Lesson 3

Welfare Economics.
 Welfare Properties of Walrasian Equilibrium: The
two Fundamental Theorems of Welfare.
z The existence of a Walrasian Equilibrium is interesting as
a positive result insofar as we believein the behavioral
assumptions –primarily that agents are price takers-
which underly the model. However, we may still be
interested in WE for their normative content:
z Are WE optimal or efficient in some sense?
z Recall Pareto efficiency:
z Pareto Efficiency: An allocation of goods is Pareto
efficient if no one can be made better off without making
someone else worse off. Formally:
z Definition: A feasible allocation x is Pareto optimal (or
Pareto Efficient) if there is no other feasible allocation y
such that:
1) ui(yi)≥ ui(xi) for all i, and
2) uj(yj)> uj(xj) for at least some j

©2005 Pearson Education, Inc. Chapter 16 2

 Welfare Properties of Walrasian Equilibrium: The
First Fundamental Welfare Theorem.
Recall the alternative definition of WE taking into account the
Equilibrium allocation:
Definition (alternative): A pair allocation-price (x*, p*) is a WE if:
z 1) ∑i x*i =∑i wi (x* is feasible), and
z 2) If ui (xi)> ui (x*i), then p* xi > p* wi (x is not affordable).

Proposition FTW : If (x*, p*) is a WE for the initial endowment w, then

x* is Pareto efficient.

Proof: Suppose on the contrary that x* is not Pareto efficient. Then

there is an allocation x such that for all i, ui (xi)> ui (x*i), and
∑I xi =∑i wi (x is feasible), →p* ∑i xi =p*∑i wi (1)
As x* is a WE, then by definition, for all i
p* xi > p* wi and adding over all i’s :
p* ∑i xi >p*∑i wi , which contradicts (1) (= ∑i xi )
Then x* is Pareto efficient.

©2005 Pearson Education, Inc. Chapter 16 3

 Welfare Properties of Walrasian Equilibrium: The
First Fundamental Welfare Theorem.

Recall Edgewoth’s box: It looks like that in the text book case:
Every WE is a PO allocation (1º TW) and every PO is a WE (2º TW )

0B
The text book case
relies on many
assumptions:
1.Convex Preferec.
x* 2. No satiation
3. Perfect divisibility
4…

W
0A

©2005 Pearson Education, Inc. Chapter 16 4

 Welfare Properties of Walrasian Equilibrium:
The First Fundamental Welfare Theorem.

Is every WE always a PO allocation? Under the assumptions of

our model the answer is YES, but, in general, we can find problems
when relaxing two axioms: 1) Satiation→allowing points of satiation
inside the box and 2) Good indivisibilities.
Example 1. B has a point of maximal satiation inside the box:
OP
0B Indifference curves are convex
but not strictly convex.
IA4 Suppose first that there is no
satiation point inside the box.
X* Pareto optimal allocations:
OP IA3 IB1 North and west sides of the box
.x 0
IA2 Suppose that the price budget
IB2 line coincides with IA2 y W=X0.
IB3
IB6 IB5 IB4 IA1 WE=X* and WE=PO
0A

©2005 Pearson Education, Inc. Chapter 16 5

 Welfare Properties of Walrasian Equilibrium:
The First Fundamental Welfare Theorem.

Example1 (cont.). Suppose now that B has a point of maximal

satiation inside the box. Let xS be such point a recall that x0 is the
Initial endowment.
B is completely satiated over the straight line xS-x0

xS OP
0B Pareto optimal allocations: from xS to
OB in the north side of the box.
IA4
Suppose, as before, that the price
budget line coincides with IA2
IA3 IB1 and that W=X0.
.x 0
IA2
By satiation:
IB2
IB3 WE=X0 and WE is not PO, since A
IB6 IB5 IB4 IA1 Is better off in xS without B being
0A worse off.

