Four Essays on Technology Licensing and Firm Innovation

Moreira, Solon Doctoral Thesis Four Essays on Technology Licensing and Firm Innovation PhD Series, No. 1.2014 Provided in Cooperation with: Copenhagen Business School (CBS) Suggested Citation: Moreira, Solon (2014) : Four Essays on Technology Licensing and Firm Innovation, PhD Series, No. 1.2014, ISBN 9788793155039, Copenhagen Business School (CBS), Frederiksberg, https://hdl.handle.net/10398/8868 This Version is available at: https://hdl.handle.net/10419/208879 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/3.0/ Solon Moreira The PhD School of Economics and Management PhD Series 01.2014 PhD Series 01.2014 Four Essays on Technology Licensing and Firm Innovation copenhagen business school handelshøjskolen solbjerg plads 3 dk-2000 frederiksberg danmark www.cbs.dk ISSN 0906-6934 Print ISBN: 978-87-93155-02-2 Online ISBN: 978-87-93155-03-9 Four Essays on Technology Licensing and Firm Innovation 1 Four Essays on Technology Licensing and Firm Innovation Solon Moreira PhD School in Economics and Management Copenhagen Business School Solon Moreira Four Essays on Technology Licensing and Firm Innovation 1st edition 2014 PhD Series 01.2014 © The Author ISSN 0906-6934 Print ISBN: 978-87-93155-02-2 Online ISBN: 978-87-93155-03-9 “The Doctoral School of Economics and Management is an active national and international research environment at CBS for research degree students who deal with economics and management at business, industry and country level in a theoretical and empirical manner”. All rights reserved. No parts of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or by any information storage or retrieval system, without permission in writing from the publisher. 3 Acknowledgements The completion of this dissertation has been a long and personal journey. The life as a graduate student, as we all the other steps that have to be taken to obtain the PhD, sometimes may sound like a miserable process: endless hours reading papers and running models in front of the computer; come up with original research topics and find the best way to translate then into interesting papers; spend several weeks, or months, in finding the most appropriate way to operationalize a theoretical construct. However, my experience as a PhD student has been anything but miserable. Those years that I have spent at Copenhagen Business School have been some of the most challenging and rewarding of my life. I have no doubt that the days I spent at CBS will always be important for me not only as a researcher but also as a person in general. The people that I thank here are a few of the many who made this dissertation possible, and who made my time as a PhD student an enjoyable and enriching experience. To begin, I am deeply grateful to my two supervisors Keld Laursen and Toke Reichstein. I doubt that any other combination of supervisors could have been more fruitful than the one I had. In the first place I am thankful to them for believing in me and devoting their time and efforts in training me. I am also very thankful for their guidance and demand for rigor with my research. I would like also to thank Mark Lorenzen for the guidance and supervision, especially on the beginning of my PhD. I am sure that I will never find the words to express all my gratitude for the support that I received from Mark. Another important person during the PhD was Thomas Rønde. I am very thankful for his insightful comments on early versions of the papers in this dissertation and also for accepting to be the Chair of my Committee. Along this line, I also want to thank Bo Nielsen for 4 accepting to be part of the pre-defense committee, all his comments were very useful to improve the papers in this dissertation. I would like also to thank Atul Nerkar and Marco Giarratana for accepting to be part of my PhD committee. Their comments and suggestions on how to improve the essays in this dissertation were very opportune and valuable. I also want to thank the other PhD students in my cohort: Arjan, Gouya, Maggie, Maria, Milan and Virgilio. We went through so many things together that is hard to imagine how the PhD would have been without those guys. I would also like to thank my dear friends Jasim, Raphael and Stefano. Leaving Cambridge was not an easy decision for me and Lorena, but our friendship has been a great way to remain in touch with that wonderful time in our lives. Once someone told me that it is not easy to find true friends after a certain age. After having met both Marlon and Esron I can certainly say it is not true. I am very thankful to you two for always being supportive and encouraging at the moments I thought I could not make it. Please never ask me to be your “supervisor”, I like you guys too much to do that. Looking a little further back in time, other very important person during my academic journey has been Professor Mario Amin. He deserves special mention for his contributions during a time when I was first learning many of the issues examined here. I am also thankful to my father for being always supportive and enthusiastic on each of the steps towards the conclusion of this PhD. Thank you dad for understanding my absence over those 4 years, we will recover this time together. And finally, I would like to thank to my wife. I cannot express my love and gratitude for all the support that she has been giving me over those 15 years together, without her I would never make it. I dedicate this PhD to you Lorena. 