Innovation as a Systemic Phenomenon: Rethinking - Semantic Scholar

Enterprise & Innov a tion M a na gement Studies, Vol. 1, No. 1, 2000, 73 ± 102

Innova tion a s a Sys temic P henomenon: Rethinking the Role of P olicy

K EIT H SM ITH

Abstract This pa per looks a t the polic y im plic a tions of v iew ing innov a tion a s a system ic phenomenon. The ® rst sec tion prov ides a brief ov erv iew of conc eptua l a pproa c hes used in the rec ent litera ture on innov a tion system s. The sec ond pa rt of the paper looks a t lea rning a nd tec hnologic a l know ledge a t the ® rm -lev el, a nd explores the w a ys in w hich different theoretica l a pproa c hes a ffect our understa nding of innov a tion processes. This discussion foc uses on the contra st betw een `system s’ m odels of lea rning a nd the conc epts of knowledge w hich underpin the current `m a instrea m ’ ra tiona le for public polic y in this a rea . The third sec tion disc usses polic y problem s a rising from this broa d ® eld of study, focusing on tw o issues: the ra tiona le for policy interv ention; a nd policy c a pa bilities a nd `knowledge ba ses’. Key w ords: Innovation; Innovation systems; Innovation policy.

1. S ys tem s Appr oaches to Innova tion 1.1. The Ba sics of `System s’ Approa c hes `S ys tems’ approaches to innovation are founded on one of the most pers istent themes in modern innovation s tudies, namely the idea that innovation by ® rms cannot be unders tood purely in terms of independent decis ion-making at the level of the ® rm. Rather, innovation involves complex interactions between a ® rm and its environment. On one level the environment cons is ts of interactions between ® rms ± especially between a ® rm and its network of cus tomers and suppliers . Here the argument is that inter-® rm linka ges often involve sus tained quasi-cooperative relations hips which shape learning and technology creation rather than being arms -length market relations hips . On a second level the environment involves broader factors shaping the behaviour of ® rms: the social and perhaps cultural context; the institutional and organizational framework; infras tructures ; the proces ses which create and distribute scienti® c knowledge, and so on. Environmental This paper is based on work undertaken for a research project funded under the F ourth F ramework Programme, E uropean Commis sion, Targeted Socio-Economic Research Programme, Project no. S OE1-CT95-1004, Ìnnovation Systems and European Integration’ . Keith Sm ith, STEP Group, Storga ten 1, 0155 Oslo, Norwa y; e-m a il: keith.sm [email protected] Enterprise and Innova tion M ana gem ent Studies ISSN 1463± 2446 print/ISS N 1469± 5863 online # 2000 Taylor & F rancis L td http://www.tandf.co.uk/journals/routledge/14632446.html

74

K. Smith

conditions are often seen as speci® c to regional or national contexts, but they are als o dynamic: their forms of operation change with political conditions, changing technological opportunities, economic integration process es and so on. The basic argument of systems theories is that s ys tem conditions have a decis ive impact on the extent to which ® rms can make innovation decis ions, and on the modes of innovation which are undertaken. The relevance of `s ystems ’ approaches springs from two hypotheses concerning innovation. The ® rs t is that innovation is central to competitivenes s among ® rms; the evidence for this hardly needs documentation here. The second is that innovation is perv a siv e: it is in no s ense a marginal phenomenon, and it underpins economic growth at the national level. The processes which effect innovation thus s hape overall trajectories of economic development. S ome empirical evidence of the interactive `s ys temic’ processes which underpin innovation will be presented below. Systems theories involve a very strong overall hypothes is that differences in macroeconomic performance can be traced to underlying system differences . What factors s uggest that a focus on na tiona l s ys tem s and national policy levels might be relevant? On the one hand we have pers istent differences in national economic performance. The world economy es sentially divides into two groups of countries, rich and poor, with little convergence between them; inside the rich group there is convergence in real income levels , productivity levels and so on ± for a recent overview, see Dowrick (1991). Hower, even where economies are converging in terms of macroeconomic indicators , it should be noted that this is occurring on the bas is of differenc es in growth rates of output and productivity which can be marked and persistent between the national economies involved. Thes e differences appear agains t the background of underlying structural differences . Two important areas of structural difference are identi® able at a national level. F irs tly, there are pers istent va riations in s ystem s of governance: both narrowly in the s ense of formal regulatory s ystems of corporate governance, and more broadly in the s ense of rules of the game for corporate behaviour (L azonick, 1991). S econdly, many countries construct and maintain quite specialized technological capabilities, re¯ ected in patterns of R&D expenditure, patenting, scienti® c publication and so on (Archibugi and Pianta, 1992; Patel and Pavitt, 1994), such that ® rms appear to develop competencies and capabilities within speci® c national contexts , even when they are multinational in term s of production and operations (Patel and Pavitt, 1991). It is s ometimes argued that globalization is rendering the state (at whatever level) obsolete, and that the integration of product and capital markets removes the poss ibility of effective policy interventions by government. However, the continuing nationally s peci® c character of much economic functioning sugges ts that, although the foundations and scope of policy may by changing, government will remain important in setting the context and framework for economic behaviour (Hirs t and T hompson, 1996). 1.2. The Litera ture on Innov a tion System s T he innovation s ystems literature is an evolving ® eld; moreover it is one with s trong connections to other theories and ® elds of study, both historically and in contemporary research. F or example, s ys tems theories often return us to long-s tanding debates in economic theory. Thes e may be to do with the importance of national policy frameworks in economic development (often deriving from ideas of F riedrich

Rethinking the Role of Polic y

75

L is t (F reeman, 1994)), or of ins titutional conditions (where the very extensive ins titutional economics literature remains important). More generally, they return us to the broad conceptions of the economy as a s ocial proces s deriving from M arx; in fact M arx is one of the few really important theorists to attempt to combine a theory of technological change with a theory of development. Thes e dis parate frameworks are applied to develop an unders tanding of quite s peci® c trajectories or proces ses of innovation and, s ince systems theories tend to argue that innovation and technological change fundamentally shape economic growth, what we have is an emerging empirical and historical approach which (potentially or actually) integrates a theory of innovation into a theory of growth and development. This kind of integration is a s tep beyond Schumpeter, whose treatment of innovation processes is rudim entary or partial. Regardless of the genealogy of the approach, it stands in strong contrast to that of neo-class ical mains tream in economics. This is particularly the case for the general equilibrium approaches which underpin the s o-called new clas sical m acroeconomics; thes e approaches rely on optimization behaviour. But there is als o a contrast with recent industrial organization theory. M ains tream industrial economics and organization theory has moved a long way from the models of perfect competition, determinis tic environments, perfect information, cons tant returns to s cale which characterize many textbooks and which are often used to caricature mainstream economics . F or more than twenty years the dominant analys is has s tress ed strategic interdependence between ® rms , uncertainty, asymmetric information and increasing returns , and the literature on thes e topics is now very large. Nevertheless this literature has not addressed ins titutional iss ues , it has a very narrow concept of uncertainty, it has no adequate theory of the creation of technological knowledge and technological interdependence amongst ® rms , and it has no real analys is of the role of government. In addres sing these core features of reality, the s ys tems approach takes us, for all its pos s ible limitations , into a more promis ing arena for policy analys is . The modern literature, focused around na tiona l s ys tems of innovation, but at leas t in recent years also acknowledging both localized and trans national dimens ions , begins in the 1980s (Edquist, 1997; McK elvey, 1991). Quite apart from the broad con¯ icts between neo-class ical and more heterodox theories , there appear to have been two basic underpinnings to the national sys tems approach, both rooted in s tudies of innovation. On the one hand, there were ® rm-level studies of inter-dependence between producers and users of technology, emphasizing sustained user-producer interactions in technology creation. These relationships were facilitated by indus trial s pecialization and common cultural and policy environm ents . Teubal (1977) was probably the ® rs t to use the term ùs er-producer interaction’ , but this notion was developed into a systems approach by L undva ll, Andersen and others. Their studies of export specialization in Denmark showed the importance of the Danis h agroindustrial complex, and s uggested that competitivenes s in agricultural products was related to strong links between agricultural producers and specialized equipment s uppliers (Anders en and L undvall, 1987). Their analysis focus ed very much on s earch s trategies and learning process es within this key area of specialization, without linking this s peci® cally to a national level of ins titutional organization. In this approach, and in an important 1988 paper by L undvall (1988), the national s ys tem was es sentially seen in terms of inter-industry technological inter-dependence, based on the suppliers and users of capital and intermediate goods in areas of

76

K. Smith

competitive trade specialization. G roups of users and producers, engaged in interactive learning, create different complexes or clusters of technological capability which when taken as a whole de® ne the divers e components of the national s ys tem. This is in effect an evolutionary approach, looking at the co-development of learning process es and competitive s pecialization. On the other hand, at around the same time, F reeman (1987) was developing a s omewhat different approach, based on an attempt to unders tand Japanes e post-war ind ustrial and innovative performance. F reeman focused on national-level policies, and social and institutional factors shaping ® rm behaviour. These included the role of MITI in forming economic development strategies, in forecas ting, and in a range of actions related to technology acquis ition from abroad. F reeman also stress ed s peci® c features of the internal organization and objectives of Japanese companies (es pecially large companies) with res pect to the role and methods of innovation ins ide the ® rm, and to their ability to inves t for the long term ± in effect drawing attention to the national characteristics and effects of Japanes e corporate governance. F inally F reeman placed great weight on the scale and character of education and training in Japan. Rather than having the evolutionary characteristics of the L undvall approach, F reeman stres s ed dis cretionary decision-making, arguing that the combination of public policy, corporate governance, and education and training s haped the rate and ¯ exibility of innovation in Japan, and underpinned is extraordinary pos t-war performance. What we have, then, in these `founding’ texts is different (although not necess arily incompatible) approaches : one bas ed on the evolution of specialization and its ass ociated patterns of interaction and learning, and one based on economy-wide features of corporate behaviour, policy and s upport proces ses such as education. T hes e differences in emphasis were carried through into the two major s tudies on national sys tems publis hed in the early 1990s: the ® rst, a major extens ion of the work of L undvall and his collaborators in Aalborg (L undvall, 1992), and the s econd an international comparative s tudy of fourteen economies under the leaders hip of Nels on (1992). The difference between these volumes can probably best be s ummed up in terms of two approaches to national sys tems, described by L undvall hims elf. According to L undvall a distinction can be made between a narrow and a broad de® nition of an innovation s ystem res pectively: The narrow de® nition would include organisations and ins titutions involved in searching and exploring ± s uch as R&D departments, technological ins titutes and univers ities . The broad de® nition . . . includes all parts and aspects of the economic s tructure and the ins titutional set-up affecting learning as well as s earching and exploring ± the prod uction sys tem, the marketing system and the system of ® nance present thems elves as s ubs ystems in which learning takes place (L undvall, 1992, p. 12). Nelson’ s Na tiona l Innov a tion System s ess entially follows the narrow de® nition. It cons is ts of fourteen national studies of three groups of countries: large high-income countries, smaller high-income countries, and lower-income countries (three of which are however rapid developers ). The studies are primarily descriptive, and are in effect detailed studies of R&D organization structures and allocations over time; as such they are very valuable but they often go to the conceptual core of the s ystem s approach only in somewhat partial ways. In Na tiona l System s of Innov a tion, L undvall and his collaborators focus ed much more on a conceptual account of the characteris tics and effects of learning. T heir


