Callander, S., and C.R. Plott. 2005. Principles of Network Development and Evolution: An Experimental Study. Journal of Public Economics. Vol. 89. pp. 1469-1495.

Principles of Network Development and Evolution:

An Experimental Study

A review of the paper by Callander and Plottby

Kash Barker

Networkin’ it

• Networks used to describe many physical systems and theoretical environments– paper cites labor market participation,

internal organization of firms, social interactions

– what principles guide the unregulated development and evolution of networks?

Questions addressed here

• Do networks converge to steady state outcomes and what are the properties of this state?

• What principles underlie network development and evolution?

• How can these principles be understood in the context of individual decisions and behavior?

Preliminaries

• Network is a set of connected nodes– each node is a separated decisionmaking

agent

– links represent a transfer of “stuff”• node B connects to node A to receive a benefit

at some cost

– flow of stuff is “without decay”

– network structure is common knowledge to each agent

Principles

• Three principles of network formation– Nash equilibrium

• no individual can improve personal gains by some unilateral change of action

– efficiency• proportion of gains received relative to

potential gains

– focalness• refers to geographic location in the experiment

Types of networks

wheel

counterclockwise wheel

Types of networks

an efficient nonfocal wheel

an efficient star

Sets of parameters• Authors used four sets of parameters for link

connection cost and for the value of information

Model predictions• For each set of parameters, the following

network types were predicted for each principle

Comparisons

Types of networks

wheel

counterclockwise wheel

Types of networks

an efficient nonfocal wheel

an efficient star

Experimental background• Six participants (nodes)• Consisted of “rounds” during which

subjects could make connection with another subject at a cost– convergence occurred if same

configuration appeared in three consecutive rounds

• Two “series” performed– series 1 used parameter set 1– series 2 used parameter sets 2, 3, 4

Result 1

• Networks happen, with an appreciation of externalities is incorporated into agent decisions– in each experiment, a network

instantaneously formed

– given appropriate conditions, social or economic network will emerge

Result 2

• (i) Networks tend to converge to stationary configurations, (ii) Nash equilibrium is necessary condition for stationarity– (i) eight of 12 networks converged

– (ii) all eight convergent networks are Nash equilibria

Result 3

• Each of focalness, efficiency, and strict Nash equilibrium can be rejected as being necessary for configuration to become stationary– one experiment converged to nonfocal,

inefficient configuration in rounds 16-18

– one experiment converged to nonfocal wheel in rounds 17-19

– one experiment converged to weak Nash equilibrium configuration

Result 4

• Principle of Nash equilibrium does not prove a sufficient condition for configuration to be stationary– weak Nash configurations that did not

prove stable were played in six experiments

– strict Nash configurations not proving stable were played in 3 experiments

Result 5

• Nonconvergent networks do not exhibit increasing efficiency

– dynamics of network evolution and change are not guided by a principle of efficiency seeking

Al fin