arrow's theorem for incomplete...

190
Arrow’s Theorem for Incomplete Relations R. D. Maddux Department of Mathematics Iowa State University Ames, Iowa, USA Ames, Iowa Tuesdays, October 9 and 16, 2012 R. D. Maddux Arrow’s Theorem for Incomplete Relations

Upload: others

Post on 25-Jun-2020

7 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem for Incomplete Relations

R. D. Maddux

Department of MathematicsIowa State University

Ames, Iowa, USA

Ames, IowaTuesdays, October 9 and 16, 2012

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 2: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 3: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Abstract

Let U be a set with three or more elements, let W be the set ofweak orderings of U, let T be the set of total orderings of U, andlet f be an n-ary function mapping Wn to W. Arrow’sImpossibility Theorem asserts that if f satisfiesArrow’s Condition P (“Pareto”) and Condition 3 (“independenceof irrelevant alternatives”) then f is a projection function on totalorderings, i. e., there is some k ∈ {1, . . . , n} such thatf (R1, . . . ,Rn) = Rk for all total orderings R1, . . . ,Rn ∈ T .Th. If a transitive-valued multivariate relational operator fsatisfies versions of Arrow’s Conditions P and 3, and maps allprofiles from a diverse set R of binary relations on U to transitiverelations on U, then f must be the unanimous consent function forsome set of input variables, and if R is very diverse then f is aprojection function.Cor. Characterizations of intersection and projection functions;Arrow’s Theorem.

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 4: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

History

Kenneth J. Arrow (Nobel, 1973, Econonics)

learned the Calculus of Relations from Alfred Tarski in 1940

applied it to his problems in economics

obtained Arrow’s Impossibility Theorem (Arrow 1950, 1963)

created Social Choice Theory

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 5: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

History

Kenneth J. Arrow (Nobel, 1973, Econonics)

learned the Calculus of Relations from Alfred Tarski in 1940

applied it to his problems in economics

obtained Arrow’s Impossibility Theorem (Arrow 1950, 1963)

created Social Choice Theory

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 6: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

History

Kenneth J. Arrow (Nobel, 1973, Econonics)

learned the Calculus of Relations from Alfred Tarski in 1940

applied it to his problems in economics

obtained Arrow’s Impossibility Theorem (Arrow 1950, 1963)

created Social Choice Theory

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 7: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

History

Kenneth J. Arrow (Nobel, 1973, Econonics)

learned the Calculus of Relations from Alfred Tarski in 1940

applied it to his problems in economics

obtained Arrow’s Impossibility Theorem (Arrow 1950, 1963)

created Social Choice Theory

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 8: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

History

Kenneth J. Arrow (Nobel, 1973, Econonics)

learned the Calculus of Relations from Alfred Tarski in 1940

applied it to his problems in economics

obtained Arrow’s Impossibility Theorem (Arrow 1950, 1963)

created Social Choice Theory

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 9: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

History

Kenneth J. Arrow (Nobel, 1973, Econonics)

learned the Calculus of Relations from Alfred Tarski in 1940

applied it to his problems in economics

obtained Arrow’s Impossibility Theorem (Arrow 1950, 1963)

created Social Choice Theory

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 10: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

History, cont.

spring semester of 1940

Tarski was a Visiting Professor of Philosophy at the CityCollege of New York

class attended by Kenneth Arrow:

“It was a great course, Calculus of Relations. His organizationwas beautiful — I could tell that immediately — and he wasthorough. In fact what I learned from him played a role in myown later work — not so much the particular theorems butthe language of relations was immediately applicable toeconomics. I could express my problems in those terms”

Arrow proofread Tarski’s Introduction to Logic for no pay atTarski’s request:

“I also owe many thanks to Mr. K. J. Arrow for his help inreading proofs.”

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 11: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

History, cont.

spring semester of 1940

Tarski was a Visiting Professor of Philosophy at the CityCollege of New York

class attended by Kenneth Arrow:

“It was a great course, Calculus of Relations. His organizationwas beautiful — I could tell that immediately — and he wasthorough. In fact what I learned from him played a role in myown later work — not so much the particular theorems butthe language of relations was immediately applicable toeconomics. I could express my problems in those terms”

Arrow proofread Tarski’s Introduction to Logic for no pay atTarski’s request:

“I also owe many thanks to Mr. K. J. Arrow for his help inreading proofs.”

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 12: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

History, cont.

spring semester of 1940

Tarski was a Visiting Professor of Philosophy at the CityCollege of New York

class attended by Kenneth Arrow:

“It was a great course, Calculus of Relations. His organizationwas beautiful — I could tell that immediately — and he wasthorough. In fact what I learned from him played a role in myown later work — not so much the particular theorems butthe language of relations was immediately applicable toeconomics. I could express my problems in those terms”

Arrow proofread Tarski’s Introduction to Logic for no pay atTarski’s request:

“I also owe many thanks to Mr. K. J. Arrow for his help inreading proofs.”

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 13: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

History, cont.

spring semester of 1940

Tarski was a Visiting Professor of Philosophy at the CityCollege of New York

class attended by Kenneth Arrow:

“It was a great course, Calculus of Relations. His organizationwas beautiful — I could tell that immediately — and he wasthorough. In fact what I learned from him played a role in myown later work — not so much the particular theorems butthe language of relations was immediately applicable toeconomics. I could express my problems in those terms”

Arrow proofread Tarski’s Introduction to Logic for no pay atTarski’s request:

“I also owe many thanks to Mr. K. J. Arrow for his help inreading proofs.”

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 14: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

History, cont.

spring semester of 1940

Tarski was a Visiting Professor of Philosophy at the CityCollege of New York

class attended by Kenneth Arrow:

“It was a great course, Calculus of Relations. His organizationwas beautiful — I could tell that immediately — and he wasthorough. In fact what I learned from him played a role in myown later work — not so much the particular theorems butthe language of relations was immediately applicable toeconomics. I could express my problems in those terms”

Arrow proofread Tarski’s Introduction to Logic for no pay atTarski’s request:

“I also owe many thanks to Mr. K. J. Arrow for his help inreading proofs.”

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 15: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

History, cont.

spring semester of 1940

Tarski was a Visiting Professor of Philosophy at the CityCollege of New York

class attended by Kenneth Arrow:

“It was a great course, Calculus of Relations. His organizationwas beautiful — I could tell that immediately — and he wasthorough. In fact what I learned from him played a role in myown later work — not so much the particular theorems butthe language of relations was immediately applicable toeconomics. I could express my problems in those terms”

Arrow proofread Tarski’s Introduction to Logic for no pay atTarski’s request:

“I also owe many thanks to Mr. K. J. Arrow for his help inreading proofs.”

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 16: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Goals

Recast Arrow’s Theorem as a contribution to the Calculus ofRelations

What does the proof of Arrow’s Theorem show?

Avoid Axiom I

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 17: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Goals

Recast Arrow’s Theorem as a contribution to the Calculus ofRelations

What does the proof of Arrow’s Theorem show?

Avoid Axiom I

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 18: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Goals

Recast Arrow’s Theorem as a contribution to the Calculus ofRelations

What does the proof of Arrow’s Theorem show?

Avoid Axiom I

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 19: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

References

Arrow, K.J. (1950), ”A difficulty in the concept of social welfare”,Journal of Political Economy 58:328-246.Arrow, K.J. (1951), Social Choice and Individual Values, 1stedition (Wiley, New York).Arrow, K.J. (1963), Social Choice and Individual Values, 2ndedition (Wiley, New York) (all subsequent quotations from here)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 20: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

References

Arrow, K.J. (1950), ”A difficulty in the concept of social welfare”,Journal of Political Economy 58:328-246.Arrow, K.J. (1951), Social Choice and Individual Values, 1stedition (Wiley, New York).Arrow, K.J. (1963), Social Choice and Individual Values, 2ndedition (Wiley, New York) (all subsequent quotations from here)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 21: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

References

Arrow, K.J. (1950), ”A difficulty in the concept of social welfare”,Journal of Political Economy 58:328-246.Arrow, K.J. (1951), Social Choice and Individual Values, 1stedition (Wiley, New York).Arrow, K.J. (1963), Social Choice and Individual Values, 2ndedition (Wiley, New York) (all subsequent quotations from here)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 22: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

References

Arrow, K.J. (1950), ”A difficulty in the concept of social welfare”,Journal of Political Economy 58:328-246.Arrow, K.J. (1951), Social Choice and Individual Values, 1stedition (Wiley, New York).Arrow, K.J. (1963), Social Choice and Individual Values, 2ndedition (Wiley, New York) (all subsequent quotations from here)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 23: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — informally

Th. There is no fair social welfare function.Th. There is no multivariable relational operator on weak orderingsof a set with at least three elements that is non-dictatorial,independent of irrevelevant alternatives, and Pareto.

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 24: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — informally

Th. There is no fair social welfare function.Th. There is no multivariable relational operator on weak orderingsof a set with at least three elements that is non-dictatorial,independent of irrevelevant alternatives, and Pareto.

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 25: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — informally

Th. There is no fair social welfare function.Th. There is no multivariable relational operator on weak orderingsof a set with at least three elements that is non-dictatorial,independent of irrevelevant alternatives, and Pareto.

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 26: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — more details

Hypotheses:

U is a set with three or more elements

W is the set of weak orderings of U

T is the set of total orderings of U

2 ≤ n < ω

f is an n-ary function mapping Wn to W(f is a social welfare function)

f satisfies Arrow’s Condition P (“Pareto”)and Arrow’s Condition 3 (“independence of irrelevantalternatives”)

Conclusion:

f is a projection function on total orderings, i. e.,for some “dictator” k ∈ {1, . . . , n},

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn ∈ TR. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 27: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — more details

Hypotheses:

U is a set with three or more elements

W is the set of weak orderings of U

T is the set of total orderings of U

2 ≤ n < ω

f is an n-ary function mapping Wn to W(f is a social welfare function)

f satisfies Arrow’s Condition P (“Pareto”)and Arrow’s Condition 3 (“independence of irrelevantalternatives”)

Conclusion:

f is a projection function on total orderings, i. e.,for some “dictator” k ∈ {1, . . . , n},

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn ∈ TR. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 28: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — more details

Hypotheses:

U is a set with three or more elements

W is the set of weak orderings of U

T is the set of total orderings of U

2 ≤ n < ω

f is an n-ary function mapping Wn to W(f is a social welfare function)

f satisfies Arrow’s Condition P (“Pareto”)and Arrow’s Condition 3 (“independence of irrelevantalternatives”)

Conclusion:

f is a projection function on total orderings, i. e.,for some “dictator” k ∈ {1, . . . , n},

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn ∈ TR. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 29: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — more details

Hypotheses:

U is a set with three or more elements

W is the set of weak orderings of U

T is the set of total orderings of U

2 ≤ n < ω

f is an n-ary function mapping Wn to W(f is a social welfare function)

f satisfies Arrow’s Condition P (“Pareto”)and Arrow’s Condition 3 (“independence of irrelevantalternatives”)

Conclusion:

f is a projection function on total orderings, i. e.,for some “dictator” k ∈ {1, . . . , n},

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn ∈ TR. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 30: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — more details

Hypotheses:

U is a set with three or more elements

W is the set of weak orderings of U

T is the set of total orderings of U

2 ≤ n < ω

f is an n-ary function mapping Wn to W(f is a social welfare function)

f satisfies Arrow’s Condition P (“Pareto”)and Arrow’s Condition 3 (“independence of irrelevantalternatives”)

Conclusion:

f is a projection function on total orderings, i. e.,for some “dictator” k ∈ {1, . . . , n},

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn ∈ TR. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 31: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — more details

Hypotheses:

U is a set with three or more elements

W is the set of weak orderings of U

T is the set of total orderings of U

2 ≤ n < ω

f is an n-ary function mapping Wn to W(f is a social welfare function)

f satisfies Arrow’s Condition P (“Pareto”)and Arrow’s Condition 3 (“independence of irrelevantalternatives”)

Conclusion:

f is a projection function on total orderings, i. e.,for some “dictator” k ∈ {1, . . . , n},

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn ∈ TR. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 32: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — more details

Hypotheses:

U is a set with three or more elements

W is the set of weak orderings of U

T is the set of total orderings of U

2 ≤ n < ω

f is an n-ary function mapping Wn to W(f is a social welfare function)

f satisfies Arrow’s Condition P (“Pareto”)and Arrow’s Condition 3 (“independence of irrelevantalternatives”)

Conclusion:

f is a projection function on total orderings, i. e.,for some “dictator” k ∈ {1, . . . , n},

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn ∈ TR. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 33: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — more details

Hypotheses:

U is a set with three or more elements

W is the set of weak orderings of U

T is the set of total orderings of U

2 ≤ n < ω

f is an n-ary function mapping Wn to W(f is a social welfare function)

f satisfies Arrow’s Condition P (“Pareto”)and Arrow’s Condition 3 (“independence of irrelevantalternatives”)

Conclusion:

f is a projection function on total orderings, i. e.,for some “dictator” k ∈ {1, . . . , n},

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn ∈ TR. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 34: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — more details

Hypotheses:

