1.1.1. Structure and mechanism of an ecosystem

Conceiving a purely local integration, however, raises some objections. An ecosystem does not develop in isolation: it is part of an industry, and, in this respect, it cannot be analyzed as an independent unity unrelated to the national and international industrial system [BRE 11]. According to these authors, a historical and dynamic analysis re-associated with the national and international perspective of the development of this industry may be useful if we want to explain the non-emergence of certain ecosystems. Their cross relationships are essential to understand the development paths of these forms of organization, once we acknowledge that it is possible for a CEO, a highly qualified employee or a company to belong to a dynamic industrial community without remaining in the same position throughout the lifecycle of this company. We should not be so naïve as to think that the ecosystems we analyze are not subjected to forces of social fragmentation and hierarchical relationships.

When an ecosystem is progressively implemented, the growing number of individuals involved increases its legitimacy in relation to two aspects [ALD 94]: a cognitive one, referring to the extent of knowledge about this activity, which becomes meaningful for producers and consumers, and a socio-political one, referring to the public acceptance of this activity, the public aids it benefits from and the prestige of its leaders. Cognitive legitimization in the eyes of the producers reinforces the trend of industries, especially high-tech sectors, to be geographically grouped together, adopting an organizational structure called regional cluster⁴. This suggests both that the proximity of the knowledge sources that represent R&D inputs has actual significance [AGR 06] and that newcomers will more likely copy an existing organizational form rather than have to test it⁵. Besides, the creation of research consortia or joint ventures in a region simultaneously facilitates the production of localized knowledge and their protection by means of intellectual property rights. More generally, once a collective action is directed at a common goal, the legitimate activities are appropriated by the members of organizations and mutually interpreted, thus becoming customary.

Nonetheless, if the multiplication of components and of their relationships – therefore, the increase in the diversity of these environments – may represent a response to the imperative need for innovation by favoring exploratory abilities, compromises about other characteristics concerning performance (cost, environment, quality of the relationships, etc.) are required. Diversity comes at a price [STI 07]. Some options are inevitably constrained in terms of available resources and they require: either downscaling or more limited types of learning or growing costs, increased complexity or a loss of coherence when the number of components becomes larger and the transaction costs increase (information overload, high coordination and communication costs). The difficulty lies in finding a balance between the advantages of diversity and the contrary forces that emerge when governing knowledge, innovation and choices concerning development.

The literature of evolutionism, thanks to the concept of related variety, underlines the importance of assembling different pieces of knowledge that, however, complement one another. The extent of diversity is then limited: knowledge is transferred from one activity to another when it is complementary in terms of skills. This phenomenon may take place in a sector or between sectors as long as there is a technological relationship. This analysis, carried out on a regional level [ASH 11], can also be applied to those clusters believed to be able to produce diversity in related activities. Budding takes place through mechanisms of knowledge transfer: spinning off, diversification of companies, labor mobility and networking of actors. The authors quoted admit that these transfer mechanisms enjoy a local advantage: spin-off firms and the mother firms are in the same position, new divisions of existing firms develop in organizations that are already in place, jobs are taken up by new individuals in a localized work market and knowledge networks are often organized by actors who are socially and geographically close.

This analysis, which attaches more importance to the mechanisms of production and knowledge transfer as well as the necessary diversity of actors, is quite different from the approach of agglomeration economies as a significant determining factor of geographic location [ELL 99]. We know that Marshall forces that lead to agglomeration include three categories of transport costs: goods, people and ideas. In this context, the costs are decreased when companies become closer, as they share certain features. The same can be said about the co-agglomeration of industries that may be similar in some aspects and differ in others [ELL 10]: for example, two industries may exchange goods (input–output relationships) but employ individuals with different qualifications or hire employees with the same qualifications without exchanging goods. Empirical tests unambiguously show that the Marshall forces affect the models of co-agglomeration of industries, the natural advantages offered by a given location. The input–output flows represent the most significant factor, whereas the flows of technological knowledge have a far less important effect.

