternatives, then it is the winning alternative,
according to a particular method of group
voting, and (2) if a particular alternative
loses in a pairwise comparison against
other alternatives, then it cannot be the
winning alternative of a given group voting method.
10 An example of the paradox which
leads to a tie is the situation where three
decision-makers A, B, and C have the
185
following preferences regarding three alternatives , and ,respectively.
11 Group aggregation method is individually manipulable if the decision-maker
changes his/her preferences purposefully
and as a result achieves the desired outcome of the aggregation.
12 The aggregation method is agenda manipulated if, e.g. the inclusion of a new
alternative or the change of a given voting
order changes the result of the aggregation.
13 If the decision-makers are in full agreement which alternative is the best and
which is the worst, then the particular aggregation method should only refer to the
remaining alternative (if we want a complete ordering (also the weak one) of all
alternatives).
14 Irrelevant alternatives are alternatives
outside the alternatives whose ordering
we are considering.
WSB University in Wroclaw Research Journal I ISSN 1643-7772 I eISSN 2392-1153 I Vol. 16 I No. 3
we can see, none of the aggregation
methods satisfy each axiom.
In this case introducing veto is equivalent to establishing a dictator whose
decisions are irrevocable (veto of
the first type) or conditional (veto of
the second type). In none of the sets
of axioms proposed for specifying
the theorems about the existence or
non-existence of group aggregation
method the dictator is admissible
(veto of the first type). The conditional veto (veto of the second type), in
the plurality method (where only the
alternatives ranked first in the individual orderings by the decision-makers
are of importance), implies raising
the limit on the amount of necessary
first rankings for a particular alternative and almost inevitably means that
a plurality-at-large method will have
to be applied (Ramsey, Mercik, 2015).
This, however, does not affect the
fulfilment of specific axioms by this
method.
criteria for weak preferences, and Gibbard (1973) and Satterthwait (1975)
demonstrated that every rational, in
Arrow’s sense, aggregation method is
individually manipulable.
As the subject of discussion in the paper
is the aggregation method called the
plurality method, we must at once see
that this is not the method which could
satisfying Arrow’s postulates. May
(1952) named necessary and sufficient
conditions (unanimity15, duality16,
and strong monotonicity17) whose
fulfilment allows one to chose one alternative in unanimous way by applying the plurality method.
In Mercik’s work (1990) we may see
that the plurality method of aggregation satisfies the following axioms:
• Condorcet’s axiom I on choosing
the wining alternative in a pairwise
comparison,
• monotonicity axiom,
• consensus axiom18,
• axiom of simplicity and easy application, and
• Pareto axiom
This represents 6 out of 9 of the
axioms under discussion. “The best”
of the aggregation methods (if all the
axioms are treated equally) satisfies
7 axioms (approval method19). As
15 If all decision-makers regard a particular alternative as the best (worst), then a given aggregation method also retains it.
16 If we assume that a given alternative
which is individually the best becomes
individually the worst (reversal of preferences), then as a result of aggregation
this alternative will be the worst with the
previous individual preferences.
17 If two alternatives have been ordered
by all decision-makers individually, then
the aggregation method retains this order.
18 This axiom states that if in two separate groups of decision-makers a given
alternative is the best in each of the group,
then after combining these groups this alternative continues to be the best.
19 The missing axioms of the approval method are represented by Condorcet’s criteria, which, to some extent,
Axiomatization of
Decisions in Class K2
Let us recall, the second class, K2, is
made up of the decisions in which an
𝑁 number of decision-makers take
part (|𝑁|=𝑛≥2), having to accept
a single alternative.
For n=2 the application of the plurality method means that it is necessary
for alternatives to be accepted by both
decision-makers simultaneously. The
right of veto does not have any impact
on the outcome of the aggregation.
Significant changes occur when we
raise the number of decision-makers
(𝑛>2), which is what we will discuss
presently.
We are focusing on the question of accepting an alternative based on group
186
disqualifies the very method. However,
the incidence of paradoxes in this method of aggregation is so low (Nurmi, Uusi-Heikkilä, 1985) that it can practically
be disregarded.
