The direction of firm innovation: The contrasting roles of strategic alliances and individual scientific collaborations

Research Policy 44 (2015) 1473–1487

Contents lists available at ScienceDirect

Research Policy journal homepage: www.elsevier.com/locate/respol

The direction of firm innovation: The contrasting roles of strategic alliances and individual scientific collaborations Jan Hohberger a,∗ , Paul Almeida b,1 , Pedro Parada c,2 a b c

University of Technology, Sydney P.O. Box 123 Broadway, NSW 2007, Australia McDonough School of Business, Georgetown University, Washington, DC, USA ESADE Business School, Av. Pedralbes 60-62, 08034, Barcelona, Spain

a r t i c l e

i n f o

Article history: Received 28 March 2014 Received in revised form 27 April 2015 Accepted 27 April 2015 Keywords: Alliances Research collaboration University–industry interactions Innovation Knowledge search

a b s t r a c t In dynamic and uncertain technological environments, the focus of industry innovative activity changes over time and the position of each firm with respect to the industry’s innovative focus changes as well. Drawing upon insights from evolutionary economics, we derive hypothesis on the role of R&D alliances and individual scientific collaborations in influencing a firm’s innovative direction and its position relative to the industry’s innovation focus. The analyses of patent and alliance data show that biotechnology firms that rely on external individual scientific collaborations are likely to grow closer to the future focus of innovation, while firms that emphasize R&D alliances grow more distant from the future industry focus. Thus, the use of collaborative mechanisms influences the position of firms in innovative space over time. Additionally, the effect of collaborative mechanisms on the direction of innovation is influenced by the technological specialization of the firm. © 2015 Elsevier B.V. All rights reserved.

1. Introduction In science and technology driven industries, the direction of innovation is often unpredictable. Firms competing in knowledge based industries face the complex challenge of identifying and recognizing the ever changing set of problems and solutions that may be relevant to their own technological and scientific strategies and, when necessary, building innovative capabilities and expertise along the emerging innovative areas in the field (Deeds et al., 2000; Grant and Baden-Fuller, 2004; Powell et al., 1996; Wilden and Gudergan, 2014; Zander and Kogut, 1995). Building innovative capabilities in new areas is not easy. Prior research has shown that firms tend to search for knowledge locally-in the neighborhood of their past practices and current capabilities and expertise (Benner and Tushman, 2003; Rosenkopf and Almeida, 2003; Stuart and Podolny, 1996). The path dependent nature of technology development and innovation makes adjusting the direction of innovation particularly challenging when critical knowledge inputs needed for this process lie in numerous and uncertain locations outside the

∗ Corresponding author. Tel.: +61 2 9514 3522. E-mail addresses: [email protected] (J. Hohberger), [email protected] (P. Almeida), [email protected] (P. Parada). 1 Tel.: +1 202 687 3822. 2 Tel.: +34 932 806 162. http://dx.doi.org/10.1016/j.respol.2015.04.009 0048-7333/© 2015 Elsevier B.V. All rights reserved.

firm. Hence, keeping close to the forefront of science and technology is a daunting challenge for most firms. These firms may not fully understand where to go (in innovative space) and, even if they do, given their internal inflexibility, they may find it difficult to get there. Prior research has highlighted the important role played by search processes beyond the borders of the firm in sourcing knowledge (Cassiman and Veugelers, 2006; Laursen and Salter, 2006). Numerous papers have highlighted the role of R&D alliances in enabling firms to access external knowledge and this research often relates these alliances to positive innovation outcomes (Hagedoorn and Duysters, 2002; Rosenkopf and Almeida, 2003; Rothaermel and Deeds, 2004; Srivastava and Gnyawali, 2011; Stuart and Podolny, 1996). In addition, a number of studies have pointed to the importance of external individual scientific level collaborations to knowledge access and innovation and have suggested that while R&D alliances and external individual scientific level collaborations are both collaborative mechanisms, they also have distinctive characteristics (Almeida et al., 2011; Cockburn and Henderson, 1998; Fabrizio, 2009). Thus, we examine and compare the role of these two mechanisms of external knowledge search and analyze their influence on the direction of future innovation. We define the focus of innovation in the industry as the set of technological areas along which a plurality of innovations is produced by the firms in the industry at a point of time. We see this focus of innovation as changing as firms in the industry innovate in different sets of

1474

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487

technological areas across time. We suggest that both R&D alliances and scientific collaborations of individuals across firms can foster knowledge exchange and innovation. However, individual collaborations are particularly suitable for knowledge exchange associated with scientific activities and often result in the publication of scientific papers co-authored by researchers belonging to different organizations (Carnabuci and Operti, 2013; Paruchuri, 2010; Singh, 2005). Since individual collaborations appear to often focus on scientific activities that benefit the circulation of new or emerging knowledge, we argue that external individual scientific collaborations are more likely to enable firms to innovate at the forefront of science and become aligned with the emerging innovative focus of the wider industry. (Almeida et al., 2011; Cockburn and Henderson, 1998; Jiang et al., 2011; Liebeskind et al., 1996; Narin et al., 1997; Zucker et al., 2002), On the other hand, R&D alliances are not only less frequently focused on scientific knowledge and they are often the result of firm-level management decisions oriented to the application of knowledge and therefore limited in scope and reach (Doz, 1996). Prior research in the area of biotechnology supports this idea – while interpersonal networks of inventors are seen as important mechanisms through which exchanges of new scientific knowledge takes place, alliances were most often used to apply and commercialize the knowledge developed (Liebeskind et al., 1996; Liebeskind et al., 1996Liebeskind et al., 1996; Oliver and Liebeskind, 1998; Zucker et al., 1996). Though R&D alliances also act as conduits of learning, therefore, they may not be the best mechanisms for the development and circulation of newly emerging knowledge and firms engaging in these collaborations may find themselves more distant from the emerging innovative focus of the field. Additionally, as the scientific nature of knowledge is an important factor influencing innovation outcomes in high technology industries and in explaining the differences between individual collaboration and R&D alliances (Almeida et al., 2011; Cockburn and Henderson, 1998; Fabrizio, 2009; Hess and Rothaermel, 2014; Jiang et al., 2011), we explore the influence of scientific orientation of both collaborative mechanisms. Understanding both what leads firms to innovate close to the focus of innovation in an industry and what leads firms to deviate from that focus is important because, both innovating close to the focus and distant from the focus can be attractive. Competition often takes place within particular technological domains or sub-fields (Dosi and Nelson, 1994; Stuart and Podolny, 1996). The most rapidly evolving and potentially attractive technological areas, while presenting opportunities for innovation with the associated economic benefits, could also attract attention from other players, making the economic and technological gains of competing in these spaces questionable (Dosi and Nelson, 1994). On the other hand, competing in technological spaces that appear to present less innovative opportunity, may attract fewer other players and hence present greater rewards (Dosi and Nelson, 1994). Given the high costs of technological development, and the vast and continuously emerging array of technological areas along which firms innovate, it is important to understand the influences on the direction of innovation. While prior literature has investigated the influence of collaborative mechanisms on innovative performance (Almeida et al., 2011; Rothaermel and Hess, 2007; Srivastava and Gnyawali, 2011) and the direction of firm innovation or search activities (Jiang et al., 2011; Lavie et al., 2011; Phelps, 2010; Rosenkopf and Almeida, 2003), this paper makes a novel contribution as it focuses on the mechanisms that facilitate knowledge exchange to explain the movement of the firm relative to the industry in the evolving innovative space. The firm’s innovative focus may, or may not, be distant from past practice and exploratory search may not necessarily lead the firm toward the industry’s focus of innovation. Depending on the direction of the movement of the field at any given

time, the firm’s innovative capabilities may involve varying levels of exploitation and exploration. Our study finds that firms with increasing numbers of external individual scientific collaborations are likely to become more closely aligned to the emerging focus of innovation in biotechnology, while firms with increasing number of R&D alliances are likely to become more distant from the innovative focus of the field. We also find that technological specialization negatively affects the match with the future innovation focus and it reinforces the effect of R&D alliances and internal publications on moving away from the innovation focus of the field while it reduces the effect of external individual collaborations. Thus our paper builds on the existing literature on the mechanisms that aid the external search for knowledge by: (a) highlighting the contrasting roles those mechanisms play, (b) exploring the outcomes of the search processes, (c) acknowledging the dynamics of the innovative field and, (d) looking at the direction of the changes in innovative expertise. Firms may find it useful to understand how their position in future innovative space is affected by the extent of their use of different types of collaborative mechanisms.

2. Theory and hypothesis The behavioral theory of the firm (Cyert and March, 1963) suggests that individuals are boundedly rational. In the face of uncertainty and complexity, individuals do not rationally evaluate the complete range of choices before them. They are, instead, strongly influenced by current practices when making decisions about future actions. Individuals select actions that tend to be in the neighborhood of current practice rather than those that may be the most attractive in terms of future success. Evolutionary theorists, like Nelson and Winter (1982), make a similar point when they suggest that organizations, like individuals, are bounded in their decision processes. Using these insights to explain the evolution of organizations, they suggest that firms are path dependent – actions (including technology development and innovation) tend to be along well established and familiar paths. They ascribe this to the formation of routines within the organization. These routines favor local search processes that make it difficult for the firm to adapt to any changes that depart from past practices and trajectories (Nelson and Winter, 1982). In complex and dynamic environments, local search routines may fail to identify the best solution to a problem (Fleming and Sorenson, 2004). Similarly, Leonard-Barton (1992) suggests that in dynamic environments, existing capabilities may become core rigidities that prevent firms from changing and adjusting to external needs. Levinthal and March (1993) argue that prior experience could be a ‘poor teacher’ – often leading to myopia of learning and an inability to incorporate new knowledge to address changes in the external environment. Since innovation involves a recombination of existing knowledge, access to a diverse set of knowledge sources is crucial for innovative success (Fleming and Sorenson, 2004; Henderson and Clark, 1990; Nelson and Winter, 1982). Yet we know that firms have the tendency to recombine familiar knowledge (Levinthal and March, 1993; March, 1991) making it more likely that they will perform incremental innovations close to the existing trajectory of firm. Experimenting with novel and diverse knowledge components, on the other hand, allows the firm to develop new and different innovation outcomes (Ahuja and Lampert, 2001; Laursen, 2012). However, firms find it difficult to search for, and utilize knowledge in areas that are distant from their existing areas of expertise and may find it challenging to move in new directions even when these could be related to organizational success. Even in dynamic and rapidly evolving innovative environments, firms often tend to exploit and build existing capabilities and continue