The conceptual framework thus described is not suitable for the analysis of innovation and production ecosystems. Naturally, input–output relationships are in line with the existence of value chains that justify the presence in a certain location of suppliers belonging to different industries. Similarly, the creation of companies is linked to the presence of other industries, whose employees and human capital can be highly valued by new entrepreneurs [GLA 08]. On a more general level, research works confirm that there are human capital externalities that may benefit all companies, including the existing ones. By contrast, the weak technological flows shown by R&D spending and patents do not account for richer forms of production and knowledge transfer on which the dynamics of an innovation and production are based. Such an ecosystem relies on the interaction between companies, institutional actors (universities, public laboratories, research bodies) and different kinds of service providers: audit, training, advising, finance and so on.

1.1.2. Economic behaviors and social relationships

Some innovation and production ecosystems must be conceived to emerge on a global basis (this is the case for traded clusters in the United States), as a large part of users are located outside national borders, regardless of the activity considered. This assumes a high level of cognitive legitimacy in the eyes of the users. For example, the production of software for timber houses in Sweden and Germany represents agglomerations of service companies, productive unities that are downstream of the timber sector and training institutes that have built, based on specific needs and thanks to essentially regional funding, European-sized markets. In this context, issues concerning scaling and marketing are crucial; globalization is multiplying the number of markets and intensifying their segmentation in increasingly more specific kinds of demand, leading to “semi-customized” applications of technologies of generic products. The process that allows us to reach an effective minimum production size requires public policies oriented toward the creation, in relation with private actors, of resource centers designed to produce public or semi-public goods that the market cannot create on its own: information about technologies and products, creation of instances of risk-pooling and risk reduction (insurance mechanisms), testing and certification committees and committed coordination procedures, headed by either private companies or public bodies [MIT 13] able to take on the role of anchor tenant. It is as if the selective infusion of the communication, control and normalization functions, enabled by the access to shared resources⁶, transformed the traditional hierarchical relationship.

Top-down economic approaches are ineffective in these forms of organization, because it is impossible for a single actor to identify all the relevant knowledge and mobilize all the adequate types of expertise. Clusters offer significant competitive edges in relation to firms vertically integrated in three domains⁷. Productivity is increased thanks to reduced transaction costs and interdependencies that are not market oriented. Innovation is amplified by the interactive exchange of knowledge between different actors, especially due to the proximity required by the exchanges of tacit knowledge. The creation of new companies is facilitated by the assistance provided in terms of “monitoring, role-model provision, learning, communication, and commercialization gains that arise from operating in a cluster setting” [COO 02, pp. 9–10].

We will add to this the definition of fiscal and financial procedures (namely, via specialized bodies) that facilitate the widespread dissemination of information technologies and the regional funding allocated to the production of qualifications.

Overall, the relational aspect becomes predominant if we want to ensure a key function of governance [POW 12]: the interactions between the participants enable the emergence of specific investments in physical and human assets, the definition of new practices and their internalization, while also avoiding opportunistic behaviors. An ecosystem organizes, in a particular way, the social distribution of risks and rewards (not only monetary) between the participants. Besides, the procedures and definitions that at first required several efforts, explanations and translations are progressively codified. The equivalent of a phenomenon of “compression” takes place and allows those involved to rapidly understand meanings and nuances [COL 06]. Consequently, the process that leads to the institutionalization of an innovation and production ecosystem is not accounted for by the summary analysis of the behaviors exploited and brought about by monetary incentives, or, at the very least, it makes it more complex.

The approach suggested leads us to put less emphasis on the exceptional role that may be played by some particularly qualified and innovative individuals, such as star scientists⁸ [AND 13], insofar as the effectiveness of the device depends on the coordination of the whole. It also forces us to consider more than the mere monetary incentives that limit the reflection about what employees and companies profit from or lose by belonging to a cluster. The works focused on Great Britain clearly show that a location in a cluster is not only determined by knowledge spillovers with the potential to reduce costs, for example, increasing the productivity of R&D or decreasing the costs involved in its implementation. Spillovers may involve different aspects: becoming involved in informal networks, facilitating face-to-face interactions, building formal collaborations, obtaining expertise and guidance, benefitting from the creation of university spin-offs and so on. [ABR 07]. These authors somewhat relativize the importance of human capital by combining it with the role played by institutions. In the pharmaceutical and chemical sectors, the supply of trained students or scientific professionals plays a less important role than university research, especially when the latter is of a high standard, as a location factor of companies spending on internal R&D. The social infrastructure of clusters frames the behaviors of economic agents and simultaneously increases the innovative ability of companies. In line with the neo-institutional theory, the actors are integrated in institutional environments: it is as if a script defined the roles and linked the actors to a repertoire of actions, interests and goals to be reached in a specific context [DRO 09].