©2005 Pearson Education, Inc. Chapter 16 6

 Welfare Properties of Walrasian Equilibrium:
The First Fundamental Welfare Theorem.
z Example 2: A variation of the above argument. A feels satiated
over and from IAS. Each allocation in the area from IAS gives him
the same utility.

0B
IAS
.C Let W be the initial endowment.
With the budget constraint
IA
.P through W, the allocation
C=WE.
.W But C is not a PO allocation,
IB0 since B is better off in allocation
IB1 P, without A being worse off.
IB2
0A

©2005 Pearson Education, Inc. Chapter 16 7

 Welfare Properties of Walrasian Equilibrium:
The First Fundamental Welfare Theorem.
Example 3: Indivisible goods. Preferences are defined over bundles in
R2, but the consumer can only choose points in the grid. Let x0 be the
initial endowment.

0B

The WE is allocation xWE, but this

allocation is not PO, since A
is better off in x1 without B being
xWE worse off.
x1
A1
A0
0A x0

©2005 Pearson Education, Inc. Chapter 16 8

 Welfare Properties of Walrasian Equilibrium:
The Second Fundamental Welfare Theorem.
The normative content of WE comes from the 2º Theorem of Welfare:
P0→WE
Notice that at each PO allocation: all the baskets preferred by A to
this PO have an empty intersection with those preferred by B, that
is, the agents’s preferred sets to a given PO are disjoint sets.
Then, it is possible to draw a straight line (a hyperplane) passing
between the two sets and “separating” them, and going through
the PO allocation as well.
Then:
0B
Baskets
This PO could be Baskets
X PO
preferred
“supported” by a preferred by A to xPO
descentralized price by B to xPO uA
system. uB
0A

©2005 Pearson Education, Inc. Chapter 16 9

 Welfare Properties of Walrasian Equilibrium:
The Second Fundamental Welfare Theorem.
z There is a Theorem giving the sufficient conditions for the
existence of a hyperplane “separating” sets: Theorem of the
Separating Hyperplane.
z Hyperplane: Let a be in R, p in Rn. A hyperplane H(p,a) in
Rn is a set such that: H(p,a)={x in Rn : px=a}
H(p,a) is a n-1 dimensional set: in R2 is a straight line, in R3 a
plane.
Example: In a model with two goods, the budget line is the
hyperplane H(p,M)={x en R2 : px=M}, where M is the agent’s
wealth, p are the prices and x the agent’s consumptions.
Separating Hyperplane: H(p,a)={x in Rn : px=a} separates (or
strictly separates) the non-empty sets S1 and S2 in Rn if:
x1 in S1 implies that p x1≥ a (> a) If H exists, then S1 and S2
x2 in S2 implies that p x2 ≤ a (< a) are separable.

©2005 Pearson Education, Inc. Chapter 16 10

 Welfare Properties of Walrasian Equilibrium:
The Second Fundamental Welfare Theorem.
Theorem of the Separating Hyperplane : (Minkowski)
Let S1 and S2 in Rn be convex, disjoint and non-empty sets, then
there exists a hyperplane separating them, that is: there exists
H(p,a)={x in Rn : px=a} , such that
x1 in S1 implies that p x1≥ a p x1≥ p x2
¨
x2 in S2 implies that p x2 ≤ a
The theorem gives sufficient conditions for existence:

separable No disjoint
S1
S1
S1 S2
S2 S2 S1 no convex
Non-separable

©2005 Pearson Education, Inc. Chapter 16 11

 Welfare Properties of Walrasian Equilibrium:
The Second Fundamental Welfare Theorem..
The logic-foundations of the 2º TW is the Separation Theorem
¨convex preferences are needed in order the agents’
preferred sets are convex and can be separated by a
hyperplane.

2º Theorem of Welfare: Suppose that x* a PO allocation with

x*i >>0, for all i=1,2..,n, and that the agents’ preferences are
convex, continuous and monotone. Then, x* is a WE for the
initial endowments wi=x*i, for all i=1,2..,n.

