3.1.1. An environment which fosters innovation

The first element highlighted by research in this subject is that clusters do not form part of a centralized industrial policy. They form and develop via the densification of a local industrial fabric, based on entrepreneurial initiatives, groupings of companies or industrial networks. The pivots of these forms of organization, as discussed in Chapter 1, are either large groups in close collaboration with a number of smaller companies or cells produced through the close interweaving of a network of neighboring small businesses. In some cases, they may be the product of a local environment (agricultural or industrial) and/or of public support, linked to the presence of a university, research centers or government agencies.

In this context, entrepreneurial actions, policies and strategies adopted by companies play a key role in the creation and development of clusters, more than public/private partnerships. American ecosystems represent an efficient mode of organization for industrial activity (in the broadest sense of the term) but the creation of an industrial cluster is a means to an end and not an end in its own right [FEL 14a]. Spotting an opportunity before others is, for the author, at the very heart of the notion of an entrepreneurial advantage. This creates a change in perspective which is particularly significant in American works on the subject: instead of considering pre-existing clusters and the benefits which companies may derive from their localization, writers have taken the way in which company initiatives transform their host territories as a starting point. This transformation has an effect on different institutions and organizations (see below) and on the composition of industrial networks. American clusters are often constructed on the basis of entrepreneurial behaviors.

Entrepreneurial initiatives can also be observed in the cluster extension phase. Companies joining the cluster benefit from both economic and non-financial benefits in the form of knowledge transmitted by the firms already present in the cluster. This “intangible” benefit is hard to evaluate but its existence is undeniable. In this context, Feldman notes that companies may choose to behave in one of two ways: opportunistically, aiming to draw maximum benefit from their localization as quickly as possible, or by investing in the creation of resources and the reinforcement of institutions in order to derive long-term advantages from their geographic location. American clusters “develop following an endogenous logic (local spillovers, spinoffs and the accumulation of human and social capital), triggering processes of self-reinforcement and self-development, but equally involve exogenous aspects, obtaining increased visibility via the development of partnerships and key contracts” [ALC 08]. The primary purpose of these ecosystems, according to the report’s authors, is to contribute to local economic development, the creation of value and of jobs.

The actors in these clusters profit from the action of public and private structures which, via their support and initiatives, contribute to the self-reinforcement process mentioned above. Building on the industrial fabric, they play a role as “growth catalysts, supporters and accelerators for high-tech companies, in concertation, and in interface management. They enable structuring and formalization of the network” [ALC 08, p. 39]. The most widely cited organizations include government agencies, economic development agencies at local or regional levels, innovation-support organizations and councils. These organizations are heavily involved in cluster development, providing a certain number of more or less specialized and essential services. Those which deal with innovation problems (cf the presentation of CONNECT at the end of this chapter) contribute to the creation and growth of high-tech companies and to the commercialization of new products and recent technologies. Note also that American clusters are not subject to any form of certification as their governance is not covered by any recognized entity at state or federal level.

3.1.2. Solid foundations

The Alcimed study for the DGE highlights certain key aspects of American clusters.

First, innovation is seen to justify the existence and exploitation of clusters. Innovation is not limited to the creation of new elements but also covers their commercial application. The network effect means that the various components involved must be articulated in an efficient manner and the established connections must allow knowledge exchanges. This knowledge often originates in universities and research centers and the image of the researcher-entrepreneur is a prestigious one. The creation of new companies as a result of this contributes to the development of clusters. The pattern is further strengthened by spinoff dynamics; from this perspective, many American clusters have evolved along similar lines. In creating new businesses, former employees of major established groups extend existing product chains relating to a specific technology or create new offshoots.

Why then is this process so intense and what is it that former employees have learned which allows them to create new spinoffs? In American culture, the creation of new products or services is at the very heart of entrepreneurial behavior. However, while innovation involves novelty, this novelty does not, in and of itself, constitute innovation. Novelty is important in relation to existing products and services and consequently in relation to certain usages. However, the means of improving existing objects and practices are increasingly dependent on the possibility of exploiting accumulating knowledge and of combining fragmented or dispersed knowledge. Learning refers principally to tacit knowledge which is hard to obtain without prior experience. Firms created as spinoffs draw on established companies which already possess a large part of the knowledge required to organize their production activities. This does not imply that firms entering a cluster through diversification within a similar technological domain or as spinoffs are ipso facto competent. Skills obtained from established companies are a necessary but not sufficient condition [KLE 11]. Studies carried out by this author simply indicate that the most skilled companies produce more successful spinoffs. The success rate (longer lifespan, larger size at the time of creation, etc.) is higher than that found in spinoffs outside of clusters or for startups in the same region. By densifying the fabric of businesses, the creation of spinoffs within the clusters studied by Klepper enabled these clusters to progressively appropriate an increasing portion of activity within an industry.