77

de® nition of a system is as follows : . . . A system of innovation is cons tituted by elements and relations hips which interact in the production, diffusion and us e of new and economically useful, knowledge . . . a national system encompasses elements and relationships, either located within or rooted inside the borders of a national s tate (L undvall, 1992, p. 2). In the L undvall framework innovation is conceptualized as learning, since innovation is ± by de® nition ± novelty in the capabilities and knowledges which make up technology. The Aalborg school uses three basic concepts to understand the nature and dynamics of learning: the organized market, interactive learning, and the ins titutional framework. The s tarting point for the analys is is the prevalence of vertical linkages between ® rms supplying interm edia te and capital goods to each other through the production chain. The argument is that these market relationships are not arms -length, but ins tead pers ist through time and involve inter-® rm co-operation in the development and design of products. Cooperation means that there are no independent dem and and s upply functions : product speci® cations are jointly developed. The markets are òrganized markets ’ in that they are bas ed on explicit or tacit agreements to collaborate over time. User-producer interaction in technology development means that learning proces ses are interactive ± there are ¯ ows of data and information, and feedbacks concerning both needs and product performance between agents . Such arrangements require trust, and involve underlying cultural contexts which go beyond the contractual relationships of the pure market. It is here that institutions become important, in the sens e of culturally or politically es tablis hed `rules of the games’ . The ins titutional framework is wider than the economic sphere, but it has powerful impacts on the internal organization of ® rms, and on ® rm inter-relationships . Ins titutions imply routinized behaviour and actions: they `reduce uncertainties , coordinate the use of knowled ge, media te con¯ icts and provide incentive systems’ (L undvall, 1992, p. 26). Within this overall framework, learning involves the creation of both tacit and codi® ed knowledge concerning not only technical characteristics of production and innovation, but also knowledge concerning how and why to search in particular ways, including knowledge of key (problem-s olving) people within the relevant networks (L undva ll and Johnson, 1994). F rom both of the perspectives overviewed above, proces ses of learning and knowledge creation emerge as central iss ues in innovation capability. Within the s ys tem s literature there is one major attempt to conceptualize knowledge creation and dis tribution at the system level in an important paper by David and F oray (1995). The paper s eeks to des cribe the multi-dimensional character of s cienti® c and technological knowledge. It s hould be noted that they do not look at all forms of knowledge or interactions relevant to ® rm-level economic perform ance: they leave aside knowledge related to ® nance, marketing, and design, for example, which are elsewhere s een as relevant to innovation. The David and F oray approach focus es explicitly on `learning s ys tems for s cienti® c and technological knowledge’ , which is s een in a highly differentia ted way, both in terms of its characteris tics and functions , and its ins titutional features. F irs tly, relevant knowledge is clas s i® ed in terms of its objects , and related actions , dis tinguishing between: knowledge of factual propositions; knowledge

78

K. Smith

which constitutes explanations and understanding; operative knowledge for performance of tas ks ; and knowledge of relevant actors. T echnological knowled ge bas es are s een as either generic, infratechnological (meaning primarily methodological), applied, and product-proces s relevant. All of thes e types of knowledge can be either codi® ed or tacit, and are producer under different modes of organization which shape different disclosure regimes. Thes e considerations lead to a concept of `knowledge-product s pace’ , which is ess entially a way of categorizing different forms of knowled ge by placing them with res pect to three different dimens ions : . from completely tacit to fully codi® able; . from fully disclosed to fully res tricted; and . from privately owned to publicly ava ilable. T he argument is that within this complex structure of differentiated knowledges , what determines performance is not s o much knowledge creation as the `d istribution power’ of the system: the s ystem’ s `capability to ensure timely access by innovators to the relevant s tocks of knowledge’ . The distribution power of the s ys tem affects risks in knowledge creation and use, speed of access to knowledge, the amount of s ocially was teful duplication and s o on. David and F oray identify ® ve process es of knowled ge dis tribution relevant for innovation: . the distribution of knowledge among universities, res earch institutions and industry; . the distribution of knowledge within a market, and between s uppliers and users; . the re-use and combination of knowledge; . the distribution of knowledge among decentralized R&D projects ; and . dual technological developments (es pecially civil and military). T his kind of approach clearly has policy signi® cance, since several of the organizational channels identi® ed by David and F oray ± particularly the ® rst and last ± are in practice strongly shaped by policy decisions . 1.3. Other `System s’ Approa c hes T he approaches des cribed above are important as analys es of contemporary learning proces ses and patterns of innovation. But as noted above, they have been developed agains t the background of wider bodies of work which explore interactive phenomena, at the level of ® rms, industries , regions or even national contexts.

1.3.1. Aproa ches from the History of Technology. This ® eld has developed rapidly in recent years es pecially around analys es which conceptualize technologies not as artefacts but as integrated s ys tem s of components and supporting managerial or s ocial arrangements. A particularly in¯ uential body of work has been that of T homas Hughes , whos e his tory of electrical power generation and dis tribution emphas izes ® rstly that the development of this core technology of the `s econd ind ustrial revolution’ , mus t be understood in terms of `s ystems , built by s ystem s builders’ . This work includes study of the electri® cation of the US A, UK and G ermany between the 1880s and 1930s. As Hughes s hows, the evolution of electric power systems was different in each country, despite the common pool of knowl-


79

edge to draw on. Reasons for these differences are found in the geographical, cultural, managerial, engineering and entrepreneurial character of the regions involved. The `networks ’ which he s tudies refer not only to the technology but also to the institutions and actors involved. This s ystemic approach has also been used in general studies of the history of technology, particularly thos e of Mokyr (1992) and G ille (1978).

1.3.2. Approa c hes from `S c ienc e a nd Tec hnology Studies’. The term `s cience and technology studies’ refers to a wide body of primarily s ociological res earch that sees the development and use of technology within a social framework: thus differences in technological performance between s ocieties have at least s ome of their roots in s ocial structure and cultural forms (Jas onoff et a l., 1995; Williams and Edge, 1996). The s ocial system makes economic and political choices (via for example the evolution of cons umption patterns) which in¯ uence the development and spread of technologies, and which ± through education, training and general culture ± develop the skills needed to operate technologies. This approach has been applied to s uch disparate particular technologies as electric cars, miss ile guid ance systems, ¯ uorescent lighting and the S ONY discman. An important policy application drawing on aspects of the STS approach has been made by Bell and Callon (1994).

1.3.3. Approa c hes from Businesses Orga niza tion Studies a nd the Theory of the Firm . The pioneering work of Chandler in this ® eld has focus ed on the evolution of business organizations ± particularly large ® rms ± in terms of vertical integration and the s ys temic integration of intra-® rm functions (Chandler, 1962, 1977, 1990). Chandler analys es the growth of ® rms in terms of the ability of management to undertake major programmes of long-term inves tment in three areas : production, dis tribution and management itself. This work is important from a `s ystems’ pers pective in that it relates inter-country differences ± shaped by policy, ® nancial s ys tem s and governance regulations ± to major differences in economic performance. Perhaps more than the national s ys tems literature its elf, it draws the link between system differences at ® rm level, and differences in macroeconomic trajectories . Approaches developed from thes e ideas have explored thes e national performance differences explicitly: the work of William L azonick is particularly important here, both in challenging the notion of àrms -length’ market relations as a prevalent form of economic organization, and in relating actual ® rm and market organizations to economic performance (Elbaum and L azonick, 1986; L azonick, 1991).

1.3.4. The Regula tion Sc hool. The work of Boyer and his colleagues is particularly noteworthy for a systems approach based on the concept of `® lieÁre’ (Boyer, 1988). A ® lieÁ re is made up by a s peci® c set of infrastructures, technologies, institutions, practices and actors . Behind the notion of a ® lieÁ re is the idea that technologies are best unders tood not as individual techniques, but as integrated systems . T his view of the technology of a ® rm implies strong interdependence, because relevant technological knowledges are located in different ® rms , with interactions between ® rms in terms of technological capability. That is, the capabilities of any individual ® rm are shaped in part by its historical experience and its dynamic development of

80

K. Smith

competence, but als o by accompanying developments in related ® rms . The development of s pecialization, accompanied by inter- and intra-industry ¯ ows of technology, implies that we should think of the technological structure of an economy not as an agglomeration of independent micro-level decis ions, but as an integrated s ys tem shaped partly by the input-output relations between ® rms, and partly by intra-® rm specializaton of tasks .

1.3.5. Industrial Cluster Approa ches. Clos ely related to the ® lieÁre concept, thes e are analys es which explore the performance of industrial sectors in terms of the integration of different types of ® rms and ind ustries, s ometimes around key technologies , and which emphas ize environm ental conditions and inter-industry interactions in creating dynamic clusters or blocks of industry characterized by high growth of output, productivity and ± sometimes ± trade shares. The bes t-known example of this approach in recent years is the work of Porter (1990), but this approach actually has many antecedents ; in particular the work of DahmeÂ n (1970) on `d evelopment blocks’ and Hirschman (1958) on linkage effects. Although s uch work is strongly systemic in character, it is not necess arily focus ed on the speci® c dynamics of innovation and technology creation. The latter is taken up in what might be called `technological s ys tem’ approaches to the technological level its elf. Advanced-economy technologies do not exist as individual artefacts: they usually take the form of integrated technological systems , in which component elements are incorporated into overall s ys tems . F or such key technologies as cars , computers , and aircraft, but als o for a hos t of less spectacular products, there is in a sens e no uni® ed knowledge base at all: product producers are in effect s ys tem managers, whose competence relies primarily on the ability to s pecify and integrate diverse inputs. A recent major s tudy in this area, looking at the integration of electronics , adva nced manufacturing s ystems, and robotics in the context of ® rm interactions and academic infrastructures, is the work of Carlss on and his collaborators on the development and evolution of factory automation technology in S weden (Carls son, 1995). To sum up, we might argue that sys tems approaches have three bas ic conceptual underpinnings, and we can dis tinguish among the approaches accord ing to the emphas is which they place on the different underpinnings . They are: . the idea that economic behaviour res ts on ins titutional foundations, in the s ense of legally or customarily established `rules of the game’ which evolve because of the advantages they offer in reducing uncertainty. Different modes of institutional set-up lead to differences in economic behaviour and outcomes. . the idea that competitive advantage res ults from va riety and specialization, and that this has path-dependence-inducing effects. That is, successful specializations are self-replicating, with s ystem-creation as an outcome ± particularly around speci® c industrial structures. . the idea that technological knowledge is generated by interactive learning, and that technological knowledge in general takes the form of `dis tributed’ knowled ge bas es among different types of economic agents who must interact in some way if technologicl knowledge is to be applied. T his overview is a neces sarily brief one, and the bodies of literature which are relevant might well be extended , and might als o be categorized in different ways . T he general point here is that interactive `s ys tems’ approaches are not con® ned to


