CC BY
173
KNOWLEDGE FLOWS IN HIGH-TECH INDUSTRY CLUSTERS: DISSEMINATION MECHANISMS AND INNOVATION REGIMES
БО КАРЛСОН. ПОТОКИ ЗНАНИЙ В ИНДУСТРАЛЬНЫХ HIGH-TECH КЛАСТЕРАХ: МЕХАНИЗМЫ РАСПРОСТРАНЕНИЯ И ИННОВАЦИОННЫЕ РЕЖИМЫ. Этот документ исследует потоки знаний, то есть создание и распространение знаний в трех типах кластеров, с целью заложить концептуальную основу для анализа кластеров в промышленности и в области технологий. Проводится различие между двумя режимами различных инноваций: discovery-driven инновацией, представленной Силиконовой долиной и Кембриджем (Великобритания) в области полупроводников и Бостоном/ Кембриджем (Сан-Франциско Bay Area и Medicon долина) в области биотехнологий; и design-driven инновацией, представленной Boeing в Сиэтле, Bombardier в Монреале, Airbus в Тулузе и Saab в Линчё-пинг в авиационной промышленности. В каждом кластере рассматривается роль университетов и других создателей знаний. Также анализируются характер распространения знаний, различия между рынком опосредованной передачи знаний и нерыночной опосредованной и неориентированной передачей («true» spillove...).
This paper explores knowledge flows, i.e., creation and dissemination of knowledge, in three types of clusters in order to lay a conceptual foundation for analysis of knowledge-based industry clusters and for technology policy. Distinction is made between two different innovation regimes: discovery-driven innovation, represented by Silicon Valley and Cambridge, UK, in semiconductors, and by Boston/Cambridge, the San Francisco Bay Area and Medicon Valley in biotechnology; and design-driven innovation as represented by Boeing in Seattle, Bombardier in Montreal, Airbus in Toulouse, and Saab in Linkoping in the aircraft industry. In each cluster, the role of universities and other creators of knowledge is examined. The nature of knowledge dissemination is also analyzed, distinguishing between market-mediated transfers of knowledge and non-market mediated and undirected transfers ("true" spillovers). The role of new start-ups versus incumbent firms in knowledge dissemination and cluster growth is also examined. Revised September, 2011
Bo Carlsson,
Case Western Reserve University, Cleveland, Ohio, U.S.A.
Introduction
New knowledge is an important driver of economic growth. Much of the economic growth literature in the last couple of decades has focused on the role of knowledge creation and diffusion . The theory of "endogenous growth" (see for example Romer, 1986 and 1990, and Lucas, 1988) is based on the idea of knowledge spillovers emanating from R&D . Although this is a useful contribution to our understanding of economic growth, it has led to an overly simplistic focus of public policy on knowledge creation. The theory specifies neither the nature nor the mechanisms of spillovers, with the
unfortunate result that the term "knowledge spillover" has come to be used much too frequently and loosely. The purpose of the present paper is to show that it matters where and by whom new knowledge is created as well as how it is disseminated. Focus needs to be on the process, not only on the outcome (Werker and Athr-eye, 2004). What is needed is an evolutionary approach rather than a traditional (orthodox) one .
A basic idea in traditional theory is that new knowledge is a public good which gives rise to economic growth when it is applied in economic activity and that its benefits increase the more widely it is applied. But in a world of uncertainty, lack of appropriability, indivisibilities, and externalities there are insufficient incentives to engage in R&D This results in market failure, leading to underinvestment in knowledge creation and weak mechanisms for knowledge transfer
(Metcalfe, 1994) . The implication for public policy in standard theory is that knowledge creation and knowledge diffusion (R&D) should be stimulated and intellectual property rights protected . However, this general policy prescription needs considerable adjustment in a world characterized by large differences among firms, bounded rationality, and various other asymmetries. The claim of this paper is that a more solid basis for technology policy requires a better understanding of the processes of knowledge creation and dissemination as well as how "spillovers' relate to these processes. Evolutionary theory provides an appropriate analytical framework for such a discussion. (See e . g . Metcalfe 1994, 1995; Cantner and Pyka, 2001; Smits et al , 2010) The main components of the necessary theoretical framework are incorporated in the theory of innovation systems (Carlsson & Stankiewicz, 1991).
It is useful for the purposes of this analysis to examine knowledge flows in knowledge-based (hightech) industry clusters which may also be referred to as technological innovation systems (Carlsson, 1995, 1997, 2002) .1 However, while there are many studies of industry clusters or innovation systems, they are often more descriptive than analytical, focusing more on their geographic and institutional dimensions than on the nature of knowledge and knowledge flows that are at their core . In this paper the focus is on knowledge flows in particularly knowledge-intensive industry clusters . Having reviewed the literature, I chose to examine three domains which are represented by several studies covering more than one geographic area, namely biotechnology, semiconductors, and aerospace. I wanted to see whether there are common features of knowledge creation and dissemination in these knowledgeintensive clusters . What emerged from this explorative study were distinct patterns of knowledge flows with respect to the mechanisms for knowledge dissemination and innovation regime . The idea is to generate, not to test, new theory. For considerations of space, the evidence is presented here according to the patterns that emerged
The central questions in this paper are: Where does the knowledge come from that constitutes the core of knowledge-based industry clusters (innovation systems)? How is the knowledge disseminated — via market mechanisms (such as technology transfer) or non- market mechanisms (spillovers), and what are the implications for the organizational structure of the clusters and for public policy?
1 For a survey of the literature on innovation systems, see Carlsson (2007).
It is demonstrated that the sources of knowledge and the vehicles of dissemination of knowledge differ among knowledge-intensive (high-tech) industry clusters depending on whether innovation is design-driven or discovery-driven. In design-driven systems or clusters such as in the aerospace industry, technology sharing and transfer is typically market-mediated; new knowledge tends to be created in large firms, and the role of universities is primarily to supply skilled labor. By contrast, in discovery-driven innovation such as in the biotechnology and semiconductor industries, universities play a much more prominent role as creators of new knowledge, and technology sharing usually involves both market-mediated transfer and true spillovers . As a consequence, the two types of clusters are organized differently, and there are important implications for the role of public policy
The paper is organized as follows I begin with a brief review of the knowledge spillover literature and a discussion of dissemination mechanisms . The next section discusses various types of clusters, how and why they have evolved, and the knowledge flows within each The fourth section reviews the literature on knowledge flows in high-tech industry clusters characterized by different innovation regimes. The fifth section examines the sources of knowledge and mechanisms of knowledge diffusion and how these have evolved over time in two types of discovery-driven clusters: (1) the microelectronics clusters in Silicon Valley and in the Cambridge (UK) area; and the biotechnology clusters in Boston/Cambridge (Massachusetts), the San Francisco Bay Area, and Medicon Valley (spanning the Copenhagen-Malmo region in Denmark and Sweden) Design-driven clusters in the aircraft industry are also examined: Boeing (Seattle), Bombardier (Montreal), Airbus (Toulouse), and Saab (Linkoping, Sweden) . In each case I am interested primarily in the sources of knowledge, how knowledge flows differ and especially the role of university research versus industrial R&D, how these interrelate, what the mechanisms of interaction are, the character and importance of "anchor tenants' located within the cluster and their links to outside entities, and how these relationships have evolved over time Where (inside or outside the cluster) and by whom (academia or business) does knowledge generation take place? To what extent is it appropriate to speak of "spillovers' (unintended, non- market mediated) in reference to knowledge flows? The final section summarizes the argument, draws conclusions, and states the policy implications
Knowledge Spillovers And Economic Growth
In economics, an externality or spillover of an economic transaction is defined as an impact on a party that is not directly involved in the transaction. It was noted long ago by Abramovitz (1956) and Solow (1956) that only a small fraction of macroeconomic growth in the United States can be attributed to increased inputs of labor and capital, the rest (the "residual") being attributable to other factors, particularly technological change — "a measure of our ignorance" as Abramovitz put it . The contribution of endogenous growth theory is to include technology explicitly in the production function, arguing that the remaining unexplained residual is due to R&D spillovers (and measurement errors). This residual constitutes the benefits reaped elsewhere in the system in addition to those appropriated by those who made the R&D investment. As pointed out by Griliches (1992), this results in problems in measuring spillover effects: it is difficult to distinguish between true externalities in the form of (unappropriable) knowledge spillovers and market-mediated knowledge transmissions that are hard to price accurately and that therefore result in measurement errors
Griliches reviewed the basic model of R&D spillovers (based on the knowledge production function) and focused on the empirical evidence for their existence and magnitude. He found that "taken individually, many of the studies are flawed and subject to a variety of reservations, but the overall impression remains that R&D spillovers are both prevalent and important"(Griliches, 1992, p S29) He distinguished between two types of R&D "spillovers': One represents knowledge embodied in capital equipment and involves the problem of measuring capital equipment, materials and their prices correctly. The foremost example here is the computer industry As computers have improved and their price has come down, different industries have benefited differentially, depending on their rate of computer purchases But according to Griliches, "these are not real knowledge spillovers . They are just consequences of conventional measurement problems True spillovers are ideas borrowed by research teams of industry i from the research results of industry j. It is not clear that this kind of borrowing is particularly related to input purchase flows' (Griliches, 1992, p . S36) . The second type of spillovers, "true spillovers," involves disembodied knowledge . "The assumption is made that two firms that are active in the same technological areas, as indicated by their taking out patents in the same patent classes, are more likely to benefit from each other's research results " (p S39)
In an influential paper, Jaffe (1989) brought the analysis of spillovers from the macroeconomic to the regional level He studied spillovers from university research to commercial innovation using state-level time-series data on corporate patents, corporate R&D, and university research . He found a significant effect of university research on corporate patents at the state level in a few industries, particularly in drugs and medical technology, electronics, optics, and nuclear technology Subsequently, Jaffe et al. (1993) found that patent citations are geographically concentrated to local metropolitan areas
Anselin et al. (1997) estimated knowledge production functions at both the state and the metropolitan statistical area (MSA) levels. They found strong evidence of local spillovers at the state level. At the MSA level, they distinguished between industrial research and development activities and university research in the MSA and in the surrounding counties and found evidence of local spatial externalities between university research and high technology innovative activity, both directly and indirectly via private research and development
Feldman (1999) reviewed four separate strains in the empirical spillover literature: innovation production functions, the linkages between patent citations, the mobility of skilled labor based on the notion that knowledge spillovers are transmitted through people, and knowledge spillovers embodied in traded goods. Feldman then examined the composition of agglomeration economies, the attributes of knowledge, and the characteristics of firms . She found that knowledge spillovers from science-based activities are localized and contribute to higher rates of innovation, increased entrepreneurial activity, and increased productivity within geographically bounded areas The main mechanisms of knowledge spillover are patent citations and movements of people and traded goods There is evidence that knowledge spillovers are limited in the spatial dimension in some domains but not necessarily in others (Feldman, 1999, pp . 20-21).
