1
EC9A4EC9A4 Social Choice and Social Choice and
VotingVoting Lecture 2Lecture 2
Prof. Francesco SquintaniProf. Francesco Squintani
2
Summary from previous Summary from previous lecturelecture
We have defined the general set up of the We have defined the general set up of the socialsocial
choice problem.choice problem.
We have shown that majority voting is We have shown that majority voting is particularlyparticularly
valuable to choose between two alternatives.valuable to choose between two alternatives.
We have proved Arrow’s theorem: the onlyWe have proved Arrow’s theorem: the only
transitive complete social rule satisfying weaktransitive complete social rule satisfying weak
Pareto, IIA and unrestricted domain is a Pareto, IIA and unrestricted domain is a
dictatorial rule.dictatorial rule.
3
Preview of the lecturePreview of the lecture
We will extend Arrow’ theorem to social We will extend Arrow’ theorem to social choicechoice
functions.functions.
We will introduce the possibility of We will introduce the possibility of interpersonalinterpersonal
comparisons of utility.comparisons of utility.
We describe different concept of social We describe different concept of social welfare: welfare:
the utilitarian Arrowian representation the utilitarian Arrowian representation
and the maximin Rawlsian representation.and the maximin Rawlsian representation.
4
Social Choice FunctionsSocial Choice FunctionsThe Arrow theorem concerns the The Arrow theorem concerns the
impossibility to impossibility to find a social preference order that find a social preference order that
satisfies minimalsatisfies minimalrequirements.requirements.
But the object of social choice may be But the object of social choice may be restricted torestricted to
find simply a socially optimal find simply a socially optimal alternative ratheralternative rather
than to rank all possible alternatives.than to rank all possible alternatives.
We will show that a similar We will show that a similar impossibility resultimpossibility result
holds when considering social choice holds when considering social choice functions.functions.
5
Let Let RR be a subset of the set of all be a subset of the set of all profile of preferences profile of preferences RR = R(1),… = R(1),…R(N).R(N).
A social choice function f defined on A social choice function f defined on RR assigns a assigns a
single element to every profile single element to every profile RR in in RR..
The function f is The function f is monotonicmonotonic if, for any if, for any R R and and R’ R’ s.t.s.t.
the alternative x = f (the alternative x = f (RR) maintains its ) maintains its position position
from from R R to to R’ R’ we have that x = f (we have that x = f (R’R’).).
The alternative x The alternative x maintains its maintains its positionposition from from R R to to R’R’,,
if x R(i) y implies x R’(i) y for all i and if x R(i) y implies x R’(i) y for all i and y. y.
6
The social choice function f is The social choice function f is weakly weakly ParetianParetian if if
y = f (y = f (RR) whenever there is x s.t. x R(i) ) whenever there is x s.t. x R(i) y for all i.y for all i.
An individual n is An individual n is dictatorialdictatorial for f if, for f if, for every for every RR
f (f (RR) ) {x : x R(n) y for all y in X}.{x : x R(n) y for all y in X}.
Theorem.Theorem. Suppose that there are at Suppose that there are at least three least three
distinct alternatives. Then any weakly distinct alternatives. Then any weakly Paretian Paretian
and monotonic social function defined and monotonic social function defined on theon the
whole domain of preferences is whole domain of preferences is dictatorial.dictatorial.
7
Proof.Proof. The proof shows the result as a The proof shows the result as a corollary of corollary of
Arrow impossibility theorem. We Arrow impossibility theorem. We derive a social derive a social
rule F(rule F(RR) from f, using the properties ) from f, using the properties of f on all of f on all RR. .
We show that F satisfies all Arrow We show that F satisfies all Arrow axioms but ND.axioms but ND.
Step 0. Given a subset X’ of X and Step 0. Given a subset X’ of X and RR, , we say that we say that
R’R’ takes X’ to the top from takes X’ to the top from RR, if for , if for every i,every i,
x P’(i) y for all x in X’ and y not in X’;x P’(i) y for all x in X’ and y not in X’; x R(i) y if and only if x R’(i) y for all x, x R(i) y if and only if x R’(i) y for all x,
y in X’.y in X’.