U is a set with three or more elements

W is the set of weak orderings of U

T is the set of total orderings of U

2 ≤ n < ω

f is an n-ary function mapping Wn to W(f is a social welfare function)

f satisfies Arrow’s Condition P (“Pareto”)and Arrow’s Condition 3 (“independence of irrelevantalternatives”)

Conclusion:

f is a projection function on total orderings, i. e.,for some “dictator” k ∈ {1, . . . , n},

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn ∈ TR. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 35: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — more details

Hypotheses:

U is a set with three or more elements

W is the set of weak orderings of U

T is the set of total orderings of U

2 ≤ n < ω

f is an n-ary function mapping Wn to W(f is a social welfare function)

f satisfies Arrow’s Condition P (“Pareto”)and Arrow’s Condition 3 (“independence of irrelevantalternatives”)

Conclusion:

f is a projection function on total orderings, i. e.,for some “dictator” k ∈ {1, . . . , n},

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn ∈ TR. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 36: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — more details

Hypotheses:

U is a set with three or more elements

W is the set of weak orderings of U

T is the set of total orderings of U

2 ≤ n < ω

f is an n-ary function mapping Wn to W(f is a social welfare function)

f satisfies Arrow’s Condition P (“Pareto”)and Arrow’s Condition 3 (“independence of irrelevantalternatives”)

Conclusion:

f is a projection function on total orderings, i. e.,for some “dictator” k ∈ {1, . . . , n},

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn ∈ TR. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 37: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem — more details

Hypotheses:

U is a set with three or more elements

W is the set of weak orderings of U

T is the set of total orderings of U

2 ≤ n < ω

f is an n-ary function mapping Wn to W(f is a social welfare function)

f satisfies Arrow’s Condition P (“Pareto”)and Arrow’s Condition 3 (“independence of irrelevantalternatives”)

Conclusion:

f is a projection function on total orderings, i. e.,for some “dictator” k ∈ {1, . . . , n},

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn ∈ TR. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 38: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definitions from the Calculus of Relations

A binary relation R ⊆ U2 = {〈x , y〉 : x , y ∈ U}The diversity relation 0

,= {〈x , y〉 : x , y ∈ U, x 6= y}

The identity relation 1,= {〈x , x〉 : x ∈ U}

xRy iff 〈x , y〉 ∈ RThe complement of R ⊆ U2 is

R = {〈x , y〉 : x , y ∈ U ∧ 〈x , y〉 /∈ R}

for all x , y ∈ U, xRy iff ¬xRyThe relative product of R and R ′ is

R ;R ′ = {〈x , z〉 : ∃y (xRy ∧ yR ′z)}

The converse of R is

R−1 = {〈y , x〉 : xRy}

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 39: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definitions from the Calculus of Relations

A binary relation R ⊆ U2 = {〈x , y〉 : x , y ∈ U}The diversity relation 0

,= {〈x , y〉 : x , y ∈ U, x 6= y}

The identity relation 1,= {〈x , x〉 : x ∈ U}

xRy iff 〈x , y〉 ∈ RThe complement of R ⊆ U2 is

R = {〈x , y〉 : x , y ∈ U ∧ 〈x , y〉 /∈ R}

for all x , y ∈ U, xRy iff ¬xRyThe relative product of R and R ′ is

R ;R ′ = {〈x , z〉 : ∃y (xRy ∧ yR ′z)}

The converse of R is

R−1 = {〈y , x〉 : xRy}

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 40: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definitions from the Calculus of Relations

A binary relation R ⊆ U2 = {〈x , y〉 : x , y ∈ U}The diversity relation 0

,= {〈x , y〉 : x , y ∈ U, x 6= y}

The identity relation 1,= {〈x , x〉 : x ∈ U}

xRy iff 〈x , y〉 ∈ RThe complement of R ⊆ U2 is

R = {〈x , y〉 : x , y ∈ U ∧ 〈x , y〉 /∈ R}

for all x , y ∈ U, xRy iff ¬xRyThe relative product of R and R ′ is

R ;R ′ = {〈x , z〉 : ∃y (xRy ∧ yR ′z)}

The converse of R is

R−1 = {〈y , x〉 : xRy}

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 41: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definitions from the Calculus of Relations

A binary relation R ⊆ U2 = {〈x , y〉 : x , y ∈ U}The diversity relation 0

,= {〈x , y〉 : x , y ∈ U, x 6= y}

The identity relation 1,= {〈x , x〉 : x ∈ U}

xRy iff 〈x , y〉 ∈ RThe complement of R ⊆ U2 is

R = {〈x , y〉 : x , y ∈ U ∧ 〈x , y〉 /∈ R}

for all x , y ∈ U, xRy iff ¬xRyThe relative product of R and R ′ is

R ;R ′ = {〈x , z〉 : ∃y (xRy ∧ yR ′z)}

The converse of R is

R−1 = {〈y , x〉 : xRy}

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 42: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definitions from the Calculus of Relations

A binary relation R ⊆ U2 = {〈x , y〉 : x , y ∈ U}The diversity relation 0

,= {〈x , y〉 : x , y ∈ U, x 6= y}

The identity relation 1,= {〈x , x〉 : x ∈ U}

xRy iff 〈x , y〉 ∈ RThe complement of R ⊆ U2 is

R = {〈x , y〉 : x , y ∈ U ∧ 〈x , y〉 /∈ R}

for all x , y ∈ U, xRy iff ¬xRyThe relative product of R and R ′ is

R ;R ′ = {〈x , z〉 : ∃y (xRy ∧ yR ′z)}

The converse of R is

R−1 = {〈y , x〉 : xRy}

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 43: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definitions from the Calculus of Relations

A binary relation R ⊆ U2 = {〈x , y〉 : x , y ∈ U}The diversity relation 0

,= {〈x , y〉 : x , y ∈ U, x 6= y}

The identity relation 1,= {〈x , x〉 : x ∈ U}

xRy iff 〈x , y〉 ∈ RThe complement of R ⊆ U2 is

R = {〈x , y〉 : x , y ∈ U ∧ 〈x , y〉 /∈ R}

for all x , y ∈ U, xRy iff ¬xRyThe relative product of R and R ′ is

R ;R ′ = {〈x , z〉 : ∃y (xRy ∧ yR ′z)}

The converse of R is

R−1 = {〈y , x〉 : xRy}

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 44: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definitions from the Calculus of Relations

A binary relation R ⊆ U2 = {〈x , y〉 : x , y ∈ U}The diversity relation 0

,= {〈x , y〉 : x , y ∈ U, x 6= y}

The identity relation 1,= {〈x , x〉 : x ∈ U}

xRy iff 〈x , y〉 ∈ RThe complement of R ⊆ U2 is

R = {〈x , y〉 : x , y ∈ U ∧ 〈x , y〉 /∈ R}

for all x , y ∈ U, xRy iff ¬xRyThe relative product of R and R ′ is

R ;R ′ = {〈x , z〉 : ∃y (xRy ∧ yR ′z)}

The converse of R is

R−1 = {〈y , x〉 : xRy}

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 45: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definitions from the Calculus of Relations

A binary relation R ⊆ U2 = {〈x , y〉 : x , y ∈ U}The diversity relation 0

,= {〈x , y〉 : x , y ∈ U, x 6= y}

The identity relation 1,= {〈x , x〉 : x ∈ U}

xRy iff 〈x , y〉 ∈ RThe complement of R ⊆ U2 is

R = {〈x , y〉 : x , y ∈ U ∧ 〈x , y〉 /∈ R}

for all x , y ∈ U, xRy iff ¬xRyThe relative product of R and R ′ is

R ;R ′ = {〈x , z〉 : ∃y (xRy ∧ yR ′z)}

The converse of R is

R−1 = {〈y , x〉 : xRy}

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 46: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definitions from the Calculus of Relations

A binary relation R ⊆ U2 = {〈x , y〉 : x , y ∈ U}The diversity relation 0

,= {〈x , y〉 : x , y ∈ U, x 6= y}

The identity relation 1,= {〈x , x〉 : x ∈ U}

xRy iff 〈x , y〉 ∈ RThe complement of R ⊆ U2 is

R = {〈x , y〉 : x , y ∈ U ∧ 〈x , y〉 /∈ R}

for all x , y ∈ U, xRy iff ¬xRyThe relative product of R and R ′ is

R ;R ′ = {〈x , z〉 : ∃y (xRy ∧ yR ′z)}

The converse of R is

R−1 = {〈y , x〉 : xRy}

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 47: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

More definitions from the Calculus of Relations

Binary relation R ⊆ U2 is

transitive if ∀x ,y ,z∈U (xRy ∧ yRz ⇒ xRz) iff R ;R ⊆ R

co-transitive if R is transitive

reflexive if ∀x∈U (xRx) iff 1, ⊆ R

connected if ∀x ,y∈U (x 6= y ⇒ xRy ∨ yRx) iff 0, ⊆ R ∪ R−1

complete if ∀x ,y∈U (xRy ∨ yRx) iff 1 = R ∪ R−1

symmetric if ∀x ,y∈U (xRy ⇒ yRx) iff R = R−1

anti-symmetric if ∀x ,y∈U (xRy ⇒ (x = y ∨ yRx)) iffR ∩ R−1 ⊆ 1

,

an equivalence relation if R is transitive, symmetric, andreflexive

a weak ordering if R is transitive, reflexive, and connected

a total ordering if R is transitive, reflexive, connected, andanti-symmetric

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 48: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

More definitions from the Calculus of Relations

Binary relation R ⊆ U2 is

transitive if ∀x ,y ,z∈U (xRy ∧ yRz ⇒ xRz) iff R ;R ⊆ R

co-transitive if R is transitive

reflexive if ∀x∈U (xRx) iff 1, ⊆ R

connected if ∀x ,y∈U (x 6= y ⇒ xRy ∨ yRx) iff 0, ⊆ R ∪ R−1

complete if ∀x ,y∈U (xRy ∨ yRx) iff 1 = R ∪ R−1

symmetric if ∀x ,y∈U (xRy ⇒ yRx) iff R = R−1

anti-symmetric if ∀x ,y∈U (xRy ⇒ (x = y ∨ yRx)) iffR ∩ R−1 ⊆ 1

,

an equivalence relation if R is transitive, symmetric, andreflexive

a weak ordering if R is transitive, reflexive, and connected

a total ordering if R is transitive, reflexive, connected, andanti-symmetric

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 49: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

More definitions from the Calculus of Relations

Binary relation R ⊆ U2 is

transitive if ∀x ,y ,z∈U (xRy ∧ yRz ⇒ xRz) iff R ;R ⊆ R

co-transitive if R is transitive

reflexive if ∀x∈U (xRx) iff 1, ⊆ R

connected if ∀x ,y∈U (x 6= y ⇒ xRy ∨ yRx) iff 0, ⊆ R ∪ R−1

complete if ∀x ,y∈U (xRy ∨ yRx) iff 1 = R ∪ R−1

symmetric if ∀x ,y∈U (xRy ⇒ yRx) iff R = R−1

anti-symmetric if ∀x ,y∈U (xRy ⇒ (x = y ∨ yRx)) iffR ∩ R−1 ⊆ 1

,

an equivalence relation if R is transitive, symmetric, andreflexive

a weak ordering if R is transitive, reflexive, and connected

a total ordering if R is transitive, reflexive, connected, andanti-symmetric

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 50: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

More definitions from the Calculus of Relations

Binary relation R ⊆ U2 is

transitive if ∀x ,y ,z∈U (xRy ∧ yRz ⇒ xRz) iff R ;R ⊆ R

co-transitive if R is transitive

reflexive if ∀x∈U (xRx) iff 1, ⊆ R

connected if ∀x ,y∈U (x 6= y ⇒ xRy ∨ yRx) iff 0, ⊆ R ∪ R−1

complete if ∀x ,y∈U (xRy ∨ yRx) iff 1 = R ∪ R−1

symmetric if ∀x ,y∈U (xRy ⇒ yRx) iff R = R−1

anti-symmetric if ∀x ,y∈U (xRy ⇒ (x = y ∨ yRx)) iffR ∩ R−1 ⊆ 1

,

an equivalence relation if R is transitive, symmetric, andreflexive

a weak ordering if R is transitive, reflexive, and connected

a total ordering if R is transitive, reflexive, connected, andanti-symmetric

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 51: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

More definitions from the Calculus of Relations

Binary relation R ⊆ U2 is

transitive if ∀x ,y ,z∈U (xRy ∧ yRz ⇒ xRz) iff R ;R ⊆ R

co-transitive if R is transitive

reflexive if ∀x∈U (xRx) iff 1, ⊆ R

connected if ∀x ,y∈U (x 6= y ⇒ xRy ∨ yRx) iff 0, ⊆ R ∪ R−1

complete if ∀x ,y∈U (xRy ∨ yRx) iff 1 = R ∪ R−1

symmetric if ∀x ,y∈U (xRy ⇒ yRx) iff R = R−1

anti-symmetric if ∀x ,y∈U (xRy ⇒ (x = y ∨ yRx)) iffR ∩ R−1 ⊆ 1

,

an equivalence relation if R is transitive, symmetric, andreflexive

a weak ordering if R is transitive, reflexive, and connected

a total ordering if R is transitive, reflexive, connected, andanti-symmetric

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 52: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

