Complexity and the Economy Read online

  Complexity and the Economy

  Other Titles by W. Brian Arthur:

  The Nature of Technology: What It Is and How It Evolves

  The Economy as an Evolving Complex System II

  (co-edited with D. Lane and S. Durlauf)

  Increasing Returns and Path Dependence in the Economy

  Complexity

  and the Economy

  W. Brian Arthur

  1

  1

  Oxford University Press is a department of the University of

  Oxford. It furthers the University’s objective of excellence in research,

  scholarship, and education by publishing worldwide.

  Oxford New York

  Auckland Cape Town Dar es Salaam Hong Kong Karachi

  Kuala

  Lumpur Madrid Melbourne Mexico

  City Nairobi

  New

  Delhi Shanghai Taipei Toronto

  With offices in

  Argentina Austria Brazil Chile Czech

  Republic France Greece

  Guatemala Hungary Italy Japan Poland Portugal Singapore

  South

  Korea Switzerland Thailand Turkey Ukraine Vietnam

  Oxford is a registered trademark of Oxford University Press

  in the UK and certain other countries.

  Published in the United States of America by

  Oxford University Press

  198 Madison Avenue, New York, NY 10016

  © Oxford University Press 2015

  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, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by license, or under terms agreed with the appropriate reproduction rights organization.

  Inquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above.

  You must not circulate this work in any other form

  and you must impose this same condition on any acquirer.

  Library of Congress Cataloging-in-Publication Data

  Arthur, W. Brian.

  Complexity and the economy / W. Brian Arthur.

  pages cm

  Includes bibliographical references and index.

  ISBN 978–0–19–933429–2

  1. Evolutionary economics. 2. Economics–Psychological aspects.

  3. Economics. 4. Technological complexity. I. Title.

  HB97.3.A78 2015

  330.01′511352—dc23

  2014012789

  9 8 7 6 5 4 3 2 1

  Printed in the United States of America

  on acid-free paper

  To Brid, Sean, Niamh, and Ronan

  CONTENTS

  Preface ix

  Acknowledgments xxiii

  1. Complexity Economics: A Different Framework

  for Economic Thought 1

  2. Inductive Reasoning and Bounded Rationality: The

  El Farol Problem 30

  3. Asset Pricing under Endogenous Expectations in an Artificial

  Stock Market 39

  W. Brian Arthur, John H. Holland, Blake LeBaron, Richard Palmer,

  and Paul Tayler

  4. Competing Technologies, Increasing Returns, and Lock-In

  by Historical Events 69

  5. Process and Emergence in the Economy 89

  W. Brian Arthur, Steven N. Durlauf, and David A. Lane

  6. All Systems Will Be Gamed: Exploitive Behavior in Economic

  and Social Systems 103

  7. The Evolution of Technology within a Simple Computer Model 119

  W. Brian Arthur and Wolfgang Polak

  8. The Economy Evolving as Its Technologies Evolve 134

  9. On the Evolution of Complexity 144

  10. Cognition: The Black Box of Economics 158

  11. The End of Certainty in Economics 171

  12. Complexity and the Economy 182

  An Historical Footnote 189

  Other Papers on Complexity and the Economy 193

  Index 195

  [ viii ] Contents

  PREFACE

  Every so often a discipline gets thrown into a period of upheaval where its

  old ideas once taken for granted seem no longer so reliable, and its practitioners search for what to put in their place. Economics is in such a period now.

  This is partly due to the financial crisis of 2008, but the rethinking goes back to well before this. Slowly, over the last three or more decades, a feeling has grown among economists that their key assumptions of perfect rationality,

  equilibrium, diminishing returns, and of independent agents always facing

  well-defined problems are somehow not trustworthy, too restrictive, some-

  how forced. Now in the air are ideas of behavioral rationality, nonequilibrium, increasing returns, and of interconnected agents facing fundamental uncertainty in problems of decision-making. Economics has opened up to other

  approaches besides the standard neoclassical one.

  I have been heavily involved in one of the new approaches, complexity eco-

  nomics, so I decided this would be a good time to put together several of my

  earlier papers and bring them out in a collected volume. This collection, on

  complexity and the economy, dates from the mid-1980s to the present, and

  it follows my earlier one on increasing returns and path dependence in the

  economy.1

  None of these “new” ideas of course are really completely that new.

  Separately and in various forms they have been mooted by economists for

  years, sometimes even for a century or more. But what has been missing

  was the means to handle them, not just raw techniques but the mindset that

  would go with them, that the world is not perfect, that it isn’t machine-like, and that much of it cannot be reduced to simple equations—to variations in

  the number or level of entities. And missing too was a coherent framework for

  economics based on these new ideas.