Breschi and Lissoni (2001) reviewed the literature of knowledge spillovers in relation to industry clusters . They distinguished between three kinds of externalities: economies of specialization, labor market economies, and knowledge spillovers (Breschi and Lissoni, 2001, p. 978) The first two of these are pecuniary externalities (knowledge flows mediated by the market mechanism); the third is a technological externality (pure spillover) to the extent that it involves unintended and non-market mediated transfer of knowledge
Thus, Breschi and Lissoni take a critical view of the literature:
"The major limitation of the empirical literature we have reviewed... is that virtually no contribution has explored the ways in which knowledge is actually transferred among people located in the same geographic area... We need to explore the price and non-price mechanisms through which knowledge may be traded between universities and firms (or individuals therein), as well as between firms. First and foremost, we observe that much of knowledge transmitted from universities to firms has nothing to do with the public results of basic science, but consists of consultancy services to firms . Rather than providing innovation opportunities, such knowledge transfer may enhance the customer firms' appropriation capabilities." Breschi & Lissoni, 2001, p. 994)
Local academic institutions and public research institutes often provide critical inputs for firms' innovative activities, such as training and consultancy, even if their current research is not directly relevant to those activities By producing graduates and offering services (or tolerating their staff doing so), universities contribute to enhancing absorptive capacity of firms even if their research is not on the frontier. Hiring of skilled personnel increases the absorptive capacity of the firm and thus enables the firm to take advantage of spillovers
Arikan (2009) studied inter-firm knowledge exchanges and the knowledge creation capability of clusters He found that these exchanges typically take place through frequent interactions among cluster firms and that they that take various forms, from vertical supplier-buyer relations to horizontal alliances, licensing agreements, and research consortia — all of which are market-mediated . In addition, geographic proximity increases the frequency of interactions among cluster firms as well as the effectiveness of knowledge exchanges through these interactions; face-to-face contact between firm members contributes to the building of inter-firm trust and institutional norms of cooperation. " (Arikan, 2009, p . 658)
"A cluster that has a high level of knowledge creation capability is one where knowledge held by individual firms is effectively shared among cluster firms through interfirm knowledge exchanges and amplified by individual firms' knowledge spirals, leading to enhanced knowledge creation by individual firms. " (Arikan, p. 660) In other words, a high level of absorptive capacity increases the probability of identifying and benefiting from spillovers .
The problem with much of the literature on knowledge spillovers and economic growth is that it fails to distinguish between knowledge transfers (targeted sharing or dissemination of knowledge) and true spillovers (externalities) . Only a fraction of new knowledge is economically useful, and only a small fraction of economically useful knowledge is commercialized (via new products in existing firms, licenses, or new start-ups) (Carlsson & Fridh, 2002) . Some of the new knowledge created in academic institutions is published, but the bulk of it is embodied in the students who carry it into the labor force (Carlsson et al. , 2009) . Most R&D is targeted; about 60 % of total R&D in the United States involves development, 22 % is applied, and only 18 % is basic R&D, mostly untargeted (source: NSF). Most of the basic R&D is carried out in academic institutions And while basic R&D has shifted increasingly towards universities (away from business) in recent decades, it is only a few top universities that are capable of producing basic R&D of sufficient quality to give rise to new business opportunities (Mansfield, 1995). Most of the basic R&D carried out by business firms tends to enhance their absorptive capacity rather than pushing out the knowledge frontier; the utilization of new skills acquired through hiring of new PhDs is part of this process. These are certainly important knowledge transfers, but they take place via (admittedly imperfect) market mechanisms. They are not spillovers in a true sense. True spillovers are not market-mediated; they are the result of externalities in the form of knowledge transferred or acquired from outside sources without intent or direction on the part of the inventor and without compensation. Knowledge acquired from publications or patents or via employees or students leaving to start a new firm without payment to the employer represents spillovers . Buying a license or a piece of equipment, hiring of skilled workers, or acquiring knowledge via joint ventures, alliances, or mergers and acquisitions are intentional transfers mediated via markets; they do not involve true knowledge spillovers. 2 As will be shown below, true spillovers are the main raison d'être for some hightech clusters . In other clusters in which knowledge is disseminated via transfers, the reasons for co-locating may be more conventional
Types Of Industry Clusters
The term "industry cluster" became prevalent in the economic literature around 1990 (see e g Krugman, 1991,
2 This is not to suggest that market-mediated knowledge transfers are unimportant - on the contrary. They are the dominant mechanisms of knowledge diffusion. But they are not true spillovers.
and Porter, 1990 and 1998), but it has been used rather loosely. Porter (1990) defined an industry cluster as a geographically proximate group of firms and associated institutions in related industries, linked by economic and social interdependencies. This is the definition most commonly used in the literature . Gordon & McCann (2000) distinguished between three different interpretations of industry cluster: the classic model of "pure agglomeration" based on the (neo-) classical tradition in economics, the industrial complex model of tight integration and stable relationships among firms, and the social network model built on interpersonal trust and relationships transcending firm boundaries There are many different types of clusters, depending on the type of economic activity involved as well as stage of development . Much of the literature refers to Alfred Marshall's Principles of Economics (1920), the first edition of which was published in 1890
As Marshall pointed out, many industrial activities tend to cluster in certain geographic regions Marshall distinguished between regional agglomerations and "industrial districts. " He referred to the former as "elementary localization of industry" which "gradually prepared the way for many of the modern developments of division of labor in the mechanical arts and in the task of business management" that characterize industrial districts (Marshall, 1920, p . 268). According to Marshall, there are three primary causes of localization of industries: non-tradable inputs (physical conditions such as climate, soil, and access to raw materials), "patronage of a court" (demand for goods of high quality), and "the presence of a town" (urbanization economies, i. e . , a sufficient number of customers) (ibid., pp . 268-269). Once an agglomeration has emerged, it may be transformed over time into an industrial district if certain advantages are acquired: a local market for special skills that can be passed on to the next generation (mysteries of the trade are no longer mysteries but are "in the air"); growth of subsidiary trades; and use of highly specialized machinery (Belussi and Caldari, 2009, p . 337) . The resulting industrial district is the locus of economic activity that makes up a large fraction of an industrial economy; it represents the ordinary growth process — what Schumpeter would refer to as "economic growth" in the stationary state . Universities, government policies, and public laboratories play a modest role in these districts; they are self-organized agglomerations of private firms competing in similar markets, together with specialized suppliers of equipment and services (Niosi and Zhegu, 2005, p . 3). There are not many knowledge externalities (true spillovers)
associated with these districts, since they are not knowledge-based .
To get to a more dynamic stage ("economic development" in Schumpeter's terminology) two additional factors are needed — emphasized by Marshall in both his Principles and Industry and Trade (1923), although not specifically in connection with his discussion of industrial districts: what he calls "industrial leadership" (i e , entrepreneurship), and "introduction of novelties' (i e , innovation). These additional elements make it possible to break out of mere "organic" growth into a more dynamic phase, transforming an industrial district into what we may call a rapidly growing technology-based industrial cluster. These are the types of clusters with which this paper is concerned
According to this interpretation of Marshall, there are two key elements to look for in the formation of clusters: a pre-existing local or regional agglomeration ("industrial district") of economic activity and a scaling-up of that activity through entrepreneurship and innovation
In discussing the organization of industry clusters, Markusen (1996) distinguishes between (1) Marshal-lian "industrial districts' consisting mainly of locally owned SMEs; (2) hub-and-spoke districts characterized by a small number of large, vertically integrated firms surrounded by many small suppliers; (3) "state-anchored districts' which are similar to hub-and-spoke districts, but with the "hub" being a public or nonprofit organization, such as a university, government laboratory, or defense plant, rather than a large firm; and (4) satellite industrial platforms consisting of the branch facilities of multi-plant firms that are headquartered outside the cluster
Maskell (2001) argues that any economic theory of clusters must provide an explanation for the existence and growth of the cluster and identify its boundaries Once a cluster exists, focusing specifically on knowledge-based clusters, he distinguishes between the horizontal and vertical dimensions of clusters He finds that "while suppliers and customers simply need to interact with each other in order to do business, competitors don't Most relationships in the cluster will therefore be along the vertical dimension " (Maskell, 2001, p 930)
Maskell also emphasizes the role of heterogeneity of firms in a cluster. He asks, What are "the advantages of N co-localized firms of size S undertaking related activities that are not transferable to a single firm of size S X N doing the same? This is arguably the single most important question for understanding the existence of the cluster, yet largely ignored in discussions on the
subject . " (p . 927) Maskell argues that clustering reduces the costs of co-ordination and helps in overcoming problems of asymmetrical information, leading to further specialization so that a higher level of knowledge creation is obtained. "The main advantages are not the ease of intra-cluster interaction as such..., but the deepening of the knowledge base that it enables. Only by a steady increase in the number of firms in the cluster would it be possible to create knowledge simultaneously by variation and by the division of labor. " (p . 932) This introduces a time dimension to the analysis of clusters. Growth of clusters occurs by relocation of existing firms, by attracting (e. g . via existing dominant firms) entrepreneurs to start new firms, and by spin-offs from existing firms . If and when new entry no longer occurs, the cluster stops growing
Following up on Maskell's analysis, Bathelt, Malmberg & Maskell (2004) discuss the idea of different types of knowledge flows, distinguishing between "local buzz" and "global pipelines " Buzz refers to the information and communication flows within the same industry and place or region It consists of "specific information and continuous updates of this information, intended and unanticipated learning processes in organized and accidental meetings, the application of the same interpretative schemes and mutual understanding of new knowledge and technologies, as well as shared cultural traditions and habits within a particular technology field, which stimulate the establishment of conventions and other institutional arrangements . Actors continuously contribute to and benefit from the diffusion of information, gossip and news by just "being there" . " (p . 38) "Global pipelines," on the other hand, refers to the linkages between anchor tenants within a cluster and similar entities outside the cluster such as the sharing of designs and technical specifications among aircraft manufacturers and their suppliers of major sub-systems . Ernst & Kim's (2002) concept of "global production networks' is similar
Design Space, Innovation Regimes, And Knowledge Flows
There are several dimensions of knowledge creation and dissemination that we need to understand before proceeding to empirical analysis of clusters The sources, nature, and diffusion of knowledge differ among industry clusters Knowledge flows vary dependent on design space and innovation regime
Design space is defined as a cluster of complementary technical competencies Its boundaries shift constantly due to scientific discovery (serendipitous or
purposive), leading to new combinations. The design space is potentially influenced by academic research (concepts, theories, research methods and tools) as well as by industrial R&D (changes in absorptive capacity). (Stankiewicz, 2002)
It is useful to distinguish between two types of innovation regimes: discovery-driven and design-driven .
