NETWORKS, NATIONAL INNOVATION SYSTEMS AND SELF- ORGANISATION.

Transcription

1 InnoSystSelfOrg/Curitiba/23/08/00/1 NETWORKS, NATIONAL INNOVATION SYSTEMS AND SELF- ORGANISATION. Pier Paolo Saviotti, INRA-SERD, Université Pierre Mendés-France, BP 47, Grenoble Cédex 9, France. Tel: ; Fax: ; saviotti@grenoble.inra.fr To be presented at the 4 th International Conference on Technology Policy and Innovation, to be held in Curitiba, August 28-31, First draft, please do not quote without the autor s agreement.

2 InnoSystSelfOrg/Curitiba/23/08/00/2 NETWORKS, NATIONAL INNOVATION SYSTEMS AND SELF- ORGANISATION. Pier Paolo Saviotti, INRA-SERD, Université Pierre Mendés-France, BP 47, Grenoble Cédex 9, France. Tel: ; Fax: ; saviotti@grenoble.inra.fr 1) INTRODUCTION. The concept of National Innovation System (NSI) acquired a considerable importance due to the work of Lundvall (1988), Freeman (1987) and Nelson (1992). The NSI can be defined as the set of institution and organisations responsible for the creation and adoption of innovations in a country. The origin of the concept of NSI can be related to the observation that innovations are not created in particular places due to greater R&D expenditures by firms and by public research institutions in socio-economic systems that are otherwise identical. We can observe that individual countries show persistent asymmetries at two levels: 1) The structure of output. The composition of the output of different countries and the sectors or sub-sectors in which they are competitive tend to be stable for long periods of time. 2) The institutions. The institutions and organisations used in different countries to achieve comparable aims can be very different and these differences can persist for very long periods of time. Examples of the first type of asymmetry are Germany s strength in chemicals and luxury cars, Japan s strength in cars and electronics, Italian strength in textiles, footwear etc.(for these and other examples see Porter, 1990). Examples of the second type are the organisation of research institutions in different countries, the involvement of government departments and ministries in the formulation of industrial and technology policies etc.. The persistence of these asymmetries seems to be incompatible with the existence of a particular pattern of R&D expenditures or of an institutional configuration that makes the country adopting it or coming closer to it more efficient than the others.if such best solution existed and if it could be unambiguously related to he differential performance of one or more countries, then the solution itself would be imitated thus leading to a convergence of different countries on the same policies and institutional configurations. The persistence of the above mentioned asymmetries implies that, even if the optimal solution existed, either it would not be perceived by different countries or the obstacles to its achievement would be so great that most countries would never get there. Furthermore, the fact that several countries can achieve comparable rates of growth with very different policies and institutional configurations raises considerable doubts about the uniqueness of the optimal solution. On the basis of the previous discussion it seems that two different concepts of the productive system of a country can be compare and contrasted: on the one hand we have a bunch of firms producing internationally comparable outputs but characterised by a differential efficiency in the production of these outputs. On the other hand we have individuals organised by institutions and assembled in organisations that affect their overall performance for what concerns both efficiency and creativity. Such institutions show a very high specificity with respect to the circumstances of the country and a high resistance to change. They are thus

3 InnoSystSelfOrg/Curitiba/23/08/00/3 profoundly affected by the history of the country. Innovations are thus not introduced by individual isolated organisations, be they firms or research institutions, but by the complex interactions of these and other organisations, that together form a system. It seems as if the particular patterns of institutional interaction constituting the NSI display self-organisation. Thus an NSI is capable of adapting to some changes in its external environment while preserving its identity and fundamental properties, that is, it runs. However, an innovation system can also learn and, under certain conditions, undergo changes in structure. Such changes do not occur only by means of adaptive learning. In addition to learning the changes in its extern environment and adapting to them, the system can perform search activities that can lead to partly endogenous structural changes. The previous considerations focused exclusively on national innovation systems. However, some of them would be equally applicable to other types of innovation systems, such as regional or sectoral ones. For all these systems we can observe asymmetries equivalent to those previously described for NSIs. Regional innovation systems can display equivalent output and institutional asymmetries, although we can expect the national system to impose a common framework on regional systems. Sectoral systems of innovation can be expected to arise when the constraints internal to the sector outweigh those inherent in the nation state. This does not mean that a sectoral system of innovation is completely independent of its location, but that the variability of its mode o utilisation amongst different countries is smaller than the differences amongst different sectoral innovation systems. Of course, while these systems are equivalent in the sense that they are all analysable as systems, they are related to one another in ways that are not yet known and very little investigated. This question is very interesting but it will not be pursued any further in this paper. The main problem analysed here will be instead the nature of the theoretical explanations appropriate to the systems investigated. The properties of all these systems are difficult to justify based on equilibrium theories. In this paper it is argued that evolutionary and complexity based theories provide a more appropriate framework to understand innovation systems. These two approaches are not just different technically, but represent two alternative world views. 2) COMPLEXITY AND EVOLUTIONARY THEORIES. 2.1) A CHANGE IN WORLDVIEW. During the past century our worldview has changed considerably. While in this paper we are not going to take into account a very large number of aspects of these changes in worldview, some need to be examined. This is true mostly for what concerns our expectations about the knowledge that we can possibly gain of the external world, be it physical or social. By the mid XIX century we had become convinced that we were in principle able to know the physical world. It was thought that both the basic components of the universe and the interaction amongst them were known. It is important to stress that such knowledge, while considered possible in principle, could be difficult to achieve in practice, because computation costs could be very high. The world was also seen as deterministic, reversible and in principle calculable. This conception of the world, deterministic, reversible an in principle calculable, was later called the Laplacian dream, from Pierre Simon Laplace, who, at the beginning of the XIXth century, attempted to give a systematic mathematical formulation of Newton s theories (Mirowski, 1989; Prigogine, Stengers, 1984, p. 28). This worldview has long since been abandoned in physics, though not without traumas in the profession, but it still holds a great

