Computational complexity of competitive equilibria in exchange - - PowerPoint PPT Presentation

computational complexity of competitive equilibria in
SMART_READER_LITE
LIVE PREVIEW

Computational complexity of competitive equilibria in exchange - - PowerPoint PPT Presentation

Computational complexity of competitive equilibria in exchange markets Katarna Cechlrov P. J. afrik University Ko ice, Slovakia Budapest, Summer school, 2013 Outline of the talk n brief history of the notion of competitive


slide-1
SLIDE 1

Computational complexity

  • f competitive equilibria

in exchange markets

Katarína Cechlárová

  • P. J. Šafárik University

Košice, Slovakia

Budapest, Summer school, 2013

slide-2
SLIDE 2

Outline of the talk

n brief history of the notion of competitive

equilibrium

n model computation for divisible goods n indivisible goods – housing market n Top trading cycles algorithm n housing market with duplicated houses

¨ algorithm and complexity ¨ approximate equilibrium and its complexity

  • K. Cechlárová, Budapest 2013

2

slide-3
SLIDE 3

First ideas

  • K. Cechlárová, Budapest 2013

3

n Adam Smith: An Inquiry into

the Nature and Causes of the Wealth of Nations (1776)

n Francis Ysidro

Edgeworth: Mathematical

Psychics: An Essay on the Application

  • f Mathematics to the Moral Sciences

(1881)

n Marie-Ésprit Léon

Walras: Elements of Pure

Economics (1874)

n Vilfredo Pareto: Manual of

Political Economy (1906)

slide-4
SLIDE 4

Exchange economy

n set of agents, set of commodities n each agent owns a commodity bundle

and has preferences over bundles

n economic equilibrium: pair

(prices, redistribution) such that:

¨ each agent owns the best bundle he can

afford given his budget

¨ demand equals supply

n if commodities are infinitely divisible and

preferences of agents strictly monotone and strictly convex, equilibrium always exists

  • K. Cechlárová, Budapest 2013

4

Kenneth Arrow & Gérard Debreu (1954)

slide-5
SLIDE 5

Example: two agents, two goods

  • K. Cechlárová, Budapest 2013

5

n agent 1: n agent 2: n prices (1,1) 2 1 2 1 1 1

) , ( ); 1 , 2 ( x x x x u = = ω

2 2 1 2 2

) , ( ); , 1 ( x x x u = = ω

1

x

2

x ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ = 2 3 , 2 3

1

x

1

x

2

x

( )

1 ,

2 =

x

prices (1,1) are not equilibrium, as supply ≠demand

⎟ ⎠ ⎞ ⎜ ⎝ ⎛ = 2 5 , 2 3

i

x

( )

= 1 , 3

i

ω

slide-6
SLIDE 6

Example - continued

  • K. Cechlárová, Budapest 2013

6

n agent 1: n agent 2: n prices (1,4) 2 1 2 1 1 1

) , ( ); 1 , 2 ( x x x x u = = ω

2 2 1 2 2

) , ( ); , 1 ( x x x u = = ω

1

x

2

x ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ = 4 3 , 3

1

x

1

x

2

x ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ = 4 1 ,

2

x

Equilibrium!

( )

= 1 , 3

i

x

( )

= 1 , 3

i

ω

slide-7
SLIDE 7

Economy with indivisible goods

  • K. Cechlárová, Budapest 2013

7

Equlibrium might not exists!

  • X. Deng, Ch. Papadimitriou, S. Safra

(2002): Decision problem: Does an economic equilibrium exist in exchange economy with indivisible commodities and linear utility functions? NP-complete, already for two agents

slide-8
SLIDE 8

Housing market

  • K. Cechlárová, Budapest 2013

8

n n agents, each owns one unit of

a unique indivisible good – house

n preferences of agent: linear

  • rdering on a subset of houses

n Shapley-Scarf economy (1974) n housing market is a model of:

¨ kidney exchange ¨ several Internet based markets

slide-9
SLIDE 9
  • K. Cechlárová, Budapest 2013

9

acceptable houses strict preferences ties trichotomous preferences

slide-10
SLIDE 10
  • K. Cechlárová, Budapest 2013

10

a1 a2 a7 a6 a4 a5 a3

slide-11
SLIDE 11
  • K. Cechlárová, Budapest 2013

11 11

Lemma. Definition.

not equilibrium: a6 not satisfied

a1 a2 a7 a6 a4 a5 a3

slide-12
SLIDE 12

Top Trading Cycles algorithm for

Shapley-Scarf model (m=n,ω identity)

  • K. Cechlárová, Budapest 2013

12

Step 0. N:=A, round r:=0, pr=n. Step 1. Take an arbitrary agent a0. Step 2. a0 points to a most preferred house, in N, its

  • wner is a1 . Agent a1 points to the most preferred house

a2 in N etc. A cycle C arises. Step 3. r:=r+1, pr= pr-1; Cr:=C, all houses on C receive price pr, N:=N-C. Step 4. If N ≠∅, go to Step 1, else end.

n Shapley & Scarf (1974): author D. Gale n Abraham, KC, Manlove, Mehlhorn (2004): implementation

linear in the size of the market

slide-13
SLIDE 13

Top Trading Cycles algorithm for

Shapley-Scarf model (m=n,ω identity)

  • K. Cechlárová, Budapest 2013

13

Step 0. N:=A, round r:=0, pr=n. Step 1. Take an arbitrary agent a0. Step 2. a0 points to a most preferred house, in N, its

  • wner is a1 . Agent a1 points to the most preferred house

a2 in N etc. A cycle C arises. Step 3. r:=r+1, pr= pr-1; Cr:=C, all houses on C receive price pr, N:=N-C. Step 4. If N ≠∅, go to Step 1, else end.

Theorem (Gale 1974).

slide-14
SLIDE 14
  • K. Cechlárová, Budapest 2013

14

Theorem (Fekete, Skutella , Woeginger 2003). Theorem (KC & Fleiner 2008).

slide-15
SLIDE 15
  • K. Cechlárová, Budapest 2013

15

h2 h4 h1 h3 a1 a4 a2 a3 a5 a6

p1 > p2

a7

slide-16
SLIDE 16
  • K. Cechlárová, Budapest 2013

16

h2 h4 h1 h3 a1 a4 a2 a3 a5 a6 a7

Theorem (KC & Schlotter 2010). Theorem (KC & Schlotter 2010).

Definition.

slide-17
SLIDE 17
  • K. Cechlárová, Budapest 2013

17

Approximating the number of satisfied agents

Definition.

slide-18
SLIDE 18
  • K. Cechlárová, Budapest 2013

18

Theorem (KC & Jelínková 2011).

slide-19
SLIDE 19
  • K. Cechlárová, Budapest 2013

19

Theorem (KC & Jelínková 2011).

slide-20
SLIDE 20
  • K. Cechlárová, Budapest 2013

20

Theorem (KC & Jelínková 2011).

slide-21
SLIDE 21
  • K. Cechlárová, Budapest 2013

21

Theorem (KC & Jelínková 2011).

1 2 3 4 5 6 7 8 9

slide-22
SLIDE 22
  • K. Cechlárová, Budapest 2013

22

Theorem (KC & Jelínková 2011). Theorem (KC & Jelínková 2011).

slide-23
SLIDE 23

Thank you for your attention!