Technological Forecasting & Social Change

Transcription

1 Technological Forecasting & Social Change 78 (2011) Contents lists available at ScienceDirect Technological Forecasting & Social Change The typology of technology clusters and its evolution Evidence from the hi-tech industries Jiang He, M. Hosein Fallah Howe School of Technology Management, Stevens Institute of Technology, Hoboken, NJ 07030, USA article info abstract Article history: Received 22 November 2009 Received in revised form 29 November 2010 Accepted 16 January 2011 Available online 21 February 2011 Keywords: Technology clusters Agglomeration Patent analysis Regional planning Clustering is one of the key drivers for regional economic growth. Development of clusters is a dynamic process shaped by a variety of internal and external factors such as availability of skilled labor, presence of functioning networks and partnerships, technological changes, and market competition, etc. As a result, the patterns of cluster growth may differ from one another. Although each cluster is unique in some way, previous research has attempted to identify few simplified models of evolution of clusters. In this study, we briefly reviewed the literature on a variety of models of clusters. Based on these models, we investigated 15 hi-performing metropolitan-based clusters in the United States, covering communications equipment manufacturing, information technology, and biopharmaceutical industries, in order to find out the similarities and differences between real-world clusters. Specifically, by examining the composition of these high-tech clusters, we attempted to find out the following: 1) What are the typologies of these technology clusters? 2) Whether different industries tend to support different cluster typologies? and 3) How do clusters change their typologies over time? Our analysis results suggest that the real-world clusters rarely feature any single type of typology; a mixed type of typology is much more prevalent in reality. We also found that different industries tend to support different types of cluster typologies. In other words, an individual cluster's typology is to some extent shaped by the industry group it belongs to. In addition, we note that, as a cluster goes through different stages of its lifecycle, its typology may change significantly Elsevier Inc. All rights reserved. 1. Introduction Clusters are geographic concentrations of interconnected companies, specialized suppliers, service providers, firms in related industries, and associated institutions (for example, universities, standard agencies, and trade associations) in particular fields that compete but also cooperate [1]. We noted that, from literature research, there are many different ways of defining a cluster. In the earliest study in this field, conducted by Alfred Marshall [2], the notion of industrial districts was developed and it was referring to agglomerations of firms operating in one industry sector in a well-defined and relatively small geographic area. Building on Marshall's notion, other regional innovative models, such innovative milieu and technology districts have been offered as frameworks for discussing the pattern of agglomerations. What distinguishes clusters from these other models has not been clearly defined [3]. Indeed, these concepts clusters, industrial districts, innovative milieu, and technology districts are often used almost interchangeably despite having origins in different conceptual contexts [4]. Although we've recognized that there are studies attempting to distinguish the industrial clusters from technology clusters [5], the differences identified by certain authors were not consistently acknowledged by other studies/researchers. In this study, we follow Porter's [1] definition of clusters, mainly because of the influence of Porter's research in this field. Corresponding author. Tel.: addresses: jhe@stevens.edu (J. He), hfallah@stevens.edu (M.H. Fallah) /$ see front matter 2011 Elsevier Inc. All rights reserved. doi: /j.techfore

2 946 J. He, M.H. Fallah / Technological Forecasting & Social Change 78 (2011) Despite the potential confusion caused by the inconsistence in terminology, the contribution of clustering to regional economic development has been well accepted and documented [6,7]. Companies cluster together to benefit from the availability and quality of local labor pools, well-developed intermediate input suppliers, and better information flows which can facilitate the generation of new ideas. Among those factors, knowledge spillovers between economic actors are believed to be a particular important one promoting the development of modern clusters [8 10]. Because of the significant benefits brought by clustering, both academic researchers and regional planners seek to understand the factors shaping the evolution of clusters. In many of the real-world cases, clusters have emerged unplanned and then gone through a dynamic lifecycle process, which typically consists of four stages: embryonic, established, mature and declining. It is noteworthy that, when a cluster goes through its own evolution path, the forces that foster the subsequent growth of the cluster are not necessarily those that provide the cluster its initial advantage [11]. From a perspective of regional planning, it seems to be rather challenging, if not impossible, to develop a cluster from scratch by planning. But policy interventions, when appropriate, can facilitate the development of an existing cluster [12]. In order to come up with appropriate cluster policy interventions, understanding the driving forces of individual clusters would be fundamental, as clusters are all different from one another. Despite the differences between clusters, there are some important features shared by certain groups of real-world clusters. Previous research has attempted to model the evolution of real-world clusters with those important cluster characteristics. In the following, we will review the literature on cluster characteristics and a variety of cluster typologies. Following the literature review, we will examine 15 U.S.