4 InnoSystSelfOrg/Curitiba/23/08/00/4 place in economics. An alternative, even if not yet as complete the Laplacian dream purported to be during the XIXth century, is today provided by what can generally be called theories of complexity. It has to be admitted from the start that such theories are by no means unified except for some extremely broad generalisations. To begin with, our certainty to be able in principle to know the world has been severely dented. First quantum mechanics pointed out that our knowledge of the physical world was intrinsically limited: observations of an unperturbed physical world were in principle impossible. The knowledge that we could gain of the physical world could at best be statistical. Later theories of non-linear systems and of chaos demonstrated that even a system represented by a deterministic equation could behave in a chaotic and non reproducible way. The description given so far of changes in our view of the possibilities and limits of human knowledge seems to imply that we were highly conceited and that we thought we had achieved a complete knowledge of the universe, while it turns out that such view was simply an exaggeration and that we had finally to scale down our ambitions and to recognise that we can know less than we hoped. While there is some element of truth in this conclusion, it is not correct to say that either we know less or can know less than we hoped. In fact our knowledge of the physical world has immensely increased during the XXth century. The fact that more scientific publications have been produced during it than in all previous history by itself does not prove the claim. It would in fact be possible for the basic principles of knowledge to have been established previously and for most of the work done in the XXth century to have been simply an extension of these basic principles to new situations, or, in other words, an extension of the paradigm embodied in classical mechanics. In fact, during the XXth century not only we accumulated an enormous amount of knowledge about new situations, but our worldview changed to a very considerable extent. The emerging worldview is not complete, but it amounts to say that the world is far more complex than we previously thought and that our possibility to know it is limited. Uncertainty and complexity are two often recurring words in today s literature. To represent this change we could say that we previously defined the universe as the set of events contained in a given range, within which we expected to be able to know everything. Today we see the universe as being a much wider range than before, but we do not expect to be able to know all of this range. Yet what we know of it might easily be much greater than the initial range. Thus we know a lot more although the perception we have of our limits has become clearer. For what concerns the social sciences, and economics in particular, the model of knowledge creation and validation was represented by physics. This was not peculiar to the social sciences. Physics was in general considered the model of scientific knowledge, that is of a knowledge that could be tested in ways impossible for other types. Most epistemology was essentially epistemology of physics. A more recent example of the powerful influence exerted by physics on other disciplines is the emergence and rise of molecular biology. Physics played a very important role in the development of neo-classical economics, as it was pointed out by Georgescu-Roegen (1971) and by Mirowski (1989). Of course, the physics from which neoclassical economists took inspiration in the construction of their system of knowledge was XIXth century physics, and in particular the theory of electromagnetic fields(mirowski, 1989). Thus, as within physics, the view of a deterministic, reversible and in principle calculable universe was gradually giving way to that of a complex, uncertain, irreversible and limitedly knowable universe, the same transition was not occurring in economics, or at least, not in neo-classical economics. The research tradition in economics that recently started to take these changes in world view into account is evolutionary economics (See Nelson, Winter, 1982; Dosi, Nelson, 1994; Nelson, 1995; Saviotti 1996). Before starting to enter into any