Proof: Let Ρi={x in Rk: ui(xi)> ui(x*i)}, (the set of baskets

preferred by i to x*i) and let

Ρ=∑i Ρi ={z : z=∑i xi, xi in Ρi }, (the set of all aggregate

bundles that can be distributed among the n agents so as to
make all of them better off).

©2005 Pearson Education, Inc. Chapter 16 12

 Welfare Properties of Walrasian Equilibrium:
The Second Fundamental Welfare Theorem..
Proof (cont.) Since the agents’ preferences are convex ¨each
Ρi is un a convex set.
Since the sum of convex sets is convex¨ Ρ is a convex set.

Let w=∑ix*i be the current aggregate bundle

Since x* is a PO allocation, then there is no redistribution of
x* that makes everyone better off ¨
w does not belong to Ρ and w ∩ Ρ=∅ (empty intersection).

We have then two non-empty, convex and disjoint sets

w=∑ix*i and Ρ: then there exists a p such that
pz ≥ p ∑ix*i = pw, for all z in Ρ,
or rearranging,
p(z-∑ix*i ) ≥ 0, for all z in Ρ.

©2005 Pearson Education, Inc. Chapter 16 13

 Welfare Properties of Walrasian Equilibrium:
The Second Fundamental Welfare Theorem.
z It has to be shown now that p is in fact an equilibrium price
vector ¨ that the pair (x*, p) is a WE.

z By the (alternative) definition of WE, it tranlates to showing

that:
z 1) ∑i x*i =∑i wi (x is feasible), and
z 2) If ui (xi)> ui (x*i), then p* xi > p* wi (x is not affordable).

1) Is trivially satisfied by hypothesis. Then, it remains to show

2).
The proof consists of three steps: (we do not prove them here):

a) p es non-negative
b) If ui (yi)> ui (x*i), then p* yi ≥ p* wi , for all i
c) If ui (yi)> ui (x*i), then p* yi > p* wi, for all i.

©2005 Pearson Education, Inc. Chapter 16 14

 Welfare Properties of Walrasian Equilibrium:
The Second Fundamental Welfare Theorem.

z Implications of the 2º TW :
z Distribution problems can be separated from
efficiency problems.
z The market enables to achieve any resource
allocation: it is neutral from a distributive
viewpoint.
z Normative content: All PO allocations can be
sustained by a price-system when there is a
right redistribution of initial endowments¨
implications for Polítical Economy.

©2005 Pearson Education, Inc. Chapter 16 15

 The Second Fundamental Welfare Theorem :
Examples where PO allocations cannot be
descentralized by a price system

z Example 1: No convexity of preferences: A’s preferences

are not convex.
0B
Let X* be PO, and let W= X* .

The price-vector p=(p1,p2)

X* cannot support X* as a WE,
A3
C since A would prefer
A2 allocation C to X*:
-p1/p2
A1
X* is not a WE.
B1
B2
0A

©2005 Pearson Education, Inc. Chapter 16 16

 The Second Fundamental Welfare Theorem :
Examples where PO allocations cannot be
descentralized by a price system
z Example 2: Relaxation of non-satiation. The points of
maxima satisfaction (bliss points) of A and B are in the
Edgeworth’ box. Let XA and XB be such points for A and B.

0B PO allocations= tangency points

between XA y XB..
.XB
Allocation X0=PO and let W=X0
and B(p) the budget line.
X0
Is X0 a WE? NO if prices are
.X A
B positive, since both A and B are
better off in XA and XB and are
A B(p) feasible for them.

0A

©2005 Pearson Education, Inc. Chapter 16 17

 The Second Fundamental Welfare Theorem :
Examples where PO allocations cannot be
descentralized by a price system
z Example 2 (cont). Notice that if both XA and XB change of
place instead, each agent in his bliss point is not feasible:

0B PO allocations= tangency
points between XA y XB.
.XA Allocation X0=PO and let W=X0
X0
and B(p) the budget line.

