Evolution Kurf PDF

Economics, Evolution and the State
Economics, Evolution
and the State
The Governance of Complexity
Edited by
Kurt Dopfer
Professor of Economics, University of St Gallen, Switzerland
Edward Elgar
Cheltenham, UK • Northampton, MA, USA
© Kurt Dopfer 2005
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, electronic,
mechanical or photocopying, recording, or otherwise without the prior
permission of the publisher.
Published by
Edward Elgar Publishing Limited
Glensanda House
Montpellier Parade
Cheltenham
Glos GL50 1UA
UK
Edward Elgar Publishing, Inc.

136 West Street
Suite 202
Northampton
Massachusetts 01060
USA
A catalogue record for this book

is available from the British Library
ISBN 1 84542 438 7
Printed and bound in Great Britain by MPG Books Ltd, Bodmin, Cornwall
Contents
List of contributors vii
Introduction 1
Kurt Dopfer
PART I ECONOMIC EVOLUTION AS OPEN PROCESS
1 Heterogeneity and evolutionary change – concepts and

measurement 13
Uwe Cantner and Horst Hanusch
2 Is the notion of progress compatible with an evolutionary
view of the economy? 43
C. Christian von Weizsäcker
3 Reconciling evolutionary economics with liberalism 58
Gerhard Wegner
4 Historical economics and evolutionary economic policy –
Coasean perspectives 78
Matthias Klaes
PART II SELF-ORGANISATION AND NETWORKS
5 The concept of network organisation – biotechnology-based

industries as exemplar 99
Andreas Pyka and P. Paolo Saviotti
6 Sociodynamics – an integrated approach to modelling in the
social sciences 120
Wolfgang Weidlich
7 The concept of space in trade – some evolutionary basics 140
Carsten Herrmann-Pillath
8 Economic policy – a process of communication 168
Lambert T. Koch
9 Why endogeneity is not enough to explain technological
change – a critique of Paul Romer 190
Malcolm H. Dunn
v
vi Contents
PART III THE POLITICAL ECONOMY OF COMPLEXITY
10 Innovation and the learning policy maker – an evolutionary

approach based on historical experience 215
Joachim Schwerin and Claudia Werker
11 The national German innovation system – its development in
different governmental and territorial structures 239
Hariolf Grupp, Icíar Dominguez Lacasa and Monika
Friedrich-Nishio
12 Emergence and diffusion of disastrous innovations – a case
study 274
Reiner Peter Hellbrück
13 Applying evolutionary economics to public policy – the
example of competitive federalism in the EU 296
Wolfgang Kerber
14 Can evolutionary economics make a billion $ difference for
60 per cent of the world’s poor in Asia? 325
Hans-Peter Brunner
Index 351
Contributors
Hans-Peter Brunner, Asian Development Bank, Manila
Uwe Cantner, Friedrich-Schiller-Universität Jena
Kurt Dopfer, Universität St Gallen
Malcolm H. Dunn, Universität Potsdam
Monika Friedrich-Nishio, Institute for Economic Policy Research (IWW),
Karlsruhe University
Hariolf Grupp, Fraunhofer Institute for Systems and Innovation Research
(ISI) and Karlsruhe University
Horst Hanusch, Universität Augsburg
Reiner Peter Hellbrück, Fachhochschule Würzburg-Schweinfurt
Carsten Herrmann-Pillath, Universität Witten-Herdecke
Wolfgang Kerber, Philipps-Universität Marburg
Matthias Klaes, Keele University
Lambert T. Koch, Bergische Universität Wuppertal
Icíar Dominguez Lacasa, Fraunhofer Institute for Systems and Innovation
Research (ISI), Karlsruhe
Andreas Pyka, Universität Augsburg
P. Paolo Saviotti, INRA, Université Pierre Mendes-France, Grenoble
Joachim Schwerin
Gerhard Wegner, Universität Erfurt
Wolfgang Weidlich, Universität Stuttgart
C. Christian von Weizsäcker, Frontier Economics, Köln
Claudia Werker, Max-Planck-Institut für Ökonomik, Jena
vii
Introduction
Kurt Dopfer
A NEW BRANCH: EVOLUTIONARY ECONOMIC

GOVERNANCE
The last two decades have seen an explosion of research in evolutionary

economics and related fields. There has been an upsurge in the publications
addressing evolutionary themes in economics, paralleled by the foundation
of new journals, new academic societies and, last but not least, a reorien-
tation of the publication programmes of major publishing houses.
The potential of this knowledge stock with respect to its application to
issues of economic governance – in the line of an evolutionary economic
governance – has as yet received little attention in the literature. Evolutionary
economic governance can be defined as the study of the possibilities and
consequences of any external intervention into structure and processes of
an evolving economic system. Specifically, evolutionary economic policy is
the application of evolutionary governance to the economy as a whole.
The notion of ‘evolutionary’ acknowledges, on the one hand, that con-
tinual change in economic systems is an empirical fact, and it calls, on the
other hand, for a theoretical account of this dynamic on the basis of an
evolutionary rather than mechanistic explanatory schema. Thus, both
explanandum and explanans are evolutionary.
The notion of ‘governance’, in turn, means that the analysis is not confined
to an economic system in isolation from external instances of mutual or uni-
directional influences. Evolving systems can be, and often are, embedded in
a hierarchy of decision-making strata. Evolutionary economic policy is a
special case of governance that applies to the highest level of aggregation
and decision-making authority in an economic system. In its interdisciplin-
ary and long-run perspective evolutionary policy governance resembles
much classical political economy.
The 14 contributions to this volume address the issue of evolutionary
economic governance at various levels of aggregation and of decision
making. Inspired by evolutionary thinking a range of models emerges that
highlight the variety and the complexity of evolutionary governance of eco-
nomic systems.
1
2 Introduction
The contributions were selected from a set of papers presented at

two annual conferences of the ‘Expert Group for the Study of Evolu-
tionary Economics’ (Ausschuss für Evolutorische Ökonomik) devoted to
the topics ‘Evolutionary Economics: Methodological, Econometric and
Mathematical Foundations’ and ‘Evolutionary Economic Policy’. The com-
plete set of papers of the two conferences was published in the German lan-
guage as volumes VII and VIII of the series ‘Studien zur Evolutorischen
Ökonomik’ (Dopfer, K.(ed.) (2003, 2004), Berlin: Duncker and Humblot).
INSTITUTIONAL SETTING
Before giving a summary introduction of the various contributions to the

volume, a few remarks on the institutional setting in which they came about
will be made. The research group mentioned is an institution within the
so-called Verein für Socialpolitik which is the professional association of
German-speaking economists – the German Economic Association – with
members coming from all over the world, but mostly from Germany,
Austria and Switzerland.
The Verein was founded in the year 1873. The activities of the new association
focused on opposing laissez-faire in social policy and the revolutionary social
ideas of emerging socialism. The founders of the association wanted ‘to raise,
educate and reconcile the lower classes on the basis of the existing order’ as
Gustav Schmoller, the chairman of the Verein for many years (1890–1917), put
it. In Schmoller’s time, the Verein für Socialpolitik started to develop into a polit-
ically neutral, interdisciplinary society. (www.socialpolitik.org)
The Verein voluntarily disbanded in 1936 to avoid having to bow to the

ruling political party. It was refounded in 1948.
While its over 3000 members discuss their scientific findings at the society’s
meeting annually (with 700 papers at the 2004 meeting in Dresden), the most
concentrated research efforts are probably made in the regular meetings of
the expert groups. The Expert Group for the Study of Evolutionary Economics
has emerged from an unofficial working group organised by Ulrich Witt in
1989 and became an official Ausschuss of the Verein in 1991. It is the youngest
and presently largest and fastest developing of the 23 expert groups of the
Verein.
The various expert groups reflect the variety of the scientific activities of
its members and mirror its history, but the dominant doctrine of the Verein
(if there is any) revolves around neoclassical economics. The likelihood that
a majority of its members may contradict such a statement may signify the
particular kind of neoclassical economics that is being followed.
Introduction 3
The contributions to this volume are grouped around three major topics:
Economic evolution as open process

Self-organisation and networks
The political economy of complexity.
ECONOMIC EVOLUTION AS OPEN PROCESS
Directedness in Economic Evolution
In their chapter on ‘Heterogeneity and evolutionary change – concepts and

measurement’, Uwe Cantner and Horst Hanusch deal with an empirical
approach to heterogeneity as both the force and the result of evolutionary
change, that is the result of as well as the source for local technological
change. Heterogeneity here is understood as differences in the technological
performance of productive units. Those differences are due to the local appli-
cation and local innovation of specific production techniques resulting in
structures of asymmetric productive efficiency and/or in structures of variety
in the production technologies in use. Investigating those structures and their
development empirically requires a specific measure for technologically het-
erogeneous performances and a specific methodology which allows the
detection of this heterogeneity. With respect to the former the authors
propose total factor productivity (TFP) and changes of TFP as measures for
technological heterogeneity and its dynamics. From a methodological point
of view a non-parametric empirical approach appears appropriate to iden-
tify heterogeneous structures and to track them over time. Structures of
asymmetric performances as well as of variety can be identified by non-
parametric best-practice production or efficiency frontiers. Computing the
Malmquist-productivity index allows one to track those structures over time
and to identify local changes. Discussing these concepts and procedures the
chapter concludes by raising a number of unresolved issues.
C. Christian von Weizsäcker in his chapter is concerned with the ques-
tion: Is the notion of progress compatible with an evolutionary view of the
economy? The measuring rod for the performance of an economy is the
fulfilment of the needs of its members as evaluated by themselves accord-
ing to their preferences. Neoclassical economics has developed policy
advice making the assumption of exogenously given preferences. This
assumption is incompatible with an evolutionary approach to economics.
In evolutionary economics it is assumed that preferences ‘evolve’, hence
are endogenously determined. Is then policy advice still possible, if the
measuring rod changes with the object to be measured? Von Weizsäcker
4 Introduction
develops a theory of adaptive preferences which fulfils two requirements at

the same time: (1) adaptive preferences can be seen as the ‘laws of evolu-
tion’ of preferences, thereby being compatible with an evolutionary
approach in economics; (2) a theorem can be derived which ensures that
welfare economics is possible with adaptive preferences, hence endogen-
ously determined preferences. The central idea is to show that normative
comparability of different allocations A and B can be obtained consistently
by means of ‘improvement paths’ going from A to B (or vice versa, but not
both) even if preferences induced by A would make A preferred to B and
preferences induced by B would make B preferred to A, hence would make
A and B normatively incomparable.
Gerhard Wegner’s contribution to this volume aims at reconciling evo-
lutionary economics with liberalism. Although evolutionary economics
highlights the innovation process as the driving force of economic welfare,
its political implications are far from being unambiguous and have not
resulted in a new conceptualisation of economic policy. It is an open ques-
tion whether evolutionary economics should stick to the logical structure
of deductive normative economics, that is to deduce welfare-promoting
policies from a revised ‘non-stationary’ welfare concept. This chapter casts
doubts on such deductivism and advocates a liberal interpretation of
evolutionary economics. Namely, economic policy should not be concep-
tualised to promote welfare directly but to ensure the institutional precon-
ditions for a rise in welfare. It is argued that evolutionary economics
supports the view that economic welfare results from experiments which are
ultimately based on economic ‘world views’ or beliefs and lack objectivity;
for that reason, their ongoing assessment by a competitive environment is
required. Generally, economic policy is well advised to foster economic
freedom as well as competition in order to make sure that rival economic
ideas cannot outperform each other except through competition. From an
epistemological point of view, evolutionary economics denies that a com-
prehensive description of economic opportunities can be made from the
perspective of a theoretical observer. For that reason alone economic
policy should desist from ‘constructing’ superior allocative states in con-
crete terms.
In ‘Historical economics and evolutionary economic policy – Coasean
perspectives’, Matthias Klaes focuses on evolutionary aspects in Coase’s
thinking. The recent literature evaluating Coase’s work in the light of evo-
lutionary economics has come to distinguish between two Coasean tradi-
tions, one appropriated by the economic mainstream, the other identifying
a significant heterodox potential. It is the aim of Klaes’ paper to take the
latter reception of Coase’s legacy as the starting point for exploring his
contribution in the light of recent debates on an evolutionary approach to
Introduction 5
economic policy. The author defends the view that contrary to attempts to
label Coase as an arch neoliberal, one can identify a commitment to the
primacy of institutional direction rather than decentralised allocation of
resources across markets as the core building block of his outlook on eco-
nomic policy. This leads to an interpretation of Coase’s approach in terms
of a historico-empirical method of comparative institutional analysis,
which calls both for a hermeneutical approach to economic policy evalua-
tion, and constitutes a form of historical economics that offers an ambi-
tious research agenda for attempts to move the conceptualisation of history
in evolutionary economics beyond affirmations of the relevance of path-
dependent processes. What Coase ultimately offers to evolutionary eco-
nomics is the prospect of a morphological analysis that exhibits interesting
parallels to recent developments in evolutionary ‘Evo-Devo’ biology that
have re-established the relevance of ontogenetic approaches to develop-
mental processes.
Self-organisation and Networks
Andreas Pyka and P. Paolo Saviotti in their chapter examine ‘The concept
of network organisation – biotechnology-based industries as exemplar’.
Technological progress in the biological sciences is now advancing across
such a wide range and at such a pace that no firm can hope to keep up in all
the different areas. Participating in innovation networks offers an alterna-
tive to extremely expensive go-it-alone strategies. This imbalance between
the rate of growth of the biotechnology knowledge base and the capability
of individual firms to access it can explain the persistence of cooperative
R&D in the biotechnology-based sectors at the end of the 1990s. Such
imbalance is not due any more only to the lack of absorptive capacity of
large pharmaceutical firms, because they have meanwhile developed con-
siderable competencies in that field. The authors are proposing that a new
role, that of explorers scanning parts of the knowledge space that LDFs
(large diversified firms) are capable of exploring but unwilling to commit
themselves to in an irreversible way, can be played by DBFs (dedicated
biotechnology firms) in innovation networks. The authors’ simulation
approach attempts to represent the emergence of these two roles as endo-
genous changes in the motivation for participating in innovation networks,
allowing them to become an important and long-lasting organisational
device for industrial R&D.
The contribution by Wolfgang Weidlich deals with ‘Sociodynamics – an
integrated approach to modelling in the social sciences’. Sociodynamics – an
important branch of the emergent field of econophysics – aims at providing
a general strategy for designing mathematical models for the quantitative
6 Introduction
description of a broad class of dynamic collective phenomena within human

society. These design principles have been applied in the past to models of
non-equilibrium economics as well as to those of group dynamics, political
opinion formation, migration and urban evolution. Weidlich’s contribution
highlights the essential formalism of the so-called master-equation, and dis-
cusses specifically aspects of the choice of variables, transition rates and the
evolution of socioconfigurations. The socioeconomic method is applied to
the case of interregional migration of socioculturally different populations
and to that of the evolution of a city with hinterland connections. Solutions
of the model equations for a range of meaningful trend parameters are
offered for both applied models.
Carsten Herrmann-Pillath discusses in his paper ‘The concept of space
in trade – some evolutionary basics’. Recent research in the theory of inter-
national trade shows the serious limitations of its theoretical foundations.
The chapter proposes alternative foundations based on the concept of
transaction and the proposition that the capabilities of economic agents to
realise transactions cannot be priced within a market system. It is thus
impossible to include the trading competences of international traders in
the general concept of factor, and to apply neoclassical approaches, such
as the factor-proportions theory, on the basic endowments of countries and
of agents that would enable them to transact successfully across the border.
The chapter proposes a totally new concept of factor, based on the prop-
erties of storability, non-rivalness of use in time, non-producibility and
non-tradability. These factors determine country-specific competences to
trade, which are classified in a threefold way as organisational capital, col-
lective human capital, and social capital.
The aim of Lambert T. Koch’s ‘Economic Policy – A Process of
Communication’ is to understand and analyse policy and economic policy
in particular as an evolutionary process. Agents in this process are by no
means only those who are in the public eye as responsible parties for policy
and decision making. Every policy has its roots beyond the official politi-
cal structures, in areas where there is ‘private’ dissatisfaction, where there is
a need seen for action, where this kind of perception is shared with others,
where potential policy content is defined and, finally, where critical masses
are created which can generate pressure for policy action. The view of the
policy-making process to be taken in this chapter goes beyond traditional
approaches by emphasising the relevance of cognition and communication.
It moves away from the implicit idea currently prevailing, even in well-
informed circles, that there are ‘born’ state tasks, ‘right’ packages of meas-
ures to implement ‘objective’ aims and scientific methods to find out what
‘good’ policies are. The basic credo of an evolutionary theory of economic
policy according to the author is to see policy as being the product of
Introduction 7
communicated, related perceptions. These perceptions – which influence

evaluations of policy conditions, formulation of objectives and packages of
measures – are themselves transformed by the results of running political
processes. This doubly recursive structure in itself leads to innovations,
giving the phenomenon its evolutionary character.
The chapter by Malcolm H. Dunn, ‘Why endogeneity is not enough
to explain technological change – a critique of Paul Romer’, analyses the
inner logic of one of the most prominent models of endogenous growth:
Paul Romer’s contribution ‘Endogenous Technological Change’. Although
Romer’s model captures important features of technological advance, such
as the relevance of externalities and imperfect competition, it also neglects
many aspects that are equally important. The paper argues that Romer’s
growth theory is based on a production function which gives an inadequate
account of the dynamics between the existing stock of capital and know-
ledge on the one hand and the factors leading to the emergence of new prod-
ucts on the other hand. The chapter then attempts to show that technological
change can be better interpreted as an open process which is driven by entre-
preneurs acting under true uncertainty. Drawing on the rich pool of insights
provided by the tradition of evolutionary economics a more complex story
behind the emergence of innovation behaviour is suggested. As for the
methodology the chapter suggests that more emphasis should be placed on
forces which escape an easy formalisation, for instance the achievement
motivation and competence of market participants and the system of prop-
erty rights.
Political Economy of Complexity
In their contribution on ‘Innovation and the learning policy maker – an evo-

lutionary approach based on historical experience’, Joachim Schwerin and
Claudia Werker discuss how to design, implement and perform a learning
economic policy approach in line with the principles of evolutionary theory.
Socio-economic systems evolve in ways that cannot be fully anticipated.
These dynamics pose a challenge to policy makers, who have to devise a
framework that needs to co-evolve with the socio-economic system. The
knowledge stock on which politicians base their decisions to modify the set
of policy rules plays a pivotal role in this process. Here, learning is key to
success: Based on past experience, politicians need to constantly improve
their knowledge by turning learning into a routine systematically linked with
modifications of policy rules. The chapter starts with a methodological
analysis that identifies types of rules on which policy should be based to
institutionalise learning. On this basis the authors present a practically fea-
sible concept of learning policy built upon the identification of historical
8 Introduction
invariant patterns. They apply this concept to the analysis of innovation and
growth, where they demonstrate that the openness of the set of identified
patterns towards a changing environment is a prerequisite for learning.
Their results, which lead to some policy guidelines, serve as a reference
system for innovation policy that meets evolutionary standards.
‘The national German innovation system – its development in different
governmental and territorial structures’ by Hariolf Grupp, Icíar
Dominguez Lacasa and Monika Friedrich-Nishio provides several long
time series of R&D indicators bridging more than 100 years of German
economic history. The ups and downs are put into perspective and com-
pared to important events in the global innovation system. The pertinence
found for the national German innovation system seems to rest in educa-
tional, cultural and language traditions rather than in governance and
territorial coverage, which changed many times. This is possibly due to path
dependence of technological paradigms. The suggested range of indicators
on a national or sectorial level gives a detailed impression of both the extent
and the contents of innovation activities. The empirical base which evolu-
tionary researchers, interested in innovation and economic history-related
questions, can rely on, has been broadened to a large extent, so that there
is no longer a serious empirical gap.
Reiner Peter Hellbrück’s ‘Emergence and diffusion of disastrous inno-
vations – a case study’ examines diabetes care in Germany. There is a con-
sensus among all specialists that diabetes care needs to be improved. Hence,
there has been steady development, improvement, and implementation of
programmes for diabetes care in Germany. In virtually all German lands
there is at least one (legal) health insurance fund which is offering a diabetes
programme. For this purpose special contracts between at least one health
insurance fund and some health care providers have to be signed. These
contracts contain an extra budgeting payment to the health care providers.
Those contracts have the charm for the health care providers of earning,
supplementary to their regular budgets, extra money. Health insurance
funds have in turn been hurt by concluding such contracts when evaluated
on a regulation ruling from 1993 up to 2000. The chapter tries to answer
the following questions: What reasons led to the conclusion of such con-
tracts? Why are these contracts under today’s laws, rules, and regulations a
hazard to health insurance funds and their business? Why did those disas-
trous contracts diffuse at all? Which strategies are chosen to reduce the
hazard to insurance business? Why are these business-harming innovations
not simply stopped?
In his chapter ‘Applying evolutionary economics to public policy – the
example of competitive federalism in the EU,’ Wolfgang Kerber states that
due to the Hayekian knowledge problem it seems often to be very difficult
Introduction 9
to derive positive policy proposals from evolutionary reasoning; he argues

in favour of a pragmatic approach, as to how evolutionary argumentation
should be used in policy discussions as important complements to other
(mainly neoclassical) arguments. Starting with the hotly discussed policy
question of centralisation vs decentralisation of public policies in the EU,
it is shown that the traditional economic theory of federalism is mainly
based upon neoclassical theories, which do not take into account the cre-
ation and diffusion of new policies in a satisfactory way. Therefore the
theory of federalism should be supplemented by important evolutionary
reasoning about the innovation and imitation of public policies. The main
evolutionary argument is that decentralised experimentation can lead to
more policy innovations and a greater degree of mutual learning from
the experiences with public policies (laboratory federalism). Therefore the
application of evolutionary (and innovation) economics leads to the con-
clusion that a more decentralised EU (with a system of competitive feder-
alism) can be expected to lead to more innovation in regard to public goods,
legal rules and regulation.
‘Can evolutionary economics make a billion $ difference for 60 per cent
of the world’s poor in Asia?’, Hans-Peter Brunner asks in his contribution
to this volume. The use of evolutionary and institutional economics in the
analysis of increasing returns, transaction costs, and agglomeration and
network effects is important because of their determining impacts on the
processes of trade, growth and development. Institutional analysis facili-
tates definition of the role of governments, and of international agencies, in
the design of projects and provision of services in sectors such as transport,
communications, and financial services. The application of institutional
methods and evolutionary models that relate positive externality and
linkage effects to actual representations of institutional and physical infra-
structures results in a powerful tool for the economic development of
financial sectors and trade corridors. The question posed is whether wide
application of institutional and evolutionary methods would yield higher
payoffs for growth, equity, and poverty reduction. The cases of Nepal’s
financial sector reform, and of the design of West Bengal’s trade and trans-
port corridor, show the huge difference evolutionary economics can make.
It is worth the effort. Why then is evolutionary analysis not employed more
often? The answer appears to lie in unfamiliarity with the methods and
results. The chapter ends with practical suggestions as to how evolutionary
and institutional economics could become a household tool for multilateral
development institutions.
PART I
Economic evolution as open process

1. Heterogeneity and evolutionary
change – concepts and measurement
Uwe Cantner and Horst Hanusch
1. INTRODUCTION
In this chapter we present a methodological discussion on the identification

and the tracking of structures over time. In this respect the problem under
consideration is central to evolutionary economics, which puts much effort
into the explanation of how structures come into existence and how they
develop and change over time. The structures we are looking at are based
on the heterogeneity of actors – where actors are seen in a rather broadly
defined sense including also e.g. countries. To make things more easily
accessible we constrain this heterogeneity to that which is closely related to
technological and innovative abilities on the one hand, and to imitative
behaviour and technological learning on the other. In theoretical work this
so-called technological heterogeneity is discussed in a dynamic context
focusing on its sources as well as on its consequences for innovative devel-
opment and technological change. Among other things this work deals
mainly with the question of how knowledge spillovers arising out of tech-
nologically different actors do influence the direction and intensity of tech-
nological change,1 the structural and competitive development within
sectors,2 and the cross-fertilisation effects between sectors.3
This chapter, however, will discuss empirical issues related to hetero-
geneity and its change over time. The main goal we want to pursue is to
coming up with a rather general empirical approach or procedure allowing
one to identify and to track technological structures among actors. For this
we suggest a three-step procedure:
1. select the variables you consider relevant for identifying technological

structures and their change over time;
2. apply a methodology – to be explained below – and identify those
structures and their change;
3. test whether the identified structures and their change can be explained
by variables related to innovative activities, research output etc.
13
14 Economic evolution as open process
This chapter will be dealing partly with the first and mainly with the second
step. For the third step we refer to the literature.
The methodology suggested is of a rather general character in the
sense that it can be applied on several levels of aggregation, i.e. firms,
sectors or countries. This rather common conception about technological
heterogeneity is applied with respect to the productive structure of a sample
of observations under consideration. We suggest that those structures can
be dealt with looking at (a) the differences in total factor productivity
and (b) the differences in input-intensities, output-intensities and/or output
coefficients.
In order to identify such heterogeneous structures and to track their
development and their change over time we suggest a specific empirical
approach. For this purpose we introduce a two-step analysis consisting of
the non-parametric procedure to construct production or efficiency frontiers
and the Malmquist-productivity index to track the change of this structure
over time.
The plan of the chapter is as follows. In section 2 we will briefly refer to
some theoretical issues describing, firstly, the role heterogeneity plays in the
evolutionary framework; secondly, we ask for the criteria to be applied in
order to determine the analytically relevant heterogeneity in empirical data
and put forward three requirements for the empirical approach to be
employed. In section 3 we refer to the measure total factor productivity, dis-
tinguish our procedure from the traditional way of computing this measure
and briefly introduce the main methodological approach we have chosen in
order to detect heterogeneity and to track its development over time. The
final section 4 summarises our approach and discusses some as yet unre-
solved issues and problems.
2. THE CONCEPT OF HETEROGENEITY
In this section, what we attempt to briefly discuss is the concept of hetero-

geneity both in a more general way and more specifically with respect to
the economic theory of technological change and innovation. For the
further discussion this initial step allows us to propose a rather well-
defined analytical frame constructed on the basis of a conception of het-
erogeneity which aims at differences in the technological performance of
agents. This, of course, is an extraction out of all possible sources and
instances of heterogeneity, but it enables us to focus on that kind of het-
erogeneity we consider analytically relevant in the economic theory of tech-
nical change.
Heterogeneity and evolutionary change 15
2.1 Theoretical Issues
Heterogeneity
In economics, there are a number of distinguishing elements of the
neoclassical and the evolutionary approach. Probably one of the most
important ones is the heterogeneity of behaviour, attitudes, characteristics
etc. of agents.4 Thus, what heterogeneity is all about is asymmetry among
the agents in a set. However, it is not at all obvious whether this asymmetry
matters for the description of the state of this set or for its development
over time. In neoclassical approaches one would deny this in general with
the consequence that the theoretical models suggested are characterised by
symmetry of agents or even by a representative agent.5 All the approaches
designed in this way are justified by the attitude that for the final outcome
of a certain process the differences in agents’ behaviour during this process
do not matter – it is just average behaviour which is determining the result
and which analytically is relevant and interesting. Hence, heterogeneity is
of an only temporary nature and as such it is a phenomenon only showing
up during transitory dynamics.
Heterogeneity or asymmetry is a fundamental principle in the theories of
economic evolution. Selectionist approaches, synergetic approaches and
developmental approaches rely on it and discuss how the system’s nature or
structure as well as how its (structural) dynamics is affected or driven by it.
However, in each approach the way heterogeneity affects evolutionary
development is quite specific: in the selectionist approach it is heterogeneity
which is reduced by competition and generated by innovation. In the syn-
ergetic approach it is heterogeneity which brings about specific structural,
self-organising features with respect to learning, cooperation etc. In the
developmental approach, finally, heterogeneity is a matter of stages of
development (to be) passed.
Heterogeneity as asymmetry and variety

Heterogeneity is a concept which refers to the degree of difference within
a population of observations, be they households, firms, sectors or even
regions or countries which differ with respect to their efforts, behaviours
and/or success due to – among other things – the artifacts they consume or
produce, the modes of production they employ, the direction and intensity
of innovative activities they pursue, or the organisational setting they
choose. This heterogeneity of agents is, on the one hand, considered the
result of technological change, i.e. of different innovative/imitative/adap-
tive activities and differential innovative/imitative/adaptive success; on the
other hand, it also serves as a source for further progress in the sense that
this heterogeneity puts pressure on technologically backward actors to
improve performance when the gaps become too large and on leaders when
the gaps become too close; and that it provides for different kinds of learn-
ing processes (imitative and adaptive learning, cross-fertilisation etc.).
In order to account for the heterogeneity of agents driven by and driving
technological change one draws on the close relationship between the char-
acteristics and the behaviour of agents, on the one hand, and the kind of
inputs which, on the other hand, they transform into outputs. In fact, in the
theory of technological change the actors are characterised by the nature,
the level and the degree of their innovative activities – either on the input
or on the output side.
In this respect heterogeneity can be accounted for by the conception of
variety (Saviotti 1996). This concept is based on the number of distin-
guishable elements of a set of artifacts.6 In this sense Saviotti (1996, p. 94)
distinguishes output and input variety, the former being the number of
distinguishable outputs and the latter taking account of the number of dis-
tinguishable types of processes.
However, heterogeneity in general and within the context of innovative-
ness in particular is not only a matter of simply counting distinguishable
elements. Any innovator attempts to perform better than competitors, and
this ‘better’ may show up in providing goods and services with superior
price–quality ratios – compared to those of the competitors. Thus, more
often than not one would like to have a conception which allows for a quan-
tification of the differences on which heterogeneity rests. Hence, with respect
to output variety one would be interested whether the variety observed is
also built upon measurable quality differences (i.e. ‘higher’ and ‘lower’
quality) or whether this variety is found within a more narrow or more broad
range of the specific characteristic under consideration (i.e. ‘more’ or ‘less’
built-in features). Equivalently, with respect to input or production variety
we should have an account of whether the several techniques in use are
rather similar or far different with respect to their efficiency (i.e. ‘more’ or
‘less’ efficient) or their relative input requirements (e.g. ‘more’ or ‘less’
capital-intensive).
An appropriate conception in this respect is found in Dosi (1988,
pp. 1155–7) who is concerned with the asymmetry of activities and distin-
guishes variety as a special case of asymmetry. Both are to be seen in a
context where firms engaged in innovative activities are affected differently
by technological change in terms of their process technologies and quality
(or kind) of output. Whenever firms can be ranked as ‘better’ or ‘worse’
according to their distance to some technological frontier he refers to as
asymmetry. The degree of asymmetry of an industry is then its dispersion
of input efficiencies for a given (homogeneous) output and price-weighted
performance characteristics of firms’ (differentiated) products. For all
differences or asymmetries among firms which cannot be ranked as

unequivocally better or worse he refers to variety. This may be the case
when (a) firms producing the same good with identical costs employ
different production techniques or (b) when firms search for their product
innovations in different product spaces, embodying different product char-
acteristics and aim at different corners of the markets.
The concept of heterogeneity

Based on this discussion, in the following we will be concerned with hetero-
geneity which is as closely as possible related to technological performances
and their differences – thus, heterogeneity is meant to be technological het-
erogeneity and it is based on the local application of certain technologies. As
such it, first, includes performances which can be compared directly to each
other and by this means be ranked – i.e. producing a certain product with a
higher/lower quality or running a specific production process more/less
efficiently. Second, this conception comprises also technological perform-
ances which cannot be compared to each other directly – i.e. producing
different products in the sense of old and new or running quite different
production techniques. These latter performances cannot be compared
directly (in terms of physical measures) and one has to rely on some other
measures such as the comparative economic success of those performances
(as measured in terms of profitability, market share, growth rates, etc.).
Technological heterogeneity on higher aggregation levels

Moreover, the technological heterogeneity we are concerned with is not
only confined to the technological performance of individual actors. It is
also applicable to higher levels of aggregation such as the sectoral, the
regional or the national level. By way of aggregation, of course, any sub-
level heterogeneity gets covered and only some – here no matter how
defined – average characterises the higher-level unit. Despite this inevitable
loss of information involved here, we nevertheless expect considerable and
relevant heterogeneity in technological performance of sectors, regions or
economies with respect to the product and quality range produced (e.g.
agricultural products, Germany compared to India) and/or the kind and
degree of certain production techniques employed (e.g. cotton production,
US compared to Pakistan). Accordingly, what we mean by locally applied
technologies also – with a loss of specific description – refers in a more
aggregate sense to sectors, regions and countries.
Heterogeneity and local progress

Heterogeneity as just introduced can be considered as a snapshot descrip-
tion of a sample of observations. Especially in the context of technological
progress it is quite obvious and has to be expected that heterogeneity is

subject to change. One could easily think of exogenous forces which quite
equally affect all agents. However, for endogenous changes which are pro-
voked by individual action it is just as much obvious that the respective
changes are to a considerable degree specific to a certain agent or group of
agents – that is the change we are concerned with is local technological
change.7 And even if we considered a number of agents as behaving rather
similarly, e.g. in catching up to the technology leaders, such progress is local
as well in the sense that only a subgroup of the agents under consideration
achieved that.
Equivalently to our discussion of heterogeneity the concept of local tech-
nological change is applicable to several levels of aggregation. In this
respect, technological change is specific to a certain country (e.g. the US
compared to Togo), to a certain region (e.g. East Asian Tigers compared to
Western Europe), or to a certain sector (e.g. in machinery, Germany com-
pared to Japan). Of course, and again equivalent to the above, local change
on higher levels of aggregation hides local changes on lower aggregation
levels, so that only an ‘average’ change shows up.
2.2 Empirical Issues
To state the importance of heterogeneity is one side of the coin, the other
one is to clearly specify in which unit we should measure or observe het-
erogeneity and this in a way that it is also analytically relevant.
To identify heterogeneity in empirical work is not an easy and straight-
forward task at all. In principle, one is facing a similar problem to that the
typological approach is confronted with: what is essential for analysing
the issue under consideration? Whereas the typological approach searches
for some reliable average characteristic, the population perspective is con-
fronted with the task of finding characteristics which are diverse, i.e. het-
erogeneous, and as such essential for performance and the progress of the
population under consideration. Looking for variables which can render
this, one more often than not has to be engaged in rather detailed analyses
of a nearly case-study type. Although the results are often very illuminat-
ing and interesting, it is very often not possible to transfer the methodology
and the results of one study to another one; the aggregation of several
results is often not possible, because the relevant variables are not of the
same type, and so on.
Heterogeneity and innovation

Let us now look more carefully at the theory of technological change and
innovation. How can we measure technological performance? To give an
answer we start with another question: What does technological progress

provoke, how do we distinguish an innovation from a well-known old
artifact?
Here it is quite obvious that innovations provide for heterogeneity
because something new – a ‘new combination’ in Schumpeter’s (1935/1912)
words – is introduced into the market. This may be a better technique of
production, a better organisational structure, a better product quality, or
an entirely new product. Hence, the innovator introducing this new combin-
ation can be distinguished from competitors just by his/her innovation.
Thus, more generally, technological heterogeneity is just the consequence
of differential innovative success accumulated up to the present. In a
dynamic context, with respect to several features of the process of techno-
logical change such as path dependency, cumulativeness etc., this hetero-
geneity can also serve as an indicator for the direction and success probability
of further innovative activities – such as innovation, imitation, adoption etc.
The central question arising out of this is concerned with the measure we
should apply in order to account for this innovation and technology related
or determined heterogeneity.
Specific versus general measures

Of course, we could have a long list of possible characteristics or features
which perform the task of detecting the effects of technological change and
innovations. All of the characteristics used in technometrics are based on
rather technical issues. Look for example at the technological development
of helicopters so well studied by Saviotti: technological progress here is
represented by the development in technical characteristics such as engine
power, rotor diameter, number of engines etc. Or, look at computer chips
where we are informed about technological progress by steadily increasing
storage capacities. Or, finally look at automobile production where tech-
nological progress or organisational progress shows up in the increased
number of cars assembled within one hour (Fordism) or in the decreased
number of bad quality cars assembled (Toyotism).
However, for all its merits this quite specific and quite exact technical
measurement has also its drawbacks:
1. Despite the respective specific characteristics of a technology and

despite the fact that its development can be represented relatively
exactly, it also brings it about that the more exactly one measures
specific features the less a comparison between different observations
will be possible and meaningful.
2. Whenever different technologies and their respective progress are
analysed, a comparison of the results is less likely to be possible.
3. Any aggregation from the business unit to the firm, to the sector and
industry, to the regional or even to the macro-economic level is not pos-
sible anymore. The reason for that is quite obvious, because to aggre-
gate the products of different firms in a sector, the products of different
sectors in an economy cannot be accomplished when technical attrib-
utes are used such as pieces, kg, Mb etc.
On the basis of the following four central requirements we suggest and

introduce a measure and empirical procedure which attempts to circumvent
the problematic issues just raised and which allows to analyse empirical
observations within a theoretical framework aiming at locally applied tech-
nologies and local technological change. For each requirement we give a
brief suggestion here – a fuller discussion is found in section 3.
Requirement 1 The task therefore is to detect a measure which may help to

overcome these problems. Thus, what one has to look for in this context is a
measure which on the one hand is exact enough and on the other hand is not
that specific so that the above-mentioned deficit will not show up. Hence,
what we look for is a measure which serves this purpose and is applicable to
a broad range of innovative phenomena on different levels of aggregation.8
Suggestion 1 In order to show the way for setting up a broadly applicable

taxonomy we suggest the measure of total factor productivity (TFP) and
its change over time to play a major or even pivotal role in this endeavour.
This suggestion, at first glance, might look somewhat old-fashioned as the
concept of total factor productivity has been criticised intensely in the
past – mainly in the context of growth-accounting exercises where its con-
struction is based on equilibrium assumption and conditions of traditional
production theory combined with the notion of the same production to be
applied to all observations. This leads us to a second requirement.
Requirement 2 The way TFP is measured should differ considerably from

standard procedures. As such, as a first step, it should allow one to distin-
guish innovators from imitators and account for better and for worse tech-
nological performance. Moreover, it should deliver a quantitative account
of these differences or asymmetries.
Suggestion 2 With respect to requirement 2 we suggest applying a frontier

analysis where the frontier function or technology frontier is set up by the
best-performing observations. All worse-performing observations are at
some distance from this technology frontier where this distance can be used
as a measure for different technological performance.
Requirement 3 Related to the need to distinguish better from worse per-

formance is the requirement that, following the evolutionary approach, the
empirical analyses should not be restrictive in the sense that a functional
relationship, e.g. a specific production function, is a priori assumed to hold
for all observations. One should rather allow for an open number of those
relationships and take into account also variety in productions functions or
output mixes.
Suggestion 3 For satisfying requirement 3 we suggest the computation of

TFP measures by a non-parametric procedure to determine technology fron-
tiers which – at least compared to the traditional approaches of TFP index
numbers, parametric production functions and parametric production
frontiers9 – are rather unrestricted in the functional form employed for the
aggregation of inputs and outputs, respectively. In principle, there are as
many functional forms allowed for as a sample contains observations.
Requirement 4 The measure applied should be tracked over time. The

respective measure of the change in TFP should be able to take account of
local technological change.
Suggestion 4 In this respect we suggest employing the procedure to

compute the Malmquist productivity index which just measures the change
in TFP. The important feature of this measure is that it allows one to iden-
tify local technological change at both the technology frontier as well as the
below-best-practice observations.
3. TFP, TECHNOLOGICAL PROCESS

AND EVOLUTIONARY THEORY
3.1 TFP as a Measure of Technological Performance
Referring to requirement and suggestion 1 we consider total factor pro-

ductivity and its change over time as an appropriate measure for techno-
logical performance and technological change. This, of course, requires
some qualifications.
Generality
As already claimed we are interested in a generally applicable measure which
allows one to track technological change on several levels of aggregation
and in several fields of application. Thus, what we have to accept is a loss of
specificity especially found if one applies the analysis on lower levels of
aggregation often coming close to pure case studies. The loss of specificity,
however, is counterbalanced (and in our view even overcompensated) by the
opportunity to detect more general insights into structure and change
whose driving elements are found on the individual level of actors and firms,
whose collective outcome then shows up in a characteristic manner on the
next level of aggregation and so forth. In this respect the measure of total
factor productivity is applicable to all levels and areas of aggregation when-
ever we have at hand appropriate data on outputs and inputs. The latter we
assume to be satisfied – although we are readily aware of the problem of
whether the data we use in the analysis are at all the appropriate measures
and/or proxies. We do not want to discuss this issue further here.
Construction
Index numbers for total factor productivity (TFP) have found a prominent
application in growth-accounting exercises. There it is aggregate output Y,
prominently GDP, set into relation to an aggregate I of various input
factors, prominently labour and capital:
Y
TFP
I
One can easily apply this measure to lower levels of aggregation such as to
the sectoral level10 and to the firm level. Any change in total factor product-
ivity, in the sense that this indicator rises, is considered as the effect of tech-
nological progress, i.e. that change in output which cannot be accounted for
by a change in aggregate inputs. It is this so-called residual which attracted
so much research especially in the analysis of economic growth. And it is
also this residual that Abramovitz called ‘our measure of ignorance’.
A first question arising in this context refers to whether TFP can be taken
as a measure of technological performance and whether a change in total
factor productivity can adequately account for technological change. Let
us take up this issue accordingly.
TFP as performance indicator

In order to account for the performance of an observation the indicator
applied is to be interpreted always as a relative measure, either with respect
to some known optimal performance or with respect to the best perform-
ance observed. In empirical work it is always the latter relativisation that is
employed. For this comparison to work, however, one has to provide for
1. the categories used for measuring inputs and outputs and

2. the respective way of aggregating of inputs and outputs in order to
compute the TFP.
being identical among the observations. Otherwise the comparison is

inadequate.
To cope with the first problem, one has to look for measures which allow
for homogeneous input and output categories. This is certainly not the case,
but by some degree of abstraction or cleverly chosen units of measure-
ment – in the sense of real units (e.g. hours worked, kg, etc.) – one might
be able to cope with this problem – at least partly.
As to the aggregation functions for inputs and outputs, with respect to
inputs it is just the production function that is searched for and which has
a number of specific problems. We do not want to go into detail here but
only remind the reader that on the theoretically founded perceptions of
techniques applied locally and local technological change an aggregation
or production function identical for all observations cannot be expected a
priori – contrariwise heterogeneity is to be expected.
With respect to outputs the problem is similar whenever we are not in the
lucky situation of having to consider only one homogeneous output.
Again, this is not the normal case and among the observations we normally
have to expect both differences in the quality of the output as well as
differences with respect to the number of outputs produced. A common
way to deal with this is to accept product prices as weights which account
for quality differences as well as differences in kind.11 This leads to output
measures such as GDP, sectoral sales or firm sales. Besides this, however,
one might also be interested in dealing with output variety in a disaggre-
gated way such as splitting up GDP into the output of various sectors or
of firms, sales into the sales of different products. A possible way of per-
forming this is presented below.
Change of TFP as measure of technological change

Interpreting the change of TFP as a measure for technological progress
faces the same problems as just stated. Whenever we consider process
innovations allowing the given resources to produce more of a homoge-
neous output, the change in TFP appropriately takes account of this.
However, dealing with quality improvements or new products, whenever
quantity and price changes account for this in a proper way we can use the
aggregate output. But if we were interested in the development beneath the
level of aggregation it would be helpful to have the respective TFP change
determined on the basis of a disaggregated TFP index.
Other influences on TFP

A final remark here refers to differences in the TFP which are not due to
the respective technological performance of the observation. Proper can-
didates are vintage structures as well as economies of scale.12 For the
change in TFP we should additionally be aware of substitution effects at

work. With respect to substitution effects according to Rosenberg (1976)
substitution along a traditional isoquant is to be considered as applying a
technique not applied yet and this could also be considered as technical
change. Moreover one may argue here that even in a purely neoclassical
context the distinction between substitution, scale and progress effects is
not possible without imposing rather restrictive assumptions.
Having given some justification and qualifications on the TFP measure
as a response to requirement 1 we now want to go one step further and
introduce a method for taking care of requirements 2 and 3.
3.2 Structure: a Non-parametric Frontier Function Approach
Requirements 2 and 3 ask for a method which allows one to determine TFP
in such a way that technological heterogeneity in the sense of asymmetry
and variety shows up. For this purpose we suggest a non-parametric fron-
tier function approach.
Unrestricted performance measure

The non-parametric frontier function approach (or DEA for data envelop-
ment analysis) basically relies on index numbers to measure total factor pro-
ductivity in a fashion similar to the one used in more standard productivity
analysis. In a sample of n observations for each observation j (j 1, . . ., n)
a productivity index hj is given by:
uTYj
hj (1.1)
vTXj
Here Yj is a s-vector of outputs (r 1, . . ., s) and Xj a m-vector of inputs

(i 1, . . ., m) of observation j. The s-vector u and the m-vector v contain
the aggregation weights ur and vi respectively.
The hj in (1.1) is nothing else than an index of total factor productivity.
The respective aggregation functions (for inputs and outputs respectively)
are of a linear arithmetic type as also employed in the well-known
Kendrick-Ott productivity index.13 There, however, by special assumptions
the aggregations weights, ur and vi, are given exogenously. The non-para-
metric approach does not rely on such assumptions – in particular, it is not
assumed that all observations of the sample have a common identical pro-
duction function. With this – at least to a certain degree – unconstrained way
of aggregating both inputs and outputs we are able to account for require-
ment 3 above. The parameterisation of the aggregation functions and thus
the aggregation weights which may be specific to a certain observation are
determined endogenously. They are the solution to a specific optimisation

problem (as discussed below), and therefore they are dependent on the
empirical data of the sample. Critics often argue that a linear arithmetic
aggregation nevertheless presupposes at least a special type of production
function,14 such as the Leontief-type production function.15 However, those
arguments do not take into account the following:
1. Since the aggregation weights are determined endogenously and can be

different between observations, there ultimately exists a number of
parametrically different possible aggregation functions although they
are all of the same type.16
2. Moreover, the linear aggregation function holds for a Leontief type of
production function as well as a linear one.
3. For the input side, moreover, the fact that the Leontief production
function fits well into this framework suits well the widely held assump-
tion of short-run limited substitutability of production factors when-
ever techniques employed are of a local character.
4. The procedure does not have to rely on production functions at all but
fits well into a framework dealing with production correspondences.
This interpretation is also quite appealing to an evolutionary frame-
work where a neoclassical production function is not considered suit-
able for an evolutionary analysis.
This unrestricted form of the total factor productivity measure is central to an

application of this method to evolutionary analysis and to detecting hetero-
geneity in particular. For computing this index we can include all different
kinds of inputs and different types of outputs. This implies also that new
products can be taken care of and equivalently new production factors. Since
the non-parametric approach does not require all inputs to be employed or
outputs to be produced by each observation we are readily able to take into
account both product innovations and new techniques of production.
Having found a rather unrestricted mode for measuring the performance
of an observation we would like to provide also a comparison of this per-
formance with those of the other observations in the sense that we find
statements about ‘unequivocally better’, or ‘unequivocally worse’ or even
‘not comparable’.
Comparison of performance
For doing this, the basic principle of the non-parametric approach is just
to determine the indices hj in such a way that they can be interpreted as
efficiency ratings which implies a comparison of each observation with the
best observation(s). The (relatively) most efficient observations of a sample
are evaluated by h1, less efficient observations by h1. Hence, by com-

paring all observations with each other we arrive at an account of different
technological performance where the differences are quantified in the
measure h – this is just what requirement 2 asked for.
The following constrained maximisation problem is used to determine
such an h-value for a particular observation l,l{1,. . .,n}:
uTYl
max hl
vTXl
uTYj
st: 1; j 1, . . ., n; (1.2)
vTXj
u,v 0.
Problem (1.2) determines hl of observation l subject to the constraint that
the hj of all observations (including l itself) of the sample are equal or less
to 1. The constraints provide that h is indexed on (0,1). Moreover the ele-
ments of u and v have to be positive. This requirement is to be interpreted
that for all inputs used and outputs obtained there must exist at least a pos-
itive efficiency value.17
Best-practice or frontier functions

Since we employ linear arithmetic aggregation functions for inputs and
outputs, (1.2) is a problem of linear fractional programming.18 To solve
such optimisation problems, there exist a number of methods the best-
known of which is Charnes and Cooper (1962). They suggest transforming
(1.2) into a standard linear program which then can be solved with the well-
known simplex algorithm. Performing this step and transforming the
resulting primal to its corresponding dual problem, one arrives at the well-
known Charnes/Cooper/Rhodes (1978)19 envelopment form of the non-
parametric approach:
min l
st:
Yl Yl
lXl Xl 0 (1.3)
l 0
Yl and Xl are the s- and m-vectors of outputs and inputs respectively of

observation l, Y and X are the s n-matrix of outputs and m n-matrix of
inputs of all observations of the sample. The parameter l to be minimised
expresses the percentage level to which the inputs of observation l can be
reduced proportionally, in order to have this observation producing on
the production frontier representing the best practice technologies – it is

identical to hl and is a relative measure of technological performance.
With l 1 the respective observation belongs to the efficient observations
on the frontier. Proceeding in this way and solving (1.3) for all observations
in the sample, the non-parametric approach determines an efficiency frontier
or technology frontier constructed by the best-practice observations. The
efficiency rating of each observation is measured relative to this frontier.
Figure 1.1a states this result for a sample of observations which produce
with two inputs, x1 and x2, one unit of output y. The technology frontier
determined is DAB. The technological performance is the relative distance
of an observation from the technology frontier. In the case of observation
C, the measure C is given by the ratio OC
to OC.
The n-vector l states the weights of all (efficient) observations which serve
as reference for observation l. For the efficient observation l (with l 1), we
obtain 1 for the lth element of l and 0 for all other elements. Grouping all
observations according to their respective reference observations allows one
to detect regions of similarity. These regions are distinguished by different
input intensities, output intensities or input coefficients. In terms of Figure
1.1a, for observation C the reference observations are A and B.
Consequently only A C and C are different from 0. The respective values state
B
the degree to which A and B are respectively used to construct C

.
In figure 1.1b we show how the problem looks in output space when one
considers more than one output. Here, the observations produce with one
x2/y
C
A C
Ft
B
O
x1/y
Figure 1.1a Technology frontier and the measure c

y2/x
D
A
C
C
B
Ft
O y1/x
Figure 1.1b Technology frontier in output space
unit of input x two outputs, y1 and y2. The frontier function is a transform-
ation curve. The performance of an observation C is again evaluated
against this frontier here given by the ratio OC
to OC.20
So far, the observations of sample are considered as using all the same
categories of inputs and to produce all the same kinds of output – although
to a different degree. In addition to that the procedure suggested is readily
able to deal also with more diversity such as observations which use an add-
itional type of input or produce an additional kind of output. This latter is
appropriate if we deal with product innovations. Figures 1.1c and 1.1d
show the kind of technology frontier we get in such cases in the input space
(1.1c) and in the output space (1.1d).
A characterisation of the structure of a sample

So far the discussion has delivered an account of a sample which allows one
to detect and quantify heterogeneity in productive performance. With
program (1.3) we are now readily able to characterise the structure of a
sample of observations:
1. as a measure of performance indicating and quantifying whether an

observation is best practice or below best practice;
2. as a measure of structural (dis)similarity (Cantner 1996).
However, modifying program (1.3) some measures can be computed which

shed additional light on the structure of a sample.
x2/y
C
A
C
Ft
B E
O E
x1/y
Figure 1.1c Extreme input variety and technology frontier in input space
y2/x
D
A
C
Ft
E B
O y1/x
E
Figure 1.1d Extreme output variety and technology frontier in output

space
Comparison of best practice

Since the frontier function is quite regularly constructed by several best-
practice observations which cannot be ranked as better or worse, one might
additionally be interested in a comparison between them. The following
modification of program (1.3) allows for this where now the observation
under consideration l is not a member of the reference set:
min *l
st:
Yl l Yl
*lXl Xl l 0 (1.4)
l 0
The matrices Yl and Xl contain the outputs and inputs of all n observa-
tions except observation l. The modified efficiency measure is *l . For all
below-best-practice observations it is identical to l determined by program
(1.3). However, for all best-practice observations *l is different. It holds
*l 1 and the difference *l 1 can be interpreted as the buffer or lead
observation l holds compared to certain other observations. This *l is a
measure to distinguish observations which with program (1.3) are deter-
mined as not comparable (Cantner/Westermann 1998).
Figure 1.2 states the result of program (1.4) for observation A. The
respective frontier for A in this case is DB and the * A is equal to the ratio
OA
to OA which is larger than 1.
x2/y
Ft
B
O x1/y
Figure 1.2 Comparison of best-practice observations

Besides this mode of comparing best practice observations an alternative

or additional way is to look at the dynamic performance, i.e. their compar-
ative ability to shift the frontier function (by technological progress). This
issue will be taken up below.
Accounting for scale effects

Finally, since the programs used so far have been discussed under the
assumption of constant returns-to-scale technologies one might be inter-
ested in taking into account size effects. This is done by first setting up a
program allowing for non-constant returns to scale. This leads to a formu-
lation where the elements of the l vector have to sum up to 1 (eT is a vector
containing only elements 1):
min vl
st:
Y l Yl
l Xl X l 0
v (1.5)
eTl 0
For the efficiency measure determined by program (1.5) we get vl l.
Taking the ratio of these two measures, l l vl, states the level of
efficiency which is due to scale with 1 l accounting for that degree of
below best practice which is caused by a size different to the minimum
efficient scale size.
Besides these measures the non-parametric frontier approach does
deliver a number of other measures allowing one to deal with allocative
efficiency, non-radial inefficiencies, specific forms of returns to scale etc.
These are of minor importance in the context of this chapter. More inter-
esting, however, is the dynamic extension of the analysis.
3.3 Structural Dynamics: Local Technological Change, Catching up and

Falling behind
The structures identified so far are of course interesting in their own

respect. However, in a dynamic evolutionary context the static performance
of observations is only part – probably the minor part – of the whole story.
This distinction is rather appealing when one thinks of static and dynamic
efficiency (Stiglitz 1987). For this reason a dynamic measure of perform-
ance is required.
Therefore, the following discussion refers to requirement 4 asking for an
appropriate way of dealing with localised technological change and thus
the structural dynamics induced.
Dynamic analysis
In order to track the structure – determined by the measures introduced
above – it is by no means sufficient to compare the structural results of con-
secutive periods because for each period these measures are of an only rela-
tive type such a comparison makes no sense. Consequently, consecutive
periods have to be set into relation implying that we have to compute rela-
tive measures which compare period t with t 1 and vice versa. Doing this
pairwise for all consecutive periods allows one to track structural change
over time. The procedure chosen for this purpose is based on the Malmquist
index which states a specific observation’s change in productivity between
two periods. A quite interesting feature of this index is that it can be decom-
posed into a measure for technological change and one for catch-up – or, of
course, technological regress and falling behind.
Malmquist index The theoretical basis of the Malmquist productivity

index is found in the work of Malmquist (1953), Solow (1957) and
Moorsteen (1961). For productivity measurement this index has been
applied by Caves/Christensen/Diewert (1982a, 1982b). Färe/Grosskopf/
Lindgren/Roos (1994) have shown how the efficiency measure above can be
used to compute the Malmquist index. We will follow this line of reasoning.
In order to explain what the Malmquist productivity index measures
we refer to Figure 1.3 which contains a simple example of two non-
parametric production frontiers Ft and Ft 1 pertaining to period t and t 1.
x2/y
At+1 d
At
b
e
Ft
c
Ft+1
O x1/y
Figure 1.3 Malmquist-productivity index

For measuring the productivity change of observation A from At to At 1

consider the following. First evaluate At and At 1 towards the frontier Ft
and compute the ratio of the two results. For this we get Ob/OAt divided by
Od/OAt 1; if this ratio is less than 1 the productivity of A increased.
Second, and in addition to that we could also evaluate At and At 1 towards
the shifted frontier Ft 1; again we determine the ratio, here Oc/OAt divided
by Oe/OAt 1; this ratio less than 1 implies a productivity improvement. In
a final step the geometric mean of these two computations is taken.
The resulting index

0,5
ObOAt OcOAt
A
Mt 1
OdOAt 1 OeOAt 1
the Malmquist-productivity index, states the productivity change of A

between t and t 1. In a general way, the Malmquist-productivity index
Mt 1
l measuring the productivity change of observation l from t to t 1 is
defined as follows:
lt,t t,t 1

0,5
Mt 1 l
(1.6)
l
t 1,t
l t 1,t 1
l
t,s, t,sT, is the efficiency of observation l in period t whenever the fron-

tier function of period s serves as reference measure.21
Decomposition of the Malmquist index With some manipulation we can

develop (1.6) to the following expression for the Malmquist index:
lt,t t,t 1

0,5
Mt 1 l
l
t 1,t
l t 1,t 1
l
lt,t
t 1,t 1 t,t 1

0,5
l l
l
t 1,t 1 lt 1,t t,t
l
MC· MT. (1.7)
The second line in (1.7) states the decomposition of the productivity change
into technological progress MT and change in the technology gap MC.
Whenever MC1 (MC 1) we find catch-up (falling behind). The second
term is MT and indicates the movement of the frontier. This is measured
twice: first with the factor intensities of l in t, and a second time with
those of l in t 1. With MT1 (MT 1) we have technological progress
(technological regress) at the frontier. Looking at our example in Figure 1.3

this decomposition is given by the following ratio of distances:

ObOAt OeOAt 1 OcOAt 0.5
A
Mt 1
OeOAt 1 OdOAt 1 ObOAt
MC MT.
From this, we can state that the first bracket term measures the change in
the distance of A towards the frontiers Ft and Ft 1. The second term in
brackets takes account of the (geometric) mean change of the frontier part
pertaining to A. In this example both terms will be smaller than 1 indicat-
ing that observation A performed technological progress and was able to
catch up to the frontier.
Local change As is readily apparent from Figure 1.3 the productivity

change in (1.6) is local in the sense that it is specific to the observation under
consideration. In this respect the degree of this local change depends (a) on
the observation’s ability to shift in direction to the origin and (b) on the
behaviour of the frontier. As to (b) the respective change is also local in the
sense that for observation l it is only relevant how the respective part
of the frontier assigned to l (by the way of the elements of the -vector)
shifts. The decomposition of the index allows one to distinguish these two
movements.
Moreover, the decomposition allows one also to evaluate best-practice
observations in a dynamic context by comparing them by the way of the
index MT and thus by the ability to locally shift the frontier function.
3.4 Summary of the Issue
In face of the theoretical and empirical requirements stated in section 2 we

have suggested measuring total factor productivity by a procedure which is
as unconstrained as possible but nevertheless allows one
1. to systematise heterogeneity and

2. to track its change accomplished by technological progress in general
and local technological progress in particular.
For this purpose we apply a non-parametric procedure to determine frontier

functions. These consist of the best-practice observations in a sample and
do not rely on any common a priori parametrically given production func-
tion. We thus dispense with any notion of the neoclassical production func-
tion and rely entirely on production techniques which in the short run show
no substitutability among production factors, i.e. which could be described

by a Leontief-type relationship between output and input.
For the dynamics we apply the Malmquist index measuring productivity
change by comparing the non-parametric production frontiers and obser-
vations of consecutive periods. In this way we dispense with the notion that
technological progress shifts the entire production frontier and instead we
allow for (a) parts of the frontier to shift and for (b) this shift not to be
proportional.
With respect to heterogeneity and its change this two-step procedure
performs or detects the following. The first step of this two-step procedure
allows one to detect heterogeneity – here technological heterogeneity – and
classify the observations into the following categories:
1. heterogeneity in the performance of running a specific technique, class

or range of techniques
2. heterogeneity in applying a specific technique out of a larger range of
possible techniques.
The second step then tracks this heterogeneity over time and allows for the
following:
1. measuring local technological change

2. distinguishing between progress of the best-practice techniques or
forging ahead and dynamics of catching up, falling behind
Taken literally, the procedure suggested does classify the observations in a

specific way in both a static and a dynamic context. In this way we do not
have to rely a priori on restrictive assumptions or constraints which force
the observations to behave in a certain way, e.g. to obey to the same para-
metrically given production function.
4. CONCLUSION
This chapter deals with empirical analysis in evolutionary economics in

general and innovation economics as a prominent application of evolu-
tionary ideas in particular. Within the latter, heterogeneity in the sense of
different innovative activities, different production processes employed,
different qualities or goods produced, is a major analytical element – the
more so because innovative actors aim at creating heterogeneity and imita-
tors attempt to reduce it again. This heterogeneity has an additional feature
to be accounted for, the performance of the different techniques, activities,
goods, etc. under consideration. Thus, it is not only a counting of different

elements in a set but also the evaluation of these elements due to their static
and their dynamic performances.
The task to be performed by empirical analyses contains three steps or
problems: (1) defining the heterogeneity which is analytically relevant;
(2) evaluating the performance of the heterogeneous entities; (3) testing
whether the structural development of the entities can be explained by evo-
lutionary conceptions.
In this chapter we focus mainly on the two first steps. The third one
requires much more space and cannot be presented in an appropriate way
here. With respect to steps (1) and (2) we suggest a measure and procedure
which are applicable to all levels of aggregation – micro, meso and macro –
and which rests on a comparison of total factor productivities of the enti-
ties under consideration. The procedure we suggest is as unrestricted as
possible: in the static analysis of the non-parametric frontier function
approach aiming at the identification of structures there is no restriction
on the production technique employed or the output mix produced. In the
dynamic analysis performed by the computation of Malmquist-produc-
tivity indexes the local character of technological change is allowed to
work and to be identified. By this ‘twin procedure’ the heterogeneity and
the differences in performance, so central to innovation, can be accounted
for.
Of course, the twin procedure presented provides opportunities for
further development; several shortcomings have to be overcome:
1. Stochastic elements could be included or the frontier conception

chosen could be switched. An example of the latter is found in Cantner/
Hanusch (1996) where we investigate a frontier function in the sense of
best practice up to period t.
2. The relations of the aggregated and the disaggregated analyses have to
be investigated in much more detail.
3. A major problem is also the rather unrestricted form of the procedure
which by definition allows as many production functions or output
mixes as observations. Does this imply that a representative sample
cannot be used to explain the behaviour of the whole population?
4. Obviously, all the results presented and the future research agenda are
dependent on the quality and the number of the data available:
● The coverage of the data with respect to the time period under con-
sideration is one point.
● Another one refers to the degree to which the respective variables
are an appropriate measure for the technology, the activity the
output mix under consideration. Related to this is the problem of
using data which represent economic values – on the output side for
example if one uses sales: then we do not measure the effects on
technology but the economic evaluation of these effects.
● Measurement errors are of considerable influence on the results as
there is no way in which the procedure can deal with this.
5. Finally, the research we attempt to follow aims at shedding some light
on the phenomenon of total factor productivity and its development.
In many applications this is still a black box or residual. To achieve a
better understanding of this residual the procedure suggested might be
an promising way to go.
Finally we want to refer to some empirical applications of this approach such

as Bernard/Cantner/Westermann (1996), Cantner/Hanusch/Westermann
(1997), Bernard/Cantner (1999), Krüger/Cantner/Hanusch (2000). An
analysis of how the structural results found are related to technological
change is performed in Krüger/Cantner/Hanusch (2000), Bernard/Cantner
(1998) and Cantner (1996).
NOTES
1. E.g. Cantner/Hanusch/Pyka (1998).

2. E.g. Cantner/Pyka (1998a, 1998b) Cantner (2000).
3. E.g. Cantner (1996).
4. Other distinguishing features are the different conception of uncertainty, of rationality,
etc.
5. Analyses dealing with asymmetric information are an exception to this.
6. An obvious candidate would be this number itself. According to Saviotti (1991, p. 177)
in information theory the variety of a set is just the logarithm in base 2 of the number
of distinguishable elements.
7. See Atkinson/Stiglitz (1969); a good overview is found in Antonelli (1994).
8. Saviotti (1996, p. 52) states: ‘. . . a taxonomy at all levels of aggregation in such a way
that the relationships of the various units of analysis within and between each level of
aggregation can be analysed’.
9. For a discussion and comparison of the non-parametric approach with more traditional
methods see Cantner/Krüger (1999).
10. E.g. Wolff (1992).
11. This comes close to what Dosi (1988, pp. 1155–7) claims to be price-weighted perform-
ance characteristics of firms’ (differentiated) products.
12. Differences might occur also due to scale effects and/or vintage structures (Dosi 1988,
p. 1156)
13. See Ott (1959).
14. See also Chang/Guh (1991, p. 217).
15. Leontief (1947) has shown that a linear aggregation exists for a Leontief-type production
function. Instead of a Leontief function one could also use a linear production function.
16. Employing parametric methods, e.g. the COLS or the EM-algorithm a specific produc-
tion function is assumed. The coefficients of this function are estimated using the
available data and the resulting production function is used to determine technical
(in)efficiencies of all the firms in the sample. This procedure, however, suggests that there
is only one ‘best-practice’ technology. With the non-parametric approach a number of

‘best-practice’ technologies can be determined.
17. This procedure is also known from activity analysis.
18. An overview of linear fractional programming is given in Böhm (1978).
19. There obviously exists a range of possible model specifications where the one chosen is
known as CCR. Applying this one has to keep in mind that possible scale inefficiencies
are included in the technical inefficiency measure.
20. Here the efficiency measure is allowed to take values out of [1,[ with 1 as best practice.
This is due to an output-oriented design of the method which starts not with a product-
ivity index as in (1) but with the inverse ratio.
21. For the respective programs required to compute the several -measures see Appendix
to this chapter.
REFERENCES
Antonelli, C. (1994), The Economics of Localized Technological Change and

Industrial Dynamics, Dordrecht et al.: Kluwer Academic Publishers, 1994.
Atkinson, A. and J.E. Stiglitz (1969), ‘A new view of technological change’,
Economic Journal, 79, 573–78.
Bernard, J. and U. Cantner (1998), ‘Retombées technologiques, effects de rattrapage
et dynamique intra-sectorielle,’ in J. Bernard and M. Catin (eds), Les Conditions
Économiques du Changement Technologique, Paris, Montreal: L’Harmattan,
pp. 95–116.
Bernard, J. and U. Cantner (1999), ‘French regional performance and variety – A non-
parametric frontier approach, in P.Y. Badillo and J.C. Paradi (eds), La Méthode
DEA- Analyse des Performances, Paris: Hermes Science Publications, pp. 255–73.
Bernard, J., U. Cantner and G. Westermann (1996), ‘Technological leadership and
variety – a data envelopment analysis for the French machinery industry’, Annals
of Operations Research, 68, 361–77.
Böhm, K. (1978), Lineare Quotientenprogrammierung – Rentabilitätsoptimierung,
Frankfurt a.M.: Haag & Herchen Verlag.
Cantner, U. (1996), ‘Heterogenität und technologische Spillovers – Grundelemente
einer ökonomischen Theorie des technologischen Fortschritts’, Habilitations-
schrift, Universität Augsburg.
Cantner, U. (2000), ‘Heterogenität, technologischer fortschritt und spillovers’, in
M. Lehmann-Waffenschmidt, Studien zur Evolutorischen Ökonomik V, Berlin:
Duncker & Humblot, pp. 15–40.
Cantner, U. and H. Hanusch (1996), ‘Intrasektorale technologische Dynamik’, in
E. Helmstädter, G. Poser and H.J. Ramser (eds.), Beiträge zur angewandten
Wirtschaftsforschung, Festschrift K.H. Oppenländer, Berlin: Duncker &
Humblot, pp. 187–216.
Cantner, U. and H. Hanusch (1999), ‘Productivity development – a macro-meso
approach’, mimeo, University of Augsburg.
Cantner, U., H. Hanusch and A. Pyka (1998), ‘Pushing technological progress
forward: a comparison of firm strategies’, in J. Lesourne and A. Orlean (eds),
Advances in Self-organization and Evolutionary Economics, London, Paris,
Geneva: Economica, pp. 114–45.
Cantner U., H. Hanusch and G. Westermann (1997), ‘Technological performance
and variety – the case of the German electronics industry’, in K. Nielsen and
B. Johnson (eds), Institutions and Change – New Perspectives on Markets, Firms

and Technology, Cheltenham, UK and Lyme, USA: Edward Elgar, pp. 116–40.
Cantner, U. and J. Krüger (1999), ‘An empirical account of technological progress
on the macrolevel – DEA versus Growth Accounting’, mimeo, University of
Augsburg.
Cantner, U. and A. Pyka (1998a), ‘Absorbing technological spillovers – simulations
in an evolutionary framework’, Industrial and Corporate Change, 7 (2), 369–97.
Cantner, U. and A. Pyka (1998b), ‘Technological evolution – an analysis within the
knowledge-based approach’, Structural Change and Economic Dynamics, 9 (1),
85–108.
Cantner, U. and G. Westermann (1998), ‘Localized technological progress and
industrial dynamics – an empirical approach’, Economics of Innovation and New
Technology, 6, 121–45.
Caves, D.W., L.R. Christensen and W.E. Diewert (1982a), ‘Multilateral compari-
sions of output, input and productivity using superlative index numbers’,
Economic Journal, 92, 73–86.
Caves, D.W., L.R. Christensen and W.E. Diewert (1982b), ‘The economic theory of
index numbers of the measurement of input, output and productivity’,
Econometrica, 50, 1393–414.
Chang, K.-P. and Y.-Y. Guh (1991), ‘Linear production functions and the data
envelopment analysis’, European Journal of Operations Research, 52, 215–23.
Charnes, A. and W.W. Cooper (1962), ‘Programming with linear fractional func-
tionals’, Naval Research Logistics Quarterly, 9, 181–6.
Charnes, A., W.W. Cooper and E. Rhodes (1978), ‘Measuring the efficiency of deci-
sion making units’, European Journal of Operational Research, 2, 429–44.
Dosi, G. (1988), ‘Sources, procedures, and microeconomic effects of innovation’,
Journal of Economic Literature, 26, 1120–71.
Färe, R., S. Grosskopf and C.A.K. Lovell (1994), Production Frontiers, Cambridge:
Cambridge University Press.
Krüger, J., U. Cantner and H. Hanusch (2000), ‘Total factor productivity, the East
Asian miracle and the world production frontier’, Weltwirtschaftliches Archiv,
136, 111–36.
Leontief, W.W. (1947), ‘Introduction to a theory of the internal structure of func-
tional relationships’, Econometrica, 15, 361–73.
Malmquist, S. (1953), ‘Index numbers and indifference surfaces’, Trabajos de
Estatística, 4, 209–42.
Moorsteen, R.H. (1961), ‘On measuring productive potential and relative
efficiency’, Quarterly Journal of Economics, 75, 451–67.
Ott, A.E. (1959), ‘Technischer Fortschritt’, in Handwörterbuch der Sozialwissen-
schaften, 10, Stuttgart, Tübingen, Göttingen: Fischer, Mohr, Vandenhoeck &
Ruprecht, pp. 302–16.
Rosenberg, N. (1976), Perspectives on Technology, Cambridge: Cambridge
University Press.
Saviotti, P.P. (1991), ‘The role of variety in economic and technological develop-
ment’, in P.P. Saviotti and J.S. Metcalfe (eds), Evolutionary Theories of Economic
and Technological Change, Chur: Harwood Academic Publishers, pp. 172–208.
Saviotti, P.P. (1996), Technological Evolution, Variety and the Economy,
Cheltenham, UK and Brookfield, USA: Edward Elgar.
Schumpeter, J.A. (1935/1912), Theorie der Wirtschaftlichen Entwicklung, five
volumes, Berlin: Duncker & Humblot.
Solow, R.M. (1957), ‘Technical change and the aggregate production function’,
Review of Economics and Statistics, 39, 312–20.
Stiglitz, J.E. (1987), ‘Learning to learn, localized learning and technological
progress’, in P. Dasgupta and P. Stoneman (eds.), Economic Policy and
Technological Performance, Cambridge: Cambridge University Press, pp. 125–51.
Wolff, E.N. (1992), ‘Productivity growth and capital intensity on the sectoral and
industry level: specialization among OECD countries, 1970–1988’, working
paper, New York University.
APPENDIX
A1 The Enhanced Linear Program of Non-parametric Frontier Analysis
The version of the envelopment form including possible excess inputs and
output slacks reads as
min l eTsl eTsl

s.t.:
Yl sl Yl
lXl Xl sl 0 (A.1)
l, sl , sl 0
A proportional reduction of inputs (as given by l) does not necessarily lead
to efficiency in the Pareto-Koopmanns sense. To correct for this the remain-
ing excess inputs (s ) and output slacks (s) are taken into account in the
objective function. Vector eT contains only elements 1. (Of course, one
should here distinguish two vectors eT for inputs and output respectively
which contain s and i elements respectively. To ease notation we do not take
account of this. The further analysis is not affected.) is a positive non-
Archimedean small number. Thus, additionally to program (A.1) takes
into account the remaining output slacks or excess inputs. Only then is a
clear-cut selection of efficient and inefficient observations possible.
A2 Linear Programs Required for Computing the Malmquist Index
In computing the Malmquist-productivity index, for each observation l and

for each periodical change four different linear programs have to be solved.
In the case of t,t and t 1,t 1 the programs are just the ones given by (1.3)
and we will always get results obeying 1. In the case of t,t 1 the obser-
vation in period t will be compared to the frontier function of period t 1;
and in the case of t 1,t the observation in t 1 will be compared with the
frontier in t. In both cases the efficiency values are not necessarily con-
strained to the interval [0,1] but they may be larger than 1. In this case tech-
nical progress would be detected.
For these four computations different linear programs are required. They
are given as follows with t as the period under consideration and s as the
period of the reference frontier:
min lt,s
s.t.:
Ysl Ylt
lXtl Xsl 0 (A.2)
l 0
With these programs T-1 index numbers can be computed for all observa-
tions, with T being the length of the period under investigation.
2. Is the notion of progress compatible
with an evolutionary view of the
economy?
C. Christian von Weizsäcker
1. THE NATURALISTIC FALLACY
A hundred years ago G.E. Moore published Principia Ethica, dedicated to

his teachers and friends at the University of Cambridge.1 John Maynard
Keynes was one of his closest friends2 in the group who called themselves
‘Apostles’. Later on Moore as professor of philosophy at the University of
Cambridge was the successor of John Neville Keynes, the father of the
economist.
In his Principia Ethica Moore criticises Herbert Spencer who had devel-
oped an evolutionary ethics.3 For Spencer evolution has a direction: from
the lower to the higher. Spencer derives from this directedness of evolution
that acting in accordance with evolution is ethically valuable action. Moore
shows that the construction of Spencer is a tautology. The measure of the
good with Spencer is identical with the results of evolution. Moore talks of
the naturalistic fallacy: by putting certain phenomena of nature (for
example that which is successful in the survival of the fittest) as identical
with ‘the good’ one believes oneself able to derive ethical measures from
nature; but one obtains those ethical values from nature only because one
has put them into nature in the first place.
In the social Darwinism of Spencer social progress is identical with the
results of evolution. If we understand this to be a ‘naturalistic fallacy’ then
the concept of progress becomes much more problematical.
2. NORMATIVE INDIVIDUALISM
Neoclassical orthodoxy has developed a welfare economics which can be

used to evaluate policy measures. Economics thereby has the theoretical
instrument to derive a judgement like the following. The money equivalent
of the advantages of this measure with some people is higher (lower) than
43
the money equivalent of the disadvantages of this measure with other

people. Hence this measure should (should not) be implemented. I do not
want to discuss in this chapter the debate within neoclassical economics
concerning compensation principles which, of course, are behind the state-
ment just written down. I have done this elsewhere4. The important point
is the logical coherence between positive neoclassical theory and neoclassi-
cal welfare economics. This coherence is the main reason that economics in
the form of neoclassical economics has become the ‘Queen of the Social
Sciences’, that economics has become the valued consultant of politics.
If evolutionary economics wants to compete in earnest with neoclassical
economics then it needs a comparable coherence between its positive state-
ments and its normative statements. What evolutionary economics needs is
an evolutionary welfare economics. Without this the law makers will simply
ignore evolutionary economics.
In traditional economics we have an attitude which has been called ‘nor-
mative individualism’.5 Value judgements in society should if possible be
derived from value judgements of individuals who are part of this society.
There are different variants of this normative individualism which I do not
want to discuss here. They result from different approaches to the difficulty
of the transition from the level of the individual to the level of society, to
the level of the volonté générale, to speak with Rousseau. We have one line
of thinking in the tradition of Immanuel Kant. I believe that Rawls6 as well
as perhaps Sen7 and Buchanan8 can be associated with this. And then there
is a more utilitarian line of thinking which is associated with mainstream
economics: authors like Bergson, Samuelson,9 Arrow,10 or Mirrlees.11 Then
there is the fundamental scepticism against the concept of justice as it has
been proposed by Hayek12 who of course has been one of the pioneers of
modern evolutionary thinking in the social sciences.
The traditional question has been: how do we derive political decisions
from individual preferences? In solving this difficult problem authors have
used a methodological simplification: the assumption that individual pre-
ferences are exogenous. They assume that there is no feedback from the
state of society upon the preferences of individuals. Accepting this simpli-
fying assumption and accepting the general norm in social philosophy of
the inherent rationality of the desired total process almost by necessity
implies the assumption of rational behaviour of individuals. For no clev-
erly designed social decision process can derive rational social decisions out
of irrational individual preferences. Thus we are led to the standard
assumption of fixed and transitive preferences, i.e. of homo oeconomicus.
In summary: The program of normative individualism and the assump-
tion of fixed preferences compel us to make the assumption of homo
oeconomicus.
Notion of progress 45
3. THE ASSUMPTION OF FIXED PREFERENCES IS

ANTI-EVOLUTIONARY
In a sense it is obvious that the assumption of fixed preferences does not fit
in an evolutionary theory. But I believe it pays to pursue this topic a little
bit further. Obviously biological evolutionary theory in the tradition of
Darwin takes certain things as given. These are the basic natural laws, for
example those of physics. We can wait as long as we want: nature will never
invent the perpetuum mobile because it contradicts basic laws in physics.
But to assume fixed preferences of individuals goes much beyond the
assumption of given physical or chemical natural laws. If we follow Darwin
then there is a great commonality between the nature of humans and the
nature of animals. Behaviour of animals follows certain instincts of those
animals which have turned out in the evolutionary process to be fit for the
survival of that particular species or the genes of that particular species.
Human behaviour in prehistoric times (the ‘Stone Age’) must have followed
the same Darwinian laws. The preferences which have been inherited from
this time have been fit for the survival of the species homo sapiens in these
prehistoric times.13
But one of the conditions for survival of the human species was a certain
flexibility of the human individual. Characteristics specific to the human
species are their ability to plan forward, to use instruments, to produce
instruments, the ability to cooperate on a higher level using developed lan-
guages. These characteristics have been the process of a co-evolution
together with a rising life expectancy, a longer period of infancy which is
characterised by intense learning. From the view of the species as compared
to other species the species has invested much more in individuals. The sur-
vival of the individual has a higher value for the human species than for most
other species because of the substantial investment in this individual. The
mental abilities of the human species enabled it to adapt to quite different
environmental conditions. They meant a particular flexibility which implied
that homo sapiens was able to inhabit a wide variety of climatic zones. All
these factors imply that the share of the total adaptability to changed envir-
onmental conditions which was borne by the individual rather than by the
genes would have to increase. The premium for a higher ability of the indi-
viduals to survive after changing circumstances must have increased.
Under these specific conditions of homo sapiens in prehistoric times it is
highly implausible that preferences, for example preferences for food, have
been fixed once and for all. Much more plausible is the hypothesis that
those individuals had a higher survival value whose preferences were able
to adapt to the specific circumstances, i.e. individuals whose instincts were
sufficiently flexible to adapt behaviour to changed circumstances even
beyond the rational capabilities of the individuals. The hypothesis of adap-

tive preferences is therefore much more plausible in terms of evolutionary
biology than the hypothesis of fixed preferences.
I want to repeat the argument in different words to make it clear. We
make a distinction between an adaptability to the environment by means of
Darwinian evolution of the genes and an adaptability by means of indi-
vidual behaviour. Plants and animals also have an individual adaptability
for given sets of genes. The genes determine the degree to which individu-
als are flexible. Individual flexibility is a characteristic of the phenotype
which in itself is determined by the genotype. That has an evolutionary
advantage. Other things being equal a higher individual adaptability helps
the survival of the genes of this particular species. But other things are not
equal, for a higher adaptability or flexibility pays a price in terms of advan-
tages of specialisation. Therefore in Darwinian evolution there will always
be a compromise between individual adaptability and the advantages of
specialisation of the species. We can expect that the optimal degree of indi-
vidual adaptability is positively correlated with the life expectancy of the
individuals and with the breadth of the ecological niche which is occupied
by that particular species.
Within individual adaptability of humans there is a trade-off between
cognitive abilities and will power on the one side and flexibility of prefer-
ences on the other side. To obtain a given degree of adaptability an indi-
vidual with fixed preferences needs a higher intelligence and will power
than an individual with adaptive preferences. It is highly implausible that
within the range measuring the degree of adaptability of preferences an
adaptability degree of zero (i.e. fixed preferences) has an evolutionary
advantage over a positive degree of adaptability.
4. ADAPTIVE PREFERENCES: AN ALTERNATIVE

TO FIXED PREFERENCES WHILE
MAINTAINING THE APPROACH OF
NORMATIVE INDIVIDUALISM
I introduce a theory of endogenously changing preferences. It is the theory

of adaptive preferences: preferences of economic agents adapt to their
environment. To make precise what is meant we need some formal theory.
We look at stationary states A, B, C. . . . . There is a preference ordering
between these states. But these preferences depend on the state the person
is in. By
(A; B) ( ) (C; B)
we mean: given that the actual state of the past has been B the agent
prefers A to C. The preference inequality makes sense, because A and C are
compared, given the preferences induced by state B. An inequality like
(A; B) ( ) (C; D) does not make sense, because in evaluating A preferences
are induced by B whereas in evaluating C preferences are induced by D.
Now the axiom of adaptive preferences can be formulated in the follow-
ing way: Let (B; A) ( ) (A; A). Then (B; B) ( ) (A; B).
What is ruled out by this axiom is: (B; A) ( ) (A; A) and at the same time
(A; B) ( ) (B; B) the latter would mean: if the agent is adapted to A he/she
prefers B and if the agent is adapted to B he/she prefers A. In that case the
agent is structurally dissatisfied. It is always better to be where you are not.
Note that you can move the agent between A and B and at each move he/she
feels improved.
Now we observe that in the inequality defining adaptive preferences there
is an asymmetry between A and B. B is preferred to A no matter whether
the past was A or B. We then tend to view a move from A to B to be progress
for the agent. In terms of progress not every pair of choices may be com-
parable. We may have (A; A) ( ) (B; A) and (B; B) ( ) (A; B). Thus it is
not automatically ensured that we can define progress consistently.
Here we now develop a theory. In the next two sections I outline a formal
theory. For adaptive preferences under certain assumptions we can find a
concept of progress which is universally applicable as we are used to it in
a world of fixed preferences. The performance of the economy remains
measurable even though the measuring rod changes with the state of the
economy.
5. THE MODEL14
There is one person who is a consumer of commodity baskets x in the pos-
itive orthant of the n-dimensional Euclidean space RN. We work in a con-
tinuous time model. At each moment t the consumer maximises an
instantaneous utility subject to a budget constraint. But the utility function
U(x) itself depends on past experience as expressed by an n-dimensional
vector q which stands for past consumption. We formalise this dependence
on past consumption in the following way
dq .
q (x q) (2.1)
dt
where 0 is a real number. Thus, if consumption x remains constant over

time, then q converges towards x.
The utility function then can be written
UU (x; q)
Utilities with different values of q are not comparable; they simply repre-
sent different preferences.
We assume that U is continuously differentiable with respect to x and q.
We now consider consumption paths through time. Among paths x(t)
which are piecewise continuous we define a subset which we call ‘improv-
ing paths’. This description is motivated by an axiom, which allows a
minimum of comparability between different preferences.
Consider a path of vector x through time such that x(0) x0 is the start-
ing point, x(T)xT is the end point and there is a set J with a finite number
of moments of time, called jump points, t1, t2, . . . . . tN such that 0t1
t2 . . . . . . . tN T. Let IT be the interval [0, T] of real numbers. (We
admit the possibility that T 0). Then we consider paths x(t) such that x(t)
is continuous in IT J. We, of course, admit the possibility that the set J of
jump points is empty.
For such paths q(t) is well defined by the differential equation (2.1), if
q(0) q0 is given. Indeed we find the unique solution

t
q(t) e t q0 x(z)ezdz
0
The integral is well defined for piecewise continuous functions x(z).

.
For any point where the vector x exists we can define the following
expression
n
U .
U xixi
i1
Note that this is different from the time derivative of U which is

n
U . n
U .
U
i1 xi
xi
i1 qi
qi
Given that the comparison of utility values for different values of q does
not have an economic meaning the expression U is economically meaning-
less. On the other hand the expression U does have an economic meaning.
If U is positive it means that ‘real income’ increases. For example, if x
comes about by maximisation of U with respect to x subject to a budget
constraint then U 0 implies that either the budget rises or the price index
(in terms of the Divisia index) falls, i.e. that real income rises. Therefore it
is plausible to talk about an improvement if U 0.
We use the following notation for piecewise continuous paths x(t). By
{x(t); q0; T} we mean a path of the piecewise continuous consumption
vector x(t) in the time interval [0,T] such that preferences are determined
by q(t) e t [q0 x(z)ezdz] with the initial value q0.
t
0
We then can talk of an improvement path. It is basically a path in time
{x(t); q0; T} such that U 0 along the whole path and there are either
improvement jumps or there is a subinterval in which U 0 is strictly
positive. A path is called weakly improving, if U 0, wherever it is
defined and utility does not decline at jump points. Note that an improve-
ment path is then weakly improving. But there may be paths which are
weakly improving but not improving. One example is a path along which
U 0.
I now introduce the improvement axiom. Since preferences change
through time welfare economics would become impossible unless we had
some way of normatively comparing consumption paths that do not have
the same preferences. We need some kind of ‘metapreferences’. But I want
to restrict metapreferences to a minimum. The metapreferences in our case
are encapsulated in the following axiom.
Improvement Axiom: Let x0 be a consumption vector and let preferences
be adapted to x0, i.e. q0 x0. Let X1 be a path such consumption remains
constant through time: x1(t)x0 and therefore q1(t)q0. Let X2 be a con-
sumption path also starting with preferences q0 x0 and starting with the
same consumption bundle x0. Let x2(t) be such that X2 is an improving
path. Then the person prefers to walk along the improvement path X2 over
walking along the stationary path X1.
The Improvement Axiom is highly plausible: starting from the same
tastes (as represented by q0) adapted to initial consumption and the same
consumption basket x0 the consumer prefers improvement over constant
consumption even if he or she is aware that tastes adapt to the evolving situ-
ation through time.
If we accept the Improvement Axiom we can, as will be shown, maintain
the concept of progress such that it is consistent with welfare economics –
even with endogenously changing preferences.
6. THE MAIN THEOREM

I now report a theorem, which indicates the existence of ‘laws of motion’
of preferences which are consistent with a reasonable concept of ‘progress’
in the tradition of welfare economics.
The following notation will be used: x*( )q x means that x* is preferred

over x with preferences corresponding to past consumption vector q. It is
equivalent to the expression U(x*;q) U(x;q). We will consider price
vectors p, which are, as usual, n-dimensional vectors of nonnegative real
numbers.
I first introduce a set of five assumptions.
Assumption 1 (Existence of a demand function) Preferences are such
that a direct demand function xf(p, q) exists, where it is always
assumed that the budget p x1.
Discussion of Assumption 1: by assuming a unique demand vector for
each q and each price vector p we restrict our analysis to convex preference
structures. We need some such restriction because the process of gradual
improvement which we describe with our improvement path only leads to
local optimisation. We cannot expect global optimisation in the case of
non-convex structures.
Assumption 2 (Non-saturation) If x*i xi for i1, 2, . . . n, then
U(x*; q) U(x; q) for all q.
Discussion of Assumption 2: it is my guess that this assumption can be
weakened. For the time being I find it convenient for my proof. It provides
a uniform ‘direction’ in which to look for improvement.
I now introduce the concept of a constant budget path. It is a path which
is generated by a constant budget constraint so that changes in the con-
sumption basket are only due to changes in preferences which are induced
by changed past consumption.
Assumption 3 (Existence of long run demand) For each budget con-
straint p 0 there exists a unique convergence point x F(p) of a con-
stant budget path; i.e. the convergence point is independent of the initial
value q0.
Discussion of Assumption 3: the independence of ‘asymptotic’ behav-
iour from the initial preferences is unlikely to hold, if the preferences them-
selves exhibit important non-convexities. Thus, in a sense, Assumption 3 is
an extension of Assumption 1. An assumption like Assumption 3 is neces-
sary for a ‘global’ theorem of the type to be shown here. We cannot expect
global optimisation from purely local optimisation procedures as discussed
in this chapter, unless we make an assumption like Assumption 3.
Assumption 4 (Continuity) For each triple of vectors {x*; x; q} such
that x* ( )qx there exist neighbourhoods M(x*), M(x), M(q) such that
z* ( )rz for z* M(x*), z M(x), r M(q).
Discussion of Assumption 4: it is an extension of the continuity of fixed
preferences to changing preferences. The continuity assumption is essential
for my proof. This I know from the fact that I cannot simply carry over my
theorem to a model where commodity quantities are restricted to integer
numbers. The theory of adaptive preferences for non-Euclidean commod-

ity spaces has yet to be developed.
Assumption 5 (Adaptive preferencesnon-circularity of improving
paths) Let {x(t); q0; T} be an improving path and let x(0) q0. Then
x(T)x(0).
Discussion of Assumption 5: this is the assumption of adaptive prefer-
ences. Here it takes a form which is different from the intuitive introduction
in section 4. As will be discussed in section 7 the two formulations of adap-
tive preferences are equivalent.
Definition For any given x0 let A(x0) be the set of vectors x such that there
exists an improving path {x(t); q0; T} with q0 x0; x(0) x0; x(T) x.
In other words: the set A(x0) is the set of vectors, which can be reached
from x0 by an improving path, given that initially the preferences are
‘adapted’ to x0.
Remark Obviously, by Assumption 5, it follows that x0 A(x0).
Main Theorem 1. Part A The ‘long-run demand function’ F(p)satisfies
the strong axiom of revealed preference.
Part B There exists a utility function V(x), which generates the demand
function F(p). Let B(x0) be the set of vectors such that V(x) V(x0). Then
B(x0)A(x0).
Here I do not give a proof of the theorem. The interested reader is
referred to my paper (von Weizsäcker 2004).
The meaning of the Main Theorem is, as has already been said above,
that we have an ordering of the commodity space which makes it possible
to speak in a consistent manner about progress even though migrations
through this space change the preferences which are the measuring rod for
the meaning of progress. Indeed, take any two consumption vectors x and
z in the n-dimensional space. Then there are three possibilities. Either – with
preferences adapted to x – there exists an improvement path from x to z so
that we can consider a (perhaps gradual) movement from x to z to be
‘progress’. Or the reverse is true: with preferences adapted to z there exists
an improvement path from z to x so that we can consider a (perhaps
gradual) movement from z to x to be ‘progress’. Or there is neither an
improvement path from z to x nor from x to z (with initial preferences
adapted to the starting point). But then for any vector with equal compo-
nents 0 we can find an improvement path from x (with preferences
initially adapted to it ) to z, and vice versa, so that in a sense x and z are
equivalent in a similar sense to that in which two points on the same
indifference curve are in the case of fixed preferences.
With adaptive preferences we can avoid the un-evolutionary assumption
of fixed preferences and we can avoid at the same time ‘welfare chaos’.
Welfare economics remains possible.
7. EQUIVALENCE OF THE TWO DEFINITIONS

OF ADAPTIVE PREFERENCES
Thus we have shown that there is a chance to develop a welfare economics

under the assumption of adaptive preferences. In my Thünen Lecture
I have given an example how we can apply such welfare economics.15 If we
want to continue in this direction then we need a deeper understanding of
the hypothesis of adaptive preferences. In section 3 above I have already
tried to make clear that an adaptation of individual preferences to changed
circumstances is a much better hypothesis from the point of view of evolu-
tionary biology than the assumption of fixed preferences.
In the mathematical formulation as developed in sections 5 and 6 the
assumption of adaptive preferences takes a form where it is not immedi-
ately obvious how it relates to evolutionary biology. Therefore we need to
clarify this connection. Intuitively we would say that the adaptability of
preferences in a modern market system implies that the long-run demand
is more price-elastic than short-run demand. If a good becomes more
expensive then demand for it goes down. This reduction in consumption of
the good induces a change in preferences such that this good is demanded
less at given prices. So this will in the long run amplify the primary effect of
the higher price on demand.
A higher price elasticity by means of adaptation of preferences corres-
ponds to a higher elasticity of substitution in traditional theory. But the
elasticity of substitution is nothing else but a measure of the adaptability
of behaviour to changed circumstances. Adaptive preferences therefore are
preferences which adapt in such a way that the substitution effect in the
demand function which is induced by a change in relative prices becomes
larger than it would be without this change in preferences.
The connection with the assumption of non-circularity of progress paths
which was the way in which we formulated adaptive preferences can be
made plausible in the following way. The theorem says that the long-run
demand under the condition of non-circularity of progress paths is com-
patible with a utility function V(x). The indifference hyper-surface corres-
ponding to this utility function V(x) which goes through point x0 is the
lower boundary of the area A(x0) which can be reached from x0 by means
of a progress path. But this area contains the set of all points which accord-
ing to the short-run adapted preferences are preferred to point x0, i.e. the
set of all points for which the following inequality holds
U(x; x0) U(x0; x0).