We can see that it is difficult, while analyzing ecosystems, to separate the geographic dimension from the aspects related to the social structure: proximity and location in a network are closely interconnected [WHI 09]. More generally, the research carried out by these authors shows that the innovation factors in the Boston, San Diego and San Francisco Bay regions are more relational than geographic. The network effect is more significant than the effect due to location. A substantial amount of research had already underscored how, in regions that are significantly structured by clusters, networks of social relationships are the main source of new knowledge for the companies located there. The dissemination of knowledge results from high labor mobility. The analysis of the San Diego technological cluster reveals this phenomenon [CAS 07]. A high labor mobility, which allows experienced managers to leave prestigious careers to work in lucrative, if high-risk, start-ups, cannot exist without a social structure formed by firmly established and effectively organized links. A social network, once set up, decreases the risk for these qualified individuals of joining these companies again and, possibly, leaving them should difficulties arise. The social networks that establish connections between experienced managers must be regarded as institutions, and this institutional infrastructure is at the root of the development of some technological clusters in the United States.

Overall, the appearance and development of clusters rely on two interconnected explanatory factors.

The factor-focused approach draws up a list of criteria that range from the quality of academic research and the education system to the presence of venture capitalists and historical luck.

The structure-focused approach focuses less on the specific features of the factors and gives more weight to the structure of the relationships between the agents, be they companies, individuals, institutions or public authorities [BRE 09]. If these social conditions are not there, clusters will never grow, regardless of the availability of the different factors. Networks allow technological knowledge in particular to be disseminated thanks to the collective learning that has taken shape and the exchange of information between the agents⁹.

Let us also point out that the relation process is in line with a shift toward more open types of governance in terms of innovation and, in this respect, it eludes the traditional analysis of the market–hierarchy distinction and its support in terms of transaction costs, which remains relevant if we want to tackle problems concerning the structure of production [FEL 14b]. There are several governance procedures that cannot be identified as exclusively related to either hierarchical management or the market. The function of an innovation and production system does not involve solving a possible hold-up problem, but facilitating the flows of information and knowledge, especially tacit and living, which are crucial for the success of innovation¹⁰.

In this context, the integrated organization of research within a firm, as it makes it possible to exploit accumulated and diversified technological knowledge, is a recent invention that dates back to the 20^th Century. Nowadays, it tends to be replaced by partnerships and alliances, which represent at once federated structures and forms of governance that determine technological courses of action, create economies of scale and consequence and favor the exchange of knowledge. They are buffer institutions that create a neutral area between academic and industrial research and leverage their respective abilities [AND 13].

The relational design that underlies the dynamics of an innovation and production system can be established on the basis of research that has focused more specifically on firms [GIL 09]. It is becoming common to think that a key innovation involves implementing a system where the regular and mutual supply of information about everyone’s skills and desire to cooperate shows how to collaborate more effectively and create mutual types of learning in common projects. Over time, these projects are going to become more precise in relation to the reciprocal predictions and expectations of each participant whose organizational borders are open enough to allow an entity to gain access to technological knowledge situated in other entities. At first, the nature of an innovation and production ecosystem makes it impossible to qualify it: it progressively builds and elaborates mechanisms and tools of governance suitable for the context (characterized by incertitude) where it operates and the tasks it intends to carry out.

More generally, innovation is becoming more collaborative and the involvement of public bodies has changed over the last few centuries. The works carried out on a sample of 1,200 innovations chosen randomly among those considered by The R&D 100 Awards are proof of this evolution. Block and Keller [BLO 11] pointed out that in the 1970s 67% of the innovations noted were produced by private sector companies operating on their own. This figure fell to 27% in the 2000s. On the contrary, the number of innovations produced by public institutions (universities, federal laboratories, public agencies) has increased significantly, from 14% in 1975 to 61% in 2006. Similarly, the growing importance of inter-organizational collaboration, especially between private companies and universities, confirms the central role of networks in innovation processes: in 1971, 18% of rewarded innovations involved collaboration between organizations. This figure increased to 68% in 2006. If we only consider funding for innovations as a factor, the results obtained for the same sample indicate that the percentage of innovations financed with federal public funds has increased from 41% in 1975 to 77% in 2006.