.X B
A Is X0 a WE? YES, since now
XA andXB are not feasible for
B B(p) the agents.

0A

©2005 Pearson Education, Inc. Chapter 16 18

 Examples where PO allocations cannot be
descentralized by a price system: Arrow’s
exceptional case.

z Example 3. Arrow’s exceptional case. Let us assume

satiation: B is satiated of x1 in x1B. Let W=X0
X1B
A2 0B In W, A has nothing but good 1.
A0
A1 A3 A4
X0=PO, Is X0 a WE?

The unique price-vector tangent

to X0 is p1/p2=0, and then p1=0.

But if p1=0, A will maximize in

B0 C, since x1 is a free good. .
B2 B1
0A B4 C Then X0 is not a WE.
X0
Preferred
Set of B
©2005 Pearson Education, Inc. Chapter 16 19
 Arrow’s exceptional case.
z Notice that in X0, the value of A’s consumption basket is
zero and B’s preferred set is outside the box (not feasible)
z Arrow’s exceptional case can be summarized in any of the
following statements:
z 1. There is no other state in the economy where the value of
good 2 for A is lower than in X0. Then, for p1=0, the value of
goods in X0 for A, is the minimum one (=0).
z 2. In X0, the MgU of x1 for B is not positive (B is satiated)
z 3. In X0, A has nothing desired by B ( then it is not possible
to trade).

z In order any PO is achievable as a WE the above 1), 2) y 3)

statements have to be eliminated, that is, all the agents
posses some units of a good that are desired by someone
else ¨ nobody can be excluded from trading.

©2005 Pearson Education, Inc. Chapter 16 20

 The maximization of Social Welfare.
z Up to now we have only considered individual decisions to discuss
the existence and optimality of WE.
z However, when studing the 2º TW some problems of collective
decision seem to appear: for instance, some criterion is needed to
decide which Pareto optimum is going to be decentralized¨ how
to distribute welfare in the society.
z Collective decision problems are related back with:
z 1. Bentham and Mills’ studies on personal welfare.
z 2. Voting Theory of Condorcet and Borda.
z Modern formulation: start with Bergson (1938), who defined a
social welfare function (swf) and has evolved with
z Arrow (1951, 1963), who changed the viewpoint (SWF).
z Collective decision problems are known nowadays as the
“Theory of Social Choice”, whose aim is to design
evaluation rules gathering individual preferences: “to design
aggregation criteria of individual preferences in order to
obtain social preferences”. ¨he same criteria are the SWF’s

©2005 Pearson Education, Inc. Chapter 16 21

 Paretian judgement values
z 1. Independence of the process: the process by which a
particular allocation is achieved is not important.
z 2. Individualism: Under the Paretian criterion the only
important aspect of an allocation is its effect on the
individuals of a asociety.
z 3. No paternalism: Individuals are the best judges of their
own welfare.
z Doubts: drugs, child pornography, etc.
z 4. Benevolence: The Paretian criterion is benevolent with
individuals since an increase (caeteris paribus) in the utility
of any individual is considered a welfare improvement.
z Doubts: an increase in the utility of the richest person of a
society is considered welfare improving regardless of some
other people dying by starving.

©2005 Pearson Education, Inc. Chapter 16 22

 Paretian criterion:
Set of utility possibilities:
The set of utility vectors assigned to the feasible
allocations.
U={(u1, u2, …, un) in Rn: there is a feasible allocation x
such that ui≤ ui(xi), for all i=1,2,3,…,n}.

Utility Frontier or Pareto Frontier:

UF={(u1, u2, …, un) in Rn: there is no other vector (u’1, u’2,
…, u’n) in U such that u’j ≥ uj for all j=1,2,..,n, and u’i > ui
for some i}.

The utility frontier is all the utility vectors assigned to the

Pareto-efficient allocations.