More definitions from the Calculus of Relations

Binary relation R ⊆ U2 is

transitive if ∀x ,y ,z∈U (xRy ∧ yRz ⇒ xRz) iff R ;R ⊆ R

co-transitive if R is transitive

reflexive if ∀x∈U (xRx) iff 1, ⊆ R

connected if ∀x ,y∈U (x 6= y ⇒ xRy ∨ yRx) iff 0, ⊆ R ∪ R−1

complete if ∀x ,y∈U (xRy ∨ yRx) iff 1 = R ∪ R−1

symmetric if ∀x ,y∈U (xRy ⇒ yRx) iff R = R−1

anti-symmetric if ∀x ,y∈U (xRy ⇒ (x = y ∨ yRx)) iffR ∩ R−1 ⊆ 1

,

an equivalence relation if R is transitive, symmetric, andreflexive

a weak ordering if R is transitive, reflexive, and connected

a total ordering if R is transitive, reflexive, connected, andanti-symmetric

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 53: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

More definitions from the Calculus of Relations

Binary relation R ⊆ U2 is

transitive if ∀x ,y ,z∈U (xRy ∧ yRz ⇒ xRz) iff R ;R ⊆ R

co-transitive if R is transitive

reflexive if ∀x∈U (xRx) iff 1, ⊆ R

connected if ∀x ,y∈U (x 6= y ⇒ xRy ∨ yRx) iff 0, ⊆ R ∪ R−1

complete if ∀x ,y∈U (xRy ∨ yRx) iff 1 = R ∪ R−1

symmetric if ∀x ,y∈U (xRy ⇒ yRx) iff R = R−1

anti-symmetric if ∀x ,y∈U (xRy ⇒ (x = y ∨ yRx)) iffR ∩ R−1 ⊆ 1

,

an equivalence relation if R is transitive, symmetric, andreflexive

a weak ordering if R is transitive, reflexive, and connected

a total ordering if R is transitive, reflexive, connected, andanti-symmetric

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 54: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

More definitions from the Calculus of Relations

Binary relation R ⊆ U2 is

transitive if ∀x ,y ,z∈U (xRy ∧ yRz ⇒ xRz) iff R ;R ⊆ R

co-transitive if R is transitive

reflexive if ∀x∈U (xRx) iff 1, ⊆ R

connected if ∀x ,y∈U (x 6= y ⇒ xRy ∨ yRx) iff 0, ⊆ R ∪ R−1

complete if ∀x ,y∈U (xRy ∨ yRx) iff 1 = R ∪ R−1

symmetric if ∀x ,y∈U (xRy ⇒ yRx) iff R = R−1

anti-symmetric if ∀x ,y∈U (xRy ⇒ (x = y ∨ yRx)) iffR ∩ R−1 ⊆ 1

,

an equivalence relation if R is transitive, symmetric, andreflexive

a weak ordering if R is transitive, reflexive, and connected

a total ordering if R is transitive, reflexive, connected, andanti-symmetric

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 55: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

More definitions from the Calculus of Relations

Binary relation R ⊆ U2 is

transitive if ∀x ,y ,z∈U (xRy ∧ yRz ⇒ xRz) iff R ;R ⊆ R

co-transitive if R is transitive

reflexive if ∀x∈U (xRx) iff 1, ⊆ R

connected if ∀x ,y∈U (x 6= y ⇒ xRy ∨ yRx) iff 0, ⊆ R ∪ R−1

complete if ∀x ,y∈U (xRy ∨ yRx) iff 1 = R ∪ R−1

symmetric if ∀x ,y∈U (xRy ⇒ yRx) iff R = R−1

anti-symmetric if ∀x ,y∈U (xRy ⇒ (x = y ∨ yRx)) iffR ∩ R−1 ⊆ 1

,

an equivalence relation if R is transitive, symmetric, andreflexive

a weak ordering if R is transitive, reflexive, and connected

a total ordering if R is transitive, reflexive, connected, andanti-symmetric

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 56: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

More definitions from the Calculus of Relations

Binary relation R ⊆ U2 is

transitive if ∀x ,y ,z∈U (xRy ∧ yRz ⇒ xRz) iff R ;R ⊆ R

co-transitive if R is transitive

reflexive if ∀x∈U (xRx) iff 1, ⊆ R

connected if ∀x ,y∈U (x 6= y ⇒ xRy ∨ yRx) iff 0, ⊆ R ∪ R−1

complete if ∀x ,y∈U (xRy ∨ yRx) iff 1 = R ∪ R−1

symmetric if ∀x ,y∈U (xRy ⇒ yRx) iff R = R−1

anti-symmetric if ∀x ,y∈U (xRy ⇒ (x = y ∨ yRx)) iffR ∩ R−1 ⊆ 1

,

an equivalence relation if R is transitive, symmetric, andreflexive

a weak ordering if R is transitive, reflexive, and connected

a total ordering if R is transitive, reflexive, connected, andanti-symmetric

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 57: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definition of a social welfare function

A social welfare function is a multivariate operator that mapsweak orderings to weak orderings:

f : Wn →W

i.e., for all R1, . . . ,Rn ∈ W,

R = f (R1, . . . ,Rn) ∈ W

Interpretations:

U is a set of alternative social states

input Ri ∈ W is a ranking of alternative social states bysociety member i ∈ {1, . . . , n}the output R is the social ordering

Why use weak orderings?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 58: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definition of a social welfare function

A social welfare function is a multivariate operator that mapsweak orderings to weak orderings:

f : Wn →W

i.e., for all R1, . . . ,Rn ∈ W,

R = f (R1, . . . ,Rn) ∈ W

Interpretations:

U is a set of alternative social states

input Ri ∈ W is a ranking of alternative social states bysociety member i ∈ {1, . . . , n}the output R is the social ordering

Why use weak orderings?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 59: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definition of a social welfare function

A social welfare function is a multivariate operator that mapsweak orderings to weak orderings:

f : Wn →W

i.e., for all R1, . . . ,Rn ∈ W,

R = f (R1, . . . ,Rn) ∈ W

Interpretations:

U is a set of alternative social states

input Ri ∈ W is a ranking of alternative social states bysociety member i ∈ {1, . . . , n}the output R is the social ordering

Why use weak orderings?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 60: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definition of a social welfare function

A social welfare function is a multivariate operator that mapsweak orderings to weak orderings:

f : Wn →W

i.e., for all R1, . . . ,Rn ∈ W,

R = f (R1, . . . ,Rn) ∈ W

Interpretations:

U is a set of alternative social states

input Ri ∈ W is a ranking of alternative social states bysociety member i ∈ {1, . . . , n}the output R is the social ordering

Why use weak orderings?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 61: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definition of a social welfare function

A social welfare function is a multivariate operator that mapsweak orderings to weak orderings:

f : Wn →W

i.e., for all R1, . . . ,Rn ∈ W,

R = f (R1, . . . ,Rn) ∈ W

Interpretations:

U is a set of alternative social states

input Ri ∈ W is a ranking of alternative social states bysociety member i ∈ {1, . . . , n}the output R is the social ordering

Why use weak orderings?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 62: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definition of a social welfare function

A social welfare function is a multivariate operator that mapsweak orderings to weak orderings:

f : Wn →W

i.e., for all R1, . . . ,Rn ∈ W,

R = f (R1, . . . ,Rn) ∈ W

Interpretations:

U is a set of alternative social states

input Ri ∈ W is a ranking of alternative social states bysociety member i ∈ {1, . . . , n}the output R is the social ordering

Why use weak orderings?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 63: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definition of a social welfare function

A social welfare function is a multivariate operator that mapsweak orderings to weak orderings:

f : Wn →W

i.e., for all R1, . . . ,Rn ∈ W,

R = f (R1, . . . ,Rn) ∈ W

Interpretations:

U is a set of alternative social states

input Ri ∈ W is a ranking of alternative social states bysociety member i ∈ {1, . . . , n}the output R is the social ordering

Why use weak orderings?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 64: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Definition of a social welfare function

A social welfare function is a multivariate operator that mapsweak orderings to weak orderings:

f : Wn →W

i.e., for all R1, . . . ,Rn ∈ W,

R = f (R1, . . . ,Rn) ∈ W

Interpretations:

U is a set of alternative social states

input Ri ∈ W is a ranking of alternative social states bysociety member i ∈ {1, . . . , n}the output R is the social ordering

Why use weak orderings?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 65: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings?

(p. 129–15). “It is assumed further that . . . the chooser considers inturn all possible pairs of alternatives, say x and y , and for eachsuch pair he makes one and only one of three decisions: x ispreferred to y [xPy ], x is indifferent to y [xIy ], or y is preferred tox [yPx ]. The decisions made for different pairs are assumed to beconsistent with each other, so, for example, if x is preferred to yand y to z then x is preferred to z ; similarly, if x is indifferent to yand y to z , then x is indifferent to z .”Arrow assumes

P ∪ I ∪ P−1 = 0,

P, I , P−1 are pairwise disjoint

P and I are transitive

Arrow replaces P and I with R = P ∪ I

Axiom I. R is complete, reflexive, and connected: 1 = R ∪R−1

Axiom II. R is transitive: R ;R ⊆ R

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 66: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings?

(p. 129–15). “It is assumed further that . . . the chooser considers inturn all possible pairs of alternatives, say x and y , and for eachsuch pair he makes one and only one of three decisions: x ispreferred to y [xPy ], x is indifferent to y [xIy ], or y is preferred tox [yPx ]. The decisions made for different pairs are assumed to beconsistent with each other, so, for example, if x is preferred to yand y to z then x is preferred to z ; similarly, if x is indifferent to yand y to z , then x is indifferent to z .”Arrow assumes

P ∪ I ∪ P−1 = 0,

P, I , P−1 are pairwise disjoint

P and I are transitive

Arrow replaces P and I with R = P ∪ I

Axiom I. R is complete, reflexive, and connected: 1 = R ∪R−1

Axiom II. R is transitive: R ;R ⊆ R

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 67: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings?

(p. 129–15). “It is assumed further that . . . the chooser considers inturn all possible pairs of alternatives, say x and y , and for eachsuch pair he makes one and only one of three decisions: x ispreferred to y [xPy ], x is indifferent to y [xIy ], or y is preferred tox [yPx ]. The decisions made for different pairs are assumed to beconsistent with each other, so, for example, if x is preferred to yand y to z then x is preferred to z ; similarly, if x is indifferent to yand y to z , then x is indifferent to z .”Arrow assumes

P ∪ I ∪ P−1 = 0,

P, I , P−1 are pairwise disjoint

P and I are transitive

Arrow replaces P and I with R = P ∪ I

Axiom I. R is complete, reflexive, and connected: 1 = R ∪R−1

Axiom II. R is transitive: R ;R ⊆ R

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 68: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings?

(p. 129–15). “It is assumed further that . . . the chooser considers inturn all possible pairs of alternatives, say x and y , and for eachsuch pair he makes one and only one of three decisions: x ispreferred to y [xPy ], x is indifferent to y [xIy ], or y is preferred tox [yPx ]. The decisions made for different pairs are assumed to beconsistent with each other, so, for example, if x is preferred to yand y to z then x is preferred to z ; similarly, if x is indifferent to yand y to z , then x is indifferent to z .”Arrow assumes

P ∪ I ∪ P−1 = 0,

P, I , P−1 are pairwise disjoint

P and I are transitive

Arrow replaces P and I with R = P ∪ I

Axiom I. R is complete, reflexive, and connected: 1 = R ∪R−1

Axiom II. R is transitive: R ;R ⊆ R

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 69: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings?

(p. 129–15). “It is assumed further that . . . the chooser considers inturn all possible pairs of alternatives, say x and y , and for eachsuch pair he makes one and only one of three decisions: x ispreferred to y [xPy ], x is indifferent to y [xIy ], or y is preferred tox [yPx ]. The decisions made for different pairs are assumed to beconsistent with each other, so, for example, if x is preferred to yand y to z then x is preferred to z ; similarly, if x is indifferent to yand y to z , then x is indifferent to z .”Arrow assumes

P ∪ I ∪ P−1 = 0,

P, I , P−1 are pairwise disjoint

P and I are transitive

Arrow replaces P and I with R = P ∪ I

Axiom I. R is complete, reflexive, and connected: 1 = R ∪R−1

Axiom II. R is transitive: R ;R ⊆ R

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 70: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings?

(p. 129–15). “It is assumed further that . . . the chooser considers inturn all possible pairs of alternatives, say x and y , and for eachsuch pair he makes one and only one of three decisions: x ispreferred to y [xPy ], x is indifferent to y [xIy ], or y is preferred tox [yPx ]. The decisions made for different pairs are assumed to beconsistent with each other, so, for example, if x is preferred to yand y to z then x is preferred to z ; similarly, if x is indifferent to yand y to z , then x is indifferent to z .”Arrow assumes

P ∪ I ∪ P−1 = 0,

P, I , P−1 are pairwise disjoint

P and I are transitive

Arrow replaces P and I with R = P ∪ I

Axiom I. R is complete, reflexive, and connected: 1 = R ∪R−1

Axiom II. R is transitive: R ;R ⊆ R

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 71: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings?

