140 Foster, HilMn, and Adler US Census Bureau (2001). 1997 Economic Census Manufacturing Summary Series. www.census. gov/prod/ec97197m31 s-gs.pdf [27 December 2004). ~_ (2002). 'Pollution Abatement (osts and Expenditures: 1999.' Issued November 2002. www.census.gov/prod/2002puhs/ma20o-99.pdf[27 December 2004]. US National Science Foundation (2004). www.nsf.gov/sbe/srs/iris/tables.dm?pub-year= NSF%2001-30S [15 December 2004]. VDP, the German Pulp and Paper Association (2004). Facts on Paper 2003. Available at www.vdp~ online.de [29 September 2004]. Vogel, D. (1986). National Styles of Business Regulation: A (ase Study of Environmental Poficy. Ithaea, NY: Cornell University Press. (Reprinted 2003, Beard Books.) World Business Council for Sustainable Development (2005). Ecoefficiency. www.wbcsd.org/ templates/TemplateWBCSD5/layout.asp?type=p&:Menuld=NzA&:doOpen=1&:ClickMenu= LeftMenu [22 September 2005]. 7 From Theory to Practice: The Use of the Systems of Innovation Approach in Innovation Policy Cristina Chaminade and CharIes Edquist Introduction Since the seminal work of Freeman (1987) on the ]apanese national innovation system, the number of contributions to the systems of in- novation approach at anational, sectoral, and regional level has grown (Lundvall 1992 Carls- son and ]acobsson 1993 Cooke, et al. 1997 Edquist 1997a Edquist and ]ohnson 1997 Lundvall, ]ohnson, et al. 2002 Malerba 2004 Malerba and Orsenigo 1997 Nelson 1993). The academic discussion started in the sphere in the 1990S thanks to the ~.~~:l~~l:~:t of Economic Cooperation and t (OECD), which played a prom- role in promoting the use of the SI ap- in the design and implementation of ihrl0\'ation poliey in the OECD countries 2004). Among the diverse initiatives place in the OECD during the 1990S, seven����yelar projeet on National Systems of tfff[��v'ation (NSIs) (1995-2002) is of special rele- OECD had a great influenee in the countries and same of the govern~ adopted the innovation system their innovation policy. However, by Mytelka and Smilh 2002, the SI not been entirely successful in of designing poliey and pro- instruments easier. ' hlapter proposes a way of dealing wilh q1l'Lpl',x reality. By breaking down the operation of the SI into 'activities,' the role of the government and the interplay between pri- vate and public actors can be discussed, and specific recommendations on how and when public actors should intervene can be made. The point of departure of this ehapter f��r the discussion of innovation policy is the 'generic' SI approach, as discussed briefly in its second section. This section also identifies the main components of the SI approach. The third sec~ tion presents different approaches to classify- ing the activities in a SI in the fourth section, the authors propose ten activities that capture the operation of an innovation system. The role of the public sector in each activity is then discussed, and a new research agenda is proposed the final section draws some conclu- sions. Systems of innovation 1 There are almost as many definitions of SIs as authors, but most relate in some way to the definition of a system. According to Ingeistam (2002): (a) a system consists of two kinds of constitu- ents: there are, first, some kinds of com~ ponents and, second, there are relations among them. The components and rela R tions should form a coherent whole

142 Chaminade and Edquist (which has properties different from the properties of the constituents) (h) the system has a 1imction-that is, it is performing or achieving something (e) it must be possible to discriminate be M tween the system and the rest of the warld that is, it must be possible to iden- tify the boundaries of the system. If W���, fm example, want to make empirical studies of specific systems, we must, of course, know their extension. 2 A systemic approach is the point of departure fm the literature on technological systems (Dosi 1982 Gille 1978 Hughes 1983 Rosenberg 198z), industrial systems (Hirschman 1958 Porter 1992), and innovation systems. Within this last group, and according to the level of analysis, it is possible to distinguish between (Edquist 1997): ��� National���� Innovation Systems (Freeman 1987 Lundvall 1992 Nelson 1993) ��� Regional Innovation Systems (Camagni '99' Cooke et al. 1997 B1aczyk et al. 1998 eooke 2001 and Asheim and Isaksen 2002) ��� sectoral and 'technological innovation systems' (Breschi and Malerba 1997 Carls- son 1995 Carlsson and Stankiewicz 1991 Malerba 2004). For the purpose of the discussion here, we propose that an SI includes 'all-important eco- nomic, sodal, political, organizational, institu- tional and other factols that influence the development, diffusion and use of innov- ations' (Edquist 1997).3 If all factors that influ- ence innovation processes are not included in adefinition, one has to argue which potential factors should be excIuded-and why. This is quite difficult, since, at the present state of the art, we do not know the determinants of in- novations systematically and in detail. What are the components of an S1? Organizations and institutions are often con- sidered to be the main components of SIs, al- though it is not always clear what is meant by these terms. Let us, therefore, specify what organizations and institutions mean here (Edquist 1997). Organizations are 'formal structures that are consciously created and have an explicit pur- pose' (Edquist and Johnson r997). They are 'players or actors.'4 Some important organiza- tions in SIs are firms (normally considered to be the most important organizations in SIs), universities, venture capital organizations, and public agencies responsible for