Comparing a Pharmaceutical and an Agro-food Bioregion: On the Importance of Knowledge Bases for Socio-spatial Patterns of Innovation more

Industry and Innovation, Vol. 13, No. 4, 393–414, December 2006 Research Paper Comparing a Pharmaceutical and an Agro-food Bioregion: On the Importance of Knowledge Bases for Socio-spatial Patterns of Innovation LARS COENEN*, JERKER MOODYSSON**, CAMILLE D. RYAN{, BJØRN ASHEIM* & PETER PHILLIPS{ *Centre for Innovation, Research and Competence in the Learning Economy (CIRCLE) & Department of Social and Economic Geography, Lund University, Sweden, **Department of Social and Economic Geography, Lund University, Sweden, {Faculty of Communication and Culture, University of Calgary, Calgary, Canada, {Department of Political Studies, University of Saskatchewan, Saskatoon, Canada The aim of this paper is to compare the socio-spatial patterns of innovation and knowledge ABSTRACT linkages of a biopharmaceutical and an agro-food biotech cluster. Dissimilarities can be expected based on differences in terms of historical technological regimes and sectoral innovation system dynamics between the agro-food and pharmaceutical industries in general and particularly the distinctive analytical (science-based) knowledge base of biopharmaceuticals in contrast with the more synthetic (engineering-based) knowledge base of agro-food biotechnology. Drawing on bibliometric data and case material the study compares two representative bioregions: a biopharmaceutical cluster in Scania, Sweden and an agro-food biotech cluster in Saskatoon, Canada. The empirical study supports the theoretical expectations and shows that knowledge dynamics in the agro-food cluster are more localized than in the biopharmaceuticals cluster. It is important, however, to acknowledge that these differences are relative. Both sectors display local and non-local patterns of collaboration following the general pattern for biotechnology. KEY WORDS: Biotechnology, clusters, regional innovation systems, knowledge bases Introduction Many policy-makers and scientists view biotechnology as the next big thing in the knowledge economy. On a regional level, this is illustrated by various policy programs Correspondence Address: L. Coenen, Centre for Innovation, Research and Competence in the Learning Economy (CIRCLE) & Department of Social and Economic Geography, Lund University, Sweden. Tel.: 46 46 2227747. Email: lars.coenen@circle.lu.se 1366-2716 Print/1469-8390 Online/06/040393–22 # 2006 Taylor & Francis DOI: 10.1080/13662710601032937 394 L. Coenen et al. specifically aimed at strengthening regional biotechnology activities such as ‘BioRegio’ in Germany. Also generic regional development policy often targets biotechnology as a strategic sector to leverage future regional competitiveness (for example EUs Objective 2 program supporting Genomic Technology and Informatics in Scotland). Empirical research acknowledges the importance of regions as prime sites of innovation through a strong geographical concentration of biotechnology firms and organizations in a handful of ‘megacentres’ (Cooke, 2002) or nodes of excellence (Feldman, 2000), such as San Diego, Boston and Munich. However, global knowledge connections are indispensable to maintain at the cutting edge of research and development (R&D). This has resulted in the typical ‘local nodes, global networks’ configuration (Gertler and Levitte, 2005). These observations feed into discussions on the role of geographical proximity for interactive innovation, firm-based learning and knowledge transmission (e.g. Amin and Cohendet, 2004; Morgan, 2004). Even though agreement can be found in the assertion that globalization and localization are complementary processes that tend to enhance instead of erode each other, accusations of under- or over-exaggerating the role of proximity for innovation are frequently heard. Acknowledging that innovation is an interactive process between economic agents that is socially and territorially embedded and culturally and institutionally contextualized (Lundvall, 1992), it can be argued that a discussion on the importance of proximity for innovation evolves around the construction of relational proximity (Torre and Gilly, 2000; Coenen et al., 2004). It refers to closeness in terms of relations (e.g. through organizations and networks), reference and knowledge (e.g. norms, values, rules of thought and action). Admittedly, relational proximity is a somewhat vague concept, yet it recognizes that interactive learning does not need to be territorially confined as the actual explanatory power of proximity does not pertain to its quality of being physically close together as such (Amin and Cohendet, 2004). Instead it implies that social interactions, whether local or non-local, have to be actively constructed (Morgan, 2004). We argue that the actual spatial distribution of an industry’s innovation network and its transformation through processes of globalization depends on its knowledge base. For example, previous research has demonstrated the importance of heavily localized innovation networks based on engineering knowledge in a Danish furniture district as opposed to more distant knowledge links for a more science-based Norwegian electronics cluster (Asheim and Coenen, 2005). This paper focuses specifically on the role of proximity for knowledge dynamics in two biotech clusters set against different industrial contexts. One of the reasons for the popularity of biotechnology in the knowledge economy can be found in its paradigmatic and pervasive character. Potentially it affects all living organisms as its domain of application. This broad application span has also caused problems. Some studies focus nearly exclusively on one particular industry while others claim that it should be seen as a generic technology (Brink et al., 2004). The sector that so far has been most extensively influenced and analyzed is the pharmaceutical industry. Due to the extensive economic and societal impact of pharmaceutical applications this niche has become the largest, now representing about 70% of all biotechnology sales (Cooke, 2005). At the same time is a drastic and sudden shift suggested in the often-coined term ‘biotech revolution’ questionable. On an aggregate level the adoption of biotechnological knowledge, techniques and tools in industry is following a familiar pattern of slow but consistent diffusion rather than dramatic upheaval (Nightingale and Martin, 2004). Notwithstanding this, substantial qualitative Socio-spatial Patterns of Innovation 395 changes are taking place with respect to innovative practices in a range of industries underpinned by biotechnology (OECD, 2004). It is common to make a distinction between red biotechnology (applied to pharmaceuticals), green biotechnology (applied to agro-food) and grey biotechnology (applied to industrial processes) (Fuhrer, 2003). Each category produces distinctive economic geographies and innovation patterns. Whereas biopharmaceuticals is based on triadic R&D trade among universities and research institutes, pharmaceutical companies and small, dedicated biotechnology firms (DBFs), agro-food is organized around dual basic knowledge transfer between large agro-food companies and large public research institutes (Cooke, 2004). This paper seeks to follow up on this distinction by offering an explanation drawing on differences in the knowledge base of biopharmaceuticals and agro-food biotechnology. More concretely, its aim is to compare the socio-spatial patterns of innovation linkages and knowledge flows between a biopharmaceutical cluster, i.e. the Swedish part of Medicon Valley1 (Scania), and an agro-food biotechnology cluster, i.e. Innovation Place in Saskatoon, Canada. The remainder of the paper is structured as follows: The following section discusses the conceptual framework. Section three provides a general comparison between pharmaceutical and agro-food bioregions. Section four presents the empirical work specifically on the biopharmaceutical cluster in Scania while section five analyzes the agro-food cluster in Saskatoon. Finally, the conclusions compare these results and provide recommendations for future research. Innovation Systems, Knowledge Bases and Communities This paper has its conceptual vantage point in the systems approach to innovation and cluster literature. In short, innovation systems as well as clusters capture the notion that innovations are carried out through a heterogeneous network of various actors, often comprising the triple helix of industry, universities, and (multi-level) government (Etzkowitz and Leydesdorff, 2000), underpinned by an institutional framework (Edquist, 1997; Porter, 1990). In this, institutions, understood as sets of common habits, routines, established practices, rules or laws, regulate the relations and interactions amongst these actors (Edquist and Johnson, 1997). Central to both is the emphasis on inter-connectivity, between actors as well as institutions, providing scope for complementarities or systemicness. Dependent on its boundaries, many types of innovation systems can be distinguished, such as local, regional, national, sectoral and technological. A cluster, on the other hand, refers to firms in the same or adjacent industrial sector(s) concentrated in a small geographic area such as a region. This explains the considerable overlap that exists between regional innovation systems (RIS) and cluster literature. They should not, however, be conflated. A RIS can, in principle, stretch across several sectors (and thus, clusters) in the regional economy, provided that firms and knowledge organizations interact in a systemic way. As such, a RIS relates to the boundaries of a territorially bounded, administrative region (Cooke and Leydesdorff, 2006) while a cluster primarily refers to functionally interrelated industries (Malmberg, 2003). 