  In the last few decades this has changed. The missing pieces have begun

  to fill in and techniques have slowly become available that can deal with the

  new assumptions. Among these are nonlinear dynamics, nonlinear stochastic

  1. Increasing Returns and Path Dependence in the Economy, W. B. Arthur, University of Michigan Press, 1994.

  processes, agent-based computation, and computational theory itself. The mindset too has changed. A feeling now runs across the sciences, and economics too, that the world is not a perfectly ordered system reducible in principle to mathematical equations, but is to a large extent organic and algorithmic—

  it proceeds by building on what is there already and it builds and changes step by step. Slowly, as a result of these occurrences, economics is developing an

  approach based on these more realistic assumptions. It is developing a new

  framework for economic thought.

  The collected papers in this book reflect my part in the development of

  this new framework. Taken together they view the economy as a system not

  necessarily in equilibrium, but as one where agents constantly change their

  actions and strategies in response to the outcome they mutually create, a sys-

  tem where agents are constantly creating an “ecology” of behaviors they must

  mutually adapt to. This viewpoint has roots of course in complexity thinking

  as it developed in the 1970s in groups in Brussels, Stuttgart, and Ann Arbor.


  And it has roots in the work of individual researchers in universities such as Stanford and MIT. But in its current economic form it grew largely from work

  at the Santa Fe Institute. In the late 1980s a small group of researchers at

  the nascent Santa Fe Institute began systematically to look at the economy as

  an evolving complex system. I headed that group for its first two years, and

  have been associated with it ever since, and in this collection of papers I want to show how these ideas developed and how the economics they led to came

  about.

  The papers in this volume were not the outcome of some planned process.

  They arose haltingly and over several years, and were heavily influenced by

  my colleagues and by thinking in general at Santa Fe. Several appeared in

  well-known journals, others appeared in places more obscure. Many were

  written in Santa Fe, others were written at Stanford. The papers present fin-

  ished thinking but not why or how that thinking came about, so it will be

  helpful to the reader to understand the background to them and the context

  in which they arose.

  Most of them started with a single incident.

  In April 1987 I was walking toward my office in Stanford when a helmeted

  Kenneth Arrow swung round me on his bicycle and stopped. He was putting

  together a group of economic theorists in September to exchange ideas with

  a group of scientists that his counterpart, physicist Philip Anderson, would

  propose. The venue would be a small institute in the Rockies just starting up.

  It was in Santa Fe. Would I like to come? I said yes immediately without being sure of what I was committing myself to. The idea looked promising.

  The conference in Santa Fe when I got there a few months later turned out

  to be a more heavyweight affair than I’d imagined. Among the ten or so econ-

  omists Arrow chose were Larry Summers, Tom Sargent, Jose Sheinkman, and

  [ x ] Preface

  William (Buz) Brock. Among the ten or so scientists Phil Anderson chose were John Holland, David Ruelle, Stuart Kauffman, and David Pines. The

  meeting was held in the chapel of a convent the new institute was renting

  and there was nothing rushed about it. A participant would talk in the morn-

  ing and we would discuss, another participant would talk in the afternoon

  and again we would discuss. We were learning not just solutions to problems

  in the others’ disciplines, but about what each discipline saw as a problem,

  and how it thought about these, and what mindset it brought to bear on

  these problems. Questions not normally raised within economics were

  raised—why do you guys cling onto perfect rationality? Why do you assume

  so much linearity? And questions were asked of physics too. Why is a prob-

  lem “solved,” say in spin glasses, when it has not settled to a steady state?

  Chaos theory and nonlinear dynamics were discussed in both economics and

  physics. Modeling of positive feedbacks and of interactions, again in both

  disciplines, was discussed. People would meet at night in twos and threes to

  talk over ideas and problems.

  The meeting was exhilarating—and exhausting. Nothing had quite been

  solved by the end of the ten days, yet the physics side was left with a respect for the sheer complicatedness of the economy—the elements in the economy

  (people), unlike the ions in a lattice, could decide what to do next not just

  based on the current situation of themselves and other elements, but on what

  they thought those other elements might do given what they might do. And the economists were left with a feeling for modern physics, for its interactions and nonlinearities, its multiple possible end states, its lack of predictability—

  indeed for its complicatedness.