1.2.1. Locations, sources of skills

The flows of knowledge received inevitably influence the activities performed within these hubs, which require a diversified range of skills in not only research and design but also production, marketing and training. It is locations themselves, enriched in this sense, that become the source of these skills. By helping the creation of these hubs, economic policy reminds us that competitiveness is not exclusively of a microeconomic nature and that competition moves toward intermediate levels made up of localized groups of companies and institutions. The rules of the meso-economic game imposed by globalization are as follows: the technologies, knowledge and skills enjoyed by a location must be necessarily different from those of other hubs, lest they become “commodities” [ZYS 07]. First, it is a matter of building different research outputs, products and services. Just like for companies, skills vary widely from location to location and represent neither a coincidence nor the exclusive outcome of a historical process. Naturally, the positions obtained in terms of assets (knowledge, abilities, etc.) have a certain weight, as do the processes implemented: the governance structures, the qualification systems as well as the degree of internal connectedness that facilitates the integration in networks [BEN 11]. What matters even more are the paths taken, as knowledge and skills are cumulative and develop over time through a series of coordinate investments [PIS 15]. The strategic choices made by companies (developing general or specific skills, focusing more on applications or widening the range of domains to which these skills will be assigned) will be all the more effective as public policies have been able to identify the most promising paths. Hence, the need for screening and selection mechanisms that avoid mere window dressing.

1.2.2. Long-term decisions

Globalization is making it harder to think long term. The policies in favor of innovation and production ecosystems involve long-term perspectives by means of irreversible decisions that need meticulous planning and a larger amount of information than reversible decisions. Globalization tends to shorten the lifecycle of technologies, narrowing the margins of opportunity. We should add to this the instability of the marginal conditions that affect the environment of the decision: quality of data and prediction models, degree of investment spillover, difficulties involved in financing the first models of a series and so on. The result is that decision makers, due to time constraints, have to lower the essential standards of rationality. In other words, decision makers cannot plan their long-term decisions, which, by becoming less rational, run the risk of being faced with unpredicted organization failures. Thus, upstream, it is necessary to increase the abilities and expertise of the public actors in charge of playing a part in some projects that are particularly physical capital, infrastructure and knowledge intensive. Nonetheless, “as the number of options increases, the peoples’ ability to accurately evaluate the different options declines” [WOR 15, p. 181].

A type of failure that is often encountered involves the quality of partnership relations. From this perspective, public policies may re-define the relationships between companies by establishing legal-level binding rules of conduct that large-size operators cannot get around. They must also aim to help small and medium-size companies to cross the growth thresholds by allowing them to gain access to funding tailored to their stage of development (venture capital in the initial phases), to improve their R&D abilities in order to access common resource centers and, consequently, to seize technological opportunities by bettering the quality of the mechanisms of technological transfer.

However, we should not forget that universities in the United States have played a major part in the development of fundamental knowledge, applied science and engineering sciences by producing useful or instrumental knowledge, namely knowledge that is integrated in final products or manufacturing technologies [ROS 13]. In particular, empirical observation and the usefulness of carrying out tests, without necessarily resorting to an abstract scientific model, represent a set of latent opportunities for instrumental knowledge for some SMEs. Knowledge transfer requires inter-organizational structures (e.g. as ad hoc laboratories) that favor active collaboration between researchers and practitioners. Therefore, spatial proximity is likely to facilitate the transfer of knowledge from research toward small structures. The advantage is that, besides gaining access to a wide range of scientific skills, companies can benefit from knowledge, whose production costs are lower, as academic researchers simultaneously aspire to monetary and non-monetary rewards (reputation through publications).