1 Previously Medicon Valley has been analyzed as a cross-border cluster covering both the Danish and Swedish side. However, based on an analysis of intra-regional, cross-border publications it was concluded that the cluster is embedded in two distinct (but related) regional innovation systems (Coenen et al., 2004). 396 L. Coenen et al. In contrast to traditional, neo-classical approaches that have reduced knowledge to ubiquitous free-flowing information, the RIS and cluster literature conceptualizes learning between economic actors as an inherently social process (Lundvall, 1992). It can be interpreted as situated action in which the organizational and institutional context provides structures and shared meanings for action and communication in which people are able to learn (Nooteboom, 2000). In terms of systemicness the RIS approach draws substantially on complementarities following from frequent interactions and untraded interdependencies on the regional level in the form of conventions, informal rules and habits (Storper, 1997). Through this focus it can be distinguished from a national innovation system written small (Howells, 1999). The main rationale for applying a regional perspective is that it allows for a bottom-up, grounded approach in analyzing innovation interaction and inter-organizational learning processes (Doloreux and Parto, 2005). It emphasizes the role of embeddedness, accounting for the importance of personal relations and networks for economic action and outcomes ingrained in a social and cultural context (Granovetter, 1985). It is, however, important to note that the RIS approach refers to localized but not to exclusively local learning. It acknowledges the need to combine both local and non-local knowledge, skills and competences (within the spatial context of the region) in order to go beyond the limits of the region (Doloreux, 2004). Similarly, the ‘local buzz, global pipeline’ metaphor stresses the importance of non-local connections that clusters need in order to tap into new and valuable knowledge created in other parts of the world and to prevent cognitive and economic lock-in (Bathelt et al., 2004). As mentioned in the introduction, the industrial knowledge base provides an important novel dimension to unpack and compare socio-spatial patterns of clusters and regional innovation systems. Within the innovation systems family, sectoral innovation systems (SIS) are probably most pronounced in their focus on the impact of specific knowledge bases on innovation processes. As this paper compares two different applications of biotechnology the following provides a closer look at how the sectoral approach differs, overlaps and complements the regional innovation system framework. A SIS is defined as ‘‘a system (group) of firms active in developing and making a sector’s products and in generating and utilizing a sector’s technologies’’ (Breschi and Malerba, 1997: 131). Similar to RIS, actors, networks and institutions are its main building blocks (Malerba, 2005). Contrary to the territorial basis of RIS, however, its boundaries are specified by a certain product group characterized by a common knowledge base. A knowledge base refers to the area of knowledge itself as well as its embodiment in techniques and organisations (Brink et al., 2004). Upon closer observation of the broad concept of institutions, another interesting difference is noted which is linked to the threefold categorization by Scott (1995) of regulative, normative and cognitive rules. Studies of regional innovation systems seem to be mainly geared towards analyses of regulative (e.g. formal rules, laws, standards and governance structures) and normative rules (e.g. values, norms and codes of conduct). In comparison, studies of sectoral innovation systems pay more attention to cognitive institutions such as problem-agendas, paradigms, categories, classifications and search heuristics. Nelson and Winter (1977) argue for example that engineering practices in firms are less sensitive to price-related changes on the market than to technical ideas and beliefs about where to go, what problems to solve and what sort of knowledge to draw on (Kemp et al., 1998). Thus SIS tends to be more specifically concerned with the impact of different technological regimes on innovation processes than RIS. Socio-spatial Patterns of Innovation 397 Malerba (2004) relates three dimensions of knowledge to the notion of a technological regime: opportunity, appropriability and cumulativeness. Opportunity refers to the likelihood of successful innovation, appropriability to the possibility of protecting innovations from imitation and of reaping profits from innovative activities and cumulativeness represents the degree to which the generation of new knowledge builds upon previous knowledge. Based on these dimensions he distinguishes various types of sectoral innovation systems. Agrofood is generally characterized by rather low degrees of opportunity, appropriability and firm-level cumulativeness. Opportunities for innovation are typically aimed at lowering production costs by introducing new (raw) materials or process machinery and at improving quality control schemes (Lagnevik et al., 2003). Assets and production inputs based on such price competition are easy prey for processes of ubiquification. Over time they become everywhere available at more or less the same cost, gradually eroding the competitive advantage of the innovator (Maskell and Malmberg, 1999). Moreover, the ability to innovate depends strongly on the adoption of knowledge developed elsewhere (Wilkinson, 2002). This hampers the endogenous development of firms’ technological competencies. Malerba (2004) thus concludes that an agro-food sectoral innovation system, in general, consists of few innovative firms and high degrees of geographical dispersion. This spatial pattern is explained by the relatively simple and codifiable nature of the knowledge base making spatial proximity not relevant for knowledge transmission. Notwithstanding these general tendencies, Breschi and Malerba (1997) also note that exceptions in the form of geographical concentrations can be found. Case studies have shown that particular food clusters have emerged (Porter, 1990; Nilsson et al., 2002; Onsager and Aagen, 2003) due to natural geographic factors, specific historical industrial trajectories, the role of institutions and specific competencies of firms. However, the above description of the agro-food innovation system does not take full account of the recent introduction of biotechnological principles in agro-food. Alluding to section five, the case of Saskatoon provides a clear example of some of the changes taking place in post GMO (genetically modified organisms) agro-food production and innovation. A sectoral innovation system perspective on the pharmaceutical industry clearly provides a different outlook. Here, arguably, the technological regime is characterized by very high opportunity conditions. The knowledge base for drug discovery and drug development is highly fragmented, dispersed and in transition, which allows for a high level of heterogeneity and new entry in the population of innovating firms (Malerba, 2004). The extent of appropriability is determined by two factors: the extent of intellectual property protection and collaborative relations with firms and organizations possessing complimentary knowledge and competencies. With respect to intellectual property rights, Malerba and Orsenigo (2001: 11) argue that ‘‘pharmaceuticals has historically been one of the few industries where patents provide solid protection against imitation. Because small variants in a molecule’s structure can drastically alter its pharmaceutical properties, potential imitators often find it hard to work around the patent’’. Secondly, because of substantial advances in various disciplines of life science such as genetics and molecular biology since the 1970s, the potential applications of life science in pharmaceuticals have increased dramatically. Neither dedicated biotechnology firms nor traditional large pharmaceutical firms are able to master all the capabilities needed to bring new, biotechnology-based pharmaceutical innovations to market and are thus dependent on universities, publicly funded research institutes and each other for complementary competencies (Nilsson, 398 L. Coenen et al. 2001). Such symbiotic relationships allow individual firms to accumulate specialized knowledge within their particular technological niche, further reinforcing heterogeneity within the innovation system. On the basis of these characteristics of the technological regime Malerba (2004) concludes that, the pharmaceutical industry displays a dual spatial pattern in terms of knowledge dynamics: local nodes and global networks. The SIS approach points to some key differences between agro-food and pharmaceuticals. Thus, it complements a predominantly territorial perspective on innovation systems in that it explicitly accounts for different industrial dynamics on the basis of its typical knowledge characteristics. On the other hand, it is a very broad and general differentiation that typically dichotomizes low-tech from high-tech. Thus it runs the risk of conflating knowledge-intensive industries and