81

the national systems of innovation literature. On the contrary, they are a prevalent feature of research which relies not on ass umptions about the nature of technology but rather on empirical analys is of its real character, and which is therefore relevant to thinking about policy against the background of a systems framework. 1.4. Innov a tion a nd the Agenda of Gov ernm ent: Em pirica l Ev idence Although this is not the place for an empirical overview of s ystem s, it is important to take up one empirical point of evidence before turning to a discuss ion of policy foundations . In order for any particular approach to be cons idered seriously in policy term s, it is important that its practical relevance should be demonstrated. We have noted above that innovation systems approaches tend to be founded on two s trong hypothes es . T he ® rst is that innovation is a pervas ive phenomenon ± central rather than marginal to the operations of ® rms. The s econd is that interactions between ® rms, and between ® rms and other knowledge-producing agencies , are central to innovation performance. Both of thes e propos itions receive s ome support from available data on innovation. Data from the Community Innovation Survey (CIS ) of 1992 for four countries ± Denmark, the Netherlands, Germany and Norway ± shows that a s izeable proportion of ® rms have new products (introduced to the market within the past three years ) in their product mix, the number of new products increas ing with ® rm s ize (OECD, 1992). The ® gures imply rather rapid changes in product mixes in innovating ® rms. The data also shows that a s ubs tantial proportion of sales are coming from new products in all the countries examined and that, furthermore, innovation is not con® ned to `high-tech’ ind ustries by appears to be pervas ive across s ectors. Interaction among ® rms can only be examined in a preliminary way with CIS data, however the data is sugges tive. The CIS Survey as ked ® rms whether they undertook R&D collaboration, and about external s ources of information for innovation. S imply looking at technical R&D collaboration (which is of cours e likely to understate the general scope of collaboration) and as king whether collaborating ® rms are more innovative shows s ubs tantial differences across countries between co-operating and non co-operating ® rms . L ooking at differences across s ectors for one economy, Norway, co-operating ® rm s once again appeared to be considerably more innovative. F igure 1 indicates that cooperating ® rms dominate turnover among innovators in 13 of the 17 clas ses of industry surveyed. Although the data needs deeper examination and more formal testing procedures , there does appear to be an empirically-s upported cas e to be made for s ome of the major propositions of the systems approach. We turn now to a dis cus sion of the nature of the learning and knowledge creation process es which underpin innovation, and which are central to the policy issues.

2. The Na ture of Technologica l Knowledge If innovation is conceptualized as learning, then policy must concern its elf with the nature of learning and the knowledge which res ults from learning. In fact this is a core is sue for policy: as we s hall see, current rationales for policy are intimately bound up with as sumptions about the nature of technological knowledge. In this part of the paper, therefore, we addres sed the problem of unders tanding technological knowledge, and in particular exploring those characteris tics of

82

K. Smith

Source: NaÊ s and S mith (1994)

F igure 1. S hare of turnover from new products , Norway, ® rms with (n ˆ 165) and without (N ˆ 226) R&D cooperation, by indus try class (% ).

technological knowledge which imply: (a) the existence of `s ys tem s’ ; and (b) the need for new approaches to policies aimed at innovation. We begin with a critique of the ways in which production and technological knowledge are conceptualized in neo-class ical theory, and the approaches to policy which are derived from that theory. T hen we turn to ideas about technological knowledge drawn from modern innovation theory and analys is, and the ways in which this implies s ystemic relations between ® rms, and between ® rms and public ins titutions. The paper then moves to a discuss ion of policy iss ues . 2.1. Neo-Cla ssica l Produc tion Theory a nd Tec hnologic a l Knowledge a t Firm Lev el Insofar as innovation policies have had a theoretical rationale in the past, it derives from ideas within neo-clas sical prod uction theory concerning the nature of technological knowledge; these ideas have been powerfully in¯ uential in s tructuring views about the appropriate s cope, objectives and instruments of policy. Although they have been frequently criticized in the modern innovation literature, especially when it comes to policy discus sion (Metcalfe, 1994; Smith, 1991), it is worth looking in some detail at the debate here, since it has important implications both for the foundations of policy, as well as for the empirical operations of policymakers. Neo-clas sical production theory is built on the idea that ® rms face a dual prod uction decision. F irstly, they must decide what to produce. This decision is bas ed on rates of return: potentia l prod uct lines are known, and ® rms will allocate and reallocate capital among them in search of the highest returns . Then the problem is the choice of production technique: ® rms within an industry face a given and known array of production technologies and are as sumed to have the competence to operate all ava ilable production methods. Armed with this knowledge, and with a knowledge als o of present and future factor prices, and pres ent and future


83

product prices , ® rms can make a pro® t-maximizing choice of technique. In this context technology is seen as knowledge, and ® rms are able to acces s knowledge in a relatively rapid and costless way. With these types of underlying as sumptions, the technological dimens ions of production are clearly relatively unproblematic. The problem of technological change is also unproblematic, both with respect to adaptation to already-exis ting technologies , and to (exogenous ly-given) new technologies . This type of competitive theory res ts on the ideas of rapid subs titution poss ibilities across well-de® ned choice s ets in production. In this framework ® rms move smoothly to new production con® gurations in res pons e to changed environmental conditions . These environmental cond itions are factor, input and output prices , with the ® rm adjusting its technology (that is , adjusting its capital-labour ratio) in response to changed factor or input prices; when prices change the ® rm moves rapidly, perhaps instantaneously, to a new pos ition within the choice s et. But the environment also includes technology itself. Within growth theory, ris ing productivity follows ± in Solow’ s famous distinction ± from movements of the production function as well as movements along the production function. This implies that ® rms are adjusting ins tantaneously and optimally to changes in the choice s et its elf, although these changes are seen as exogenous to the system. In this type of approach, economic ef® ciency rests very much on ¯ exibility, both at economy-wide level (where free entry and exit to activities are central to allocative ef® ciency) and at ® rm level (where the ability to change the technical con® guration of production is central to pro® t maximization); these notions have had rather powerful policy effects. Both allocative and technological ef® ciency res t on freeing markets , removing barriers to entry (and not being too concerned about exit), removing barriers to change within the ® rm, and increasing competitive pres sures as a form of generating incentives to optimize. However both these types of adjus tment, and hence the policies which are built on them, res t on an implied form of technological knowledge with very particular characteristics. What exactly are the underlying as sumptions about the characteristics of technological knowledge? One of the key points about neo-class ical theory has been that although it deals with the economic characteristics of knowledge as a commodity, it does not contain any differentiated concept of knowledge itself. But of course it is pos sible to des cribe the characteris tics that knowledge pos sess es in the neo-class ical approach, even though they are implicit within the analysis . We could argue that in neo-class ical analyses, technological knowledge must have the following attributes if the production theory is to work: . It is generic. That is to say, an item of knowled ge, or a particular advance in knowledge, can be applied widely among ® rms and perhaps among industries . . It is codi® ed. Transmitability implies that knowledge is written or otherwis e recorded in fairly complete us eable form. . It is costless ly acces sible. On the one hand this can involve the idea that transmiss ion cos ts are negligible, but it can also mean that ® rms do not face differential cost barriers in accessing knowledge or bringing it into production. . It is context independent. That is, ® rms have equal capabilities in transform ing such knowledge into production capability. With these kinds of tacit as sumptions about knowledge, ® rms can readily make optimal pro® t-m aximizing choices . This is basically because, within this context, the

84

K. Smith

prod uction problem of the ® rm is ess entially a problem of calculation rather than a problem of technological capability and organization. This calculation activity has two further elements which are important to note. T he ® rs t is that if the feas ible industry technologies are available to all then the prod uction/innovation decision of any ® rm is independent of decisions made by others; interdependence or interaction between ® rms is simply not an is sue ± as Andersen (1992) has pointed out, neo-class ical analys is normally presupposes an extreme degree of ¯ exibility in the rela tionships of the economic sys tem which can only be founded in independence. Secondly, if market forces move rapidly to weed out ® rms who fail to make optimal choices, then there will es sentially be only one optimal way for ® rms within an industry to produce, and inter-® rm differences will be negligible or non-existent. However, if thes e kinds of as sumptions and analytical procedures make the acquis ition and operation of technologies unproblematic within equilibrium theory, they rais e acute problems when it comes to the development of technology, and in particular to the invention of new technological principles . Perhaps the most in¯ uential approach to bus iness -s ector R&D, and hence to policy, derives from two class ic papers by Nelson (1959) and Arrow (1962) res pectively. Although different in objectives, the papers have close analytical s imilarities. Both argue that technological knowledge has distinctive features, which lead business ® rms in a market economy to perform les s R&D than is socially optimal. Arrow begins by identifying technology with knowledge: technology in the most general sens e is `know-how’ , and therefore the proces s of invention ìs interpreted broadly as the production of knowledge’ . This ques tion then is, what are the technical and economic characteris tics of knowledge, and how do these characteristics affect the amount of new productive knowledge which ® rms might s eek to produce? The ® rst problem is that of uncertainty, which in this case means that knowled ge outputs are not predictable from inputs: producers must commit res ources to a knowledge production process without knowing the res ults with any accuracy. Arrow’ s ® rst point is that although market economies have a number of mechanisms for sharing risks ± s uch as ins urance, contingent markets, or equities ± these rarely apply to res earch activities. Ins urance, for example, would be impractical because it would weaken incentives to s ucceed; only the exis tence of large companies, with s izeable portfolios of relatively small projects , resolves this problem (because the companies act, in effect, as their own insurance bodies). Then there is the problem of appropria bility: it is dif® cult or even impos s ible to create a market for knowledge once it is produced, s o it is dif® cult for producers of knowledge to appropriate the bene® ts which ¯ ow from it. F irs tly, `there is a fundamental paradox in the determination of demand for information; its value for the purchaser is not known until he has the information, but then he has in effect acquired it without cos t’ . S econdly, any purchaser of the knowledge can in effect destroy the market, s ince he can reproduce the knowledge at very low, perhaps even zero, cost. If producers cannot appropriate the bene® ts of knowledge, then they have no incentive to produce it, and market economies will therefore under produce which would be s ocially bene® cial if it were produced. A ® nal characteris tic of technological information is indivisibility. That is , the underlying knowled ge must exist on a certain minimum scale before any production at all take place, and this necessary minimum is ind ependent of the rate of prod uction. A familiar example of such indivis ibility would be a railway, which