Clearly, this process does not necessarily apply to all activities. In sectors which already present high levels of geographical concentration (radio, TV, etc.) in large cities, a shift toward geographical de-concentration has been observed following the departure of most producers from these cities (due to reorganization problems, real estate prices, etc.). In this context, spinoffs have mostly attenuated the de-concentration movement. In contrast, they have amplified the movement toward concentration in the automobile and semiconductor industries. Finally, note that established companies only rarely sponsor spinoff operations, fearing competition from incoming firms. This justifies the action of public authorities and, more generally, of organizations which provide specialized services to facilitate the creation of new companies and the passage of growth thresholds.

The second axis at work in the polarization of American clusters relates to human capital. The skills required are of a technical or technological nature and also cover notions of implication (referred to as political citizenship), voluntary participation and managerial capacities. The cited study summarizes the essential factors as follows [ALC 08, p. 49]:

• “– the identification of reservoirs of skills relating to technological developments and to future needs;
• – the provision of education and the creation of awareness of technological change from an early age;
• – broad-based actions for requalifying the labor force, initiated at federal or state level;
• – a strong willingness to both attract the best personnel, but also to retain them within the territory in question”.

In addition to these vertical actions, the creation of social networks centering on scientific, technical and managerial knowledge should also be taken into account. In regions structured by clusters, social networks are the main source of new knowledge for companies. The diffusion of knowledge results from a high level of mobility in the labor force. For example, experienced managers may abandon prestigious careers in order to work in lucrative but high-risk startups; this requires the existence of a solid and efficiently organized web of social connections. Once a social network has been established, it reduces the risks taken by qualified personnel in joining these companies or in leaving them in case of difficulties. Social networks which create connections between qualified personnel should be considered as institutions in their own right and this institutional infrastructure forms the basis for the development of certain technological clusters in the USA.

The third pathway for polarization relates to investor funding. This notion covers five aspects: providing continuity in funding for innovation, attracting venture capital funding cells and “business angels” to facilitate face-to-face contact between funders and investors, and promoting the implication of companies, notably in the earliest stages of technological development. It also relates to actions taken by public authorities to ensure the continuity of funding and to compensate for the insufficiencies of private funding sources, alongside the fact that clusters must be sufficiently attractive to attract new investors on a regular basis. Reference is also made to the possibility of breaks in the funding chain. According to the “valley of death” representation (see Chapter 1), when private investors do not maintain strong connections with public laboratories they may lack expertise, both in scientific and commercial terms, when deciding to fund projects at an intermediate stage of development; their expertise and R&D investment efforts also tend to be concentrated in the later stages of the innovation process.

The analysis of American clusters thus provides us with the following information:

Public policies relate, essentially, to the support of the development of clusters. A cluster cannot be created from nothing; it must be allowed to germinate from an entrepreneurial initiative. However, when firms choose to put down roots within an ecosystem, they do so in anticipation of economic – and not financial – gains which they aim to internalize while increasing these benefits for companies already established in the area. This final aspect is not taken into consideration in localization decisions. The social benefits thus outweigh private yields. This is the main benefit of clusters, one which “may justify proactive policies” [KLE 11, p. 141]. In certain clusters, this is manifested in the involvement of public authorities in connection with key themes. Government agencies intervene in the domain of life sciences and healthcare via the NIH, for example, or in the field of advanced manufacturing procedures through the NIST. The public/private articulation is thus essential in understanding the operation of the most innovative clusters. “Dynamic clusters demonstrate high levels of interaction between businesses, public authorities, local organizations and investors in collaboratively establishing different support policies and in ensuring rapid adaptation” [ALC 08, p. 67].

The second finding is that companies are not motivated by short-term profit maximization and that the relational aspect has a strong influence on ecosystem dynamics (see Chapter 1). In a similar vein, we see that company strategies have significant consequences on the performance dynamics of certain geographic locations. This does not mean that clusters can be classified by performance. The cluster effect is hard to measure, partly due to issues of endogeneity relating to the creation of resources and company growth (a classic chicken-and-egg problem) and partly due to the difficulty of establishing proven connections between location and performance (do companies which form part of a cluster not, by definition, perform better than others?).