-based hi-tech geographical clusters, covering information technology, communications equipment manufacturing, and biopharmaceutical sectors. By examining the typologies of the select high-performing clusters, we aim to answer the following 3 questions: 1) What are the typologies of these industrial clusters? 2) Whether different industries tend to support different cluster typologies? and 3) How do clusters change their typologies over time? The reminder of this paper is organized as follows. Section 2 reviews the literature on industrial clusters and cluster characteristics. Section 3 describes our methodology for analyzing the select clusters. In Section 4, we present and interpret the data results. In Section 5, we conclude with a discussion about the implications of this study for cluster researchers and practitioners. 2. Cluster characteristics and the variety of cluster typologies The studies of clusters date back to Marshall [2], who developed the agglomeration concept, based upon the cost-saving scale effects brought by industrial localization. According to Marshall, companies in a particular industry tend to cluster together for taking advantage of external economies brought by co-locating, including reduced transaction costs, increased specialization, and efficient information flows. Although Marshall's original model successfully explained some earliest industrial clusters, observations from recent clusters suggest that the phenomenon of clustering appears to be much more complex than the original Marshallian agglomeration model has suggested. For example, case studies on many recent clusters suggested that social interactions as well as knowledge flows between cluster members play an important role in the development of many modern clusters [8 10,13]. Social relations, however, were not considered an important element in Marshall's original cluster model. Following Marshall's work, Becattini [14], Brusco[15], and Sforzi [16] identified a newer type of cluster by examining the economic success of several Italian cities and regions. The modified Marshallian model not only acknowledges Marshall's original agglomeration concept, it also recognizes the roles of social relations as well as inter-firm cooperation/competition within industrial clusters. While Marshall stressed the direct benefits of co-locating (e.g., input cost reduction, availability of labor, etc.); the modified Marshallian model (sometimes referred to as Italian version of Marshallian model in the literature) emphasized the role of social interactions between co-located firms/individuals when explaining the success of clusters. The modified Marshallian model is important as it provides a good foundation to explain many successful modern clusters including Silicon Valley and Orange County [9,10]. Even with its Italian version, the Marshallian formulation wasn't able to explain the flourishing of some clusters and demise of others. In order to address the increasing complexity and variety of real-world clusters, Markusen [17], through inductive inquiries, broadened the picture by introducing additional models of clusters. Besides the Marshallian formulation, three additional models of clusters were proposed by Markusen hub-and-spoke, satellite platform, and state-centered. Fig. 1 illustrates the typology for each of the cluster models. The links depicted in this chart reflect regular inter-firm business activities such as supplier customer transactions. In the categorisation shown in Fig. 1, Markusen[17] stressed the distinct roles played by different groups of cluster members, as well as the way different players interact with each other. We noted there are other categorisations proposed by different researchers; based on our understanding, other models are not as comprehensive as the Markusen's model when it comes to capturing the structures of real-world clusters. For example, Lorenzoni [18] proposed a categorisation based on asymmetries among the cluster members and introduced the concept of leader firms and surrounding firms. According to Markusen [17], a hub-and-spoke cluster, is structured around one or few dominant firms supporting the regional cluster. In this type of cluster, suppliers and other entities spread around and are connected to the hubs like wheel spokes. In a simple version as depicted in Fig. 1, agianthubfirm trades with local small firms as well as external suppliers and sells mainly to external customers, which could be large or small ones. In reality, single-hub-based clusters include the Seattle-based aircraft manufacturing cluster supported by Boeing, and the Toyota City-based car manufacturing cluster supported by Toyota. A multi-hub-based cluster can be found in Detroit, where the Big Three auto manufacturers dominate the regional economy and are surrounded by a number of small and large suppliers. In such a cluster, substantial trades take place between hub firms and local smaller suppliers. However, cooperative activities among competitors within this type of cluster are remarkably lacking. Cooperation may occur between hubfirms and surrounding suppliers, but the terms of cooperation are often set by the hub-firms. Due to its very nature, the dynamics of a hub-and-spoke cluster are largely determined by the hub firm(s), either via its success or failure.