©2005 Pearson Education, Inc. Chapter 16 23

 Utility Frontier:
Graphically:
02
u1B C
u1C
B
u1A Feasible allocations and
u2C
A Contract curve.
u2B
01 u2A
u201 A
u2A
Set of utility possibilities, U and
u2B B UF Utility frontier, UF (or Paretian frontier)
C
u2C
U
02
u1A u1B u1C u1

©2005 Pearson Education, Inc. Chapter 16 24

 Utility Frontier:
z The utility frontier as a function of utility levels can be
calculated from the optimization problem characterizing
the efficient allocations:
z Max u1(x1)
z s.t. u2(x2)≥u2=c (utility constraint)
z s.t. x11+x21=w1 and x12+x22=w2 (feasibility constraint)

z Solution: x1*(u2, w1,w2) and x2*(u2, w1,w2), and substituing

into agent 1’s utility function:
z u1(x1*(u2, w1,w2))=u1 (u2, w1,w2),

z Or implicitely:
z F(u1 , u2 )=0

©2005 Pearson Education, Inc. Chapter 16 25

 Social Welfare functions and Social
Optima.
z Problem: The Pareto efficiency criterion cannot generate a
complete order over the feasible allocations. Even some
pairs of allocationas cannot be compared.

u2 .α 2 α* and α2 cannot be compared:

Neither α* is Pareto superior to α2,
nor is α2 Pareto superior to α* .

.α* α1 is inefficient but α2 is not

.α
1
UF
Pareto superior to α1.

u1

©2005 Pearson Education, Inc. Chapter 16 26

 Social Welfare functions and Social
Optima.
z Bergson’s social welfare function (swf) is a function
assigning utility values to the feasible allocations of an
economy, and generating a complete, transitive and
reflexive ordering on the set of feasible allocations.
z W: Rn→R, W(u1(x1),u2(x2),…,un(xn))
z Any swf defined as a Bergsonian’s swf has some underlying
distributive principles.
z Paretian swf: swf with paretian judgement values:
z 1. Independende of the process
z 2. Individualism →W(x1, x2,…,xn)
z 3. No paternalism→W(u1(x1),u2(x2),…un(xn))=W(u1,u2,…un)
z 4. Benevolence (Monotonicity): W increasing in each uj
δWu
( 1,u2,..,un)
=Wj >0, for all j =1,2,..,n
δuj

©2005 Pearson Education, Inc. Chapter 16 27

 Social Welfare functions and Social
Optima.
z Consequences of benevolence (or monotonicity).
Assume two agents. The swf is W(u1,u2)
z 1. The indiference curves of welfare or isowelfare curves
more far away from the origin represent higher welfare
levels. .
z 2. Isowelfare curves have negative slope. Let
W(u1,u2)=W0 be an isowelfare curve.
z dW=W1 du1+W2 du2=0
du2 W1
=− <0
du1 W2
z since W1>0 y W2>0

©2005 Pearson Education, Inc. Chapter 16 28

 Social Welfare functions and Social
Optima.

z Isowelfare curves in the utility space.

u2

W’’

W’

W0 UF
u1

©2005 Pearson Education, Inc. Chapter 16 29

 Social optima.
z A social optimum maximizes the paretian swf over the set
of feasible allocations.
z As the set of feasible allocations can be expressed as the set of
utility possiblities, the maximization can be written as:
z Max W(u1,u2,…un) s.a. F(u1 , u2,…,un)=0.
z For two agents: Max W(u1,u2), s.a. F(u1 ,u2)=0
z Associated Lagrangian: L(u1,u2, λ)= W(u1,u2)- λ F(u1 ,u2)
δ L δW δ F
= − λ = 0
δ u1 δ u1 δ u1
δ L δW δ F
= − λ = 0
δ u2 δ u2 δ u2
δ L
= F (u1 , u 2 ) = 0
δλ

©2005 Pearson Education, Inc. Chapter 16 30

 Social Optima.
z The F.O.C. imply (for interior solutions):
δW δ F
δ u1 δ u1 du du
= → 2 en W = 2 en F → RMSW = RMS F
δW δ F du1 du1
δ u2 δ u2

u2

SO
At the SO, the isowelfare W’’
W’’
is tangent to the UF:
MRSW=MRSF
W’