8
Step 1. If both Step 1. If both R’R’ and and R’’R’’ take X’ to the take X’ to the top from top from
RR, then f(, then f(R’R’) = f() = f(R’’R’’).).
In fact, by WP, f(In fact, by WP, f(R’R’) is in X’. Because ) is in X’. Because f(f(R’R’) maintains its position from ) maintains its position from R’R’ to to R’’R’’, f(, f(R’R’) = f() = f(R’’R’’) by monotonicity.) by monotonicity.
Step 2. For every Step 2. For every RR, we let `x F(, we let `x F(RR) y’ if ) y’ if x = f(x = f(R’R’) )
when when R’ R’ is any profile that takes {x,y} is any profile that takes {x,y} to the top of to the top of
profile profile R.R.
By Step 1, F is well defined. By Step 1, F is well defined.
9
Step 3. Because f is weakly Paretian, if Step 3. Because f is weakly Paretian, if R’R’ takes {x,y} takes {x,y}
to the top from to the top from RR, then f(, then f(RR) in {x,y}. By ) in {x,y}. By step 1 and step 1 and
2, either x F(2, either x F(RR) y or y F() y or y F(RR) x (but not ) x (but not both).both).
Hence F is complete.Hence F is complete.
Step 4. Suppose that x F(Step 4. Suppose that x F(RR) y and y F() y and y F(RR) ) z. z.
If If R’’R’’ takes {x,y,z} to the top of takes {x,y,z} to the top of RR, then , then F(F(R’’R’’) in ) in
{x,y,z}, because f is weakly Paretian.{x,y,z}, because f is weakly Paretian.
Say by contradiction that y = f(Say by contradiction that y = f(R’’R’’). ).
Consider Consider R’R’ that takes {x,y} to the top that takes {x,y} to the top from from R’’R’’..
By monotonicity, f(By monotonicity, f(R’R’) = y.) = y.
10
But But R’R’ also takes {x,y} to the top from also takes {x,y} to the top from RR..
This contradicts x F(This contradicts x F(RR)) y. In sum, y = f y. In sum, y = f ((R’’R’’).).
Similarly z = f (Similarly z = f (R’’R’’), by contradiction ), by contradiction with y F(with y F(RR) z.) z.
Thus x = f (Thus x = f (R’’R’’).).
Let Let R’R’ take {x,z} to the top from take {x,z} to the top from R’’R’’..
By monotonicity, x = f (By monotonicity, x = f (R’R’). ).
But But R’R’ also takes {x,z} to the top from also takes {x,z} to the top from RR,,
and hence x F(and hence x F(RR)) z.z.
Hence F(Hence F(RR)) is transitive.is transitive.
11
Step 5. The social rule F(.) satisfies Step 5. The social rule F(.) satisfies Weak Pareto.Weak Pareto.
In fact, if x R(i) y for all i, then x = f In fact, if x R(i) y for all i, then x = f ((R’R’) whenever ) whenever R’R’ takes {x,y} to the takes {x,y} to the top from top from RR..
Thus x F(Thus x F(RR) y, by step 1.) y, by step 1.
Step 6. The social rule F(.) satisfies IIA.Step 6. The social rule F(.) satisfies IIA.
In fact, if In fact, if RR and and R’R’ have the same have the same ordering for x,y, and ordering for x,y, and R’’R’’ takes {x,y} to takes {x,y} to the top of the top of RR, then it also takes {x,y} to , then it also takes {x,y} to the top of the top of R’R’..
Hence, x = f (Hence, x = f (R’’R’’) implies: x F() implies: x F(RR) y and ) y and x F(x F(R’R’) y.) y.
12
Step 7. The social choice function f is Step 7. The social choice function f is dictatorial.dictatorial.
By Arrow theorem, there is an agent i By Arrow theorem, there is an agent i such thatsuch that
for every profile for every profile RR, we have x F(, we have x F(RR) y ) y when x R(i)y.when x R(i)y.
Thus if x R(i) y for all y, then x = f(Thus if x R(i) y for all y, then x = f(RR).).