3 J. He, M.H. Fallah / Technological Forecasting & Social Change 78 (2011) Fig. 1. Categorization of technology clusters. Another type of cluster typology proposed by Marsuken is the satellite platform. This type of cluster typically consists of a congregation of branch facilities of externally based multi-plant firms. The Research Triangle Park located in North Carolina represents a high-end example of satellite platform cluster, where a number of high-tech multinational firms have established their R&D centers there. In such a cluster, the remotely located parent companies make important decisions for the local branches thus influence the regional cluster. A key characteristic for this type of cluster is the absence of any sort of connections or networks within the regional cluster, as the local branches maintain regular business relations mainly with their remote headquarters or partners. The future growth for such a cluster is largely determined by the region's ability of keeping its competitive advantage. For high-end satellite platforms, the most challenging issue is to maintain and develop the pool of skilled professionals; while low-end platforms are more vulnerable to increases in the cost of running business. Given its nature, a satellite platform is less sustainable over time if the cluster fails to develop a more diversified local economy. The last type of cluster typology proposed by Marsuken [17] features a state-centered structure. The local business structure in this case is dominated by the presence of one or few large public or non-profit entities, such as universities, public research institutions, or military bases. The key public entities are typically surrounded by smaller firms/organizations, thus forming a hubbased structure similar to the previously described hub-and-spoke typology. In fact, many fast-growing cities in the U.S. and overseas, to a large extent, owe their successes to the presence of government-supported entities. For example, cities such as Madison, Wisconsin and Ann Arbor, Michigan benefit greatly from the presence of the state universities in region. Over the long run, the growth outlook for such a cluster is very much dependent on two factors: 1) the size and power of the dominating public entities; and 2) the region's capability of supporting start-ups and entrepreneurship. Actually, the second would be an important factor facilitating the region to transform itself into a more dynamic and diversified regional economy. It is noteworthy that the categorization proposed by Markusen [17] provides only a much simplified framework to examine cluster structures. It helps cluster researchers/practitioners understand how a cluster comes about and what it thrives on, but it may not always reflect the realities of cluster development. As acknowledged by Markusen [17], real-world clusters often feature characteristics of a mix of multiple typologies. In addition, clusters evolve over time; therefore one may change its typology from one to another as it moves through different stages of its lifecycle. In the following sections, we examine 15 real-world strong clusters to find out how well the above-mentioned cluster models works in reality and discuss the implications of cluster typologies for regional planning. 3. Methodology Based on the objective of this study and the research questions proposed in Section 1, the first step of our investigation is to identify clusters. Using the data from Cluster Mapping Project, which is led by Professor Michael Porter at Harvard University, we selected 15 strong U.S. metropolitan city-based clusters covering information technology, communications equipment manufacturing, and

4 948 J. He, M.H. Fallah / Technological Forecasting & Social Change 78 (2011) biopharmaceutical sectors. We decided to focus on the hi-tech industries because these newer industries are more dynamic than traditional ones and have generated some of the country's strongest clusters such as Silicon Valley. The Cluster Mapping Project assembles a detailed account of the location and performance of industries in the United States, with a special focus on the linkages or externalities across industries that give rise to clusters. Economies are analyzed at various geographic levels, including states, economic areas, metropolitan areas, and counties. The economic performance of a technology cluster in a particular region for a particular year is measured by multiple indicators including number of employment, number of job creation, and level of wages. For each of the three industries being investigated, we identified the top-5 strongest