(p. 129–15). “It is assumed further that . . . the chooser considers inturn all possible pairs of alternatives, say x and y , and for eachsuch pair he makes one and only one of three decisions: x ispreferred to y [xPy ], x is indifferent to y [xIy ], or y is preferred tox [yPx ]. The decisions made for different pairs are assumed to beconsistent with each other, so, for example, if x is preferred to yand y to z then x is preferred to z ; similarly, if x is indifferent to yand y to z , then x is indifferent to z .”Arrow assumes

P ∪ I ∪ P−1 = 0,

P, I , P−1 are pairwise disjoint

P and I are transitive

Arrow replaces P and I with R = P ∪ I

Axiom I. R is complete, reflexive, and connected: 1 = R ∪R−1

Axiom II. R is transitive: R ;R ⊆ R

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 72: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings?

(p. 129–15). “It is assumed further that . . . the chooser considers inturn all possible pairs of alternatives, say x and y , and for eachsuch pair he makes one and only one of three decisions: x ispreferred to y [xPy ], x is indifferent to y [xIy ], or y is preferred tox [yPx ]. The decisions made for different pairs are assumed to beconsistent with each other, so, for example, if x is preferred to yand y to z then x is preferred to z ; similarly, if x is indifferent to yand y to z , then x is indifferent to z .”Arrow assumes

P ∪ I ∪ P−1 = 0,

P, I , P−1 are pairwise disjoint

P and I are transitive

Arrow replaces P and I with R = P ∪ I

Axiom I. R is complete, reflexive, and connected: 1 = R ∪R−1

Axiom II. R is transitive: R ;R ⊆ R

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 73: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings?

(p. 129–15). “It is assumed further that . . . the chooser considers inturn all possible pairs of alternatives, say x and y , and for eachsuch pair he makes one and only one of three decisions: x ispreferred to y [xPy ], x is indifferent to y [xIy ], or y is preferred tox [yPx ]. The decisions made for different pairs are assumed to beconsistent with each other, so, for example, if x is preferred to yand y to z then x is preferred to z ; similarly, if x is indifferent to yand y to z , then x is indifferent to z .”Arrow assumes

P ∪ I ∪ P−1 = 0,

P, I , P−1 are pairwise disjoint

P and I are transitive

Arrow replaces P and I with R = P ∪ I

Axiom I. R is complete, reflexive, and connected: 1 = R ∪R−1

Axiom II. R is transitive: R ;R ⊆ R

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 74: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings? (cont.)

Arrow: a weak ordering is a relation satisfying Axioms I and II.(p.146–7) “. . . we ordinarily feel that not only the relation R butalso the relations of (strict) preference and of indifference aretransitive.”Arrow gets these “usually desired properties . . . by definingpreference and indifference suitably in terms of R”.Definition 1. P = R−1

Definition 2. I = R ∩ R−1

(p.1915–19) “. . . it will be assumed that individuals are rational, bywhich is meant that the ordering relations Ri satisfy Axioms Iand II. The problem will be to construct an ordering relation R forsociety as a whole that will also reflect rational choice-making so Rmay also be assumed to satisfy Axioms I and II.”

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 75: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings? (cont.)

Arrow: a weak ordering is a relation satisfying Axioms I and II.(p.146–7) “. . . we ordinarily feel that not only the relation R butalso the relations of (strict) preference and of indifference aretransitive.”Arrow gets these “usually desired properties . . . by definingpreference and indifference suitably in terms of R”.Definition 1. P = R−1

Definition 2. I = R ∩ R−1

(p.1915–19) “. . . it will be assumed that individuals are rational, bywhich is meant that the ordering relations Ri satisfy Axioms Iand II. The problem will be to construct an ordering relation R forsociety as a whole that will also reflect rational choice-making so Rmay also be assumed to satisfy Axioms I and II.”

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 76: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings? (cont.)

Arrow: a weak ordering is a relation satisfying Axioms I and II.(p.146–7) “. . . we ordinarily feel that not only the relation R butalso the relations of (strict) preference and of indifference aretransitive.”Arrow gets these “usually desired properties . . . by definingpreference and indifference suitably in terms of R”.Definition 1. P = R−1

Definition 2. I = R ∩ R−1

(p.1915–19) “. . . it will be assumed that individuals are rational, bywhich is meant that the ordering relations Ri satisfy Axioms Iand II. The problem will be to construct an ordering relation R forsociety as a whole that will also reflect rational choice-making so Rmay also be assumed to satisfy Axioms I and II.”

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 77: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings? (cont.)

Arrow: a weak ordering is a relation satisfying Axioms I and II.(p.146–7) “. . . we ordinarily feel that not only the relation R butalso the relations of (strict) preference and of indifference aretransitive.”Arrow gets these “usually desired properties . . . by definingpreference and indifference suitably in terms of R”.Definition 1. P = R−1

Definition 2. I = R ∩ R−1

(p.1915–19) “. . . it will be assumed that individuals are rational, bywhich is meant that the ordering relations Ri satisfy Axioms Iand II. The problem will be to construct an ordering relation R forsociety as a whole that will also reflect rational choice-making so Rmay also be assumed to satisfy Axioms I and II.”

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 78: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings? (cont.)

Arrow: a weak ordering is a relation satisfying Axioms I and II.(p.146–7) “. . . we ordinarily feel that not only the relation R butalso the relations of (strict) preference and of indifference aretransitive.”Arrow gets these “usually desired properties . . . by definingpreference and indifference suitably in terms of R”.Definition 1. P = R−1

Definition 2. I = R ∩ R−1

(p.1915–19) “. . . it will be assumed that individuals are rational, bywhich is meant that the ordering relations Ri satisfy Axioms Iand II. The problem will be to construct an ordering relation R forsociety as a whole that will also reflect rational choice-making so Rmay also be assumed to satisfy Axioms I and II.”

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 79: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Why weak orderings? (cont.)

Arrow: a weak ordering is a relation satisfying Axioms I and II.(p.146–7) “. . . we ordinarily feel that not only the relation R butalso the relations of (strict) preference and of indifference aretransitive.”Arrow gets these “usually desired properties . . . by definingpreference and indifference suitably in terms of R”.Definition 1. P = R−1

Definition 2. I = R ∩ R−1

(p.1915–19) “. . . it will be assumed that individuals are rational, bywhich is meant that the ordering relations Ri satisfy Axioms Iand II. The problem will be to construct an ordering relation R forsociety as a whole that will also reflect rational choice-making so Rmay also be assumed to satisfy Axioms I and II.”

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 80: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Conditions 1′, P, and 5 on a social welfare function

(p. 96) Condition 1′. “All logically possible orderings areadmissible” (the domain of f is Wn)

(p. 96) Condition P. N = {1, . . . , n}, Pi = Ri−1, P = R−1,

∀R1,...,Rn∈W ∀x ,y∈U (∀i∈N (xPiy) ⇒ xPy)

(p. 30) Condition 5. The social welfare function is not dictatorial

¬∃k∈N ∀x ,y∈U (xPky ⇒ xPy)

Conditions P and 5 are about P, not R!

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 81: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Conditions 1′, P, and 5 on a social welfare function

(p. 96) Condition 1′. “All logically possible orderings areadmissible” (the domain of f is Wn)

(p. 96) Condition P. N = {1, . . . , n}, Pi = Ri−1, P = R−1,

∀R1,...,Rn∈W ∀x ,y∈U (∀i∈N (xPiy) ⇒ xPy)

(p. 30) Condition 5. The social welfare function is not dictatorial

¬∃k∈N ∀x ,y∈U (xPky ⇒ xPy)

Conditions P and 5 are about P, not R!

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 82: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Conditions 1′, P, and 5 on a social welfare function

(p. 96) Condition 1′. “All logically possible orderings areadmissible” (the domain of f is Wn)

(p. 96) Condition P. N = {1, . . . , n}, Pi = Ri−1, P = R−1,

∀R1,...,Rn∈W ∀x ,y∈U (∀i∈N (xPiy) ⇒ xPy)

(p. 30) Condition 5. The social welfare function is not dictatorial

¬∃k∈N ∀x ,y∈U (xPky ⇒ xPy)

Conditions P and 5 are about P, not R!

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 83: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Conditions 1′, P, and 5 on a social welfare function

(p. 96) Condition 1′. “All logically possible orderings areadmissible” (the domain of f is Wn)

(p. 96) Condition P. N = {1, . . . , n}, Pi = Ri−1, P = R−1,

∀R1,...,Rn∈W ∀x ,y∈U (∀i∈N (xPiy) ⇒ xPy)

(p. 30) Condition 5. The social welfare function is not dictatorial

¬∃k∈N ∀x ,y∈U (xPky ⇒ xPy)

Conditions P and 5 are about P, not R!

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 84: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Conditions 1′, P, and 5 on a social welfare function

(p. 96) Condition 1′. “All logically possible orderings areadmissible” (the domain of f is Wn)

(p. 96) Condition P. N = {1, . . . , n}, Pi = Ri−1, P = R−1,

∀R1,...,Rn∈W ∀x ,y∈U (∀i∈N (xPiy) ⇒ xPy)

(p. 30) Condition 5. The social welfare function is not dictatorial

¬∃k∈N ∀x ,y∈U (xPky ⇒ xPy)

Conditions P and 5 are about P, not R!

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 85: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Condition 3 on a social choice function

(p. 15) “In terms of the relation R, we may now define the conceptof choice,. . . ”Definition 3. C (S) = {x : x ∈ S ∧ ∀y∈S (xRy)} if S ⊆ U

Lemma 2. ∀x ,y∈U

(xPy ⇔ {x} = C ({x , y})

)(by Axiom I)

(p. 27) Condition 3. (“independence of irrelevant alternatives”)

∀R1,...,Rn,R′1,...,R

′n∈W ∀S⊆U

(∀i∈N ∀x ,y∈S (xRiy ⇔ xR ′

i y)

⇒ C (S) = C ′(S)).

(p.97) “Theorem 2: Conditions 1′, 3, P, and 5 are inconsistent.”To eliminate C , use x ∈ C ({x , y}) ⇔ xRx ∧ xRy (by logic alone)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 86: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Condition 3 on a social choice function

(p. 15) “In terms of the relation R, we may now define the conceptof choice,. . . ”Definition 3. C (S) = {x : x ∈ S ∧ ∀y∈S (xRy)} if S ⊆ U

Lemma 2. ∀x ,y∈U

(xPy ⇔ {x} = C ({x , y})

)(by Axiom I)

(p. 27) Condition 3. (“independence of irrelevant alternatives”)

∀R1,...,Rn,R′1,...,R

′n∈W ∀S⊆U

(∀i∈N ∀x ,y∈S (xRiy ⇔ xR ′

i y)

⇒ C (S) = C ′(S)).

(p.97) “Theorem 2: Conditions 1′, 3, P, and 5 are inconsistent.”To eliminate C , use x ∈ C ({x , y}) ⇔ xRx ∧ xRy (by logic alone)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 87: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Condition 3 on a social choice function

(p. 15) “In terms of the relation R, we may now define the conceptof choice,. . . ”Definition 3. C (S) = {x : x ∈ S ∧ ∀y∈S (xRy)} if S ⊆ U

Lemma 2. ∀x ,y∈U

(xPy ⇔ {x} = C ({x , y})

)(by Axiom I)

(p. 27) Condition 3. (“independence of irrelevant alternatives”)

∀R1,...,Rn,R′1,...,R

′n∈W ∀S⊆U

(∀i∈N ∀x ,y∈S (xRiy ⇔ xR ′

i y)

⇒ C (S) = C ′(S)).

(p.97) “Theorem 2: Conditions 1′, 3, P, and 5 are inconsistent.”To eliminate C , use x ∈ C ({x , y}) ⇔ xRx ∧ xRy (by logic alone)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 88: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Condition 3 on a social choice function

(p. 15) “In terms of the relation R, we may now define the conceptof choice,. . . ”Definition 3. C (S) = {x : x ∈ S ∧ ∀y∈S (xRy)} if S ⊆ U

Lemma 2. ∀x ,y∈U

(xPy ⇔ {x} = C ({x , y})

)(by Axiom I)

(p. 27) Condition 3. (“independence of irrelevant alternatives”)

∀R1,...,Rn,R′1,...,R

′n∈W ∀S⊆U

(∀i∈N ∀x ,y∈S (xRiy ⇔ xR ′

i y)

⇒ C (S) = C ′(S)).

(p.97) “Theorem 2: Conditions 1′, 3, P, and 5 are inconsistent.”To eliminate C , use x ∈ C ({x , y}) ⇔ xRx ∧ xRy (by logic alone)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 89: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Condition 3 on a social choice function

(p. 15) “In terms of the relation R, we may now define the conceptof choice,. . . ”Definition 3. C (S) = {x : x ∈ S ∧ ∀y∈S (xRy)} if S ⊆ U

Lemma 2. ∀x ,y∈U

(xPy ⇔ {x} = C ({x , y})

)(by Axiom I)

(p. 27) Condition 3. (“independence of irrelevant alternatives”)

∀R1,...,Rn,R′1,...,R

′n∈W ∀S⊆U

(∀i∈N ∀x ,y∈S (xRiy ⇔ xR ′

i y)

⇒ C (S) = C ′(S)).