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487

to innovate in areas close to their past knowledge base. Empirical research (in the form of in-depth case studies and large scale quantitative analyses) supports the idea of local search even in the face of significant environmental change (Helfat, 1994; Martin and Mitchell, 1998; Siggelkow, 2001; Stuart and Podolny, 1996). There is evidence that collaborative mechanisms can be used to access and absorb knowledge distant from a firm’s current expertise. For example Rosenkopf and Almeida (2003) show that firms can acquire knowledge from geographically or technologically distant domains through alliances thus helping them overcome the tendency for local search. Srivastava and Gnyawali (2011) demonstrate that the characteristics of partners in alliances can positively influence the creation of breakthrough innovations and thus help them overcome competency traps. Rothaermel (2001) finds that in the face of technological change, incumbents use alliances to access knowledge related to new product development. Jiang et al. (2011) shows that a firm’s innovation output increases by exploring novel technologies via technologically distant collaboration partners and particularly scientific knowledge. This also helps firms gain a better understanding of emerging industries. Hence, prior research suggests that external knowledge sources may be useful in helping organizations adjust their technological trajectories. This paper builds on this research by focusing on the roles of different collaboration mechanisms (that facilitate knowledge access) and showing how these mechanisms relate to the movement of the firm in innovative space relative to the innovation activities of the broader industry. 2.1. External individual scientific collaborations Multiple studies have shown the importance of science based knowledge to the development of industries and to firm innovative performance (Narin et al., 1997; Zucker et al., 2002) and how research collaborations by scientists can serve as mechanisms that facilitate the flow of knowledge (Almeida et al., 2011; Carnabuci and Operti, 2013; Cockburn and Henderson, 1998; Fabrizio, 2009; Paruchuri, 2010; Singh, 2005). Scientists can be seen to belong to both organizational and scientific communities, and hence have the opportunity to facilitate the flow of knowledge between these communities (Murray, 2002). Since many scientific interactions focus on emerging technologies, approaches, and ideas, the flow of knowledge within the community influences which future innovations will develop in the field. Membership in this community therefore allows for an understanding of, and movement toward, both existing and future innovative areas. This community of scientists shares not just scientific knowledge but also knowledge about the application and commercialization of innovation through involvement in patenting, consulting, advisory boards and entrepreneurship (Crane, 1972; Murray, 2002; Saxenian, 1991). This sharing of knowledge leads to a development of a common view of areas that are rich with innovative opportunity and the exposure to shared knowledge permits individuals and firms to innovate. Membership in scientific communities often gives rise to research collaborations of scientists (often informal) across organizational boundaries and results in the publication of co-authored scientific papers (Almeida et al., 2011; Cockburn and Henderson, 1998; Murray, 2002). Individual-level scientific collaborations do not just lead to an increase in knowledge available to a firm but, importantly, also facilitate insights and access to the knowledge from a different spectrum than may otherwise be possible (Almeida et al., 2011; Cockburn and Henderson, 1998). Firms that have a large number of scientists engaged in external knowledge exchange are likely to be infused with a broad set of new ideas, decreasing the myopia that is so often a part of the learning process and consequently decreasing the resistance to change. Collaborative activity with those outside

1475

the organization promotes the development of new capabilities at the individual level, and cumulatively at the organizational level, along directions of emerging interest. Scientific knowledge inputs and exchanges provide individuals with an early and clearer picture of the emerging scientific and technological landscape and this expands the view of scientific and technological possibilities available to the firm (Fleming and Sorenson, 2004). Fleming and Sorenson (2004) point out that scientific investigation allows one to develop better maps of the innovative landscape of (scientific and technological) possibilities being faced by an individual or organization. The scientific knowledge obtained is often broader in it application, so it can be applied in wider range of technological areas and can help a firm move within the associated technological space. In terms of collaborating with scientists in academic institutions, Jiang et al. (2011) show that such collaborations have less organizational constrains and are less threatened by the emergence of new technologies than for-profit organizations. Scientific academic research also provides scientists relatively more freedom to select and change their research directions. Engagement in scientific collaboration allows firms to better understand the impact of the new developments through the exchange of tacit knowledge. This is particularly important during technological changes, as emerging knowledge is tacit in nature and its acquisition requires individual interactions (Hohberger, 2014; Zucker et al., 2002). Additionally the relatively informal nature of individual collaborations makes their use fast and flexible and scientists can engage in them more actively than in classical R&D alliances (Almeida et al., 2011). In sum, a firm whose employees engage in an increasing number of external individual scientific collaborations can be seen to gain wider and more detailed views of possible avenues for research, clearer ideas of the relative merits of alternative search processes within the organization, and greater access to stocks of emerging insights, techniques, and knowledge that could enable it to pursue attractive options. This firm will therefore grow better equipped to overcome some of the limitations of local search, and also grow better able to innovate in newly emerging and developing scientific and technological areas. Hence, the knowledge shared by scientists from different organizations, can push innovation singly and collectively in certain directions that help define the focus of innovation in the field. Hence, Hypothesis 1. With increasing numbers of external scientific individual collaborations, a firm will become more aligned to the emerging focus of innovation in the field. 2.2. R&D alliances The idea that alliances can lead to inter-firm learning is well documented in the strategic management literature (Hamel, 1991; Inkpen, 2002; Powell et al., 1996). Several studies show that alliances can be used to acquire different types of knowledge and thus can influence the direction of firm innovation (Bercovitz and Feldman, 2007; Colombo et al., 2006; Hohberger, 2014; Inkpen, 2002; Rosenkopf and Almeida, 2003; Stuart and Podolny, 1996). In contrast to prior studies, our interest is in examining the role of R&D alliances in facilitating a firm’s direction of innovation toward or away from the broader field rather than comparing it its own previous innovative position. We argue that several characteristics of R&D alliances suggest that they will tend to increase the distance of the organization from the future focus of innovation. Alliances often involve organizational and legal processes that are time consuming, costly to maintain and difficult to implement (Gulati and Singh, 1998). Even once legally established, alliance formation is itself evolutionary (Doz, 1996) involving the gradual establishment of routines and trust across firms before the alliance can be an effective in

1476

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487

knowledge sharing (Zollo et al., 2002). Hence, there is a significant time lag between the decision to target knowledge through the formation of an alliance, and the point at which the alliance is useful in facilitating knowledge flows between firms. Additionally, firms can form only a limited number of R&D alliances (Almeida et al., 2011; Rothaermel and Hess, 2007; Stuart et al., 2007) and the numbers of alliances are significantly less than the numbers of individual level collaborations in an organization (Almeida et al., 2011). R&D alliances are also more likely created with other firms rather than with universities and research institutions (Almeida et al., 2011; Stuart et al., 2007). Through interviews Almeida et al. (2011) showed that for a scientist, R&D alliances are often unattractive as their formal and planned nature makes them relatively slow to implement and costly to establish and maintain. Additionally, research particularly in biotechnology has shown that interpersonal networks of inventors are the primary mechanism through which exchanges of new knowledge and breakthroughs take place, whereas formal inter-organizational ties are used mainly to commercialize the knowledge developed (Liebeskind et al., 1996; Liebeskind et al., 1996Liebeskind et al., 1996; Oliver and Liebeskind, 1998, Zucker et al., 1996). In comparison to external individual collaboration, R&D alliances may not provide scientists with the same opportunity to reassess and recalibrate mental maps. They are more likely to reinforce existing innovation directions along established trajectories rather than facilitate experimentation into emerging fields and approaches. Overall, given that R&D alliances are restricted in their number and breadth, and given that there are significant lags between the formation decision and the availability of knowledge from these collaborations, R&D alliances focus frequently on the application of existing knowledge. They are not useful mechanisms for keeping abreast of emerging ideas, lines of inquiry, methods and innovations in a dynamic field where the sources of new science and technology are diverse and changing. Thus, while R&D alliances enable to access external knowledge and enhance a firm’s innovativeness, and its ability to commercialize these innovations1 , they are more likely to reinforce the trajectories of existing innovations and this will result in the firm growing more distant from the emerging focus of innovation. Hence, Hypothesis 2. With increasing numbers of R&D alliances, a firm will become less aligned with the emerging focus of innovation in the field. 2.3. Internal scientific publication In addition to the use of external collaborative mechanisms like R&D alliances and informal scientific collaborations, the internal collaborative practices of inventors are likely to affect the ease and direction of innovation. Early research on inventors and scientists showed that the knowledge that flows between groups and individuals within the firm is crucial to its innovation activities (Allen and Cohen, 1969; Tushman, 1977). The fact that internal R&D activities are embedded in organizational structure and routines fosters problem solving and mutual consultation across members of the organization and this eases the internal flow of knowledge (Grant, 1996; Kogut and Zander, 1992). Scientists and inventors

1 It is important to note that previous research has shown that alliances can be exploitative and/or explorative activities (Bercovitz and Feldman, 2007; Colombo et al., 2006; Rothaermel and Deeds, 2004). Frequently R&D alliances are associated with exploration as they provide new knowledge or knowledge that differs from the previous knowledge base of the firm (Colombo et al., 2006; Hoang and Rothaermel, 2010 Lavie et al., 2011). In this study, our focus is not on the general innovation benefits of R&D alliances but on the extent to which they are able to align a firm with the moving focus on innovation of the industry.

are, in general, more familiar with the knowledge of their own firm than any other knowledge, and searching for knowledge created within the organization is a habitual or routinized activity (Nerkar and Paruchuri, 2005). As a result, individuals tend to build more on their own firm’s knowledge and are often efficient in doing so (Helfat, 1994; Levinthal and March, 1993; Nerkar and Paruchuri, 2005). Through internal collaborations, individuals develop shared representations and interpretations of the meaning and usefulness of technology and this helps create common strategies on how to use and apply technologies within the organization. Through the familiarity with complex technologies and their application, internal individual collaborations can help the firm explore ways, and develop capabilities, related to the recombination of existing knowledge, thus facilitating innovation (Carnabuci and Operti, 2013). However the benefits of successful internal R&D practices, including internal collaboration, can come at a cost. argue that that successful experiential learning encourages an organization to overly rely on knowledge that is familiar and close to its current experience and expertise. Technologies and knowledge closely related to existing and established firm practices and domains will be highly valued, and this will often be at the expense of more distant, and perhaps useful, knowledge. Levinthal and March (1993) called this the ‘myopia of learning’. Internal collaborations between scientists, in particular, provide an opportunity for greater interaction and discussion between individuals of the same firm. This interaction can lead to the reinforcement of existing mental models and world views that enhance the value placed on what is already known within the firm. A strong sense of identification with the firm, and increased by internal collaboration, can result in stringent cognitive and normative rules and approaches and this can discourage the openness to any knowledge and practice that does not fit easily with current schemas (Day, 1994). Thus, the benefits of internal collaboration in fostering a shared language, communication, and understanding between scientists of a firm, may come at the cost related to the difficulty in recognizing, assimilating and exploiting external knowledge (Carnabuci and Operti, 2013). Thus internal collaborations can increase the ‘inward-looking absorptive capacity’ at the expense of ‘outward-looking absorptive capacity’ (Cohen and Levinthal, 1990). In this context, Carnabuci and Operti (2013) showed individuallevel collaborations of inventors decreases the salience and the utilization of external knowledge and lower the firm’s ability to create new combinations between previously unrelated technologies (Carnabuci and Operti, 2013). Additionally, individual internal collaborations can consume resources and mindshare that could otherwise be oriented to searching for knowledge beyond the boundaries of the firm. A firm that emphasizes internal collaborations may miss signals that the external environment provides about new and attractive areas of innovation and may have lower levels of flexibility and motivation to explore new technology directions. Thus, based on the theoretical arguments from Levinthal and March (1993) and initial empirical results of internal inventor networks (Carnabuci and Operti, 2013), we argue that with higher levels of internal publication activities, a firm will become less aligned with the emerging focus of innovation in the field. Hypothesis 3. With increasing numbers of internal publication, a firm will become less aligned with the emerging focus of innovation in the field. 2.4. Technological specialization A key firm characteristic that affects the innovation process is the extent to which the firm is technologically specialized (or