clusters based on number of cluster employment in year Geographically, the top-performing clusters were identified at the level of metropolitan area, a geographic measure used by U.S. Census as well. After identifying the 15 high-performing clusters, we then observed the composition of each cluster and its evolution with respect to typology over time. As suggested by Markusen [17], a crucial factor determining the typology of a cluster is the asymmetries between cluster members in their roles of supporting the regional cluster. Therefore, in the analysis, we are particularly interested in finding the importance of leader-firms relative to that of surrounding smaller firms. To measure the relative importance of different firms in supporting a regional cluster, we examined the individual firms' patent contribution relative to the total patent output of the particular cluster. We acknowledge that number of patents cannot be an ideal measurement for company performance, because only a portion of business activities are patented. Nevertheless, number of patents is a reasonable indicator for company innovation activities [19], which are fundamental for the hi-tech enterprises' success. Individual companies' historical patent output in each year can be obtained from the USPTO (United States Patent and Trademark Office) database [20]. For the leader-firms, in this study defined as the top-10 patent-generating firms/organizations in a regional cluster, we checked their background individually in order to find out their nature (i.e., locally headquartered company, externally headquartered company, or state-supported institution). Then, for each cluster, we calculated the relative patent contribution by each group of cluster members (i.e., locally headquartered large firms, externally headquartered large firms, state-supported large organizations, and local small/medium firms). The relative patent contribution by different categories of cluster members enabled us to observe whether a cluster is close to any of the typical typologies depicted in Fig. 1. Patents within USPTO database are originally classified by main classes, rather than by broadly defined industries. In order to identify patents in each of the three industries, we followed Jaffe's [21] approach of categorizing patents. In Table 1, we list the main patent classes belonging to each industry (i.e., information technology, communications equipment manufacturing, and biopharmaceutical). 4. Data analysis Using the methodology described in Section 3, we identified 15 top-performing US-based clusters, covering information technology, communications equipment manufacturing, and biopharmaceutical sectors. The select clusters for each industry are listed in Table 2, and the ranking of cluster performance is based on number of cluster employment measured in year In Fig. 3, we use a clustered-stacked column chart to demonstrate the composition of each cluster, and different types of firms (i.e., locally headquartered large firms, externally headquartered large firms, state-supported large organizations, and local small/medium firms) are represented by different colors. From this chart, we observed a few interesting characteristics for these strong clusters. First, it appears that local small/medium firms dominated most of the clusters. This observation is consistent with the literature which indicates that an Italian version of Marshallian cluster is more sustainable over the long run when compared with other types of typologies. Second, we observed hub-and-spoke clusters in few cases, and the most significant one is the Seattle-based IT cluster ( IT- Cluster-2 ). Given Microsoft's presence in Seattle, WA and its influence, our finding is not surprising at all. A similar cluster composition was observed for Communications-Cluster-4, representing the communications equipment manufacturing cluster located in the metropolitan area of New York North New Jersey Long Island. While in the latter case, the hub firms' power was less significant than the one's in Seattle. The situation observed for Communications-Cluster-4 is understandable as well, because Bell Laboratory (later changed to Lucent Technologies and Alcatel Lucent), located in North New Jersey, has long been an important hub for innovations in telecommunications. Table 1 Categorization of patents by industry. Industry Classifications by Jaffe Main patent classes by USPTO Information technology Cat.22 Computer hardware and software 341, 380, 382, 395, 700, 701, 702, 704, 705, 706, 707, 708, 709, 710, 712, 713, 714 Cat.23 Computer peripherals 345, 347 Cat.24 Information storage; 360, 365, 369, 711 Cat.41 Electrical devices 174, 200, 327, 329, 330, 331, 332, 334, 335, 336, 337, 338, 392, 439 Cat.46 Semiconductor devices 257, 326, 438, 505 Communications equipment Cat.21 Communications 178, 333, 340, 342, 343, 358, 367, 370, 375, 379, 385, 455 Biopharmaceuticals Cat.31 Drugs 424, 514 Cat.33 Biotechnology 435, 800 Cat.39 Miscellaneous drug and med 351, 433, 623