This final step concludes the proof.This final step concludes the proof.
We have shown how to construct the We have shown how to construct the social rule social rule
F(F(RR) starting from the social choice ) starting from the social choice function f.function f.
Applying Arrow’s theorem, we have Applying Arrow’s theorem, we have shown that shown that
there is a dictator for F(there is a dictator for F(RR), and so f is ), and so f is dictatorial.dictatorial.
13
Interpersonal Interpersonal comparisonscomparisons
We will assume that all social choice We will assume that all social choice functions ffunctions f
satisfy welfarism, i.e. U, P and IIA, satisfy welfarism, i.e. U, P and IIA, and continuity.and continuity.
Hence there is a continuous function Hence there is a continuous function W such that:W such that:
V(x) V(x) >> V(y) if and only if W( V(y) if and only if W(uu(x)) (x)) >> W(W(uu(y)).(y)).
The welfare function depends only on The welfare function depends only on the utility the utility
ranking, not on how the ranking ranking, not on how the ranking comes about.comes about.
14
Under Arrow axiom, utilities are Under Arrow axiom, utilities are measured along measured along
an an ordinal scale, ordinal scale, and are and are non-comparablenon-comparable across across
individuals. individuals.
Specifically, the function W aggregates Specifically, the function W aggregates thethe
preferences (upreferences (uii))i=1,…,ni=1,…,n if and only if W if and only if W aggregates aggregates
the preferences (v(uthe preferences (v(uii))))i=1,…,ni=1,…,n, for all , for all increasing increasing
transformation vtransformation vi i (u(uii), for any i, ), for any i, independently across i.independently across i.
We modify the framework to allow for We modify the framework to allow for cardinal cardinal
comparisons of utility, and comparability comparisons of utility, and comparability acrossacross
individuals.individuals.
15
Suppose that preferences are fully Suppose that preferences are fully comparable but comparable but
measured on the ordinal scale.measured on the ordinal scale.The social ranking V must be invariant to The social ranking V must be invariant to
arbitrary,arbitrary,but common, increasing transformations but common, increasing transformations
vvii applied applied
to every individual’s utility function uto every individual’s utility function uii..
Specifically, the function W aggregates Specifically, the function W aggregates thethe
preferences (upreferences (uii))i=1,…,ni=1,…,n if and only if W if and only if W
aggregates the preferences (v(uaggregates the preferences (v(uii))))i=1,…,ni=1,…,n, , for all for all
increasing transformation vincreasing transformation vi i (.), such that (.), such that v’v’i i is is
constant across i.constant across i.
16
Suppose that preferences are fully Suppose that preferences are fully comparable andcomparable and
measured on the cardinal scale.measured on the cardinal scale.The social ranking V must be invariant to The social ranking V must be invariant to
increasing,increasing,
linear transformations v(ulinear transformations v(uii)= a)= aii+bu+buii, , where b is where b is
common to every individual.common to every individual.
Specifically, the function W aggregates Specifically, the function W aggregates thethe
preferences (upreferences (uii))i=1,…,ni=1,…,n if and only if W if and only if W aggregates theaggregates the
preferences (v(upreferences (v(uii))))i=1,…,ni=1,…,n, for all , for all transformation vtransformation vii(.),(.),
such that vsuch that vii(u(uii)= a)= aii+bu+buii, with b>0., with b>0.
17
HE: Let HE: Let uu and and u’u’ be two distinct utility be two distinct utility vectors. vectors.
Suppose that uSuppose that uk k = u’= u’k k for all k other than i for all k other than i and j.and j.
If uIf ui i > u’> u’i i > u’> u’j j > u> uj j , then W(, then W(u’u’) > W() > W(uu).).
Condition HE states that the society has a Condition HE states that the society has a preferencepreference
towards decreasing the dispersion of towards decreasing the dispersion of utilities acrossutilities across
individuals.individuals.
Rawlsian FormRawlsian Form
18
AN: The social rule W is AN: The social rule W is anonymousanonymous if for every if for every
permutation p, W(upermutation p, W(u11, …, u, …, uNN)=F(u)=F(up(1)p(1),…u,…up(N)p(N)).).
