ECONOMICS 001

Microeconomic Theory

Instructor: Ram Singh

Social Choice Theory

Not for circulation.

1 Preferences and Choices

Results in this section are general and therefore hold for individual as well as social
preference relations. Let1

• X be the set of alternatives; X = {x1 , x2 , ...xm }. Assume, there are at least

three alternatives. That is, m ≥ 3 or ]X ≥ 3.
We allow a very broad interpretation of the set of alternatives, i.e., X. For
instance, it could be a set of feasible allocations, or a set of feasible distributions
of given resources. It can be thought of as a set of policy options available to
the society, or the set of political candidates in the fray, etc. The alternatives in
the set X could be standard (risk-free) allocations of goods, but also we can be
thought of as as the set of ‘risky’ alternatives, such as, lotteries or investment
projects.

• Let R be a ‘weak’ preference relation defined over X. Assume, the preference

relation R has a strict preference relation associated with it, which is denoted
by P - that is, xP y iff xRy and ∼ yRx. Let I be the indifference relations
associated with R - that is, xIy iff xRy and yRx.

• Let R be the set of all possible preference relations. That is, if R is a (weak)
preference relation then R ∈ R. Note that at this stage R is a general preference
relations. It may or may not be ordering. It could represent preference relation
for an individual or it could do so for society/group of individuals.

• Let O be the set of individual preference relations that are orderings; i.e., if
R ∈ O, then R is reflexive, complete and transitive. Clearly, O ⊂ R.

Take any S ⊆ X. Consider any given preference relation R defined over S. Let
P be the strict preference relation associated with R. An element x ∈ S is called a
maximal element of S iff there is no element in S that is strictly preferred to x,
according to the preference relation R. That is, x ∈ S is called a maximal element of
S iff the following holds: ∼ (∃y ∈ S)(yP x).

1
This document has not been proof-read. Please bring typos to my notice.

1
Maximal Set. M (S, R): Let S ⊆ X. A set M (S, R) is the set of Maximal elements
of S w.r.t. preference relation R iff: For all x ∈ S,
[x ∈ M (S, R)] ⇔ [∼ (∃y ∈ S)(yP x)].
As the following example shows, depending on R and S, there can be more than
one maximal elements.

Example 1: Let S = {x, y}. Suppose, the preference relation R is reflexive but is
such that ∼ xP y and ∼ yP x. In such a situation, we will have M (S, R) = {x, y},
i.e., M (S, R) = S.

Again, take any S ⊆ X and preference relation R defined over S. Assume that
according to this preference relation, element x ∈ S is at least as good as any other
element of S. Suppose the decision maker -an individual or a society- has to make
a choice from the elements of S on the basis of the preference relation R. Since x
is at least as good as any other element of S, it is plausible to consider x as one
of the desirable (best) candidates. Formally, for any x ∈ S, we will call x to be a
‘best’ elements of S iff (∀y ∈ S)[xRy] holds. If x is a best candidate, clearly it will
belong to the choice set C(S, R) generated by the preference relation R defined over S.

Best Set. C(S, R): Let S ⊆ X. A set C(S, R) is the set of ‘best’ elements of S iff
[x ∈ C(S, R)] ⇔ (∀y ∈ S)[xRy].
Depending on R and S, there may or may not be a best element. In some cases,
there can be more than one best elements.

Example 2: Let S = {x, y}. Suppose, the preference relation R is reflexive and is
such that xRy and yRx, i.e., xIy. In such a situation, we will have C(S, R) = {x, y},
i.e., C(S, R) = S.

Given the preference relation R and the set S, how does a Maximal element com-
pare with a Best element? In Example 1, we do not know whether R is complete or
not. Suppose, in Example 1 R is not complete; i.e., suppose ∼ xRy and ∼ yRx. Still,
M (S, R) = {x, y}. That is, x as well as y is a maximal element. However, neither x
nor y is a best element. Check that C(S, R) = ∅. Therefore, x ∈ M (S, R) does not
mean that for all y ∈ S, xRy holds.