  Word began to leak out after the conference that something interesting

  had happened at Santa Fe and the new institute’s Science Board decided it

  would follow the conference up by initiating a long-term research program

  on the Economy as an Evolving Complex System. John Holland and I were

  asked to come to Santa Fe the following year to head this. I had a sabbatical

  coming from Stanford and accepted, John found it harder to get away from

  Michigan and declined. So I found myself heading up the Santa Fe Institute’s

  first research program; it would start in August the following year, 1988.

  My immediate problem of course, working from Stanford, was to put

  together a team of first-rate people for the new program and to decide its

  direction. Some people I already knew from the conference. John Holland

  promised to come for a couple of months, and the physicist Richard Palmer for

  much longer than that. Stuart Kauffman would be in residence. From my own

  network I was able to bring in David Lane and Yuri Ermoliev, both excellent

  probability theorists. Arrow and Anderson helped greatly. Where I found it

  hard to cajole people to join in, Arrow or Anderson, both Nobel Prize winners, could simply lift the phone and quickly get people to join us. As to direction I was less sure. Early on, the physicist Murray Gell-Mann suggested to me that Pr eface [ xi ]

  we come up with a manifesto for doing economics differently. I didn’t quite have the confidence for that; in fact I didn’t yet know what topics we would

  go after. I had done quite a bit of work already on complexity and the econ-

  omy, but now we had a much broader reach in what topics we might research.

  From the conference it was assumed that chaos theory would be central, but

  the idea somehow didn’t appeal to me. Vaguely I thought that we should look

  at increasing returns problems, which I was more than familiar with, at how

  some of the physics methods could be transferred into economics, and at non-

  linear dynamics in the economy. Also we might be able to do something inter-

  esting with computation in economics.

  When the program opened finally in 1988 we discussed directions further,

  still groping for a way forward. I phoned Ken Arrow from Santa Fe and asked

  for his advice and Phil Anderson’s. They got in touch with the funder of the

  program, John Reed of Citibank, and the word came back: Do what you want,

  providing it deals with the foundations of economics, and is not conventional.

  For me and the others on the team, this directive seemed like a dream. We

  had carte blanche to do what we wanted, and at Santa Fe we wouldn’t have

  colleagues from the discipline looking at us and asking why we were doing

  things differently.

  In fact, outside our small team the few colleagues we did have were from

  physics or theoretical biology. Stuart Kauffman was one, and we immediately

  included him in the program. There was little else in the way of researchers the new institute could offer. It was in its earliest days and was all but unknown, an experiment, a small startup in the Rockies set up to have no students, no

  classes, no departments, and no disciplines—no discipline, the wags said.

  We had discussions, mainly in the convent’s kitchen, and I remember in

  an early one Kauffman said, Why do you guys do everything at equilibrium?

  What would it be like to do economics out of equilibrium? Like all economists

  I had thought about that, but not seriously. In fact the question took me

  aback, and it did so with the other economists. I had no good answer. It fell

  into the category of questions such as w
hat would physics be like if the gravitational force were suspended, something that seemed perfectly thinkable as

  a thought experiment, but strange. And yet Kauffman’s question stuck. We

  retained the question but we were still looking for a direction ahead.

  One of the directions that interested me was still half formed. It had come

  out of the conference the previous year. In an after-lunch talk the first day of that conference, John Holland had described his work on classifier systems,

  basically systems that are concatenations of condition-action rules. One rule

  might say that if the system’s environment fulfills condition A, then execute action R. Another might say, if it fulfills condition D, execute action T. A third might say that if A is true, and R-being-executed is not true, then execute action Z. And so on. The actions taken would change the environment, the

  [ xii ] Preface

  overall state of the system. In this way you could string such if-then rules together to get a system to “recognize” its environment and execute actions

  appropriately, much as an E. coli bacterium “recognizes” a glucose gradient in its environment and swims in an appropriate direction. Moreover, you could

  allow the system to start with not-so-good rules and replace these with better ones it discovered over time. The system could learn and evolve.

  As Holland talked about this I found myself deeply excited, and I checked

  the room to see if other economists were similarly taken with these ideas.

  There was no evidence; in fact one of them was taking a post-lunch nap. A feeling grew in me that somehow, in some way, this was an answer and all we

  had to do was find the question. Somehow Holland was describing a method

  whereby “intelligence” or appropriate action could automatically evolve

  within systems. I quizzed John later about his ideas. We were sharing a house

  in Santa Fe for two months at that time in 1987, but in several conversations

  neither of us could work out what these ideas might directly have to do with

  economics.

  I had gone back to Stanford, where I was teaching a course in economic