5 English Summary Licensing contracts represent one of the most widely used mechanisms to exchange technologies and transfer know-how between firms. Due to the opportunities that licensing creates for firms operating on both sides of the markets for technology, it has increasingly become an integral part of firms’ R&D strategies. On the supply side, the existing literature has been focused on understanding how technology licensing can be used by firms as a mechanism to recover investments in innovative activities and to foster learning opportunities. On the demand side, it has been shown that licensing is an important source that firms can tap into to feed their internal needs for innovative knowledge. While several studies have examined technology licensing through the lens of the licensor, research on how firms rely on licensing contracts to acquire knowledge and improve their innovation performance still leaves much to be investigated. Furthermore, with few exceptions, neither organizational nor contractual characteristics related to the licensing deals have received enough attention as determinants of the capacity of the acquiring firm to benefit from licensing in a new technology. The purpose of this dissertation is to investigate the relationship between technology licensing and firm innovation, also examining how the characteristics of the acquiring firm and the use of specific contractual clauses affect this main relationship. The papers in this dissertation build on a different set of theoretical perspectives connected to the licensing literature. The dissertation consists of a general introduction, four papers, and a conclusion. Although all the papers build on the same main dataset related to licensing contracts in the global pharmaceutical industry, supplementary information from different data sources was connected to the licensing contracts to answer the specific research questions. Indeed, each paper, from a different perspective, contemplates and contributes to the existing literature by 6 examining the relationship between technology licensing and specific dimensions of firm innovation. Understanding how licensing deals affect the performance of licensees and licensors is critical to understanding how markets for technology function. 7 Danish Summary Licenskontrakter repræsenterer en af de mest udbredte mekanismer til udveksling af teknologier og overførsel af knowhow mellem virksomheder. På grund af de muligheder, som licensudstedelsen skaber for virksomheder, der opererer på begge sider af markederne for teknologi, er licensudstedelsen i stigende grad blevet en integreret del af virksomhedernes F&U- strategier. På udbudssiden har den eksisterende litteratur været fokuseret på at forstå, hvordan teknologien licensudstedelse kan bruges af virksomheder som en mekanisme til at inddrive investeringer i innovative aktiviteter og fremme læringsmuligheder. På efterspørgselssiden er det blevet påvist, at licensudstedelse er en vigtig kilde, som virksomhederne kan bruge til at brødføde deres interne behov for innovativ viden. Mens flere studier har undersøgt teknologilicensudstedelse fra licensgiverens perspektiv, er der behov for mere forskning, der undersøger, hvordan virksomhederne er afhængige af licensaftaler for at tilegne sig viden og forbedre deres innovationsresultater. Derudover har der med få undtagelser været nok opmærksomhed på de organisatoriske eller kontraktmæssige egenskaber i relation til de licensaftaler, som determinanter for kapaciteten af den overtagende virksomhed. Formålet med denne afhandling er at undersøge forholdet mellem teknologilicenser og virksomhedsinnovation. Dette bliver gjort ved at undersøge de særlige kendetegn ved den overtagende virksomhed og anvendelse af særlige kontraktbestemmelser, som påvirker dette forhold. Essayene i denne afhandling bygger på et andet sæt af teoretiske perspektiver forbundet med licenslitteraturen. Afhandlingen består af en generel introduktion, fire essays og en konklusion. Alle essayene bygger på det samme datasæt, der omhandler licensaftaler i den globale farmaceutiske industri, men supplerende oplysninger fra forskellige 14 by 63% between 1996 and 2006 (OECD, 2009). Furthermore, a 2003 OECD survey covering firms located in Europe, North America, and Asia-Pacific revealed that almost 60% of the firms in the sample reported a significant increase in licensing activities during the 1990s (Arora & Gambardella, 2010). This consistent expansion in licensing activities has major implications for firms’ corporate strategies (Arora, Fosfuri, & Gambardella, 2001), dissemination of new technologies (Arora and Fosfuri, 2003), and the way that the production and use of technologies are organized between firms (Ceccagnoli & Jiang, 2013). Previous studies have examined several reasons that firms on both sides of markets for technology (technology suppliers and buyers) have to engage in technology licensing. On the supply side, it has been shown that through licensing contracts firms are able to generate significant income (Arora et al., 2001), benefit from learning opportunities (Leone & Reichstein, 2012) and maximize return on investment in R&D activities (Atuahene-Gima, 1993). In reality, explaining the reasons behind firms’ decisions to trade their technologies has been the main focus of the extant licensing literature, with several papers approaching technology licensing under the lens of the licensor. Nevertheless, a small number of studies focusing on the demand side of markets for technology has provided consistent evidence that licensing can be used by the acquiring firm to speed the innovation process (Leone & Reichstein, 2012), gain strategic flexibility (Ceccagnoli & Jiang, 2013), and explore new technological areas (Laursen, Leone, & Torrisi, 2010). Although those studies have shed light on important dimensions of technology licensing, several questions concerning the way that firms manage their licensing activities and its implications for firm performance and strategy remain unaddressed. Looking at some of those questions, the essays that follow examine four main points related to technology licensing and firm innovation: 15 1) How do individual and group level characteristics within firms affect the ease of knowledge absorption and recombination of licensed-in technologies? 