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487

diversified). Greater technological specialization (or focus on relatively few technologies areas) allows a firm to build capabilities along well defined and narrow trajectories and helps concentrate the application of often scare resources. Investment in a few, related technological areas can enable a firm to benefit from scale economies and move quickly down the experience curve, permitting the development of expertise and technological outputs more quickly than may have otherwise been possible. However, while specialization may enhance a firm’s technological productivity along existing areas of expertise, it may also hamper a firm’s ability to explore new territories or move in new technological directions should the need arise (Kim and Kogut, 1996). One of the important insights arising from Cohen and Levinthal (1990) seminal work on absorptive capacity is that a firm’s investments in R&D influence its ability to recognize, absorb, and utilize external knowledge that could be useful to its research. Technological diversification should help a firm recognize and absorb useful knowledge pertaining to a wider range of fields since it increases the likelihood of having knowledge overlaps with a broader range of organizations. Technological diversification also helps a firms to utilize external knowledge more effectively since it increases the opportunities to make useful knowledge combinations given the breadth of its existing knowledge base (Cohen and Klepper, 1996). Since the extent of technological diversification is not just a function of technical knowledge but also dependent on an organization’s human resources, structure, and processes, it is not surprising that Cantwell (1993) observe that the extent of specialization changes only gradually over time. Thus we argue that a firm’s technological diversity provides greater opportunities to move into emerging technological areas given the greater opportunities for cross-fertilization of knowledge (Granstrand, 1998; Suzuki and Kodama, 2004). On the other hand, higher levels of specialization are likely to lead to greater path dependence and increased difficulty in overcoming local and narrow search even in a dynamic environment. Therefore, Hypothesis 4. With increasing specialization, a firm will become less aligned with the emerging focus of innovation in the field. Our previous arguments on the relationship between specialization and absorptive capacity suggest that the degree to which a firm is specialized will moderate the relationship between the various collaborative mechanisms utilized by the firm and the organization’s ability to move into emerging technological areas. If specialization reduces a firm’s ability to recognize and value external knowledge that is distant from current practice, it is also likely to have a negative effect on the utilization of knowledge absorbed from a wide variety of sources by informal scientific collaborations. The scientists of more specialized firms are likely to collaborate with colleagues from a narrower spectrum of technological areas and the resultant cumulative knowledge brought in by collaboration is also likely to be less diverse (than it would be in a technologically diverse firm). Not only is the knowledge absorbed likely to be less diverse in specialized firms, but firms ability to absorb and leverage external knowledge, is also likely to be limited given its narrow knowledge base. We had argued earlier that R&D alliances and internal collaborations are a product of path dependent and often backward looking decisions within the firm. These modes of collaboration further enhance the rigid focus of firms along fixed technological trajectories and reduce the chances of the firm moving in new technological areas as opportunities emerge in the field. An increase in technological specialization Hypothesis 5a. reduces the positive relationship between the increase of external individual collaborations and the alignment with the emerging focus of innovation in the field.

1477

Hypothesis 5b. An increase in technological specialization increases the negative relationship between the increase in R&D alliances and the alignment with the emerging focus of innovation in the field. An increase in technological specialization Hypothesis 5c. increases the negative relationship between the increase in internal individual collaborations and the alignment with the emerging focus of innovation in the field. 3. Methods We test our hypothesis in the context of the biotechnology industry. This industry is characterized by constant advances in science and technology. The development and commercialization of new drugs requires a range of skills and capabilities including basic and applied research, clinical testing, production and manufacturing, marketing, distribution, and management of the regulatory process (Powell et al., 1996). The complexity of the innovation process makes it a very costly process with long time horizons (Powell et al., 1996). Since no organization can independently assemble all the resources and capabilities needed to innovate and compete successfully across a wide range of continuously emerging scientific and technological fields, firms often engage in a variety of collaborative agreements (Rothaermel and Deeds, 2004). Since the early days of the biotechnology industry, the industry has been characterized by a strong interplay between basic sciences and applied research firms. Firms, therefore, have seen the need to establish collaborations with research scientists (Powell et al., 1996) through R&D alliances with other firms, laboratories and universities. In addition, firms allow their scientists considerable autonomy to work on their own projects and to publish and participate in the scientific community (Powell et al., 1996). We use the BioScan database to create an unbalanced panel data set publicly traded, stand-alone biotechnology firms from the US and Europe for the years 1990–2005. To ensure comparability, private firms, biotechnology divisions of large pharmaceutical companies and research institutes are excluded from the sample. The final sample includes 147 firms (115 with the primary R&D location in the US and 34 with the primary R&D location in the EU). This sample is slightly larger than those in studies with a similar industry and time focus (Hess and Rothaermel, 2014; Fabrizio, 2009; Phene et al., 2006). Despite the fact that we focus only on publicly traded biotechnology firms, we believe that our sample includes a broad set of firms since biotechnology firms often become public when they are relatively small and young. Also, since financial reports provide company information for up to three years before they go public, we are able to incorporate data from firms that are still private. Our sample therefore includes firms in early stages of development, which reduces possible sample bias that could arise from the exclusive focus on listed firms. Nevertheless, it is important to mention the possibilities of sample biases through this selection process. For instance, we expect our sample firms to be larger than private firms and we acknowledge that publicly traded firms may have better access to capital markets and can better fund their research activities. 3.1. Dependent variable Our dependent variable measures the change in the extent of alignment between a firm and the field (or industry) in innovative space. Our study suggests that each firm occupies a position in innovative space that can be observed from the technological areas in which the firm innovates at any point in time (Breschi et al., 2003; Phene et al., 2010). Like firms, the overall industry can also be seen to occupy a technological (innovative) position at a point

1478

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487

in time and to follow an evolutionary trajectory. An aggregation of the technological areas of the patented innovations of all the firms in the industry reveals the innovative focus of the industry and the changes in the technological areas of innovations over time indicates the movement of the innovative focus. We can therefore use data on patented innovations, to study the position of a firm relative to the industry focus at any point of time and compare the trajectory of the firm relative to the moving field. The comparison of the relative innovative position of the firm to the field across time allows us to observe whether a firm is increasing or decreasing its alignment with the innovative focus of the field. We use the international patent classification (IPC) codes to capture the technological positions (and the change in position across time) of firms in the industry. IPC codes have been frequently used to assess firms’ positions in technological space (Rosenkopf and Almeida, 2003; Song et al., 2003; Stuart and Podolny, 1996). The IPC codes are extracted from patent families2 in the Derwent Innovation Index (DII). Our analysis of patent data uses patent families rather than individual patents. Using information on patent families allows us to effectively consolidate patent data from various countries. Additionally, it gives us a better measure of patent output since we group patents belonging to one innovation into one family instead of counting each as an independent innovation. We follow Rosenkopf and Almeida (2003) in calculating the technological distance between a firm’s patent portfolio and the patent portfolio of the whole industry.3 We first tabulate the technological classes (at a sub-class level) for all patents in the sample (which is also our approximation of the industry). We then summarize the percentage of patents in each patent class for each firm and for the whole industry. The Euclidean distance (D) between each is calculated between the respective patent firm and the industry  (pkit − pkt )2 where pkit represents the proclass vectors: Dit = portion of a patenting activity for a firm (i) in a given subclass (k) in year (t), and pkt estimates the proportion of patenting activity in a given subclass (k) of the whole industry in the year (t). K is the number of dimensions (patent classes). This distance measure ranges from a low of zero (very high similarity between the firm and industry) to a value greater than one. The theoretical maximum of this measurement is slightly above 1.414 (the square root of two). This measure has the advantage that it weights the different IPC classes based on their size and use and thereby takes their relative importance into account. Thus, our dependent variable ‘Alignment with Focus of Innovation’ is the difference in the Euclidean distance between the firm and the innovative focus (field) across two points in time (Dit and Dit+n) and gives us a picture of whether a firm’s innovative position is growing closer to, or further away from, the innovative focus of the evolving field. The value of the dependent variable can range from −1.414 to +1.414. A positive value indicates an increase in alignment between the firm and the focus of the field, with a higher value indicating a greater increase in alignment. As innovative changes need time to be reflected in patent activities, we use a three year time period to evaluate the changes in the distance between the firm and the field (n = 3).4

2 A patent family is a group of patents filed by the same assignee(s) based on the claim of an original or priority patent. It includes the original patent and every subsequent patent based on the original. A patent family may include multiple applications from several countries, since there are differences in national regulations defining the breadth of intellectual property. 3 Rosenkopf and Almeida (2003) use the measurement to calculate the distance between patent portfolios of firms, but the logic is transferable to any two kind of patent portfolios. 4 As in previous related studies, we use the patent application date to identify the time of the invention since the application date is closer to the time of knowledge generation (Almeida and Phene, 2004 Lahiri, 2010; Phene et al., 2006; Rothaermel and Hess, 2007). The “first-to-file” approach of most patent systems (including the

The three year lags are based on the assumption that even in fast moving high technology and science driven industries like biotechnology it is difficult for firms to change or redirect their technological trajectories in short time periods. To test this assumption we calculated the summary statistics for change in firm technological profiles with lags from 1 to 5 years. The means and the standard deviation increase successively with increasing lags, which supports the idea of greater change and movement over time. Our analysis shows that the differences between the variances become significant if the lags are greater than two years. Additionally we run sensitivity analyses on the lags from 1 to 5 years and the results remain very stable for the 3 and 4 year lags. As expected the models for the 1 and 2 year lags are not significant due to the lower variance in the dependent variable. In regard to the insignificance of the 5 year model one has to consider that with increased lags, the influence of the collaboration activity declines. Additionally the larger time window reduces the use sample considerably (sample size n = 430). For this reason we believe that the models using three year lags present the most accurate description of the underlying phenomenon.