Take any arbitrarily given S and R. Let, x, y, z ∈ S. Suppose, x ∈ C(S, R) and

z ∈ C(S, R). Since x ∈ C(S, R), for all y ∈ S, xRy holds. That is, x is at least as
good as any other element of S. In particular, xRz will hold. By the same logic, zRx
must hold. Therefore, we must have xIz. In general, all elements in C(S, R) will be
indifferent to each other.

2
 Suppose, for given S, we want to make choice using R. Note that the context
could be of individual decision making ( i.e., R could be an individual preference
relation), or of social choice (if R happens to be the social preference relation). The
question is should C(S, R) or M (S, R) be used as basis of final decision making? By
now it is clear that if x ∈ C(S, R) then x is at least as good w.r.t. R as any other
element of S. But, as we have seen already x ∈ M (S, R) does not mean that for all
y ∈ S, xRy holds. Therefore, it seems plausible to argue that to the extent possible,
the decision/choice from among the set of alternatives in S should be an element of
C(S, R), rather than of M (S, R). However, depending on the choice context, C(S, R)
may be empty. Therefore, we may be restricted to choose from elements of M (S, R),
if it happens to be non-empty.

In any case, the question is: What can we say about the relationship between sets
C(S, R) and M (S, R)? As the following examples demonstrate, depending on S and
R, various possibilities exist.

Example 3: Both C(S, R) and M (S, R) can be non-empty and equal. This is the
case when S = {x, y} and the preference relation R is reflexive and is such that xRy
and yRx.

Example 4: C(S, R) can be empty but simultaneously M (S, R) can be non-empty.

E.g., let S = {x, y} and ∼ xRy and ∼ yRx. Note here R is not ‘complete’.

Example 5: Both C(S, R) and M (S, R) can be empty. For example, when S =
{x1 , x2 , ..., xn , ...} is an infinite set. And, the preference relation R defined over S is
such that:
x2 P x1 , x3 P x2 , ..., xn P xn−1 , ...
However, you can easily prove that C(S, R) will always be a subset of M (S, R).
That is,
Proposition 1 For any give S ⊂ X and preference relation R,2

C(S, R) ⊂ M (S, R).

That is, whenever M (S, R) is empty so will be the choice set C(S, R); whenever
C(S, R) is non-empty so will be the set M (S, R). However, in principle, M (S, R) can
be non-empty even when C(S, R) is empty.

Examples 2-4 show that when the choice is between two alternatives, say x and
y, i.e., when S = {x, y},
• Reflexivity and Completeness of R is a necessary and sufficient conditions for
the choice set C(S, R) to be non-empty.
2
S ⊂ X means that S is a subset of X. It admits the possibility that S = X.

3
 • Reflexivity and Completeness of R together are sufficient but not a necessary
condition for the set M (S, R) to be non-empty.

In contrast, Example 5 shows that when S is infinite the choice set C(S, R) as
well as M (S, R) can be empty, even when the preference relation R is an ordering.
So, for infinite sets the condition ‘R is an ordering’ is not a sufficient for these sets to
be non-empty. When will the sets M (S, R) and C(S, R) be non-empty?

Before answering this question, note that if R is transitive, then the corresponding
strict preference P as well the indifference relation I will be transitive. Alternately,
if P and I are both transitive, then R will also be transitive. However, if R is such
that only P is transitive, then we say that R is quasi-transitive.

Quasi-transitivity: A preference relation R defined over set X is quasi-transitive if

(∀x, y, z ∈ X)[xP y & yP z ⇒ xP z].

When choice is to be made from a finite set, for the choice set to be non-empty it
is not necessary to insist that R be an ordering. As the following Proposition shows,
a weaker condition on R give us non-empty choice set.