2) What is the effect of recoverable slack and organizational myopia on the firm’s capacity to deal with licensed-in technologies? 3) How do the characteristics of licensed technologies (e.g., unfamiliarity, complexity and uncertainty) affect firm performance? 4) Under what circumstances will certain contractual clauses related to the evolution and application of the licensed technologies be used in licensing contracts? While the connecting point of the four papers lies in technology licensing literature, the papers in this dissertation also use different theoretical perspectives to integrate technological licensing within different analytical frameworks. The first paper focuses on the licensee’s point of view, and builds on the absorptive capacity and network analysis literatures to examine the effects of network structure and composition on firm capacity to deal with the challenges of unfamiliarity related to licensed-in technologies. The second paper, also focusing on the licensee’s perspective, builds mainly on the organizational learning literature to propose that by engaging in technology, licensing-in firms can increase their capacity to produce innovations mainly due to learning effects resulting from the access to new knowledge. The third paper considers both the licensee and the licensor within the same analytical framework, building on the extended resource based view of the firm to predict firm behavior in terms of contractual preferences. Finally, the fourth paper follows the classic tradition in the licensing literature and builds on industrial economics studies to look at the competitive implications experienced by licensors commercializing core-technologies. 16 In order to approach the different research questions empirically, the four essays rely on the same main dataset: Recombinant Capital’s Biotech Alliance (Recap). This database is one of the most accurate sources of information regarding partnerships and technology exchange in the pharmaceutical industry (Audretsch & Feldman, 2003; Schilling, 2009). More specifically, this database offers the possibility to access the original licensing contracts, from which it was possible to extract precise information regarding the characteristics of the licensed technologies, contractual specifications, and information related to the identification of licensees and licensors (e.g., firm name, address, and operating segment). Furthermore, the fact that the licensing deals are restricted to the Pharmaceutical industry allowed me to investigate in more detail the licensing dynamics specifically related to knowledge creation. There are at least three main reasons that make this context (pharmaceutical firms) particularly appropriate to test the hypotheses proposed in the four essays. First, the pharmaceutical industry is characterized as technology driven and R&D intensive, which makes technological knowledge a critical component to develop and sustain competitive advantages (Roberts, 1999). Second, R&D collaboration with other firms and universities represents an important driver of technology development (Arora & Gambardella, 1990). Third, firms in this industry routinely and systematically protect and document their inventions through patents (Hagedoorn & Cloodt, 2003). In connection with the last point mentioned above, the strong reliance that pharmaceutical firms have on patents to protect innovation was particularly important for me to successfully integrate the Recap database with other databases related to firm patenting activity. The novel combination of existing databases was critical for me to aim at addressing questions that previous studies left open. For example, using patent data allowed me to reconstruct intrafirm inventor networks, which made it possible to connect technology licensing to different 17 analytical levels within firms. Indeed, the first paper of this dissertation is the first attempt that I am aware of at considering the effect of licensing on firm innovation in light of group level characteristics within the acquiring firm. Beyond patenting activity, additional data related to firm-level information was obtained from COMPUSTAT. This database was important to obtain consistent financial information about the firms listed in the licensing database. In total, the papers in this dissertation explored four different data sources. 1.1 Papers on Technology Licensing and Firm Innovation Several specific characteristics found in technology licensing contracts make them a particularly useful mechanism for investigating the relationship between external knowledge acquisition and firm innovation. For example, an important and distinguishing characteristic of licensing contracts regards the contractual nature of the exchanged knowledge. While ex-ante contracts such as research alliances involve higher uncertainty about their potential outcomes, in ex-post contracts such as licensing, the traded technology can be more easily defined. Furthermore, most licensing contracts signed between firms involve technologies that have already been proven (Atuahene-Gima, 1993; Leone & Reichstein, 2012). These characteristics were important for the development of the papers in this dissertation, both empirically and conceptually. On the empirical side, the fact that licensing contracts usually trade well-defined and identifiable technologies was fundamental for me to compute most of the measures used to test the proposed hypotheses. On the conceptual side, this dissertation sheds light on important dimensions of knowledge acquisition such as the use of contractual instruments to shape knowledge flows between firms that would not be easily developed without the analytical framework that is provided by technology licensing contracts. Accordingly, the four papers summarized below 18 develop conceptual and empirical applications of different phenomenon of interest for innovation scholars within the scope of technology licensing. 