5 InnoSystSelfOrg/Curitiba/23/08/00/5 details of complexity or evolutionary theories let us point out that while the scope for knowledge may now seem more limited, there is a substantial advantage for economics and for the social sciences. XIXth century physics studied a world in which the only possible changes were represented by motion, by displacements of particles with respect to one another. No change in composition, as could be represented by the emergence of new types of entities qualitatively different from those that preceded them, could easily be taken into account. The changes we observe in economic systems are quite often discontinuities, radical technological innovations, new types of organisations, new activities. That is, qualitative change is an essential feature of economic development, especially in the long term. Theories of complexity and evolutionary theories provide an a priori more appropriate framework for the understanding of such long term development. These theories will be briefly described in the following section. They do not constitute a unified body of knowledge, but they are the result of research carried out in different disciplines. Thus, we do not expect that the new worldview will give rise to a general theory, capable of explaining all observable events in any discipline. In a sense this general theory would be in contradiction with the newly emerging awareness of the limits of our knowledge. However, sharing a basic framework constituted by some general concepts, metaphors and tools can both give inspiration to ask new questions and provide co-ordination between different disciplines. It must be stated here that the transfer of concepts, models and tools between different disciplines can only lead to interesting and novel questions, but not provide answers from discipline A (e.g. biological answers) to problems arising in discipline B ( e.g. economic problems). The only common element amongst the theoretical developments that will be described below is the attempt to go beyond the Laplacian dream, which in economics is represented by the equilibrium approach, and to develop the theoretical treatment of a world that is complex, changeable and continuously affected by true novelty. In other words, the presence of qualitative change and the growing complexity of the reality that we observe and study are common components of the different theoretical developments that will be described in the following section. 2.2) SYSTEMS THEORY AND OUT OF EUIIBRIUM THERMODYNAMICS A very fundamental distinction is that between closed and open systems (Nicolis and Prigogine, 1989; Prigogne and Stengers, 1984; Von Bertalannfy, 1950). The former cannot exchange anything with their environment while the latter can exchange matter, energy and information with it. These two types of systems have some strikingly different properties, some of which are particularly interesting for the purposes of this paper. Thus closed systems eventually achieve a state of equilibrium, that corresponds to the maximum possible degree of disorder, randomness and homogeneity. Examples of such systems and of their equilibrium is given by a mixture of two gases, obtained by injecting the second gas in a corner of the space containing the first one, or by a drop of ink deposited on the surface of a glass of water. In both cases we expect the heterogeneity constituted by the addition of the second component to the system gradually to disappear as the second component diffuses in the first one. The state of equilibrium will then be constituted by a homogeneous gas in the first case and by a homogeneous ink water solution in the second case. As long as the systems are kept isolated from their environment we expect the equilibrium state to be stable. We do not expect the second gas to separate spontaneously from the first one or the drop of ink to separate from water. We can imagine to obtain an open system from a closed one by gradually increasing the rates of the exchanges that the system has with its environment. As we start increasing the exchanges the equilibrium of the system will be disturbed leading to a number of possible outcomes. Instability or turbulence is one of the possible outcomes. Another outcome is the achievement of one or more stationary states, that is states characterised by the invariance of a

6 InnoSystSelfOrg/Curitiba/23/08/00/6 number of their properties. We can consider the rates of the exchanges with the environment as a measure of the distance from equilibrium. While these steady states could be confused with the equilibrium states attained by closed systems, they are in reality very different. Both equilibria and steady states show a temporal invariance of their properties, but while an equilibrium is characterised by the maximum possible disorder or randomness, that is not necessarily the case for a steady state. As we move away from equilibrium by increasing the rates of exchange of a system with its environment the system itself may undergo a number o transitions, leading to qualitative changes in the structure of the system. The concept of structure itself involves the existence of separable and distinguishable components within the system, that is of a system heterogeneity. Let us notice that if we imagine to create an open system by gradually increasing the interactions of a closed system with its environment, we start from a homogeneous state, in which individual components cannot easily be separated or distinguished, at least not at an aggregate state. Conversely, when we add a second gas or a drop of ink to the first gas or to water respectively, we move from a heterogeneous to a homogeneous state. That is, in moving towards equilibrium a closed system looses structure while an open system may acquire a structure as it moves away from equilibrium. This difference is very important for what concerns the social sciences. The systems we observe show a high degree of structure and the changes that they undergo in the course of time often lead to changes in the structure, represented by the emergence of new components and by the disappearance of pre-existing ones. This is especially true if we observe long tem developments, but we have to take into account that the rate at which innovations are introduced into socio-economic reality may be increasing sharply, thus shortening the time scale on which these qualitative changes can be observed. If both the creation of structure and changes in it are only observed for open systems then it is very likely that a large number of the systems involved in socio-economic reality are open. Open systems can undergo transitions after which their structure and properties change radically or qualitatively. As the interactions with the environment increase, when the distance from equilibrium becomes large enough, the systems may undergo bifurcations, that is transitions to new and different states, but in which the number of states after the transition is greater than before it. In the vicinity of the transition the system shows fluctuations, or it is highly indeterminate. The particular state in which the system ends up after the transition is not rigidly determined a priori. The transition can have multiple outcomes. We can at best predict the probabilities of the different states after the transition. There is thus an element of indeterminacy in the evolution of open systems. Furthermore, the transitions of open systems are both irreversible and path dependent. So far we have been comparing the analysis of equilibrium in closed systems with that of steady states, and of possible transitions between them, in open systems. In fact, as Allen (2000) points out, a model needs some simplifying assumptions in order to reduce the complexity of reality to a manageable simplicity. The more restrictive the assumptions we make, the simpler and apparently more predictive the models are. The five most common assumptions made in models are: 1) the possibility of defining the boundary separating the system studied from its environment. 2) The possibility and stability of a taxonomy describing and classifying the components of the system. 3) The assumption of homogeneity. All individual sub-components are identical to each other or have a diversity that is at all times distributed normally around the average.

