Policy Analysis from First Principles
CPM Report No.: 01-83
By: Scott Moss
Date: 31 August 2001
Published as: Policy Analysis from First Principles
Publication information: Proceedings of the US National Academy of Sciences
(forthcoming) -- an invited paper for the Arthur M. Sackler Colloquium on
Adaptive Agents, Intelligence and Emergent Human Organization: Capturing
Complexity Through Agent-Based Modeling, October 5 and 6, 2001.
Abstract
The Sackler Colloquium on on Adaptive Agents, Intelligence and Emergent
Human Organization is intended to force a re-examination of current social
theory by means of an exploration of adaptive agent models. The primary
issue addressed by this paper is why it should be assumed that adaptive
agent models are suitable for that purpose. The argument is predicated
on the view that social science should start with observation and the specification
of a problem to be solved and then reason from that basis to define the
appropriate properties and conditions of application of relevant tools
of analysis. Accordingly, evidence is adduced from data for sales
volumes and values of a disparate range of goods to show that, in all presented
cases, frequency distributions are fat tailed. This result implies
that, if there is a suitable and stable population distribution, it will
generally have infinite variance and perhaps undefined mean. Agent
based models with agents that reason about their behaviour and specifically
do not conform to the rational choice model and are influenced by, but
do not imitate, other agents known to them will typically generate fat
tailed time series data. There is no reason to suppose that this
data is drawn from any stable distribution. An agent based social
simulation model of intermediated exchange is reported that has the same
type fat tailed time series and cross sectional data that is found in data
for fast moving consumer goods and for retail outlets. This result
supports the proposition that adaptive agent models of markets with agents
that reason and are socially embedded have the same statistical signatures
as real markets. This statistical signature precludes any conventional
hypothesis testing or forecasting – a point which pertains both to intermediated
(e.g. retail or financial) markets in general and to the models.
However, these agent based social simulation models offer unique opportunities
for validation on the basis of domain expertise and qualitative data.
While they cannot be used for prediction or forecasting the specific consequences
of any policy or commercial action, they can be used both to identify differences
in stakeholders’ perceptions and to investigate the properties of effective
responses to unpredictable events.
Accessible as: