G Model RESPOL-2410; No.

of Pages 12

ARTICLE IN PRESS
Research Policy xxx (2010) xxxxxx

Contents lists available at ScienceDirect

Research Policy
journal homepage: [Link]/locate/respol

Learning at the boundaries in an Open Regional Innovation System: A focus on rms innovation strategies in the Emilia Romagna life science industry
Fiorenza Belussi a, , Alessia Sammarra b,1 , Silvia Rita Sedita a,2
a b

Department of Economics and Management, University of Padova, Via del Santo 33, 35123 Padova, Italy Faculty of Economics, University of LAquila, Via Giovanni Falcone 25, 67100 Coppito (AQ), Italy

a r t i c l e

i n f o

a b s t r a c t
The paper investigates the existence of an Open Regional Innovation System (ORIS model). This model is characterised by the rms adoption of an open innovation strategy, which overcomes not only the boundaries of the rms but also the boundaries of the region. Using data collected in a sample of life science rms, our research provides the evidence that the Emilia Romagna RIS has evolved towards an ORIS model, where rms innovation search strategy, despite being still embedded in local nets (involving several regional public research organisations PROs), is open to external-to-the-region research networks and knowledge sources. It also shows that innovation openness inuences signicantly the rms innovative performance. 2010 Elsevier B.V. All rights reserved.

Article history: Received 28 February 2007 Received in revised form 6 October 2009 Accepted 20 January 2010 Available online xxx Keywords: Life science sector Learning at the boundaries Regional Innovation Systems Networks Open innovation model

1. Introduction Over the last decades several new concepts have emphasised the systemic character of innovation, which overcomes the old view of innovation being just a specialised activity that depends directly on the amount of resources involved in subsidised science and in internal-to-the rms R&D laboratories (Audretsch et al., 2002; Metcalfe, 2005). This has been elaborated along different directions, looking at innovation as: an interactive linked-chain connecting numerous departments (R&D, design and engineering) and activities within a rm (Kline and Rosenberg, 1986); a complex modality (Carlsson, 1994) involving private and public actors, through industryuniversity collaborations, claimed to be fruitful for both the actors involved in the relationship, within a national system of innovation or a triple-helix approach (Etzkowitz, 1998; Freeman, 1987; Lundvall, 1992); a specic system characterised by its sectoral dimension, where innovation and technological change show different rates, types, and trajectories, depending on the sector in which they take place,

and where agents and institutions of a sector exert a major inuence (Malerba, 2002, 2004); a distributed process (Coombs and Metcalfe, 2000; Robertson and Smith, 2008) that integrates components, skills and knowledge from several organisations especially in the elds characterised by increasing modularisation of complex products (Brusoni and Prencipe, 2001) and decomposability of innovation-related tasks (Valentin and Jensen, 2003); a geographically bounded spatial system, whose extension can be either national3 or regional, or local (Cooke et al., 1997; Cooke, 1998; Lundvall, 1992; Nelson, 1993; Saxenian, 1994). In a globalised (and interconnected) world, the examination of spatial

Corresponding author. Tel.: +39 049 827 4051; fax: +39 049 827 4211. E-mail addresses: [Link]@[Link] (F. Belussi), sammarra@[Link] (A. Sammarra), [Link]@[Link] (S.R. Sedita). 1 Tel.: +39 0862 336671; fax: +39 0862 336481. 2 Tel.: +39 049 827 4236; fax: +39 049 827 4211. 0048-7333/$ see front matter 2010 Elsevier B.V. All rights reserved. doi:10.1016/[Link].2010.01.014

3 As argued by Carlsson and Stankiewicz (1995), nation-state constitutes a natural boundary of many technological systems, but sometimes, it makes sense to talk about a regional or local system. The elements of a shared culture (including conventions, which are tacit, and rules of the game, which are more codied elements embedded in micro-institutions), common administrative and political dimension are another parameter discussed in Cooke et al. (1997): RISs form a mosaic within a single national system of innovation. A key question with respect to RISs is the extent to which they are systemic. The basic point is that there are a number of discrete elements connected by specic relationships. The RIS can be specied in abstract including organisational element and linkages among them (public actors: like universities, research institutions, skills-development agencies, technology-transfer agencies, science park and incubators, public funding entities, patent ofces, etc., and private actors: rms, venture capital organisations, banks, consultants, legal consultants, etc.). Cooke et al. (1997) further hypothesise that there are weak/intense, regular or irregular interactions which shape the system; so we can draw many potential empirically-founded typologies of systems.

Please cite this article in press as: Belussi, F., et al., Learning at the boundaries in an Open Regional Innovation System: A focus on rms innovation strategies in the Emilia Romagna life science industry. Res. Policy (2010), doi:10.1016/[Link].2010.01.014

G Model RESPOL-2410; 2 No. of Pages 12

ARTICLE IN PRESS
F. Belussi et al. / Research Policy xxx (2010) xxxxxx

scale seems quite a difcult task, being compressed between the existence of overlapping multiple scales and spatial ows. In fact, while contemporary reections deal with the issue of the internationalisation of national innovation systems (Carlsson, 2006), others reect upon the local-global connections (Bathelt et al., 2004). Assuming the complex theorising about the nature, evolution, and impact of the innovation systems, this paper investigates the existence and the performance of an Open Regional Innovation System (ORIS model). This model is characterised by the rms adoption of an open innovation strategy, which overcomes not only the boundaries of the rms but also the boundaries of the region. Our empirical analysis is grounded on the study of 78 life science rms in Emilia Romagna, which was described as a RIS.4 Using primary data collected at the rm level, through eld interviews, the study explores in detail innovation openness in the life science RIS, focusing on the rms willingness to overcome both the organisational boundaries, through the use of external sources of innovation, and the regional boundaries, through long distance research collaborations. The paper applies quantitative research methodologies. In particular, it presents some descriptive analyses on the innovation sources (internal and external) used for innovation and it applies the social network analysis to map the geographical distribution of R&D linkages. The study then offers some indications on the effect of innovation openness, testing the impact of internal and external (regional and global) innovation sources on rms patenting activity. The paper proceeds as follows: Section 2 focuses on the theoretical and empirical background and introduces the research issues and objectives. Section 3 describes the method and data used in the analysis. Section 4 gives descriptive results, while Section 5 contains an econometric analysis. Section 6 proposes some concluding remarks and hints for further research. 2. Theoretical background and empirical context 2.1. RIS and open innovation models RIS is one of the most inuential concepts developed in the context of regional science studies, which has grown rapidly since the middle of the 1990s (Braczyk et al., 1998; Cooke et al., 1997; Iammarino, 2005). The notion of RIS lies on the crossroads of two main bodies of literature: evolutionary theories of economic and technical change, which conceptualise innovation as the result of complex, non-linear social processes, stimulated and nurtured by several actors and factors within and outside the rm (Edquist, 2005), and theories of regionalisation and clustering, which emphasise that economic growth and innovation do not take place in abstract spaces, but are locally rooted, thanks to the advantages of spatial proximity, social embeddedness, interaction with local institutions, and knowledge spillovers (Camagni, 1991; Cooke, 2002; Maskell and Malmberg, 1999; Porter, 1998; Storper, 1997; Belussi and Sedita, 2009). According to the RIS approach, regions, especially when they have developed industrial clusters, and an appropriate administrative framework for supporting innovative enterprise, are meaningful loci of innovation, fostered by direct and indirect linkages, cooperation and synergies among local eco-

4 In Cooke et al. (1997), in the Emilia Romagna, the autochthonous emergence of industrial districts, characterised by collaborative and competitive business practices and nanced by local banks, led the region, in partnership with banks and other organisations, to support localised innovation centres in promoting systemic interactions around the use of new technologies in traditional industries. In doing so, the region has developed a collective identity and rules of the game which have resulted in the Emilian model being exported worldwide. p. 481.