We may call this area the area C(x0). For it is obvious that any point in C(x0)
can be reached from x0 by means of a progress path. The consumer simply
must jump from x0 to x and if x is contained in C(x0) then obviously this is
a progress path. Hence we have C(x0) is contained in A(x0). But this means
that the curvature of the indifference hyper-surface which is the lower
boundary of A(x0) is less or at most equal to the curvature of the corres-
ponding indifference hyper-surface which is the lower boundary of C(x0).
Therefore the long-run elasticity of substitution is never lower than the
short-run elasticity of substitution. Hence the adaptation of the prefer-
ences to changed relative prices amplifies the demand reaction for given
preferences. Thus we have built a bridge between the assumption of non-
circularity and the adaptiveness of preferences.
Another form in which make clear the connection between non-
circularity and adaptivity is the following thought experiment. Assume
that preferences are the opposite of adaptive; let us call them anti-adaptive.
This implies that the long-run reaction to a change in relative prices is
lower than the short-run reaction. Take the case of a linear demand func-
tion. The price p of a good is raised whereas all other prices remain the
same. The consumer is compensated for the price change p so that his or
her utility level according to the short-run utility function not only remains
the same but is raised slightly. The required increase in income Y hence
is higher by a level than the amount which would have been required
for keeping utility the same. But due to linearity the compensating Y is
equal to p(x1 x0)/2 where x0 is the quantity demanded before the
price was raised and x1 is the quantity demanded after the price was raised.
Thus for Y we have the following equation Y p (x1 x0)/2 . Hence
we have a slight rise in utility for this person. If now preferences change
but prices remain the same then the resulting change in consumption is
an improvement because the consumer has the choice to keep his con-
sumption as before. Let x2 be the quantity of the good to which con-
sumption converges in the long run after the price has been raised by p.
Because of the assumption of anti-adaptive preferences we have x2 x1.
A part of the primary reduction in demand is reversed. If now the
consumer is sufficiently close to this convergence point x2 we make
another price change, namely that the price is lowered by p so it is again
at its original level. Also the income is changed by Y , where Y is
negative. Y is computed in such a way that it is slightly lower in absolute
value (or slightly higher algebraically) than it would be to maintain
the utility level. The compensating income reduction can be computed as
Y p(x2 x3)2, where x3 is the quantity demanded to which the
consumer jumps because of the price reduction. Because in the long run
the consumer will converge to the original point x0 and because we have
assumed anti-adaptive preferences we know that x3 x0. Therefore we have

the inequality
Y p(x2 x3)2 p(x1 x0)2 Y 2.
This means that for a sufficiently small 0 we can obtain the inequality
Y Y 0 . Thus on an improving path we come into a budget situation
which is less favourable than the initial situation. The prices are the same
as in the initial situation but the budget is lower. Now we can finish the
improving path by now raising the budget to its original level. Then the con-
sumer has the possibility to return by an improving path to the original con-
sumption vector. Hence we have constructed an improving path which is
circular. This shows that anti-adaptive preferences do not have the property
of non-circularity of improving paths.
8. ADAPTIVE PARETO-OPTIMALITY
Is there a theorem with adaptive preferences which corresponds to the

efficiency theorem of Walras equilibria? The traditional efficiency theorem
says: a Walras equilibrium is Pareto-optimal. For reasons of simplicity of
presentation I take here the case of stationary Walras equilibria. There are
m persons in this economy with an infinite life horizon. All these m persons
have adaptive preferences. The economy is in a stationary state of a Walras
equilibrium. This in particular means that q x for each person. The rate
of interest in this equilibrium is positive.
Now I introduce the following definition:
A stationary state of the economy is called adaptive Pareto-optimal if,

beginning with the stationary state, there is no implementable path in the
economy such that for all persons this path is a weakly improving path and
such that there are persons for whom this is an improving path. The tradi-
tional concept of Pareto-optimality I denote by ‘classic Pareto-optimum’.
Now we can show the following theorem:
Theorem 2: The stationary Walras equilibrium is adaptive Pareto-

optimal.
Sketch of the proof: Since we are in a stationary Walras equilibrium our
Main Theorem tells us that every person behaves in such a way as would
maximize its long utility function V(x). Also according to established theory
the stationary equilibrium is classic Pareto-optimal, if we take the long-run
utility function as the fixed preferences of this classic Walras equilibrium.
Hence according to traditional theory and using the equilibrium prices the
allocation in the equilibrium maximises the present value of the sum of the
consumption vectors of the m consumers subject to the constraint that we
only look at implementable allocations. Now, because of our Main Theorem
a weakly improving path and an improving path for every person only
touches consumption vectors for which we have V(x)V(x0), where x0 is the
equilibrium consumption vector of the stationary equilibrium. For an
improving path we even have the strict inequality V(x) V(x0). But then in
the classical model, with fixed preferences V(x), the total utility of the
weakly improving path is greater than or equal to the total utility of the sta-
tionary path and the total utility of the improving path is greater than the
total utility of the stationary path. But because the stationary equilibrium is
classic Pareto-optimal this implies that the total paths consisting of these
improving and weakly improving paths cannot be implementable. This
shows that the stationary equilibrium is adaptive Pareto-optimal.
Hence we have built a bridge between traditional concepts and theorems
of general equilibrium theory and welfare economics on the one side and
the concepts and theorems developed with the idea of adaptive preferences
on the other side.
9. PATH DEPENDENCE
Our main theorem provides the possibility to define progress even

with endogenous preferences as long as these preferences are adaptive.
Nevertheless it must be clear that the philosophy of evolution is a different
one from the neoclassical philosophy of optimisation, in particular of global
optimisation. Our theory of adaptive preferences is a theory of local opti-
misation with an adaptation of the optimisation criterion to the changed
environment. But our theory rests on an assumption (Assumption 2) which
implies a convex structure of short-term preferences: for each consumption
vector x0 and each vector q the set of consumption vectors x which are pre-
ferred to x0 is a convex set.
But then there are preferences which do not have this convexity. If I am
indifferent between living in Munich and living in Cologne then it does not
follow in general that I would prefer to live 50 per cent of the time in
Munich and 50 per cent of the time in Cologne over living in one of the two
cities exclusively. Is the theory therefore without practical interest? No. For
there are large subsets of the commodity space so that within each subset
the preferences are indeed convex. Only the theory cannot simply be
applied if we want to compare consumption vectors which are in different
subsets which are separated by a non-convexity.
But this is what one should expect if one speaks about an evolutionary
philosophy. There are such things as evolutionary dead-ends. These are
ecological equilibria which under a more general point of view are some-
what unsatisfactory but which nevertheless are the result of a local optimi-
sation. Small mutations and random changes will not lead out of this
dead-end. We also have to take into account that speaking of a point of
view from a higher level is theoretically not well-founded as long as we don’t
have a criterion for this higher-level point of view.
But the theory developed here may help us along. The theory is able to
distinguish between two kinds of status quo conservatisms or local equi-
libria. We may have a kind of status quo which, as has been described in
section F of my Thünen Lecture,16 can be criticised and can be explained
as the result of a somewhat mistaken institutional arrangement. It is then
possible to get out of this equilibrium by means of a progress path and to
reach a new equilibrium. The new equilibrium from the point of view of
preferences adapted to the old equilibrium is inferior. But it is better in the
sense that it can be reached by a progress path and that after adaptation of
preferences to the new equilibrium it is preferred to the old equilibrium.
Then there is the other kind of equilibrium or local optimum such that two
states A and B cannot be compared because of the non-convexities which
have been mentioned. Due to these non-convexities there may be neither a
progress path from A to B nor a progress path from B to A. In this case we
must simply admit that A and B are incomparable. That French people pre-
dominantly prefer the French way of life and that Germans predominantly
prefer the German way of life does not automatically mean that we can
establish a ranking between these two countries by means of our theory.
This chapter adds to my Thünen Lecture on welfare economics with
adaptive preferences. In particular I have emphasised the evolutionary
character of this theory. I hope to have provided the basis for evolutionary
economics to develop valuation criteria so that policy advice becomes pos-
sible from the point of view of evolutionary economics. In the long run evo-
lutionary economics may thereby be able to challenge the dominant
position of neoclassic economics in policy advice.
NOTES
1. Moore (1903).
2. Keynes (1949).
3. Spencer (1892–93).
4. von Weizsäcker (1998).
5. Vanberg (1986).
6. Rawls (1971).
7. Sen (1970).
8. Buchanan (1974).
9. Samuelson (1947).
10. Arrow (1951).
11. Mirrlees (1971).
12. Hayek (1976).
13. On this topic Witt (2004).
14. This and the following section correspond to parts of a paper of mine: von Weizsäcker
(2004). There the reader can find the proof of the Main Theorem.
REFERENCES
Arrow, K. (1951), Social Choice and Individual Value, New York.

Buchanan, J. (1974), The Limits of Liberty, London.
Hayek, F.A. von (1976), The Mirage of Social Justice, London.
Keynes, J.M. (1949), Two Memoirs, London.
Mirrlees, J.A. (1971), ‘An exploration in the theory of optimal income taxation’,
Review of Economic Studies.
Moore, G.E. (1903), Principia Ethica, London.
Rawls, J. (1971), A Theory of Justice, Cambridge, MA.
Samuelson, P.A. (1947), Foundations of Economic Analysis, Cambridge, MA.
Sen, A.K. (1970), Collective Choice and Social Welfare, San Francisco.
Spencer, H. (1892–93), The Principles of Ethics, two volumes, London.
Vanberg, V. (1986), ‘Individual choice and institutional constraints: the normative
element in classical and contractarian liberalism’, Analyse und Kritik, 8.
von Weizsäcker, C.C. (1998), ‘Das Gerechtigkeitsproblem in der Sozialen
Marktwirtschaft’, Zeitschrift für Wirtschaftspolitik, pp. 257–88.
von Weizsäcker, C.C. (2002), ‘Welfare Economics bei endogenen Präferenzen:
Thünen-Vorlesung 2001’, PWP, Perspektiven der Wirtschaftspolitik, Eine
Zeitschrift des Vereins für Socialpolitik, 3, (4, S), 425–46.
von Weizsäcker, C.C. (2004), ‘Welfare economics with endogeneous preferences:
a beginning’, discussion paper, presented at the European Econometric Society
Meeting in Madrid, August; available upon request via my e-mail address:
ccvwzs@t-online.de.
Witt, U. (2004), ‘Beharrung und Wandel – ist wirtschaftliche Evolution theo-
riefähig’, Erwägen Wissen Ethik (EWE), 15, 33–45.
3. Reconciling evolutionary economics
with liberalism
Gerhard Wegner1
1. INTRODUCTION
In economic theory it is common to assess new theories by their relevance

for economic policy. However, theorists often avoid such an assessment, in
particular if new theories are not fully fledged. When theoretical work is in
progress, the practical impact of new theories can be ambiguous, some-
thing which invites different interpretations as the paradigm undergoes
further elaboration. But even if the practical relevance for policy making
looms ahead, the political consequences might lack attractiveness for
theorists; for the cleavage between economic rationality and political ratio-
nality also applies to new economic paradigms so that economic policy is
unlikely to change its course in light of new theoretical insights. Such
difficulties are commonly taken as an excuse to avoid exploring the polit-
ical impact of new economic paradigms in advance. Nevertheless, eco-
nomic theorising is ultimately oriented to welfare and by nature bears
political consequences, even if the latter are only indirect and theories do
not explicitly identify welfare-promoting policies.
Evolutionary economics is a case in point. So far it has not been linked
to one specific political conception, e.g. in terms of interventionism or,
alternatively, liberalism, although different scholars tend to adopt one of
these interpretations with regard to policy making. But before we enter
into a discussion, let us establish that evolutionary economics, even
though it represents a very broad field of heterodox research, is oriented
to economic policy and thus echoes its stationary rival, that is Walrasian
theory. The investigation of novelty and innovation shifts attention to
important sources of welfare in society. Furthermore, at least in this
respect, evolutionary economics claims superiority to Walrasian theory
since innovation is deemed to be of much higher relevance for economic
welfare in the long term than welfare maximisation under given con-
straints in the short term. For that reason alone one should hope for polit-
ical conclusions deriving from evolutionary economics. Even though only
58
Reconciling evolutionary economics with liberalism 59
tentative answers have been given so far, a systematic exploration is likely

to be within its scope.
If we wish to probe the political consequences of evolutionary econom-
ics, we can proceed along different theoretical avenues: first of all, an
analogy with Walrasian theory suggests itself. In doing so we would hold
on to an overarching goal of economic ends called ‘welfare’ and derive well-
suited political measures to increase or even maximise that goal if markets
bring forth welfare deficiencies. The welfare target would, however, undergo
revision in order to include innovation as the major driving force of welfare
in the long term. The logic behind this concept echoes ‘normative econom-
ics’ in the well-known sense while the conventional welfare concept is
revised by an evolutionary update. Although the main problem of this con-
ception derives from the operationalisation of an evolutionary notion of
welfare, attempts have followed this way of theorising which would main-
tain the ‘deductivism’ of normative economics, that is the deriving of best
courses of economic action in a logical sense from an abstract welfare goal.2
A second, less ambitious theoretical option would reject a single eco-
nomic target for the whole society to which the political sphere would have
to commit itself. Instead, the concept of welfare could be decomposed and
adapted to specific circumstances, which gives politics leeway for defining
goals. In so doing, the autonomy of political decision making in a democ-
racy would be acknowledged in a sense which comes closer to Tinbergen’s
(1952) early conception of a means-ends oriented theory of economic
policy. However, evolutionary economic theory could still ask to what
extent economic evolution may profit from intervention or even requires it.
Furthermore, as we know that political intervention can interfere with eco-
nomic evolution in terms of unintended side-effects, the political agenda
should be carefully checked and broken down if need be, something which
is not necessarily at variance with politicians’ own interests.
In the following I concentrate on this second theoretical option. But let
us mention that the potential of evolutionary economics can also be probed
when it comes to further developing a positive theory of economic policy
(public choice). By transferring basic ideas into the field of public choice,
evolutionary economics may focus on political innovation and its diffusion;
in so doing evolutionary economics could analyse preconditions of a
change of political preferences instead of taking them as given. As a result
evolutionary economics would replace the concept of the political entre-
preneur in a sense which comes closer to its market-process analogy, which
implies that political action includes aspects of experimenting and learning
from failures. Few attempts have been undertaken to further develop an
enriched behavioural concept in politics and thus work on this third option
of evolutionary political economy is scarce (Wohlgemuth, 2003).3
For reasons of space let us concentrate on the prescriptive potential of

evolutionary economics. The reasons for our subscription to the second
approach instead of the more ambitious first one will be given later.
However, let us emphasise that an analysis of the prescriptive aspects of
evolutionary economics by no means implies neglecting the discrepancy of
political and economic rationality. Evolutionary economists recommend-
ing better policies are aware of that built-in imperfection of democracies.
However, theoretical reflections on politics may not be pointless if voters
are the intended addressees of recommendations. Besides policy makers,
the principal in democracy does not attract much theoretical attention
although they have an interest in becoming informed about feasible polit-
ical targets. Otherwise, the principal could pay for ambitious or unfeasible
claims addressed to their agents. For the sake of accessibility, prescriptive
theories of economic policy could easily be read as recommendations to
voters when they shape their political preferences.
2. POLICY NORMS FROM THE PERSPECTIVE

OF EVOLUTIONARY ECONOMICS
Any liberal policy norm is strongly linked to normative methodological

individualism inasmuch as individual welfare makes up the welfare of
society. The standard conception of ‘normative economics’ operationalises
this overarching goal for state activities with a strong version of welfare
maximisation in terms of a social welfare function. In order to avoid the
difficulties of interpersonal utility comparisons, the standard conception
often adopts the criterion of Pareto-superiority. It is usually taken for
granted that this conceptualisation of economic policy reflects liberal
norms concerning politics, although critics such as Rowley and Peacock
(1975) or Cordato (1992) have called this claim into question from an
Austrian perspective. Before we assess the adaptiveness of policy norms to
a liberal conceptualisation of politics, let us briefly restate the meaning of
liberal politics.
According to our understanding of liberal politics,
1. economic policy should concern the welfare of the whole society, which
excludes politics acting systematically on behalf of some individuals or
groups (classes, ethnical or religious groups etc.);
2. the welfare of society concerns the welfare of its members so that indi-
viduals represent the genuine target of state activities;
3. individuals define the criteria of welfare according to their own value
judgments which in normal cases should not be overridden by
government (exceptions e.g. in the realm of social policies require good

reasons);
4. individuals are deemed to be competent to identify as well as realise
welfare gains by entering into co-operation and/or trade relationships
with other persons (to be sure, a liberal concept could modify this pre-
sumption to some extent, namely with respect to specific groups and
their conditions of life).
An implication of this liberal conceptualisation of politics is that individ-

uals need freedom in order to realise mutual gains from trade (freedom of
trade; freedom of movement; right of abode). As liberalism, at least in its
classical version, is a cosmopolitan idea and has no truck with national pri-
orities, let alone nationalism, it gives strong arguments for opening national
borders not only for goods but also for individuals who look for better
places in order to make use of their talents and resources.4 A second impli-
cation pertains to the orientation of governmental activities which have to
serve individual welfare and are justified only if markets as well as other
forms of non-state collective actions fail to increase individual welfare.
The latter qualification updates the well-known reservations of classical
liberalism toward state activities by taking into account a broader concept
of institutions as elaborated by New Institutional Economics (NIE).
Since institutions can also emerge from voluntary agreements in society
(internal institutions as opposed to external institutions), the provision of
institutions by the state represents only one possible alternative (Voigt and
Kiwit, 1998). Even if an increase of individual welfare requires institutions,
safeguards against state activities have to be introduced because mandatory
institutions enforced by the state may interfere with individual freedom
without justification in terms of welfare.
Even though standard welfare theory claims to be in line with such an
understanding of liberal politics, evolutionary market theory shores up lib-
eralism much better and supports doubts about ‘19th century mechanical
liberalism’ (Ladeur 2000). Evolutionary market theory does without a
theoretical observer who defines the entire ends of economic activities as
well as their means in terms of production functions. Productive activities
satisfying human wants as well as technologies result from evolutionary
processes when individuals make use of their freedom. Determining a
transformation curve from the viewpoint of a theoretical observer means
fixing the outcome of a process which by definition is open-ended, or, to put
it in Hayekian terms, to ignore the discovery procedure inherent in markets
(Hayek 1996). For that reason the overall target of productive efficiency
from the viewpoint of normative economics gives the impression that
freedom is at best only functional for achieving that goal but principally
gets second priority. It is no accident that normative economics ignores the

notion of freedom.
Compared to this static conception of welfare which points to the satis-
faction of given wants under given restrictions in the best way, evolutionary
economics stresses the experimental character of identifying welfare. In this
regard evolutionary economics differs in the description of welfare and
attaches importance to innovation. Innovations are not valued for their own
sake from the perspective of evolutionary economics, nor are individual
assessments of innovations overridden by a theoretical observer. The
importance of innovation from a welfare perspective results from the oppor-
tunity given to individuals to reassess their former definition of well-being in
view of new alternatives. Hence, it would be a complete ‘non-individualistic’
misconception to deem innovations as such to be valuable without regard
to their consequences. In reality, however, innovations are the salient phe-
nomenon in raising welfare in economies. I recommend that we hold on
to normative methodological individualism and discard any valuation of
technologies made by some outside ‘expert’. In consequence, innovation-
promoting policies can hardly refer to evolutionary economics without
introducing some external and henceforth questionable valuation of inno-
vations (see below). Nothing can be said about the social value of innovation
ex ante except that it extends the choice set, leaving their non-acceptance as
one alternative which has to be respected by a theoretical observer.
Unlike standard welfare economics, evolutionary economics cannot be
concerned with welfare itself but only with the possibility of increasing
welfare. This re-orientation of welfare economics is closely connected to the
knowledge problem inherent in any evolutionary theory: since novelties and
their welfare attributes defy prediction (otherwise they would not be novel-
ties), no ‘goal function’ (welfare function) for the society can be stated.
However, the extension of opportunities to which innovations contribute is
desirable from any welfare point of view, whether or not individuals substi-
tute in favour of new (now: better) alternatives. In view of more alternatives,
individuals have the chance to check their consumption (and production)
plans and to make themselves better off. As a consequence, the premise of
normative methodological individualism combined with the experimental
nature of identifying welfare strongly corroborates reservations against any
public interference with the extension of opportunities; this, in turn, occurs
when politics selects innovations which are to be supported and furthered.
In so doing, government would supplant the social valuation of innovations
by governmental preselection. From the aforementioned normative posi-
tion there is no justification for such a political privilege. This holds particu-
larly true if innovation policies are accompanied by some crowding-out
(distortion) effects in the sense that the promotion of specified innovations
implicitly impedes innovations in other technological fields which do not

gain public support (any financial support for specified innovations will
make for such distortions). The more politics seeks to drive the innovation
process in a selective way, the more questionable it becomes from a welfare
perspective. This normative conclusion, in turn, derives from the knowledge
problem which defies political solution but must be left to competition, that
is to the occurrence of parallel ‘welfare experiments’ to be judged by the
demand side of the market. As a precautionary clause one ought to add that
policy failure does not represent a logical outcome either, or, put differently,
that governments might increase welfare accidentally (our arguments cast
severe doubts on innovation policy but, as a matter of logic, cannot deduce
welfare-impairing effects of policy in a deterministic way).
As a first result one can conclude that evolutionary economics states the
welfare problem of society in a more appropriate way than that of norma-
tive economics in the ordinary sense: there is no welfare function to be
maximised even if the problem of constructing a social welfare function
based on individual judgments were amenable to solution. Instead, the
process of welfare generation comes to the fore. Liberty is a precondition
because it gives firms or individuals the option to enter into economic
experiments, namely to commercialise novelties and to bear the economic
consequences in terms of profits or losses. If neither a theoretical observer
nor the public nor politics is privileged to predict the outcome of experi-
mental welfare identification, liberty is the only political way to deal with
this problem. While liberty by no means guarantees that welfare will
increase (because such deterministic statements turn out to be an alien
element in any evolutionary conception), it establishes one crucial institu-
tional prerequisite of such an increase. By the same token, there is no other
feasible way to increase the welfare of society as long as society does not
articulate one consistent definition of welfare.
The absence of an objective reference criterion also applies to the supply
side of markets. Even if the demand side prefers more innovations to fewer
due to the extension of opportunity sets, one cannot conclude that a state
of more novelties compared to less should be realised. A possible conse-
quence of this idea would be likewise the public promotion of novelties, e.g.
in term of subsidies. Again, such a conclusion suffers from the knowledge
problem; since a theoretical observer cannot foresee whether novelties will
meet demand, a basis for cost-benefit analysis in order to determine the size
and structure of public subsidies for innovation is lacking. Precisely for that
reason, to recall a well-known argument given by Hayek, the economic com-
petence should be transferred from some public sphere to the market agents
who allocate resources for innovations due to their own profit expectations
by making use of their individual assessment of market opportunities.
Innovation decisions are based on beliefs instead of knowledge.5 Not only

can decision making about innovation be erroneous like scientific know-
ledge; it also entails subjective elements such as world-views or beliefs about
human needs which characterise entrepreneurship. Unlike collective deci-
sion making which requires the exchange of arguments and good reasons,
entrepreneurs can escape this type of rationality and follow their crazy ideas
if they are confident of success. Juxtaposing various entrepreneurial ideas
in a thought-experiment would hardly lead to a consistent picture of eco-
nomic opportunities (which the ‘ideal socialist planner’ (Arrow, 1962)
could exploit if private decision making is not optimal). Rather, this picture
would display inconsistencies and even contradictory entrepreneurial ideas
of how to raise welfare. Nevertheless, this inconsistency does not represent
a problem but is the solution to a problem.
Again, the institutional precondition is the non-interference of both
public and private actors with individual decision making, to use a different
word for liberty. From a welfare point of view, no valid answer can be given
to the question whether and what kinds of innovations are desirable for
society. This answer must be left to a process of experimentation and, in
particular, the constituting institutions which channel the identification of
welfare. Hence, no theoretical observer is privileged to offer a solution.
However, a political consequence can be drawn as to the institutional
setting; those institutions which increase liberty are socially desirable
because they ensure the possibility of innovations. For the innovation
process itself, no benchmark is at hand, once we accept the above-sketched
normative position. (If, e.g., one society is late in adopting a particular
innovation in comparison with others for reasons which have nothing to do
with institutional restrictions, there is no way to disqualify its tardiness or
even to term this outcome a welfare loss. Nor does liberal economic policy
have the right to force innovations on society because this would contradict
the essentials of liberalism).
This welfare conception, of course, deviates from prevalent policy norms
attached to standard welfare economics. Neither the Pareto-criterion nor
the dead-weight-loss analysis can be adopted without major qualification:
1. The process of market evolution does not lend itself to state-oriented

judgments such as the Pareto-criterion. Arguably, the common sense
behind the Pareto-criterion appears to be sensible from an evolution-
ary perspective, too: the statement that the welfare of society has been
increased (and the state of the economy is thus preferable to a former
state) reasonably demands that no individual undergoes a loss of
income when others better their well-being, which includes the possi-
bility that not all individuals participate in the rise of welfare. Although
this normative position makes sense in the medium term, it appears too
moral in the short term, as it were. In the process of market evolution,
individual assets including human capital are permanently revalued,
whereas the creative destructions (Schumpeter) of some assets as a by-
product of innovation is one potential outcome. Market evolution,
however, is fuelled by the incentive of firms (individuals) to search for
new opportunities in order to compensate a loss of income due to a
devaluation of their assets. The outlook of some external compensa-
tion to be paid by the winners would erode this very mechanism so that
the Pareto-criterion may be violated at least in the short term. Applying
this norm in a strict sense would suggest negative judgments of any
market evolution, which does not seem a sensible normative position
from the perspective of evolutionary economics if no other mechanism
of raising the welfare of society is feasible.6 A critical point in particu-
lar is the universality of the Pareto-criterion; once accepted, it is a
general norm and thus applicable to any comparisons of alternative
states concerning welfare positions, no matter whether welfare losses
are a temporary phenomenon in the course of market development. In
the same vein, no attention is given to the question whether welfare
improvements of all individuals need time to materialise. With no such
qualifications the ‘pure’ (in the sense of universal) Pareto-norm must
denounce any market process.
2. Innovators create monopolies by definition; monopoly rents prove to
be the by-product of market evolution and defy interpretation as dead-
weight losses. Again, the transfer of the standard norm ‘welfare maxi-
misation’ to the market process entails inadequacies as long as the
genesis of monopolies escapes attention. Used as a general standard
for market deficiencies, dead-weight losses can only be avoided if the
market process itself comes to a halt, which amounts to an absurd nor-
mative consequence.7 While the evolutionary conception by no means
accepts monopolies in general, it qualifies the analysis of welfare losses
from a more general perspective. Whether welfare losses caused by
monopolies are a case of market failure now depends on assessments
concerning the likelihood of maintaining monopoly profits in view of
potential competitors.
3. THE PITFALLS OF ACTIVE ECONOMIC POLICY
In the following I will make a distinction between active economic policy on

the one hand and adaptive economic policy on the other. The former
may also be termed a production approach of innovation, while the latter
is institutional or ‘ordnungspolitisch’, to refer to a popular term of the

German ‘Social Market Economy’ school. Evolutionary economics has
accumulated many insights into the nature of innovation processes, which
makes active economic policy a tempting alternative. However, the institu-
tional approach is better prepared to cope with uncertainty, while the pro-
duction approach is likely to introduce an additional element of uncertainty
into society.
By active economic policy I mean a policy which aims to further the
innovative process by intervening in the market. Measures for doing so
include subsidisation of innovation in general, the subsidisation of selected
‘key innovations’, the selection of specific paths of innovation, e.g. through
public expenditures or the subsidisation of specific patterns of transactions
which promise to bring forth new technologies, that is networks. Active eco-
nomic policy attempts to overcome delays in innovation or the diffusion of
innovation in order to attain a first-mover advantage for the economy or,
more generally, to exploit ‘idle’ innovative opportunities. Of particular
interest is the contribution to positive externalities which promote specific
innovations in regions. I exclude the promotion of basic knowledge gained
at universities or publicly financed research institutions from active eco-
nomic policy because it does not promote commercial innovations directly.
Generally active economic policy takes the institutions as given: property
rights, regulations or externalities are not changed in order to promote
innovations.
Examples of active economic policy can easily be found in many coun-
tries and are expected to become even more prevalent in future. The
European Union has not only committed itself to this type of policy but
also bestows upon it a quasi-constitutional status. In the Maastricht Treaty
of the EU, Title XVI (Industry) as well as Title XVIII (Research and tech-
nological development) define the promotion of new technologies in the
aforementioned sense as one goal of the European Union. Streit and
Mussler (1995) express their irritation about the fact that the European
Treaty attaches the same importance to the so-called industrial policy
(which is identical to active economic policy in our terms) as it does to rules
which constitute the economic order in the EU. This implies that the latter
can be traded off against industrial policy measures in case of conflict.
Active economic policy might invoke some results of the type described
by evolutionary economics.8 Its various theoretical branches emphasise the
contingencies (accidental nature) of market processes and reject a posi-
tivistic view of markets which attaches to any outcome some explanation
in terms of ex-post rationality. For instance, a multitude of technological
opportunities which have not yet filtered out one innovation path can cause
the retardation or even the overlooking of profitable innovation; private
entrepreneurship by no means guarantees the exploitation of such oppor-

tunities based on innovation. The rise in welfare also depends on the early
identification of innovative paths, which suggests a more active role for
technological policy in order to attain a first-mover advantage for the
domestic economy. Additional arguments would stress investment in
human capital, learning-curve effects or the gains in reputation of firms
when governments promote innovations. Time and again economic history
records the tardiness of entrepreneurs who have failed to perceive the
importance of key innovations.9 Other examples point to the importance
of technological clusters which can hardly be created by single entrepre-
neurs but prove to be of great importance for the continuation of innova-
tive processes. Even liberals have to concede that innovative processes are
accidental by nature, so that the effect on welfare is uncertain.
Although there is no compelling argument that active economic policy
must fail in every case, several arguments militate against the tempting
alternative of active political promotion of innovations:
1. Even if we know about the occurrence of various alternative technol-

ogy paths which make for a window of opportunity, this abstract
knowledge does not map onto concrete recommendations for policy
makers as long as a theoretical observer is not privileged to identify the
most promising innovative path in advance.
2. A wait-and-see strategy by entrepreneurs concerning innovations can
be caused by bewildering and inconsistent market signals which do not
point to one promising technological path. Active economic policy
which promotes the preselection of technological alternatives might
suppress the complexity of market signals and bet on the wrong path.
The reality of market complexity cannot be ignored by policy makers
simply by claiming superiority for political interpretation of market
signals.
3. The administrative system proceeds according to its internal require-
ments (processes) which are difficult to bring into accord with the pace
of market developments; in consequence, identifying the precise
moment for investment in innovative paths when the window of oppor-
tunity is open appears to be a difficult mission for collective decision
making in politics.
4. In the same vein doubts can be raised against the learning capacity of
the political system once policy failure becomes obvious. Unlike entre-
preneurs who act in the private sphere, policy makers have to justify a
change of action publicly. Learning from errors offers chances for the
opposition in parliament and gives reason to question the competence
of government; consequently, policy makers are well-advised to avoid
reactions to changed market signals but hold on to their former course

of action as long as possible. Decision making about innovations often
requires quick decisions as well as revision of decisions, which is not a
proper part of political competence in a democracy; here, policy
makers have a stronger incentive to ensure the legitimacy of policies
than to reassess them according to benchmarks.
5. Clusters as well as industrial networks play an important role in the
(enduring) creation of innovations but are ontologically misunder-
stood if innovation policy attempts to model them. In Nozick’s terms
they represent an invisible-hand phenomenon, which in our context
means that they emerge as a sum of entrepreneurial decisions but do
not derive from one single plan. Their ontological nature is more com-
parable to Schelling’s settlement patterns (ghettos) than to public infra-
structure planned by administrations. In consequence, our knowledge
about the importance of industrial clusters will not lead us very far in
practice because we cannot design what is to emerge spontaneously.
Rather, clusters represent an ex-post category for theoretical descrip-
tion but hardly a blueprint for politics.
6. The prospect of public support for innovations changes the incentive
structure of the economy and favours those agents who have better
access to government; unfortunately, this group is not identical with
the most competent agents in the economic sense. A distortion of com-
petition will result which tends to further the capability of communi-
cating with the political system instead of perceiving profitable
opportunity. The innovation politics of the EU offers some striking
examples of an emerging ‘political economy’, as it were.10
7. The problem of creating new interest groups as an inadvertant side-
effect of innovation policy can hardly be by-passed through expertise.
Expertise itself creates an incentive to continue a former course of
policy, in particular if scientific experts such as economists or engineers
become specialised in this field of government consulting and thus
create a comparative advantage in the attraction of resources (funds).
The chief incentive for scientific advisers and consultants will be ori-
ented to modifications instead of renouncement of innovation policy;
only the former offers the chance of receiving further orders from their
governmental clients, while the latter amounts to a self-damaging
behaviour in the competition for governmental consultive contracts.
(Exceptional examples confirm the rules of that game).
8. Public support of innovation runs counter to the rules of a market
economy and has to be investigated from that fundamental point of
view. Unlike entrepreneurs, policy makers lack liability for erroneous
decisions in the field of innovation policy. This very fact calls active
economic policy into question. In market economies it is difficult to

give reasons for compulsory charges to be allocated as risk capital, that
is for purposes which are attainable by private decision making, too. If,
in addition, failures in innovation policy escape the attention of voters,
something which is likely to occur, policy makers can avoid responsi-
bility for their decisions even in political terms. Again, this bespeaks
doubts concerning the quality of decision making.
Note that these arguments do not rest on a positivist interpretation of

market outcomes; neither reality will become rational nor rationality real,
to allude to Hegel’s well-known phrase. Rather, contingencies in evolution-
ary markets are prevalent and so, too, will be deficiencies in the welfare of
society. However, this view of markets to which evolutionary economics
has contributed does not support active economic policy which rivals
private decision making. Chief among the reasons is the experimental
nature of decision making concerning innovations which cannot be sup-
planted by politics without introducing new risks. Implicitly we would
presume the supremacy of one theoretical observer (to which politics has
access) who is able to identify ‘welfare gaps’ caused by delays in innovation
from his privileged position. Indeed, evolutionary economics denies the
existence of such supremacy as long as one takes the notion of novelty seri-
ously and holds onto a liberal concept of welfare.
4. ADAPTIVE ECONOMIC POLICY

AS AN ALTERNATIVE
This scepticism towards active economic policy, in turn, does not bespeak
the irrelevance of all economic policy with regard to innovations. In the
remainder of the chapter, I outline a concept of economic policy which is
in line with our above-stated reflections.
The idea of adaptive policy desists from supporting innovations directly.
By adaptive policy, I understand policy making which is oriented to the
institutional environment alongside the innovative processes. Such institu-
tions concern property rights, competition rules and regulations which
affect transaction costs germane to innovations. In the course of the market
process institutions can become maladapted and need to be changed.
From a property-rights perspective novelties result from a new combi-
nation of resources. As institutions channel the way in which resources
are used, they also influence the emergence of novelties, depending on the
restrictiveness of institutions. In modern economies these restrictions
usually extend those limits identified by economic liberalism, that is the
non-interference of economic activities with other private spheres (exter-

nalities). In other cases novelties give rise to new cases of externalities and
call for new rules which prevent the well-being of other individuals from
being impaired by novelties. Both cases represent important examples for
adaptive economic policy and sometimes defy easy solution. In the latter
case, for instance, the definition of negative externalities can depend on
scientific knowledge which confirms or refutes the presumption that novel-
ties such as genetically modified organisms (GMO) or mobile transmitter
stations impair the health of human beings. Even if no scientific arguments
support the negative impact of novelties so far, the question remains how
economic policy should handle the preliminary state of scientific know-
ledge if irreversible impairments are at least conceivable. From the above-
sketched liberal point of view, no reasons support the designing of
institutions conducive to novelties without regard to their potential nega-
tive externalities on others.11
If no externalities are involved, the perspective of evolutionary econom-
ics can be linked to well-known liberal arguments pointing to welfare
impediments of tight regulations which serve the interests of special
groups. This criticism holds even though the restrictiveness of regulations
is mitigated by the innovativeness of individuals. By making use of their
creativity, individuals not only bring forth novelties but also gain the ability
to circumvent institutions which restrict their behaviour. Tight regulations,
however, are not recommended for the purpose of eliciting innovations in
the economy. The welfare of society benefits more if market agents shift
their perception directly to opportunities, that means in the absence of such
restrictions which have no grounds in the protection of other individuals
for economically legitimate reasons (consumer safety, foodstuff regula-
tions, health and safety at work, environmental regulation and similar
reasons for regulation). In this regard, deregulation extends liberty and
thus has to be put onto the agenda of economic policy which subscribes to
evolutionary economics.
Of particular interest for economic policy is the selection mechanism for
novelties offered on markets. First of all, evolutionary economics changes
the priorities of political targets and attaches more importance to compe-
tition policy than to industrial policy. If we stick to the position that the
welfare impact of novelties is to be assessed by individuals (firms) on the
demand side, there is no room for public interference with the selection
process in markets. The support of allegedly efficient global players or
mergers and even its toleration through a weakening of merger control are
candidates for critique. The achievement of a dominant market position by
firms (sometimes with public support such as tax rebates) can hardly be
justified on grounds of efficiency when efficiency gets second priority in
relation to innovation. From the viewpoint of evolutionary economics, cost

savings will not compensate for the weakness of competition if, as a result,
innovative entrants are prevented from market entry.
For the same reason large firms which are deemed to be ahead of tech-
nological development should not be exempted from competition law.
Contrary to EU competition law, which holds out the possibility of a priv-
ileged treatment for European global high-tech firms, the distortionary
effects of competition should be emphasised. This revision of perspective
derives from doubts about the so-called Schumpeter-hypothesis according
to which, roughly speaking, innovations depend more on current profits
than on competition; one background assumption of this perspective can
be seen in what one might call the ‘production theory of innovation’, which
stresses the capability for innovation due to profits but downgrades the
necessity to innovate, i.e. the competitive environment of firms. Even if
current firms in oligopolistic markets drive technological progress at
present, adaptive economic policy refuses to bet on one successful indus-
trial structure for future innovations. What matters is the openness of
markets which permanently checks the suitability of industrial structures,
whereas a policy of ‘picking the winners’ orientates itself to the past.12
The denial of ends-oriented criteria for market evolution also applies to
further instruments of economic policy, namely designing or revising of
institutions. North (1990) rightly stresses the potential interference of insti-
tutions with market performance. In particular, maladapted innovations
turn out to increase transaction costs and thus impede reallocation which
is the basic operation of dynamic economies. From that perspective one can
follow Engerer and Voigt (2002) who, in accordance with a well-accepted
view held by New Institutional Economics, take high transaction costs as
an indicator for welfare losses. However, circumspection is required if we
draw a clear-cut conclusion concerning institutional policy. In doing so,
institutions have to be assessed with respect to their impact on transaction
costs, which holds out the possibility of revising institutions accordingly.
Such a conclusion would be premature as long as the causes of high trans-
action costs due to institutions are neglected. As we have seen above, trans-
action costs will rise if institutions restrict opportunities and induce
individuals (firms) to circumvent restrictions in order to realise profitable
opportunities. North (1990) made reference to such institutions (instigated
by vested interests) which have given rise to a long-term slow-down of eco-
nomic performance in history, if not a long-term decline of the country in
question.
However, transaction costs can also temporarily rise as a result of market
evolution which does not encounter such institutional interference. On the
micro level, market evolution necessarily makes for intransparency as
a result of variation (innovation). Screening innovations on the demand

side will increase transaction costs because of informational deficiencies
about novelties; for instance, quality attributes of innovations will be
unknown in advance and may raise both information and control costs of
transactions. If the economy encounters a wave of novelties in a certain
period, the total level of transaction costs is supposed to be higher in com-
parison to a more static economy. But this very situation proves to be an
impediment to the diffusion of innovations which are held up solely
because of a lack of information on the demand side.
On this account it is tempting to revise institutions politically in order to
lower transaction costs. This can be done by regulations dealing with
quality attributes so that innovators have to fulfil certain characteristics
when they intend to commercialise new products. Regulations concerning
consumer protection are candidates for the adaptation of institutions to
new products; as a consequence, consumers reduce transaction costs and
are better prepared to digest the abundance of information about novelties.
Transaction costs, however, do not depend solely on institutions set by
authorities (external institutions). Internal institutions also impinge on
transaction cost and are formed by market agents who voluntarily agree
upon quality requirements in order to promote the emergence of markets.
Not least will intransparency in markets encourage new entrepreneur-
ial opportunities, namely for brokers who specialise in the supply of
information for customers; as a result, novelties intensify the division of
labour and knowledge. These internal capabilities of markets have to be
taken into account before economic policy avails itself of the top-down
approach to institutional design. Notably, the emergence of better-adapted
internal institutions needs time and so will the entrepreneurial discovery of
new opportunities dealing with information about new products or ser-
vices. In this regard we encounter the common characteristics of self-
organisation which lends itself to the identification of encompassing
market failures.
However, such an external critique implicitly presumes an objective
benchmark for competition which is difficult to fix in the process of exper-
imenting with welfare (see above). For that reason adaptive economic policy
in our sense should be reluctant to lower transaction costs by (re)designing
external institutions. Although governments can create transparency in
markets in this way, it is likely to interfere with the welfare-generating
process. Regulations can only lower transaction costs by setting new restric-
tions on economic activities. As long as these restricted activities could meet
consumer demands without imposing external effects on others, no inter-
vention should be undertaken. In such circumstances economic policy
should attach second priority to the lowering of transaction costs, even if
the diffusion of innovations or the emergence of new markets undergoes

impediment. Put differently, economic policy should not pursue innovation
as an end in itself.13
Instead, a more liberal orientation to economic policy is better prepared
to cope with the diversity of markets. Priority should be given to the emer-
gence of internal institutions as well as to the discovery of entrepreneurial
opportunities dealing with information. Again, policy should identify regu-
lations which impair the emergence of service enterprises. For instance, civil
laws which define legal capital requirements for founding a firm may turn
out to be maladapted in view of lower capital requirements in the emerg-
ing service sector. An example of this kind of potential maladaption of civil
law is given by the German economy where entrepreneurs have begun to
found ‘Limiteds’ (limited liability companies) in accordance with British
law after the European Court opened up this possibility. In the past, stricter
German rules have not impaired the founding of firms in the economy
which was dominated by manufacturing; there were good reasons for the
legal requirements being in line with the economic requirements to launch
a new enterprise. At present, however, these rules turn out to be maladapted
and indicate a need of political revision. Although the possibility of adapt-
ing the rules of EU member states without relocating offers loopholes and
thus mitigates the need for political reform, policy should make use of the
latter option.14 Transaction costs expended in the attempt to become
acquainted with foreign rules are avoidable and can be lowered by adapt-
ing domestic rules.15
5. CONCLUDING REMARKS
Our considerations have been presented in a sketchy way in order to give a
flavour of the orientation guided by evolutionary economics. Interestingly,
no particular policy recommendations have been devoted to human capital
from an evolutionary perspective, although this notion plays an important
role in the current debate on economic growth. Two reasons can be given
for this omission. Firstly, the public support of scientific knowledge
deserves no specific emphasis inasmuch as this type of knowledge repre-
sents a public good which has limited chances to be produced by private
activities. This type of knowledge defines an indispensable part of eco-
nomic policy which aims at the promotion of economic evolution.
However, evolutionary economics subscribes to standard theory in this
respect and so far has no additional theoretical perspective. A second
reason is given by the revised perspective on human capital which evolu-
tionary economics takes towards the accumulation of knowledge alongside
the market process. Generally, evolutionary economics will not deny the
importance of knowledge but questions the possibility of endowing market
agents from outside with that specific type of knowledge which drives
market evolution. Entrepreneurial activities are based on beliefs about
market opportunities which prompt a specific compilation of knowledge in
order to pursue their economic ideas. In that respect a theoretical observer
lacks the possibility of identifying relevant knowledge which firms can suc-
cessfully apply. From the evolutionary perspective, knowledge results from
entrepreneurial alertness and thus renders more characteristics of eco-
nomic output than input. What economic policy can influence is the insti-
tutional context in which knowledge is accumulated but hardly knowledge
itself. The institutional context, in turn, pertains to the competitive envir-
onment of market agents.16
Our interpretation of evolutionary economics is inspired by Hayek’s
reflections on market evolution (Hayek 1996). The fundamental idea
emanates from the impossibility of describing the economy from the view-
point of a theoretical observer if welfare generation is conceptionalised as
a process of experimenting with welfare. Notably, this perspective leaves the
presumptions of normative methodological individualism unchanged but
arrives at different political conclusions. Evolutionary economics has good
reason to change this perspective in so far as it cannot neglect the role of
innovation for economic welfare. However, if individual choice decides on
the welfare impact of innovations, evolutionary theories have to desist from
supplanting individual choice by outside valuation.
Note that this interpretation of evolutionary economics with regard to
politics differs from that of other scholars working in this field who prefer
a more active role for the state. However, for reasons stated above I suggest
that one reconcile evolutionary economics with economic liberalism.
Besides, the evolutionary perspective holds out the possibility of taking a
fresh approach to liberalism without repeating its dogmatism. If the ele-
ments of welfare are unknown in advance, the whole process of identify-
ing welfare depends on liberty for market agents on both sides of the
market. For lack of external benchmarks, economic policy cannot avoid
accepting the preliminary results of market evolution but has to open the
process for further experiments with welfare. Liberty plays the most
important role here in contrast to ends-oriented criteria on which debates
on economic policy normally focus: the number of innovations as well as
specific technological paths do not represent ends in themselves, even if
theorists or policy makers deem them to be of highest importance for the
domestic eonomy. From an evolutionary perspective, market outcomes are
always imperfect and this very imperfection offers a chance to drive market
evolution.
NOTES
1. This contribution has profited from discussions which I had in the Evolutionary
Committee of the Verein für Socialpolitik and with my discussant Lambert Koch; I also
thank Hans Nutzinger, Sven Meth, and in particular Mark Peacock for discussions and
helpful comments.
2. See Vosskamp (2002). An early in-depth criticism of the deductive approach with appli-
cation to standard welfare economics can be found in Albert (1975).
3. Other attempts have been made by Meier/Slembeck (1998), Slembeck (1997) and Koch
(1996).
4. Even the notion of ‘liberal nationalism’ (Tamir 1993) would contradict to the cos-
mopolitan conception of liberalism.
5. This statement slightly modifies Hayek who only speaks of knowledge.
6. This insight does not exclude the support of losers in some dramatic cases of income
losses. The reader may think of older persons who are simply impeded by their age from
reacting to the creative destruction of their human capital, although they might be
willing to do so.
7. See Metcalfe (2003).
8. See, for example, Forndahl/Brenner (2003), Lembke (2002) or Pyka/Küppers (2002).
9. In the late 1970s German industrial leaders were criticised for being too late in identify-
ing the market potential of the new generation of consumer electronics. The reader may
also recall that even the star of IT, Bill Gates, was late in perceiving the overwhelming
importance of the World Wide Web in the early 90s.
10. For a critique see Streit/Mussler (1995).
11. The GMO case is difficult to judge from a liberal point of view: if GMO drive out other
plants via offspring which cannot be controlled, individuals lack choice of opportunities
no matter what economic policy decides; if, on the other hand, GMO are prohibited,
those consumers who prefer this innovation lack choice; in the reverse case, individuals
can not buy non-GMO even if they want; then they are forced into a choice and undergo
a loss of well-being compared to the status quo ante.
12 ‘Picking the winners and helping the losers’ has become the motto of industrial policy
referring to the Maastricht Treaty of the EU.
13. For a more detailed investigation of the bottom-up approach to institutional policy, see
Wegner (2003).
14. For that reason economic policy is well-advised to accept the New Systems Competition
in general; a detailed investigation can be found in Wegner (2004); Sinn (2003) presents
a critique from the perspective of equilibrium theory.
15. Another example of adaptive economic policy which investigates the adaptation of
property rights to novelties can be found in Hutter (1989).
16. A by-product of this perspective on knowledge is the modification of Arrow’s (1962)
public-good analogy which has derived a market failure problem too straightforwardly;
for an investigation see Metcalfe (2003).
REFERENCES
Albert, H. (ed.), (1975), Aufklärung und Steuerung, Hamburg: Rowohlt.

Arrow, K.J. (1962), ‘Economic welfare and the allocation of resources for invention’,
in N. Rosenberg (ed.), The Economics of Technical Change, Harmondsworth:
Penguin, 1971, pp. 164–81.
Cordato, R.E. (1992), Welfare Economics and Externalities in an Open Ended
Universe: A Modern Austrian Perspective, Boston: Kluwer Academic Publishers
Group.
Engerer, H. and S. Voigt (2002), ‘Institutionen und Transformation – Mögliche

Politikimplikationen der Neuen Institutionenökonomik’; in K.F. Zimmermann
(ed.), Neue Entwicklungen in der Wirtschaftswissenschaft, Heidelberg: Physica,
pp. 149–215.
Forndahl and Brenner (eds) (2003), Cooperation, Networks and Institutions in
Regional Innovation Systems, Cheltenham, UK and Northampton, MA, USA:
Edward Elgar.
Hayek, F.A. von (1996), Individualism and Economic Order, Chicago: Chicago
University Press.
Hutter, M. (1989), Die Produktion von Recht. Eine selbstreferentielle Theorie der
Wirtschaft, angewandt auf den Fall des Arzneimittelpatentrechts, Tübingen:
Mohr.
Koch, L.T. (1996), Evolutorische Wirtschaftspolitik: Eine elementare Analyse mit
entwicklungspolitischen Beispielen, Tübingen: Mohr.
Ladeur, K.-H. (2000), Negative Freiheitsrechte und gesellschaftliche
Selbstorganisation, Tübingen: Mohr.
Lembke, J. (2002), Competition for Technological Leadership: EU Policy for High
Technology, Cheltenham, UK and Northampton, MA, USA: Edward Elgar.
Meier, A. and Slembeck, T. (1998), Wirtschaftspolitik – ein kognitiv-evolutionärer
Ansatz, Munich: Oldenbourg.
Metcalfe, J.S. (2003), ‘Equilibrium and evolutionary foundations of competition
and technology policy: new perspectives on the division of labour and the inno-
vation process’, in P. Pelikan and G. Wegner (eds), Evolutionary Approaches to
Economic Policy, Cheltenham, UK and Northampton, MA, USA: Edward
Elgar, pp. 162–90.
North, D.C. (1990), Institutions, Institutional Change and Economic Performance,
Cambridge: Cambridge University Press.
Pyka, A. and G. Küppers (eds) (2002), Innovation Networks. Theory and Practice,
Cheltenham, UK and Northampton, MA, USA: Edward Elgar.
Rowley, C.K. and A.T. Peacock (1975), Welfare Economics – A Liberal Restatement,
London: Martin Robertson.
Sinn, H.-W. (2003), The New Systems Competition, Oxford: Basil Blackwell.
Slembeck, T. (1997), ‘The formation of economic policy: a cognitive-evolutionary
approach to policy making’, Constitutional Political Economy, 8, 225–54.
Streit, M.E. and W. Mussler (1995), ‘Wettbewerb der systeme und das binnen-
marktprogramm der Europäischen Union’ in L. Gerken (ed.), Europa zwischen
Ordnungswettbewerb 96 und Harmonisierung, Berlin, Heidelberg, New York:
Springer, pp. 75–107.
Tinbergen, J. (1952), On the Theory of Economic Policy, Amsterdam: North-
Holland.
Voigt, S. and D. Kiwit (1998), ‘The role and evolution of beliefs, habits, moral
norms, and institutions’, in H. Giersch (ed.), Merits and Limits of Markets, Berlin
et al: Springer, pp. 83–108.
Vosskamp, R. (2002), ‘Evolutorische Finanzwissenschaft: Problemstellungen und
Lösungsansätze’, in Lehmann-Waffenschmidt (ed.), Perspektiven des Wandels –
Evolutorische Ökonomik in der Anwendung, Marburg: Metropolis, pp. 427–56.
Wegner, G. (2003), ‘Evolutionary markets and the design of institutional policy’,
in P. Pelikan and G. Wegner (eds), The Evolutionary Analysis of Economic
Policy, Cheltenham, UK and Northampton, MA, USA: Edward Elgar,
pp. 46–66.
Wegner, G. (2004), Nationalstaatliche Institutionen im Wettbewerb, Berlin: de

Gruyter.
Wohlgemuth, M. (2003), ‘Democracy as an evolutionary method’, in P. Pelikan and
G. Wegner (eds), The Evolutionary Analysis of Economic Policy, Cheltenham, UK
and Northampton, MA, USA: Edward Elgar, pp. 96–127.
4. Historical economics and
evolutionary economic policy –
Coasean perspectives*
Matthias Klaes
INTRODUCTION
The work of Ronald H. Coase is typically discussed in the context of the

neoliberal approach of the younger Chicago School, in particular with
reference to certain economic policy implications of the Coase Theorem,
and to Coase’s preference for the market to coordinate economic activity.
While Stephen Medema (1996, 1999) and others have drawn attention in
the meanwhile to the difference between Coase’s work and George Stigler’s
(1966) ‘Coase Theorem’, the relationship between Coase’s institutional
analysis and ‘mainstream law and economics’ (Kerber 2000, pp. 149–56)
has remained less clear. Coase counts as one of the co-founders of this sub-
discipline, which seems to suggest a continuity with neoliberal positions
such as Richard Posner’s. However, this reading would gloss over sig-
nificant differences (Coase 1993; Williamson 1993). One can observe a
similar appropriation of Coase’s work by the orthodox wing of the New
Institutional Economics (Furubotn 1994). Given this threefold doctrinal
canonisation of Coase’s work, his reception in evolutionary economics,
which has been mostly critical, comes as no surprise.
Not all authors share this critical reading. Nicolai Foss (1994), for
example, has begun to speak of two Coasean traditions in order to draw
out Coase’s potential for evolutionary economics. In the context of present
debates on how an economic policy informed by the perspectives offered by
evolutionary economics should look (Slembeck 1997, Wegner 1997,
Hodgson 1999), this poses the question to what extent a re-evaluation of
Coase’s work may offer a useful contribution to these ongoing debates.
As a first aim, the present chapter seeks to rebut the predominant neo-
liberal reading of the Coasean research programme. Coase’s neoliberal
reputation is largely based on narrow readings of ‘The problem of social
cost’ (Coase 1960, ‘SC’ from now on). SC sums up and generalises an
78
Historical economics and evolutionary economic policy 79
argument first made in the context of a proposal to allocate the radio fre-
quency spectrum through an auction mechanism rather than administra-
tive decisions of the US Federal Communications Commission (Coase
1959). In contrast to this deregulative and seemingly anti-interventionist
potential of SC (see Posner 1975), ‘The nature of the firm’ (Coase 1937;
henceforth ‘NF’) – the second article responsible for Coase’s fame – was
written by a self-confessed socialist (Coase 1996, 206; see Campbell and
Klaes 2000), setting up a basic tension underlying Coase’s work which
forms the basis of the reading of SC defended below.
The second and main aim of this chapter is to recover this tension on the
level of economic policy and combine it with the historical perspective
offered in much of Coase’s other – unduly neglected – work in order to
clarify important aspects of evolutionary economic policy. Neither is it
thus the intention of the following discussion to establish Coase as an evo-
lutionary economist, nor to offer a detailed historical exegesis of his numer-
ous writings. Given the widespread tendency to see his work as a neoliberal
institutionalist variation of economic orthodoxy, the present chapter seeks
to open up the debate of its evolutionary potential.
Starting out with a discussion of the concept of institutional direction in
Coase’s work, which is central to its policy dimension, the second section
of this chapter looks at the role of the concept of direction in NF, paying
particular attention to the context in which it was conceived: the contem-
porary debate on socialist economic planning. The third section develops
this analysis further to establish the primacy of institutional direction as
the core proposition of Coase’s economic policy perspective. The fourth
section demonstrates how this primacy of institutional direction is the
direct outgrowth of what will be identified as Coase’s historical economics,
in which developmental thinking assumes a key role. This leads in the fifth
section to a discussion of important parallels to evolutionary economics.
The sixth section, finally, draws together the implications of the preceding
discussion for an evolutionary economic policy.
THE PRINCIPLE OF DIRECTION
NF counts today as a core contribution to the theory of the firm

(Holmström and Tirole 1989). Its main propositions regarding the
transaction-cost rationale of the firm are well-known. The crucial economic
policy dimension of Coase’s early article, by contrast, has attracted much
less attention, if it has been noticed at all. Coase was not only concerned
with providing a consistent definition of the firm concept, he also actively
engaged with the contemporary debate on socialist planning and with
Hayek’s position in that debate in particular (Campbell and Klaes 2005).

Initially Coase seems to confine his analysis in NF to the relationship
between price theory and the economic concept of the firm (NF, 387–9).
This discussion is then however translated into the terminology of the plan-
ning debate (see Hayek 1935): ‘Indeed, it is often considered to be an objec-
tion to economic planning that it merely tries to do what is already done by
the price mechanism’ (NF, 387). Coase counters Hayek’s (1933) criticism of
central planning as follows: ‘Those who object to economic planning on the
grounds that the problem is solved by price movements can be answered by
pointing out that there is planning within our economic system . . . which
is akin to what is normally called economic planning’ (NF, 387–8). Coase is
thus keen to transfer the full force of his criticism of received price theory
to the realm of the planning debate, by suggesting that an entrepreneur runs
his business according to the same principles which a central planner would
apply to the steering of the whole economy. He refers to this second coor-
dination mechanism, which he holds is different in kind from the price
mechanism and should not be confounded with the planning activity of
individuals which underlies both, as ‘direction’ (NF, 399, 404).
NF thus adds support to the socialist position in the planning debate by
underlining a fundamental inescapability of regimes of central planning in
capitalist economies. What is more, it adds a set of conditions under which
planning will be more efficient than economic coordination through prices.
This argument comes however with a twist: Coase’s transaction-cost argu-
ment rests on the premise that ultimately, the optimal mix within the eco-
nomic system between centralised planning and decentralised prices is
determined by the forces of competition.
Therefore, a more detailed look at Coase’s views on competition is nec-
essary. When he worked out the main ideas underlying NF in the early
1930s, his knowledge of economics consisted largely of the principles of
price theory as he had encountered them in Arnold Plant’s seminars of
1931. At the core of this conceptual lens was the emphasis on ‘invisible-
hand’ processes: ‘From him [Plant] I learned that producers maximise
profits, that producers compete, and therefore that prices tend to equal
costs and the composition of output to be that which consumers value most
highly’ (Coase 1991, p. 37). It is this basic perspective which underlies
virtually all of Coase’s subsequent major contributions to economics.
As an example, consider Coase’s analysis of industry structure (NF,
394–8). Having located the rationale of the firm in the existence of mar-
keting costs and conversely, the rationale of market co-ordination in the
existence of intra-firm costs of organising production, Coase initially
argues from a partial-equilibrium perspective. Assuming perfect divisibility
and homogeneity of transactions, the boundary of the firm is determined
by that transaction for which marginal marketing and organising costs are
equal.
It is often overlooked that Coase considered this partial analysis only as
a preliminary step that needs to be broadened into the kind of multimar-
ket equilibrium analysis typical of early neoclassical economics (see Blaug
1990). In addition, he explicitly considers transactional indivisibility and
heterogeneity. In this extended analysis, the size of a firm F is determined
jointly by the marketing costs of the marginal transaction, by the cost
structure of the industry to which the firm belongs, and ultimately, due to
the relevance of multiproduct firms (NF, 402–3), by the organising costs
across the whole economy. F grows to the point where the marginal trans-
action is more efficiently executed either across the market or within other
firms. In the latter case, the market is only superseded completely if F is
able, through merger, to internalise all stages of production previously
undertaken elsewhere. Coase sticks thus to the marginal principle, but
rather than doing so out of formal considerations his thinking is driven by
a concern for realistic analysis. Transactions are not assumed to be perfectly
divisible in this form of discrete marginal analysis.
Further qualifications become necessary once the heterogeneity of trans-
actions is taken into account (NF, 396). In general, heterogeneity rules out
speaking of the level of marketing or organising costs in general. Instead,
the particularities of the relevant marginal transaction under each mode of
co-ordination need to be considered. This amounts to the implicit acknow-
ledgement that industries are populated by heterogeneous firms. What is
more, the marginal transaction that determines the efficient boundary of
firm F is not found ex ante, through optimising calculus on the part of the
entrepreneur. Rather it emerges out of a process of trial and error by which
entrepreneurs across all firms continuously test the limits of profitable
expansion.
For our discussion, the following aspects of Coase’s analysis of entre-
preneurial decision making within an overall context of multimarket
moving equilibrium are of particular relevance. First, Coase refused to
follow the turn towards the concept of Pareto-efficiency that took hold of
the increasingly formalist neoclassical tradition of the early decades of the
20th century. A comparatively understood ‘value of production’, as the
yardstick of economic efficiency that he employed across all levels of analy-
sis, be it the individual firm or the economy as a whole, has kept him firmly
grounded in the Marshallian tradition.
Second, Coase did not follow the neoclassical tradition in its assumption
of a priori omniscient and faultlessly economising individuals. One can find
this commitment unambiguously expressed in the following statement,
where Coase distances himself from certain orthodox currents in the new
institutional economics: ‘[T]here is one respect in which I hold a heretical

view. Most economists make the assumption that man is a rational utility
maximiser. This seems to me both unnecessary and misleading. . . . Let
us . . . start with man as he is’ (Coase 1984, 231). Given Coase’s overall
realist commitment (Mäki 1998), this statement can plausibly be taken to
express a programmatic orientation that informs all of his work, which
largely proceeds on the basis of the multimarket equilibrium analysis dis-
cussed above, combined with the assumption of local ‘common-sense’
rationality of economic agents.
Third, while Coase’s analysis is no doubt guided by an overall com-
mitment to equilibrium thinking, the relevant concept is that of moving
equilibrium (e.g. NF, 405; see Hayek 1928; Schlicht 1997), within the
context of his local rationality concept and the continuous experimentation
on the part of economic agents that comes with it. As a fourth point, finally,
Coase’s extended analysis in the second part of section II of NF, in particu-
lar following the assumption of heterogeneous transactions between verti-
cally integrated multiproduct firms, implicitly presumes a heterogeneous
population of firms within a given industry.
THE PRIMACY OF INSTITUTIONAL DIRECTION
By pointing to the crucial role played by the debate on economic planning

in NF, the previous section laid the ground for placing Coase’s binary juxta-
position of direction and market conceptually into an economic policy
context. In NF, the tension between these alternative modes of economic
coordination is resolved on the basis of Coase’s multimarket equilibrium
analysis and traditional neoclassical price theory. The present section
extends the discussion to SC. It will be argued that SC turns on a more pro-
foundly probing elaboration of the co-ordination problem than NF, by
calling attention to the principles of direction that underlie the operation
of markets themselves.
The basic starting point of SC is a clarification of the commodity
concept, unwittingly taking up a fundamental point most forcefully pressed
by Commons (1934). Coase is adamant in stressing that the entities traded
in markets are not physical goods and services but bundles of property
rights. Importantly, the bundling of rights itself is negotiable and subject
to the market process. From this and the basic implications of price theory,
what has become known as the Coase Theorem follows immediately. In
perfect markets, initial and equilibrium allocation are independent of one
another. Therefore, the initial allocation of property rights has no effect on
their equilibrium allocation. This straightforward yet somewhat startling
result follows from Coase’s commitment to a multimarket perspective that

already informed NF: the market mechanism allocates resources on the
basis of their alternative-use values. They therefore end up in those employ-
ments where their contribution to the total value of production – Coase’s
pre-Paretian efficiency standard – is highest.
Coase’s main interest in SC is directed to an economy subject to positive
transaction costs (see Medema 1999).1 Similarly to the analysis in NF, this
leads him to embrace a form of comparative institutional analysis that pro-
ceeds from the alternatives of directing resources according to some central
plan versus their decentral allocation on the basis of market transactions,
which are now conceptualised as the result of bilateral negotiations (‘Coase
mechanism’, see Schlicht 1996). SC moves further than NF however by
drawing a distinction between direction of resources in firms, and direction
by the state as a super-firm (SC, 17). This super-firm stands apart in that it
is able to escape competition with other firms (though not necessarily with
other states), is not bound to transact solely on the basis of contract law,
and has means of enforcement of agreements that go beyond the internal
control mechanisms of conventional firms. There are thus two alternative
modes of co-ordination to consider now alongside the market, which have
to be evaluated according to their relative transaction-cost properties.
Casting the state as a super-firm exposes the central dichotomy of direc-
tion and price mechanism most clearly in Coase’s thinking. In the context of
economic policy considerations, this opens up the question of a potential
bias in favour of one of these two modes of co-ordination. Should it turn out
that a principled preference for market co-ordination can be identified, one
should not hesitate to regard Coase as a neoliberal exponent of the
(younger) Chicago School. Yet, ‘[f]rom these considerations it follows that
direct governmental regulation will not necessarily give better results than
leaving the problem to be solved by the market or the firm. But equally there
is no reason why, on occasion, such governmental administrative regulation
should not lead to an improvement in economic efficiency’ (SC, 18). As
further argued in the next section, Coase assumes an agnostic position in
respect of the relative merits of direction and market, rejecting the yardstick
of perfect markets that can be found in conventional welfare economics. The
nagging ‘on occasion’ in the above quotation indicates though that Coase’s
agnosticism is of a methodological kind, leaving room in SC for a prima facie
presumption in favour of market co-ordination. Still, the crucial point is that
no attempt is made to defend this disposition on dogmatic grounds, allow-
ing only empirical arguments as admissible for comparative considerations.
There is however a further dimension to Coase’s discussion of alternative
modes of coordination in SC which balances Coase’s prima facie disposition
towards market exchange with a principled defence of direction, reaching
beyond its antithetical position vis-a-vis the price mechanism as expressed

in NF. It is the chief concern of SC to demonstrate the relevance of the initial
distribution of property rights in an economy subject to positive transaction
costs. Due to the differential costs of using the market that result from alter-
native initial distributions of rights, different distributions will lead to
different social products. This means that both legislative processes and
court proceedings play a central economic role through the resulting reallo-
cation or rebundling of property rights.
SC addresses therefore a dimension of economic direction that is no
longer reducible to market principles. Coase regards the role of direction as
more pervasive than the background powers afforded by a libertarian under-
standing of the state as a ‘nightwatchman’, which restrict the role of direc-
tion to the definition and maintenance of the legal framework within which
market exchange operates. On the one hand, centralised direction is opera-
tive via the courts, in particular wherever cost-benefit arguments leave room
for discretion on the part of judges or juries. On the other hand, the state is
not a priori denied the potential to bring about comparatively more efficient
solutions than alternative modes of co-ordination. The relevant question in
each case is not whether but how markets should be directed.2
While those aspects of NF that answer to the debate on economic plan-
ning emphasise the principle of direction but keep it ultimately subordinate
to the optimality principle of competition, SC develops the argument
further in at least two respects. First, as outlined above, market processes
themselves rest on the principle of direction. Second, Coase acknowledges
implicitly that the optimality principle of competition must in the end be
seen as a further aspect of blackboard economics. If the forces of compe-
tition were indeed sufficient to bring about an optimal balance between the
alternative modes of economic co-ordination, SC itself would beg the ques-
tion: it would then be futile to point to the diminishing returns of an
increasingly centrally administered economy. It is precisely because in eco-
nomic policy terms, direction is primary to the market that Coase argues
for the practical relevance of a case-study based but historically oriented
research agenda that allows policy assessment in terms of a comparative
analysis of institutional trajectories. Let us explore this unduly neglected
aspect of Cosean analysis in a bit more detail.
HISTORICAL ECONOMICS: A DEVELOPMENTAL

APPROACH
In a static economy, the role of economic policy is reduced to defining

and implementing the preferred institutional framework and the rules
according to which market transactions and other economic activity

should take place. Following this initial setup, the mandate of economic
policy would consist in the monitoring and enforcement of these rules. It
would be misguided to speak in this context of the primacy of institutional
direction. Coase’s approach to economic policy must however be read in
the context of his numerous case studies on the economic implications of
various communication technologies. These case studies, which notwith-
standing Steven Medema’s (1994) pioneering contribution have only
recently begun to inform the secondary literature on Coase’s legacy (e.g.
Pratten 2001), clearly exhibit a developmental understanding of economic
institutions (e.g. Coase 1939, 1947, 1955, 1961a, 1979). A central recurrent
motive is the continuous development of the institutional structure of pro-
duction in its interaction with technical progress.
Coase explored the development of the British and US telecommu-
nications industry and various forms of technological competition. For
example, he looked at inner-city messenger services in the light of emerg-
ing broadcasting technologies, or at the competition between wire-based
and wireless broadcasting. Each emerging technology prompts existing reg-
ulatory regimes to be defended or revised, which the case studies typically
portray as a struggle between market co-ordination and state regulation
and control. Coase locates the formative factor in each case not in any tech-
nological determinants but shows that a detailed historical analysis of the
relevant influences is necessary for a thorough institutional appreciation of
the situation.3
After his emigration to the US, Coase found himself increasingly con-
tributing to contemporary questions of economic policy, notably regarding
the allocation of broadcasting frequencies in the US, including testimonies
to Congressional hearings (Coase 1959, 14–18, 1961b, 1962). His economic
policy recommendations are primarily based on his case-study research
prior to the publication of SC and do not draw on transaction-cost
arguments at all, being more concerned with the effect of alternative
specifications of property rights. The reciprocal nature of harmful side-
effects must have become clear to him as a result of his long-standing
interest in issues of broadcasting signal interference and its putative rele-
vance for arguments against a marked-based allocation of frequencies.
Nevertheless, his doubts regarding the advantages of central allocation of
frequencies are primarily based on an empirical assessment of the poten-
tial disadvantages such a solution would bring with it.
Coase’s historico-empirical method consists of three parts. His case
studies typically start out with a historical retrospective on the emergence
of a particular communication technology, tracing the economic and insti-
tutional effects of the technology in question. At this stage, Coase draws
predominantly from secondary historical literature. This step is followed by

a more detailed analysis of a controversy centring on the technology. Here,
his attention is focused on the motives of stakeholders and other interest
groups. His sources at this second stage are often hearing transcripts, court
procedings and official reports. The final stage consists of a qualitative
comparative evaluation of the various positions that the previous step
brought to light. It is mostly only here that economic arguments enter his
discussion.
Seen in this light, the historical dimension of Coase’s approach takes
precedence over economic arguments. It is thus no coincidence that his
arguably most successful direct attack on orthodox dogma next to SC
(Coase 1974) proceeded almost exclusively on the basis of historical argu-
ments (see Coase 1974). Given that Coase’s teacher Arnold Plant had ini-
tially read economic history, having been a student of Lilian Knowles, one
of the pioneers of the then emerging discipline (see Coase 1986, p. 177;
Koot 1987, p. 212 fn. 5), one should not underestimate the influence of the
English historical school on the young Coase in this respect.
As we have seen, Coase’s method is both historical and of a comparative
nature. Hence, it will encounter difficulties when assessing institutional
arrangements alternative to the ones that have established themselves. In
other words, comparative historical analysis is bound to exhibit a counter-
factual dimension when it comes to the assessment of existing arrange-
ments, an issue explicitly recognised and addressed by Coase (1964). Coase
applies a largely diachronic analysis of different regimes of economic
co-ordination at the subconstitutional level, followed by an evaluation of
plausible alternatives. Due to his rejection of the idealised points of
reference of conventional welfare analysis, this stage of his analysis is
necessarily speculative. One possibility would be to compare an existing
constellation in one industry with alternative constellations in other indus-
tries. Coase doubts however whether all relevant ceteris paribus conditions
would be fulfilled in the practical application of such an approach. His
proposed alternative consists in gaining a detailed grasp of existing
institutional arrangements in order to arrive at an understanding of the
procedural implications of any changes, such as passing of different admin-
istrative codes or statutes. ‘What is needed is an act of imaginative recon-
struction’ (Coase 1964, 195).
Coase’s empirical method focuses thus on institutional development,
both actual and potential. It seeks to gain an understanding of the effect of
alternative arrangements by studying these alternatives not in different con-
temporary industries but in the past of the same industry, therefore exhibit-
ing clear marks of a historico-hermeneutical understanding of economic
institutions. This should however not be read as a call towards descriptive
economic history at the expense of theoretical insight and explanation.

Coase’s method is not historical in any historiographic sense. His aim is not
to apply standard historiographic methods to unlock the past for its own
sake. Quite to the contrary, historical reflection takes the form of what
might be regarded as a ‘scenario analysis’ to establish alternative regimes of
economic order, which are motivated in their development and serve as
important points of reference for the ensuing comparative institutional
analysis.
COASE AND EVOLUTIONARY ECONOMICS
Coase’s approach to economics can be characterised in terms of his concept

of direction, his process-informed multimarket equilibrium analysis, the
resulting principle of the primacy of institutional direction and his
historico-empirical comparative analytical framework. Before being able to
discuss the implications of this approach for evolutionary economic policy,
it is necessary to address the following point head on. Ronald Coase’s work
is uncontroversial as a classical locus of institutional economic thought.
Due to the appropriation of some of his contributions in orthodox quar-
ters, including an important current in the new institutional economics, and
the resulting predominantly critical reception in evolutionary economics,
one must ask oneself whether Coase may serve as a fruitful source of
insight after all, for a research agenda that is diametrically opposed to the
economic mainstream.
It is not the aim of the present chapter to establish Ronald Coase as a
canonical pillar of evolutionary economics. However, given that the recep-
tion of his work has been ripe with misunderstandings of the true thrust of
his research in orthodox and heterodox fields of economic research alike, it
seems timely to call for a discussion of his contribution as one to evolu-
tionary economics as such (Foss 1994; see also Hodgson 2001, pp. 153,
347), rather than merely regarding him as a proponent of traditions that
are related to evolutionary economics but ultimately inextricably bound to
the economic mainstream.
To prepare our discussion of the evolutionary dimension of Coase’s
work, let us first summarise the points of contact that result from the dis-
cussion of the preceding sections. While Coase accepts that economic
agents have a basic competence of rational decision making, he does not
assume perfect knowledge of all contingencies. Instead he suggests we
should proceed from ‘man as he is’.4 This proposal not only implies an
explicit rejection of the concept of utility maximisation but rests on a scep-
tical stance towards the concept of utility in the first place. Coase regards
this concept as merely ornamental, without any explanatory power in its

own right. This assumption of common-sense rationality combines with
Plant’s price theory into a market-process and competitive rivalry-oriented
multimarket analysis, which is ultimately committed to the notion of equi-
librium but acknowledges the crucial role played by continuous experi-
menting on the part of economic agents.
One can see some parallels here to the more elaborate market-process
theories in evolutionary economics, notwithstanding the limits that Coase’s
welfare concept, committed as it remains to early neoclassical price theory,
is bound to encounter in any serious process-oriented evolutionary analy-
sis. The entrepreneurs of NF for example are continuously engaged in
experiments in how to best allocate transactions, but it is the forces of
economy-wide competition that bring about an optimal division of labour
and thus an allocation of transactions across firms that ends up minimis-
ing transaction costs in that the total value of production achieved through
the resulting configuration ensures realisation of the greatest possible value
of production. This result is assumed to emerge in a way reminiscent of
Alchian’s (1950) evolutionary arguments: In the long term, it is not sus-
tainable to buy dearer than one is able to sell (Coase 1998, 577). A similar
argument applies to the boundaries of the firm. Lines of production which
consistently deliver an operating deficit will be closed down, while produc-
tive lines are likely to be expanded.
In the terminology of Ullmann-Margalit (1978), Coase’s approach is
therefore based on an ‘invisible hand explanation of the functional-
evolutionary mould’ (see Vromen 1995, p. 103). Economic agents may well
have the goal to arrive at optimal decisions. However, due to their consti-
tutive incapacity to systematically perform without oversight and miscal-
culation (Coase 1993, 97), the actual determinants of the results of their
behaviour arise from systemic effects.5 This functionalist theory of selec-
tion that forms the basis of Coase’s price theory aside, it is also possible to
detect more or less clearly expressed aspects of population thinking in his
work, most notably in his implicit assumption in NF of a heterogeneous
population of firms making up an industry.
The points of contact in Coase’s work with evolutionary economic
thought mentioned so far should not be accorded too much weight since
they are not uncommon within the Marshallian neoclassical tradition of
the early 20th century which heavily influenced Coase. A more decisive indi-
cation of Coase’s position and his intellectual vicinity to current thinking
in evolutionary economics may be found in his discussion of the promise
of evolutionary theorising for economic purposes (Coase 1978), which,
while admittedly not fully elaborated, still resonates with today’s discussion
(e.g. Witt 1999). In this contribution, Coase underlines the importance of
evolutionary thought, but remains sceptical regarding any direct or analo-

gous application of neo-Darwinist approaches in economics. A theory that
would be capable of explaining both biological and economic processes
risks being formulated at too remote a level of abstraction to be practically
relevant in economics. Moreover, economic systems develop much more
rapidly than biological systems, not leaving enough time for evolutionary
processes to have any significant impact. Finally, economic institutions are,
at least in part, the result of intentional action, which cannot be found on
the level of genes. For all these reasons, Coase regards the most fruitful
application of evolutionary theorising to be in the area of sociobiology, to
the extent that it may help furthering our understanding of learning, rule
following and the formation of expectations.
Nevertheless, selection provides the basis in a metaphorical sense for
Coase’s belief that the forces of competition will lead to a distribution of
resources that realises the highest achievable social product. This tendency
towards optimal outcomes is called into question due to the primacy of
institutional direction in the operation of markets. As soon as Coase allows
for room for the state as a super-firm to withdraw from competition with
other firms, while at the same time according it a role of active direction of
markets, competitive pressures cease being able to decentrally balance plan
and market co-ordination with each other in optimal proportions. The
current extent of central direction becomes to a non-negligible extent the
result of the operation of the principle of direction itself, and is criticised
by Coase on empirical grounds for having proceeded beyond the efficient
scale. It is precisely because he regards direction as primary that he engages
in an historico-empirical agenda of comparative institutional analysis that
seeks to demonstrate that direction is operating at an inefficient scale in the
economy.
In order to come to Coase’s main contribution to evolutionary economic
policy, however, we first have to address the question of how an evolution-
ary approach in economics may be meaningfully differentiated from other
approaches. One can observe general agreement regarding the object level
of evolutionary economics, which has led commentators to speak of an
ontological unité de doctrine (Dopfer 2001, p. 5). This consensus has so far
not led to a comparable level of consensus on conceptual matters. There is
presently no commonly agreed set of core concepts or theories of evolu-
tionary economics (see Klaes 2004a for a positive evaluation of this situ-
ation). Neo-Darwinist impulses from the biological literature have to date
experienced only limited success of a direct or metaphorical transfer of the
Darwinist world-view to the social sciences (see Hodgson 1993, Witt 1999).
Alongside neo-Darwinist approaches, one also needs to consider parts of
the Austrian tradition, ‘old’ and ‘new’ institutional economics, non-linear
systems theory, synergetics, bioeconomics and other related fields, all of

which draw to a significant degree from evolutionary concepts and theories
(Witt 2001, pp. 47–8). The boundary between evolutionary economics and
related approaches that do not label themselves as ‘evolutionary’ seems
thus to be blurred.
Nevertheless, the following basic distinction has crystallised between
economic approaches that draw their inspiration from mechanistic meta-
phors, and evolutionary approaches (Hodgson 1993, pp. 18ff.; Witt 1993).
Conceptually, this distinction can be expressed in terms of the differences
between an integral and non-integral algebraic space (Potts 2000). One
of the defining characteristics of a non-integral space consists in the fact
that processes are non-ergodic. Frequently this gives rise to the catch-
phrase that ‘history matters’. Historical perspectives are therefore of
unquestionable importance within the overall programme of evolutionary
economics.
Coase would not deny the significance of historical specificity. Nor
would he doubt the importance of path-dependent processes for the under-
standing of institutional developments: ‘However fluid an organization
may be in its beginning, it must inevitably adopt certain policies and orga-
nizational forms which condition its thinking and limit the range of its poli-
cies’ (Coase 1966, 442). His contribution to evolutionary economics in
terms of a historical economics that is based on a developmental under-
standing of institutions and on a hermeneutic method of ‘imaginative
reconstruction’ however goes further than a mere acknowledgement of the
relevance of path-dependent processes. Given the present interest in evolu-
tionary economics in appreciating history as an evolutionary phenomenon
in its own right and not just as an epiphenomenon of non-ergodic processes
(e.g. Dopfer 1986, Lehmann-Waffenschmidt 1998), Coase’s work should be
regarded as a key repository of relevant concepts and methods.
In this context, it is useful to draw attention to current debates in
evolutionary biology that centre on recent developments in ‘Evo-Devo’
evolutionary biology (see Goodman and Coughlin 2000). Evo-Devo works
towards a synthesis of neo-Darwinist and developmental approaches.
Embryology and evolutionary morphology play crucial roles in these
debates, notwithstanding their long-standing marginalisation by neo-
Darwinist proponents, who regarded these fields as both committed to
descriptive data gathering and fundamentally atheoretical.
The integration of phylogenetic and ontogenetic perspectives is all the
more plausible and promising in an economic context (Hodgson 1993,
p. 50). The crucial difference between the concept of history in Evo-Devo
biology and evolutionary economics consists in the fact that in the eco-
nomic realm, the concept of history must be regarded not just as a category
employed by the analysing economist but as an epistemological concept of

economic agents themselves (e.g. Fransman 1994, 752ff.). One finds here a
promising point of contact for the adoption and further extension of the
hermeneutical method that underlies Coase’s historical economics into an
evolutionary morphology of institutional forms. While historical specificity
and developmental thinking have become important co-ordinates within
evolutionary economic thought, one can still observe a tendency towards
regarding history first and foremost as a path-dependent process. A more
deeply probing understanding of institutions and their endogenous devel-
opment will have to embrace a richer perspective on the ontological status
of history.
IMPLICATIONS FOR EVOLUTIONARY ECONOMIC

POLICY
Our starting point has been to ask the relevance of the work of Ronald
Coase for evolutionary economic policy. The aim was not to portray Coase
as an evolutionary economist, but to investigate his contribution in the
light of possible points of contact that could inform present debates on
evolutionary economic policy. It is a central tenet of the evolutionary
approach that economic policy must be regarded as endogenous to eco-
nomic processes, and this in a way that goes beyond perspectives offered
by conventional political economy (e.g. Okruch 2001). There is a clear par-
allel here to Coase’s insistence that any change desirable from the per-
spective of economic policy must proceed from changes in relevant statute
law and the rules on the basis of which government bureaucracy operates
and impinges on the economy. Coase’s policy proposals rest on the premise
that the proper mode of implementation of economic policy is not the
blackboard but the legal and administrative rules that govern the oper-
ation of markets and the effect of direction in firms: ‘In the real world, to
influence economic policy, we set up or abolish an agency, amend the law,
change the personnel, and so on: we work through institutions’ (Coase
1984, 230).
Acknowledging the primacy of institutional direction in the operation of
markets entails that every policy measure, even if it is aimed at promoting
competition and the setting up and facilitating of market mechanisms,
relies on active institutional design efforts. Procedural and developmental
thinking promotes an awareness of the difficulties associated with attempts
to arrive at institutions that perform better than the existing ones (Coase
1975, p. 60). Not only is clear and detailed understanding of the extant
institutional structure required, but also insights into ‘how a government
organisation does in fact carry out the tasks entrusted to it’ (Coase 1984,
61; orginal emphasis).
A second point of contact relates to the necessary revision of the tradi-
tional interpretation of the framework-setting task of ordo- or neoliberal
economic policy. According to this view, the state should ensure that appro-
priate legal rules and institutional frameworks would be in place to allow
free-market exchange to operate smoothly. To refer to one of the lasting
metaphors of this strategy of defining an active role of the state that goes
beyond being a ‘nightwatchman’, it would be up to the state to define the
rules of the economic game, while the choice of actual moves in the game
would be up to economic agents. The design of economic rules on the con-
stitutional level, aimed at ensuring the efficient operation of free markets,
would allow the state to largely withdraw from the subconstitutional realm
of economic co-ordination and exchange, an argument which side-steps the
traditional mutual exclusiveness of central plan and decentral exchange
(Pies 2000, p. 349–53).
It was the concern of Coase’s contributions right from the beginning to
break up this polarity on the subconstitutional level, too. His solution does
not however consist of an orthogonal positioning of the two alternative
modes according to their mode of operation as done by the German
ordoliberals. Market and direction, while seen as alternative and mutually
exclusive ways of co-ordinating economic activity, are brought into a
complex interplay with each other, interpreting the setting of the rules of
the game as a move in the game itself. On this basis, framework setting con-
stitutes at the same time active engagement in the allocation of economic
resources, and economic policy remains an endogenous factor to consider.
Finally, Coase’s proposal of a historical approach to comparative institu-
tional analysis calls into question attempts to pursue a static comparative
analysis that abstracts from the operation of ongoing economic processes.
Coase did not turn to qualitative analysis out of incompetence in quantita-
tive methods (e.g. Coase 1935, 1946; Coase and Fowler 1935, 1937). In ret-
rospect, the contributions of SC to economic theory seem to justify his
overall inductive approach. Evolutionary economists do not need to be
reminded that Coase ultimately shied away from the consequential next step:
the development of a theoretical framework to address processes of institu-
tional development themselves. Nevertheless, his historical economics pro-
vides a valuable range of avenues to explore in the development of more
sophisticated historico-morphological approaches in evolutionary econom-
ics. Coase’s suggestion of conceiving of institutional change in terms of acts
of imaginative reconstruction requires further elaboration, but invites us to
try to understand and study economic institutions on a hermeneutical basis
that takes seriously the epistemological status of history for human agents.6
NOTES
* This chapter refines arguments published earlier (Klaes 2004b). The present version has
been streamlined and sharpened up at various points, largely in response to discussions
that have taken place since the German paper has been committed to print. I should like
to acknowledge in particular recent discussions with Brian Loasby, and ongoing intellec-
tual exchange and collaboration over the past decade with David Campbell on the
Coasean tradition in law and economics.
1. For a discussion of the concept of transaction costs as such, see Klaes (2000, 2003).
2. It is worth bearing in mind at this stage that while in principle, one may reject the primacy
of direction in the operation of markets with reference to international competition
between states and economic systems, the properties of competition in this public sphere
are fundamentally different from conventional market competition, to the extent that
some commentators regard analogies between the two as ‘completely inadmissible’
(e.g. Sinn 2004, 30).
3. One should note that some of the historical details of Coase’s research have recently
attracted some criticism (e.g. Brian Simpson 1996a, 1996b; Coase 1996).
4. In this context, one should also mention the relevance of the concept of radical uncer-
tainty in NF and the resulting role of flexibility and adaptation of transactional agree-
ments (see Foss 1996, 78).
5. See Vromen’s (1995, pp. 34–40) discussion of the Becker (1962a, 1962b)–Kirzner (1962)
debate, including his criticism of the evolutionary theories of Alchian and Becker, which
also applies to Coase’s position (at least in the context of global criteria of efficiency).
6. See Herrmann-Pillath (2000) for similar conclusions in a related context.
REFERENCES
Alchian, A.A. (1950), ‘Uncertainty, evolution, and economic theory’, Journal of