R&D-intensive industries by neglecting the importance of distributed knowledge bases. In a critique on the new economy’s dominant interest for high-tech, R&D-intensive industries, Smith (2000: 19) argues that ‘‘the relevant knowledge base for many industries is not internal to the industry, but is distributed across a range of technologies, actors and industries’’. His empirical study shows how the Norwegian food processing industry hardly conducts any internal R&D but draws intensively on a broad spectrum of external technologies and knowledge areas such as biotechnology, electronics, instrumentation and engineering, reflecting the heterogeneous activities along the value chain. Similarly, pharmaceuticals sometimes rely on non-R&D-based, more mundane knowledge bases. Benneworth (2003) illustrates this for example in a case study of the biotechnology industry in the North East of the UK. Here firms have managed to niche themselves by specializing in instrumentation, diagnostic kits and analytical software and pharmaceutical manufacturing (in a strict sense). The notion of distributed knowledge bases draws attention to the importance of cross-sectoral knowledge linkages. This is especially relevant in new, emerging sectors such as agro-food biotechnology where sectoral boundaries have not yet consolidated. Following received wisdom from the philosophy of science, an epistemological distinction can be made between two different forms of knowledge bases (and related activities), analytical knowledge related to ‘natural science’ versus synthetic knowledge related to ‘engineering science’ (Laestadius, 2000; Lundvall et al., 2002; Asheim and Gertler, 2005). As an ideal type, a synthetic knowledge base refers to the knowledge required for activities involved in the design of something that works as a solution to a practical problem and is closely related to engineering. Innovation predominantly refers to the application or novel combination of existing knowledge. Activities that require an analytical knowledge base are geared to understanding and explaining features of the universe. Innovation mainly occurs through the creation of new knowledge as such. The distinction contains different mixes of tacit and codified knowledge, codification possibilities and limits, qualifications and skills required by organizations and institutions involved, as well as specific innovation challenges and pressures. It should be emphasized, however, that it refers to conceptual ideal-types. In reality, innovation processes will involve elements of both. It can, however, be suggested that the degree to which (elements of) one knowledge base dominates is highly industry-specific (Asheim and Gertler, 2005). Moreover, the dominance of one knowledge base arguably has different socio-spatial implications for the knowledge interplay between actors (Wolfe et al., 2004). Research on actual knowledge flows and linkages between actors in an innovation system or cluster seems to be relatively sparse (Archibugi et al., 1999). This calls for a shift Socio-spatial Patterns of Innovation 399 from a static analysis of innovation networks and actors as repositories of knowledge to a more dynamic position that stresses the (social) practice of knowledge creation ‘in action’ (Brown and Duguid, 2002; Amin and Cohendet, 2004). The literature on ‘communities of practice’ provides an important source of inspiration by identifying key entities driving the firm’s knowledge-processing activities (Asheim and Gertler, 2005). Communities of practice are defined by the communal (shared) practice of its members, who undertake or engage in a task, job or profession while communicating regularly with one another about their respective activities (Brown and Duguid, 2002). Its members are informally bound together by shared experience, expertise and commitment to a joint enterprise (Gertler, 2004). They are able to produce and internalize shared understandings through collaborative problemsolving. Furthermore, communities are increasingly seen as key conduits of knowledge formation and exchange that enable interactive learning which takes place in embedded inter-organizational cluster relations (Bathelt, forthcoming). Communities of practice appear to accommodate the situated, pragmatic and interactive nature of learning processes ‘in action’ within and across organizations in a better way than individual-centered or classical organization-centered approaches (Amin and Cohendet, 2004). In addition, Benner (2003) points out that the community approach recognizes that successful work practice requires continual learning which is often (and increasingly so) rooted more in occupations than in business enterprise. Coenen et al. (2004) distinguish between communities of practice that are linked to industries drawing on a synthetic knowledge base and epistemic communities (KnorrCetina, 1999; Cowan et al., 2000) that are linked to industries drawing on an analytical knowledge base. The latter are bound by their commitment to enhance a particular set of knowledge without being concerned about the application of such knowledge. Knowledge dynamics in social networks of epistemic communities can more easily involve distanced ties and relationships supported by increased mobility offered through cheap and extensive air travel, the internet, and specialized literature (Amin and Cohendet, 2004). A compelling example is provided by the Human Genome Sequencing Consortium, involving research centers in the US, UK, Germany, France, Japan and China. Also Zeller’s (2004) study of learning processes in multinational pharmaceutical companies provides a case in point. Similarly, Bathelt (forthcoming) differentiates between a community of practice and an epistemic community by arguing that autonomy and self-organization are weaker in the latter. Characteristically, epistemic communities accept some collectively accredited procedural authority (e.g. peer review) and a set of conventions to facilitate their common goal of pursuing knowledge. Such a set of conventions allow scientists to speak the same universal, scientific ‘tongue’ facilitating trans-national communication. Kuhn (1970) refers to these community-specific conventions as the ‘‘disciplinary matrix’’ (p. 182) consisting of formal components or representations (e.g. E5mc2), commitment to beliefs in particular models (e.g. that molecules of a gas behave like tiny elastic billiard balls in random motion), subscription to certain values (e.g. the accuracy of predictions) and, finally, of the wellknown Kuhnian paradigms or problem-solving exemplars. Communities of practice, on the other hand, are often concentrated around a concrete problem-solving practice or task (typical for a synthetic knowledge base) that requires frequent and specialized communication and ‘co-action’, which in turn is facilitated by co-location. Here an example is provided by a community of engineers involved in developing and testing a specific piece of equipment (Gertler, 2004). Based on this we hypothesize that industries drawing on 400 L. Coenen et al. analytical knowledge bases tend to be less sensitive to distance-decay in knowledge transfer facilitating global networks and dense local collaboration in epistemic communities. Synthetic knowledge creation, on the other hand, has a tendency to more sensitive to proximity effects between the actors involved, thus favoring local collaboration in communities of practice. The following section discusses differences between the spatial distribution of innovative collaboration in biopharmaceuticals and agro-food biotech, respectively, drawing on the above conceptual framework. Pharmaceutical and Agro-food Bioregions in a Comparative Perspective Following the SIS approach, clear differences can be identified in innovation dynamics between the food and pharmaceutical sectors on a general level. As these industries differ with respect to their dominant knowledge base this, arguably, influences the knowledge dynamics of biopharmaceuticals and agro-food biotech and its socio-spatial implications. The pharmaceutical industry has its origins in natural sciences like chemistry, biology and medicine. Since the shift from traditional chemistry-based ‘random drug design’ to modern biology-based ‘rational drug design’ in the 1970s, biotechnological research techniques have largely displaced orthodox chemical capabilities (Casper and Matreves, 2003). This radical shift of technology created strong incentives for new entrants and reduced the earlier dominance of large, chemistry-based, pharmaceutical companies (Nilsson, 2001). Over the past two decades the number of possible applications has expanded rapidly and the role of universities and small research-oriented ‘dedicated biotech firms’ (DBFs) has increased even further (Cooke, 2004). Concomitantly, there has been a shift towards a more important role for basic science as a key activity in the pharmaceutical innovation and production processes. Nightingale (2000) observes a shift from ‘wet’, laboratory-based experiments to theoretical ‘in silico’ scientific discovery referring to the increased use of computer aided molecular research. Major R&D is conducted at universities, research institutes and DBFs, and subsequently in-licensed by large pharmaceuticals that carry out manufacturing, marketing and distribution activities related to the end product (Carlsson, 2002). The agro-food industry, in contrast, does not originate from academic science but has its roots in agriculture. Traditionally, the food industry has largely drawn upon an empirical, experimental up-scaling of artisan processes and substitution strategies to replace specific raw materials by means of chemical, and more recently, biological synthesis2 (Wilkinson, 2002). Similarly, agro-food biotech follows a more trial-driven (synthetic) engineering tradition as opposed to the theory-driven (analytical) approach in biopharmaceuticals (Laestadius, 2000). Typically innovation in agro-food biotech draws on the integration of several contributory knowledge bases. R&D is generally conducted in-house by large multinational corporations fulfilling this integrative role while DBFs have a more limited role in the innovation process (Cooke, 2004). Similar to biopharmaceuticals, agro-food companies have ties with universities and research universities yet these ties are based on a distinctively different knowledge profile. 