nomic actors and institutions. In particular, as regards the evolution of RISs, recent analyses have elucidated the complex relationship that exists between the building of a coherent innovation system and the evolution of innovative local clusters. The undergoing hypothesis is that in fashion-led or engineering-based industry characterised by the presence of practical knowledge the development of a competitive regional system is the result of a pre-existing industrial cluster (Belussi, 2002), while in science-based industry where the role of scientic/analytical knowledge5 is extremely important the presence of leading research institutions (and research funds) represents the sine-qua-non precondition (Asheim and Coenen, 2006). The inuence of the RIS has grown rapidly since the middle of the 1990s (Asheim et al., 2003; Braczyk et al., 1998; De la Mothe and Paquet, 1998; Doloreux, 2002; Howells, 1999). Autio (1998) illustrates RIS as composed by two interdependent sub-systems embedded in a common regional socioeconomic and cultural setting. The industry sub-system, which includes the companies, their clients, suppliers and competitors, and the institutional sub-system, which consists of various institutions that are engaged in the production and diffusion of knowledge and skills such as public research organisations (PROs), technology mediating organisations, universities and other educational organisations. In high-performing regions, these two sub-systems are expected to share strong interactions sustaining a continuous and virtuous process of knowledge generation, diffusion, application and exploitation. As pointed out by Tdtling and Trippl (2005), RISs should not be understood as internally homogeneous systems since they can encompass and usually do several industries, clusters and/or industrial districts. Further, RISs should not be conceived as isolated entities, since they are encapsulated in national and supra-national innovation systems. RISs can also have various linkages with external actors and the importance of these external ties has been increasingly recognised as crucial in accelerating technological change (Coenen et al., 2004) and innovation processes also in many SMEs (Asheim and Isaksen, 2002). Many studies have tackled this issue referring to the role of multinational corporations (Zander, 1999; Cantwell and Iammarino, 2003). While the traditional RIS approach has above all emphasised the importance of local sources of innovation, more recent studies have underlined the crucial role of accessing knowledge and innovation from the outside (Bathelt et al., 2004; Boschma and ter Wal, 2007; Gertler and Levitte, 2005). Actually, the structure of some high-tech industries seems to conceal globalisation and regionalisation through a small world pattern of connections: some local rms often the leading enterprises play the role of knowledge gatekeepers (Owen-Smith and Powell, 2004), which search for and absorb non-local knowledge, and transmit it to other actors within the RIS. The RIS approach draws on the view that innovative dynamics are not held within organisational borders or single rms research units. Similarly to RIS, other recent research approaches have contributed to afrm a new model of open innovation that overcomes the single rms boundaries. Chesbrough (2003) observed that organisations increasingly resort to open innovation strategies to augment the variety and speed of knowledge ows essential to innovation. According to Chesbrough (2003), the old closed innovation model, which dominated most of the 20th century, is becoming obsolete today. The increasing importance recognised to external sources of knowledge and innovation is threatening the closed model, which postulates the effectiveness of vertically integrated R&D departments, aimed to develop technology in-house for their sole use. Chesbrough claims that also modern technology

For a denition see Cooke (2005).

Please cite this article in press as: Belussi, F., et al., Learning at the boundaries in an Open Regional Innovation System: A focus on rms innovation strategies in the Emilia Romagna life science industry. Res. Policy (2010), doi:10.1016/[Link].2010.01.014

G Model RESPOL-2410; No. of Pages 12

ARTICLE IN PRESS
F. Belussi et al. / Research Policy xxx (2010) xxxxxx 3

powerhouses, like Cisco and Microsoft, have dropped the do-it-allyourself approach, and pioneered a new model of open innovation, in which companies import ideas from outside, and let their own innovations enter the wider marketplace. Today, the largest rms spread around the world their R&D laboratories around new poles of excellence (Chiesa, 1995). In the closed innovation model, small rms, which clearly are not able to afford large investments in R&D, are at a disadvantage. These types of rms cannot be innovative only counting on internal knowledge, but they have to build fruitful relationships with other organisations, in a network perspective. Firm size in the emergent open innovation model is no more an obstacle or unique driver to increase innovative productivity (Belussi and Gottardi, 2000). Relational and coordination capabilities of rms and research labs allow the establishment of a positive spiral of learning at the boundaries (Belussi et al., 2008), which appear to be the crucial variable to look at when determining the degree of innovativeness (Boari and Lipparini, 1999; Lipparini and Sobrero, 1994; Lorenzoni and Lipparini, 1999; Giuliani, 2007). Crossing the boundaries of the rm and cooperating with external actors (research labs or institutions) is an opportunity to multiply the learning occasions, mostly in knowledge intensive sectors (Baba et al., 2009). These industries have witnessed the wide spread diffusion of distributed forms of innovation driven by the necessity to integrate specialised and complementary knowledge (Coombs and Metcalfe, 2000). Recent studies on modularity and system integration (Brusoni and Prencipe, 2001; Hobday et al., 2005) have pointed out that increasing modularisation of complex products has favoured rms R&D disintegration and knowledge specialisation although raising the need and complexity of interorganisational coordination. External sources represent a form of learning at the boundaries, which is rooted in the capability to enrich the rm knowledge with a network of interactions including external partners (suppliers, customers, research and market institutes). Knowledge ows are easier, thanks to the increasing adoption of international-shared languages and ICT infrastructures (Castells, 2000). Consequently the number of innovation sources for the creation of new processes and productive inputs increases, favouring economies of variety, and enhancing the heterogeneity of rms. 2.2. Innovation trajectories in the life science industry The life science concept emerged in the mid-1980s along with a novel emphasis on biotechnology, to provide a rationale for the new technological trajectory with strengthened links between the agrochemical and pharmaceutical divisions of multinational chemical companies (Tait et al., 1990). The growth of biotech gave rise to a novel technological regime with specic characteristics of appropriability (very high), opportunity (very high), cumulativeness (very low at the beginning), and differentiated knowledge base (with the shift from the chemical paradigm to the bio-molecular technology). The organisation of R&D activities in the high-tech sectors of life science rms (biomedical, biotech, pharmaceutics and computer science applied to the medical elds), typically based on the exploitation of scientic knowledge, represents one of the main components of the innovative activity of the enterprises. However, at the same time, life science rms show a systematic access to external knowledge sources. Not all of the useful knowledge for the innovative activity of the enterprises originates inside the rm, and the absorbing activity of external knowledge from outside appears equally fundamental (Cohen and Levinthal, 1989, 1990). Science-based rms produce and absorb knowledge in a continuous game of interactions that generate a hybridisation of the possessed knowledge, and give rise to a cumulative process of acquisition of knowledge and competences, becoming the quintessence of a new

model of rm (Antonelli, 1999; Dasgupta and David, 1987; Winter, 1987). On the one hand, life science rms innovation strategy develops within open models, characterised by their participation in inventive and innovative activity on an international scale (Baum et al., 2000; Niosi, 2003). The life science eld shares a strong interface with research activities placed on the frontier of scientic knowledge discovery. The innovative activities introduced are therefore strongly tied to the public research (R&D projects on basic research) developed within the scientic institutions, and collectively shared (Arrow, 1994). In this eld the importance of non-prot, public and semi-public research institutions is signicant. They are a crucial source of public knowledge and they contribute to the sponsoring of the most promising rms projects (Cooke, 2002; Nelson, 1992; Nelson and Levin, 1986). On the other hand, life science rms strategy is planned within closed systems, in order to protect their dedicated technological knowledge (tacit or codied), building intangible reservoirs for creative activity of invention and innovation (Arora and Gambardella, 1994). Science-based rms protect their capability (Penrose, 1959; Kogut and Zander, 1992) and proprietary knowledge especially through the maintenance of the industrial secret (Levin et al., 1987) and patenting activity (Foray and Hargreaves, 2002). However, the necessity of acquiring complementary knowledge, in order to guarantee the progress of the research projects, tends to stretch them towards a more open model. Complementary production abilities, technological interdependences, interactions with customers and end users, and the ability to recombine different, distant, but equally necessary, sources of knowledge militate in this direction (Penner-Hahn and Shaver, 2005). Successful rms are those which invest in strategic resources, not just in in-house R&D, but in effective relations with suppliers, sub-contractors, and service rms (Teece et al., 1997; Wernerfelt, 1984). A crucial element for accumulating strategic resources is to develop numerous channels to absorb information (meetings, participation in international fairs), codied technical knowledge or know-how (university R&D, consultants, reverse engineering techniques, strategic alliances with knowledgeable suppliers, etc.). The emergence of a complex network of cooperation, as Richardson pointed out long ago (1972: 888) is explained by the need to combine closely complementary but dissimilar activities (from R&D to marketing) that in certain circumstances cannot be allocated either straightforwardly to the market (because of the existing complementarities with rm assets), or to the rm itself (because it lacks the required capabilities). The literature on system integrators (Hobday et al., 2005) discusses in detail the relationship between the division of cognitive labour and organisational structures, providing signicant insights on the role of innovative assemblers, the latter being considered repositories of architectural knowledge (Henderson and Clark, 1990). These arguments all together tend to support the motion for an open innovation model, because The process of innovation is becoming more distributed across rm boundaries (Coombs and Metcalfe, 2002: 263). Following Chesbrough (2003), West and Gallagher dene open innovation as (. . .) systematically encouraging and exploring a wide range of internal and external sources of innovation opportunities, consciously integrating that exploration with rm capabilities and resources, and broadly exploiting those opportunities through multiple channels (West and Gallagher, 2006: 82). In life science rms, then, technological spillovers appear to be extremely frequent. They take place both inside the relationship between rms and public research centres, or in localised networks of enterprises, or again, in global international networks of enterprises (Owen-Smith and Powell, 2004). This happens in all life science clusters (or technological districts) like in the area of