7 InnoSystSelfOrg/Curitiba/23/08/00/7 4) The overall behaviour of the variables can be described by the smooth average rates of individual interaction events. 5) The stability of equilibrium. If all five assumptions are made we obtain a mechanical model, which is incapable of accounting for the possibility of feed-back processes, of non linear effects etc. If we gradually eliminate first the assumption of equilibrium and then gradually the other ones, leaving only hypothesis N 1), we move gradually from mechanical to non-linear dynamical systems to self-organising systems to evolutionary complex systems. On the way we loose (apparent) predictive power but acquire the ability to study systems that are more complex and more realistic. Models of evolutionary complex systems instead of being detailed descriptions of existing system are more concerned with exploring possible futures. What is the relevance of these considerations for the social sciences? The first answer that can be given to this question is: biological and social life cannot be explained by means of the laws applying to closed systems only. The increasingly complex structures that we observe in modern societies are incompatible with the homogenisation and randomness of the equilibria in closed systems. In fact, we can easily provide examples of open systems in socio-economic reality. Hardly any organisation can be conceived as a closed system: a firm or any other organisation continuously exchange matter, energy and information with their environment. Interestingly Chandler (1977, Ch. 8) considers the rate of throughput as the most important variable determining the organisation of a firm. Theories of complex systems and of far from equilibrium phenomena, from which the previous ideas come, are required in order to explain the existence of biological and of social life. Of course, that does not imply that all systems involved are open. Even if completely closed systems are an abstraction quite difficult to reproduce in reality, very limited degrees of openness lead to a type of behaviour closely resembling that of closed systems. Thus it is quite likely that both biological and social life are created by a mixture of systems of different degrees of openness. A number of implications follow from the previous general considerations. Social systems show self-organisation, in the sense that they can adapt to fluctuations in their environment remaining within given ranges. Adaptation implies that the system remains recognisably itself, or that it is not destroyed. This form of adaptation, called self-regulation or homeostasis, involves maintaining certain critical variables of the system within predetermined ranges (Von Bertalanffy, 1950). Examples of such forms of adaptation are the maintenance of the body temperature or of the sugar level in the blood of some biological organisms. Thus self-organisation refers to the achievement and maintenance of the structure of a system. Of course, no system needs to be eternal, and social systems are in this sense more labile than biological ones. Political systems, the organisational structures of firms and of other organisations, are quite stable with respect to the average life span of their members, but can undergo discontinuous changes leading to the collapse of the system s structure and to its replacement by another system. In terms of the previous ideas such transitions would correspond to bifurcations. Radical innovations, new paradigms (Dosi, 1982), dominant designs (Utterback, Abernathy, 1975), and technological regimes Nelson, Winter, 1977) represent discontinuities in technological and social evolution. Other implications of the previous general considerations are that the evolution of social systems may be irreversible and path dependent. These properties are also found for social systems (David, 1985; Arthur, 1989). When a technology shows increasing returns to adoption it can end up dominating other technologies even if the latter are more efficient. On