2 With respect to biological synthesis, the introduction of GMOs has led to a sharp increase in technological opportunities which are heavily debated in terms of societal impact. Socio-spatial Patterns of Innovation 401 The above advancements in biotechnology have substantially increased the degree of technological opportunity in pharmaceuticals and agro-food industries, which in turn has altered their geographical patterns of activity. In both sub-sectors a strong concentration in nodes of excellence has emerged. Moreover, these clusters tend to increase in density as the sectors grow over time. This has been demonstrated in numerous studies in various parts of the world. Among the most important nodes of excellence in biopharmaceuticals are the ‘megacentres’ Boston, San Francisco and San Diego in the US, Toronto and Montreal in Canada, and Munich, Stockholm and Oxford–Cambridge in Europe (Cooke, 2005). These regions also host world-leading universities in life sciences and large pharmaceutical companies (Feldman, 2000; Cooke, 2004). This indicates that access to highly qualified (and locally mobile) labor and knowledge spill-over effects that result from an environment of ‘open science’ are important factors that attract firms and research institutes to these nodes (Owen-Smith and Powell, 2004). In agro-food biotechnology clusters the presence of large ‘anchor firms’ (Feldman, 2000) plays a more pronounced role. Here, most clusters are located in traditionally agricultural areas such as St. Louis, Connecticut and Raleigh in the US; Guelph and Saskatoon in Canada; and Wageningen, Basel and Edinburgh in Europe (Ryan and Phillips, 2004). Due to this history of agroindustrial specialization, research institutes with close ties to industry have often been established in such regions. But contrary to the more fundamental life science knowledge infrastructures in biopharmaceutical clusters, these research institutes are characterized by ‘hands-on’ competence in more applied areas of food and agricultural technologies. The position of biotechnology is one among a heterogeneous set of technologies. In the pharmaceutical industry, however, its position appears to be more dominant. Spatial concentration (in distinct places) is thus a characteristic feature of both biopharmaceuticals and agro-food biotech. Particularly for pharmaceuticals, global connections between the clusters are of complementary value due to the global distribution of relevant and often rare knowledge (Cooke, 2005). To what extent such global connections are important for agro-food biotechnology has not been investigated to the same degree. Assumingly, the differences in scientific origins and localization are reflected in the organization of innovation activities. Based upon the introductory discussion, a more concentrated, less distributed spatial pattern of collaboration in agro-food biotech as compared to biopharmaceuticals would be expected. This expectation draws on the typically synthetic knowledge base that characterizes the industry and the related importance of communities of practice. This contrasts the typical analytical knowledge base in pharmaceuticals in which epistemic communities provide an explanation for the dual pattern of local nodes and global connections in terms of knowledge dynamics. To empirically analyze the spatial distribution of knowledge collaboration, co-operation between DBFs and partners in a biopharmaceutical cluster have been compared with cooperation between star scientists and partners in an agro-food cluster.3 As indicators for knowledge collaboration the analysis drew on joint patents and scientific publications. Data collected from the Science Citation Index (SCI) and the United States Patent and Trademark Office (USPTO) have been employed. This allows for the identification of 3 To cater for symmetry in the comparison it would have been beneficial to compare clusters against similar territorial contexts. However, agro-food biotechnology clusters in Europe have not reached the same level of maturity as is the case for its biopharmaceutical counterparts. This observation underpins the choice for a cross-Atlantic comparison. 402 L. Coenen et al. inventors and authors by organizational type (firm or PRO actor) and geographical location (region). The selection has been made by searching on authors and assignees using firm name and name of star scientists that previous research has identified as the core actors of the cluster in these regions (Ryan and Philips, 2004; Coenen et al., 2004). After this, the identification, categorization and mapping of co-authors and co-inventors has been done manually. The collaboration linkages displayed in the patent and publication data have been categorised along two main dimensions: (1) intra-organizational vs. extra-organizational,4 and (2) local vs. non-local. The non-local collaboration linkages were subsequently further specified as illustrated in Figures 1–4. Two time periods have been compared: 1990–1996 and 1997–2003. The rationale for making these distinctions is that there was little biotech activity in the studied regions before 1990. The mid 1990s can be seen as a turning point in at least two ways: (1) due to aforementioned technological breakthrough the level of activity increased significantly in the regions in the mid 1990s, with respect to both the number of actors and knowledge output, and (2) preconditions for inter-organizational collaboration changed significantly with the rise of information technology and, in particular, Internetbased communication. This bibliometric approach offers a number of advantages as well as shortcomings. Data is free and publicly available. Moreover it is well-suited to analyze knowledge collaboration related to biotechnology given that firms and organizations tend to publish and patent quite extensively compared to other sectors (Klopper and Haisch, 2006). ¨ However, the relative contribution of each co-author and co-inventor cannot be taken into account. In addition, only formal interactions are accounted for. Therefore the bibliometric analysis has been further supported by interviews with researchers at universities and research centers and research managers at DBFs, surveys as well as secondary information based on previous research on the two bioregions. The next section outlines the biopharmaceutical cluster. Section five provides the analysis of the agro-food cluster. The main differences are discussed in the conclusions. The Scanian Biopharmaceutical Cluster The Scanian biotech cluster is located in the urban area of Malmo and Lund. Second only to ¨ Stockholm/Uppsala, it is one of the largest biotechnology regions in Sweden with a strong focus on biopharmaceuticals. The region hosts Lund University, the largest university in Scandinavia with approximately 5000 researchers and 40,000 students, 10 hospitals of which two are university hospitals, and 36 DBFs (VINNOVA, 2003). Twenty of the DBFs are specialized primarily in biopharmaceutical applications while another 11 focus on bioproduction and biotech tools. The latter are predominately suppliers to the pharmaceutical industry. Most of the firms are small relative to international comparisons. Half of the firms have less than 10 employees, while only two firms exceed 100 (Biotech Sweden, 2004). One of the world’s leading pharmaceutical firms, Astra, was founded in Lund. Upon its merger with Zeneca in 1999 (becoming AstraZeneca), the company continues to head its operations out of Lund. Up until recently, Pharmacia Corporation (which merged with Upjohn in 1995 and was acquired by Pfizer in 2003) also situated one of its major research units in the region. Since July 2000, Scania has been connected to the Danish biotech 4 The patent data does not allow for this type of distinction on the level of individuals, since the inventors are listed with name and home address, not affiliation. Socio-spatial Patterns of Innovation 403 cluster in Copenhagen through the construction of the Øresund-bridge. The cross-border ‘Medicon Valley’ is by scope and scale fully comparable to established megacentres such as Boston and Munich (Cooke, 2005). Previous studies analyzing the cross-border integration of Medicon Valley consider its impact relatively limited in terms of knowledge collaboration measured through co-publications (Coenen et al., 2004). Similarly, Lundquist and Winter (2003) warn against exaggerated optimism over the economic integration of the Øresund region in the short run. None the less, the long-term perspective regarding regional integration remains more optimistic when taking on-going political and economic efforts and initiatives into consideration. For example, Medicon Valley Academy5 (a network organization