We still need to analyze the public aid allocated to R&D, which may help companies of all sizes. R&D gives birth to spillovers of two kinds: technological (or knowledge-related) spillovers with the potential to increase the productivity of other firms working in similar technological sectors within regional clusters, and negative spillovers of business competition, which lead to the subversion of some activities and losses in some parts of the market. Recent econometric works [BLO 13], carried out in the United States on a sample of 10,000 companies over a 20-year period, show that the social returns of R&D are two to three times higher than the private returns (55% vs. 20%) when we take into account net returns, namely when we consider a depreciation rate of 15%. The gap between social and private returns is much higher – and, consequently, spillovers are much more significant – for large companies than for small ones: the former have a higher degree of connectedness with other firms in their technological space, whereas small companies, which operate effectively within more limited technological niches, produce lower spillovers.

Spillovers run the risk of being less significant in proportion to the decrease in R&D spending. A recent study considering Germany focuses on this point [RAM 16]. These authors note that the development of R&D spending in SMEs does not imply a change in the sectorial composition of the productive fabric, but that it has to be analyzed in terms of its behavior, which changed in the period from 2001 to 2013. The data about German firms provided by the CIS (Community Innovation Survey) in relation to this period indicate at once a decrease in the number of companies that have introduced at least a product or a process (43.7% in 2007 and 37.01% in 2013) and a significant contraction of SMEs in terms of their R&D and innovation activities¹¹. In order to explain this, the persistence of innovation activities is associated with the size of the firm. The literature acknowledges that large companies, which can rely on complementary assets and elaborate more complex innovation strategies, are less at risk of adopting discontinuous innovation behaviors. As a result, on the sample observed, it is possible to note a strong contraction of firms that carry out R&D continuously: 28% in the sub-period from 2001 to 2007 and 19% from 2011 to 2013. The trend seems to become more accentuated after the crisis in 2008 and depends on SMEs, especially the small ones. Overall, the authors draw the conclusion that R&D behavior and the innovation of SMEs is modified in favor of a more discontinuous activity.

Persistence also affects other variables. Studies carried out on American data indicate that R&D expenditures do not necessarily boost the growth of small firms. Innovation makes it possible to cross the growth thresholds only if these companies file for patents on a regular basis. The EU’s FINNOV program reaches the same conclusion while thinking that this period, in the pharmaceutical industry, must be shorter than 5 years. All these studies converge: R&D and innovation activities only contribute to the growth of SMEs if they are constantly implemented.

This implies that public policies should target their actions more in favor of actually innovative SMEs on a sufficiently promising technological path¹². The studies that have focused on Great Britain indicate that innovative firms are fast-growing SMEs with a size of 50–170 employees. They are present in high-tech and low-tech sectors simultaneously. However, these SMEs only represent 6% of all the companies. Public intervention in favor of this 6% will yield much more in terms of growth and employment than a horizontal policy targeted at all SMEs [NES 09].

In this context, innovation and production ecosystems may be regarded as organizational innovations that allow us to counteract the trend toward the concentration of R&D and innovation activities in large companies¹³. Access to privileged funding¹⁴ and potential collaborative projects in R&D stabilize long-term perspectives and ensure the continuity of the efforts made in this direction. Rammer and Schubert aptly identified the risk involved: companies, especially small firms, with inactive R&D are likely to become occasional actors in R&D and above all less likely to carry out R&D activities on a constant basis. In relation to this, the creation of institutionalized frameworks for innovative projects favors newcomers and existing SMEs immediately, especially because these SMEs generate significant spillovers in their region. The studies conducted on the socio-economic structure of regional industries, however, underline that the effectiveness of these units is closely associated with their integration in networks of firms. We come across the well-known idea that an economic agent increases its social capital by establishing different relationships that it may use to access information and find the aspects of demand, sources of qualified labor or financial providers.

Moreover, the integration of innovative SMEs in clusters is facilitated once the latter can be legitimized socio-politically in relation to public authorities and other companies so as to be identified as partners in institutionalized practices that combine the use of technologies and rules.

1.2.3. Basic research and development of products

Innovation and production ecosystems are structured on complex problems that challenge individual solutions obtained rapidly. Easier access to information and knowledge results in the dissemination of the existing signals among a larger population of actors. Consequently, the reduction in information asymmetries increases the number of adopters and, at a second stage, innovators, besides facilitating the integration of SMEs in value chains. At an earlier stage, the reduction in information asymmetries makes it easier to coordinate knowledge in the innovation sequence. Some works have shown that a valley of death tends to appear in phase 2 (converting an idea into a potentially marketable product), which occurs after basic research, often guided by the public sphere (phase 1), and before phase 3, which involves marketing and dissemination. This may lead to the disappearance of discoveries that cannot benefit from enough funding. The less private investors act in tandem with public laboratories, the more they will lack expertise, at once scientific and commercial, to involve funding for projects in phase 2, and the more they will tend to localize their skill and investments in R&D near the final stages of the innovation sequence. Intermediate projects are characterized by a type of incertitude that is at once scientific and market related, and the sunk costs required to acquire this double skill are too high to be financed by private investors.