8 InnoSystSelfOrg/Curitiba/23/08/00/8 the other hand, the presence of increasing returns in the creation of knowledge may be responsible for the continuation of economic growth in the long range (see for example Romer, 1987, 1990). To summarise, in this section we contrasted an equilibrium approach, normally used in economics, with an approach derived from out of equilibrium thermodynamics and from systems theory, and we argued that this second approach is required in order to explain the evolution of socio-economic systems. The two approaches differ by more than their techniques or equations. In particular in economics, an equilibrium approach attempts to explain why reality is stable while an evolutionary approach attempts to explain why reality changes (Metcalfe, 1997). 2.3) IDEAS FROM BIOLOGY The general properties predicted by systems theory and by non-equilibrium thermodynamics are displayed by biological, economic and social systems. Systems theory and nonequilibrium thermodynamics provide a theoretical justification for all these disciplines/research traditions. In this sense we can say that systems theory and nonequilibrium thermodynamics are in a hierarchically more fundamental position than the other disciplines/research traditions. However, this does not imply that we can deductively infer the properties of biological or economic systems from systems theory and non-equilibrium thermodynamics. Historically the theoretical legitimation comes ex-post and each discipline/research tradition develops concepts appropriate to its observation space. Biology started much earlier than economics to deal with the problem of qualitative change: the origin of species, the emergence of new ones and the extinction of pre-existing ones, are examples of such problems. If economics intends to deal with qualitative change it seems legitimate to borrow ideas and metaphors from biology. Some o the most important biological ideas that can be applied to the study of economic evolution are variation, selection and inheritance. Variation is the process which gives rise to new species, some of which will survive the process of selection. The net number of species surviving selection determines the diversity of the system. In order to survive species have to adapt to their environment. While it was previously believed that only the fittest would survive, modern opinions imply that the tolerably fit can survive (Hodgson, 1993). In economic systems R&D, or more in general search activities, contribute to variation, while regulation and competition are the main forces responsible for selection. Inheritance refers to the conservation of some traits in the subsequent generations of a species. In turn such conservation depends on the transmission of genetic material from one generation to the next. Of course, firms and research laboratories do not have DNA, but organisational routines have been interpreted as the equivalent of the genetic heritage of biological organisms. The use of a population approach, as opposed to the typological approach commonly used in economics, is common in biology (Saviotti, Metcalfe, 1991). All these concepts and processes currently used in biology constitute a very good basis for the analysis of qualitative change and of the heterogeneity of agents, problems which are central to an evolutionary approach in economics. In other words, economics and biology have a considerable degree of similarity, both structural and of overall knowledge goals. However, no mechanical transfer of concepts and models between different disciplines/ research traditions is possible. Adaptation of general concepts is required in the specific context of each discipline/research tradition. Thus variation is blind or random in biological systems,

9 InnoSystSelfOrg/Curitiba/23/08/00/9 corresponding to darwinian evolution, while it acquires a Lamarckian character in economic systems, due to the intentionality and purposeful character of the latter ones ( Saviotti, Metcalfe, 1991; Hodgson, 1993; Nelson, 1995; Dosi, Nelson, 1994). Biology can be a very powerful source of inspiration for evolutionary economics, but in the sense of allowing us to formulate new questions and problems, and not to provide biological answers to economic problems. 2.4) Organisation theories. The term organisation theory refers here to a number of heterodox theories of the firm and to theories and concepts which have emerged in management science and in business history. These theories have two aspects in common: first, they differ from neo-classical theories because they do not assume optimising behaviour; second, they open up the black box of the firm, or of other organisations, by introducing explicitly organisational structure and internal conflicts. Satisficing behaviour and internal conflicts are emphasised by behavioural theories of the firm ( Simon, 1947, 1957; Cyert and March, 1962). The distinction between strategy and structure and the emergence of qualitatively different forms of organisational structure ( U and M form) have been studied by Chandler (1962, 1977). Competencies have been stressed for example by Penrose (1959), McKelvey (1982), Teece (1986, ), Tushman and Anderson (1986). Satisficing behaviour, routines and selection rules have been introduced into their evolutionary scheme by Nelson and Winter (1982). The growing role played by knowledge creation and utilisation in the performance of firms, a topic which has become very important in evolutionary theories of the firm, has been perceived and developed mostly within this research tradition. 2.5) Economic antecedents of evolutionary theories. In the past a number of economists have had intuitions which represent true antecedents of modern evolutionary theories. For example, Marshall is very often quoted as having said that 'the mecca of the economist lies in economic biology rather than in economic dynamics' (Marshall, 1949). Marshall clearly recognised that 'economics, like biology, deals with a matter, of which the inner nature and constitution, as well as the outer form, are constantly changing' (1949, p. 637), a relatively clear reference to qualitative and structural change in economics. However, in spite of recognising the value of a biological metaphor, Marshall did not use it and relied more on economic statics than on economic dynamics. Herbert Spencer was amongst the first to develop an evolutionary approach to social development. While some of his ideas can be interpreted in a pro-aristocratic, racist and sexist way, others are quite relevant for modern evolutionary developments. Spencer (1892, p. 10) defined evolution as 'a change from an indefinite, incoherent, homogeneity, to a definite, coherent heterogeneity through differentiation. Spencer thought that evolution necessarily involves progress, and that complexity is generally associated with fitter and more adaptable forms. These considerations anticipate the formation of structure and variety growth. Veblen made a very explicit use of a biological metaphor (Hodgson, 1993). For Veblen 'idle curiosity' was the source of diversity or mutation in the evolutionary process. The institution became the unit of selection but also in the mean time the replicator. Institutions were characterised by a relative stability and continuity through time. They could thus transmit diversity from one period to the next ensuring that selection had relatively stable