85

must be cons tructed in its entirety before any trains can use it; and it must be constructed whether it is used by one train per day or ® fty. The latter point m eans that there are scale economies in the use of indivisible capital goods , and this applies to technological inform ation. Thes e problems of risk, indivisibility and inappropria bility all sugges t that market economies will s ys tem atically underinvest in R&D, and this will, argues Arrow (1962), `lead to the conclusion that for optimal allocation to invention it would be neces sary for the government or some other agency not governed by pro® t-and-loss criteria to ® nance research and invention’ . But what kinds of knowledge really have the characteris tics described in Arrow’ s analysis? His approach points to a very narrow de® nition of knowled ge. The ® rst two characteristics clearly apply to the knowledge which res ults from fundamental s cienti® c res earch, but cannot be extended unconditionally to other forms of knowledge important to innovative activity. The other class ic s tatement of the externality problem, Nelson’ s 1959 paper, speaks speci® cally of basic science. Implicit in the Arrow approach is the idea that technological knowledge is the s ame kind of knowledge as bas ic science, perhaps, indeed, that it is s imply the application of bas ic science. This `m arket failure’ approach to knowled ge production leads to a relatively s imple s et of policy proposals. In this set-up the bas ic policy task is to encourage dis covery-oriented activities, and then to protect the use of the results. The problems of risk and indivisibility lead to straightforward under-provision of knowledge, and sugges t that the public sector should either produce knowledge directly, or provide s ubs idies to knowledge-producing institutions . The appropriability problem implies the existence of a major positive externality, and sugges ts policies either of s ubs idy, or the creation of property rights (via patents or other intellectual property protection). The bas ic problem with the approach is that it does not give any s ecure guide to how to identify areas of market failure, or the appropriate levels of public s upport which might follow from it. There appears to be a rationale for public provision, but where, and how much? It is worth noticing that this type of approach to innovation policy accords very well with what is sometimes called the `linear model’ of innovation, which is frequently contrasted with systems approaches (S oete and Arundel, 1993). This is the view that the process of innovation is es sentially a process of dis covery, in which new knowledge is transform ed into new products via a s et of ® xed (linear) s equence of activities or stages. T here is s ome debate about whether the term `linear model’ is really appropriate for characterizing S &T policies within OECD countries over the long term, but we can outline broad characteristics of a `linear’ approach, and this accords with many ideas and practices in post-war research policy. These characteristics are: . F irs t, the technological capabilities of a society are ess entially de® ned by the knowledge frontier; hence, technological adva nce depends on expans ion of the frontier by a knowledge creation proces s based on dis covery. . S econd, the knowledge which is relevant for industrial production is de® ned by principles which are es sentially scienti® c, and which have in some s ense been transferred, translated, or concretized from a more abs tract realm. . Third, the `trans lation’ proces s is bas ically sequential; there are temporally and institutionally dis crete phases in the translation proces s, and these have to occur in sequence.

86

K. Smith

. F ourth, the approach is technocratic, in the s ense that it views technological change broadly in terms of engineering development proces ses and hardware creation. However, more recent theory and applied research sugges ts that thes e characteristics are not a good guide to the nature of technological knowledge, and must therefore have limitations as a guide to the rationale and content of S&T policies. What are the alternative views , and what are their implications?

2.2. Tec hnologic a l Knowledge in the `New Growth Theory’ T he pres ence of knowledge externalities is als o a key component of the so-called `new growth theory’ . The new growth theory can be characterized as an attempt to integrate the S chumpeterian notion of endogenous knowledge creation into the formal modelling approach to growth pioneered by Solow. In these models , the basic process us ed to explain economic growth is the phenomenon of increas ing returns to scale, ¯ owing from the production of knowledge. Thes e increasing returns exis t because of an externality: knowledge ¯ ows to multiple users without being traded . F ollowing from S chumpeter, new growth theory sees knowledge as partly appropriable by the ® rm but als o, following Arrow’ s (1962) analysis , involving external bene® ts ¯ owing to all ® rms within an industry or line of business . F irms thus have s ome limited incentives to produce knowledge, but that there will als o be inter-® rm s pillovers which shape the s tock of knowledge for all ® rms. Romer (1986) models the production proces s of knowledge in optimization terms, as pro® t maximization over time. F irms generate new knowled ge in their R&D departments using a `research technology’ . They must forego other forms of prod uction today to generate knowledge that can be used to produce more tomorrow. Romer as sumes a s ingle `res earch technology’ to produce knowledge, and models knowledge as something which can be measured on a single continuous s cale, and which does not depreciate. The res earch technology is modelled as poss essing cons tant returns to scale. The growth rate of knowledge ava ilable to the ® rm is thus dependent on the amount inves ted in further knowled ge creation. However, a ® rm’ s ® nal output is not s imply depend ent on the accumulated s tock of its own knowledge: it also involves all (accumulated) spillovers from all knowledge production in the economy. Romer makes a rather s harp as sumption here that àll factors other than knowledge are in ® xed supply’ . If the labour s upply is not growing, land is in ® xed supply, and there is no increas e in the capital stock, inves tm ent behaviour is in effect con® ned to knowledge production. If the knowledge spillover component is signi® cant, then the wider effects of private investment decis ions in knowledge will shape the overall growth of output. What is being argued here is that savings behaviour, determining the investment rate, endogenously s hapes the path of output. What we have in this type of theory is the explicit introduction of knowled ge into growth theory with very little cons ideration of the real characteristics of the knowledge-creation proces s. We turn now to cons ider more extensively the realities of knowledge creation and its s ystemic features.


87

2.3. Lea rning a nd Tec hnologic a l Knowledge in M odern Innov a tion Theory The question of the nature of learning is approached here via cons ideration of the nature of technological knowledge, or more particularly the nature of the knowled ge bases which are generated by learning. Clearly all ® rms operate with some kind of technological knowledge base. Such knowledge bas es tend to be complex, in the sense that they involve the articulation of many elements. Here we dis tinguis h between three areas of production-relevant knowledge, namely ® rm-speci® c knowledge, s ector product-® eld speci® c knowledge, and generally applicable knowledge (Salter, 1969). Agains t this background we make three further distinctions, namely between the form of knowledge, the object of knowledge, and the institutional s tructure of production of knowledge. At the ® rm level, the ® rs t area of production s peci® c knowledge, the knowled ge base is highly localized and speci® c to product characteris tics. T he speci® c character of thes e knowledge bas es is not simply technical: it is als o s ocial, concerning the way in which technical proces ses can be integrated with skills , production routines , use of equipment, explicit or tacit training, management systems and s o on. In terms of the form of knowledge, the relevant technological knowledge base is likely to be informal and uncodi® ed, taking the form of skills speci® c to individuals or to groups of co-operating individuals. T he tacit and localized character of ® rm-level knowledge means that although individual ® rms may be highly competent in speci® c area, their competence has de® nite limits. T his means, ® rs tly, that they may eas ily run into problems in innovation which lie outside their area of competence, and s econdly that their ability to carry out search process es relevant to problems can also be limited; when creating technologies they must be able to access and use knowledge from outs ide the area of the ® rm. The key characteris tic of knowledge at this level is that it is bounded. F irms have a relatively restricted knowledge-base and a relatively res tricted set of technological capabilities: their technological performance at any point in time is shaped by their his tory, and by the niches which they have been able to occupy. Typically, they have a limited range of products and process es which they unders tand well, and where they can compete. The notion of bounded rationality ± limits to knowledge, cognition and therefore calculation at the level of individual economic actors ± deriving from the work of Herbert S imon, has played an important role in the evolutionary analys es of Richard Nelson and Sidney Winter (1982) which have in turn been central to m odern innovation analysis. The concept refers to two things: ® rs tly, that only a narrow band of technological s olutions or poss ibilities are known to any particular ® rm and that, s econdly, following from this there are cons traints in the ® rm’ s ability to calculate or compute the res ults of particular decis ions or choices. This helps to explain the strong uncertainty attached to decision-ma king. The idea of bounded knowledge is also useful in explaining the focused character of technological advance. F ransman (1990, p. 3) speaks of `bounded vision’ at the ® rm level: . . . The ® eld of vis ion of for-pro® t corporations is determined largely by their exis ting activities in factor and product markets , in production and in R&D and by their need in the short and medium term to generate s atisfactory pro® ts. The resulting bounded vis ion implies that new technologies emerging from neighbouring areas where the corporation does not have current activities are likely to take s ome time to penetrate the

88

K. Smith

corporation’ s ® eld of vision . . . The need to generate satis factory pro® ts in the s hort to medium term therefore further bounds the vision of the corporation, contributing in some case to a degree of `s hort-s ightedness ’ . One example is the creation of technologies for `the day after tomorrow’ where the degree of commercial uncertainty is frequently great. In view of their bounded vis ion, corporations often tend to underinves t in the creation of such technology. T here are als o knowledge-bases at the level of the industry or product-® eld. M odern innovation analys is emphas izes the fact that industries often share particular s cienti® c and technological parameters; there are shared intellectual unders tandings concerning the technical functions , performance characteristics, us e of materials and s o on of products. Nelson (1987, pp. 75± 76) calls this the `generic’ level of a technology: On the one hand a technology consists of a body of knowledge, which I shall call generic, in the form of a number of generalisations about how thinks work, key va riables in¯ uencing performance, the nature of the currently binding constraints and approaches to pushing these back, widely applicable problem-s olving heuris tics, etc. I have called this the `logy’ of technology . . . Generic knowledge tends to be codi® ed in applied scienti® c ® elds like electrical engineering, or materials science, or pharmacology, which are àbout’ technology. G eneric knowledge bas es are highly structured, and tend to evolve along structured trajectories (Dosi, 1982). This part of the indus trial knowledge bas e is public (not in the s ens e that it is produced by the public s ector, but public in the s ense that it is acces s ible knowled ge which in principle ava ilable to all ® rms ): it is a body of knowledge and practice which s hapes the performance of all ® rms in an industry. Of cours e this knowledge bas e does not exist in a vacuum. It is developed, maintained and dis seminated by ins titutions of va rious kinds , and it requires resources (often on a large scale). Tass ey (1991, p. 347) has de® ned the combination of knowledge and ins titutional bas e as the `technology infrastructure’ , in the following way: The tec hnology infra struc ture cons is ts of s cience, engineering and technological knowledge available to private industry. S uch knowledge can be embodied in human, ins titutional or facility forms . M ore s peci® cally, technology infrastructure includes generic technologies, infratechnologies , technical information, and res earch and tes t facilities, as well as less technically-explicit areas including information relevant for strategic planning and market development, forums for joint indus try-government planning and collaboration, and ass ignment of intellectual property rights . F inally, there are widely applicable knowledge bases, of which the most important technically is the general s cienti® c knowledge bas e. This is itself highly differentiated internally and of widely varying relevance for industrial production; but some ® elds ± such as molecular biology, solid-s tate phys ics, genetics or inorganic chem istry ± have close connections with major industrial s ectors . It is important not to overemphasize the role of scienti® c knowled ge in modern industrial development by presuming a one-way connection between science and technology: organized science does not evolve simply accord ing to some internal dynamic,