Political Economy, 58, 211–21.
Becker, G.S. (1962a), ‘Irrational behavior and economic theory’, Journal of Political
Economy, 70, 1–13.
Becker, G.S. (1962b), ‘A reply to I. Kirzner’, Journal of Political Economy, 70, 82–3.
Blaug, M. (1990), ‘Comment on O’Brien’s “Lionel Robbins and the Austrian
Connection” ’, Carl Menger and his Legacy in Economics (Annual Suppl. to
History of Political Economy, 22), B.J. Caldwell (ed.), Durham, NC: Duke
University Press, pp. 185–8.
Brian Simpson, A. (1996a), ‘Coase v. Pigou reexamined’, Journal of Legal Studies,
25, 53–97.
Brian Simpson, A. (1996b), ‘An addendum’, Journal of Legal Studies, 25, 99–101.
Campbell, D. and M. Klaes (2000), ‘Ronald Coase’s political views at the time
he wrote “The Nature of the Firm” ’, paper presented to EIPE Seminar, 12 April,
Rotterdam, Erasmus Institute for Philosophy and Economics, Erasmus
University Rotterdam.
Campbell, D. and M. Klaes (2005), ‘The principle of institutional direction:
Coase’s regulatory critique of intervention’, Cambridge Journal of Economics, 29,
forthcoming.
Coase, R.H. (1935), ‘The problem of duopoly reconsidered’, Review of Economic
Studies, 2, 137–43.
Coase, R.H. (1937), ‘The nature of the firm’, Economica, 4 N.S., 386–405.
Coase, R.H. (1939), ‘Rowland Hill and the Penny Post’, Economica, 6 N.S., 423–35.
Coase, R.H. (1946), ‘Monopoly pricing with interrelated costs and demand’,
Economica, 13 N.S., 278–94.
Coase, R.H. (1947), ‘The origin of the monopoly of broadcasting in Great Britain’,
Economica, 14, 189–210.
Coase, R.H. (1955), ‘The postal monopoly in Great Britain: an historical survey’,
in J.K. Eastham (ed.), Economic Essays in Commemoration of the Dundee School
of Economics 1931–1955, Coupar: Angus, Culross.
Coase, R.H. (1959), ‘The federal communications commission’, Journal of Law and
Economics, 2, 1–40.
Coase, R.H. (1960), ‘The problem of social cost’, Journal of Law and Economics, 3,
1–44.
Coase, R.H. (1961a), ‘The British Post Office and the messenger companies’,
Journal of Law and Economics, 4, 12–65.
Coase, R.H. (1961b), ‘Why not use the pricing system in the broadcasting industry’,
The Freeman, 11, 52–7.
Coase, R.H. (1962), ‘The Interdepartment Radio Advisory Committee’, Journal of
Law and Economics, 5, 17–47.
Coase, R.H. (1964), ‘Discussion’, American Economic Review, Papers and
Proceedings, 54, 194–7.
Coase, R.H. (1966), ‘The economics of broadcasting and government policy’,
American Economic Review, 56, 440–7.
Coase, R.H. (1974), ‘The lighthouse in economics’, in R.H. Coase (1988), The Firm,
the Market, and the Law, Chicago and London: University of Chicago Press,
pp. 187–213.
Coase, R.H. (1975), ‘Economists and public policy’, in R.H. Coase (1994), Essays
on Economics and Economists, Chicago: University of Chicago Press, pp. 47–63.
Coase, R.H. (1978), ‘Economics and biology: discussion’, American Economic
Review Papers and Proceedings, 68, 244–5.
Coase, R.H. (1979), ‘Payola in radio and television broadcasting’, Journal of Law
and Economics, 22, 269–320.
Coase, R.H (1984), ‘The new institutional economics’, Journal of Institutional and
Theoretical Economics, 140, 229–31.
Coase, R.H (1986), ‘Arnold Plant’, in R.H Coase (ed.) (1994), Essays on Economics
and Economists, Chicago: University of Chicago Press, pp. 176–84.
Coase, R.H (1991), ‘The nature of the firm: origin’, in O. Williamson and S. Winter
(eds), The Nature of the Firm: Origins, Evolution, and Development, Oxford:
Oxford University Press, pp. 3–47.
Coase, R.H (1993), ‘Coase on Posner on Coase’, Journal of Institutional and
Theoretical Economics, 149, 96–8.
Coase, R.H (1996), ‘Law and economics and A.W. Brian Simpson’, Journal of Legal
Studies, 25, 103–19.
Coase, R.H. (1998), ‘Comment on Thomas W. Hazlett. “Assigning property rights
to radio spectrum use: why did FCC license auctions take 67 Years?” ’ Journal of
Coase, R.H and R.F. Fowler (1935), ‘Bacon production and the pig-cycle in Great
Britain’, Economica, 2 N.S., 142–67.
Coase, R.H and R.F. Fowler (1937), ‘The pig-cycle in Great Britain: an explana-
tion’, Economica, 4 N.S., 55–82.
Commons, J.R. (1934), Institutional Economics, reprinted 1961, two vols, Madison,
WI: University of Wisconsin Press.
Dopfer, K. (1986), ‘The histonomic approach to economics: beyond pure theory

and pure experience’, Journal of Economic Issues, 20, 989–1010.
Dopfer, K. (2001), ‘Evolutionary economics: framework for analysis’, in
Evolutionary Economics: Program and Scope, Boston, MA: Kluwer, pp. 1–44.
Foss, N. (1994), ‘The two Coasian traditions’, Review of Political Economy, 6,
37–61.
Foss, N. (1996), ‘The “alternative” theories of Knight and Coase, and the modern
theory of the firm’, Journal of the History of Economic Thought, 18, 76–95.
Fransman, M. (1994), ‘Information, knowledge, vision and theories of the firm’,
Industrial and Corporate Change, 3, 713–57.
Furubotn, E. (1994), ‘Future development of the New Institutional Economics:
extension of the neoclassical model or new construct?’ Lectiones Jenenses, 1,
Jena: Max-Planck-Institut zur Erforschung von Wirtschaftssystemen.
Goodman, C.S. and B.C. Coughlin (2000), ‘The evolution of Evo-Devo biology’,
Proceedings of the National Academy of Sciences of the United States of America,
97, 4424–5.
Hayek, F.A. von (1928), ‘Das intertemporale Gleichgewichtssytem der Preise und
die Bewegungen des “Geldwertes” ’, Weltwirtschaftliches Archiv, 1, 33–76.
Translated in F.A.V. Hayek (1984) Money, Capital and Fluctuations, edited by
R. McCloughry, Chicago: University of Chicago Press, pp. 71–117.
Hayek, F.A. von (1933), ‘The Trend of Economic Thinking’, Economica, 13, 121–37.
Hayek, F.A. von (1935), Collectivist Economic Planning, London: Routledge.
Herrmann-Pillath, C. (2000), ‘Wirtschaftsordnung und Kultur aus evolutorischer
Sicht’, in H. Leipold and I. Pies (eds), Ordnungstheorie und Ordnungspolitik,
Stuttgart: Lucius & Lucius, pp. 371–400.
Hodgson, G.M. (1993), Economics and Evolution, Cambridge: Polity.
Hodgson, G.M. (1999), Economics and Utopia, London: Routledge.
Hodgson, G.M. (2001), How Economics Forgot History, London: Routledge.
Holmström, B.R. and J. Tirole (1989), ‘The theory of the firm’, in R. Schmalensee
and R. D. Willig (eds), Handbook of Industrial Organization, Amsterdam: North-
Holland, vol. 1, pp. 61–133.
Kerber, W. (2000), ‘Law and economics: Ein neuer Ansatz zur Analyse institu-
tioneller Rahmenbedingungen’, in H. Leipold and I. Pies (eds), Ordnungstheorie
und Ordnungspolitik, Stuttgart: Lucius & Lucius, pp. 145–170.
Kirzner, I.M. (1962), ‘Rational action and economic theory’, Journal of Political
Economy, 70, 380–5.
Klaes, M. (2000), ‘The birth of the concept of transaction costs: issues and contro-
versies’, Industrial and Corporate Change, 9, 567–93.
Klaes, M. (2003), ‘Residual categories and the evolution of economic knowledge’,
in H.S. Jensen, L.M. Richter and M.T. Vendelø (eds), The Evolution of Scientific
Knowledge, Cheltenham, UK and Northampton, MA, USA: Edward Elgar,
pp. 37–56.
Klaes, M. (2004a), ‘Evolutionary economics: in defence of “vagueness” ’, Journal
of Economic Methodology, 11 (3), 359–76.
Klaes, M. (2004b), ‘Das wirtschaftspolitische Programm von Ronald H. Coase’ in
Kurt Dopfer (ed), Studien zur Evolutorischen Ökronomik VIII, Berlin: Duncker &
Koot, G.M. (1987), English Historical Economics 1870–1926, Cambridge:
Lehmann-Waffenschmidt, M. (1998), ‘Kontingenz und Strukturähnlichkeit als
Charakteristika selbstorganisierter Prozesse in der Ökonomie’, in F. Schweitzer

and G. Silberberg (eds), Evolution und Selbstorganisation in der Ökonomie, Berlin:
Duncker and Humblot, pp. 187–208.
Mäki, U. (1998), ‘Is Coase a realist?’, Philosophy of the social sciences, 82 (1), 5–31.
Medema, S.G. (1994), Ronald H. Coase, New York: St Martin’s Press.
Medema, S.G. (1996), ‘More than a caricature: the Coase that nobody knows: a
review essay’, Research in the History of Economic Thought and Methodology, 14,
373–83.
Medema, S.G. (1999), ‘Legal fiction: the place of the Coase Theorem in Law and
Economics’, Philosophy and Economics, 15, 209–33.
Okruch, S. (2001), ‘Ziele, Formen und Verfahren einer evolutorischen
Wirtschaftspolitik’, paper presented Buchenbach Workshop zur Evolutorischen
Ökonomik, 23–26, May.
Pies, I. (2000), ‘Institutionenökonomik als Ordnungstheorie: Ein Ansatz für
wissenschaftliche Politikberatung in der Demokratie’, in H. Leipold and I. Pies
(eds), Ordnungstheorie und Ordnungspolitik, Stuttgart: Lucius & Lucius,
pp. 347–70.
Posner, R. (1975), ‘The economic approach to law’, Texas Law Review 53, 758.
Potts, J.D. (2000), The New Evolutionary Microeconomics, Cheltenham, UK and
Northampton, MA, USA: Edward Elgar.
Pratten, S. (2001), ‘Coase on broadcasting, advertising and policy’, Cambridge
Journal of Economics, 25, 617–38.
Schlicht, E. (1996), ‘Exploiting the Coase mechanism: the extortion problem’,
Kyklos, 49, 319–30.
Schlicht, E. (1997), ‘The moving equilibrium theorem again’, Economic Modelling,
14 (2), 271–8.
Sinn, H.-W. (2004), ‘The new systems competition’, Perspektiven der
Wirtschaftspolitik, 5, 23–38.
approach to policy making’, Constitutional Political Economy, 8, 225–54.
Stigler, G.J. (1966), The Theory of Price, New York: Macmillan.
Ullmann-Margalit, E. (1978), ‘Invisible-hand explanations’, Synthese, 39, 263–91.
Vromen, J. (1995), Economic Evolution, London: Routledge.
Wegner, G. (1997), ‘Economic policy from an evolutionary perspective – a new
approach’, Journal of Institutional and Theoretical Economics, 153, 485–509.
Williamson, O. (1993), ‘Transaction cost economics meets Posnerian Law and
Economics’, Journal of Institutional and Theoretical Economics, 149, 99–118.
Witt, U. (1993), ‘Introduction’, in Evolutionary Economics, Aldershot, UK and
Brookfield, US: Edward Elgar, pp. xiii–xxvii.
Witt, U. (1999), ‘Evolutionary economics and evolutionary biology’, in
P. Koslowski (ed.), Sociobiology and Bioeconomics, Berlin: Springer, pp. 279–98.
Witt, U. (2001), ‘Evolutionary economics: an interpretative survey’, in K. Dopfer
(ed.), Evolutionary Economics: Program and Scope, Boston, MA: Kluwer,
pp. 45–88.
PART II
Self-organisation and networks

5. The concept of network
organisation – biotechnology-based
industries as exemplar
Andreas Pyka and P. Paolo Saviotti
1. INTRODUCTION
Biotechnology is one of the fields that is at the forefront of the creation

of a knowledge-based society. This seems somewhat paradoxical, since it
could be maintained that biotechnology is one of the oldest technologies
used by mankind. Wine and cheese making constitute two typical exam-
ples. However, modern biotechnology has been substantially changed by
the advent of molecular biology, a new discipline which was founded in the
1930s based on the attempt to apply to biology the methods of physics. In
the mid 1970s two discoveries, recombinant DNA and monoclonal anti-
bodies, transformed a scientific discipline with a brilliant if distant future
into a seedbed of industrial applications. Accordingly some authors now
call this latest vintage of biotechnology third-generation, to distinguish it
from the completely empirical first generation and from the second gener-
ation, which began with the production of antibiotics. Second-generation
biotechnology used scientific methods but it did not have the knowledge
required to change the genetic make-up of organisms. Such knowledge was
only provided in a systematic way by molecular biology.
Who were the first actors involved in the development of third-generation
modern biotechnology? In the mid 1970s very few research institutions did
research in molecular biology and they were mostly in the USA. The indus-
trial firms that in principle could have exploited molecular biology did not
have the respective absorptive capacity. Their competencies and knowledge
bases were concentrated in more traditional disciplines such as organic
chemistry or microbiology. In fact, this lack of knowledge hampered firms’
recognition of the opportunities that could have been offered by molecular
biology.
With respect to applications and the scope of the research programme, one
has to consider that biotechnology is not an industrial sector but a scientific
99
100 Self-organisation and networks
discipline underlying a number of industrial sectors (pharmaceuticals,

agriculture, food, environment etc.), here called the biotechnology-based
sectors. Very importantly, industrial applications of biotechnology are
highly dependent on new scientific developments, even on those that are the
result of basic research. Although the lead-times between the discovery of
new knowledge and its final embodiment in new products may be very long,
the time between the creation of new knowledge and the funding of indus-
trial research aimed at its applications is in general very short. Thus, basic
research is not exclusively confined to public research institutions, but is also
carried out by firms.
The paradigmatic change in the underlying knowledge-bases, the fast
speed of knowledge creation and diffusion processes as well as the great
number of different actors involved in these processes necessitate new
forms of industrial organisation of research and development. In particu-
lar, collaborative arrangements between the different and heterogeneous
actors involved seemed to be the form of industrial organisation which
emerged in the biotechnology-based industries.
Thus, both in what concerns its intensity of knowledge utilisation and in
the mechanisms employed, biotechnology seems to be a very good example
of industrial organisation in a knowledge-based society. Of course, the con-
clusions reached in this chapter will depend on the specificity of biotech-
nology, but they will also have some general significance for the analysis of
a knowledge-based society.
In what follows we first have a look at the developments in the R&D organ-
isation in the biotechnology-based industries since the 1980s. We analyse
biotechnology innovation networks with special emphasis on two types of
actors, Large Diversified Firms (LDFs) and Dedicated Biotechnology
Firms (DBFs). In a numerical model we then analyse the advantages and dis-
advantages for firms of going-it-alone strategies and of networking strate-
gies, taking into account the environmental factors influencing the
formation of links between actors. Network formation will be shown to
display a dynamics going beyond a first wave of networks and leading to a
re-organisation of the partnerships involved.
2. INNOVATION NETWORKS IN THE

BIOTECHNOLOGY-BASED INDUSTRIES
The earliest analyses of innovation networks pointed to the possibility that

they are only a temporary form of industrial organisation. Such a temporary
character could be the result of discontinuities in knowledge generation, for
example of the emergence of a new technological paradigm. It was argued
The concept of network organisation 101
that LDFs were committed to the old paradigm, in which all their compe-
tencies were concentrated, and that they could not easily internalise the new
knowledge. Alternatively, LDFs did not have the absorptive capacity
required to internalise the new paradigm and they were not capable of con-
structing it rapidly. A new type of industrial actor, small high-technology
firms, arose to bridge the institutional gap between public research institu-
tions and LDFs. In the specific case of biotechnology such firms were called
Dedicated Biotechnology Firms (DBFs). DBFs were expected to act as
intermediaries between LDFs and public research institutions. In the rest of
the chapter the DBFs performing this role will be called translators. In the
course of time by collaborating with DBFs and with public research institu-
tions, LDFs could construct a knowledge-base and an absorptive capacity
in biotechnology. Once this happened the role of DBFs would have become
superfluous and industrial organisation would return to the traditional
dichotomy between the market and hierarchical organisations.
However, the rate of creation of interinstitutional collaborative net-
works steadily increased all throughout the 1980s and 1990s. In Figure. 5.1
we see e.g. the growth of international collaborations in the biotechnol-
ogy-based industries including only those cases of new collaborations
where the partners come from different countries. The continuous attrac-
tiveness of the networking strategy implies the following two alternatives:
either LDFs have not internalised the new paradigm constituted by
biotechnology or a new role for DBFs has emerged in innovation net-
works.
The analysis of this problem constitutes one of the main objectives of the
present chapter. Looking at Figure 5.2 which shows the number of actors
involved in new collaborations in Europe, this clearly suggests the second
200
150
100
50
0
1980 1985 1990 1995
Source: Science and Engineering Indicators 2000
Figure 5.1 Number of new international strategic biotechnology

collaborations
120
Biotechnology Company
100
Diversified Company
80
60
40
20
0
81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
Source: Bioability
Figure 5.2 Types of agents in European biotechnology collaborations
alternative, that a new role for the DBFs has emerged. Since the late 1980s
they have clearly dominated the founding of new alliances, often between
DBFs only. Furthermore, by the end of the 1980s, of course also the LDFs
had acquired a knowledge-base in molecular biology (see for example
Grabowski and Vernon 1994) and yet they continued to enter into collab-
orative agreements with DBFs.
The new role for DBFs we hypothesise in this chapter to be linked to the
extremely rapid rate of creation of new knowledge. Even if LDFs have
acquired an absorption capacity for it, the sheer rate of advance is such that
no LDF could keep up with it all. LDFs might thus use collaborative agree-
ments with DBFs in order to keep abreast of new developments that could
turn out to have important economic applications. The alternative course
of action for an LDF would be to invest in research in the same fields of
biotechnology. However, with a very high rate of growth of knowledge this
strategy would involve a very heavy, irreversible and risky commitment.
The collaboration with DBFs constitutes a more flexible and reversible
strategy. It is to be observed that this role does not involve a qualitative
difference in LDFs’ ability to understand molecular biology, but only the
attempt to reach a better trade-off between readiness for action if promis-
ing developments were to emerge in new subsets of the biotechnology
knowledge-space and the sunk costs that need to be faced in order to keep
these windows open. And, it must be remembered that the competitive
advantage of LDFs is not constituted by their ability to understand new
knowledge but by their capacity to combine the different competencies and
complementary assets required to produce a final product. Furthermore, a
rather important share of the new collaborations is occupied by alliances
between DBFs only, which underlies the changed trend of collaborations.

Whereas in the early 1980s we find a dominance of agreements where
knowledge flows from the DBFs to the LDFs in an exchange of R&D
funds, since the late 1980s bilateral knowledge-flows play the dominant role
in the new collaborations. This second role that can be played by DBFs will
be called that of explorers.
3. THE MODEL
Before we start explaining the basic structure1 of our model some remarks
with respect to the methodological framework will be helpful. In particu-
lar, we are going to use the methodology of the so-called history-friendly
models, originally introduced by Malerba, Nelson, Orsenigo and Winter
(1999). History-friendly models are designed to capture, in a stylised form,
the mechanisms and factors affecting industry evolution, technological
advance and institutional change detected by empirical scholars of indus-
trial economics, technological change, business organisation and strategy,
and other social scientists. Thus history-friendly models can be considered
the natural extension to modelling of qualitative and appreciative theories.
Obviously even in a neo-Schumpeterian approach simulation models
have to introduce a certain degree of abstraction and cannot reflect reality
in all its complexity. The mechanisms built into the formal model have to
be transparent enough, so that the analyst can figure out what the causes
are of the observed effects. Therefore, in the first step of our modelling
effort we have to carefully single out the relevant actors, bring together vari-
ables which are effective in the same direction and combine important
developments and possibilities of action. Nevertheless, adopting the neo-
Schumpeterian approach allows us to put emphasis on crucial features of
innovation processes, such as non-linear dynamics, heterogeneity and true
uncertainty, which are beyond the scope of traditional approaches.
In the following the basic building blocks of our modelling framework
are introduced. In particular, we focus on the way we present the different
agents in our model, the way we capture innovation processes, what we con-
sider to be the prerequisites and consequences of networking as well as the
representation of the economic realm in our simulation.
3.1 The Agents
The agents we are considering in our model are Large Diversified Firms
(LDFs) and Dedicated Biotechnology Firms (DBFs). They are described in
terms of their competencies and capabilities. DBFs possess technological
competencies while LDFs possess a mixture of economic and technologi-

cal competencies.
Competencies
Technological competencies are the components of the knowledge-base
required for building up production and innovation capabilities in a specific
technology. In other words, before firms are able to develop new marketable
outputs they have to develop the respective biotechnological competencies.
Furthermore, technological competencies alone are not sufficient to achieve
economic success with a new product. Economic competencies are neces-
sary in order to successfully produce and market a new commodity.
Examples of these economic competencies are experience in clinical trials,
distribution channels and so on. Obviously this representation is somewhat
simplified. The full range of competencies required by firms to conceive,
develop, produce and market new products is very large and heterogeneous.
However, given that most DBFs at the beginning of their life-cycle do not
possess any economic competencies and that LDFs in the 1970s were gener-
ally unable to acquire the knowledge required to use modern biotechnology,
the representation in terms of technological and economic competencies
adequately describes the difference between our two main groups of agents.
Moreover, we could consider technological competencies as the core com-
petencies (Prahalad and Hamel 1990) of firms and economic competencies
as a large part of the complementary assets (Teece 1986) required to produce
and market a product.
Figure 5.3 shows the development of technological competencies. In the
early phases the building up of the knowledge-base is a difficult process and
progress is hard to achieve. However, after having developed a certain
knowledge-base it becomes easier to learn even more (threshold effect).
Finally, marginal progress becomes progressively more difficult as the
knowledge-frontier existing at a given time is approached. A function of this
type implies variable returns to investment in the creation of new knowledge
within a given field: very low at the beginning, positive and growing in the
intermediate phase before diminishing returns set in as the potential of the
new field has been exploited. The process of building up a knowledge-base
in biotechnology is supported by co-operative arrangements with firms who
are already active in this field – an important part of the respective knowl-
edge base is transferred by networking.
The difference between LDFs and DBFs

The two populations of firms which can be observed in the biotechnology-
based industries can be distinguished on the basis of their relative techno-
logical and economic competencies. The first population of firms is that of
Competencies
Time
Co-operations
Figure 5.3 Building up of a knowledge-base
LDFs, for example the large established pharmaceutical firms. Until the
end of the 70s their research and development was mainly embedded in the
paradigm of traditional organic chemistry. The coming up of the new
biotechnological paradigm meant a ‘competence-destroying technological
progress’ (Tushman and Anderson 1986) for them, as most of their com-
petencies were threatened by the new ones. In our model this group of
agents is represented in the starting distribution with well-developed eco-
nomic competencies but with almost no technological competencies in
biotechnology.
In the second population we find small start-up companies, often uni-
versity spin-offs specialised in the biotechnology field. This group of
agents, the so-called Dedicated Biotechnology Firms or DBFs, by their
very nature have highly developed technological competencies, but almost
no economic competencies. When they start their existence DBFs depend
on external funds for research and development. Accordingly, in our start-
ing distribution they are represented just as having no economic compe-
tencies but highly developed technological competencies.
Venture capital firms and universities

In addition to these agents that we are explicitly taking into account, we
also consider two further important groups of actors in our model: public
research institutes or university labs and venture-capital firms. In order to
keep our model simple their behaviour is not explicitly analysed, but they
are considered as an important component of the external environment of
biotechnology firms. For example, in order to acquire the funds necessary

to undertake R&D a DBF can co-operate either with an LDF or with a
venture-capital firm; similarly, the cooperation of an LDF with a DBF or
with a public research institute leads to almost the same consequences for
the LDF etc.
Capabilities
Drawing on their competencies firms can accumulate technological capa-
bilities in specific fields which allow them to explore the technological
opportunity-space. The firms in our model act in an environment which
continuously forces them to be engaged in such R&D processes. Not to
innovate means to fall behind in the competitive environment of biotech-
nology. In order to increase the probability of an innovation firms accu-
mulate technological capabilities in the course of time.
Together with the technological competencies the technological capabil-
ities determine the probability of an innovation. The relationship is shown
in Figure 5.4.
To consider the intrinsic uncertainty of innovation processes the innova-
tion probability of a firm is matched in every period with a Poisson-
distributed random number whose mean value is asymptotically reached.
A firm is successful in its innovative efforts only if the innovation probabil-
ity is above the random number.2
R&D decision rules

The investment in R&D is no longer guided by an optimisation calculus, but
by a routinised behaviour, as innovation goes hand in hand with true uncer-
tainty (e.g. Nelson and Winter 1982, p. 132). Firms adopt certain rules, for
example, invest x per cent of your turnover in R&D, retain x per cent of your
Innov. prob.
Capabilities
Competencies
Figure 5.4 Innovation probability as a function of competencies and

capabilities
financial support in order to build up your own capital stock etc. In the same
way the distribution amongst different activities (e.g. between investing in
the building up of technological or economic capabilities) is captured by
referring to routines.
3.2 Networking
In order to carry out their innovation processes firms can choose different
strategies. They can either decide to go it alone, which means not to draw
on external knowledge-sources and not to share their own new know-how
with potential competitors, or they can decide to co-operate with other
actors and build up collectively the new capabilities necessary for the intro-
duction of a new commodity. Innovation networks emerge through this
mutual co-operation, which gives rise to channels for knowledge-flows
between the firms participating in the network. In particular, we are con-
sidering the evolution of innovation networks at three levels within the
model: the environmental conditions favouring or inhibiting the growth of
networks, the individual decisions of firms to co-operate or not, and a
matching process bringing together firms willing to co-operate. This
process creates a population of networks with its own dynamics. The for-
mation of any network constitutes an act of birth or entry into the popu-
lation. Conversely, the disappearance of a network constitutes an act of
death or exit. The dynamics of birth and death of networks will be deter-
mined by the specific features of each network and by some features of the
external environment.
Environmental conditions for networking

A number of environmental factors increase the probability of the birth of
innovation networks. The growing complexity of innovation processes as
well as a high degree of technological uncertainty play the most important
role. Every time a firm successfully introduces an innovation the number of
knowledge fields is assumed to grow. Given the complementary and combi-
natorial nature (e.g. Staropoli 1998, 15) of biotechnology, the technological
space, defined as the number of possible combinations of knowledge-fields,
increases in a non-linear way.
Especially in the early phases of a technological life-cycle this increasing
complexity is combined with a high technological uncertainty because
specific research techniques or heuristics – e.g. how to handle this com-
plexity – are not yet developed. In the model the phase of a technological
life-cycle is approximated very roughly by the average age of the different
commodities on the product markets.
Additionally, R&D networks are dependent on a number of core
technologies or core/central actors who play a crucial role in the establish-

ment of the networks (e.g. Saviotti 1998, pp. 36–7). In our model the pop-
ulation of LDFs is supposed to play this role.
In addition to the previous network-supporting effects, other influences
decrease the probability of network formation, thus leading to network
death. First, the degree of competition is crucial in this respect. In our model
we use as an economic framework a heterogeneous oligopoly. We consider
the degree of substitutability of the final products as a measure of the inten-
sity of competition. We use the variance of the variables describing a
firm’s relative product quality as a measure of product heterogeneity: the
higher this variance, the lower is the competitive threat between firms.
Furthermore, we can expect demand saturation to decrease the rate of
growth of the respective markets and thus the scope of co-operative R&D.
In this phase of the industry life-cycle minor improvements of the technol-
ogy could lead to considerable advantages for a single firm. To capture this
influence, we draw again on the life-cycle and assume that in later stages of
this life-cycle the rate of growth of demand is likely to decrease. Finally, the
techno-economic performance of the network members, again approxi-
mated by the relative quality compared to the average performance of all
firms, is itself an indicator of the attractiveness of joining a network. In cases
where the performance of network members is below the average perfor-
mance of the whole firm population, the networking strategy significantly
loses attractiveness.
Networking decisions
Next, the firms have to decide whether they want to co-operate or not.
Generally two forms of co-operation are possible:
1. co-operation focusing on complementary assets, i.e. firms are induced to

co-operate to acquire technological or economic competencies that they
do not possess but that they judge crucial for their economic success
2. co-operation focusing on general complementarities (i.e. the bundling
of R&D efforts in a specific direction) and synergies (i.e. detecting
potentials for cross-fertilisation by the combination of different tech-
nological capabilities). For example, it is possible to conceive of a divi-
sion of labour in which firms pursue similar objectives using similar
competencies, but they collaborate in order to speed up the innovation
process and to spread the relative uncertainty over the network.
In the form of co-operation (1) DBFs play the role of translators, while in
form (2) they play the role of explorers. Consequently, the networking
decision depends on the respective competencies and capabilities that firms
have accumulated. For example, a small start-up DBF in its early phases
is not able on its own to raise funds for R&D and necessarily has to look
for a partner in order to obtain funding. In the same way established LDFs
which want to become active in the promising fields of biotechnology but
have no internal technological competencies need collaboration partners
experienced in these fields. On the other hand, firms with highly developed
capabilities would not run the risk and share their knowledge with poten-
tial competitors in the stages immediately preceding the introduction of an
innovation.
Matching process
Finally, we have to decide on the mechanism which brings together different
firms willing to co-operate. Although different mechanisms are conceivable
we think that a mechanism which could be labelled ‘success breeds success’
is best suited to our purposes. Success breeds success means that firms
would tend to pick collaborators with the highest technological and/or eco-
nomic capabilities. We are here assuming that firms are able to advertise
their own capabilities and to evaluate those of potential co-operating part-
ners. This seems to be a realistic assumption, especially in the biotech
industry, where firms are ranked on the basis of their technological perfor-
mance, which is advertised by press announcements, publications, patents
and even by the professional standing of the scientists hired by firms,
including the Nobel prize winners present on their scientific committee.
3.3 Networking Consequences
After having introduced the way firms get together in innovation networks
we now have to focus on the consequences of networking. By entering into
a collaboration the agents are exchanging their know-how. This means that
firms can benefit from the efforts of other firms in order to build up their
own capabilities.
Absorptive capacities
The extent to which a firm can benefit from the knowledge-flow available
by co-operation depends on its absorptive capacity (see Cohen and
Levinthal 1989 and Cantner and Pyka 1998). In turn, absorptive capacity
is expected to increase with the firm’s previous experience in co-operation.
This means that external knowledge is not easily integrated within their
own knowledge-stock, but certain prerequisites have to be fulfilled and a
minimum amount of experience is necessary. This also means that the
amount of knowledge which flows within the network is severely limited.
Financial flows
Start-up DBFs with missing economic competencies cannot finance their
own R&D and are obliged to find a co-operation partner. In this case an
LDF co-operating with a DBF is supposed to provide the required research
funding. Another possibility for DBFs to acquire R&D money is to apply
for venture capital, of which we assume an exogenous supply growing at a
constant rate. Access to venture capital is competitive. Amongst the firms
applying only those which show the best record in biotechnological capa-
bilities as well as in previous co-operations are funded.
Knowledge-flows
One of the most important advantages of participating in an innovation
network is the access to channels of knowledge-flow. External knowledge
exerts an impact on the innovation probability function and depends on the
amount of absorptive capacities, as well as on the technological capabili-
ties of the co-operating firms. Participating in an innovation network exerts
a threefold influence: first, the research budget of a firm is reduced due to
co-ordination costs and, in the case of a co-operation with a DBF, by the
financial support of this firm. Second, absorptive capacities are positively
influenced by entering into a new collaboration as the experience with inte-
grating external knowledge is increasing. Finally, external knowledge
becomes available via knowledge-flows between the collaborating firms.
3.4 Competition Processes
The innovative activities of firms are undertaken in an economic environ-

ment which is characterised by a certain degree of competition. On the one
hand, the firms offering products on the final market compete with each
other in attracting demand. Also, those firms whose aim is to offer new
technological knowledge compete in a particular way with other firms in
acquiring the respective funds. Finally, firms who want to buy the respec-
tive knowledge also compete for co-operation with the most attractive
research laboratories.
The two levels of competition take place in two different markets: the
market for final products and the market for knowledge. Of course, we
know that markets for knowledge cannot exist due to their imperfections.
However, the existence of DBFs, which very often though not always func-
tion as contract research organisations, implies that in particular circum-
stances such imperfections can be reduced to a level where a market for
knowledge, although very imperfect, can exist. In fact, co-operation often
exists between firms operating in different markets (e.g. for final goods and
for knowledge) and thus having a complementary relationship. Of course,
this does not exclude the possibility that firms operating in the same market
can co-operate.
On the final markets firms compete in terms of prices and quality which
are, in a dynamic context, determined by their innovative success. Generally,
one would expect that a successful innovator will be able to attract demand
away from its competitors because consumers can choose between several
goods. These substitution effects are due to price and quality changes which
are the results of the following actions and reactions:
Introducing a new product with improved quality characteristics creates
additional demand allowing the innovator to charge higher prices.
1. In the case of an introduction of a new product by two or more verti-

cally integrated firms who co-operated in the R&D stages the increase
in demand is divided between the firms involved.
2. As a reaction to this quality-induced substitution effect, non-innova-
tors in related markets lower their prices in order to keep the loss in
demand as small as possible.
3. Exploitation of technological opportunities of an already existing
technology allows the respective innovator to reduce its price, thereby
increasing the demand for his product;
4. As a reaction, non-innovators could fight their loss in demand by also
lowering their prices, thereby, however, reducing their profit margin.
Another form of competition takes place in finding the most attractive

network partner which is described with the help of the notion of success
breeds success (see above). Firms engaged in the search for a co-operation
partner will match with those which show either the most developed tech-
nological or the most developed economic competencies.
By choosing a heterogeneous multiproduct oligopoly (see Kuenne 1992
and for an application in a simulation model Cantner and Pyka 1998) we
allow for the relationships described above. Firms are offering their goods on
a heterogeneous product market. Due to an innovation and the introduction
of a new commodity on these markets the relative market share of the
already existing goods gets eroded. In this way, we also generate the endoge-
nous incentives to the firms to engage in innovation, as they cannot survive
in the long run by relying on their original established positions which are
continuously threatened by the innovative actions of their competitors.
3.5 The Basic Structure of the Model
The following flow-chart summarises the basic structure of our model.

Starting with firms and industry characteristics of the previous round firms
Firm & industry

characteristics at time t1
Environmental
conditions for Firm decisions:
networking R&D budget
co-operation
Innovation I:
Actors’ development of Heterogeneous
knowledge- a new technology/product oligopoly
stocks
Innovation II:
Introduction of the new technology/
product to the market
Exit
Figure 5.5 Flow chart
have to decide whether to go it alone or to co-operate. They are influenced

by environmental conditions either favouring or inhibiting the growth of
networks. After having found a co-operation partner in the matching
process the firms enter the innovation stages which on the one hand
influence the industry and firm characteristics, and on the other hand the
market outcomes of the next round.
4. THE MODEL’S RESULTS
In this section we introduce some of the numerical results of the network

evolution. These results are then compared in a history-friendly manner
with real-world developments in order to see whether our model is able to
reproduce some of the developments observed in reality.
Figure 5.6 shows the development of the network density. After a first
increase in the density the dynamics of network growth comes to rest after
around ten periods, and even starts slightly decreasing until period 45.
However, after that period network density starts increasing again, until it
begins to oscillate around a value which is twice the average at the begin-
ning. This can already be interpreted as evidence for the changed role of
DBFs, which in the first periods find temporary collaboration partners in
0.06
0.04
0.02
0
1 46 91 136 Time
Figure 5.6 Network density (moving averages)
the population of LDFs. These collaborations are mainly oriented towards

bridging the gap between the new biotechnologies and the established
industry. Later on, however, the DBFs are finally considered as an exten-
sion of internal R&D facilities, allowing LDFs to explore a wider oppor-
tunity-space. Therefore, collaborations become more frequent and lasting
in more advanced states of the industry evolution.
In section 2 we argued that the persistence of innovation networks in the
biotechnology-based industries could not be explained by means of only
one role played by DBFs. Whereas in early stages the small technology-
oriented firms play the role of translators, facilitating the absorption of the
new technologies by LDFs, in later stages they become more emancipated
as collaboration partners. This means that they no longer serve solely as
institutions transferring knowledge between academic and industrial
research, but become explorers, allowing LDFs to investigate a broader
technological portfolio in an increasing complex technological opportu-
nity-space. This changed role of the DBFs accordingly has to be observed
also in the simulation as a development which endogenously takes place
within our model’s specification. In Figure 5.7 we therefore plot the specific
composition of collaborative agreements.
In the first part of the period investigated only co-operative arrange-
ments between LDFs and DBFs are found: DBFs are supposed to support
LDFs in building up their biotechnology competencies; and, as a compen-
sation for their R&D efforts, they are funded by LDFs. As soon as some
DBFs start to earn their own money they also initiate further collabora-
tions in which they are no longer playing the role of translators but that of
explorers. In the simulation we find that these collaborations between DBFs
become of increasing importance in later stages. Now, the co-operative
No.
LDF/DBF
DBF/DBF
Figure 5.7 Composition of collaborations: LDF/DBF and DBF/DBF
agreements aim at bundling know-how and joint exploration of the tech-

nological opportunities. At the end of the period studied the number of
agreements between DBFs is becoming comparable to that of agreements
between LDFs and DBFs.
This changed role played by DBFs is also mirrored in the decisions made
by firms with respect to their collaboration policy. In the model we differ
between three strategies: first, the go-it-alone strategy chosen by firms
who are either at the technological frontier and do not want to share know-
how with followers or by firms which already are engaged in several co-
operations. The second strategy aims at attracting research funds; this
strategy is adopted by the DBFs in their early phases, when they enter the
scene with highly developed technological competencies but they have no
economic competencies. Finally, the third strategy aims at the integration
of external knowledge in order to build up jointly in a network the capa-
bilities necessary for the introduction of an innovation.
Figure 5.8 shows the share of strategy choices the firms (LDFs and
DBFs) make with respect to the specific form of co-operation they want to
initiate. This decision, of course, is always influenced by the relative posi-
tion of the firm, and in particular depends on whether the firm is already
engaged in one or more co-operations. Therefore, the white area represent-
ing the proportion of firms who do not want to start a new co-operation
has to grow with the number of already existing networks because of the
increasing co-ordination costs. In the early stages almost all firms wish to
start new co-operative relationships according to the translator’s type
(black shaded area). With the growing diffusion of technological compe-
tencies within the population of LDFs and as some successful DBFs
No co-operation
Translator
Explorer
t
Figure 5.8 Willingness to start a new collaboration and Strategy choice
become vertically integrated producers, this decision shifts nearly exclu-

sively to collaborative relationships following the explorer’s type (grey
shaded area). This also means that the new collaborative agreements to be
started in later periods will almost entirely be of this latter type which is in
line with the results shown in Figure 5.7.
For an immediate comparison of our artificial world with developments
in the real world we use graph theory (e.g. Burt 1980) which offers some
measures to compare different networks from a structural perspective.
These measures describe, for example, the adjacency, the reachability and
the connectivity of a network as well as the centrality of single actors
(e.g. Freeman 1979). By comparing these figures we will get some first
insights into whether we have caught the basic mechanisms of networking
in our industries, or where we have to modify specific components of our
model in order to improve our understanding.
The empirical database contains collaborations in the biotechnology-
based industries from 1982 to 1998. For more details of the database see
Pyka and Saviotti (2002).
In the following we have applied two concepts, in particular the average
distance and the degree of centrality in order to compare time-series of
artificial networks with real networks. The computations are done with
Ucinet (Bogatti et al. 1999), a software tool designed for network analysis.
Figure 5.9 shows the development of the average distance, a measure of
the average shortest path between two nodes for our artificial network as
well as for our empirical database. This measure can be interpreted as an
indicator of the diffusion of information in a network. Both figures show a
structural similarity in a sequence of peaks which indicate a qualitative
change in the network structure. Whereas these peaks grow in magnitude in
3 10
2.5
8
2
6
1.5
4
1
0.5 2
0 0
a001 a020 a040 a060 a080 a100 a120 a140 a160 a180 a200 1982 1985 1987 1989 1991 1993 1995
Figure 5.9a Average distance in the artificial world Figure 5.9b Average distance in the real world
116
2
2.5
1.5
2
1.5 1
1
0.5
0.5
0 0
a001 a020 a040 a060 a080 a100 a120 a140 a160 a180 a200 1982 1985 1987 1989 1991 1993 1995
Figure 5.10a Degree of centrality in the artificial world Figure 5.10b Degree of centrality in the real world
the real world, their artificial counterpart stays almost on the same level and
also the second peak is unimodal in the real world compared to the bimodal
peak in the artificial world. This difference is basically caused by the different
sizes of the empirical and the artificial populations. The populations in the
simulation run are kept smaller due to computational restrictions.
To rule out the influence of network size, index-oriented measures exist.
Here we measured and calculated the degree of centrality for both of our
worlds shown in Figure 5.10. The degree of centrality measures the asym-
metry in the roles played by various actors in a network. For this aspect of
network activity we also find a broad correspondence of our artificial and
real worlds. The sequence of three peaks can be interpreted as a conse-
quence of the changing role of DBFs in the networking processes. The first
peak is caused by DBFs playing the role of translators supporting the
LDFs in their efforts to overcome the gap between their dominant knowl-
edge orientation and the upcoming new knowledge-base in biotechnology.
The second peak has to be characterised as an intermediate phase, with
some DBFs who have already become vertically integrated producers and
LDFs still mainly concerned with building up competencies in the new
field. The third wave in networking is then caused by a tremendous growth
in the technological opportunity-space, where networking is considered to
be a strategy to cope with the speed and complexity of technological devel-
opment. In this phase DBFs play the role of explorers allowing the large
and established firms to explore a wider range of technological approaches
within biotechnology.
5. CONCLUSIONS
This chapter provides a simulation analysis of the evolution of innovation
networks in the biotechnology-based industries. Since this is an applied
simulation exercise, great emphasis is placed on the characteristic features of
this industry. The results of the simulations are compared to developments
in the real world by applying concepts of graph theory which provide us with
some measurements of the overall network dynamics. Although there are
still some significant differences between the artificial evolution of network
structures and the real-world networks, the results look promising as they
are able to reproduce at least qualitatively some developments which are
observed in reality. The next steps have to be to balance the different mech-
anisms and to find relative weights in accordance with their specific impacts.
Once such weights were attributed, different scenarios could be analysed,
showing the influence of different environments as well as of policy measures
aiming at the establishment of these new biotechnology-based industries.
One final remark with respect to the stochastic influence of innovation

processes on results seems to be necessary. By repeating the simulation
experiment several times the Poisson-distributed random number respon-
sible for the innovation event leads to varying relationships between the
firms in our sample. However, although collaboration partners change the
overall network dynamics do not depend on stochastic influences, but
remain rather stable over a large number of simulation experiments per-
formed in a Monte-Carlo-method fashion.
To summarise, in our research we started from the empirical literature
and from the existing case studies on the biotechnology-based sectors and
developed a formal representation of innovation networks that, while
abstract, matched a number of the observed features of innovation in these
sectors. Going through this analytical exercise has significantly sharpened
our theoretical understanding of the key factors behind the development
of networking in the biotechnology-based sectors and contributed to a
more general understanding of innovation networks in other sectors.
NOTES
1. The description of the model is given in a non-equation form, not explicitly introducing
into details. Interested readers may obtain a comprehensive description of the model by
writing to the authors.
2. In this respect, a major methodological advantage of simulation studies shows up in the
construction of the innovation processes. Whereas in traditional optimisation models
there is no difference between the modeller and the modelled agents, simulation analysis
allows programming random numbers whose statistical distribution is unknown to the
agents in the model, see Pyka (1999, pp. 189ff.).
REFERENCES
Bogatti, S.P., M.G. Everett and L.C. Freeman (1999), Ucinet 5 for Windows:
Software for Social Network Analysis, Natick, MA: Analytic Technologies.
Burt, R.S. (1980), ‘Models of network structure’, Annual Review of Sociology,
6, 79–141.
Cantner, U. and A. Pyka (1998), ‘Absorbing technological spillovers: simulations in
an evolutionary framework’, Industrial and Corporate Change, 7, 369–97.
Cohen, W.M. and D. Levinthal (1989), ‘Innovation and learning: the two faces of
R&D’, The Economic Journal, 99, 569–96.
Freeman, L.C. (1979), ‘Centrality in social networks, conceptual clarification’,
Social Networks, 1, 215–39.
Grabowski, H. and J. Vernon (1994), ‘Innovation and structural change in phar-
maceuticals and biotechnology’, Industrial and Corporate Change, 3, 435–49.
Kuenne, R.E. (1992), The Economics of Oligopolistic Competition, Cambridge, MA:
Blackwell.
Malerba, F., R.R. Nelson, L. Orsenigo and S.G. Winter (1999), ‘History friendly
models of industry evolution: the computer industry’, Industrial and Corporate
Change, 8, 3–40.
Nelson, R.R. and S.G. Winter (1982), An Evolutionary Theory of Economic Change,
Prahalad, C.K. and G. Hamel (1990), ‘The core competencies of the corporation’,
Harvard Business Review, 90, 79–91.
Pyka, A. (1999), Der kollektive Innovationsprozeß – Eine theoretische Analyse
absorptiver Fähigkeiten und informeller Netzwerke, Berlin: Duncker & Humblot.
Pyka, A. and Saviotti, P.-P. (2002), ‘Innovation networks in the biotechnology-
based sectors’, in A. Pyka and G. Küppers (eds), Innovation Networks – Theory
and Practice, Cheltenham UK and Northampton, MA, USA: Edward Elgar.
Saviotti, P.-P. (1998), ‘Industrial Structure and the Dynamics of Knowledge
Generation in Biotechnology’, in Jacqueline Senker (ed.), Biotechnology and
Competitive Advantage: Europe’s Firms and the US Challenge, Cheltenham, UK
and Northampton, MA, USA: Edward Elgar.
Staropoli, C. (1998), ‘Cooperation in R&D in the pharmaceutical industry’,
Technovation, 18, 13–24.
Teece D. (1986), ‘Profiting from technological innovation’, Research Policy,
15, 285–305.
Tushman, M.L. and P. Anderson (1986), ‘Technological discontinuities and orga-
nizational environments’, Administrative Science Quarterly, 31, 439–65.
6. Sociodynamics – an integrated
approach to modelling in the
social sciences
Wolfgang Weidlich
INTRODUCTION
Sociodynamics has the ambitious aim of providing a general strategy,

which means a frame of theoretical concepts for designing mathematical
models for the quantitative description of a rather broad class of collective
dynamical phenomena within human society.
In the meantime these design principles have been applied to models of
migration of interacting populations, group dynamics, political opinion
formation, non-equilibrium economics and urban evolution.
A full acount of the sociodynamic method and its applications can be
found in a recent book of the author (Weidlich 2002).
The present chapter sketches the main ideas and, as an example, their
application to problems of migration and urban evolution.
Since sociodynamics provides an open frame for model design, it can be
expected that previous models will not be the end but rather the beginning
of further progress.
1. PRINCIPLES
1.1
Sociodynamics aims at describing the dynamics of processes in human

society – i.e. not only stationary states – on the macro level by an overar-
ching mathematical concept of model design.
120
Sociodynamics 121
1.2
The micro level of individuals and the macro level of collective evolu-
tion are mutually interdependent. The formalism takes this fact into
account.
1.3
Concepts of social science, notably the motivations, decisions and actions

of individuals under the influence of collective trends are integrated from
the start. Therefore, sociodynamics is not at all a ‘physicalism’.
1.4
Transition rates between states in the space of configurations of the social

system (e.g. of its demographic, economic, social or political sector) are
the central concept of sociodynamics. They comprise individual motiva-
tions and collective trends as well and engender the dynamics of the
system.
1.5
Equations of evolution for system states are derived starting from the tran-
sition rates, either on the probabilistic level (including all fluctuations) or
on the deterministic level (for the mean evolution).
1.6
Simulations of scenarios represent possible evolutions according to the

models. After appropriately adjusted calibration of trend parameters
the real evolution can also be interpreted and forecast in terms of the
models.
2 FORMAL FRAME
2.1 State and Variables of State
The state of a society is characterised:

on the one side by a set of personal variables
n {. . . ni . . .} (6.1)
where the meaning of ni is the following:
ni number of members of a subpopulation P having an ‘attitude’ (i.e.

opinion, or kind of activity) ‘i’.
on the other side by a set of material variables
m{m1, . . .,mk . . . mK} (6.2)
where the meaning of mk can for instance be: number of produced com-
modities per unit of time; price of one unit of a commodity; number of
buildings on a site of a city; etc. The multiple of variables
v {n; m} (6.3)
is an element of the space of socioconfigurations. It describes the personal

and material state of the society, or of one of its sectors, e.g. of the demo-
graphic, urban, economic, sociopolitical sector, respectively.
2.2 Transitions and Transition Rates
The dynamics of the society is described by transitions from any one given
initial state va to a neighbouring state vb.
Transition rates (by definition probabilities per unit of time proceeding
from state va to state vb) engender the transitions, i.e. they are the ‘driving
forces’ of the transitions. They are induced by the activities of individuals
(e.g. on the migratory, social, economic or political sector).
Since transition rates are by definition always positive definite quantities
they can be written in the form
wba(vb; va)napba na0 exp{Mba(vb, va)} (6.4)
where na is the number of individuals initiating the transition va ⇒ vb. The

individual transition rates pba can be mathematically represented in a stan-
dard form:
pba(vb; va)(vb,va)exp{ub(vb) ua(va)} (6.5)
Interpretation of transition rates

From the form of equation (6.5) of the individual transition rates there
follows their general sociodynamic meaning:
The motivation potential Mba(vb, va) depends on the motivation of the
individuals initiating the transition. Their motivation arises from com-
paring and evaluating the initial state va versus the intended state vb.
Sociodynamics 123
The mobility (vb, va)(va, vb) is a measure (symmetric in va and vb)

of the speed of the transition.
The attractiveness ua(va) and ub(vb) is the respective measure of the utility
of the initial state va and intended state vb, valuated by the individuals
initiating the transition. If ub(vb) exceeds ua(va), the transition va ⇒ vb is
favoured against the inverse transition vb ⇒va according to equation
(6.5). The functions ua(va) and ub(vb) are also denoted as ‘dynamic utility
functions’ in order to distinguish them from utility functions used in neo-
classical modelling concepts.
2.3 Evolution of the Socioconfiguration
The sequence of elementary transitions between neighbouring states even-

tually leads to the long-term evolution of the social system. The descrip-
tion of this evolution can take place on two levels of exactitude:
If the probabilistic nature of all stepwise actions of individuals is fully
taken into account, their composition will also lead to a probabilistic evo-
lution of the collective macrovariables introduced in section 2.1. We thus
obtain a probabilistic description of the evolution of the society on the
macro level.
If instead only the mean evolution of a bundle of ‘stochastic trajectories’
is considered one obtains by neglection of fluctuations a deterministic
description of the evolution of the society or its sectors in terms of ‘quasi-
mean values’.
The probabilistic description

This description is based on the master equation, i.e. the central equation
for the evolution of the probability distribution over all collective
macrovariables of the social system.
Let P(v; t) be the probability of finding realised the configuration v of
the social system at time t. Since any one of the – mutually exclusive – con-
figurations must be realised, the total probability of all configurations is
equal to one:
P(v; t) 1
v
(6.6)
The master equation, which is compatible with (6.6), reads in its general form:
dP(va; t)
dt
wba(vb; va)P(va; t)
vb
(6.7)
˛

vb
wab(va; vb)P(vb; t)
It can easily be understood that the change with time of the probability va,
i.e. the l.h.s. of (6.7), comes about by two contraproductive processes,
namely:
1. the outflux from probability P(va; t) into the probabilities of neigh-

bouring configurations vb (first term of the r.h.s. of (6.7))
2. the influx from probabilities P(vb; t) of neighbouring configurations
vb into the probability of configuration va (second term of the r.h.s.
of (6.7)).
It should be noted that the master equation, in spite of its easy interpreta-
tion, is not easily solved!
The deterministic description

The concrete single social system traverses a ‘stochastic trajectory’. It is
the macroscopic resultant of all accidential, but motivation-guided actions
of individuals. Considering, however, a bundle of such trajectories start-
ing – under equal model assumptions – from the configuration v(t0), there
results within the small interval of time a ‘preferred’ mean evolution to
configuration v(t0 ). If the sequence of ‘preferred mean evolutions’ is
composed, one obtains deterministic differential equations for the ‘quasi-
mean values’ v(t).
In the case of v {. . . ni . . .; . . . mk . . .}, which means of the personal
and material variables of the socioconfiguration (3), the quasi-mean value
equations have the form:
dmk
wk(mk , m) wk(mk, m) (6.8)
dt
dni
dt
wij(nij, n) wji(nji, n)
j j
(6.9)
where nij , nji are personal configurations adjacent to n, and mk , mk

are material configurations adjacent to m. And wij (nij , n), wji (nji , n),
wk(mk , m), wk(mk, m) are the transition rates for the transitions
n ⇒ nij ; n ⇒ nji ; m ⇒ mk ; m ⇒ mk , respectively.
Sociodynamics 125
3. SCENARIOS OF SELECTED MODELS
We now consider the general procedure for designing selected models and
we present some of their results in terms of characteristic scenarios.
3.1 Migration
As a first model we consider the interregional migration of socioculturally

differentiated populations.
The procedure of model design consists of the following steps:
1. Introduction of variables: numbers of P ‘subpopulations’ distributed

over L regions
2. Choice of transition rates for migration between regions: by introduc-
ing plausible ‘dynamic utility functions’ taking into account migratory
trends
3. Setting up of equations of evolution (master equation; quasi-mean
value equations)
4. Calculation of characteristic scenarios after explicit choice of trend-
parameters.
The model comprises globally different cases of migratory dynamics:
● the development of a homogeneous mixture

● the development of ghettos
● the development of a ‘penetration-evasion dynamics’.
and there exist phase-transitions between these cases, which can be studied
in terms of the model.
The simplest case

Two populations 1, 2 migrating between two regions a, b.
We now skip steps 2 and 3 and only present characteristic scenarios,
namely solutions of the quasi-mean value equations and the master equa-
tion for selected trend-parameters in the two populations–two regions
case.
Variables:
1 x 1: Scaled distribution of population 1 over regions a and b.

1 y 1: Scaled distribution of population 2 over regions a and b.
Case. 1 Trend towards homogeneous population mixture for weak inter-

nal agglomeration trend within, and weak symmetrical segregation trend
between both populations.
0.5
y 0
0.5
1
1 0.5 0 0.5 1
x
Figure 6.1a Fluxlines of the quasi-mean values
The stable homogeneous population mixture x0; y 0 of both
populations over both regions is approached.
Figure 6.1b The stationary unimodal probability distribution of both

populations over both regions
The maximal probability is at x0; y 0.
Sociodynamics 127
Case. 2 Trend towards ghetto formation for moderate internal agglomer-

ation trend within, but strong symmetrical segregation trend between both
populations.
0.5
y 0
0.5
1
1 0.5 0 0.5 1
x
Figure 6.2a Fluxlines of the quasi-mean values

Stable stationary ghetto situations are approached, where either
population 1 is concentrated in region a and population 2 in region b, or
vice versa.
Figure 6.2b Bimodal stationary probability distribution with probability

peaks at the ghetto situations
Case. 3 Trend towards homogeneous population mixture for moderate

internal agglomeration trend within, and strong assymmetric segregation
trend between both populations.
0.5
y 0
0.5
1
1 0.5 0 0.5 1
x
Figure 6.3a Fluxlines of quasi-mean values

They approach the stable focus x0; y 0 of homogeneous population
mixture.
Figure 6.3b The stationary unimodal probability distribution with

maximum at x 0; y 0
Sociodynamics 129
Case. 4 Evolution of a penetration-evasion dynamics for strong internal

agglomeration trend within, and strong assymmetrical segregation trend
between both populations.
0.5
y 0
0.5
1
1 0.5 0 0.5 1
x
Figure 6.4a Fluxlines of quasi-mean values

They approach a limit-cycle, whereas the point x0, y 0 is now an
unstable focus.
Figure 6.4b Stationary probability distribution with a ‘probability ridge’

along the limit-cycle of the quasi-mean values
Interregional migration in the Federal Republic of Germany between 1957

and 1983
Another application of the migration model consisted in the comparison
and correlation between the model and empirical data for the interregional
migration in the Federal Republic of Germany from 1957 to 1983.
For this purpose the following procedure was implemented:
1. The yearly population numbers of the federal states of Germany

during the past 25 years were available.
2. The regional utility functions were determined year by year for this
period so that the results of the model equations coincided with the real
evolution.
3. The correlation of the regional utility functions with regional socio-
economic key variables was determined by regression analysis.
The result of the determination of the time-dependent regional utility func-

tions by comparison between model and empirical data is exhibited in
Figure 6.5.
The result of the correlation of regional utility functions ui(t) with

regional socio-economic key variables by regression analysis is exhibited in
Figure 6.6.
3.2 Integrated Evolution of a City and its Population
As a second model we consider the interrelated evolution of a city and its

population. The ‘sociodynamic’ procedure of model design is now more
complex and consists of the following steps:
Introduction of variables
The set of variables consists of material and personal variables.
The material variables are formed by first tessellating the area of city c
and hinterland h into a lattice of sites with co-ordinates i(i1, i2), j(j1, j2); . . ..
Thereupon on each site ‘i’ several units mi(K) of types . . . K . . . L . . . of build-
ings can be erected until eventually the capacity limit Ci is reached. The
multiple
m {. . . mi(K) . . . m(L) (K) (L)

i . . . mj . . . mj } (6.10)
then forms an ‘occupation number representation’ of a concrete material

state of city and hinterland which is an element of the ‘space of city and
hinterland configurations’.
1
0.50
131
0.50
1
1955 1960 1965 1970 1975 1980 1985
Schleswig-Holstein Rheinland-Pfalz
Hamburg Baden-Württemberg
Niedersachsen Bayern
Bremen Saarland
Nordrhein-Westfalen Berlin
Hessen
Figure 6.5 The estimated regional utilities ui(t) of the Federal Republic of Germany
1
0.50
132
0.50
1
1955 1960 1965 1970 1975 1980 1985
Schleswig-Holstein Rheinland-Pfalz
Hamburg Baden-Württemberg
Niedersachsen Bayern
Bremen Saarland
Nordrhein-Westfalen Berlin
Hessen
Figure 6.6. Key-factor analysis of the regional utilities of the Federal Republic of Germany
Sociodynamics 133
The personal variables describe how several subpopulations P are

distributed over the sites ‘i’, so that Ni members of P live on site ‘i’. The
multiple
N {. . . Nj . . .} (6.11)
then characterises this distribution and is an element of the ‘space of pop-
ulation configurations’.
Choice of transitions and transition rates

The integrated material and personal evolution of city and hinterland is
engendered by the following elementary transitions:
● construction and tearing down processes of building units of type K

on site ‘i’
i . . .} ⇒ mi {. . ., (mi 1), . . .}.

m {. . . mK (K) (K) (6.12)
● birth-death processes of members of P on site i
N {. . . Ni . . .}⇒Ni {. . ., (Ni 1), . . .}. (6.13)
● migration of members of P from site ‘i’ to site ‘j’
N {. . . Nj . . . Ni . . .}⇒ Nji {. . . (Nj 1) . . ., (Ni 1), . . .}. (6.14)
To these elementary transitions there belong the following transition rates,

respectively
wK
i↑
(m, N) ; wK
i↓(m, N) (6.15)
wi↑(m, N) ; wi↓(m, N) (6.16)
wji(m, N) (6.17)
According to their standard form (6.4), (6.5), the transition rates can and
must be further specified in order to obtain a concrete integrated city and
population evolution model. This has been done in the author’s book men-
tioned in the introduction, from which we take the further concrete results.
Setting up of equations of evolution

We restrict further consideration to the quasi-mean value equations which
have, according to equation (6.8) and equation (6.9), the following general
form:
dmj(K)
wj(K)
↑
(m, N) w(K)
j↓ (m, N) (6.18)
dt
for the evolution of the material variables, and
dNj()
dt
wji(m, N) wij(m, N)
i i
(m, N) w (m, N) (6.19)
wj↑ j↓
for the evolution of the personal variables.
Scenarios
Integrated models of course contain several sectors which may be treated
separately under simplifying assumptions. In the present model, the inter-
action between city- and population-evolution essentially depends on how
city and hinterland provide capacities Cc and Ch, respectively, for a
(perhaps growing) total population.
If on the other hand the capacities Ci of each site i of the city are (for-
mally) kept constant, the evolution of the city configuration must come to
an end when all capacities are exhausted. This case is treated in the follow-
ing scenario.
The first sector: a simple model city

For a simple ‘model city’ a square lattice of 1515 sites is assumed
with a ‘natural’ constant capacity distribution over the sites determined by
a landscape which contains a river. Only apartments and commerical
buildings are assumed to exist and the utility functions in the transition
rates take neighbourhood relations between both kinds of buildings into
account. Starting from a ‘natural’ initial condition of a not yet developed
city the concrete explicit quasi-mean value equations then eventually
lead to a final distribution of ‘apartments’ and ‘commerical buildings’
over the sites of the city. (It should be noted that the end configuration
depends on some details of the initial configuration. Hence the evolution
is ‘path-dependent’ due to the non-linear structure of the quasi-mean
value equations.)
The final distribution of apartments on the lattice of 1515 sites of a
‘model city’ is shown in Figure 6.7.
The final distribution of commercial buildings on a lattice of 1515
sites of a ‘model city’ is shown in Figure 6.8.
The second sector: migration between hinterland and city

The second sector of the integrated model consists in the migration of the
population between hinterland (h) and city (c), which is treated in the next
scenario.
Sociodynamics 135
0.1
x(p,q)
0.05
0
15
13 15
11 13
11
9 9
q 7 7 p
5 5
3 3
1 1
Figure 6.7a Perspective representation of distribution of apartments
15 15
13 13
11 11
q 9 p
9
7
7
5
5
3
3
1
1
Figure 6.7b Topographic representation of distribution of apartments

0.06
0.04
y(p,q)
0.02
0
15
13 15
13
11
11
9
q 9 p
7
7
5 5
3 3
1 1
Figure 6.8a Perspective representation of distribution of commercial

buildings
15 15
13 13
11
11
q 9 p
9
7
7
5
5
3
1 3
1
Figure 6.8b Topographic representation of distribution of commercial

buildings
Sociodynamics 137
In this scenario we let the capacities Cc and Ch of city and hinterland

depend on the respective time-dependent population numbers. The model
equations are essentially simplified, if only aggregated population numbers
are considered as follows:
Nc ncP city population

Nh nhPhinterland population
Nc Nh P total population
nc, nh: population shares, where (nc nh)1
n (nc nh): majority variable, where1 n 1 (6.20)
In the – now presumed – case of equal net birth-rates in city and hinterland
the population equations (6.19) are decisively simplified to two equations
for the population shares nc, nh only:
dnc
phcnc pchnh
dt
(6.21)
dnh
phcnc pchnh
dt
or, because of (nc nh)1, to one equation for the majority variable n(t)
only:
dn
(pch phc) (phc pch)n (6.22)
dt
The transition rates have the standard form:
phc v exp [vc vh] v exp [a(t) b(t)n]

(6.23)
pch v exp [vh vc] v exp [a(t) b(t)n]
where vc and vh are measures for the attractiveness (‘dynamic utility’) of

city and hinterland, respectively, and where it turns out, according to
plausible assumptions of the form of vc and vh, that a(t) and b(t) are
co-efficients depending on the (in general time-dependent) total popu-
lation P(t) and on the capacities Cc(t) and Ch(t) provided by city and
hinterland.
Equation (6.22) for the evolution of the majority variable can be written
in the form
dn V(n; t)
(6.24)
dt n
after introduction of the ‘evolution potential’

V(n; t) v e(a(t) b(t)n) b(t)n 1
2
b (t)

1
b(t) (6.25)
e (a(t) b(t)n) b(t)n 1
b2(t)

1
b(t)
0
2
4
V(n^,t)
6
8
1 0
0.5 100
0 200
n^ 0.5
1 400 300 t
500
Figure 6.9 Development with time of the evolution potential V(n, t) for
growing total population and worsening balance of capacities
between hinterland and city
5
10
V(n^,t)
15
20
25
1 100 0
0.5 400 300 200
0 0.5 600 500
1 700 t
n^
Figure 6.10 Evolution potential V(n, t) with ‘system point’, representing

the evolution of the majority variable n(t) with time
Sociodynamics 139
Among the possible cases we consider the dramatic case of a growing

total population and a simultaneously worsening balance of capacities
between hinterland and city favouring the city. Then the slowly time-
dependent potential V(n; t) develops from a two-minima potential into
a one-minimum potential. Correspondingly, the majority variable n(t),
i.e. the solution of equation (6.24), shows a first slow, then dramatically fast
evolution exhibiting a rush of the population from hinterland into the city.
Such a ‘migratory phase transition’ can take place in developing countries
with a rural hinterland unable to feed and maintain its growing population
which escapes into the neighbouring metropole.
The evolution potential V(n; t) and the evolution of the majority vari-
able n(t) are exhibited in Figures 6.9 and 6.10.
REFERENCES
Weidlich, Wolfgang (2002), Sociodynamics – A Systematic Approach to Mathematical

Modelling in the Social Sciences, London: Taylor & Francis.
7. The concept of space in
trade – some evolutionary basics
Carsten Herrmann-Pillath1
1. THE CHALLENGE: EXPLAINING THE SPATIAL

DIRECTEDNESS OF TRADE
The theory of international trade is one of the most developed contempor-

ary applications of neoclassical equilibrium theory. Its political impact is
tremendous, for it provides the intellectual foundation of the free-trade
doctrine shaping the so-called ‘neoliberal’ approach to globalisation. Its
academic rank and file obeys a tough intellectual discipline and receives a
demanding, high-quality education.2 The high scholarly appreciation refers
to the elegance and coherence of the formal approach as well as to the
tremendous efforts in empirical testing and application. The general
significance of this strand of research is related to its very strong norma-
tive claims and, in particular, to the fact that the basic premise of trade
theory, the theory of comparative advantage, is the most cogent demon-
stration of why markets are beneficial to all participants.
So far, the evolutionary economics of international trade is not a major
concern. There are two salient topics, namely the impact of technology on
trade (the seminal study is Dosi et al. 1990) and the role of corporate strat-
egies and industrial policies in global competition (for example, see the
volume edited by de la Mothe and Paquet 1996). However, there is no sys-
tematic attempt at deconstructing the neoclassical approach. Of course,
one may hope that both approaches can co-exist, however, as they have
fundamentally conflicting foundations, I think that this challenge should
not be avoided. One obstacle to a reasonable debate is the fact that the
debate is always inextricably linked with the free-trade issue, so that evolu-
tionary approaches, unwillingly or not, are pushed into the corner of
potentially protectionist positions.
Facing the impressive record of neoclassical trade theory, deconstruct-
ing its theoretical basics is by no means an easy endeavour. The normal
scientific way to tackle a task of these dimensions is to focus on unresolved
issues and anomalies of a theory in order to accomplish an alternative
140
The concept of space in trade 141
explanation that, at the same time, includes the mainstream theory as a

special case.
1.1 Some Unresolved Issues in Trade Theory
There are two main anomalies in neoclassical trade theory (NTT). First,
NTT cannot exhaustively explain the direction of trade in space (or, what
the same thing amounts to, the location of production); second, the explan-
ation of the increasing role of multinational corporations in trade remains
difficult. The latter issue is linked with the former, because one major
research topic is the investigation into the patterns of foreign direct invest-
ment, which also involves spatial patterns of economic activity of those
firms (e.g. exports of foreign subsidiaries). In the past two decades, investi-
gating these unresolved issues has triggered a burgeoning literature, mostly
exploiting synergies between New Trade Theory and (New) Economic
Geography (Krugman 1995; Helpman 1998). The trigger for this effort is
the recognition of the blatant fact that real space does not actually exist in
trade theory (frequently assuming even zero transport costs), which in its
mainstream versions is a theory of the international allocation of produc-
tion as a result of international equilibrium exchange (for a penetrating
criticism, see Dunning 1995).
Let me summarise the main facts, beginning with the problem of spatial
directedness of trade.
1. The factor-proportions theory leaves a significant share of trade-flows

unexplained, i.e. given the current relative endowments of countries in
the world, trade-flows should be significantly larger than the observed
ones (Trefler 1993); at the same time, the general indeterminacy in
multifactor, multigoods models can only be solved by adding exoge-
nous costs of trade to the picture, hence disintegrating parameters in
an otherwise perfectly integrated market system (Davis et al. 1997;
Bernstein and Weinstein 1998).
2. The atheoretic gravity model is as useful as neoclassical trade theory in
explaining trade-flows, however, at the same time this model shows a
strong impact of singular spatial determinants on trade (Hummels
and Levinsohn 1995; Helliwell 1997).
3. The law of one price is systematically invalid in the long run, including
very strong ‘border effects’ across countries, although neoclassical
theory leads to the expectation that violations should be short-run phe-
nomena (Rogers and Jenkins 1995; Haskel and Wolf 1999); this issue is
related to the widespread occurrence of ‘pricing to market’ phenomena
in international trade, which renders the purchasing power parity
approach to real exchange rates problematic (Goldberg and Knetter

1997).
If we take a closer look at the arguments that are proposed to resolve these
empirical riddles, so far the most important ones are:
1. Taking country-specific productivity differentials into consideration,

differences either between observed and factual trade-flows or between
expected and factual production structures of countries could be
shown to reflect level-effects of the production function whilst still
being influenced by relative factor prices (Harrigan 1997; Hanson and
Slaughter 1999); in the most general terms, this amounts to an inter-
ference among relative and absolute advantages in the determination
of trade-flows.
2. The gravity model can be fit into a factor-proportions approach, if it
is implemented as a device to identify impediments to trade via the
demand side of the model, that is to add multidimensional ‘distances’ to
the neoclassical model, which at the same time reduces the indetermin-
acy problem in multifactor, multigoods models (Davis and Weinstein
1998).
3. Such kind of generalised ‘distances’ can also be referred to in order to
explain observed deviations from the law of one price in both the gravity
model and pricing-to-market models of imperfect competition, with
the singular ‘Armington bias’ as a determinant of directedness serving
as an econometric ‘catch-all’ parameter (Trefler 1995; Deardorff 1998;
Engel and Rogers 1999).
Whereas mainstream trade economists interpret these approaches to

a solution as supporting the neoclassical model, the particular argu-
ments actually create leeway for switching to other theoretical approaches,
simply by means of shifting the emphasis. That is, we could argue, for
example, that a technology-based theory of trade is the basic model,
which also shows the impact of relative factor endowments, without
necessarily including the fully fledged neoclassical equilibrium theory
as a reference. Similarly, generalised distances may be regarded as an
empirical element of an institutional approach to trade, sticking to
the gravity model without trying to adapt it to a general equilibrium
framework.
In the ‘pricing-to-market’ literature, a major focus is the role of corpor-
ate strategies and competences in market entry (Clark et al. 1999). Hence,
the aformentioned issues are also linked to the phenomenon of the trans-
national corporation, in the empirical respect at least.
Turning our attention to the second major field of research into open
questions in international economics, the main empirical issues are:
1. There is a significantly positive interdependence between trade (goods

with factor content) and investment (factor-flows), so that the implied
convergence in relative factor-endowments is not reflected in trade-
flows, implying that the factor-proportions theory is violated (Collins
et al. 1997).
2. Worldwide FDI is mostly among nations on the same level of devel-
opment and of the horizontal kind, which implies that investment
flows cannot be influenced by relative factor-endowments and, hence,
factor prices (Braunerhjelm and Lipsey 1998).
3. There is a strong impact of industry-specific determinants as opposed
to country-specific determinants, so that countries may not be the
relevant units of theorising (Markusen 1995).
4. Intra-industry trade is closely enmeshed with intrafirm trade, which
raises the question whether trade-flows are strongly influenced by the
institutional determinants of the ‘market versus firm’ alternative (Klau
1995).
The answers so far proposed by mainstream theory are, amongst others:
1. The factor-proportions approach may be empirically underdetermined

because firm-specific knowledge-stocks are neglected, which also implies
that observed trade might incompletely reflect intrafirm trade of the
respective services (‘headquarter services’) (Ekholm 1998a; Markusen
and Maskus 1999).
2. Equilibrium theory can be modified by including monopolistic com-
petition so that market access becomes an important motive for intra-
industry investment-flows, with market access being influenced by the
attraction of larger markets as a location of production with increas-
ing returns to scale, and with transport costs as a countervailing force
of dispersion of production (Zhang and Markusen 1997; Markusen
and Venables 1998).
3. Industry-specific effects may be related to differences in terms of the
use of knowledge and human capital, which implies different market
structures and firm behaviour, for example in terms of exploiting
specific managerial competences (Ray 1995).
4. The ‘market versus firm’ framework may lead to an enlargement of the
equilibrium theory by the neoclassical theory of the firm, so that the
firm is a solution to market imperfections, which also impact the direc-
tion of trade (Anderson and Marcouiller 1999); hence, observed
patterns of trade and investment reflect the complex interaction of

imperfections and costs in the market for goods and the market for
direct investment (Ekholm 1998b).
As we see at first glance, these results can also be interpreted in terms of a

change of paradigm, with the neoclassical arguments becoming a special
case of a knowledge-oriented theory of the transnational corporation.
Most of the contributions are useful and significant independently of the
general framework of equilibrium theory, so that the naive question can be
raised whether it is indeed worth the intellectual effort to fit all this into the
Procrustean bed of pure theory. This is the more justified, if we consider
the very intricate methodological issues involved in the empirical testing of
the models; validity is frequently more dependent on econometric method-
ology than on sound construction of hypotheses (Leamer 1992). Hence, the
door is wide open for entering the evolutionary alternative on the scene.
1.2 Opening the Way for a Change of Paradigm in the Theory

of International Trade
On the theoretical level, evolutionary trade theory (ETT) so far emphasises

the impact of technology on trade and the role of institutions in the broad-
est meaning of the term. However, as we have seen, there is no direct con-
tradiction between a technology-based reasoning and the most recent NTT
models, because the latter have no difficulties with integrating technology,
either as country-specific productivity differentials in labour in trade models
or in terms of introducing a special factor component akin to human
capital, i.e. a variant of a ‘knowledge-stock’, in the theory of the trans-
national firm. In both cases, the ultimate causes of the patterns of trade
remain outside the model, however, the mechanics of trade still obey the
neoclassical laws. That is, as long as ETT does not directly question the equi-
librium framework, even its own empirical results may be easily absorbed
by the latter, thereby being re-interpreted as supporting statements of NTT.3
At first sight, the same seems to be true for institutional approaches,
which may also leave the neoclassical description of the basic mechanisms
unchanged, simply because there is an exploding literature on neoclassical
institutionalism. However, trying to relate the comparative advantage argu-
ment to institutions can result in a major rethinking of the theory of trade
proper. The question is whether it is sufficient to separate the NTT from the
analysis of the institutional framework, which is then in turn explained via
a neoclassical theory, or whether precisely the attempt at establishing an
encompassing framework of both the market and its institutions will lead
to a crumbling of the theoretical building blocks.
Obviously, transactions in international trade face the same and even

more difficult practical problems as all other kinds of transactions, which
arise from legal uncertainty, informational asymmetries and so forth. High
transaction costs might induce the agents to discover particular institu-
tional solutions, such as for example substituting firms for trade. However,
since all these solutions are related to common characteristics of both
exporter and importer, it seems to be difficult to reconcile this approach
with the comparative-advantage argument in terms of factor endowments,
which by definition are exclusive properties of one of the agents. It makes
no sense to talk about the relatively ‘larger’ endowment of one country with
institutions, as far as the common transactions are concerned: the real issue
is the common institutional framework, or the common capability to
co-ordinate transactions successfully. That means, even if the explanation
of the institutions rests upon the neoclassical presumptions, difficulties
arise from the fact that the impact of institutions on trade falls back on a
relational property that cannot be ascribed to either of the sides exclusively.
The only way to think about this is to include comparative advantages of
pairs of traders as compared to other traders, which, however, explodes the
framework of relative characteristics of sets of countries. This difficulty
reveals the astonishing fact that trade theory does not explain the transac-
tions of trade proper, which are simply given (Schmidtchen 1984).
There is one important mainstream discourse where these issues gained
prominence recently, that is the social-capital debate (for a survey, see
Dasgupta and Serageldin 2000). Social capital of countries could be con-
ceived as another factor of production, as far as the effects of internal
transaction costs on comparative costs are concerned. But social capital of
country A may be completely insignificant for enabling the traders to trade
with country B, unless they are able to build new common social capital
with traders of this country. This new social capital cannot be simply added
to the social capital of country A relative to the endowment of other coun-
tries, because it might be insignificant there. This relational property of
social capital may also become salient when considering the nonsense argu-
ment that countries richly endowed with social capital should specialise in
exporting goods with an intensive use of social capital. Although this seems
partly reasonable when thinking, for instance, about the importance of
social capital for the institutional framework of the production of complex
technologies, this would immediately imply that countries with less social
capital would stay in this state, specialising in ‘lawlessness’. Obviously, lack
of social capital is an absolute disadvantage, which might render all other
relative advantages useless under any circumstances.
This observation on social capital opens our eyes to a fundamental logical
problem of NTT: the basic equilibrium premise in factor-proportions
theory is that the impact of factor endowments on trade works via the
pricing of factors which in turn affects goods prices, eventually closing the
total equilibrium system. This raises the question whether the capabilities
of traders to trade can also be treated as a part of that endowment, given
the fact that capabilities to trade can only be measured relative to pairs of
traders who realise the transaction. This is tantamount to asking whether a
relational property of the traders can be priced within the system which
comprises that trading relationship, i.e. the market.
The point of departure in the recent literature is the problem of how to dis-
tinguish between production cost and transaction cost, because the theory
of international trade is exclusively based on the concept of comparative
production-cost advantages. We adopt the conceptual distinction between
transformation cost and transaction cost introduced by North (1990,
pp. 27ff.) who defined the latter as adding up to total production cost. North
himself identified the problem that it is very difficult to distinguish neatly
between both categories on the theoretical level as well as (and even more so)
on the empirical level. The foremost example are the costs associated with
measurement problems in transactions where goods characteristics might
give rise to transaction costs which could be lowered via an appropriate
redesign of the production process, thereby changing transformation costs.
Looking at the problem from the opposite angle, Langlois and Foss (1999)
complain about the neglect of production cost – transformation costs in
North’s terms – in the modern theory of the firm based on transaction cost.
If the concept of production is rooted in the general concept of putting
knowledge to use, there are co-ordination problems in the production
sphere that have nothing in common with the behavioural co-ordination
problems emphasised by transaction-cost economics, namely opportunism.
Therefore, the boundary between the market and the firm cannot be fully
explained by transaction costs.
Hence, our problem can be restated in the straightforward question,
whether the theory of comparative advantage can include production costs
in terms of the sum of transformation costs and transaction costs, which
would imply that the capability to trade is a part of general comparative
advantage.
2. THE LOGICAL INCONSISTENCY

OF PRICING THE CAPABILITY TO TRADE
IN AN EQUILIBRIUM TRADE THEORY
Within the neoclassical framework, different capabilities to trade can be

treated as being reflected in different transaction costs of the transactions
which are realised by different agents. A gifted trader will realise the trans-
actions at lower cost, thereby making higher profit and possibly appropri-
ating the transaction potential of other, less efficient traders (Wegehenkel
1981). Starting out from this, could we think of that capability in terms of
a ‘factor’ in the sense of neoclassical trade theory?
Structurally, to pose the issue in this way is homogeneous with the
general Gödel problem in the foundation of formal systems, related to the
question whether true statements in a formal language can be decided
within that system. This proved to be impossible because of the self-refer-
ential structure so that decidability is only possible in new, higher formal
language where we can talk about the propositions of the former language.
In the same vein, we ask whether pricing of the transactions forming a price
is possible in the market, with ‘pricing of the transactions’ meaning that the
underlying institutional quality of the transaction is priced, or the capabil-
ities of the agents to realise the transaction. This is the same as asking
whether capabilities to trade can be traded in the market system which is
formed by just these trading relations.4
In NTT, this problem never arises because equilibrium theory presup-
poses the successful arrangement of the entire set of transactions outside
the system. Traders in NTT act autistically. That this assumption is by no
means innocuous becomes immediately obvious when we realise that the
inclusion of the capabilities of traders to trade into the equilibrium system
leads to severe inconsistencies.
2.1 Non-tradability as a Necessary Condition for Competitive Advantage
Our starting point is the common distinction between comparative and

competitive advantage. At any rate, the common ground of NTT and ETT
is that both claim to analyse a competitive economy. We elaborate on the
simple but fundamental idea that competitive advantage can only rest upon
factors that remain exogenous to the market system in the sense of being
non-tradable,5 because otherwise these factors might easily be transferred
(imitated, appropriated, purchased etc.) among the competitors, therefore
leading toward their erosion.6 The common example in the evolutionary
setting is technology. Competitive advantages in technology would easily
erode if the technology can be imitated easily and would therefore diffuse
rapidly across all competitors. Apart from the patent system, the mainte-
nance of a competitive lead therefore presupposes the dominant role of
technological knowledge that cannot be imitated due to inherent causes, as
for example, tacit knowledge in applying the technology. This by definition
cannot be traded via the market, and hence cannot be directly priced inde-
pendently of the entire set of capabilities that make up the technology.
Therefore, it remains exogenous to the market in a fundamental way. The

structural evolution of the system of competitive markets and firms will be
determined by the degree of separability of the different constituent units
of the technology and their respective degree of tradability.
Now we can turn this argument upside down and consider the question
whether in the long run the market system will develop toward the wither-
ing away of those non-tradable factors, just resulting from the ongoing
process of learning and transforming implicit knowledge into explicit.
Obviously, this would be possible if there were no novelty. However, the
very workings of a competitive market system will foster the re-emergence
of new non-tradable entities because only these confer competitive advan-
tages. Whether this happens either unintentionally or by intention does not
matter at all. That is, competition and non-tradability, or non-pricibility,
are necessary correlates.
As we realise immediately, this is tantamount to a variation-and-selec-
tion approach to competitive advantage: Since only non-tradable factors
confer sustainable competitive advantages, any structural and behavioural
variation leading to the emergence of such advantages will resurrect the
strong impact of non-tradable properties on the market system. Hence,
the market system itself will always trigger the emergence of forces outside
the market driving its own evolution, constantly ‘telescoping’ minor varia-
tions into larger differential pattern formation. As long as these new ele-
ments are non-tradable, they cannot be priced and, therefore, do not form
an element in the equilibrium system working according to the compara-
tive advantages of agents and countries. That means, if novelty occurs in
the system, competition itself produces the incapability of the system to
endogenise its driving force, or, the market cannot price the ultimate causes
of competitive advantage. Comparative advantage refers to a system where
all information has been dispersed across all agents, and where all relevant
factors of competitiveness are priced, whereas competitive advantage
emerges in a system with exclusive knowledge of single agents, and with
essential determinants of competitiveness bearing no price-tag.
Subsequently, I will show that this argument can be made in more
detailed and specific terms when analysing the transaction-cost approach
as a possible-venue toward pricing the quality of a trading relationship. By
this means I wish to show that there is a fundamental distinction between
a New Institutional Economics approach to trade and an evolutionary one,
and that the problem of making transactions possible necessarily explodes
any equilibrium approach to trade. The standard argument to be pondered
here is that in equilibrium the difference between the equilibrium level of
transaction costs and the disequilibrium level reflects the market-priced
value of the efforts of the traders to realise the transaction. The crucial
question therefore is what happens in disequilibrium. In disequilibrium,

trading opportunities might exist which still await discovery, hence the dis-
tinction between tradable and non-tradable goods remains endogenous to
the system. To understand sustainable competitive advantages, we must
therefore understand the mechanism leading the traders to discover other,
more efficient ways to trade.
This is important for the analysis of comparative advantages related to
production costs, because there is obviously an interaction between trans-
formation costs and transaction costs. For example, high transaction
costs might prevent traders from realising the transaction with traders
operating with relatively lower transformation costs, however, production-
cost advantages might still prevail for the entire set of activities. The ques-
tion is, therefore, whether the market process allows traders to compare
these two kinds of costs. This step presupposes a separation between trans-
action costs and transformation costs in the traders’ profit calculus.
Here we can apply Buchanan’s (1969) classical treatment of opportunity
costs, since the opportunity costs of realising a certain transaction are the
lost efficiency gains of the possible, yet unaccomplished alternative trans-
action (i.e. way of organising the transaction) (we follow Kiwit’s, 1994,
ideas). As we see, the problem is that the trader evidently can only compare
their selection with the known alternatives, which might or might not
include the most efficient one. This is precisely the meaning of entrepre-
neurship: the opportunity costs of the entrepreneur are the lost, yet
unknown profit opportunities, which, however, might be discovered by
other entrepreneurs who thereby make the opportunity costs visible and
even calculable for the competitors. Competition is about discovering the
value of errors. As Buchanan has therefore argued forcefully, any equation
between opportunity costs and costs is only possible in equilibrium where
all profit opportunities have been uncovered and exploited, and where there
is no longer any difference in the state of knowledge across the agents.7
From this we may conclude that observed patterns of comparative advan-
tage will reflect a certain state of knowledge in the economy where entre-
preneurs have not yet discovered all the possible ways to transact, and where
these undiscovered opportunities remain outside the system, hence being
non-tradable. Competitive advantages result from the discovery of these
unknown opportunities, and they will only be sustainable as long as they
continue to be non-tradable. This, of course, seems to be a tautological state-
ment, yet this statement supports the crucial insight that entrepreneurial
capabilities cannot be priced via the market process. Following Buchanan,
pricing can only be possible in equilibrium, where, however, the entrepre-
neurial function has lost any significance. Hence, the question emerges
whether such a state can be imagined as arising from an evolutionary market
process. This would, at least, presuppose that observers of entrepreneurial

action are able to distinguish exactly between the entrepreneurial input and
all other inputs.
2.2 The Fundamental Impossibility Theorem on a Unified Price System

for Transactions and Transformations
The final sentence in the last paragraph means that we would still assume
that the separation between transformation and transaction is viable, so
that in equilibrium traders can optimise the mix of transaction and trans-
formation costs. We cast serious doubts on that. Obviously, the separation
between transformation and transaction is a transaction in turn. Yet, this
new transaction gives rise to transaction costs, too. The difficulties in sep-
arating analytically and empirically between transformation costs and
transaction costs reflect the transaction costs of just that operation in the
real market system. We cannot separate these as observers because the
market does not give rise to such a separation.
Obviously, here we are running into the trap of infinite regress. I do not
wish to overstretch that point, but this kind of infinite regress is sufficient
to conclude that there can be no empirically exact way to measure trans-
action costs unless the separation already exists in the market, so that there
is a price-tag attached to the services needed to realise the transaction. In
particular, we are not able to assess the relative role of the transaction
costs of the original transaction and the transaction costs preventing the
separation between transaction and transformation. That means, for
example, that the boundary between firm and market is indeterminate in
observational terms, because we cannot identify exactly the differential
impact of the relative transaction costs within the firm and across
the boundary as compared to the transaction costs linked with the oper-
ation of just setting that boundary. Even if we were to believe that there is
a clear relation working in reality, we would never be able to observe it
unequivocally.
This infinite regress could never be solved if there is a type of transaction
that cannot be separated from transformation in principle. In order to find
such a kind of transaction, we can start with the most simple case of selling
a good. If we leave legal issues aside, the transaction is only complete if the
buyer is indeed able to use it. That means, the good actually consists of a
good and a service, namely the good proper and the service that leads to
the accomplishment of the transaction.8 This amalgam can be dissected in
many ways, as reality demonstrates, displaying a manifold of arrange-
ments, for example, with the separation between logistics, insurance, retail
activities and other services from the trading transaction proper. Are there
any services complementary to the physical features of the good that

cannot be separated from the transfer of the good?
Obviously, this depends on whether there are entrepreneurs who invent
a method of separating the service and who are able to sell it profitably. At
first sight, again we are back with the problem of infinite regress because
selling the service includes the service of organising that transaction, which
is by no means a minor issue (for example, the informational asymmetries
in both transactions are necessarily different). On a second look we can
conclude that the crucial property involved in that process is tradability. In
that regard Casson (1997, pp. 81ff., 108ff.) argues that the core property of
the firm lies in the fact that the entrepreneur owns a resource that cannot
be traded, hence if the service generated from using that resource cannot
be separated from the good produced by the entrepreneur, the only viable
solution is the establishment of a firm.9 The most simple unit of a firm
therefore is a dyad of one act of production and one related entrepreneur-
ial resource that cannot be traded, which, in our Northian terminology, is
the same as saying that a minimal firm consists of a transformation and
a complementary transaction, with the latter being inseparable and, hence,
non-tradable.
There are many possible entrepreneurial resources that cannot be traded,
yet there seems to be one common property they can be reduced to: this is
the utilisation of contingent, singular and mostly tacit knowledge of the
entrepreneur (cf. Hallwood 1997). There is no price aside from the price set
by the owner, i.e. the entrepreneur, for the peculiar good they offer. Nobody
aside from the owner is able to assess its value which is determined by
the subjective opportunity costs of the entrepreneur. Why is this kind of
knowledge not tradable, given that at least the entrepreneur might set a
price? Since in order to be tradable, buyers must be able to know in advance
the possible uses of that knowledge. But if they already knew the possible
uses, they would no longer need to buy the service but simply act them-
selves. Hence the capability to transact would only become an object of
market transactions if there were a minimum degree of knowledge about
its uses and consequences on the part of the buyer. But precisely that con-
dition would render the service superfluous. Hence, the special core capa-
bility to transact must be a non-tradable.
There are many additional arguments supporting this general proof of
non-tradability of entrepreneurial core capabilities to transact. Suffice it to
mention, first, the relevance of Akerlof’s ‘lemons’ story, given the fact that
we think about a state of asymmetric information between seller and buyer.
The entrepreneur offering the separate service of realising the transaction
would always reap higher profit if he realises the business himself, so that
finally a market for that service would collapse. Since possible buyers
estimate the entrepreneur’s capability according to some average trader’s

type on the market, true entrepreneurs would decline the buyer’s price offer
because they know better. Another approach may be to ask whether there
is insurance possible for entrepreneurial failure. Obviously, if knowledge
underlying the entrepreneurial action is unique, there can be no insurance
based on statistical risk and hence no market valuation of entrepreneurial
risk, notwithstanding that other entrepreneurs might offer an insurance
based on their own singular knowledge.
Leaving those more specific and detailed arguments aside, the essential
point is the time asymmetry involved in knowing how to transact and
knowing about knowing how to transact (i.e. knowing that).10 On the part
of the hypothetical buyer, there must be fundamental ignorance in terms of
the latter if the respective service is of interest for them, because knowing
about knowing how means knowing how by knowing that. Yet, if that fun-
damental ignorance prevails, no market transactions can occur, so that the
capability to transact is a non-tradable.
All these arguments lead us to conclude that the market system cannot
price the capability to transact, given a state of incomplete and imperfect
knowledge about the entire set of possible transactions.11 Therefore, if
competitive advantages rest upon a peculiar mix of transformation and
transaction costs, they can never be fully measurable within a market
system. We have now demonstrated that this is also true for a hypothetical
system where the division of labour would be based on comparative advan-
tages, unless all the transactions had already been arranged outside the
system, like, for example, by the Walrasian auctioneer or an omniscient
social planner.
This result has direct consequences for the issue of spatial directedness of
international trade. If we cannot separate the transaction competences and
the related services from the transaction proper, we can never unequivocally
reduce the observed transactions on the relative endowments of traders dis-
tributed in space, so that neither can we explain the flows via the relative
prices of these endowments. There remains always one component that is
not priced, although exerting influence on trade-flows. By implication, the
observed spatial patterns of trade can never be explained completely by ref-
erence to the distribution of pricible factors in space. Hence, the observed
violations of the factor proportions theorem are very plausible empirical
phenomena, or, the many results of the fundamental role of singular deter-
minants of trade. Similarly, our argument shows that non-separability of
transaction and transformaton is the most basic reason why firms exist,
which implies that in certain transactions the firm itself must be moved in
space to make transactions possible. This is, in principle, the explanation
of the rise of the role of the transnational corporation in international
trade, and, at the same time, of the difficulties of fitting this phenomenon
into the common factor-proportions-framework.
3. A RESTATEMENT OF THE THEORY

OF COMPETITIVE ADVANTAGE BASED
ON ENDOWMENTS WITH NON-TRADABLE
FACTORS
Now we can turn to the positive consequences of our abstract reasoning.