3.2. Independent variables 3.2.1. R&D alliances We identify R&D alliances between two firms through the patent database by examining each case where two or more organizations are listed as joint assignees (or owners) of the patent at the time it was granted. We take into account merger and acquisition activities, subsidiary relationships, and name changes which could have resulted in the listing of multiple assignees but did not represent R&D Alliances between independent firms. Previous research has shown that co-patenting is a not uncommon outcome of alliances and is more likely in industries with strong intellectual property regimes (Hagedoorn, 2003).5 Additionally, studies have highlighted the benefits and collaborative nature of co-patenting (Almeida et al., 2011; Belderbos et al., 2014). For example, Hussler and Rondé (2007) argue that co-patenting inventors frequently need to adjust and alter their behavior to create mutual understanding with their partners, which is essential for the exchange of knowledge and fruitful collaboration. Similarly, qualitative investigations have argued that co-patenting normally involves face-to-face interaction, extensive discussion between firms and individuals, exchange of ideas and joint problem solving (Almeida et al., 2011) and this is necessary to the exchange of tacit knowledge (Maggioni and Uberti, 2009). As a consequence co-patenting is costly and time-consuming and can be considered a strong form of collaboration (Porter et al., 2005). Though the use of co-patenting to identify alliances has the disadvantage that it captures only a limited set of activities as collaborations (since not all collaborative activities lead to copatenting), it has the advantage that it (1) focuses very precisely on innovation related alliances, (2) links the innovation generating activity accurately to the participants (authors, firms), (3) provides a finer grained picture of collaborations by including those not available from other sources, and (4) enables us to calculate input and performance measures of collaborative innovation activity. For

USPTO) incentivises the patent applicant to file relatively quickly for a patent, making the application date an appropriate indicator of the time of invention. 5 In our sample, co-patenting is more frequent than standard R&D alliances announcements. Within the time period, the sample firms engaged in an average of 1.4 co-patented patents (S.D of 2.4 and a maximum of 22). In contrast, the same firms engaged in average in 0.6 technological alliances (S.D of 1.5 and a maximum of 11).

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487

these reasons, we believe that co-patenting is a good measure for alliance based activities for this study.6 3.2.2. External individual collaborations This variable captures the extent to which scientists of a firm engage in collaborative activities with scientists from other organizations. It is measured as the number of articles (in scientific journals) co-authored by employees of the focal firm with employees of another organization. Publication data is frequently used to estimate the scientific activities of individuals and is seen as a reliable source of information because they are subject to the critical review of colleagues and have gained approval in a peer review process (Gittelman and Kogut, 2003). We use the ISI Science Citation Index (SCI) to identify all publications authored by at least one employee of the sample firms. We then code for whether a publication is created in collaboration with an individual from a different organization. It is important to check if the publication or patent is in fact created by two different organizations, so to properly identify firms, we followed a careful process of matching firm names and accounting for M&A activity, name changes and corporate and affiliate relationships. When comparing external individual collaborations and R&D alliances, it is important to examine the assumption that these collaborative activities are independent of each other. In this context, Almeida et al., (2011) empirically examined the extent to which R&D alliances of biotechnology firms are linked to individual collaborations and vice-versa. They investigated how many biotechnology firms are linked to the same partner via both collaborative mechanisms. The authors found that only 1.15% of firms that were linked by individual collaborations were also linked by R&D alliances. This suggests that a vast majority of individual collaborations are between organizations that do not have a R&D alliance relationship and it is therefore useful to examine these mechanisms as distinct entities. Murray (2002) showed that scientific and technical progress arises in two distinctive networks and that there are only very few overlaps between patenting and scientific activities as indicated by cross-citations and co-publications. These findings are also supported by our data that show that scientific publications and R&D alliances connect different sets of actors.7 3.2.3. Internal publications This variable measures the total number of articles published by individuals within the firm. External individual collaborations and internal publications together make up the complete set of publications by employees of a firm. To distinguish between collaborative and non-collaborative internal publications, we introduce the variables internal individual collaborations and internal indi-

6 As a robustness test we collected the data on R&D alliance announcement from the Bioscan Database and performed the main regression models with this alternative variable. In one set of regression models, we run the regression with the new variable instead of the co-patenting variable and in another set of models with use the new variable alongside the co-patenting variable. In both cases the alliance announcement variable is not significant and does not change the direction or significance of the other variables in a meaningful manner. 7 From a conceptual point of view, previous research has a relatively clear position that R&D alliances are formed and executed by individuals within an organization but usually involve a broader set of entities within the organization and therefore can be considered an organizational level mechanism (e.g., Powell et al., 1996; Stuart, 2000 Rothaermel and Hess, 2007). On the other hand, research on science and innovation has frequently highlighted the importance of individuals in scientific collaboration within the biotechnology (Powell et. al., 1996; Gittelman and Kogut, 2003). Recent field research (Almeida et. al, 2011) highlights the differences between the two collaborative mechanisms and argues “that individual collaborations are much more informal in nature and are usually managed and initiated by an individual based on personal relationships” (p. 1594). Contrasted with this, R&D alliances are seen as “more formal in nature and are planned and executed within the organizational bureaucracy” (p. 1594).

1479

vidual publications. Internal individual publications measures the number of single authored publications within the firm and internal individual collaborations measures the number of publications co-authored by two or more scientists in the same organization. 3.2.4. Technological specialization We measure technological specialization of the firm by calculating the Gini coefficient and using IPC technology sub-classes for every

year.8

The Gini coefficient is calculated as Gi =

n−1 CPij j=1  n

2

(n−1)(

P ) j=1 ij

where n is the total number of technology domains in which a firm is patenting, j is the technological class defined by three digit codes, CPij is the cumulative sum of patents by firm i in technology field j, ranked in increasing order, and Pij is the number of patents in each technological class. 3.3. Control variables To control for heterogeneity in our sample, we include several control variables that have previously been shown to influence to innovation, including: intellectual capital, science orientation, relative technological advantage, star scientists, number of patents, R&D intensity, marketing alliances and M&A activity. the intellectual strength of the researchers involved in research activities could influence their capability to predict future directions of the field and so we control for intellectual capital measured as the average number of times the academic articles published by each firm has been cited. The science orientation of the firm could influence the direction and level of specialization of subsequent innovation, so we measure this as the ratio of the numbers of papers published to R&D Investments. To provide a better picture of the overall alliance activities of a firm, we also measure the downstream or marketoriented alliances (including those dealing with marketing, sales and distribution).9 The strength of a firm in a technological area could affect its subsequent technological trajectory and therefore we measure the relative technological advantage (RTA) of each firm (Zhang et al., 2007). RTA measures the technological strength of a firm in a specific patent class in a given year. The numerator is the number of patents ‘p’ of firm ‘i’ in a technology class ‘k’ in a given time period ‘t’ divided by total number of patents of the firm ‘i’ across all patent classes. The denominator is the number of patents ‘p’ in technology class ‘k’ for every firm divided  by all patent  activity in the sector at time ‘t’: RTAit = ((pikt / pikt )/(pkt / pkt )). To obtain firm level constructs RTAit has to be aggregated for each firm and year. This measure also indicates the share of a firm’s innovative activities of the total innovation in the field. Several previous studies have highlighted the role that star scientists play in the innovation process (Zucker et al., 2002). Therefore we control for the number of star scientists for each firm in a given year. Similar to Rothaermel and Hess (2007), we identify star scientists based on their publication activity. We first identify every author within our sample of publications and then count the number of citations received per researcher. To account for the fact that in biotechnology, articles have a large number of authors, we adjust for the number of authors per article. Finally, we define star scientists as those scientists whose total number of citations are more than three standard deviations above the mean. Our procedure identifies 906 star scientists (1.04% of the total scientists), who are involved in 30.9% of all publications and accounted for

8 Zeebroeck et al. (2006) argue that the Gini coefficient is the most reliable measure of technological specialization using patent data. 9 The alliances are identified using the BioScan database.

1480

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487

30.8% of all citations. These ratios are comparable to previous studies (Rothaermel and Hess, 2007; Zucker et al., 2002). To account for the overall patent activities of firms we control for the number of patents generated in a given year and include a patent dummy, which controls if a firm is patenting at all in a given year. In addition, we account for a firm’s R&D intensity, measured by R&D investment divided by the number of employees, as this firm characteristic has been shown to influence cooperative activities (Fritsch and Lukas, 2001).10 Since acquisitions can be used as a mode of accessing external knowledge, we measure if the focal firm acquired another biotechnology firm in a given year. We also account for the number of marketing (downstream) focused alliances. Finally, due to the non-normal distribution of R&D intensity, star scientists, number of patent and all publication variables, we normalized these variables using a logarithmic scale and to ease the interpretation of the results and to reduce potential co-linearity, interaction effects are mean-centered. 3.4. Estimation We estimate all models with an OLS (unbalanced) panel regression with robust standard errors.11 To capture unobserved time variant heterogeneity between firms, all models include firm fixed effects (Greene, 2003) and firm clustered standard errors. Our parameters are thus identified by the within-firm variation. Furthermore, we also control for time-varying factors that influence all firms via year fixed effects. Year fixed effects capture secular movements, including macroeconomic conditions, in the dependent variable and attenuate any right truncation effect. Additionally, we reduce potential simultaneity bias by calculating the dependent variable using a lag of 3 years. As with the dependent variable, we use alternative lags (2 years and 4 years) to test for robustness. 4. Results Table 1 present the descriptive statistics and the correlation matrix. The data reveal strong heterogeneity across firms. As expected the data show that individual collaborations are more numerous than R&D alliances and single authored publications. The average number of external individual collaborations per firm per year is 14,2 compared to 4.3 internal publications and 1.4 R&D alliances. Of the 698 firms-year observations, 8 have no individual collaborations, 32 have no internal publications and 23 have no R&D alliances. This suggests that collaborative research activity is commonplace in the industry. The firm with the most publication activities over the period studied is Amgen with more than 2700 individual collaborations and 660 internal publications. Of the internal publications for the sample, 26.9 percent are single authored and 73.1 percent are co-authored.12

10 The number of employees is preferable to alternative measures of size such as total assets or sales in this industry. Total sales is a poor measure for this study since biotechnology firms often do not have positive revenue streams (or have very volatile ones) and accounting measures may not capture the size of small firms in high technology industries. 11 The dependent variable could suggest the application of a Tobit regression as it has an upper and lower limit. However, for this study, truncation is not a critical issue, as the data is, as expected, relatively normally distributed with a mean close to zero (mode = 0.002; skewness = 0.650; kurtosis = 15.415). Thus the upper and lower limits represent theoretical limits with little practical relevance. When we run our OLS regressions, the predicted results are within the limits of the upper and lower boundaries. Additionally, Tobit models have the disadvantage of being sensitive to the violation of their underlying assumptions (particularly normality) and they also prevent the use of fixed-effect models to control for unobserved heterogeneity Greene (2003). Since the ability to control for unobserved heterogeneity via fixed effects is crucial for our paper, we believe that OLS approach is most suitable. 12 In comparison to other studies, the sampled firms are on average larger (481.42) than in the study of Tzabbar and Kehoe (2014) which includes private firms. How-