dedicated to life sciences) has initiated several promotional measures to facilitate cross-national interaction among firms and to promote further growth of the region. The bibliometric analysis of knowledge collaboration covers all DBFs located in the region. The selection of firms is based on a list compiled by Medicon Valley Academy. At the start of this study in early 2003, the list included 36 Scanian firms. After a crosscheck with VINNOVA’s (2003) selection and a manual verification of the company websites, three firms were excluded.6 The reason for starting the analysis from a firm perspective is that DBFs conduct the core of the innovation process in biopharmaceuticals. By adopting basic science from universities, DBFs ‘‘make the technology ready for use’’ (Carlsson, 2002: 110). Subsequently, large pharmaceuticals take care of manufacturing, marketing and distribution. Although interesting as a benchmark for the Scanian regional innovation system, the use of university and big pharma bibliometrics would stretch the boundaries too much from a more narrowly defined biopharmaceuticals cluster perspective. The data revealed that 18 of the DBFs had scientific publications listed in SCI by the close of 2003, and 14 had patents granted and registered through the USPTO. In sum, 21 firms were either identified as patent assignee or appeared in the scientific publications producing in total 61 patents and 202 publications during the period 1990–2003. Since the paper’s main interest concerns knowledge collaboration, single publications and patents were excluded. This has resulted in an overall total of 183 publications and 58 patents. The findings are illustrated graphically in Figures 1–4. With respect to patents (Figures 1 and 2), the geographical distribution of non-local collaborations in the period 1990–1996 (involving seven Scanian DBFs) is concentrated in two nodes outside Scania: Stockholm and Frankfurt. No less than 61% of all collaborations are local. In the second period, 1997–2003, the geographical distribution of collaboration (involving 12 Scanian DBFs) is more dispersed but, again, concentrated in a few particular nodes. The share of local collaborative activity in the second period was about the same as in the previous period. The major nodes are found in Stockholm and Frankfurt, but with growing relevance of London, Munich, Gothenburg and Oslo. Comparing this with the general picture of biopharmaceutical megacentres as described by Cooke, it appears that scale of activity is relevant: four of six nodes qualify as megacentre candidates, while Gothenburg is the third largest Swedish bioregion. The outsider in this context is Oslo. These collaborations are, however, represented by one single firm which had an employee working part-time at the University of Oslo. 5 6 http://www.mva.org Two of the firms that were excluded only had a sales department in the region while the third had no activity at all. 404 L. Coenen et al. The distribution of co-publications (Figures 3 and 4) demonstrates some striking differences. Although a large share of the collaborations in 1990–1996 are also local and the non-local collaborations are concentrated to a few nodes, co-authors are far more distributed than co-inventors. The megacentres Stockholm/Uppsala, London and Paris are identified as important as are peripheral regions such as Umea (north of Sweden) and ˚ Minnesota (US). There is some collaboration with Copenhagen, at that time not yet connected by bridge. In the second period the pattern becomes considerably more diffused, with significant intercontinental linkages. Still, most of the collaboration activitiy is concentrated with counterparts in close-by Scandinavia and northern Europe. More strikingly though is the fact that 90% of all collaborations are with universities and/or other public research organizations. Another important finding is that the average number of inventors is about 3.5 in both periods, while the average number of authors has increased from 5 to about 5.5. The average number of addresses in the publication material was about 3 in both periods studied, which indicates some but not particularly high degrees of intraorganizational co-authorship. This quantitative analysis yields a general indication of the spatial distribution of collaborative innovation efforts among the clustered firms. However, it provides no answer to questions concerning the role of space, communities of practice and epistemic communities in various collaborative relations. In semi-structured interviews with research managers at nine of the 33 DBFs in Scania, more detailed explanations to the observed patterns were sought. In sum, 16 interviews have been conducted. Even though the firms differ considerably with respect to size and scope, they share some basic characteristics in the organization of their innovation activities as well as in their patterns of external collaboration. Innovation activities are mainly organized in cross-functional project teams with a limited number of external partners upon whom the firms are highly dependent. Face-toface meetings are considered important only in certain parts of the innovation process. This is especially the case in the early phases of development when no strict division of labor has been established yet. However, since the number of external partners is limited these meetings are handled relatively easily across long distances through frequent travels. In later phases the process becomes more strictly divided between highly specialized experts (or groups of experts), and most communication within the project group is about coordination rather than common problem-solving. Spatial proximity is less crucial for knowledge exchange. What matters is a high degree of shared cognitive understanding and affinity in terms of professional skills irrespective of geographical location. None the less, proximity is seen as an additional benefit for the collaboration in that it makes it easier to arrange project meetings, and it also makes it easier to retain administrative control over the projects. When identifying and selecting new partners, scientific profile is considered far more important than geographical location. Due to the extensive involvement of key researchers, scientific partners are often identified through the R&D manager’s professional network. Other sources are scientific publications, trade fairs and scientific conferences. Another growing source of information used in search for new partners is the Internet. Despite the decreasing weight of geographical proximity for knowledge collaboration, more than 40% of all co-authors are still located in Scania in the second period studied. A closer look at these collaborations reveals that nearly all are partners at Lund University. According to the interviewed R&D managers this is the case because they have been employed at these Socio-spatial Patterns of Innovation 405 Figure 1. Distribution of co-inventions 1990–1996 Figure 2. Distribution of co-inventions 1997–2003 particular departments themselves and feel confident collaborating with them. They know their specific area of competence and their scientific qualifications. Several of the studied firms have been founded by researchers at Lund University, and some still have part-time professorships and act as supervisors for PhD-students. The Agro-food Biotech Cluster of Saskatoon Ryan and Phillips (2004) conclude that the only biotechnology clusters in the world exclusively oriented to agro-food are in North America. The Saskatoon cluster represents 406 L. Coenen et al. Figure 3. Distribution of co-authorship 1990–1996 Figure 4. Distribution of co-authorship 1997–2003 one of the most advanced agricultural-dependent examples. It grew out of a series of events beginning with the establishment of the College of Agriculture and the University of Saskatchewan in the early part of the 20th century. The development of Canola in the 1970s—a low erucic acid and low glucosinolates variety of rapeseed—served as the defining event that launched the evolution of the cluster as it is reputed today. Saskatoon is acknowledged as ‘‘a key starting point for Canadian rapeseed efforts’’ (Khachatourians et al., 2001: 41). In the mid 1940s, the Canadian Agriculture Research Station began a continuous program on rapeseed research which, by 1998, was named the Centre of Socio-spatial Patterns of Innovation 407 Excellence in Canola. Over the years, researchers from a number of public institutions, in Canada and abroad, worked to develop better varieties of rapeseed. Using traditional breeding methodologies, key efforts led by Dr. Keith Downey at Agriculture Canada in Saskatoon resulted in the development of a low erucic and nutritionally superior variety of rapeseed—modern-day canola. Saskatoon-based public sector research served to advance the widespread acceptance of the canola-quality oil for consumption. The public sector has historically dominated R&D-based activity within the region. In fact, both the National Research Council-Plant Biotechnology Institute (NRC-PBI) and Agriculture and Agri-Food Canada (AAFC) have been identified as lead actors or ‘hubs’ in the Saskatoon cluster. However, it is the collective activities of both the public and private sectors throughout the 1980s and 1990s that led to the commercialization of the first transgenic canola crop in the mid 1990s (Khachatourians et al., 2001). More recently, the federal government’s injection of funding into genomics-based research activities (and public–private collaborations) in the region