A decade ago, Michael Porter (1990, pp. 73ff.) developed the argument that
competitive advantage cannot rest on the standard factors of comparative
advantage unless they manifest the fundamental property of being singu-
lar, non-imitable and immobile. This is directly related to our theoretical
point on the non-separability and hence non-tradability of entrepreneurial
competences to transact in the market system. When we accept this idea,
we reach the conclusion that there is a need to fundamentally restructure
the common concept of factor that underlies by far the largest part of
mainstream trade theory.12
3.1 A New Definition of ‘Factor’
For reasons of space, I cannot go into the details. Suffice it to say that the
concept of factor belongs to the most elusive categories of economic rea-
soning, because its use results in the construction of highly aggregate
entities that produce ‘factor services’ which are the real substance of inter-
national trade, beyond the surface of the traded goods. If we take the stand-
ard Heckscher-Ohlin definition as a starting point, factors are defined as
being internationally immobile factors of production. Adding related
definitions in the theory of growth, this rather broad definition may be more
circumscribed if we add the property of non-producibility (cf. Woodland
1987).
All these definitions, of course, are highly contentious because in many
uses of the factor concept other, often contradicting definitions inhere. For
example, capital continues to be treated as ‘factor’ even though it is regarded
to be perfectly mobile. Consequently, there is a plethora of possible trade
models including, for example, single-factor models if the criterion of
immobility is applied strictly (see, for example, ‘skills only’ models à la Wood
1994), and, on the other extreme, multifactor models which, for example,
analyse the impact of sector-specific factors on trade. However, precisely
this looseness of definiton results in serious troubles when trying to sep-
arate ‘factors’ from ‘goods’ analytically, if the latter concept encompasses
both final and intermediate goods. This is especially intricate when

considering the distinction between ‘producer services’ in general and
‘factors services’.
Furthermore, another conceptual pair becomes relevant here, that is, the
distinction between stocks and flows. Obviously, the idea is close to sense-
less to assume that intermediate goods change their essence when a part of
a stock, so that ‘capital’ would be a different thing when being accumulated
and stored as compared to capital being traded. On the other hand,
‘services’ are frequently treated as being non-storable which implies that
no stocks can be built up. Hence, the entire semantic field seems to be in a
confused state.
The concept of a stock presupposes the property of storability. If a
certain good or service cannot be stored, there is no way to accumulate it.
For example, labour in the theory of growth cannot be accumulated per se,
because it cannot be stored, so that it is a pure flow. On the other hand, accu-
mulation presupposes the concept of producibility (at least beyond the
hunter-gatherer economy), which is a more exact concept than repro-
ducibility. Producibility is related to intentional human action, whereas
reproduction can also result from other natural processes. In that sense, pro-
ducibility is related to the exogenous/endogenous distinction. Producibility
is a property that cannot be related to entities which are in principle
exogenous to the system and which therefore cannot be influenced by
human action at all. Reproducibility can depend on either exogenous or
endogenous processes. For example, climate is decisive for the reproductive
use of agricultural land and still remains exogenous to the system, i.e. is
non-producible, whereas human capital can be regarded as producible
today.
If an entity cannot be produced, then the question arises whether the use
of it leads to its final consumption. In the classical uses of ‘nature’ as a
factor there is the implicit idea that an eternal reproduction takes place, so
that the factor is never consumed. Independent of this idea there is a stand-
ard way to assume eternal usability of an entity, namely that the uses are
conceived as non-rival in time. This is precisely the property ascribed to
knowledge in the most recent theories of trade, especially in the theory of
the multinational firm, where the knowledge-stock is treated as a pure
public good, whose services, however, can be appropriated within the firm
(the related argument is made in the context of country-specific knowledge-
bases). Hence, if we regard non-consumability as a necessary property
of sustainable competitive advantages, this can be cast into the positive
characterisation of non-rival use in time.
Non-producibility entails another twist of the argument if we con-
sider the special role of intention in our definition. We can surmise that
non-producibility refers to all unintended results of human action, that is

to all external effects which are either unobserved or cannot be avoided by
intentional action. Of course, this shows that non-producibility is eventu-
ally no property that can be given an empirical meaning a priori, because
it depends on the state of knowledge, technology and of the institutional
framework. We realise that our ideas begin to close the circle, because unin-
tended consequences are also called ‘externalities’. By definition, external-
ities are phenomena that impact the outcome of the market system, yet
cannot be priced via market transactions.
These seemingly scholastic reasonings achieve significance if we link
them to the general idea of sustainable competitive advantages. Evidently,
if we search for factors that provide the basis for such advantages, this is
immediately the case if the relevant entity is non-producible and immobile.
This, for example, is the property of land as the most ‘classical’ factor in
international trade because land cannot be moved in space and could only
to a very limited extent be produced in those early times of economic the-
orising. Non-producibility is the basic precondition of non-imitability,
because imitation presupposes the capability to produce and generate the
respective property intentionally. At the same time, the relevant entity
should be storable, so that the distinction between ‘stocks’ and ‘flows’ can
be applied meaningfully. Finally, the entity should be outside the system in
the sense that there is no way to attach a separate price-tag to it, i.e. in our
terminology, it is non-tradable.
Drawing all these arguments together, we reach a new and cogent
definition of a factor.
A factor is an aggregate stock with the properties of non-producibility, stor-
ability and non-tradability, as well as non-rivalness of use in time. Factor uses
result in a flow of factor services that inhere in trade in goods.
If we link this definition with our discussion of competitive advantage,
we can start our search for factors with the question of which kind of
service is the prime source of these advantages. As we have seen, the prim-
ordial service is the entrepreneurial making of transactions.
3.2 Three Dimensions of Entrepreneurial Competence as Factor

Generating Entrepreneurial Action as Complementary Service
to Goods
Which kind of entity might be the stock out of which entrepreneurial ser-
vices flow? The most simple answer, of course, would point toward the
entrepreneurs as a group. However, if we wish to lend more detail to the
analysis of the sources of entrepreneurial capabilities, I propose to treat as
‘factors’ the following three entities, the services of which constitute the
flow of entrepreneurial activity that is inseparably linked with the transac-

tions of trade:
● organisational capital
● collective human capital
● social capital.
All these three entities have one structural feature in common, namely the
fact that relational patterns in neighbourhoods are involved, with neigh-
bourhoods frequently, but not necessarily, referring to spatially close inter-
actions (closeness can result from ‘similarity’ in any other respect, for
example, kinship). This idea refers to certain arrangements of people in
space and time, which are stable and recurrent, and which because of their
complexity cannot be straightforwardly imitated (‘produced’) and trans-
ferred (‘traded’) to another place. This basic insight has already been
achieved, amongst others, by Michael Storper (1995), who argued that
competitive advantages of locations are rooted in non-traded interdepend-
encies between people at certain places. Our concept broadens this insight
to the more abstract structural aspects of social relations between agents,
which are regarded as ‘capital’ that generates entrepreneurial services.13
This capital is the non-intended result of human action, therefore is non-
producible, yet storable, because of its recurrent unintended reproduction,
and it also has the property of non-rivalness of use in time.
Organisational capital is the crystallisation of the competences of the
firm, that are rooted in certain organisational routines, tacit knowledge,
and non-tradable assets which are only reflected in the market value of the
firm as compared to the market price of its constituents. Although this is a
patterned neighbourhood, obviously the firm as a whole can be traded and
transferred in space. However, this presupposes very special competences,
too, as is evident from the many failures, for example, of mergers, acquisi-
tions and foreign direct investment. Hence we may speculate that one of the
crucial ingredients of organisational capital of the firm is the competence
to trade itself, namely to grow into other markets and regions in space-time.
Firms, hence, differ in terms of this special competence which can therefore
be regarded as especially relevant for international trade.
The concept of organisational capital provides a new foundation for the
theory of the multinational corporation, because we regard the stock of
firms in an economy as a base of its competitive advantage in trade. This
is directly mirrored in the fact that trade and FDI are closely enmeshed,
and that intra-firm and intra-industry trade dominate trade the more devel-
oped the nations involved, hence the more complex and diversified their
stock of firms becomes. The country- or location-specific accumulation of
organisational capital is therefore one of the determinants of directedness

of trade, and the patterns depend on the special capabilities to transfer
organisational capital in space, which is a recursive property of organisa-
tional capital. At the same time, that means that the firm drives a logical
wedge into the market system, as we have demonstrated in section 2. From
this point of view, the violations of the factor-proportions theorem are
natural reflections of the existence and the operations of the firm in the
international economy.
Collective human capital is a set of competences and knowledge that
inheres in networks of agents and not the agents proper (cf. Foss 1999).
This, of course, also applies to organisational capital, but we try to distin-
guish between human capital that is intentionally organised in the firm and
another kind that arises from spontaneous interaction both within and
beyond the firm. The latter is most salient from the study of industrial dis-
tricts and other forms of clustering of economic activity (compare Loasby
1999). In other words, collective human capital rests upon stable patterns
of spillovers among agents putting their knowledge into use. Stability of
patterns presupposes certain social structures which are maintained inde-
pendently of the specific transaction. This has been called ‘embeddedness’
in other theoretical contexts. Obviously, these patterns may be highly
specific and cannot be directly imitated, because they go back, for example,
to historical constellations of social change, or to properties of agents
which are not easily imitable, such as religious affiliation. The most salient
example is the specific role of ‘settings’ for certain service industries like
entertainment and advertisement (for a penetrating case-study, see Grabher
2001), yet similar considerations may be valid for financial centres. These
places are sources of directed flows of services in the world economy, pre-
cisely because the complex interaction between lifestyles, professional com-
munication and local organisation of complementary services render those
places unique. Traders are attracted to these places even though the costs
of entering the hypercompetitive environment are rapidly increasing, yet
still being outweighed by the very strong effects on labour productivity.
Collective human capital, therefore, stays in the background of the NTT
observation that local level-effects play an important role in interregional
trade.
Social capital is a property of social networks that is reflected in the
effects of transactions on the successful realisation of other transactions,
that is, it emerges from chains and networks of transactions that are being
perceived and interpreted by agents (Herrmann-Pillath 2000a). Among our
three dimensions, social capital most directly refers to the interaction
between the two sides of a transaction, hence it is a relational category.
However, our concept transcends the simple idea, for example, that both
sides belong to the same ethnic network and therefore can lower transac-
tions costs because of higher levels of trust (e.g. Landa 1994). Social capital
may also emerge from social structures where indirect channels of commu-
nication and social perception are existent, possibly restricting deviant
behaviour even without a closer relation between the agents proper.
Furthermore, if externalities across transactions are taken into consider-
ation, social capital may be an emergent property of networked transac-
tions, which, however, shows a distinct pattern in space. Those patterns can
be self-reinforcing, because there are direct effects on observed transactions
costs, so that the externalities are eventually internalised in stable patterns
of repetitive transactions.
As we see, all these three forms of capital are to a large extent non-
producible in the sense that there is no direct relation between input, output
and intentional action. The organisational capital of the firm, for example,
is based on tacit knowledge which arises from the accumulation of infor-
mal routines. Since this cannot be imitated easily, it provides the founda-
tion of competitive advantages. The same is true for a singular network
pattern forming social capital, like particular social relations among
traders which cannot be simply rebuilt by other persons who do not have
the same personal history.
In what sense are the three kinds of factors related to the phenomenon
of directedness? This becomes clear if we distinguish three dimensions of
distance that impact on transactions, and if we assign one of the factors to
the special capability to overcome the respective distance.
● Spatial distance can be overcome (amongst other ways) via the use of
organisational capital that enables agents either to shift their focus of
activity (direct investment) and/or supports trade relations. Hence we
expect the pattern of trade to be strongly influenced by the direction
of investment flows and different corporate strategies of market
entry. The most salient empirical phenomenon is the growth of
intrafirm trade, however, one should also pay attention to network-
ing across firms, like strategic alliances.
● Technological distance can be overcome via collective human capital.
Here we refer to tacit differences in the use of technologies across
locations which imply that trade (especially in intermediate prod-
ucts) will be influenced by these indirect characteristics of space.
Technological distance can be measured by comparing absorptive
capacities across locations, if this is not simply reduced to ‘stocks’ of
human capital, but is related to qualitative and structural differences
between knowledge systems. An important empirical case in point are
national systems of innovation which foster the emergence of localised
technological trajectories. These trajectories influence the patterns of

trade, for example, via the technological networks implied by the flows
of intermediate products.
● Interpersonal distance is overcome via the use of social capital. This
refers to all kinds of unfamiliarity with the social, legal and institu-
tional environment across locations. The most important empirical
example is cultural distinctness, which may induce low levels of trust
because of possible misunderstandings. However, social capital
allows agents to overcome the impediments caused by cultural dis-
tinctness because highly personal relations are developed which
include an idiosyncratic environment of the transactions.
If we look back on the list of unresolved issues of NTT, we immediately

realise the relevance of our approach. In the most abstract terms, our analy-
sis of factors results in the insight that the directedness of trade can only be
explained within a framework that integrates the impact of absolute and
relative advantages, the latter working via the price system, and the former
working via the singular impact of factors on trade. In empirical terms, these
become evident in the strong influence of historical determinants, culture or
corporate structures on trade. This by no means precludes the role of prices
and relative endowments in trade, however, we expect a complex interaction
of the different determinants, just as is being actually observed.
One example of the move of mainstream analysis in this direction is the
recent research by Feenstra and associates (1997, 1999) into the role of
networks in East Asian trade. They demonstrate how different kinds of
networks in Korea, Taiwan and Japan result in different patterns of com-
petitiveness that are reflected in the product composition of their respective
exports, because organisation of the markets for intermediate products
affects price competitiveness and range of qualities. In our terms, the
differences between networks in these countries can be easily described as
differences in the structure of organisational capital and different mixes
between organisational and social capital. Hence, there is much room for
an integration between mainstream and evolutionary arguments, provided
that the claim is given up of achieving a complete and full equilibrium
system of international trade.
4. PROSPECTS FOR AN EVOLUTIONARY THEORY

OF INTERNATIONAL TRADE
We have tried to set up a general and abstract framework for ETT. The main
achievement seems to be that we do not reject the empirical insights
obtained from NTT, yet put them into a completely different theoretical
approach. We also accept one feature of the NTT approach to trade, that
is, to analyse trade-flows as surface phenomena of deeper structures. These
deep structures refer to properties of locations in space, conventionally
treated as ‘factor endowments’ which generate certain flows of services.
These services we interpret either as factor content or as inseparable, com-
plementary services enabling agents to trade. Our main hypothesis, then, is
that competitive advantages rely on entrepreneurial inputs that are outside
the market system, and which generate the factor contents and services.
More specifically, these are related to technology and transactions.
In the most general terms, I wish to pinpoint the difference between ETT
and NTT by means of Figure 7.1. The theory of comparative advantage
explains trade under the condition of complete knowledge, with the price
system endogenising all relevant determinants of trade. The evolutionary
theory of competitive advantage explains trade as a process generating
knowledge about comparative advantage. This difference is easily illus-
trated in Figure 7.1.
Circle A–E depicts the circular causation of NTT, which is a closed equi-
librium system starting out from points A and B, with the price system
directly revealing the complete information about comparative advantages
in production of different locations. Via the trading activities, which are
basically arbitrage processes, countries are led to further specialise in their
Comparative advantage
of production
Adaptation Opportunity costs as
A reflected in the price
of production
E B system
structures e
d
f Discovery of
C arbitrage
Knowledge about c g opportunities
comparative
b a D
advantage
Realised transactions
of trade
New information
revealed by trade
Figure 7.1 Competing patterns of causation in trade theory

comparative advantages, so that in equilibrium optimal production struc-

tures are realised. In ETT, there is a wider loop a–g starting at a. Here,
traders firstly realise transactions that generate information about com-
parative advantages. That is, the adaptation of the production structure
is contingent on the determinants of the transactions.
As we have seen, these determinants lie outside the market system. We
have identified organisational, collective human and social capital as
theoretical aggregates by which we may grasp and systematise these deter-
minants. If we look at these in more abstract terms, we may imagine trade-
flows to follow invisible networks of relations between agents which are
rooted in their endowments with the three kinds of capital. These networks
become observable via the impact of costs of transactions and costs of
co-ordinating knowledge among the agents. This vision is closely related to
recent efforts of evolutionary theorists to reformulate the theory of the
market as a theory of open, non-integral systems (foremost, Potts 2000).
Methodologically, the research strategy of trade theory would move toward
the integration of sociology and economics, which is already salient in the
recent upsurge of interest in business networks and other forms of non-
price co-ordination in international trade (see the survey by Rauch 2001).
I think that ETT can achieve a similar role as theoretical pinnacle to that
which NTT plays with reference to equilibrium theory. Whether there are
important policy conclusions beyond the free-trade paradigm, cannot be
answered here and remains a task for future work.
NOTES
1. This chapter is based on my German book Herrmann-Pillath (2001a), presenting a con-
densed version of some crucial theoretical arguments. Related papers in English are
Herrmann-Pillath (2000b), (2001b).
2. To get an impression, the reader is recommended to visit the homepage of the ‘Academic
Consortium on International Trade’ (www.spp.umich.edu/rsie/acit/), which reflects
the public role of trade economists. The intellectual esprit de corps is reflected in the
fact that there is even a ‘family tree’ of trade economists available in the internet, see
www.econ.lsa.umich.edu/~alandear/tree/index.htm
3. I should emphasise that I regard New Trade Theory as a part of the mainstream
approach, since it sticks to the equilibrium framework, though achieving, at least partly,
an endogenisation of technological change, as in Grossman and Helpman (1991).
4. In economics, Hayek has made the similar point when arguing (1952, see Herrmann-
Pillath 1992) that a system with a certain degree of complexity cannot understand itself
except via a system of higher complexity that allows one to reproduce the system of
lower complexity as well as including at least the additional self-referring relation.
Loasby (2001) also stresses the importance of this fundamental insight for constructing
economic theories.
5. Our use of ‘tradability’, ‘non-tradable’ and related expressions differs a bit from the
common use because we refer to the general property of a good being transferable via
market transactions, both internal and external. In trade theory, these expressions are
normally restricted to transferability via international markets. In our use, a ‘non-

tradable’ is a good that cannot be sold via any market, whereas in the common use it is
a good that can only be sold via internal markets.
6. This is common sense in modern theories of the competitiveness of firms, as in resource-
based approaches (e.g. Barney 1991) or competence-based theories (Foss 1993). I should
add that in the New Trade Theories, there is also a crucial role of non-tradables in the
sense of public goods and externalities, which by definition cannot be priced. In these
models, like Grossman and Helpman’s (1991) now-classic one, sustainable advantages
arise from externally imposed restrictions on the reach and scope of externalities and the
use of public goods, such as, for example, national borders.
7. There are exact approaches to this problem within an equilibrium setting. One very inter-
esting one is Greenhut et al. (1987, especially chapter 19) who model an imperfectly
competitive economy with endogenous determination of the location of firms (which is
a problem closely related to international trade). They introduce a residual return
remunerating entrepreneurs for coping with uncertainty. This return accrues to ‘non-
functional inputs’, which cannot be measured with conventional accounting procedures
(which, hence, bear no price-tag). Furthermore, they distinguish between ‘differential
rents for managerial skills’ and ‘fixed costs of behavioral uncertainty’. Both kinds of
rents are conceived as elements of the long-run average cost curve, and they are being
determined by the subjective opportunity costs of the entrepreneurs (which is equivalent
to Buchanan’s point of view). In their model, differences in capabilities will be reflected
in different firm sizes, with the zero-profit condition applying. Because of the size
differences, the subjective opportunity costs of the entrepreneurs will converge on a
common equilibrium value, which, however, is not directly observable in the market, but
is only indirectly visible in terms of the size distribution of firms. As we see, if all trans-
action costs result from uncertainty and lack of information, the Greenhut et al.
approach can be applied straightforwardly to demonstrate that transaction costs are
in fact imputed rents accruing to differential capabilities to transact.
8. This approach has been proposed by Hirsch (1989) for the analysis of international
trade, yet without reference to transaction costs. Hirsch argues that there is always a
service component in trading goods, if only the service of trading proper. As is well
known, services generally feature a lower degree of international tradability than goods,
hence, the international tradability of goods partly depends on the tradability of the
complementary services.
9. This argument is related to Barzel’s (1989, chapter 4) explanation of the distribution of
ownership rights in the firm. In the efficient state, the residual claimant will become that
person who affects the variability of the outcome most, and for whom this effect can be
measured to the least degree (which, as in Greenhut’s approach mentioned in note 7, is
the same as saying that it has no price-tag). Only the residual claimant has the incentives
to monitor himself, if others cannot monitor him because of the obstacles to
measurement.
10. The argument follows the structure of Streit and Wegner (1992) who demonstrate that
transaction costs are fundamentally different from transport costs because they result
from fundamental uncertainty about the future course of events. If transaction costs
result from incomplete information about the future, an optimisation calculus cannot
apply because of the information paradox. Hence, transactions need to be realised in
order to learn about their costs, and we cannot choose the most efficient one ex ante.
On the importance of the knowing how/knowing that distinction for the theory of the
firm, firstly proposed by the philosopher Ryle, see Grant (1996).
11. This argument can be stated in other terms without changing its substance. For example,
Anderlini and Felli (1998a, b) posit a similar impossibility theorem when analysing the
question whether agents can write complete ex ante contracts in which the costs of real-
ising the contract are also included. Obviously, this is impossible to achieve ex ante,
without having already realised all the activities necessary to launch the transaction.
To conclude a contract, both sides have to incur sunk costs specific to the transaction,
and there is no way to reach a sharing agreement ex ante in which, for example, one side
would share its own search costs with the other side. Hence, in a fundamental way con-
tracts can never be complete.
12. In modern writings on international trade, these definitional issues have been shifted into
the background, with encompassing monographs like Wong (1995) even doing without
any definition of ‘factors’ in general. There were times when this question received a lot
of attention, as, for example, Caves (1967, pp. 93ff.) illustrates. In the early days of for-
malising the equilibrium trade theory, intellectual giants like Samuelson (1948, 181)
raised very serious doubts on the theoretical role of ‘factors’ in the factor-proportions
theory. All this is simply forgotten today.
13. Related ideas have been developed in the most recent foundational research in evolu-
tionary economics, which focuses on the role of networks as embodying knowledge, with
the seminal work by Potts (2000, 2001).
REFERENCES
Anderlini, Luca and Leonardo Felli (1998a), ‘Costly Coasian contracts’,

Discussion Paper no. 362, London School of Economics and Political Science.
Anderlini, Luca and Leonardo Felli (1998b), ‘Costly bargaining and renegotia-
tion’, Discussion Paper no. 361, London School of Economics and Political
Science.
Anderson, James E. and Douglas Marcouiller (1999), ‘Trade, insecurity, and home
bias: an empirical investigation’, NBER Working Paper 7000.
Barney, Jay (1991), ‘Firm resources and sustained competitive advantage’, Journal
of Management, 17 (1), 99–120.
Barzel, Yoram (1989), Economic Analysis of Property Rights, Cambridge:
Bellet, Michel and Corine L’Harmet (eds.) (1999), Industry, Space and Competition,
Bernstein, Jeffrey R. and David E. Weinstein (1998), ‘Do endowments predict the
location of production? Evidence from national and international data’, NBER
Working Paper 6815.
Borchert, Manfred, Ulrich Fehl and Peter Oberender (eds) (1987), Markt und
Wettbewerb. Festschrift für Ernst Heuß zum 65. Geburtstag, Berne/Stuttgart:
Haupt.
Braunerhjelm, Pontus and Karolina Ekholm (eds) (1998), The Geography of
Multinational Firms, Boston, MA: Kluwer.
Braunerhjelm, Pontus and Robert E. Lipsey (1998), ‘Geographical specialization of
US and Swedish FDI activity, in Braunerhjelm and Ekholm (1998), pp. 33–58.
Buchanan, James M. (1969), Cost and Choice. An Inquiry in Economic Theory,
Chicago: University of Chicago Press.
Casson, Mark (1997), Information and Organization. A New Perspective on the
Theory of the Firm, Oxford: Clarendon.
Caves, Richard E. (1967), Trade and Economic Structure. Models and Methods,
Cambridge, MA: Harvard University Press.
Clark, Terry, Masaaki Kotabe and Dan Rajaratnam (1999), ‘Exchange rate pass-
through and international pricing strategy: a conceptual framework and research
propositions’, Journal of International Business Studies, 30 (2), 249–68.
Collins, William J., Kevin H. O’Rourke and Jeffrey G. Williamson (1997), ‘Were
trade and factor mobility substitutes in history?’, NBER Working Paper 6059.
Dasgupta, Partha and Ismail Serageldin (eds) (2000), Social Capital. A Multifaceted
Perspective, Washington, DC: World Bank.
Davis, Donald R. et al. (1997), ‘Using international and Japanese regional data to
determine when the factor abundance theory of trade works’, American
Economic Review, 87 (3), 421–46.
Davis, Donald R. and David R. Weinstein (1998), ‘An account of global factor
trade’, NBER Working Paper 6785.
de la Mothe, John and Gilles Paquet (eds) (1996), Evolutionary Economics and the
New International Political Economy, London/New York: Pinter.
Deardorff, Alan V. (1998), ‘Determinants of bilateral trade: Does gravity work in a
neoclassical world?’, in Frankel (1998), pp. 7–22.
Dosi, Giovanni, Keith Pavitt and Luc Soete (1990), The Economics of Technical
Change and International Trade, New York: Harvester Wheatsheaf.
Dunning, John H. (1995), ‘What’s wrong – and right – with trade theory?’
International Trade Journal, 9 (2), 163–200.
Ekholm, Karolina (1998a), ‘Headquarter services and revealed factor abundance’,
Review of International Economics, 6, 545–53.
Ekholm, Karolina (1998b), ‘Proximity advantages, scale economies, and the loca-
tion of production’, in Braunerhjelm and Ekholm (1998), pp. 59–76.
Engel, Charles and John H. Rogers (1999), ‘Violating the law of one price: should
we make a federal case of it?’, NBER Working Paper 7242.
Feenstra, Robert C., Deng-Shing Huang and Gary Hamilton (1997), ‘Business
groups and trade in East Asia, Part I: Networked equilibria’, NBER Working
Paper 5886.
Feenstra, Robert C., Maria Yang and Gary G. Hamilton (1997), ‘Business groups
and trade in East Asia, Part II: Product variety’, NBER Working Paper 5887.
Feenstra, Robert C., Tzu-Han Yang and Gary G. Hamilton (1999), ‘Business
groups and product variety in trade: evidence from South Korea, Taiwan and
Japan’, Journal of International Economics, 48, 71–100.
Foss, Nicolai J. (1993), ‘Theories of the firm: contractual and competence perspec-
tives’, Journal of Evolutionary Economics, 3, 127–44.
Foss, Nicolai J. (1999), ‘Networks, capabilities, and competitive advantage’,
Scandinavian Journal of Management, 15 (1), 1–15.
Frankel, Jeffrey A. (ed.) (1998), The Regionalization of the World Economy,
Chicago: University of Chicago Press.
Goldberg, Pinelopi Koujianou and Michael M. Knetter (1997), ‘Goods prices and
exchange rates: what have we learned?’, Journal of Economic Literature, 35,
1243–72.
Grabher, Gernot (2001), ‘Ecologies of creativity: the village, the group, and the het-
erarchic organization of the British advertising industry’, Environment and
Planning, 33, 351–74.
Grant, Robert M. (1996), ‘Toward a knowledge-based theory of the firm’, Strategic
Management Journal, 17 (Winter special issue), 109–22.
Greenhut, Melvin L., George Norman and Chao-shun Hung (1987), The Economics
of Imperfect Competition: A Spatial Approach, Cambridge: Cambridge University
Press.
Grossman, Gene M. and Elhanan Helpman (1991), Innovation and Growth in the
Global Economy, Cambridge, MA/London: MIT Press.
Grossman, Gene M. and Kenneth Rogoff (eds.) (1995), Handbook of International
Economics, vol. 3, Amsterdam: Elsevier.
Hallwood, C. Paul (1997), ‘Competencies as private information: an efficient capital

asset pricing theory of the firm’, Journal of Theoretical and Institutional
Economics, vol. 153, 532–44.
Hanson, Gordon H. and Matthew J. Slaughter (1999), ‘The Rybczynski Theorem,
factor-price equalization, and immigration: evidence from U.S. states’, NBER
Working Paper 7074.
Harrigan, James (1997), ‘Technology, factor supplies and international specializa-
tion: testing the neoclassical model’, American Economic Review, 87, 475–94.
Haskel, Jonathan and Holger Wolf (1999), ‘Why does the “law of one price” fail?
A case study’, CPER Discussion Paper 2187.
Hayek, Friedrich August von (1952), The Sensory Order, Chicago: University of
Chicago Press.
Helliwell, John F. (1997), ‘National borders, trade and migration’, NBER Working
Paper 6027.
Helpman, Elhanan (1998), ‘Explaining the structure of foreign trade: where do we
stand?’, Weltwirtschaftliches Archiv, 134 (4), 573–89.
Herrmann-Pillath, Carsten (1992), ‘The brain, its sensory order and the evolution-
ary concept of mind, on Hayek’s contribution to evolutionary epistemology’,
Journal for Social and Biological Structures, 15 (2), 145–87.
Herrmann-Pillath, Carsten (2000a), ‘Culture and observation in the study of eco-
nomic systems,’ Discussion Paper 52, Fakultät für Wirtschaftswissenschaft der
Universität Witten/Herdecke, March 2000.
Herrmann-Pillath, Carsten (2000b), ‘Indeterminacy of international trade:
methodological reflections on the impact of non-economic determinants on the
direction of trade and absolute advantage,’ Aussenwirtschaft, 55 (2), 251–89.
Herrmann-Pillath, Carsten (2001a), Kritik der reinen Theorie des interna-
tionalen Handels, Band 1: Transaktionstheoretische Grundlagen, Marburg:
Metropolis.
Herrmann-Pillath, Carsten (2001b), ‘A general refutation of the law of one price as
empirical hypothesis’, in A. Wagner and H. Strecker (eds), Jahrbücher für
Nationalökonomie und Statistik, 221 (1), pp. 45–67.
Hirsch, Seev (1989), ‘Services and service intensity in international trade’,
Weltwirtschaftliches Archiv, 125, 45–59.
Hummels, David and James Levinsohn (1995), ‘Monopolistic competition and
international trade: reconsidering the evidence’, Quarterly Journal of Economics,
110 (3), 799–836.
Kiwit, Daniel (1994), ‘Zur Leistungsfähigkeit neoklassisch orientierter Transa-
ktionskostenansätze’, ORDO, 45, 105–35.
Klau, Arne Ragnar (1995), Firmeninterner Handel: Eine theoretische und empirische
Analyse des unternehmensinternen Außenhandels multinationaler Unternehmen,
Berlin: VWF.
Krugman, Paul (1995), ‘Increasing returns, imperfect competition and the positive
theory of international trade’, in Grossman and Rogoff (1995), pp. 1243–77.
Landa, Janet T. (1994), Trust, Ethnicity, and Identity. Beyond the New Institutional
Economics of Ethnic Trading Networks, Contract Law, and Gift-Exchange, Ann
Arbor, MI: Michigan University Press .
Langlois, Richard N. and Nicolai J. Foss (1999), ‘Capabilities and governance: the
rebirth of production in the theory of economic organization,’ Kyklos, 52,
201–18.
Leamer, Edward E. (1992), ‘Testing trade theory’, NBER Working Paper 3957.
Loasby, Brian (1999), ‘Industrial districts as knowledge communities’ in M. Bellet

and C.P. Harmet (eds), Industry, Space and Competition, Cheltenham, UK and
Northampton, MA, USA: Edward Elgar, pp. 70–85.
Loasby, Brian (2001) ‘Time, knowledge and evolutionary dynamics: why connec-
tions matter’, Journal of Evolutionary Economics, 11 (4), 393–412.
Markusen, James R. (1995), ‘The boundaries of multinational enterprises and
the theory of international trade’, Journal of Economic Perspectives, 9 (2),
169–89.
Markusen, James R. and Anthony Venables (1998), ‘Multinational firms and the
new trade theory’, Journal of International Economics, 46, 183–203.
Markusen, James R. and Keith E. Maskus (1999), ‘Multinational firms: reconcil-
ing theory and evidence’, NBER Working Paper 7163.
North, Douglass C. (1990), Institutions, Institutional Change, and Economic
Performance, Cambridge: Cambridge University Press.
Porter, Michael E. (1990), The Competitive Advantage of Nations, London:
Macmillan.
Potts, Jason (2000), The New Evolutionary Microeconomics, Cheltenham, UK and
Potts, Jason (2001), ‘Knowledge and markets’, Journal of Evolutionary Economics,
11 (4), 413–32.
Rauch, James E. (2001), ‘Business and Social networks in international trade’,
Journal of Economic Literature, xxxiv, 1177–1203.
Ray, Edward John (1995), ‘Old myths and new realities: foreign direct investment in
the United States’, International Trade Journal, 9 (2), 225–46.
Rogers, John H. and Michael Jenkins (1995), ‘Haircuts or hysteresis? Sources of
movements in international exchange rates’, Journal of International Economics,
38, 339–60.
Samuelson, Paul A. (1948), ‘International trade and the equalisation of factor
prices’, Economic Journal, 58, 163–84.
Schmidtchen, Dieter (1987), ‘Unsichtbare-Hand-Erklärung und die Theorie kom-
parativer Kosten’, in Borchert et al. (1987), pp. 287–309.
Storper, Michael (1995), ‘The resurgence of regional economies, ten years later: the
region as a nexus of untraded interdependencies’, European Urban and Regional
Studies, 2 (3), 191–221.
Streit, Manfred and Gerhard Wegner (1992), ‘Information, transaction, and catal-
laxy – reflections on some key concepts of evolutionary market theory’, in Witt
(1992), pp. 125–49.
Trefler, Daniel (1993), ‘International factor price differences: Leontief was right!’,
Journal of Political Economy, 101 (6), 961–87.
Trefler, Daniel (1995), ‘The case of missing trade and other mysteries’, American
Economic Review, 85 (5), 1029–46.
Wegehenkel, Lothar (1981), Gleichgewicht, Transaktionskosten und Evolution. Eine
Analyse der Koordinierungseffizienz unterschiedlicher Wirtschaftssysteme,
Tübingen: Mohr Siebeck.
Witt, Ulrich (ed.) (1992), Explaining Process and Change: Contributions to
Evolutionary Economics, Ann Arbor, MI: Michigan University Press.
Wong, Kar-yiu (1995), International Trade in Goods and Factor Mobility,
Cambridge, MA/London: MIT Press.
Wood, Adrian (1994), ‘Give Heckscher and Ohlin a chance!’, Weltwirtschaftliches
Archiv, 130, 20–49.
Woodland, A.D. (1987), ‘Entry “tradeable and non-tradable commodities” ’, in

John Eatwell, Murray Milgate and Peter Newman (eds), The New Palgrave.
A Dictionary of Economics, London/New York/Tokyo: Macmillan/Stockton
Press/Maruzen.
Zhang, Kevin H. and James R. Markusen (1997), ‘Vertical multinationals and host-
country characteristics’, NBER Working Paper 6203.
8. Economic policy – a process of
communication
Lambert T. Koch
1. INTRODUCTION
In the attempt to provide an analytical structure for the process of eco-

nomic policy making, the literature on the subject often differentiates
between three basic categories, i.e. determining the status quo, setting aims,
selecting measures.1 Here it is presumed that this three-step process should
programmatically guide investigations into economic policy making and
that it should thus be the original task of a positive theory of economic
policy to pose the question as to the most suitable tools for a target-
performance comparison on an aggregated level when faced with different
types of situations and target constellations. Suitable measures would be
those which lead to a corresponding balance in the most efficient way.
Doubts about this idea, which is part of the core of orthodox theory on
economic policy, are nothing new.2 Here we are dealing with the basic
pattern of a procedure which theorists as far back as Albert already criti-
cised as being instrumentalistic-technocratic (cf. Albert 1967, p. 113), seeing
the danger of false methodological concepts resulting in excessive expecta-
tions of feasibility in economic policy (cf. Homann 1980). An instrumen-
talistic-technocratic programme suggests that the process of economic
policy making can be understood using a machine model as described by
Heinz von Förster (cf. Förster on 1984). This would mean that sufficient
knowledge of an initial situation and of currently valid transformation
rules would be enough to deduce the ‘right’ measures.3 In this way, plan-
ning and forecasting problems in economic policy would to a great extent
be reduced to technical-organisational factors – with correspondingly
beneficial consequences regarding general control possibilities.
Criticism of this kind of understanding of economic policy in the past
few decades goes in many different directions. The difficulties which are
created when knowledge necessary for economic policy is to be centralised
were dealt with (cf. Hayek 1983). The ideas of ‘bounded rationality’ of eco-
nomic agents and the relevance of information costs were introduced,
168
Economic policy 169
which prevent voters and the elected from including all information rele-
vant in a situation in their decisions (cf. Simon, 1955). And from the per-
spective of public choice theory, self-interest and the power games that go
with it for politicians and interested parties were put forward as grounds
for the necessity of changing perspectives; because, independent of their
role in the economic policy-making process, the choice and decision-
making behaviour of economic subjects does not by any means lead to the
preference for continuously ‘correct’ decisions – quite apart from the fact
that the determination of what could be ‘correct’ is itself confronted with
insurmountable theoretical and practical obstacles (cf. e.g. Frey and
Kirchgässner 1994, pp. 362ff.).
It is however only recently that a problem area which develops at an
earlier stage has increasingly come to the fore. This problem area becomes
obvious when moving away from the customary common-knowledge
assumption which is an implicit part of the formation of theory and which
assumes a predefined econopolitical event or situation. Before the situation
is assessed and alternatives chosen at a macro-economic level, the question
as to how an ‘econopolitical situation’ originates needs to be posed. Here it
can be shown that the results gained from determining the situation, setting
an aim and selecting measures are very dependent on the individual and
societal perception and communication processes which come before them.
It is in these processes that certain problem areas crystallise, the contours
of which gradually become clear at a later stage. This means, in contrast to
the assumptions of many models of economic policy theory, that there are
neither ‘ready-made’ solutions to problems in advance, nor, and this is a
particularly important point, are there ‘ready-made’ problems which just
need to be identified by individual politicians or parties and, depending on
their political couleur, integrated into respective agendas. Very complex
phenomena play a role in the complicated interaction process which deter-
mines at what point in time and at what ‘stage in its development’ an eco-
nomic problem which is initially only perceived by a minority actually
advances to become a problem of economic policy. At the beginning of the
problem process, this includes for example the degree of homogeneity of
perceptions of reality existing in a society, the individual and collective
perception of certain information, the way problems are ‘offered’, i.e.
processed or presented in the media, as well as the effect of historically
grounded path dependencies, inponderable events or accelerating/retarding
frequency dependency effects in the course of the genesis of a problem.
The contexts mentioned here will be developed in four steps focusing in
particular on cognitive aspects: firstly, a short look will be taken at the
general significance of epistemological factors in the analysis of economic
policy, followed by an investigation into the relevance of ‘cognitive maps’
which guide problems originating at an individual level. Then, an attempt

will be made to describe and make sense of the problem-diffusion process
and the corresponding generation of problems in economic policy making
at a societal level as an important preliminary phase to the problem treat-
ment process. Finally, in the last step, selected cognitive aspects of the
process of problem treatment in economic policy will be examined.
2. ECONOMIC POLICY IN THE LIGHT OF THE

PROBLEM OF COGNITION
For further analysis the first question which should be asked is why instru-
mentalistic-technocratic thought patterns have established themselves so
firmly in many areas of economics. The idea that we can analytically separ-
ate the possibilities and limitations of cognition from the conditions of
action seems to play a role here.4 Methods based on this regarding the
partial analysis of problems will not occasion much criticism. However
the same is not true when dealing with such a complex phenomenon as the
process of economic policy. For this subject matter, the integration of
results of modern perception, cognition and communication research
seems of the essence, and a central demand of this chapter is to include the
construction process of ‘econopolitical reality’, together with the relevant
constructional content, as an issue in a positive theory of economic policy.5
To achieve this, first of all a short overview of the mainstream of modern
cognitive sciences6 will be given. There has been a clear departure from the
idea of a passive, externally determined perception of the environment.
Today, it is normally considered as proven that our brain, which processes
environmental stimuli, links these external impulses with already existing
cognitive structures in an actively constructive way. Every piece of incom-
ing information absorbed only receives its specific identity in our brain and
mechanisms such as perception, inference, attribution and memory in par-
ticular act as mental operations.7 Cognitive psychology provided sound
theoretical and empirical proof for these contexts as early as the sixties and
seventies (cf. Neisser 1976) and this later provided strong support for the
development of a perspective which was to become known as cognitive
constructivism. The following ideas in this chapter are based on this
approach, which is present in almost all areas of psychology today (for an
overview cf. Nüse et al. 1991). In this epistemological model it is important
to understand cognition also as discovery and invention, without having to
commit to a decision as to which parts of human perception actually
provide information about the existence of an outside world which is
subject-independent. Without negating a ‘neutral’ environment, there is an
Economic policy 171
unavoidable veil of uncertainty surrounding any statements made about it.

The main general consequences which result from the integration of these
contexts into an analysis of economic policy can be outlined in three points,
as follows.
2.1 Subject-specific Reception of Information as a Basic Variable of

Economic Policy
First of all, the idea of ‘information’ as the basic variable of planning in

economic policy needs to be rethought. Here, the basic significance of
information within this process is not being questioned, however a
modification of the way information is understood is required. Modern
cognitive sciences emphatically prove that information about facts does not
exist independently of individuals and thus cannot simply be passed on to
others. Information in the actual sense of the word is created at the moment
of reception of external signals in the apparatus of human perception.
Thus, in the context of economic policy, the prevailing conditions of this
genesis or transformation of information need to be investigated more
intensively. Information transformation and information transfer can be
taken as phenomena which belong together and it is of particular import-
ance to realise that each piece of transferred information gains a new iden-
tity in the act of being transferred. The main question to be posed here is
which factors, either intended or non-intended, influence the process of
gaining meaning.
2.2 Individual Information Processing and the Problem of Predictability

of Results
Secondly, these aspects throw a different light on the problem of forecast-

ing, both at a scientific and an economic policy level. In information trans-
formation the way our perceptive apparatus functions seems to hinder any
use of transformation rules. Research into perceptive physiology has
proved that the brain works in a kind of temporal rhythm, always combin-
ing several physiological events into one perception, then to be replaced
autonomously or spontaneously by another. Here the content of a later per-
ception can be changed even if there has been no change in the sensory data
(for an extensive description see Pöppel 1985; Oeser 1987). Suddenly, com-
pletely new interpretations of situations can be created, which may then
lead to very unexpected and new reactions.8 This means, ceteris paribus,
that it is already for purely cognitive reasons that the possibilities of fore-
casting economic, social and political cycles decrease with increasing com-
plexity of information networking in a society and between societies
(globalisation), especially as there are basic difficulties in connection with

the attempt to observe and evaluate internal personal processes (cf. Little
1972; Nisbett and Ross 1980; Witt 1992).
2.3 The Communicative Act in its Significance for the Form of

Econopolitical Situation
A third consequence of integrating perception and cognition theories into

the formation of a theory on economic policy originates in the central idea
that no economic subject thinks, plans and acts against the background of
an objective reality, but rather always within the framework of their own
subjective reality which only experiences a partial comparison with other
individual ‘realities’ via an act of communication. On the one hand ‘cogni-
tive maps’ play a role here, a concept which will be dealt with in more detail
later. On the other hand the function of social interaction as a reality
shaper comes into play. Living together in a society and economic policy
are therefore only imaginable if, at least in certain areas, common relevant
realities are created by way of communication as described by Berger and
Luckmann (1979). This means that, from a cognitive-evolutionary per-
spective, societal communication has a much more important function
than merely the exchange of information, taking the term ‘information’
here in the traditional, more technical sense (cf. Habermas 1984).
3. ELEMENTS OF AN EVOLUTIONARY THEORY OF

ECONOMIC POLICY: ON THE ORIGINS OF
PROBLEMS
3.1 The Individual Level
Here, a closer look will be taken at the actual phenomenon, the origin of
problems in the context of economic policy. Firstly, in accordance with
methodological individualism, what happens on an individual level is of
interest – especially considering the fact that there is a deficit in explanation
which needs to be dealt with that comes up in well-known attempts at soci-
ological explanations, which are mainly based on collective phenomena in
the political process (cf. Scharpf 1991; Searle 1995). To examine the indi-
vidual level the concept of ‘cognitive maps’ will be addressed here. This
concept, the development of which is linked to Piaget’s work (Sugarman
1987), combines the above-mentioned concepts with a series of subphe-
nomena in information processing which have not yet been addressed. For
example, an important point is that cognitive maps are highly selective,
Economic policy 173
meaning that pieces of knowledge are linked within them under the same
label in the form of terms for characteristics, prototypes, limiting cases and
examples. At the same time, so-called tacit knowledge is arranged, know-
ledge which is constantly developing and being applied, thus changing in
action and at the same time influencing the human ability to act without
the actor himself or an external observer being able to systematically
describe its structure (for an application-based overview see for example
Hirschfeld et al. 1994; Vowe 1994).
Cognitive maps which help master the complex range of sensory stimuli
play a central role in any cognition and memory process: they form the
framework, so to say, into which newly constructed information is inserted,
to be then complemented by already-organised pieces of knowledge (cf.
Eysenck 1984, p. 324). Thus a referential network is created within which
experiences are condensed into patterns, which themselves then shape expe-
riences. All evaluations, expectations, plans and decisions made by actors
are based on this interdependent linkage of present and past cognitions
(cf. Koch 1996). All economic subjects, whatever role they are playing
at any moment in the process of economic policy, are affected by this sub-
jective prestructuring of perception. Politicians, bureaucrats or voters,
members of whatever social class, the old or the young: in their reception
of information, events and complex situations everyone is subject to the
biological restriction of only being able to perceive what their brain offers
them on the basis of their personal (stored) ‘history’.9 In contrast to the
concept of ‘bounded rationality’, the concept of ‘cognitive maps’10 thus
emphasises not so much the lack of quantity in information gathering as
the unavoidability of path-dependent quality changes in the reception of
information.
Should there be considerable contradictions to the previous subjective
‘reality’ in an individual perceptive situation in the face of certain ‘infor-
mation conditions’, these cognitive dissonances can at the same time
awaken the wish to reduce them. The strength of the stimulus depends on
the estimation of the expected costs and benefits of corresponding activi-
ties.11 The costs can be, for example, to analytically structure and then to
communicate the initially subjectively defined problem (for example, in
discussions or publications), they can express themselves in altered choice
behaviour and they can (this option involves the highest costs) imply a trig-
gering of activity in citizens initiatives, parties or government authorities.
‘Information conditions’ in which cognitive dissonance occurs can on the
one hand be newly created. But on the other hand they can also remain
unchanged and the ambiguity results solely from a recombination of estab-
lished consciousness content which is produced by the apparatus of per-
ception.12 While the latter possibility can occur at any time, the reception
of new target parameters (for example ideas or values) or new knowledge

about actual parameters (causal contexts and restrictions) must be added
for the former.
The discussion so far implies an altered image of the actors in the politi-
cal process. At first we have the politician in mind as the economic policy
maker, but the assumption of the ubiquitous influence of cognitive maps
emphasises that any idea of an exogenous controller having a complete cata-
logue of assumptions about the behaviour of actors in a system is doomed
to failure. Their perception and their resulting behaviour are rather insepar-
ably and reflexively coupled with the specific social context which they are
a part of. Every economic politician is himself or herself ‘just’ an endoge-
nous variable in the model of economic policy. In the process of changing
cognition, knowledge components and evaluations, their own cognitive
maps are just as inconsistent as those of any other individual agent, so that
in principle no ‘objective’ decision-making field can exist in the area of eco-
nomic policy (cf. Scherer et al. 1988, pp. 803ff.). Concerning the group
voters, there are important conclusions to be drawn from the point of view
of cognitive science, which are complementary to corresponding reflections
in political economics: in the case of democratic votes on actors and
programmes in economic policy, for example, it must seem obsolete to look
for voting procedures which could represent an ‘objective’ portrayal of
voters’ preferences. Also the idea of the ‘voters’ mandate’ which is often
diagnosed is nothing more than a point-in-time-based, open structure
which changes its importance in a continuous process of reflexive interpre-
tation of reality.13 Each elected person interprets the ‘voters’ mandate’ just
as contextually as the voter evaluates the information available to them con-
cerning a social situation.14 Voters and the elected will change their ideas of
target and performance mutually dependent on each other in the course of
a legislative period and will thus also change their interpretation of problem
situations, leading to a co-evolution of relevant decision-making fields.
These thoughts already show clearly why an evolutionary understanding of
economic policy is being supported here and at the same time, they lead to
the analysis of the collective level of the process of economic policy.
3.2 The Collective Level
At the level to be discussed here, the question needs to be dealt with as to

which kinds of subjectively perceived discontent (problem variation) – for
example if faced with an expected under-supply of certain public goods –
actually end up becoming a problem of economic policy (problem selec-
tion). To explain this, the constitutive function of interaction and commu-
nication for societies needs to be examined. When individuals come into
Economic policy 175
contact with each other, they ‘automatically’, so to speak, alter reality for
the person they are interacting with. They observe reactions to their own
actions and, if necessary, conceptualise recurrent new experiences together
with previous knowledge as cause–effect relationships in new variations or
cognitive maps. The fact that in this situation each individual acts accord-
ing to their own, subjective, schematically coded expectations, aims and
purposes is also termed ‘double contingency’. As we can assume that within
social communities the subjective realities of interactive partners are at
least partly synreferential in advance, actors can also always work towards
orienting the cognitive area of others towards a certain object (cf. Rusch
1994, p. 63). This point represents an important interface where it is
decided whether individually perceived problems at a certain point in
time ever become intersubjective or societal problems. The mobilisation
resources of individual economic subjects play a decisive role here and
there is a wide range of possibilities. Firstly, it depends on the strength of
the intrinsic potential, i.e. on the degree of subjectively perceived dissatis-
faction and on the will to mobilise. Secondly, there must be sufficient mobil-
isation abilities such as the correct selection of communication partners
and the necessary argumentation skills. Here the social position taken on
by the mobiliser is important, as well as the economic and/or political
resources and power potential available to them and the experience already
had with them in similar situations in the past.
When the spark of problem perception jumps from the individual level
to the collective level, a problem subsequently gains a new piece of identity
with each further communication, i.e. its quality by no means remains con-
stant during the diffusion process. Nor is there any guarantee that every
person concerned will interpret a communicated problem in a similar way.
Here the complexity of the subject matter and the quality of the commu-
nication process play an essential role. ‘The more successful orientation
interactions there are and the more reliable they are, the stronger will the
actors be convinced of the common and equal nature of reality’ (cf. Rusch
1994, p. 71). In other words, depending on the situation there can be a
homogenisation of how people interpret reality and problems.
Whether a problem which is perceived and discussed in parts of society
actually becomes an issue in economic policy depends in essence on this
homogenisation process. The more homogenous the perception of a certain
economic target/performance discrepancy and the larger and more power-
ful the dissatisfied group, the higher the probability, all else being equal, that
the problem will be admitted to the political level. In spite of this, no gen-
erally valid if-then criteria can be derived for this step. Historical constella-
tions play a role which need to be evaluated according to what kind of
macro-economic situation a society is in, what other problems are on the
issues list and how urgent they are, and thus what problem perception
and problem treatment capacities are still available at that point in time
(cf. Meier and Slembeck 1994, p. 75). A further very important point to
consider is what institutionally dictated mechanisms facilitate or impede the
problem being admitted to the political level (e.g. constitutional require-
ments for referendum petitions) – a question which will not be dealt with
further here. But even if a certain degree of homogeneity is necessary as a
kind of minimum ‘critical value’ for admitting a problem to the level of eco-
nomic policy, the relativity of this context must be considered. Because in
increasingly untransparent, pluralistic and multicultural societies it will
become generally more difficult to install homogenous, powerful interpre-
tations of problems which are at the same time stable in time and content.
With the acceleration of technological progress and the global diffusion of
knowledge (decreasing ‘knowledge half-life’) as well as the internationali-
sation of the economic and interjurisdictional competition that goes with
it, the relevant problem areas and at the same time the interpretation of
individual problem areas are fanning out more and more. However the
more heterogeneous the prevailing and competing realities are, the more
difficult it is, all else being equal, to make any qualified statement on the
course of problem-admittance processes and thus projections about future
developments in the process of economic policy as a whole.
In the face of this development, one thing is becoming more and more
important: to what extent are initially alien problem areas and interpreta-
tions of problems communicated to the other groups in a society in order
to create a pool of relevant information in common knowledge (intended
reduction of ambiguity)? From the perspective proposed here, such a pool
seems to be a conditio sine qua non for constructive sociopolitical discus-
sion on the solution of problems which come up. It combats to a certain
extent a ‘derationalisation’ of the process of economic policy making. In
accordance with this, the economic policy-making process would be ‘ratio-
nal’ in that what is being communicated about is also what every individual
concerned understands as being what they want to express; therefore, the
more intensively a theme-based process of mutual orientation gets going,
the higher the probability that a problem recognised by many will later be
treated ‘correctly’ or ‘rationally’ – in the view of many.
Let us now take a closer look at the mechanisms and patterns which play
a role in the process of problem diffusion in economic policy. This includes,
above all, the observation that cost–benefit relations which economic
agents are confronted with, if they want to spread an individual view of a
problem, are highly dependent on the established communication struc-
tures within a society. These form the ‘resonance space’ in which problem
interpretations have to prove themselves (cf. Siegenthaler 1994, pp. 177ff.).
Economic policy 177
Here, communication structures should be understood as the total number

of valid possibilities and limits of communication between economic sub-
jects in a society. In the modern ‘information economy’ the various media
are of particular importance (cf. Löffelholz and Altmeppen 1994). They do
not simply function as ‘neutral’ communication channels (information
transfer), they rather play a decisive controlling role in problem-origination
processes by spreading new possibilities for interpretation (information
transformation). The abstract object of public opinion as an obstacle or
selector for individual models and interpretations is to a great extent a
product of the media. A careful evaluation of public opinion can show
individuals or groups of actors what costs of mobilisation and thus what
level of material and personal resources are to be expected to ensure that a
problem actually becomes an issue in economic policy (cf. Merten and
Westerbarkey 1994, pp. 200ff.; Kuran 1995).
In this context it can be shown that the problems by no means always orig-
inate in the same way. Ideal-typically, there are two basic patterns (cf. Meier
and Mettler 1988, pp. 40ff.). The first one, the bottom-up formation of
opinion, represents the ‘ideal democratic situation’: a problem is perceived
by a group of individuals relatively homogeneously, and potential support-
ers – i.e. interest groups, governments and parliamentarians – have to be
‘triggered into action’ on this basis. The extensive overlapping of individual
cognitive maps which is necessary in advance of this15 means that this type
seldom occurs in its purest form, especially considering the fact that there
are many structural obstacles which can have an effect. Of particular impor-
tance is the fact that as a rule the so-called agenda setters are indispensable,
i.e. those who take a problem and make it a political issue, whereby the
mobilisation of effective media is potentially very ‘expensive’ in this case.
Therefore a large part of the formation of opinion in society is a more top-
down situation, i.e. it is triggered by influential actors. For politicians and
representatives of powerful interest groups – for example trade-union func-
tionaries – the access to relevant publicity channels and the focused
influence on public opinion that goes with it, mostly incur relatively low
transaction costs.16 ‘They have capacity available to them to ensure a media-
friendly processing of issues and therefore enjoy good access to the mass
media’ (Meier and Mettler 1988, p. 41). Thus it is relatively easy for them to
produce certain problems or interpretations of problems in a focused way,
i.e. to chose the right moment in time, in order to instrumentalise them in
the process of economic policy making. This can be observed at the time of
writing in the discussion about the introduction of a common European
currency or in the debate on the consequences of economic globalisation.
The question as to what events or constellations kick off a problem
process and in which economic, social and political context plays an
important role here. If there is latent discontent which has been predomin-
ant in a larger social group for a long time, the course of problem diffusion
will be different than if only a small peripheral societal group sees need for
econopolitical action on a certain issue. Based on a categorisation in
Bennett (1980, pp. 113ff.), various categories of initiation of econopolitical
problems can be differentiated according to the number of people affected
and the extent to which they are affected (cf. Meier and Slembeck 1994,
pp. 68ff.). According to this, one can speak of an interest-group problem if
only a minority in society is affected but this minority organises itself for-
mally on the basis of the high willingness to pay of the individuals con-
cerned. On the basis of the German coal problem in the Ruhr area, the
cognitive determinants which are relevant here become clear: normally this
concerns groups of economic subjects whose individual realities have a high
level of synreferentiality in advance, as they have economic (industry, pro-
fession etc.), social (class, ethnicity etc.) and/or geographical (estate com-
munity etc.) features in common. This means that the costly setting up of
functioning communication structures and the spread of common inter-
pretations of reality are rendered superfluous. In the case of structural prob-
lems, however, there is a large number of affected actors but the individual
is affected to a lesser degree. Even if the motivation to form a formal organ-
isation is weaker for this reason, this type of problem triggering is still just
as significant as the former one. To go back to the planned introduction of
the euro, for example, the political significance associated with it becomes
clear. In the run-up to the Bundestag elections in 1998, the election cam-
paign strategists of the largest German parties saw an above-average level
of uncertainty concerning this issue. From a cognitive-evolutionary point
of view, the reason for this was that in the course of the currency debate each
potential voter had an opinion, but that these opinions were based on com-
pletely different backgrounds concerning knowledge, experience and inter-
est. As a result, it seemed dangerous to commit to a certain strategy at too
early a stage, as public opinion was extremely unstable under the circum-
stances and still is today in connection with this theme.
Only the latter of the ideal types of elite and crisis problems which are
less frequent in reality will be looked at here as this is a particularly inter-
esting phenomenon for the context under discussion. A chaotic situation is
characteristic, after which a coincidental individual event can very quickly
completely change the societal assessment of the situation and thus also the
relevance of possible previously existing problems (interest group, structure
or elite problems) (cf. Kuran 1995). For example, taking a look at public
reaction to a media-effective accident at a chemical production plant,
extremely intense ‘avalanche-like’ protest can be observed in certain cases
– especially if there have been no similar accidents in the recent past. It may
Economic policy 179
even happen that all chemical production plants in a country become time
bombs in public opinion. The decisive consideration here is that the sud-
denly altered assessment of the situation can lead to huge pressure on polit-
ical actors. A social problem is ‘born’ which results in a political necessity
to immediately formulate and implement new aims for a certain area which
has been at most a peripheral issue for a long time as regards social inter-
est. It is obvious what kind of economic consequences such coincidental
individual events can have, for example, for a whole industry. From a cog-
nitive-evolutionary point of view, the interesting aspect to such an example
is that in certain cases the homogenisation of cognitive maps can take place
very quickly and unpredictably. When a large majority has swung towards
a certain view of reality, decisions can be taken which would never have had
a chance of implementation in the past. This is also an aspect which again
makes very clear the significance of the factors ‘coincidence’, evolutionary
self-dynamics and path dependency for the results of the process of
economic policy – phenomena which are often neglected in traditional
approaches.
4. ELEMENTS OF AN EVOLUTIONARY THEORY OF

ECONOMIC POLICY: ON THE TREATMENT OF
PROBLEMS
The above reflections lead us to an analysis of the phase of problem treat-

ment. In the process of economic policy, this follows the phase of problem
origination, provided that this has led to successful admittance of the
problem to the political level. The following sketch (figure 8.1) is an attempt
to recapitulate on the evolutionary view which forms the basis for this and
which has been discussed above. As the arrows indicate, changing private
and/or public interpretations accompany the process of economic policy
from the first phase of problem origination right up to problem treatment –
and indeed throughout all its phases respectively.
As the psychological-cognitive phenomena which occur in the process of
problem treatment have already been dealt with to a much greater extent in
economic literature than the subject matter discussed here so far (cf. Jones
1984; Kuran 1987; Hutter 1989; Lindenberg 1993; Frey and Eichenberger
1994; Meier and Slembeck 1994), this section will be kept relatively brief.
Just a few points will be selected which clarify the cognitive-evolutionary
dimension of the context of effect on the level of political action.
In general, current power constellations in the competition of interests
and interpretation are also the most important factor in the problem-
treatment stage. Even if a problem interpretation has already established
Individual level Collective level

Subjective problem Intersubjective
origination problem origination
Public & private

institutions which
channel
communication and
influence opinion
(esp. the media)
Political level
Problem admittance
Problem treatment
Figure 8.1 From the origins of a problem to its treatment – sketch of the
process of economic policy making
itself or become concrete at a collective level to such an extent that it has

achieved admittance to economic policy, individuals and groups will con-
tinue to attempt to change things to their advantage. Here, there are many
different formal and informal options on various decision levels of
problem treatment depending on the specific structure of the political
system. For example it is well known that the preventive measures taken
against the state’s abuse of power in the more recent constitutional democ-
racies (‘checks and balances’), result in many veto and influence possi-
bilities which do not exist in other systems. However, there are also
considerable differences between the various forms of democracy. For
example, the British Westminster model or its New Zealand offshoot
enable a determined head of government or an ideologically homo-
genous ‘policy conspiracy’ to act much faster on the basis of their own
Economic policy 181
interpretation of a problem than is possible in corporatist systems (cf.

Wallis 1997).
It should be kept in mind that regardless of to what extent they can
influence and affect the situation, actors will attempt up to the last moment
to use all their power, resources, and powers of persuasion in order to
ensure the success of their own interpretation of a problem and normally
also the implementation of their own solution model. This continuous
competition between alternative interpretations means that the observer is
constantly being offered numerous – prima facie unsystematic or even
seemingly contradictory – partial interpretations of the process of eco-
nomic policy. Such apparent ‘irrationalities’ can be observed in particular
when very different reality models are almost forcibly combined into one
perspective in the problem-treatment phase. This is because from this polit-
ical reality certain aims are formulated and certain measures are chosen
solely due to the fact that this ‘artificial’ current perspective spreads very
definite causalities – even if the latter often seem very unsuitable to the
external observer (cf. Luhmann 1989). In the background there is an
important basic assumption which motivates the political process. This
assumption aims at overcoming societal conflicts which hinder action by
taking the wide variety of individual realities into consideration.17 So it is
no surprise from a theoretical point of view that in this hither and thither
of competing ideas about econopolitical cause-effect contexts, chains of
action are created which are difficult to comprehend taken at face value.
Furthermore a cognitive-evolutionary approach lends credibility to the
assumption that the idea which could be labelled systematical ‘muddling
through’ (cf. Lindblom 1959, 79ff.) is not a contradiction in itself. It should
be much more the rule in economic policy – at least in a democratic
market-economy system which is conceived as an open and pluralist
society.18 This is because with the institutionally secured admittance of
variety, the unavoidable necessity of making compromises on the public
level of action is created, implying a loss of logical consistency of the
subarguments.
An extensive analysis of different psychocognitive subphenomena of
the problem-treatment process in economic policy could now be added to
these general reflections, and there are various systematisation methods
available for this. For example, the systematics widespread in political eco-
nomics could be followed and specific ‘anomalies’ could be worked out
for groups of actors (personal differentiation), for voters, politicians, par-
liamentarians and bureaucrats respectively (cf. Frey and Eichenberger
1994). Another possibility is, as shown below, to differentiate subphe-
nomena according to their occurrence in different phases of political
control (temporal differentiation), for diagnosis, prognosis, programming,
decision, implementation and evaluation (cf. on this sequence Meier and

Slembeck 1994). For this discussion it will, however, suffice to give an
example of several selected cognitive factors which play a role in attach-
ing a sometimes ‘irrational’ character to the real problem-treatment
process in economic policy.
4.1 Persistence of the Status Quo as a Result of Diagnosis
Once a problem has become an issue of economic policy, it must then be

operationalised. To do this an actual situation must be determined by
searching for suitable variables. Depending on the specific subject matter,
this can be done quite easily, for example, if it is a routine case to which
experience values can be applied. But it could also be a new kind of
problem constellation for which there is no solution grid yet and whose
cause-effect context is evaluated extremely heterogeneously by the decision
makers. In this situation there are two basic possibilities: on the one hand
there could be a compromise-led diagnosis result as described above, so
that the problem-treatment process can run its course. Or the treatment
process is abandoned after the unsuccessful attempt to demarcate the
problem in content and form. This possibility is enhanced if the problem
has also gradually lost significance in public discussion – whether by force
of politics or not. From the point of view of political economics, for
example, this form of significance loss could be based on the fact that the
problem was instrumentalised solely for election-campaign purposes,
the agenda setters lose almost all interest in it after the election and the
problem falls victim to the fact that actors and voters forget relatively
quickly. From the point of view of cognitive science, and taking an indi-
vidual perspective, there is the additional aspect that in specific cognitive
structures the ‘frame goals’ change from time to time.19 This can help
‘solve’, for example, long-term contradictions; because, as neurologists
have proved, our nervous system contains qualitative functions which work
as a filter in cognition, so that logical contradictions cannot always or con-
tinuously be perceived as cognitive dissonance. Thus this brain function is
a kind of protection mechanism for the reduction of environmental com-
plexity and for stabilisation of cognitions (cf. Ciompi 1988). At the collec-
tive level these kinds of phenomena can become so strong that in case of
a long-lasting ambiguity of problem interpretations which cannot be elim-
inated by communication, the previously existing conviction, both on the
part of politicians and those concerned, that there is a basic need for action
is gradually reduced. The result would then be the continuance of the for-
merly criticised status quo.
Economic policy 183
4.2 The Inherent Value of Motives in the Prognosis Phase
If on the other hand there is an agreement on a majority ‘political reality’

in the diagnosis phase as a result of a successful reduction in ambiguity as
described above, a problem-based target situation needs to be determined.
Here there can also be typical ‘irrationalities’, for example when certain
individual variables are overvalued in the context of describing the situa-
tion. Often the result of this is that an attempt is made to solve an urgent
problem in area A, but that a new problem is created in area B with just as
many consequences. This kind of one-sided orientation towards individual
variables can easily lead to a long-term process of swinging between
different problem situations, a fact which is proved by the well-known
example of the political activity-cycle. In addition to established explana-
tions based on public-choice theory (cf. Frey 1983), explanation factors
based on cognitive science can also play an important role here: on the one
hand, when temporarily overvaluing certain motives in the context of for-
mulating targets, consequential problems may be consciously reckoned
with; on the other hand non-intended processes may also be crucial. Again
the decisive point is the consideration of the cognitive context of the action
situation, from which the dominance of certain ‘frame goals’ with tempo-
rary suppression of other motives can be explained. In addition, it seems
important that the sometimes dominating motives of economic subjects
change at certain intervals, something which is difficult to forecast and
observe from the outside. If certain individual ‘frame goals’ assert them-
selves in the societal communication process, they will also influence the
determination of collective aims in the problem treatment process of
economic policy. From this perspective, it is very possible that ‘rational’
events form the basis for neglecting motives which ‘make sense’ to the
benefit of one-dimensional formulation of aims.
4.3 The ‘Wrong’ Dosage of Measures in the Programming Phase
Once comparable performance and target variables are available as a result

of the above-mentioned processes of determining interpretations, the pro-
gramming of bundles of measures can begin, which then have to be decided
on politically and then possibly implemented and tested for effectiveness
(evaluation). For the analysis of this phase, cognitive phenomena again play
an important role alongside econopolitical phenomena. The cognitive phe-
nomena help to explain the consequences of actions which are basically to
be evaluated as failures on the part of political actors. Thus the result is, for
example, faulty programming because, due to the make-up of the human
perception apparatus, evidently contradictory data are incorporated into
established solution patterns. The reason for this is that, against the back-
ground of certain cognitive structures, frequency distributions and proba-
bilities are systematically disregarded (cf. Tversky and Kahneman 1982).
Other investigations show that the ability to correctly estimate effect mech-
anisms and reaction-times from the application of the instrument to the
intended effect is also very different for different individuals in complex
socio-economic systems (cf. Dörner 1991). As already mentioned, no
simple input-output prognosis modes can be constructed for these pur-
poses, so the only alternative is to take experience values from comparable
situations. Here, intuition plays an important role; from a scientific point of
view, we are dealing with implicit heuristics that the perception apparatus
offers spontaneously (cf. Six 1987). Typical for the investigation of system-
atically distorted cognition performance is the incorrect assumption that
there are linear cause-effect relations and a disregard of feedback (e.g. self-
destroying and self-fulfilling prophecies), regression and contradictory
effects as well as the creation of consequential problems in other areas of
the system (cf. Dörner 1991; Nisbett and Ross 1980, pp. 124ff.). On the
other hand, causal contexts or intentionalities are often searched for where
the assumption of coincidence and thus the application of probability con-
cepts would be more suitable (cf. Ross 1977, 175ff.). In the end, time-frames
are often wrongly estimated and the present is overvalued in relation to the
future.20 While the above-described ‘anomalies’ generally lead to wrong
dosages when applying instruments, this cognitive failure may explain in
particular overdosage in planning measures. This is because the fact that
measures take effect at a later stage is often misunderstood and it is assumed
that the measure has no effect at all (cf. Dörner 1991, pp. 156ff.). In con-
nection with all of these observations there is empirical evidence that more
experienced people systematically act more successfully in complex deci-
sion-making situations than the less experienced. However, this is only true
under one condition; that both groups of test subjects receive the same
initial information on the concrete situation in their respective experiments.
5. RESUMÉ
This quite cursory analysis (for the reasons stated) of the process of
problem treatment will now be followed by a summary of the results pre-
sented. It was shown that, for an understanding of the process of economic
policy, it can be helpful to take a look at the origin of the problem from a
cognitive-evolutionary perspective, analytically differentiating between the
individual and collective levels, before going over to the treatment of the
problem. Phenomena which can be explained from a psychocognitive
Economic policy 185
perspective have the effect that both the process of problem origination as
well as the process of problem treatment in economic policy often seem to
the observer to be systematically ‘unsystematic’. As problem origination
and problem treatment are always accompanied by complex competition
for influence and power between different interpretations of reality, what is
left over as the final result of this process on a collective level can still only
be an ambiguous and partially contradictory compromise. This results
from the collision of many different ways of thinking, reality models and
ideologies whose foundations, even if they seem incommensurable, are set
in relation to each other. Because it is the motivation of the decision makers
in economic policy to reduce ambiguity on all levels – at least in a democ-
ratic system – in order to become or remain able to act. The extent to which
the left-over residue of ambiguity or cognitive heterogeneity is present in
the phase of problem treatment depends not least on the informal and
formal givens of the sociopolitical system of a country. In an individual
case analysis, the quality of the installed communication structures and the
structure of the democratic involvement and decision-making rules would
need to be investigated here.
In addition, it should be kept in mind that the appearance of such
problem interpretations born out of the societal communication process
and the solution patterns which follow is in no way long-lasting; it rather
shifts according to the extent to which certain events, discoveries or power
shifts help other or new realities to become temporary ‘competitive
successes’. Therefore, a positive theory of economic policy must also the-
matise influential factors of the variability of its categories. It should
understand the societal communication process as an all-encompassing
medium in which econopolitical interests, models and ideologies are trans-
ported, strengthened but also put in their place, thus at the same time
sketching again the process-oriented framework concept, within which
more light can be shed on the respective individual phenomena.
NOTES
1. Molitor, for example, demands: ‘the logic of the econo-political situation entails a divi-
sion into aim, situation and measure. Whether dealing with political tasks concerning
order, stability or distribution, policy should always follow the three steps of interpreta-
tion of aim, analysis of situation and programming of measures’ Molitor (1988), pp. 29.
2. For an overview of the historical genesis of this programme cf. Albert (1977), pp. 181ff.
3. Here it is of no importance whether the machine is trivial or non-trivial: in a non-trivial
machine the output sequence is not only determined by the input sequence and the trans-
formation rule fy(x → y), it is also determined by the rule ( fz), according to which the
transformation function ( fy) is altered. In this way, very complicated input-output
sequences of a higher order are conceivable.
4. Or, as Schiepek (1990), p. 199, criticizes common practice: the attempt is often made ‘to
free scientific descriptions from their cognitive preconditions, i.e. to eliminate the condi-
tions of their feasibility’.
5. Cf. Albert (1977), p. 205, who demands the inclusion of epistemological aspects in a
theory of econopolitical action ‘which should be taken seriously from a realistic point of
view’.
6. ‘Cognitive sciences’ include all disciplines which deal with the phenomena of perception
and cognition, including neurobiology, chemistry, physiology and psychology, other
fields of biology and psychology as well as some areas of philosophy.
7. Perception is the absorbtion of stimuli by sensory cells and organs, inference is the for-
mation of implicit conclusions and attribution is the subconscious allocation of cause
by the brain.
8. In this context Hesse’s reflections on ‘cognitive creativity’ as a basic variable in the theory
of action and economics as well as the consequences of its integration for the explana-
tion of innovation are particularly significant; cf. Hesse (1990). For corresponding con-
sequences in economic policy cf. Koch (1996).
9. In a similar context to the one discussed here, Frey and Eichenberger (1991), p. 75, under
the point ‘Reference Point Effects’, refer to the example of the fear of inflation some
Germans have. It seems to actually make a difference in the evaluation of econopolitical
scenarios whether actors still have (unpleasant) memories of inflation they experienced
themselves or not. People with experience of certain events and developments will sub-
consciously evaluate situations differently than individuals without the same history.
10. These are sometimes called ‘scripts’, or ‘frames’. In all cases these are special types of
complex structural organisations (the terms are also partly synonomous). For example
‘frames’, which will be dealt with later, normally describe organised knowledge struc-
tures that order conventionally determined knowledge of laws and standards which
plays a role in social situations. Cf. Dijk (1997).
11. Cognitive assimilation and accommodation processes play a role in connection with the
motivation to reduce cognitive dissonance or ambiguity. Cf. Meier and Slembeck (1994),
pp. 54ff.
12. In addition to the above-mentioned sources, cf. Pöppel and Edingshaus (1994), pp. 166ff.
on the cognitive-scientific preconditions for this possibility.
13. In this context it can be of interest to analyse the mechanisms of reflexive reality model-
ling, also in advance of elections and votes. Spontaneous actions and institutionalised
procedures (cf. Blondel, 1990) ensure that the actual decisions have been made long
before the formal election or vote takes place. In spite of this, this cognitive-evolutionary
aspect of preparation for votes is often undervalued in expert literature on the subject,
an exception being Slembeck, (1997).
14. Kuran (1995), pp. 189ff. analyses in detail the extent to which private preferences for
example are in a continuous change process depending on private knowledge and public
discussion and the ‘preference distortions’ which result.
15. The word ‘overlapping’ here should not be understood from a set theory or homological
perspective, but rather in an abstract way which refers back to the above-described
reflections on communication theory.
16. Cf. Mueller (1989). The general idea is described that the transaction costs of turning
social or political problems into issues (in the widest sense) are basically communication
costs, using the example of patent-law production in Hutter (1989), p. 187f, examples for
politically motivated ‘problem generation’ (top-down) can be found, for example, in
Scharpf (1991).
17. However this does not exclude the possibility that political oppositions produce ambi-
guity on purpose to hinder the success of the actions of decision makers or to manipul-
ate society’s evaluation of the results of actions.
18. A reference can be made here to Popper’s advocacy of a social technique of ‘piecemeal
social engineering’, which he explains on the basis of the characteristics of such a society
concept. Cf. Popper (1995).
Economic policy 187
19. See, for example, Lindenberg (1993) for an explanation of the framing concept which is
based on the above-described cognitive-scientific results. From an empirical point of
view, Lindenberg refers in particular to investigations carried out by Goffman (1986) as
well as by Tversky and Kahneman (1981). The latter proved in experiments that indi-
vidual decisions for different alternative actions can be extremely dependent on how the
respective possibilities are described. In the experiment the descriptions ‘objectively’ con-
tained the same information, but in some cases the possible profit was emphasised, in
others the danger of loss, cf. ibid, 453ff.
20. Here, the elementary human characteristic of setting every action and experience on a
time-scale can be counterproductive. Because by constructing action sequences, interac-
tion chains and contexts that ‘make sense’ from a procedural perspective, thus turning
singularities into elements of procedures (cf. Vowe 1994, 438), there is always the possi-
bility that incorrect temporal structures will continue to exist as consequential errors in
action planning.
REFERENCES
Albert, H. (1967), Marktsoziologie und Entscheidungslogik, Neuwied: Verlag

Luchterhand.
Albert, H. (1977), ‘Individuelles Handeln und soziale Steuerung. Die ökonomische
Tradition und ihr Erkenntnisprogramm’, in H. Lenk (eds), Handlungstheorien –
interdisziplinär, 4, Munich: Fink, pp. 177–226.
Bennett, W.L. (1980), Public Opinion in American Politics, New York: Harcourt
Brace Jovanovich.
Berger, P.L. and T. Luckmann (1979), The Social Construction of Reality: A Treatise
in the Sociology of Knowledge, Harmondsworth: Penguin.
Blondel, J. (1990), Comparative Government, Hemel Hempstead: Philip Allan.
Ciompi, L. (1988), The Psyche and Schizophrenia. The Bond between Affect and
Logic, Cambridge, MA: Harvard University Press.
Dijk, T.A. van (1997), Discourse Studies: A Multidisciplinary Introduction,
Thousand Oaks, CA, US: Sage Publications, Inc.
Dörner, D. (1991), Die Logik des Mißlingens. Strategisches Denken in komplexen
Situationen, Reinbek bei Hamburg: Rowohlt.
Eysenck, M.W. (1984), A Handbook of Cognitive Psychology, London: Erlbaum.
Förster, H. von (1984), Observing Systems, Seaside, CA: Intersystems Publ.
Frey, B.S. (1983), Democratic Economic Policy: A Theoretical Introduction, Oxford:
Robertson.
Frey, B.S. and R. Eichenberger (1991), ‘Anomalies in political economy’, Public
Choice, 68, 71–89.
Frey, B.S. and R. Eichenberger (1994), ‘Economic incentives transform psycholog-
ical anomalies’, Journal of Economic Behavior and Organization, 23, 215–34.
Frey, B.S. and G. Kirchgässner (1994), Demokratische Wirtschaftspolitik: Theorie
und Anwendung, Munich: Vahlen.
Goffman, E. (1986), Frame Analysis: An Essay on the Organization of Experience,
Boston, MA: Northeastern University Press.
Habermas, J. (1984), The Theory of Communicative Action, London: Heinemann
Education.
Hayek, F.A. von (1983), Knowledge, Evolution, and Society, London: Adam Smith
Institute.
Hesse, G. (1990), ‘Evolutorische Ökonomik oder Kreativität in der Theorie’, in

U. Witt (ed.), Studien zur evolutorischen Ökonomik I, Berlin Duncker & Humblot
pp. 49–73.
Hirschfeld, L. et al. (eds) (1994), Mapping the Mind, New York: Cambridge
University Press.
Homann, K. (1980), Die Interdependenz von Zielen und Mitteln, Tübingen: Mohr.
Hutter, M. (1989), Die Produktion von Recht: Eine selbstreferentielle Theorie der
Wirtschaft, angewandt auf den Fall des Arzneimittelpatentrechts, Tübingen:
Mohr.
Jones, S.R.G. (1984), The Economics of Conformism, Oxford: Blackwell.
Koch, L.T. (1996), Evolutorische Wirtschaftspolitik: Eine elementare Analyse mit
entwicklungspolitischen Beispielen, Tübingen: Mohr.
Kuran, T. (1987), ‘Chameleon voters and public choice’, Public Choice, 53, 53–78.
Kuran, T. (1995), Private Truths, Public Lies: The Social Consequences of Preference
Falsification, Cambridge, MA, US, London, UK: Harvard University Press.
Lindblom, C.E. (1959), ‘The science of muddling through’, Public Administration
Review, 19, 79–98.
Lindenberg, S. (1993), ‘Framing, empirical evidence, and applications’, in
P. Herder-Dorneich et al. (eds), Jahrbuch für Neue Politische Ökonomie 12: Neue
Politische Ökonomie von Normen und Institutionen, Tübingen: Mohr.
Little, B.R. (1972), ‘Psychological man as scientist, humanist and specialist’,
Journal of Experimental Research in Personality, 6, 95–118.
Löffelholz, M. and K.-D. Altmeppen (1994), ‘Kommunikation in der
Informationsgesellschaft’, in K. Merten, S.J. Schmidt und S. Weischenberg (eds),
Die Wirklichkeit der Medien: Einführung in die Kommunikationswissenschaft,
Opladen, Westdeutscher Verl. pp. 570–91.
Luhmann, N. (1989), ‘Politische Steuerung’, in H.-H. Hartwich (ed.), Macht und
Ohnmacht politischer Institutionen, Opladen, Westdeutscher Verl. pp. 12–16.
Meier, A. and D. Mettler (1988), Wirtschaftspolitik: Kampf um Einfluß und
Sinngebung, Stuttgart: Haupt.
Meier, A. and T. Slembeck (1994), Wirtschaftspolitik: Ein kognitiv-evolutionärer
Ansatz, Munich: Oldenbourg.
Merten, K. and J. Westerbarkey (1994), ‘Public opinion and public relations’, in
K. Merten, S.J. Schmidt und S. Weischenberg (eds), Die Wirklichkeit der Medien:
Einführung in die Kommunikationswissenschaft, Opladen, Westdeutscher Verl.
pp. 188–211.
Molitor, B. (1988), Wirtschaftspolitik, Munich: Oldenbourg.
Mueller, D.C. (1989), Public Choice II, Cambridge: Cambridge University Press.
Neisser, U. (1976), Cognition and Reality, San Francisco: W.H. Freeman.
Nisbett, R.E. and L. Ross (1980), Human Inference, Englewood Cliffs, NJ: Prentice-
Hall.
Nüse, R. et al. (1991), Über die Erfindung/en des radikalen Konstruktivismus:
Kritische Gegenargumente aus psychologischer Sicht, Weinheim: Dt. Studien-Verl.
Oeser, E. (1987), Psychozoikum, Berlin: Parey.
Pöppel, E. (1985), Grenzen des Bewußtseins, Stuttgart: Dt. Verlagsanst.
Pöppel, E. und A.-L. Edingshaus (1994), Geheimnisvoller Kosmos Gehirn, Munich:
Bertelsmann.
Popper, K.R. (1995), The Open Society and its Enemies, London, New York:
Routledge.
Economic policy 189
Ross, L. (1977), ‘The intuitive psychologist and his shortcomings’, in: L. Berkowitz
(ed.), Advances in Experimental Social Psychology 10, New York: Academic
Press, pp. 173–220.
Rusch, G. (1994), ‘Kommunikation und Verstehen’, in K. Merten, S.J. Schmidt und
S. Weischenberg (eds), Die Wirklichkeit der Medien: Einführung in die
Kommunikationswissenschaft, Opladen: Westdeutscher Verl. pp. 60–78.
Scharpf, F.W. (1991), Crisis and Choice in European Social Democracy, Ithaca, NY,
London, UK: Cornell University Press.
Scherer, K.R., C. Vogel and K. Immelmann (1988), ‘Psychobiologie und Politik’, in
K. Immelmann, K.R. Scherer, C. Vogel und P. Schmoock. (eds), Psychobiologie,
Grundlagen des Verhaltens, Stuttgart and New York: Free Press, pp. 803–32.
Schiepek, G. (1990), ‘Selbstreferenz in psychischen und sozialen Systemen’, in:
K.W. Kratky und F. Wallner (eds), Grundprinzipien der Selbstorganisation,
Darmstadt: Wiss. Buchges.
Searle, J.R. (1995), The Construction of Social Reality, London, UK, New York,
US: Free Press.
Siegenthaler, H. (1994), ‘Wege zum Wohlstand: Das Beispiel der USA, der Schweiz
und Brasiliens’, in W. Fischer (ed.), Lebensstandard und Wirtschaftssysteme,
Frankfurt am Main: F. Knapp, pp. 173–212.
Simon, H. A. (1955), ‘A behavioral model of rational choice’, Quarterly Journal of
Economics, 69, 99–118.
Six, B. (1987), ‘Attribution’, in D. Frey und S. Greif (eds), Sozialpsychologie,
Munich: Psychologie-Verlags-Union.
approach to policy-making’, Constitutional Political Economy, 8, 225–54.
Sugarman, S. (1987), Piaget’s Construction of the Child’s Reality, Cambridge, New
York: Cambridge University Press.
Tversky, A. and D. Kahneman (1981), ‘The framing of decisions and the psychol-
ogy of choice’, Science 211, 453–58.
Tversky, A. and D. Kahneman (1982), ‘Causal schemas in judgements under uncer-
tainty, in D. Kahneman, P. Slovic und A. Tversky (eds), Judgement under
Uncertainties, Cambridge: Cambridge University Press, pp. 117–28.
Vowe, G. (1994), ‘Politische Kognition: Umrisse eines kognitionsorientierten
Ansatzes für die Analyse politischen Handelns’, Politische Vierteljahresschrift,
35, 423–47.
Wallis, J. (1997), ‘Conspiracy and the policy process: a case study of the New
Zealand experiment’, Journal of Public Policy, 17, 1–30.
Weick, K.E. (1986), ‘Organizations as cognitive maps: charting ways to success and
failure’, in: H.P. Sims et al. (eds), The Thinking Organization, Dynamics of
Organizational Social Cognition, San Francisco: Jossey-Bass.
Witt, U. (1992), ‘The endogenous public choice theorist’, Public Choice, 73, 117–29.
9. Why endogeneity is not enough
to explain technological change –
a critique of Paul Romer
Malcolm H. Dunn1
For ten years now, we have been experiencing a renaissance of growth the-
ories in the form of so-called ‘endogenous growth theories’. These claim to
supply us with an empirically substantial explanation of the driving forces
behind the growth-rate of an economy. On the one hand, proponents argue
that, in contrast to the older neoclassical growth theory of the Solow-type,
technological change is no longer treated as a residual value, calculated
by determining the difference between the growth of inputs to the empiri-
cally observed national growth of outputs, i.e. to the social product. Instead
of an assumed exogenous rate of technological change, i.e. a rate that is not
explicitly explained, we have a rate which is attributed to the intentional
actions of profit-maximising agents of an economy, i.e. an endogenous rate.
Another advance, proponents argue, is that endogenous growth theory
coincides better with the observable facts than does the older theory. While
the older neoclassical theory still concluded that per-capita growth rates
had a tendency to converge, explaining this among other things with the
public goods character of technical advances, newer theories assume that
technical knowledge is not ubiquitous. This is why a stable gap in per-capita
growth rates can occur, the closure of which gives rise to the need for eco-
nomic policy. At least some proponents of endogenous growth theory view
their reasoning as a justification for extensive state intervention in the form
of industrial and technological policy measures.
There is no doubt that endogenous growth theory is en vogue and this not
only because one of its founders was awarded the Nobel prize for his work
on rational expectations, but also because the deficits in the area of the
micro-economic foundations of macro-economics are felt more strongly
than was the case 30 or 40 years ago. In spite of or because of the public
appreciation of endogenous growth theory, it must be ‘permissible’ to ask
wherein the advancement of endogenous growth theories actually lies.
Addressing this question is the objective of this chapter. The present inquiry
190
Why endogeneity is not enough to explain technological change 191
will be made from the perspective of both evolutionary economics and

the very closely related New Institutional Economics, another school of
thought which has also experienced a renaissance in the past few years.
The present contribution is structured as follows: the actual inquiry is
preceded by a description of technological change from the perspective of
evolutionary economics (second section). The purpose here is not to review
the already numerous contributions to this school of thought, but rather to
develop questions which are important for the ensuing analysis. In the third
section one of the models which has received the most attention, namely
Romer’s growth model, ‘Endogenous Technological Change’, introduced
in 1990, is discussed. Again, the purpose here is not to presume to give an
exhaustive account of all endogenous growth theories, but to use the model
to illustrate some fundamental problems of endogenous growth theories.
In the ensuing fourth section, it will be demonstrated that, in spite of the
advances of this model in comparison to those of traditional neoclassical
growth theory, one cannot speak of an ‘endogenisation’ in the strict sense
of the term, because important aspects of the market, representing the
driving forces behind technological change, are ignored. The chapter closes
with some deliberations on the conclusions to be drawn from the deficits of
the Romerean growth model (fifth section). In this context, brief reference
to other endogenous theories of growth will be made.
TECHNOLOGICAL CHANGE FROM THE

PERSPECTIVE OF EVOLUTIONARY ECONOMICS
Both the perspective of evolutionary economics and the New Institutional
Economics assume ‘that the institutional and organisational structure of a
national economy and not merely market mechanisms must be considered
for resources and income to be distributed effectively’ (Grupp 1998, p. 66).
In addition to the question of how institutions are formed and how insti-
tutional change is to be explained, defendants of the evolutionary para-
digm dedicate themselves to explaining technological change. However,
instead of taking these contributions as a point of departure, another
approach will be chosen here. If it is assumed that technological change, as
the growth generator of an economy, is the product of the interaction of
various institutions, then it makes sense to first have a look at how these
institutions are characterised within the context of evolutionary economics
in order to then analyse the extent to which these same institutions are
modelled within the framework of newer growth theories.
Institutions, as they are understood here, are defined as a system
of deliberate (formal) and indeliberate (informal) rules including the
provisions for their enforcement. ‘Institutions together with the people

taking advantage of them are called organisations . . .’ (Furubotn and
Richter 2000, p. 7). The most important institutions involved in techno-
logical change are no doubt firms, private households and the state as an
intervening authority that sets up laws and implements economic policy.
How do these organisations come into the field of view of the analysis of
technological change?
The way in which institutional economics discusses the firm dates, as we
know, back to Coase’s contribution ‘The nature of the firm’ (1937). In this
article, Coase pursues the question of how the existence of firms can be
explained. On the other hand, firms as producers of technological change
are not considered. Coase does, however, introduce the concept of trans-
action costs, a concept unknown to neoclassical theory. The neglect of
transaction costs in neoclassical theory leads, according to Richter and
Furubotn (1998, p. 9f.), to the conclusion that the question of how pro-
duction is organised is unimportant. For this reason, the existence of firms
must remain unexplained. Not only this, but it is noteworthy ‘that, in the
zero-transaction-cost-world of neoclassical economics, the two economic
systems . . . , capitalism and socialism, can be modelled by general equilib-
rium systems’ (ibid., p. 14).
This assertion is also important for the explanation of technological
change. For it is not at all inconsequential for the explanation of techno-
logical change if the economic system in question is characterised by cap-
italist firms or not. The incentive for technical progress arises from the
firm’s objective to realise the economic benefit that technical progress
affords. The profit motive induces firms to continually reduce their pro-
duction costs by means of process innovations and to attract the ever lim-
itedly solvent demand to its own supply via product innovations. Both
processes take place, not by coincidence, but as a result of deliberate plan-
ning. Following Schumpeter’s earlier contributions, Freeman, e.g., empha-
sises that technological change is less something produced by obsessive
tinkerers working somewhere in a garage and more the result of the work
of professional research teams. Freeman explains the professionalisation
of innovation work by pointing to the growing complexity of scientific
and technological tasks and the increasing division and specialisation that
go with it.
Firms as innovators promote technological change out of self-interest
and not because the state demands it of them. Competition, however, does
serve as an additional driving force. Due to the fact that at least some
firms are actively innovative so that they can attain innovation rents, the
question of technical progress becomes, within the context of competi-
tion, an existential one: as soon as some firms succeed in minimising their
production costs, other firms will be forced to follow. As imitators they

contribute to the diffusion of innovations. The process is similar for
product innovations. Product innovations will be copied and improved by
other firms. Product markets develop which in turn go through a similar
cycle (see Heuss 1965).
A constitutive element for the appearance and diffusion of product and
process innovations is the uncertainty of market outcomes, as all true incal-
culable uncertainties are prerequisites for functioning markets.2 The indi-
vidual who gives the impetus for a technical innovation does not know
beforehand if the new process or product justifies the costs that were
incurred for it. Technological change is not the same as economic success,
the latter will only make itself felt on the market. This is the point where
another economic agent of the market economy is addressed, namely the
‘private household’. Just like the firm, the private household appears on the
market and decides if a technological advance is also to be considered an
economic advance.
Private households are also important in another way. They serve as
suppliers of the most important production factor, namely human capital,
acquire skills and knowledge during their working life and, by virtue of
their work, contribute to a continuous improvement of the implemented
technologies and processes. What then happens is what evolutionary econ-
omists call incremental technical progress: while the technological paradigm
as defined by Dosi is a stock of knowledge with public-goods character –
examples are provided by the research results of the natural sciences –
incremental technical progress is less fundamental. It is related to the
specific use of technologies and results mostly from step-by-step improve-
ments in efficiency due to learning effects.3 For this reason, Dosi also
describes this type of change as ‘localised technological progress’, the pace
and direction of which depends on the technical level that has already been
reached in the firm.4 This localised technological progress is important
especially because it is based on firm-specific know-how and with this, takes
on the characteristics of a private good.5
This differentiation of technical progress as a public or a private good
bridges the gap to another field of research within the framework of evo-
lutionary economics, namely to property rights-theory and with this, indi-
rectly to the role of government in the innovation process. As Richter
confirms: ‘Property rights matter!’ (1994, p. 13), arguing that control
rights are an incentive in a world in which there are latent transaction
costs. In connection with technical progress, the question arises as to
whether the innovator will also succeed in obtaining the profits of his
investment. This is by definition a problem for the type of technical
progress which can be viewed as a public good, because the non-rivalry
and non-excludability of public goods means that innovators cannot

exclude others from using their know-how. It is possible, when competing
firms imitate an innovation, that the innovator will find himself deprived
of his innovation profit. He even suffers a competitive disadvantage in
comparison to the competition, because of the costs he has incurred pro-
ducing the innovation. In this case, the market provides little or no incen-
tives for innovation. The institutional deficit of the market is only
attenuated by the creation of another institution, namely of the govern-
ment-regulated protection of patents. This may maintain the incentive
system for innovations, because it gives the innovator the opportunity to
realise monopoly revenues temporarily. However, since externalities
cannot be completely excluded by the protection of the patent, many
authors see it as the responsibility of the state to intensify its own efforts
in the production of new knowledge.
Evolutionary economists were not the first to object to this conclusion
concerning industrial policy by noting that technical progress can also take
on the characteristics of a private good. Furthermore, it is noted that the
existence of positive externalities does not automatically lead to a decrease
in the propensity to innovate.6 On the contrary, technological spillovers can
be related to synergy effects which lead to more, and not less, innovation
activity.7 From this it follows that the state’s role is not, as industrial policy
makers have repeatedly demanded of the state, that of promoting certain
technologies and sectors. What really matters is developing networks
between firms and public research institutions to promote the diffusion of
technical knowledge in a sector or between different sectors. (see Hanusch
and Cantner 1993, p. 35).
PAUL ROMER’S GROWTH MODEL (1990) – AN

OUTLINE
A common characteristic of the newer growth theories is that they attempt

to explain continuous economic growth without having recourse to exoge-
nous values. Instead, growth is explained by market participants’ prefer-
ences, especially derived from their savings, consumption and investment
behaviour and is, in this sense ‘endogenised’.
It not difficult to make out where the assumptions of endogenous growth
theories come from.8 First of all, new growth theories assume, just like the
older neoclassical ones, profit-maximising economic agents guided by ratio-
nal expectations. Second, it is also true for the new growth theories that the
goods and factor markets’ price mechanism perfectly co-ordinates the plans
of firms and private households. They are, without exception, equilibrium
models. Third, the question of consumers’willingness to forgo consumption

and to save is also a relevant question for the new growth theories. This
depends, assuming optimisation behaviour, on the profit-rate on invest-
ments and the time preference-rate of private households.
A key difference between the older and the newer growth theories is
related to the marginal productivity of capital. In traditional growth theory
it is assumed that the marginal productivity of the accumulative factor
sinks at a constant rate. For this reason, at a certain point it becomes foolish
for consumers to continue forgoing consumption. This is why they will
reduce their savings, with the consequence that net investments are not
made, so that the economy converges towards a zero growth-rate. Only the
exogenously introduced technical progress prevents this.9 In contrast,
endogenous growth theories are intent on identifying mechanisms and
factors that prevent the continuous decrease of the marginal productivity
of the accumulative factor and which in this way create an incentive for
market participants to forgo present consumption in favour of future
consumption.
The essence of the new growth theory lies therefore in the fact that the
marginal productivity of capital must remain constant, so that an endoge-
nous steady-state growth of per-capita income is possible.10 An endogeni-
sation of the growth-rate takes place in such a way that the factor identified
as the respective growth generator is not a given. On the contrary, produc-
tion factors are considered to be a result of investment decisions made for
economic reasons, i.e. the firms produce these themselves.
Figure 9.1 gives us an idea of Romer’s multisector model. Technical
knowledge is produced in the research sector. Romer calls this kind of
knowledge ‘designs’. These designs are needed to produce intermediary
products. These represent capital goods, necessary for the production of
consumption goods. In the centre of Romer’s model we have technical
knowledge, regarded as a public good with the capacity to grow bound-
lessly. An example would be natural laws which can be used by different
engineers in order to produce innovations. (The private-good character of
localised technological progress is still not considered in this model!)
Technical knowledge is produced in this sector. In order for this to take
place, the costs of producing this knowledge would have to be covered.
However, as Schreiter (1997, p. 95) points out, if perfect competition is
assumed, there is nothing left for covering the costs of non-rivalling knowl-
edge, unless excludability is possible. Excludability, though, is not compat-
ible with the assumption of perfect competition.
In order to solve the financing problem, Romer makes three assump-
tions. First it is assumed that the innovators in the research sector receive a
patent which is infinitely lived. This makes their knowledge marketable.
Consumer goods sector

HY, L Y
Y(HY, L, x) HYL x(i)1di
0
HYLAx1
↑↓ Intermediate goods are sold to the con-

sumer goods sector and used there.
Consumer goods which were not consumed
are at the intermediate goods sector’s dis-
posal and can be used as inputs.
Intermediate goods sector

Y Ki x or Xi
xi x 1Y(HiY, Li, xi)
↑↓ The research sector develops designs

and sells these to the intermediate
goods sector. This enables the research
sector to fully cover its costs.
Research sector
HA Designs
dAdt · HAA
↑↓ Research enlarges the public stock of

knowledge (general knowledge), which
represents a production factor of the
research sector. This makes research on
the whole more productive.
Public stock of knowledge
Notes:
Variables of the model:
A Quantity of designs, dA/dt increase in designs
HA Human capital of the research sector; HY of the consumer sector
L Simple labour
K Capital, measured in units of the end product
Y Standard consumer good
x Equilibrium quantity on the intermediate goods market
Productivity factor effective in the research sector
Quantity of end products ‘sacrificed’ for an intermediate product
Figure 9.1 Romer’s growth model ‘Endogenous Technological Change’

Each individual producer in the intermediate-goods sector makes a single

payment to purchase exactly one design and produces, using capital, an
intermediary good of a certain quality, which he then sells to the consumer-
goods sector. The second assumption is that intermediate goods producers
exert market power over consumption goods producers. Since the produc-
ers of intermediate goods have a monopoly on their products, they can
exercise their market power over the consumption market, which is com-
petitive, and realise a monopoly profit. However, they lose this quasi-rent.
This is because of the third assumption, namely that a large number of pro-
ducers of intermediate goods compete with each other to acquire licences
from the research sector. The research sector’s quasi-rents are spent
completely to obtain patent rights from that same sector. In this way, the
research sector’s costs for the production of knowledge are financed by the
sale of usufructuary rights for a discovery.
How do spillovers or positive external effects occur in this model?11
These occur because in the research sector new knowledge is not only pro-
duced in the form of designs but also in the form of general knowledge. In
spite of patents, research increases the public stock of knowledge. The
research sector produces a factor as a by-product, which it can, in a certain
sense, use for free. Due to this, research activity becomes cheaper and, with
this, the number of new intermediate goods increases, on which, according
to the assumptions of the model, the productivity of the consumer goods
sector depends.
PAUL ROMER’S GROWTH MODEL: A CRITIQUE

Richard Nelson (1997) posed, in a widely acclaimed article, the question of
how new the new growth theory is. He then proceeded to answer the ques-
tion by noting that many components of endogenous growth theory can be
found in older contributions to growth theory. What appears to be more
important than this finding, however, is the criticism he expresses against a
degree of formalism these models have. A formalism which rules out, from
the onset, being able to take important factors which determine technical
progress into account. In the following, this criticism will be discussed using
a few leading questions: First: Is technological change modelled in a way
suited to production theory? Second: Are market forces identified as the
motor behind technological change? Third: How is the endogenisation of
technical progress and economic growth achieved? In other words, is the
endogenisation successful and, if so, to what degree? Let us begin with the
first question.
Is Technological Change Modelled in a Way Suited to Production Theory?
Let us begin with a look at the production function (equation 9.1). The pro-
duction of consumption goods (Romer, for reasons of simplification,
assumes a single consumption good (Y)) takes place using the inputs
unskilled labour (L), skilled labour, i.e. human capital (HY) and intermedi-
ate or capital goods (x).