Table 2 shows that most correlations are at a moderate level. It is important to notice that not all variables shown in the correlation table appear in the same regression model. The variable, internal publications, is relatively highly correlated to internal individual collaborations and internal individual publications. This is because internal individual collaborations and internal individual publications add-up to internal publications. However, the variable internal publications is not entered in the same model as internal individual collaboration and internal individual publication. To evaluate the remaining correlations, we calculate the Variance Inflation Factor (VIF). All VIF’s are at acceptable levels (<5) indicating that the multi-co-linearity does not lead to a bias in our results.13 4.1. Main regressions Table 3 shows the results for the OLS firm and year fixed-effects regression model. In model 1, we estimate a baseline model including only the control variables.14 In models 2–4 we test the main independent variables separately and in models 5 and 6 we test the combination of the main independent variables but without testing external individual collaborations and internal publication within model to avoid possible bias through the high correlation of these variables. Supporting Hypothesis 1 and 2, R&D alliances (p < 0.01) has a negative and significant while external individual collaborations has a positive and significant effect (p < 0.05).15 However, the coefficient for internal publications is not significant. These results indicate that collaborations with individuals from other organizations help align firms with the (moving) focus of future innovation in the field, while internal publications and R&D alliances have the opposite effect. We next add the degree of technological specialization to the analysis (models 7–12). In line with Hypothesis 4, technological specialization decreases the alignment with the innovative focus (p < 0.01) across all models. The incorporation of technological specialization reduces the level of significance of R&D alliances (p < 0.1) and external individual collaborations (p < 0.1). In models 12–14, we include the individual interaction effects between the search mechanisms and technological specialization and in models 15 and 16 we show the models combining the main independent variables but without testing external individual collaborations and Internal Publication in one model to avoid issues of high correlations between these variables. The results indicate that technological specialization decreases the negative effect of R&D alliances and Internal Publications (both p < 0.5) but only slightly influences external individual collaborations (p < 0.1). Thus we do not have support for Hypothesis 5a–c. 4.2. Robustness checks & extended analysis on knowledge and collaboration types To examine the robustness of the results, we run further analyses. First, we use the ratio of external individual collaboration publications to internal publications to examine the effect of

ever the average amount of R&D investment in our sample is (47.46, SD = 97.86) is smaller but comparable to the study of Hess and Rothaermel (2014), which also focuses on publicly traded firms. 13 We also performed robustness test for specific variable pairs, e.g., external individual collaboration and internal publication (see Section 4.2). 14 The F-statistics are significant for all presented models. 15 The OLS coefficients describe the change in the dependent variable that would be predicted by a unit increase in the independent variable. The transformation (Ln) of some of the key variables changes the interpretation slightly. For example, the coefficient of 0.037 for external individual collaboration in model 5 means that a 100 percent increase in external individual collaboration is related to a 0.037 units closer to the emerging focus of innovation (based on model 5).

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487

1481

Table 1 Descriptive statistics. Variable

Min

q25

q50

q75

Mean

Max

Std. dev.

Focus of innovation R&D intensity Acquisition Patenting dummy Marketing alliances Science orientation (x1000) Relative technol. advantage Intellectual capital Star scientist Number of patents Technological specialization External individual collaboration R&D alliances Internal publication Internal individual collaboration Internal individual publication

−0.86 0.18 0.00 0.00 0.00 0.00 0.67 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

−0.03 4.37 0.00 1.00 0.00 0.14 1.26 0.00 0.00 2.00 0.17 1.00 0.00 0.00 0.00 0.00

0.00 4.98 0.00 1.00 0.00 0.37 2.06 0.00 0.00 5.00 0.30 5.00 1.00 1.00 0.00 0.00

0.03 5.34 0.00 1.00 0.00 0.78 3.24 3.80 0.00 10.00 0.40 12.00 2.00 4.00 3.00 1.00

0.01 4.78 0.10 0.90 0.12 0.67 2.39 3.29 4.61 12.10 0.29 14.17 1.36 4.29 3.36 0.99

0.85 7.51 3.00 1.00 5.00 15.67 8.55 168.28 293.00 265.00 0.85 321.00 22.00 87.00 80.00 23.00

0.15 0.96 0.37 0.31 0.47 1.04 1.34 9.09 24.08 26.80 0.18 34.26 2.40 10.46 8.67 2.27

individual collaboration intensity. This variable provides a complementary perspective on the use of the collaborative mechanisms as it focuses on the relative use of external individual collaboration publications. The variable also reduces concerns of the high correlation between the two variables. Model 17 show that firms relying more on individual collaborations are more aligned with the focus of innovation (p < 0.05), which is in line with Hypothesis 1 and 3, however the interaction of this variable with technological specialization is not significant. Second, we recognize that a few firms that patent a lot could have a disproportional influence on the location of the innovation focus and the trajectory of the field. Even though we believe that this effect is small due to the sufficient sample size, we run an additional set of regressions excluding any firm that is responsible for more than 10 percentage of the industry patent output in a given year.16 While our investigation focuses on the difference between individual collaborations and R&D alliances as regards their influence on innovation direction outcomes, we run additional models (Table 4) to further explore the effect of the different knowledge and collaboration types. 4.2.1. Types of internal publications An interesting question is whether the negative relationship between the internal publications variable and the dependent variable is driven by the internal collaborative publications or by sole authored publications. To investigate this, we run models splitting internal publications into two parts – internal individual collaborations (capturing the number of co-authored publications by individuals within the same firm) and internal individual publications (the number of sole-authored papers in the same firm). Model 19 and 20 shows the effect of the two variables independently and in model 21 both variables are tested together. The results, indicates that internal individual collaborations has a marginally significant (p < 0.1) negative effect on the dependent variable but not internal individual publications. Following the logic of the previous table, we also run a model including technological specialization (model 22) and the interaction term between technological specialization and the two

16 Due to the similarity of the results, these models are not included in the paper, but they can be requested from the authors. Additionally, we also run all models with firm and year fixed-effect logit specification, where the dependent variable is specified as a dummy indicating increasing or decreasing alignment with the innovation focus. This analysis is more conservative than the OLS model, because the specification of the dependent variable only measures the direction of change but not its magnitude. As expected the results broadly support the previous findings.

new variables (model 23). In line with the previous models, the technological specialization variable is negative and significant (p < 0.01) and the interaction term that includes internal individual collaborations is significant (p < 0.01). These results suggest that collaborative activities – both external and internal – significantly influence the dependent variable. While external collaborations at the individual level increase the alignment of the firm with the focus of innovation of the field, internal collaborations have the opposite effect.

4.2.2. Local vs. non-local external collaborations and partners Even in a globalizing world, knowledge often remains localized within countries or regions (Bartholomew, 1997; Lundvall et al., 2002). Researchers have argued, therefore, that geographically spread knowledge inputs can provide firms with a diverse set of knowledge inputs which positively influences innovation outcomes (Lahiri, 2010; Phene et al., 2006). Additionally, several studies have shown collaborative mechanisms can be used to access non-regional or international knowledge (Almeida et al., 2002; Inkpen, 1998; Simonin, 2004). For example, Almeida et al. (2002) shows that alliances are a superior solution to international knowledge transfers than market transactions. Hohberger (2014) shows that alliances foster the search for emerging knowledge and this is particularly true for non-regional and international knowledge search. Colombo et al. (2009) shows that: (a) the number of countries in which R&D alliance partners are located and, (b) the closeness of these countries to worldwide knowledge sources, positively influence firm performance. In one of the few investigations of internationality of individual collaborations However, Zaheer and George (2004) highlight the fact that alliances are also beneficial to the access of regional knowledge. Their study suggest that the access of complex and regional knowledge is best organized through local collaboration as it is not sufficient for firms to be located within in a region in order to harness knowledge spillovers (Zaheer and George, 2004). Thus, in the context of this study it is important to account for the extent to which firms engage in non-regional individual collaborations (or technological alliances). Non-regional collaborations (or alliances) can provide firms with a diverse and new set of knowledge inputs and might enable firms to innovate closer to the emerging focus of innovation. Therefore, we incorporate a variable that examines whether collaboration partners (of individual collaborations and alliances) are located within the same state (or country for European firms) as the focal firm. We then calculate the ratio of regional partners to nonregional partners for individual collaborations and alliances, and include them in the regression analysis (models 24 and 25). We use the ratio of the pure collaboration or alliance counts of regional

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487

0.4106 0.4061 0.440 0.479 0.211 0.141 0.036 −0.149

0.161

0.240

0.044

−0.133

0.251

0.626

0.481

0.6218 0.755 0.649 0.191 0.178 0.016 −0.042

0.151

0.157

0.187

−0.100

0.429

0.452

0.397

0.969 0.357 0.693 0.503 0.257 0.186 0.075 −0.058

0.117

0.179

0.127

−0.180

0.412

0.396 0.739 0.388 0.522 0.098 0.280 0.162 0.187 −0.015 0.063 −0.097 −0.061

0.106 0.138

0.124 0.193

0.071 0.160

−0.122 −0.176

0.263 0.441

1.000 1.000 0.566 1.000 0.285 0.181 0.271 −0.092 −0.170 −0.253 −0.240 0.015 0.178 0.006 0.301 0.039 0.120 0.148 0.189 0.062 −0.021 0.081 0.035 −0.028 −0.038 −0.099 −0.029

0.011 0.074 0.178 0.146

0.102 0.078 0.540 0.202

1.000 −0.006 1.000 0.033 -0.058 1.000 0.047 −0.015 −0.147 1.000 −0.139 0.121 -0.059 −0.293 −0.203

1.000 0.044 0.097 −0.049 −0.018

5

16

15

14

12 13

8 9 10 11

2 1

1.000 0.037 −0.024 −0.090 −0.026 −0.010 0.106

Focus of innovation R&D intensity (Ln) Acquisition Patenting dummy Marketing alliances Science orientation Relative technological advantage Intellectual capital Star scientist (Ln) Number of patents (Ln) External individual collaboration (Ln) R&D alliances (Ln) Internal publication (Ln) Internal individual collaboration (Ln) Internal individual publication (Ln) Technological specialization

Table 2 Correlation matrix.