Jacek Mercik | On Axiomatization of Plurality Decisions with Veto
The fourth axiom is called a transfer
axiom:
decision. In this case, it is of significance to try and specify the possibilities to exert influence on such a decision, so that it would be in accordance
with a particular decision-maker’s
preference. Naturally, it is not possible to examine how a specific decision-maker will behave (apart from
a post factum analysis), yet, we could
employ here an a priori analysis, assuming that every decision-maker
knows whether they are for or against
a particular alternative.
Let us investigate what axioms are
linked to an a priori power index, regardless of its form. As we have already mentioned, a priori power of
a given decision-maker can be measured using the so called power indices. The following representation
we call power index: 𝜑:𝑆𝐺→𝑅𝑛.
For every 𝑖∈𝑁 and 𝑣∈𝑆𝐺, the 𝑖-th
coordinate 𝜑(𝑣)∈𝑅𝑛, 𝜑(𝑣)(𝑖), is interpreted as the decision-maker’s
power i in the game with a characteristic function 𝑣. One may expect
that the decision-maker endowed
with the right to veto 𝑖𝑣𝑒𝑡𝑜 should have
a priori power at least not smaller
than without this right. This leads
to the first axiom called a value-added axiom: 𝜑(𝑣)(𝑖𝑣𝑒𝑡𝑜)≥𝜑(𝑣)(𝑖). For
the veto of the first type we get:
𝜑(𝑣)(𝑖𝑣𝑒𝑡𝑜)>𝜑(𝑣)(𝑖) 20. Of course,
for certain situations with the
veto of the second kind we get:
𝜑(𝑣)(𝑖𝑣𝑒𝑡𝑜)=𝜑(𝑣)(𝑖).
The second axiom is the so called
gain-loss axiom: if 𝜑(𝑣)(𝑖)>𝜑(𝑤)(𝑖)
for 𝑣,𝑤∈𝑆𝐺 and 𝑖∈𝑁 there exists
𝑗∈𝑁 such that 𝜑(𝑣)(𝑗)<𝜑(𝑤)(𝑗). Let
us notice that if 𝜑(𝑣)(𝑖𝑣𝑒𝑡𝑜)>𝜑(𝑤)(𝑖)
𝜑(𝑣∨𝑤)(𝑖)+𝜑(𝑣∧𝑤)(𝑖)=
=𝜑(𝑣)(𝑖)+𝜑(𝑤)(𝑖),
for 𝑣,𝑤 ∈𝑆𝐺 and it does not change its
properties for games with or without
veto. Its equivalent can be expressed
(Dubey et al. 1981) as follows: consider
two pairs of games 𝑣, 𝑣’ and 𝑤, 𝑤’ in
SG, each game with veto. Let us assume
that the transition from 𝑣’ to 𝑣 and 𝑤’
to 𝑤 denotes the same winning coalitions,i.e.≥𝑣′,𝑤≥𝑤′ oraz 𝑣−𝑣′=𝑤−𝑤′.
Hence, the equivalent transfer axiom
is as follows:
𝜑(𝑣)(𝑖)−𝜑(𝑣′)(𝑖)=
=𝜑(𝑤)(𝑖)−𝜑(𝑤′)(𝑖). This implies
that the change in power depends
only on the change in the game itself
The fifth axiom is called symmetry
axiom and denotes that 𝜑(𝜋𝑣)(𝑖)=
=𝜑(𝑣)(𝜋(𝑖)) for every permutation
of players (decision-makers) with or
without veto. What is more, changing
the way decision-makers are ordered
has no effect on the value of their
power index.