Proposition 2 Suppose X is finite. If the preference relation R defined over X is

reflexive, complete and quasi-transitive, then for every non-empty S ⊂ X, C(S, R) is
non-empty.

Proof: (Optional reading) Take any non-empty S ⊂ X. Let S = {x1 , x2 , ..., xn }.

By reflexivity of R, we have xi Rxi . Take x1 . Let y1 = x1 . we have y1 Rx1 . Now,
consider a choice from the pair (x1 , x2 ). From above, we know that when R reflexive
and complete there is best element in the (x1 , x2 ). Call this element y2 (if both are
best, take any element and call it y2 ). So, we have

y2 Rx1 and y2 Rx2 . (1)

Now, consider the set {x1 , x2 , x3 }. We have

either x3 P y2 and y2 Rx3 . (Why?) (2)

If y2 Rx3 , then y2 is a best element of {x1 , x2 , x3 }. If x3 P y2 , then x3 is a best element
of {x1 , x2 , x3 }. To see why, suppose x3 is not a best element of {x1 , x2 , x3 }. This
means, either x2 P x3 or x1 P x3 . If x2 P x3 , then x2 P x3 and x3 P y2 together imply x2 P y2
(remember R is quasi-transitive). But this is a contradiction, since x2 P y2 implies y2
is a not best element of {x1 , x2 }. Similarly, we x1 P x3 gives us a contradiction.
Therefore, the set {x1 , x2 , x3 } has a best element. If we repeat this argument for
the sets {x1 , x2 , x3 , x4 }, ..., {x1 , x2 , ..., xn }, we can prove that S = {x1 , x2 , ..., xn } has
a best element.

4
 
Note that a transitive preference relation R is also quasi-transitive (the converse
is not true necessarily). Therefore, we have the following result:
Proposition 3 Suppose X is finite. If the preference relation R defined over X is
an ordering (reflexive, complete and transitive), then for every non-empty S ⊂ X,
C(S, R) is non-empty.
The set M (S, R) can be shown to be non-empty even under milder conditions.
However, M (S, R) is non-empty, if we assume that R is quasi-transitive.3 To see
the importance of quasi-transitivity for the set M (S, R), suppose, S = {x, y, z} and
the preference relation R is such that xP y and yP z, and also zP x. That is, the
strict preference relation, P , associated with R is not transitive - therefore R is not
quasi-transitive. In this case, you can check that the set M (S, R) is empty. There-
fore, when R is not quasi-transitive, at least for some S, M (S, R) can be empty. In
contrast, Example 4 suggests that completeness of R is not necessary for M (S, R) to
be non-empty. It turns out that M (S, R) is non-empty, when R is quasi-ordering.

Quasi-ordering: A preference relation R defined over set X is called a quasi-ordering

if R is reflexive and transitive.

Proposition 4 If S ⊂ X is finite and and preference relation R is quasi-ordering

over S, then M (S, R) is non-empty.
Proof: Since S is finite, let S = {x1 , ..., xn }. Since R is reflexive, therefore fore each
xi , we have xi Rxi , i.e., ∼ xi P xi . Let a1 = x1 . We rename the other alternatives as
follows: 
x2 , if x2 P x1
a2 =
a1 , otherwise.
That is, we compare a1 = x1 with x2 . If x2 P x1 we rename and call x2 as a2 . On the
other hand, if ∼ x2 P x1 we call a1 as a2 . Note that ∼ x2 P x1 can hold either because
x1 P x2 or because R is not complete and therefore cannot compare x1 and x2 . In any
case, note that in a comparison of x1 and x2 , we have chosen the alternative that is
not inferior to the other.

xj+1 , if xj+1 P aj
aj+1 =
aj , otherwise.
That is, at each stage of comparison we choose the alternative that is not inferior to
any of the alternatives considered so far (in the previous rounds). We carry on with
process. In the last round, we will get an alternative, named an . You can verify that
there cannot be any y ∈ S such that yP an holds. Therefore, an is a maximal element.