has represented a new, more evolved approach to knowledge generation activity and the building of renewed regional capacity. The success of the Saskatoon region lies in its ability to be responsive to changing market demands and for its actors (both private and public sector) to occupy a number of stages in the innovation chain (Phillips, 2002). This has meant cultivating and preserving connections with downstream markets (i.e. producers). The cluster remains flexible through its capacity for collective action through public–private partnerships and in its ability to accumulate and leverage know-how and know-why knowledge and research capacities to bring new products to the market. In this case, generation of synthetic knowledge is realized through multiple actors carrying out multiple activities over time, subject to technological and industrial change. Typically, basic research and many of the proprietary technologies are imported, assembled with locally produced germplasm into new crop varieties and then exported as intermediate product to global markets. As previously outlined, synthetic knowledge creation refers to the application or combination of existing knowledge. Field trials are a major integrating activity in the Saskatoon region, and throughout the province of Saskatchewan in general, in terms of agro-food and biotechnology. Respondents to Innovation Systems Research Network (ISRN) survey conducted in the Saskatoon area in 2001–2002 suggest that the region offers companies and organizations access to an arable land base to assess the physiological and agronomic traits of canola. In the process of bringing novel crop varieties to market, plants are grown in confined field areas and evaluated by government scientists (CFIA 2005). Saskatchewan is home to a significant portion of the field trials conducted in Canada. Over the period of 1990–1996, there were almost 3000 field trials conducted across Canada. On average per year, 47% of these field trials were located in Saskatchewan. This average decreased to 30% from 1997 to 2003 although the aggregate number remains at close to 3000 for this same period. From 1990 to 1996, an average of 89% of these Saskatchewan-based field trials involved canola varieties. This average drops to 53% in the latter period of 1997–2003. Reductions in terms of regional output and a deemphasis on canola varieties in Saskatchewan relative to national numbers signals a shift in the agro-food biotechnology market focus with a reduction in post-commercialization activity, particularly in canola. During the latter period, localized activity appears to switch to field trials in varieties of wheat, alfalfa and flax. 408 L. Coenen et al. The source population for the Saskatoon cluster was estimated in 2002 to include about 110 actors, including firms, associations, research institutes, government programs, departments, and venture capitalists. Research institutes represent the greatest portion of the source population (35%) and are the largest employers of biotechnology-dedicated workers. They include university colleges, departments and extension divisions that supply the local cluster with various research outputs, and federal, provincial and regional research organizations. Firms, representing 34% of the source population, include enterprises active in development and commercialization activities related to life science. All major agricultural biotechnology multinational enterprises had operations in Saskatoon in 2002 (many just 1–2 employees while others had up to 100 workers at some time over the period), in addition to some smaller DBFs. A further 15% of the source population was made up of venture capital operations. The remaining 17% of actors were government agencies and industry associations, all of which have aggressively positioned Saskatoon as a centre for global agro-food innovation (Phillips et al., 2004). As anticipated by Malerba (2004), private-sector activity is not as significant in agrofood biotech as most efforts are led by public-sector counterparts. This goes also for the Saskatoon cluster. Given this public-sector dominance in the region, as represented by actors as National Research Council-Plant Biotechnology Institute (NRC/PBI), Agriculture and Agri-Food Canada (AAFC) and the University of Saskatchewan (U of S), one would expect that the output (particularly in publications) over the 1996–2003 period would be substantial. In order to narrow the search to obtain a more manageable data set, the analysis of knowledge collaborations has been focused in two ways: (1) by research foci,7 and (2) by lead authors and inventors.8 The former are fair representations of regional market focus and the latter have been identified as significant contributors or ‘stars’ in terms of knowledge output for the Saskatoon region. The list of authors and inventors also includes individuals that work in the private sector, so although primary output in terms of patents and publications appears to be largely led by the public sector, we also want to account for any private-sector activities. Based on our search criteria, we found that regional activity in terms of publications amounted to 80 for the period 1990–1996 and 143 for 1997–2003, whereas patenting activity output totaled 31 for the period 1990–1996 and 24 for 1997–2003. Again, as collaborative activity is of major interest, sole publications and sole patents were excluded leaving a total of 138 publications and 50 patents. The geographical distribution of collaborative activity for 1990–1996, in terms of patents (Figure 1), was limited to North America. Local activity accounted for 83% of all linkages with another 7% attributed to collaborations with other Canadian regions (i.e. Calgary/Alberta, Guelph/Ontario, Ottawa/Ontario, Gatineau/Quebec). The remaining linkages (10%) were with counterparts in the United States from California, New York and Nebraska. Although patent activity marginally diminished in the second period (1997– 2003), from 27 to 23, the number of collaborators per patent increased slightly to 3.7 per patent from 3.2 per patent in 1990–1996. Local activity still accounts for the most significant 7 Search parameters Brassica5rapeseed, canola and brassica. Search parameters Animal-related5mucosal, immunity and vaccine. 8 Search parameters Brassica lead authors/inventors5Taylor, Zhou, Potts, Rakow, Keller, Datla, Georges, Dormann, Wang, Oelck, Robert, Ripley, Abrams, Gusta and Reaney. Search parameters Animal-related lead authors/ inventors5Bolton, Fontaine, Potter, Babiuk, Rioux, Shryvers, Mittal, Khachatourians and Acres. Socio-spatial Patterns of Innovation 409 collaborative activity at 81% and Canadian collaborations still accounting for 7% of activity. South of the border, collaborative activity diminishes slightly to 9%. However, these slight reductions in collaborative activity in North America are replaced with linkages overseas; one with the U.K. and one with India. In the latter time period, it appears that Saskatoonbased patent collaborations are showing signs of dispersing internationally (Figure 2). Consistent with the Scanian data, co-authorship between the Saskatoon region and its counterparts appeared far more distributed than in incidences of co-patenting activity (Figures 3 and 4). Local collaborations accounted for 69% of all collaborative activity across both time periods. Collectively, local collaboration and collaborations with other Canadian counterparts accounted for a total of 89% in 1990–1996 and 81% in 1997–2003. Overall, co-publishing activity appears to be dispersing significantly in the latter period. However, activity dispersed further in 1997–2003 in that the cluster actors collaborated with more countries and regions worldwide which also included Ethiopia, Australia, France, Switzerland, the Netherlands, England, India, South Korea, Belgium, China, Poland and Italy. In this latter period, collaborative activity dispersed and increased particularly with the United States. Publishing activity increased significantly in the second period (1997–2003) from a total of 80 articles to a total of 143 as does the number of collaborators per publication from 1.8 in 1990–1996 to 3.6 in 1997–2003. Overall, the public sector dominated all collaborative activity in publications and patents. In terms of collaborative activity in publications, the public sector—including all actors across all regions—accounted for 74% of all linkages in the time period from 1990 to 1996 and for 59% of same for 1997–2003. Private-sector activity in terms of publications more than doubled in the second time period, from 3% to 8%. Remaining collaborative activity (23% for 1990–1996 and 33% for 1997–2003) was led by the ‘quasi’ organization VIDO.9 Conclusions The aim of this paper has been to compare the socio-spatial patterns of innovation and knowledge linkages of a biopharmaceutical and an agro-food biotech cluster. Dissimilarities can be expected based on substantial differences in terms of historical technological regimes and sectoral innovation system dynamics between the agro-food and pharmaceutical industries in general. A distinction has been made between the more analytical knowledge base of biopharmaceuticals (consisting predominantly of natural sciences such biology, chemistry and medicine) and the more synthetic knowledge base of agro-food biotechnology (consisting of more engineering-based domains such as food technology and agriculture). Based on the theoretical discussion, it can be expected that knowledge dynamics in the agro-food cluster are more localized than in the biopharmaceuticals cluster. When comparing the clusters’ distribution of collaborative relations, the theoretical assumptions are supported. Three major results should be highlighted. First, the agro-food biotech cluster displays a less spatially distributed pattern of collaboration than the biopharmaceutical cluster, as measured in average number of inventors per patent and 9 VIDO is renowned for the research, development and commercialization of products used by producers in the food animal industry and represents a unique organizational configuration. It is a non-profit organization owned by the University of Saskatchewan, however, it is considered financially self-reliant. 