Please cite this article in press as: Belussi, F., et al., Learning at the boundaries in an Open Regional Innovation System: A focus on rms innovation strategies in the Emilia Romagna life science industry. Res. Policy (2010), doi:10.1016/[Link].2010.01.014

G Model RESPOL-2410; 4 No. of Pages 12

ARTICLE IN PRESS
F. Belussi et al. / Research Policy xxx (2010) xxxxxx

Cambridge and Oxford in the UK, or in Boston and Minneapolis in the US, or in the life science scientic parks, like Sophia Antipolis, in France (Longhi, 2002). The degree of decomposability of complex problems, introduced by Simon, represents another way to underline the cooperative nature of inter-organisational relations. This concept has been further applied by Valentin and Jensen (2003), in the realm of biotech in food processing. In order to investigate the role of research collaborations on rm innovativeness, performance and growth, many theoretical and empirical studies have focused on the type and location of external partners (Dahlander and McKelvey, 2005; McKelvey et al., 2003). With respect to type of partners, the literature on innovation systems (Cooke, 2001; Lundvall, 1992; Nelson, 1993) and the triple-helix model (Etzkowitz and Leydesdorff, 1997) have both stressed the fundamental role in boosting innovativeness of close interactions among heterogeneous actors such as large rms and SMEs, venture capitalists, end users, universities and other public and private research institutions. With respect to the spatial distribution of partners, some authors suggest that external research collaborations should be more frequent and effective among co-located partners, while others emphasise the importance of research links with geographically distant partners in order to get access to global circuits of knowledge creation and diffusion. The former argument is based on the idea that the complexity of the innovation process requires direct contacts between partners for the success of research collaborations (Pisano et al., 1988). Spatial proximity should enable the transfer of tacit knowledge and facilitate the exploitation of knowledge spillovers (Malmberg and Maskell, 2005; Maskell, 2001). The geography of spillovers has been also analysed in the context of patent citation. Comparing the geographical location of patent citations with that of the cited patents, Jaffe et al. (1993) found evidence that knowledge spillovers are geographically localised. In opposition with this view, the latter theoretical perspective sustained the importance of global sources of external knowledge, claiming that, in high-tech industries, innovation requires knowledge that is both global best and diverse (Dahlander and McKelvey, 2005). 2.3. Research issues and objectives The theoretical overview presented above underlines that in many sectors, and especially in the life sciences, two intertwined phenomena are modifying the traditional model of innovation: (i) the increasing relevance of external to the rm knowledge sources, which foster the adoption of a new open/distributed innovation model and (ii) the need of exploiting the advantages of local knowledge spillovers while getting access to non-local sources of knowledge and information. Whereas the rst phenomenon is perfectly consistent with the traditional RIS model typically based on the view of innovation as an interactive process the second phenomenon implies the evolution towards a new model of open RIS (ORIS). On the basis of this consideration, the objective of the present paper is to examine the empirical existence and the performance of the ORIS model in the Emilia Romagna life science industry. In order to do so the study focuses on three research questions: (i) External sources of innovation: To what extent do rms within the RIS rely on external innovation sources? What are the most important sources according to rms evaluation? (ii) Geographical scale of innovation relationships: Do rms prefer to engage in collaborative research relationships with proximate partners (within the RIS) or external ones (national, European or extra-European)?

(iii) Effectiveness of open innovation model: Is the open innovation model more effective than the closed one in terms of patenting activity? Which is the impact of internal sources (R&D efforts) and external sources (variety of external channels and number of research ties) on rms innovativeness? The rst two issues aim to assess the openness of the life science RIS (towards an ORIS model). In order to do so, the study examines which are the external channels more frequently activated by the interviewed rms, and the importance they attribute to each source for innovation. The study also provides a descriptive analysis of research collaborations, useful to capture the geographical scale of the inter-organisational division of innovative tasks in the life science RIS. This analytical framework is designed to provide a comprehensive and bottom-up understanding of the existence of the ORIS model. When local rms are strongly oriented towards the use of a breadth search strategy, cultivating multiple channels for innovation (both institutional and market-based) and are engaged not only in local research collaborations, but also in extraregional ones, the RIS ts what we have here called the ORIS model. The dominant mode of innovation here is largely inuenced by the rms capability of learning at the boundaries. The third research question is concerned with testing the effectiveness of the open innovation model. In order to do so, the study applies multivariate analysis to test the impact of internal sources (R&D investments) and external sources (variety of external channels and number of research ties) on rms innovativeness. The statistical test presented in the study is a negative binomial regression, where the dependent variable is innovative performance (measured as the number of patents registered in the period 20002004), which is explained by R&D expenditure, number of external research collaborations, and variety of external sources used for innovation search. The model also includes a number of control variables, such as size, age, and rms sub-sector. 3. Empirical setting, data source, and sampling procedure The empirical context of this study is the life science industry in Emilia Romagna.6 Our denition of the sector includes the following specialisations: biomedical, biotechnology, pharmaceutics and computer science industry applied to the medical elds. Therefore, our study does not focus only on dedicated biotech enterprises, including all rms active in the knowledge areas of the modern life science industry. Several data sources were used to build a systematic database of the population of rms involved in the Emilia Romagna life science sector. Our research work started with the inspection of the Cerved archive, an existing database provided by the Italian Chamber of Commerce, which includes all the registered rms not only legally founded but also operative. We updated and completed this archive using information gathered through other sources: websites inspection, interviews with regional experts and queries of rm associations archives, like Consobiomed, the association of small rms belonging to the biomedical district of Modena. We identied three productive segments: (1) biomedical rms, which produce medical appliances and disposables for diagnosis and therapeutic aims; (2) pharmaceutical and biotech rms, even if the latter are little developed in the Emilia Romagna region and in Italy at large; (3) information science rms that have developed specic medical applications, mainly in the eld of distance telemedicine.

An extended work related to our empirical survey is published in Belussi (2005).

Please cite this article in press as: Belussi, F., et al., Learning at the boundaries in an Open Regional Innovation System: A focus on rms innovation strategies in the Emilia Romagna life science industry. Res. Policy (2010), doi:10.1016/[Link].2010.01.014

G Model RESPOL-2410; No. of Pages 12

ARTICLE IN PRESS
F. Belussi et al. / Research Policy xxx (2010) xxxxxx 5 Revised Cerved population Firms N % 2.34 20.66 30.60 41.13 2.73 2.53 100.00 Employees N 31 3,306 1,917 1,947 4,264 169 11,643 % 0.27 28.42 16.48 16,74 36.65 1.45 100.0 Sampled rms Firms N 5 11 28 21 3 10 78 % 6.41 14.10 35.90 26.92 3.85 12.82 100.00 Employees N 28 2,535 892 680 165 169 4,469 % 0.62 56.73 19.97 15.21 3.69 3.78 100.0

Table 1 Sectoral classication of life science rms: universe and sample of rms. Manufacturing rms of the Emilia Romagna life science sector

Diagnostic Therapy and rehabilitation Disposables Other appliances Pharmaceutical and biotech Computer science applied to medicine Total

12 106 157 211 14 13 513

Our elaboration on Cerved data, website and interviews with sector experts.

In turn, the biomedical segment has been divided into four subareas of specialisation: (i) diagnostics, (ii) therapeutics (complex machinery), (iii) disposable, and (iv) other electro-medical or nontherapeutic machinery (other apparatus and appliances).7 Table 1 provides the distribution of rms and employees for the population of life science rms identied. Overall, the Emilia Romagna life science sector appears quite signicant more than 500 rms and about 11,600 employees (data refer to all productive and commercial rms). The main sectors are, respectively, therapeutic and rehabilitation, other appliances, pharmaceutical and biotech. From a spatial point of view, the regional life science system is composed of a clear-cut industrial district, centred in the province of Modena, and a mosaic of niches of dispersed producers, mainly localised along the ModenaBologna axis, where there are also numerous regional clinical institutions and universities (important university centres and medical clinics are also in the nearby cities of Ferrara and Parma). A simple random sampling technique was applied, and 78 rms entered the sample.8 We assigned the sampled rms to each subsector using the information obtained directly during the interview, or using the Cerved archive (which describes the sector of activity in some dedicated records). From a spatial point of view, the interviewed enterprises are mainly concentrated in the province of Modena (42 cases corresponding to 53.8% of the rms, and 84% of the employees), where the only Italian biomedical industrial district is located.9 22 rms are based in Bologna, while the others are dispersed within the region. Semi-structured interviews were conducted in the period MarchSeptember 2005 by a group of four researchers (Alessia Sammarra, Massimo Gastaldon, Alberto Corazza and Tito Casali) directed by Fiorenza Belussi. Interviews were mainly organised with the entrepreneur, owner of the rm, or with the manager delegated to deal with innovation activities. Interviews lasted 12 h and were focused on the history of the rm, products, innovation capability, R&D investments, patents, markets, number of competitors, recent growth trend, external R&D cooperation, and the relationships with the regional supporting institutions. Interview data on the number of employees, rm sales, and number of international

patents registered by the rms were controlled through the inspection of rms web sites and on-line ofcial sources (Chamber of Commerce, European Patent Ofce). The utilisation of multiple sources to collect data allowed us to reduce the risk of the variables used for descriptive and multivariate analyses being affected by single-source bias.