Our third finding relates to the analysis proposed by Porter, in which a large part of company strategy boils down to actions in favor of competition. The “strategic desire” of companies is to activate localized competition in order to foster sustained innovation efforts [FEL 14a]. As the author suggests, this strategy might be defined as “…the pattern for decisions that … defines the kind of economic and human organization it is or intends to be and the nature of the economic and non-economic contributions it intends to make to its shareholders, employees, customers and communities” [AND 99, p. 52]. The spectrum of behaviors to take into account is thus extended, notably covering communities, in the form of innovation and production ecosystems or clusters.

Finally, the analysis of American clusters highlights both the essential role of entrepreneurial behaviors and the viability of these forms of organization, characterized by renewed creation of productive resources and products, efficient institutions, a climate of openness and risk tolerance, the acceptance of diversity, and confidence in the generation of mutual benefits for both public and private actors.

3.2.1. The network effect

This effect can be analyzed on the basis of an observation. The three high-performance clusters all display robust social connections between scientists, funders and managers. The idea that cluster growth is related to the density and quality of social connections has been put forward by a number of authors, notably Saxenian [SAX 94]. For this author, the success of Silicon Valley is due to the development of a decentered social structure, fostering the creation of informal connections between scientists, engineers and managers in the region. Furthermore, high levels of mobility in the labor force contribute to the creation of high-density social networks, creating links between employees in different firms in the area. The study carried out in 2009 [CAS 09] analyzed social networks, mapping the career trajectories of senior managers and highlighting the intensity of these relationships in the San Francisco and San Diego structures; it established a connection between looser social connections among managers and scientists and the limited growth of the Los Angeles cluster¹. A partial consideration of social connections (excluding relationships formed through professional associations or informal networking) provides a solid explanation for the growth of clusters in California and for their different trajectories. In contrast, biotechnology clusters in Europe, notably in Cambridge (UK) and at certain sites in Germany, are “smaller and less dense” in terms of the extent and density of connections, with more limited mobility of qualified labor over the course of individual careers. In this context, the effects of geography are less important than the implantation of qualified personnel within a community of social relationships (see Chapter 1). Individuals possess an intellectual capital made up of scientific, technological and managerial knowledge, and make use of professional networks to increase their mobility between companies at a single location. The extent and density of social networks for the San Francisco and San Diego clusters are shown in Table 3.3.

Both clusters can be seen to have employed a significant number of senior managers: at the end of the study period, 2096 senior managers (1,229 + 867) were in activity across the two locations. The density of the network is measured in terms of the size of the largest group of individuals connected by affiliation networks. The level of connectivity observed was 80% in the 1990s and in excess of 90% between 2000 and 2005 (over 1,000 in San Francisco and more than 800 in San Diego). Based on a calculation of the mean pathway length between these individuals, Casper considered that, in 2005, a senior manager in San Francisco would be able to contact any of the 1,120 other individuals in the network thanks to the density of these connections.

Two mechanisms may be seen to explain the development of a “self-organizing” social structure and the growth of companies within a cluster. First, the social connections developed between scientists, engineers and managers enable the diffusion of knowledge between companies. “In particular, embeddedness within a decentralized social structure may provide a competitive advantage for technology-intensive firms in market segments in which technological volatility is high” [CAS 09, p. 5]. In the case of the biotechnology industry, informal connections between companies may provide technological intelligence and increased understanding of markets, assisting informed decision-making with regard to technological choices. Along the same lines, companies are able to react to market changes more rapidly than their competitors.

Second, social structures of this type may assist companies in recruiting highly qualified personnel. The success of startups in this area is partially dependent on their capacity to persuade experienced managers and qualified employees to quit existing, well-paid and often secure positions in established companies or universities, in order to join the new organization. The decision to work for a startup in a cluster, within which social networks exist and promote mobility, becomes a rational option; in a way, the mobility of the labor force creates guarantees of employment.

One form of social organization which has been particularly well analyzed is that of the networks of social founders who form the “backbone” of biotechnology clusters. Authors have highlighted the role of “serial founders” in the ecosystems in question, as shown in Table 3.4.