3
Extra idea: Actually, if the preference relation is acyclic, M (S, R) will be non-empty.

5
2 Social Choice Rules
In this section, we will consider procedures/rules to derive Social Preference Relation
from individual preference relations. Let, N be the set of individuals in the society;
N = {1, ..., n}. We assume that there are at least two individuals, i.e., ]N ≥ 2. Each
individual in set N has a preference relation defined over X. Let,
Ri be the ‘weak’ preference relation for individual i, defined over X; i = 1, ..., n.
Corresponding preference relation Ri , let Pi be the strict preference relation for
individual i. Let Ii be the indifference relation associated with the ‘weak’ preference
relation Ri for individual i.
(R1 , ..., Rn ) denotes a profile of individual preference relations - one for each indi-
viduals. Let,
Rn be the set of all possible profiles. That is,

Rn = {(R1 , ..., Rn )|Ri ∈ R for each i = 1, ..., n}

To distinguish the notations for social and individual preference relations, we use
the symbol R to denote a ‘weak’ Social preference relation. Again, R ∈ R. Like
individual preference relation, every social preference relation R has associated with
it a strict preference relation, P, and an indifference preference relation, I, associated
with it. By definition, xPy ⇔ xRy and ∼ (yRx). Similarly, xIy ⇔ xRy and yRx.

To repeat, at this stage the social preference relation R may or may not be an
ordering. However, we will assume that each individual’s preference relation is an
ordering, i.e., we assume that

(∀i ∈ N)[Ri ∈ O].

A Social Choice Rule (SCR) is a rule to derive Social Preference Relation from
individual preference relations. Formally, a SCR is a functional relation specified by
the function: f : Rn 7→ R, such that,

(∀(R1 , ..., Rn ) ∈ On )[f (R1 , ..., Rn ) = R],

where R ∈ R. That is, for any given profile of individual orderings (R1 , ..., Rn ) ∈ On ,
function f produces one and only preference relation R, called social preference rela-
tion. As mentioned above, R may or may not be an ordering.

Decisive SCR: A SCR f is decisive if the social preference relation generated by f

is complete, i.e., if ∀(R1 , ..., Rn ) ∈ Rn , f (R1 , ..., Rn ) = R is complete.

Rational SCR: A SCR is rational if the social preference relation generated by f is

an ordering. That is, if ∀(R1 , ..., Rn ) ∈ Rn , f (R1 , ..., Rn ) = R ∈ O.

6
 2.1 Pareto Criterion as a SCR
Suppose we want to use Pareto criterion as a SCR. By definition, x is Pareto superior
to y if: (∀i ∈ N)(xRi y) and (∃j ∈ N)(xPj y). Besides, it is meaningful to treat x
to be ‘Pareto as good as’ y, as long as (∀i ∈ N)(xRi y) holds. Furthermore, we can
say that x is Pareto indifferent to y if: (∀i ∈ N)(xRi y) and (∀i ∈ N)(yRi x), that is,
(∀i ∈ N)(xIi y). Note all these inferences follow from the Pareto criterion. Let us
define a preference relations, R̄, P̄ and Ī, based on these inferences. Formally, for
any given x, y ∈ X, we define:

xR̄y ⇔ [(∀i ∈ N)(xRi y)] (3)

xP̄y ⇔ [(∀i ∈ N)(xRi y) & (∃j ∈ N)(xPj y)] (4)
xĪy ⇔ [(∀i ∈ N)(xRi y) & (∀i ∈ N)(yRi x)] (5)

Consider the statement (4). This statement, i.e., [(∀i ∈ N)(xRi y) & (∃j ∈ N)(xPj y)]
implies that xR̄y holds. However, when [(∀i ∈ N)(xRi y) & (∃j ∈ N)(xPj y)] is true,
the statement (∀i ∈ N)(yRi x) cannot be true; hence, y R̄x cannot hold. So, we can
say that xP̄y, i.e., x is Pareto superior to y. Indeed, we can restate statements (4)
and (5) as follows:

xP̄y ⇔ [xR̄y & ∼ (y R̄x)]

xĪy ⇔ [xR̄y & y R̄x]

You can easily prove the following proposition.