Discovery-driven regimes are characteristic of fields with poorly articulated or structured design spaces . The limited extent to which functions are clearly identified and mapped on the known structures and processes means that the solutions to problems have to be discovered rather than designed. Typical for discovery regimes is that innovation is driven by opportunity rather than demand . Technological advances, particularly the radical ones, tend to be triggered by serendipitous discoveries The search processes that follow these discoveries are usually massively parallel (various forms of screening) . Product performance requirements are hard to fully specify and operationalize early in the process . Hence the scope for vicarious testing is limited, and there is often strong dependence on some form of field trials. (Stankiewicz, 2002, pp . 40-41)
By contrast, in well-developed engineering fields, technical problems are typically attacked through "analytical design" — presupposing a well-articulated design space
The search processes taking place in that space are sequential and iterative rather than parallel. The relatively high efficiency of the development processes reflects the fact that the design space utilized is strongly bounded and the performance requirements well defined and easy to operationalize . Design-oriented innovation processes are demand rather than opportunity driven. Mechanical engineering, electrical engineering, and software are examples of technologies operating predominantly under the design regime (Stankiewicz, 2002, p 40)
This means that even though knowledge-based clusters are dependent on new knowledge, the organization of knowledge creation and diffusion varies from one cluster to another and may also change over time
Laursen and Salter (2004) investigate what types of firms use universities as a source of innovation. They find that firms that adopt "open" search strategies (firms that use many external sources of knowledge such as competitors, suppliers and customers, private research institutes, fairs and trade associations) and invest in R&D are more likely than other firms to draw from universities. They also find that only a limited number of firms draw directly from universities as a
source of information or knowledge for their innovative activities The results imply that the direct contribution of universities to industrial practice is likely to be highly concentrated in a small number of industrial sectors (Laursen & Salter, 2004, pp. 1211-12).
Studying bioscience-based clusters, Cooke (2004) notes the rise of specialist research firms, dedicated biotechnology firms or DBFs ("discovery companies") in the life sciences, along with university and other research labs, in proximity to which knowledge-intensive firms tend to cluster
"Hence we see a highly globalized, hierarchical knowledge generation model in which leading-edge research is initiated by multi-disciplinary DBFs in clusters linking with (often many) large pharmaceutical firms, research institutes and other DBFs as developers. It is plain that the clusters are increasingly the locus of knowledge generation... The rise of research over science explains the rise of DBFs over big pharma in new knowledge generation But DBFs still need large drugs firms to fund their discovery programmes " (Cooke, 2004, p. 1115)
Universities and research institutes create basic scientific knowledge that is commercialized in clusters of DBFs, with the support of venture capitalists and other business and legal services . At the same time, multinational pharmaceutical companies fund the research in exchange for future licenses and acquisitions
Powell et al. (1996) discuss "learning through networks' in biotechnology-based clusters. They argue that when knowledge is broadly distributed, the locus of innovation is found in networks of inter-organizational relationships. To be able to benefit, firms must be directly involved in the research process . "Passive recipients of new knowledge are less likely to appreciate its value or to be able to respond rapidly In industries in which know-how is critical, companies must be expert at both in-house research and cooperative research with such external partners as university scientists, research hospitals, and skilled competitors " (Powell et al., 1996, p . 119)
Owen-Smith and Powell (2004) distinguish between channels and conduits. They see channels as diffusely and imperfectly directing transfers between nodes, facilitating information spillovers that benefit both loosely connected and centrally positioned organizations Conduits, on the other hand, are more closed; they are characterized by legal arrangements (e g , nondisclosure agreements and exclusive licensing contracts that transfer intellectual property rights) designed to ensure that only the specific parties to a given connection
benefit from the information that is exchanged They also find that both the geographic location of organizations connected by formal ties and the institutional characteristics of nodes in a network can alter the character of information flows . New knowledge flows out of universities much more readily than it does from commercial organizations (Owen-Smith and Powell, 2004, pp. 5-7).
Knowledge Flows In Three Types Of Clusters
Having thus laid the foundations — distinction between market-mediated and non-market mediated knowledge dissemination and between design-driven and discovery-driven innovation — we now proceed to an analysis of knowledge flows in three types of knowledge-based industry clusters. We examine the sources of new knowledge and the mechanisms of knowledge transfer in the semiconductor-based clusters in Silicon Valley and Cambridge (UK), the biotechnology clusters in Boston and the San Francisco Bay Area, as well as Medicon Valley, and the aerospace clusters formed around Boeing, Bombardier, Airbus, and Saab . As mentioned earlier, these clusters were chosen because they represent a spectrum of high-tech industrial activity located in different geographic regions with different institutions and because there is a relatively rich literature on each . To my knowledge, this is the first systematic analysis of knowledge flows in a cross-section of clusters The review of the literature revealed that the knowledge creation processes in semiconductors and bioscience are essentially discovery-driven, whereas that in aerospace is design-driven. This has implications for the role of universities and entrepreneurial activity in cluster formation and growth and also for the structure and organization of the clusters
Discovery-Driven Innovation: Semiconductors
Silicon Valley
The evolution of Silicon Valley is discovery-driven. The invention that gave rise to Silicon Valley (and similar clusters elsewhere) was the transistor The invention was made at Bell Labs in New Jersey around 1950 by a team led by William Shockley. The fact that the new technology was commercialized in what later became known as Silicon Valley can be attributed to both Shock-ley's (partly incidental) decision to re-locate to the area and start his Shockley Semiconductor Laboratory there (in 1956) and the prior existence of the beginnings of an industrial agglomeration near Stanford University.
There were several electronics companies already in place: Litton Engineering Laboratories (founded in 1932), Hewlett-Packard (1937), Varian Associates (1948), Westinghouse, Philco-Ford, and IBM (1950s), and Lockheed Aerospace Co. research lab (1956). There were also important institutions such as Stanford Industrial Park (founded in the late 1940s) and Stanford Research Institute (1950s)
In analyses of the evolution of Silicon Valley, Stanford is typically featured as a paradigm of universities generating innovations that lead to new technology-based firms . See e . g. , Saxenian (1994) and Bresnahan et al. (2001) . While it is clear that Stanford has indeed played an important role in shaping Silicon Valley, it is also important to note that at the same time there have been many external factors influencing research and other activities at Stanford — such as the federal government (particularly the Department of Defense) and many other actors such as business firms and inventors It is a matter of co-evolution of institutions, academic and business R&D, and new technology As Lenoir et al. (2003) point out,
[t] he key to understanding these dynamic flows between the Valley and Stanford is the role of Federal support of research and development at major universities as well as the stimulus provided by federal R&D... Creating and sustaining an entrepreneurial culture has been crucial to developing this synergistic feedback between federally supported research and research problems of industry, and it has positioned Stanford researchers to make major advances in science and engineering A further crucial element in this synergism is the presence at Stanford of an engineering school, a medical school, and an environment that encourages interdepartmental and cross-school collaborative work . Such collaborations have been fundamental in producing startup companies focusing on convergent technologies (such as computing and biotechnology, or nanotechnol-ogy and communications) that have been crucial to generating new waves of technological innovation. (Lenoir et al., 2003, p . 1)
Many studies of Silicon Valley have emphasized the role of Fred Terman, the Dean of Engineering and subsequently Provost at Stanford, who joined the University in 1946 Part of Terman's vision was to build Stanford's research capabilities through close alliances with industry, similar to what MIT had done before the war. But he was aware of the desire on the part of industrial sponsors of academic research to control the direction of research and to ensure exclusive access to the research results Therefore, he built his research programs with
government grants funding the research of doctoral students who would then become attractive candidates for hiring by industry. Terman was also one of the drivers behind the building of infrastructure Stanford Industrial Park was a part of Terman's strategy of building a strong university center for research and graduate instruction in electronics . (Lenoir et al., p . 5)
Upon his arrival in California, Shockley hired a set of extraordinarily talented engineers for his Semiconductor Laboratory. Within a year these engineers ("The Traitorous Eight") left the company to form their own firm, Fairchild Semiconductor. A decade later, several of these engineers spawned another set of their own individual companies: Intel, Advanced Micro Devices, Inc . (AMD), Kleiner-Perkins venture capital company, and Fairchild Semiconductor Corp . (Moore and Davis, 2001) Stanford did not play a direct role in creating the knowledge that gave rise to the first generations of Silicon Valley firms, but it has certainly done so subsequently, as exemplified by Sun Microsystems (founded 1982), Cisco (1984), Yahoo! (1994), and Google (1998), all founded by Stanford graduate students
Lecuyer (2005) has argued that while the Department of Defense dictated the intellectual contours of academic science and engineering during the Cold War, American science was also deeply influenced in important ways by industry. He has shown that between 1955 and 1985 Stanford University benefited from industrial innovation in solid state technology (transistors, integrated circuits, and VLSI systems) and that these transfers enabled Stanford engineers to make significant contributions to the expanding fields of microelectronics and computing
Along similar lines, Kenney and Patton argue that the primary source of entrepreneurs for Silicon Valley start-ups has been other firms, not university institutions . It is an indirect process: it is still true that many of the "defining firms' (pioneers) in individual sectors originated in universities and corporate laboratories For example, in addition to the Stanford spin-offs already mentioned, 3Com, Seagate, and Cadence are directly linked to Bay Area corporate research institutes and universities (Kenney and Patton, 2006, pp . 39-40) . There were (and are) close links between these corporate research institutes and universities in the area But it is these firms rather than Stanford per se that have spawned most of the new firms . The Silicon Valley pattern seems to have been for a university spin-off to start a new line of business in the semiconductor industry and then in turn spin off new firms, each specialized in a new business . Sometimes the mechanism was the
start-up of a firm to design and market new integrated circuits that would then contract for manufacturing from existing producers who happened to have spare capacity As advances were made in existing design by Stanford or Berkeley faculty, these faculty would form new start-ups . As the software improved, many IC firms abandoned their in-house software and purchased software from design software vendors The standardization of the design software facilitated the rise of the fabless semiconductor firms as they were able to purchase their design tools, eliminating the need for them to create their own software The design software became the interface between the designers and the manufacturers (Kenney and Patton, 2006, p. 48).