Y(HYL,x) HYL x(i)1di HYLAx1 (9.1)
0
The striking thing about this production function is the assumption that
skilled labour (HY) remains qualitatively unchanged while the production
process undergoes change because of the increase in capital goods input.
In other words, the fact is ignored that a change in the quantity of capital
goods also implies a qualitative change in the production process, which, in
turn, means other skills will be required, skills more appropriate to the
changed work process. In this equation, technical progress has no influence
on the quality of skilled labour. Progress in productivity takes place, one
could say, without so much as touching the human factor of production,
even though this same factor becomes more productive due to the increased
input of intermediate goods.12
Let us look at the last factor in the total economy’s production function,
capital or intermediate goods (x). It is assumed that every increase in
capital or intermediate goods also means an increase in output and that no
productivity relation exists between capital or intermediate goods. Both
assumptions represent a strange abstraction within the context of an expla-
nation for technical progress, because technical progress does not consist
only in the quantitative increase of those intermediate goods used for the
production of a consumption good but also in their qualitative change. In
addition, technical progress can make an intermediate good superfluous or
lead to the replacement of several intermediate goods by a new one.
A decreasing number of intermediate goods can therefore be just as much
an indicator of technical progress as the development of new intermediate
goods (see Hayek 1984).
The fact that – due to technical progress – old products are replaced by
new ones is reversed here: every capital good ever made retains its
effectiveness regardless of technical progress.13 In this way, technological
change appears as a mere increase of the capital goods that go into the
production process, the elimination of technically obsolete capital goods or
their replacement with more advanced capital goods does not take place.
Thus, a very important factor in technological change is left out.
Another objection refers to how capital goods implemented in the pro-

duction process come to develop their efficiency. Romer assumes that
capital goods are efficient independently. The implementation of a capital
good in combination with another or of several capital goods together has,
according to the assumption of the model, no effect on efficiency. Again, a
surprising assumption, since this means that capital goods are not com-
bined at all, but used separately. The difference between a production
process in which technical progress is taken into account and one in which
it is not is merely that, in the former case, more intermediate goods are
implemented.14
The qualitative relation between the human capital incorporated in
workers on the one hand and the quantitative as well as qualitative com-
position of capital goods on the other hand, all so important to the under-
standing of the production relation, are theoretically deleted in the
assumed modelling. This is, by the way, also true for the influence that the
technically changed production process has on the constitution of the end
product. In Romer’s model, the quality of the end product remains likewise
completely unaffected by technical progress.15
Now, one could raise the objection that the production function for the
whole economy is not the actual innovation of the Romer model and that
the more important thing is the way in which the accumulation and use of
technical knowledge is modelled. This addresses the production function of
the intermediate or capital-goods sector and the knowledge-production
function. The former is expressed in equation (9.2). For the purpose of sim-
plification, it is assumed that the production of the intermediate good
proves to be forgone consumption. After a design is purchased from the
research sector, a certain amount of the output of good Y can presumably
be transformed into a capital-good i. The consumption goods which are
not consumed are available to the capital-goods sector. In the correspond-
ing equation (9.2) this is expressed by saying that units of Y are needed
to produce one unit of capital good i.
Ki x. (9.2)
Let us address the question of how the technical knowledge that goes into
intermediate-goods production is produced. This technical knowledge is, in
contrast to human capital, the non-rivalling factor. This is justified by the
assumption that technical knowledge exists separately from every and any
individual and – as is common in endogenous growth theories – can grow
boundlessly. It is assumed, mainly for the purpose of simplifying the model,
that each unit of technical knowledge corresponds to exactly one unit of an
intermediate good. For this reason, technical knowledge can be described
numerically by the number of patents. The production relation behind the

production of technical knowledge is contained in equation (9.3).
dAdt HAA. (9.3)
As can be derived from this equation, a surplus of technical knowledge

results from a surplus of human capital working in the research sector, the
already available stock of knowledge A and a productivity factor ,
assumed to be constant, and which, when HA 1, indicates the relation

between the new design and the already available ones ( AA).
How well does the above cover the process of creating new technical
knowledge? If one leaves out the influence of human capital on knowledge
production, it follows from equation (9.3) that a ‘surplus’ of the existing
knowledge, regardless of the content, generally increases the production
of new knowledge, also regardless of the content. First of all, one must
object that technical innovations are, as a rule, path-dependent. They are
i.e. dependent on a certain set of scientific findings from the preceding
periods: Progress in technical knowledge in the aviation industry has
nothing to do with advances in cancer research or the pharmaceutical
industry. Second, technical knowledge depreciates constantly and, this, in
fact, not only in the economic sense, but also in the technical sense.
Advances in the entire stock of knowledge do not consist in a mere ‘surplus’
of ideas. It is often the case that new ideas replace old ones with regard to
their technical relevance. There is also a process of ‘creative destruction’ of
outdated technical knowledge within the natural and engineering sciences,
which, by the way, does not exclude the possibility that in individual cases
and within the context of a new paradigm, a forgotten concept will be
‘rediscovered’.16
The simplifying idea of conceiving the production of knowledge as a
function of the existing stock of knowledge finds an extension in the way
in which human capital is taken into consideration. Connecting the two
variables by multiplication means that no advance in technical knowledge
can be made without a bearer of that knowledge. Who would want to chal-
lenge this? What is striking, however, is the idea that the stock of knowledge
exists externally and separately from human capital or, expressed in another
way, that the stock of knowledge is not at all incorporated in human capital:
According to this specification, a college-educated engineer working today and

one working 100 years ago have the same (!) human capital, which is measured
in terms of years of forgone participation in the labor market. The engineer
working today is more productive because he or she can take advantage of all
the additional knowledge accumulated as design problems were solved during
the last 100 years. (Romer 1990, p. 84f.; emphasis: M.D.)
Romer’s specification contradicts all experience: even a forgotten patent

will be grasped by someone when its usefulness is rediscovered. In contrast,
an engineer’s great-grandfather, likewise trained as an engineer, would
have, because of his poorer knowledge, great difficulties grasping his
great-grandchild’s scientific findings because these were gained on the basis
of a much broader stock of knowledge. He would be forced to learn new
skills in order to understand current patents. In other words, the stock of
knowledge is precisely the central component of the qualification. One
finds oneself asking what the engineer’s level of knowledge consists in, if
not in the acquired knowledge of his trade?
With the above, we have gathered the elements of Romer’s production
function: advances in technical knowledge result, as if we were dealing
with a recipe, directly from the multiplication of a stock of knowledge
with researchers whose qualification does not include this knowledge at
all. Regardless of the problem of choosing the factors involved in the
achievement of technical knowledge and examining the way they interact,
the functional relation is not really of the kind described in Romer’s model.
The idea that advances in technical knowledge can be represented by a
steady monotone production function misses the characteristic feature of
technical progress and that of any and every advance in knowledge.17
Are Market Forces Identified as the Motor Behind Technological Change?
Process and product innovations increase the welfare of any society.

Process innovations make the labour necessary to produce a good cheaper
and increase the number and quality of capital goods, while labour remains
constant. In addition, they contribute to the creation and satisfaction of
new needs. What separates market systems from other systems is the driving
force of competition which forces its participants, especially the firms, to
constantly lower their prices and to develop new products in order to
attract limited solvent demand to their own supply. In this way, innovation
and its diffusion become a question of the continuing existence of a firm in
competition, an endeavour entailing a number of risks and imponderabil-
ities for market participants. Research has to be done where it crosses the
line into a zone of uncertainty, where patents are developed even when their
economic use is never realised because better or cheaper solutions exist. Yet
other patents make the production of new consumption articles possible
but these do not meet with adequate solvent demand.
In short, just like the large supply of goods, patents as well as those con-
sumer and intermediate goods produced with the help of new knowledge
are also confronted with competition and have to stand the test of time. It is
inevitable that there will also be losers in this process. It is not seldom that
production and research costs as well as costs for qualifying employees

prove themselves afterwards to be bad investments. On the other hand, it is
possible, in other cases, for firms to increase and expand their monopoly
position because of technical progress.
One finds very little of this in the Romer model: producers of new know-
ledge in the research sector develop new designs, which they can, without
any apparent complications, sell to the producers of the intermediate- or
capital-goods sector. Furthermore, every design obtains the same price.
Patents which turn out afterwards to be bad investments or are replaced by
other patents do not exist in this model. Future research findings, which
cannot be known beforehand, are modelled in such a way as if they were
already known. Both in the technical and the economic sense, certainty
concerning the production and exploitation of patentable knowledge exists.
Capital goods producers who buy individual designs and in this way
become exclusive suppliers of a capital good also compete with each other
in a surprising way.
The economic incentive to produce capital-goods, i.e. to achieve an extra
profit over interest on capital by means of a monopoly price, is completely
lost because of the intensive competition of new firms entering the market.
At equilibrium, monopoly rents are just enough to finance fix costs, used
to purchase designs. The research sector therefore completely assimilates
the monopoly profit of the capital-good producers. Finally, conditions of
perfect competition in the consumer-goods sector prevail anyway. All firms
produce the same consumer good, i.e. they do not compete with each other
by diversifying their supply.
The exclusion of the competition of firms that anticipate trends and those
that follow suit, i.e. those who take on, imitate or develop innovations and
in this manner bring about the diffusion of technological innovations, gives
rise to the question of how technical progress can be endogenised. Let us
therefore address the third question: How is an endogenisation of technical
progress and economic growth achieved in Romer’s model in the first place?
How is the Endogenisation of Technical Progress and Economic Growth

Achieved?
Growth and technical progress appear in Romer’s model in three sectors.

Let us have a look at each sector in turn. The growth of consumer goods
(Y), with skilled (HY) and unskilled labour (L) remaining constant, is exclu-
sively a result of the increased input of intermediate or capital goods. It is
assumed in addition that the quantity of intermediate goods of each and
every type i is the same. This means: the more qualitatively ‘new’ interme-
diate goods, without any change in their quantity, the higher the output
of consumer goods. What is striking here is that technical progress in

Romer’s model is oblivious to the replacement of worn and technically
obsolete capital goods. ‘New’ capital goods are simply placed side by side
with ‘old’ capital goods, which are, by the way, operated by the same number
of skilled workers. In a nutshell, both assumptions contradict the essence of
technological change: neither does the number and the qualifications of
workers remain untouched by technological change, nor does this change
consist in constantly adding new intermediate goods to older ones.
Growth in the consumer-goods sector is based on growth in intermedi-
ate or capital goods. This requires a production process, with the input of
capital goods and specially qualified workers – a process which differs from
that of the consumer-goods sector. This is also abstracted from in Romer’s
model, ‘to keep things simple’. Instead, it is assumed that the intermediate-
goods sector acquires designs from the research sector and, using the same
(!) technology used in the consumer-goods sector, produces capital goods.18
In other words, ‘the formal specification here describes the sector that
produces capital goods as a black box that takes final output in on one side
and gives capital goods out of the other side’ (Romer 1990, p. 81).
It is the assumption that the production function of the intermediate-
goods sector is equal to that of the consumer-goods sector which makes
possible the interpretation of the production of capital goods as forgone
consumption. One is reminded of a statement made by Hofmann (1971,
p. 269) about the Solowean growth theory when he points out what he calls
a peculiarity of that theory. Hofmann comments that one comes to wonder
how a loaf of bread as a single good can be invested or, vice versa, how a
certain kind of machine is to be consumed.
The objections made in connection with the capital-theory debates of the
last decades concerning the ‘well-defined’ macro-economic production
function seem to interest no one at all anymore. For this reason, it is pointed
out again here that capital goods and intermediate goods differ in quality,
which is why their factor-input ratios differ. Robinson (1953) and Sraffa
(1960) pointed this out in their critiques. Thanks to Garegnani (1970 and
1990) we know that, in the light of this objection, a ‘well-defined produc-
tion function’ is no longer tenable even when interpreted as a ‘parable’.19
In the Romer model, the increase in capital goods is due to an increase
in the input of designs, which are produced in the research sector. The
function decisive for growth in Romer’s model is, in fact, the knowledge-
accumulation equation (9.3). From this equation it follows that, as a factor
of production, the production of designs for one period goes into the pro-
duction of designs for the subsequent periods. The increase in designs
depends on the stock of designs of the preceding periods.20 The productiv-
ity of the research sector, reflected in the increase of designs, increases in
proportion to how knowledge was accumulated and more qualified workers

were employed in the preceding periods.21 However this does not suffice to
describe equilibrium growth.
Let us assume, e.g., that A in equation (9.3) is replaced by a concave A
function. This would mean that the marginal productivity of the human
capital in the research sector would not increase in the same proportion as
the designs. The larger the number of designs and the stock of knowledge,
the more qualified workers would leave the research sector in order to work
in the consumer-goods sector. In this way, the economy’s growth would be
weakened and, eventually, come to a halt. To prevent this in the analysis,
Romer assumes a linear relation: qualified workers’ marginal product in the
consumer-goods sector grows in the same proportion as the increment in
designs. This assumption of a linear relation is what makes unbounded
growth in this model possible. How is growth achieved in the end? Romer
(1990, p. 84) himself gives us the answer:
Linearity in A is what makes unbounded growth possible, and in this sense,

unbounded growth is more like an assumption than a result of the model.
The quote, placed at the end of the critical explanation of the model, is
revealing: unbounded growth is not a result of the inherent argumentation
of the model, but is simply a result of a scantily justified assumption of lin-
earity. Romer regards the assumption of linearity as ‘feasible’, ‘because
there is no evidence from recent history to support the belief that opportu-
nities for research are diminishing’ Romer (1990, p. 84). If one compares
Romer’s model with other endogenous growth models, one will find similar
linearity relations.
Linearity, however, is not exactly typical for technological change, as
long as the change is path-dependent. New technological paradigms in
certain areas of knowledge stimulate an increase in certain areas of
research and lead temporarily to a marked increase in patentable know-
ledge until the paradigm is finally exhausted. Characteristic for techno-
logical change are temporary increases in patents on the basis of a
technological paradigm, which then slowly decrease in number and impor-
tance. In many cases, new paradigms arise to replace the old ones, in turn
triggering a new wave of research activities. At the same time, older patents
are depreciated, both in the technical and in the economic sense. In yet
other cases and because of technical and economic developments, the par-
ticular relevance of ‘older’ patents will be rediscovered. In any case, the
assumption that technical knowledge, like every type of knowledge, is infi-
nite, in no way implies linear relations. Even interruptions and periods of
less innovation activity cannot be ruled out.
Romer’s model then, does not give us an ‘endogenisation’ of economic

growth, in the sense that proof is given as to why empirical evidence shows
that growth is not weakened and a convergence in the Solowean sense does
not occur. Romer admits this himself when he notes: ‘Whether opportuni-
ties in research are actually petering out, or will eventually do so, is an
empirical question that this kind of theory cannot resolve’ (1990, p. 84).
But what is left then of the claim to endogenisation, if the explanation
of technological change is based on a linear-relations assumption, which
has not been tested – what is more, would probably not withstand an empir-
ical test?
CONCLUSIONS OF AN EVOLUTIONARY
ECONOMIST
Models of endogenous growth theory are different from Solow-type

growth models in many ways, but not in the sense that they do not interpret
growth as a deterministic process. They remain deterministic even when, in
other endogenous growth models, the role of the ‘dynamic entrepreneur’ is
given more emphasis or the process of innovation is modelled stochasti-
cally.22 More research does not necessarily lead to more innovation.23
Modelling technological change as a deterministic process ignores the insti-
tutional context in which technical progress takes place: competition as an
open process which brings forth, and must bring forth, not only winners,
but losers as well.24 However, there is yet another reason why it is incorrect
to regard technical progress as a deterministic process. The reason lies in
the very nature of technological change: even before the evaluation of the
results of research and development is to be given thought to, those
same results would already have to be available.25 Exactly because of their
far-reaching informational assumptions, traditional models lead to an
erroneous characterisation of technological change.
The institutional deficits of the newer growth theory are of a similar con-
sequence. A definition of a growth equilibrium based on the assumption of
a deterministic process of knowledge accumulation is indifferent to the
underlying economic system.26 In actual fact, since market uncertainties
have no role to play in neoclassical growth models, these models resemble
more the ideal of a planned economy than that of a decentralised market
economy.27 If the whole point of technological change were to ‘produce’ it
with the help of the available stock of knowledge, then, instead of the
process of competition, one could have the figure of a central planner who,
in order to increase the stock of technical knowledge, would allot the avail-
able resources of a society to the research sector. It is not a coincidence that
even growth theorists confirm that the steady-state growth-rates of cen-

trally planned economies do not differ substantially from those of market-
economy systems.28
However, controversy exists concerning the conclusions to be drawn from
this criticism. As is generally known, Solow (1991, pp. 15ff.) uses the occur-
rence of true uncertainty, necessarily connected to technological change,
to argue against making technical progress endogenous and to ask ‘why
I should buy a car with so much horsepower’ – meaning the models of
endogenous growth theory – ‘to drive on such a dark and winding road’. In
the same article, Solow points to the fact that there is an ‘exogenous element’
to technological change in the sense that the result of human decisions and
actions have an inherent moment of true and incalculable uncertainty.29
Does this mean that science should refrain from investigating technological
change as an explanation for economic growth? Hardly! Even if it is true
that chance represents a constitutive element of technical progress and that
neither the technical nor the economic results of a research and develop-
ment investment are known a priori, this does not mean that we should
exclude the investigation of technological change. This is exactly what the
contributions of many evolutionary economists make clear.
Characteristic of these works, which stand in the Schumpeterean tradi-
tion, is their focus on the entrepreneur as the agent of the innovation
process. Instead of the firm as an input/output relation, we have the entre-
preneur who recognises and implements new opportunities on the economic
plane.30 Following Röpke (1982), innovative behaviour can be seen as a
function of rights of action, competence and entrepreneurial motivation.
Rights of action determine the degree in which one’s own actions are per-
mitted to infringe on others. These rights encompass every and any type of
normative restriction on action (laws, rules and regulations, norms and
sociocultural values). Since innovative possibilities of action do not neces-
sarily represent permissible possibilities of action, transaction costs ensue
for the entrepreneur, when they try either to enforce those rights which are
propitious for them or have to recompense someone for damages caused by
innovations. Both diminish the incentive to carry out innovations. But even
when rights of action promote the development and implementation of
technical innovations, innovative behaviour requires a certain degree of
competence. Innovative behaviour is tied to persons who have to have the
capability to develop or imitate new ideas and to carry these new ideas out.
These personality traits are both cognitive and non-cognitive. They include
abilities, aptitudes, experience, intuition and creativity.
Ability, however, must be paired with inclination, i.e. with ‘industry’
(Adam Smith). As everyone knows, agents’ motivation to act does not only
depend on personality traits. Motivation is a cultural phenomenon that is
influenced by social environment. As Röpke’s work has shown, which

causes are attributed to which results is not irrelevant for innovation behav-
iour. If the probability of an aspired-to goal is not to be influenced by one’s
own performance, there is no incentive to be innovative. Conversely, one
can observe that the incentive for innovative activity is particularly high if
carrying out an innovation is perceived to be neither trivial nor extremely
difficult. Comparative research on innovation behaviour in different Asian
countries supports Röpke’s speculation that the environment provides an
incentive for innovation behaviour when the challenges are in a middle
range of difficulty, i.e. neither too easy nor too difficult with respect to the
individual entrepreneurial ability (on this, see Röpke 1982).
Both Röpke’s deliberations, as well as those of other evolutionary econ-
omists, make clear that more attention has to be paid to qualitative factors
which, though not easily formalised, are nevertheless important for under-
standing the process of technological change.31 The conclusion which
Nelson (1997, 39) draws in his commentary on endogenous growth theory
can therefore also serve as the concluding remark to this chapter: ‘The
general conclusion suggested by this survey of the factors controlling the
vigor of enterprise is that a vast deal of emphasis must be placed on forces
that, in the ordinary conception of the bounds of economics, would have
to be classed as political, psychological, or sociological.’
NOTES
1. I thank Ulrich Fehl, Carsten Schreiter, Rainer Klump and Markus Braun for helpful
comments on the German version of this article, which was first published under the
title ‘Wachstum und endogener technologischer Wandel – Eine Kritik des
Wachstumsmodells von Paul Romer aus der Perspektive der Evolutorischen Ökonomik’
in ORDO (2000), 51, 277–99.
2. As is generally known, Morgenstern (1972, 1184) pointed out that the firm, as it is pre-
sented in the textbooks, could easily be replaced by a computer. The same author makes
reference to the fact that the competition of neoclassical equilibrium economics has little
to do with the competition firms have to deal with in the real world. On this point, see
also Dunn (1998).
3. These learning effects do not always take place on the knowledge frontier. Rosenberg
(1994), e.g., confirms that technical progress often only then ensues, when previous
knowledge is rediscovered and related to new contexts.
4. Innovation theory speaks in this context of the fact that technological change is path-
dependent and develops continually along so-called natural trajectories. Hanusch and
Cantner (1993, p. 24) point out that the more technological progress is based on firm-
specific know-how, the more one may characterise it as a private good and the easier it
is to appropriate its (temporarily accrued) rents. As self-enforcing mechanisms that
account for the path-dependencies, North (1990, p. 94) following Arthur (1988) men-
tions: ‘(1) large setup or fixed costs, which give the advantage of falling unit costs as
output increases; (2) learning effects, which improve products or lower their costs as their
prevalence increases; (3) coordination effects, which confer advantages to cooperation
with other economic agents taking similar action; and (4) adaptive expectations, where
increased prevalence on the market enhances beliefs of further prevalence’.
5. One of the findings of empirical innovation economics is that ‘The simple equation of a
private stock of intellectual property with total excludability and rivalry, contrasting
with a public stock of intellectual property without excludability or rivalry, (with respect
to the factor-embodied and disembodied knowledge), only holds in a few instances’
(Grupp 1998, p. 462)
6. The occurrence of external effects cannot be equated with market failure. They are rather
an ubiquitous characteristic of market economies. On this, see Sälter (1989).
7. This has to do with the fact that technical knowledge, in the form of acquired skills, flows
from one firm to the other when employees change their workplace or firms co-operate
in R&D projects.
8. See, for a description, Walz (1999).
9. This holds true even when, as in Solow’s model, a constant savings rate is assumed.
However, if the savings rate is dependent on the interest rate, the growth process will
come to a standstill much sooner.
10. As Ramser (1995, p. 238) correctly notes, increasing returns to scale are not necessary
for endogenous growth. It is necessary for one or a group of cumulative factors to exist
which can be produced with constant returns to scale and without the direct or indirect
use of non-reproductive factors. On the same topic, see also Schreiter (1997, p. 89),
Klump and La Grandville (2000) as well as Segerstrom (1999).
11. The relation between the introduction of technical knowledge and the occurrence of
increasing returns to scale can be illustrated in the following way. Let Y denote the output
of any type of good, X the input bundle and A the stock of knowledge, then Y AX.
From this, the relation aYaAX follows, when there is an increase in factor inputs by
the incremental factor a. If the stock of knowledge itself also now increases by factor a,
we have Y*a2AX, where Y* aY . This production function obviously exhibits
increasing returns to scale.
12. This also becomes clear when we consider the fact that human capital is treated as a
constant. The argument that workers cannot be trained ad infinitum is not valid because
the level of human capital is not measured primarily by the duration of training but
rather by the content and skills acquired during training.
13. It has to be noted here, however, that there are, in the meantime, endogenous growth
theories in which qualitative changes, e.g. in intermediate products, are modelled along
a so-called ‘quality ladder’. The improvements in quality due to technical progress cause
a decrease in the costs of consumer-goods production and an increase in per-capita
production (see also Grossman and Helpman 1991, Chapter 4). An endogenous model
of growth which treats, following Schumpeter, the case of ‘creative destruction’, has been
developed by Aghion and Howitt (1992).
14. Romer (1990, 81) expresses this as follows: ‘Equation (9.1) expresses output as an addi-
tively separable function of all the different types of capital goods so that one additional
dollar of trucks has no effect on the marginal productivity of computers. . . . The model
here considers the case in which all durables have additively separable effects on output.
An investigation of complementarity as well as of mixtures of types of substitutability
is left for future work.’
15. The remark that the quality of the end product remains unaffected by technological
change, does not refer to all endogenous growth models. As has already been mentioned
in note 13, the concept of quality ladders is an attempt to formalise the improvement of
goods as a result of technical progress. See also Maussner and Klump (1996, pp. 238–41
and 262–68).
16. Equating technical knowledge with patentable designs overlooks the fact that the know-
ledge process is also based on experimental research within firms. Essential impulses for
patentable knowledge result precisely from the engineering sciences’ analysis of the
industrial production process. It is true that Romer indicates that one could model the
research sector as a research department of the business sector. This is not enough
though, if the knowledge accumulation equation (9.3) is retained.
17. One has to agree with Solow (1991, p. 15f., and 1994, 51ff.), when he expresses criticism
aimed at some endogenous growth theories by saying that the scientific research findings,
be they patentable or not, are, for the most part, uncertain. Tietzel (1985, pp. 18f.) points
out this concept’s inconsistency by saying that this would require the predictability of
unknown nomological knowledge, which is, even if infinite information costs could be
lavished on it, totally illogical and impossible, because we would already have to be in
possession of that same knowledge. See also Lachmann (1976, pp. 127f.).
18. ‘The resources that would have been used to produce the forgone output are used instead
to manufacture capital goods’ (Romer 1990, p. 81).
19. Garegnani’s criticism (1970 and 1990) is aimed explicitly at Samuelson’s (1971) attempt
at interpreting the production function as a mere ‘parable’, in which the possibility of a
continual substitution between labour and capital is interpreted as a sign of a change
between different production processes for the production of a qualitatively unchanged
end product with respectively different capital goods.
20. Romer (1990, p. 84) expresses this in the following way: ‘The crucial feature of the spec-
ification used here is that knowledge enters into production in two distinct ways. A new
design enables the production of a new good that can be used to produce output. A new
design also increases the total stock of knowledge and thereby increases the productiv-
ity of human capital in the research sector.’
21. As has been mentioned above, this assumption contradicts a characteristic of empirically
observable technological change, which does not result from the already accumulated
stock of knowledge, particularly because advances in knowledge include the devaluation
of former knowledge.
22. In this sense, the already mentioned contribution of Grossman and Helpman (1991) as
well as that of Aghion and Howitt (1992), which follow Schumpeter’s ‘process of cre-
ative destruction’, must also be described as deterministic. Unfortunately, they cannot
be discussed here. A much more positive assessment of those endogenous growth theo-
ries which cite Schumpeter is given by Klump (1996).
23. On this topic, see especially Heuss (1965), Röpke (1970 and 1977) and Dosi (1988).
24. Competition means, as Morgenstern (1972) asserts ‘struggle with others, . . . fight, . . .
attempting to get ahead, or at least to hold one’s place’. One feels – according to
Morgenstern – absolutely nothing of this true character of competition in equilibrium
economics: ‘The contrast with reality is striking’.
25. Rosenberg (1994, p. 5) puts it succinctly when he says: ‘Instead of proceeding from a
natural starting point where firms possess little or no information and acquire informa-
tion through experience and investment, most economic models of technological change
assume that firms are aware of all the technological options available to them (leading
to a well-defined production function).’ It should be called to mind here that the strong
information assumptions are necessary. This is why Tietzel asserts that there can be no
surprises in equilibrium and Streissler (1980, p. 40) sums up that neoclassical theory has
always been, in principle, a theory of perfect information and certainty.
26. Nelson (1997, 33) expresses himself in a similar way: ‘The new neo-classical growth
models, . . . treat firms in a highly simplified way and barely address institutions, aside
from the “competitive” (or monopolistically competitive) market.’ In another contribu-
tion, Nelson (1994) himself has proven that organisational structures are relevant for
production and distribution.
27. Nelson (1997, 33) sees the continued adherence to the equilibrium approach as the
reason why uncertainties play no role in endogenous growth models. On the topic of
alternative market processes see also Lachmann (1976), Fehl (1986), Gutmann (1995),
and Schreiter (1997).
28. Now and then the necessity of industrial policy interventions is justified with the occur-
rence of positive externalities, which underlie the assumption of constant marginal
returns of capital inputs. However, this justification of the need for government action
rings hollow since decision makers are assumed to have doubtful preference structures
and an unrealistic information level. Weder and Grubel (1993, 491f.) point out that the
use of instruments of industrial policy is often due to particular interests which have
nothing at all to do with the occurrence of positive externalities. In addition, it is possi-

ble for private agents to internalise positive externalities without using government inter-
vention. The occurrence of positive externalities does not therefore necessarily justify
government intervention (see also Lucas 1988, 12, Ramser 1991; Maussner and Klump
1996, p. 287; Schreiter 1997, p. 105).
29. A. Wagner (1998, p. 244) sees in the quality of innovation behaviour, which he calls
other-than-average behaviour of individuals, the reason for the fact that an explanation
of technological change and with this, also the explanation of growth caused by tech-
nical progress necessarily remain ‘insufficient’ in macro-economic models. In the same
tenor, Krelle (1985, 214) comments that a great part of technical progress is undoubt-
edly coincidental and can only be taken into account via an exogenous value. Private
inventors, he goes on to say, experiment and try this and that, without systematic under-
standing – this is how the alchemists developed a basic understanding of chemistry,
Gutenberg developed the printing press and Lilienthal the airplane. See also H. Wagner
(1993, p. 88).
30. Entrepreneurial tasks include recognising the advantage of new combinations, planning
the necessary production process thereof, recruitment, training and direction of employ-
ees and the securing of funds to finance all of these measures. However, entrepreneurial
activity, as defined by Schumpeter, does not require entrepreneurs to do all these tasks
themselves – they can also delegate tasks. It is more important for the entrepreneur to
succeed in influencing and organising the process of innovation in such a way that a new
combination results (see Röpke 1982, p. 33).
31. Nelson’s objection to the new growth theories is that they rely on formalisation without
being aware of the dangers of an exaggerated formalisation: ‘In particular, there is a
danger that the goal of formalisation, per se, focuses efforts on understandings that are
relatively easy to formalise or on formalisations of these understandings that are rela-
tively easy but miss or deform important parts of those understandings.’ (1997, 54)
REFERENCES
Aghion, P. and P. Howitt (1992), ‘A model of growth through creative destruction’,

Econometrica, 60, 324–51.
Arthur, W.B. (1988), ‘Self-reinforcing mechanisms in economics’, in P.W. Anderson,
K.J. Arrow and D. Pines (eds), The Economy as an Evolving Complex System,
Reading, MA Addison-Wesley: pp. 9–27.
Coase, R.H. (1937), ‘The nature of the firm’, Economica, 4, 386–405.
Dosi, G. (1988), ‘Sources, procedures, and microeconomic effects of innovation’,
Journal of Economic Literature, 26, 1120–71.
Dunn, M.H. (1998), Die Unternehmung als ein soziales System: Ein sozialwis-
senschaftlicher Beitrag zur Neuen Mikroökonomie, Berlin: Duncker & Humblot.
Fehl, U. (1986), ‘Spontaneous order and the subjectivity of expectations: a contri-
bution to the Lachmann-O’Driscoll problem’, in I. Kirzner (ed.), Subjectivism,
Intelligibility and Understanding. Basingstoke, London: pp. 72–86.
Furubotn, E. and R. Richter (2000), Institutions and Economic Theory. The
Contribution of the New Institutional Economics, Ann Arbor, MI: Michigan
University Press.
Garegnani, P. (1970), ‘Heterogeneous capital, the production function and the
theory of distribution’, Review of Economic Studies, 37, 407–36.
Garegnani, P. (1990), ‘Quantity of capital’, in J. Eatwell, M. Milgate and
P. Newman (eds), The New Palgrave, Basingstoke: Macmillan, pp. 1–78.
Grossman, G.M. and E. Helpman (1991), Innovation and Growth in the Global
Economy, Cambridge, MA: MIT Press.
Grupp, H. (1998), Foundations of the Economics of Innovation Theory. Measurement
and Practice, Cheltenham, UK, and Lyme, USA: Edward Elgar.
Gutmann, G. (1995), ‘Wettbewerbstheorie und Wachstumstheorie: Chance für eine
Synthese?’, ORDO, 46, pp. 149–63.
Hanusch, H. and U. Cantner (1993), ‘Neuere Ansätze in der Innovationstheorie
und der Theorie des technischen Wandels: Konsequenzen für eine Industrie- und
Technologiepolitik’, in F.M. Kramer (ed.), Innovationsökonomie und Techno-
logiepolitik. Heidelberg: Springer, pp. 11–45.
Hayek, F.A. von (1984), Der Strom der Güter und Leistungen. Tübingen: Mohr.
Heuss, E. (1965), Allgemeine Markttheorie. Tübingen: Mohr.
Hofmann, W. (1971), Theorie der Wirtschaftsentwicklung. Berlin: Duncker &
Humblot.
Klump, R. (1996), ‘Wachstum und Wandel: Die “neue” Wachstumstheorie als
Theorie wirtschaftlicher Entwicklung’, ORDO, 47, pp. 101–20.
Klump, R. and O. La Grandeville (2000), ‘Economic growth and the elasticity of
substitution: two theorems and some suggestions’, American Economic Review,
90, 281–91.
Krelle, W. (1985), Theorie des wirtschaftlichen Wachstums. Berlin: Duncker &
Humblot.
Lachmann, L.M. (1976), ‘On the central concept of Austrian economics: market
process’, in E.G. Dolan (ed.) The Foundations of Modern Austrian Economics,
New York: University Press, Columbia, pp. 126–32.
Lucas, R.E. (1988), ‘On the mechanics of economic development’, Journal of
Monetary Economics, 22, 3–42.
Maussner, A. and R. Klump (1996), Wachstumstheorie. Berlin: Springer.
Morgenstern, O. (1972), ‘Thirteen critical points in economic theory: an interpre-
tation’, Journal of Economic Literature, 10, 1163–89.
Nelson, R. (1994), ‘What has been the matter with neoclassical growth theory?’, in
G. Silverberg and L. Soete (eds), The Economics of Growth and Technical Change,
Aldershot: Brookfield, pp. 290–324.
Nelson, R. (1997), ‘How new is new growth theory?’, Challenge, 90 (5), 29–58.
North, D.C. (1990), Institutions, Institutional Change and Economic Performance,
Ramser, H.J. (1991), ‘Industriepolitik und Wachstumstheorie’, in B. Gahlen,
H. Hesse and H.J. Ramser (eds), Wachstumstheorie und Wachstumspolitik: Ein
neuer Anlauf. Schriftenreihe des Wirtschaftswissenschaftlichen Seminars
Ottobeuren, 20, 305–27.
Ramser, H.J. (1995), ‘Wachstumstheorie’, in N. Berthold (ed.), Allgemeine
Wirtschaftstheorie, Munich: Vahlen, pp. 235–50.
Richter, Rudolf (1994), Institutionen ökonomisch analysiert, Tübingen: Mohr/
Siebeck.
Robinson, J. (1953), ‘The production function and the theory of capital’, Review of
Economic Studies, 21 (2), 81–106.
Romer, P.M. (1990), ‘Endogenous technological change’, Journal of Political
Economy, 98 (5), pp. 71–102.
Röpke, J. (1970), Primitive Wirtschaft, Kulturwandel und die Diffusion von
Neuerungen, Tübingen: Mohr.
Röpke, J. (1977), Strategie der Innovation, Tübingen: Mohr.
Röpke, J. (1982), Die unterentwickelte Freiheit: Wirtschaftliche Entwicklung und

unternehmerisches Handeln in Indonesien. Göttingen: Vandenhoeck.
Rosenberg, N. (1994), Exploring the Black Box: Technology, Economics, and
History, Cambridge: Cambridge University Press.
Samuelson, P. (1971), ‘Parable and realism in capital theory: the surrogate produc-
tion function’, Review of Economic Studies, 29, 193–206.
Sälter, P.M. (1989), Externe Effekte: Marktversagen oder Systemmerkmal?
Heidelberg: Physika.
Schreiter, C. (1997), ‘Humanvermögen und Wirtschaftsordnung: Konsequenzen der
neuen Wachstumstheorie für die Entwicklungsländer?’, in S. Paraskewopoulus
(ed.), Wirtschaftsordnung und wirtschaftliche Entwicklung. Schriften zu
Ordnungsfragen der Wirtschaft, 53, 72–119.
Segerstrom, P.S. (1999), ‘Endogenous growth without scale effects’, Amercian
Economic Review, 88 (5), pp. 1290–310.
Solow, R.M. (1991), ‘New directions in growth theory’, in B. Gahlen, H. Hesse and
H.J. Ramser (eds), Wachstumstheorie und Wachstumspolitik: Ein neuer Anlauf,
Tübingen: Mohr, pp. 3–17.
Solow, R.M. (1994), ‘Perspectives on growth’, Journal of Economic Perspectives, 8,
45–54.
Sraffa, P. (1960), Production of Commodities by Means of Commodities, Cambridge:
Streissler, E. (1980), ‘Kritik des neoklassischen Gleichgewichtsansatzes als
Rechtfertigung marktwirtschaftlicher Ordnungen’, in E. Streissler and C. Watrin
(eds), Zur Theorie marktwirtschaftlicher Ordnungen, Tübingen: Mohr, pp. 38–69.
Tietzel, M. (1985), Wirtschaftstheorie und Unwissen, Tübingen: Mohr.
Wagner, A. (1998), Makroökonomik, Stuttgart: Lucius & Lucius.
Wagner, H. (1993), Wachstum und Entwicklung, Munich and Vienna: Oldenbourg.
Walz, U. (1999), ‘Wissensakkumulation, endogenes Wachstum und Implikationen
für Entwicklungstheorie und-politik’, in R. Schubert (ed.), Neue Wachstums und
Außenhandelstheorie: Implikationen für die Entwicklungstheorie und-Politik.
Berlin: Duncker & Humblot, pp. 47–62.
Weder, R. and H. Grubel (1993), ‘The new growth theory and Cosean economics:
institutions to capture externalities’, Weltwirtschaftliches Archiv, 129, 488–513.
PART III
The political economy of complexity

10. Innovation and the learning policy
maker – an evolutionary approach
based on historical experience1
Joachim Schwerin and Claudia Werker2
1. INTRODUCTION
Economic policy should be based on an appropriate analysis of the eco-

nomic system it tries to influence and it should be carried out as much as
possible in a non-disruptive manner: if one was to propose guiding princi-
ples for the conduct of economic policy, these two statements would hardly
be contested. In particular the second principle reflects an evolutionary
policy approach that adopts a systemic perspective. Rather than disrupting
the internal working mechanisms, political actions should build upon exist-
ing systemic processes and modify them as little as possible. In practice,
however, such ‘harmonious’ policy making becomes the more difficult to
achieve the more turbulent the part of socio-economic life targeted by
policy proves to be. Especially in fields characterised by complex dynamics
shaped by continuous change and a multitude of overlapping causal
factors, sound policy making requires politicians to be able to learn.
Learning places them in a position to fit their measures to the current state
of the evolving socio-economic system at each point in time.
The field of innovation and growth constitutes a prototypical example of
such a dynamically changing part of economic life. In this chapter, we will
explore the possibility to design, implement and perform a learning inno-
vation policy. We start from the observation that learning cannot occur in
a world where everything is the result of pure chance. In such an environ-
ment, the past offers no guidance for present and future decisions. Learning
requires secured knowledge, and this knowledge depends on the existence
of general characteristics of economic and technological change as well as
the ability to identify these. This is difficult, as historical processes are not
deterministic. However, neither are they purely random. Therefore we will
argue that socio-economic change is contingent in the sense that what we
call history is the outcome of a complex mixture of regular behaviour
215
216 The political economy of complexity
based on rules and unpredictable events. If this holds true, economic policy
need not be haphazard; instead it can harmoniously build upon those pat-
terns of technological and organisational change that can be identified.3
Our analysis will proceed in four steps. We will first discuss in greater
detail the dynamically desirable framework for a learning policy from the
meta-point of view of the theory of economic policy. These methodolog-
ical considerations lead to the identification of types of rules on which
policy should be based if learning is to be institutionalised (section 2).
Building on these results we will then present a practically feasible concept
of learning policy. It depends on the identification of invariant patterns
within the recorded experience of the past and can be interpreted as a
modification of the traditional concept of stylised facts (section 3). In a
third step we will apply our method to one particular example, the
identification of invariant patterns of innovation and growth. This set of
stylised facts serves to describe the ‘regular’ part of socio-economic change
with regard to our concrete topic. In this context we will lay special empha-
sis on the question how open the set of identified patterns is towards a
changing environment or, in other words, what political learning can look
like in this context (section 4). These considerations lead to our fourth step,
in which we discuss how these results can serve as a reference system for
practical innovation policy. We will argue why we view our approach as a
sound basis for the conduct of an evolutionary innovation policy, and we
will discuss some policy guidelines that encompass the notion of continu-
ous learning (section 5). Short conclusions will serve to highlight aspects
for which we see the need for further research (section 6).
2. THE METHODOLOGICAL BASIS OF LEARNING

INNOVATION POLICY
Technological and organisational change is a complex phenomenon. It is
the outcome of a flow of innovations which overlap, interact and in any
case are hardly predictable. In the following it will prove useful to define
innovation in a broad way that encompasses its different forms. We will thus
define an innovation as the implementation of new combinations, which
includes product, process and organisational innovations as well as the new
access to markets of suppliers or consumers.4
The traditional linear innovation model depicts innovations as the
central part of a sequence of steps that follow one another in a linear
manner. This sequence starts with an invention resulting from basic
research and proceeds with the outcome of applied research, that is results
which are then used to market an idea. It is this second step that is usually
Innovation and the learning policy maker 217
called innovation. The last step consists of the diffusion of this idea and its
adoption by other economic agents.5
This linear innovation model is based on the technology-push approach,
which assumes that technological change is supply-driven. However, con-
vincing empirical results show that this linear innovation model does
not correctly reflect economic reality.6 The major reason is that the sim-
plifications within this basic model neglect feedback effects between the
different stages of this process. This is a severe shortcoming, as such effects
are indispensable for understanding the generation and selection of inno-
vations, because the demand side can considerably contribute to both.7 In
particular the problems faced by first-fast adaptors can lead to renewed
questions for basic or applied research. As a result innovation processes are
much more complex in practice.
This insight raises an important question: How can policy cope with the
complexity inherent in processes of technological and organisational
change? Answering this question is difficult, even more so in ever-changing
environments driven by a multitude of parallel processes. To tackle this
problem it seems advisable to first study in greater detail the nature of
dynamic change. Most notably one needs to clarify whether such change
consists of structural elements which obey general rules, whether it is
purely driven by random effects or whether a particular mixture of both
elements dominates. The stronger the rule-driven part turns out to be, the
more feasible becomes an analytical understanding of these processes,
which is a prerequisite for policy making.
In order to assess the characteristics of innovation-driven dynamic
change, we apply the concept of ‘contingency’. This concept originates
from sociology and has recently been introduced into the theory of system
dynamics.8 It is based on an analysis of the dynamic characteristics of
evolving socio-economic systems and encapsulates the idea that the com-
plexity of such systems can be reduced by the systematic evaluation of
past behaviour. Economic agents do not behave erratically, because selec-
tion mechanisms (such as markets) force them to behave in the ‘best pos-
sible’ way. As such selection mechanisms are the same for many individuals
and, moreover, prove to be relatively stable in time and space,9 different
individuals tend to behave in similar ways if faced with comparable
situations.10
Based on these principles one can formulate the hypothesis that similar
behaviour of individuals and groups of individuals – ‘organisations’ such
as firms or other decision-making bodies – translates into similar patterns
of the evolution of different socio-economic systems. The extent to which
this holds true is an empirical question. For the moment it suffices to leave
this extent open (we will however provide an answer in a specific context;
see section 4). We therefore define contingency as the phenomenon that, at

each point in time of a system’s evolution, the number of future paths is
greater than one but not infinitely large, because it is restricted by the fact
that each path must obey the causal logic of socio-economic dynamics.11
It follows from these considerations that the concept of contingency
replaces the antagonistic concepts of determinism and chance by stating
that every economic action which can be observed in reality is a mixture of
both. There exists a rule-driven element of behaviour (to whatever extent),
but random elements inherent in the concrete historical situation in which
this behaviour occurs exert an influence (again to whatever extent). For the
analysis of contingent processes it is now essential to understand that the
element which is based on rules (‘general laws’ or ‘guiding principles’) can
be identified by repeated observations, but that predictions for any individ-
ual historical process we encounter in reality are prone to errors, since the
element of chance interferes with the invariant part. And yet, as we will
demonstrate in sections 3 and 4, the existence of an invariant part offers a
good starting point for rule driven (and thus non-erratic) innovation
policy.
If one accepts that the conduct of policy based on general rules is in
principle possible, one needs to ask what kind of rules should be applied.
It is generally conceived that the conduct of economic policy can be based
either on per-se rules or on rules of reason.12 Per-se rules specify causal if-
then relations ex-ante. This has the advantage that economic agents know
the political reactions to the outcomes of their economic activities in
advance. Per-se rules thus create an expectation-stabilising framework of
institutions by reducing uncertainty. In addition they allocate economic
control in favour of private agents, as they largely reduce the discretionary
power of political boards, which have no opportunity to decide on market
results ex-post in a discretionary and therefore unpredictable manner. For
these reasons per-se rules are often considered as superior to rules of
reason. This particularly holds in the context of long-term economic
activities such as investment decisions linked to the marketing of innova-
tions. Here, a framework that stabilises entrepreneurial expectations is
indispensable.
On the other hand, a general disadvantage of per-se rules consists in their
inflexibility: the rules are specified at one moment in time and are then left
unaltered for a period during which the economic environment changes,
due to the innovation-driven dynamics of socio-economic change.13
Another problem consists in the complexity of socio-economic relations,
which might render impossible the ex-ante specification of all possible out-
comes and thus the formulation of government responses towards them.
Interpreted in this light a policy based on rules of reason might still be
considered as an advantageous alternative, as these rules offer politicians

or bureaucratic bodies the opportunity to decide for every individual case
anew which market result is ‘useful’ and thus allowed (given a set of polit-
ical goals) and which is not.
This view however neglects the information problem. As information is
embedded in markets, the substitution of inflexible per-se rules, which put
decentralised behaviour centre stage, by discretionary rules of reason,
which open the door for ex-post state intervention based on secondhand
information at most, loses its appeal. To put this critique of governments’
‘Anmaßung von Wissen’ in Hayek’s words:
If we can agree that the economic problem of society is mainly one of rapid
adaptation to changes in the particular circumstances of time and place, it would
seem to follow that the ultimate decisions must be left to the people who are
familiar with these circumstances, who know directly of the relevant changes
and of the resources immediately available to meet them. We cannot expect that
this problem will be solved by first communicating all this knowledge to a central
board which, after integrating all knowledge, issues its orders. We must solve it
by some form of decentralization.14
The theory of economic policy further underpinned Hayek’s rejection of

central planning, especially in the positive theory of regulation. A major
result is that even in a neoclassical world information-saving adaptive
central planning usually fails to establish a social optimum. More import-
antly, this co-ordination mechanism becomes the less satisfactory the more
goods and services are traded and the more changes in the socio-economic
environment occur.15 It follows that although decentralised decision
making is far from reaching a social optimum in every case, it becomes the
more convincing as an alternative the more dynamic and complex the eco-
nomic sphere proves to be.
It would however go too far to conclude that this critique based on
Hayek’s thinking denied any role for policy at all. As in classical econom-
ics, the role of policy lies in shaping framework conditions but not in
interference with the socio-economic process that emerges and evolves
within this framework. This applies for innovation processes as well (see
section 5). Here, such a framework is mainly necessitated by the existence
of positive externalities in knowledge accumulation and in the importance
of publicly financed basic research in the innovation process discussed
above.
Evolutionary economics broadly mirrored Hayek’s critique and stressed
that policy makers operate under constraints of imperfect information and
bounded rationality. For this reason they do not reach solutions that are
valid once and for all times. Politicians thus share the constraints faced by
individual consumers and firms, for instance in the conduct of innovation

policy:16
The evolutionary policy maker adapts rather than optimizes, and his central
concern is the innovation system, the operation of the set of institutions within
which technological capabilities are accumulated. . . . [J]ust as individuals
operate under the constraints of localized, imperfect and uncertain information
so does the adaptive policy maker.17
And yet, while this view on policy making seems appropriate in the light
of the information problem, it does not grasp the full story. There is no
need to restrict policy making to purely adaptive behaviour in the sense
that it should be strictly reactive to socio-economic change. Method-
ological considerations derived on the basis of Hayek-type arguments in
fact primarily serve as a demonstration of the comparative disadvantages
of political actions which violate the principle of the superiority of decen-
tralised decision making and selection of economic results. They do not
necessitate an adaptive approach. In terms of our discussion of per-se
rules versus rules of reason, the implication is that – while both have com-
parative advantages and disadvantages – from a system perspective the
deficiencies of rules of reason seem particularly severe. We will therefore
focus on a solution which encompasses a compromise between the two
that gives greater relative weight to ex-ante specified per-se rules. In order
to address the latter’s disadvantages we enrich their application with a
dynamic element: in order to avoid the increasing imbalances in time
between unchanging, rigid per-se rules and the continuously evolving
socio-economic process, we institutionalise a learning routine for policy
makers.
Why is the difference between adaptive behaviour and learning so
important and what does it imply? The answer is twofold. Firstly, policy
making can be interpreted as a transformation of – both old and newly
arising – knowledge on socio-economic change into incentives that
influence future behaviour of individuals and firms. It thus contains by
definition a forward-looking perspective, as it tries to shape future events.
Therefore it is backward-looking only to the extent that the accumulation
of knowledge (as a prerequisite of policy making) requires the learned
analysis of past experience. Secondly, policy is in itself best regarded as a
process of trial and error. The stimulation of innovation is subject to great
uncertainty, and policy can fail.18 It thus seems advisable to interpret policy
making as part of an experiment, a process of trial and error that serves to
accumulate knowledge. As a consequence, policy has an important func-
tion even if it fails – as long as politicians learn by their mistakes. ‘Learning’
refers to improved knowledge about the nature of change in the policy field
concerned, which is more than passive observation of change without an

understanding of its driving forces.
These considerations do of course not imply that the margin of political
error is unimportant. Quite to the contrary, one can hardly contest that the
margin for error should be minimised. In the following, we will argue that
political learning can achieve this aim if politicians change the set of rules
that govern behaviour if and only if scientific progress in economics has
advanced enough to enable a better understanding of the mechanisms that
drive socio-economic change (it is this ‘understanding’ that distinguishes
learning from adaptation).
In section 3 we suggest a concept of learning policy making that is
general enough to be applied in various contexts. Its centrepiece consists of
scientific learning by gaining improved knowledge of past experience
through the analysis of historical processes. Such analysis could for
instance be conducted by a panel of experts, who for the policy field in
question continuously evaluate scientific research on past processes for
which data exist. These experts then condense their findings into a set of
patterns of socio-economic change. This set describes the structured part
of socio-economic change outlined earlier in this section. It forms the
starting point for policy making, namely the formulation of rules that
govern the structured (and thus in principle predictable) part of the
behaviour of economic agents. An in-built feedback loop institutionalises
learning: If the continuous evaluation of the latest research leads to
modifications of the knowledge base by adding new knowledge or falsify-
ing past insights, the existing set of per-se rules needs to be changed as well.
This learning process ultimately bases economic policy on the best avail-
able knowledge of the socio-economic process to be influenced at every
point in time.
3. THE MODIFIED CONCEPT OF STYLISED

FACTS AS A BASIS FOR LEARNING
INNOVATION POLICY
We will now specify a practically feasible concept of learning policy making

that incorporates the preceding methodological considerations. The start-
ing point is the question how the analysis of historical processes can result
in new knowledge, which forms the foundation for learning. To answer this
question it is useful to assess one of the most commonly used research
methods in economics, the concept of stylised facts. It was originally intro-
duced into growth theory by Nicholas Kaldor when he tried to identify
invariant patterns between different historical growth processes.19
Kaldor postulated that every model should be able to explain the chara-
cteristic features of the economic process analysed as observed in reality.20
The problem of the scientist was to find that kind of abstraction which best
explained this observed reality; thus, he faced a problem of choice between
different theories. To find a solution Kaldor stated that the researcher had
to identify these characteristic features first. In a second step these features
could serve as a reference system for the evaluation of competing theories.
On this basis he could then choose the ‘best’ theory in a third and final step.
Kaldor summarises what has since become the ‘classical’ idea of stylised
facts as follows:
[T]he theorist, in choosing a particular theoretical approach, ought to start off

with a summary of the facts which he regards as relevant to his problem. Since
facts, as recorded by statisticians, are always subject to numerous snags and
qualifications, and for that reason are incapable of being accurately summa-
rized, the theorist, in my view, should be free to start off with a ‘stylized’ view of
the facts – i.e. concentrate on broad tendencies, ignoring individual detail, and
proceed on the ‘as if’ method, i.e. construct a hypothesis that could account for
these ‘stylized’ facts, without necessarily committing himself on the historical
accuracy, or sufficiency, of the facts or tendencies thus summarized.21
Kaldor thus suggests the identification of the invariant part of socio-

economic change (‘ “stylised” facts without . . . committing . . . on the his-
torical accuracy’) in the light of historical experience. However he remains
opaque on how to achieve this in practice. Even more unsatisfactory is the
fact that his concept turns out to be neither methodologically sound nor
operational.22 The main problem is that Kaldor provides no rules that
explain how to generate stylised facts. In his own work as well as in the pub-
lications of most scientists who used his concept, stylised facts are listed at
the very beginning of the analysis, but it is not explained how they were
derived.
Such an approach renders stylised facts useless. One of the key problems
is that such a way of conducting research offers the researcher (and also any
politician who bases his measures on it) broad space for manipulation.
Consider a scientist who developed a theory and tries to find out whether it
corresponds to the economic process in reality. Based on Kaldor’s approach
this scientist would formulate a set of stylised facts with which he was to
compare his theoretical results. If he did not lay open how he derived these
stylised facts, however, there would be no possibility to objectively check
whether they are methodologically sound or whether he preselected only
those ‘historical facts’ that fitted to his theory. In the latter case the analy-
sis would obviously become tautological. As there undoubtedly exist incen-
tives for scientists to (deliberately or not) suppress facts that contradict their
theories, it is important to ensure that they cannot be exploited, both in

theoretical research and for the purposes of policy making.
Based on a detailed critique of Kaldor’s approach one of us has there-
fore constructed a modified concept of stylised facts which offers strict
guidelines for the generation of such facts.23 This new approach for
instance emphasises the need to state on what empirical basis stylised facts
are derived and what methods are used to analyse the historical data, so
that any reader possesses all the pieces of information to reproduce these
results independently. Moreover the identification of stylised facts should
be systematically based on the broadest accessible amount of other experts’
research in the field in question. The scientist checks this research for
sufficient consensus. If such consensus on a given hypothesis existed, this
hypothesis would qualify as a stylised fact.24 As stylised facts have to be
open to scientific progress, they regularly have to be reassessed in the light
of the latest research.
Figure 10.1 summarises the procedure that leads to the identification of
a set of stylised facts. It is of particular relevance here that this modified
concept can be directly applied by panels of economic advisors, because the
use of stylised facts is not restricted to theory evaluation: they can also
serve as a foundation for economic policy, as their aim is to give a thorough
understanding of economic processes in a condensed yet detailed way. The
modified concept of stylised facts thus renders the practical conduct of a
learning economic policy possible.
How should a board of policy advisors apply this concept? The process
starts with the identification of a practical problem for which policy meas-
ures might need to be taken. This problem then has to be transposed into
a clear scientific question which is suited for tests on the basis of empirical
evidence. As most questions can be tested in various ways based on
conflicting theories, the choice of a single test method already includes the
implicit use of normative statements. As a matter of consequence one
therefore needs to avoid such an approach, which would directly lead to
political dispute (given that alternative procedures exist that could lead to
alternative results). As a consequence the modified concept of stylised facts
takes a different route and starts by using the broadest available amount of
experts’ knowledge on the relevant question. After the formulation of the
scientific question, a panel of scientists systematically collects and assesses
all empirical studies which currently exist for a given topic, deliberately
including varying approaches.25 This also implies that the panel bases its
analysis on a set of studies that contains all recently discussed variables and
hypotheses of the debated question.
The next step marks the core of the concept, the consensus analysis. For
all hypotheses put forward in these empirical studies, this analysis reveals
Practical problem
Formulation of
scientific question
Collection and assessment of question-related

empirical studies
Consensus analysis
Set of stylised facts
Applications
(theory evaluation, policy advice)
Figure 10.1 The modified concept of stylised facts26

the percentage of experts who agree to them. It is important to confine this

group of experts to the authors of the empirical studies analysed. In addi-
tion only topic-related statements in already-published material qualify as
‘opinions’. These requirements minimise observation effects and any direct
influence between policy advisors and scientists. Those hypotheses for
which a sufficient consensus exists constitute the set of stylised facts for the
topic in question.27 This set of stylised facts then serves as a reference
system for theory evaluation and – here the more interesting aspect – as a
basis for the guidance of economic policy, most notably the formulation of
policy measures.
The upward arrow in Figure 10.1 reflects the crucial aspect for a learn-
ing policy: the modified concept of stylised facts is dynamically open in the
sense that the facts – and thus the justification for any specific policy
measure – have to be regularly reassessed in the light of the latest empirical
research. This implies that always if new studies emerge, the question
whether a scientific consensus exists or not has to be raised anew. Provided
that this process of the generation of stylised facts is institutionalised and
will be performed in regular intervals, politicians’ learning about the topics
they have to decide upon can thus be institutionalised as well.
4. PAVING THE WAY FOR LEARNING INNOVATION

POLICY: AN APPLICATION OF THE MODIFIED
CONCEPT OF STYLISED FACTS
It is advisable to demonstrate potential strengths and weaknesses of a

concept – and any policy based on it – in the context of a concrete appli-
cation. The modified concept of stylised facts has been applied first in an
analysis of the dynamics of growth paths in transition economies.28 The
key results were firstly that this analysis revealed a surprisingly large
number of stylised facts on which consensus existed and, secondly, that
these results taken in their entirety described a coherent process of innova-
tion-driven growth. It was thus possible to use this analysis not only for the-
oretical purposes, but also to postulate concrete policy advice that satisfied
the methodological desiderata developed in section 2. In the following we
will first present the invariant patterns with regard to innovation and
change. We will then proceed by discussing how to use these patterns as a
reference system for innovation policy (section 5).
The question how long-term growth emerges within economies and what
politicians can do to foster this process is an interesting research topic not
only because of its political relevance, but also because of the existence of
a spectrum of competing theories. These include neoclassical approaches,
the factor-accumulation and the innovation-driven models of New Growth

Theory, Keynesian models, evolutionary approaches and institutional
theory frameworks. From an empirical point of view, numerous works in
the tradition of quantitative economic history have recorded historical
growth processes and have assessed a set of hypotheses that link potential
causes with the effects observed. In order to derive a set of stylised facts by
applying the method summarised in section 3, an evaluation based on the
use of bibliometric methods evaluated the hypotheses contained in all these
works and checked them for sufficient consensus to qualify as a stylised
fact.29 This approach followed the suggestion of Robert Whaples, who
defined speaking of a consensus if two thirds of all scientists agree on a
certain topic, and added the condition that a sufficient number of diverg-
ing methods had been used.30
If one applies this method to historical research on growth processes
published in refereed journals between 1988 and 1997, the set of stylised
facts contains nine results that fulfil the Whaples criterion. This set looks
as follows:31
1. The generation of growth is a regional phenomenon.

2. Growth is caused by a set of various factors, each of which proves to
be a necessary condition.
3. An unspecific accumulation of human capital does not prove to be
growth-enhancing. The accumulation of technological knowledge (as
a specific variant of human capital) can nevertheless foster growth in
the context of the generation of technological progress.
4. The generation and diffusion of technological and organisational
progress is a crucial element within this set of factors which cause long-
term growth.
5. The increase of demand triggered by geographical market expansion
fosters growth at least indirectly by inducing technological and organ-
isational progress.
6. Within the process of technological and organisational progress,
the growth effects of innovations are more important than those of
inventions.
7. Growth processes are initiated by a clustering of basic innovations and
are later maintained by a stream of minor improvements, which espe-
cially reflect learning effects.
8. Knowledge networks facilitate the continuation of already existing
growth paths. No general statement can be derived for the optimal size
of these networks.
9. The emergence of a long-term, self-sustaining growth path within a
so-far stagnating economy requires at least three decades.
The first two stylised facts describe general features of growth processes:
their regional embeddedness and the fact that they are subject to multiple
causality. This second aspect is particularly important for any policy based
on these results. Politicians need to avoid the trap of monism32 and devise
a consistent bundle of measures that reflects the interdependencies between
the individual causal elements which interact and mutually depend on one
another. In order to facilitate a systemic policy approach that takes account
of multiple causality, it is important to assess all candidates for such causal
elements.
This assessment leads to the third to eighth stylised facts. While no
scientific consensus emerged on the role of physical-capital accumulation
or labour, the empirical base gave rise to a negative statement on human-
capital accumulation. A clear majority of experts refuses the hypothesis
that an increase in the rate of human capital accumulation in itself creates
a growth effect. They point however towards an indirect effect. An increase
in human capital proves important if it is employed in basic or applied
research to generate innovation. As a consequence there exists a direct link
between structural variables (the stock of human capital) and flow vari-
ables (the rate of innovation). It is therefore important to take a closer look
at the process of technological and organisational change. This necessity is
underpinned by the result (which was achieved independently from the pre-
vious one, because it was based on a different empirical basis) that the gen-
eration and diffusion of innovation is an indispensable part of the bundle
of causal elements that shape growth paths over time.
We stated in section 2 that the demand side played a significant role in
the complex process of innovation and change. The fifth stylised fact
confirms this view. While a statement that increased demand generally
induces growth is too general to be agreed upon by most economists,
a more refined hypothesis meets the Whaples criterion: market expansion –
notably in geographical terms – fosters innovation, which in turn promotes
growth. Here again, as with regard to human capital accumulation, the
causal chain leads via the innovation process, which is the actual link
between demand and growth.
The following three statements assess this innovation process in greater
detail. It turns out that innovation has a bigger impact on growth than
invention. This is prima facie not surprising, given that innovation implies
the successful marketing of new solutions, whereas invention stresses the
technological dimension of novelty. However there is more to this state-
ment than that. A further implication is that transfer of knowledge
between regions or enterprises that primarily invent to regions or enter-
prises that primarily innovate (and imitate) has good chances of success in
terms of market results. In other words, for successful innovation to take
place it is not a necessary condition that the innovator is the inventor

as well.
The seventh stylised fact complements this view. Growth processes
evolve in several phases. Firstly, breakthrough innovations pave the way for
the emergence of new markets. If several come together a critical mass
emerges that pushes the economy on a growth path. This (supply-side) push
does not suffice, however, to keep the momentum. The longer the market
exists the more important become minor improvements, which are mostly
pulled by the demand side, to maintain growth. Radical and incremental
innovation, supply-driven push and demand-driven pull effects therefore
serve different functions at different points in time, but in essence they are
complements, not substitutes.
Knowledge networks support this process in time.33 Such networks pro-
pagate knowledge, and this diffusion seems to be so indispensable that the
degree of support for this stylised fact among experts is greater than for any
other one: There is virtually full consensus on the fact that a continuous
flow of technological and organisational improvements depends on the
existence of a full-fledged knowledge network or, in other terms, an inno-
vation system. Only the ‘optimal’ geographical size of such networks is dis-
puted. Surprisingly, a majority of scientists that conduct historical analyses
of innovation systems seem to favour an international dimension compared
to a national or regional one.34 This view is somewhat contradictory to the
emphasis within evolutionary economics on national or regional systems of
innovation.
The last stylised fact contains an important statement on the time dimen-
sion by emphasising the long-term nature of innovation-driven growth
processes. This result turns out to be of particular significance for policy
making. Processes of technological and organisational change that are to
exert macro-economic effects are time-intensive and cannot be artificially
accelerated by significant increases of resources. The reasons for this are
not fully clear. They can possibly be best explained by taking an institu-
tional approach, that is by analysing the time-frames required for the
changes of formal rules and especially informal behaviour. This behav-
ioural aspect would merit further analysis, as it is directly linked to the fact
that the accumulation of tacit knowledge is bounded by individuals’ limited
capabilities, which cannot be substituted by other (political) actions.
If one summarises all these findings by looking at them in their entirety,
it is striking that the features of dynamic growth processes described here
seem surprisingly consistent. Note that the research question was not to
explore the link between innovation and growth, but between any perceiv-
able causal factor – labour, factor accumulation, technological change,
infrastructure provided by the government, institutions and others – and
growth. It is therefore a strong result that the mechanics of growth

processes hinges on a complex yet clearly discernable innovation process,
while similar statements cannot be made for alternative factors.
Equally intriguing, the set of stylised facts mirrors recent advances eco-
nomic thinking and thus underlines the openness of our method. It is
impossible to conceive that 30 years ago a bibliometric analysis such as the
one described above would have created the same results. Quantitative his-
torical research on growth processes started from scratch in the early 1960s
and slowly built up the tools to carry out research that looked inside the
‘black box’35 of endogenous growth generation. This was a prerequisite for
the identification of patterns of change as described here. It is thus clear
that the modified concept of stylised facts is indeed historically open in the
sense that new scientific evidence leads to a reassessment of old knowledge:
new research on a given hypothesis can either increase or decrease consen-
sus, and if this affects whether the Whaples criterion is met or not, the set
of stylised facts changes as well. If such changes were to occur, it is only
logical that any policy that bases itself on these ‘facts’ would have to be
reconsidered as well, and political learning would occur.
From a theoretical perspective, these results highlight the relevance of
approaches that endogenously explain the generation and diffusion of
knowledge. Such approaches primarily include evolutionary theorising on
systems of innovation and the innovation-driven models within New
Growth Theory (including neo-Schumpeterian models, which incorporate
some principles of evolutionary economics). Furthermore, economic
thinking on institutional change provides valuable additional insights.
From a political perspective, the stylised facts discussed here provide start-
ing points for an innovation policy approach that bases itself on clear
per-se rules whilst using the systemic nature of innovation processes. The
following section further explores the possibilities to conduct such policy.
5. IMPLICATIONS FOR INNOVATION POLICY IN

PRACTICE
The final step within the flow chart depicted in Figure 10.1 consists of the
formulation of policy advice based on the set of stylised facts identified
before. As indicated in the previous section, the fact that the patterns of
innovation and change are complementary in the sense that they mutually
reinforce one another facilitates this task. It is therefore possible to inte-
grate them into a common reference system for a consistent economic
policy, which proves to be in line with the systemic approach adopted by
evolutionary economists in their analyses of innovation systems. Although
this great extent of consistencies is not a prerequisite for the conduct of

policy, it is an important asset that increases the likelihood of success.
Before we come to more specific policy implications, we find it instruc-
tive to briefly assess the main characteristics of the evolutionary approach.
Although evolutionary economics cannot be considered as a well-defined
paradigm, a number of features combine to distinguish this approach from
theoretical alternatives.36 First of all, it provides a systemic approach
towards innovation and change that builds on the analysis of a population
of entities and their interaction among one another as well as with their
environment. This approach is aptly summarised as the variation-selection
paradigm: it builds on two mechanisms, one of which is the creation of
differences (the innovation process) and the other one the choice between
the different outcomes on the systemic level.
This emphasis on the differences between economic agents can be
considered as a second characteristic of evolutionary economics. Such
differences are crucial for competition and change, because heterogeneity
spurs innovation in the wake of individuals’ attempts at coping with their
environment.37 While many agents will display similar behaviour provided
that their sets of objectives do not differ too much, some will deviate from
the average and experiment. While their primary goal is survival in their
market, that is to fulfil the external constraints imposed upon them by the
selection mechanism (thus their behaviour is in line with our discussion of
contingency in section 2), there are several reasons why they innovate.
These reasons include differences in information, differences in the percep-
tion of their relevant environment and how it works and, last not least, the
simple fact that the tacit nature of some knowledge leads to individualised
solutions to general problems. It is important to endogenise this diversity –
and its useful economic function – into any discussion of policy making.
Thirdly, evolutionary economics views endogenous innovation processes
as the driving force of change.38 The fourth, fifth, sixth and seventh stylised
facts mirror this perspective as they point at the importance of technolog-
ical and organisational change for growth and thus fill the ‘black box’ of
growth generation with life. Even more importantly they show that the
sources of innovation change in time. Combined with the ninth stylised
fact, which reveals that processes of technological and organisational
change take several decades to bring about macro-economic effects, these
results imply that it is indispensable for policy to take a dynamic, long-term
approach. Any short- or medium-term actions can influence the way
change occurs, but ultimately not its result. Lasting success requires a
lasting, yet not too rigid, approach.
Two other features of evolutionary economics combine to provide further
insights. Evolutionary theorising takes into account the institutional and
organisational context within which production processes take place,39 and

it focuses on the role of knowledge, as it models the difficulties to find, trans-
fer and store knowledge.40 Again, the empirical findings on innovation and
change support these views. This becomes particularly clear from the eighth
and third stylised facts. The interplay between different economic agents
needs an institutional framework in which it operates. Knowledge networks
provide this framework and can be considered as the core of any innovation
system. While stylised facts on the macro-economic level are for the moment
too aggregated to allow for a detailed assessment of the working mecha-
nisms within such networks, they point at specific elements. For instance,
‘traditional’ inputs for growth, such as human capital, need to be embedded
in the innovation processes within such networks in order to create effects.
This is the essence of the third stylised fact.
The five characteristics of an evolutionary systems approach depicted
above provide a foundation for policy making on which more specific policy
suggestions can be based. For instance, the lack of empirical support both
for monocausal solutions and for unspecified increases in total factor input
imply that common government expenditure programmes will not work.
Any untargeted increase of demand or supply factors cannot be considered
as corresponding to the results of the above analysis. Targeting factors such
as human-capital accumulation can only be successful on a sectoral level in
the context of knowledge creation. As an example, political measures could
foster learning effects by improving on-the-job training. At the same time,
targeted support for decentralised information exchange within knowledge
networks, for instance through scientific or sectoral engineering societies,
access to trade fairs and local or regional gate keepers will complement
firm-specific efforts at individual and collective learning.
In this context it is important to note that the sources of knowledge
change over time in the evolution of a market. These sources include basic
research activities as well as the experience and needs of suppliers of inputs
and buyers of output, but also more indirect spillover effects. However not
all sources are of equal importance at each point in time. In particular the
seventh stylised fact hints at the fact that there exist stages within the
dynamics of sectoral change, which are largely caused by the different
stages within innovation processes as described in section 2. This finding is
in line with the evolutionary product-life-cycle approach, which distin-
guishes between different market phases. As a consequence policy measures
derived from an evolutionary perspective need to mirror these changes
in time.41
In new markets, for instance, the construction of a suitable institutional
framework is of key importance. Because of their greater legitimacy and
better co-ordination abilities, policy makers can quickly provide a set of
general formal rules that reduce transaction costs.42 At the same time,
policy must not attempt to select market outcomes as long as there do not
exist any stylised facts that support specific selection criteria. This means
that policy support should not be dependent on strict criteria that can only
be met by a limited number of economic agents, such as support only for
firms that possess an R&D department.43 The usual critique of untargeted
support does not apply in very early market phases, because the enlarge-
ment of technological and organisational variety is a crucial prerequisite
for the selection of outcomes by the market later on and thus for change
and growth. It therefore reflects the systemic approach described above.
In later stages of market evolution, factors such as increasing returns to
scale in production unfold their effects and lead, in connection with the
stabilisation of consumer tastes, to the reduction of variety and, ultimately,
a certain standardisation of product designs. Routinisation of production
processes and greater formalisation of the organisation within firms and
their supporting innovation system support this maturation. Here as well
these processes need time. Firstly, institutional settings that for instance
shape the flow of information within innovation systems depend not only
on formal laws, but at least as much on informal conduct based on trust
and shared objectives. This process cannot be accelerated without limits, as
it is linked to the change of attitudes within the persons involved, which is
a slow process. Secondly, minor improvements gain greater relative weight
in the innovation process. As their sources differ from those of the bigger
innovative steps that usually coincide with the emergence of new sectors,
the focus of attention shifts as well, towards the diffusion of best practice
and the creation of joint standards.
In the end, markets consolidate and become mature. At this stage, any
policy that aims at fostering innovation reaches its limits. Well-established
ties between the different market players and relatively low uncertainty may
now provide an environment with a functioning market mechanism in
theory that in practice becomes prone to collusion. In other words, the
focus of political attention shifts from fostering the ability to innovate
towards fostering incentives to innovate, which basically means preventing
anticompetitive behaviour. Here, for instance, competition policy might
turn into an effective tool. This insight highlights the need for a policy
mix that goes beyond the realms of a single policy field and calls for
co-operation not only between market players, but also different political
bodies within and beyond innovation systems.
And yet, no matter how sophisticated the design of such a dynamic
policy mix might be, policy based on this systemic approach can fail, too.44
What is important to note, however, is the function of such failures within
our approach. Failures are in themselves an element of learning policy,
because they contribute to the stock of historical experience and, as a

matter of consequence, to a better understanding of socio-economic
change. This particularly holds true if every policy measure is evaluated
independently. Evaluation may accompany the policy measure or may be
executed ex post.45 While we do not focus here on this issue, which goes
beyond the basic elements of the concept of learning policy as discussed
above, a few words on it will conclude this overview.
The aim of an evaluation is to find out whether the policy measure sets
the right incentives or adequately influences more directly the targeted
group of economic agents according to the policy aims as specified ex ante.
It thus includes an assessment of intended and unintended effects of any
policy measure. It also contains an efficiency control of the implementation
of the policy measures. The stylised facts derived in section 4 only give
hints, but do not provide a full picture, on where policy measures are
promising and where not. Current political practice does not base itself on
consistent standards for evaluation, and some measures are not evaluated
at all. If improvements could be made in this area, for instance through
greater international exchange of experience and co-ordination, the results
of policy evaluation would turn into a powerful tool for improvements of
the empirical basis on which future decisions are based. If innovation
policy is performed as a learning policy, even failures may be a part of its
success. In such a case, the main requirement for successful innovation
policy would not be to exist without any failures, but to identify these fail-
ures as soon as possible, avoid them as much as possible in the future and
store them in the policy knowledge stock.46
6. CONCLUSIONS
This chapter discussed a coherent approach towards the design, imple-
mentation and practical conduct of learning policy. Learning policy is a
method of conducting policy that intelligently builds upon the best avail-
able knowledge at each point in time and uses this knowledge to dynam-
ically modify the set of rules that govern economic behaviour. After an
in-depth discussion of the methodological requirements for such a policy
approach and the development of a practically feasible concept, we focused
our attention on innovation policy to demonstrate how this concept could
be applied in practice. We chose the topic of innovation because of its
pivotal role for socio-economic change and because it is a truly dynamic
process. This particularly enabled us to highlight the openness of our
approach towards changes in the knowledge stock of society. Of course,
this concept can be applied to other policy fields as well.
We consider this approach as a first step towards improving the conduct

of practical policy. However, further steps will have to develop important
aspects in greater detail. It seems of particular importance to carefully
analyse the incentives and incentive effects that govern the interaction
between policy makers and scientific boards. Traditional ways of conduct-
ing economic policy often allocate too much discretionary power to polit-
icians, who abuse their powers by putting their private agenda first. Political
boards normally tend to facilitate such unproductive behaviour: experts
often disagree among themselves to such an extent that politicians have too
much room for manoeuvre. One reason results from the fact that an adviser
who holds a political function has incentives for expressing his or her views
in a way that differs from what he or she would publish in a purely scientific
context. Another problem consists in the lack of mechanisms that trans-
form the divergent views of (politically elected) board members into a joint
position of the board itself.
The concept of learning policy reduces these problems by providing
filters. Firstly, one layer of discretion vanishes as only previously published
material that fulfils an objective criterion – publication in refereed jour-
nals – is taken into account regardless of who the author is. Secondly, the
Whaples criterion offers a clear-cut rule for what can be interpreted as
‘knowledge’ and what not. As a result the scope of advisors’ discretion
further decreases. The board of advisors thus becomes enabled to provide
a filter in the sense that the current state of knowledge in any particular field
is defined as a scientific consensus, not a political one. This filter takes much
discretion out of economic policy and minimises effects of lobbyism – at
the very least it will become more difficult for politicians to deviate from
such advice than before.
We are totally aware of the fact that boards of advisors already exist in
many fields of policy and that their composition and role differs. However,
we do not think that it is unfair to state that the working mechanisms of
these boards usually do not meet the stringent criteria set above. As a con-
sequence they leave too much room for discretionary political decisions.
The way forward we would suggest is therefore not the abolishment of
boards of advisors or immense institutional changes, but rather an evolu-
tion towards more stringency in the internal working mechanisms of such
boards. Indeed, already-existing boards may easily use the modified
concept of stylised facts. Whether they have the incentives to do so and
whether the ensuing results would meet the expectations are important
questions that merit further research.
Last but not least the performance of economic policy in itself needs to
be evaluated. By definition policy will have discretion even if our approach
is applied – experts can exert pressure by presenting unified and more
scientific views, but they cannot force politicians to follow their advice. It
thus remains perfectly possible that unintelligent policy prevails, that is
policy that fails to learn from past experience. Only the political sovereign,
the electorate, could effectively sanction such behaviour. It is a question
that still needs to be explored in greater detail which metamechanisms
might be introduced to make ‘unintelligent’ political behaviour, that is
the refusal to learn, costly enough for politicians to trigger changes of
behaviour.
NOTES
1. We are grateful for helpful comments by Friedrich Breyer, Uwe Cantner and Stan
Metcalfe on earlier versions of this paper. The usual disclaimer applies.
2. Corresponding author: Dr Claudia Werker, Max-Planck-Institut für Ökonomik,
Abteilung Evolutionsökonomik, Kahlaische Straße 10, D-07745 Jena, Germany; email:
werker@econ.mpg.de
3. Economic agents’ as well as politicians’ expectations of such patterns may turn out to be
wrong. However, the concept we propose below takes this into account by means of a
feedback loop that translates failed expectations into new knowledge.
4. Cf. Schumpeter (1911/1987), pp. 100f.
5. See, for example, Steinmueller (1994), p. 54, or Fritsch (1991), 198.
6. Cf. e.g. Grupp (1997), pp. 17–20.
7. Cf. Lundvall (1992).
8. See Fulda, Lehmann-Waffenschmidt and Schwerin (1998), pp. 340ff.
9. This statement applies if one takes a functional approach, that is if one does not focus
on superficial historically grown differences between different laws, customs and so on
that govern behaviour but on the (economic) function these institutions perform.
10. For a profound discussion of the concept of contingency, cf. Schwerin (2001), pp. 75ff.
Also note that traditional economics takes this insight to the ahistorical extreme by mod-
elling identical economic agents (‘homo oeconomicus’), whereas chaos theory turns out
to be blind in the other eye by denying any similarity between different individual actions
and the respective outcomes on the system level.
11. Cf. Schwerin (2001), p. 79.
12. Cf. Schmidt (1996), pp. 148ff.
13. A good example is the German antitrust law, which is altered every three to nine years
but nevertheless hardly keeps pace with the changes of competitive behaviour in the
economy; cf. Schmidt (1996), pp. 155ff.
14. Hayek (1945), 524, emphasis in orginal.
15. See Bernholz and Breyer (1984), chapter. 5.
16. Cf. Hanusch and Cantner (1993); Metcalfe (1995); Metcalfe and Georghiou (1997).
17. Metcalfe (1995), p. 418. See also Teubal (1997).
18. For approaches that discuss such a systemic view of policy; cf. Metcalfe (1995); Lipsey
and Carlaw (1996); Metcalfe and Georghiou (1997); and Teubal (1997).
19. Kaldor widely used ideas first developed by Arthur Okun; cf. Kaldor (1985), p. 8.
20. For an introduction into Kaldor’s concept, see Kaldor (1968), pp. 177ff.
21. Kaldor (1968), p. 178.
22. See Schwerin (2001), pp. 92ff.
23. Schwerin (2001), pp. 98ff. Parts of the methodological requirements are omitted here in
order not to overload this introduction to the new concept.
24. Given that, for a certain topic, several mutually exclusive and contradictory theories
exist for which no metacriterion exists which states ex ante what theory is ‘true’, such
a consensus is the only way to base practical actions on a sound scientific foundation.
The absence of such a metacriterion in science is obvious, thus the use of stylised
facts as a reference system for theory and policy evaluation serves as a second-best
mechanism.
25. The use of quality filters is advisable, as this restricts the studies to those which have been
put forward in a period of for instance the last ten years. Such a filter could consist in
the exclusive use of publications in top-ranked scientific journals.
26. Cf. Schwerin (2001), p. 114.
27. The consensus analysis becomes even more methodologically sound if it considers only
those hypotheses which have been tested in several studies by using different methods.
28. Schwerin (2001), pp. 133ff.
29. On the comparative advantages of bibliometric research compared to questionnaires
and other direct means of assessing experts’ views, cf. ibid., pp. 118 ff.
30. Cf. Whaples (1995), 139.
31. See Schwerin (2001), p. 179.
32. Landes (1994), 653.
33. For a more detailed discussion of this stylised facts and the implications thereof
cf. Schwerin and Werker (2003), 397ff.
34. Cf. Schwerin (2001), p. 174.
35. Cf. the approach taken in Rosenberg (1994).
36. As indicated in section 4, these features may also be part of other theoretical approaches,
such as innovation-driven models within New Growth Theory.
37. Cf. Metcalfe (1995), p. 418.
38. Cf. Nelson (1995), 67.
39. Ibid.
40. Cf. for instance Nelson and Winter (1982), pp. 76–82.
41. Cf. the following Werker (2003), pp. 285–90.
42. Cf. Metcalfe and Georghiou (1997), p. 25.
43. Cf. Werker (forthcoming).
44. See Metcalfe and Georghiou (1997), 25.
45. Cf. Kuhlmann (1992), p. 125.
46. Cf. Lipsey and Carlaw (1996), p. 269.
REFERENCES
Bernholz, Peter and Friedrich Breyer (1984), Grundlagen der Politischen Ökonomie,
second edition, Tübingen: Mohr (Paul Siebeck).
Fritsch, Michael (1991), ‘Innovation und Strukturwandel’, Das Wirtschaftsstudium,
20, 195–200.
Fulda, Ekkehard, Marco Lehmann-Waffenschmidt and Joachim Schwerin (1998),
‘Zwischen Zufall und Notwendigkeit – zur Kontingenz ökonomischer Prozesse
aus theoretischer und historischer Sicht’, in Gerhard Wegner and Josef Wieland
(eds), Formelle und informelle Institutionen – Genese, Interaktion und Wandel,
Marburg: Metropolis, pp. 327–77.
Grupp, Hariolf (1997), Messung und Erklärung des technischen Wandels. Grundzüge
einer empirischen Innovationsökonomik, Berlin and Heidelberg: Springer.
Hanusch, Horst and Uwe Cantner (1993), ‘Neuere Ansätze in der
Innovationstheorie und der Theorie des technischen Wandels – Konsequenzen
für eine Industrie- und Technologiepolitik’, in Frieder Meyer-Krahmer (ed.),
Innovationsökonomie und Technologiepolitik. Forschungsansätze und politische
Konsequenzen, Heidelberg: Physica, pp. 11–46.
Hayek, Friedrich August von (1945), ‘The use of knowledge in society’, American
Economic Review, 35, 519–30.
Kaldor, Nicholas (1968): ‘Capital accumulation and economic growth’, in Friedrich
A. Lutz and Douglas C. Hague (eds), The Theory of Capital. Proceedings of a
Conference Held by the International Economic Association [1958], London:
Macmillan, pp. 177–222.
Kaldor, Nicholas (1985): Economics without Equilibrium. The Okun Memorial
Lectures at Yale University, Armons NY: M.E. Sharpe.
Kuhlmann, Stefan (1992), ‘Evaluation von Technologiepolitik: Zur Analyse der
Wirksamkeit politischer Techniksteuerung’, in Klaus Grimmer et al. (eds),
Politische Techniksteuerung, Schriften des Instituts Arbeit und Technik 5, Opladen:
Leske und Budrich, pp. 119–35.
Landes, David S. (1994), ‘What room for accident in history? explaining big changes
by small events’, Economic History Review, 47, 637–56.
Lipsey, Richard G. and Kenneth Carlaw (1996), ‘A structuralist view of innovation
policy’, in Peter Howitt (ed.), The Implications of Knowledge-Based Growth for
Micro-Economic Policies, Calgary: University of Calgary Press, pp. 255–337.
Lundvall, Bengt-Åke (1992), ‘User–producer relationships, national systems of
innovaton and internationalisation’, in Bengt-Åke Lundvall (ed.), National
Systems of Innovation and Interactive Learning, London: Pinter Publishers,
pp. 45–67.
Metcalfe, J. Stanley (1995), ‘The economic foundations of technology policy: equi-
librium and evolutionary perspectives’, in Paul Stoneman (ed.), Handbook of the
Economics of Innovation and Technological Change, Oxford and Cambridge, MA:
Blackwell, pp. 409–512.
Metcalfe, J. Stanley and Luke Georghiou (1997), ‘Equilibrium and evolutionary
foundations of technology policy’, CRIC Discussion Paper no. 3, University of
Manchester.
Nelson, Richard R. (1995), ‘Recent evolutionary theorizing about economic
change’, Journal of Economic Literature, 33, 48–90.
Nelson, Richard R. and Sidney G. Winter (1982), An Evolutionary Theory of
Economic Change, Cambridge, MA and London: Belknap Press of Harvard
University Press.
Rosenberg, Nathan (1994), Exploring the Black Box. Technology, Economic, and
History, Cambridge, MA: Cambridge University Press.
Schmidt, Ingo (1996), Wettbewerbspolitik und Kartellrecht, fifth edition, Stuttgart:
Lucius & Lucius.
Schumpeter, Joseph A. (1911/1987), Theorie der wirtschaftlichen Entwicklung: Eine
Untersuchung über Unternehmergewinn, Kapital, Kredit, Zins und den Konjunk-
turzyklus, seventh edition, published first in 1911, unchanged reprint of the
fourth edition of 1934 Berlin: Duncker & Humblot.
Schwerin, Joachim (2001), Wachstumsdynamik in Transformationsökonomien.
Strukturähnlichkeiten seit der Industriellen Revolution und ihre Bedeutung für
Theorie und Politik, Köln: Böhlau.
Schwerin, Joachim, and Claudia Werker (2003), ‘Learning innovation policy based
on historical experience’, Structural Change and Economic Dynamics, 14,
385–404.
Steinmueller, W. Edward (1994), ‘Basic research and industrial innovation’, in Mark
Dogson and Roy Rothwell (eds), The Handbook of Industrial Innovation,
Aldershot, UK and Brookfield, US: Edward Elgar, pp. 54–66.
Teubal, Morris (1997), ‘A catalytic and evolutionary approach to horizontal tech-

nology policies (HTPs)’, Research Policy, 25, 1161–88.
Werker, Claudia (2003), ‘Market performance and competition: a product life cycle
model, Technovation, 23, 281–90.
Werker, Claudia (forthcoming), ‘Technologischer Wandel und Innovationspolitik’,
in Carsten Herrmann-Pillath and Marco Lehmann-Waffenschmidt (eds),
Handbuch zur Evolutorischen Ökonomik, Band II: Evolutorische Ökonomik in der
Anwendung, Heidelberg and New York: Springer.
Whaples, Robert (1995), ‘Where is there consensus among American economic his-
torians? The results of a survey on forty propositions’, Journal of Economic
History, 55, 139–54.
11. The national German innovation
system – its development
in different governmental
and territorial structures
Hariolf Grupp, Icíar Dominguez Lacasa
and Monika Friedrich-Nishio1
It is always difficult to record the history of events that have not yet run their course
and whose outstanding players are all still living. . . . Events appear different, once
they are concluded; different again, while they are still developing. In both
instances, the aims of the reporter also differ. Gustav Struve (1849/1980, p. 290)
1. METHODOLOGICAL INTRODUCTION
The general appreciation of innovation corresponds with a typically