89

but is als o shaped by policy or funding decisions which usually have economic, industrial or military objectives . Thes e different types of knowledge bas e are not s eparate but integrated with one another, often in complex ways. Moreover they evolve over time: that is to s ay, technological knowledge tends not to result from generalized processes of search, but rather builds on pas t achievements. This gives an evolutionary character both to artefacts and knowledge, but it als o implies that knowledges are both s tructured and cumulative over time. The capabilities of any knowledge producing institution, at a point in time, tends therefore to be a product of its pas t his tory. This introduces a proces s of path dependence both into institutions and to the s ys tem as a whole. Against this background, forms of innovation and learning can be clas si® ed along three main dimensions . Thes e concern the following characteris tics of innovation: . F irs tly, the foc us of innovation in terms of the broad types of change which are sought or undertaken at any particular time. . S econdly, a dimens ion concerning the degree of c ha nge in the underlying knowledge bas es , and the extent to which knowledge bases evolve on the basis of exis ting capabilities or are fundam entally changed. . Thirdly, dimens ions concerning the m odes of lea rning through which innovation occurs. 2.4. Cha ra cteristic s a nd Form of Sea rc h Processes What are the bas ic cons traints and modes of s earch and learning proces ses? F irms, and other actors involved in technological change, do not generally have good knowledge either of all available technical pos sibilities, or of how to s earch for s olutions to technical problems . Rather, economic dynamics rest on highly differentiated search process es and technology development efforts by ® rms under conditions of s erious technological and economic uncertainty. Such technological learning process es result in the creation of a great deal of technological variety, and the s election of speci® c technologies by market and non-market process es . The s election proces s narrows down the overall amount of technological variety and leads to the cumulative development of technological knowledge along more or less well-de® ned `trajectories’ . Thes e highly incom plete forms of knowledge are usually specialized around areas of previous experience, and involve s ubs tantial tacit dimens ions. This has a number of implications . An abs olutely central point here is that tacit knowledge is pers on-embodied; any ® rm-level strategy for the development of knowledge must therefore be an employment strategy. The bounded and relatively limited character of ® rm-based knowledge means that ® rms seeks to innovate on the bas is of the cumulatively-developm ent knowledge which they already poss ess : they s eek to learn along the trajectory they are familiar with (K line and Rosenberg, 1986). This m eans that ® rms rarely us e s cienti® c discovery as the bas is of innovation, and depend heavily on the skills portfolio which they have built over time. But it als o implies that they are always likely to confront technological problems which lie outside their existing boundaries of competence. What is it which ® rms must learn and know about in order to innovate and s urvive? L undvall and Johnson (1994) de® ne relevant economic knowledge along four main dimens ions as follows:

90

K. Smith

Spec i® c fa ctua l inform a tion (`know-what’ ). T his type of knowledge tends to be relevant in s pecialized areas of expertis e, such as law and medicine, but it can be also vital to innovation activities : knowledge of relevant regulatory iss ues, for example. S uch knowledge can be s tored and supplied from outs ide the ® rm, via cons ulting companies, databas es, and s o on. Knowledge of ba sic sc ienti® c princ iples (`know-why’ ). Scienti® c knowledge is increas ingly important in the s olution of particular innovative efforts, both in terms of speci® c results and in terms of search heuristics. Although there is no direct connection between scienti® c capabilities and innovative performance, s uch knowledge is increas ingly important as a problemsolving input Spec i® c a nd selec tiv e social knowledge (`know-who’ ). In the context of human-embodied knowledge and interactive learning and interactive problem solving, access to key personnel is a new res ource. A range of studies have demonstrated the role of s uch human contacts in innovative earning (von Hippel, 1989). Pra c tica l skills a nd ca pa bilities (`know-how’ ). This covers skills , and all aspects of production capabilities and marketing. T hes e types of learning and knowledge mus t cover at leas t three dis tinct areas : technological competences and capabilities, organizational capabilities, and `s ys tem’ capabilities in terms of interactive links . 2.5. Degree of Cha nge in Underlying Knowledge Ba ses: From Inc rementa l Innov a tion to Cha nges in Tec hnologic a l Regime L earning implies change in knowledge bases, but the degree of s uch change may vary cons iderably over time. On the one hand, there are continual small changes in design, components, materials and so on, which are made on the basis of existing s kills , and which over a period add up to very fundamental changes in technical characteristics , productivity and so on. On the other hand, there are from time to time very radical changes in underlying knowledges bases , which force major changes both on ® rms, sectors and whole economies. These are s ometimes referred to as `paradigm shifts’ , a term which will be more extensively dis cuss ed below. Abernathy and Clark (1985) have distinguished a number of areas of s uch change, and the degrees of change which might be involved. They remark that it is important to note that the product features thems elves, and the ® rm’ s position within them, are not in and of thems elves the fundamental source of adva ntage. S uch a pos ition is the immediate, outward manifes tation of a more fundamental, internal reality. The foundation of a ® rm’ s position res ts on a set of material res ources, human skills and relationships, and relevant knowled ge. Thes e are the competencies or competitive ingredients from which the ® rm builds the product features that appeal to the m arketplace. Table 1 s hows their divis ion of competitive components , along a s cale showing the potential range of impacts of innovation. In term s of public policy, the most important of these learning effects are probably those on the radical dimension: instead of enhancing and strengthening, innovation of this sort disrupts and destroys . It changes the technology of process and product in a way that imposes requirements that the exis ting res ources , s kills and knowl-


91

Table 1. Innovation and changes in ® rm competence Domain of innovative activity

Range of impact of innovation

Technology/production Design/embodiment of technology Production systems/organization

improves/perfects established design strengthens existing structure

S kills (labour, managerial, technical) extends variability of existing skills M aterials/supplier relations reinforces application of current materials/ suppliers Capital equipment extends existing capital K nowledge and experience base

builds on and reinforces applicability of existing knowledge

$ offers new design/radical departure from past embodiment $ makes existing structure obsolete demands new system, procedures, organization $ destroys values of existing expertise $ extens ive material substitution; opening new relations with new vendors $ extens ive replacement of existing capital with types of equipment $ establishes links to whole new scienti® c discipline/destroys value of existing knowledge base

Market/cus tomer Relationship with customer base

strengthens ties with established customers Customer applications improves service in established application Channels of distribution and service builds on and enhances the effectiveness of established distribution network/service organization Customer knowledge uses and extends customer knowledge and experience in established product

$ attracts extensive new customer group/creates new market $ creates new set of applications/ new set of cus tomer needs $ requires new channels of distribution/new services, after market s upport

$ intensive new knowledge demand of customer; destroys va lue of customer experience totally new modes of M odes of customer communication reinforce existing modes/ $ communication required (e.g. ® eld methods of communication sales engineers)

edge satis fy poorly or not at all. The effect is to reduce exis ting competence, and in the extreme case to render it obsolete. (Abernathy and Clark, 1985, p. 6) Such change is usually unders tood through the concept of change in a `technological paradigm’ of `technological regime’ . As noted above, the basis of this notion is that just as ® rms have a de® nite set of technological capabilities , so do industries . As Dosi (1988b, p. 1128) puts it, à crucial implication of the general paradigmatic form of technological knowledge is that innovative activities are s trongly selective, ® nalized in quite precise directions , cum ulative in the acquisition of problem-s olving activities . This accounts for the relatively ordered patterns of innovation’ (Dosi, 1988b). The concept of technological paradigm refers to the comm on features of the technology shared by the ® rms in an industry, or even a group of industries. Most ® rms in an industry share a core s et of approaches to the technological problems they face (although this does not de® ne a unique approach to solving those problems ). This idea of a core technological framework for industries has been widely used in modern innovation theory. On the one hand, this shared framework can be

92

K. Smith

thought of in terms of s hared knowledge, and a s hared understanding of problems . It is this approach which underlies Dosi’ s concept of `technological paradigm ’ . However, for reasons which are developed later, we prefer to use the notion of `technological regime’ because we want to go beyond an approach which focus es on s hared ideas , towards one which places more emphasis on real technological and economic factors. In a s tudy of Georghiou et a l. (1986) on pos t-innovation improvements and competition the concept of a technological regime is also us ed. A technology regime is de® ned as a s et of des ign parameters which embody the principles which will generate both the phys ical con® guration of the product and the process and materials from which it is to be cons tructed. The basic des ign parameters are the heart of the technological regime, and they constitute a framework of knowledge which is shared by the ® rms in the industry. In a s imilar way, Sahal (1985) s peaks of `technological guidepos ts’ charting the cours e of innovation activities and ìnnovation avenues’ that designate pathways of technological evolution. We would emphas ize that the ways in which a technology can develop are cons trained: cons trained by more than the consens us of engineering ideas about how to approach problems . They are constrained by ava ilable method s and techniques, by the speci® c characteris tics of technologies, by patterns of infrastructure and consumer demand, and s o on. Accordingly we de® ne a tec hnologic a l regim e in the following way: A technological regime is the overall complex of scienti® c knowledges , engineering practices, production proces s technologies, product characteris tics, s kills and procedures , ins titutions and infrastructure which make up the totality of a technology. When it comes to dynamics , most technological change theoris ts use the concept of a `technological trajectory’ to describe the continuous process of change along typical paths. T hes e trajectories can be de® ned as the activity of technological progres s limited by the economic and technological trade-offs born of the technological paradigm or regime (Dosi, 1988). Shifts in technological regim es clearly involve both major shifts in underlying technological knowledge, but also a wide range of managerial, organizational and even s ocial changes . A more fruitful line of approach is the F reeman and Perez concept of techno-economic paradigm, which refers . . . to a combination of interrelated product and proces s , technical, organizational and managerial innovations , embodying a quantum jump in potential prod uctivity for all or more of the economy and opening up an unusually wide range of investment and pro® t opportunities . S uch a paradigm change implies a unique new combination of decisive technical and economic adva ntages. Such change is by no means rare, although these forms of very radical change are often m uch slower than is commonly repres ented. However, they do involve fundamental changes in learning modes and functions and are usually as sociated with major changes also in the organization and management of ® rms. 2.6. M odes of Lea rning: From Indiv idua l to Intera ctiv e to System ic It has been argued above that innovation s hould not be seen in terms of ind ividual acts of learning or discovery. Recent theory ins tead s ees such learning as interactive. Within a ® rm, for example, success ful innovation results from multi-directional feedbacks between the various forms of competence and skill on which a business relies: between marketing, ® nance, and product-proces s development, for ins tance. Innovation is thus a process of interactive knowledge-creation, in which, for