After each new discovery in a university or corporate lab, a new company was spun off and then spawned new spin-offs as new applications of the technology were found. An example is in the magnetic storage industry whose origins can be traced to research conducted in IBM's San Jose Laboratories. As new discoveries were made, people left IBM to establish firms to exploit new market opportunities of supplying storage devices for the new entrants. Similarly, in computer networking the pioneer was Xerox's Palo Alto Research Center (PARC) which created a networked system of small computers, laser printers, and data storage devices At the end of the 1980s, computers were proliferating and entrepreneurs began forming firms to design and produce networking equipment (Kenney and Patton, 2006, pp . 50-52).
"A business model emerged in which venture capitalists funded start-ups that were established with acquisition as an exit strategy Cisco pioneered a new corporate strategy of using the Silicon Valley start-up ecosystem to identify the new technologies that would affect its business As firms competed and grew and yet others were formed, Silicon Valley increasingly became the knowledge center for computer networking This deep knowledge meant that Silicon Valley firms, entrepreneurs, and venture capitalists would be uniquely positioned to see the next big thing. " (Kenney and Patton, 2006, p 53)
Another important part of the Silicon Valley model is the openness and flexibility of the labor market It was commonplace for people to change jobs between firms in the Valley, so that over time, participating in a start-up has become a career path (Saxenian, 1994, pp 30-37)
Cambridge, UK
The Cambridge area provides an example similar to that of Silicon Valley of endogenous formation of a
high-tech cluster through spin-off, agglomeration and institutional adaptation, based on a discovery-driven process of innovation.
Endogenous developments in Cambridge encompass the founding of companies by current and former members of the university, clustering stimulated by serial spin-outs from originator firms, the rise of local suppliers and, especially significant, the emergence of specialist labour markets. These developments depended on demand for high-tech output and exerted attraction effects through business services drawn to the area, through the implantation of international subsidiaries, inward investment via acquisition and the attraction of venture capital funds Together these processes, endogenous and exogenous, contributed to the development of local competence and capabilities resulting in the formation and success of many new firms . (Garnsey & Heffernan, 2007, p. 44. )
Another endogenous determinant of clustering involves local supply chain benefits . Similar to Silicon Valley, high-tech firms in the Cambridge area make use of value chain complements or substitutes for the firms' internal activities by outsourcing to local legal and business services These, in turn, have been attracted to the area by the presence of high-tech firms Access to specialized labor is a key factor. It is not only the supply of new university graduates that is an important local asset but also a labor market of experienced specialized professionals that has accumulated over time . Mobility of highly skilled workers, facilitated by social networks, have contributed to technology transfer and fostering of interfirm links (Waters and Lawton Smith, 2008) . In Cambridge, clustering is closely related to an inter-generational spin-out process The firms are connected locally by mobile people and knowledge to a greater extent than by supply relations, and they operate in value chains that have global reach. Their production networks are more international than local
Both in Silicon Valley and in Cambridge the primary mechanism of knowledge transfer in electronics has been inventors leaving a university or corporate laboratory to start a new firm in order to commercialize a new application Stanford has played an important role primarily as institution-builder but also as a source of knowledge, along with industrial R&D. The openness and high degree of labor mobility in the industry, both in Silicon Valley and in Cambridge, has made it easy for new firms to attract skilled labor from existing companies and thus build their absorptive capacity. This process involves both market-mediated technology transfer and pure spillovers. The main vehicle for continued
growth has been proliferation of new products via startups and spin-offs
Garnsey & Heffernan (2007) point out that success in Cambridge has been achieved in spite of significant obstacles . New firms in the area have had to struggle to obtain investment capital, reflecting a short-term focus of UK capital markets and higher rates of return in other, less innovative activity elsewhere in the UK economy. Until the late 1990s, venture capital in the area consisted of only three funds investing in about five ventures each among all Cambridge high-tech companies . Local and central government have also been unsupportive of business expansion in Cambridge . Waters and Law-ton Smith (2002) argue that there is a need for more locally tailored policies rather than a local application of top-down central policy. Inadequate public transport and shortages of housing and skilled technical labor are particularly noteworthy constraints on growth.
Discovery-Driven Innovation: Biotechnology
The biotechnology industry is another discovery-driven industry. Its origin is the discovery by James Watson and Francis Crick of the structure and operation of the DNA molecule in Cambridge, UK, in 1953. Over the next couple of decades, basic research was conducted in university and government laboratories, as well as in a few large oil and chemical companies The first commercial biotechnology firm in the United States was Cetus Corporation, founded in Berkeley, CA, in 1971, by Ron Cape and Peter Farley who brought scientific experience from both academic and business laboratories Cetus was looking for a wide spectrum of applications, ranging from genetically engineered bacteria for alcohol production and oil-spill cleanups to vaccines and therapeutic proteins for the prevention and treatment of human disease . Genentech, founded in 1976, was the first commercial biotechnology firm to focus specifically on the development of pharmaceutical products using biotechnology techniques . It was founded by Bob Swanson, a venture capitalist with Kleiner and Perkins who had been an early investor in Cetus, and Herbert Boyer, a biochemist at the University of California, San Francisco . (Romanelli & Feldman, 2006, pp . 88-89)
Thus, the biotech industry originated in academic research. Similarly to the semiconductor industry, the evolutionary process is clearly discovery-driven, only even more so An important difference between the two sectors is that biotechnology relies more heavily on basic science than does microelectronics, and university research has therefore played a more prominent
role. Another important difference is that the process of converting a new scientific discovery into a new product ready for commercialization takes much longer, is riskier, and requires much more investment and scientific expertise than in microelectronics . As a result, intermediaries between scientific research and commercial application have emerged in the form of dedicated biotechnology firms (DBFs) . In a few cases (e . g. , Genentech and Amgen), the DBFs produce and market the new products themselves, but in most cases the deeper pockets and greater resources and expertise in production, marketing, and distribution of large pharmaceutical firms are needed .
An important feature of discovery-driven processes is that researchers are typically looking for applications of new discoveries in new domains . In the early days, firms experimented in broad categories of human diagnostics and therapeutics, agricultural biotechnology, and industrial and environmental biotechnology. Today firms tend to focus instead on quite specific techniques for the production of bioengineered drugs, plants, and chemicals . (Romanelli and Feldman, 2006, p . 90) Because of the costs and risks involved, the experimentation is carried out by numerous small, specialized firms (DBFs) rather than by large, established firms
The DBFs represent the "hard core" of commercial agents in biotechnology, exclusively selling science-based knowledge as inputs to other industries, especially pharmaceuticals, but increasingly also to such diverse industries as medical diagnostics, food production and agriculture, bio-environmental remediation and chemical processing Incumbents in pharmaceuticals have had to acquire and assimilate biotechnology capabilities and to engage in cooperative relations with DBFs, universities and other research institutions in order to survive . (Christensen, 2003, p . 224)
As the design space in biotechnology has become both denser and more diverse, involving knowledge from a growing variety of disciplines, inventive and innovative activities have increasingly come to require both specialized knowledge from many different sources and competencies to integrate these diverse knowledge inputs . It is beyond the capacity of even large firms to master all the required competencies As a result, DBFs have come to play the role of "experimenters' and "explorers' of scientific and technological opportunities for large pharmaceutical companies . Alliances between DBFs and large pharmaceutical corporations have become a prevalent feature of the modern pharmaceutical industry. Over time, pharmaceutical firms have also increasingly acquired small DBFs. While in the past
pharmaceutical companies have always relied primarily on in-house R&D, complex innovative networks have emerged involving pharmaceutical companies, DBFs, public research institutions, as well as public authorities Such networks have become the predominant mode of organizing innovation and may prove to become an enduring alternative to the historically vertically integrated innovation processes (Christensen, 2003).
In the early phase of the industry, and especially in biotechnology narrowly defined, i. e. , human therapeutics and diagnostics, the transfer of knowledge from universities to new start-ups (DBFs) was tied tightly to "star scientists' and was therefore confined to quite limited geographic areas. See e . g. Zucker and Darby, 1996; Zucker, Darby and Armstrong, 1998; Zucker, Darby and Brewer, 1998 According to these studies, the positive impact of research universities on nearby firms was related to identifiable market exchange between particular university star scientists and firms, not to generalized knowledge spillovers . There was simply insufficient capacity outside the university laboratories to absorb the new technology Much of the knowledge development and transfer still takes place via DBFs that commercialize technology
However, it may be that the findings of Zucker and colleagues about the role of star scientists with impact only locally pertain to the beginning of the industry but not necessarily to later periods Feldman (2003) points out that at the very beginning of the industry, universities were quite aggressive in intellectual property licensing, and that the importance of university research may decline over time
"Science, the pursuit of new knowledge, occurs primarily within the domain of the research university and is characterized by a priority-based reward system that emphasizes scientific publication. Technology, on the other hand, develops ideas from science for commercial markets It is characterized by the pursuit of economic returns and its venue is rent seeking firms While it is appropriate to consider patents, publication and the location of star scientists in the earliest stages of firm formation — the science stage — we may expect that as an industry develops and science is translated into commercial applications, the locational dynamics may change to emphasize industrial and technological attributes . While science resources may be most important in the earliest stages of the industry development, technology resources may become more important as the industry develops . " (Feldman, 2003, p . 321)
Boston/Cambridge and the San Francisco Bay Area Feldman's hypothesis is borne out, at least in part, in studies by Owen-Smith & Powell (2004 and 2007) .