Theorem: Suppose that preferences are fullyTheorem: Suppose that preferences are fully
comparable and measured on the ordinal scale. comparable and measured on the ordinal scale.
The social welfare function W satisfies Weak The social welfare function W satisfies Weak
Pareto, Anonymity and Hammond equality Pareto, Anonymity and Hammond equality
if and only if it takes the Rawlsian form if and only if it takes the Rawlsian form
W(u)=min{uW(u)=min{u11, …, u, …, uNN}.}.
19
ProofProof. It is easy to see that the function. It is easy to see that the function
W(u)=min{uW(u)=min{u11, …, u, …, uNN} satisfies Weak } satisfies Weak
Pareto, Anonymity and Hammond Pareto, Anonymity and Hammond equality.equality.
To show the converse, we will see only To show the converse, we will see only the case the case
for N=2.for N=2.
Consider aConsider autility index utility index uu,,
with uwith u1 1 > u> u22
u1
u2
u
u*I
II
III
IV
20
By anonymity, the utility profile By anonymity, the utility profile u* must be ranked
in the same way as u: hence W(u) = W(u*). By Weak Pareto, all u’ such that u’ > u or u’ > u* must be such that W(u’) > W(u).
Hence the whole areain blue is s.t.W(u’) > W(u).
u1
u2
u
u*
I
II
III
IV
21
By Weak Pareto, all u’ such that u’ < u or
u’ < u* must be such that W(u’) < W(u).
Hence the whole area in green is such that
W(u’) < W(u).
u1
u2
u
u*I
II
III
IV
22
Pick a point u’ in zone III. To be in III, it must be
that u2 < u’2 < u’1 < u1
Every linear transform v such that vi(ui)=ui
yields: u2 < v2(u’2) < v1 (u’1) < u1
This concludesthat all pointsin III are rankedthe same waywrt to u. u1
u2
u
u*I
IIIII
IV
u’
23
To be in III, it must be that u2 < u’2 < u’1 < u1
Hammond equality implies that W(u’) > W(u).
u1
u2
u
u*I
IIIII
IV
u’
24
By anonymity, the ranking of each u’ in III relative
to u must be the same as the ranking of any utility
vector u’’ in II: W(u’’)>W(u).
u1
u2
u
u*I
IIIII
IV
u’
25
Any linear transform v such that v1(u’1) = u1,
v2(u’2) = u2, yields:
W (v(u)) < W (v1 (u’1), v2(u’2)) = W (u).
Hence all the utility vectors u’’ in IV are ranked
opposite to all utility vectors u’ in III, relative to u. u1
u2
u
u*I
IIIII
IV
u’
u’’
26
Hence W(u) > W(u’’) for all u’’ in IV, and, by
anonymity, W(u) > W(u’’) for all u’’ in I.
u1
u2
u
I
IIIII
IV u’’
u’’
27
We conclude that V(II) and V(III) > W(u) > V(I) and
V(IV).We are left to consider the boundaries
of these sets.
u1
u2
u
u*I
IIIII
IV
u’
28
Because W is continuous, the boundaries opposite
to each other, relative to u must be indifferent to u’.
Thus, the boundaries between II and III and the
blue set must be better than u in W terms.The boundariesbetween I and IVand the green setmust be indifferent to u. u1
u2
u
u*I
IIIII
IV
u’
29
We have obtained the Rawlsian indifference
curves, where W (u) = min {u1 , u2}.
u1
u2
u
u*
IIIII
IV
u
30
Utilitarian FormUtilitarian Form
Theorem: Suppose that preferences are Theorem: Suppose that preferences are
fully comparable and measured on the fully comparable and measured on the
cardinal scale. cardinal scale. The social welfare function W satisfies The social welfare function W satisfies
Weak Weak Pareto and Anonymity if and only if it Pareto and Anonymity if and only if it
takes the takes the utilitarian form:utilitarian form:
W(u)= uW(u)= u11 +… + u +… + uNN..