93

example, s kills in marketing are used to channel relevant information about user needs into the development process es which shape the technical and performance attributes of products . In this context, innovation is far from being simply the `transfer’ of knowledge which has been developed els ewhere. A second interactive dimension follows from the boundedness of ® rms ’ knowledge bas es , referred to above. Succes sful innovative ® rms are usually those which are open to their environments. That is, they engage in interactive learning involving other institutions: partners , rivals , and a wide range of other knowledge-creating and knowledge-holding institutions . The ability of ® rms to engage in the interactive learning described above is s haped by the structure, types , scale and communication proces s es between relevant knowledge-creating or knowledge-holding ins titutions. Such institutions have a tangible location in speci® c regional or national s paces , and are thus shaped by regional or national political cultures , legal s ystems, modes of corporate governance, or wider social values. Among the most important of these ins titutions are those which create or maintain scienti® c knowledge. These institutional underpinnings lead to the idea that innovation involves s ys temic interactions. Hence we have the idea of ìnnovation sys tems’ as the bas ic context for innovation performance. To s ummarize this dis cus sion, we can argue that both the learning processes and the knowledge bas es of industrial ® rms have the following characteris tics: . They are differentia ted, multi-layered, involving the s ystemic integration of many different types of knowledge. . They are highly s peci® c, organized around a relatively limited set of functions which ® rms unders tand well, and the system is thus characterized by boundedness , bounded rationality and `bounded vis ion’ . . They involve s igni® cant tacit com ponents , embodied in the s kills of engineers , R&D staff, workers and managers . . They are cum ulative, developing through times as ® rms build up experience with particular technologies; this in turn implies that technological knowledge is path dependant. . They are developed through costly process es of s earch, through processes of learning and adaptation. . They are internally systemic in the sense of being part of an overall production and marketing s ys tem which has many components . . They are externally systemic, relying on interactions between ® rms and other agents , and relying also on infras tructural support. A primary foundation of sys tem phenomena therefore lies in the nature of technological knowledge its elf. This view of learning and knowledge is radically different from that of the neo-clas s ical approach, both in the long-s tanding Arrow formulation, and the newer formulations found in the new growth theory. We turn now to a dis cus sion of the implications for public policy.

3. P olicy Is s ues We now turn to s ome of the policy issues arising from these complexities concerning the nature of innovation and underlying learing proces ses . The discussion here focuses on the implications of systems approaches for two broad is s ues:

94

K. Smith

. The ra tiona le for polic y a c tion What is the underlying justi® cation for policy intervention, and do these justi® cations throw any light of the general scope, objectives and methods of policy? . Polic y ca pa bilities In a sys tem context, what competences , skills and resources do policy-makers need; to what extend do thes e differ from current views ? 3.1. The Ra tiona le(s) for Policy Action T he standard rationale for policy action with respect to learning and innovation follows from the m arket failure analys is of Arrow (1962). It simply argues that a completely competitive, decentralized market system will provide a sub-optimal level of knowledge, and that this leads to a case for either public subs idies to knowledge creation, or to creation of intellectual property rights. T his links up with `linear model’ approaches, and leads in practice to policies consis ting of s ubsidies to R&D (although the market failure approach is particularly weak in identifying where those s ubs idies s hould go, and what their level should be). Systems approaches would not necess arily drop such policies: as noted above, they certainly recognize the existence of generic knowledge bases , and would make provis ion for the s upply of non-appropria ble generic knowledge. In addition, as we s hall indicate below, systems approaches have a greater potential for identifying where public support s hould go. The most important distinction between the s tandard rationale for policy intervention and the s ystems approach is that marketbas ed sys tems not only suffer from an under-s upply of knowledge, but are likely to actually generate areas of s ys tem atically weak performance. These areas of `s ystemic failure’ may call for actions contrary to conditions of perfect competition, for example, cooperation and collaboration between ® rms to facilitate knowled ge ¯ ows, government regulation and the creation of incentives . Areas of systemic failure include: . . . .

failures in infras tructural provis ion and inves tment; `transition failures ’ ; lock-in failures; and institutional failures.

3.2. Fa ilures in Infra struc tura l Prov ision a nd Inv estm ent We have emphas ized above that s ys tems theories often stress the importance of infrastructures, a concept very prevalent in the systems literature, but more or less abs ent from the neo-classical m ainstream . Two types of interaction between ® rms and infras tructures s eem to be important: ® rstly, with physical infrastructures us ually related to energy and communications, and second ly with s ciencetechnology infrastructures s uch as universities , publicly-supported technical ins titutes , regulatory agencies , libraries and databanks , or even government ministries. The author has argued els ewhere that thes e infras tructures have a number of s peci® c technical characteristics which lead to serious problems of inves tment apprais al (S mith, 2000). T he features include very large s cale, indivis ibilities, and very long time horizons of operation; they lead to major problems in the ® nancing of infrastructural inves tment, and are very unlikely to produce adequate returns within the context of s tandard ROI inves tment appraisal techniques. T his is a s erious problem, since virtually all s tudies of major technology creation, or of the nature of


95

industry knowledge bases , indicate an important role for knowledge developed within the kinds of infras tructure developed above. Thes e problems indicate a role for public sector s upports , for which there are three bas ic modes ± regulation to set up incentives and controls for private provision; subsidies to private provis ion; or direct public provision. T his entire area is problematic at the present time, s ince increas ing pres sures on public expenditure have led to s trategies of privatization and/or marketization which have serious implications for infrastructural operations. F rom a s ys tem s pers pective, the problem of infras tructural investment failure becomes a signi® cant jus ti® cation for public s ector actions . 3.3. `Tra nsition Fa ilures’ Trans itions and dynamics are an important part of any innovation-bas ed theory of the economic process , and this is especially true of s ystems theories . At the s ame time the notions of ® rm-level knowledge and learning underlying systems theories imply s erious problems for ® rms and sectors in adpating to transitions It has been emphasized that in adjusting to technological change an important consideration is the fact that ® rms , especially s mall ® rms , are necessarily quite limited in their technological horizons . F irms almost always concentrate on what they know best: they focus on products and technologies where they have experience and s kills, and they try to bring a high level of expertis e to the technologies which exploit those s kills. This produces a situation in which ® rms have strong competence within their area of technological knowledge, but relatively limited capabilities even in clos ely related areas. Three kinds of problems arise as a res ult: F irs tly, even in the norm al cours e of innovative activity it is almost certain that ® rms will frequently encounter technological problems outside their existing capabilities. S econdly, there may be changes in technological opportunities or patterns of demand which push the market into new areas of technology: that is to s ay, there may be dis continuous shifts in technology. T here is considerable evidence to sugges t that even relatively minor s hifts can provide serious problems for ® rms who have no background in the new technology. This is particularly a problem for s mall economies which poss es s relatively s mall numbers of players in many sectors; relatively minor discontinuous s hifts can provoke major changes in the industrial structure. F inally, there can be major shifts in technological regimes or paradigms. T hes e transitions can be particularly dif® cult since they often imply development of or adaptation to completely new generic technologies , where the relevant capabilities (which are usually not technical but organizational) lie quite outside the exis ting structure of capabilities . There are thus likely to be what we might call transition failures; many public policies are in fact aimed at thes e iss ues , frequently without any explicit rationale. This rationale s hould be made more explicit, since it has important implications for policy capabilities and objectives . 3.4. Loc k-in Fa ilures A strong feature of s ystems theories is the notion of path dependence, or `lock-in’ to existing technologies. One of the bas ic reas ons for path dependence is the exis tence of s ys tem or network externalities, combined with the fact that technologies exist in close links with their s ocial and economic environment. This means that

96

K. Smith

technological alternaives mus t com pete not only with components of an existing technology, but with the overall system in which it is embedded. Technological regim es or paradigms pers ist because they are a complex of s cienti® c knowledge, engineering practices, process technologies , infrastructure, product characterisics , s kills and procedures which m ake up the totality of a technology and which are exceptionally dif® cult to change in their entirety. Just as ® rms are not able to switch away from their existing technologies , so ind ustries and indeed the whole s ocio-economic sys tem can be `locked-in’ to a particular technological paradigm. It is very unlikely that movement away from s uch a paradigm can be induced by, for example, tax policies on a particular input. The elements of a technological paradigm interlock with each other, and with a s ocial and technical infras tructure built up over a long period. A change in a technoeconomic paradigm must involve a complex and integrated proces s of change in s cience, engineering practice, phys ical infrastructure, s ocial organization, plant design and so on. This does not mean that regimes never change, but it does lead to s erious problems of lock-in. Perhaps the most important case at the pres ent time concerns the role of the hydrocarbon-based energy sys tem in greenhouse gas emis sions and global warming. The problem lies ® rs tly in the ubiquity of thes e technologies : hydrocarbon-using energy technologies are used as inputs to virtually all economic and indeed social activities throughout the world. Secondly there is the complexity of the relevant energy technologies. We are not discussing here a s ingle technology, but rather an exceptionally complex system of integrated technologies for the prod uction, distribution and use of energy. The ques tion then aris es : how can technological change happen? The ques tion of systemic change seems central to any transition away from hydrocarbon-bas ed energy technologies, yet the system is unlikely to be changed by such initiatives as carbon taxes , or the developm ent of individual alternative techniques. That is to s ay, actions at the level of individual agents are unlikely to overcome lock-in. External agencies , with powers to generate incentives , to develop technological alternatives , and to nurture emerging technological s ystems are required. This is therefore an important rationale for public action in a s ys tems context; it is by no means a rationale which is likely to be frequently used, but on the occas ions it is relevant it is likely to be of exceptional importance. 3.5. Institutiona l Fa ilures S ystems approaches emphasize the ins titutional context as a de® ning and structuring element in the system. A key aspect of this is the framework of regulation: at a national level this involves technical standards, ris k-management rules, health and s afety regulations, and s o on. The regulatory system also includes the general legal s ys tem relating to contracts , employment, intellectual property rights (patent and copyright law) within which ® rms operate. F inally there is the wider context of political culture and s ocial va lues , which s hapes public policy objectives and particularly the macroeconomic policy environment. T aken together, this integrated s et of public and private ins titutions, regulatory systems and the policy sys tem makes up an overall context of economic and technical behaviour which shapes the technological opportunities and capabilities of ® rms. It thereby shapes ® rms’ economic performance and the macroeconomic evolution of the economy as a whole.


