Outline In-Class Experiment on a Coordination Game Test of Equilibrium Selection I :Van Huyck, Battalio, and Beil (1990) Test of Equilibrium Selection.

OutlineIn-Class Experiment on a Coordination Game

Test of Equilibrium Selection I :Van Huyck, Battalio, and Beil (1990)

Test of Equilibrium Selection II :Van Huyck, Battalio, and Beil (1991)

Test of Equilibrium Selection III : Copper, DeJong, Forsythe, and Ross (1990)

The Median Action Game

eemedianM

cbaceMbMa,M)(en

ii},...,{

0,, ;][1

2

}e,...,,{e(e,...,e)mequilibriuNash a is

21 with tuple-nAny

n (odd) players

Strategy space =

e,...,, }21{

Game : a=$0.1, b=$0.05, c=$0.6

eemedianM

cbaceMbMa,M)(en

ii},...,{

0,, ;][1

2

Hypotheses: Deductive vs. Inductive Principles

Payoff Dominance

Security (Maximin}

History DependentFor t > 1, Median (t) =

eemedianM n )}1(),...,1({)1( 1

Game : a=$0.1, b=$0.0, c=$0.6

eemedianM

cbaceMbMa,M)(en

ii},...,{

0,, ;][1

2

Game : a=$0.0, b=$0.05, c=$0.7

eemedianM

cbaceMbMa,M)(en

ii},...,{

0,, ;][1

2

Hypotheses: Deductive vs. Inductive Principles

gamein {4,...,4} and gamein }3,...,3{

and gamesin {7,...,7}Payoff Dominance:

Security (Maximin}:

For t > 1, median (t)

eemedian n )}1(),...,1({ 1

Experimental Design

n=9 * Only median was announced after every round

Test of Deductive Selection Principles in Period 1

Hypotheses: Deductive vs. Inductive Principles

gamein {4,...,4} and gamein }3,...,3{

and gamesin {7,...,7}Payoff Dominance:

Security (Maximin}:

For t > 1, median (t)

eemedian n )}1(),...,1({ 1

Observations

In game , neither payoff-dominance or security is salient: 15% chose 7 and 15% chose 3.

In game , payoff-dominance is salient: 52% chose 7

In game , security is salient: 44% chose 4Seven times as many subjects played above 4 than below 4

Test of Deductive Selection Principles in Period 1

Influence of Initial Median (M(1))

Experimental Design

n=9 * Only median was announced after every round

Salience of Last Median in a Related Game

Did Selection Principle Change?: Period 1 vs. Period 11

Summary

• In games where neither payoff-dominance nor secure equilibrium was salient, both predictions were infrequently played.

• History-dependent principle works: The initial median is a perfect predictor for subsequent behaviors.

• Experience increases the salience of payoff-dominance