They analyzed strategic alliance networks in human therapeutic and diagnostic biotechnology during the period 1988-1999 in the Boston/Cambridge (Massachusetts) metropolitan area and in the San Francisco Bay Area They found that
[d] uring the very early years of the industry, from the early 1970s to the late 1980s, most biotech firms were very small start-ups that relied, of necessity, on external support . Lacking the skills and resources needed to bring new innovations to market, they became involved in elaborate lattices of relationships with universities and large pharmaceutical firms. Lacking a knowledge base in the new scientific field of molecular biology, large companies were drawn to start-ups by the latter's capabilities in basic and translational science (Owen-Smith & Powell, 2007, p . 62)
Studying bilateral links between entities in the Boston area, they found that at the beginning of the period (1988) by far the dominant part of the linkages were between public research organizations (PROs, such as Harvard, MIT, Tufts, and Massachusetts General Hospital) and DBFs There were only a small number of ties between biotech firms or between biotech firms and local VC firms . These ties grew as the network expanded during the 1990s and dominated the commercial ties at the end of the period Thus, the Boston network grew from origins in the public sector. Public science formed the foundation for commercial application. Early in its evolution, the Boston biotechnology community was linked together by shared connections to academic research These connections have remained an important part of the network, but over time the number of DBF to DBF and DBF to VC ties has increased relative to university linkages (Owen-Smith & Powell, 2007, p . 67).
The trajectory in the Bay Area is quite different from that in Boston . In 1989-1990,
the Bay Area community was composed entirely of ties linking DBFs to local VC firms. Where the stability and technical diversity of Boston PROs anchored that network and fostered a more open technological trajectory., the Bay Area relied heavily on the prospecting and matchmaking efforts of venture investors Later years witnessed the increasing importance of VCs, a smattering of ties involving PROs, and — most importantly — dramatic growth in DBF-DBF connections. Both Boston and the San Francisco Bay Area evolved from dependence on a non-DBF organizational form to a state where significant portions of the network were made coherent by direct connections among science-based biotechnology firms In other words, similar endpoints in the evolution of the networks were reached
through different routes While both relied on the inclusion of organizations different from biotechnology firms, Boston was anchored in the public sector, whereas the Bay Area was dominated by venture capitalists " (pp 67-68)
Boston companies were often started by MIT and Harvard professors, who were typically senior professors with established reputations, who maintained their university affiliations, and who tended to serve the new companies primarily as scientific advisors In contrast, founders in the Bay Area were much more likely to come from VC or other biotech firms . When Bay Area faculty were involved in founding, they tended to be younger and much more likely to take a leave from their university positions Whereas almost all founders in Boston came from within the region, founders in the Bay Area came from a variety of locations, including faculty from Yale, Columbia, and Duke who came to California to start companies, (p . 70)
While Boston/Cambridge and the San Francisco Bay Area followed quite different trajectories during the 1990s, they ended up with rather similar structures by the end of the decade As the density of links between DBFs increased in both clusters, the relative dependence on PROs and VCs, respectively, declined. OwenSmith and Powell contend that networks dominated by PROs and "open science" will result in innovations that rely less heavily on internal R&D and that draw more on research conducted in organizations other than DBFs
There are several implications of these studies. Among these are (1) that the sources of knowledge (especially the role of universities) may vary from one location to another as well as over time, depending on institutional factors (co-evolution); (2) that the geographic boundaries of the cluster may shift over time; (3) that "true" technological spillovers may increase over time as absorptive capacity increases; and (4) that as a result of these complexities, public policy-making in biotechnology is extraordinarily difficult .
Medicon Valley
Medicon Valley refers to the biotechnology cluster located on both sides of the Oresund straight that separates Denmark and Sweden The region has a long tradition in the agricultural, brewing, and pharmaceutical industries The Swedish pharmaceutical firm Astra and the Danish firm Lundbeck started their activities in the region around World War I and were joined later by Novo Nordisk, Leo, Ferrosan, and Ferring (Denmark) . Together with the universities of Copenhagen (founded 1479) and Lund (founded 1660) and several smaller universities, these companies formed an industrial
agglomeration in the region that was already in place when an initiative was taken by Professor Sture Forsen at Lund University and Nils Horjel, the county governor, in 1983, to start the Ideon Science and Technology Park adjacent to Lund University Both Bioinvent and Biora, the first Swedish pure biotech firms, originated in different research projects at Lund University in the 1980s .
The initiation of Ideon sparked a wave of research parks in the region, including Symbion Science Park in Copenhagen. In 1995, five universities in the region began discussing cooperation among the universities in the two countries to strengthen the scientific knowledge base . This resulted in a joint effort by nine regional universities to create what is now called Oresund University. The completion of the bridge between Copenhagen and Malmo in 2000 tied the two sides together physically. (Braunerhjelm and Helgesson, 2006)
Thus, the universities took the lead in creating Medi-con Valley They were soon followed by policymakers and local governmental bodies that started to market the region in order to attract both national and international investment. The number of service providers and VC firms started to increase There were nine VC firms in the region in 1995; the number increased to thirty-three in 2002 (most on the Danish side). By 2002 there were 116 biotech firms in the cluster (82 in Denmark and 34 in Sweden) with a total employment of nearly 3,000 There were also 71 pharmaceutical firms (including large firms such as Astra Zeneca, NovoNordisk, H . Lun-dbeck, and LEO Pharma) and 129 medical technology firms in the region The research infrastructure included 26 hospitals (11 of which were university hospitals) and 12 universities . The research output in the region places it among the leading regions in the world: in per capita terms, the number of biotechnology-related articles and citations in scientific journals in the region ranked slightly above other regions in Europe and not far behind Boston and the San Francisco Bay area in the United States . (Braunerhjelm and Helgesson, 2006; Coenen et al., 2004)
It is clear that universities (especially Lund University) played an important role in forming a biotechnology cluster in the area, drawing on the pre-existing regional agglomeration of pharmaceutical firms and research institutions But where did the knowledge come from?
Coenen et al. (2004) have studied the knowledge flows in the region. They found that the knowledge dynamics of the cluster exhibit a dual local-global knowledge flow pattern. The sector is characterized by strong spatial concentration around nodes of excellence that are interconnected through a global network. Their
study highlights the significance of proximity within epistemic communities (rather than other relational or physical proximities) in shaping innovation processes across multi-spatial scales The study is based on a database-survey on collaboration in scientific publication by 109 biotechnology firms in Medicon Valley. Examining interpersonal knowledge interaction as reflected in scientific publications by all DBFs located in the region, they find that a large share (58 %) of the firms can be found in the Science Citation Index, with a total of 846 publications . About 40 % of the Danish firms and 50 % of the Swedish firms are involved in international co-publication. A vast majority of the firms' joint publications are with different types of PROs, whereas firm-firm co-publication seems to be quite rare; this applies to firms in both countries . The co-authors in international joint publications are scientists in a variety of countries, dominated by Germany, the UK and the US. About 1/3 of the firms have one or more publications with co-authors from outside Europe while only 1/5 of the firms are involved in cross-border Danish-Swedish co-publications (Coenen et al , 2004, p 1013)
Thus it seems as though the collaborations that these firms have are more influenced by epistemic community (common scientific background) than by spatial or relational proximity Many of the biotech firms in the Swedish part of the region are spin-offs from Lund University, but the common educational and professional background seems to play a greater role than the relational proximity between the researchers at the firm and their former colleagues at the university (Coenen et al., 2004, p . 1014) .
In a similar study, McKelvey et al. (2003) have studied knowledge collaboration among Swedish entities in the biotechnology-pharmaceutical sector (not just in Medicon Valley but in the country as a whole) and with entities outside Sweden. They identified 215 R&D collaborations by 67 Swedish firms or Swedish research institutes and 137 foreign partners . Among these 215 R&D collaborations there were 52 agreements between two Swedish actors and between Swedish and foreign partners. Similarly to Coenen et al., the authors concluded that the degree of interconnection among Swedish firms is quite low and that no firm or group of firms plays a central role. Instead, alliances and collaborations with entities in the United States are much more important than with entities in Sweden or elsewhere in Europe They also found that the Swedish parts of the large Swedish pharmaceutical firms Pharmacia, Astra-Zeneca and Amersham Pharmacia Biotech have very different spheres of R&D collaboration, both nationally
and internationally. Thus, while the major MNCs have little formal collaboration within the country, they are also interested in different types of partners (McKelvey et al, 2003, p . 495) . However, geographic co-location does appear to be important for smaller biotech-phar-ma firms located in regions of strong medical research
Thus, it appears that Swedish firms interact more internationally and especially with entities in the U S than domestically or in other European countries. While the existing literature suggests that the reason for this is to access American research and American biotechnology firms, this appears to be valid mostly for large European pharmaceutical firms . This Swedish study shows instead that there is also a reciprocal flow, i. e. , that international partners do deals to access knowledge at small to medium sized Swedish firms and Swedish research organizations (McKelvey et al., 2003, p. 496).