410 L. Coenen et al. average number of authors per publication. The agro-food case displays an increased degree of co-authorship in the period 1997–2003 (from 1.8 to 3.6), but still much lower than the biopharmaceuticals case (5–5.5). Hence, while there is significant inter-organizational exchange of knowledge in biopharmaceuticals, research and innovation in agro-food biotech seems to be more concentrated and integrated within large organizational bodies. This should, however, be seen in a comparative perspective given that the findings on the agro-food biotech cluster also point at the importance (and validity) of a distributed knowledge base. Second, the agro-food biotech cluster of Saskatoon displays more locally oriented patterns of collaboration than the Scanian biopharmaceutical cluster, both in terms of patents and publications. While the share of local co-authorship decreased sharply in the biopharmaceuticals case (from 65% to 42%) the agro-food figures remained constant at 69%. If other collaborations within Canada were included, the agro-food figures are 81%. In the biopharmaceuticals case only 55% of all collaborations are within Sweden. With respect to co-inventions, the agro-food figures of local collaboration (81%) remain much higher than in the biopharmaceuticals cluster (66%). None of the regions displayed any obvious change in spatial distribution over time. When including domestic collaborations, the Saskatoon figure rises to 90%, while the corresponding figure for Scania is 73%. Third, the non-local linkages of the agro-food biotech cluster of Saskatoon are spatially more diffused than the non-local linkages of the Scanian biopharmaceutical cluster. The Scanian linkages are first and foremost connected to Europe and North America, in large part to other well-known biopharmaceutical nodes like Stockholm, Gothenburg, Munich and California. Many of the Saskatoon non-local linkages are connected to places like Guelph, Ottawa, Winnipeg and Dusseldorf. This illustrates the difference in search and collaboration patterns for relevant ¨ knowledge between biopharmaceuticals and agro-food biotech. Furthermore, our in-depth studies of biopharmaceutical DBFs confirmed the importance of individual scientists at world leading universities with a suitable scientific profile. Collaboration partners were almost exclusively chosen on the basis of this criterion, regardless of geographical location. Spatial proximity was appreciated for easing collaboration, but long distance was not seen as a prime obstacle. In the agro-food biotech cluster most innovation activity is locally mediated. One hundred percent of the germplasm used to create new crop varieties is locally derived and assembled within the region. Most of the production and commercialization (66%) of new plant varieties are conducted locally (Phillips, 2002). However, Saskatoon’s innovation activity is concentrated within a segment of the broader food value chain. Its network of actors rely heavily upon intellectual properties and knowledge captured from non-local or imported sources (almost 65%) to create innovative raw or intermediate product (synthetic knowledge) that may be then exported (up to 80%) for further upstream processing (Phillips, 2002). Taken together, these three findings support our theoretical expectation that spatial patterns of knowledge collaboration and innovation differ between biopharmaceuticals and agro-food biotech due to dissimilarities in their knowledge bases. However, it is important to acknowledge that these differences are relative. Both sectors display local and non-local patterns of collaboration following the general pattern for biotechnology. This becomes particularly evident in the period after 1997. In sum, what this paper primarily shows is that biotechnology should not be conceived as one homogeneous industry. Instead a distinction based on ‘carrier’ industries (in this case pharmaceuticals and agro-food) is essential to more fully understand the way that biotechnology is influencing our economies and societies. From a policy point of view, a Socio-spatial Patterns of Innovation 411 more demand-inclined perspective would complement the currently dominating technology push rationale. Even though our empirical results support our theoretical assertions, differences and similarities between biopharmaceutical and agro-food biotech clusters need to be investigated further, also based on more qualitatively inclined case studies of innovation projects, in order to gain more insight on the suggested differences between epistemic communities and communities of practice. In both cases learning and innovation remain inherently social processes. While locally available ‘untraded interdependencies’ and frequent (face-to-face) interaction remain crucial territorial anchors for knowledge production and transmission in communities of practice, epistemic communities allow for complimentary, extra-local knowledge conduits based on sociality embedded in professional affinity and scientific practice. It needs to be stressed that communities of practice and epistemic communities should not be seen as each other’s antipodes but rather as two substantially different but nonetheless complimentary social circuits for knowledge production and innovation. Thus, acknowledging that clusters need local as well as extralocal knowledge, it is important to allow for the possibility of multiple embeddedness (Hess, 2004) where it is not so much a question of either/or. Similar to the multi-level governance perspective in innovation systems, the interconnectedness of various geographical scales ought to be stressed as a starting point for analyzing (biotechnology) clusters. Acknowledgements An earlier version of this paper has benefited from being presented at the 2005 Annual Meeting of the AAG (American Association of Geographers), Denver, Colorado, 5–9 April and the ‘‘Bringing Science to Life’’ conference held at the University of Toronto, Canada, 29 April–1 May 2005. The authors wish to thank the guest editor of this special issue, Maryann Feldman, as well as Meric Gertler, Proinnsias Breathnach and two anonymous referees for valuable comments. The maps have been developed at the Canada Rural Economy Research Lab (C-RERL)–University of Saskatchewan. Financial support is gratefully ¨ acknowledged from the Swedish Science Council and Oforsk. References Amin, A. and Cohendet, P. (2004) Architectures of Knowledge: Firms, Capabilities and Communities (Oxford: Oxford University Press). Archibugi, D., Howells, J. and Michie, J. (Eds) (1999) Innovation Policy in a Global Economy (Cambridge: Cambridge University Press). Asheim, B. T. and Coenen, L. (2005) Knowledge bases and regional innovation systems: comparing Nordic clusters, Research Policy, 34(8), pp. 1173–1190. Asheim, B. T. and Gertler, M. S. (2005) regional innovation systems and the geographical foundations of innovation, in: J. Fagerberg, D. Mowery & R. Nelson (Eds), The Oxford Handbook of Innovation (Oxford: Oxford University Press). Bathelt, H. (forthcoming) Geographies of production: growth regimes in spatial perspective 2—Knowledge creation and growth in clusters, Progress in Human Geography, 29(2), pp. 204–216. Bathelt, H., Malmberg, A. and Maskell, P. (2004) Clusters and knowledge: local buzz, global pipelines and the process of knowledge creation, Progress in Human Geography, 28(1), pp. 31–56. Benner, C. (2003) Learning communities in a learning region: the soft infrastructure of cross-firm learning networks in Silicon Valley, Environment and Planning A, 35, pp. 1809–1830. Benneworth, P. (2003) Breaking the mould: new technology sectors in an old industrial region, International Journal of Biotechnology, 5, pp. 249–268. Biotech Sweden, (2004) The Swedish Biotech Industry Guide (Stockholm: Biotech Sweden). 