One-sided

OVERCOMING
COORDINATION FAILURE
THROUGH
NEIGHBORHOOD CHOICE
~AN EXPERIMENTAL STUDY~
Maastricht University
Arno Riedl
Ingrid M.T. Rohde
Martin Strobel
Introduction: Why ?


Laboratory: people fail to coordinate on surplus maximizing
equilibria
Reality: people able to agree on rather efficient arrangements
∆?
 Interaction structure
Possibility to choose interaction partners
Largely neglected in the experimental and theoretical economic literature
Our hypothesis
Endogenous choice of interaction partners is a powerful
force to achieve coordination on the payoff dominant
equilibrium
Coordination game
B
G
B
95,95
5,90
G
90,5
75,75
Payoff dominant equilibrium
Risk dominant equilibrium
Representation during experiment
your colour
blue
blue
green
green
colour of your
neighbour
blue
green
blue
green
your earnings
95 points
5 points
90 points
75 points
earnings of your
neighbour
95 points
90 points
5 points
75 points
Experiment: Repeated Network Game

group size = 6 players
two time blocks of 30 periods
full information

Decisions player i:


# neighbors
I.
Network formation 
1.
2.
Two parameters:
One-sided vs. two-sided link formation
Costless vs. costly link formation
II.
Coordination game: 
1 strategy
simultaneously
Implications Network Game

Link formation
6 players
Total possible # links proposed: 30 links
Total possible # links formed: 15 links

Coordination game
B
G
B
95,95
5,90
G
90,5
75,75
Blue very risky
complete network
 need all 5 other players to play B in
order for you to play B as well
Experiment: Treatments

Experimental treatments
costless
Link formation
onesided
twosided
I
II
Linking cost
costly
III
Cost forming link = 80 points

Control treatment : exogenously fixed complete network
 search “stable” equilibrium strategy & network structure
Repeated network game: our predictions
One-sided, costless
complete network
risk dominance
no threat punishment
hunt
One sided, costly
small & unstable
networks
Additional coordination problem
no threat punishment
Two-sided, costless
complete network
payoff dominance
Punishment: efficiency enforcers
trade off rd versus pd action
Control treatment
risk dominance
Actual Experiment


Computerized experimental sessions at BEE laboratory Maastricht
210 subjects: Students Maastricht University
Each treatment : 3 sessions
1 session == 18 subjects
1 group == 6 subjects
One sided costly: 8 observations

Each session took approximately two hours

Average earning per subject: 26.05 Euro
9 independent
observations
Results: what is actually played
Round 1
No significant ∆ among
the four treatments
in payoff dominant play
Frequency payoff dominant play
Frequency
.2
.4
.6
.8
1
part 1
Overall (1)
Significant ∆ treatments
with endogenous
network structure
vs. control treatment
0
10
Round
one sided costly
two sided costless
20
30
one sided costless
full fixed
payoff dominant equilibrium / total play
Overall (2)
Increase played payoff
dominant eq played with
endogenous network
structure, decrease in
control treatment
Network: link structure
Excess links
Part I
Part I
Frequency
.4
.6
Frequency
0 .1 .2 .3 .4 .5
.8
1
Network density
0
10
20
Round
one sided costly
two sided costless
one sided costless
full fixed
Overall:
One sided costless: complete network
30
0
10
20
Round
one sided costly
two sided costless
one sided costless
Overall:
One sided costless: all links proposed
Two sided costless: no full reciprocation
Two sided costless: punishment
efficiency enforcers
One sided costly: additional coordination One sided costly: add.coordination
problem
problem not solved
30
Results: play
Total play/ possible play
0
0
.2
.2
.2
.4
.4
.4
.6
.6
.6
.8
.8
.8
11
1
Actual
Actual
play/
# total
Play actual
/ 15 play
ffff
2snoc
2snoc
frpdpd
frpdpd
frrdrd
frrdrd
1snoc
1snoc
1sc
1sc
frprfrpr
frnoplay
Overall:
Two sided costless:CG
Endogenous
treatments:not
surplus
always played
max. eq.
One sided
Costly
highest
costly:coordination
CG not always
surplus
played
max. eq.
ff
2snoc
frpdpd
frrdrd
1snoc
1sc
frpr
Overall:
One sided costly: coordination problem
Two sided costless: much eq. play
less off eq. play
Further tests: Why these results?

Network structure vs action choice

Action choice history vs. link proposal

Individual data on:
risk preferences: ten paired lottery (Holt Laurey, 2002)
 Generalized preferences on equality & efficiency
 personality traits: Brandstaetter 32 questions

Conclusion
Our hypothesis
Endogenous choice of interaction partners is a
powerful force to achieve coordination on the
payoff dominant equilibrium
 strong support for our hypothesis!
or