Computational Techniques for Modelling Learning in Economics

391 Pages·1999·13.679 MB·English

by  Witold Kwaśnicki (auth.), Thomas Brenner (eds.)

COMPUTATIONAL TECHNIQUES FOR MODELLING
LEARNING IN ECONOMICS
Advances in Computational Economics
VOLUME 11
SERIES EDITORS
Hans Amman, University ofA msterdam, Amsterdam, The Netherlands
Anna Nagumey, University ofM assachusens at Amherst, USA
EDITORIAL BOARD
Anantha K. Duraiappah, European University Institute
John Geweke, University ofM innesota
Manfred Gilli, University ofG eneva
Kenneth L. Judd, Stanford University
David Kendrick, University of Texas at Austin
Daniel McFadden, University of California at Berlceley
Ellen McGrattan, Dulce University
Reinhard Neck, University ofK lagenfurt
Adrian R. Pagan, Australian National University
John Rust, University ofW isconsin
Berc Rustem, University ofL ondon
Hal R. Varian, University ofM ichigan
The titles published in this series are listed at the end oft his volume.
Computational Techniques
for Modelling Learning
in Economics
edited by
Thomas Brenner
Max-Planck-Institute for Research
into Economic Systems
~.
"
Springer Science+Business Media, LLC
Library of Congress Cataloging-in-Publication Data
Computational techniques for modelling learning in economics / edited
by Thomas Brenner.
p. cm. --(Advances in computational economics; v. 11)
Includes bibliographical references and index.
ISBN 978-1-4613-7285-1 ISBN 978-1-4615-5029-7 (eBook)
DOI 10.1007/978-1-4615-5029-7
I. Economics, mathematical--Study and teaching. 2. Econometrics
-Study and teaching. 3. Computationallearning theory.
I. Brenner, Thomas, 1968- 11. Series.
HB135.C632 1999
330'.0I'51--dc21 99-25823
CIP
Copyright © 1999 Springer Science+Business Media New York
Originally published by Kluwer Academic Publishers, New York in 1999
Softcover reprint of the hardcover 1s t edition 1999
All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system or transmitted in any form or by any means, mechanical, photo-copying, recording,
or otherwise, without the prior written permission of the publisher, Springer Science+
Business Media, LLC.
Printed on acid-free paper.
Table of Contents
Preface vii
List of Contributors xi
Part One:
Simulating in Economics
Evolutionary Economics and Simulation
Witold Kwasnicki 3
Simulation as a Tool to Model Stochastic Processes in Complex Systems
Klaus G. Troitzsch 45
Part Two:
Evolutionary Approaches
Learning by Genetic Algorithms in Economics?
Frank Beckenbach 73
Can Learning-Agent Simulations Be Used for Computer Assisted
Design in Economics?
Tony Curzon Price 101
On the Emergence of Attitudes towards Risk
Steffen Huck, Wieland Maller and Martin Strobel 123
Interdependencies, Nearly-decomposability and Adaptation
Koen Frenken, Luigi Marengo and Marco Valente 145
Part Three:
Neural Networks and Local Interaction
Neural Networks in Economics
Ralf Herbrich, Max Keilbach, Thore Graepel, Peter
Bollmann-Sdorra and Klaus Obermayer 169
vi
Genetic Algorithms and Neural Networks: A Comparison Based
on the Repeated Prisoners Dilemma
Robert E. Marks and Hermann Schnabl 197
Local Interaction as a Model of Social Interaction?
Dorothea K. Herreiner 221
Part Four:
Boundedly Rational and Rational Models
Memory, Learning and the Selection of Equilibria in a Model with
Non-Uniqueness
Emilio Barucci 243
A Behavioral Approach to a Strategic Market Game
Martin Shubik and Nicolaas J. Vriend 261
Bayesian Learning in Optimal Growth Models under Uncertainty
Sardar M. N. Islam 283
Part Five:
Cognitive Learning Models
Modelling Bounded Rationality in Agent-based Simulations Using
the Evolution of Mental Models
Bruce Edmonds 305
Cognitive Learning in Prisoner's Dilemma Situations
Thomas Brenner 333
A Cognitively Rich Methodology for Modelling Emergent
Socioeconomic Phenomena
Scott Moss 363
Index 387
Preface
Learning has become an increasingly recognised topic within economics with
the number of works applying learning models to economic contexts increasing
tremendously in the last few years. Three directions of research can be distin
guished; the experimental study of individual learning processes, the analysis of
the characteristics of existing learning models, mainly in the context of games, and
the application of learning processes to an economic context which often involves
simulations. The present book presents an overview on recent developments in
the last of these research areas. This area is characterised by a enormous hetero
geneity of approaches. The main techniques are evolutionary algorithms, neural
nets, and cellular automata which are all well-defined with respect of their math
ematical features. In addition, however, boundedly rational and rational models
as well as models of cognitive learning are used to describe learning in economic
contexts. These latter models share a problem with other research into learning in
economics: the heterogeneity of approaches with each author favouring a different
model.
As a consequence, it is difficult for someone not familiar with all these models
to figure out what they are for or which one to use. What is missing is a presen
tation and discussion of the most important models within one book. The present
book fills this gap by offering a critical overview on the computational techniques
that are frequently used for modelling learning in economics. The main focus is
to describe these techniques, give some examples of applications, and discuss the
advantages and disadvantages of their use. The book contains a collection of ar
ticles where each article, except the first two, represents the application of one
computational technique to an economic context. Besides this, each article intro