Once again there is an axiom which
is equivalent to the symmetry axiom,
and that is the equal treatment axiom
(the sixth axiom): if 𝑖,𝑗∈𝑁 are players
in game 𝑣∈𝑆𝐺 with veto, then:
𝑆⊂𝑁\{𝑖,𝑗} 𝑣(𝑆∪{𝑖})=𝑣(𝑆∪{𝑗}),
tℎen 𝜑(𝑣)(𝑖)=𝜑(𝑣)(𝑗),
being also valid for 𝑖𝑣𝑒𝑡𝑜 and 𝑗𝑣𝑒𝑡𝑜 ∈𝑁
for 𝑣,𝑤∈𝑆𝐺 and 𝑖𝑣𝑒𝑡𝑜∈𝑁 there exists
𝑗∈𝑁 such that 𝜑(𝑣)(𝑗)<𝜑(𝑤)(𝑗).
The third axiom refers to the normalization of the power index value:
Σ
𝜑(𝑣)(𝑖)=1
𝑖∈𝑁
20
Let
us
note
that
value
𝜑(𝑣)(𝑖𝑣𝑒𝑡𝑜)−𝜑(𝑣)(𝑖) reflects well the net
value of veto
187
All the above axioms are valid for simple majority voting games with veto.
However, things look different with
the group of further axioms which,
although being valid for simple majority voting games without veto, are no
longer satisfied when veto has been
added to them (Mercik, 2015).
The sixth axiom refers to the so called
null player: if 𝑖∈𝑁 and 𝑖 is a null
player in 𝑣, i.e. 𝑣(𝑆∪{𝑖})=𝑣(𝑆) for
every 𝑆⊂𝑁\{𝑖}, tℎen 𝜑(𝑣)(𝑖)=0.
It can be demonstrated that since
𝜑(𝑣)(𝑖𝑣𝑒𝑡𝑜)>0,than the “null” player
axiom can be violated.
WSB University in Wroclaw Research Journal I ISSN 1643-7772 I eISSN 2392-1153 I Vol. 16 I No. 3
acceptance of outcomes obtained
on the basis of pair comparisons.
4. Endowing the decision-maker (or
decision-makers) with the right
of veto leads to further violation
of the rules accepted as a rational
behaviour of decision-makers.
5. The application of veto of the
second type, that is, the one
which can be overruled, is equivalent to raising the threshold in
the plurality decision-making,
however, in principle, it does not
change the characteristics of the
plurality method itself.
It should be assumed that this situation concerning the satisfaction of
the axioms of rational decisions with
respect to the plurality method will
not change. Nor does it look as if
some other new power indices were
to emerge to measure the influence
a particular decision-maker exerts
on the plurality method concerning
the acceptance (or the lack of it) of
a single alternative. Hence, a rational
solution to the aggregation issue of
individual preferences should be rejecting the plurality method (including
the plurality method with veto) and
to explore other and, in this sense,
better and more effective methods of
aggregation. From the point of view of
axioms describing group decisions and
power indies, the approval method
appears to be a candidate for the aggregation method.
The seventh axiom (dummy) seemingly refers to the same as the “null” player axiom: if 𝑣∈𝑆𝐺 and 𝑖 is a dummy
in game 𝑣, 𝑣(𝑆∪{𝑖})=𝑣(𝑆)+𝑣(𝑖)
for every coalition 𝑆⊂𝑁\{𝑖}, then
𝜑(𝑣)(𝑖)=𝑣({𝑖}). Here, too, applying
veto violates this axiom.
The eighth axiom is concerned with
the local monotonicity (or at least the
local one): if the weight of the 𝑖-th
player is greater than that of the 𝑗-th
player, then the power index of the
𝑖-th player cannot be smaller than the
power index of the 𝑗-th player. Neither
this axiom is satisfied in the case of
simple plurality games with veto.
Summary
The analysis of decision-making situations involving the aggregation of preferences of individual decision-makers
by the plurality method leads to the
following conclusions as regards the
axioms involved in it:
1. There is no common set of axioms
which could be useful in all decision-making situations with plurality aggregation.
2. Individual rationality of decision-makers does not guarantee
that there will be group rationality
in the decision-making.
3. The plurality method, although
simple to use and commonly applied, violates, in many instances,
the basic axiom concerning the
Bibliography
Arrow K. J. (1951): Social Choice and
Individual Values, Wiley, New York.
analysis, Rutgers Law Review 19,
317-343.