UF W0
u1
©2005 Pearson Education, Inc. Chapter 16 31
 Social optima.
Let (u1SO, u2SO), be the pair of utilities at a SO. These utilities
are associated to an allocation x*=(x1*,x2*) in the Egeworth’s
box that maximizes a swf. Then
Proposition: If x* maximizes a swf, then x* will be Pareto
efficient.
Proof: Trivial, by monotonicity of W. Suppose, on the contrary,
that x* is not Pareto efficient. Then, it will exist another
feasible allocation x’ such that:
ui(xi’)> ui(xi) for all i=1,2,..,n, and by monotonicity of W
W(u1(x1’),u2(x2’),…, un(xn’))> W(ui(x1*),u2(x2*),…, un(xn*)),
that contradicts that x* maximizes the swf W(u1,u2,…, un).
Conclusion: Pareto efficiency is necessary for the SO.

©2005 Pearson Education, Inc. Chapter 16 32

 Social Optima.
z Consequences: The welfare maximizing allocation are PO.
z Question: Is any PO achievable as the maximum of a swf?
NO in general, but:
z Proposición: Let x* be a PO allocation, with xi*>>0,
i=1,2,..,n. The individual utility functions ui, i=1,2,..,n, are
concave, continuous and monotone. Then, there exists
achoice of parameters ai* such that x* maximizes ∑i ai*ui(xi)
subject to the feasibility constraint.

z Proof: For two agents the proof is very simple and can be
graphically explained.
z Construct the set of U of utility possibilities. Since the ui are
concave, U is a convex set.

©2005 Pearson Education, Inc. Chapter 16 33

 Social Optima.
Isowelfares: W0=a1u1+a2u2, o u2=W0/a2-(a1/a2)u1, du2/du1=-(a1/a2) o
MRSW=a1/a2

u2 Let P be the PO allocation to

P=SO be achieved as a SO, and let
MRSF en P=α.
α Then, as at theSO:
MRSW=MRSF, then choose a
ratio a1/a2= α , and the
UF tangency of the isowelfare
curve with the UF is in P.
Hence P=OS.
u1

©2005 Pearson Education, Inc. Chapter 16 34

 Social Choice: Arrow’s Impossibility
Theorem
z Problem: With the Pareto’s criterion there are many efficient
allocations associated with different distributions of welfare
among individuals.
z Question: How to choose a socially optimal allocation?

z Change of approach: From individual preferences, some

social preferences could be defined such that they order the
set of feasible allocations.

z Let <i be the preference relationship of individual i defined

over set A: set of all feasible allocations. Notice that <i is
now defined over allocations instead of over individual
consumption baskets:
u1(x1,x2,…,xn), u2(x1,x2,…,xn),…, un(x1,x2,…,xn).

©2005 Pearson Education, Inc. Chapter 16 35

 Social Choice: Arrow’s Impossibility
Theorem
z Is there any mechanism to obtain from the so
defined individual preferences {<i} a social
preference relationship <, guaranteeing that the
social orderings satisfy some desiderable
properties?

z If such a mechanism exists, then it will be called

a Social Welfare Function (SWF).

z Thus, a SWF is a mechanism or aggregation

rule of individual preferences, obtaining a social
ordering of the distinct allocations.

©2005 Pearson Education, Inc. Chapter 16 36

 Social Choice: Arrow’s Impossibility
Theorem
z Clarification:
z A Bergsonian social welfare function (swf) is a function on
the set of utility possibilities and associates (in some precise
way) a real number to each vector of utilities belonging to it,
thus generating an ordering of this set.
z A Social Welfare Function (SWF) is a function on the set of
individual preferences over the set of the possible social
states, and associates a social preference to each possible
configuration of individual preferences.
z The SWF concept is more general than that of swf.
Changes in the individual preferences for a given SWF will
change the social preferences and hence the swf. And, a
different SWF on a given set of individual preferences will
produce a different social ordering and hence a distinct swf
as well.