5 J. He, M.H. Fallah / Technological Forecasting & Social Change 78 (2011) Table 2 Top-performing US-based technology clusters in Rank Information technology Communications equipment manufacturing Biopharmaceuticals 1 San Jose Sunnyvale Santa Clara, CA (57,149) Los Angeles Long Beach Santa Ana, CA (28,576) New York Northern New Jersey Long Island, NY-NJ-PA (68,309) 2 Seattle Tacoma Bellevue, WA (41,503) San Jose Sunnyvale Santa Clara, CA (17,150) Chicago Naperville Joliet, IL-IN-WI (25,263) 3 San Francisco Oakland Fremont, CA Chicago Naperville Joliet, IL-IN-WI (11,968) Los Angeles Long Beach Santa Ana, CA (17,181) (38,105) 4 Boston Cambridge Quincy, MA-NH (35,619) New York Northern New Jersey Long Island, NY-NJ-PA (11,298) Philadelphia Camden Wilmington, PA-NJ-DE-MD (12,108) 5 Los Angeles Long Beach Santa Ana, CA (28,945) Dallas Fort Worth Arlington, TX (9,522) San Francisco Oakland Fremont, CA (11,153) Third, when making comparisons between different industries, we noted that externally headquartered large firms played an important role in the communications equipment manufacturing clusters; while this type of firms were less important for the other two industries. In addition, we observed that state-supported institutions constituted an important force supporting the biopharmaceutical clusters, a situation not observed for the other two industries. This finding may reflect the fact that many inventions in biopharmaceutical industry rely heavily on R&D resources made available by public organizations. Fig. 2 suggests that the type of industry may be a factor influencing cluster typologies, which is an important finding for our second research question. In order to confirm that different industries tend to support different cluster typologies, we performed the discriminant function analysis. The discriminant analysis helps us analyze differences between groups by distinguishing those variables on which groups differ [22]. This is an appropriate technique for our analysis because here we want to confirm clusters in different industries differ with respect to their typology. Specifically, by using the discriminant analysis, we attempted to quantify the degree to which the continuous variables (i.e., relative patent contribution by locally headquartered large firms, relative patent contribution by externally headquartered large firms, and relative patent contribution by state-supported large organizations) can be used to discriminate between the groups (i.e., three types of industries). To perform the discriminant analysis, we measured three types of ratios (Ratio-1 represents the relative patent contribution by locally headquartered large firms, Ratio-2 represents the relative patent contribution by externally headquartered large firms, and Ratio-3 represents the relative patent contribution by state-supported large organizations) for each of the select clusters, and we used GroupID as the dependent variable to represent different industries (i.e., 1 represents IT, 2 represents communications equipment manufacturing, and 3 represents biopharmaceutical). Fig. 2. Breakdown of the top-performing US-based technology clusters.

6 950 J. He, M.H. Fallah / Technological Forecasting & Social Change 78 (2011) Table 3 Group statistics (discriminant analysis). GroupID Mean Std. Deviation 1 IT Ratio Ratio Ratio Communications equipment manufacturing Ratio Ratio Ratio Biopharmaceuticals Ratio Ratio Ratio Table 4 Tests of equality of group means (discriminant analysis). Wilks' Lambda F Sig. Ratio Ratio Ratio The discriminant analysis results are reported in Tables 3, 4, and 5. To find out whether the three types of industries differ with respect to cluster typology, we first examined the results of Group Statistics and Tests of Equality of Group Means. The Group Statistics in Table 3 suggest there are noticeable differences between means of different groups for Ratio-2 and Ratio-3. The Tests of Equality of Group Means (Table 4) further confirm this situation by rejecting the null hypothesis that all group means are equal. The Wilks' Lambda value reported in Table 4 is a multivariate measure of group differences over several variables, it is concerned with the ratio between the within group variance and the overall