Aumann R. J. (1985): On the
Non-Transferable Utility Value: A Comment on the Roth-Shafer Examples,
Econometrica, Vol. 53, No.3, 667-678
Bury H., Wagner D. (2008): Pozycyjne
oceny grupowe dla obiektów równoważnych [in:] Badania operacyjne i systemowe (J. Owsiński, Z. Nahorski, T.
Szapiro – red.) PTBOiS, 53-64.
Banzhaf J. F. III (1965): Weighted voting doesn’t work: a mathematical
188
Jacek Mercik | On Axiomatization of Plurality Decisions with Veto
Mercik J. W. (2009): A priori veto power
of the president of Poland. Operations
Research and Decisions, 19(4): 61-75.
Condorcet, de Marquis (1785): Essai
sur l’application de l’analyse a la probabilite des elections rendues a la pluralite de voix, Imprimerie Royale, Paris.
Mercik J.W. (2011): On a priori evaluation of power of veto, [in:] Consensual
Processes (Herrera-Viedma et al. eds),
Springer, Studies in Fuzziness and Soft
Computing, 267, 145-158
Dubey A., Neyman R. J., Weber R.
J. (1981): Value theory without efficiency. Mathematics of Operations Research vol. 6, 122-128.
Gärdenfors P. (1976): Manipulation
of social choice function, J. Economic
Theory, vol. 13, 217-228.
Mercik J. (2015): Classification of committees with vetoes and conditions for
the stability of power indices, Neurocomputing, vol. 149 (2015), Part C,
1143-1148.
Gibbard A. (1973): Manipulation of
voting schemes: a general result,
Econometrica, vol. 41, 587-601.
Nurmi H., Uusi-Heikkilä Y. (1985):
Computer simulations of approval and
plurality voting: the frequency of weak
Pareto violation and Condorcet loser
choices in impartial cultures, European
J. Political Economy, vol. 2/1, 47-59.
Gilboa I., Postlewaite A., Schmeidler
D. (2012): Rationality of belief or: why
savage’s axioms are neither necessary
nor sufficient for rationality, Synthese,
Vol. 187, Issue 1, 11-31.
Ramsey D., Mercik J. (2015): A formal
a priori power analysis of the Security Council of the United Nations [w:]
Group Decisions and Negotiation Letters (B. Kamiński et. al. – eds), Warsaw
School of Economics Press, Warsaw,
215-224.
Johnston R.J. (1978) On the measurement of power: Some reactions
to Lawer, Environment and Planning
A, vol. 10, 907-914.
Lissowski G. (2008): Zasady sprawiedliwego podziału dóbr, Scholar,
Warszawa.
Satterthwait M. A. (1975): Strategy
proofness and Arrow’s conditions:
existence and correspondence theorems for voting procedures and social
welfare functions, Journal of Economic Theory, vol. 10, 187-217.
May K. O. (1952): A set of independent
necessary and sufficient conditions for
simple majority decisions, Econometrica, vol. 20, 680-684.
Mercik J. (1990): Wybrane problemy
decyzji grupowych, Wydawnictwo Politechniki Wrocławskiej, Wrocław.
Shapley, L. S., Shubik, M. (1954),
A method of evaluating the distribution of power in a committee system,
American Political Science Review
48(3): 787-792.
O aksjomatyzacji decyzji większościowych z wetem
Abstrakt
W artykule przedstawiono analizę aksjomatów związanych z większościową
metodą agregacji preferencji indywidualnych zarówno wtedy, kiedy konieczny jest wybór jednego z wielu wariantów jak i wtedy, kiedy konieczne jest
189
zaaprobowanie danego pojedynczego wariantu. Rozpatrzono także wpływ na
podawane aksjomaty wprowadzenie nowego atrybutu jakim jest prawo weta
(bezwzględnego jak i względnego). W konkluzji podkreślono, że stosowana
powszechnie metoda agregacji, tj. metoda większościowa nie jest w tym sensie metodą najlepszą.
Słowa kluczowe: agregacja preferencji, aksjomaty, indeksy siły