McKelvey et al. conclude:
"There are two large MNCs in the pharmaceutical sector, which have strong Swedish heritages These two actors are not engaged in formal knowledge collaboration with the rest of the national firm population, and they are also reducing their involvement with Swedish universities over time. For the rest of the small and medium sized Swedish biotech-pharma firms, the propensity to collaborate with geographically co-located partners differs depending on whether the collaboration is firm to firm, firm to university, or university to university The overall finding is that geographical co-location is less important for firm to firm deals or for university to university co-authored papers than for firm to university deals. In other words, a large number of Swedish firms tend to collaborate with Swedish universities rather than international universities . " (McKelvey et a.l, 2003, p 499)
It is apparent that the knowledge flows in the three regional biotechnology clusters (Boston/Cambridge, San Francisco Bay Area, and Medicon Valley) have evolved quite differently. In Boston, major research institutions such as Harvard, M. I. T. , and Massachusetts General Hospital played a crucial role both as generators of new knowledge and as launching pads for new start-ups . In the Bay Area, venture capitalists served as major sources of linkages between academic research and its commercialization and as sources of funding In Medicon Valley, the primary sources of knowledge are outside the region; the main conduits are research collaboration with universities, particularly in the United States, and the global pipelines supplied by multinational firms There does not appear to have been much knowledge spillover in a true sense; the vast majority
of knowledge transfers have been intentional and market-mediated However, in recent years the increasing presence of research activities of major pharmaceutical firms in each of the three regions suggests that absorptive capacity is increasing to the point where true knowledge spillovers may become important
Similarly to the semiconductor industry, the main vehicle of growth in biotechnology is start-up of new firms, typically based on academic research, applying new knowledge to new products
Design-Driven Innovation: Aircraft Industry
The aircraft industry has evolved from humble beginnings as erstwhile assemblers of simple mechanical components and parts into perhaps the most knowledge-intensive integrators of complex systems known to mankind . But knowledge creation and dissemination in the aircraft industry clusters follows a different pattern than in other knowledge-based industries . While a significant portion of the knowledge is created and disseminated within local clusters, the main hubs of knowledge creation are the anchor tenants ("global network flagships' in the terminology of Ernst & Kim, 2002), not universities . As the terminology implies, these system integrators are connected to global knowledge networks and depend more on such networks than on local suppliers. Consequently, local knowledge spillovers are of a different nature than in other knowledge-based clusters
Before we discuss knowledge generation and knowledge flows in the industry, a brief history of four aircraft companies is instructive
Boeing (Seattle)
Boeing was founded in 1917 by William E . Boeing who had studied at Vevey (Switzerland) and Yale University but did not graduate . He worked initially in the timber industry He became interested in airplanes and decided he could build a better plane than the existing biplanes. In 1927 Boeing created an airline and in 1933 introduced the first modern airliner (a 10-seater) . The Air Mail Act of 1934 prohibited airlines and aircraft manufacturers from being under the same corporate umbrella, so the company split into three: Boeing Airplane Company, United Airlines, and United Aircraft Corporation (later United Technologies) . Shortly thereafter an agreement was reached with Pan American World Airways to develop and build a commercial airliner able to carry passengers on transoceanic routes The first flight of the Boeing 314 Clipper took place in 1938 It was the largest civilian aircraft of its day with
a capacity of 90 passengers. In the same year Boeing completed work on the Model 307 Stratoliner, the world's first pressurized-cabin transport aircraft During World War II, Boeing built a large number of bombers . The company also designed the B-17 bomber which was also assembled by the Lockheed and Douglas aircraft companies and the B-29 that was assembled also by Bell Aircraft Co and the Glenn L Martin Company After the war Boeing developed military jets such as the B-47 Stratojet and the B-52 Stratofortress as well as the KC-135 tanker aircraft that was adapted as the Boeing 707 civilian jetliner, the first commercial jet airliner in the United States In the 1960s and 1970s the Boeing 727, 737, and 747 were added to the product line, in the1980s the 757 and 767, and in the 1990s the 777 (Wikipedia)
Boeing dominated the large commercial aircraft industry for over 50 years It is still the world's largest producer of both military and civilian aircraft and is also the largest aerospace company in the world Its main assembly plants are located in Seattle, Washington. In 2001 its headquarters moved to Chicago . Boeing is somewhat different from other aircraft manufacturers in that for several decades it manufactured its main structural parts in-house. As a result, it became much more vertically integrated than its competitors In the last few decades the company has dispersed its manufacturing and supplier system throughout the world in order to increase market penetration and reduce design and production costs . (Niosi and Zhegu, 2005)
Bombardier (Montreal)
The production of aircraft in Montreal started in the 1920s, when several American, British, and Canadian producers competed to produce small propeller aircraft In 1944, a group of employees of the Canadian subsidiary of British Vickers founded Canadair. After World War II and during the cold war, Canadair produced mostly military aircraft Dozens of companies were spun off from Canadair or were attracted to Montreal to supply parts and components . In 1976, the company moved into civilian aircraft by acquiring the exclusive rights to the blueprint for a business jet (Learjet 600) designed by Learjet Corporation of Wichita, Kansas (USA) In 1986, Bombardier Corporation of Montreal bought Canadair and entered the regional aircraft market The company developed several new regional jets and also bought de Havilland in Toronto. By the early 2000s Bombardier Aerospace was the world's third largest producer of aircraft, with 15,000 employees in Montreal and 28,000 world-wide . (Niosi and Zhegu, 2005, p 11)
Bombardier Aerospace is the largest but certainly not the only company in the aircraft cluster in Montreal. As early as the 1920s, Pratt & Whitney Canada, a subsidiary of U.S . — based United Technologies, started overhauling and repairing American-designed and built aircraft engines Its production expanded and new products entirely designed and manufactured in Montreal were added. In the mid-1980s, Bell Helicopter of the U. S. transferred its production (but not design) of its civilian helicopters to Montreal Several other companies (including subsidiaries of British and French firms) are also located in Montreal There are now over 250 small and medium-sized manufacturing companies in the Montreal aerospace cluster. (Niosi and Zhegu, pp . 12-13)
Airbus (Toulouse)
The aircraft cluster in Toulouse is centered on Airbus Industrie, a European consortium founded on government initiative in 1969 with Aerospatiale of France and Deutsche Airbus of Germany each taking a leadership role and with British (Hawker Siddeley, later acquired by British Aerospace) and Dutch (Fokker-VFW) companies also participating. Each company would deliver its sections as fully equipped, ready-to-fly components . In 1971 the Spanish company CASA also acquired a small share of Airbus Industrie
Today Airbus is rivaling Boeing as the world's largest producer of commercial aircraft Airbus assembles six different models of aircraft in Toulouse with parts and components coming from 1,500 contractors in 30 different countries The United States is the largest provider with over 800 suppliers . Meanwhile, Toulouse has become a major aerospace cluster, with hundreds of firms. These include a French-Italian manufacturer of turboprops, manufacturers of turbines, landing gear, and small aircraft. Toulouse has also attracted producers of other aerospace-related products such as Matra and Acatel (satellite communications) (Niosi and Zhegu, pp . 17-18)
Saab (Linkoping)
Svenska Aeroplan AB (SAAB) was founded in 1937 in Trollhattan in western Sweden but soon moved its headquarters to Linkoping near the east coast about 100 miles southwest of Stockholm With World War II looming, the Swedish Air Force needed aircraft . When the war broke out in 1939, Saab was producing bombers and fighters, mainly copies of German and American designs . The first aircraft designed in-house was a light bomber that rolled off the line in 1941 In 1943 a fighter bomber aircraft was ready. At the end of the war, Saab converted seven U S B-17 Flying Fortress
bombers into passenger aircraft . It also developed a small passenger plane (Saab 19) of its own, as well as a small plane for private use . As the Cold War intensified, the Swedish government wanted Saab to concentrate on military aircraft . Consequently, the production of the Saab 19 was discontinued in 1954 and transferred to the Dutch company Fokker. The fighter J-29 was introduced in 1948, followed by the J-32 in 1952, the J-35 in 1955, the J-37 in 1967, and the JAS-39, a multi-purpose aircraft which entered service in 1997. Saab has continued to produce all the aircraft needed by the Swedish air force and has also exported these aircraft to other countries (Eliasson, 2010; Swedecar com; Wikipedia org)
Thus, Saab started out as a producer of military aircraft, diversified into civilian aircraft but was forced to revert to a primary focus on being a system integrator and producer of military aircraft and a supplier of advanced subsystems to Boeing and Airbus Organization of the Aircraft Industry The dominant characteristics of the aircraft industry are helpful in explaining why the industry is organized the way it is and why the knowledge flows differ from those in other knowledge-based clusters
"Aerospace is a high value-added sector, strongly affected by scale and timing The industry success depends on rapid technological progress; government support for corporate R&D is essential. Their activity depends on components and parts which can be widely dispersed in terms of both industry and location . Transportation costs of these components are not relevant in overall aircraft costs Also, demand (market) is not geographically bounded... [T] he primary centripetal force has been the regional pool of skilled and semi-skilled labor Less important factors have been the location to the original industries of the cluster (often engineering sectors close to aircraft such as railway manufacturing) and the entrepreneurial talent. The persistent increase of R&D costs has been the major centrifugal force for the aircraft global decentralization: in order to reduce R&D costs, the industry has been gradually implementing strategies of international cooperation. " (Niosi & Zhegu, 2005, p 6)
The large aerospace clusters typically consist of one or several OEMs (original equipment manufacturers) surrounded by hundreds of small and medium-sized suppliers of components and parts There are two types of suppliers: higher-tier lead suppliers that deal directly with several OEMs and lower-tier suppliers that usually deal with the higher-tier suppliers, not directly with the OEMs . The higher-tier suppliers are usually located
outside the local cluster, often overseas . Aerospace regions tend to specialize in different parts of the value chain. They manufacture high-value products in batches from a few hundred to several thousand. For example, there are civilian aircraft assembly clusters (such as in Seattle, Montreal, and Toulouse) and engines clusters (such around GE's engine plants in Cincinnati, Ohio, and Lynn, Massachusetts) . With Boeing as a major assembler, Seattle is specialized in engineering and production of large commercial aircraft . Toulouse (France) is the major production site of Airbus and ATR . (Niosi & Zhegu, 2005)
Knowledge Generation and Knowledge Dissemination in the Aircraft Industry
Airplane manufacturers are essentially system integrators; they provide strategic and organizational leadership in designing complex systems In the increasingly modularized global production system, the technology of the most advanced engineering firms often involves development of concepts, integration, and systems architecture rather than manufacturing (Eliasson, 2010). Manufacturing is instead outsourced to various suppliers in the value chain . The OEMs are powerful carriers of knowledge . They are primarily global pipelines to major sub-system suppliers but they also transfer technical and managerial knowledge to local suppliers so that they can meet the technical specifications. For example, the Boeing 787 Dreamliner is assembled in Seattle using components developed and produced by an international team that includes Rolls-Royce in the UK (engines), General Electric in Ohio (engines), Kawasaki Heavy Industries in Japan (main landing gear), Dassault Systèmes in France (software), and Saab Aerostructures in Sweden (cargo doors) and dozens of other suppliers of components and sub-systems, plus hundreds of local suppliers of parts In the case of Saab, the core technologies for the JAS-39 Gripen aircraft (other than platform development, systems integration, and aircraft control system which are Saab's own responsibility), the engine is manufactured by VolvoAero based on the General Electric F404 engine and the radar, computer, and electronic systems are developed by Ericsson. Other subsystems are developed by a variety of major aerospace contractors in the U S , U K , France, and Germany Only one sub-system is contracted to a Swedish company, but there are many Swedish suppliers of components (Eliasson, 2010) . Clearly, in terms of knowledge flows, the linkages to other advanced firms are much more important than to the local firms in the cluster
Modern aircraft integrate advanced mechanical technology with electronics, sensor technology,
hydraulics, new materials, and communications systems, among others. The system integration involves overall design, safety and reliability, availability and maintainability, monitoring and diagnostics, survivabili-ty, and produceability Military aircraft are designed and developed in collaboration between government (military) agencies and aircraft manufacturers; for civilian aircraft, airlines play the role of competent customers The bulk of R&D expenditures in advanced firms is devoted to identifying internationally available complementary technology to integrate with their existing knowledge base, and only a small fraction is allocated on genuinely new technology development . The multinational firms are specialists in this field . It is noteworthy that distributed and integrated production became the mode of operation in engineering industries only after the micro processor resulted in the integration of computing and communications technology in the 1990s (Eliasson, 2010)
Knowledge Transfer Mechanisms Given the tiered structure of the aircraft industry, it is useful to examine knowledge flows at two levels The knowledge flows between the system integrator and tier 1 (sub-system) contractors are bilateral; there is a great deal of learning, but the system integrators must necessarily take a leadership role The knowledge flows are based on contracts Historically, such contracts have typically been of a cost-plus nature . 3 These knowledge flows are large and have great economic impact as the participants apply advanced technology in their own businesses, with ripple effects to their sub-contractors . But it is important to note that they are market-mediated; they are not spillovers .