(p.97) “Theorem 2: Conditions 1′, 3, P, and 5 are inconsistent.”To eliminate C , use x ∈ C ({x , y}) ⇔ xRx ∧ xRy (by logic alone)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 90: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Condition 3 on a social choice function

(p. 15) “In terms of the relation R, we may now define the conceptof choice,. . . ”Definition 3. C (S) = {x : x ∈ S ∧ ∀y∈S (xRy)} if S ⊆ U

Lemma 2. ∀x ,y∈U

(xPy ⇔ {x} = C ({x , y})

)(by Axiom I)

(p. 27) Condition 3. (“independence of irrelevant alternatives”)

∀R1,...,Rn,R′1,...,R

′n∈W ∀S⊆U

(∀i∈N ∀x ,y∈S (xRiy ⇔ xR ′

i y)

⇒ C (S) = C ′(S)).

(p.97) “Theorem 2: Conditions 1′, 3, P, and 5 are inconsistent.”To eliminate C , use x ∈ C ({x , y}) ⇔ xRx ∧ xRy (by logic alone)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 91: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Condition 3 on a social choice function

(p. 15) “In terms of the relation R, we may now define the conceptof choice,. . . ”Definition 3. C (S) = {x : x ∈ S ∧ ∀y∈S (xRy)} if S ⊆ U

Lemma 2. ∀x ,y∈U

(xPy ⇔ {x} = C ({x , y})

)(by Axiom I)

(p. 27) Condition 3. (“independence of irrelevant alternatives”)

∀R1,...,Rn,R′1,...,R

′n∈W ∀S⊆U

(∀i∈N ∀x ,y∈S (xRiy ⇔ xR ′

i y)

⇒ C (S) = C ′(S)).

(p.97) “Theorem 2: Conditions 1′, 3, P, and 5 are inconsistent.”To eliminate C , use x ∈ C ({x , y}) ⇔ xRx ∧ xRy (by logic alone)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 92: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem restated, “weak” vs. “total”

Arrow’s Th. Conditions 1′, 3, and P imply the negation ofCondition 5:

∃k∈N ∀x ,y∈U (xPky ⇒ xPy)

i. e., by Axiom I, there is some “dictator” k ∈ {1, . . . , n} such that

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn (f is a projection function on T )

Arrow’s theorem characterizes projection functions on T asfunctions f : T n → T satisfying Conditions P and 3 (however, verymany non-projection functions f : Wn →W satisfy Conditions Pand 3)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 93: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem restated, “weak” vs. “total”

Arrow’s Th. Conditions 1′, 3, and P imply the negation ofCondition 5:

∃k∈N ∀x ,y∈U (xPky ⇒ xPy)

i. e., by Axiom I, there is some “dictator” k ∈ {1, . . . , n} such that

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn (f is a projection function on T )

Arrow’s theorem characterizes projection functions on T asfunctions f : T n → T satisfying Conditions P and 3 (however, verymany non-projection functions f : Wn →W satisfy Conditions Pand 3)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 94: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem restated, “weak” vs. “total”

Arrow’s Th. Conditions 1′, 3, and P imply the negation ofCondition 5:

∃k∈N ∀x ,y∈U (xPky ⇒ xPy)

i. e., by Axiom I, there is some “dictator” k ∈ {1, . . . , n} such that

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn (f is a projection function on T )

Arrow’s theorem characterizes projection functions on T asfunctions f : T n → T satisfying Conditions P and 3 (however, verymany non-projection functions f : Wn →W satisfy Conditions Pand 3)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 95: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Theorem restated, “weak” vs. “total”

Arrow’s Th. Conditions 1′, 3, and P imply the negation ofCondition 5:

∃k∈N ∀x ,y∈U (xPky ⇒ xPy)

i. e., by Axiom I, there is some “dictator” k ∈ {1, . . . , n} such that

f (R1, . . . ,Rn) = Rk

for all total orderings R1, . . . ,Rn (f is a projection function on T )

Arrow’s theorem characterizes projection functions on T asfunctions f : T n → T satisfying Conditions P and 3 (however, verymany non-projection functions f : Wn →W satisfy Conditions Pand 3)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 96: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Condition 3 for preference relations

Arrow’s characterization and its proof should only involvepreference relations, but Condition 3 applies to R, not P, so use a“unidirectional” property of R ⊆ P(U2), called IIA:

∀R1,...,Rn,R′1,...,R

′n∈R ∀v ,w∈U

(

v 6= w ∧ ∀i∈N

(vRiw ⇔ vR ′

i w)⇒ (vRw ⇔ vR ′w)

)

IIAimplies Cond. 3 if R = W, butCond. 3 implies IIA if R = T

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 97: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Condition 3 for preference relations

Arrow’s characterization and its proof should only involvepreference relations, but Condition 3 applies to R, not P, so use a“unidirectional” property of R ⊆ P(U2), called IIA:

∀R1,...,Rn,R′1,...,R

′n∈R ∀v ,w∈U

(

v 6= w ∧ ∀i∈N

(vRiw ⇔ vR ′

i w)⇒ (vRw ⇔ vR ′w)

)

IIAimplies Cond. 3 if R = W, butCond. 3 implies IIA if R = T

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 98: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Condition 3 for preference relations

Arrow’s characterization and its proof should only involvepreference relations, but Condition 3 applies to R, not P, so use a“unidirectional” property of R ⊆ P(U2), called IIA:

∀R1,...,Rn,R′1,...,R

′n∈R ∀v ,w∈U

(

v 6= w ∧ ∀i∈N

(vRiw ⇔ vR ′

i w)⇒ (vRw ⇔ vR ′w)

)

IIAimplies Cond. 3 if R = W, butCond. 3 implies IIA if R = T

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 99: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Condition 3 for preference relations

Arrow’s characterization and its proof should only involvepreference relations, but Condition 3 applies to R, not P, so use a“unidirectional” property of R ⊆ P(U2), called IIA:

∀R1,...,Rn,R′1,...,R

′n∈R ∀v ,w∈U

(

v 6= w ∧ ∀i∈N

(vRiw ⇔ vR ′

i w)⇒ (vRw ⇔ vR ′w)

)

IIAimplies Cond. 3 if R = W, butCond. 3 implies IIA if R = T

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 100: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Condition 3 for preference relations

Arrow’s characterization and its proof should only involvepreference relations, but Condition 3 applies to R, not P, so use a“unidirectional” property of R ⊆ P(U2), called IIA:

∀R1,...,Rn,R′1,...,R

′n∈R ∀v ,w∈U

(

v 6= w ∧ ∀i∈N

(vRiw ⇔ vR ′

i w)⇒ (vRw ⇔ vR ′w)

)

IIAimplies Cond. 3 if R = W, butCond. 3 implies IIA if R = T

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 101: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Arrow’s Condition 3 for preference relations

Arrow’s characterization and its proof should only involvepreference relations, but Condition 3 applies to R, not P, so use a“unidirectional” property of R ⊆ P(U2), called IIA:

∀R1,...,Rn,R′1,...,R

′n∈R ∀v ,w∈U

(

v 6= w ∧ ∀i∈N

(vRiw ⇔ vR ′

i w)⇒ (vRw ⇔ vR ′w)

)

IIAimplies Cond. 3 if R = W, butCond. 3 implies IIA if R = T

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 102: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Diverse sets of relations

R ⊆ P(U2) is diverse if

∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

))∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

))∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

) )∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

) )R is very diverse if R is diverse and also

∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

)∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

)2 (of 8) conditions are missing—the ones that can’t happen fortransitive, co-transitive relations

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 103: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Diverse sets of relations

R ⊆ P(U2) is diverse if

∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

))∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

))∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

) )∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

) )R is very diverse if R is diverse and also

∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

)∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

)2 (of 8) conditions are missing—the ones that can’t happen fortransitive, co-transitive relations

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 104: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Diverse sets of relations

R ⊆ P(U2) is diverse if

∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

))∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

))∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

) )∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

) )R is very diverse if R is diverse and also

∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

)∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

)2 (of 8) conditions are missing—the ones that can’t happen fortransitive, co-transitive relations

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 105: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Diverse sets of relations

R ⊆ P(U2) is diverse if

∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

))∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

))∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

) )∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

) )R is very diverse if R is diverse and also

∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

)∀x ,y ,z∈U

(|{x , y , z}| = 3 ⇒ ∃R∈R

(xRy ∧ xRz ∧ zRy

)2 (of 8) conditions are missing—the ones that can’t happen fortransitive, co-transitive relations

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 106: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 107: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 108: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 109: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 110: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 111: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 112: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 113: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 114: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 115: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 116: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 117: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 118: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 119: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Examples of diverse sets of relations

Let R = {{〈x , y〉 , 〈x , z〉 , 〈z , y〉} : x , y , z ∈ U, |{x , z , y}| = 3}R contains only transitive relations and is very diverseSome very diverse sets of relations:

transitive relations

co-transitive relations

reflexive relations

connected relations

symmetric relations

anti-symmetric relations

weak orderings Wtotal orderings T

The set of equivalence relations on U is diverse but not very diverse

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 120: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Intersection Theorem

Th. 1. Assume

U is a set with at least three elements

R ⊆ P(U2) is diverse

2 ≤ n < ω, f : P(U2)n → P(U2)

Trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is transitive

)Unan ∀x ,y∈U ∀R1,...,Rn∈R

(x 6= y ∧ ∀i∈N (xRiy) ⇒ xf (R1, . . . ,Rn)y

)IIA ∀x ,y∈U ∀R1,...,Rn,R′

1,...,R′n∈R

(x 6= y ∧ ∀i∈N (xRiy ⇔ xR ′

i y)

⇒(xf (R1, . . . ,Rn)y ⇔ xf (R ′

1, . . . ,R′n)y))

Then f (on diversity relations) is just intersection over some set ofinput variables

∃D0⊆N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 121: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Intersection Theorem

Th. 1. Assume

U is a set with at least three elements

R ⊆ P(U2) is diverse

2 ≤ n < ω, f : P(U2)n → P(U2)

Trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is transitive

)Unan ∀x ,y∈U ∀R1,...,Rn∈R

(x 6= y ∧ ∀i∈N (xRiy) ⇒ xf (R1, . . . ,Rn)y

)IIA ∀x ,y∈U ∀R1,...,Rn,R′

1,...,R′n∈R

(x 6= y ∧ ∀i∈N (xRiy ⇔ xR ′

i y)

⇒(xf (R1, . . . ,Rn)y ⇔ xf (R ′

1, . . . ,R′n)y))

Then f (on diversity relations) is just intersection over some set ofinput variables

∃D0⊆N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 122: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Intersection Theorem

Th. 1. Assume

U is a set with at least three elements

R ⊆ P(U2) is diverse

2 ≤ n < ω, f : P(U2)n → P(U2)

Trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is transitive

)Unan ∀x ,y∈U ∀R1,...,Rn∈R

(x 6= y ∧ ∀i∈N (xRiy) ⇒ xf (R1, . . . ,Rn)y

)IIA ∀x ,y∈U ∀R1,...,Rn,R′

1,...,R′n∈R

(x 6= y ∧ ∀i∈N (xRiy ⇔ xR ′

i y)

⇒(xf (R1, . . . ,Rn)y ⇔ xf (R ′

1, . . . ,R′n)y))

Then f (on diversity relations) is just intersection over some set ofinput variables

∃D0⊆N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 123: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Intersection Theorem

Th. 1. Assume

U is a set with at least three elements

R ⊆ P(U2) is diverse

2 ≤ n < ω, f : P(U2)n → P(U2)

Trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is transitive

)Unan ∀x ,y∈U ∀R1,...,Rn∈R

(x 6= y ∧ ∀i∈N (xRiy) ⇒ xf (R1, . . . ,Rn)y

)IIA ∀x ,y∈U ∀R1,...,Rn,R′

1,...,R′n∈R

(x 6= y ∧ ∀i∈N (xRiy ⇔ xR ′

i y)

⇒(xf (R1, . . . ,Rn)y ⇔ xf (R ′

1, . . . ,R′n)y))

Then f (on diversity relations) is just intersection over some set ofinput variables

∃D0⊆N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 124: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Intersection Theorem

Th. 1. Assume

U is a set with at least three elements

R ⊆ P(U2) is diverse

2 ≤ n < ω, f : P(U2)n → P(U2)

Trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is transitive

)Unan ∀x ,y∈U ∀R1,...,Rn∈R

(x 6= y ∧ ∀i∈N (xRiy) ⇒ xf (R1, . . . ,Rn)y

)IIA ∀x ,y∈U ∀R1,...,Rn,R′

1,...,R′n∈R

(x 6= y ∧ ∀i∈N (xRiy ⇔ xR ′

i y)