31
ProofProof. It is easy to see that the . It is easy to see that the functionfunction
W(u)= uW(u)= u11+ …+ u+ …+ uNN, satisfies Weak , satisfies Weak Pareto and Pareto and
Anonymity.Anonymity.To show the converse, we will see only To show the converse, we will see only
the case the case for N=2.for N=2.
Consider aConsider autility index utility index uu,,
with uwith u1 1 = u= u22
u1
u2
u’
u*
u
32
Define the constant k = uDefine the constant k = u11 + u + u22..
Consider the locus k(u) = {(uConsider the locus k(u) = {(u11, u, u22) : k ) : k = u= u11 + u + u22}. }.
For any vector u’ on k(u), the vector For any vector u’ on k(u), the vector u* such that u* such that
u* = (u’u* = (u’2 2 , u’, u’11) is also on k(u).) is also on k(u).
By Anonymity,By Anonymity,
W(u’)=W(u*).W(u’)=W(u*).u1
u2
u’
u*
u
33
Suppose now that W(u) > W(u’). Suppose now that W(u) > W(u’).
Under CS/IC, this ranking must be Under CS/IC, this ranking must be invariant toinvariant to
transformation vtransformation vii(u(uii) = a) = ai i + bu+ buii
u1
u2
u’
u*
u
34
Let vLet vii(u(uii) = (u) = (uii - u’ - u’ii)+ u)+ ui i for i=1,2.for i=1,2.
Hence, (vHence, (v11(u’(u’11), v), vii(u’(u’ii)) = u and (v)) = u and (v11(u(u11), ), vvii(u(uii)) = u*.)) = u*.
If W(u) > W(u’), then W(u*) > W(u), If W(u) > W(u’), then W(u*) > W(u), which which
contradicts W(u’) = W(u*). contradicts W(u’) = W(u*).
u1
u2
u’
u*
u
35
If W(u) < W(u’), then W(u*) < W(u), If W(u) < W(u’), then W(u*) < W(u), which which
contradicts W(u’) = W(u*). contradicts W(u’) = W(u*).
Hence we conclude that W(u)=W(u’) Hence we conclude that W(u)=W(u’) for all for all
vectors u’ on k(u).vectors u’ on k(u).
u1
u2
u’
u*
u
36
By Weak Pareto, each vector u’’ to the By Weak Pareto, each vector u’’ to the north-east north-east
of a vector u’ on k(u) is strictly of a vector u’ on k(u) is strictly preferred to u.preferred to u.
Thus W(u’’) > W(u) for u’’ such that Thus W(u’’) > W(u) for u’’ such that
u’’u’’11+ u’’+ u’’2 2 > u> u11+ u+ u2 .2 .
u1
u2
u
37
Similarly, W(u’’) < W(u) for u’’ such Similarly, W(u’’) < W(u) for u’’ such that that
u’’u’’11+ u’’+ u’’2 2 < u< u11+ u+ u2 2
u1
u2
u
38
We concluded that the indifference We concluded that the indifference curve of anycurve of any
vector u is k(u) = {(uvector u is k(u) = {(u11, u, u22) : k = u) : k = u11 + + uu22}. }.
HenceHence
W(u) = uW(u) = u11+ u+ u2 2
IndifferenceIndifference
curves arecurves are
straight lines straight lines
of slope -1.of slope -1.
u1
u2
uu
39
If we drop the axiom of Anonymity, the If we drop the axiom of Anonymity, the full range full range
of generalized utilitarian orderings is of generalized utilitarian orderings is allowed. allowed.
These are the linear social welfare These are the linear social welfare functionsfunctions
W(u)= aW(u)= a1 1 uu1 1 +…+ a+…+ aN N uuNN, with a, with ai i >> 0 for 0 for all i, all i,
and aand aii > 0 > 0
for some i.for some i.
IndifferenceIndifference
curves are linescurves are lines
of negativeof negative
slope.slope.
u1
u2
u
40
The maximin Rawlsian form and the utilitarian The maximin Rawlsian form and the utilitarian form both belong to constant elasticity classform both belong to constant elasticity classwith the formula:with the formula:
W = ( uW = ( u11 +… + u +… + uNN
) )
where 0 = where 0 = and and is the constantis the constantelasticity of social substitution between any elasticity of social substitution between any
pair pair of individuals.of individuals.