At tier 2 and lower levels, knowledge flows in the aircraft industry are usually more unilateral in nature and take place through more formal contracts Flagship companies transfer knowledge in the form of blueprints and technical specifications, mostly free of charge, to ensure that products and services produced by the suppliers meet the necessary specifications Sometimes these knowledge transfers are bilateral, i e , systems and sub-systems evolve through collaboration between the integrator and the suppliers Knowledge may also be transferred informally, without a contract and without any payment involved, particularly through technical assistance to local suppliers The flagship company may exercise significant control over the way in which knowledge is disseminated and used, or it may play a more
3 The Saab JAS-39 Gripen project is an exception; it is based on fixed-price contracts.
passive role with little influence on how local suppliers take advantage of the knowledge . Even though these transfers may not involve direct payment from the supplier to the OEM, the benefits are appropriated primarily by the OEM in the form of purchased products that meet the specification Only to a limited extent should they be regarded as knowledge spillovers But to the extent that local suppliers can develop their absorptive capacity, they can effectively absorb knowledge disseminated by global network flagships This requires both individual and organizational learning. (Ernst & Kim, 2002) These knowledge flows are not market-mediated, but they are directed to specific users, not generally to all the firms in the cluster
"Flagships typically provide the local suppliers with encoded knowledge, such as machinery that embodies new knowledge, blueprints, production and quality control manuals, product and service specifications, and training handouts. This is done to assist the suppliers in building capabilities that are necessary to produce products and services with the expected quality and price " (Ernst & Kim, p. 1425)
In contrast to electronics and biotechnology, aerospace clusters, even though they are knowledge-based, are not based on local knowledge spillovers. They rely mostly on global pipelines . For example, Saab's technological prowess as a developer and producer of military aircraft has depended in large measure on access to U .S . technology, notably advanced electronics . In return for building a strong air force capable of preventing Soviet anti-submarine aircraft from crossing Swedish airspace, the U S made advanced military technology available to Sweden, even though Sweden is not a member of NATO . (Eliasson, 2010) The clustering of economic activity in this sector is due primarily to agglomeration effects (externalities) in the form of pools of skilled labor and local suppliers of parts, components, and services. Knowledge spillovers from universities do not play a very important role. Niosi and Zhegu argue that local knowledge spillovers are less significant, of a different nature, and make less contribution to explaining the geographical agglomeration of firms in the aircraft industry than in other knowledge-based clusters On the other hand, international transfers of technology help to explain the dispersion of industry across nations The fact that the industry is geographically clustered is due to the anchor tenant effects as creators of labor pools and owners of very large manufacturing plants creating regional inertia.
Even though most of the technology transfers in the aircraft industry are market-mediated (i e , not true
spillovers), they still have enormous economic impact . According to calculations made by Eliasson, the economic effects of aerospace R&D anchored by Saab in Sweden are very large, at least 2-4 times the original investment in R&D This includes not only the core technologies integrated in military aircraft but also related technologies in the engineering industries more generally For example, Eliasson argues that were it not for its collaboration with Saab on military aircraft, Ericsson — currently the world's largest supplier of telecommunications equipment — would not have survived as an independent company Other Swedish companies have also been able to develop more advanced products as a result of collaborating with Saab The diffusion of technology rarely occurs in the form of transfers of well-defined and patentable technology packages; there is much learning on the part of both user and supplier The main diffusion channel is people with knowledge and experience who move on through internal careers in firms or over the labor market (Eliasson, 2010)
Conclusions And Policy Implications
In this paper I have tried to draw together several strands of literature, both theoretical and empirical, in order to analyze knowledge flows in various types of knowledge-based industry clusters. Where does the knowledge come from, and what mechanisms are used to disseminate knowledge? In particular, to what extent is it appropriate to use the term "spillover" to refer to the diffusion of knowledge?
The sources of knowledge and the vehicles of dissemination of knowledge differ among high-tech clusters . In clusters characterized by discovery-driven innovation, such as biotechnology and semiconductors, universities play a much more important role as creators of knowledge than in design-driven clusters . In biotechnology, the new knowledge tends to be basic science that needs to be developed and "translated" before it can be commercialized This is typically accomplished via dedicated biotechnology firms, the new products being manufactured and marketed via existing firms . The transfer from DBFs to large firms such as pharmaceutical companies is typically market-mediated (via license, acquisition, or joint venture), while the transfer from university to DBF may be either market-mediated (via license or joint ownership) or spillover. Universities are anchors in the early phase of discovery-driven innovation Their role remains important as the technology matures, but other linkages increase in number and importance as the cluster grows Large incumbent firms locate subsidiaries (listening posts) in the cluster in order to pick up new ideas .
In electronics there is typically no intermediate stage similar to DBFs; new knowledge results in new start-ups that often spawn new spin-offs These typically involve spillovers . Firms may eventually grow large, and some become dominant creators and distributors of new technology (usually through market-mediated processes), but the vitality of the cluster depends on new applications of technology typically innovated by new firms spun off from existing firms or from universities
By contrast, in design-driven processes, large incumbent firms are the main creators of new technology They do so by combining and integrating components and subsystems co-designed and co-developed with major suppliers . In addition to co-ordination of research done elsewhere, this requires vast amounts of in-house research Most of the technology sharing and transfer is market-mediated. By challenging local suppliers to meet high technical standards the system integrators also elevate the absorptive capacity and thus contribute to technology spillovers in the local cluster. Universities have not been important in the early phase of design-driven clusters but have become more important as suppliers of researchers, engineers, and other skilled labor (although not new technology), as technology has become more complex. Design-driven clusters grow primarily by expanding linkages with existing companies both globally and locally rather than through the formation of new entities
There are several policy implications of this analysis It is necessary to distinguish between sectors characterized by design-driven innovation and those characterized by discovery-driven innovation . In the former, new knowledge creation tends to take place in large firms rather than universities These firms tend to be connected to other large firms (suppliers of sub-systems and components), often via international networks through which knowledge is both created and shared via market-mediated processes The role of universities is to supply skilled labor Public policy can promote the building of a strong knowledge base by supporting higher education and by instituting policies and mechanisms for public procurement of advanced technology. Successful implementation of public procurement may require prior investment in competence and absorptive capacity The primary functions of public policy are to identify the domain, thereby providing legitimacy and reduced uncertainty and risk in order to promote resource mobilization and experimentation, helping to establish a market, and creating positive externalities in related industries such as venture capital and services . It can also directly provide funding to promote knowledge creation (Bergek et al , 2008)
In sectors characterized by discovery-driven innovation, universities play a much more important role as creators of new knowledge as well as suppliers of skilled labor Serendipity is key; pure knowledge spillovers are important As a result, targeted public procurement is unlikely to be successful Instead, the role of public policy is to support R&D and to promote entre-preneurship, particularly via spin-offs from universities . Promoting connectivity, both globally and locally, is also important, both for knowledge flows and for capital flows (especially via well-functioning venture capital and exit markets)
References
1. Abramovitz, Moses, 1956. "Resource and Output Trends in the United States Since 1870," American Economic Review, 46 (2): 5-23.
2. Acs, Z J. , Audretsch, D . B . , Braunerhjelm, P. , and Carlsson, B . , 2004. "The Missing Link: The Knowledge Filter and Entrepreneurship in Economic Growth, CEPR working paper no 4358
3. Acs, Z. J. , Audretsch, D. B. , Braunerhjelm, P. , and Carlsson, B , 2009 "The Knowledge Spillover Theory of Entrepreneurship," Small Business Economics, 32 (1): 15-30.
4. Acs, Z J. , Audretsch, D . B . , Braunerhjelm, P. , and Carlsson, B , 2005 "Growth and Entrepreneur-ship — An Empirical Assessment, working paper, April
5. Anselin, Luc, Varga, Attila, and Acs, Zoltan, 1997. "Local Geographic Spillovers between University Research and High Technology Innovations," Journal of Urban Economics, 42: 422-448
6. Arikan, Andag T. , 2009. "Interfirm Knowledge Exchanges and the Knowledge Creation Capability of Clusters," Academy of Management Review, 34 (4): 658-676.
7. Bathelt, H. A. , 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): 31-56.
8. Belussi, F. and Caldari, K. , 2009. "At the Origin of the Industrial District: Alfred Marshall and the Cambridge School. " Cambridge Journal of Economics 33: 335-355.
9. Bergek, A. , Jacobsson, S . , Carlsson, B. , Lindmark, S. , and Rickne, A . , 2008. "Analyzing the Functional Dynamics of Technological Innovation Systems: A Scheme of Analysis . " Research Policy 37 (4): 407-429.
10. Braunerhjelm, Pontus, and Feldman, Maryann (eds . ), 2006. Cluster Genesis: Technology-Based Industrial Development, Oxford: Oxford University Press
11. Braunerhjelm, Pontus and Helgesson, Christian, 2006. "The Emergence of a European Biotechnology Cluster: The Case of Medicon Valley," in Braunerhjelm and Feldman (eds. ), op. cit., 133-147.
12. Bresnahan, T. , Gambardella, A . , and Saxenian, A. , 2001. "Old Economy Inputs for New Economy Outcomes: Cluster Formation in the New Silicon Valleys," Industrial and Corporate Change, 10 (4): 835-860.
13. Breschi, Stefano and Lissoni, Francesco, 2001. "Knowledge Spillovers and Local Innovation Systems: A Critical Survey," Industrial and Corporate Change, 10 (4): 975-1005.
14. Cantner, Uwe and Pyka, Andreas, 2001. "Classifying Technology Policy from an Evolutionary Perspective," Research Policy, 30 (5), 759-775
15. Carlsson, Bo and Stankiewicz, Rikard, 1991 . "On the Nature, Function, and Composition of Technological Systems," Journal of Evolutionary Economics, 1 (2), 93-118.