Proposition 5 Relation R̄ is a quasi-ordering. That is, it is reflexive and transitive.

Doubt- why is SCR a quasi-ordering.
The Pareto criterion becomes a SCR if social preferences are based only on this cri-
terion. Formally, iff the preference relation R̄ serves as the social preference relation,
i.e., R = R̄. In that case have: For all x, y ∈ X:

xRy ⇔ xR̄y;
xPy ⇔ xP̄y;
xIy ⇔ xĪy.

You can check that the preference relation R̄ is a quasi-ordering. (Is R̄ a com-
plete preference relation?) Therefore, if Pareto criterion is used as a SCR, the social
preference relation R = R̄ is a quasi-ordering. Now, in view of Proposition 4, we get
the following result.

Proposition 6 Take any given finite S ⊂ X. If R = R̄ as defined above, i.e., if the

SCR is the Pareto criterion, then the set M (S, R) is non-empty.

That is, if Pareto criterion is used as a SCR and the set S is finite, then there will
be at least one maximal element. However, when Pareto criterion is used as a SCR,

7
there may not be any best element even if the set S is finite, i.e., the set C(S, R) can
be empty. Consider the following example.

Example 6:. Let N = {1, 2, ..., n} and S = {x, y}, where x 6= y. Suppose, individual
preferences over S are as follows: xP1 y, yP2 x and for all other i = 3, 4, ..., N , xIi y
holds. Assume that the Pareto criterion is used as a SCR, i.e., R = R̄.

For the setting as in Example 6, you can check that following statements are true:

• x and y are both Pareto optimum

• neither xPy nor yPx holds, so M (S, R) = {x, y}.

• neither xRy nor yRx, so C(S, R) is empty.

Question: Can you change individual preferences in Example 6 and still make the
above claims?

Note in Example 6, the choice set, C(S, R), is empty on account of the fact that
Pareto criterion is not decisive, i.e., R = R̄ is not complete. The following Proposi-
tion 7 gives the condition for the Pareto criterion to be decisive.

Note: Proposition 7, its proof and the rest of Subsection 2.1 are strictly
optional readings.

Proposition 7 Pareto criterion is a decisive SCR iff

(∀x, y ∈ X)[(∃i ∈ N)(xPi y) ⇒ (∀j ∈ N)(xRj y)]

Proof of if (sufficiency) part: Given that individual preferences are rational, we

have: (∀i ∈ N)[xRi y or yRi x]. Further, the following two cases are possible.
Case 1: (∀i ∈ N)[xIi y] in that case, the condition in the statement of the propo-
sition is trivially satisfied and the xRy holds. That is, the social preference relation
is complete and hence decisive.
Case 2: If (∃i ∈ N)(xPi y) then (∀j ∈ N)(xRj y) holds and therefore, xPy. That
is, the social preference relation is complete. So, the condition is sufficient.

To see why the condition is necessary, suppose the above condition is violated,
i.e., ∃i ∈ N such that xPi y, and at the same time ∃j ∈ N such that yPj x. In that
case, we have
∼ xR̄y and ∼ (y R̄x), i.e.,
∼ xRy and ∼ (yRx), i.e.,
the SCR is non-decisive. Therefore, the condition is necessary.

8
Pareto-inclusive SCR: A SCR is Pareto inclusive, i.e., satisfies the Pareto Criterion
if: For all x, y ∈ X

xR̄y ⇒ xRy
(∀i ∈ N)(xRi y) ⇒ xRy.

Moreover,
xR̄y and ∼ y R̄x ⇒ xPy.