innovation policy, competition policy, or drug regulation. Institutions are 'sets of common habits, norms, routines, established practices, rules or laws that regulate the relations and inter- actions between individuals, groups and or- ganizations,' (Edquist and Johnson 1997). They are the mies of the game. Examples of important institutions in SIs are patent laws, as weIl as rules and norms influencing the re~ lations between universities and firms. Obvi- ously, these definitions are of a Northian character (North 1990), discriminating be- tween the rules of the game and the players in the game. Which institutions and organizations are in- cluded within the boundaries of the system of innovation is a matter of discussion. Lundvall (1992) distinguishes between a narrow and a broad definition of an SI. The narrow one in- cludes only the organizations and institutions involved in research activities (searching and exploring). This embraces universities} R&D departments in firms, and technological insti- tutes. The broad definitionj on the other hand} refers to all 'parts and aspects of the economic structure and the institutional set-up affecting learning as weIl as searching and exploring' Lundvall (1992: r2). This chapter adopts this broader perspective. Implications of the SI approach for innovation po1icy Innovation policy is public actions that influ:" ence innovation processes: that is, the opment and diffusion of (product and pHJeE"S) innovations. The objectives of innovation icy are often economic ones, such as eOJnorr��c growth, productivity growth, increased em- ployment and competitiveness. However} they mayaiso be of a non-economic kind, such as cultural, social} environmental, or military. The objectives are determined in a political process} and not by researchers. They must} however, be specific and unambiguously formulated in relation to the current situation in the country and/or in comparison to other countries. Understanding innovation as a complex interactive learning process has important im- plications for the design and implementation of any kind of policy to support innovation. It affects the focus of the policy} the instruments, and the rationale for public policy. This chap- ter will deal mainly with the first two issues, whilst the third will be discussed in detail in Chaminade and Edquist 2005. The implications of the SI approach for pub- He policy are better understood when its basic assumptions are compared to those of main- stream economics (Lipsey and Carlaw 1998 Smith 2000). Knowledge, leaming, and innovation in mainstream economics One of the basic assumptions of neoclassical economic theory is perfeet information: that is, all economic agents can maximize their profits because they have perfeet information about the different options available to them. Knowledge is equal to information: that is, it is codified, generic, accessible at no cost, and easily adaptable to the firm's specific condi- tions. These tacit assumptions about the properties of knowledge are included in the discussion about the process of invention. For Nelson (r959) and Arrow (1962), the knowledge eman- ating from research has some specific proper- ties: uncertainty, unappropriability, and indivisibility (Smith 2000). 'UncertaintY refers to the impossibility of knowing apriori the outcomes of the research process and the risk associated with it. 'Unappropriability' refers to firms' being unable fully to appropriate the benefits which derive from the invention. As From Theory to Practice 143 knowledge is information, freely accessible to all economic agents} this means that there is no incentive for the research activity. FinaIly, 'indivisibility' implies that there is a minimum scale of knowledge needed before any new knowledge can be created: that is, new know- ledge is created on the basis of an existing pool of knowledge (inside or outside the firm). Therefore, it is difficult to separate what con- stitutes new knowledge from the knowledge that already exists. For neoclassical economics, the innovation process is narrowed down to research (and in- vention). How to transform the results of the research activity into products or processes that can be traded in the market is a black box (Rosenberg 1982, 1994). For the neoclassical theorists, the process of innovation is a fixed sequence of phases, where some research ef- forts will automatically turn into new prod- ucts. These three characteristics of scientific knowledge (uncertainty, unappropriability, and indivisibility) will lead to an underinvest- ment in R&D activities. This constitutes the main rationale for public intervention in re- search activities. Policy makers have to inter- vene because of a market failure: private actors in the economies will systematically underin- vest in R&D, not reaching the optimal alloca- tion of resources for invention. As argued by Smith (2000), the neoclassical approach, despite its many shortcomings, can be useful for understanding basic science, but it is very limited when trying to explain innov- ation activities, especially those with doser links to the market. The policyimplications that emerge from the market failure theory are, from a practical and specific point of view, not very helpful for pol- icy makers. They are too blunt to provide much guidance. They do not indicate how large the subsidies or other interventions should be, or within which speciflC areas one should inter- vene. They say almost nothing about how to intervene: that is} which policy instruments should be used and the process through which they should be implemented. Standard eco- nomic theory is not of much help when it