©2005 Pearson Education, Inc. Chapter 16 37

 Social Choice: Arrow’s Impossibility
Theorem
z Let A be a set of alternatives or social states, and {<i} be
the individual preferences over A. Consider a criterion or
aggregation rule of {<i}, generating some social
preferences <.
z Desiderable properties of the aggregation criterion.
z 1. Completness,
z 2. Reflexivity, The rule generates an order
z 3. Transitivity,

z 4. Universality or condition of unrestricted domain:

For all {<i}, there exists a social preference
< = Σ(<1, <2, …., <n),
where Σ denotes the aggregation rule or mechanism of
individual preferences.

©2005 Pearson Education, Inc. Chapter 16 38

 Social Choice: Arrow’s Impossibility
Theorem
z Universality or condition of unrestricted domain (cont):
This property says that from any set of individual
preferences {<i}, a social preference relationship < can
be derived.
z This property has a clear logical content since it is a
completeness property with regards the obtention of <. It
also has a clear political content: the aggregation
mechanism is permissive enough to admit any system of
values and/or rules of individual behaviors.

z We will see next that this property is not generally

satisfied:
z Example:The Vote-Paradox

©2005 Pearson Education, Inc. Chapter 16 39

 The Vote-Paradox
z Example:The Vote-Paradox:
z Suppose three agents and three social
states (or alternatives): {a,b,c}
z Aggregation rule: mayority rule: state “a”
is socially preferred to state “b”, if “a” is
preferred by the mayority of the
individuals.
z Individual preferences over social states:
z (a,b,c)1, (b,c,a)2, (c,a,b)3

©2005 Pearson Education, Inc. Chapter 16 40

 The Vote-Paradox
z The Vote-Paradox (cont):
z Social preferences: (pair-comparisons)
a: 2 votes
(a,b) b: 1 vote a afb or [a,b]
b: 2 votes
(b,c) c: 1 vote a bfc or [b,c]
Transitivity implies that afc, or [a,c].
Let us compare now the pair (a,c)
a: 1 vote
(a,c) c: 2 votes a cfa or [c,a]
The rule is not transitive!

©2005 Pearson Education, Inc. Chapter 16 41

 The Vote-Paradox
z The Vote-Paradox (cont): then, this mechanism may create
a problem: it can generate social orderings that are not
transitive.
z The mayority rule fails for this particular individual
preferences, but it would produce a transitive social ordering
for identical preferences, for example, (a,b,c)i, i=1,2,3.
z Another example: (a,c,b)1, (b,a,c)2, (a,b,c)3
z What is desired is that the social choice rule works out for
any type of individual preferences.
z Types of possible preferences for three social states:
z (a,b,c)i, (a,c,b)i , (b,a,c)i, (b,c,a)i, (c,a,b)i, (c,b,a)i
z Each one of them can be combined with each of all the
others, so that there are: 63=216 possible types of
preferences.
z The social choice rule has to be valid for all of them: the
domain of the function transforming a set of individual
preferences in a social ordering is not restricted.
©2005 Pearson Education, Inc. Chapter 16 42
 Social Choice: Arrow’s Impossibility
Theorem
5. Unanimity or Pareto rule:
For any pair a and b in A, if a fi b for all i, then
a f b.

z This condition is either too weak or too strong

z Too weak in the sense that any social preference

must consider a better than b if all the individual
so consider it.

z Too strong in the sense that to consider unanimity

as the unique criterion of the social rule implies
that the orderig relationship is going to almost
never order.
©2005 Pearson Education, Inc. Chapter 16 43
 Social Choice: Arrow’s Impossibility
Theorem
6. Independence of the irrelevant alternatives:
For any pair a and b in A0 (A0 j A), if a<i b and
a<i* b, for all i, then a<b and a<* b, for all A0.

z A0 is any subset of A, and {<i} y {<i*} are two

sets of individual preferences.
z If individual preferences change but leave
unchanged each i’s individual preferences
between a y b, then social preferences must
keep that a is socially preferred to b.