1 2 3 4 5 6 7

3

4

6

7

1.000

8

9

1.000 0.354 0.522

11 10

12

1.000 0.380

13

1.000

14

1.000

15

1.000

16

1.000

1482

and non-regional collaboration mechanisms due to the high correlation of these variables (r = 0.73). The results show that both ratios are not significant. Additionally, the direction and significance of the main coefficient remains nearly unchanged (in comparisons to models 10 and 11 in Table 3), which supports the robustness of our initial analysis. It is important to notice that despite the fact that previous research highlights the regional aspects of collaboration, we do not much find support for this association. The reason for this could be that while earlier research largely used firm-specific dependent variables, this study uses a dependent variable which is calculated in relation to the industry. Non-regional knowledge might influence the innovation direction of a firm as it often provides diverse knowledge to the firm. However, it is less clear how non-regional knowledge can influence the movement of the firm relative to the industry, as the industry is comprised of firms that span regional boundaries. 4.2.3. Scientific nature of external collaborations and alliances The scientific nature of individual collaboration partners is a key factor explaining a firm’s alignment with the focus of innovation, as scientific partners such as universities operate with different governance principles, institutional constrains, and within different knowledge regimes (Almeida et al., 2011; Jiang et al., 2011; Murray, 2002; Sorenson and Fleming, 2004). However, individual collaborations can vary in the degree of scientific orientation (this is equally true for technological alliances). To explore this heterogeneity, we investigate the influence of science based partners in external individual collaborations and technological alliances on the alignment with the focus of innovation. We code whether individual collaboration partners and alliance partners are employed in universities (or public research institutions) or in for-profit firms. We use the ratio of science-oriented partners to non-science oriented partners. Similar, to the analysis of local vs. non-local collaboration partners, we prefer the ratio to raw counts due to the high correlation of collaboration partner types (r = 0.92). Model 26 and 27 (Table 4) shows that these ratios are not significant and similar to the previous robustness test, the direction and significance of the main coefficients remain nearly unchanged. 4.2.4. Limitations While we have used different models and approaches to ensure the robustness of our results, some limitations remain. First, this study focuses on a sample of publicly traded firms in a single industry, which raises questions about the generalizability of the results. The biotechnology industry is special given its heavy reliance on basic scientific research and its unique product development and approval process. Second, patent and article based measures are limited in the extent to which they can capture firm innovation and collaboration behavior. These measures most likely under-represent collaborative activity of companies, since they only capture the collaborations that lead to publications or patents. Nevertheless they provide one of the best measures currently available and we have applied and interpreted them with the necessary care and rigor. Third, we believe that our sample is representative of a wide range of firms in the industry, as it includes small and private firms (see description Section 3), but the definition of the industry or technological sector can potentially influence the results. For example we deliberately exclude university patenting from our sample since Trajtenberg et al. (1997) show that firm and university patenting is different along several dimension including basic science orientation and appropriability. Including universities and research institution should provide a rather different technological landscape, which is more focused on basic science rather than technological application. However, we believe that our results on the underlying collaborative mechanisms are potentially transferable

Table 3 The influence of R&D alliances, individual scientific collaborations and internal collaboration on the direction of innovation. Model Variables R&D intenstity (Ln) Acquisition Patenting dummy Marketing alliances Science orientation

Intellectual capital Star scientist (Ln) Number of patents (Ln)

−0.053*** 0.020 −0.013 0.013 −0.018 0.034 −0.008 0.009 −3.899 6.815 0.011 0.008 0.000 0.000 −0.002 0.003 −0.001 0.008

External individual collaboration (Ln)

2 B/SE −0.058*** 0.020 −0.016 0.014 −0.013 0.033 −0.008 0.009 −8.929 5.826 0.010 0.008 0.000 0.000 −0.006* 0.003 −0.009 0.008 0.026** 0.012

3 B/SE −0.052*** 0.019 −0.013 0.013 −0.028 0.034 −0.009 0.009 −4.183 6.678 0.011 0.008 0.000 0.000 −0.003 0.003 0.006 0.008

4 B/SE −0.049** 0.021 −0.013 0.013 −0.021 0.034 −0.008 0.009 −2.181 7.651 0.011 0.008 0.000 0.000 −0.002 0.003 0.002 0.009

−0.035*** 0.011

R&D alliances (Ln)

5 B/SE −0.058*** 0.019 −0.016 0.013 −0.023 0.033 −0.008 0.009 −9.302 5.746 0.010 0.008 0.000 0.000 −0.007* 0.004 −0.002 0.008 0.027** 0.012 −0.036*** 0.011

−0.018 0.015

Internal publication (Ln) Technological specialization

6 B/SE

7 B/SE

−0.047** −0.056*** 0.020 0.019 −0.012 −0.013 0.013 0.013 −0.031 −0.026 0.034 0.033 −0.008 −0.009 0.009 0.009 −2.352 −10.268* 7.504 5.856 0.011 0.010 0.008 0.008 0.000 0.000 0.000 0.000 −0.002 −0.006 0.004 0.004 0.009 0.008 0.009 0.010 0.022* 0.012 −0.036*** 0.011 −0.019 0.015 −0.226*** 0.060

8 B/SE −0.050*** 0.018 −0.011 0.013 −0.034 0.033 −0.009 0.009 −5.808 6.565 0.011 0.008 0.000 0.000 −0.003 0.003 0.016* 0.009

−0.023* 0.013

−0.186*** 0.070

9 B/SE

10 B/SE

−0.047** −0.056*** 0.019 0.016 −0.010 −0.013 0.013 0.017 −0.032 -0.031 0.034 0.026 -0.009 −0.009 0.009 0.014 −4.371 −10.128 7.266 8.438 0.011 0.010 0.008 0.006 0.000 0.000 0.000 0.001 −0.002 −0.006 0.004 0.008 0.016 0.010 0.010 0.012 0.022** 0.011 −0.022* 0.012 −0.016 0.014 −0.213*** −0.196*** 0.060 0.055

Individual collaboration (Ln) × tech. specialization R&D alliances (Ln) × tech. specialization Internal publication (Ln) × tech. specialization Constant Year dummies Firm level effect R-square R-square adj. Observations Groups *** ** *

0.258** 0.099 Yes Fix E. 0.031 0.018 698 147

0.262*** 0.098 Yes Fix E. 0.041 0.027 698 147

0.269*** 0.097 Yes Fix E. 0.046 0.032 698 147

0.250** 0.100 Yes Fix E. 0.035 0.021 698 147

0.274*** 0.097 Yes Fix E. 0.057 0.042 698 147

0.261*** 0.099 Yes Fix E. 0.051 0.035 698 147

0.297*** 0.097 Yes Fix E. 0.067 0.052 698 147

0.294*** 0.097 Yes Fix E. 0.066 0.051 698 147

0.285*** 0.099 Yes Fix E. 0.063 0.048 698 147

0.300*** 0.081 Yes Fix E. 0.073 −0.199 698 147

11 B/SE −0.046*** 0.016 −0.011 0.017 −0.037 0.027 −0.009 0.014 −4.156 8.275 0.011* 0.006 0.000 0.001 −0.002 0.008 0.019 0.012

12 B/SE −0.053*** 0.019 −0.010 0.013 −0.018 0.033 −0.007 0.009 −9.380 5.772 0.010 0.008 0.000 0.000 −0.006 0.004 0.003 0.009 0.005 0.019

13 B/SE −0.046*** 0.017 −0.012 0.013 −0.030 0.034 −0.004 0.009 −4.733 6.525 0.011 0.008 0.000 0.000 −0.001 0.003 0.013 0.009

14 B/SE −0.046** 0.019 −0.007 0.013 −0.021 0.034 −0.008 0.009 −3.211 7.381 0.011 0.008 0.000 0.000 0.000 0.004 0.012 0.010

−0.024* −0.086** 0.012 0.036 −0.017 −0.049** 0.012 0.024 −0.181*** −0.319*** −0.257*** −0.283*** 0.055 0.096 0.073 0.076 0.070* 0.039 0.180** 0.076 0.108** 0.042 0.287*** 0.295*** 0.292*** 0.289*** 0.082 0.097 0.095 0.099 Yes Yes Yes Yes Fix E. Fix E. Fix E. Fix E. 0.069 0.074 0.080 0.072 −0.204 0.058 0.064 0.056 698 698 698 698 147 147 147 147

15 B/SE −0.050*** 0.018 −0.012 0.013 −0.021 0.033 −0.004 0.009 −8.400 5.724 0.011 0.008 0.000 0.000 −0.004 0.004 0.004 0.009 0.009 0.020 −0.078* 0.039

−0.320*** 0.102 0.048 0.044 0.156* 0.084

0.296*** 0.095 Yes Fix E. 0.090 0.071 698 147

16 B/SE −0.042** 0.018 −0.009 0.013 −0.025 0.034 −0.004 0.009 −2.091 7.234 0.011 0.008 0.000 0.000 0.001 0.004 0.014 0.010

−0.081** 0.038 −0.043* 0.025 −0.293*** 0.083

0.162** 0.081 0.078* 0.045 0.287*** 0.096 Yes Fix E. 0.090 0.071 698 147

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487

Relative technological advantage

1 B/SE

p < 0.01. p < 0.05. p < 0.1.

1483

Model Variables R&D intenstity (Ln) Acquisition Patenting dummy Marketing alliances Science orientation Relative technological advantage Intellectual capital Star scientist (Ln)

18 B/SE

−0.051*** 0.018 −0.011 0.013 −0.035 0.033 −0.009 0.009 −6.656 6.368 0.011 0.008 0.000 0.000 −0.004 0.003 0.015 0.009

−0.050*** 0.018 −0.011 0.013 −0.034 0.033 −0.009 0.009 −6.509 6.315 0.011 0.008 0.000 0.000 −0.004 0.003 0.014 0.009

−0.023* 0.013

−0.023* 0.013

−0.188*** 0.070 0.003** 0.001

−0.199** 0.081 0.002 0.003 0.005 0.007

External individual collaboration (Ln) R&D alliances (Ln)

19 B/SE −0.053*** 0.020 −0.015 0.013 −0.029 0.034 −0.008 0.009 −6.994 6.304 0.010 0.008 0.000 0.000 −0.006 0.004 0.003 0.009 0.026** 0.013 −0.025*** 0.011

20 B/SE −0.055*** 0.019 −0.016 0.013 −0.024 0.034 −0.009 0.009 −7.222 6.110 0.011 0.008 0.000 0.000 −0.006 0.004 0.000 0.008 0.021* 0.012 ***

0.011

21 B/SE −0.052** 0.020 −0.015 0.013 −0.028 0.034 −0.008 0.009 −6.391 6.572 0.010 0.008 0.000 0.000 −0.006 0.004 0.003 0.009 0.027** 0.012 −0.025*** 0.011

22 B/SE

23 B/SE

24 B/SE

−0.051*** 0.019 −0.013 0.013 −0.034 0.034 −0.009 0.009 −8.757 6.270 0.010 0.008 0.000 0.000 −0.006 0.004 0.013 0.010 0.030** 0.012 −0.024* 0.013

−0.050*** 0.019 −0.009 0.013 −0.021 0.034 −0.007 0.009 −8.162 6.364 0.010 0.008 0.000 0.000 −0.005 0.004 0.007 0.009 0.034*** 0.012 −0.061* 0.012

−0.058*** 0.019 −0.015 0.013 90.029 0.033 −0.009 0.009 −11.834** 5.801 0.010 0.008 0.000 0.000 −0.008** 0.004 0.008 0.009 0.030** 0.012 −0.023* 0.013

−0.191*** 0.066

−0.268*** 0.080

−0.198*** 0.068

−0.024* 0.012 −0.013 0.015

−0.061** 0.026 −0.033 0.040 0.115** 0.053 0.060 0.085

Internal publication (Ln) Technological apecialization Ratio external individual collaboration to internal publication Ratio external ind. collaboration to int. publication × techn. specialization Internal individual collaboration (Ln)

−0.025* 0.013 −0.016 0.015

Internal individual publication (Ln)

−0.025* 0.013 −0.016 0.015

Internal individual collaboration (Ln) × tech. specialization Internal individual publication (Ln) × tech. specialization Ratio regional to non-regional ext. ind. collaboration Ratio regional to non-regional R&D alliances Ratio science to application oriented ext. ind. collaboration Ratio science to application oriented techn. alliances Constant Year dummies Firm level effect R-square R-square adj. Observations Groups *** ** *

p < 0.01. p < 0.05. p < 0.1.