10 InnoSystSelfOrg/Curitiba/23/08/00/10 units on which to operate. Variation, selection and inheritance were thus present in Veblen's analysis. Schumpeter defined economic development as the carrying out of new combinations of productive means by entrepreneurs (Schumpeter, 1912, 1934, p66). For him these new combinations are new products, new processes, new markets, new sources of raw materials and new organisational forms. All these new combinations give rise to products, processes etc which are qualitatively different from those that preceded them. In more modern terms one would say that Schumpeter attached a great importance to radical innovations as ingredients of economic development. Thus in his view qualitative change and the generation of economic diversity are central to long term economic development. Furthermore, Schumpeter stressed the non equilibrium aspects of capitalist development. The creative destruction which 'incessantly revolutionises the economic structure from within, incessantly destroying the old one, incessantly creating a new one' is one of the fundamental mechanisms of capitalist economic development (Schumpeter, 1943). Curiously Schumpeter rejected the use of a biological metaphor in economics. Hayek attached a great importance to the role of rules. He spoke of the 'genetic primacy of rules of conduct'(1982, Vol. 3, p. 199, cited in Hodgson, 1993, p. 164). A rule is defined by Hayek as a regularity of conduct of individuals. The durability of rules is due to replication through imitation. This mechanism accounts for the much faster rate of cultural evolution compared to the sluggish biotic process of genetic change and selection ( Hodgson, 1993, p. 165). The selection procedure for rules, however, is quite interesting. Rules are selected on the basis of their human survival value, that is, they are indirectly selected through the association with a particular group. Also, the idea of spontaneous order, which he compared to the concepts of autopoiesis, cybernetics, homeostasis, self-organisation, synergetics ( Hayek, 1988, p. 9, in Hodgson, 1993, p. ) was central for Hayek. In support of spontaneous order he quoted Prigogine and his school (Hayek, 1982 Vol. 3, p 200, cited in Hodgson, 1993, p. ). The ideas of all these economists were compatible with and sometimes anticipated further evolutionary developments. As it was previously pointed out, evolutionary theories are still at an initial stage, and they have not achieved the articulation of neo-classical economics. Their comparative advantage, which could justify the effort of further articulation, lies in the explanation of situations characterised by qualitative change, radical uncertainty and by the heterogeneity of agents and techniques. Qualitative change will be one of the main topics in the rest of this paper. Thus variety will be analysed as the concept that allows us to treat analytically qualitative change. On the other hand, qualitative change can only occur by means of the creation and utilisation of new forms of knowledge. Such phenomena are fundamental for the performance of NSIs. While evolutionary theories can give some advantages in the analysis of qualitative change, of radical uncertainty and of heterogeneous agents, they do not necessarily escape some of the tensions inherent in economics and in the social sciences in general. Thus in neoclassical theory the economic system was determined to go towards equilibrium and to stay there, except for temporary displacements, leaving agents only the freedom to optimise ( Hodgson, 1988). Even in evolutionary theories path dependence may be considered to compel agents to stay within a path that they have not chosen. However, such determinism is never

11 InnoSystSelfOrg/Curitiba/23/08/00/11 complete in evolutionary theories. First, even after having chosen a given path or trajectory, agents still have a considerable amount of residual freedom, which influences their performance. Such freedom does not allow them to redesign radically the technological or conceptual system on which they base their competitive capabilities, but it can manifest itself in terms of incremental innovation. Second, in the vicinity of transitions leading to qualitative change fluctuations lead to a very high uncertainty, destroy previously accumulated competencies (Tushman, Anderson, 1986) and temporarily disrupt path dependence. In these conditions agents' freedom is very considerable. Of course, agents are not necessarily aware of being in a transition phase. In these conditions uncertainty usually means greater risk and greater opportunities than in a mature, stable market. Transition phases represent conditions more favourable to entrepreneurial than to routine behaviour (Winter, 1984). Thus, if anything evolutionary theories leave a greater room for uncertainty, intentionality and individual freedom than neo-classical ones. We can now summarise the main differences between neo-classical and evolutionary theories. 1) Qualitative change, or change in the composition of the system, resulting from the balance of variation, creating new 'species', and selection, based on differential adaptation. Inheritance too affects the rate and type of qualitative change. 2) Uncertainty, path dependency and multistability, all features arising from the out of equilibrium nature of systems and processes. 3) Heterogeneity of agents, requiring a population approach, emphasising not only representative agents and mean values of properties, but also their distribution within a a population. Such differences can both provide a more realistic analysis of innovation systems and justify some of their main properties, such as historical specificity and the multiplicity of institutional configurations, which are impossible to justify in terms of neo-classical theory. In what follows the implications of these general concepts for innovation systems will be discussed in greater detail. 3) IMPLICATIONS OF EVOLUTIONARY THEORIES FOR THE NSI. 3.1) Variety. Economic development not only allows us to produce greater quantities of the goods and services that were previously existing more efficiently, but also to create new goods and services qualitatively different from pre-existing ones. The composition of the economic system thus changes in the course of economic development. Most of the objects of consumption that we can observe in every day life and the activities required to produce them have been created within the last one hundred years. The importance of this changing composition of the economic system for economics depends on whether composition is only an effect of previous economic development or also a determinant of future economic development. There is considerable evidence that not only the composition of an economic system is a determinant of its future economic development, but that such relationship is perceived by policy makers. The efforts made and the resources allocated by several countries to create new industrial sectors either buy supporting the relevant search activities or by