⇒(xf (R1, . . . ,Rn)y ⇔ xf (R ′

1, . . . ,R′n)y))

Then f (on diversity relations) is just intersection over some set ofinput variables

∃D0⊆N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 125: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Intersection Theorem

Th. 1. Assume

U is a set with at least three elements

R ⊆ P(U2) is diverse

2 ≤ n < ω, f : P(U2)n → P(U2)

Trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is transitive

)Unan ∀x ,y∈U ∀R1,...,Rn∈R

(x 6= y ∧ ∀i∈N (xRiy) ⇒ xf (R1, . . . ,Rn)y

)IIA ∀x ,y∈U ∀R1,...,Rn,R′

1,...,R′n∈R

(x 6= y ∧ ∀i∈N (xRiy ⇔ xR ′

i y)

⇒(xf (R1, . . . ,Rn)y ⇔ xf (R ′

1, . . . ,R′n)y))

Then f (on diversity relations) is just intersection over some set ofinput variables

∃D0⊆N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 126: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Intersection Theorem

Th. 1. Assume

U is a set with at least three elements

R ⊆ P(U2) is diverse

2 ≤ n < ω, f : P(U2)n → P(U2)

Trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is transitive

)Unan ∀x ,y∈U ∀R1,...,Rn∈R

(x 6= y ∧ ∀i∈N (xRiy) ⇒ xf (R1, . . . ,Rn)y

)IIA ∀x ,y∈U ∀R1,...,Rn,R′

1,...,R′n∈R

(x 6= y ∧ ∀i∈N (xRiy ⇔ xR ′

i y)

⇒(xf (R1, . . . ,Rn)y ⇔ xf (R ′

1, . . . ,R′n)y))

Then f (on diversity relations) is just intersection over some set ofinput variables

∃D0⊆N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 127: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Intersection Theorem

Th. 1. Assume

U is a set with at least three elements

R ⊆ P(U2) is diverse

2 ≤ n < ω, f : P(U2)n → P(U2)

Trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is transitive

)Unan ∀x ,y∈U ∀R1,...,Rn∈R

(x 6= y ∧ ∀i∈N (xRiy) ⇒ xf (R1, . . . ,Rn)y

)IIA ∀x ,y∈U ∀R1,...,Rn,R′

1,...,R′n∈R

(x 6= y ∧ ∀i∈N (xRiy ⇔ xR ′

i y)

⇒(xf (R1, . . . ,Rn)y ⇔ xf (R ′

1, . . . ,R′n)y))

Then f (on diversity relations) is just intersection over some set ofinput variables

∃D0⊆N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 128: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Intersection Theorem

Th. 1. Assume

U is a set with at least three elements

R ⊆ P(U2) is diverse

2 ≤ n < ω, f : P(U2)n → P(U2)

Trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is transitive

)Unan ∀x ,y∈U ∀R1,...,Rn∈R

(x 6= y ∧ ∀i∈N (xRiy) ⇒ xf (R1, . . . ,Rn)y

)IIA ∀x ,y∈U ∀R1,...,Rn,R′

1,...,R′n∈R

(x 6= y ∧ ∀i∈N (xRiy ⇔ xR ′

i y)

⇒(xf (R1, . . . ,Rn)y ⇔ xf (R ′

1, . . . ,R′n)y))

Then f (on diversity relations) is just intersection over some set ofinput variables

∃D0⊆N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 129: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Intersection Theorem

Th. 1. Assume

U is a set with at least three elements

R ⊆ P(U2) is diverse

2 ≤ n < ω, f : P(U2)n → P(U2)

Trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is transitive

)Unan ∀x ,y∈U ∀R1,...,Rn∈R

(x 6= y ∧ ∀i∈N (xRiy) ⇒ xf (R1, . . . ,Rn)y

)IIA ∀x ,y∈U ∀R1,...,Rn,R′

1,...,R′n∈R

(x 6= y ∧ ∀i∈N (xRiy ⇔ xR ′

i y)

⇒(xf (R1, . . . ,Rn)y ⇔ xf (R ′

1, . . . ,R′n)y))

Then f (on diversity relations) is just intersection over some set ofinput variables

∃D0⊆N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 130: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Limits to the theorem

D0 may not be a proper subset of N (perhaps D0 = N), e.g.,if R be the set of equivalence relations on U

and ∀R1,...,Rn∈P(U2)

(f (R1, . . . ,Rn) =

⋂i∈N Ri

)then the hypotheses of the Intersection Theorem all hold—

R is diverse

Unan holds trivially

IIA obviously also holds

f preserves transitivity, reflexivity, and symmetry, so

applied to equivalence relations in R, f produces a (transitive)equivalence relation, so

Trans holds

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 131: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Limits to the theorem

D0 may not be a proper subset of N (perhaps D0 = N), e.g.,if R be the set of equivalence relations on U

and ∀R1,...,Rn∈P(U2)

(f (R1, . . . ,Rn) =

⋂i∈N Ri

)then the hypotheses of the Intersection Theorem all hold—

R is diverse

Unan holds trivially

IIA obviously also holds

f preserves transitivity, reflexivity, and symmetry, so

applied to equivalence relations in R, f produces a (transitive)equivalence relation, so

Trans holds

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 132: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Limits to the theorem

D0 may not be a proper subset of N (perhaps D0 = N), e.g.,if R be the set of equivalence relations on U

and ∀R1,...,Rn∈P(U2)

(f (R1, . . . ,Rn) =

⋂i∈N Ri

)then the hypotheses of the Intersection Theorem all hold—

R is diverse

Unan holds trivially

IIA obviously also holds

f preserves transitivity, reflexivity, and symmetry, so

applied to equivalence relations in R, f produces a (transitive)equivalence relation, so

Trans holds

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 133: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Limits to the theorem

D0 may not be a proper subset of N (perhaps D0 = N), e.g.,if R be the set of equivalence relations on U

and ∀R1,...,Rn∈P(U2)

(f (R1, . . . ,Rn) =

⋂i∈N Ri

)then the hypotheses of the Intersection Theorem all hold—

R is diverse

Unan holds trivially

IIA obviously also holds

f preserves transitivity, reflexivity, and symmetry, so

applied to equivalence relations in R, f produces a (transitive)equivalence relation, so

Trans holds

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 134: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Limits to the theorem

D0 may not be a proper subset of N (perhaps D0 = N), e.g.,if R be the set of equivalence relations on U

and ∀R1,...,Rn∈P(U2)

(f (R1, . . . ,Rn) =

⋂i∈N Ri

)then the hypotheses of the Intersection Theorem all hold—

R is diverse

Unan holds trivially

IIA obviously also holds

f preserves transitivity, reflexivity, and symmetry, so

applied to equivalence relations in R, f produces a (transitive)equivalence relation, so

Trans holds

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 135: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Limits to the theorem

D0 may not be a proper subset of N (perhaps D0 = N), e.g.,if R be the set of equivalence relations on U

and ∀R1,...,Rn∈P(U2)

(f (R1, . . . ,Rn) =

⋂i∈N Ri

)then the hypotheses of the Intersection Theorem all hold—

R is diverse

Unan holds trivially

IIA obviously also holds

f preserves transitivity, reflexivity, and symmetry, so

applied to equivalence relations in R, f produces a (transitive)equivalence relation, so

Trans holds

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 136: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Limits to the theorem

D0 may not be a proper subset of N (perhaps D0 = N), e.g.,if R be the set of equivalence relations on U

and ∀R1,...,Rn∈P(U2)

(f (R1, . . . ,Rn) =

⋂i∈N Ri

)then the hypotheses of the Intersection Theorem all hold—

R is diverse

Unan holds trivially

IIA obviously also holds

f preserves transitivity, reflexivity, and symmetry, so

applied to equivalence relations in R, f produces a (transitive)equivalence relation, so

Trans holds

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 137: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Limits to the theorem

D0 may not be a proper subset of N (perhaps D0 = N), e.g.,if R be the set of equivalence relations on U

and ∀R1,...,Rn∈P(U2)

(f (R1, . . . ,Rn) =

⋂i∈N Ri

)then the hypotheses of the Intersection Theorem all hold—

R is diverse

Unan holds trivially

IIA obviously also holds

f preserves transitivity, reflexivity, and symmetry, so

applied to equivalence relations in R, f produces a (transitive)equivalence relation, so

Trans holds

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 138: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Limits to the theorem

D0 may not be a proper subset of N (perhaps D0 = N), e.g.,if R be the set of equivalence relations on U

and ∀R1,...,Rn∈P(U2)

(f (R1, . . . ,Rn) =

⋂i∈N Ri

)then the hypotheses of the Intersection Theorem all hold—

R is diverse

Unan holds trivially

IIA obviously also holds

f preserves transitivity, reflexivity, and symmetry, so

applied to equivalence relations in R, f produces a (transitive)equivalence relation, so

Trans holds

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 139: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Limits to the theorem

D0 may not be a proper subset of N (perhaps D0 = N), e.g.,if R be the set of equivalence relations on U

and ∀R1,...,Rn∈P(U2)

(f (R1, . . . ,Rn) =

⋂i∈N Ri

)then the hypotheses of the Intersection Theorem all hold—

R is diverse

Unan holds trivially

IIA obviously also holds

f preserves transitivity, reflexivity, and symmetry, so

applied to equivalence relations in R, f produces a (transitive)equivalence relation, so

Trans holds

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 140: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem

Use the Intersection Theorem to prove Arrow’s: assumingf : Wn →W satisfies Conds. P, 3, let with R = T and check—

R is diverse because T is very diverse

f : Wn →W, so f produces only transitive output relations

IIA follows from Condition 3 because R = T (not W!)

Unan follows from Condition P by Axiom I and R = T (pf?)

therefore, by Th. 1, for some subset D0 ⊆ N,

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)D0 cannot have two or more elements, because we can feed twototal orderings to f that disagree on a pair of alternatives, and getan output relation from f that is not connected—

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 141: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem

Use the Intersection Theorem to prove Arrow’s: assumingf : Wn →W satisfies Conds. P, 3, let with R = T and check—

R is diverse because T is very diverse

f : Wn →W, so f produces only transitive output relations

IIA follows from Condition 3 because R = T (not W!)

Unan follows from Condition P by Axiom I and R = T (pf?)

therefore, by Th. 1, for some subset D0 ⊆ N,

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)D0 cannot have two or more elements, because we can feed twototal orderings to f that disagree on a pair of alternatives, and getan output relation from f that is not connected—

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 142: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem

Use the Intersection Theorem to prove Arrow’s: assumingf : Wn →W satisfies Conds. P, 3, let with R = T and check—

R is diverse because T is very diverse

f : Wn →W, so f produces only transitive output relations

IIA follows from Condition 3 because R = T (not W!)

Unan follows from Condition P by Axiom I and R = T (pf?)

therefore, by Th. 1, for some subset D0 ⊆ N,

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)D0 cannot have two or more elements, because we can feed twototal orderings to f that disagree on a pair of alternatives, and getan output relation from f that is not connected—

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 143: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem

Use the Intersection Theorem to prove Arrow’s: assumingf : Wn →W satisfies Conds. P, 3, let with R = T and check—

R is diverse because T is very diverse

f : Wn →W, so f produces only transitive output relations

IIA follows from Condition 3 because R = T (not W!)

Unan follows from Condition P by Axiom I and R = T (pf?)

therefore, by Th. 1, for some subset D0 ⊆ N,

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)D0 cannot have two or more elements, because we can feed twototal orderings to f that disagree on a pair of alternatives, and getan output relation from f that is not connected—

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 144: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem

Use the Intersection Theorem to prove Arrow’s: assumingf : Wn →W satisfies Conds. P, 3, let with R = T and check—

R is diverse because T is very diverse

f : Wn →W, so f produces only transitive output relations

IIA follows from Condition 3 because R = T (not W!)

Unan follows from Condition P by Axiom I and R = T (pf?)

therefore, by Th. 1, for some subset D0 ⊆ N,

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)D0 cannot have two or more elements, because we can feed twototal orderings to f that disagree on a pair of alternatives, and getan output relation from f that is not connected—

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 145: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem

Use the Intersection Theorem to prove Arrow’s: assumingf : Wn →W satisfies Conds. P, 3, let with R = T and check—

R is diverse because T is very diverse

f : Wn →W, so f produces only transitive output relations

IIA follows from Condition 3 because R = T (not W!)

Unan follows from Condition P by Axiom I and R = T (pf?)

therefore, by Th. 1, for some subset D0 ⊆ N,

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)D0 cannot have two or more elements, because we can feed twototal orderings to f that disagree on a pair of alternatives, and getan output relation from f that is not connected—

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 146: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem

Use the Intersection Theorem to prove Arrow’s: assumingf : Wn →W satisfies Conds. P, 3, let with R = T and check—

R is diverse because T is very diverse

f : Wn →W, so f produces only transitive output relations

IIA follows from Condition 3 because R = T (not W!)