−0.001 0.003 0.002 0.003

0.293*** 0.097 Yes Fix E. 0.068 0.052 698 147

0.293*** 0.097 Yes Fix E. 0.069 0.051 698 147

0.262*** 0.099 Yes Fix E. 0.059 0.042 698 147

0.268*** 0.097 Yes Fix E. 0.053 0.037 698 147

0.258** 0.100 Yes Fix E. 0.061 0.043 698 147

0.285*** 0.099 Yes Fix E. 0.081 0.062 698 147

0.288*** 0.099 Yes Fix E. 0.096 0.074 698 147

0.302*** 0.097 Yes Fix E. 0.079 0.060 698 147

25 B/SE −0.046** 0.019 −0.011 0.013 90.037 0.034 −0.009 0.009 −4.199 7.264 0.011 0.008 0.000 0.000 −0.002 0.004 0.019* 0.010

−0.024* 0.013 −0.017 0.014 −0.181*** 0.068

26 B/SE −0.058*** 0.019 −0.014 0.013 −0.028 0.033 −0.008 0.009 −11.980** 5.789 0.009 0.008 0.000 0.000 −0.007* 0.004 0.008 0.009 0.028** 0.012 −0.023* 0.013

27 B/SE −0.048** 0.019 −0.012 0.012 −0.035 0.034 90.008 0.009 −5.484 7.396 0.010 0.008 0.000 0.000 −0.002 0.004 0.019* 0.010

−0.199*** 0.069

−0.024* 0.012 −0.019 0.014 −0.185*** 0.069

0.002 0.003 −0.004 0.006 0.305*** 0.096 Yes Fix E. 0.080 0.061 698 147

0.006 0.004 −0.005 0.006 0.290*** 0.097 Yes Fix E. 0.073 0.054 698 147

0.000 0.003 0.001 0.003

0.286*** 0.099 Yes Fix E. 0.069 0.050 698 147

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487

Number of patents (Ln)

17 B/SE

1484

Table 4 Robustness checks & extended analysis on knowledge and collaboration types.

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487

to other industry definitions (including the broader pharmaceutical industry). Fourth, even though we have accounted for key partner characteristics (e.g., organizational type and their location) the volume and international nature of these organizations inhibit us to calculate accurate technology profiles of the partner organizations. Fifth, even with a large set of controls, time lags in conjuncture with fixed effects and multiple robustness tests does not completely resolve concerns of unobserved time-varying characteristics, and the related issue of endogeneity. For example we cannot discount the role of scientists who make choices to collaborate internally and hence could enhance the level of specialization. Thus, caution should be exercised in interpreting the observed relationship between collaborative activities the alignment within technological space. Finally, while we focus on collaborative mechanism and control for M&A activities, there are other mechanisms to source knowledge externally which could potentially influence the alignment with the emerging focus of innovation in the field (e.g., mobility). Future studies could expand our analyses to other learning mechanisms.

5. Discussion & conclusion Our study sheds light on the issue of how firms can keep abreast of continuously evolving, complex, and dispersed knowledge through the use of collaborative mechanisms and how the usage of these mechanisms influences their technological trajectory relative to the field. Building upon concepts from evolutionary economics (Nelson and Winter, 1982), we suggest that firms can potentially choose from a range of mechanisms to acquire and utilize knowledge that not only affects their innovativeness but also, the direction of their innovation. The study compares the effect of three key collaborative mechanisms related to knowledge acquisition and innovation. This paper seeks to build upon, and contribute to, the extant literature in strategy, technology and innovation and focuses not on testing the impact of external knowledge on the quality and the quantity of innovation (a question that has been well researched) but rather on the influences of this knowledge on the direction of innovation. Further, it moves beyond investigating whether external knowledge can be used for exploration by examining the mechanisms (and conditions) that can lead to adjustments in the innovation profile of a firm relative to the broader field in a dynamic environment. We suggest that the collaborations of individual scientists across organizations gives firms a diverse set of knowledge inputs from a range of external sources and this provides early signals about various innovative directions of the field. The knowledge inputs so obtained and incorporated into research at the individual level are less impacted by the constraining elements of the organization’s routines and can thus help to redirect the mindset of decision making within the firm and reorient the innovative directions of the firms toward emerging innovative areas. Our empirical findings support this idea. We argue that two other collaborative mechanisms – R&D alliances and internal collaborations – are products of existing firm systems, expertise, and world views, and serve to harness knowledge that matches, or is close to, existing activities and trajectories. They are, therefore, likely to enhance local search and help harden existing innovation trajectories. In a dynamic environment, as new areas of innovation unfold, firms that employ these mechanisms will tend to drift away from the evolving focus of innovation in the field. Our empirical analysis confirms these expectations largely for R&D alliances. While we do not find supports for our expectations for the case of internal publications our extended analysis shows that our argumentation is supported for internal collaborations.

1485

One of the main contributions of this study is that, while acknowledging that various collaborative mechanisms can be similarly useful in determining the extent of knowledge acquisition and application, there are important differences in how they influence the direction of innovation. R&D alliances and internal research collaborations may enhance innovation and be useful in extending existing innovative abilities, but they tend to move the firm away from the emerging innovative focus. Therefore, R&D alliances and internal collaborations are likely to be useful collaborative tools in fairly stable technological and innovative environments where firms can choose and build their abilities along predictable trajectories. These collaborative mechanisms could be useful when firms do not want to compete in the most crowded innovative areas of the future. External collaborations at the individual level may be particularly useful in dynamic and uncertain environments since they appear to facilitate and enhance the flexibility of firms to innovate in areas of emerging opportunities. Of course, innovating in the most crowded areas has its perils in terms of increased competition in technology and related products. Our paper suggests that using mechanisms of knowledge acquisition to position a firm in innovative space is difficult. First, the spectrum of technological and scientific areas across which innovations occur is continuously changing, often in unpredictable ways. Competition shapes the innovative field and it is hard to gage the most attractive innovative areas for a particular firm. Second, even if firms can clearly see the most attractive innovative areas for the future, it is hard to move any firm in new technological and innovative directions. This research offers an important insight – even if the firm does not know where (which technologies or precise areas of innovation) the firm itself is going, or where exactly the field is going, it can, to some extent still influence, whether it will innovate closer to the focus of the field or further away from it in the future. The mechanisms it employs to acquire knowledge will guide the firm away from or toward the innovative focus of the future. Our research builds on recent papers that emphasize the role of individuals in knowledge transfer and innovation (Rothaermel and Hess, 2007; Song et al., 2003; Zucker et al., 2002). It is interesting to note that one of the core ideas in organizational design is that various intra-organizational tools and levers are used to align the incentives and actions of individual employees toward the main objectives of the organization. Yet we find here that the value of individual collaborations across organizational boundaries emerges, in part, from the fact that individuals search a diverse knowledge base and incorporate knowledge and insights that may not be fully aligned with the organization’s current way of thinking and practices. In doing so, these individual collaborations can help redefine the perspective of the firm, allow the firm to rethink the landscape of innovative possibilities, and move the innovative trajectory in new directions. It is very important to emphasize that our analysis does not make any assumption regarding the relationship between technological position in the field and organizational performance. Rather than examine the performance effects of moving toward the focus of industry innovation, we focus on the mechanisms which may influence a firm’s position within the field. Firms could influence their future position in the technological space by choosing and incentivizing the extent of use of different types of collaborative mechanisms. Acknowledgments The authors would like to thank two anonymous reviewers for the constructive and thoughtful feedback. We also wish to thank Francesco DiLorenzo and audiences at several seminars and conferences for helpful critiques and suggestions. We are responsible for any errors or omissions.