4. Descriptive results 4.1. Innovative search strategy In order to assess the degree of openness of rms search strategy, rms were asked to indicate which external sources they use for their innovation activities, and evaluate the importance of each source on a scale from 1 to 10. The interview questionnaire contained a list of 16 possible sources (see Table 2), grouped together under three different headings (market-based, institutional-based, and semi-public). Overall, our ndings indicate that rms use on average about 5 external knowledge sources out of the total of 16. The most frequently used source is clients and customers (or end users). Indeed, 65% of respondents gathered some information useful for innovation activities from their clients. Among all the market sources, clients and customers is also considered the most relevant, receiving an average score of 8.57. This result is consistent with the Innovation Systems approach, which emphasises the importance of interactions among actors in the industry subsystem as crucial driver of innovation, especially with regards to close relationships with customers and end users. The other most frequently used sources are the semi-public ones, which are in order the Internet (56%), scientic publications (51%) and fairs and exhibition (47%). The important role played by semi-public sources can be explained by taking into account their relative greater accessibility and lower cost compared to the other external sources of knowledge or information. It is interesting to notice that scientic publication (a form of indirect interaction with PROs) is, in absolute terms, the most relevant source according to rms evaluation, receiving the highest average score in the list (9.20). This result clearly indicates the strategic importance of keeping pace with scientic discoveries for rms operating in the life science sector, which is typically more dependent on the exploitation of scientic knowledge in comparison to other industries. The number of rms relying on direct interactions with PROs (institutional sources) is also quite relevant, although with some degree of variation within the category (which includes not mutually exclusive options). Indeed, this percentage rises up to 24% for regional universities, while it decreases to only 9% for the National Research Council (CNR). These sources are generally ranked as highly important, with an average score of 8.19.

7 See Appendix A for a short description of the sub-sectors of activity of the enterprises inserted in the sample. 8 During the sampling procedure we excluded rms involved only in assistance services, because they are not rms endowed with innovation and technological capabilities. 9 The district enterprises are located in a small bunch of municipalities which show high spatial contiguity, like Camposanto, Cavezzo, Concordia, Finale Emiliano, Medolla, Mirandola, S. Felice, S. Possidonio and S. Prospero. The biomedical district of Modena counts about 80 rms and 5000 employees, quite a small cluster if compared with the Medicon Valley, localised between Lund and Copenhagen with 1000 rms and 34,000 employees.

Please cite this article in press as: Belussi, F., et al., Learning at the boundaries in an Open Regional Innovation System: A focus on rms innovation strategies in the Emilia Romagna life science industry. Res. Policy (2010), doi:10.1016/[Link].2010.01.014

G Model RESPOL-2410; 6 No. of Pages 12

ARTICLE IN PRESS
F. Belussi et al. / Research Policy xxx (2010) xxxxxx Knowledge source Occurrence N % 29.5 11.5 12.8 21.8 10.2 65.4 19.2 10.2 34.6 24.00 9.0 30.8 29.5 19.2 22.12 47.4 56.4 51.3 51.70 Importance (score: 110) Mean 7.91 5.22 5.40 6.35 4.89 8.57 7.47 7.87 7.74 6.82 8.29 8.25 8.09 8.13 8.19 6.95 7.25 9.20 7.80

Table 2 Sources of information and knowledge for innovation activities, year 2004 (N = 78). Type

Market-based

External R&D labs Regional rms imitation National rms imitation Foreign rms imitation Research agreements with other rms Clients and customers Suppliers of intermediary goods Patent acquisitions Distribution network Average Market sources

23 9 10 17 8 51 15 8 27 18.78 7 24 23 15 17.25 37 44 40 40.33

Institutional

CNR (National Research Centre) R&D Regional Universities R&D National Universities R&D Foreign Universities Average Institutional sources

Semi-public

Fairs, exhibitions Internet Scientic publications Average Other sources

Elaboration from our survey.

Table 3 Occurrence of external research collaborations within the 78 sampled rms and total number of ties; absolute value (row percentage in brackets). Type of relation Occurrence Yes Total external relationships of which: Firm-to-rm relations Firm-to-PRO relations Our survey. 45 (58%) 8 (10%) 44 (56%) No 33 (42%) 70 (90%) 34 (44%) Total 78 (100%) 78 (100%) 78 (100%) 170 17 153 Total number of ties

Finally, quite an interesting result concerns the use for innovation activities of research agreements with other rms. A very small percentage of respondents (10%) draw on this external source, which is also considered not very relevant, with an average score of 4.89. This nding indicates a clear divide between rm-to-PRO and rm-to-rm research collaborations, the latter being a concern of a small minority of rms in our sample. This nding is further conrmed by the descriptive analysis of the occurrence of research collaborations presented in the following section. 4.2. The role of external research collaborations Given the relevance of research collaborations in the modern life science industry, we investigated in further detail the role played by this particular external source of knowledge and information for innovation. Following Dahlander and McKelvey (2005), we gathered relational data on the occurrence and spatial distribution of research collaborations among the sampled rms. Specically, respondents were asked to list the number of research collaborations they have established with other rms and PROs in the period 20002004 and to indicate the geographical location of each partner. Fig. 1 illustrates the research network established by the respondent rms with regional, national and foreign partners.10 Table 3 shows that 58% of rms in the sample are involved in external research collaborations. Indeed, out of the 78 respondents, 45 have established external ties, reaching a total number of 170 collaborations. Table 3 clearly indicates that rm-to-rm

10

The software used to analyse relational date is Ucinet 6 (Borgatti et al., 2002).

relationships have a marginal diffusion, being only 17 out of the 170 total collaborations. Indeed, only 10% of the sampled rms have established research collaborations with other enterprises. This result is quite remarkable, since it indicates a lower number of rm-to-rm relationships than could be expected, given the emphasis assigned by the literature on this type of collaborations in the life science industry. The low frequency of rm-to-rm relationships in the Emilia Romagna sample appears quite striking also with respect to the empirical ndings reported by Dahlander and McKelvey (2005) for the Gothenburg population of biotech enterprises. Indeed, in terms of occurrence, the authors found that 43% of the rms were involved in rm-to-rm relations, measured through formal arrangements. One possible reason for the difference may lie in the fact that we focused exclusively on research collaborations, while the results obtained by Dahlander and McKelvey are based on a broader set of inter-organisational relations, which include also marketing and distribution alliances. Further, compared to the typical biotech model, our results may testify a specic form of distributed innovation in the Emilia Romagna life science RIS, whose main peculiarity resides in the prominent role of PROs in R&D collaborations. Overall, our analysis shows that rms in the Emilia Romagna life science RIS are much more strongly linked to public research organisations. Indeed, in terms of occurrence, 56% have established collaborative relationships with research organisations, reaching a total number of 153 ties. In terms of the spatial distribution of research collaborations, Table 4 shows some interesting variations across the two sets of relationships analysed. Firm-to-rm collaborations occur primarily with European enterprises (41% of the ties), followed by national (29%) and extra-European (18%), while collaborations with

Please cite this article in press as: Belussi, F., et al., Learning at the boundaries in an Open Regional Innovation System: A focus on rms innovation strategies in the Emilia Romagna life science industry. Res. Policy (2010), doi:10.1016/[Link].2010.01.014

G Model RESPOL-2410; No. of Pages 12

ARTICLE IN PRESS
F. Belussi et al. / Research Policy xxx (2010) xxxxxx 7

Fig. 1. The research network within and across the RIS boundaries.

enterprises localised in the Emilia Romagna region are the least frequent (12% of the whole ties). The analysis of the spatial distribution of collaborations with PROs offers a quite different picture. Firms in the sample establish half of their external ties with regional PROs, and another half with trans-regional PROs. Our data partially align with Dahlander and McKelvey (2005), because, despite indicating that geographical location matters, they also reveal that distant research collaborations with PROs heavily characterise the rms behaviour. Overall, our results show that rms in the Emilia Romagna life science sector establish the largest part of their research collaborations with PROs localised outside the region, supporting the shift towards an ORIS model. The majority of these ties are national (31%); local rms have shown a lower propensity or capability to establish transnational research collaborations (23%). Several studies have provided empirical evidence on the positive correlation between collaborative ties and company performance measures (Ahuja, 2000; Powell et al., 1999; Stuart and Podolny, 1999). We conducted a correlation analysis between research collaborations with both PROs and rms and companies innovativeness, measured as the number of patents registered in the period 20002004. The total number of external research relations shows a signicant correlation at the 0.05 level (two-tailed) with the number of patents owned by the rm, suggesting that innovativeness is positively correlated with the rms capacity to

be engaged in multiple external relationships. In terms of type of actors involved, both rm-to-rm and rm-to-PROs relationships are signicant and positively correlated with rms innovativeness. Finally, and what is most relevant for the present study, when geographical location is taken into account, only the relationships established with foreign partners appears signicant and positively related to the rms patenting activity. 5. Econometric analysis The econometric analysis presented here investigates the role and relevance for rms innovativeness of the various components of the open innovation strategy that are respectively: (a) the internal innovative efforts (R&D investments), (b) the external sources of knowledge used for innovation and, last but not least, and (c) the capabilities of building innovative networks. The statistical method applied is a negative binomial regression. The variables entered in the model and the regression results are described in the following sections. 5.1. Measures Since we are interested in the evaluation of the determinants of rms innovative capabilities, our dependent variable (N PATENTS) is a measure of the rms patenting activity. N PATENTS measures