variance. A ratio close to 1 is an indication of equality of group means, whereas a lower ratio indicates a larger difference between the various group means. We observed that the Wilks' Lambdas for Ratio-2 and Ratio-3 are relatively low and the results are statistically significant. Table 5 reports the eigenvalues and Wilks' Lambdas for the discriminant functions. As we have three groups for analysis, two discriminant functions were produced in this case. The eigenvalues and canonical correlations reported in Table 5 are used to measure the discriminating power of the discriminant functions. A high eigenvalue along with a high canonical correlation indicate a more discriminating function. Based on this, we concluded that the discriminating power of the first discriminant function is high (eigenvalue=1.383, canonical correlation=0.762). In addition, the Wilks' Lambda reported in Table 5 indicates the significance of the first discriminant function. Accordingly, we concluded that the three different industries tend to support different cluster typologies. As discussed earlier, clusters evolve over time and may change their typology from one to another. Next we examine the evolution of the select clusters by observing how compositions of the clusters changed from one time-window to another. In Fig. 3, we show the composition for each of the 15 clusters. For each cluster, its composition was examined on 3 different time-windows 1995, 2000, and Based on this chart, we observed that few clusters indeed went through significant changes in their typology over time. The Communications-Cluster-1 provides us with a good example to observe such changes. The Communications- Cluster-1 in this study represents the communications equipment manufacturing cluster based in Los Angeles Long Beach Santa Ana metropolitan area. In early 1990s, that LA-based cluster was more close to the satellite platform typology, as much of its performance was attributed to the presence of a few large externally headquartered defense contractors including Hughes Electronic Devices, TRW, and Boeing. Over the past decades, the cluster has successfully converted itself into a mixed typology. For the recent period, year 2005, we observed that small/medium firms and locally headquartered firms became increasingly important in supporting the regional cluster. Table 5 Eigenvalues and Wilks's Lambdas for discriminant functions. Function Eigenvalue % of variance Cumulative % Canonical correlation (a) (a) Test of function(s) Wilks' Lambda Chi-square df Sig. 1 through

7 J. He, M.H. Fallah / Technological Forecasting & Social Change 78 (2011) Fig. 3. Compositions of the select clusters in different periods. 5. Conclusions and policy implications Our analysis results suggest that the real-world clusters rarely feature any single type of typology; a mixed type of typology is much more prevalent in reality. Second, we found that different industries tend to support different types of cluster typologies. In other words, an individual cluster's typology is to some extent shaped by the industry group it belongs to. In this regard, cluster practitioners/planners would be better off examining a cluster's evolution by comparing it with peer clusters in the same industry. Third, as a cluster goes through different stages of its lifecycle, its typology may change as well. Our study indicates that a cluster may characterize one of the prototypical models proposed by Markusen [17], for example a satellite platform type, at its initial stage of development. Given the fact that most of the high-performing clusters feature a mixed typology, we argue that, over the long run, the success of a cluster is largely determined by the growth potential of its small-and-medium size firms. Therefore, for cluster policy makers, one of the top priorities is to promote local entrepreneurship and enable smaller firms, we refer to these as toddlers, to benefit from the available resources within the cluster. In this regard, the significant success of Silicon Valley and the recent trouble for Detroit confirmed our observation. This study comes with certain limitations which are noteworthy. The study attempted to capture the typology for each of the select clusters. We acknowledge that the typology of a cluster should be measured with multiple indicators, due to the complexity of real clusters. However, in the study we focused on the asymmetries among the cluster members, and only examined the relative patent contribution by different cluster groups. Other factors shaping a cluster's typology, for example how the cluster members network with each other, were