duces the specific kind of modelling it applies, including a description of the basic
features of the technique as well as a discussion of its usefulness for modelling
learning in economic contexts. Hence, the book offers some guiding in the field
of modelling learning in computational economics. In addition, it presents some
state-of-the-art applications of learning models and some innovative examples of
the use of computation devises for understanding economic dynamics.
The book is divided into five sections. The first section focuses on the use of
simulations in economics. Witold Kwasnicki discusses the advantages and disad-
viii
vantages of simulations compared with mathematical analyses and the study of
real systems and presents different approaches and different platforms for running
simulations. In addition, Klaus G. Troitzsch discusses the mathematical aspects
of simulations, including a review on the history of simulations, a discussion of
the different purposes for running simulations and a discussion on drawing con
clusions from simulations.
The second section focuses on approaches that are inspired by biological evo
lution. The first two papers by Frank Beckenbach and Tony Curzon Price ad
dress the use of genetic algorithms as a description of learning processes. Frank
Beckenbach analyses the usually claimed correspondence between learning pro
cesses and the processes represented by evolutionary algorithms on a quite gen
erally level. Tony Curzon Price, instead, discusses, based on simulating bidding
behaviour with the help of evolutionary algorithms, the insights we may expect
to gather from modelling human behaviour with computational techniques. The
indirect evolutionary approach is described by the third paper, written by Steffen
Huck, Wieland Muller and Martin Strobel who use this approach for explaining
attitudes towards risk. Finally, Koen Frenken, Luigi Marengo and Marco Valente
use the evolutionary idea to discuss in a very innovative way the decomposition
of tasks.
The third section addresses the use of neural nets and local interaction mod
els in economics. The first paper, written by Ralf Herbrich, Max Keilbach, Thore
Graepel, Peter Bollmann-Sdorra and Klaus Obermayer, provides a detailed de
scription of neural nets and their applications in economics, including a state-of
the-art application of a specific net to the learning of preferences. Robert E. Marks
and Hermann Schnabl connect genetic algorithms and neural nets with each other
by using them to describe learning in a repeated prisoner's dilemma and com
paring the results. An overview on the approaches to model social interactions
is given by Dorothea K. Herreiner. This contribution discusses the impact of the
local structure on the establishment of cooperation within a popUlation.
Section four contains some examples of boundedly rational and rational learn
ing models that are frequently used in economics. The first approach by Emilio
Barucci discusses different versions of least squares learning and applies these
models to the development of rational expectations. Martin Shubik and Nico
laas J. Vriend use a behavioural approach describing learning by a mixture of
a classifier system and an evolutionary algorithm and discuss their approach in
the view of simple consumption decisions. The third paper by Sardar Islam de
scribes Bayesian learning and applies this model to growth theory including the
aspect of uncertainty.
Section five is the most innovative part of the book because it contains three
papers on cognitive learning. Since cognitive learning has been widely neglected
in the economic literature there are no common models for cognitive learning.
Thus, the three papers of this section present three very different proposals of
ix
how cognitive learning can be captured in an computational approach. The paper
by Bruce Edmonds uses genetic programming as the starting point and modifies
this technique to describe cognitive processes in the EI Faro problem. Thomas
Brenner develops a specific kind of Moore automata, describes cognitive learn
ing as a dynamic processes of choice between these automata and applies this
approach to the repeated prisoner's dilemma. The third paper, written by Scott
Moss, describes economic agents as individuals using specific routines that are
changed over time and applies this approach to the problem of economic transi
tion.
The starting point for the creation of the present book was the workshop
"Agent-based and Population-based Modelling of Learning in Economics" which
was held at the Max-Planck-Institute for Research into Economic Systems in
March 1998. Therefore, I am indebted to the Max-Planck-Society which provided
me with the opportunity to organise this workshop and bring scientists working
on modelling learning in economics together to discuss their ideas.
Thomas Brenner
Evolutionary Economics Unit
MPI for Research into Economic Systems
Jena
List of Contributors
Emilio Barucci
Dipartimento di statistic a e matematica, applicata all' economia, Universita di Pisa,
Italy
Frank Beckenbach
Department of Social Sciences, University of Osnabrueck, Germany
Peter Bollmann-Sdorra
Statistical Research Group, TU Berlin, Germany
Thomas Brenner
Evolutionary Economics Unit, Max-Planck-Institute for Research into Economic
Systems, lena, Germany
Tony Curzon Price
ELSE, Department of Economics, University College London, UK
Bruce Edmonds
Centre for Policy Modelling, Faculty of Management and Business, Manchester
Metropolitan University, UK
Koen Frenken
INRNSERD, University Pierre Mends France, Grenoble, France
Thore Graepel
Neural Information Processing Group, TU Berlin, Germany
Ralf Herbrich
Statistical Research Group, TU Berlin, Germany