Unan follows from Condition P by Axiom I and R = T (pf?)

therefore, by Th. 1, for some subset D0 ⊆ N,

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩⋂

i∈D0

Ri

)D0 cannot have two or more elements, because we can feed twototal orderings to f that disagree on a pair of alternatives, and getan output relation from f that is not connected—

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 147: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

Suppose i 6= j and i , j ∈ D0

Pick distinct x , y ∈ U, pick Ri ,Rj ∈ T so that xRiy , yRix , yRjx ,and xRjy(this is easy with total orderings, which can be anti-symmetric, butit cannot be done with equivalence relations, because they aresymmetric)Let R = f (. . . , Ri , . . . ,Rj , . . . ) = Ri ∩ Rj

with missing input relations chosen arbitrarily from {Ri ,Rj}Then xRy and yRx , i. e., R is not connected, contradicting theassumption that f produces weak (hence connected) orderingsThis shows |D0| < 2Is D0 = ∅?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 148: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

Suppose i 6= j and i , j ∈ D0

Pick distinct x , y ∈ U, pick Ri ,Rj ∈ T so that xRiy , yRix , yRjx ,and xRjy(this is easy with total orderings, which can be anti-symmetric, butit cannot be done with equivalence relations, because they aresymmetric)Let R = f (. . . , Ri , . . . ,Rj , . . . ) = Ri ∩ Rj

with missing input relations chosen arbitrarily from {Ri ,Rj}Then xRy and yRx , i. e., R is not connected, contradicting theassumption that f produces weak (hence connected) orderingsThis shows |D0| < 2Is D0 = ∅?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 149: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

Suppose i 6= j and i , j ∈ D0

Pick distinct x , y ∈ U, pick Ri ,Rj ∈ T so that xRiy , yRix , yRjx ,and xRjy(this is easy with total orderings, which can be anti-symmetric, butit cannot be done with equivalence relations, because they aresymmetric)Let R = f (. . . , Ri , . . . ,Rj , . . . ) = Ri ∩ Rj

with missing input relations chosen arbitrarily from {Ri ,Rj}Then xRy and yRx , i. e., R is not connected, contradicting theassumption that f produces weak (hence connected) orderingsThis shows |D0| < 2Is D0 = ∅?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 150: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

Suppose i 6= j and i , j ∈ D0

Pick distinct x , y ∈ U, pick Ri ,Rj ∈ T so that xRiy , yRix , yRjx ,and xRjy(this is easy with total orderings, which can be anti-symmetric, butit cannot be done with equivalence relations, because they aresymmetric)Let R = f (. . . , Ri , . . . ,Rj , . . . ) = Ri ∩ Rj

with missing input relations chosen arbitrarily from {Ri ,Rj}Then xRy and yRx , i. e., R is not connected, contradicting theassumption that f produces weak (hence connected) orderingsThis shows |D0| < 2Is D0 = ∅?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 151: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

Suppose i 6= j and i , j ∈ D0

Pick distinct x , y ∈ U, pick Ri ,Rj ∈ T so that xRiy , yRix , yRjx ,and xRjy(this is easy with total orderings, which can be anti-symmetric, butit cannot be done with equivalence relations, because they aresymmetric)Let R = f (. . . , Ri , . . . ,Rj , . . . ) = Ri ∩ Rj

with missing input relations chosen arbitrarily from {Ri ,Rj}Then xRy and yRx , i. e., R is not connected, contradicting theassumption that f produces weak (hence connected) orderingsThis shows |D0| < 2Is D0 = ∅?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 152: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

Suppose i 6= j and i , j ∈ D0

Pick distinct x , y ∈ U, pick Ri ,Rj ∈ T so that xRiy , yRix , yRjx ,and xRjy(this is easy with total orderings, which can be anti-symmetric, butit cannot be done with equivalence relations, because they aresymmetric)Let R = f (. . . , Ri , . . . ,Rj , . . . ) = Ri ∩ Rj

with missing input relations chosen arbitrarily from {Ri ,Rj}Then xRy and yRx , i. e., R is not connected, contradicting theassumption that f produces weak (hence connected) orderingsThis shows |D0| < 2Is D0 = ∅?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 153: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

Suppose i 6= j and i , j ∈ D0

Pick distinct x , y ∈ U, pick Ri ,Rj ∈ T so that xRiy , yRix , yRjx ,and xRjy(this is easy with total orderings, which can be anti-symmetric, butit cannot be done with equivalence relations, because they aresymmetric)Let R = f (. . . , Ri , . . . ,Rj , . . . ) = Ri ∩ Rj

with missing input relations chosen arbitrarily from {Ri ,Rj}Then xRy and yRx , i. e., R is not connected, contradicting theassumption that f produces weak (hence connected) orderingsThis shows |D0| < 2Is D0 = ∅?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 154: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

Suppose i 6= j and i , j ∈ D0

Pick distinct x , y ∈ U, pick Ri ,Rj ∈ T so that xRiy , yRix , yRjx ,and xRjy(this is easy with total orderings, which can be anti-symmetric, butit cannot be done with equivalence relations, because they aresymmetric)Let R = f (. . . , Ri , . . . ,Rj , . . . ) = Ri ∩ Rj

with missing input relations chosen arbitrarily from {Ri ,Rj}Then xRy and yRx , i. e., R is not connected, contradicting theassumption that f produces weak (hence connected) orderingsThis shows |D0| < 2Is D0 = ∅?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 155: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

Suppose i 6= j and i , j ∈ D0

Pick distinct x , y ∈ U, pick Ri ,Rj ∈ T so that xRiy , yRix , yRjx ,and xRjy(this is easy with total orderings, which can be anti-symmetric, butit cannot be done with equivalence relations, because they aresymmetric)Let R = f (. . . , Ri , . . . ,Rj , . . . ) = Ri ∩ Rj

with missing input relations chosen arbitrarily from {Ri ,Rj}Then xRy and yRx , i. e., R is not connected, contradicting theassumption that f produces weak (hence connected) orderingsThis shows |D0| < 2Is D0 = ∅?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 156: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

If D0 = ∅ then ∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, )but f (R0, . . . ,R0) = R0 by Condition P and the reflexivity ofrelations in W, hence D0 6= ∅, the only possibility left is D0 = {k}for some k ∈ N, hence

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩ Rk

)However, since all (output and input) relations in W (and T ) arereflexive, f is actually a projection function (onto coordinate k):

∀R1,...,Rn∈T(

f (R1, . . . ,Rn) = Rk

)applying converse-complement and notational conventions gives

∀R1,...,Rn∈T(

P = Pk

)so Condition 5 fails (because there is a “dictator” k)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 157: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

If D0 = ∅ then ∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, )but f (R0, . . . ,R0) = R0 by Condition P and the reflexivity ofrelations in W, hence D0 6= ∅, the only possibility left is D0 = {k}for some k ∈ N, hence

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩ Rk

)However, since all (output and input) relations in W (and T ) arereflexive, f is actually a projection function (onto coordinate k):

∀R1,...,Rn∈T(

f (R1, . . . ,Rn) = Rk

)applying converse-complement and notational conventions gives

∀R1,...,Rn∈T(

P = Pk

)so Condition 5 fails (because there is a “dictator” k)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 158: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

If D0 = ∅ then ∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, )but f (R0, . . . ,R0) = R0 by Condition P and the reflexivity ofrelations in W, hence D0 6= ∅, the only possibility left is D0 = {k}for some k ∈ N, hence

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩ Rk

)However, since all (output and input) relations in W (and T ) arereflexive, f is actually a projection function (onto coordinate k):

∀R1,...,Rn∈T(

f (R1, . . . ,Rn) = Rk

)applying converse-complement and notational conventions gives

∀R1,...,Rn∈T(

P = Pk

)so Condition 5 fails (because there is a “dictator” k)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 159: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

If D0 = ∅ then ∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, )but f (R0, . . . ,R0) = R0 by Condition P and the reflexivity ofrelations in W, hence D0 6= ∅, the only possibility left is D0 = {k}for some k ∈ N, hence

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩ Rk

)However, since all (output and input) relations in W (and T ) arereflexive, f is actually a projection function (onto coordinate k):

∀R1,...,Rn∈T(

f (R1, . . . ,Rn) = Rk

)applying converse-complement and notational conventions gives

∀R1,...,Rn∈T(

P = Pk

)so Condition 5 fails (because there is a “dictator” k)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 160: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

If D0 = ∅ then ∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, )but f (R0, . . . ,R0) = R0 by Condition P and the reflexivity ofrelations in W, hence D0 6= ∅, the only possibility left is D0 = {k}for some k ∈ N, hence

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩ Rk

)However, since all (output and input) relations in W (and T ) arereflexive, f is actually a projection function (onto coordinate k):

∀R1,...,Rn∈T(

f (R1, . . . ,Rn) = Rk

)applying converse-complement and notational conventions gives

∀R1,...,Rn∈T(

P = Pk

)so Condition 5 fails (because there is a “dictator” k)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 161: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

If D0 = ∅ then ∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, )but f (R0, . . . ,R0) = R0 by Condition P and the reflexivity ofrelations in W, hence D0 6= ∅, the only possibility left is D0 = {k}for some k ∈ N, hence

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩ Rk

)However, since all (output and input) relations in W (and T ) arereflexive, f is actually a projection function (onto coordinate k):

∀R1,...,Rn∈T(

f (R1, . . . ,Rn) = Rk

)applying converse-complement and notational conventions gives

∀R1,...,Rn∈T(

P = Pk

)so Condition 5 fails (because there is a “dictator” k)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 162: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Proof of Arrow’s theorem, cont.

If D0 = ∅ then ∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, )but f (R0, . . . ,R0) = R0 by Condition P and the reflexivity ofrelations in W, hence D0 6= ∅, the only possibility left is D0 = {k}for some k ∈ N, hence

∀R1,...,Rn∈T(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩ Rk

)However, since all (output and input) relations in W (and T ) arereflexive, f is actually a projection function (onto coordinate k):

∀R1,...,Rn∈T(

f (R1, . . . ,Rn) = Rk

)applying converse-complement and notational conventions gives

∀R1,...,Rn∈T(

P = Pk

)so Condition 5 fails (because there is a “dictator” k)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 163: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Projection Theorem

Is connectedness really the key property that produces a dictator?No, because—Th. 2. If R ⊆ P(U2) is very diverse, |U| > 2, 2 ≤ n < ω,f : Rn → P(U2), IIA, Unan, Trans, and f produces co-transitiveoutputs from inputs in R,

Co-trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is co-transitive

)then

∃k∈N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩ Rk

)Pf (of Arrow’s Th.) Given Arrow’s hypotheses, let R = T . ThenTrans, Unan, IIA hold (shown above), T is very diverse, and everytotal ordering is co-transitive (but not conversely). By Th.2, f is aprojection function when restricted to T , hence there is a dictator.

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 164: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Projection Theorem

Is connectedness really the key property that produces a dictator?No, because—Th. 2. If R ⊆ P(U2) is very diverse, |U| > 2, 2 ≤ n < ω,f : Rn → P(U2), IIA, Unan, Trans, and f produces co-transitiveoutputs from inputs in R,

Co-trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is co-transitive

)then

∃k∈N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩ Rk

)Pf (of Arrow’s Th.) Given Arrow’s hypotheses, let R = T . ThenTrans, Unan, IIA hold (shown above), T is very diverse, and everytotal ordering is co-transitive (but not conversely). By Th.2, f is aprojection function when restricted to T , hence there is a dictator.

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 165: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Projection Theorem

Is connectedness really the key property that produces a dictator?No, because—Th. 2. If R ⊆ P(U2) is very diverse, |U| > 2, 2 ≤ n < ω,f : Rn → P(U2), IIA, Unan, Trans, and f produces co-transitiveoutputs from inputs in R,

Co-trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is co-transitive

)then

∃k∈N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩ Rk

)Pf (of Arrow’s Th.) Given Arrow’s hypotheses, let R = T . ThenTrans, Unan, IIA hold (shown above), T is very diverse, and everytotal ordering is co-transitive (but not conversely). By Th.2, f is aprojection function when restricted to T , hence there is a dictator.

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 166: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Projection Theorem

Is connectedness really the key property that produces a dictator?No, because—Th. 2. If R ⊆ P(U2) is very diverse, |U| > 2, 2 ≤ n < ω,f : Rn → P(U2), IIA, Unan, Trans, and f produces co-transitiveoutputs from inputs in R,

Co-trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is co-transitive

)then

∃k∈N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩ Rk

)Pf (of Arrow’s Th.) Given Arrow’s hypotheses, let R = T . ThenTrans, Unan, IIA hold (shown above), T is very diverse, and everytotal ordering is co-transitive (but not conversely). By Th.2, f is aprojection function when restricted to T , hence there is a dictator.

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 167: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Projection Theorem

Is connectedness really the key property that produces a dictator?No, because—Th. 2. If R ⊆ P(U2) is very diverse, |U| > 2, 2 ≤ n < ω,f : Rn → P(U2), IIA, Unan, Trans, and f produces co-transitiveoutputs from inputs in R,

Co-trans ∀R1,...,Rn∈R(f (R1, . . . ,Rn) is co-transitive

)then

∃k∈N ∀R1,...,Rn∈R(

0, ∩ f (R1, . . . ,Rn) = 0

, ∩ Rk

)Pf (of Arrow’s Th.) Given Arrow’s hypotheses, let R = T . ThenTrans, Unan, IIA hold (shown above), T is very diverse, and everytotal ordering is co-transitive (but not conversely). By Th.2, f is aprojection function when restricted to T , hence there is a dictator.