European method of thinking which is not found in all cultures. ‘The posi-
tive evaluation of new findings, the esteem for innovation, the idolisation
of inventors, as well as inventions and patents, are achievements of the
modern world dominated by European-American influence, which, from a
historical point of view, are relatively young’ (Dohrn-van Rossum 1999,
p. 39). However, even in the Christian Occident, the presently predominant
emphasis on innovation results from the manifold historical changes of the
past centuries. Initially, inventions and discoveries were not considered as
an act of creation but only represented the rediscovery of natural phe-
nomena created by God. This change of consciousness – which took place
prior to the period investigated by this chapter (1850–2000) – should be
dealt with in order to better localise innovation-critical opinions in the
present; however, this cannot be done here.
A practicable way to measure innovation could be the elaboration of
definitions and measurement methods by starting from a historical point
of view, with the objective of recording the enormous change characteris-
ing innovation activities. However, this chapter takes the opposite point
of departure: starting from today’s definitions, an investigation of the
239
comprehensive statistical material including related indicators is carried

out, followed by the attempt to trace and complete these back before the
foundation of the German Empire. This means that the presently achieved
level of theory and methodology serves as a point of departure for the
following retrospective.
Consequently, this chapter tries to include a considerable number of
quantitative variables, preferentially in the form of time series. Therefore,
this analysis can be included in the field of cliometrics, the ‘new’ kind of
economic history, which is based on quantitative methods including econo-
metrics, aimed at reconstructing and interpreting the past (Bannock et al.
1998, p. 61). This method is regularly criticised since indicators cannot be
facts; however, according to some points of view, narrative historiography
cannot distinguish facts from interpretation either. According to these, no
fundamental difference exists between the description of facts and their
interpretation, since every description already represents a certain inter-
pretation which, moreover, depends on the definitions presently available
to the describing person (Lorenz 1997, p. 32).2
Even if there is no basic cognitive difference between the (widespread) nar-
rative approach and the (less common) quantitative approach, the objective
of the quantitative approach is the statement of a relationship between vari-
ables, based on many cases, and thus generalisation (Fogel 1964, p. 237
onwards). In contrast, the qualitative approach aims to compare case by
case, that is, this contribution has an analytical or holistic character (Lorenz
1997, p. 238). It seems that most historians prefer an approach by case
(qualitative approach), whereas most social scientists choose the variable
approach, verifying hypotheses for a whole series of cases.
To be provided with variables for many cases on the aggregated level, con-
ceptional ideas are needed which regulate inclusion and exclusion. Whereas,
in general, brief experiences can be clearly delimited, the application of
specific selection criteria is often difficult, especially in the case of long time
series, due to the fact that the individual investigation of all relevant indica-
tions is absolutely impossible. Moreover, indicators have a selective rather
than objective quality, so their undeniable status only applies to a specific
disciplinary context: according to the present status of empirical economic
research, more or less ideal indicators are available for theoretical constructs.
The process of ‘statistical adequation’ means the ‘tailor-made adaptation’
of measurement concepts, the result of which is not entirely satisfactory in
view of the theoretical constructs, but which at least corresponds with the
descriptive framework used as a basic structure for measurement.3
This is aggravated by the fact that the theoretical constructs of innovation
research are not clearly defined. Up to the present, rival and irreconcilable
innovation theories still exist in several disciplines (Grupp 1998). In addition,
The national German innovation system 241
linear models are widespread, presuming a sequential succession of

innovation-oriented phases, the point of departure of which is an unpre-
dictable serendipity in basic research or exogenous technical progress which
falls like manna from heaven. Orthodox approaches have developed, which
try to subordinate or marginalise alternative approaches in order to find the
‘truth’ with the help of their own theory. From an empirical perspective such
attempts must be considered with scepticism since modern epistemology
actually tries to erase any efforts to find the real truth (Hoyningen-Huene
1999). Therefore, the research on science and technology indicators should
ideally begin from a heterogeneous level of theory and definitions in order to
find historically solid indicators. Empirical standardisation of theoretically
heterogeneous constructs must take the individual contexts into consider-
ation, but it should lead to ‘adequate’ indicators; for this reason, empirical
statistical adaptation often remains incomplete, and yet only a minimum
of discrepancy should be left (one only has to refer to the discussion about
real and ideal concepts, Machlup 1960/61). Using an economic-historical
approach, definitions must be fixed as an orientation structure for both the-
oretical and empirical analyses; these definitions should stand iterative
modification, they should apply on an interdisciplinary level and they should
be valid over time.
The standard German textbook for ‘businessmen and students’ by
Roscher (1886), which was reprinted about 20 times by the second half of
the 1880s, and which also shaped Schumpeter’s theory of economic devel-
opment (1911/1964), distinguishes six different economic activities, listing
Invention and Discovery first (sic!) – ahead of mining, agriculture, the pro-
cessing industry, the distribution of goods and the service sector last
(excluding wholesale). Based on this, the result-oriented concept by
Schumpeter (1942/1975, pp. 136ff.) defines criteria in such a way that
innovation represents all that yields a profit for the entrepreneur by being
the first, (the so-called quasi-rent or innovation rent). Quasi-rents of innov-
ation tend to be neutralised by time due to the effects of concurrent
processes. Innovations can be seen as new goods or services, new methods
of production or transportation, new markets or new organisations.
In Germany, the term ‘novelty’ (Neuerung) was used as a definition of
innovation for a long period. The word ‘innovation’ was unknown for a long
time: it only reached the German-speaking regions after Schumpeter’s emi-
gration to the United States, where he published English texts; here, the
English word ‘innovation’ was maintained as the Germanised ‘Innovation’
instead of being translated back to the term ‘novelty’ originally used by
Schumpeter. The definition of innovation was probably adopted around
1960. Consequently, it is evident that the definition of innovation as it is
used nowadays cannot be considered as an anchor for the investigation
period since 1850. Prior to the 1960s, innovation phenomena were described
using other definitions: archives, libraries and research institutions as well
as documents from management, personnel departments or production
centres show terms which are different from those used according to present
standards (for example, ‘laboratory’ might be described as ‘establishment’,
‘experimental factory’).
According to today’s view, the concept of a specific research process
which leads to measurable innovation and which requires personnel and
financial expenditure, is based on Bernal’s (1939) farsighted and clearly
analysing work. Bernal distinguished the role of public research expend-
iture from that of civil research and, – as things stood – from that of the
war industry. The first statistics on expenditure for ‘industrial research’
by British companies are found in the annexes to his works. As reported
by Freeman (1992, p. 3), the definitions used by Bernal during his lectures
at the London School of Economics were brought to international com-
mittees (by Freeman himself as well as by others), which, in the 1960s,
worked for another standardisation of definitions, which led to a first
paper about the measurement of output of research and development
(Freeman 1969).
Consequently, the empirical framework underlying this chapter will be
determined using the current definitions and concepts. These may have had
other meanings in the past, but this ‘anachronism’ must be accepted.
Language is used by any historiography including the hermeneutic one;
however, as language develops over time, definitions may arise, disappear,
or change their meanings. Consequently and independently of the method
applied, any historiography is anachronistic to a certain degree (Lorenz
1997, p. 364). Former innovators were not masters of today’s historical
knowledge, nor of information about present innovation processes –
knowledge and information which we have acquired from our observa-
tion post.
Therefore, the definitions found in the leading OECD manuals4 from the
1990s will be used. Cliometrics are also more concerned with anomalies
than with constantly ongoing, inconspicuous processes. If a structural
breakage is found in a time series, this could point to a statistical artifact
arising from the change of definitions and conventions used. Consequently,
structural breakages5 found must always be interpreted and categorised in a
qualitative way. The problem of anachronism – if not avoided – can thus at
least be moderated. Intertemporal shifts of emphasis as an explanation of
structural breakage are all the more permissible since a functional innov-
ation model serves as an additional basis (see Grupp 1998), working on the
assumption that different innovation-oriented processes can be influenced
by all types of research and development (R&D).
2. NATIONAL INNOVATION SYSTEM
This chapter is divided into a national (this section) and a sectoral level
(see the next section). For the definition of ‘national innovation systems’,
see Grupp 1998, p. 244. From a historical point of view, modifications of
the territorial situation (population, and so on) may not be ignored.
Therefore, from an empirical point of view it is essential to consider, for
example, the size of the Empire or of each federal territory. Not only is the
German Democratic Republic considered here, but also Saarland, the
Corridor, East Prussia and other areas. Territorial changes which took
place can be considered on the basis of today’s statistical procedures, so
that such data series, a priori, do not have to be absolutely consistent with
a territory (also refer to Hoffmann 1965, pp. 2 f.). However, it must be
pointed out that the omission of smaller districts (such as Alsace-Lorraine
from 1871 to 1917) in most cases brings in its train less important errors of
estimation than the big variances in the series of the whole territory of the
German Reich (Hoffmann 1965, p. 3).
2.1 Public Expenditure for National Science and Technology
Traditionally, the development of science and technology is measured by

the number of scholars. In this way, for example, Gascoigne submitted a
historical demography (1992) of the scientific community between 1450
and 1900, by listing the nationality and age of all the scientists. According
to this study, Italy was the leading scientific country at the beginning of
modern times in the late 15th century, representing about half of all the sci-
entists in the world. This had remained almost unchanged during the entire
Middle Ages before that century; then exponential growth with a doubling
period of approximately 50 years took place.
Detailed and complete statistics are available about scientific staff in
Germany since the foundation of the Empire, accessible via today’s elec-
tronic means. However, generally accessible statistical material about R&D
personnel in Germany has only been recorded since the 1960s (in the frame-
work of the Federal Research Report which has been published since 1965).
Another traditional route to the empirical definition of the importance
of an innovation system is scientific expenditure (the sum of R&D funds
and those for training, teaching, maintenance and diffusion of knowledge).
Whereas the evaluation of expenditure for purely educational and R&D
institutions is quite simple, this is more complicated in the case of institu-
tions engaged in both research and teaching. Quotas were adopted to cope
with the individual fields of specialisation as well as with the individual
types of universities. However, it is questionable whether these reflected the
right proportions between the percentage of research and that of teaching

at all historical points in time; in addition, not only is the historical con-
sideration problematic but also the consideration of the present time.
Nevertheless, it is common statistical practice in all OECD countries to
work with such quotas (Hetmeier 1990, and Irvine et al. 1990).
Pfetsch (1982) calculated scientific expenditure between 1850 and 1975,
so rough estimates of the degree of R&D financing can be derived from
this; however, these data records only include public expenditure, disre-
garding the private sector. Consequently, industrial innovation indicators
must be researched separately (see below).
In order to avoid dealing with the difficulty of different currencies, the
development of scientific expenditure can be best evaluated by its percent-
age of the total expenditure of public budgets (Figure 11.1). According to
this, scientific expenditure in the German regions prior to the foundation
of the Empire was approximately 1 per cent (see Figure 11.1). After this,
the percentage reached more than 2 per cent, but dropped to almost 1.5 per
cent between the 1880s and the First World War. The Republic of Weimar
attained a doubling of scientific financing which, however, was lost again
due to the worldwide economic crisis. In West Germany, the support of
science was increased dramatically to reach 6.5 per cent of all public
6
Percentage of public budget
5
FRG
4
United Germany
3
Empire
2
1 GDR
0
18 0
55
18 0
65
18 0
18 5
18 0
18 5
18 0
95
19 0
05
19 0
19 5
19 0
19 5
19 0
35
19 0
45
19 0
19 5
19 0
19 5
19 0
75
19 0
85
19 0
95
00
5
7
7
8
8
9
1
1
2
2
3
5
5
6
6
7
9
18
18
18
19
19
19
19
19
19
20
Note: The numbers for United Germany before the unification in 1990 are artificially
composed by weighted average values between West Germany and the GDR.
Figure 11.1 Development of scientific expenditure in proportion to the

total expenditure of public budgets(per cent)
budgets by the 1970s (university expansion), to fall off to approximately 5

per cent by the time of German reunification. Finally, due to reunification,
the level dropped even further. These indications are based on the numbers
of Empire or Federal institutions and those of regions and states.
Besides the above-mentioned data records, and also based on an analysis
of the older part of these, Pfetsch (1974) submitted an extensive analysis of
German scientific policy between 1850 and 1914. For example, Pfetsch
(1974, p. 60 and p. 171) cannot confirm the thesis that the state spends more
money on science and technology in times of increasing economic wealth.
In spite of some references to anti-cyclical research policy, overall an irregu-
lar economic attitude is shown by government policy.
Surprisingly, expenditures after World War II start at around the same
level as at the beginning of the last century (after World War I) which is at
the same level as the endpoint of records in the 1940s war years, and increases
in a similar way after World War II as after World War I. This points to quite
stable and persistent institutional structures underlying the financial totals.
The financial support of Research and Development is typical for post-
war Germany. Until 1945, the financing share for R&D only played a sub-
ordinate role in total scientific expenditure. Although the research share6
was 20 to 30 per cent during the first period after the foundation of the
Empire, it dropped to less than 20 per cent by the beginning of the first
World War (see Figure 11.2). In addition, it is important to know that
1.0
0.9 R&D expenditures
0.8
0.7
0.6
0.5
0.4
Other science expenditures
0.3
0.2
0.1
0.0
18 0
18 5
18 0
18 5
18 0
18 5
18 0
18 5
18 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
95
5
5
6
6
7
7
8
8
9
9
0
0
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
9
18
Figure 11.2 Government expenditure for science, divided into the publicly
financed R&D share and other scientific tasks
a great deal of scientific expenditure by the Empire was used for defence
tasks shortly after the foundation of the Reich. During the Weimar
Republic and the Third Empire, the R&D share of the total of scientific
expenditure continued to fluctuate at around 20 per cent (industrial
research not included).
A quick increase in the R&D share of scientific expenditure occurred
when research in certain areas was allowed again in the young Federal
Republic, after the signing of the Treaty of Paris in 1955: at times it reached
70 per cent and has only declined due to the recent reunification.
Prior to the Second World War, the relatively insignificant role played by
R&D within the scientific world is also shown by the distribution of funds
to the different institutions. During the decades preceding and following
the foundation of the Empire, the lion’s share of approximately 70 per cent
of the total public scientific expenditure is accounted for by all types of uni-
versities. Empire agencies and other institutions were established over time,
so that the universities’ share of scientific expenditure decreased to about
35 per cent at the beginning of the 20th century. Now, a little more than 20
to 30 per cent goes to these institutions charged with varied tasks, and
almost 10 per cent is accounted for by pure R&D institutions.
During the entire period before and between the two World Wars, a small
but significant and strongly varying amount of public expenditure went to
functions other than institutional support. The non-institutional support
is described as ‘project-specific scientific expenditure’ by Pfetsch (1982,
p. 113); this could be misunderstood, since part of it consisted of public
grants and support given to a wide range of projects and not only R&D
projects. They ranged from ‘Scientific efforts for opening up Central
Africa’, the publication of archives, international contributions to the sur-
veying of the earth, measures to combat typhoid fever or infant mortality,
to financial support for congresses.
Until the First World War more attention was generally given to scientific
support on an industrially relevant level, as well as to scientific application
(Pfetsch 1974); however, it would be incorrect to conclude that a major part
of these funds were granted to private companies, as can be shown for the
period following the Second World War. Industry was more interested in a
proportional increase in public support of production-relevant branches of
science (consequently, in the creation of external effects upon science-based
industries, Pfetsch 1974, p. 107) than in the support of their own R&D.
Of course scientific expenditure was borne exclusively by the German
states until the foundation of the Empire; afterwards, the central power
moderately supported the total public science budget by 20 per cent. Only
during the Republic of Weimar did this share grow considerably (see
Figure 11.3). After the occupation of Germany at the end of the Second
1.0
0.9
0.8 Financial share of States/Länder
0.7
0.6
0.5
0.4
0.3
0.2
0.1 Financial share of Empire/Federal

government
0.0
18 0
18 5
18 0
18 5
18 0
18 5
18 0
18 5
18 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
19 5
19 0
95
5
5
6
6
7
7
8
8
9
9
0
0
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
9
18
Figure 11.3 Development of the financing of scientific expenditure: the

proportion of federal and state financing
World War the Federal Government could not play its prior role again, all
the more so since several research areas were prohibited (research in the fields
of armament, nuclear science, chemistry and aviation). However, federal
institutions systematically increased their influence on science until the level
during the Third Reich was reached again. Since the recent reunification,
federal administration has been slowly but surely withdrawing from
scientific support.
Consequently, it must be noted that, following critical upheavals such as
the foundation of the Empire, and the First and the Second World Wars, the
Federal states had always begun to take over important tasks of scientific
support and were later relieved by the central power. The events following
the foundation of the Empire, which were called ‘getting empired’ by Pfetsch
(1974, p. 105), occurred in a hesitant way on complex scientific-political
levels. The results were manifold forms of co-operation between private,
mixed and individual government institutions and Empire authorities. The
same can be observed since 1945 under completely different political cir-
cumstances: the Federal Government only slowly became a dominant and
central supporter and organizer of the scientific system, its scientific expend-
iture diminishing both on an absolute and on a relative level only after the
reunification. Again this points to persistent basic structures in the national
innovation system. The only historical exception is the reunification of 1990:
whereas the Federal Government played a dominant role as a central
supporter and organiser during the reunification of the two scientific

systems, its scientific expenditure diminished afterwards both on an absolute
and on a relative level.
Regarding the R&D expenditure of the German Democratic Republic
(GDR) (see Figure 11.4), note that the individual statistics were centrally
maintained and are comprehensive. However, the conditions which were
applied do not fully comply with those used by OECD countries and often
show exaggerated values. Following reunification the relevant statistics
were revised and adapted to Western standards; however, the conversion
problem of the East German Bank’s Mark (M) persists. Due to the non-
convertibility of this currency the reliable purchasing-power parity values
of OECD countries cannot be applied.
Figure 11.4 compares the R&D expenditure of the German Democratic
Republic with that of West Germany. In order to be certain, a pessimistic
and an optimistic variation can be applied in order to show a range of uncer-
tainty due to conversion. The first possibility is based on the purchasing-
power parity (PPP) of so-called baskets of commodities; in the second
model the subsidies included in GDR commodity prices are taken into
consideration and deducted (anonymous, 1986, pp. 259–268). It is shown
in both estimations that the national R&D expenditure of the German
Democratic Republic could not equal the West German level (per head of
the population) but the general upward trends resemble each other some-
what. This may come as a surprise to those who point to the inefficiencies
of the communist part of Germany, but again, the underlying institutional
1400
1200
Former FRG
1000
800
DM
GDR (PPP)
600
GDR (PPP minus subventions)
400
200
0
1971197219731974197519761977197819791980198119821983198419851986198719881989
Figure 11.4 Development of the national R&D expenditure per head in

East and West Germany (DM, 1989)
structures remained basically the same as before the war, requiring similar
amounts of public support.
2.2 Development of Scientific Activities
It is impossible to achieve an insight into the development of non-

codified and thus ‘tacit’ knowledge and experience of the scientific staff.
For this reason the historical development of an innovation system is
often shown by the personnel statistics, or by statistics showing monetary
expenditure. However, only expenditure is measured by this method,
instead of the fruit of scientific activities. Efficiency measurements are
particularly impossible. Consequently, modern innovation statistics
make regular use of yield measures; regarding scientific work, statistics
of publications are a typical output indicator. Analyses of the degree of
publication activities have been maintained for centuries; however, it
must be noted that the publication media chosen by scientists may differ
from one faculty to another, as well as over time (Wagner-Döbler and
Berg 1996, p. 289). Only during the 19th century did scientific journals
achieve the same degree of significance as books, the dominating publi-
cation media until then.
Regarding publication activities in selected areas, only a few but informa-
tive historical time series for selected areas are available (see below).
However totals are studied first. Due to the known difficulties of aggrega-
tion, only limited sources of publication activities are available on this level
of analysis. An analysis of the Catalogue of Scientific Papers for the 19th
century shows that the output of scientific papers had been growing con-
stantly since 1800; and it accelerated tremendously from 1884 onwards
(anonymous, 1925). This analysis is not limited to Germany but refers to
worldwide publication output.
From 1900 onwards the availability of data improved worldwide. The
growth rates of periodicals were evaluated in Ulrich’s Periodical Database
(CD-ROM version) by Mabe and Amin (2001); if one takes the example of
peer-reviewed academic journals from this catalogue, a remarkable expo-
nential growth is shown up to the end of the 20th century, which then
slightly decreases (Figure 11.5). A detailed statistical analysis covering the
period from 1900 until 1944 shows an almost constant increase in the inven-
tory of journals of 3.2 per cent per year, followed by a phase of expansion
with an almost constant growth rate of 4.8 per cent per year until 1974,
ending with a lower rate of growth of 3.7 per cent per year, similar to the
first half of the century (‘century standard’).
This development seems to be linked with historical structural breakages
on a worldwide level. Following its academisation during the 19th century,
Cumulated number (logarithmic) 100 000
10 000
1000
100
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
Note: The thin drawn lines represent the exponential growth rates for the main growth
periods as explained in the text.
Figure 11.5 Cumulated number of scientific journals peer-reviewed

worldwide (shown as a semi-logarithmic figure) from 1900–2000
science was characterised by constant expansion until the Second World

War. This can be called the normal development. After the Second World
War, the growth rate of scientific output increased and reached almost
5 per cent for almost three decades (the so-called boom phase), due to the
economic miracle, the armament race and reinforced industrial research and
development activities, as well as an expansion of scientific activities in
general. This phase was followed by a normal phase of expansion similar to
pre-war conditions; it can be linked to the ending of university expansion
and the consequences of a severe recession due to the oil crisis, as well as with
a general decrease of economic growth rates (‘limits to growth’). Using a
completely different timeseries, Maddison (1982, p. 92) found phases
showing a high degree of similarity (his phases are 1930–50, 1950–73,
after 1973).
Analysing the situation in Germany, the Bibliography of German
Periodicals Literature represents an alternative regarding historical data.
This bibliography covers the German-speaking periodical literature from
1896 until 1964, and the international periodical literature from 1965
onwards. However, no information about the author’s nationality or the
institute’s location is found in this source; probably most of the older
records come from German authors, but ratios and changes in these are
also unknown.
The Science Citation Index (SCI), which was available as an online

version as early as 1964 (see below), has a printed version listing the publi-
cations from 1945 until 1974. Although no indications are found regarding
the authors’ nationalities or the institutes’ locations, the listing of period-
icals is classified by the countries editing and printing them. Since Mabe
and Amin (2001) have demonstrated a highly positive correlation between
the number of periodicals and the number of journal articles, the number
of German-language articles listed in this data record can be counted for
the period since 1945. The regular announcement by the SCI that records
would be completed back to 1900 was withdrawn,7 so there is no hope for
the early publication of a century’s inventory.
Taking into account that the total volume of publications has grown
enormously, it is surprising to see that the share of German periodicals has
been constant since 1945, that is, it has augmented in line with the world-
wide volume. The average proportion of German periodicals, which is 8.9
per cent of the total SCI inventory, shows only minimal fluctuation. Since
1974 an equally constant proportion of German authors is shown in the
SCI online version; it is obviously sensible to link these two data records,
all the more so since they overlap each other over a long period (from 1975
to 1984) so that the corresponding factor of extrapolation from periodicals
to publications can be determined.
The SCI online version shows the nationality, a corresponding field being
encoded where a German author or an institute located in Germany are
concerned (independently from the medium’s language). The strongest
growth of the (extrapolated) publication numbers is stated from the middle
of the 1960s and during the 1970s (Figure 11.6). This matches perfectly
with the observation of a worldwide expansion of the scientific system,
even though the German scientific world showed a delay of almost 20 years
resulting from the special situation of reconstruction as well as from the
Allied Forces’ restrictions regarding certain research areas.
At the end of the 1980s the growth rate decreases not only in Germany
but on a worldwide level; after 1989 the total number of German articles
shows a dramatic decrease. It must be noted that the statistics cover both
West and East Germany, and that this decrease in publication activities
could principally represent the decay or dissolution of the East German
scientific system. The publication level of 1987 was only reached again
during the publication year 1993, which was characterised by strong
growth anticyclical to the worldwide slowdown. A comparison of research
activities both on a disciplinary and on a qualitative level is suggested
below. Figure 11.6 also shows a simulation of the course of the three
periods when growth rates were constant, which points out the exponential
growth.
80 000
70 000
Number of publications
60 000
Simulation with exponential growth
50 000
40 000
30 000
20 000
10 000
0
45 50 55 60 65 70 75 80 85 90 95
19 19 19 19 19 19 19 19 19 19 19
Figure 11.6 Development phases of publication activities in Germany

since 1945
From 1974 to 1990, SCI publications from West and East Germany can
be compared electronically.8 This period is characterised by moderate
growth of publication activities (‘century standard’). In the 1970s, the share
of East German publications was approximately 16 to 17 per cent.
However, if one compares East and West Germany, both the proportional
shares of population and the proportion of R&D staff is almost 30 per
cent, so that scientific publications from the German Democratic Republic
are less represented in the US-based database. The proportion of East
German publications had constantly diminished to reach 13 per cent by the
end of the 1980s; and there is no answer to the question as to whether the
representation in the database was worse or if the output efficiency of East
German research activities continued declining until reunification.
Measured by its publication output, the profile of GDR research resem-
bles that of the former Federal Republic. In proportion to worldwide
average shares, researchers of both parts of Germany published much more
than a pro rata share in the research areas of energy and nuclear technol-
ogy, chemistry, solid-state physics and microbiology. A weaker level than
the worldwide average was shown in information science, engineering,
environmental research and public health, as well as in other biomedical
subjects. According to our estimation, this structural similarity could be the
reason for such a strong diminution of publication activities on an all-
German level following reunification. Integration did not affect differently
specialised East and West research systems, but only research systems with
the same principal orientation, which led to the deplorable ‘re-allocation

and consolidation’ process in East Germany. Independently from a polit-
ical evaluation of the organisation of GDR research institutes, this struc-
tural similarity must be pointed out; obviously 40 years of division were not
sufficient for a differentiated development of the basic specialisation pat-
terns of research in both parts of Germany. To a great extent, and in the
sense of path dependency, research is still based on the (common) prefer-
ences which existed prior to the division. This unique historical situation
could be understood as an unintended experiment: basic patterns of
scientific specialisation change only slowly, even in times of great political
system change (Hinze and Grupp 1995, p. 65).9
Another method of comparing the two scientific systems is to observe
the frequency of citations. More frequently cited scientific studies are con-
sidered to be more significant in some way, for example because they
include either important methods to which many successive authors return,
or especially important results (or errors which are repudiated later). GDR
literature shows a lower rate of citations per scientific publication than
West German literature; there are two possible reasons for this: on the one
hand, the scientific value of GDR publications could have been minor; on
the other hand, the periodicals in which these articles appeared were
received to a lesser degree on an international level, in particular in the
English-speaking realm. This could also be due to the lower circulation
figures of the corresponding periodicals (partly of Soviet origin).
In the bibliometric statistics these two effects can be viewed separately,
by referring the rate of citations either to the worldwide average or to the
average of the individually selected periodicals (Grupp et al. 2001). The fol-
lowing research results emerge: GDR research publications show an almost
general lack of worldwide communication of results (particularly in the
Anglo-Saxon linguistic area). This explains the low rate of citations.
However, a correlation between the frequency of citations and the publica-
tion organs (which are typically less read in the Anglo-Saxon area) selected
by GDR authors shows that a favourable rate of citations is found in com-
parison with articles from countries other than the GDR which are also
published in these periodicals. Consequently, GDR publications are con-
sidered to an above-average degree once the citation ratio due to the lack of
international spread is mathematically corrected. The highest regard (ratio
of citations) is found in the areas of neuro science and internal medicine,
as well as dietetics and agriculture. Following the correction of the ratio of
citations, the results of GDR environmental research were the least noticed
by the professional world.
A division of science into 27 subsections (Hinze and Grupp 1995) leads
to a surprising correlation between the degree of internationalisation and
that of attention received: it was found that those research areas which are
not internationally spread are significantly more noticed by the profes-
sional world than the publications with a higher degree of international-
isation. No such correlation can be found for West German research. It is
possible that hierarchy and cadre selection mechanisms played a role
regarding access to Western periodicals.
2.3 Industrial Research and Development in Germany
Since the foundation of the Empire, economic growth of industrial coun-

tries, in particular in Europe, has increasingly been based on the innovation
energy of the knowledge-based industry. ‘This is undeniably true for the
impulses of growth immediately released by these industries, starting with
carbon chemistry and electrical technology’ (Wengenroth 1997). There is
hardly a clearer and more distinct way to describe the conducive effects of
industrial research on the culture and efficiency of innovation.
It is still difficult to demonstrate the companies’ increasing R&D
expenditure for such an undeniable success. In particular, no complete data
records are available about monetary expenditure or research personnel
prior to the end of the Second World War, that is, the data record estab-
lished by Pfetsch (1982) regarding public scientific expenditure has no
counterpart for industry. Today’s systematic statistics about R&D expendi-
ture and personnel of the Federal Republic start from the year 1962; certain
presumptions allow the reconstruction of the corresponding indicators
starting from 1948–49 (Figure 11.7). According to this, industry has con-
tinuously increased its R&D budgets to a higher degree than government,
the share of the latter being at present approximately 35 per cent.
The reconstruction of the corresponding indicators prior to the Second
World War is only possible if one starts from the individual companies or
branches. This was tried for the areas of chemistry and electrical technol-
ogy (see the next section). The investigation of the history of R&D in
leading companies of two selected industrial branches shows that the exe-
cution of case studies in order to represent time series is possible in princi-
ple and, moreover, makes sense. During the whole investigation period
many large and leading companies in Germany maintained archives which
can be considered as complete in spite of difficult conditions at times in
German history.
Presently the representativity of company-related R&D time series cannot
be plausibly proved for the development of a whole branch and that of all
branches for the whole national economy. A more detailed analysis should
include other important branches, in particular the metal industry. However,
the results found so far show that the existence of different ‘cultures’of R&D
100%
80%
Business expenditures for R&D
60%
40%
R&D expenditures by governments
20%
0%
48
50
52
54
56
58
60
62
64
66
68
70
72
74
76
78
80
82
84
86
88
90
92
94
96
98
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
Figure 11.7 Development of government and industrial R&D expenditure
in relation to each other from 1948–2000
is possible in the different branches, which influences not only the forms of
organisation but also researchers’ liberties and practical work.
2.4 Development of Innovation Activity in Germany
The observation of the development of innovation activity is important in

itself in order to establish R&D results, mostly on a technological or appli-
cation level. Adopted methods are statistics on patent applications (a figure
representing successful innovation activity seen from the innovators’ or
applicants’ subjective perspective) on the one hand, and, on the other hand,
statistics on the number of granted patents (as a figure representing suc-
cessful innovation activity, seen from the objective perspective of patent
examiners). Statistics on patents make even more sense if one takes into
consideration that only fragments of industrial R&D expenditure are
known prior to the Second World War. Instead of inputs, industrial R&D
activities can be measured by their patent outputs, and this even more pre-
cisely from a technological perspective than by monetary indicators. This
also explains our interest in both patent grants and patent applications: if
no patent is granted after verification of the novelty, the inventive step and
its commercial usefulness, for example due to a lack of novelty, the apply-
ing company had nevertheless invested R&D efforts – even if these led to an
objectively already known result. Consequently, the ‘subjective’ perspective
of a successful invention is closely linked to the R&D performance which

was in fact realised. Statistics on patent applications as a proxy variable for
R&D expenditure may ignore whether the object of the invention was a
world novelty or not. R&D expenditure also includes the costs of unsuc-
cessful inventions or those which were belated in comparison with com-
petitors (imitations).
The period to be considered is fully included in the statistics of patents.
In some German regions, patents were applied for as early as 1820, start-
ing from the South due to the influence of the Napoleonic legislation.
From 1 July 1877, a patent act for the German Empire standardised pro-
cedures. Thus, the creation of patent acts in Germany follows the scientific-
technological innovation push of the 19th century, at the end of which
Germany was one of the leading industrial nations. In about the middle of
the century the local, largely secluded markets were dissolved, and the
German economy was integrated into the quickly expanding world
economy (Ziegler 2000, p. 198, and North 2000b, p. 13).
For several years, patent statistics back to 1879 have been available using
machine readable methods. On the one hand, electronic data records since
1970 are more informative than those of former periods, leading to a largely
increased importance and use of these patent data records by modern
studies in science and technology. On the other hand, if one makes the
effort to bring the individually valid patent classifications together, and to
link together different patent data records for the appropriate historical
sequences, assembled patent statistics can be established for the whole
period. Moreover, regarding the assignment of priority years prior to 1969,
these must be established according to the reference system of the individ-
ual patent authority.
Considering global patent activities in Germany (Figure 11.8), an obvious
difference is found between the dynamics of chronological development and
that of scientific activities (publication statistics). The strongest growth with
an average annual rate above 9 per cent on a low level takes place from 1820
to the foundation of the German Empire; the total growth rate for German
regions is shown to be constant except for the setback due to the war of
1870–71. Following the introduction of the countrywide German patent
act, the number of applications and grants rises rapidly within a few years,
and continues growing at a constant rate, which is lower compared with the
period preceding 1870 (annual average rate 6.7 per cent). This growth, which
lasted for almost one century, was abruptly stopped by the First World War,
the annual patent production being halved. From approximately 1920 to
2000, an eventful development pattern nevertheless shows very little growth.
For almost one century the number of annual patent applications is approxi-
mately 50 000 to 60 000. Consequently, German patent productivity per
90 000
80 000
Simulation with exponential
growth
70 000
Number of patent applications
60 000
50 000
40 000
30 000
20 000
10 000
0
1812182218321842185218621872188218921902191219221932194219521962197219821992
Figure 11.8 Development of patent applications from 1812
person reaches one of the highest levels in comparison with the United
States, Japan and the European Union.
Diverging from this rough rule, growth is observed during the Weimar
Republic phase until the beginning of the Third Reich, followed by a very
deep setback during the Second World War, which is distinctly more serious
than that during the First World War, and a return to the average rate by
approximately 1960. Another boom follows until 1975 when deep recession
takes place which is only overcome in the mid 1990s.
No investigation has yet discovered whether these growth cycles only
have economic causes. The economic boom after the foundation of the
Empire is well-known (Ziegler 2000, p. 201); the same is true for the serious
recession following the oil crisis in 1973 straight after the economic
miracle. The question remains as to whether the reduction of innovation
activities at the beginning of the Third Reich was only due to economic
reasons or to a modified practice of patenting (for example by stronger
observance of secrecy due to the early war economy, or by the expulsion
or migration of Jewish scientists). Further, the question is asked as to why
the growing R&D budgets granted after the Second World War did not
lead to an increase in patent activities. Obviously this decrease of patent
efficiency is due to an economic calculation which is not exclusively driven
by R&D inputs.
Up to now, global indications have been made about patent documents
of national and international actors. By international actor it is understood
that either the inventors’ residence or the applying company is located

outside Germany. From 1881 to 1913, the share of foreign patent grants was
extremely high, showing an average of 35 per cent: until 1933, Germany’s
reputation as the leading scientific country attracted many young scientists
from abroad. Especially Americans came to the German Empire in order to
benefit from practice-oriented education for their degrees, and possibly even
to experience some years of active industrial research (Erker 1990 and Smith
1990). After the First World War and the efforts to achieve self-sufficiency
in the 1930s, the share of foreigners was reduced by almost 10 per cent but
remained a significant figure in spite of all war speculations. Since the recon-
struction of the German patent administration following the Second World
War, the share of foreign patent grants has constantly increased, reaching
more than 60 per cent at this so-called globalisation time.
Although most patent applications originate from industry, universities
and other public research institutions are increasingly also involved in
patent production. This can be linked to their stronger orientation towards
applied research, as well as to their acceptance of outside resources from
private industry. Prior to the reunification in 1990, universities attained a
share of 3 per cent of all West German patent applications, which grew to
more than 4 per cent after reunification. Similar contributions are shown
by all other public research institutions and companies, above all the
Helmholtz Association (big national labs) and the Fraunhofer Institutes. In
view of their limited human resources the patent productivity especially of
the Fraunhofer Society must be more highly valued than that of univer-
sities (Schmoch et al. 2000).
Compared with Western conditions, certain deviations in the patent law
conditions of the former GDR were caused by the socialist spirit of own-
ership. Consequently, the national patent applications at the former GDR
Authority of Invention and Patent Administration (AfEP) can hardly be
compared with those submitted in the West (Hinze and Grupp 1995, p. 42
onwards). Therefore another method was chosen for this analysis, which is
now based on GDR patent activities in West European foreign countries.
With the help of this method all the particularities related to patent law
specifications are circumvented, enabling comparison with Western coun-
tries. GDR inventors were mostly interested in the economic sector of the
former Federal Republic, so that the foreign applications submitted for this
target market can be referred to (independently of whether the application
was submitted to the German Patent and Trademark Office, to the
European Patent Office, or to the International Patent Authority WIPO
designating the Federal Republic of Germany).
The basic framework conditions for GDR activities in view of indus-
trial property rights are fixed in the patent law of 6 September 1950
(Albrecht et al. 1991, p. 4). Nevertheless, GDR patent activities according

to Western legislation are hard to ascertain during the first years. This is
linked to the various forms of recognition of the GDR as an autonomous
state by different nations. Some GDR inventors operated from Federal
Republic addresses. In spite of these imponderabilities an increase in
patent activities by GDR inventors is seen until approximately 1983–84.
Then the figures are characterised by stagnation, and since 1987 they
have been decreasing. The same tendency is shown on a higher level by
national patent applications in the German Democratic Republic, so that
the drop of innovation activities prior to reunification is undoubtedly
proved (Hinze and Grupp 1995, p. 47). The causal and significant explan-
ation of this development is the fact that since 1981 the share of R&D
personnel in the economic sector of the GDR had continuously declined,
and financial resources for R&D in the economic sector were also
reduced.
A comparison of the specialisation of GDR patent portfolios with
those of West Germany is very interesting. According to a division of the
whole GDP technology area into 28 fields, particular strength is found in
the fields of paper and print, textiles, machine-tool manufacture, hand-
ling, optical instruments and metrology. Distinct weakness is shown in the
fields of chemistry, electrical technology and electronics and information
technology, as well as traffic and transportation. This specialisation profile
was constant over time. In particular, the eastern regions’ patent profile of
the 1990s (including East Berlin) corresponds largely with that of the
GDR of the 1980s (Schmoch and Sass 2000). In addition, there is an
amazing correlation with that of West Germany. In spite of completely
different economic conditions (Stolper et al. 1964), large fields of tech-
nology show a correspondence between East and West Germany
until reunification (Grupp and Schmoch 1992, p. 118 onwards). This
was also found for the area of basic research (publication statistics) and
explained by path dependencies and persistent structures in both parts of
Germany despite their different political regimes (see above). Since
reunification, the new federal regions have expanded their top level tech-
nology (semiconductors, biotechnology, surface technology) starting from
a low level.
Consequently, the question is whether a similar influence of scientific
development on technology can be found in both parts of Germany. An
evaluation of the scientific dependency of technology is usually based on
the fact that patent specifications include responses to former inventions in
the form of citations to patent documents. If an invention is directly based
on science which was published but not patented, the patent engineers
annotate references to the corresponding scientific literature. It could be
shown that the frequency of such scientific hints included in patent

specifications is a valid measure of the dependency on science inherent in
a field of technology (Grupp and Schmoch 1992).
Regarding the degree to which technology was based on science, the
German Democratic Republic remained constantly behind the world
average. Although the same is true for West Germany, the gap was distinctly
smaller. During the 1980s, both West and East Germany significantly
increased their orientation to science-based technology – the GDR on a
lower level but with a higher growth rate compared to West Germany.
Comparison between the two countries is interesting in that the East-West
distance was reduced in this area.
When Germany was reunified in 1990, two almost identically specialised
technology systems came together. It was not possible to integrate the
strength of one side and the weakness of the other one; instead, the fields
characterised by strength were the same on both sides and the weaker fields
were equally neglected.
The limited use of science by GDR technology is clear in international
comparison; the same is true for the Federal Republic, but on a different
level. In view of the extended scientific activities, which were most significant
in proportion to the size of the GDR, it is surprising to find that technologi-
cal development did not benefit from the use of science and therefore
remained on a relatively science-poor level.
3. SECTORAL INNOVATION SYSTEMS:

ELECTRICAL TECHNOLOGY AND CHEMISTRY
So far, the national German innovation activities have been analysed in
the varying territories. In the framework of this article, it is not possible
to consider all the university and industrial sectors individually. Therefore,
sectoral analysis of the two areas chemistry and electrical technology will
be carried out as representative studies. In the literature, chemistry and
electrical technology are considered perfect examples of the science-based
industries which came into existence in the second half of the 19th
century. These industries are characterised by a rapid transfer of research
results to production. Due to the intense exchange activities on both sides
one can also talk of ‘industry-based science’ (König 1995, p. 283), since
science probably benefits more from industry than inversely in certain
phases (for example in the case of the newly emerging academic electrical
engineering).
Another motive for the choice of these two specific sectors is the power
of innovation in German industry today, which, roughly speaking, is
considered as contrasting: chemistry with its brilliant innovation and

export performance compares with international standards, and electrical
technology (or its subsumption into information technology) is considered
as a weak point in the German economy. According to our thesis, an exem-
plary analysis of these two sectors within the national German innovation
system should reveal essential facts to explain their different characters at
the end of the 20th century.
3.1 Sectoral Expenditure for Science
Statistics on science expenditure until 1945 include a classification by

sectors. However, only natural science and engineering are identified in
totals, so that these data records cannot be used for the study of electrical
technology and chemistry. The establishment of a data base with sufficient
capacity from an institutional level (ministries, universities, and so on) also
seems impossible. This has remained unchanged since the Second World
War. However, personnel figures from universities are available for selected
years in both chemistry and electrical technology.
3.2 Development of Scientific Activities
First, the publication history of the two selected areas, electrical technol-
ogy and chemistry, will be summarised. One problem is the tracking of
publication activities in the fields of electrical technology and chemistry
prior to 1974, the period for which no online data base is available. To
resolve this problem, the corresponding monographs were submitted
to manual analyses, which made the period from approximately 1924
onwards accessible. Because this process was undertaken beyond the year
1974, the manually obtained figures can be extrapolated to the data base
level. First analyses support the earlier hypothesis: during the 1960s and
early 1970s, the growth of publication activities in chemistry was higher
than it was later on, so that a synchronism of scientific publication activ-
ities and the economic success of mass chemistry (base) could be
confirmed (Figure 11.9).
An extension of the historical period shows a rapid growth of publication
activities in the field of electrical technology since the 1950s, reaching its peak
from the middle to the end of the 1960s, then again in the mid-seventies. This
development is followed by a decrease of publications to reach a relatively low
level in 1981, then by another period of growth. Consequently, the present
study of data records is substantially supported and completed by manual
tracking up to the Weimar Republic. The manual method probably has an
even higher explanatory power, since the continuous completion of inventory
250 000
200 000
Chemistry
150 000
100 000
50 000
Electrical engineering
0
24
29
34
39
44
49
59
64
69
84
99
5
9
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
Figure 11.9 Development of German publications in the fields of electrical
technology and chemistry, from 1924 to 1999 (early figures
extrapolated)
data records (like in the case of the SCI) could lead to artefacts. Although the
SCI strives for transparency with the selection procedures of periodicals
entering the data base, quantitative conclusions for individual subjects are
not always possible (Testa 1997).
Publication activities in the fields of electrical technology and chemistry
show interesting subject-specific deviations from the average trends of
publication activities in all scientific disciplines in Germany (Figure 11.9).
In contrast to the upward development shown by electrical technology
prior to and following reunification, which takes an anticyclical course
compared with global publication activities, the growth of publication
activities in chemistry is reduced synchronously with the publication
activities in all scientific fields in the middle of the 1970s. Presumably the
anticyclical development of publication activities in the area of chemistry
can be linked to the division of basic material chemistry into specialised
and fine chemistry which took place in the 1970s, followed by bio-
technology. The anticyclic boom of publication activities in the area of
electrical technology could be linked to the appearance of the present
information and communications technology, and could be interpreted as
its scientific predecessor. First analyses of the development of R&D
expenditure in chemistry and electrical technology industries support this
thesis.
3.3 Industrial R&D Expenditure and R&D Personnel
In the scope of investigations about industrial R&D expenditure, special

attention was given to the two industrial branches selected in this analysis.
In the fields of chemistry and electrical technology, a small sample of large
industrial companies could be personally interviewed regarding the exist-
ence of corresponding data records. Sometimes the relevant archives could
be accessed. The corresponding industrial federations were involved.
Interesting series of figures are available from the federations, but no figures
were found regarding R&D expenditure or R&D personnel. In contrast to
this, time series can be found relating to the status of memberships, about
lectures and other things.
Comprehensive archive material concerning the number of chemists and
physicists is available from the successor organisations of IG Farben such
as BASF. These indications correlate strikingly with the patents held by
these companies; the hypothesis according to which a lack of input figures
can be substituted by patent statistics is thus supported. Besides this,
employment figures of other technical professions can be constructed.
A comparison between the total number of chemists employed by BASF,
Hoechst and Bayer, and the chemists employed by universities and techno-
logical universities (see Pfetsch 1974, p. 158), shows that, since approxi-
mately 1880, the contribution of industry can no longer be ignored. Prior
to the foundation of the Empire only a single-digit percentage of all
chemists were employed by industry; equality was reached in 1885, and in
1890 the number of industrial personnel exceeded that of university staff.
Prior to the beginning of the First World War the three companies already
employed three times as many chemists as German universities.
Seen from a quantitative point of view, it is found that even prior to the
First World War industrial R&D efforts parallel to public engagement
could not be neglected either.
The most important German companies in the electrical technology
branch were recorded on the basis of the list of patents granted in 1928,
which was established by the German Empire Patent Office (in total 1121
patents, classified by companies). According to this record, the contribution
to patents in the area of electrical technology was 20 per cent by the leading
company Siemens-Schuckert-Werke, 15 per cent by AEG, 9 per cent by
Siemens und Halske and 5 per cent by the subsidiary of the latter,
Telefunken. However, the large scale to which electrical technology was
embodied in the German economy during this period is shown by the fact
that the ten most important patent assignees produced only little more than
60 per cent of all patents; almost 40 per cent is accounted for by other com-
panies. Such a limited degree of concentration means that it would be almost
hopeless to register complete R&D input figures by individually interview-

ing all companies of this branch. Since 80 per cent of all patents are attained
by the 50 most ‘patent-active’ companies the number which were also
involved in R&D could have exceeded 100 according to a rough estimate.10
The number of scientists employed in research institutes by the above-
mentioned large companies can be ascertained; however, no indications
about monetary expenditure for R&D are found in the business reports.
Even the annual reports of the research institutes were not fertile. In the area
of electrical technology it seems almost impossible to take an inventory of
relevant R&D input indicators, covering both the broad industry structure
and the corresponding period of time. However, as in the case of chemistry,
a study based on a selection of companies could be carried out to find a pos-
sible correlation between personnel data and patent data. In any case, stat-
istics on patents will serve as a significant method of investigation.
3.4 Development of Sectoral Invention Activities
Due to very detailed patent classifications, the inventory of patents in the

area of electrical technology and chemistry can easily be recorded.
Although partly contestable delimitations are necessary for this investiga-
tion, these can be documented; moreover, a study based on different delimi-
tations is possible.
It is shown that high-voltage technology (electricity, electrical energy)
and low-voltage technology (audio-visual technology and telecommunica-
tions or communication engineering) can be distinguished for the whole
period. On an analytical level, a strong growth of high-voltage engineering
is shown since approximately 1900 (the more modern subject in times of
electrification), as well as its stagnation after 1950, whereas low-voltage
engineering shows moderate growth until 1944, and becomes stronger after
1950 (Figure 11.10).
In total, the very different development dynamics are reflected in global
statistics by a 4 per cent share of all German patents held by electrical tech-
nology until approximately the end of the 19th century; then the share
shows an almost exponential growth to more than 20 per cent during the
Second World War (Figure 11.11). In spite of an increased growth of low-
voltage engineering, the sum of patent shares is reduced from almost 20 per
cent (1958) to approximately 13 per cent (1998); this is in accordance with
the internationally common image of a lack of specialisation in the area of
German information technology.
At the beginning of the observation period, chemistry had already
achieved a patent share of 10 per cent, which remained almost unchanged
until 1945 (except for short-term cyclical fluctuations over a few years).
2500
Low-voltage technology
2000
High-voltage technology
1500
1000
500
0
1882 1892 1902 1912 1922 1932 1942 1952 1962 1972 1982 1992
Figure 11.10 Moving average line of the first publications of national

patent applications submitted to the German or European
Patent Office (with destination Germany)
25%
20%
15%
Chemistry
10%
5%
Electrical engineering
0%
1885 1895 1905 1915 1925 1935 1945 1955 1965 1975 1985 1995
Note: For Patent Offices and averaging see Figure 11.10.
Figure 11.11 Share of the total patent publications about national patent
applications held by electrical technology and chemistry
During the period of reconstruction after World War II, the number of
patents in the area of chemistry increased, reaching more than 20 per cent,
but was hit by a severe crisis following 1967. Since approximately 1970 –
when the proportion had dropped to less than 10 per cent – the share of the
total number of patents held showed moderate growth to approximately
12 per cent. These high shares of national patent inventory reveal that the
fields of electrical engineering and chemistry were and remain leading indus-
trial sectors. They show above-average expansion, and their global economic
importance is characterised by continuous growth (Ziegler 2000, p. 240).
Analogous with electrical technology, chemistry can also be divided into
patent-statistical classifications (for example organic chemistry, plastics,
pharmaceutical substances, biochemistry, detergents and agrarian chem-
istry. Innovation activities in these six fields show extremely different
degrees of development. Both of the first two fields show a continuous
increase in the total number of chemistry patents (in spite of cyclical
differences in detail), whereas the smaller fields of chemistry, described in
the category ‘other fields’, which was very significant from approximately
1880, started to grow from about 1970 (Figures 11.12 and 11.13). This only
points to different definitions concerning ‘other fields’.
An attempt to subdivide these ‘other fields’ shows the spectacularly high
share of biochemistry in the total number of chemistry patent applications
prior to the turn of the century. This surprising development is explained
45%
40%
Organic chemistry
35%
All other
30%
25%
20%
15%
10%
Resins
5%
0%
1883 1893 1903 1913 1923 1933 1943 1953 1963 1973 1983 1993
Note: For sources and averaging see Figure 11.10.
Figure 11.12 A selection of patent shares of the whole field of chemistry

by a comparison of patent classes included in the definition biochemistry

in the years 1900 and 1998: according to our definition, biochemistry was
dedicated to the food industry in 1900. Between the turn of the century and
the First World War, the food industry based on biochemistry reached its
peak: living organisms were used in the most varied ways. At that time,
biotechnological procedures competed with chemical-synthetic procedures
in the field of production.
According to Marschall (1999, p. 280), prior to the First World War
a ‘path dependency’ arose in German chemistry, which supported the evolu-
tionary choice for chemical synthesis. This path dependency was based
on subsidised investments in high pressure technology, as well as on its use by
the chemical industry. Therefore, inventions in biochemistry were repressed.
Nevertheless, it must be noted that brewing beer accounted for almost
50 per cent of all inventions in the area of biochemistry around 1900 (only a
few per cent today), whereas second place was held by the extraction, refining
and conservation of fat (today, also only a few per cent). In 1998, however,
the dominating area included devices for enzymology or microbiology, and
second place was held by microorganisms or enzymes (Figure 11.14).
The difference between electrical technology and chemistry also comes
from the significance of foreign inventors. According to the same criteria
as above, patents can also be classified according to their country of origin.
35%
30%
25%
20%
Biochemistry
15%
Pharmaceutical
10%
Agro-chemistry
5%
Detergents
0%
1883 1893 1903 1913 1923 1933 1943 1953 1963 1973 1983 1993
Note: For sources and averaging see Figure 11.10.
Figure 11.13 Breakdown of the ‘other fields’ in the total number of patent
publications in the area of chemistry
100%
Fat, oil and wax
90%
Alcoholic beverages
80% Vinegar and its processing
Processing of enzymes or
70%
micro-organisms
60% Side products of fermented solutions
Micro-organisms or enzymes
50%
Pitching
40%
Processing of alcoholic beverages
30%
Processing of fat
20%
Beer brewing
10%
0%
Biochemistry 1900 Biochemistry 1998
Figure 11.14 Breakdown of patent shares within biochemistry in 1900

and 1998
The results show that more important foreign inventions were realised in
the area of electrical technology than in chemistry. The proportion of
foreign inventors in the area of electrical technology corresponds approxi-
mately with the average of all technical areas; that is, it regressed from more
than 35 per cent (1902) to approximately 20 per cent by 1928 and to
10 per cent by 1942. In the area of chemistry, the situation is different:
foreign inventors always played a minor role (1902: 20 per cent), however,
this ratio was slightly reduced (1942: 13 per cent).
Regarding the historical statistics of the selected sectors, it is concluded
that, under conditions of a relatively low level of knowledge exchange with
foreign countries, a branch suffers less from periods of crises and autarky
and that, on the other hand, intense knowledge exchange with foreign
countries leads to correspondingly sensible change. This will certainly have
important effects on the organisation of innovation activities, which should
be analysed on the level of individual institutions (companies as well as
research institutes).
The patent statistics of the German Democratic Republic can also be
divided both on a technical and on an institutional level. For example,
studies exist showing the extent to which the Humboldt University of East
Berlin applied for patents in the areas of chemistry and electrical technol-
ogy. According to this, chemistry had become one of the most prominent
research areas of the GDR in approximately 1990, whereas electrical
technology lost significance (Albrecht et al. 1991, p. 107). Identical analy-

ses were submitted for central institutes, other centres and companies
(Albrecht 1991, Hinze and Grupp 1995, p. 59). The function of universities
in relation to technological development was very important not only for
West Germany but also in the GDR.
4. DISCUSSION AND CONCLUSIONS
This look back into the historical times of innovation reveals many inter-
esting perspectives: for instance the present globalisation trends in R&D
may now be interpreted as a renaissance of the times around 1900: before
the autarky and war situations in national socialist Germany the innovation
system was internationalised in a similar way as today but possibly not to
the same quantitative extent. Yet the logistic and travel possibilities for
exchange of knowledge are much better today than a hundred years ago.
Most astonishingly the German innovation system was very stable
although it witnessed several political system changes in the past century.
The total amount of government spending on science and innovation fol-
lowed similar quantitative tracks after its formation in the 19th century,
after the First World War and after the Second World War. The respective
central powers were no strong pillars in science and technology. On the
contrary, the science and technology operation was maintained and recon-
structed by the German states before the central power found ways to estab-
lish itself as dominating. However, considerable differences are observed
when regarding the strong role of enterprises in innovation after the Second
World War which was – in pecuniary terms – not as visible before. Only
after reunification in 1990 was the acting power the federal government at
a time when enterprises were largely dominating the financing of R&D.
This was definitely different a hundred years ago.
In terms of the basic sectoral structures in science and technology, the
strong and the weak sides were almost the same whatever regime and terri-
torial boundaries existed. This persistence of the innovation system points
to a resistant innovation culture in and around Germany which may not be
influenced too much by external shocks or incentives be it in monetary or
institutional form. If technology and innovation policy intends to change
the German innovation culture in its basics one probably needs other gov-
ernment methods than those being used up to today. Even the isolation of
the former GDR and its subjection under the communist regime could not
change much.
There seems to be a specific German understanding of the opening and
development of technology trajectories. The industrial research system in
Germany was one of the first in the world to be formed and developed.
Other countries followed that pattern more or less closely. Yet the subjects
of research seemed to be different between the countries and remained
largely constant over long periods. Obviously the technical and scientific
elites in Germany succeeded in following their interests in any political
system collectively. For the research and education policy this means that
soft factors like group identity, schools of thought and personal exchange
are more reliable and more efficient government instruments than the trad-
itional monetary incentive systems. This sustainable culture imprint can
only be analysed and detected in historical time series.
Innovation is no ‘reality’ whatsoever, and it is not considered as existing
‘in itself’; instead, innovation activities are only constituted by the specific
method of scientific approach. An historical object of investigation about
innovation activities is not ‘given’, instead, it is ‘imposed’ on the researcher,
that is, he/she must establish a measurement concept. An operationalisa-
tion considered as ideal could lead to an exorbitant cost of collection,
whereas a less adequate method could have favourable effects on both the
process and the results of measurement. This applies particularly to evolu-
tionary innovation research, where results are not always shown by a
formal mathematical model. Here, the keyword of an ‘appreciative’ theory
was created. Based on such an approach, absolutely no immediate, con-
stituent measuring directives can be derived either. Consequently, it is
necessary to construct corresponding indicators for insufficient ‘tailor-
made’ theoretical constructs. Although this procedure could be considered
wearisome, there is no other option.
The suggested range of indicators on a national or sectoral level gives a
detailed impression of both the extent and the contents of innovation activ-
ities during more than the past hundred years. The empirical base which
evolutionary researchers interested in innovation and economic history-
related questions can rely on, was broadened to a large extent, so that there
is no longer the possibility of a serious empirical gap. On the one hand,
many of the questions raised by evolutionary theory cannot yet be studied
on an empirical level. On the other hand, however, the inverse is equally
correct: empirical findings have become available, discovering structures
which could be picked up by the theoretical side for further construction of
theoretical thought.
On the future research agenda should be many more such longitudinal
studies of innovation systems. The basic findings for Germany should be
compared to other countries which possibly suffered less from territorial
and political changes. The data used in this article should exist in other
countries as well and may be brought to the surface. Also we need more sec-
toral studies in order to work out typologies of innovation development
over long periods. Altogether results achieved so far should encourage

more cliometric research all over the world.
NOTES
1. This chapter was drafted and presented to the International Joseph A. Schumpeter
Conference in April 2002. An earlier abridged version was published in Shavinina (2003).
The chapter is based on a much more detailed volume by the same authors in German
language (Grupp et al. 2002).
2. See also the contributions included in Müller and Rüsen (1997).
3. Cf. Machlup (1960/61), Grohmann (1988) and Grupp (1998).
4. According to these (OECD 1992, 1993), technological innovations comprise new prod-
ucts and processes and significant technological changes of products and processes.
5. Maddison (1982, p. 2 and p. 83) talks about system shocks. Gerschenkron (1943) points
to the pervasive institutional powers that may overcome external shocks for decades.
6. More precisely, ‘research share’ means the ‘R&D share’ of the total expenditure for
science and technology.
7. Personal communication Dr. Eugene Garfield, founder of the Institute for Scientific
Information, Philadelphia, 14 October 2000.
8. Due to the delay in appearance of scientific publications following submission, no quan-
titative cutback in literature production by the researchers of German Democratic
Republic institutes can be perceived until the end of 1990 (Weingart et al. 1991, p. 4).
9. Recently, surprisingly similar results were also found using other methods.
10. In 1928, 358 member companies were registered in the central association of elec-
trical technology and electrical industry (Zentralverband Elektrotechnik- und
Elektroindustrie, ZVEI).
REFERENCES
Albrecht, E., O. Dohnert, M. Schneider and H. Bourcevet (1991). ‘DDR-forschung
im internationalen vergleich unter zugrundelegung der patentstatistik’, Report
for the Science Council, Sektion Wissenschaftstheorie und Wissenschaftsorgan-
isation, Berlin: Humboldt University.
Anonymous (E. W. H.) (1925). ‘Output of scientific papers’, Nature 116 (2908),
pp. 129–30.
Anonymous (1986). ‘Das kaufkraftverhältnis zwischen d-mark und mark in der
DDR 1985’, DIW-Wochenbericht 21/86, 53, pp. 259–68.
Bannock, G., R.E. Baxter, and E. Davis (1998). Dictionary of Economics, 6th
edition, London: Penguin.
Bernal, J.D. (1939). The Social Function of Science, London: George Routledge &
Sons.
Boch, R. (ed.) (1999). Patentschutz und Innovation in Geschichte und Gegenwart,
Frankfurt a. M.: Peter Lang.
Dohrn-van Rossum, G. (1999). ‘Erfinder und erfinderschutz im spätmittelalter und
in der frühen neuzeit’, in Boch (1999), pp. 39–50.
Erker, P. (1990). ‘Die verwissenschaftlichung der industrie: zur geschichte der indus-
trieforschung in den europäischen und amerikanischen elektrokonzernen
1890–1930’, Zeitschrift für Unternehmensgeschichte 35, pp. 73–94.
Fogel, R.W. (1964). Railroads and American Economic Growth, Baltimore: The
Johns Hopkins Press.
Freeman, C. (1969). ‘Measurement of output of research and experimental devel-
opment’, Statistical Reports and Studies, Paris: UNESCO.
Freeman, C. (1992). The Economics of Hope, London: Pinter Publishers.
Freimann, K.-D. and A.E. Ott (eds) (1988). Theorie und Empirie in der
Wirtschaftsforschung, Tübingen: J.C.B. Mohr.
Gascoigne, R. (1992). ‘The historical demography of the scientific community’,
1450–1900, Social Studies of Science, 22, pp. 545–73.
Gerschenkron, A. (1943). Bread and Democracy in Germany, Ithaca: Cornell
University Press.
Grohmann, H. (1998). ‘Die statistische Adäquation als Postulat einer
sachgerechten Abstimmung zwischen Theorie und Empirie’, in Freimann und
Ott (1988), pp. 25–42.
Grupp, H. (1998). Foundations of the Economics of Innovation – Theory,
Measurement and Practice, Cheltenham, UK and Lyme, USA: Edward Elgar.
Grupp, H. and U. Schmoch (1992). Wissenschaftsbindung der Technik, Heidelberg:
Physica-Verlag.
Grupp, H., U. Schmoch and S. Hinze (2001). ‘International alignment and scientific
regard as macroindicators for international comparisons of publications’,
Scientometrics 51(2), 359–80.
Grupp, H., I. Domínguez Lacasa and M. Fredrich–Nishio (2002), Das Deutsch
Innovationssystem seit der Reichsgründung, Heidelberg: Springer–Physica
Publishers.
Hetmeier, H.-W. (1990). ‘Öffentliche ausgaben für forschung und experimentelle
entwicklung 1987’, Wirtschaft und Statistik 2/1990, pp. 123–29.
Hinze, S. and H. Grupp (1995). ‘Ein rückblick auf wissenschaft und technik in der
ehemaligen DDR: ostdeutschlands forschungs- und entwicklungspotential’, in
Holland and Kuhlmann (1995), pp. 41–86.
Hoffmann, W.G. (1965). Das Wachstum der Deutschen Wirtschaft seit der Mitte des
19. Jahrhunderts, Berlin: Springer.
Holland, D. and S. Kuhlmann (1995). Systemwandel und industrielle Innovation,
Heidelberg: Physica.
Hoyningen-Huene, P. (1999). ‘Kommt die physik der wahrheit immer näher?’
Physikalische Blätter 55(3), 56–8.
Irvine, J., B.R. Martin, and A. Isard, Ph. (1990). Investing in the Future, Aldershot,
UK and Brookfield, US: Edward Elgar.
König, W. (1995). Technikwissenschaften – Die Entstehung der Elektrotechnik aus
Industrie und Wissenschaft zwischen 1880 und 1914, Chur: G+B Verlag Fakultas.
Lorenz, Ch. (1997). Konstruktion der Vergangenheit, Köln: Böhlau.
Mabe, M. and M. Amin (2001). ‘Growth dynamics of scholarly and scientific jour-
nals’, Scientometrics 51(1), 147–62.
Machlup, F. (1960/61). ‘Idealtypus, wirklichkeit und konstruktion’, Ordo – Jahrbuch
für die Ordnung von Wirtschaft und Gesellschaft 12, 21–57.
Maddison, A. (1982). Phases of Capitalist Development, Oxford: Oxford University
Press.
Marschall, L. (1999). ‘Industrielle Biotechnologie im 20. Jahrhundert –
Technologische Alternative oder Nischentechnologie?’ Technikgeschichte 66, (4),
277–93.
Müller, K.E. and J. Rüsen, (eds.) (1997). Historische Sinnbildung, Reinbek: Rowohlt.
North, M. (ed.) (2000a). Deutsche Wirtschaftsgeschichte, München: C.H. Beck.
North, M. (2000b) Einleitung, in North (2000a), pp. 11–14.
OECD (1992). Oslo Manual: Proposed Guidelines for Collecting and Interpreting
Technological Innovation Data, Paris: OECD.
OECD (1993). Frascati Manual 1992. 5th Revision, Paris: OECD.
Pfetsch, F.R. (1974). Zur Entwicklung der Wissenschaftspolitik in Deutschland
1750–1914, Berlin: Duncker & Humblot.
Pfetsch, F.R. (1982). Datenhandbuch zur Wissenschaftsentwicklung, Köln: Zentrum
für historische Sozialforschung.
Roscher, W. (1886). Grundlagen der Nationalökonomie, 18th ed., Stuttgart:
J.G. Cotta’sche Buchhandlung.
Schmoch, U., G. Licht, and M. Reinhard (2000). Wissens- und Technologietransfer
in Deutschland, Stuttgart: Fraunhofer IRB Verlag.
Schmoch, U. and U. Sass (2000). ‘Erfassung der technologischen Leistungsfähigkeit
der östlichen Bundesländer mit Hilfe von Patentindikatoren’, Report for the
Federal Ministry BMBF, Karlsruhe: FhG-ISI.
Schumpeter, J.A. (1911/1964). Theorie der wirtschaftlichen Entwicklung, München
and Leipzig, quoted from the 6th edition, Berlin: Duncker & Humblot.
Schumpeter, J.A. (1942/1975). Capitalism, Socialism and Democracy, New York,
quoted from the German translation, 4th edition, München: Francke.
Shannina, L.V. (2003) (ed). International Handbook on Innovation, New York:
Pergamon–Elsevier, pp. 1018–43.
Smith, J.K. jr (1990). The Scientific Tradition in American Industrial Research,
Technology & Culture 31, 121–31.
Stolper, G., K. Häuser, and K. Borchardt (1964). Deutsche Wirtschaft seit 1870,
Tübingen: J.C.B. Mohr.
Struve, G. (1849/1980). Geschichte der drei Volkserhebungen in Baden 1848/1849,
Bern: von Jenni; quoted from the revised reprint, Freiburg: Rombach.
Testa, J. (1997). The ISI Database: The Journal Selection Process, Essay No. 1,
Philadelphia, Institute of Scientific Information.
Wagner-Döbler, R. and J. Berg (1996). ‘Nineteenth-century mathematics in the
mirror of its literature: a quantitative approach’, Historia Mathematica
23, 288–318.
Weingart, P., Strate, J. and M. Winterhager (1991). Bibliometrisches Profil der
DDR, Report for the Science Council, Universitätsschwerpunkt Wissenschafts-
forschung, University of Bielefeld.
Wengenroth, U. (1997). Der Beitrag des Staates zu Forschung und Entwicklung in
Deutschland seit der Reichsgründung, Manuscript.
Ziegler, D. (2000). Das Zeitalter der Industrialisierung, in North (2000a),
pp. 102–281.
12. Emergence and diffusion of
disastrous innovations – a case
study
Reiner Peter Hellbrück1
INTRODUCTION
The central statements are the following: (a) If the ‘circumstances corres-
pond with each other’, then and only then there can be innovations.2 Thesis
(a) says that there have to be various conditions present and met so that
their combination is sufficient for an innovation. The following conditions
are met in the case study: (b) the increase of knowledge is essential,
nonetheless it is not yet sufficient to explain the emergence and diffusion of
the organisational innovation; (c) the organisational innovation corres-
ponds to the preferences of decision makers and supporters; (d) the danger
of disastrous consequences for competing firms adopting an organisational
innovation are not sufficient to prevent its diffusion; (e) regarding the
broader consensus about the societal advantages of an organisational inno-
vation, the potential harm to business can be neutralised and moreover
transformed into a competitive edge.
Case studies are only rare in economics. Usually the method of the par-
ticipating observation is chosen. This case is different; it relates to the
author’s own personal work experience in the German health-care system.
This working experience relates to the years 1998 until 2000 and the author
also participated actively in some projects. The time-frame before July
1998, which is the main focus of this analysis, has neither been actively
influenced nor affected by the author.
The focus is an organisational innovation and accompanying improve-
ment innovations in diabetes care in Germany between the years 1993 and
2000. This care needs, according to experts’ opinion, to be improved (Mehl,
Becker-Berke und Müller-de-Cornejo 2000, p. 6 and Bertram 1999). An
impression of problems caused by diabetes shows the following facts:
approximately 1800 persons become blind annually, 9000 diabetes patients
need to undergo enduring dialysis, 28 000 patients need amputations, and
274
Emergence and diffusion of disastrous innovations 275
there is a three to four times higher risk of coronary heart diseases. Eighty
per cent of the people with diabetes receive orally administered antidiabetic
medical treatment (Laschet 1998).
Due to the problems of diabetes, in early 1993 contracts concerning
diabetes patients were concluded between some health insurance funds
and Regional Associations of Statutory Health Insurance Physicians
(Kassenärztliche Vereinigung) which can be seen as precursors of rules
and regulations to the current contracts (Mehl, Becker-Berke und Müller-
de-Cornejo 2000, p. 9). Nowadays, contracts concerning better supply for
diabetes patients can be found in each of Germany’s federal states: between
the years 1993 and 1998, 23 diabetes contracts were counted (Landgraf et al.
1999, ii). However, contracts can be distinguished between old and new
types where the latter cover more knowledge about the long-run problems
and consequences of diabetes. Information about the disease diabetes as well
as the characteristics of different kinds of diabetes contracts will be summed
up in the upcoming section ‘Diabetes Contracts’.
In the year 1997, additional regulations were introduced by the Statutory
Health Insurance Reorganisation laws (GKV3-Neuordnungsgesetze, GKV-
NOG1 and GKV-NOG2). These revisions particularly concerned the
so-called ‘Modellvorhaben’ (model projects) and ‘Strukturverträge’ (struc-
ture contracts) whereby health insurance funds and Regional Associations
of Statutory Health Insurance Physicians were permitted projects to try
new organisational forms which were not allowed under the regular law.
These changes did not however trigger the new diabetes contracts. These
changes in regulation will be described later on in the ‘Regulation’ section
of the chapter.
Surprisingly, the organisational innovation was not introduced by the
Allgemeine Ortskrankenkasse (AOK) (a big health insurance fund) which
had the main influence in the development and introduction of the ‘second-
generation’ diabetes contracts. The model was brought up first by an AOK
competitor, the Barmer Ersatzkasse (BEK). The introduction of this new
idea will be further discussed and explained in the ‘Emergence’ section.
Additional health insurance funds, the innovator health insurance fund
AOK included, are still practising in the ‘second-generation’ diabetes con-
tracts. In the ‘Diffusion’ section the question of why the diabetes contracts
were disastrous will be discussed and why those types of contracts still
diffused. The section ‘Continuation’ is dedicated to the question why the
diabetes contracts were not simply cancelled and which strategies were
adopted to avoid harmful consequences or even to switch the disadvantage
into a competitive edge. The most important insights are summed up in the
section ‘Conclusions’.
DIABETES CONTRACTS
Diabetes is an illness present in virtually every family. Approximately

5 per cent of the German population suffers from diabetes. Diverse illness
patterns are the reason for distinguishing at least two types of diabetes:
insulin-dependent diabetes mellitus (IDDM,) and non-insulin-dependent
diabetes mellitus (NIDDM). The last mentioned is the most common type
of diabetes and generally known as ‘diabetes of the ageing’ because this
type emerges at about 35 years of age. In this case the disproportionate level
of needed and available insulin usually has its reason in patients being over-
weight. One treatment is the reduction of the overweight condition.
IDDM has its origin in the malfunction of the human pancreas. This
means that the demolished pancreas does not produce any insulin at all and
it is speculated that this is caused by an auto-immune reaction. Accordingly,
diabetes patients are dependent on the supply of insulin. It could be that a
person with NIDDM becomes dependent on insulin as well but this person
will not be filed as IDDM.
In general, IDDM is mostly diagnosed in the younger years: from child-
hood age up to 40 years; NIDDM manifests between the ages of 35 and
older. Around 20 per cent of the population between the ages of 65 and
75 have diabetes. Diabetes is a chronic sickness which is correlated with the
following diseases: cerebral apoplexy, cardiac infarct, blindness, kidney
failure and amputation of toes, feet, and legs. For example, the most
common main diagnosis of the AOK Magdeburg (a health insurance fund
belonging to the community AOK)4 is diabetes mellitus with approximately
3 per cent of all hospital cases (Robra et al. 1996). Diabetes mellitus is seen
as an illness where methods for an effective treatment are available to delay
or to avoid the expensive treatment of subsequent diseases. This is one
reason why competing health insurance funds trying to cut costs are inter-
ested in improving diabetes care.
In the beginning, diabetes contracts contained only additional services
of local doctors and/or hospitals, which were concluded between Regional
Associations of Statutory Health Insurance Physicians (Kassenärzlichen
Vereinigungen) and health insurance funds. The reason for this kind of
contract was the insight concerning NIDDM patients that there is an
inequality between needed and produced insulin. Hence, programmes were
developed to improve the metabolism of diabetic patients (first-generation
diabetes contracts). Usually, in these contracts, an additional amount had
to be paid for such services.
Occasionally, so-called transfer fees were negotiated as well which means
that in the case of a patient’s transfer to a diabetes specialist, the referring
family doctor will receive an additional compensation. Transfer lump sums
were negotiated because family doctors feared to lose patients to specialists

which would result in income losses. To compensate this fear, transfer fees
were paid.
Second-generation diabetes contracts have the following characteristics.
Additionally to the transfer fees of family doctors, return transfer fees for
diabetes specialists have been negotiated. As the fear of family doctors was
revealed to be real, return transfer fees should give diabetes specialists
the incentive to refer back to the family doctor. On the other side, these
fees should only be paid if certain transfer criteria, defined in the diabetes
contract, were achieved. The transfer of a diabetes patient from the
family doctor to a specialist in case of further vital treatment (training
and examinations of the patient) should be obligatory. For example, it
should be immediate as soon as the HbA/1c ratio shows a higher rate than
7.5 per cent within two or three sequential three-month periods or if the
blood pressure exceeds certain limits determined in the diabetes contract.
On the other side, there should also be a transferral of the patient from the
specialist back to the family doctor as soon as the indispensable treatment
was completed. Furthermore, criteria and instructions for treatment by
family doctors and diabetes specialists respectively were contractually
determined, so-called treatment instructions, which corresponded with
the transfer criteria. Transfer criteria and treatment instructions created
the treatment corridors (Scherbaum 1997, p. 45 presents a version differing
considerably from the ones actually used). First-generation and second-
generation diabetes contracts differ from each other in two factors. First,
return transfer fees which should secure the patient’s return from the
specialist to the family doctor. The other factor is the definition of the
treatment corridors.
REGULATION
The organisation of health insurance funds and health-care providers, the

claims of the insured, as well as the relationships between insured, health
insurance fund, and health-care provider are regulated by law and add-
itional contracts. Health insurance funds, Regional Associations of
Statutory Health Insurance Physicians and licensed hospitals are only
allowed to close contracts based on laws and degrees (Verordnungen) as
well as skeleton agreements (Rahmenvereinbarungen) concluded between
their federal or regional associations. The most important source is the fifth
book of the social legislation (i.e. Sozialgesetzbuch V, abbreviated by
SGB V); for in-patient treatment additional legislation being relevant: the
hospital financing law (Krankenhausfinanzierungsgesetz (KHG)) and the
decree to regulate hospital financing in detail (Bundespflegeversatzordnung

(BpflV)), especially the regulation of hospital budgets and fees.
Within the last few years, the most important rules and regulations
in relation to diabetes contracts are the health-structure law (Gesund-
heitsstrukturgesetz (GStruktG)) of 12 December 1992,5 the first Statutory
Health Insurance Reorganisation law (GKV-NOG1) of 23 June 1997, the
second Statutory Health Insurance Reorganisation law (GKV-NOG2) of
23 June 1997, the Statutory Health Insurance Solidarity Stabilisation Law
(GKV-Solidaritsstärkungsgesetz) of 19 December 1998, as well as
the Statutory Health Insurance Health reform law 2000 (GKV-
Gesundheitsreformgesetz) of 22 December 1999. Especially the sections
§§63–65 and §73a SGB V 1997 were the pillars for concluding diabetes
contracts. Starting from 1 January 2000, a new type of contracts was made
possible with paragraphs §§140a–h SGB V 2001. Since this change does not
lie in the inspection period, this chapter will not further explain this
voluminous regulation.
As early as 1993, health insurance funds had the opportunity to try new
models and research.
To improve and develop statutory health insurance, health insurance funds can
examine benefits, measures, and procedures (even as models) based on the regu-
lation of this section. Its goal should be to implement incentives to supply
benefits at a reasonable price and not to waste scarce resources. (§63, SGB V
1993)
For the most part, the legislator had the implementation of elements of
private health insurance in mind. For example, according to §65 SGB V
1993, an opportunity for health insurance funds was given, saying they
could choose reimbursement of expenses instead of benefits in kind6 to
their insured.
Sections §§63 to 64 of the SGB V were completely restated in 1997 by the
GKV-NOG2. In particular, the objective had changed through this legal
amendment.
To improve the quality of care and the economical supply of health care, health
insurance funds and their associations can either carry out or arrange models
according to §64 concerning procedures, organisation, financing, and reim-
bursement within the framework of its legal task. (§63 SGB V 1997)
In contracts which were concluded on the basis of §§63 and/or 64 SGB V

1997 deviations from Chapter 4 SGB V 1997, the hospital financing law and
all regulations based on them were allowed. The financing of out-patient
treatment as well as the medication area are regulated in Chapter 4 SGB V
1997. Compensation of in-patient treatment is also regulated in Chapter 4

SGB V 1997, the hospital financing law (KHG 2000), and in a decree called
‘Bundespflegesatzverordnung’ (BpflV 2000).
However, section §§63 SGB V 1997 is linked to contribution-rate stabil-
ity (Beitragsatzstabilität) which had its validity in models based on §§63 and
64 SGB V 1997. The principle of contribution-rate stability states that:
Concerted action gives recommendations to the treatment sectors and also to the
variation of compensation. Recommendations can be given for the duration of
one year and for a longer period too. The recommendations have to be formed
such that an increase of the contribution rate will be avoided except where the
necessary medical care cannot be guaranteed even if scarce resources are being
used effectively (principle of contribution-rate stability). (§141, 2 SGB V 1997)
An additional characteristic of regulation §§63 and 64 SGB V 1997 was that

models were limited to a time-frame of eight years. On the other side, struc-
ture contracts (Strukturverträge) introduced by the GKV-NOG2 1997
opened the possibility to restructure out-patient medical care permanently.
Regional Associations of Statutory Health Insurance Physicians are able to

agree upon supply and compensation structures with Regional Associations of
Health Insurance Funds and Associations of Statutory Substitution Health
Insurance Funds (Verbände der Ersatzkassen) with the aim to give the family
doctor who was chosen by the patient responsibility for the quality of medical
care and economical supply as well as responsibility for prescribed and disposed
services. . . ., . . . . (§73a SGB V 1997)
This paragraph was included in the SGB V 1997 due to some activities of
the Federal Association of Health Insurance Funds BKK (BKK means
firm health insurance fund). In the mid 90s, this association set up a ‘phys-
ician network with combined budgets’ (Müller, Richard, and Schönbach
1999). Its aim was the realisation of cost savings through the co-ordination
of medical activities.
In 1993, the health-structure law brought an additional important reform
for health insurance funds. The responsibilities of health insurance funds
and members’ voting rights were changed so that future competition
between health insurance funds became more and more fierce. However, this
competition between health insurance funds did not take place until the
year 1997. Without additional regulation it was feared that health insurance
funds would compete for insurants with low risk of sickness. To prevent this,
the Risk Structure Compensation Scheme (Risikostrukturausgleich abbre-
viated by RSA) was introduced (§266 SGB V 1993). No matter what char-
acteristics a person admissible to the German Statutory Health System has,
health insurance funds should have no incentive to prefer low-risk insurants.
With the GKV-NOG2, competition among health insurance funds has

intensified. Regulation §175, 4 sentence 3 SGB V was changed into the fol-
lowing: ‘If a health insurance fund increases its contribution rate or
changes benefits as part of its conduct, a termination of the membership
for the following month is possible, . . ., with a time period of a month
beginning with the end of the following month of the day of the increase
or the service change’( SGB V 1997). Moreover, if a health insurance fund
increased its contribution rate, their insurants had to pay higher co-
payments too. However, the legislator also defined that co-payments of the
insurant would shrink, if their health insurance fund was going to reduce
its contribution rate (§221 SGB V 1997). Regulation §221 SGB V 1997 was
cancelled by the Solidarity Stabilisation Law of 19 December 1998.
Anyway, due to fierce competition among health insurance funds, the
contribution rate became the most important decision variable for them.
Nevertheless, migrations from health insurance funds with high contribu-
tion rates to those with low contribution rates appeared and took place very
quickly. Some persons of the Federal Association of Health Insurance
Funds AOK and others (Andersen and Schwarze 1999) feared to attract
bad risks as a result of managed care.
EMERGENCE
Diabetes contracts of the first generation were signed and closed before and
after the year 1997, when competition between health insurance funds
intensified considerably (Meierjürgen 1994). Mehl, Becker-Berke and
Müller de Cornejo (2000) talk about the precursor of the second-generation
diabetes contracts: diabetes contracts in Brandenburg and Saxonia for the
year 1993. These contracts already included extra budgeting payments to
furnish diabetic services and training achievements. The first diabetes con-
tract of the second generation took effect in 1998; it was concluded between
the BEK (innovator health insurance fund) and the Associations of
Statutory Health Insurance Physicians in Westphalia-Lip. Shortly after
that, the innovator fund AOK concluded a second-generation diabetes con-
tract with a minor health insurance fund (agricultural health insurance
company) and the Associations of Statutory Health Insurance Physicians
in Thuringia.
However, why did the diabetes contracts come into existence at all? There
were several reasons. First of all, diabetes was back in everyone’s mind and
interest due to the St Vincent Declaration. Moreover, through the fall of
the Berlin Wall in East Germany a development was unleashed which led
to the first-generation diabetes contracts. Insufficiencies and inefficiencies
of first-generation diabetes contracts gave enough reasons to revise the con-

tracts. The new contracts were founded on evidence-based medicine
(EbM), which means medical treatment upon medical, economic and
scientific evidence. The basic idea of the second-generation diabetes con-
tracts cannot be assigned retroactively to a certain person, but individual
concern was a strong stimulus for the great personal engagement in the
development process of diabetes contracts. The origin of those contracts
can be described in detail as follows.
1. In 1989 the so-called ‘St Vincent Declaration’ was published. In this

declaration goals were formulated to be achieved within five years for
improving diabetes care. The goals were discussed, accepted and pub-
lished by European government representatives, diabetes physicians,
and representatives of patient associations in the town of St Vincent,
Italy in 1989. For example, this proclamation demanded that within
five years late consequences of diabetes should decline by one-third.
These demands were supported by the World Health Organisation
(WHO) and in Germany were also included in political discussions
concerning health issues (Hellbrück 1997, pp. 27–8).
2. In former East Germany, the care of diabetic patients took place in
special medical units, called ‘Dispensaires’ (not to be confused with dis-
pensaries); family doctors were not involved in diabetes care. In the old
states of West Germany, diabetes care was and still is a common task
of local family doctors. After the fall of the Berlin Wall, the Western
health-care system was spread over the new German states; now dia-
betes care were also delivered by family doctors and not any more by
Dispensaires. Due to the introduction of this new system, health insur-
ance funds feared a decline in quality of diabetes care in the new lands.
This fear led to the first-generation diabetes contracts between the
AOK and Associations of Statutory Health Insurance Physicians in
the states of Brandenburg and Saxonia in 1993. The structure of dia-
betes care of the former East Germany was copied and additional dia-
betes centres were established.
3. However, a lack of co-operation between family physicians and spe-
cialists for diabetes care as well as between the out-patient and
in-patient area was found (Mehl, Becker-Berke und Müller-de-Cornejo
2000). One reason for this lack of collaboration between the two kinds
of physicians lay in the fear of the family doctor to lose patients
through a transferral to a speciality physician which could result in a
permanent income loss for the family doctor. To avoid this, family
doctors referred diabetic patients to hospitals if patients’ metabolic
attitude visibly deteriorated or because of illnesses complications or for
pure economic reasons. Then, family doctors felt certain not to lose
patients for long periods to more highly qualified persons (specialist
physicians).
4. Parallel to this development, alternatives for NIDDM treatment were
internationally examined for years in a long-term study (especially in
Great Britain) in the framework of the famous UKPDS (start of the
recruiting of test persons in 1977). This study showed that a better
metabolism leads automatically to a significant reduction of subse-
quent diabetes diseases. For example, with the exception of cataracts
it showed that diabetes-dependent complications could be decreased
as a result of a person’s lower blood pressure (Adler, Stratton, Neil
and Yudkin 2000). This study confirmed hypotheses held for a long
time and based on epidemiological studies (Ärzte-Zeitung 1998). In
1993, a guideline for the treatment of IDDM and a standard for the
treatment of NIDDM already existed. At the beginning of the 90s and
based on requests by a board of experts (Sachverständigenrat für
die konzertierte Aktion im Gesundheitswesen) this progress was
handed to German scientific-medical research institutes (AWMF) to
develop medical guidelines. Since then, the AWMFs have already
published several hundred guidelines. Due to the diverse quality of the
established AWMF guidelines, the Federal Medical Association
(Bundesärztekammer) and the Federal Association of Statutory
Health Insurance Physicians (Kassenärztliche Bundesvereinigung)
founded the Centre for Quality Assurance (Ärztliche Zentralstelle
Qualitätssicherung) in the mid 90s to ensure the quality of medical
guidelines (Ärztliche Zentralstelle Qualitätssicherung 1998). This
health-political development resulted in an increasing interest by
German physicians in assuring quality care (see also Hellbrück 1997 for
additional information concerning quality assurance from an eco-
nomic point of view).
5. On the initiative of the Federal Association of Health Insurance
Funds AOK (AOK-BV) the Medical Advisory Council on Diabetes
(Wissenschaftsrat Diabetes) was founded. However, a National
Association of Health Insurance Funds may not become active
without legal permission. According to §217 SGB V 2000, a National
Association of Health Insurance Funds has to fulfil its legal tasks. The
legislator firmly determined in §73, 1c SGB V 1993 that the content and
extent of the medical care of family doctors is to be defined by the
Federal Associations of Health Insurance Funds and the Federal
Association of Statutory Health Insurance Physicians. This was the
basis for the Federal Association of Health Insurance Funds AOK for
defining family doctors’ competencies. This paragraph was not only
used as an acting base but also as an essential impulse for the foundation
of the Medical Advisory Council on Diabetes whereby the Federal
Association of Statutory Health Insurance Physicians became, contro-
versially to the legal order, not directly involved.
To found the Medical Advisory Council on Diabetes, some prepara-
tory work started in 1996. Representatives of self-help groups, the pres-
ident and additional representatives of the German Diabetes Society,
the president of the Scientific Society of Family Doctors, representa-
tives of Associations of Statutory Health Insurance Physicians, repre-
sentatives of hospitals and family doctors and diabetes experts, and
representatives of the AOK formed the Medical Advisory Council on
Diabetes. Other health insurance companies were not represented.
With this composition it is still active today, i.e. the organisations men-
tioned send further representatives, however it came to personnel
changes. In January 1997 the Advisory Council on Diabetes began with
the definition of the treatment corridors. The definition of the treat-
ment corridors, which first included the family doctors and diabetic
specialists and which were expanded to the the in-patient sector in
1999, were based on insights of EbM.
6. Without the large stake of the persons involved, the development and
diffusion of the diabetes contracts of the second generation would not
have been possible. In the AOK there were personalities who, due to
personal concern, made the improvement of diabetes care their task.
As diabetes is present in virtually every family, the same incentive may
be present in other organisations too. For many years a representative
of a Health Insurance Fund AOK strove for the improvement of dia-
betes care. For training a representative of the Federal Association of
Health Insurance Funds AOK was sent into the Health Insurance
Fund AOK and made subordinate to the representative mentioned.
Both representatives laid claim to ‘copyright’ in the diabetes agree-
ments of the second generation. From this co-operation and/or from
co-operation of the Federal Association of Health Insurance Funds
AOK with representatives of the Eastern Health Insurance Funds
AOK, the philosophy of the second-generation diabetes contracts was
developed. Hesse’s (1990) idea according to which the emergence of
knowledge passes several selection stages, thus approaches reality.
7. Within the Federal Association of Health Insurance Funds AOK the
diabetes contracts were disputed. The department manager responsible
refused the conclusion of the diabetes contracts of the second-
generation, however without success. How is this to be explained? The
governing board of the Federal Association of Health Insurance
Funds AOK (Verwaltungsrat des AOK Bundesverbandes) had taken
an unequivocal stand on intensified efforts to improve prevention

(on the self-understanding of members of the governing board, see
Henkel (1999) and Becker (1999)). The diabetes contracts, in particu-
lar the diabetes contracts of the second-generation, were welcomed on
the part of the governing board (Nachtigall 1997, p. 18). The manag-
ing committee of the Federal Association of Health Insurance Funds
AOK as well as the managing committee of the Regional Association
of Health Insurance Funds AOK are elected by their respective gov-
erning boards AOK (§§215, 209, 209a SGB V 2000 and §35a, 5 SGB IV
2000).
In particular the governing board has to make all decisions which
are of fundamental importance for its health insurance fund.
Additionally, it has to represent the health insurance fund in relation
to the managing committee. Furthermore, as the members of the gov-
erning board are representatives of the employer or of the employee
side, they have an incentive to represent the interests of the insurants
which explains the preference of the governing board for prevention.
Due to this preference, it is clear why the department manager respon-
sible did not gain a hearing at the managing committee AOK. The gov-
erning board of the Federal Association of Health Insurance Funds
AOK consists of representatives of the governing boards of all
Regional Associations of Health Insurance Funds AOKs. Thus it may
be speculated that the influence of the governing boards on the man-
aging committees of their regional AOKs was similar to that on the
federal level.
Besides, the responsible employee of the Federal Association of
Health Insurance Funds AOK also electioneered outside of his regular
work-time for the diabetes contracts of the second generation. In the
end, virtually all relevant health politicians at federal level and almost
all influential persons in the German Health Care System were
informed about and unusually interested in the new possibilities in
health-care management.
8. As the foundation of the Medical Advisory Council on Diabetes
started in 1996, that is about one year before the Statutory Health
Insurance Reorganisation laws passed in the parliament, it can fairly
be alleged that legal conditions changed, but these changes cannot be
regarded as the cause of the emergence of the diabetes contracts of the
second generation. In 1996 the persons participating in the develop-
ment process of the second-generation diabetes contracts could not
know the concrete contents of the amendment happening one year
later. However the amendment, the so-called third stage of the German
health-care reform, was expected, so that the BKK could operate its
‘physician network with combined budgets’ beyond its testing phase.