97

Whether or not thes e institutions/regulatory process es develop through cons cious choice or through the evolution of cooperation, they are inva riably discus sed and implemented through policy agencies. The operation of this system ought to be itself a respons ibility of policy makers. An example is the operation of m arkets for corporate control, which are shaped by a major part of the regulatory environment, namely the corporate governance s ys tem. Given that corporate take-overs appear to have impacts on overall R&D performance, they als o impact on the innovation s ys tem . T here are powerful arguments to sugges t that regulatory differences indeed s hape innovation and economic performance. The need for monitoring and ass ess ment of regulatory performance, and if necess ary changes in regulatory s ys tem s, provides a rationale therefore public action. 3.6. Policy Ca pa bilities a nd Knowledge Ba ses What competences or capabilities must policy-makers poss ess, what do they need to know, if they are to develop and implement policy actions within the overall framework suggested in this paper? Operating policies within a s ys tem framework would seem to imply new demands for knowledge and ass ess ment for policy-makers themselves . Here we look at ® ve problem areas : . . . . .

the ass essment of system speci® cities; unders tanding of relevant knowledge bas es ; as sess ment of s ystem dynamics; system co-ordination; and identi® cation of untraded knowledge ¯ ows.

3.7. Identifying System Spec i® c ities It follows from the analysis pres ented above that s ystems are likely to be quite s peci® c in term s of ins titutional frameworks, industrial structures and technological bases. Empirical analys is s eems to support this , both at national levels and regional levels : technological specialization is a pervasive phenomenon, and wide va riations exists between s ystems. Archibugi and Pianta’ s work is especially important with res pect to national s ystems; they show the existence of s ubs tantial specialization patterns with res pect to R&D investments , technological ® elds of patent grants , and s cienti® c publication (Archibugi and Pianta, 1992). The CIS survey has clearly s uggested major ìnnovation structure’ differences acros s Europe. We know also that there are s igni® cant institutional differences with res pect to governance of ® rms and sectors (Hollings worth, 1994). F inally, the regional literature also shows that s pecialization and diversity are pervas ive at regional levels Ð according to S torper and S cott (1995, p. 513) à new ``heterodox’ ’ economic policy framework has emerged in which signi® cant dimens ions of economic policy at large are being reformulated in term s of regional policies ’ . This is partly the result of the economic s ucces s stories of teritorially agglomerated clus ters of SM Es and partly the res ult of the new political initiatives towards a È urope of regions ’ . Carlss on and S tankiewicz (1991, p. 115) have gone s o far as to argue that s ometimes it seems more accurate to refer to a regional technological sys tem (in their words) than to a national one às high technological density and diversity are properties of regions rather than countries. They are the res ults of local agglomeration of industrial, technological and s cienti® c activities ’ .

98

K. Smith

Thes e speci® cities m ust be taken account of in policy design. T heir existence s eems to sugges t that `neutral’ policies are likely to be so abs tract that they have little effect within the distinct s tructures of speci® c systems. Policies therefore should be designed with sys tem s peci® cities in mind . This impos es fairly s ubs tantial analytical demands (both s tatis tical and otherwis e), regardless of whether policies are being made at central level or at regional level. There are a range of attempts to develop this kind of s ys tem knowledge bas e at the pres ent time, for a number of countries, mainly through the OECD’ s Na tiona l Innov a tion System s project, which is an attempt to produce quantitative analys es of s ystem s which go beyond a primary focus on the R&D s ystem (Chaminade, 1995; den Hertog et a l., 1995; L aurs en and Chris tens en, 1996; Numminen, 1996; Smith and Nas , 1995). Certainly there are many methodological problems here, but pos sibilities clearly exis t to improve knowledge bases for sys tems structures. 3.8. Identifying a nd M a pping the Tec hnologic a l Knowledge Ba ses of System s T here is clearly a need for policy-relevant knowledge concerning the broad knowledge inputs which are relevant for a s ystem. Central to the systems approach is the view that the key res ource of a ® rm, or an industry, is the knowledge ba se from which it draws its competence in re® ning, developing, creating and s elling new products. K nowledge in s peci® c areas underpins the capabilities and speci® c competences of the ® rm, and may hence be seen as the competitive bas is of ® rms and industries and one of the major factors that creates ® rm differences and hence ® rm competitiveness. As well as diversity between systems , there is considerable structural differentiation within systems . Policies in support of innovation and technological change need rather precis e identi® cation of how system knowledge bas es are actually cons tructed. This is not at all a s imple matter: even the knowledge bas es of apparently simple industries res t on the articulation of quite different knowledges , and interactions between quite different ins titutional forms. Apparently s imple ind ustries such as ® shing or timber products can have very complex s cienti® c knowledges underlying them, and an important part of the policy problem is to identify thes e knowledges and provide s upport for their development: many s uch knowledge ® elds (s uch as wave dynamics or GPS in ® shing, or algorithms for optimal cuts in timber) ful® l clas sic criteria for public provis ion ± they are codi® ed, non-appropria ble and have public-good properties . How can these interactive knowledge bases be mapped and unders tood by policy-makers? It seems in principle pos sible to map the complexity of s ectoral knowledge bas es by identifying the following: (1) The key activities in the industry and key pers onnel performing these kinds of activities . What are the main technical components of production activity within the sector concerned? What must a ® rm do to be a viable operator in the industry? (2) The key techniques ± meaning capital inputs, equipment, ins truments and production routines ± being utilized to perform these activities. What are the techniques which the ® rm must master in ord er to be able to undertake the activities des cribed above? (3) The knowledge bases ± in terms of engineering and s cienti® c knowledges ± s upporting thes e techniques. What are the codi® ed knowledges with which the technical operations are des igned, analys ed, and produced ? Note of course that


99

the knowledge bases behind capital and intermedia te inputs are likely to be cons iderably more complex than those which are produced directly within most industries. (4) What are the organizational forms ± in terms of companies , research institutes, univers ities and so on ± through which thes e knowled ges are produced and diss eminated? Concretely, who develops the relevant knowledge inputs, and on what res ource bas is? Here are identifying the level at which knowledge is shared among ® rms , or is common to them; that is , the level which policy can addres s. Although there also exists a dimens ion of knowledge which is ® rm speci® c , there is clearly a need for an approach of this type, which identi® es both the direct and indirect knowledge inputs for a sector in a s peci® c country or region, an approach which is necess ary s ince policy makers are likely to need a much enhahnced view of both the depth and complexity of knowledge bases us ed in s ectors and activities which are relevant in their economies (Balconi, 1993). 3.9. Assessing System Dyna m ics S ys tems are dynamic; it might even be claimed that they are turbulent. We have already indicated one dimension of this above, nam ely pervasive product innovation, but various other quantitative indicators could be mentioned. We have also outlined a number of potential `s ys temic failures’ which call for policy intervention. `L ock-in’ failures, for example, imply a role for policy in adapting to or generating s hifts in technological regimes, the question is, how can we identify when such shifts are happening? When s hould policy-makers be attempting to support the existing s ys tem ± with its historically accumulated adva ntages ± and when s hould they be helping to create a new system? S imilarly, how can transitions which may be occurring anyway be identi® ed within the `noise’ of the normal dynamics of the s ys tem ? It is clearly dif® cult to answer thes e questions in a practical way, but solving them is an important part of generating knowledge bases for systems-oriented policies. 3.10. System Co-ordina tion A key policy issue arising from sys tems approaches is the need to identify and perhaps support nodal points in the creation and dis tribution s ys tems, keeping in mind that thes e are likely to be changing over time: the innovation sys tem is not a s tructure, but a dynamic proces s. At the s imples t level, the task would be to identify key points or functions within the s ystem where public support would improve the overall dis tribution capability. S ince knowled ge systems are complex in practice (even in small s ocieties), and usually managed by quite separate ins titutions , there is a need for policy co-ordination and for adequate information systems to ensure that s uch co-ordination is possible. Actually, this problem is pres ent even when a linear approach to policy is adopted. Although such policies are fundam entally dis coveryoriented, they tend in practice also to involve other elements: to combine bas ic res earch policies with policies aimed at developing commercial applications, at diffus ion, at training, and so on. The general ques tion of policy co-ordination is a long-standing one in all advanced economies, and more or less unres olved. A s ys tem s approach s uggests that developing information sys tems for policy coordination is a core priority.

100

K. Smith

3.11. Identifying Externa lities: Untra ded Flows of Knowledge T he bas ic rationale for public policy is that there are necess ary activities and functions which are insuf® ciently ful® lled by private initiative: this usually implies an externality, and much public policy is concerned with externalities. The characteristics of knowledge described above have a number of implications for the externality question. On the one hand they imply that the externality effects emphas ized by Arrow are not a s igni® cant obs tacle to the production of knowled ge in that knowledge is not always cheap to transmit or easily appropriated (Carls son and Jacobs en, 1993). On the other hand, the knowledge s ys tem approach emphas izes a wide range of interactions , some of which will take the form of non-traded ¯ ows of economically useful knowledge. There are potentia lly large externalities , the identi® cation of which might be central to policy formation and operation. What forms can such externalities take? Given the general characteris tics of industrys peci® c and ® rm-s peci® c knowledge bases sketched above, we can sugges t a range of forms of external knowledge. Thes e certainly include generic `public domain’ s ources of scienti® c technological information, but they could also include knowledge from other ® rms in an industry (through marketing relationships, co-operative knowledge exchange, trade literature etc.); acquis ition of skilled personnel; acquis ition of process technologies; regulations and standards, and so on. Any identi® cation of externalities requires some form of overall system mapping, with particular reference to ìntangible’ knowledge ¯ ows.

4. Conclus ion As we suggested in the ® rs t s ection of this paper, sys temic approaches have established a powerful case for thinking about economic performance in terms of `s ys tems’ . However, at this stage discus sion of policy in the context of innovation s ys tems can only be conducted at a rather abstract and general level. T here is nothing particularly coherent about s ys tems approaches at the pres ent time, and there remains much to be done in re® ning the theory and giving it empirical relevance. Even then, it is in the nature of s ys temic approaches that the details of policy must va ry widely to suit particular national, regional, and local needs. Having s aid this, it remains clear that public policies have had a central role in the evolution of national sys tems, both in development of underlying knowledge bas es , and in provis ion of the phys ical and knowledge infras tructures on which technological regim es rest. It is also clear that observation and analysis of the dynamics of particular s ystems will be neces sary to identify the places for policy action. Conceptualizing the appropria te rationales for and tas ks of policy are precursors to the much more demanding ques tion of policy ins truments and implementation methods.