From this table, we see that 46% (123/269) of founders in San Francisco are serial founders, compared with 25% in San Diego. The total number of founders is highest in San Francisco, with some creating multiple businesses; 69, that is, 26%, created at least 2 companies, with 11 individuals creating 5 or more firms. The significant percentage of founders with prior experience in the industry in question may thus be seen to have a strong impact on cluster growth. During a first wave of movement, accumulated experience in the bio-pharmaceutical sector played an important role in the history of the San Diego cluster, when a number of employees left Hybritech after its sale to Eli Lilly in 1986. This seminal case led to the definition of a “general model” of behavior, diffused throughout the industry as a whole.

A second wave of movement has resulted in revised views of the “organizational legacy” [FEL 06]². Founders with prior experience in the pharmaceutical industry contributed to the spinoff process, leaving their companies to found biotechnology firms. More generally, within clusters, anchor companies can be seen to play a crucial role in promoting the creation of new companies. For example, 25 experienced managers (16% of the total) left Genentech to create 22 different biotechnology firms and four senior managers left Amgen to create three other companies.

3.2.2. High growth rates in clusters with heterogeneous populations

The three clusters in question involve individuals and organizations with a wide range of skills and experiences. In the first chapter, we stressed the importance of social capital in the cluster institutionalization process. When the process reaches a stable state, the cluster becomes sustainable and the location benefits from increased vitality. This means that both the intensity of interactions between actors and the legitimacy of the site increase, leading to higher levels of intensity in innovation. Social capital is a form of knowledge which takes a relational form, rather than being a personal attribute. It is created through human interactions and is shared by several individuals, forming a basis for action. Its properties make it comparable to a public good, insofar as it cannot be appropriated by a private entity.

However, the results of social capital are hard to evaluate; they cannot be strictly quantitative, as “capital suggests the existence of an asset while the qualifier social suggests that benefits accrue to being connected to a network or community” [FEL 14, p. 15]. The social interactions which this asset enables create confidence, limit moral risk in contractual activities, and reduce transaction costs.

Compared with individuals, incidences of social capital mean that those who form part of social networks benefit from higher yield on their ideas and investments. Their mobility between companies is based on, and reinforces, a network of belonging. This same movement fosters knowledge transfer. More generally, social capital forms the basis for the formation of communities which are essential for the appropriation of emerging technologies. These communities strengthen the learning process and contribute to assessments of the potential of technologies, creating forms of consensus. Technology development pathways evolve progressively, shared languages emerge and meanings are created, strengthening interactions and information exchanges which favor the commercialization of new products.

Non-commercial actors (universities, hospitals, research centers, etc.) play an active role alongside their commercial counterparts in this process of legitimizing localizations and recognizing innovative products.

3.2.3. Clusters and the development of market effects

The existence of a dynamic labor market, notably for scientists and managers with prior industrial experience, conditions both the creation and growth of new companies. Grass-roots investments earlier in the process, intended to increase the skill level of the labor force, do not imply that priority will be given to local actors. Labor markets, corresponding to different levels of qualification, are activated by significant levels of arrivals and departures (Casper also notes that the labor market for experienced managers in Germany is more sluggish). Companies implement policies which are designed to attract, and where possible to retain, human resources from further afield. These resources then circulate within ecosystems. In the San Francisco cluster, for example, 58% of movements observed between 1976 and 2005 by individuals leaving a biotechnology company were toward a different field of activity. These movements may benefit biotechnology companies in another region, or companies within the same cluster with a different focus. Of the reported movements, 42% were lateral, toward other biotechnology companies within the cluster.

In contrast, the incoming and outgoing movement of labor in the Los Angeles cluster is much more limited. Taking the example of Amgen, a major player in the biotechnology field, we see that the company has developed a “house culture”, which fosters long-term employment and favors internal promotion policies when vacancies become available. This self-sustaining policy does not contribute to the development of social networks within the cluster.

There are active labor markets for different skill profiles, but the consequences are specific to the three US clusters in the study. Notably, high-profile scientists working in the San Diego cluster in 2005 had backgrounds in biotechnology companies (65%) or the pharmaceutical industry (18%). Academic qualifications do not allow scientists to reach the upper echelons of management. The market effect, as a guiding principle, therefore has a strong influence on company direction.

In the study of biotechnology clusters, Casper noted that 40% of scientists working in their clusters had prior industrial experience, compared with a bare 10% in Germany. Casper and Murray concluded that “German biotechnology firms appear unable to systematically recruit senior scientists from the several large pharmaceutical companies active in the country. While we have not addressed the performance of companies within clusters, our analysis strongly implies that the lack of industry expertise within Munich firms should result in weak performance” [CAS 05, p. 70].