Table 4 Spatial distribution of the overall relations of the 78 sampled rms; absolute value (row percentage in brackets). Type of relation Spatial distribution Regional National (Italy) Foreign EU Firm-to-rm relations Firm-to-PRO relations All Our survey. 2 (12%) 76 (50%) 78 (46%) 5 (29%) 48 (31%) 53 (31%) 7 (41%) 24 (16%) 31 (18%) Extra-EU 3 (18%) 5 (3%) 8 (5%) 17 (100%) 153 (100%) 170 (100%) All

Please cite this article in press as: Belussi, F., et al., Learning at the boundaries in an Open Regional Innovation System: A focus on rms innovation strategies in the Emilia Romagna life science industry. Res. Policy (2010), doi:10.1016/[Link].2010.01.014

G Model RESPOL-2410; 8 Table 5 Variables description, year of analysis: 2004. Role Explanatory variables Phenomenon No. of Pages 12

ARTICLE IN PRESS
F. Belussi et al. / Research Policy xxx (2010) xxxxxx Variable RD EXP N SOURCE N REL Description R&D expenditure log External search breadth (number of external sources used for innovation activities) Research networking (number of collaborations established with PROs and rms for research purposes) Firm size (number of employees log) Firm age Sectoral dummies (diagnostic, therapy and rehabilitation, disposables, other appliances, pharmaceutical and biotech, information science applied to medicine i = 1, 2, 3, 4, 5, 6) Number of patents owned by the rm

Internal innovation effort absorptive capacity Openness of rms innovation strategy

Control variables

SIZE AGE SECi

Dependent variables

Innovative capacity

N PATENTS

Table 6 Descriptive statistics, year of analysis: 2004. Variable 1. N PATENTS 2. N REL 3. N SOURCE 4. RD EXP (log) 5. SIZE (log) 6. AGE N 78 78 78 78 78 78 Mean 2.10 2.18 4.67 7.68 2.88 16.04 Std. Dev. 4.30 3.21 3.25 6.03 1.46 12.20 Min 0 0 0 0 0 1 Max 22 19 16 15.12 6.68 73 1. .273* .309* .261* .508*** .374** 2. 3. 4. 5.

.341** .346** .213 .109

.434*** .277* .096

.270* .008

.344**

Note: The R&D expenditure varies from 0 to 3,700,000 D , the mean value being 290,437 D . The size of the rm varies from 1 to 800 employees, the mean value being 57.29 employees. * p < .05. ** p < .01. *** p < .001. p < .10.

the number of patents owned by the single rm in the sample. Data were gathered from the European Patent Ofce Database and eld interviews.11 Earlier studies have suggested, and assumed, that patents are a fairly good indicator of the inventive output of a rms research department and a measure of the output or success of R&D (Bound et al., 1982; Hausman et al., 1984), although they have only been able to prove a simultaneous effect of R&D on patents, without considering any lagged effect (Hall et al., 1986). The regression model includes three explanatory variables, which are listed and briey described in Table 5. The rst variable (RD EXP) measures the amount of investments in R&D allocated by the rm during the year 2004. In the innovation literature, this variable is usually used to capture two intertwined although distinct phenomena. On the one hand, the amount of internal R&D investments is considered as a proxy to measure rms internal innovation effort. On the other hand, it is used as an indicator of the rms absorptive capacity (Cohen and Levinthal, 1989). In this view, doing R&D internally should increase rms ability to integrate internally the knowledge gathered from the outside with the prior knowledge base. We used two variables, N SOURCE and N REL, to measure external innovation efforts. Specically, N SOURCE is a proxy variable similar to the one used by Laursen and Salter (2004, 2006), which indicates the number of different external sources of knowledge and information normally used by the rm for innovation activities. This indicator is calculated from the list of 16 external sources reported in Table 3. The external sources listed in the questionnaire were not mutually exclusive. They have been coded as binary variables (0 = not used; 1 = used) and then added up to create the proxy variable N SOURCE, which varies from 0 if none of the sources is

used to 16 if all knowledge sources are used by the respondent rm. Therefore, this indicator reects the rms external search breadth. The variable N REL is a proxy which measures the intensity of rms networking activity for research purposes. This indicator corresponds to the total number of external collaborative relationships established for research purposes with both rms and PROs within and across the RIS boundaries. This indicator is used to assess the impact of research relationships on the rms capability to produce innovations. It is a proxy of the rms propensity to invest in research activities through the establishment of interorganisational networks. In addition to the three explanatory variables discussed above, we included three variables to take into account the structural dimension of the rms in the sample, controlling for size, age and sub-sector. Specically, we used the number of employees (expressed in logarithms) as the measure of size (SIZE), and ve sub-sector dummies (SECi) to control for different propensities to produce innovations across the life science sub-sectors identied.12 Table 6 provides descriptive statistics and simple correlations for our variables. In particular, we can notice the signicant and positive correlation between the dependent variable (N PATENTS) and the selected explanatory variables (N REL, N SOURCE, RD EXP). As a preliminary comment, we can claim that the patent activity is positively related to the main components of an open innovation model: the internal efforts absorptive capacity (in terms of R&D expenditure), the breadth of innovation search (in terms of number of sources for innovation) and the intensity of networking activities (as number of external research collaborations).

11 We decided to integrate the information derived from the European Patent Ofce Database with the entrepreneurs declarations, to avoid the risk of not considering the patents not signed by the rm, but by the individual researcher regularly employed by the rm.

12 We are aware of the limitation given by the sample size on the regression analysis estimation (Green, 1991), nevertheless, since we do not want to infer heavily on the generalisability of the results to the population, and we are working in the eld of the social science, we adopted the simplest rule of thumb for determining the minimum number of subjects required to conduct multiple regression analyses, that is N >50 + m, where m is the number of predictors (Harris, 1975).

Please cite this article in press as: Belussi, F., et al., Learning at the boundaries in an Open Regional Innovation System: A focus on rms innovation strategies in the Emilia Romagna life science industry. Res. Policy (2010), doi:10.1016/[Link].2010.01.014

G Model RESPOL-2410; No. of Pages 12

ARTICLE IN PRESS
F. Belussi et al. / Research Policy xxx (2010) xxxxxx 9 Model 1 Coeff. Std. Err. .033 Model 2 Coeff. .245*** .337* .001 .128* .030** .023 .239* 3.717*** 3.343*** 2.764** 3.085*** 1.526 6.757*** 78 138.857 77.876*** .280 Std. Err. .050 .163 .006 .053 .010 .011 .117 1.011 .903 .867 .852 1.130 1.130

Table 7 Negative binomial regression, rm analysis (dependent variable = N PATENTS).

RD EXP (log) N REL N REL (squared) N SOURCE RD REL AGE SIZE (log) SEC1 Diagnostic SEC2 Therapy and rehab. SEC3 Disposables SEC4 Other appliances SEC5 Pharm. and biotech SEC6 Informatics Constant No. of observations Log likelihood Chi-square McFaddens R-square
* ** ***

.145***

.022 .441** 3.340** 2.817** 2.053* 2.420** .607 Benchmark 4.976*** 78 110.799 56.117*** .202

.013 .133 1.014 .896 .839 .823 1.251 .970

p < .05. p < .01. p < .001. p < .10.