©2005 Pearson Education, Inc. Chapter 16 44

 Social Choice: Arrow’s Impossibility
Theorem
Indep. irrelevant alternatives (Cont) →Implications:
Consider three alternatives (a,b,c) and two individuals. A change
only in the position of c in the individual orderings (a change of the
preferences) does not affect the social ordering between a
and b (c is the irrelevant alternative in the choice between a and b).

Example: Individual one: (a,b,c)1 → a f1 b

individual two: (b,c,a)2 → b f2 a
Consider the aggregation rule: Voting through orderings: an integer
(a number of points) is assigned to each alternative with the
property that the more preferred alternatives are assigned the
smaller integers. Points are aggregated to compare any pair of
alternatives and the alternative with less points is the social winner.

©2005 Pearson Education, Inc. Chapter 16 45

 Social Choice: Arrow’s Impossibility
Theorem
Example (cont): Then (a=1,b=2,c=3)1 and (b=1,c=2,a=3)2 .
Comparing the pair of alternatives (a,b): a=1+3=4 and b=2+1=3 →
b f a ( b is socially preferred to a).
Consider a change of the individual preferences:
Individual one: (a,c,b)*1 → a f1* b Individual preferences between
individual two: (b,a,c)*2 → b f2* a a and b remain the same

Then: (a=1,c=2,b=3)1 and (b=1,a=2,c=3)2

Socially choosing between (a,b): a=1+2=3 and b=3+1=4 →
a f b (a is now socially preferred to b).
The individual preferences between a and b have not changed but the
social preferences have changed.

The aggregation rule: Voting through orderings is not independent

of the irrelevant alternatives.

©2005 Pearson Education, Inc. Chapter 16 46

 Social Choice: Arrow’s Impossibility
Theorem
Another example showing that the aggregation rule of Voting
through orderings is not independent of the irrelevant
alternatives:
Let A=(a,b,c) and consider again that:
Individual one: (a,b,c)1 → a f1 b
individual two: (b,c,a)2 → b f2 a
and remember that since (a=1,b=2,c=3)1 and (b=1,c=2,a=3)2,
such an aggregation rule produces, when comparing between
(a,b): a=1+3=4 y b=2+1=3 → b f a (b is soc. preferred to a).

Consider the subset of A, A0=(a,b), then (a,b)1 and (b,a)2, and

this rule says that since (a=1,b=2)1 and (b=1,a=2)2, then
a=1+2=3 and b=2+1=3, and socially: a v b.

©2005 Pearson Education, Inc. Chapter 16 47

 Social Choice: Arrow’s Impossibility
Theorem

7. No dictatorship:
There is no individual i* such that for all a and b in
A: a<i*b, implies that a<b.

This property avoids that an individual is

fundamental (decisive) in all choices, regardless of
the other individual preferences.

©2005 Pearson Education, Inc. Chapter 16 48

 Social Choice: Arrow’s Impossibility
Theorem
z Arrow’s Impossibility Theorem:
z If a mechanism of social choice generates a social
ordering satisfying properties 1-6, then it will be a
dictatorship: all social orderings are those of a unique
individual

z The Theorem shows that the desiderable properties of a

social ordering coming from a social choice rule are not
compatible with democracy: there is no “perfect system”
to take social decisions.
z If we use some system, then we will loose some of the
properties defined in 1-7.

©2005 Pearson Education, Inc. Chapter 16 49

 Social Choice: Arrow’s Impossibility
Theorem

How the Social Choice Theory keeps

developing in spite of the non-existence
result?

1. To relax the condition of universality or not

restricted domain
2. To ask only for no-cyclicity, instead of
transitivity.
Etc.

©2005 Pearson Education, Inc. Chapter 16 50