16. Carlsson, Bo (ed . ), 1995, Technological Systems and Economic Performance: The Case of Factory Automation, Boston, Kluwer Academic Publishers
17. Carlsson, Bo (ed . ), 1997. Technological Systems and Industrial Dynamics, Boston, Kluwer Academic Publishers
18. Carlsson, Bo (ed. ), 2002. Technological Systems in the Bio Industries: An International Study, Boston, Kluwer Academic Publishers .
19. Carlsson, Bo, 2007. "Innovation Systems: A Survey of the Literature from a Schumpeterian Perspective," in Horst Hanusch and Andreas Pyka (eds . ), Elgar Companion to Neo-Schumpeterian Economics, (Cheltenham: Elgar), 2007: 857-871.
20. Carlsson, Bo and Fridh, Ann-Charlotte, 2002 "Technology Transfer in United States Universities: A Survey and Statistical Analysis . " Journal of Evolutionary Economics, 12 (1-2): 199-232.
21. Carlsson, B. , Acs, Z. , Audretsch, D. B. , and Braunerhjelm, P , 2009 "Knowledge Creation, Entrepreneurship, and Economic Growth: A Historical Review," Industrial and Corporate Change, 2009, http://icc. oxfordjournals. org/cgi/content/full/dtp043? ijkey=Ayz62ZKvXWykosm&keytype=ref .
22. Christensen, Jens Fr0slev, 2003. "Introduction: The Industrial Dynamics of biotechnology: New Insights and New Agendas," Industry and Innovation, 10 (3): 223-230.
23. Coenen, Lars, Moodysson, Jerker, and Asheim, Bj0rn, 2004. "Nodes, Networks and Proximities: On the Knowledge Dynamics of the Medicon Valley Biotech Cluster," European Planning Studies, 12 (7): 1003-1016.
24. Cooke, Philip, 2003. "Networks and Hierarchies in Bioscientific Knowledge Management," paper presented at the DRUID summer conference, 2003.
25. Cooke, Philip, 2004. "Life Sciences Clusters and Regional Science Policy," Urban Studies, 41 (5/6): 1113-1131.
26. Dosi, Giovanni, 1988. "Sources, Procedures and Microeconomic Effects of Innovation," Journal of Economic Literature, 26 (3): 1120-1171.
27. Eliasson, Gunnar, 2010. Advanced Public Procurement as Industrial Policy—Aircraft Industry as a Technical University, New York: Springer.
28. Eliasson, Gunnar and Eliasson, Ása, 2006. "The Pharmacia story of entrepreneurship and as a creative technical university — an experiment in innovation, organizational break up and industrial renaissance," Entrepreneurship & Regional Development, 2006, 18: 393-420.
29. Ernst, Dieter and Kim, Linsu, 2002. "Global Production Networks, Knowledge Diffusion, and Local Capability Formation," Research Policy, 31: 1417-1429.
30. Feldman, Maryann P. , 1999. "The New Economics of Innovation, Spillovers and Agglomeration: A Review of Empirical Studies," Economics of Innovation & New Technology, 8 (1): 5-25 .
31. Feldman, Maryann P. , 2003. "The Locational Dynamics of the US Biotechnology Industry: Knowledge Externalities and the Anchor Hypothesis," Industry and Innovation, 10 (3): 311-328
32. Frenken, Koen, 2007. Applied Evolutionary Economics and Economic Geography (Cheltenham, UK: Elgar)
33. Garnsey, Elizabeth, and Heffernan, Paul, 2007. "The Cambridge High-Tech Cluster: An Evolutionary Perspective", ch . 2 in Frenken, K. (ed . ), 27-47.
34. Gordon, I. R . , and McCann, P. , 2000. "Industrial Clusters: Complexes, Agglomeration and/or Social Networks?" Urban Studies, 37 (3): 513-532.
35. Griliches, Zvi, 1992. "The Search for R&D Spillovers," Scand. J. of Economics, 94, Supplement, 29-47
36. Jaffe, Adam, 1989. "Real Effects of Academic Research," American Economic Review, 79: 957-970.
37. Jaffe, Adam B . , Trajtenberg, Manuel, and Henderson, Rebecca, 1993. "Geographic Localization of Knowledge Spillovers as Evidenced by Patent Citations," The Quarterly Journal of Economics, 108 (3): 577-598.
38. Kenney, Martin, and Patton, Donald, 2006. "The Coevolution of Technologies and Institutions: Silicon Valley as the Iconic High-Technology Cluster," in Braunerhjelm and Feldman (eds . ), op . cit . , 38-60
39. Krugman, Paul, 1991. "Increasing Returns and Economic Geography," Journal of Political Economy, 99: 483-499.
40. Laursen, Keld, and Salter, Ammon, 2004. "Searching High and Low: What Types of Firms Use Universities as a Source of Innovation?" Research Policy 3: 1201-1215.
41. Lecuyer, Christophe, 2005. "What Do Universities Really Owe Industry? The Case of Solid State Electronics at Stanford," Minerva, 43: 51-71.
42. Lenoir, T. , Rosenberg, N. , Rowen, H. , Lecuyer, C . , Colyvas, J. , and Godlfarb, G. , 2003. "Inventing the Entrepreneurial University: Stanford and the Co-Evolution of Silicon Valley," http://www. siepr . stanford. edu/programs/SST_seminars/ Lenoir pdf
43. Lucas, Robert, 1988. "On the Mechanics of Economic Development," Journal of Monetary Economics, 22: 3-39
44. McKelvey, Maureen, Alm, Hakan, and Riccaboni, Massimo, 2003 "Does Co-Location Matter for Formal Knowledge Collaboration in the Swedish Biotechnology-Pharmaceutical Sector?" Research Policy, 32:483-501
45. Markusen, Ann, 1996. "Sticky Places in Slippery Space: A Typology of Industrial Districts," Economic Geography, 71 (3): 293-313.
46. Marshall, Alfred, 1920. Principles of Economics (8th edition), vol. 1. London: Macmillan and Co , Ltd
47. Marshall, Alfred, 1923. Industry and Trade. London: Macmillan and Co Ltd
48. Maskell, Peter, 2001. "Towards a Knowledge-Based Theory of the Geographical Cluster," Industrial and Corporate Change, 10 (4): 921-943.
49. Metcalfe, J. S . , 1994. "Evolutionary Economics and Technology Policy," The Economic Journal, 104 (425), 931-944.
50. Metcalfe, J S , 1995 "The Economic Foundations of Technology Policy: Equilibrium and Evolutionary Perspectives," in P. Stoneman (ed. ), Handbook of the Economics of Innovation and Technological Change. Oxford: Blackwell.
51. Moore, G. and Davis, K . , 2001. "Learning the Silicon Valley Way. " Stanford Institute for Economic Policy Research discussion paper 00-45.
52. Niosi, Jorge, and Zhegu, Majlinda, 2005. "Aerospace Clusters: Local or Global Knowledge Spillovers?" Industry and Innovation, 12 (1):1-25
53. Owen-Smith, Jason, and Powell, Walter W. , 2004. "Knowledge Networks as Channels and Conduits: The Effects of Spillovers in the Boston Biotechnology Community," Organization Science, 15 (1):5-21
54. Owen-Smith, J. and Powell, W. , 2007. "Accounting for Emergence and Novelty in Boston and Bay Area Biotechnology," in Braunerhjelm and Feldman, op. cit., 61-83.
55. Porter, Michael E , 1990 The Competitive Advantage of Nations. New York: Free Press .
56. Porter, Michael E , 1998 "Clusters and the New Economics of Competition," Harvard Business Review, November-December, 77-90.
57. Powell, Walter W. , Koput, Kenneth W. , and Smith-Doerr, Laurel, 1996. "Interorganizational Collaboration and the Locus of Innovation: Networks of Learning in Biotechnology," Administrative Science Quarterly, 41 (1): 116-145.
58. Romanelli, Elaine and Feldman, Maryann, 2006. "Anatomy of Cluster Development: Emergence and Convergence in the US Human Biotherapeu-tics, 1976-2003," in Braunerhjelm & Feldman, op. cit., 87-110 .
59. Romer, Paul, 1986 "Increasing Returns and Economic Growth," American Economic Review, 94, 1002-1037
60. Romer, Paul, 1990. "Endogenous Technical Change," Journal of Political Economy, 98: 71-102.
61. Saxenian, AnneLee, 1994. Regional Advantage. Culture and Competition in Silicon Valley and Route 128. Harvard University Press, Cambridge, MA
62. Smits, R. , Kuhlmann, S, and Shapira, P. , 2010 . The Theory and Practice of Innovation Policy Cheltenham: Elgar.
63. Solow, Robert M , 1956 "A Contribution to the Theory of Economic Growth," The Quarterly Journal of Economics, 70 (1): 65-94
64. Stankiewicz, Rikard, 2002. "The Cognitive Dynamics of Biotechnology and the Evolution of Its Technological System" in B . Carlsson (ed. ), op.
cit., 35-52. 68.
65. Waters, Rupert, and Lawton Smith, Helen, 2002. "Regional Development Agencies and Local Economic Development: Scale and Competitiveness in High-technology Oxfordshire and Cambridgeshire," European Planning Studies, 10 (5), 633649. 69.
66. Waters, Rupert, and Lawton Smith, Helen, 2008. "Social Networks in High-Technology Local Economies: The Cases of Oxfordshire and Cambridgeshire," European Urban and Regional Studies, 15 70. (1), 21-37.
67. Werker, Claudia and Athreye, Suma, 2004. "Marshall's Disciples: Knowledge and Innovation
Driving Regional Economic Development and Growth," Journal of Evolutionary Economics, 14 (5), 505-523
Zucker, Lynne G. , and Darby, Michael R. , 1996. "Star Scientists and Institutional Transformation: Patterns of Invention and Innovation in the Formation of the U. S . biotechnology Industry," Proceedings of the National Academic of Science, 93: 706-717.
Zucker, Lynne G. , Darby, Michael R. , and Armstrong, Jeff, 1998 "Geographically Localized Knowledge: Spillovers or Markets?" Economic Inquiry, 36 (1): 65-86.
Zucker, Lynne G and Brewer, M , 1998 "Intellectual Human Capital and the Birth of US Biotechnology Enterprises," American Economic Review, 88 (1): 290-306.