5.2. Statistical method and results Since the dependent variable (number of patents N PATENTS) is a count of scores (nonnegative integers), ranging from zero to many, rather than continuous, a negative binomial model is applied as the means of estimation (Bound et al., 1982; Hausman et al., 1984; Zimmerman and Schwalbach, 1991; Crpon and Duguet, 1997; Licht and Zoz, 1998).13 This type of regression analysis is applied to deepen our understanding of open innovation models, assessing the relative importance of internal R&D investments, search activities and networking abilities, on the innovative rms performance. The results of the models estimation are reported in Table 7. To interpret the negative binomial regression estimated coefcients, we must take in account the logarithmic transformation of the dependent variable that is the count of patents (Liao, 1994). The estimated model (1), which is representative of the closed innovation model, reveals the signicant positive impact of R&D expenditures on the patenting activity of rms. A doubling of the R&D investment, in fact, is associated with a 11% increase in the number of patents.14 The estimated model (2) tests exactly what we have called the open innovation model. Here R&D expenditure and networking activities are taken into account, together with the search breadth strategy and the interaction between R&D expenditure and networking activity, the latter explaining how the impact on patents of the external relationships is dependent on the internal R&D. The model, which shows the highest tness (R-square of 0.28), capturing 30% of the variability of the dependent variable, indicates that engaging in external research relationships has a positive and signicant impact on the rms innovative output (a 10% increase in the number of relations is associated with a 40% increase in

the number of patents15 ), as well as the breadth strategy does (a 10% increase in the number of sources is associated with a 14% increase in the number of patents). Similarly R&D investments signicantly and positively affect the number of patents owned by the rm (a doubling of the R&D investment is associated with a 18% increase in the number of patents16 ). Interestingly, the interaction term is signicant and negative, informing on the costs of the joint management of networks and internal R&D. As the complexity of the innovation strategy increases, the individual positive effect of networks and R&D is mitigated. Besides, creating and maintaining networks could lead to knowledge leaking and thus diminish the returns on R&D investment efforts. Moreover, the number of relations seems to inuence positively the innovative output in a non-linear way, suggesting the existence of an inverse U-shaped relationship. The size of the rm is signicant for the determination of the innovative output, showing a higher innovative capability of larger rms, while the age of the rm seems not to be relevant. We controlled also for the sub-sectors that the rms belong to, which appear to inuence the results. Model 2 estimation suggests that an open innovation strategy, which combines an investment in internal efforts absorptive capacity (by means of R&D expenditure), a breadth innovation search, and a networking activity, is better performing, in terms of patenting activity, despite the costs of managing multiple innovative inputs. 6. Conclusions The aim of this paper was to provide some empirical evidence on a new model of open RIS (ORIS). The empirical evidence presented focused on a representative sample of life science rms operating in the Emilia Romagna RIS. The innovation system approach describes innovation as a systemic and interactive process that crosses the rms boundaries to include several external sources of knowledge and information

13 The Poisson regression model assumes that the variance of the counts is equal to the mean, which appears not to hold in our situation, where we witness an overdispersion phenomenon. The negative binomial succeeds in accommodating this problem. The negative binomial model is an extension of the standard Poisson model where the Poisson parameter for each rm has an additional random component, accounting for (unobserved) heterogeneity, not yet accounted for by the regressors that determine the individual mean function. 14 Given the coefcient of 0.145 for the R&D investment, a doubling of the R&D investment would multiply the rate by exp (0.145 log(2)), which is 1.11.

15 Given the coefcient of 0.337 for the number of external relationships, a 10% increase in the number of external relationships would multiply the rate by exp (0.337), which is 1.40. 16 Given the coefcient of 0.245 for the R&D investment, a doubling of the R&D investment would multiply the rate by exp (0.245 log(2), which is 1.18.

Please cite this article in press as: Belussi, F., et al., Learning at the boundaries in an Open Regional Innovation System: A focus on rms innovation strategies in the Emilia Romagna life science industry. Res. Policy (2010), doi:10.1016/[Link].2010.01.014

G Model RESPOL-2410; 10 No. of Pages 12

ARTICLE IN PRESS
F. Belussi et al. / Research Policy xxx (2010) xxxxxx

available in both the industry and institutional sub-systems. In order to explore this aspect, our analysis focused on the breadth of rms external search with respect to an ample list of possible external sources of knowledge and information, distinguished in market-based, institutional-based, and semi-public. Our results showed that several rms in our sample do not follow the closed innovation model as they employ various external sources of innovation. The most frequently used are clients and customers and semi-public sources (51.7%) while the most relevant are the institutional ones. In the ORIS model, innovation is not only a systemic and open process which crosses the rms boundaries but also a trans-local phenomenon which overcomes the regional border. In order to explore this aspect in the Emilia Romagna life science RIS, the study presented a detailed analysis of the spatial distribution of R&D collaborations. Our results showed that rms in our sample establish the largest part of their research collaborations with actors localised outside the region (54%). These ties link local rms with national, European and extra-European circuits of knowledge creation and diffusion. Interestingly, the propensity of establishing research collaborations outside the region is not conned to a small minority of leading enterprises. This evidence is particularly relevant because it testies a shift towards the ORIS model, where rms innovation search strategy, despite being still embedded in local nets (involving several regional PRO), is open to external-to-the-region research networks and knowledge sources. The analysis of R&D collaborations also offered some interesting indications on the form of distributed innovation in the Emilia Romagna life science ORIS. We indeed found that rms in the Emilia Romagna life sciences are more likely to establish research collaborations with PROs than with other rms, suggesting a different form of inter-organisational division of innovative tasks compared to the typical biotech model. This nding is consistent with the study conducted by Valentin and Jensen (2003) on R&D collaborations in food biotechnology, which showed a prominent role of PROs compared to other rms specialised in biotech research. Considering that rms in our sample belongs to several sub-sectors of the life science industry, our ndings corroborates the argument made by Valentin and Jensen (2003): it is misleading to expect that all elds in the life sciences or in high-tech sectors in whatever country should emulate or reect the typical biotech model of distributed innovation. In this respect, our study provides new empirical evidence useful to challenge some taken-for-granted assumptions about the life science industry. Finally, our study tested the effectiveness of the open innovation model. The open innovation model better explains rms innovative performance compared to the closed one, which is based on the sole use of internal R&D. We also found that research collaborations contribute signicantly to the determination of rms innovative output. In our sample, internal R&D investment is signicant, but the use of research collaborations can compensate low levels of internal investment. This result indicates a substitution effect between these two sources of innovation. A similar effect was found in the study of Laursen and Salter (2006), concerning the interaction between the openness to external search activities and the internal R&D investments. The authors explained this substitution effect as a manifestation of the Not Invented Here syndrome. One likely interpretation of the substitution effect is related to the costs of simultaneous network building and internal R&D investment. This paper has some limitations that should be acknowledged. The rst one is connected to the choice of the specic empirical setting. To what extent the Emilia Romagna life science ORIS can be elected as representative of a new model?

Further empirical evidence from different contexts (in other countries or industries) is necessary to fully understand the implications of the open innovation approach for RIS evolution. While the present study examined RIS openness in one period, another future research challenge is to investigate the evolution towards a model of open RIS using a longitudinal research design. Still, our methodological choice allowed gathering negrained information on the role played by research collaborations in rms external search processes, strengthening the importance of the learning at the boundaries mechanism in a new ORIS model.

Acknowledgements We gratefully acknowledge nancial support from the Programma di Azioni Innovative of the Regione Emilia Romagna and ICaTSEM (Institutional Changes and Trajectories of SocioEconomic Development Models), EU-7th Framework Programme. Thanks are due to Silvano Bertini of the Assessorato alle Attivit Produttive Sviluppo Economico and Maria Grazia Zucchini of ASTER (Associazione Scienza e Tecnologia Emilia Romagna), who provided invaluable assistance. Any remaining errors are our own.

Appendix A. Description of the business model of the sampled enterprises We describe here in greater detail some characteristics of the sampled rms working in the life science eld. 50 interviewed enterprises are exclusively productive, 23 are mainly commercial (but they have developed on order large quantities of specic products) and 5 have a fty/fty mixed activity, as we can see in the following table.

Classication of the Emilia Romagna life science sector rms based on their main product. Manufacturing enterprises in the life science eld (production and commercialisation) Diagnostic Bio-image Clinical diagnosis Functional evaluation Therapy and rehabilitation Machinery for dialysis and respiration Articial organs Rehabilitation and support Surgical therapy Orthopaedics and prostheses Other Non-durable materials Dental materials Firms hospital materials Other equipment Aesthetic and stimulators Dental equipment Hospital equipment Electro-medical equipment Various machinery Pharmaceutical and biotech rms Pharmaceutical enterprises Biotech enterprises Computer science enterprises applied to telemedicine Source: Our elaborations on 78 interviews. Firms in the sample

5 5 11 4 1 1 1 2 2 28 1 27 21 3 3 12 3 3 2 1 10

Please cite this article in press as: Belussi, F., et al., Learning at the boundaries in an Open Regional Innovation System: A focus on rms innovation strategies in the Emilia Romagna life science industry. Res. Policy (2010), doi:10.1016/[Link].2010.01.014