not taken into consideration in this study. Therefore, we suggest that future research should seek to understand the roles of social interactions in the transformation of cluster typology. References [1] M.E. Porter, Clusters and the new economics of competition, Harv. Bus. Rev. (Nov./Dec. 1998). [2] A. Marshall, Principles of Economics, 8th Ed.Macmillan, London, [3] E.M. Bergman, E.J. Jeser, Industrial and Regional Clusters: Concepts and Comparative Applications, The Web Book of Regional Science, Regional Research Institute, West Virginia University, [4] D. Newlands, Competition and cooperation in industrial clusters: the implications for public policy, Eur. Plan. Stud. 11 (2003) [5] C.H. St. John, R.W. Pouder, Technology clusters versus industry clusters: resources, networks, and regional advantages, Growth Change 37 (2) (2006) [6] M.E. Porter, D. van Opstal, US Competitiveness 2001: Strengths, Vulnerability and Long-Term Priorities, Council on Competitiveness, Washington, [7] T. Bresnahan, A. Gambardella, A. Saxenian, Old economy inputs for new economy outputs: cluster formation in the New Silicon Valleys, Ind. Corp. Change 10 (4) (2001) [8] A.J. Scott, A.S. Paul, Collective order and economic coordination in industrial agglomerations: the technopoles of Southern California, Environ. Plann. C Gov. Policy 8 (2) (1990) [9] A. Scott, High tech industry and territorial development: the rise of the Orange County Complex, , Urban Geogr. 7 (1986) [10] A. Saxenian, Regional Advantage: Culture and Competition in Silicon Valley and Route 128, Harvard University Press, Cambridge, MA, [11] M. Enright, in: U. Staber, N. Schaefer, B. Sharma (Eds.), Regional Clusters and Economic Development: A Research Agenda, Business Networks: Prospects for Regional Development, Walter de Gruyter, Berlin, 1996, pp [12] DTI, A Practical Guide to Cluster Development, England's Regional Development Agencies, UK, 2000.

8 952 J. He, M.H. Fallah / Technological Forecasting & Social Change 78 (2011) [13] J.K. Larsen, E.M. Rogers, Silicon Valley Fever, George Allen and Unwin, London, [14] G. Becattini, The Marshallian industrial districts as a social-economic notion, in: F. Pyke, G. Becattini, W. Sengenberger (Eds.), Industrial Districts and Inter- Firm Cooperation in Italy, International Institute for Labour Studies, Geneva, 1990, pp [15] S. Brusco, in: F. Pyke, G. Becattini, W. Sengenberger (Eds.), The Idea of the Industrial District: Its Genesis, Industrial Districts and Inter-Firm Cooperation in Italy, International Institute for Labour Studies, Geneva, 1990, pp [16] F. Sforzi, in: E. Goodman, J. Bamford (Eds.), The Geography of Industrial District in Italy, Small Firms and Industrial Districts in Italy, Routledge, London, 1989, pp [17] A. Markusen, Sticky places in slippery space: a typology of industrial districts, Econ. Geogr. 72 (3) (1996) [18] G. Lorenzoni, L'architettura di sviluppo delle imprese minori, Il Mulino, Bologna, [19] J. Singh, Collaborative Networks as Determinants of Knowledge Diffusion Patterns,, Working Paper, INSEAD. [20] USPTO, Patents BIB: Selected Bibliographic Information from U.S. Patent Grant Publications 1969 to Present and U.S. Patent Application Publications 2001 to Present, U.S. Department of Commerce, Washington, DC, [21] A.B. Jaffe, M. Trajtenberg, M.S. Fogarty, Knowledge spillovers and patent citations: evidences from a survey of inventors, Am. Econ. Rev. 90 (2000) [22] B.G. Tabachnick, L.S. Fidell, Using Multivariate Statistics, 3rd EditionHarper Collins College Publishers, Northridge, USA, Dr. Hosein Fallah is an Associate Professor of Technology Management at Stevens Institute of Technology. He teaches graduate courses in Telecom Management and Technology Management programs. His research interest is in the area of innovation management with a focus on the communications industry. Prior to joining Stevens, Dr. Fallah was Director of Network Planning and Systems Engineering at Bell Laboratories. He has over 30 years of experience in the areas of systems engineering, product/service realization, software engineering, project management, and R&D effectiveness. Dr. Fallah has a B.S. in Engineering and a MS and a Ph.D. in applied science from the University of Delaware. Jiang He has received his Ph.D. in Technology Management from Stevens Institute of Technology. His research interest is in the area of technology clusters and social networks with a focus on the telecommunications industry. He had worked in the telecommunications industries for over 3 years. Dr. He also holds a MS degree in Telecommunications from State University of New York Institute of Technology at Utica.