In other words, the valley of death appears as the producers of knowledge in stage 1 produce more outputs than the localized agents in stage 2 can process. Their objectives differ markedly: maximizing well-being through scientific and technological progress goes against the profitability-centered behaviors of private companies, which we analyze here as a behavior that involves waiting for public science to validate future technologies. As a result, firms must invest in scientific research to widen their knowledge base, and the creation of suitable organizational forms (public–private partnerships, joint ventures, specialized bodies) reduces information asymmetries, decreases sunk costs and shortens the amount of time required to go from theory to application. In this context, it is likely that transposition becomes an essential operation thanks to which knowledge and abilities are transferred to new domains where they are recombined with existing practices [POW 12]. Through this mechanism, the identities of research institutions in the United States, in Life Science Clusters, have changed, as individuals and organizations have moved from one domain to another and cognitive frameworks have been modified, for example, due to the blurring of the boundaries between basic and applied research or between public and private science.

In the United States, the National Institute of Standards and Technology (NIST) has just taken the interesting initiative of funding two specialized institutes belonging to the National Network for Manufacturing Innovation (NNMI). These institutes group industrial producers, universities, higher-education communities, federal agencies and public organizations with the aim of bridging the gap between basic research and the development of a product. Innovation lies in how the NIST (see Box 1.1) has not pre-determined at the beginning a precise sector for each ecosystem: suggestions may concern any project related to industrial robotics and biopharmaceutical manufacturing, as these two fields are identified as critical technologies likely to meet national needs and strengthen competitiveness.

As for the NIST, public intervention aims to invest in production technologies at a precompetitive stage that are relevant to the national needs and whose applications involve a wide range of activities. Operators can rely on shared assets that encompass cutting-edge technological equipment and training opportunities for employees immediately.

In other countries, innovation programs are offered in technological sectors. This is the case for Singapore, which favors the following four technological sectors: healthcare, advanced manufacturing, digital services and economy, urban solutions and sustainability. These fields have a largely transversal and inter-sectorial nature, involve several activities and meet societal and global needs immediately. The city-state of Singapore seems a place where innovation is managed in various ways.

1.2.4. Innovation and production ecosystems and the choice of location

Due to globalization, innovation and production ecosystems may have to make choices about the location of production. In the optoelectronic sector, the manufacturers of components organized in clusters had to tackle the following problem: developing new integration technologies on-site or outsourcing to developing countries the production of non-integrated existing technology. Integrated technology costs less when it is produced in the United States, whereas existing technology is more competitive when produced abroad [FUC 10]. Two elements have favored outsourcing: the uncertainty of the domestic market and the exchangeable nature of the service provided, which manifests itself as a more favorable wage/productivity ratio (for equal skills) in emerging countries. Choosing a technology ultimately depends on the context, historical circumstances and the actors’ behaviors rather than any kind of optimizing rationality. We know that inferior options may prevail over apparently superior technologies.

Nonetheless, outsourcing production creates some holes in the productive fabric (loss of information and knowledge, severed links with providers) and implies an endogenous obstruction of innovation and a questioning of the dynamics of an innovation and production ecosystem. When faced with this, public authorities have reacted in two ways. First, they promoted the creation of consortia that favored the entrance of technologically similar firms driven by the desire to carry out research in and manufacture integrated technologies. Then, they resorted to Defense Agencies so as to obtain funding for these technologies.

Overall, we can see after a few years that technological and productive offsetting effects have been brought about by larger innovation and production ecosystems. Thus, the make-up of an innovation and production ecosystem can change once links outside the cluster have been established, especially by large companies. Similarly, the anchor tenant changes his identity when new technological opportunities arise. We can also see that, in the context of globalization, choosing a technology is affected by the location of the production.