412 L. Coenen et al. Breschi, S. and Malerba, F. (1997) Sectoral innovation systems: technological regimes, schumpeterian dynamics and spatial boundaries, in: C. Edquist (Ed.), Systems of Innovation: Technologies, Institutions and Organisations (London: Pinter). Brink, J., McKelvey, M. and Smith, K. (2004) Conceptualizing and measuring modern biotechnology, in: M. McKelvey, A. Rickne & J. Laage-Hellman (Eds), The Economic Dynamics of Modern Biotechnology (Cheltenham: Edward Elgar). Brown, J. S. and Duguid, P. (2000) The Social Life of Information (Boston: Harvard Business School Press). Carlsson, B. (2002) Institutions, entrepreneurship, and growth: biomedicine and polymers in Sweden and Ohio, Small Business Economics, 19, pp. 105–121. Casper, S. and Matreves, O. (2003) Institutional frameworks and innovation in the German and UK pharmaceutical industry, Research Policy, 32(10), pp. 1865–1879. CFIA Canadian Food Inspection Agency, Regulatory Initiatives. Available online at: http://www.inspection.gc.ca/english/reg/consultation/ reginite.shtml (accessed 15 June 2005). Coenen, L., Moodysson, J. and Asheim, B. T. (2004) Nodes, networks and proximities: on the knowledge dynamics of the Medicon Valley biotech cluster, European Planning Studies, 12(7), pp. 1003–1018. Cooke, P. (2002) Rational drug design, the knowledge value chain and bioscience megacentres., Paper prepared for International Workshop ‘Clusters in High-Technology Aerospace, Biotechnology and Software Compared’, Department of Management and Technology, Universite du Quebec a Montreal, November 2002. ´ ´ ` ´ Cooke, P. (2004) Editorial: the accelerating evolution of biotechnology clusters, European Planning Studies, 12(7), pp. 915–920. Cooke, P. (2005) The evolution of biotechnology in bioregions and their globalisation., Paper prepared for The ‘Unlocking Biotechnology’ Sub-theme of the EGOS Colloquium 30 June–2 July 2005, Berlin. Cooke, P. and Leydesdorff, L. (2006) Regional development in the knowledge-based economy: the construction of advantage, Journal of Technology Transfer, 31(1), pp. 5–15. Cowan, R., David, P. and Foray, D. (2000) The explicit economics of knowledge codification and tacitness, Industrial and Corporate Change, 9(2), pp. 211–253. Doloreux, D. (2004) Regional innovation systems in Canada: a comparative study, Regional Studies, 38, pp. 479–492. Doloreux, D. and Parto, S. (2005) Regional innovations systems: current discourse and unresolved issues, Technology in Society, 27, pp. 133–153. Edquist, C. (1997) Introduction, in: C. Edquist (Ed.), Systems of Innovation: Technologies, Institutions and Organisations (London: Pinter). Edquist, C. and Johnson, B. (1997) Institutions and organizations in systems of innovation, in: C. Edquist (Ed.), Systems of Innovation: Technologies, Institutions and Organisations (London: Pinter). Etzkowitz, H. and Leydesdorff, L. (2000) The dynamics of innovation: from national systems and ‘mode 2’ to a triple helix of university–industry–government relations, Research Policy, 29, pp. 109–123. Feldman, M. P. (2000) Where science comes to life: university bioscience, commercial spin-offs, and regional economic development, Journal of Comparative Policy Analysis: Research and Practice, 2, pp. 345–361. Fuhrer, B. (2003) The NEXT biotechnology—beyond the attempts at bloning biotech clusters., PhD thesis, Hrsg. Institut fur Geographie ¨ Unviversitat Hamburg. Hamburger Beitrage zur Geographischen Forschung, Band 4. ¨ ¨ Gertler, M. (2004) Manufacturing Culture: The Institutional Geography of Industrial Practices (Oxford: Oxford University Press). Gertler, M. and Levitte, Y. (2005) Local nodes in global networks: the geography of knowledge flows in biotechnology innovation, Industry and Innovation, 13, pp. 487–507. Granovetter, M. S. (1985) Economic action and social structures: the problem of embeddedness, The American Journal of Sociology, 91(3), pp. 481–510. Hess, M. (2004) ‘Spatial’ relationships? Towards a reconceptualization of embeddedness, Progress in Human Geography, 28(2), pp. 165–186. Howells, J. (1999) Regional systems of innovation?, in: D. Archibugi et al. (Eds), Innovation Policy in a Global Economy (Cambridge: Cambridge University Press). Kemp, R., Schot, J. and Hoogma, R. (1998) Regime shifts to sustainability through processes of niche formation: the approach of strategic niche management, Technology Analysis and Strategic Management, 10, pp. 175–196. Khachatourians, G., Sumner, A. and Phillips, P. (2001) An introduction to the history of canola and the scientific basis for innovation, in: P. Phillips & G. Khachatourians (Eds), The Biotechnology Revolution in Global Agriculture: Invention, Innovation and Investment in the Canola Sector (Wallingford: CABI publishing). Socio-spatial Patterns of Innovation 413 Klopper, C. and Haisch, T. (2006) Satellite or star—the evolution of the Swiss biotech and medtech industry and the influence of the ¨ pharma-shaped innovation system. Paper presented at the AAG Annual Meeting, Chicago, 7–11 March. Knorr-Cetina, K. (1999) Epistemic Cultures: How the Sciences Make Knowledge (Cambridge, MA: Harvard University Press). Kuhn, T. (1970) The Structure of Scientific Revolutions, 2nd edn (Chicago: University of Chicago Press). Laestadius, S. (2000) Biotechnology and the potential for a radical shift of technology in forest industry, Technology Analysis and Strategic Management, 12(2), pp. 193–212. ¨ Lagnevik, M., Sjoholm, I., Lareke, A. and Ostberg, J. (2003) The Dynamics of Innovation Clusters: a Study of the Food Industry ¨ (Cheltenham: Edward Elgar). ¨ Lundquist, K-J. and Winther, L. (2003) Between Denmark and Sweden: the industrial dynamics of the Oresund region., Paper prepared for RSAs (Regional Science Association) annual conference in Pisa, 12–15 April. ˚ Lundvall, B-A. (Ed.) (1992) National Systems of Innovation: Towards a Theory of Innovation and Interactive Learning (London: Pinter). ˚ Lundvall, B-A., Johnson, B. and Lorenz, E. (2002) Why all this fuss about codified and tacit Knowledge?, Industrial and Corporate Change, 2, pp. 245–262. Malerba, F. (Ed.) (2004) Sectoral Systems of Innovation: Concepts, Issues and Analyses of Six Major Sectors in Europe (Cambridge: Cambridge University Press). Malerba, F. (2005) Sectoral systems: how and why innovation differs across sectors, in: J. Fagerberg, D. Mowery & R. Nelson (Eds), The Oxford Handbook of Innovation (Oxford: Oxford University Press). Malerba, F. and Orsenigo, L. (2001) Innovation and market structure in the dynamics of the pharmaceutical industry and biotechnology: towards a history friendly model. Paper presented at the DRUID Nelson and Winter Conference, Aalborg, 12–15 June 2001. Malmberg, A. (2003) Beyond the cluster—local milieus and global connections, in: J. Peck & H. Yeung (Eds) (2003) Remaking the Global Economy (London: Sage Publications). Maskell, P. and Malmberg, A. (1999) Localised learning and industrial competitiveness, Cambridge Journal of Economics, 23, pp. 167–185. Morgan, K. (2004) The exaggerated death of geography: learning, proximity and territorial innovation systems, Journal of Economic Geography, 4, pp. 3–21. Nelson, R. and Winter, S. (1977) In search of useful theory of innovation, Research Policy, 6, pp. 36–76. Nightingale, P. (2000) Rethinking pharmaceutical innovation: risk, institutional legacies and the integration of specialised knowledge. Paper for the Schumpeter Conference, Manchester UK, June 2000. Nightingale, P. and Martin, P. (2004) The myth of the biotech revolution, Trends in Biotechnology, 22(11), pp. 564–569. Nilsson, A. (2001) Biotechnology firms in Sweden, Small Business Economics, 17, pp. 93–103. Nilsson, M., Svensson-Henning, M. and Wilkenson, O. (2002) Skanska kluster och profilomraden—en kritisk granskning (Malmo: Region ˚ ˚ ¨ Skane). ˚ Nooteboom, B. (2000) Learning and Innovation in Organisations and Economies (Oxford: Oxford University Press). OECD, (2004) Science, technology and innovation for the 21st century., Final Communique meeting 29–30. Onsager, K. and Aasen, B. (2003) The case of ‘‘Rogaland Regional Food Cluster’’, in: B. T. Asheim, L. Coenen & M. Svensson-Henning (Eds), Nordic SMEs and Regional Innovation Systems—Final Report (Oslo: Nordic Industrial Fund). Owen-Smith, J. and Powell, W. W. (2004) Knowledge networks as channels and conduits: the effects of spillovers in the Boston biotechnology community, Organization Science, 15(1), pp. 5–21. Phillips, P. (2002) Regional systems of innovation as a modern R&D Entrepot: the case of the Saskatoon biotechnology cluster, in: J. ˆ Chrisman et al. (Eds), Innovation, Entrepreneurship, Family Business and Economic Development: A Western Canadian Perspective (Calgary: University of Calgary Press). Phillips, P., Parchewski, P., Procyshyn, T., Ryan, C., Karwandy, J. and Kihlberg, J. (2004) Agricultural and life-science clusters in Canada: an empirical and policy analysis. Report on the Saskatoon ag biotech cluster for Agriculture and AgriFood Canada (AAFC), March. Available online at: http://www.agbio-management.org/. Porter, M. E. (1990) The Competitive Advantage of Nations (Worchester: Billing and Sons Ltd). Ryan, C. and Phillips, P. (2004) Knowledge management in advanced technology industries: an examination of international agricultural biotechnology clusters, Environment and Planning C: Government & Policy, 22, pp. 217–232. Scott, W. R. (1995) Institutions and Organisations (London: Sage Publications). Smith, K. (2000) What is ‘The Knowledge Economy’? Knowledge-intensive Industries and Distributed Knowledge Bases. Paper presented at the DRUID Summer Conference, Aalborg, Denmark, June 2000. 414 L. Coenen et al. Storper, M. (1997) The Regional World: Territorial Development in a Global Economy (London and New York: Guilford Press). Torre, A. and Gilly, J-P. (2000) On the analytical dimension of proximity dynamics, Regional Studies, 34(2), pp. 169–180. VINNOVA, (2001) The Swedish Biotechnology Innovation System (Stockholm: VINNOVA). VINNOVA, (2003) Swedish Biotechnology—Scientific Publications, Patenting and Industrial Development (Stockholm: VINNOVA). Wilkinson, J. (2002) The final foods industry and the changing face of the global agro-food system, Sociologia Ruralis, 4, pp. 329–246. Wolfe, D., Davis, C. and Lucas, M. (2004) Global networks and local linkages: an introduction, in: D. Wolfe & M. Lucas (Eds), Global Networks and Local Linkages: The Paradox of Cluster Development in an Open Economy (Montreal and Kingston: McGill-Queen’s University Press). Zeller, C. (2004) North Atlantic innovative relations of Swiss pharmaceuticals and the proximities with regional biotech arenas, Economic Geography, 80(1), pp. 83–112.