2.2 Social Welfare Functions

Recall, a SCR in its most general form can be defined as a function f : Rn 7→ R, such
that,
(∀(R1 , ..., Rn ) ∈ Rn )[f (R1 , ..., Rn ) = R], where R ∈ R.
In particular, neither the individual preferences nor the social preference needs to
be an ordering. However, all along we will assume that the individuals are rational,
i.e., the individual preferences are orderings: For all i = 1, .., n, Ri ∈ O. We will
see that the rationality of individuals does not guarantee that the social preferences
relation will also be rational. If the social preference relation generated by a SCR, f ,
is an ordering, we say that f is a ‘social welfare function’.

Social Welfare Function (SWF): Let D ⊂ On . A SCR, f : D 7→ O, is called a

Social Welfare Function, if f such that:

(∀(R1 , ..., Rn ) ∈ D)[f (R1 , ..., Rn ) ∈ O].

.
Note that the set of SWFs is a subset of the set of SCRs. In the previous section
we saw that when the Pareto criterion used as a SCR the induced preference relation
R is not complete for ∀(R1 , ..., Rn ) ∈ On . (See example 6.) That is, R = R̄, as
defined above, is not an ordering for ∀(R1 , ..., Rn ) ∈ On . Therefore, when D = On ,
the Pareto criterion as a SCR is not a SWF. In the next section, we will introduce
some SCRs that are SWFs in that the social preference relations produced by them
are orderings.

3 SCRs: Desirable Features

By definitions, a SCR gives us a social preference relation that depends on the in-
dividual preferences. We want the social preference relation to reflect individual
preference/tastes in a meaningful way. For this, the SCR must satisfy some plausible
conditions related to the relationship between the profile of individual preferences
and the social preference relation produced by the a SCR.

9
 Let f denote the SCR under consideration. First of all, we may want our SCR
to be SWF (Why?) That is, we may want the SCR to generate a social preference
relation that is an ‘ordering’. Formally, we may want f to satisfy the following con-
dition:

Condition O: A SCR, f , satisfies condition O if it is a SWF. That is, if the range

of function f is the set O.

Note the the Pareto criterion used as a SCR is not a SWF. (See Section 2.1 and
Propositions 3 above.)

Next, we want to allow for diversity in tastes, opinions, beliefs and priorities of
individuals in the society. More precisely, we want to allow individuals in the society
to have different ranking of the alternatives in the set X. Still, we want the SCR
to work. That is, we want the SCR to be a guide for social decision making. Since
the SCR provides guidance by generating a social preference relation. Therefore, we
want the SCR to produce a social preference relation, for every possible profile of
individual orderings. That is, we want to make sure that the SCR generates a social
preference relation, for every possible profile of individual orderings. Formally, we
want to the SCR, f , to satisfy

Condition U. Unrestricted Domain: A SCR f satisfies condition of ‘unrestricted

domain’, if its domain is On , i.e., D = On

So, if a SCR f satisfies Condition U, f generates a social preference relation for

every possible profile of rational individual preferences. Note, as such we do not insist
that the social preference relation be an ‘ordering’ - of course, it will be be useful if
the social preference relation generated by the SCR happens to be an ordering.

Suppose, the society has to choose between two different alternatives, say x and
y. Assume that there is total agreement among individuals that x is better than y -
i.e., (∀i ∈ N)(xPi y) holds. In such a scenario, the social preference relation produced
by a sensible SCR would be expected to reflect this agreement. That is, it would be
expected that the social preference relation also say that x is better than y. Moreover,
we would expect the social preference relation to behave like this in every instance of
perfect agreement. Formally, we want the SCR to satisfy the following condition.

Condition P. Weak Pareto Criterion/Principle: A SCR f satisfies condition P, if

(∀x, y ∈ X)(∀i ∈ N)[xPi y ⇒ xPy].