1486

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487

References Ahuja, G., Lampert, C.M., 2001. Entrepreneurship in the large corporation: a longitudinal study of how established firms create breakthrough inventions. Strategic Manage. J. 22, 521–543. Allen, T.J., Cohen, S.I., 1969. Information flow in research and development laboratories. Admin. Sci. Quart. 14, 12–19. Almeida, P., Hohberger, J., Parada, P., 2011. Individual scientific collaborations and firm-level innovation. Ind. Corp. Change 20, 1571–1599. Almeida, P., Song, J., Grant, R.M., 2002. Are firms superior to alliances and markets? An empirical test of cross-border knowledge building. Org. Sci. 13, 147–161. Bartholomew, S., 1997. National systems of biotechnology innovation: complex interdependence in the global system. J. Int. Bus. Stud. 28, 241–266. Belderbos, R., Cassiman, B., Faems, D., Leten, B., Van Looy, B., 2014. Co-ownership of intellectual property: Exploring the value-appropriation and value-creation implications of co-patenting with different partners. Res. Policy 43, 841–852. Benner, M.J., Tushman, M.L., 2003. Exploitation, exploration, and process management: the productivity dilemma revisited. Acad. Manage. Rev. 28, 238–256. Bercovitz, J.E., Feldman, M.P., 2007. Fishing upstream: firm innovation strategy and university research alliances. Res. Policy 36, 930–948. Breschi, S., Lissoni, F., Malerba, F., 2003. Knowledge-relatedness in firm technological diversification. Res. Policy 32, 69–87. Cantwell, J., 1993. Corporate Technological Specialisation in International Industries. In: Casson, M.C., Creedy, J. (Eds.), Industrial Concentration and Economic Inequality. Edward Elgar, Aldershot. Carnabuci, G., Operti, E., 2013. Where do firms’ recombinant capabilities come from? Intraorganizational networks, knowledge, and firms’ ability to innovate through technological recombination. Strategic Manage. J. 34, 1591–1613. Cockburn, I.M., Henderson, R.M., 1998. Absorptive capacity, coauthoring behavior, and the organization of research in drug discovery. J. Ind. Econ. 46, 157–182. Cohen, W.M., Klepper, S., 1996. Firm size and the nature of innovation within industries: the case of process and product R&D. Rev. Econ. Stat. 78, 232–243. Cohen, W.M., Levinthal, D.A., 1990. Absorptive capacity: a new perspective on learning and innovation. Admin. Sci. Quart. 35, 128–152. Colombo, M.G., Grilli, L., Murtinu, S., Piscitello, L., Piva, E., 2009. Effects of international R&D alliances on performance of high-tech start-ups: a longitudinal analysis. Strategic Entrepren. J. 3, 346–368. Colombo, M.G., Grilli, L., Piva, E., 2006. In search of complementary assets: the determinants of alliance formation of high-tech start-ups. Res. Policy 35, 1166–1199. Crane, D., 1972. Invisible Colleges. Diffusion of Knowledge in Scientific Communities. University of Chicago Press, Chicago, IL. Cyert, R.M., March, J.G., 1963. A Behavioral Theory of the Firm. Prentice-Hall, Englewood Cliff, NY. Day, D.L., 1994. Raising radicals: different processes for championing innovative corporate ventures. Org. Sci. 5, 148–172. Deeds, D., DeCarolis, D., Coombs, J., 2000. Dynamic capabilities and new product development in high technology ventures: an empirical analysis of new biotechnology firms. J. Bus. Venturing 15, 211–229. Dosi, G., Nelson, R.R., 1994. An introduction to evolutionary theories in economics. J. Evol. Econ. 4, 153–172. Doz, Y.L., 1996. The evolution of cooperation in strategic alliances: initial conditions or learning processes? Strategic Manage. J. 17, 55–83. Fabrizio, K.R., 2009. Absorptive capacity and the search for innovation. Res. Policy 38, 255–267. Fleming, L., Sorenson, O., 2004. Science as a map in technological search. Strategic Manage. J. 25, 909–928. Fritsch, M., Lukas, R., 2001. Who cooperates on R&D? Res. Policy 30, 297–312. Gittelman, M., Kogut, B., 2003. Does good science lead to valuable knowledge? Biotechnology firms and the evolutionary logic of citation patterns. Manage. Sci. 49, 366–382. Granstrand, O., 1998. Towards a theory of the technology-based firm. Res. Policy 27, 465–489. Grant, R.M., 1996. Toward a knowledge based theory of the firm. Strategic Manage. J. 17, 109–125. Grant, R.M., Baden-Fuller, C., 2004. A knowledge assessing theory of strategic alliances. J. Manage. Stud. 41, 61–81. Greene, W., 2003. Econometric Analysis. Prentice Hall, Englewood Cliffs. Gulati, R., Singh, H., 1998. The architecture of cooperation: managing coordination costs and appropriation concerns in strategic alliances. Admin. Sci. Quart. 43, 781–814. Hagedoorn, J., 2003. Sharing intellectual property rights—an exploratory study of joint patenting amongst companies. Ind. Corp. Change 12, 1035–1050. Hagedoorn, J., Duysters, G., 2002. External sourcing of innovation capabilities: the preferences for strategic alliances or mergers and acquisitions. J. Manage. 39, 168–188. Hamel, G., 1991. Competition for competence and inter partner learning within international strategic alliances. Strategic Manage. J. 12, 83–104. Helfat, C.E., 1994. Evolutionary trajectories in petroleum firm research-and-development. Manage. Sci. 40, 1720–1747. Henderson, R.M., Clark, K.B., 1990. Architectural innovation-the reconfiguration of existing product technologies and the failure of established firms. Admin. Sci. Quart. 35, 9–30.

Hess, A.M., Rothaermel, F.T., 2014. When are assets complementary? Star scientists, strategic alliances, and innovation in the pharmaceutical industry. Strateg. Manage. J. 32, 895–909. Hohberger, J., 2014. Searching for emerging knowledge: the influence of collaborative and geographically proximate search. Eur. Manage. Rev. 11, 139–157. Hussler, C., Rondé, P., 2007. The impact of cognitive communities on the diffusion of academic knowledge: evidence from the networks of inventors of a French university. Res. Policy 36, 288–302. Inkpen, A.C., 1998. Learning and knowledge acquisition through international strategic alliances. Acad. Manage. Executive 12, 69–80. Inkpen, A.C., 2002. Learning, Knowledge Management, and Strategic Alliances: So Many Studies, So Many Unanswered Questions. In: Contractor, F.J., Lorange, P. (Eds.). Cooperative strategies and alliances, Pergamon, Oxford. Jiang, L., Tan, J., Thursby, M., 2011. Incumbent firm invention in emerging fields: evidence from the semiconductor industry. Strategic Manage. J. 32, 55–75. Kim, D.J., Kogut, B., 1996. Technological platforms and diversification. Org. Sci. 7, 283–301. Kogut, B., Zander, U., 1992. Knowledge of the firm, combinative capabilities, and the replication of technology. Org. Sci. 3, 383–397. Lahiri, N., 2010. Geographic distribution of R&D activity: how does it affect innovation quality? Acad. Manage. J. 53, 1194–1209. Laursen, K., 2012. Keep searching and you’ll find: what do we know about variety creation through firms’ search activities for innovation? Ind. Corp. Change 21, 1181–1220. Lavie, D., Kang, J., Rosenkopf, L., 2011. Balance within and across domains: the performance implications of exploration and exploitation in alliances. Org. Sci. 22, 1517–1538. Leonard-Barton, D., 1992. Core capabilities and core rigidities - a paradox in managing new product development. Strategic Manage. J. 13, 111–125. Levinthal, D.A., March, J.G., 1993. The myopia of learning. Strategic Manage. J. 14, 95–112. Liebeskind, J.P., Oliver, A.L., Zucker, L., Brewer, M., 1996. Social networks, learning and flexibility: sourcing scientific knowledge in new biotechnology firms. Org. Sci. 7, 428–443. Lundvall, B.A., Johnson, B., Andersen, E.S., Dalum, B., 2002. National systems of production, innovation and competence building. Res. Policy 31, 213–231. Maggioni, M.A., Uberti, T.E., 2009. Knowledge networks across Europe: which distance matters? Annal. Region. Sci. 43, 691–720. March, J.G., 1991. Exploration and exploitation in organizational learning. Org. Sci. 2, 71–87. Martin, X., Mitchell, W., 1998. The influence of local search and performance heuristics on new design introduction in a new product market. Res. Policy 26, 753–771. Murray, F., 2002. Innovation as co-evolution of scientific and technological networks: exploring tissue engineering. Res. Policy 31, 1389–1403. Narin, F., Hamilton, K.S., Olivastro, D., 1997. The increasing linkage between us technology and public science. Res. Policy 26, 317–330. Nelson, R.R., Winter, S., 1982. An Evolutionary Theory of Economic Change. Belknap Press/Harvard University Press, Cambridge. Nerkar, A., Paruchuri, S., 2005. Evolution of R&D capabilities: the role of knowledge networks within a firm. Manage. Sci. 51, 771–785. Paruchuri, S., 2010. Intraorganizational networks, interorganizational networks, and the impact of central inventors: a longitudinal study of pharmaceutical firms. Org. Sci. 21, 63–80. Phelps, C.C., 2010. A longitudinal study of the influence of alliance network structure and composition on firm exploratory innovation. Acad. Manage. J. 53, 890–913. Phene, A., Fladmoe Lindquist, K., Marsh, L., 2006. Breakthrough innovations in the US biotechnology industry: the effects of technological space and geographic origin. Strategic Manage. J. 27, 369–388. Phene, A., Tallman, S., Almeida, P., 2010. When do acquisitions facilitate technological exploration and exploitation? J. Manage. Porter, K., Whittington, K.B., Powell, W.W., 2005. The institutional embeddedness of high-tech regions: relational foundations of the Boston biotechnology community. Cluster Network Innov. 261, 296. Powell, W., Koput, K.W., Smith-Doerr, L., 1996. Interorganizational collaboration and the locus of innovation: network learning in biotechnology. Admin. Sci. Quart. 41, 116–145. Rosenkopf, L., Almeida, P., 2003. Overcoming local search through alliances and mobility. Manage. Sci. 49, 751–766. Rothaermel, F.T., Deeds, D.L., 2004. Exploration and exploitation alliances in biotechnology: a system of product development. Strategic Manage. J. 25, 202–221. Rothaermel, F.T., Hess, A.M., 2007. Building dynamic capabilities: innovation driven by individual-, firm-, and network-level effects. Org. Sci. 18, 898–921. Saxenian, A.L., 1991. The origins and dynamics of production networks in silicon valley. Res. Policy 20, 423–437. Siggelkow, N., 2001. Change in the presence of fit: the rise, the fall, and the renaissance of liz claiborne. Acad. Manage. J. 44, 838–857. Simonin, B.L., 2004. An empirical investigation of the process of knowledge transfer in international strategic alliances. J. Int. Bus. Stud. 35, 407–427. Singh, J., 2005. Collaborative networks as determinants of knowledge diffusion patterns. Manage. Sci. 51, 756–770. Song, J., Almeida, P., Wu, G., 2003. Learning-by-hiring: when is mobility more likely to facilitate interfirm knowledge transfer? Manage. Sci. 49, 351–365.

J. Hohberger et al. / Research Policy 44 (2015) 1473–1487 Sorenson, O., Fleming, L., 2004. Science and the diffusion of knowledge. Res. Policy 33, 1615–1634. Srivastava, M.K., Gnyawali, D.R., 2011. When do relational resources matter? Leveraging portfolio technological resources for breakthrough innovation. Acad. Manage. J. 54, 797–810. Stuart, T.E., Ozdemir, S.Z., Ding, W.W., 2007. Vertical alliance networks: the case of university-biotechnology-pharmaceutical alliance chains. Res. Policy 36, 477–498. Stuart, T.E., Podolny, J.M., 1996. Local search and the evolution of technological capabilities. Strategic Manage. J. 17, 21–38. Suzuki, J., Kodama, F., 2004. Technological diversity of persistent innovators in japan-two case studies of large japanese firms. Res. Policy 33, 531–549. Trajtenberg, M., Henderson, R., Jaffe, A., 1997. University versus corporate patents: A window on the Basicness of invention. Econ. Innovation New Technol. 5 (1), 19–50. Tushman, M.L., 1977. Special Boundary Roles in Innovation Process. Admin. Sci. Quart. 22, 587–605.

1487

Wilden, R., Gudergan, S.P., 2014. The impact of dynamic capabilities on operational marketing and technological capabilities: investigating the role of environmental turbulence. J. Acad. Market. Sci., 1–19. Zaheer, A., George, V.P., 2004. Reach out or reach within? Performance implications of alliances and location in biotechnology. Manage. Decis. Econ. 25, 437–452. Zander, U., Kogut, B., 1995. Knowledge and the speed of the transfer and imitation of organizational capabilities- an empirical-test. Org. Sci. 6, 76–92. Zhang, J., Baden-Fuller, C., Mangematin, V., 2007. Technological knowledge base, R&D organization structure and alliance formation: evidence from the biopharmaceutical industry. Res. Policy 36, 515–528. Zollo, M., Reuer, J.J., Singh, H., 2002. Interorganizational routines and performance in strategic alliances. Org. Sci. 13, 701–712. Zucker, L.G., Darby, M.R., Armstrong, J.S., 2002. Commercializing knowledge: university science, knowledge capture, and firm performance in biotechnology. Manage. Sci. 48, 138–153.