12 InnoSystSelfOrg/Curitiba/23/08/00/12 providing resources and environmental conditions required for their creation are proofs that there is an implicit admission that the composition of an economic system matters for its future economic development. Yet our theoretical understanding of how composition affect economic development is very limited. Important studies by Salter (1960) and by Cornwall (1977) showed that structural change could make a very important contribution to economic growth. However, more recent studies by Fagerberg (2000) and by Fagerberg and Verspagen (1999) show that the role played by structural change in economic growth in the period is radically different from that of the periods studied by Salter and by Cornwall ( and 1950s to 1960s respectively). A limit of these studies pointed out by the authors is the use of statistical information that was not at the outset designed to measure changes in the composition of the economic system. Without getting into a treatment that would be excessively detailed for the purposes of this paper we can observe that there is evidence that the composition of the economic system can be an important determinant of its growth and development, but that our theoretical understanding of the mechanisms by which this happens is very limited. In order to improve our understanding of the role played by the composition of the economic system we need first to define one or more variables capable of measuring such composition and its changes in the course of time. The variable used in this paper is variety. Qualitative change modifies the composition of the system. New entities, qualitatively different from the old ones, appear and the old ones disappear. Variety measures the net number of these entities. If we adopt a simplified representation of the economic process as a set of actors ( individuals and institutions) performing transformation activities ( transforming inputs into outputs) thus generating new types of output, we can define variety as the number of actors, activities and objects necessary to describe the system ( Saviotti, 1991, 1994, 1996). Several aspects of variety need to be distinguished. For example, we can define output variety (V q ) as the number of distinguishable types of output, process variety (V p ) as the number of distinguishable types of processes used to produce V q, institutional variety (V i ) as the number of distinguishable types of institutions, etc. The overall variety of the economic system, as defined above (previous paragraph) is the combination of all these types of variety. Another relevant distinction is that between national (V N ) and world (V w ) variety. These distinctions need to be introduced because the types of variety behave differently in the course of economic development. For example, process variety can fall at certain times by means of standardisation or modularization, while in the meantime output variety is increasing. Here it is to be stressed that an important component of process variety comes from the competencies used in firms to produce the required output. This type of variety is discussed much more explicitely by Cohendet and Llerena (1997). 3.2) VARIETY AND ECONOMIC DEVELOPMENT. Most theories of growth, while not denying the presence of qualitative change, cannot take it into account. Most such theories are macroeconomic and the composition of the system is not one of the variables that they consider. Schumpeter s theories, pointing to the fundamental role played by radical innovations, imply a very close relationship between qualitative change and economic development. Of course, economic development can be conceived as comprising a large number of aspects, such as demand, competition, trade etc.. Variety is in principle related to each of these aspects of economic behaviour. However, in this paper the discussion will concentrate on a limited number of aspects. For a more extended discussion