Refer ences Abernathy, W. and Clark, K . (1985) Innovation: mapping the winds of creative destruction, Researc h Policy, 14(1), pp. 3± 22. Andersen, E. and L undvall, B.-AÊ . (1987) Small national s ystems of innovation facing technological Ê . L undvall and C. F reeman (eds) Sm a ll Countr ies Fa cing th e revolution: an analytical framework, in B.-A Tec hnologic a l Rev olution (L ondon: Pinter).


101

Andersen, E.S . (1992) Approaching national systems of innovation from the production and linkage Ê . L undvall (ed.) Nationa l System s of Innov a tion. Towa rds a Theory of Innova tion a nd structure, in B.-A Intera c tiv e Lea rning (L ondon: Pinter) pp. 68± 92. Archibugi, D. and Pianta, M. (1992)The Tec hnologica l Specia liza tion of the Adv a nced Countr ies, (Dordrecht/Boston/L ondon: Kluwer). Arrow, K . (1962) Economic welfare and the allocation of resources for invention, in R. Nelson (ed.) The Ra te a nd Direction of Inv entiv e Ac tiv ity (Princeton: Princeton University Press) pp. 609± 625. Balconi, M. (1993) The notion of industry and knowledge bases: The evidence of steel and minimills, Industria l a nd Corpora te Cha nge, 2, (3), Bell, G. and Callon, M. (1994) Techno-economic networks and science and technology policy, STI Rev iew, 14, pp. 59± 118. Carlsson, R. (1995) Tec hnologica l Systems a nd Econom ic Perform a nce: The Ca se of Fa c tory Autom a tion (Dordrecht: K luwer). Carlsson, B. and Jacobson, S . (1993) Technological sys tems and economic performance: the diffusion of factory automation in Sweden, in D. F oray and C. F reeman (eds) Technology a nd the Wea lth of Nations. The Dyna mic s of Constructed Adv a nta ge (L ondon: Pinter). Carlsson, B. and S tankiewicz, R. (1991) On the nature, function and composition of technological systems, J ourna l of Ev olutiona ry Ec onomics, 1. Chaminade, C. (1995) Assessing the Distribution Power of Nationa l System s of Innov a tion: Proposed Indica tors for th e New Inform a tion Tec hnologies Industries in Spa in (Madrid: Universidad Autonoma De Madrid). Chandler, A.D. (1962) Stra tegy a nd Structure: Cha pters in the History of America n Industria l Enterprise (Cambridge, MA.) Chandler, A.D. (1977) The Visible Ha nd: The M a na geria l Rev olution in Am erica n Business (Cambridge, MA: Bellknop Press). Chandler, A.D. (1990) Sca le a nd Scope. The Dyna m ics of Industria l Ca pita lism (Cambridge, MA). DahmeÂn, E. (1970) Entrepreneuria l Activ ity a nd the Dev elopment of Swedish Industry 1919± 1939 (Illinois: Irwin). David , P. and F oray, S. (1995) Assess ing and expanding the science and technology knowledge base, STI Rev iew, 16, pp. 13± 68. Den hertog, P., Roelandt, T.J.A., Boekholt, P. and Van Der Gaag, H. (1995) Assessing th e Distribution Power of Nationa l Innov a tion Systems Pilot Study: the Netherla nds (Apeldoorn: TNO Centre for Technology and Policy Studies). Dosi, G . (1982) Technological paradigms and technological trajectories, Resea rc h Policy, 11(3), pp. 147± 162. Dosi, G. (1988) The nature of the innovative process, in G. Dosi, C. F reeman, R. Nelson, G. S ilverberg and L . S oete (eds) Tec hnic a l Cha nge a nd Econom ic Theory (L ondon: Pinter). Dowrick, S . (1991) Technological catch-up and diverging incomes: patterns of economic growth 1960± 1988, Ec onomic J ournal, 102(412), pp. 600± 610. Edquist, C. (1997) Introduction: systems of innovation approachesÐ their emergence and characteristics, in C. E dquist (ed.) System s of Innov a tion ± Technologies, Institutions a nd Orga niza tions (L ondon: Pinter). Elbaum, B. and L azonick, W. (1986) The decline of the British Economy (Clarendon Press: Oxford). Boyer, R. (1988) Technical change and the theory or `ReÂgulation’ , in Dosi, G. (ed.) Tec hnic a l Cha nge a nd Ec onomic Theory (L ondon: Pinter). F ransman, M. (1990) The M a rket a nd Beyond. Coopera tion a nd Competition in Informa tion Technology in the J a pa nese System (Cambridge: CUP). F reeman, C. (1994) The economics of technical change, Ca m bridge J ournal of Ec onomics, 18, pp. 463± 514. F reeman, C. (1988) Japan: a new national system of innovation, in G . Dosi (ed.) Tec hnic a l Cha nge a nd Ec onom ic Theory (L ondon: Pinter) pp. 330± 348. F reeman, C. (1987) Technology Polic y a nd Ec onom ic Performa nce: Lessons From J a pan (L ondon, Pinter). G eorghiou, L ., Metcalfe, J.S ., Gibbons, M ., Ray, T. and Evans , J. (1986) Post-Innova tion Perform a nce: Tec hnologic a l Dev elopment a nd Competition (L ondon: MacM illan). G ille, B. (1978) Histoire des Tec hniques (Paris: G allimard). Hirschman, A. (1958) The Stra tegy of Ec onomic Dev elopm ent (New Haven: Yale UP). Hirst, P. and Thompson, G. (1996) Globa liza tion in Question (Polity Press: Cambridge). Hollingsworth, J. Gov erning Ca pita list Econom ies. Performa nce a nd Contr ol of Econom ic Sectors (Oxford: OUP). Jasanoff, S. (1995) Ha ndbook of Scienc e a nd Technology Studies (L ondon: Sage). K line, S . and Ros enberg, N. (1986) An overview of innovation, in R. L andau et a l. (eds) The Positiv e Sum Stra tegy (Washington: National Academy Press).

102

K. Smith

L aursen, K. and Christensen, J.L . (1996) The Creation, Distribution a nd Use of Knowledge (Aalborg: University of Aalborg). L azonick, W. (1991) Business Orga niza tion a nd the M yth of th e M a rket Ec onomy (Cambridge: CUP). L azonick, W. (1991) Business organization for competitive advantage: capitalist transformations in the twentieth century, in G . Dosi, R. Gia nnetti and P. A. Toinelli (eds) Technology a nd Enterprise in a Historica l Perspectiv e (Oxford: Oxford University Press). L azonick, W. and O’ Sullivan, M. (1996) Organization, ® nance, and international competition, Industrial a nd Corpora te Cha nge, 5(1) pp. 1± 48. Ê (1992) Nationa l Systems of Innov a tion. Towa rds a Theory of Innov a tion a nd Intera c tiv e L undvall, B.-A Learning (L ondon: Pinter). Ê . (1998) Innovation as an interactive process: from user-producer interaction to the L undvall, B-A national system of innovation, in G. Dosi (ed.) Tec hnic a l Cha nge a nd Econom ic Theory (L ondon: Pinter), pp. 349± 369. L undvall, B.-AÊ . and Johnson, B. (1994) The learning economy, J ournal of Industry Studies, 1(2), pp. 23± 42. M ckelvey, M . (1991) How do national systems of innovation differ? A critical analysis of Porter, F reeman, L undvall and Nels on, in G. Hodgson and E. S crepanti (eds) Rethinking Econom ics ± M a rkets, Tec hnology a nd Econom ic (Aldershot: Edward Elgar). M etcalfe, J.S . (1994) Evolutionary economics and public policy, Ec onomic J ournal, 104(425), pp. 931± 944. M okyr, J. (1993) The Lev er of Ric hes (Oxford: OUP). NaÊ s, S . O., Sandven, T. and Smith, K. (1994) Innovasjon og teknologi i norsk industri: en oversikt (Innov a toin a nd New Technology in Norwegia n Industry: An Ov erv iew) STEP report 4/94, STEP G roup, Oslo. Nelson, R. (1992) Nationa l System s of Innov a tion. A Com para tiv e Study (L ondon: OUP). Nelson, R. (1987) Understa nding Tec hnologic a l Cha nge a s a n Ev olutiona ry Proc ess (North Holland: Amsterdam). Nelson, R. (1959) The simple economics of basic scienti® c research, J ourna l of Politica l Econom y, 67, pp. 297± 306. Nelson R. and Winter, S . (1982) An Ev olutiona l Theory of Ec onomic Cha nge (Cambridge: Harvard University Press). Numminen, S . (1996) Nationa l Innov a tion System s: Pilot Ca se Study of the Knowledge Distribution Power of Finland (Group for Technology Studies ). OECD (1992) propos ed guidelines for collecting and interpreting technological innovation data, oslo ma nual, OECD/G D(92)26 (Paris: OECD). Patel, P. and Pavitt, K . (1994) National innovation systems: why they are important and how they might be meas ured and compared, Econom ics of Innov a tion a nd New Technology, 3, pp. 77± 95. Patel, P. and Pavitt, K. (1991) L arge ® rms in the production of the world’s technology: an important case of `non-globalization’ , J ourna l of Internationa l Business Studies, 22, pp. 1± 20. Porter, M. (1990) The Com petitiv e Adv a nta ge of Nations (L ondon: Macmillan). S alter, W.E.G. (1969) Productiv ity a nd Technica l Cha nge (Cambridge: CUP). S mith, K. (2000) Economic infrastructures and innovation systems, in C. Edquist (ed.) Systems of Innov a tionÐ Tec hnologies, Institutions a nd Orga nizations (L ondon: Pinter). S mith, K. (1991) Innovation policy in an evolutionary context, in J. S . Metcalfe and P. P. S aviotti (eds) Ev olutiona ry Theories of Econom ic a nd Technologica l Cha nge (Harwood: L ondon), pp. 256± 275. S mith, K . and NaÊ s, S . O. (1995) The Norwegian national innovation system: mapping the system of knowledge production, distribution and use, STEP Group Working Pa per, 16/95, Oslo. S oete, L . and Arundel, A. (1993) An Integra ted Approa ch to Europea n Innov a tion a nd Tec hnologic a l Diffusion Polic y (Maastricht: European Commis sion). S torper, M . and Scott, A. (1995) The wealth of regions, Futures, 27(5), p. 505. Tassey, G. (1991) The functions of technology infrastructure in a competitive economy, Resea rch Polic y, 20(4), pp. 345± 362. Teubal, M. (1977) Innov a tion Perform a nce, Lea rning a nd Gov ernm ent Policy. Selec ted Essa ys (Madis on: University of Wisconsin Press). Von hippel, E. (1989) Sourc es of Innova tion (Oxford: OUP). Williams, R. and Edge, D. (1996) The social s haping of technology, Resea rch Polic y, 25(6), pp. 865± 900.