As the diabetes contracts can be understood as a special physician
network, that is a diagnosis-specific practice network, justified hope
existed of being able to implement the organisational innovation.
The reasons leading to the emergence of the diabetes contracts of the

second generation can be summarised as follows. Calling attention to a
problem of great importance (St Vincent Declaration, fear of a deterior-
ation of diabetes care in the new lands of Germany); development of a
solution with consideration of existing structures (diabetes agreement of
the first-generation); discovery of the weak points of this solution and
efforts to solve the problems; infusion of new knowledge (new scientific
insights and foundation of the Medical Advisory Council on Diabetes);
large stake of persons concerned with diabetes care and similar preferences
of decision makers, whereby it may be that the preferences of the decision
makers were affected by the efforts of some few persons. Though the chang-
ing law was surely a necessary condition for the innovation, it can hardly
be regarded as its cause.
DIFFUSION
Now the diffusion of the second-generation diabetes contracts is outlined

without laying claim to completeness. In 1998 in Thuringia (AOK and
Agricultural Health Insurance Fund), Suedwuerttemberg (AOK, other
health insurance funds are not known), in Westphalia Lip (BEK), North
Rhine (BKK, later joined by the AOK), Saxonia and Bavaria (all health
insurance funds) second-generation diabetes contracts were concluded.
Why did the innovation diffuse at such a great speed? The fast propaga-
tion within the AOK community is due to the short information paths and
due to the problems of diabetes care in the new lands of Germany. Every
health insurance fund is a public corporation (§4 SGB V 2000), furthermore
it is however, due to §207 SGB V 2000, a conscripted member of a Regional
Association of Health Insurance Funds, and this is in turn an obligatory
member of a Federal Association of Health Insurance Funds (§212 SGB V
2000). One task of the federal associations is, among others, consultation
with and information to its members (§217 SGB V 2000). To fulfil these
obligations, work meetings regularly took place in federal associations, in
which representatives of the federal and regional associations participated,
whereby the information about the organisational innovation spread fast.
How could the information about the new type of diabetes contracts
arrive at other health insurance funds? The innovation could not remain
hidden. In the Medical Advisory Council on Diabetes there were beside sci-
entists and representatives of the AOK also representatives of the Regional
Associations of Statutory Health Insurance Physicians, who were the only
contracting parties of health insurance funds at that time in out-patient
treatment. Furthermore, representatives of family doctors were on the
Medical Advisory Council on Diabetes, and they had expressed a great
interest in a demarcation of the family doctors from medical specialists.
On what legal basis were diabetes agreements concluded? A playing field
was opened to the health insurance funds by allowing model projects
in order to accomplish such projects or to conclude contracts with
Associations of Statutory Health Insurance Physicians for supplying
benefits, for example prevention programmes, which were not statutory
benefits (§63, 2 SGB V 1997). Diabetes contracts of the second generation
were usually concluded on this legal basis. There were two reasons for this:
(1) the financing risks of the health insurance funds were limited to eight
years by law (this rule based on the GKV-NOG2 in 1997); (2) additionally,
without extra budgeting payments, the conclusion of diabetes contracts
would not have been possible and §63 SGB V (1997) created this very pos-
sibility.
Nevertheless contribution-rate stability was to be considered. This
meant that health insurance funds and Regional Associations of Statutory
Health Insurance Physicians had to shape their agreements about remu-
neration in such a way that a rise of the contribution rate was usually
excluded. If necessary medical supply was not possible even after ‘exhaus-
tion of economic reserves’, increasing contribution rates were however
allowed. The reason for this rule was increasing contribution rates in the
past (i.e. before 1997) which, as regards the policy, were sometimes dammed
up by drastic means (for example by the strict budgeting of the in-patient
sector by the Health Structure Law (Tuschen and Quaas 1995).
Contribution-rate stability was however not violated, if there were ‘add-
itional expenses caused by a model project being balanced by savings due
to that model project. This balance has to be proven’ (§63, 3 sentence 2 SGB
V 1997). This formulation permits two interpretations: (a) increasing
expenses and savings temporarily fallen apart or (b) increasing expenses
and savings taking place at the same time. Since the primary reason for the
diabetes contracts was to avoid subsequent diseases, it was to be assumed
that additional expenses result today and possible savings arise tomorrow.
The conclusion of contracts based on §§63–65 SGB V 1997 with extra
budgeting payments would not have been possible with a strict interpret-
ation of the legal basis. The contracting parties (in particular the health
insurance funds) had despite the possibility of the cancellation of regula-
tions laid down in Chapter 4 SGB V 1997 to consider contribution-rate
stability. But the formulation in §63, 3 sentence 2 SGB V 1997 left open
whether savings have to be proven ex post or ex ante. In the diabetes con-
tracts extra budgeting payments were agreed upon that were to be disbursed
during the run-time of the model projects. Thus the financing risk of the dia-
betes contracts had to be borne by the contracting health insurance funds.
On the other hand, to convince their member physicians, the Associations
of Statutory Health Insurance Physicians had to insist on extra budgeting
payments, otherwise the conclusion of second-generation diabetes contracts
would have not been possible. As in 1998 health insurance funds could only
conclude diabetes contracts with Associations of Statutory HealthInsurance
Physicians on the basis of §§63–65 and 73a SGB V (1997), extra budgeting
payments were inevitable for the conclusion of diabetes contracts. Though
§64, 2 SGB V 1997 permitted the conclusion of model projects between
health insurance funds and contract physicians (i.e. physicians allowed to
treat insurants of health insurance funds) without participation of of the
Associations of Statutory Health Insurance Physicians (the so-called quorum
regulation), the condition for this was the definition of the principles for exe-
cuting model projects between Federal Associations of Statutory Health
Insurance Funds and the Federal Association of Statutory Health Insurance
Physicians (§64, 2 sentence 1 SGB V 1997) which was never realised.
Why, however, were diabetes contracts concluded at all although the
financing risk was borne exclusively by the contracting health insurance
funds? It was referred to the large amount of diabetic persons (5 to 6 million
are diagnosed in Germany) and the associated treatment costs as well as
on the associated savings potentials (Mehl, Becker-Berke und Müller-
de-Cornejo 2000, p. 5 and Ahrens 1997). As early as in November 1997,
before the negotiations of the AOK for diabetes contracts in Thuringia
were finished, the chairman of the managing committee of the Federal
Association of Health Insurance Funds AOK did not leave any doubt
during a public event that the AOK would conclude diabetes contracts of
the second-generation (Ahrens 1997, p. 51).
Were the diabetes contracts of the second generation disastrous at all if
evaluated from the viewpoint of the Health Insurance Funds as announced
in the heading of this contribution? The answer to this question is simplified
by defining the term ‘disastrous’. Disastrous innovations are in our context
business-damaging. Being business-damaging has two dimensions: on the
one hand the competition of health insurance funds into private health
insurance companies and on the other hand the competition within the
group of health insurance funds are affected. A health insurance fund
behaves in a business-damaging way, if its competition position worsens
compared to other health insurance funds or to private health insurance
companies.
In order to discover what is business-damaging, it is worthwhile to

differentiate strictly between costs and expenditures. Additionally, it is
necessary to make clear, who incurs costs, who has cost-cuts, whose expen-
ditures increase and whose expenditures are reduced. Diabetes is a costly
illness because of associated subsequent costs. As explained earlier in the
chapter, the most important subsequent diseases are diabetic feet, kidney
failure, cardiac infarct, going blind and cerebral apoplexy. These subse-
quent diseases develop within ten years due to bad metabolic adjustment
and were often treated in hospitals. If a second-generation diabetes con-
tract was to be successful, the improvement of the metabolic adjustment
and accompanying reduction of subsequent diseases and in-patient costs
would take place after several years.
Second-generation diabetes contracts are a means to strengthen preven-
tion to either avoid subsequent diseases or to postpone their development.
Prevention is part of out-patient treatment. Diabetes contracts of the
second generation not only consist of a job-sharing idea between home
doctor and diabetes specialist, they comprise also quality management
(Hellbrück 2000). The development of the necessary structures requires
extensive settlements which haven’t been concluded yet. The settlements
bear on the development of a new database, the implementation of neces-
sary software in medical practices, the compatibility of the software used,
and changes in the working organisation in doctors’ offices. An improved
metabolism causes additional costs in the drugs area. Because of this
doctors, Associations of Statutory Health Insurance Physicians, and
Health Insurance Funds today face costs.
Associations of Statutory Health Insurance Physicians are often ready
to bear a part of the costs for their internal reorganisation, but these costs
are usually not accompanied by expenses. In addition, extra budgeting pay-
ments lead today to higher expenses and costs for the Health Insurance
Funds. Though in-patient costs will be reduced by diabetes contracts, the
existing regulation does not allow health insurance funds to change these
cost savings into expense reductions for health insurance funds. As a result,
diabetes contracts lead to higher contribution rates, worsening their com-
petition position: these contracts are disastrous. This is especially true for
health insurance funds in the new lands; their financial situation between
1997 and 2000 was very fraught.
A prerequisite for the implementation of model projects was the alter-
ation of the statute of health insurance funds, i.e. benefits supplied by
model projects were so-called ‘statute benefits’. Expenses for statute
benefits, as well as for such benefits for which patients have no right to
claim, are left out in the Risk Structure Compensation Scheme. This means
additional expenses from diabetes contracts not being balanced by the Risk
Structure Compensation Scheme which make it even harder to compete

with others.
Furthermore, from the perspective of a health insurance fund a diabetic
person represents a high risk even if he/she has a good metabolism; that is
the amount of coverage of diabetic persons is negative. Nevertheless, the
diabetes contracts were published in the form of press briefings, press state-
ments and brochures. It is however not known to what extent diabetic
persons were attracted by those contracts.
The attraction of bad risks by diabetes contracts is related to the number
and the contribution rate of the participating insurance companies. If an
insurance company makes a contract in a particular area without the par-
ticipation of another insurance company, it has to carry all the risk by itself.
If all insurance companies from one area are contracting partners in a dia-
betes contract, the competition doesn’t change within this group but in rela-
tion to private health insurance companies; nevertheless, a good marketing
campaign for the diabetes contract by a health insurance fund with high
contribution rates can leave health insurance funds with low contribution
rates much affected by diabetic insurants.
CONTINUATION
At the end there is still the question, why diabetes contracts were not can-
celled by health insurance funds as the negative effects were identified. The
answer to that question cannot be given with complete certainty. It is pos-
sible that the harm to the companies hasn’t been recognised. Against this
interpretation is the fact, that within the Federal Association of Health
Insurance Funds AOK and BEK criticism arose, which included the effects
of diabetes contracts on competition.
The following interpretation is more likely. The Regional Associations of
Statutory Health Insurance Physicians were, without any doubts, interested
in continuing these contracts. If a diabetes contract had been cancelled by the
Health Insurance Funds, a public éclat initiated by the Regional Associations
of Statutory Health Insurance Physicians would certainly have happened.
Additionally, the personal interests of the decision makers were not nega-
tively affected, and/or the harm to the business was evaluated as too small.
This point of view is supported by the fact that the first AOK diabetes
contract of the second generation in Thuringia was over on 31 March
2001. An ensuing contract is being negotiated and till its conclusion the
existing contract will be used. Another fact is that diabetes contracts of
the second generation in general replaced diabetes contracts of the first
generation. It looks as though existing extra-budgeting payments paid for
first-generation diabetes contracts were replaced by higher-quality second-

generation diabetes contracts. The Regional Associations of Statutory
Health Insurance Physicians secure extra payments and the representatives
of health insurance funds could justify the continuation of extra budgeting
payments by higher-quality care. This explains also why there was no limi-
tation on some regions, for example the new German states, within the AOK
community.
It also needs to be stated that the damaging of business has a time dimen-
sion: what does damage in the short run, might not impair in the long run.
Health insurance funds concluding diabetes contracts damage their busi-
ness in the short run, as it has to afford extra-budgeting payments now-
adays, without getting them paid back at the next day. Contracts based on
§§63–65 SGB V will not be considered in the Risk Structure Compensation
Scheme and the danger of additional bad risks and of accompanying
increasing contribution rates is possible. Any health insurance fund taking
part in a diabetes contract has to face these consequences. Subject to the
relative number of diabetic patients participating in a diabetes contract, the
market position of the contracting health insurance funds changes even
within this group. The Federal Insurance Office (Bundesversicherungsamt)
makes it clear: ‘A health insurance fund which has an excellent quality man-
agement of expensive diseases could economically commit suicide’ (quoted
according to Jacobs et al. 2001, p. 46, footnote 2).
Health insurance funds have more possibilities than cancelling a con-
tract. One strategy is to keep up a good relationship with the legislator, for
example acting on behalf of insurants with chronic diseases. To build a
good relationship with state institutions is very important for health insur-
ance funds insuring a lot of bad risks compared to their competitors, e.g.
health insurance funds AOK, BEK. Moreover, some persons who devel-
oped the second-generation diabetes contracts as representatives of health
insurance funds changed their job in 1998/99 and obtained new jobs as rep-
resentatives of physicians’ organisations. For example, one person became
managing director of an association of physicians, another gained an
executive position at the Federal Association of Statutory Health
Insurance Physicians. Though these job changes changed their arguments,
they tried furthermore to improve diabetes care but with significant room
for manoeuvre. Sometimes it takes some time until important views are
accepted by health politicians and decision makers. Once in a while they
need to be pushed to revise their old convictions.
This push came from experts’ opinions on the Risk Structure
Compensation Scheme. One advisory opinion was commissioned by the
Federal Government (Jacobs et al. 2001), the other (Lauterbach and Wille
2001) from Federal Associations of Health Insurance Funds (that is
AOK-BV, VDAK/VEK, and IKK-BV) whose insurants represent bad

risks. Both statements complain about insufficient quality management of
chronic diseases, especially the advisory opinion commissioned by some
Federal Associations of Health Insurance Funds who point out the
deficits in this particular area.
Because of these statements a revision of the Risk Structure Compen-
sation Scheme was required (Ahrens 2001). Those demands and the advi-
sory opinions induced the government to revise the Risk Structure
Compensation Scheme. The Ministry of Health wanted to open the possi-
bility of chronic disease-management programmes. Since 1 January 2002,
30 per cent of the costs should be borne by the health insurance fund which
insures the person (that is a ‘co-payment’ of the health insurance fund) and
the other 70 per cent of the treatment costs should be paid by the ‘Solidarity
Community’ (that is all health insurance funds). All these disease-manage-
ment programmes have to be certified by a neutral, nationwide certification
institution and all participating patients must register in the programmes.
In the long run the introduction of a morbidity oriented Risk Structure
Compensation Scheme is wanted.
To evaluate forthcoming disease-management programmes based on the
new regulation of the Risk Structure Compensation Scheme is not yet pos-
sible, as there is too little information about this. We can however say that
the Risk Structure Compensation Scheme may turn out to be an expend-
iture compensation scheme; all the better will morbidity be mapped (Rürup
2001). Three possible scenarios can be distinguished: (a) health insurance
funds insuring especially bad risks get a competitive edge, other health
insurance funds are the losers; (b) the opposite of case (a) will happen; (c)
the competition between health insurance funds does not change. In the
long run it turns out that only in the borderline case (c) could the harmful
effects on health insurance funds be avoided. It is however unclear whether
extra budgeting payments will pay for themselves and thus the position in
the competition process of contracting health insurance funds improve
compared to private health insurance.
CONCLUSIONS
Altogether we can say, firstly, that the knowledge increase was not sufficient
for the innovation (thesis (b)). Secondly, the changing legal conditions in
the Statutory Health Insurance Reorganisation laws in 1997 cannot be
regarded as the cause either for the emergence or the diffusion of the
second-generation diabetes contracts as these changes were after the first
developments of the second-generation diabetes contracts. On the contrary
the existing first-generation contracts caused the Regional Associations of

Statutory Health Insurance Physicians to take an interest in continuing the
contracts, which was not contrary to the personal interests of the decision
makers in health insurance funds. If someone would like to see some kind
of path dependency, it could be accepted.
Thirdly, to argue that these diabetes contracts were initiated as part of a
general strategy of some health insurance funds to change the Risk
Structure Compensation Scheme is not convincing. Diabetes contracts were
developed and implemented to avoid subsequent diseases and their costs,
however, without differentiating between costs and expenditures. These
goals have not changed since the statements of Lauterbach and Wille
(2001). Besides that the first-generation diabetes contracts were concluded
before the Risk Structure Compensation Scheme were invented and imple-
mented at all. The immense improvements to these contracts realised by the
second-generation contracts were a nice possibility to continue the old con-
tracts. An even fiercer competition between the health insurance funds who
competed for good risks made necessary a revision of the legal conditions,
especially of the Risk Structure Compensation Scheme. Additionally, the
change in the German government in 1999 and the accompanying shift in
health politics (especially the wish to improve quality medical care, see
Bundesministerium für Gesundheit 2000) could not be foreseen some
years ago.
Fourthly, every contract partner of diabetes contracts usually had the
alternative to cancel the contract even during running-time. The short-term
business-damaging effects would thus have been avoided. In public discus-
sions which would most likely have taken place, the companies could have
pointed out allocative distortions and could have initiated a change in the
statutory conditions. This is further evidence that personal preferences of
some representatives were crucial in this process (thesis (c)). In addition,
systematic business-damaging behaviour is no exclusion criterion for an
innovation (thesis (d)). To prove this, one example has to be given; and we
just did so. Though the innovation might be disastrous from the perspec-
tive of a health insurance fund, all circumstances corresponded to each
other (particularly the preferences) in that the innovation was evaluated as
socially desirable (thesis (d)).
NOTES
1. I would like to thank Frank Engelmann for his helpful comments. The first translation of
the German original (Hellbrück 2003) was made by some busy students: Christian Haas,
Patricia Wlodarczak, Dirk Enskat, and Ina Wunderlich; nevertheless, I am responsible for
the remaining errors. This work reflects solely the opinion of the author. The chapter
relates to the developments up to May 2001.
2. Statement of Albrecht Dietz, a ‘Schumpeter’ entrepreneur.
3. ‘GKV’ is the abbreviation for statutory health insurance.
4. In Germany we distinguish various types of health insurance funds (§§143–171
SGB V 1997): Allgemeine Ortskrankenkassen (AOK), Betriebskrankenkassen
(BKK), Innungskrankenkassen (IKK), See-Krankenkassen, Landwirtschaftliche Kran-
kenkassen, Bundesknappschaft, Ersatzkassen. Each health insurance fund is a member
of exactly one type which is determined by law.
5. The date given shows the time of the publication in the federal law paper. In most
cases, the day of publication and the date of introduction and enforcement are not the
same.
6. Benefits in kind are one principle of the German statutory health system. Everybody
should pay according to his abilities and everybody should get equal access to the health
care system and equal health care. To accomplish this, benefits in kind are seen as funda-
mental; reimbursements would break the principle as healthier people would not pay
anymore according to their abilities.
REFERENCES
Adler, A., I. Stratton, H. Neil and J. Yudkin (2000), ‘Association of systolic blood
pressure with macrovascular and microvascular complications of type 2 diabetes
(UKPDS 36): prospective observational study’, British Medical Journal, 321,
412–19.
Ahrens, H.J. (1997), ‘Ökonomische Anreizsysteme – Herausforderungen und
Chancen’, in AOK-Bundesverband (ed.), AOK im Dialog; Qualität und
Wirtschaftlichkeit – Neue Wege gemeinsamer Verantwortung, Bonn, pp. 49–54.
Ahrens, H.J. (2001), Pressekonferenz zu AOK im Dialog ‘Weiterentwicklung des
Risikostrukturausgleichs’, Berlin, 15 March 2001, AOK-Bundesverband.
Andersen, H. and J. Schwarze (1999), ‘Kassenwahlentscheidungen in der GKV’,
Arbeit und Sozialpolitik, 53, 10–23.
Ärzte-Zeitung (15 September, 1998): ‘Endlich Klarheit für die Behandlung von
Typ-II-Diabetikern in der Praxis’.
Ärztliche Zentralstelle Qualitätssicherung (1998), ‘Checkliste “Methodische
Qualität von Leitlininen” 1. Version: Gültigkeit: 1.2.1998–31.1.1999’, Deutsches
Ärzteblatt, 95, B-2014-B-20116.
Becker, J. (1999), ‘Drei-Klassen-Medizin darf es nicht geben’, Gesundheit und
Gesellschaft, 2, 44–5.
Bertram, B. (1999), ‘Zusammenarbeit von Hausarzt und Augenarzt in der
Diabetikerbetreuung. Kommunikation unerläßlich’, Deutsches Ärzteblatt, 96,
B-2450-B-2454.
Bundesministerium für Gesundheit (2000), ‘Begründung zum Entwurf eines
“Gesetzes zur Strukturreform der gesetzlichen Krankenversicherung” im Jahr
2000’.
Hellbrück, R.P. (1997), Qualität und Ausgaben in der medizinischen Versorgung – Von
Qualitätssicherung und Kosteneffizienz zu mehr Wettbewerb im Gesundheitswesen?,
Berlin: Duncker & Humblot.
Hellbrück, R.P. (2000), ‘EDV-gestütztes Qualitätsmanagement in Praxisnetzen’,
Das Gesundheitswesen, 62, 511–15.
Hellbrück, R.P. (2003), ‘Entstehung und Ausbreitung geschäftsschädigender
Innovationen: ein Fallbeispiel’, in Kurt Dopfer (ed), Studien zur Evolutorischen

Ökonomik Band VII, Berlin: Duncker & Humblot, pp. 219–41.
Henkel, H. (1999), ‘Selbstverwaltung ist ein Modell mit Zukunft’, Gesundheit und
Gesellschaft, 2, 42–3.
Hesse, G. (1990), ‘Evolutorische Ökonomik oder Kreativität in der Theorie’, in
U. Witt (ed.), Studien zur Evolutorischen Ökonomik I, Berlin: Duncker &
Jacobs, K., P. Reschke, D. Cassel amd J. Wasem (2001), Zur Wirkung des Risiko-
strukturausgleichs in der gesetzlichen Krankenversicherung: Eine Untersuchung im
Auftrag des Bundesministeriums für Gesundheit. Endbericht, Berlin, Duisburg,
Greifswald: IGES, Gerhard-Mercartor-Universität Duisburg, Ernst-Moritz-
Arndt Universität Greifswald, 15 February.
Landgraf, R. et al. (1999), ‘Diskussionsentwurf Regelungsbedürftige Vertrag-
sinhalte von Diabetesvereinbarungen – eine Initiative des DQI-Ausschusses der
DDG’, Diabetes und Stoffwechsel. Zeitschrift für angewandte Diabetologie, 8 ii-
vi.
Laschet, M. (1998), ‘Erstmals flächendeckende und optimale Diabetikerversorgung
in Bayern’, Bayerisches Ärzteblatt, 53 (8).
Lauterbach, K. and W. Wille (2001), Modell eines fairen Wettbewerbs durch den
Risikostrukturausgleich. Sofortprogramm ‘Wechselkomponente’ und ‘solidarische
Rückversicherung’ unter Berücksichtigung der Morbidität. Abschlußbericht,
expert’s opinion commissioned by Verband der Angestellten-Krankenkassen
e-V. (VdAK), Arbeiter-Ersatzkassenverband (AEV), AOK-Bundesverband
(AOK-BV) and IKK-Bundesverband (IKK-BV); obtainable at www.vdak.de
Mehl, E., S. Becker-Berke and G. Müller-de-Cornejo (2000), Einführung eines
Disease-Managements am Beispiel Diabetes, Bonn: KomPart.
Meierjürgen, R. (1994), ‘Gesundheitsförderung und Krankenkassenwettbewerb.
Die Auswirkungen des Gesundheitsstruktur-Gesetzes’, Arbeit und Sozialpolitik,
48, 58–64.
Müller, M., S. Richard and K. Schönbach (1999), ‘Erste Jahresbilanz zeigt:
Kombiniertes Budget hat sich bewährt’, in Die BKK, pp. 585–90.
Nachtigall, G. (1997), ‘Rationale Medizin – Ein Beitrag zur Zukunftssicherung der
sozialen Krankenversicherung’, in AOK-Bundesverband (ed.), AOK im Dialog;
Qualität und Wirtschaftlichkeit – Neue Wege gemeinsamer Verantwortung, Bonn,
pp. 13–18.
Robra, B. et al. (1996), Krankenhausfallanalyse. Die stationäre Versorgung von
Versicherten der AOK, Magdeburg and Sankt Augustin: Asgard-Verlag Dr
Werner Hippe KG.
Rürup, B. (2001), Die Reform des Risikostrukturausgleichs im Rahmen der
Vorstellungen der ‘Fünf Weisen’ zur Gesundheitspolitik. Berlin: AOK-Kongress:
Reform des RSA, 15 March.
Scherbaum, W. (1997), ‘Schnittstellenmonitoring durch Qualitätsindikatoren:
Neue Konzepte zur Verbesserung der Versorgung von Diabetikern’, in AOK-
Bundesverband (ed.), Qualität und Wirtschaftlichkeit – Neue Wege gemeinsamer
Verantwortung, Bonn, pp. 33–47.
SGB IV, 2000, Sozialgesetzbuch IV vom 23.12.1976 (BGBl. fI S. 3845), zuletzt geän-
dert durch Gesetz vom 22.12.1999 (BGBl. I S 2601) und 20.12.1999 (BGBl. I S. 2),
2000.
SGB V 1993, Sozialgesetzbuch V – Gesetzliche Krankenversicherung – vom 20.
Dezember 1988, BGBl. I S. 2477 zuletzt geändert durch das Gesund-
heitsstrukturgesetz vom 21. 12. 1992, BGBl. I S. 226. 1992.

Dezember 1988, BGBl. I S. 2477 zuletzt geändert durch das 1. Und 2. GKV
Neuordnungsgesetz vom 23.6.1997 (BGBl I S. 1518 u. S. 1520).
Dezember 1988, BGBl. I S. 2477, Artikel 1), zuletzt geändert durch Gesetz vom
21.7.1999 (BGBl. I S. 1648), 22.12.1999 (BGBl. I S. 2534), 22.12.1999 (BGBl. I S.
2626) und 22.12.1999 (BGBl. I S. 2657), 2000.
Dezember 1988, BGBl. I S. 2477, Artikel 1), zuletzt geändert durch Gesetz vom
21.7.1999 (BGBl. I S. 1648), 22.12.1999 (BGBl. I S. 2534), 22.12.1999 (BGBl. I S.
2626) und 22.12.1999 (BGBl. I S. 2657), 2000.
SGB V 2001, Sozialgesetzbuch V Gesetzliche Krankenversicherung [SGB V] vom
20.12.1988 (BGBl. I p. 2477, Artikel 1), zuletzt geändert durch Gesetz vom
21.12.2000 (BGBl. I S 1971), 2001.
Tuschen, K. and M. Quaas (1995), Bundespflegesatzverordnung Kommentar mit
einer umfassenden Einführung in das Recht der Krankenhausfinanzierung,
Stuttgart, Berlin, Cologne: W. Kohlhammer.
13. Applying evolutionary economics
to public policy – the example of
competitive federalism in the EU
Wolfgang Kerber
1. INTRODUCTION
One of the major problems of evolutionary economics is its difficult and

unclarified relationship to public policy questions.1 Often the problem
arises that evolutionary reasonings can be used rather well for a sound cri-
tique of a wide range of policy recommendations. But due to the evolu-
tionary argument of the openness of economic processes and the resulting
Hayekian knowledge problem it seems often very difficult to derive positive
policy proposals from evolutionary reasonings – leading to the danger of a
general retreat from policy discussions. In this chapter it is suggested that
evolutionary economics should take a pragmatic approach to economic
policy questions:
1. Evolutionary economics should be pragmatic in a methodological

sense, i.e. that in an ever-changing world economic policy can and has
to be made despite the impossibility of eliminating all uncertainties in
regard to its effects.
2. Evolutionary economics should not restrict itself to theoretical and
basic research but should also do research about the application of evo-
lutionary reasonings to the solving of real-world problems including
participation in policy discussions.
3. For applying evolutionary arguments to policy questions we have to
find a pragmatic way to combine evolutionary with neoclassical argu-
ments, which to a certain degree will remain indispensable for many
real-world problems.
Instead of discussing these demands on evolutionary economics on a

methodological or theoretical level, I would like to show in regard to a
specific, hotly discussed policy question, how evolutionary economics can
296
Applying evolutionary economics to public policy 297
be productively applied to economic policy. The policy problem is the ques-

tion of the future institutional development of the EU in regard to central-
ization and decentralization. The allocation of competencies to the EU level
on one hand and on the national level of the EU member states on the other
hand is one of the most important problems in the process of European inte-
gration. From an economic point of view the theory of fiscal federalism
offers criteria for the optimal vertical allocation of competencies. In this
chapter I will argue that the perspective of evolutionary economics will
provide crucial additional arguments in favour of a greater decentralization
of competencies, because from an evolutionary perspective a more decen-
tralized system will be characterized by more innovativeness and adaptabil-
ity than a more centralized federal system. On the theoretical level, the
implications will be that the originally neoclassical theory of fiscal federal-
ism has to be supplemented by evolutionary reasonings in regard to the inno-
vation and diffusion of public policies. This will lead to the proposal of an
evolutionary concept of competitive federalism as the appropriate theoreti-
cal basis for deciding on the vertical allocation of competencies in the EU.
The chapter is structured as follows: In section 2 a short introduction is
given into the problem of centralization and decentralization in the EU.
Section 3 presents a short overview of the well-established neoclassical the-
ories of federalism and interjurisdictional competition before they are crit-
ically examined from an evolutionary perspective in section 4. Particularly
important is section 5, in which both explicit and implicit evolutionary con-
tributions to the theory of federalism and interjurisdictional competition
are presented. These reasonings are used for the sketching of a pragmatic
approach to competitive federalism (and the problem of the appropriate
vertical allocation of competencies in a federal multilevel system) in section
6, which entails both neoclassical and evolutionary criteria. In section 7
some brief implications are deduced for our policy problem of centraliza-
tion and decentralization in the EU, before in section 8 some general
conclusions are drawn as to how policy questions should be dealt with from
the perspective of evolutionary economics.
2. THE POLICY PROBLEM: CENTRALIZATION

VERSUS DECENTRALIZATION WITHIN THE EU
The problem of the future institutional development of the European

Union is one of the most important topics in economic policy in Europe.
The specific problems of the enlargement of the European Union and
the inevitable reform of European institutions and decision processes
(European Commission, majority decisions etc.) are well known. But at
least as important as these topics is the question how centralized or decen-

tralized the European Union should be in the long run.
The debate on centralization and decentralization in the EU is charac-
terized by the tension between two tendencies:
1. There can be no doubt that the large efforts of the EU to complete the
European internal market by breaking up many barriers to national
markets and to introduce competition in many industries, which for
decades were monopolized in the member states (as e.g. telecommuni-
cation and energy markets), were a huge success. In that respect the poli-
cies of the European Commission (internal-market programme) and
also the European Court of Justice (‘Cassis de Dijon’ judgment) to
enforce the four basic freedoms (freedom of goods, services, labour, and
capital) by removing national non-tariff barriers to trade were very
helpful. This was carried out partly by introducing the principle of
mutual recognition of national regulations and partly by (minimum)
harmonization of regulations. To some degree also European Monetary
Union with the introduction of the Euro as single currency can be
seen in this perspective of promoting the European internal market.
Consequently, many efforts on the EU level that simultaneously implied
a pushing back of the competencies of the member states, were impor-
tant for enforcing competition of European markets.
2. But the observed tendencies to centralization in the EU are also evalu-
ated very critically both by European citizens and politicians on one
hand and by economists and legal scholars on the other. Increasing
dangers of intransparency and lack of monitoring by the citizens,
bureaucratization, and mounting rent-seeking problems are suggested
(Vaubel 1995, 1997). There are widespread preferences in the European
population for a high degree of decentralization that limit considerably
the acceptance of additional shiftings of competencies to the Euro-
pean level. The increasing heterogeneity within the EU – due to the
enlargement by mostly much less developed Middle and East
European countries – will reinforce this problem. So the introduction
of the principle of subsidiarity in the Treaty of Maastricht can be seen
as an important political sign against too quick and often too
unreflective tendencies of centralization and harmonization in the EU.
But the problem of the best vertical allocation of competencies between the
EU and the member states is still unsolved. How is it possible to maintain
a high degree of decentralization and heterogeneity within the EU without
impeding the achievement of the aim of European integration? This has
remained an open question both in public and academic discussions.
Neither political scientists, legal scholars nor economists have developed

convincing elaborated concepts of how European integration can be real-
ized without giving up decentralization and variety to a large degree. To
answer this question a careful balancing of the advantages and disadvan-
tages of centralization and decentralization of competencies is required.
3. ECONOMIC THEORIES OF FEDERALISM AND

INTERJURISDICTIONAL/REGULATORY
COMPETITION AS A POSSIBLE ANSWER
From a theoretical perspective, the problem of centralization and decen-
tralization in the EU can be conceived as the question of how the compe-
tencies of the state should be allocated within a federal multilevel system of
jurisdictions (with the levels EU, the member states, regions, and munici-
palities). For solving this problem of the optimal vertical allocation of
competencies, we can use the criteria of the well-established economic
theory of fiscal federalism, which has been developed in the public finance
literature since the 1960s.2 From this perspective, policies which attempt to
achieve aims of macro-economic stabilization or redistribution should be
allocated to the central federal level. For the efficient provision of public
goods and their financing by taxes several criteria have been developed, on
which the decision for a centralized or decentralized provision depends:
regional scope of the utility of public goods (spillover effects, interregional
externalities), economies of scale in regard to the provision of public goods,
the extent of regional homogeneity/heterogeneity of preferences of the citi-
zens, and decision and transaction costs, which might vary in differently
structured federal systems.
But two central problems, which are important for the current discussion
in regard to the EU, have not been investigated in this traditional theory of
fiscal federalism:
1. One is the problem of competition between states, regions, or munici-

palities (interjurisdictional competition). If mobility of persons, firms,
and production factors (particularly capital) between jurisdictions
exists to some extent, processes of interjurisdictional competition for
these mobile resources emerges in federal systems. Therefore the issue
of the positive or negative effects of these competition processes arises.
Since the removal of all mobility barriers is one of the central aims of
the EU, this is particularly important for the European discussion.
2. The theory of fiscal federalism neglected the problem of the appropri-
ate vertical allocation of legal rules and regulations – due to its focus
on public goods and taxation. Implicitly it was assumed that the legal
framework was the task of the central (national) level. But within the
EU it has become a crucial question, to what extent legal rules and
regulations have to be centralized/harmonized or can also remain on a
more decentralized level. This also raises the problem of the extent and
effects of processes of regulatory competition.
The long-existing reluctance to integrate interjurisdictional competition

into the economic theory of federalism is surprising, because through
Tiebout’s theory of the competitive provision of local public goods, a well-
formulated model for interjurisdictional competition existed from the
1950s, leading to a broad theoretical and empirical literature based upon
Tiebout’s approach (Tiebout 1956, Dowding, John and Briggs 1994). But
only since the 1990s have the issue of interjurisdictional and regulatory
competition (also called systems competition, locational competition,
competition among governments) been discussed broadly. This encom-
passed general studies on competitive federalism (in contrast to coopera-
tive federalism), the problem of tax competition, and the specific European
discussion on the advantages and disadvantages of harmonization or com-
petition of regulations within the EU.3 All these discussions are deeply
influenced by the dispute between two opposing views, being either
extremely optimistic or extremely pessimistic in regard to the effects of
interjurisdictional or regulatory competition.
The proponents of interjurisdictional competition argue that it would
induce the governments of these jurisdictions to offer bundles of public
goods, regulations, and taxes that better fulfil the preferences of the citizens,
because interjurisdictional competition restricts the discretionary scope of
governments, leading to a more efficient provision of public goods and less
redistribution due to rent-seeking activities. Furthermore, interjurisdic-
tional competition will lead to more innovations in regard to the public
services of the jurisdictions – an argument to which we will later come back
in more detail. In contrast to that, the opponents of interjurisdictional com-
petition show that there are sound arguments why interjurisdictional com-
petition can suffer from considerable market failures implying inefficient
outcomes and therefore considerable welfare losses (Sinn 1997). In theoret-
ical models it has been attempted to show, for example, that tax competition
might lead to an under-provision of public goods and/or to an inefficiently
low degree of redistribution. But the overall results of the theoretical models
in regard to the positive or negative effects of tax competition depend very
much on the specific assumptions of the models. The same is true for the
empirical literature. Empirical studies of Switzerland, however, which due
to its far-reaching competencies for their regional units (‘Kantone’) can be
seen as a kind of a model for competitive federalism, show that no system-

atic negative effects can be observed (Feld 2000).
Closely linked to this dispute about interjurisdictional competition is the
discussion about the advantages and disadvantages of regulatory competi-
tion.4 In Europe, this question has been much discussed in the context of
the introduction of the principle of ‘mutual recognition’ of national regu-
lations by the European Court of Justice (‘Cassis de Dijon’ judgment). This
transition to the so-called ‘rule of origin’, which implies that it is sufficient
for firms that their products fulfil the requirements of the regulations in
their home country in order to export freely to all other member states of
the EU, can lead to an indirect form of competition among the regulations
of different member states, because within the EU firms with different
product regulations compete with one another (Sun and Pelkmans 1995).
In the US, competition among the corporate laws of the federal states is
known as a famous example of regulatory competition. Similar to the dis-
cussion on interjurisdictional competition, the proponents and opponents
of the concept of regulatory competition hold different opinions, whether
the positive effects in form of restrictions for rent-seeking activities and
more legal innovations are more important than potential problems of
regulatory competition, particularly in form of ‘race to the bottom’ prob-
lems. Here too neither the theoretical models nor the few empirical studies
have come up with clear results. For the well-studied competition among
corporate laws within the US, however, most experts have come to the con-
clusion that the positive effects of this regulatory competition are much
larger than the negative ones (Romano 1998).
Although much research work has been done on the question of the
appropriate vertical allocation of competencies in a federal multilevel
system of jurisdictions including the analysis of interjurisdictional and
regulatory competition, which might emerge in decentralized systems, a
consistent, integrated theory of (competitive) federalism that can give a
well-founded orientation for the future development of the EU has not
been developed yet. In the following, the theoretical foundations of the
central arguments in the economic theories of federalism and interjuris-
dictional/regulatory competition are critically analysed.
4. CRITIQUE OF NEOCLASSICAL ARGUMENTS IN

FEDERALISM THEORY FROM THE
PERSPECTIVE OF EVOLUTIONARY ECONOMICS
The traditional economic theory of fiscal federalism can be characterized

as a typical neoclassical equilibrium approach, which normatively is based
upon Paretian welfare theory. The dominant criteria for the vertical allo-
cation of competencies, regional scope of the utility of public goods
(spillover effects/externalities), economies of scale-effects and the extent of
regional homogeneity/heterogeneity of preferences, can be consistently
deduced from the neoclassical theories of public goods and taxation – as
well as the dominant aim of efficient allocation follows from Paretian
welfare theory. The same is true for most approaches to interjurisdictional
and regulatory competition. Tiebout’s model of the competitive provision
of local public goods was the exact transfer of the model of perfect com-
petition to the market for local public goods.5 Most of the theoretical dis-
cussion about interjurisdictional competition (and, particularly, tax
competition) as well as regulatory competition are such welfare-theoretic
neoclassical analyses. This implies that the outcome of interjurisdictional
and regulatory competition is compared with the perfect theoretical
solution of allocative efficiency. And according to market-failure theory,
any allocative inefficiency and therefore any deviation from the optimal
outcome of the perfect competition model leads to the diagnosis of a
failure of interjurisdictional or regulatory competition. Thus it seems to be
the obvious conclusion that interjurisdictional (or regulatory) competition,
and therefore also competitive federalism, is no viable approach, leading to
the policy recommendations of the centralization of competencies or the
harmonization of regulations.
One kind of fundamental critique to this neoclassical market-failure
argumentation is the well-known critique of Demsetz (1969), comparing
situations in the real world with the fictitious approach of the model of
perfect competition (‘nirvana approach’). This method leads to the diag-
nosis of ubiquitous market failure and therefore the suggestion of ubiqui-
tous state intervention or the entire abolishment of competition. Rather, a
more differentiated methodological approach along the lines of Demsetz’s
concept of ‘comparative institutional analysis’ seems to be appropriate:
1. After the diagnosis of some kind of inefficiency in regard to interjuris-

dictional/regulatory competition, firstly, it is necessary to examine
whether these problems of market failure might not be reduced or even
eliminated by modifications of the institutional framework for these
competition processes, because it is well known that many problems of
market failure can be traced back to defective institutional arrange-
ments (or property rights structures). This raises the crucial issue, what
kind of institutional framework (sets of rules) is necessary for the
proper working of interjurisdictional/regulatory competition and
competitive federalism in general. In most neoclassical studies this
import dimension of the problem is not discussed at all.
2. But even if we do not succeed in eliminating or reducing the allocative

inefficiencies by changing the institutional framework, these disadvan-
tages of decentralized federal systems have to be compared with the
real disadvantages of the centralization of competencies or the har-
monization of regulations. In that respect also the disadvantages of the
institutional alternatives ‘centralization’/‘harmonization’ have to be
studied, before well-founded policy recommendations can be given.
For their analysis political-economic models are appropriate. From the
perspective of economic policy, this is the well-known problem of
weighing market failure against state failure.
Despite its fundamental character, this first kind of critique of neoclassical

welfare-theoretic reasonings about the working of decentralized federal
systems is primarily based upon argumentations from institutional eco-
nomics and political-economic literature.6 The main critique from the
perspective of evolutionary economics can start from basic ideas of
Schumpeterian and Hayekian thinking, which emphasise the importance of
innovations and knowledge problems. Our argumentation begins with the
knowledge problems from the perspective of Hayek.7 In the traditional
theory of fiscal federalism it is – explicitly or most often implicitly –
assumed that the policy makers on the central level of a federal system have
perfect knowledge of which combination of public goods, taxes and regu-
lations are most suitable for fulfilling the preferences of the citizens and
therefore ensuring efficiency. If we make the assumption of an ‘omniscient
dictator’ (Breton 1996) on the central level (and exclude knowledge prob-
lems), and combine this with the additional assumption of a ‘benevolent
dictator’ (and therefore exclude also rent-seeking problems), it is not sur-
prising that centralized solutions seem to be superior to decentralized solu-
tions. But reality is different. Starting from basic ideas of Hayek we can
distinguish two fundamental knowledge problems that are of great impor-
tance for our problem of the appropriate extent of centralization and
decentralization in the EU.
In the discussion about federalism, it is often argued that policy makers
on lower-level jurisdictions have a better knowledge about the specific prob-
lems of their citizens, and are therefore better able to solve these problems.
Theoretically, this problem can be characterized with the Hayekian notion
of ‘dispersed knowledge’ about the particular circumstances of time and
space (Hayek 1945) or as the problem of ‘local knowledge’. Thus the
Hayekian argument about the impossibility of the centralization of all rele-
vant knowledge in society, which he developed as a critique of the concept
of a centrally planned economy, can also be applied to the provision of
public goods and regulations by jurisdictions. If it is true that the knowledge
about the specific regional and local problems is to a considerable extent

local knowledge, which cannot be centralized, then decentralized federal
systems are able to use more of this relevant local knowledge than more cen-
tralized systems. But to be able to use this knowledge, lower-level jurisdic-
tions have to have competencies for deciding freely on their policies.
Another variant of this ‘local knowledge’ argument uses reasonings from
public choice theory. It is maintained that a greater decentralization
increases the capability of citizens to monitor the decisions of politicians
leading to less welfare losses due to rent-seeking activities.
At least as important as the problem of decentralized knowledge is the
general problem that there is no perfect knowledge of what kind of public
goods and what kind of regulations are the best solutions for the problems
of the citizens. In neoclassical theories of public goods and regulations, it
is implicitly assumed that the knowledge about the optimal solution in
terms of the provision (and production) of public goods, of regulations,
and taxation already exists. But just as nobody – not even scientists – knows
the optimal solution for building cars or other private goods, we also
cannot assume that we already know the best way how jurisdictions can
provide security, establish systems of higher education, organize their judi-
cial system, or fulfil the cultural preferences of their citizens. Neoclassical
theories, however, always presuppose that the optimal public good and the
optimal way for its production are already known. But there are no good
reasons why we should assume that governments are better informed about
the best solutions for public-good problems or regulatory problems than
private firms are in regard to private goods. Rather in both cases, the
problem solutions which are offered by jurisdictions and private firms can
only be seen as fallible hypotheses about what the best problem solutions
might be. The Hayekian knowledge problem implies that we have to accept
that the knowledge of the policy makers can always be only fallible know-
ledge that can be more or less wrong. Therefore innovative improvements
of these solutions are always possible.
In respect of private goods, it is the task of market competition to help to
solve this knowledge problem. In competition firms experiment with
different problem solutions (innovations), test them on the market, and
learn from their experiences and those of their competitors. Therefore
Hayek (1948, 1978) rejected the neoclassical concept of market competition,
because it presupposes that the firms already know what the best products
or production technologies are. In his notion of ‘competition as a discovery
procedure’, which conceives competition as a process of parallel experi-
mentation, knowledge about the best products or production technologies
is the central outcome of market competition – and not a precondition. This
implies that all theoretical approaches which use this neoclassical concept of
competition for analyzing the effects of interjurisdictional or regulatory

competition (as the Tiebout model and most of the literature do) fail to
grasp a crucial part of the problem, i.e. that the knowledge problem cannot
be assumed to be solved, rather that its solving is one of the central tasks that
have to be tackled with by federal multilevel systems and interjurisdic-
tional/regulatory competition.
Thus a theory that wants to give well-founded policy recommendations
about the appropriate degree of centralization or decentralization of public
policies in a multilevel system of jurisdictions (like the EU) has to take this
knowledge problem seriously. From this perspective one question arises that
has not been addressed in the traditional theory of fiscal federalism or in
neoclassical theories of interjurisdictional or regulatory competition: How
can new knowledge about the improvement of public goods and regulations
be produced? Or: What do we know about innovation and technical progress
in regard to all of those public services that jurisdictions provide to their cit-
izens? It is a fundamental deficit of the traditional literature in public finance
that the dimension of innovation in the public sector has not been taken into
account in a systematic way. The neoclassical approaches to fiscal federal-
ism as well as the neoclassical theories of interjurisdictional and regulatory
competition do not ask for the effects of different degrees of centralization
or decentralization on the innovation of public goods and regulations. But
since in the public sector product and process innovations are as important
as in the private sector, it is necessary to ask for the capability of federal
multilevel systems of jurisdictions to generate innovations and to adapt
quickly to exogenous change. This introduction of the innovation perspec-
tive opens the door for applying arguments from evolutionary economics to
the analysis of the advantages and problems of competitive federalism.
5. EVOLUTIONARY CONTRIBUTIONS TO THE

THEORIES OF FEDERALISM AND
INTERJURISDICTIONAL/REGULATORY
COMPETITION
The discussion about centralization and decentralization does not consist

only of neoclassical arguments as they were shown in the last section. Also,
important arguments emerged which cannot be derived from the neoclas-
sical theory of fiscal federalism. A crucial part of them can be traced back
to basic concepts of evolutionary economics.8 Some of these arguments
refer explicitly to evolutionary economics, whereas other arguments have
been developed independently of evolutionary thinking, but can be inter-
preted as evolutionary.
5.1 Explicit Evolutionary Arguments: Competition and Decentralization

as Processes of Experimentation
Particularly in the discussion about systems competition, interjurisdic-

tional and regulatory competition, a number of authors, who are rela-
tively optimistic about the potential positive effects of these competition
processes, base their more positive assessment explicitly on evolutionary
reasonings.9 The starting point of their argumentation is the Hayekian
knowledge problem and his critique of neoclassical concepts of com-
petition, as was shown in the last section. They suggest using evolution-
ary concepts of competition (Schumpeter, Hayek) for analysing the effects
of interjurisdictional and regulatory competition. Particularly, Hayek’s
concept of ‘competition as a discovery procedure’ has been applied to
show that interjurisdictional and regulatory competition can be charac-
terized also as an evolutionary process of variation and selection of
policies and institutions/legal rules, in which new knowledge is generated
of how the jurisdictions can improve their bundles of public goods, regu-
lations and taxes.10 From this evolutionary perspective decentralization
and interjurisdictional/regulatory competition can lead to advantages in
the form of innovations, which were not seen from a purely neoclassi-
cal point of view. What would a more detailed application of this argu-
ment of the innovative effects of experimentation to federalism look
like?11
Starting from the Hayekian knowledge problem, jurisdictions have only
fallible hypotheses about the best bundles of public goods, legal rules and
taxes. In their endeavour to attract mobile individuals, firms and produc-
tion factors, they will experiment with different bundles of public goods,
legal rules and taxes. The resulting competition process can be seen as a
process of experimentation, in which the mobile factors have the task of
deciding on the relative quality of the public services of the competing
jurisdictions. The basic idea is that in this process of variation and selection
the superior hypotheses about the best and/or cheapest way in which juris-
dictions can offer public services will be sifted out. Since successful juris-
dictions will flourish, and their less successful competitors will suffer from
losses of mobile resources, there will be incentives for the latter to attempt
to catch up with the leading jurisdictions, either by imitation (mutual learn-
ing) or by innovating themselves. In that respect, an evolutionary concep-
tualization interprets interjurisdictional (or regulatory) competition as a
‘discovery procedure’, in which new knowledge is generated and spread, or,
to put it in Schumpeterian terms, as a dynamic process of innovation and
imitation, in which technical progress in the form of product and process
innovations in regard to public goods and legal rules is created. Of course,
the appropriate working of those competition processes depends on a

number of conditions.
If we apply this evolutionary concept of interjurisdictional/regulatory
competition to a federal multilevel structure of jurisdictions (as, for example,
the EU, member states, regions, and municipalities), those competition
processes emerge on each jurisdictional level. Since the vertical delimitation
of competencies between the different federal levels defines the policies,
which the jurisdictions can use to compete with other jurisdictions on the
same level, this vertical allocation of competencies determines the action
parameters of the competing jurisdictions. The positive innovative effects of
interjurisdictional competition can emerge only in regard to those policies
on which the respective jurisdictions can decide freely. Therefore the vertical
allocation of competencies determines on which levels innovation of new
policies and therefore experimentation and mutual learning are possible.
What does more or less centralization imply for the extent of this positive
innovation effect due to interjurisdictional competition? Each shifting of
competencies to a higher jurisdictional level (or each harmonization) elimi-
nates competition among lower-level jurisdictions, and therefore discards the
advantages of experimenting with policy innovations. Vice versa, the shifting
of tasks to lower-level jurisdictions increases the scope of interjurisdictional
competition, leading to additional processes of parallel experimentation. If
competencies are allocated to the central level or a complete harmonization
(of regulations) takes places, interjurisdictional (or regulatory) competition
is completely eliminated and the positive innovative effects of experimenta-
tion and mutual learning cannot emerge.12 Consequently, it can be suggested
that the extent of centralization and decentralization has a crucial impact on
the extent of variation selection processes and the capability of a federal
system in creating and spreading innovations in regard to the performance of
jurisdictions. Since evolutionary innovation economics has shown that het-
erogeneity and variety is a positive resource for the innovation capability of
systems (Röpke 1977; Metcalfe 1998), variety within a decentralized federal
system can be expected to enhance the long-term development of the perfor-
mance of jurisdictions.
But the application of these arguments to the analysis of particular
processes of interjurisdictional or regulatory competition requires a more
differentiated approach in regard to the transmission mechanisms of inno-
vation and imitation of policies. It is necessary to distinguish between
different types of interjurisdictional (or regulatory) competition in depend-
ing on different assumptions in regard to mobility (Kerber 2000a, Kerber
and Budzinski 2003, Heine 2003). Advantages from mutual learning due to
parallel processes of experimentation can even emerge, if there is no mobil-
ity of goods, services, individuals, firms or production factors at all. If the
citizens of different jurisdictions can observe the performance of policies in

other jurisdictions, e.g. the German population can observe the labour-
market policies and their relative success and failure in other countries, they
can use this information as a benchmark for the evaluation of the perfor-
mance of their own government in regard to domestic elections as well as
the domestic government can learn from the positive and negative experi-
ences in other countries. Therefore parallel experimentation processes with
mutual learning are possible without any direct interjurisdictional competi-
tion in the form of gains and losses due to the migration of individuals, firms
and production factors and not even through an indirect interjurisdictional
competition via international trade. Only the mobility of information is nec-
essary for this so-called yardstick competition.13 Consequently, pure decen-
tralization within a federal system can already lead to these positive
innovation effects, even if interjurisdictional competition does not work –
for example, due to relatively high migration costs.
But if additionally to that there is also considerable mobility of goods
and services or even of firms and production factors, interjurisdictional
competition processes emerge, which increase the pressure on governments
to offer attractive bundles of public goods, regulations and taxes. It
depends crucially on the specific conditions of mobility, and therefore also
on the institutional framework, what type of interjurisdictional or regula-
tory competition processes arises. If, for example, firms can choose directly
between legal rules of different countries without having to migrate to
another jurisdiction (choice of law), regulatory competition among juris-
dictions in regard to these particular legal rules can be very intensive, as in
the case of e.g. competition among corporate laws in the US.
5.2 Implicit Evolutionary Arguments in Theories of Federalism and

Regulatory Competition
After these explicit evolutionary contributions, it will be shown that this

evolutionary argument about the innovation advantages of experimenta-
tion and variety has been used by many authors in the discussion about
federalism, centralization/decentralization and interjurisdictional/regula-
tory competition without explicit reference to evolutionary economics.
This argument was partly used as a – theoretically not integrated – addi-
tional argument in economic discussions about federalism or as an impor-
tant argument in isolated discussions about special problems, which have
gained almost no broad attention.
Oates – as one of the most important economists working on fiscal feder-
alism – had already mentioned the additional advantages of experimentation
with public policies, which are possible in a decentralized federal system, in
his classical standard book Fiscal Federalism (Oates 1972), although he did
not integrate this effect into his consequently elaborated welfare-theoretic
neoclassical approach to federalism. The same is true for other new contri-
butions to federalism as, for example, Breton’s Competitive Governments
(1996, p. 233), who, for example, used the external benchmark mechanism of
Salmon (1987), which goes back to the theory of rank-order tournaments
and is linked to the concept of yardstick competition.14 Thus the importance
of the innovation dimension of public policies and its reference to federalism
is increasingly emphasized in economic theories of federalism.
Oates (1999) took the opportunity in his comprehensive survey article on
fiscal federalism in the Journal of Economic Literature to stress the advan-
tages of experimentation in decentralized federal systems and claims the
necessity of an entire research programme for the study of innovation and
imitation processes of public policies in federal systems. After the diagno-
sis of the problem of imperfect knowledge about appropriate public pol-
icies, he develops the perspective of federalism as ‘laboratory federalism’.
‘In a setting of imperfect information with learning-by-doing there are
potential gains from experimentation with a variety of policies for address-
ing social and economic problems. And a federal system may offer some
real opportunities for encouraging such experimentation and thereby pro-
moting “technical progress” in public policy’ (Oates 1999, 1132).
It is particularly remarkable that Oates, who in the first part of his survey
article argues consistently within a neoclassical welfare-theoretic frame-
work, assesses the advantages of decentralized experimentation in some
cases as so large that they can dominate neoclassical arguments for cen-
tralization. As an example, he uses redistribution policy. First, he repeats
the neoclassical argument of traditional federalism theory that redistribu-
tion policies, aiming to help the poor, should be allocated to the central
level, because in the case of decentralized redistribution policies a ‘race to
the bottom’ problem should be expected due to the mobility of the rich and
poor between lower-level jurisdictions, leading to an inefficient low level of
redistribution. Although Oates does not deny this argument, he is in favour
of a decentralized redistribution policy (Oates 1999, 1131), because he esti-
mates the advantages of decentralized experimentation with different redis-
tribution policies as so high that they overcompensate potential negative
effects of ‘race to the bottom’ problems.
But these implicitly evolutionary arguments have been developed
without an explicit reference to the established literature on evolutionary
economics. Oates neither uses Hayek as a reference for the knowledge prob-
lems he wants to alleviate by his ‘laboratory federalism’ nor does he apply
an evolutionary concept of interjurisdictional competition, which he could
have used for integrating experimentation and innovation, imitation
processes in regard to public policies. On the contrary, his critical reason-

ings about interjurisdictional competition – in another section of his article
(1999, 1134) – are purely neoclassical in the Tieboutian tradition. Therefore
the theoretical perspective to integrate his implicitly evolutionary ‘labora-
tory federalism’ argument about the advantages of experimentation by
applying an explicitly evolutionary concept of interjurisdictional (or regu-
latory) competition (instead of a neoclassical concept of interjurisdictional
competition) into an integrated theory of federalism, which includes both
interjurisdictional competition and the innovation dimension, has not been
elaborated in his broad survey article by Oates.
But Oates (1999, 1133) develops a clear perspective for applying innovation
economics to the problem of the creation and spreading of new public poli-
cies and linking it to his concept of ‘laboratory federalism’. He mentions a
number of studies of political scientists, who analysed horizontal and verti-
cal diffusion processes of successful policy innovations within the US federal
system.15 He also applies typical arguments from innovation economics to
federalism. Is it possible that in a decentralized federal system the positive
information externalities from new public policies might lead to free-rider
problems and therefore to an underprovision of experimentation (appropri-
ability problem of innovations)?16 Should this problem be addressed by sub-
sidizing experimentation on lower jurisdictional levels? Or should the central
government undertake limited experiments without committing the nation to
an untested and risky policy measure’ (1133)? But Oates also sees ‘that rela-
tively independent efforts in a large number of states will generate a wider
variety of approaches to public policy than a set of centrally designed exper-
iments’ (1133). Thus the question is raised whether experimentation should
be organized ‘top-down’ or allowed to develop ‘bottom-up’. A very interest-
ing simulation model of the effects of decentralized experimentation was pre-
sented by Kollman, Miller, and Page (2000). In their model innovation and
imitation of policies is analysed for different cases of centralized and decen-
tralized political decision making. Their simulation results suggest a trade-off
between the advantages of experimentation and the advantages of the poten-
tial larger ability of the central level to solve problems.
Although Oates primarily uses arguments from neoclassical approaches
to innovation economics, it has to be seen as considerable progress that he
opens up the perspective of applying innovation economics systematically
to federalism and therefore to the problem of centralization and decentral-
ization. ‘More generally, we need a lot more work on the implications of
fiscal decentralization for both the amount of policy experimentation and
innovation’ (1999, 1134).
The problems of the advantages and disadvantages of centralization and
decentralization are also discussed in regard to the legal system. Particularly
within the EU, the question increasingly arises whether legal rules should
be harmonized with the long-term perspective of a uniform European law
or whether there also might be advantages from the diversity of legal rules.
In this discussion on uniformity vs diversity of law, legal scholars also devel-
oped the argument that a decentralized legal system, which allows for
variety and experimentation with new legal rules, can have considerable
advantages in comparison to an entirely centralized, uniform legal system.17
Also, in regard to legal rules – as institutional solutions for solving transac-
tion and cooperation problems – it cannot be assumed that the best legal
solutions have already been found. And from an evolutionary perspective,
a permanent stream of new problems emerge, which require a high adapt-
ability of the legal system. Consequently, searching for legal innovations is
necessary. Therefore a decentralized legal system (legal federalism), in
which parallel processes of experimentation with legal innovations, mutual
learning and a quick correction of inappropriate legal solutions are pos-
sible, might be much more innovative in the long run than a centralized legal
system. This can be true, in particular, if it also allows for competition
among legal rules (regulatory competition).
One famous example of regulatory competition was already mentioned
in section 3 – competition among corporate laws in the US.18 The federal
states have the competence in corporate law in the US. Since within the US,
the so-called ‘incorporation theory’ determines that the federal states
mutually each accept the corporate law of the others, firms have the choice
to incorporate in any of the federal states without having any restrictions
in regard to their location within the US. So firms can choose freely
between the corporate laws of 50 federal states. This situation has led to an
intensive regulatory competition between the corporate laws of the federal
states, which it has been possible to observe for decades. The incentives of
the states are ‘franchise taxes’, which incorporated firms have to pay.19 In
the US, there have been many studies of this regulatory competition among
corporate laws, focusing on the question whether it has led to a negative
‘race to the bottom’ or to a positive ‘race to the top’. Although there are
adherents of both claims, nowadays most scholars hold the opinion that
the positive effects prevail. Particularly interesting is that this regulatory
competition can be primarily described as a dynamic process of innovation
and imitation of new legal rules leading to a step-by-step improvement of
the corporate laws. Therefore in this special discussion about the evolution
of US corporate laws, evolutionary arguments from innovation economics
were applied to regulatory competition. These arguments encompassed
the advantages of experimentation with different legal rules, but also
Schumpeterian argumentation patterns on dynamic processes of innova-
tion and imitation, path-dependency argumentations due to critical-mass
phenomena and dynamic economies of scale-effects (Romano 1985,

Heine 2003).20
6. COMPETITIVE FEDERALISM FROM THE

PERSPECTIVE OF NEOCLASSICAL AND
EVOLUTIONARY ECONOMICS
What are the consequences for the development of an integrated theory of

federalism, which can give a better basis for answering the question of cen-
tralization and decentralization within the EU?
6.1 Knowledge, Competition, and Innovation in Federalism Theory
The traditional theory of fiscal federalism is based upon a static neoclas-

sical approach, in which the knowledge problem in regard to the appropri-
ate public goods and regulations is assumed to be already solved. Contrary
to that, an appropriate theory of federalism has to take this knowledge
problem seriously, which also implies the systematic taking into account of
the innovation dimension in regard to public goods and regulations.
Therefore the capability of a federal multilevel system of jurisdictions in
generating and spreading new problem solutions (public goods, regula-
tions) and therefore helping to solve the knowledge problem is a crucial cri-
terion for the evaluation of alternative options for the vertical allocation of
competencies. From the evolutionary perspective, a more decentralized
system is better able to exploit both the decentralized local knowledge and
the advantages that accrue from decentralized experimentation of lower-
level jurisdictions with policy innovations and their mutual learning by
imitation (diffusion of policies). Therefore a federal multilevel system of
jurisdictions also has to be seen as a system of innovation. In addition to
that, from an evolutionary point of view the main perspective changes from
a ‘top-down’ to a ‘bottom-up’ approach.
Since neoclassical concepts of interjurisdictional and regulatory compe-
tition cannot take innovation effects into account, evolutionary concepts of
competition based upon Hayekian and Schumpeterian approaches should
be applied. Due to the positive effects of interjurisdictional/regulatory
competition on innovations, and the considerable knowledge problems of
centralized decisions, allowing interjurisdictional and regulatory competi-
tion is a much more interesting policy option from an evolutionary per-
spective than from a purely neoclassical point of view. As a consequence, it
is suggested that one integrate systematically interjurisdictional and regu-
latory competition processes within a federal multilevel structure of
jurisdictions – leading to the perspective of an integrated concept of com-

petitive federalism. Of course, interjurisdictional and regulatory competi-
tion can only be permitted if they work satisfactorily in comparison to
alternative institutional solutions. But similar to private goods, the provi-
sion of the public services of jurisdictions should also be controlled by
competition as far as possible (application of the market paradigm). The
resulting reduction (or even loss) of the monopoly power of jurisdictions
in regard to individuals, firms and production factors, which leads to com-
petition among public policies, is another basic change of perspective,
which gains weight from an evolutionary point of view.
6.2 Centralization and Decentralization: Neoclassical and Evolutionary

Criteria
The consideration of evolutionary arguments might change the relative

importance of neoclassical criteria for the vertical allocation of competen-
cies. But also most evolutionary economists would not deny that the neo-
classical criteria such as spillover effects (externalities), economies of
scale-effects, and heterogeneity of preferences are relevant criteria for
deciding on the optimal degree of centralization or decentralization. They
would also not deny that interjurisdictional and regulatory competition
might fail, for example, due to ‘race to the bottom’ problems or to the lack
of incentives. Therefore the methodological problem arises that both neo-
classical and evolutionary arguments are relevant without having a
common theoretical basis. But for scholars dealing with policy questions,
this is not an unusual situation. Many policy problems are characterized by
the fact that one theoretical approach is not sufficient to grasp the com-
plexity of the problem in a satisfactory way. This leads to the consequence
of the simultaneous application of different theoretical approaches and
implies a careful assessment of the relative importance of arguments from
different theoretical approaches, each of which can explain only a part of
the complex problem. Therefore the pragmatic method I want to suggest
also for the issue of centralization and decentralization is the simultaneous
application of both neoclassical and evolutionary arguments.
Additionally, a comprehensive theory of federalism needs an institutional
economics approach by applying the principle of ‘comparative institutional
analysis’. From that perspective, the issue of centralization and decen-
tralization can be conceived as the problem of what the institutional struc-
ture of a multilevel system of jurisdictions should look like. Theoretically,
this would entail both the vertical allocation of competencies within this
multilevel system of jurisdictions and the necessary set of institutional
arrangements (in the form of metarules), which are necessary for solving
conflicts and ensuring the working of the multilevel system of jurisdictions

as a whole. For the vertical allocation of competencies, it is suggested that
the already existing catalogue of (up to now predominantly neoclassical)
criteria is extended by additional evolutionary criteria. In that respect, par-
ticularly important are the innovation advantages of experimentation (and
variety), but also the Hayekian problem of the use of decentralized know-
ledge, dynamic economies of scale and the problem of path dependencies –
as additional evolutionary arguments.21 In the same way it is necessary to
examine the working of any interjurisdictional or regulatory competition
processes which might emerge in decenteralized federal systems, in order to
recognize potential market failures and reduce or eliminate them by estab-
lishing appropriate sets of rules for interjurisdictional or regulatory com-
petition. But from the perspective of evolutionary economics these analyses
should not only be made in regard to neoclassical static efficiency but also
in regard to dynamic efficiency in the tradition of the Schumpeterian
approaches. And as in other fields of public policy, in federalism also trade-
off problems between static and dynamic efficiency cannot be excluded.
But up to now, no elaborated theory of competitive federalism exists
which encompasses both neoclassical and evolutionary analyses of the
effects of federal multilevel systems of jurisdictions. Therefore much research
work needs to be done, particularly in regard to the application of evolu-
tionary innovation economics to the provision of public goods and regula-
tions by jurisdictions. But first studies about the problem of centralization
and decentralization of regulations and regulatory competition show that no
simple solutions can be expected. It seems rather to depend very much on the
specific kind of legal rules/regulations and on the specific institutional pre-
conditions whether decentralization and/or regulatory competition might be
an appropriate solution.22 Therefore for future research, a double strategy
seems to be most suitable. Although on one hand a general theory about the
design of federal multilevel systems of jurisdictions should be developed, the
specific problems in different fields of public goods, policies, and regulations
seem to be so heterogeneous that on the other hand profound special studies
also have to be made in each of these fields for answering the question of the
particular optimal degree of centralization and decentralization.
7. CONSEQUENCES FOR EUROPEAN

INTEGRATION: COMPETITIVE FEDERALISM
INSTEAD OF CENTRALIZATION
There is a large uncertainty about the final state of the European Union,
after the process of European integration is finished. The danger is that the
traditional idea of the nation state, in which nearly all kind of economic
policies are ‘naturally’ allocated to the central (i.e. national) level, will be
simply transferred to the EU, implying that the EU is seen as a new ‘nation
state’ in the traditional sense. From this perspective, it seems evident that in
the long run the central European level should have the same competencies
for public policies as the traditional nation states had in the past. This
would lead to a much stronger centralization of competencies and harmon-
ization than it is being discussed now. Within the European Union not
much scope would remain either for decentralization or variety. It can be
suggested that this is also a consequence of the lack of alternative concepts
of European integration that do not equate European integration with uni-
formity and harmonization.
The concept of competitive federalism is an alternative concept for
European integration that aims at the development of an institutional
structure for a federal multilevel system of jurisdictions in Europe, which
attempts to combine integration with decentralization and the potential for
maintaining variety.23 If we understand integration – in the tradition of the
judgment of the European Court of Justice – as the removal of barriers to
the mobility of individuals, firms, goods, services and production factors
(‘basic freedoms’) within the EU, and want to pursue the aim of as much
decentralization as possible (principle of subsidiarity), then both aims can
only be attained simultaneously within a multilevel system of jurisdictions
that is based upon the concept of competitive federalism (Kerber 2000b).
The crucial contribution of evolutionary economics to this reasoning is
that evolutionary arguments about the necessity, mechanisms and deter-
minants of innovation processes in regard to the provision of public goods
and regulations are very important for the elaboration of the long-term
advantages of a higher degree of decentralization within the EU – in com-
parison to purely neoclassical arguments.
It is not possible here to elaborate on specific argumentations and draw
conclusions about the appropriate vertical allocation of competencies in
the EU from this perspective. But a few general implications can be briefly
sketched:
1. Heterogeneity in regard to policies and regulations within the EU

should not only be seen as potential causes for impeding the internal
market and distortions of competition, but also as a positive resource
for innovation for the whole EU. Therefore policies aiming at ‘levelling
the playing field’ are problematic in order to maintain the innovation
capability of the EU.
2. Competition among member states, regions, and municipalities should
not be seen primarily as a danger for European integration, but instead
as an opportunity to extend the principle of competition (as a core

principle of the EU) also to the provision of many public goods and
regulations – as far as this is possible. In that respect, the concept of
competitive federalism opens up the perspective for the introduction of
an internal market for competing jurisdictions as location for individ-
uals, firms, and production factors (internal market for public goods
and regulations).
3. From the perspective of competitive federalism, the EU has to fulfil the
task of providing the institutional framework for ensuring the working
of the federal multilevel system of jurisdictions as a whole. For
example, even if the competencies for specific public policies, as e.g.
labour-market regulations or social policy, are allocated to the member
states, a set of metarules for the solving of conflicts (delimitation of
these national competencies) and problems (e.g. in regard to free-rider
problems due to migration) will be necessary. Therefore decentraliza-
tion of competencies in a European multilevel system also requires the
fulfilment of important tasks on the European level. But the European
Union should limit its activities – beyond specific tasks that should be
allocated entirely on the central level – more to the provision of an
institutional framework, within which the jurisdictions on different
levels can provide public services to European citizens.
8. PUBLIC POLICY AND EVOLUTIONARY

ECONOMICS: SOME CONCLUSIONS
One of the most important insights of evolutionary economics is that both
the complexity of human society and the permanent dynamic change due
to the emergence of novelties lead to fundamental limitations of our
knowledge about the world and its functioning. Therefore from an evolu-
tionary perspective the Hayekian knowledge problem is a crucial problem
that cannot be ignored. It rather has to be the starting point for all reason-
ings about our options to influence economic processes, and therefore for
all questions of public policy.
Our inevitable ignorance about the best policy solutions, however, should
not prevent evolutionary economists from making positive policy recom-
mendations, although we cannot foresee for sure what the exact effects of
these policy options are. Two reasons can be given for this conclusion:
1. In a changing world, in which nobody can foresee the future, actions

and policies of individuals, firms, and states have always to be based
upon uncertain, fallible knowledge. Therefore any requirement of

perfect foresight of the effects of policies is not compatible with an evo-
lutionary view of the world. So evolutionary economists especially
should take a pragmatic attitude in regard to the knowledge base that
is deemed as necessary for carrying out policies.24 This argument,
however, does not deny that often a careful analysis will come to the
conclusion that many policies should not be made all due to the lack
of knowledge about their effects.
2. In specific policy discussions many arguments from different theoret-
ical approaches are brought in. Although evolutionary arguments do
not seem to be as exact as e.g. neoclassical arguments, they are a very
valuable complement, and also a corrective to other arguments. Being
too scrupulous and reluctant in policy discussions due to the insight
into one’s own ignorance only leaves the ground to other theoretical
approaches that have not solved the knowledge problem either – but
are not aware of it and therefore pretend to know the right solutions.
In addition to that, evolutionary economics open up the perspective to do

systematic research on the question how we should deal with the problem
of our always fallible and deficient knowledge about the effects of policies.
One answer is well known: the substitution of discretionary, intervention-
ist policies by rule-oriented policies – such as Hayek’s demand for a ‘rule
of law’ and the claim of the Freiburg School of Law and Economics
(Ordoliberalism) for a stable institutional framework for the market
economy (‘Wettbewerbsordnung’, ‘Ordnungspolitik’) – can be seen also as
the demand for a specific form of economic policy that economizes on the
amount of knowledge that is necessary for public policy (Eucken 1952,
Hayek 1973). The second answer is that public policy should be institu-
tionalized in such a way that new knowledge about appropriate public poli-
cies is endogenously generated – implying both the production of
appropriate policy innovations and the fast discovery and elimination of
wrong policies. This is perfectly in line with Oates’ concept of federalism as
‘laboratory federalism’. Since evolutionary economics has comparative
advantages in doing research on innovation, learning, and change, it should
develop concepts of how public policy could be organized as a systematic,
innovative, searching process in regard to how jurisdictions can improve
their public policies.
The concept of competitive federalism can be seen as an application of
this general idea. The establishment of a federal multilevel system of juris-
dictions with as much decentralization and interjurisdictional and regula-
tory competition as is feasible, implies the implementation of a whole set of
processes of experimentation with public policies on different jurisdictional
levels – including the sifting out of superior policies, mutual learning by

imitation, and the elimination of erroneous policies. Since competitive fed-
eralism can also be seen as a concept for implementing an innovation
system in regard to the provision of public goods and regulations, evolu-
tionary economics might contribute considerably to the development of an
integrated concept of competitive federalism. In particular, the investiga-
tion of the knowledge-generating effect of decentralized experimentation
with new policies, including the creation of policy innovations and the often
difficult process of policy diffusion, might be an interesting issue for apply-
ing evolutionary analyses. Therefore the question of ‘centralization vs
decentralization’ is a promising candidate for the application of evolution-
ary economics. Vice versa, it has become evident that the main advantages
of decentralization, and therefore of federalism in general, only become
clearly apparent from the perspective of evolutionary economics.
NOTES
1. See, e.g., Koch (1996), Wegner (1996, 1997), Slembeck (1997), Meier and Slembeck
(1998), Witt (2003), and the contributions in Pelikan and Wegner (2003).
2. See Oates (1972, 1999) and Breton (1996).
3. See, e.g., Oates and Schwab (1988), Siebert and Koop (1990), Sinn (1990, 1997, 2003),
Kenyon and Kincaid (1991), Sinn (1992), Vanberg and Kerber (1994), Frey and
Eichenberger (1995), Streit and Mussler (1995), Sun and Pelkmans (1995), Bratton and
McCahery (1997), Kerber (1998, 2000b), Streit and Wohlgemuth (1999), Feld (2000),
Kollman, Miller and Page (2000), Van den Bergh (2000), Wellisch (2000), Esty and
Geradin (2001), Besley and Coate (2003).
4. See, in particular, Siebert and Koop (1990), Hauser and Hösli (1991), Woolcock (1994),
Sun and Pelkmans (1995), Streit and Mussler (1995), Sinn (1997), Gatsios and Holmes
(1998), Kerber (2000b), Van den Bergh (2000), Esty and Geradin (2001), Heine and
Kerber (2002), Kerber and Budzinski (2003), Marciano and Josselin (2003).
5. The assumptions of Tiebout’s model are nearly identical to the assumptions of the
perfect competition model, e.g. in regard to the homogeneity of public goods and juris-
dictions, perfect market transparency, and atomistic structure of the jurisdictions.
6. See, e.g., Brennan and Buchanan (1980), Besley and Coate (2003), and Wilson and
Wildasin (2004).
7. See Hayek (1945, 1978, 1996) and Wegner (1996).
8. As examples for arguments from political science, see Inman and Rubinfield (1997) and
Weingast (1995).
9. See, e.g., Vihanto (1992), Vanberg and Kerber (1994), Frey and Eichenberger (1995),
Streit (1996), Kerber (1998, 2000a), Streit and Wohlgemuth (1999), Heine and Kerber
(2002), Wohlgemuth (2002); for a comparison of neoclassical and evolutionary
approaches to systems competition, see Windisch (1998). For an integrated evolutionary
concept of competition, which encompasses both Schumpeterian and Hayekian argu-
ments, see Kerber (1997).
10. These considerations are also directly linked to the general Hayekian approach of cul-
tural evolution, which suggests that a process of variation and selection of rules might
lead to the development of superior rules and institutions (Hayek 1973; Vanberg 1992).
11. For the following, see in more detail Kerber (1998, 2000a); for the general advantages of
experimentalism from the perspective of economic policy, see also Okruch (2003).
12. Please note that sequential experimentation processes, in which particular policies are
implemented, and after evaluation maintained or modified, are possible also on the
central level. The advantage of decentralised competencies is that here parallel experi-
mentation processes and simultaneous mutual learning are possible, which can be
expected to lead to a much faster growth of knowledge.
13. See, e.g., Besley and Case (1995), Bodenstein and Ursprung (2005) and for a short
overview Schnellenbach (2004); the basic ideas go back to Salmon (1987).
14. It would be worthwhile to analyse the widespread concept of yardstick competition in
regard to its implicit evolutionary character, which cannot be done here.
15. See for this mainly empirical political science literature on the diffusion of policy inno-
vations, e.g., Walker (1969), Gray (1973, 1994), Canon and Baum (1981), Berry and
Berry (1991), Berry (1994), Nice (1994), Rogers (1995), Sparer and Brown (1996),
Mintrom (1997), and Clark (2000).
16. See in more detail Rose-Ackerman (1980), Mukand and Rodrik (2002), Strumpf (2002).
17. See, e.g., Kötz (1986, p. 9), Behrens (1986, p. 26), Parisi and Ribstein (1998, p. 238); for
advantages of experimentation in regulatory competition, see Streit and Mussler (1995),
Ogus (1999), Kerber (2000a), Van den Bergh (2000), Kerber and Budzinski (2003); from
a more critical perspective, see Kieninger (2002). The basic ideas of ‘laboratory federal-
ism’ stem from early legal scholars in the US (see Oates 1999, p. 1132).
18. See, e.g., Romano (1985, 1998), Bebchuk (1992), and Easterbrook and Fischel (1996).
19. See Romano (1985). Delaware, for example, which is the market leader in this market for
corporate law, is financing a considerable part of its budget by this franchise tax.
20. Recently, this US debate on the effects of regulatory competition between corporate laws
has become relevant in regard to the EU. The so-called ‘Centros’ decision of the
European Court of Justice of 1999 has triggered off an intensive discussion among legal
scholars, of whether the ‘Incorporation theory’ should also be implemented within the
EU. Contrary to the present situation, this would imply the mutual recognition of cor-
porate laws and open up the long-term perspective that in the future firms in Europe
might choose also freely between the corporate laws of different European countries
leading to regulatory competition among the corporate laws of the member states of the
EU (Wouters 2000, Heine and Kerber 2002). For an analysis of the workability of this
regulatory competition in regard to potential path dependencies see Heine and Kerber
(2002), who apply explicitly evolutionary concepts of innovation economics (technolog-
ical paradigms of Dosi) to legal evolution and regulatory competition.
21. For preliminary catalogues of criteria for the vertical allocation of competencies in
regard to regulations, see Grundmann and Kerber (2002), Kerber and Heine (2002), Van
den Bergh (2002) and Kerber and Grundmann (2006).
22. See, for example, Grundmann and Kerber (2002) for an application on the problem of
centralization and decentralization of contract law in the EU, Van den Bergh and
Camesasca (2001, pp. 125–65) and Kerber (2003b) in regard to competition law, and
Heine and Kerber (2002) and Heine (2003) in regard to corporate law.
23. For attempts to apply competitive federalism to the EU, see e.g. Kirchner (1998), Frey
and Eichenberger (1999), and Kerber (2000b, 2003a).
24. For a plea for a more pragmatic view of economic policy (with explicit reference to
Dewey and additional literature), see Okruch (2003).
REFERENCES
Bebchuk, L.A. (1992), ‘Federalism and the corporation: the desirable limits on state
competition in corporate law’, Columbia Law Review, 105, 1435–510.
Behrens, P. (1986), ‘Voraussetzungen und Grenzen der Rechtsfortbildung durch
Rechtsvereinheitlichung’, Rabels Zeitschrift für ausländisches und internationales
Privatrecht, 50, 19–34.
Berry, F.S. (1994), ‘Sizing up state policy innovation research’, Policy Studies
Journal, 22, 442–56.
Berry, F.S. and W.D. Berry (1991), ‘Specifying a model of state policy innovations’,
American Political Science Review, 85, 573–9.
Besley, T. and A. Case (1995), ‘Incumbent behavior, tax-setting, and yardstick com-
petition’, American Economic Review, 85, 25–45.
Besley, T. and S. Coate (2003), ‘Centralized versus decentralized provision of local
public goods: a political economy analysis’, Journal of Public Economics, 87,
2611–37.
Bodenstein, M. and H.W. Ursprung (2005), ‘Political yardstick competition, eco-
nomic integration, and constitutional choice in a federation’, Public Choice
(forthcoming).
Bratton, W.W. and J.A. McCahery (1997), ‘The new economics of jurisdictional
competition: devolutionary federalism in a second-best world’, The Georgetown
Law Journal, 86, 201–78.
Brennan, G. and J.M. Buchanan (1980), The Power to Tax: Analytical Foundations
of a Fiscal Constitution, Cambridge, MA: Cambridge University Press.
Breton, A. (1996), Competitive Governments: An Economic Theory of Politics and
Public Finance, Cambridge, MA: Cambridge University Press.
Canon, B.C. and L. Baum (1981), ‘Patterns of adoption of tort law innovations: an
application of diffusion theory to judicial doctrines’, American Political Science
Review, 79, 178–94.
Clark, J. (2000), ‘Policy attributes and state policy innovation’, Southeastern
Political Review, 28, 3–25.
Demsetz, H. (1969), ‘Information and efficiency: another viewpoint’, Journal of
Dowding, K., P. John and S. Biggs (1994), ‘Tiebout: “A Survey of the Empirical
Literature” ’, Urban Studies, 31, 767–97.
Easterbrook, F.H. and D.R. Fischel (1996), The Economic Structure of Corporate
Law, Cambridge, MA: Harvard University Press.
Esty, D.C. and D. Geradin (eds) (2001), Regulatory competition and Economic
Integration: Comparative perspectives, Oxford: Oxford University Press.
Eucken, W. (1952), Grundsätze der Wirtschaftspolitik, Tübingen: Mohr Siebeck.
Feld, L. (2000), Steuerwettbewerb und seine Auswirkungen auf Allokation und
Distribution:Eine empirische Analyse für die Schweiz, Tübingen: Mohr Siebeck.
Frey, B.S. and R. Eichenberger (1995), ‘Competition Among Jurisdictions: The
Idea of FOCJ’, in L. Gerken (ed.), Competition Among Institutions, Basingstoke
and London: Macmillan, pp. 209–29.
Frey, B.S. and R. Eichenberger (1999), The New Democratic Federalism for Europe,
Gatsios, K. and P. Holmes (1998), ‘Regulatory Competition’, in P. Newman (ed.),
The New Palgrave Dictionary of Economics and the Law, vol. 3, London and
Basingstoke: Macmillan, pp. 271–5.
Gray, V. (1973), ‘Innovation in the states: a diffusion study’, American Political
Science Review, 67, 1174–85.
Gray, V. (1994), ‘Competition, emulation, and policy innovation’, in L.C. Dodd and
J. Calvon (eds), New Perspectives on American Politics, Washington, DC: CQ
Press, pp. 230–48.
Grundmann, S. and W. Kerber (2002), ‘European system of contract laws. A map
for combining the advantages of centralised and decentralised rule-making’, in
S. Grundmann and J. Stuyck (eds), An Academic Greenpaper on European