Suppose the society has to choose between two different alternatives, say x and y.
It seems plausible to assume that the social choice between x and y should depend
only on how individual feel about x vis-a-vis y. A third alternative say z - different

10
from x and y - be irrelevant for the social choice between x and y. Specifically, as long
as individual preferences over x and y remain the same, the social choice between x
and y not be affected by presence or absence of the third alternative z. Also, if some
of the individuals change their ranking of x vis-a-vis z or y vis-a-vis z, as long as
individual preferences over x and y remain the same, the social choice between x and
y should not change.
To capture this idea formally, consider a SCR, f . Take any two profile of individual
preferences over X; say, (R1 , ..., Rn ) ∈ On and (R10 , ..., Rn0 ) ∈ On . Let,

• f (R1 , ..., Rn ) = R, i.e., R is the social preference produced by the SCR when
individual preference orderings are R1 , ..., Rn , respectively.
• f (R10 , ..., Rn0 ) = R0 , i.e., R0 is the social preference produced by the SCR when
individual preference orderings are R10 , ..., Rn0 , respectively.

Now, take any two alternatives, say x, y ∈ X. Suppose, as far as the individual
ranking/ordering of x and y is concerned, the profiles (R1 , ..., Rn ) and (R10 , ..., Rn0 ) are
the same. Formally,
(∀i ∈ N)[xRi y ⇔ xRi0 y].
That is, the two profiles generate the same ranking of alternatives in set S = {x, y}.
However, the profiles (R1 , ..., Rn ) and (R10 , ..., Rn0 ) could be different from each other
in the individual ranking/ordering of x and z or y and z, where z is different from x
and y.
However, if the society has to choose between x and y, i.e. social choice to be
made from the set S = {x, y}, the social choice when the profile of individual ordering
is (R1 , ..., Rn ) should be the same as when the profile is (R10 , ..., Rn0 ). For example,
suppose x is treated to be socially at least as good as y, when the profile is (R1 , ..., Rn ).
Even when the profile is (R10 , ..., Rn0 ), x should still be treated to be socially at least
as good as y; and, vice-versa. Formally put,
(∀i ∈ N)[xRi y ⇔ xRi0 y] ⇒ [(xRy iff xR0 y) and (yRx iff yR0 x)].
(Remember that when the profile is (R1 , ..., Rn ) the social preference relation is R;
when the profile is (R10 , ..., Rn0 ) the social preference relation is R0 .) Therefore, re-
writing, we can say
(∀i ∈ N)[xRi y ⇔ xRi0 y] ⇒ [C(S, R) = C(S, R0 )].
Now, we are ready to state this condition formally for any set S ⊂ X.

Condition I. Independence of Irrelevant Alternatives : Take any S ⊂ X, and ANY

two profiles of individual orderings, say (R1 , ..., Rn ) ∈ On and (R10 , ..., Rn0 ) ∈ On .
Let f (R1 , ..., Rn ) = R and f (R10 , ..., Rn0 ) = R0 . A SCR f satisfies Condition I if the
following holds:
(∀x, y ∈ S)(∀i ∈ N)[xRi y ⇔ xRi0 y] ⇒ C(S, R) = C(S, R0 )

11
 Finally, we do not want the SCR to be a dictatorial rule. A SCR is dictatorial if
there is a dictator. An individual i ∈ N is called dictator if:

(∀x, y ∈ X)(∀(R1 , ..., Rn ) ∈ On )[xPi y ⇒ xPy], i.e.,

for every pair of alternatives x, y ∈ X and every profile of preference relations, when-
ever individual i prefers x over y, i.e., whenever xPi y holds, x is treated to be socially
preferred over y, regardless of how individuals other than i feel about x vis-a-vis y.
This is implies that in every context wherein the dictator has strict preference, his
preference will prevail over the rest of the society. We want our SCR to be non-
dictatorial, as defined below.

Condition ND. Non-Dictatorship: A SCR f satisfies condition ND, if there is no

individual i ∈ N such that

(∀x, y ∈ X)(∀(R1 , ..., Rn ) ∈ On )[xPi y ⇒ xPy].

12