13 InnoSystSelfOrg/Curitiba/23/08/00/13 the reader is referred to Saviotti (1994, 1996). Here we start the discussion with two hypotheses about the role of variety in long term economic development. HYP1): The growth in variety is a necessary requirement for long-term economic development. HYP2): Growing variety ( new sectors) and growing productivity ( in older sectors) are complementary, and not independent aspects of economic development. It is to be pointed out that these two hypotheses can be valid only in the long run and at a sufficiently high level of aggregation. At a lower level of aggregation in a number of cases variety does not grow. For example, the emergence of dominant designs (Abernathy, Utterback, 1975, 1978), of technological regimes (Nelson Winter, 1977), of technological paradigms (Dosi,1982), and of standardisation reduce the variety within a given technology. While the act of creation of a new technology increases the variety of the system, the commonality of concepts, routines and tools implied by the previous concepts leads the (not any more new) technology along a path of increasing efficiency, thus contributing to the complementarity between growing efficiency and growing variety. Thus, while variety may fall within a given technology, the growing efficiency afforded by this trend contributes to the accumulation of resources required for the generation of new species, thus leading to variety growth at the highest level of aggregation in the system. The situation here bears a strong similarity to the role of productivity growth in agriculture, which allowed the accumulation of resources to be invested in the nascent process of industrialisation (Kuznets, 1965). The second hypothesis can be reformulated as the complemetarity between routines, which lead to greater efficiency, and search activities, leading to the generation of new types of outputs and processes, and, therefore, to higher variety. Equivalently, one can reformulate it as the complementarity between the circular flow and innovations in Schumpeter's theory of economic development. The previous hypotheses about variety imply a particular interpretation of the schumpeterian process of creative destruction, in which old species disappear and new ones emerge. Creative destruction leads to qualitative change, but not necessarily to variety growth. On the other hand, variety growth can take place without creative destruction, just by addition of new species to old ones. Here it must be observed that it is the net variety, the one surviving the process of selection, which appears in hypotheses 1) and 2). These hypotheses imply that the rate of creation of new species is greater than the rate of extinction of pre-existing ones, not that such rate of extinction is zero. Variety growth and the actual nature of creative destruction can be more accurately represented by means of elementary processes in technological evolution. These are processes such as substitution (of a technology for another one), specialisation and the emergence of completely new products. In principle we can reconstruct any complex process of technological evolution as a combination of elementary processes. Each elementary process has different implications for variety: for example, substitution does not change variety at all (an old product disappears and a new one takes its place), while the emergence of completely new products gives the maximum possible contribution to variety growth. Overall variety can then only increase if variety creating elementary processes (e.g. new products) predominate over variety conserving (e.g. substitution) processes (Saviotti, 1988, 1991, 1994, 1996). Hence, creative destruction exists, but it creates more than it destroys. In other words, it has to be pointed out again that the variety discussed in the hypotheses 1) and 2) is net variety, the outcome of the combined processes of variation and selection. In economic systems variation is essentially created by search activities, all those activities which scan the environment searching for

14 InnoSystSelfOrg/Curitiba/23/08/00/14 alternatives to the existing routines. Variation creates a large number of potential species/technologies, accompanied by new routines, only some of which are sufficiently adapted to the environment. The less adapted ones are eliminated by selection. Selection is the result of a series of processes, like competition and several forms of regulation. There are several types of empirical evidence that variety has been increasing in the course of economic development ( family trees of technologies, classes of patents and of international trade etc). For a more detailed discussion of this problem the reader is referred to Saviotti (1996). Furthermore, the two hypotheses can be justified, for example, on the basis of Pasinetti's (1981, 1993) model of economic development. Let us begin with an economy constituted by a constant set of economic activities in which technical change leads to constant productivity growth, and in which demand for existing goods and services tends to saturate. In such an economy an imbalance would arise in the long run because it would be possible to produce all the demanded output using only a part of the existing resources, including labour. This imbalance, which would constitute a barrier to long term economic development, could be imputed to technological change. However, technological change plays a second role, that of creating completely new types of outputs and activities. If this happened, that is if net variety increased, the new activities could utilise the resources made redundant by the imbalance productivity growth/demand growth in old activities. On the other hand, variety can only grow if productivity growth in older activities allows the accumulation of the resources to be invested in search activities, which in turn will lead to new varieties. This treatment has been focused exclusively on the structural conditions leading to variety growth and development. Of course, even if the new goods and services and the sectors created by them could be correctly identified ex-ante, which is quite unlikely in general because uncertainty is at a maximum near transitions, the adaptation of the economic system to the new structure/composition could not be taken for granted. New human resources need to be created to produce the new goods/services leading to variety growth. The creation of these resources requires in addition to investment, institutional and organisational adaptation. While the required investment can come from the increased efficiency of pre-existing sectors and from the supra-normal profits of the temporary monopolists in the new goods/services, the required patterns of institutional and organisational adaptation are more difficult to predict. The analysis in this paper can only allow us to state that a set of development paths leading to variety growth has a higher economic development potential than another set in which variety is conserved. The study of the institutional and organisational conditions required for variety growth then becomes a subject for further research. 3.3) VARIETY AND THE INTERNATIONAL COMPETITIVENESS OF COUNTRIES. If we accept that growing variety is a necessary requirement for long term economic development, it follows that the income share of pre-existing sectors can be expected to fall gradually in the course of time. We can also expect that, however limited the extent of specialisation of any country, its national output variety will be lower than the world output variety at a given time: V j V w (1) If world output variety keeps increasing we can expect that, although individual countries tend to specialise, this specialisation cannot remain constant and must reflect the new goods and services emerging in the world economy. In general we expect national variety to increase