Contract Law, The Hague/London/New York: Kluwer Law International,
pp. 295–342.
Hauser, H. and M. Hösli (1991), ‘Harmonization of Regulatory Competition in the
EC (and the EEA)?’, Aussenwirtschaft, 46, 497–512.
Hayek, F.A.v. (1945), ‘The use of knowledge in society’, American Economic
Review, 35, 519–30.
Hayek, F.A.v. (1948), ‘The meaning of competition’, in F.A.v. Hayek, Individualism
and Economic Order, Chicago: University of Chicago Press, pp. 92–106.
Hayek, F.A.v. (1973), Law, Legislation and Liberty. Vol. 1: Rules and Order, Chicago
and London: University of Chicago Press.
Hayek, F.A.v. (1978), ‘Competition as a discovery procedure’, in F.A.v. Hayek,
Studies in Philosophy, Politics and Economics, Chicago: Chicago University Press,
pp. 66–81.
Hayek, F.A.v. (1996), ‘Die Irrtümer des Konstruktivismus und die Grundlagen
legitimer Kritik gesellschaftlicher Gebilde’, in F.A.v. Hayek, Die Anmaßung von
Wissen (ed. W. Kerber), Tübingen: Mohr Siebeck, pp. 16–36.
Heine, K. (2003), Regulierungswettbewerb im Gesellschaftsrecht: Zur Funktions-
fähigkeit eines Wettbewerbs der Rechtsordnungen im europäischen Gesellschafts-
recht, Berlin: Duncker & Humblot.
Heine, K. and W. Kerber (2002), ‘European corporate laws, regulatory competition
and path dependence’, European Journal of Law and Economics, 13, 43–71.
Inman, R.P. and D.L. Rubinfeld (1997), ‘The political economy of federalism’, in
D. Mueller (ed.) Perspectives on Public Choice: A Handbook, Cambridge, MA:
Cambridge University Press, pp. 73–105.
Kenyon, D.A. and J. Kincaid (eds) (1991), Competition Among States and Local
Governments: Efficiency and Equity in American Federalism, Washington; DC:
Urbane Institute Press.
Kerber, W. (1997), ‘Wettbewerb als Hypothesentest: eine evolutorische Konzeption
wissenschaffenden Wettbewerbs’, in K.v. Delhaes and U. Fehl (eds), Dimensionen
des Wettbewerbs: Seine Rolle in der Entstehung und Ausgestaltung von
Wirtschaftsordnungen, Stuttgart: Lucius & Lucius, pp. 29–78.
Kerber, W. (1998), ‘Zum Problem einer Wettbewerbsordnung für den
Systemwettbewerb’, Jahrbuch für Neue Politische Ökonomie, 17, 199–231.
Kerber, W. (2000a), ‘Rechtseinheitlichkeit und Rechtsvielfalt aus ökonomischer
Sicht’, in S. Grundmann (ed.), Systembildung und Systemlücken in Kerngebieten
des Europäischen Privatrechts. Gesellschafts-, Arbeits- und Schuldvertragsrecht,
Tübingen: Mohr Siebeck, pp. 67–97.
Kerber, W. (2000b), ‘Interjurisdictional competition within the European Union’,
Fordham International Law Journal, 23, S217–49.
Kerber, W. (2003a), ‘Wettbewerbsföderalismus als Integrationskonzept für die
Europäische Union’, Perspektiven der Wirtschaftspolitik, 4, 43–64.
Kerber, W. (2003b), ‘An international multi-level system of competition laws:
Federalism in Antitrust’, in J. Drexl (ed.), The Future of Transnational Antitrust –
From Comparative to Common Competition Law, Berne, The Hague, London,
Boston, MA: Staempfli/Kluwer, pp. 269–300.
Kerber, W. and O. Budzinski (2003), ‘Towards a differentiated analysis of competi-
tion of competition laws’, ZWeR – Journal of Competition Law, 1, 411–48.
Kerber, W. and K. Heine (2002), ‘Zur Gestaltung von Mehr-Ebenen-
Rechtssystemen aus ökonomischer Sicht’, in C. Ott and H.-B. Schäfer (eds),
Vereinheitlichung und Diversität des Zivilrechts in transnationalen

Wirtschaftsräumen, Tübingen: Mohr Siebeck, pp. 167–94.
Kerber, W. and S. Grundmann (2006), ‘An optional European contract law code:
advantages and disadvantages’, European Journal of Law and Economics (forth-
coming).
Kieninger, E.-M. (2002), Wettbewerb der Privatrechtsordnungen im Europäischen
Binnenmarkt, Tübingen: Mohr Siebeck.
Kirchner, C. (1998), ‘The principle of subsidiarity in the Treaty of European Union:
a critique from the perspective of constitutional economics,’ Tulane Journal of
International and Comparative Law, 6, 291–308.
Koch, L.T. (1996), Evolutorische Wirtschaftspolitik, Tübingen: Mohr Siebeck.
Kollman, K., J.H. Miller and S.E. Page (2000), ‘Decentralization and the search for
policy solutions’, Journal of Law, Economics, & Organization, 16, 102–28.
Kötz, H. (1986), ‘Rechtsvereinheitlichung – Nutzen, Kosten, Methoden, Ziele’,
Rabels Zeitschrift für ausländisches und internationales Privatrecht, 50, 1–18.
Marciano, A. and J.-M. Josselin (eds) (2003), From Economics to Legal Competi-
tion, Cheltenham, UK and Northampton, MA, USA: Edward Elgar.
Meier, A. und T. Slembeck (1998), Wirtschaftspolitik. Kognitiv-evolutionärer
Ansatz, second edition, Munich/Vienna: Oldenbourg.
Metcalfe, J.S. (1998), Evolutionary Economics and Creative Destruction, London
and New York: Routledge.
Mintrom, M. (1997), ‘Policy entrepreneurs and the diffusion of innovation’,
American Journal of Political Science, 41, 738–70.
Mukand, S. and D. Rodrik (2002), ‘In search of the Holy Grail: policy convergence,
experimentation, and economic performance’, unpublished manuscript, John
F. Kennedy School of Government, Harvard University.
Nice, D.C. (1994), Policy Innovation in State Government, Ames, IO: Iowa State
University Press.
Oates, W.E. (1972), Fiscal Federalism, New York: Harcourt Brace Jovanovich.
Oates, W.E. (1999), ‘An essay on fiscal federalism’, Journal of Economic Literature,
37, 1120–49.
Oates, W.E. and R.M. Schwab (1988), ‘Economic competition among jurisdictions:
efficiency enhancing or distortion inducing?’, Journal of Public Economics, 35,
333–54.
Ogus, A. (1999), ‘Competition between national legal systems: a contribution of
economic analysis to comparative law,’ International and Comparative Law
Quarterly, 48, 405–18.
Okruch, S. (2003), ‘Knowledge and economic policy: a plea for political experi-
mentalism’, in P. Pelikan and G. Wegner (eds), The Evolutionary Analysis of
Economic Policy, Cheltenham; UK and Northampton, MA, USA: Edward
Elgar, pp. 67–95.
Parisi, F. and L.E. Ribstein (1998), ‘Choice of law’, in P. Newman (ed.), The New
Palgrave Dictionary of Economics and the Law, vol. 1, London and Basingstoke:
Macmillan, pp. 236–41.
Pelikan, P. and G. Wegner (eds) (2003), The Evolutionary Analysis of Economic
Policy, Cheltenham; UK and Northampton, MA, USA: Edward Elgar.
Rogers, E. (1995), Diffusion of Innovations, fourth edition, New York: Free Press.
Romano, R. (1985), ‘Law as a product: some pieces of the incorporation puzzle’,
Journal of Law, Economics, and Organization, 1, 225–83.
Romano, R. (1998), ‘Competition for state corporate law’, in P. Newman (ed.), The
New Palgrave Dictionary of Economics and the Law, vol. 1, London and
Basingstoke: Macmillan, pp. 364–70.
Röpke, J. (1977), Die Strategie der Innovation, Tübingen: Mohr Siebeck.
Rose-Ackerman, S. (1980), ‘Risk taking and reelection: does federalism promote
innovation?’, Journal of Legal Studies, 9, 593–616.
Salmon, P. (1987), ‘Decentralisation as an incentive scheme’, Oxford Review of
Economic Policy, 3, 24–42.
Schnellenbach, J. (2004), Dezentrale Finanzpolitik und Modellunsicherheit: Eine the-
oretische Untersuchung zur Rolle des fiskalischen Wettbewerbs als Wissen gener-
ierender Prozess, Tübingen: Mohr Siebeck.
Siebert, H. and M.J. Koop (1990), ‘Institutional competition. A concept for
Europe?’, Aussenwirtschaft, 45, 439–62.
Sinn, H.-W. (1990), ‘The limits to competition between economic regimes’,
Empirica – Austrian Economic Papers, 17, 3–14.
Sinn, H.-W. (1997), ‘The selection principle and market failure in systems competi-
tion’, Journal of Public Economics, 88, 247–74.
Sinn, H.-W. (2003), The New Systems Competition, Malden, MA: Blackwell.
Sinn, S. (1992), ‘The taming of Leviathan: competition among governments’,
Constitutional Political Economy, 2, 177–96.
approach to policy-making’, Constitutional Political Economy, 8 (3), 225–54.
Sparer, M.S. and L.D. Brown (1996), ‘States and the health care crisis: the limits and
lessons of laboratory federalism’, in R.F. Rich and W.D. White (eds), Health
Policy, Federalism, and the American States, Washington, DC: Urban Institute
Press, pp. 181–202.
Streit, M.E. (1996), ‘Systemwettbewerb im europäischen Integrationsprozess’, in
U. Immenga, W. Möschel and D. Reuter (eds), Festschrift für Ernst-Joachim
Mestmäcker: zum siebzigsten Geburtstag, Baden-Baden: Nomos, pp. 512–35.
Streit, M.E. and W. Mussler (1995), ‘Wettbewerb der Systeme und das Binnenmarkt
programm der Europäischen Union’, in L. Gerken (ed.), Europa zwischen
Ordnungswettbewerb und Harmonisierung, Berlin: Springer, pp. 75–107.
Streit, M.E. and M. Wohlgemuth (eds) (1999), Systemwettbewerb als Heraus-
forderung an Politik und Theorie, Baden-Baden: Nomos.
Strumpf, K.S. (2002), ‘Does government decentralization increase policy innova-
tion?’, Journal of Public Economic Theory, 4, 207–41.
Sun, J.-M. and J.J. Pelkmans (1995), ‘Regulatory competition in the single market’,
Journal of Common Market Studies, 33, 67–89.
Tiebout, C.M. (1956), ‘A pure theory of local expenditures’, Journal of Political
Economy, 64, 416–24.
Vanberg, V. (1992), ‘Innovation, Cultural Evolution, and Economic Growth’, in U.
Witt (ed.), Explaining Process and Change. Approaches to Evolutionary
Economics, Ann Arbor, MI: University of Michigan Press, pp. 105–21.
Vanberg, V. and W. Kerber (1994), ‘Institutional competition among jurisdictions:
an evolutionary approach’, Constitutional Political Economy, 5, 193–219.
Van den Bergh, R. (2000), ‘Towards an institutional legal framework for regulatory
competition in Europe’, Kyklos, 53, 435–66.
Van den Bergh, R. (2002), ‘Forced harmonization of contract law in Europe: com-
munication from the Commission on European Contract Law’, in S. Grundmann
and J. Stuyck (eds), An Academic Greenpaper on European Contract Law, The
Hague, London, New York: Kluwer Law International, pp. 249–68.
Van den Bergh, R.J. and P.D. Camesasca (2001), European Competition Law and
Economics: A Comparative Perspective, Antwerp: Intersentia.
Vaubel, R. (1995), The Centralisation of Western Europe: The Common Market,
Political Integration, and Democracy, London: Institute of Economic Affairs.
Vaubel, R. (1997), ‘The constitutional reform of the European Union’, Consti-
tutional Political Economy, 7, 317–24.
Vihanto, M. (1992), ‘Competition between local governments as a discovery proce-
dure’, Journal of Institutional and Theoretical Economics, 148, 411–36.
Walker, J.L. (1969), ‘The diffusion of innovation among the American states’,
American Political Science Review, 63, 880–99.
Wegner, G. (1996), Wirtschaftspolitik zwischen Selbst- und Fremdsteuerung – Ein
neuer Ansatz der Wirtschaftspolitik, Baden-Baden: Nomos.
Wegner, G. (1997), ‘Economic policy from an evolutionary perspective’, Journal of
Institutional and Theoretical Economics, 153, 485–509.
Weingast, B.R. (1995), ‘The economic role of political institutions: market-
preserving federalism and economic development’, Journal of Law, Economics
and Organisation, 11, 1–31.
Wellisch, D. (2000), Theory of Public Finance in a Federal State, Cambridge, MA:
Wilson, J.D. and D.E. Wildasin (2004), ‘Capital tax competition: bane or boon’,
Journal of Public Economics, 88, 1065–91.
Windisch, R. (1998), ‘Modellierung von Systemwettbewerb: Grundlagen,
Konzepte, Thesen’, Jahrbuch für Neue Politische Ökonomie, 17, 121–54.
Wohlgemuth, M. (2002), ‘Evolutionary approaches to politics’, Kyklos, 55, 223–46.
Woolcock, S. (1994), The Single European Market: Centralization or Competition
Among National Rules?, London: Royal Institute of International Affairs.
Wouters, J. (2000), ‘European company law: Quo Vadis?’, Common Market Law
Review, 37, 257–307.
14. Can evolutionary economics
make a billion $ difference for
60 per cent of the world’s poor
in Asia?
Hans-Peter Brunner*
1. THE POWER OF THE EVOLUTIONARY

APPROACH
The first issue of the Journal of Evolutionary Economics (1991) featured

prominent academics in the field, such as Dosi, Boulding, and Eliasson
among others, presenting and discussing key elements of an ‘evolutionary
approach’. The key notions elaborated constitute a possible evolutionary
approach to economic development. They are: irreversibility (history and
path dependence); institutions; population; restructuring and ‘creative
destruction’; disequilibrium and multiple equilibria; information and
learning; innovation and entrepreneurship; non-average behaviour by eco-
nomic agents; and, positive feedback mechanisms, linkages, and externali-
ties. The discussion in this and other parts of the literature (Witt, 1992;
Silverberg, 1994) does not provide a clear manifesto of evolutionary eco-
nomics. The ability to differentiate evolutionary economics from neoclassi-
cal or other types of economic approaches is, however, considered of less
importance when compared to the fruitfulness of such concepts in theo-
retical and empirical exploration of economic development phenomena
(Adams, 1993; Witt, 2001b). The argument is not that neoclassical eco-
nomics is wrong or unnecessary but that there are demonstrable advantages
in using institutional and evolutionary concepts. ‘Fruitfulness’ means
making a difference based on the evolutionary explanation of economic
phenomena. Does the evolutionary approach have the power to make a
difference, that is, to make a real and measurable difference, to people’s
lives, in the poor, developing economies of Asia?
This chapter explores and explains how key evolutionary concepts are
being applied in field cases to the poorest of the developing economies in
325
Asia, Nepal and the Indian state of West Bengal. This chapter does not
intend to present complete institutional economics and evolutionary math-
ematical models. For those I refer to the literature (for example Allen and
Brunner, 2001; Adams, Brunner and Raymond, 2003; Brunner and Allen,
2005). Rather it is the key institutional and evolutionary concepts that can
lead to novel solutions to empirical issues encountered in development
practice. Concepts are embedded in analytical methods. Method can be
mapped onto an evolutionary spectrum, as I will do in section 4 of the
chapter. The spectrum starts with ‘simple’ neoclassical methods with
restrictive assumptions and extends to ‘complex’ adaptive-learning models.
Even simple methods can help explore key institutional and evolutionary
concepts to act on in practice. For example in the Nepal case, we used a neo-
classical CGE model, which was modified to consider transaction-cost
reductions in the financial sector as the drive for economic change. The
Nepal information-economics issues were then examined from a different
angle with a dynamic game-theoretic model. Finally we progressed along
the evolutionary spectrum with the application of an evolutionary spatial
and non-linear feedback-loop model in the example of West Bengal. This
because the issue we felt required higher complexity of analysis, and fur-
thermore because we had the luxury and money to accumulate the requi-
site quality and quantity of data to test such a model. Models are always,
and to a degree, partial representations of the cloud of reality they depict.
Knowledge is really about knowing what simplifications can be made in
practice without compromising on results. Models should be detailed
enough to capture all the important causal real-world relationships that
explain outcomes of development. Policy actions spring from fruitful
analysis of underlying causes of underdevelopment.
Do evolutionary actions lead to significantly different outcomes from
actions that may have been taken, or not taken, on the basis of, let us say,
neoclassical concepts and empirical approaches in a field of analysis? Two
Asian Development Bank (ADB) investment examples illustrate that,
indeed, evolutionary concepts when embedded appropriately in a variety of
methods, have the power to make a big difference for economic develop-
ment in Asia and in the effectiveness of resource use by the ADB and its
borrowers. Making institutional and evolutionary economic analysis
policy-relevant also means relating different methods and concepts to the
methodological and concept frameworks of decision makers, who are still
largely steeped in neoclassical economics. The West Bengal example shows
how an essentially evolutionary analysis can be related back to traditional,
simple decision models of cost-benefit analysis.
Can evolutionary economics make a difference? 327
2. DEVELOPING COUNTRY SYMPTOMS OF POOR

INSTITUTIONS – THE CASE OF THE FINANCIAL
SECTOR IN NEPAL
The Concept of Irreversibility
Factors of production are not fungible. A factor is tied up with other

factors, and thus increases its marginal contribution beyond what it could
contribute on its own. Once factors are tied together they lose value if sep-
arated. Factors are up tied in two ways: they are tied through a specific tech-
nology and they are connected in an institutional context. They are
contextual. A simple example is cultivation where labour, animal power,
and land come together when a farmer uses a bullock-drawn plough to
prepare a wheat-field. The institutional web includes the specifics of land
tenure, input markets, irrigation rights, and credit institutions. A more
complicated example would be the organisation of a financial sector or a
national airline. In the words of Caballero and Hammour (2000, p. 7) ‘rela-
tionship specificity forms the underpinning of what institutional arrange-
ments are about; technological specificity forms the underpinning of what
restructuring is about. Most of the time, both are present simultaneously
and interact in important ways.’
Simply put, in a developing economy the productive factors, land, labour
and capital, can either remain in autarchy, or they can enter into joint pro-
duction through institutionalised co-operation. Institutions help to ensure
that factors in a joint production arrangement will get their commensurate
share of returns. When they do not, as is often the case in developing
economies like Nepal, or India, one or the other factor will stay in a situa-
tion of autarchy; that is, it will be withdrawn in whole or in part because of
the absence of incentives or rewards for participation. This means, for
example, subsistence jobs for labour will prevail or that a self-employed
informal sector will predominate. It may mean capital moving abroad, or
being diverted into consumption, rather than employed to put labour and
land jointly to work in high-yield activities.
Institutional Economics
The institutional environment as encountered in most developing

economies discourages factor co-operation. This is because the hold-up
problem is not resolved, so that co-operating factors are encouraged to pre-
commit. Institutions solve the hold-up problem by guaranteeing ex ante
the effective distribution of the various factors’ returns at the margin.
These ex-ante controls may be supported by ex-post mechanisms insuring
compliance with customary practices or legal requirements; using the term

very broadly, implicit or explicit contract enforcement is part and parcel of
a working set of institutions. Strong institutions prevent the appropriation
of returns attributed to one factor by another factor. For example, only
when capital is not appropriated by another factor, let us say labour,
through excessive market regulation in favour of labour, is it assured a
return commensurate with what is obtainable in international capital
markets. When capital can be assured its share of return it will co-operate
with labour to invest in job creation, and it will tie itself to the other factors
with efficient technologies.
Irreversibility and a weak institutional environment lead to the following
phenomena. There is reduced co-operation and at least one factor loses out
relatively, and usually it is labour, although all factors are likely to lose when
institutional co-ordination does not function. The poor institutional envir-
onment in Nepal’s financial sector combines bad corporate governance
arrangements with a lack of disclosure rules, accounting standards, and
bankruptcy procedures. This network of procedures and incentives leads to
a severe restriction on movements of capital to the more efficient sectors of
the economy and to resource underutilisation and misallocation. The inter-
mediation mechanism of financial institutions has largely not come into
being and credit remains directed to short-term commercial and construc-
tion finance, predicated on reputation or social ties. Caballero and
Hammour (2000) voice a generalisation that fits Nepal well, when they
remark, ‘The poor institutional framework is the result of a combination
of under-development of contracting and regulations and of overly pow-
erful political interest groups who have tilted the institutional balance
excessively in their favour (p. 10).’
To counter the co-operation failure, Nepal has pursued a policy of
directed lending where large portions of private loanable funds are directed
according to a government formula to the so-called deprived sections at low,
mandated rates. The chief targets are, in principle, small businesses, mar-
ginal farmers, and impoverished women. Most banks choose not to engage
in such loans because of high transaction costs, and simply follow the legal
alternative of pushing money to microcredit wholesalers. Moreover, the
quota system of directed lending leads financial institutions to see this
lending as a tax on their turnover and profits. Efforts at risk management
and evaluation of borrowers are diminished with the effect of high non-per-
forming loans. The deprived sections see directed lending as a give-away. It
has become the ideal target for rent seekers inside and outside the system.
Ultimately, the intent of lending to poorer sections of the economy is under-
mined. Directed lending in Nepal has increasingly favoured those with some
clout and good connections to avail of money give-aways.
Capital deployment in Nepal is being held up by financial-sector failures.

In this instance, capital is the appropriated factor, which gets a lower rate
of return than it should at the margin. Because two public-sector banks
control about two-thirds of the banking sector’s assets, it has been easy for
officials to command funds to cover government deficits and borrowing
needs. It has been simple, too, for the public-sector firms to cover their
losses by borrowing from public-sector commercial and investment banks.
There is thus the paradox that private banks are awash in loanable funds
but cannot pursue long-term lending in an atmosphere of reliable infor-
mation disclosure, secure contract enforcement, and prudential regulation,
while public institutions, whose assets have been raided by government and
politicians, are insolvent. A consequence is that capital in Nepal has been
reluctant to tie itself efficiently to labour and land, applying efficient tech-
nologies in commerce, manufacturing, or farming. Productivity remains
extremely low in all sectors and Nepal’s 2001 per-capita income of $235
earns membership in the club of the world’s poorest countries.
Ideally, small, but growing, amounts of credit would mobilise large
pools of underutilised labour and set off a growth process, but distortions
in the financial sector make this infeasible. Compared to the more capital-
abundant, wealthier countries, efficient investment in Nepal ought to take
the form of encouraging labour-intensive technologies but this is not hap-
pening in urban or rural areas. Theoretical and empirical work at the ADB
shows in Nepal’s case that financial-sector distortions and institutional
weakness have led to excessive capital invested in inefficient sectors of the
economy, mainly capital-intensive production in public-sector enterprises
and government services. Financial-sector repression leads to unemploy-
ment, impedes economic growth, and increases poverty.
Distortions also characterise rural lending institutions in Nepal. Despite
the nationwide system of branches of the state-owned Agricultural
Development Bank of Nepal (ADBN), offering low, subsidised loans,
informal money lending is pervasive. A key finding that has emerged in
Nepal is that informal and formal lenders offer borrowers differentiated
bundles of loans and services because of their differing opportunity costs
of funds, command of information, and the weight of transaction costs,
including under-the-table payments to formal agents (Adams et al., 2003).
This approach applies deep institutional analysis to explain the common
observation that despite the appearance that formal interest rates are well
below those of the money lenders, the informal sector is not displaced by
formal lending. The explanation is that when the full costs of a formal loan,
including transportation, lost time, documentation, and bribes, are esti-
mated, they converge upwards to the rate charged by the informal lenders,
who incorporate similar charges into their quoted interest rates.
Beck (1997) demonstrates that the emergence of formal and informal

financial markets occurs largely because of informational asymmetries
between lenders and borrowers. Gupta and Chaudhuri (1998) investigate
the effects of bribery in the formal sector. Morduch (1999) validates the
presence of variant informal and formal client risks and information costs.
Stiglitz and Weiss (1981), Broecker (1990), and Hannan (1997) investigate
the impact of information problems on interest rates and competition.
Dell’Ariccia (1998) employs a model of spatial competition in banking to
analyse the effects of product differentiation and information asymmetries
in determining barriers to entry. None of these studies treats in one rigor-
ous framework the principal features observed to typify South Asian rural
credit markets: spatial segmentation, information lacunae, transaction
costs, heterogeneous informal and formal agents, and moderately disparate
bundles of services or loan ‘products’. The Nepal work integrates each of
these facets into a competitive model of development finance with hetero-
geneous credit services, with an emphasis on how adverse selection prob-
lems associated with information asymmetries can evolve from differences
in credit services. This decidedly institutional approach hinges on the exis-
tence of information barriers across geography and customer segments.
The main features of rural finance are now well known. Most borrowers
are remote from formal sources of credit whereas informal sources are
readily accessible. Travel burdens add to the formal-sector transaction
costs. Informal lenders have good knowledge about individual credit histo-
ries in their circles of clients, while formal agents rely upon irrelevant
templates and place costly documentation encumbrances on potential bor-
rowers. Further, informal credit agents can tailor the attributes of a loan to
the length of a crop-growing season, offer a balloon payment after harvest,
and permit use of funds for consumption or weddings; the regulation-
bound formal agent can normally not make these accommodations. With
respect to enforcement, money-lenders and their ilk have immediate recov-
ery or punishment capabilities and can display flexibility when necessary.
Contrariwise, in Nepal, for example, as against tens of thousands of
defaults to the Agricultural Development Bank, only a handful of cases
have been brought to adjudication (Adams and Brunner, 1997). For the
borrower, the total loan experience entails a choice between one bundle of
services or ‘product’ and another, and a comparison to the associated total
costs. In brief, the informal loan ‘product’ is differentiated from the formal
loan ‘product’.
Although the perspectives of the informal and formal lenders differ, the
costs of determining the creditworthiness of clients, in order to address the
hidden information problem; monitoring loan use, to address the moral-
hazard problem; and, securing collection, to address the enforcement
problem, are high, while most loans are small and short-term. Levine
(1997) offers conditions for formal financial market viability, but in prac-
tice agricultural lending is very commonly non-viable for private-sector
credit institutions. Consequently, the formal credit sector is severely con-
strained in its ability to meet the requirements of rural development, at
least with current institutional arrangements, in-built incentives, and poor
information systems. The common response of governments and inter-
national donors has been to offer subsidised credit, set credit quotas by type
of loan, peg an array of interest rates, and target special groups such as
women or the poor. The outcome has usually been very low or even nega-
tive rates of social return on the one hand, and ultimate insolvency for the
institutions on the other.
What institutional analysis offers, in conjunction with careful measure-
ment of true transaction costs and informational asymmetries, is not only
an understanding of the workings of the informal and formal mechanisms,
but the hope of appropriate institutional design that can improve loan
delivery and recovery from formal agencies.
The system’s institutional and incentive failures are associated with use
of extremely primitive technologies and resistance to the adoption of new
skills and tools, oftentimes out of a misguided sense, shared by manage-
ment and labour, that their jobs are at risk. Nepal’s financial system oper-
ates largely on a manual basis. This leads to costly and cumbersome
procedures as well as poor service delivery. Payment and settlement systems
are not fully computerised and no electronic funds transfers are possible. It
may take up to 30 days to have a cheque cleared if it was issued outside
Kathmandu and as a result, the economy is largely ‘cash and carry’ based.
Moreover, the absence of information technologies in all areas of the finan-
cial sector makes information collection costly and unreliable, and hinders
effective monitoring and supervision. Nepal’s ADBN, for instance, is at
least three years behind in generating credible income statements and
balance sheets.
Development Policy Responses
Understanding irreversibility consequences for an economy and the role of

institutions in an evolutionary frame leads to sound policies for financial-
sector reform. Furthermore, there are several ways of estimating the
sizeable benefits that accrue from comprehensive reforms, buttressed by
the parallel adoption of modern technologies. The use of a modified
Computable General Equilibrium (CGE) model (Bhattarai, 1999) such
that change is driven by a reduction in transaction costs, demonstrates that
continuing down the path of rapid and comprehensive financial-sector
reform in Nepal will release extra resources for savings and investment.
‘Comprehensive reform’ means mutually reinforcing institutional and
legal developments involving the Ministry of Finance, Central Bank
(Nepal Rastra Bank), the Nepal Stock Exchange, commercial banks,
joint-venture banks, multinational banks, finance companies, and rural
development banks, including the Agricultural Development Bank of
Nepal, co-operatives, and NGOs.
Financial reform has a widespread effect on informal credit markets in the
rural sector. Nepal’s rural credit market is dominated by the informal sector.
In repression, access of rural firms or households to financial institutions is
limited and the local money-lenders charge rates of interest that may reflect
market power. Increased competition with formal-sector financial institu-
tions will, over time, increase lending-flows and offer lower rates. The
opening of rural savings windows by formal agencies will enhance savings-
flows throughout the financial system, raising Nepal’s aggregate savings rate.
In addition to the pro-poor effects on employment and income genera-
tion through more efficient capital allocation, the introduction of tech-
nologies will increase transparency and reduce rent-seeking opportunities
of individuals through monopolisation of information. Access to informa-
tion is often excessively priced for the poorest segment of the population.
The use of information technologies also allows for a more efficient fran-
chising and distribution system, and greater bargaining power for the orig-
inal producer against traditional intermediaries who often collect excessive
margins at the expense of the original producer.
In sum, the institutional immaturity of Nepal’s financial sector has pro-
duced the effects that arise from factor appropriation, in this case capital:
reduced co-operation with the other factors, labour and land, investment
tilted towards capital-intensive and unproductive projects; resulting under-
employment and poverty; and, chronic formal and informal credit market
segmentation. While the methods used in the analysis of factor appropria-
tion of returns are on the ‘simple’ side of the evolutionary spectrum, the
empirical application of institutional and evolutionary concepts to Nepal’s
financial sector problems makes a real and measurable difference.
The ADB has responded to the technological and institutional challenge
in the financial sector in Nepal. A series of technical assistance projects
amounting to over $10 million are aimed at improving financial and
corporate governance. This means improving disclosure standards, training
the accounting and auditing professions, updating and completing the legal
framework of the financial sector, and training the judiciary and attorneys
in commercial law and enforcement of contracts in the market framework.
It means modest investment of $10 to $20 million in the information infra-
structure of the financial sector, including an electronic payment system,
interbank settlements, automated teller machine connections, management

information systems, company information systems, and associated train-
ing and marketing, among other things. None of this would have been con-
ceived in this manner as a response to the co-operation and co-ordination
problem, without the power of evolutionary and institutional analysis.
An important further point, which draws on the evolutionary tradition,
is that institutional changes must move hand in hand, and in fact lead, tech-
nological innovations. The aforementioned reforms in institutional designs
and incentives must be reinforced by the introduction of computers, infor-
mation systems, and a nationwide information and communications grid.
Beginning in 2001, with ADB technical assistance, Nepal’s private banks
have undertaken an initiative to create an initial hub for payment clearances.
Long-term studies demonstrate large economic returns from a national
system that will embrace the public-sector banks, as they advance to pri-
vatisation, the rural banks, and microfinance institutions. E-commerce,
reduced telecommunications costs, and other advantages are understood to
be part of this bundle of services.
3. TECHNOLOGY RESTRUCTURING AND

POPULATION DYNAMICS – BRINGING THE
ENTREPRENEUR BACK: THE CASE OF TRADE
AND TRANSPORT IN WEST BENGAL
Evolutionary Concepts of Population Restructuring and ‘creative

destruction’
Modern economies are essentially evolutionary economies. They are in a

state of constant disequilibrium. Technology creation and diffusion
through learning is an essential feature. Entrepreneurs learn about new
technologies and are able to embody them in new ways in productive and
service activities. In the words of Caballero and Hammour, technological
specificity forms the underpinning of what restructuring is about. Resource
re-allocation is made in a market arrangement, constrained and regulated
by societal institutions. That arrangement allows for effective creative
destruction where new firms using the newest techniques are continuously
created, and outdated units are destroyed as their sales fall, their prices
become uncompetitive, and their inputs are bid away by succeeding firms.
Murrell (1990) argued strongly that the lack of creative destruction and
learning processes led to the rapid technological demise of the Eastern
bloc. The Eastern bloc was notoriously inefficient in restructuring produc-
tion factors from less productive uses into increasingly productive units.
It lacked the institutional mechanisms for competitive selection in a fluid

market and information environment. In Nepal, the institutional environ-
ment has been shown crucial for the efficiency of restructuring transactions
on a short to medium-term time-scale. Recent empirical work in transfor-
mation and developing countries has confirmed the applicability of the cre-
ative destruction concept, and the use of effective restructuring processes
for entrepreneurial learning, new employment, wealth creation, and
economic growth (Kornai, 2000; Brunner, 2001).
In the long term, the self-generating potential of an economy reforms
business organisations and increases the knowledge potential. Growth is
the result of these processes that move production constraints. Macro-
economic effects emerge from the micro-economic effects of institutional
and firm interaction as part of a sector population. Selection forces and
transition processes elevate an economy on a macro-economic level. The
interaction of firm-level process with aggregate economic performance has
not been adequately studied by evolutionary economists. Dopfer (2001)
aptly describes the nested hierarchy and systemic ‘complexification’ from
the ‘subject level’ of a firm, to the ‘plurality level’ of many sector popula-
tions of firms at the economy level.
(Positive) Feedback and Externalities
The process of creative destruction is most effective for economic develop-

ment and growth when it allows the exploitation of positive externalities,
involving increasing returns, agglomeration effects, and non-linearity or
network effects. Increasing returns are prominent in knowledge-intensive
sectors, but not only there. The winner-take-all literature relates to this,
when it shows the urgency for market players to corner market-share for
dominance through monopolistic competition. Agglomeration effects are
prominent in spatial models, where productivity increases relate to a spatial
convergence of resources in a critical mass. Overcrowding in mental or
physical space can however lead to negative agglomeration and environ-
mental effects, as many inhabitants of ugly, polluted megalopolises can
attest. Network effects play a prominent role for entrepreneurs who
produce network goods, where the utility to consumers rises with the
increased and more geographically widespread use of the product, or by
services delivered through an infrastructure.
Non-average Behaviour by Economic Agents
Firms exist of course not in the abstract but are real, run by entrepreneurs
and helped by people. Variety, for example reflected in differences in
product and service-unit costs, is constantly recreated in an evolving

economy at the subject or firm level. However to evolve, modern models of
circular and cumulative causation define the macro-environment in terms
of frequency of adoptions of a product and service by a sector population
(Dopfer, 2001, p. 34). A change in relative frequencies of varieties present
in a population comes about through frequency-dependency effects.
Members of a firm population exert an effect on each other (Witt, 2001a,
p. 55). Firm behaviour is correlated, and even more so in geography when
firms which are physically close are more likely to adapt to and learn from
each other’s behaviour. In that process, variety is reduced as at the subject
level, firms want to imitate best practice, if they learn of such practice.
(Tacit) information transfer is incomplete, and levels of know-how and
capabilities of firms to innovate and imitate are inherently different. In the
market, a selection process works on firm populations. Firms with better
products and services (unit price per quality) win market share at the cost
of others, and some firms are driven out of business, or at times new firms
enter with new knowledge, translated into products and services. The cir-
cular and cumulative interplay of creation of non-average behaviour, with
co-ordination when learning and imitation come into play, and with market
selection, provides the engine of transformation. The processes in their
interplay require the theoretical link of micro- with macrolevel analysis in
a nested hierarchy, as variety and its frequency distribution is a concept at
the macro level, and regeneration of variety is one at the micro level.
People learn and carry know-how, and are motivated to innovate. To do
so they converge on a distinct place in time and on bricks-and-mortar
facilities. Firm-level heterogeneity in plant-level productivity and com-
petitiveness is in good part driven by firm-specific locational and dist-
urbance factors (Haltiwanger, 2000). Firm facilities are served by
infrastructure; for example, roads connect to the factory and help move
people and goods in and out. It is an important question why firms locate
and invest in space when they are supposed to attract the most productive
resources, and combine them under management for a competitive service
or product to be sold in the domestic and foreign market. Surely firms will
locate only in places where they have cost-effective access to resources,
markets, know-how and so on. Governments are often in the business of
providing the public goods of access and location, learning, and know-
how to make sure the process of creative destruction is tilted towards the
creative side. Generally, research shows that firms locate most produc-
tively where they have the best access to inputs and to markets on a per
quality adjusted service or product unit/kilometre measure related
to time. This depends also on the institutional arrangements for
exchange and regional or international trade. One could of course say
that knowledge-intensive services and products are less geography- and

infrastructure-dependent. Still they are even more people-dependent, and
qualified people want to live and work in an attractive location. Also,
high-tech products tend to be more time-sensitive in delivery to cus-
tomers, and even if roads may not be necessary for their transport, these
services or products would have to be moved out and around, and across
borders by electronic means or by air.
Development-policy Responses
Here is a brief summary of a partial application of the ideas just described

as this chapter does not intend to present the complete evolutionary math-
ematical model (for the model refer to Brunner and Allen, 2005). The ques-
tion that was addressed by this ADB work concerned the probable impacts
of a large transportation infrastructure project for West Bengal. Transport
consultants gathered detailed data concerning the flows of goods through
the region, and explored the possible changes in costs that might result
from the proposed infrastructure development. This relatively complex
model is driven again by the change in transaction or transport costs, as
was the ‘simple’ model in the case of Nepal’s financial sector. Further, the
model used here set out to explore the spatial economic impacts of these
changes. The transportation consultant supplied the flows of goods and
commodities between 46 economic zones (i, j . . .), along the transport-
infrastructure grid of West Bengal. This was provided in the form of a
matrix of origin and destination flows by volume and by value, for 24
different sectors.
The provision of these data concerning flows and costs for the region
allowed a much simpler version of a general model described in Allen and
Brunner (2001). In fact the key piece of information that is needed to drive
the model is the actual cost of transporting goods between the zones, and
the manner in which this will be changed by any particular transportation
investment project. It is these changes in cost that will affect the prices
charged at the points of consumption, and will therefore confer an eco-
nomic benefit on consumers. This can then be run ‘through’ the simplified
model, demonstrating where the multipliers will lead to increased produc-
tion, and in turn how this will create employment and increased incomes.
See Figure 14.1.
From the view of spatial structure, variables in the modelling framework
are disaggregated into those of different regions and sectors. The spatial
structure that exists is generated by the interactions among regions
and sectors. The interactions in spatial structure strengthen their non-
linearities and feedback features. Therefore, the spatial structure of the
Potential
degradation
Spatial pattern of
natural resources
Resource Comparative
use advantage
Comparative Resources
advantage use
Market and costs

Spatial pattern of Spatial pattern of
population economic activity
Migration attraction
Population Investment
growth
Figure 14.1 Basic modelling framework
system is not taken as being fixed necessarily, but instead as being the result
of an ongoing evolutionary process. Most importantly, it introduces three
behaviours through the use of ‘attractiveness’ which are the demand attrac-
tiveness, migration attractiveness and investment attractiveness. These are
the mechanisms by which small differences in price, incentives and pro-
ductivity can be amplified through self-reinforcing processes, eventually
leading to a spatial restructuring of the system. Economic incentives and
investment attractiveness create the opportunity of employment, increase
the vacancies and raise labour productivity, leading to increases in employ-
ment levels and wages. The improved income-earning opportunity will
alter relative economic conditions among the regions and increase migra-
tion of the workforce, which in turn will alter relative investment behaviour
of sectors.
The model considers how the lower prices of delivery result in con-
sumers spending their gains on extra consumption that, in turn, generates
more jobs. The model therefore responds to the reduced prices by examin-
ing how much the different income classes are affected and how they may
choose to spend the resulting income. Savings will affect income groups
differently, as those that spend more save more in absolute terms, but the
poor will be able to buy more food, and also if jobs available in the indus-
trial and service sectors increase, then this will also increase the income of
some of the poor.
The model uses three sectors (k) for its operation: agriculture, industry
and services. One of the first steps therefore was to aggregate the data
concerning transport into these categories.
The flows of goods are generated by the fact that the spatial distribution
of demand does not coincide with that of supply. For example, the city of
Calcutta clearly requires large quantities of food, but has no agricultural
land available to produce it. The cost of transportation, however, acts as a
‘net loss’ to the ‘utility’ of consumers, since they must pay for it, but only
because they want the goods that it brings.
The model allows us to follow the manner in which the increased demand
from each point falls onto the different zones of production and leads to
the possibility of increased production. This in turn requires jobs to be
created and filled, which means that people must be available to fill them.
The demographic increase in each zone provides a growing number of
people seeking jobs, but this may be insufficient in some zones, and too
great in others. The actual growth of active, filled jobs therefore reflects the
changing distribution of population between the zones, and the migratory
flows that tend to reflect the job opportunities. Moreover, different wage
levels available in the sectors will also allow growth in a higher-paying
sector at the expense of a lower-paying one. Another factor that is import-
ant is the availability of land for economic activities. For example, clearly
there is a very large demand for agricultural products in Calcutta, but there
is no space to make fields and grow crops and so instead, jobs are not
created and filled (despite the demand) and food is transported into the city.
The model takes into account these different factors, and calculates
where economic growth will occur, by sector and zone. This allows us to
track the changes in employment for the 33 different spatial zones in West
Bengal, and to use this as a base for a comparison with the changes that
occur when different transport infrastructure projects are implemented.
The model runs these mechanisms over the 33 different spatial zones, and
also for the different sectors of the economy. Population is attracted to
places of perceived employment opportunity and this allows vacancies to
be transformed into jobs. This increases the local income, and hence the
demand emanating from the zone.
The non-linear dynamic model will generate changes resulting from the
reduced prices of goods in different zones, adjusting supply and demand
arriving at each zone to equality, and changing the distribution of jobs to
reflect the relative locations of the demand and of transportation costs. By
running the model with no modifications in the transportation-matrix
costs, we can generate a ‘base scenario’. This can then be compared with a
run in which the costs of transportation have been changed as a result of
some investment in an infrastructure project.
The model can then compare various differences between the two simu-
lations. First, there is the amount of ‘extra income’ available for spending
in any particular zone. This can further be broken down into that arising
for the poor, medium and rich income-groups. In addition however, we can
see the difference in the distribution of jobs over time. This changes the
income distribution within each zone and changes the proportions of the
different income groups. It also changes the overall income generated in a
zone, and hence contributes to the benefits.
The data supplied were the comparative transportation costs in 2015,
with or without the project, and again in 2025 with and without the
project. The ‘with or without’ differences reveal the spatial distribution of
consequences of the project, the impacts on employment, and on the
different income groups of the zone. For 2015, the overall gain is some
40 000 jobs. These extra jobs can immediately translate into a gain for the
poor of each zone. This is because these are ‘new jobs’ in addition to those
that would have existed without the infrastructure investment. For deci-
sion-making purposes, when we take out these results from the evolution-
ary model into the simple conceptual frame of a cost-benefit analysis, we
can find a difference in results resulting from the use of a more complex
model. If we multiply the 40 000 by the average wages per job, then we find
about 600 million Rs per year. In addition to this, there are the effects on
external trade, and also the savings in consumption costs as a result of the
cheaper transportation. These two effects add up to a nominal consumer
surplus for 2015 of about 19 billion Rs. For 2025, the extra jobs generated
are estimated at about 117 000, which are worth about 1.7 billion Rs. The
consumer surplus attributed to investment multiplies to 50 billion Rs. by
2025.
Interpolating the continuing growth of the investment benefits between
the first investment year 2002, and 2025 (a gestation period of some 20 to
30 years is normal in large infrastructure projects), one gets a cost/benefit
stream as shown in Table 14.1. The stream is discounted at 12 per cent, and
results in a net present value (NPV) of about 28.9 billion Rs. In the table,
I have separated out the benefit stream calculated for the poorer section of
West Bengal’s population. As a percentage, a share of 21.5 per cent of the
overall benefit stream accrues to the poor.
What difference does the evolutionary approach employed make to the
outcome? We would know, had a conventional cost-benefit model been
used in parallel. In fact, the transportation consultant used a conventional
linear trade, transport and logistics model, as outlined in Figure 14.2.
Linear traffic forecasts induce transport-cost savings (consumer surplus).
In addition, a reduction in transport logistics costs induces demand for
both domestically consumed and internationally traded goods. The latter
Table 14.1 Corridor evaluation level
Corridor evaluation level Rs million in 2000 prices

Year Cost/benefit Benefits to poor
2002 3500
2003 4000
2004 3500
2005 3500
2006 1700 357
2007 2140 449
2008 2690 565
2009 3390 712
2010 4270 897
2011 5370 1128
2012 6770 1422
2013 8520 1789
2014 10 720 2251
2015 13 500 2835
2016 14 573 3060
2017 15 731 3303
2018 16 981 3566
2019 18 330 3849
2020 19 786 4155
2021 21 359 4485
2022 23 056 4842
2023 24 888 5226
2024 26 865 5642
2025 29 000 6090
NPV 28 907
Sum of benefits 298 544 56 624
%poor of benefits 21.5%
EIRR 25.07% ($0.15)
results were obtained by simply using price elasticities of demand. Since

this linear model relates trade generated to single (not multiple) geographic
zones and thus to individual investment projects only, it did not estimate
the cumulative effects as well as the multiplier effects that result from a
series of investments along a trade corridor. All this model generates is
first-round cost (time) savings and demand changes.
The calculation of the benefits related to the individual investment
components yield, as is shown in Table 14.2, considerably lower total ben-
efits. These are on aggregate still good investment projects. However a
Logistics chain
performance
Logistics chain exports, base –
performance (Rs, duration)
domestic, base – – Stuffing/de-stuffing
(Rs, duration) – Waiting
– Loading/unloading – Port handling costs
– Transporting – Shipping costs
– Modal seams – Agency costs
– Damaged/stolen goods – Packing costs
Transport
infrastructure,
operational, procedural
improvements
Logistics cost
Logistics cost
savings – domestic
savings – exports
movements
Forecast Reduction in price of goods

commodity flows – Port forecasts –
domestic – base base
– by commodity – by commodity
– tons – tons
– value – value
Price elasticities Price elasticities

of demand – of demand –
Domestic Export
by commodity by commodity
Change in Change in port

commodity flows – flows – options
domestic – options – by commodity
– by commodity – tons
– tons – value
– value
Figure 14.2 Estimating trade impacts from corridor improvements

Table 14.2 Individual project evaluation level (on constant annuity

payments, scenario A)
Rs. Million at 2000 prices

Component Cost NPV EIRR
Haldia-Uluberia 5200 8000 27.9
Uluberia- Joka 7500 2458 6.2
Raiganj-Islampur 2000 2600 25.8
Shantipur 800 1800 30.7
Sum 15 500 9942
decision maker using the output of these individual projects would con-
clude that it was wise to not invest in the lowest-return project, so as not
to lower the average return on investment. This however has the effect that
one corner-stone project along the entire transport and trade chain is
eliminated, with dramatic effects, as the largest part of the multiplier gains
are lost. The evolutionary model did take into account the multiplier
gains that accrue to the whole system by combining the four investments
in one model, and achieving a combined EIRR on combined investment
of some 20 billion Rs. This is an extra NPV of $500 million (which is
equivalent to $5 billion in nominal benefits, with an estimated 21 per cent
going to the poor)!
As significant as the differences between the use of an evolutionary and
a neoclassical approach turned out to be, it should be noted here that the
evolutionary model was much less of an evolutionary model than would be
ideally wanted. Although the model used increasing returns from invest-
ment, and was based on geography and positive agglomeration and
network effects, it fell short in not modelling behavioural differences among
actors, such as employers along the trade corridor. The model lacks in
reality because it cannot generate learning among actors, and it does not
allow for qualitative systems-change based on actors’ creativity and on a
selection process taking hold. Such a model extension has been proposed
by Allen and Brunner (2001) and Brunner and Allen (2005). Would such
an extension be worthwhile, given the already large gain from using a
simpler non-linear dynamics model? Let us now discuss the issue of geo-
graphy and modelling. How close to reality can (or should) we get without
getting lost in the complexities?
4. ISSUE IN EVOLUTIONARY ECONOMICS:

GEOGRAPHY AND MODELLING – HOW CLOSE
TO REALITY CAN WE GET WITHOUT GETTING
LOST IN THE COMPLEXITIES?
Understanding reality and gaining knowledge about a problem require us

to reduce the real complexity of any particular situation to a simpler, more
understandable system by making specific simplifying assumptions or by
undertaking institutional analysis that reifies many acts of behaviour into
non-random patterns. The hope is that there exists a representation that,
while being sufficiently simple to be understood, remains sufficiently repre-
sentative of reality to be useful. These assumptions and reifications have
been discussed elsewhere (Allen, 1988, 1993, 1994, 1997, 1998) and we shall
briefly summarise them here in Figure 14.3.
We use assumptions, simplifications, and reifications in order to reduce
social and economic complexity to manageable proportions. These are:
1. That we can define a boundary between the part of the world that we
want to ‘understand’ and the rest. In other words, we assume first that
there is a ‘System’ and an ‘Environment’.
2. That we have rules for the classification of objects that lead to a rele-
vant taxonomy for the system components, which will enable us to
understand what is going on. This is often decided entirely intuitively.
3. The third assumption concerns the level of description below that
which we are trying to understand, and assumes all phenomena are
either all identical to each other and to the average, or have a diversity
that is at all times distributed ‘normally’ around the average.
4. That the individual behaviour of sub-components can be described by
average interaction parameters.
The crucial difference that marks out evolutionary (complex) systems con-
cerns assumption 3. This assumes that all firms of a given type, ‘x’ say, are
either identical, equal to the average type; or have a diversity that remains
normally distributed around the average type. This means that for such
models ‘sectors’ or statistical aggregates of ‘firms’ are used as variables
reflecting the ‘average properties’ of the included firms. But in reality, the
microdiversity of behaviours among individual firms in any particular
part of the system is the result of the history of local dynamics occurring
in the system. This local history concerns the local mechanisms by which
knowledge, skills, techniques and heuristics are passed on to new individ-
uals over time. The existence of microdiversity will necessarily lead to an
evolutionary learning process of economic development. If there are firms
Complexity Successive assumptions Simplicity
Classification Average types Average events Attractors

boundary
Reality X X
Z Z
Y Y
Evolutionary models Self-organisation Non-linear dynamics

Autopoiesis
344
Changing taxonomy Catalytic sets Stability, resilience,
Learning models… bifurcations,
Creativity + Selection Selection attractors, chaos…
Soft systems
Literature, history, The system just
descriptions… Systems change RUNS. But the
qualitatively... modeller can
Events, and Processes make experiments
Systems diverge…
Figure 14.3 Assumptions made in reducing complexity to simplicity, in order to obtain understanding, and
predictability – providing the assumptions hold
with higher and lower productivity then the higher ones will be gradually
amplified, while those with lower productivity are suppressed, and the
‘average’ for the whole population of firms increases productivity (Allen
and McGlade, 1987; Brunner, 1994; Caballero and Hammour, 2000). This
is the mechanism by which adaptation takes place. This demonstrates the
vital part played by exploratory, non-average behaviour, and shows that,
in the long term, evolution selects for firms with the ability to learn, rather
than for populations with optimal, but fixed, behaviour. Economic trans-
formation therefore should really be modelled by these fully evolutionary
models.
As mentioned before, models are always, and to a degree, partial repre-
sentations of the cloud of reality they depict. Models should be detailed
enough to capture all the important causal real-world relationships that
explain outcomes of development. Evolutionary complexity, however,
reduces predictability as non-linear causal relationships, externalities,
network effects, and geography lead to multiple equilibria. Outcomes are
path-dependent, in part matter of chance. In arguing with Krugman (1997)
on the utility of complex models, one needs to show that complex and real-
istic approaches yield results which formalised models in a simple setting
cannot. The West Bengal example shows that models that are complex, and
even those that may take into account heterogeneity of actors, can be tested
for results. Yet even simple models as applied to the analysis of the finan-
cial sector in Nepal, once enriched with institutional concepts such as
transaction costs, yield significant power to make a big difference for eco-
nomic development in Asia and in the effectiveness of resource use by the
ADB and its borrowers.
5. EVOLUTIONARY AND INSTITUTIONAL

ECONOMICS AS A HOUSEHOLD TOOL IN
MULTILATERAL DEVELOPMENT-AGENCY
ECONOMICS
Models have to incorporate sufficiently realistic assumptions in order to

be reliable and useful. The parsimonious rendering of a society or
economy in terms of a set of reified institutions must render the essential
features of ongoing interactions of thousands of disparate individual
actors, be they consumers, farmers, or firms. Otherwise one falls into the
trap of the drunkard who looks for his watch under the street-lamp,
because that is where there is light. While simple non-linear, dynamic
models may appear to predict perfectly the future path of the system, this
seemingly solid piece of knowledge is only as good as the assumptions
that underlie it (Allen, 1999). Making an unrealistic prediction may actu-

ally be misleading, and thus much worse than knowing outcomes only
with some probability. Knowledge is really about knowing what simplifi-
cations can be made without compromising on results. It is also about
knowing the limits to knowledge. Complex, self-organising systems, and
the human society and economy, can only be grasped up to a point, but
we are fortunate that in most cases, whether we are farmers, entrepreneurs,
policy makers, or international development economists the artificial con-
structions of models embedded in institutional systems are sufficient to
permit self-correcting policy recommendations, including on specific
projects.
With these caveats in mind, the chapter shows an evolutionary approach,
if widely used, can make billions of dollars of difference in incomes for the
60 per cent of the world’s poor who reside in Asia. The case studies in Nepal
and West Bengal show very large incremental impacts of well-conceived
projects on such key indicators as per-capita incomes, employment, and
poverty reduction. Decisions are being based sometimes on one approach
or another, and sometimes on a combination of approaches. The choice of
analytical method and of concepts not only helps in selecting the best use
of funds by international agencies, but facilitates institutional designs more
likely to lead to desirable outcomes. The evolutionary-economics profes-
sion should stand by its theoretical models and by the concepts derived
from them, and get involved in the practice of these models. The practice
requires us at times to couch institutional and evolutionary concepts in
simple, even neoclassical terms for a decision maker, as the two ADB cases
illustrate.
What practical steps would make use of institutional analysis and evo-
lutionary modelling as methods more common in the development profes-
sion, and hence more familiar to development policy makers? There are
several concrete measures. The first is to propose appropriate measures of
economic and social targets. Traditionally, the construction of economic
aggregates has fallen short of the measures appropriate from an evolution-
ary and institutional perspective. This situation serves as an obstacle to
applied work. Hard work needs to go into the collection of appropriate
data and information that make the parameters and variables used in evo-
lutionary models measurable and communicable. It may be necessary to use
fieldwork to compensate for the absence of relevant data. In the Nepal and
Bengal projects, teams spent many person-days in the field gathering base-
line information on such normally unmeasured variables as transaction
costs, prevalence of corruption, the implicit cost burdens on borrowers,
actual transport costs, and so forth.
Second, we need to look at the regional distribution and population dis-

tribution of economic effects. The spatial and social dimensions of projects
are rarely estimated. The frictions associated with transaction costs and
information costs have a strong impact on locational and distributional
outcomes. Quality needs to be considered alongside quantity. With respect
to financial services, this means lowering transaction and information
costs for both lenders and borrowers. Selection processes for firms would
be modelled on quality unit cost as the selection criterion. Population-
model parameters would be calibrated on sector entry- and exit-rates,
concentration-ratios, and changes in firm concentration. Learning para-
meters and variables would be incorporated on the basis of entrepreneurial
response and adjustment rates.
Third, there is a need for expanded training of the staff of development
agencies and client governments. For example, because of budget con-
straints, United States Agency for International Development (USAID)
economists working in overseas posts have not had in-service training for
over a decade. In 2000, a workshop was conducted to update economic-
growth officers on the approaches of institutional and evolutionary eco-
nomics (USAID 2000). Economists can engage in co-operative, network
arrangements within the academic community to connect with practition-
ers. Very few development practitioners take the time, or have the budget and
incentives, to participate in academically oriented conferences. Some may be
discouraged by the lack of relevance of much academic dialogue. Thought
needs to be devoted to building bridges with practitioners. The evolutionary-
economics associations throughout the world can play a role in facilitating
cross-fertilisation. Exchange of knowledge, diffusion of empirical work, and
concentration of effort in tandem with practitioners will help remove the
obstacles to applied work which exist in the academic community.
Fourth, evolutionary and institutional economists should make strong
efforts to connect to development banks and to the consulting firms that
provide expertise for their projects. Most institutionalists are well prepared
to provide evolutionary types of analysis for the practitioners in the multi-
lateral agencies and to liaise with the financial, development and business
communities in overseas assignments.
NOTES
* The author alone in his personal capacity, not the Asian Development Bank, is responsible
for the article’s content. The author thanks the participants of the Wartensee seminar and
is particularly grateful to Kurt Dopfer and John Adams as well as an anonymous referee
for their helpful comments. The usual disclaimer applies.
REFERENCES
Adams, John (1993), ‘Institutions and economic development: structure, process,

and incentive’, in Marc R. Tool (ed.), Institutional Economics: Theory, Method,
Policy, Boston, MA: Kluwer Academic, pp. 245–69.
Adams, John and Hans-Peter Brunner for Asian Development Bank (1997), Project
Performance Audit: Fifth Agricultural Credit in Nepal, ADB, PPAR, NEP 181054.
Adams, John, Hans-Peter Brunner and Frank Raymond (2003), ‘Interactions of
informal and formal agents in South Asian rural credit markets: an analysis of
information asymmetries and adverse selection with heterogeneous credit ser-
vices’, Review of Development Economics, 7 (3) (August), 431–44.
Allen, Peter (1988), ‘Evolution: why the whole is greater than the sum of its parts’,
in W. Wolff, C.J. Soeder and F.R. Drepper (eds), Ecodynamics, Berlin: Springer.
Allen, Peter (1993), ‘Evolution: persistent ignorance from continual learning’, in
R. Day and P. Chen (eds), Nonlinear Dynamics and Evolutionary Economics,
Oxford: Oxford University Press, Chapter III, 8, pp.101–12.
Allen, Peter (1994), ‘Evolutionary complex systems: models of technology change’,
in Loet Leydesdorff and Peter Van den Besselaar (eds), Evolutionary Economics
and Chaos Theory, London: Pinter.
Allen, Peter (1997), Cities and Regions as Self-Organising Systems: Models of
Complexity, London: Gordon and Breach.
Allen, Peter (1998), ‘Selbstorganisation’, in F. Scheiter and G. Silverberg (eds),
Modelling Complex Economic Evolution, Berlin: Duncker & Humblot.
Allen, Peter (1999), ‘Knowledge, ignorance and the evolution of complex systems’,
paper prepared for ‘Evolutionary Economics’ Meeting, Brisbane.
Allen, Peter (2001), ‘Knowledge ignorance and the evolution of complex systems’,
in J. Foster and S. Metcalfe (eds), Frontiers of Evolutionary Economics:
Competition, Self-Organisation and Innovation Policy, Cheltenham UK and
Allen, Peter and Hans-Peter Brunner (2001), ‘A sub-regional development
approach – transport, international trade, and investment modeled in space’,
electronic copy on SSRN http://papers.ssrn.com/paper.taf ?abstract_id258429
Allen, Peter and J.M. McGlade (1987), ‘Evolutionary drive: the effect of micro-
scopic diversity, error making and noise’, Foundations of Physics, 17 (7) (July),
723–8.
Beck, T. (1997), ‘Banking at the margin: why do formal and informal intermediaries
coexist?’, working paper, University of Virginia.
Bhattarai, Keshab (1999), ‘Welfare and distributional impacts of financial liberal-
ization in Nepal’, working paper, University of Warwick.
Boulding, Kenneth (1991), ‘What is evolutionary economics?’, Journal of Evolu-
tionary Economics, 1 (1), 9–17.
Broecker, T. (1990), ‘Credit-worthiness tests and interbank competition’, Econom-
etrica, 58 (2), 429–52.
Brunner, Hans-Peter (1994), ‘Technological Diversity, Random Selection in a
Population of Firms, and Technological Institutions of Government’, in Loet
Leydesdorff and Peter Van den Besselaar (eds), Evolutionary Economics and
Chaos Theory, London: Pinter.
Brunner, Hans-Peter (2001), ‘East German international trade during transition as
reflection of the dynamics of competitiveness,’ Eastern Economic Journal, 27 (3)
(Summer), 287–308.
Brunner, Hans-Peter and Peter M. Allen (2005), Productivity, Competitiveness and

Incomes in Asia: An Evolutionary Theory of International Trade, Cheltenham,
UK and Northampton, MA, USA. Edward Elgar.
Caballero, Ricardo and Mohammad Hammour (2000), ‘Institutions, restructuring,
and macroeconomic performance’, NBER Working Paper 7720, May.
Dell’Ariccia, G. (1998), ‘Asymmetric information and the market structure of the
banking industry’, Working Paper WP/98/92, International Monetary Fund.
Dopfer, Kurt (2001), ‘Evolutionary Economics: Framework for Analysis’, in Kurt
Dopfer (ed.), Evolutionary Economics – Program and Scope, Boston, MA and
Dordrecht: Kluwer Academic.
Dosi, Giovanni (1991), ‘Some thoughts on the promises, challenges and dangers of
an “evolutionary perspective” in economics’, Journal of Evolutionary Economics,
1 (1), 5–7.
Eliasson, Gunnar (1991), ‘Deregulation, innovative entry and structural diversity
as a source of stable and rapid economic growth’, Journal of Evolutionary
Economics, 1 (1), 49–63.
Gupta, M. and S. Chaudhuri (1998), ‘Formal credit, corruption and the informal
credit market in agriculture: a theoretical analysis’, Economica, 65, 100–111.
Haltiwanger, John (2000), ‘Aggregate growth: what have we learned from micro-
economic evidence?’, working paper, University of Maryland.
Hannan, T. (1997), ‘Foundations of the structure-conduct-performance paradigm
in banking’, Journal of Money, Credit and Banking, 23 (1), 68–84.
Kornai, Janos (2000), ‘Ten years after “The Road to a Free Economy”: the author’s
self-evaluation’, paper for the World Bank ‘Annual Bank Conference on
Development Economics’, Washington, DC, April.
Krugman, Paul (1997), Development, Geography and Economic Theory, Cambridge,
MA: MIT Press.
Levine, Ross (1997), ‘Financial development and economic growth: views and
agenda’, Journal of Economic Literature, 35 (2) (June), 688–726.
Morduch, J. (1999), ‘The microfinance promise’, Journal of Economic Literature,
37, 1569–1614.
Murrell, Peter (1990), The Nature of Socialist Economies, Princeton, NJ: Princeton
University Press.
Silverberg, Gerald (1994), ‘Evolution, Formal Models of Economic’, in Geoffrey
Hodgson, Warren Samuels and Marc Tool (eds), The Elgar Companion to
Institutional and Evolutionary Economics, two Vols, Aldershot, UK and
Brookfield, US: Edward Elgar.
Stiglitz, J. and A. Weiss (1981), ‘Credit rationing in markets with imperfect infor-
mation’, American Economic Review, 71, 393–410.
United States Agency for International Development (USAID) (2000), Economic
Growth Officers Training Workshop, folio, Charlottesville, VA, 26 November–1
December.
Witt, Ulrich (1992), ‘Evolutionary concepts in economics,’ Eastern Economic
Journal, 18 (4), 405–19.
Witt, Ulrich (2001a), ‘Evolutionary economics: an interpretative survey’, in Kurt
Dopfer (ed.), Evolutionary Economics – Program and Scope, Boston, MA and
Dordrecht: Kluwer Academic.
Witt, Ulrich (2001b), ‘The evolutionary perspective on economic policy making –
does it make a difference?’, Papers on Economics and Evolution no. 0101, Max-
Planck-Institute, Jena.
Index
active economic policy Atkinson, A. 37
commitment by EU 66 Ausschuss für Evolutorische
definitions 66 Ökonomik 2
as distinct from adaptive 65–6
pitfalls of 66–9 Bannock, G. 240
Adams, John 325, 326, 329, 330 Barney, Jay 162
adaptive economic policy Barzel, Yoram 162
definitions 69 Baum, L. 319
and institution revision 71–3 Bebchuk, L.A. 319
selection mechanism for novelties Beck, T. 330
70–71 Becker-Berke, S. 274, 275, 280, 281,
adaptive Pareto-optimal, definition 54 287
adaptive preferences Becker, G.S. 93
and evolutionary biology 52–4 Becker, J. 284
formal model 47–9 Behrens, P. 319
main theorem 49–51 Bennett, W.L. 178
and Pareto-optimality 54–5 Berg, J. 249
path dependence 55–6 Berger, P.L. 172
while maintaining normative Bernal, J.D. 242
individualism 46–7 Bernard, J. 37
ADB (Asian Development Bank) 326 Bernholz, Peter 235
Adler, A. 282 Bernstein, Jeffrey R. 141
Aghion, P. 208, 209 Berry, F.S. 319
Ahrens, H.J. 287, 291 Berry, W.D. 319
Albert, H. 75, 168, 185, 186 Bertram, B. 274
Albrecht, E. 259, 269 Besley, T. 318, 319
Alchian, A.A. 88 Bhattarai, Keshab 331
Allen, Peter 326, 336, 342, 343, 345, Bibliography of German Periodicals
346 Literature 250
Altmeppen, K.-D. 177 biotechnology/biotechnology-based
Amin, M. 249, 251 sectors 99–100
Anderlini, Luca 162 Blaug, M. 81
Andersen, H. 280 Blondel, J. 186
Anderson, James E. 143 Bodenstein, M. 319
Anderson, P. 105 Bogatti, S.P. 115
Antonelli, C. 37 Böhm, K. 38
AOK (Allgemeine Ortskrankenkasse) Bratton, W.W. 318
275 Braunerhjelm, Pontus 143
Ariccia, G. dell’ 330 Brennan, G. 318
Arrow, K.J. 57, 64, 75 Brenner, T. 75
Arthur, W.B. 207 Breton, A. 303, 309, 318
asymmetry, concept of 16–17 Breyer, Friedrich 235
351
352 Index
Broecker, T. 330 competition

Brown, L.D. 319 Coase’s views on 80–82
Brunner, Hans-Peter 326, 330, 334, among governments see
336, 342, 345 interjurisdictional/regulatory
Buchanan, James M. 57, 149, 318 competition, theories of
Budzinski, O. 318, 319 competition processes, modelling
Burt, R.S. 115 110–11
competitive advantage
Caballero, Ricardo 327, 328, 345 and non-tradability 147–50
Camesasca, P.D. 319 restatement of theory with non-
Campbell, D. 79 tradable factors 153–9
Canon, B.C. 319 competitive federalism 312–14
Cantner, Uwe 28, 30, 36, 37, 109, 194, contingency, definitions 217–18
207, 235 Cooper, W.W. 26
capitalism, and technological change Cordato, R.E. 60
192 Coughlin, B.C. 90
Carlaw, Kenneth 235 creative destruction 333–4
Case, A. 319
Cassis de Dijon judgement (European Dasgupta, Partha 145
Court of Justice) 298, 301 Davis, Donald R. 141, 142
Casson, Mark 151 DBFs (Dedicated Biotechnology
Caves, D.W. 32 Firms)
Caves, Richard E. 163 definitions 100, 101
centralised planning, Coase’s views on modelling 103–7
79–84 DEA (data envelopment analysis)
Chang, K.-P. 37 approach for TFP 24–31
Charnes, A. 26 Deardorff, Alan V. 142
Chaudhuri, S. 330 Dell’Ariccia, G. 330
Christensen, L.R. 32 demand, and adaptive preferences 52–4
Ciompi, L. 182 Demsetz, H. 302
city evolution, sociodynamic model of development-agency economics, and
130–39 evolutionary and institutional
Clark, J. 319 economics 345–7
Clark, Terry 142 developmental approach, and
cliometrics, definitions 233 heterogeneity 15
Coase, R.H. 78–93, 192 diabetes 276, 282
and evolutionary economics 87–92 diabetes contracts (Germany)
historical method 84–7 characteristics 276–7
The nature of the firm (NT) 79–82, continuation 289–91
192 diffusion 285–9
The problem of social cost (SC) 83–4 disastrous nature of 287–9
Coate, S. 318 emergence 280–85
cognition 170–72 overview 274–5, 291–2
cognitive maps 172–4 regulation 277–80
Cohen, W.M. 109 transfer fees 276–7
collective human capital, as dimension Diewert, W.E. 32
of entrepreneurial competence Dijk, T.A. van 186
157 disastrous innovations 287
Collins, William J. 143 see also diabetes contracts
Commons, J.R. 82 (Germany)
Index 353
Dohrn-van Rossum, G. 239 competitive federalism instead of

Dopfer, Kurt 89, 90, 334, 335 centralization 314–16
Dörner, D. 184 and neoclassical theories of
Dosi, Giovanni 16, 37, 140, 209 federalism 299–301
Dunn, M.H. 207 see also interjurisdictional/regulatory
Dunning, John H. 141 competition, theories of
Evo-Devo biology 5, 90–91
East Germany see German Democratic evolutionary economic governance 1
Republic evolutionary economic policy 1, 296–7,
East Prussia 243 316–18
see also German national innovation evolutionary economics
system characteristics 230–31
Easterbrook, F.H. 319 exploring political consequences
economic agents, non-average 58–60
behaviour 334–6 and welfare 62–3
economic development, evolutionary evolutionary political economy,
approach 325–6 scarcity of research 59–60
economic evolution, and heterogeneity evolutionary trade theory (ETT),
15 difference from NTT 159–61
economic policy evolutionary welfare economics 44, 49,
cognition and 170–72 56
criticisms of orthodox Expert Group for the Study of
understanding of 168–9 Evolutionary Economics 2
individual understanding and 172–4 explorers (role of DBFs), definitions
problem diffusion 174–9 102–3
problem treatment 179–84 see also DBFs
Edingshaus, A.-L. 186 Eysenck, M.W. 173
Eichenberger, R. 179, 181, 186, 318,
319 factor, concept of, new definition of
Ekholm, Karolina 143, 144 153–5
empirical analysis 35–7 factor-proportions theory, problems
see also TFP with 141–2
endogenous growth theories 190–91, Färe, R. 32
205–7 Federal Republic of Germany (West
endogenous technological change Germany)
model see Romer, P.M., growth citation levels compared to GDR
model 253
Engel, Charles 142 patenting activity compared to GDR
Engerer, H. 71 259
entrepreneurial competence, R&D expenditure compared to
dimensions of 155–9 GDR 248
Erker, P. 258 SCI publications compared to GDR
Esty, D.C. 318 252
ETT see evolutionary trade theory sociodynamic migration model 130
Eucken, W. 317 federalism see competitive federalism;
European Union interjurisdictional/regulatory
centralization vs decentralization competition, theories of
policy problem 297–9, 311, 313–14 Feenstra, Robert C. 159
commitment to active economic Fehl, U. 209
policy 66 Feld, L. 301, 318
354 Index
Felli, Leonardo 162 sectoral level 260–69

fiscal federalism see competitive territorial analysis issues 243
federalism; interjurisdictional/ Germany see Federal Republic of
regulatory competition, theories Germany
of Gerschenkron, A. 271
Fischel, D.R. 319 ghetto formation, sociodynamic model
fixed preferences, as anti-evolutionary of 127
45–6 Goffman, E. 187
Fogel, R.W. 240 Goldberg, Pinelopi Koujianou 142
Fornahl, D. 75 Goodman, C.S. 90
Förster, H. von 168 governments, competition among see
Foss, Nicolai J. 78, 87, 93, 146, 157, 162 interjurisdictional/regulatory
Fowler, R.F. 92 competition, theories of
Fransman, M. 91 Grabher, Gernot 157
Freeman, C. 209, 242 Grabowski, H. 102
Freeman, L.C. 115 Grant, Robert M. 162
Frey, B.S. 169, 179, 181, 183, 186, 318, gravity models, NTT problems with
319 141–2
Fritsch, Michael 235 Gray, V. 319
Fulda, Ekkehard 235 Greenhut, Melvin L. 162
Furubotn, E. 78, 192 Grohmann, H. 271
Grosskopf, S. 32
Garegnani, P. 203, 209 Grossman, Gene M. 161, 162, 208,
Gascoigne, R. 243 209
Gatsios, K. 318 growth paths of transition economies
GDR see German Democratic 225–9
Republic growth theories see endogenous growth
geography, and modelling 343–5 theories
Georghiou, Luke 235, 236 Grubel, H. 209
Geradin, D. 318 Grundmann, S. 319
German Democratic Republic (GDR, Grupp, H. 191, 208, 235, 240, 242, 253,
East Germany) 243 258, 259, 260, 269, 271
citation levels compared to West Guh, Y.-Y. 37
Germany 253 Gupta, M. 330
patenting activity 258–9 Gutmann, G. 209
R&D expenditure compared to West
Germany 248 Habermas, J. 172
SCI publications compared to West Hallwood, C. Paul 151
Germany 252 Haltiwanger, John 335
see also German national innovation Hamel, G. 104
system Hammour, Mohammad 327, 328, 345
German national innovation system Hannan, T. 330
cliometric research on 269–71 Hanson, Gordon H. 142
industrial R&D 254–5, 263–4 Hanusch, Horst 36, 37, 194, 207, 235
innovation activity, development of Harrigan, James 142
255–60, 264–9 Haskel, Jonathan 141
national level 243–60 Hauser, H. 318
public expenditure levels 243–9, 261 Hayek, Friedrich August von 57, 61,
scientific activities, development of 74, 80, 82, 161, 168, 198, 235, 303,
249–54, 261–2 304, 318
Index 355
Heine, K. 318, 319 innovation network model

Hellbrück, R.P. 281, 282, 288, 292 absorptive capacities 109–10
Helliwell, John F. 141 agents 103–7
Helpman, Elhanan 141, 161, 162, 208, competition processes 110–11
209 financial flows 110
Henkel, H. 284 knowledge flows 110
Herrmann-Pillath, Carsten 157, 161 matching mechanism 109
Hesse, G. 186, 283 networking conditions 107–8
heterogeneity networking decisions 108–9
as asymmetry and variety 15–17 R&D decision rules 106–7
concept of 17 results 112–17
in empirical work 18–21 structural flow-chart 111–12
and innovation 18–19 innovation networks, in biotechnology-
and local progress 17–18 based industries 100–103, 117–18
roles in economic evolution 15 institutions, in developing countries
specific vs general measures 19–21 327–33
Hetmeier, H.-W. 244 intellectual property rights see patents,
Heuss, E. 193, 209 and innovation
Hinze, S. 253, 258, 259, 269 interjurisdictional/regulatory
Hirsch, Seev 162 competition, theories of
Hirschfeld, L. 173 evolutionary contributions
Hodgson, G.M. 87, 89, 90 explicit 306–8
Hoffmann, W.G. 243 implicit 308–12
Hofmann, W. 203 evolutionary economic critique of
Holmes, P. 318 neoclassical theories 301–5
Holmström, B.R. 79 knowledge, competition and
Homann, K. 168 innovation 312–13
Hösli, M. 318 neoclassical theories as answer to
Howitt, P. 208, 209 EU centralization problem
Hoyningen-Huene, P. 241 299–301
human capital, reasons for lack of irreversibility, concept of 327
evolutionary perspective 73–4
Hummels, David 141 Jacobs, K. 290
Hutter, M. 75, 179, 186 Jenkins, Michael 141
Jones, S.R.G. 179
improvement axiom 49 Josselin, J.-M. 318
incorporation theory 311–12 jump points 48
India see West Bengal, trade and
transport Kahneman, D. 184, 187
information problem, and policy Kaldor, Nicholas 235
making 219–20 Kenyon, D.A. 318
Inman, R.P. 318 Kerber, W. 78, 307, 315, 318, 319
innovation Keynes, J.M. 56
definitions 216, 242 Kieninger, E.-M. 319
and heterogeneity 18–19 Kincaid, J. 318
historical interpretation of 239–42 Kirchgässner, G. 169
linear model, shortcomings Kirchner, C. 319
216–17 Kirzner, I.M. 93
and welfare 62 Kiwit, Daniel 61, 149
see also disastrous innovations Klaes, M. 79, 89, 93
356 Index
Klau, Arne Ragnar 143 Levine, Ross 331

Klump, R. 208, 209, 210 Levinsohn, James 141
Knetter, Michael M. 142 Levinthal, D. 109
knowledge-based society, analysis of liberal politics
100 definitions 60–61
Koch, L.T. 75, 173, 186, 318 and evolutionary economics 63–5
Kollman, K. 318 and welfare theory 61–2
König, W. 260 Lindblom, C.E. 181
Koop, M.J. 318 Lindenberg, S. 179, 187
Koot, G.M. 86 Lindgren, B. 32
Kornai, Janos 334 linear innovation model, shortcomings
Kötz, H. 319 216–17
Krelle, W. 210 Lipsey, Richard G. 235
Krüger, J. 37 Lipsey, Robert E. 143
Krugman, Paul 141, 345 Little, B.R. 172
Kuenne, R.E. 111 Loasby, Brian 157, 161
Kuhlmann, Stefan 236 local progress, and heterogeneity
Küppers, G. 75 17–18
Kuran, T. 177, 178, 179, 186 locational competition see
interjurisdictional/regulatory
La Grandville, Olivier de 208 competition, theories of
la Mothe, John de 140 Löffelholz, M. 177
laboratory federalism 309–10 Lorenz, Ch. 240, 242
Lachmann, L.M. 209 Lucas, R.E. 210
Ladeur, K.-H. 61 Luckmann, T. 172
Landa, Janet T. 158 Luhmann, N. 181
Landes, David S. 236 Lundvall, Bengt-Åke 235
Landgraf, R. 275
Laschet, M. 275 Mabe, M. 249, 251
Lauterbach, K. 290, 292 McCahery, J.A. 318
law of one price, NTT problems with McGlade, J.M. 345
141–2 Machlup, F. 241, 271
LDFs (Large Diversified Firms) Maddison, A. 250, 271
definitions 100 Mäki, U. 82
modelling 103–7 Malerba, F. 103
and new knowledge 102–3 Malmquist index 32–4, 41–2
Leamer, Edward E. 144 Malmquist, S. 32
learning innovation policy Marciano, A. 318
implications from stylised facts Marcouiller, Douglas 143
229–33 Markusen, James R. 143
methodological basis of 216–21 Marschall, L. 267
stylised facts, modified, as basis for Maskus, Keith E. 143
221–5 Maussner, A. 210
learning policy 233–4 Medema, S.G. 78, 83, 85
legal federalism 311 Medical Advisory Council on Diabetes
see also interjurisdictional/regulatory 282–3
competition, theories of Mehl, E. 274, 275, 280, 281, 287
Lehmann-Waffenschmidt, Marco 235 Meier, A. 75, 176, 177, 178, 179, 182,
Lembke, J. 75 186, 318
Leontief, W.W. 37 Meierjürgen, R. 280
Index 357
Merten, K. 177 organisational capital, as dimension of

Metcalfe, J. Stanley 75, 235, 236, 307 entrepreneurial competence 156–7
Mettler, D. 177 Orsenigo, L. 103
migration, sociodynamic model of Ott, A.E. 37
125–30
Miller, J.H. 318 Page, S.E. 318
Mintrom, M. 319 Paquet, Gilles 140
Mirrlees, J.A. 57 Pareto-optimality, and adaptive
Molitor, B. 185 preferences 54–5
Moore, G.E. 43, 56 Parisi, F. 319
Moorsteen, R.H. 32 patenting activity in Germany 255–60
Morduch, J. 330 patents, and innovation 193
Morgenstern, O. 207, 209 Peacock, A.T. 60
Mueller, D.C. 186 Pelkmans, J.J. 301, 318
Mukand, S. 319 per-se rules, vs reason 218–20
Müller-de-Cornejo, G. 274, 275, 280, Pfetsch, F.R. 244, 245, 246, 247, 254,
281, 287 263
Müller, K.E. 271 Pies, I. 92
Müller, M. 279 political economy see evolutionary
Murrell, Peter 333 political economy, scarcity of
Mussler, W. 66, 75, 318, 319 research
Pöppel, E. 171, 186
Nachtigall, G. 284 Popper, K.R. 186
national innovation systems, population restructuring, evolutionary
definitions 243 concept of 333–4
naturalistic fallacy 43 Porter, Michael E. 153
Neil, H. 282 Posner, R. 79
Neisser, U. 170 Potts, J.D. 90, 161, 163
Nelson, Richard R. 103, 106, 197, 207, Prahalad, C.K. 104
209, 210, 236 Pratten, S. 85
neoclassical approach, and Principia Ethica 43
heterogeneity 15 private households, importance in
neoclassical trade theory (NTT) technological change 193
and evolutionary economics 140–41 production theory, and technological
unresolved issues in 141–4 change 198–201
Nepal, financial sector 327–33 Pyka, A. 37, 75, 109, 115, 118
Nice, D.C. 319
Nisbett, R.E. 172, 184 Quaas, M. 286
non-average behaviour, by economic
agents 334–6 R&D decision rules 106
non-parametric frontier function Ramser, H.J. 208
approach for TFP 24–31 Rauch, James E. 161
normative individualism 43–4 Rawls, J. 56
North, Douglass C. 71, 146, 207 Ray, Edward John 143
Nüse, R. 170 Raymond, Frank 326
regulatory competition 311
Oates, W.E. 309, 310, 318, 319 see also interjurisdictional/regulatory
Oeser, E. 171 competition, theories of
Ogus, A. 319 research, definitions 242
Okruch, S. 91, 318, 319 Rhodes, E. 26
358 Index
Ribstein, L.E. 319 Schwarze, J. 280

Richard, S. 279 Schwerin, Joachim 235, 236
Richter, Rudolf 192, 193 Science Citation Index 251
risk structure compensation scheme Searle, J.R. 172
(Risikostrukturausgleich, RSA) Segerstrom, P.S. 208
279, 288–9, 290 selectionist approach, and
Robinson, J. 203 heterogeneity 15
Robra, B. 276 Sen, A.K. 57
Rodrik, D. 319 Serageldin, Ismail 145
Rogers, E. 319 Shannina, L.V. 271
Rogers, John H. 141, 142 Siebert, H. 318
Romano, R. 301, 312, 319 Siegenthaler, H. 176
Romer, P.M., growth model (1990) Silverberg, Gerald 325
critique 197–205 Simon, H. A. 169
outline 194–7 Simpson, A.W. Brian 93
Roos, P. 32 Sinn, H.-W. 75, 93, 300, 318
Röpke, J. 206, 207, 209, 210, 307 Sinn, S. 318
Roscher, W. 241 Six, B. 184
Rose-Ackerman, S. 319 Slaughter, Matthew J. 142
Rosenberg, Nathan 24, 207, 209, 236 Slembeck, T. 75, 78, 176, 178, 179, 182,
Ross, L. 172, 184 186, 318
Rowley, C.K. 60 social capital, as dimension of
RSA see risk structure compensation entrepreneurial competence
scheme 157–8
Rubinfeld, D.L. 318 sociodynamic models
Rürup, B. 291 evolution of a city and its
Rusch, G. 175 population 130–39
Rüsen, J. 271 migration 125–30
sociodynamics
Saarland 243 frame of concepts 121–4
see also German national innovation master equation 123–4
system principles 120–21
Salmon, P. 309, 319 Solow, R.M. 32, 206, 209
Sälter, P.M. 208 Sparer, M.S. 319
Samuelson, Paul A. 57, 163, 209 spatial directedness of trade
Sass, U. 259 NTT problem of 141–4
Saviotti, P. Paolo 16, 37, 108, 115 and separability of transaction
Scharpf, F.W. 172, 186 competences 152–3
Scherbaum, W. 277 Spencer, Herbert 43, 56
Scherer, K.R. 174 Sraffa, P. 203
Schiepek, G. 186 Staropoli, C. 107
Schlicht, E. 82, 83 Steinmueller, W. Edward 235
Schmidt, Ingo 235 Stigler, G.J. 78
Schmidtchen, Dieter 145 Stiglitz, J.E. 31, 37
Schmoch, U. 258, 259, 260 stocks, and flows 154
Schnellenbach, J. 319 Stolper, G. 259
Schönbach, K. 279 Storper, Michael 156
Schreiter, C. 195, 208, 209, 210 Stratton, I. 282
Schumpeter, J.A. 19, 235, 241 Streissler, E. 209
Schwab, R.M. 318 Streit, M.E. 66, 75, 162, 318, 319
Index 359
Strumpf, K.S. 319 trade theory

Struve, G. 239 and pricing of capability to trade
stylised facts (classical concept) 146–53
221–3 and social capital 145–6
stylised facts (modified concept) transaction costs, and transformation
221–3 costs 146, 150–53
applied to growth paths of transition transformation costs, vs transaction
economies 225–9 costs 146, 150–53
definition 223–5 transition economies, growth paths
‘success breeds success’ matching 225–9
mechanism 109 translators (role of DBFs)
Sugarman, S. 172 definitions 101
Sun, J.-M. 301, 318 see also DBFs
synergetic approach, and heterogeneity transport, model for West Bengal
15 336–42
systems competition see Trefler, Daniel 141, 142
interjurisdictional/regulatory Tuschen, K. 286
competition, theories of Tushman, M.L. 105
Tversky, A. 184, 187
Tamir, Y. 75
technical knowledge, and human Ullmann-Margalit, E. 88
capital 199–201 universities, role in innovation network
technological change model 105–6
evolutionary perspective 191–4 Ursprung, H.W. 319
index for 32–4
technological heterogeneity 13, 17 Van den Bergh, R. 318, 319
see also heterogeneity; TFP Vanberg, V. 56, 318
technometrics 19–20 variety, concept of 16
Teece, D. 104 Vaubel, R. 298
Testa, J. 262 Venables, Anthony 143
Teubal, Morris 235 venture capital firms, role in innovation
TFP (total factor productivity), as network model 105–6
measure of technological Verein für Socialpolitik 2
heterogeneity Vernon, J. 102
change of TFP as progress 23–4 Vihanto, M. 318
construction 22 Voigt, S. 61, 71
dynamic analysis 31–4 von Weizsäcker, C.C. 56, 57
further development opportunities Vosskamp, R. 75
36–7 voters’ mandate 174
generality and aggregation 21–2 Vowe, G. 173, 187
index of 24–5 Vromen, J. 88, 93
non-parametric frontier function
approach 24–31 Wagner, A. 210
as performance indicator 22–3 Wagner, H. 210
requirements for 20–21 Wagner-Döbler, R. 249
Tiebout, C.M. 300 Walker, J.L. 319
Tietzel, M. 209 Wallis, J. 181
Tinbergen, J. 59 Walz, U. 208
Tirole, J. 79 Weder, R. 209
total factor productivity see TFP Wegehenkel, Lothar 147
360 Index
Wegner, G. 75, 162, 318 Wille, W. 290, 292

Weidlich, W. 120 Williamson, O. 78
Weingart, P. 271 Wilson, J.D. 318
Weingast, B.R. 318 Windisch, R. 318
Weinstein, David E. 141, 142 Winter, Sidney G. 103, 106, 236
Weiss, A. 330 Witt, U. 57, 88, 90, 172, 325, 335
welfare economics see evolutionary Wohlgemuth, M. 59, 318
welfare economics Wolf, Holger 141
Wellisch, D. 318 Wolff, E.N. 37
Wengenroth, U. 254 Wong, Kar-yiu 163
Werker, Claudia 236 Wood, Adrian 153
West Bengal, trade and transport Woodland, A.D. 153
333–42 Woolcock, S. 318
West Germany see Federal Republic of Wouters, J. 319
Germany
Westerbarkey, J. 177 Yudkin, J. 282
Westermann, G. 30, 37
Whaples, Robert 236 Zhang, Kevin H. 143
Wildasin, D.E. 318 Ziegler, D. 256, 257, 266

Evolution Kurf PDF

Uploaded by

Copyright:

Available Formats

Evolution Kurf PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Evolution Kurf PDF

Uploaded by

Copyright:

Available Formats

What is the main topic of the book?

What is the main topic of the book?

What are some of the major sections and chapters covered in the book?

What are some of the major sections and chapters covered in the book?

Economics, Evolution and the State

All rights reserved. No part of this publication may be reproduced, stored in

Edward Elgar Publishing, Inc.

A catalogue record for this book

ISBN 1 84542 438 7

PART I ECONOMIC EVOLUTION AS OPEN PROCESS

1 Heterogeneity and evolutionary change – concepts and

PART II SELF-ORGANISATION AND NETWORKS

5 The concept of network organisation – biotechnology-based

PART III THE POLITICAL ECONOMY OF COMPLEXITY

10 Innovation and the learning policy maker – an evolutionary

A NEW BRANCH: EVOLUTIONARY ECONOMIC

The last two decades have seen an explosion of research in evolutionary

The contributions were selected from a set of papers presented at

Before giving a summary introduction of the various contributions to the

The Verein voluntarily disbanded in 1936 to avoid having to bow to the

Economic evolution as open process

ECONOMIC EVOLUTION AS OPEN PROCESS

Directedness in Economic Evolution

In their chapter on ‘Heterogeneity and evolutionary change – concepts and

develops a theory of adaptive preferences which fulﬁls two requirements at

Self-organisation and Networks

description of a broad class of dynamic collective phenomena within human

communicated, related perceptions. These perceptions – which inﬂuence

Political Economy of Complexity

In their contribution on ‘Innovation and the learning policy maker – an evo-

to derive positive policy proposals from evolutionary reasoning; he argues

Economic evolution as open process

In this chapter we present a methodological discussion on the identification

1. select the variables you consider relevant for identifying technological

2. THE CONCEPT OF HETEROGENEITY

In this section, what we attempt to briefly discuss is the concept of hetero-

2.1 Theoretical Issues

Heterogeneity as asymmetry and variety

diﬀerences or asymmetries among firms which cannot be ranked as

The concept of heterogeneity

Technological heterogeneity on higher aggregation levels

Heterogeneity and local progress

progress it is quite obvious and has to be expected that heterogeneity is

2.2 Empirical Issues

Heterogeneity and innovation

answer we start with another question: What does technological progress

Specific versus general measures

1. Despite the respective specific characteristics of a technology and

On the basis of the following four central requirements we suggest and

Requirement 1 The task therefore is to detect a measure which may help to

Suggestion 1 In order to show the way for setting up a broadly applicable

Requirement 2 The way TFP is measured should diﬀer considerably from

Suggestion 2 With respect to requirement 2 we suggest applying a frontier

Requirement 3 Related to the need to distinguish better from worse per-

Suggestion 3 For satisfying requirement 3 we suggest the computation of

Requirement 4 The measure applied should be tracked over time. The

Suggestion 4 In this respect we suggest employing the procedure to

3. TFP, TECHNOLOGICAL PROCESS

Referring to requirement and suggestion 1 we consider total factor pro-

min l eTsl eTsl

where 0 is a real number. Thus, if consumption x remains constant over