As As approachesapproachesW approaches the W approaches the utilitarian formutilitarian form
As As approaches –infinity,approaches –infinity,W approaches the W approaches the Rawlsian form.Rawlsian form.
41
A Theory of JusticeA Theory of Justice
Behind a ``veil of ignorance’’, an individual Behind a ``veil of ignorance’’, an individual doesdoes
not know which position she will take in a not know which position she will take in a society.society.
Will she be rich or poor, successful or Will she be rich or poor, successful or
unsuccessful?unsuccessful?
If she assigns equal probability to any of theIf she assigns equal probability to any of the
possible economic and social identities that possible economic and social identities that
exist in the society, a rational evaluation wouldexist in the society, a rational evaluation would
42
evaluate welfare according to the expected utilityevaluate welfare according to the expected utility
[u[u11(x) + … + u(x) + … + uNN(x) ]/N(x) ]/N
This is equivalent to adopt the utilitarian This is equivalent to adopt the utilitarian criterion:criterion:
W(u(x)) = uW(u(x)) = u11(x) + … + u(x) + … + uNN(x).(x).
But the approach is also consistent with every But the approach is also consistent with every CES CES
form, embodying different degrees of risk form, embodying different degrees of risk aversion.aversion.
Consider the positive transformation Consider the positive transformation
vvii(x) = - u(x) = - uii(x)(x)-a -a with a>0. with a>0.
43
Suppose that uSuppose that uii(x) represents utility over social (x) represents utility over social
states “with certainty,” whereas vstates “with certainty,” whereas vii(x) (x) represents represents
utility over social states “with uncertainty.”utility over social states “with uncertainty.”
In the form vIn the form vii(x) = - u(x) = - uii(x) (x) -a -a , a>0 represents the , a>0 represents the
degree of risk aversion.degree of risk aversion.
Suppose that the social welfare function is Suppose that the social welfare function is given by given by
the expected utility:the expected utility:
W = [vW = [v11(x) +…+ v(x) +…+ vNN(x)]/N (x)]/N
= [-u= [-u11(x)(x)-a -a - …-u- …-uNN(x)(x)-a-a]/N]/N
44
Because the monotonic transformation Because the monotonic transformation
W = ( -uW = ( -u11(x)(x)-a -a - …-u- …-uNN(x)(x)-a -a ))-1/a-1/a
equivalently represents welfare, we obtain equivalently represents welfare, we obtain that anythat any
CES form is compatible with the expectedCES form is compatible with the expectedutility formulation behind a veil of ignorance.utility formulation behind a veil of ignorance.
The extreme risk aversion case of CES is The extreme risk aversion case of CES is
W = min {uW = min {u11(x), …, u(x), …, uNN(x)}, the Rawlsian form(x)}, the Rawlsian formthat describes a social concern for the agent that describes a social concern for the agent
with with the lowest utility.the lowest utility.
45
ConclusionConclusion
We have extended Arrow’ theorem to social We have extended Arrow’ theorem to social
choice functions.choice functions.
We have introduced the possibility ofWe have introduced the possibility of
interpersonal comparisons of utility.interpersonal comparisons of utility.
We have described different concept of We have described different concept of social social
welfare: the utilitarian Arrowian welfare: the utilitarian Arrowian representation andrepresentation and
the maximin Rawlsian representation.the maximin Rawlsian representation.
46
Preview of the next Preview of the next lecturelecture
We will introduce single-peaked We will introduce single-peaked utilities.utilities.
We will prove Black’s theorem: We will prove Black’s theorem: Majority voting isMajority voting is
socially fair when utilities are single-socially fair when utilities are single-peaked.peaked.
We will prove Median Voter We will prove Median Voter Convergence in theConvergence in the
Downsian model of elections.Downsian model of elections.
We will introduce the probabilistic We will introduce the probabilistic voting model.voting model.