G Model RESPOL-2410; No. of Pages 12

ARTICLE IN PRESS
F. Belussi et al. / Research Policy xxx (2010) xxxxxx 11 Cohen, W.M., Levinthal, D.A., 1989. Innovation and learning: the two faces of R&D. Economic Journal 99, 569596. Cohen, W.M., Levinthal, D.A., 1990. Absorptive capacity: a new perspective on learning and innovation. Administrative Science Quarterly 35, 128152. Cooke, P., 2002. Knowledge Economies. Clusters, Learning and Cooperative Advantage. Routledge, London. Cooke, P., 2005. The asymmetric knowledge problem, regional innovation and geographical knowledge systems industry organisation. Paper Presented at the Druid Conference Dynamics of Industry and Innovation: Organisations, Networks and Systems, 27th29th June. Cooke, P., Uranga, M., Etxebarria, G., 1997. Regional innovation systems: institutional and organisational dimensions. Research Policy 26, 475491. Cooke, P., 2001. New economy innovation systems: biotechnology in Europe and the USA. Industry and Innovation 8, 267289. Cooke, P., 1998. Global clustering and regional innovation. Systemic integration in Wales. In: Braczyk, H.-J., Cooke, P., Heidenreich, M. (Eds.), Regional Innovation Systems. UCL Press, London, pp. 245262. Coombs, R., Metcalfe, J.S., 2000. Organising for innovation: co-ordinating distributed innovation capabilities. In: Foss, N.J., Mahnke, V. (Eds.), Competence, Governance and Entrepreneurship: Advances in Economic Strategy Research. Oxford University Press, Oxford, pp. 209231. Coombs, R., Metcalfe, J.S., 2002. Innovation in pharmaceuticals: perspectives on the co-ordination, combination and creation of capabilities. Technology Analysis & Strategic Management 14, 261271. Crpon, B., Duguet, E., 1997. Research and development, competition and innovation. Pseudo maximum likelihood and stimulated maximum likelihood methods applied to count data models with heterogeneity. Journal of Econometrics 79, 355378. Dahlander, L., McKelvey, M., 2005. The occurrence and spatial distribution of collaboration: biotech rms in Gothenburg, Sweden. Technology Analysis & Strategic Management 17, 409431. Dasgupta, P., David, P.A., 1987. Information disclosure and the economics of science and technology. In: Feiwel, G. (Ed.), Arrow and the Ascent of Modern Economic Theory. New York University Press, New York, pp. 519542. De la Mothe, J., Paquet, G. (Eds.), 1998. Local and Regional Systems of Innovation. Kluwer Academic Publishers, Amsterdam. Doloreux, D., 2002. What we should know about regional systems of innovation. Technology in Society 24, 243263. Edquist, C., 2005. Systems of innovation. perspectives and challenges. In: Fagerberg, J., Mowery, D., Nelson, R. (Eds.), The Oxford Handbook of Innovation. Oxford University Press, Oxford, pp. 181208. Etzkowitz, H., 1998. The norms of entrepreneurial science: cognitive effects of the new universityindustry linkages. Research Policy 27, 823833. Etzkowitz, H., Leydesdorff, L., 1997. Universities in the global economy: a triple helix of universityindustrygovernment relations. Cassell Academic, London. Foray, D., Hargreaves, D., 2002. The development of knowledge in different sectors: a model and some hypotheses. Paper Presented at the Druid Conference, Industrial Dynamics of the New and Old Economy: Who is Embracing Whom?. Copenhagen, 68 June. Freeman, C., 1987. Technology Policy and Economic Performance: Lessons from Japan. Pinter, London. Gertler, M.S., Levitte, Y.M., 2005. Local nodes in global networks: the geography of knowledge ows in biotechnology innovation. Industry and Innovation 12, 487507. Giuliani, E., 2007. The selective nature of knowledge network in clusters: evidence from the wine industry. Journal of Economic Geography 7, 139168. Green, S.G., 1991. How many subjects does it take to do a regression analysis? Multivariate Behavioral Research 26, 499510. Hall, B., Griliches, Z., Hausman, J., 1986. Patents and R&D: is there a lag? International Economic Review 27, 265283. Hausman, J., Hall, B., Griliches, Z., 1984. Econometric models for count data with an application to the patentsR&D relationships. Econometrica 52, 909 938. Howells, J., 1999. Regional systems of innovation? In: Archibugi, D., Howells, J., Michie, J. (Eds.), Innovation Policy in a Global Economy. Cambridge University Press, Cambridge, pp. 6793. Harris, R.J., 1975. A Primer of Multivariate Statistics. Academic, New York. Henderson, R.M., Clark, K.B., 1990. Architectural innovation: the reconguration of existing product technologies and the failure of established rms. Administrative Science Quarterly 35 (1), 930. Hobday, M., Davies, A., Prencipe, A., 2005. Systems integration: a core capability of the modern corporation. Industrial and Corporate Change 14 (6), 1109 1143. Iammarino, S., 2005. An evolutionary integrated view of regional systems of innovation. Concepts, measures, and historical perspectives. European Planning Studies 13 (4), 495517. Jaffe, A., Trajtenberg, M., Henderson, R., 1993. Geographic localisation of knowledge spillovers as evidenced by patent citations. The Quarterly Journal of Economics 108, 577598. Kline, S.J., Rosenberg, N., 1986. An overview of innovation. In: Landau, R., Rosenberg, N. (Eds.), The Positive Sum Strategy. National Academy Press, Washington, pp. 275305. Kogut, B., Zander, U., 1992. Knowledge of the rm, combinative capabilities, and the replication of technology. Organization Science 3, 383397. Laursen, K., Salter, A., 2004. Searching low and high: what types of rms use universities as a source of innovation? Research Policy 33, 12011215.

References
Ahuja, G., 2000. Collaboration networks, structural holes, and innovation: a longitudinal study. Administrative Science Quarterly 45, 425455. Antonelli, C., 1999. The Microdynamics of Technological Change. Routledge, London. Arora, A., Gambardella, A., 1994. The changing technology of technical change: general and abstract knowledge and the division of innovative labour. Research Policy 23, 523532. Arrow, K., 1994. The production and distribution of knowledge. In: Silverberg, G., Soete, L. (Eds.), The Economics of Growth and Technical Change: Technologies, Nations, Agents. Edward Elgar, Aldershot, pp. 919. Asheim, B.T., Isaksen, A., 2002. Regional innovation systems: the integration of local sticky and global ubiquitous knowledge. Journal of Technology Transfer 27, 7786. Asheim, B.T., Coenen, L., 2006. The role of regional innovation systems in a globalizing economy: comparing knowledge bases and institutional frameworks of Nordic clusters. In: Vertova, G. (Ed.), The Changing Economic Geography of Globalization. Routledge, New York, pp. 148165. Asheim, B.T., Isaksen, A., Nauwelaers, C., Todtling, F. (Eds.), 2003. Regional Innovation Policy for SmallMedium Enterprises. Edward Elgar, Cheltenham. Audretsch, D.B., Bozeman, B., Combs, K.L., Feldman, M., Link, A.N., Siegel, D.S., Stephan, P., Tassey, G., Wessner, C., 2002. The economics of science and technology. Journal of Technology Transfer 27, 155203. Autio, E., 1998. Evaluation of RTD in regional systems of innovation. European Planning Studies 6, 131140. Baba, Y., Shichijo, N., Sedita, S.R., 2009. How do collaborations with universities affect rms innovative performance? The role of Pasteur Scientists in the advanced materials eld. Research Policy 38 (5), 756764. Bathelt, H., Malmberg, A., Maskell, P., 2004. Clusters and knowledge: local buzz, global pipelines and the process of knowledge creation. Progress in Human Geography 28, 3156. Baum, J.A., Calabrese, T., Silverman, B.S., 2000. Dont go it alone: alliance network composition and startups performance in Canadian biotechnology. Strategic Management Journal 21, 267294. Belussi, F., 2002. The Italian system of innovation: the gradual transition from a weak mission-oriented system to a regionalised learning system. In: Borras, S., Biegelbauer, P. (Eds.), Innovation Policies in Europe and the US: the New Agenda. Ashgate, Aldershot, pp. 233252. Belussi, F. (Ed.), 2005. Il settore delle Life Science in Emilia Romagna. Edizione Regione Emilia Romagna, Bologna. Belussi, F., Gottardi, G., 2000. Model of localised technological change. In: Belussi, F., Gottardi, G. (Eds.), Evolutionary Patterns of Local Industrial Systems. Ashgate, Aldershot, pp. 115129. Belussi, F., Sedita, S.R., 2009. Life cycle vs. multiple path dependency in industrial districts. European Planning Studies 17 (4), 505528. Belussi, F., Sedita, S.R., Pilotti, L., 2008. Learning at the boundaries for industrial districts between exploitation of local resources and exploration of global knowledge ows. In: Leoncini, R., Montresor, S. (Eds.), Dynamic Capabilities between Firm Organization and Local Systems of Production. Routledge, London, pp. 181215. Boari, C., Lipparini, C., 1999. Networks within industrial districts: organising knowledge creation and transfer by means of moderate hierarchies. Journal of Management Governance 3, 339360. Borgatti, S.P., Everett, M.G., Freeman, L.C., 2002. Ucinet 6 for Windows. Analytic Technologies, Harvard. Boschma, R.A., ter Wal, A.L.J., 2007. Knowledge networks and innovative performance in an industrial district: the case of a footwear district in the South of Italy. Industry and Innovation 14, 177199. Bound, J., Cummins, C., Griliches, Z., Hall, B., Jaffe, A., 1982. Who does R&D and who patents? NBER Working Paper Series. Braczyk, H.J., Cooke, P., Heidenreich, M. (Eds.), 1998. Regional Innovation Systems: the Role of Governance in a Globalized World. UCL Press, London. Brusoni, S., Prencipe, A., 2001. Unpacking the black box of modularity: technologies, products and organizations. Industrial and Corporate Change 10, 179205. Camagni, R., 1991. Local milieu, uncertainty and innovation networks: towards a new dynamic theory of economic space. In: Camagni, R. (Ed.), Innovation Networks: Spatial Perspectives. Belhaven Press, London, pp. 121144. Cantwell, J., Iammarino, S., 2003. Multinationals Corporations and European Regional Systems of Innovation. Routledge, London. Carlsson, B., 2006. Internationalisation of innovation systems: a survey of the literature. Research Policy 35, 5667. Carlsson, B., Stankiewicz, R., 1995. On the nature, function and composition of technological systems. In: Carlsson, B. (Ed.), Technological Systems and Economic Performance: the Case of Factory Automation. Kluwer, Dordrecht, pp. 2156. Carlsson, B., 1994. Technological systems and economic performance. In: Dodgson, M., Rothwell, R. (Eds.), The Handbook of Industrial Innovation. Edward Elgar, Aldershot, pp. 1324. Castells, M., 2000. The Rise of the Network Society, second ed. Blackwell, Cambridge, MA. Chesbrough, H.W., 2003. Open Innovation: the New Imperative for Creating and Proting from Technology. Harvard Business School Press, Boston. Chiesa, V., 1995. Globalising R&D around centres of excellence. Long Range Planning 28 (6), 1928. Coenen, L., Moodysson, J., Asheim, B.T., 2004. Nodes, networks and proximities: on the knowledge dynamics of the Medicon Valley biotech cluster. European Planning Studies 12, 10031018.