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 168: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Characterizations

Arrow’s Theorem says that a multivariate operator on T is aprojection function iff Conditions 3 and PLet R be the reflexive, transitive, co-transitive relations on UThe Projection Theorem says that a multivariate operator on R isa projection function iff IIA and UnanT ⊂ W ⊂ R but W and R are “almost” the same—

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 169: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Characterizations

Arrow’s Theorem says that a multivariate operator on T is aprojection function iff Conditions 3 and PLet R be the reflexive, transitive, co-transitive relations on UThe Projection Theorem says that a multivariate operator on R isa projection function iff IIA and UnanT ⊂ W ⊂ R but W and R are “almost” the same—

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 170: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Characterizations

Arrow’s Theorem says that a multivariate operator on T is aprojection function iff Conditions 3 and PLet R be the reflexive, transitive, co-transitive relations on UThe Projection Theorem says that a multivariate operator on R isa projection function iff IIA and UnanT ⊂ W ⊂ R but W and R are “almost” the same—

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 171: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Characterizations

Arrow’s Theorem says that a multivariate operator on T is aprojection function iff Conditions 3 and PLet R be the reflexive, transitive, co-transitive relations on UThe Projection Theorem says that a multivariate operator on R isa projection function iff IIA and UnanT ⊂ W ⊂ R but W and R are “almost” the same—

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 172: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Characterizations

Arrow’s Theorem says that a multivariate operator on T is aprojection function iff Conditions 3 and PLet R be the reflexive, transitive, co-transitive relations on UThe Projection Theorem says that a multivariate operator on R isa projection function iff IIA and UnanT ⊂ W ⊂ R but W and R are “almost” the same—

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 173: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Transitive, co-transitive relations

Assume R is transitive and co-transitive, e. g., R = ∅ (not refl.)Let I = 1

, ∪ (R ∩ R−1), J = 1, ∪ (R ∩ R−1), E = I ∪ J

Then I , J, E are equivalence relations, I ∩ J = 1,

Let X , Y be disjoint equivalence classes of E , x ∈ X , y ∈ YEither xRy , yRx , X × Y ⊆ R, and (Y × X ) ∩ R = ∅or else yRx , xRy , Y × X ⊆ R, and (X × Y ) ∩ R = ∅The E -classes are totally ordered by ≤ where

X ≤ Y ⇔ ∃x∈X∃y∈Y (xRy)

If any J-class has two or more elements, then R is incomplete(hence not in W) since there are x , y ∈ U with xRy and yRx (Rcomplete ⇔ J = 1

,)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 174: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Transitive, co-transitive relations

Assume R is transitive and co-transitive, e. g., R = ∅ (not refl.)Let I = 1

, ∪ (R ∩ R−1), J = 1, ∪ (R ∩ R−1), E = I ∪ J

Then I , J, E are equivalence relations, I ∩ J = 1,

Let X , Y be disjoint equivalence classes of E , x ∈ X , y ∈ YEither xRy , yRx , X × Y ⊆ R, and (Y × X ) ∩ R = ∅or else yRx , xRy , Y × X ⊆ R, and (X × Y ) ∩ R = ∅The E -classes are totally ordered by ≤ where

X ≤ Y ⇔ ∃x∈X∃y∈Y (xRy)

If any J-class has two or more elements, then R is incomplete(hence not in W) since there are x , y ∈ U with xRy and yRx (Rcomplete ⇔ J = 1

,)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 175: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Transitive, co-transitive relations

Assume R is transitive and co-transitive, e. g., R = ∅ (not refl.)Let I = 1

, ∪ (R ∩ R−1), J = 1, ∪ (R ∩ R−1), E = I ∪ J

Then I , J, E are equivalence relations, I ∩ J = 1,

Let X , Y be disjoint equivalence classes of E , x ∈ X , y ∈ YEither xRy , yRx , X × Y ⊆ R, and (Y × X ) ∩ R = ∅or else yRx , xRy , Y × X ⊆ R, and (X × Y ) ∩ R = ∅The E -classes are totally ordered by ≤ where

X ≤ Y ⇔ ∃x∈X∃y∈Y (xRy)

If any J-class has two or more elements, then R is incomplete(hence not in W) since there are x , y ∈ U with xRy and yRx (Rcomplete ⇔ J = 1

,)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 176: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Transitive, co-transitive relations

Assume R is transitive and co-transitive, e. g., R = ∅ (not refl.)Let I = 1

, ∪ (R ∩ R−1), J = 1, ∪ (R ∩ R−1), E = I ∪ J

Then I , J, E are equivalence relations, I ∩ J = 1,

Let X , Y be disjoint equivalence classes of E , x ∈ X , y ∈ YEither xRy , yRx , X × Y ⊆ R, and (Y × X ) ∩ R = ∅or else yRx , xRy , Y × X ⊆ R, and (X × Y ) ∩ R = ∅The E -classes are totally ordered by ≤ where

X ≤ Y ⇔ ∃x∈X∃y∈Y (xRy)

If any J-class has two or more elements, then R is incomplete(hence not in W) since there are x , y ∈ U with xRy and yRx (Rcomplete ⇔ J = 1

,)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 177: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Transitive, co-transitive relations

Assume R is transitive and co-transitive, e. g., R = ∅ (not refl.)Let I = 1

, ∪ (R ∩ R−1), J = 1, ∪ (R ∩ R−1), E = I ∪ J

Then I , J, E are equivalence relations, I ∩ J = 1,

Let X , Y be disjoint equivalence classes of E , x ∈ X , y ∈ YEither xRy , yRx , X × Y ⊆ R, and (Y × X ) ∩ R = ∅or else yRx , xRy , Y × X ⊆ R, and (X × Y ) ∩ R = ∅The E -classes are totally ordered by ≤ where

X ≤ Y ⇔ ∃x∈X∃y∈Y (xRy)

If any J-class has two or more elements, then R is incomplete(hence not in W) since there are x , y ∈ U with xRy and yRx (Rcomplete ⇔ J = 1

,)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 178: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Transitive, co-transitive relations

Assume R is transitive and co-transitive, e. g., R = ∅ (not refl.)Let I = 1

, ∪ (R ∩ R−1), J = 1, ∪ (R ∩ R−1), E = I ∪ J

Then I , J, E are equivalence relations, I ∩ J = 1,

Let X , Y be disjoint equivalence classes of E , x ∈ X , y ∈ YEither xRy , yRx , X × Y ⊆ R, and (Y × X ) ∩ R = ∅or else yRx , xRy , Y × X ⊆ R, and (X × Y ) ∩ R = ∅The E -classes are totally ordered by ≤ where

X ≤ Y ⇔ ∃x∈X∃y∈Y (xRy)

If any J-class has two or more elements, then R is incomplete(hence not in W) since there are x , y ∈ U with xRy and yRx (Rcomplete ⇔ J = 1

,)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 179: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Transitive, co-transitive relations

Assume R is transitive and co-transitive, e. g., R = ∅ (not refl.)Let I = 1

, ∪ (R ∩ R−1), J = 1, ∪ (R ∩ R−1), E = I ∪ J

Then I , J, E are equivalence relations, I ∩ J = 1,

Let X , Y be disjoint equivalence classes of E , x ∈ X , y ∈ YEither xRy , yRx , X × Y ⊆ R, and (Y × X ) ∩ R = ∅or else yRx , xRy , Y × X ⊆ R, and (X × Y ) ∩ R = ∅The E -classes are totally ordered by ≤ where

X ≤ Y ⇔ ∃x∈X∃y∈Y (xRy)

If any J-class has two or more elements, then R is incomplete(hence not in W) since there are x , y ∈ U with xRy and yRx (Rcomplete ⇔ J = 1

,)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 180: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Transitive, co-transitive relations

Assume R is transitive and co-transitive, e. g., R = ∅ (not refl.)Let I = 1

, ∪ (R ∩ R−1), J = 1, ∪ (R ∩ R−1), E = I ∪ J

Then I , J, E are equivalence relations, I ∩ J = 1,

Let X , Y be disjoint equivalence classes of E , x ∈ X , y ∈ YEither xRy , yRx , X × Y ⊆ R, and (Y × X ) ∩ R = ∅or else yRx , xRy , Y × X ⊆ R, and (X × Y ) ∩ R = ∅The E -classes are totally ordered by ≤ where

X ≤ Y ⇔ ∃x∈X∃y∈Y (xRy)

If any J-class has two or more elements, then R is incomplete(hence not in W) since there are x , y ∈ U with xRy and yRx (Rcomplete ⇔ J = 1

,)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 181: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Prospects

Other theorems, Gibbard?, Sen?,. . . ,Other operations, converse?, union?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 182: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Prospects

Other theorems, Gibbard?, Sen?,. . . ,Other operations, converse?, union?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 183: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Prospects

Other theorems, Gibbard?, Sen?,. . . ,Other operations, converse?, union?

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 184: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Condition 3

Arrow’s Condition 3 is

∀R1,...,Rn,R′1,...,R

′n∈W ∀S⊆U(

∀i∈N ∀x ,y∈S (xRiy ⇔ xR ′i y) ⇒ C (S) = C ′(S)

)Instantiate to S = {v ,w} with v 6= w to get Cond. 3 for 2-elementsets (pairs):

∀R1,...,Rn,R′1,...,R

′n∈W ∀v ,w∈U(

v 6= w ∧ ∀i∈N ∀x ,y∈{v ,w} (xRiy ⇔ xR ′i y) ⇒ C ({v ,w}) = C ′({v ,w})

)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 185: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Condition 3 for pairs

Eliminate “ ∀x ,y∈{v ,w} ” to get this equivalent form of Cond. 3 forpairs:

∀R1,...,Rn,R′1,...,R

′n∈W ∀v ,w∈U

(v 6= w ∧ ∀i∈N

((vRiv ⇔ vR ′

i v) ∧ (vRiw ⇔ vR ′i w)∧

(wRiv ⇔ wR ′i v) ∧ (wRiw ⇔ wR ′

i w))

⇒ C ({v ,w}) = C ′({v ,w})

)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 186: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Condition 3 for pairs

By the definition of C (S), x ∈ C ({x , y}) ⇔ xRx ∧ xRy , so Cond. 3for pairs is equivalent to

∀R1,...,Rn,R′1,...,R

′n∈W ∀v ,w∈U(

v 6= w ∧ ∀i∈N

((vRiv ⇔ vR ′

i v) ∧ (vRiw ⇔ vR ′i w)∧

(wRiv ⇔ wR ′i v) ∧ (wRiw ⇔ wR ′

i w))

⇒ ((vRv ∧ vRw) ⇔ (vR ′v ∧ vR ′w))∧

((wRv ∧ wRw) ⇔ (wR ′v ∧ wR ′w))

)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 187: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Condition 3 for pairs

All the relations in W and all the output relations are reflexive, soCond. 3 for pairs is equivalent to

∀R1,...,Rn,R′1,...,R

′n∈W ∀v ,w∈U(

v 6= w ∧ ∀i∈N

((vRiw ⇔ vR ′

i w) ∧ (wRiv ⇔ wR ′i v))

⇒ (vRw ⇔ vR ′w) ∧ (wRv ⇔ wR ′v)

)

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 188: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

restricted Condition 3 for pairs

Restrict Cond. 3 for pairs from W to T ⊆ W (from weak to totalorderings)Note that, since 0

, ∩ R = 0, ∩ R−1 whenever R ∈ T , if v 6= w then

(vRiw ⇔ vR ′i w) ⇔(vRiw ⇔ vR ′

i w) by logic

⇔(vRi−1w ⇔ vR ′

i−1

w) Ri ,R′i ∈ T , v 6= w

⇔(wRiv ⇔ wR ′i v) def. of −1

so the Cond 3. for pairs, restricted to total orderings, is equivalentto

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 189: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

Condition 3 restricted to pairs and total orderings

∀R1,...,Rn,R′1,...,R

′n∈T ∀v ,w∈U

(

v 6= w ∧ ∀i∈N

(vRiw ⇔ vR ′

i w)⇒ (vRw ⇔ vR ′w) ∧ (wRv ⇔ wR ′v)

)

We only need the consequence obtained by deleting the finalconjunct (equivalent if R and R ′ are also total orderings):

R. D. Maddux Arrow’s Theorem for Incomplete Relations

Page 190: Arrow's Theorem for Incomplete Relationsorion.math.iastate.edu/.../13Models/maddux.arrow.talk.pdf · 2013-11-08 · History Kenneth J. Arrow (Nobel, 1973, Econonics) learned the Calculus

A consequence of Cond. 3 restricted to pairs and T

∀R1,...,Rn,R′1,...,R

′n∈T ∀v ,w∈U

(

v 6= w ∧ ∀i∈N

(vRiw ⇔ vR ′

i w)⇒ (vRw ⇔ vR ′w)

)

This is IIA with R = T .

R. D. Maddux Arrow’s Theorem for Incomplete Relations