Please cite this article in press as: Belussi, F., et al., Learning at the boundaries in an Open Regional Innovation System: A focus on rms innovation strategies in the Emilia Romagna life science industry. Res. Policy (2010), doi:10.1016/[Link].2010.01.014

G Model RESPOL-2410; 12 No. of Pages 12

ARTICLE IN PRESS
F. Belussi et al. / Research Policy xxx (2010) xxxxxx Owen-Smith, J., Powell, W.W., 2004. Knowledge networks as channels and conduits: the effects of spillovers in the Boston biotechnology community. Organization Science 15, 521. Penner-Hahn, J., Shaver, J.M., 2005. Does international research and development increase patent output? An analysis of Japanese pharmaceutical rms. Strategic Management Journal 26, 121140. Penrose, E., 1959. The Theory of the Growth of the Firm. Blackwell, Oxford. Pisano, G.P., Shan, W., Teece, D.J., 1988. Joint ventures and collaboration in the biotechnology industry. In: Mowery, D.C. (Ed.), International Collaborative Ventures in US Manufacturing. Ballinger, Cambridge, MA, pp. 183222. Porter, M., 1998. On Competition. Harvard Business School Press, Boston. Powell, W.W., Koput, K.W., Smith-Doerr, L., Owen-Smith, J., 1999. Network position and rm performance: organizational returns to collaboration in the biotechnology industry. Research in the Sociology of Organizations 16, 129159. Richardson, G., 1972. The organisation of industry. Economic Journal 82, 883896. Robertson, P., Smith, K., 2008. Distributed knowledge bases in low-and mediumtechnology industries. In: Hirsh-Kreinsen, H., Jacobson, D. (Eds.), Innovation in Low-Tech Firms and Industries. Elgar, Cheltenham. Saxenian, A., 1994. Regional Advantage: Culture and Competition in Silicon Valley and Route 128. Harvard University Press, Cambridge, MA. Storper, M., 1997. The Regional World. The Guilford Press, New York. Stuart, T.E., Podolny, J.M., 1999. Positional consequences of strategic alliances in the semiconductor industry. Research in the Sociology of Organizations 16, 161182. Tait, J., Chataway, J., Jones, S., 1990. The status of biotechnology-based innovations. Technology Analysis and Strategic Management 2, 293305. Teece, D., Pisano, G., Shuen, A., 1997. Dynamic capabilities and strategic management. Strategic Management Journal 18, 509533. Tdtling, F., Trippl, M., 2005. One size ts all? Towards a differentiated regional innovation policy approach. Research Policy 34, 12031219. Valentin, F., Jensen, R.L., 2003. Discontinuities and distributed innovation: the case of biotechnology in food processing. Industry and Innovation 10 (3), 275 310. Wernerfelt, B., 1984. The resource-based view of the rm 10 years after. Strategic Management Journal 16, 171174. West, J., Gallagher, S., 2006. Patterns of open innovation in open source software. In: Chesbrough, H., Vanhaverbeke, W., West, J. (Eds.), Open Innovation: Researching a New Paradigm. Oxford University Press, Oxford, pp. 82106. Winter, S., 1987. Knowledge and competence as strategic assets. In: Teece, D. (Ed.), The Competitive Challenge. Ballinger, Cambridge, MA. Zander, I., 1999. Where to the multinational? Evolution of technological capabilities in the multinationals networks. International business Review 8, 261291. Zimmerman, K.F., Schwalbach, J., 1991. Determinanten der Patentaktivitt. IfoStudien 37, 201227.

Laursen, K., Salter, A., 2006. Open for innovation: the role of openness in explaining innovation performance among UK manufacturing rms. Strategic Management Journal 27, 131150. Levin, R., Klevorick, A.K., Nelson, R.R., Winter, S.G., 1987. Appropriating the returns from industrial R&D. Brookings Papers on Economic Activity 17, 783820. Liao, T.F., 1994. Interpreting Probability Models: Logit, Probit, and Other Generalized Linear Models, Sage University Paper Series on Quantitative Applications in the Social Sciences, 07-101. Sage, Newbury Park, CA. Licht, G., Zoz, K., 1998. Patents and R&D, an econometric investigation using applications for German, European and US patents by German companies. Annales dEconomie et de Statistique 4950, 329360. Lipparini, A., Sobrero, M., 1994. The glue and the pieces: entrepreneurship and innovation in small-rm networks. Journal of Business Venturing 2, 125140. Longhi, C., 2002. From exogenous to endogeneous local development: the cases of Toulouse and Sophia Antinopolis technopoles. Paper Presented at the meeting Clusters in High-Tech Sectors. Erasmus University, Rotterdam, 34 May. Lorenzoni, G., Lipparini, A., 1999. The leveraging of interrm relationships as a distinctive organizational capability: a longitudinal study. Strategic Management Journal 20, 317338. Lundvall, B.-. (Ed.), 1992. National Systems of Innovation: Towards a Theory of Innovation and Interactive Learning. Pinter, London. Malerba, F., 2002. Sectoral systems of innovation and production. Research Policy 31, 247265. Malerba, F. (Ed.), 2004. Sectoral Systems of Innovation. Cambridge University Press, Cambridge. Malmberg, A., Maskell, P., 2005. Localized learning revisited. DRUID Working Paper No. 05-19. Maskell, P., 2001. Towards a knowledge-based theory of the geographical cluster. Industrial and Corporate Change 10, 919941. Maskell, P., Malmberg, A., 1999. Localised learning and industrial competitiveness. Cambridge Journal of Economics 23, 167185. McKelvey, M., Alm, H., Riccaboni, M., 2003. Does co-location matter for formal knowledge collaboration in the Swedish biotechnologypharmaceutical sector? Research Policy 32, 483501. Metcalfe, J.S., 2005. The evolution of industrial dynamics. Paper Presented at DRUID Tenth Anniversary Summer Conference 2005, Copenhagen, DK, June 2729. Nelson, R., 1992. What is commercial and what is public about technology, and what should be? In: Rosenberg, N., Landu, R., Mowery, D. (Eds.), Technology and the Wealth of the Nations. Stanford University Press, Stanford. Nelson, R. (Ed.), 1993. National Innovation Systems. Oxford University Press, Oxford. Nelson, R., Levin, R., 1986. The inuence of science, university, research and technical societies on industrial R&D and technical advance. Policy Discussion Paper Series n. 3, Research Program on Technological Change, Yale University. Niosi, J., 2003. Alliances are not enough explaining rapid growth in biotechnology rms. Research Policy 32, 737750.

Please cite this article in press as: Belussi, F., et al., Learning at the boundaries in an Open Regional Innovation System: A focus on rms innovation strategies in the Emilia Romagna life science industry. Res. Policy (2010), doi:10.1016/[Link].2010.01.014