1.1.1 Essay 1: All for one and one for all: How intrafirm inventor networks affect the speed of external knowledge recombination The first paper of this dissertation examines a relatively unexplored dimension of knowledge acquisition regarding the speed with which firms are able to recombine external and internal knowledge in order to produce innovations. Despite the fact that innovation speed has been suggested to be critical for firms to establish first-mover advantages, achieve or sustain technological leadership and overtake rivals, very few empirical studies have aimed at understanding the speed dimension related to recombination of external knowledge (e.g., Leone & Reichstein, 2012; Tzabbar et al., 2013). This paper also offers a relevant contribution to the absorptive capacity research by focusing on the interaction patterns between individuals within firms as an appropriate unit of analysis to understand the process of knowledge recombination at the firm level. We find that the higher the unfamiliarity with the licensed technology, the longer it takes the acquiring firm to successfully recombine it with internal knowledge elements. However, the results also indicate that structural and compositional characteristics of the firms’ intra inventor networks related to diversity and closure ameliorate the negative effects of unfamiliarity on recombination speed. This finding provides insight into how individual-level network formation affects a firm’s ability to quickly recombine and integrate external knowledge. In particular, it theoretically and empirically substantiates the idea that intraorganizational inventor networks affect the speed dimension of a firms’ absorptive capacity. 19 1.1.2 Essay 2: A Longitudinal Study of the Influence of Technology Licensing on Firm Innovation: The Moderating effect of Slack and Organizational Myopia The second paper in this dissertation focuses on how licensing-in of technology affects firms’ subsequent capacity to produce innovations. Although licensing has repeatedly been acknowledged to be a major vehicle for firms to acquire external knowledge (Ceccagnoli & Jiang, 2013), surprisingly little is known about how firms use licensing as part of their overall inventive efforts. Furthermore, with the exception of absorptive (Laursen et al., 2010) capacity, the organizational determinants that facilitate or constrain firms' ability to deal with licensed-in technologies have received little attention. This paper starts investigating in a longitudinal setting the effect of technology licensing on the number of patents produced by the licensee within the three years subsequent to the technology acquisition. I further develop the idea that licensing is a complementary part of firms’ innovation efforts using organizational learning lenses. The findings indicate that technology licensing is positively related to the number of inventions produced by the licensee in the years subsequent to the licensing deal. Subsequently, I investigate the moderating effect that organizational slack and myopia have on this main relationship. The findings also suggest that high levels of Organizational Slack (available financial resources) strengthen the positive effect of licensing on innovation. However, higher levels of Organizational Myopia (the extent to which a firm draws on its own knowledge) can decrease the main effect of licensing. Those findings go in the same direction of previous studies that have suggested the relationship between knowledge acquisition and firm innovation cannot be taken for granted but should be considered in light of specific organizational characteristics. 20 1.2.3 Essay 3: Exploring the boomerang effect: The role of core technologies and uncertainty in explaining the use of the grant-back clause in technology licensing The third paper concerns the use of the technology-flow back provision (grant-back clause) in technology licensing. This clause has been described in previous studies as a relevant contractual specification with significant implications for both licensor and licensee. Despite the potential mutual benefits for firms from entering into licensing deals, previous studies have also indicated that licensing might create undesirable competition due to the transfer of the firms’ knowledge capabilities related to cutting-edge technologies (Choi, 2002). In this context, the grant-back clause can be used as an instrument to ensure that the licensee will grant back to the licensor the rights to any improvement in the licensed technology (Schmalbeck, 1974). In other words, the grant-back clause can have substantial influence on the nature and amount of knowledge that will be transferred between the firms after entering in a deal. Despite this evidence, to the best of my knowledge no previous empirical study has attempted to explain the conditions under which this clause is used. This paper looks into the contingencies related to the technological aspects of the licensing deal that make it more or less likely that the grant-back clause will be used in the contract. One of the main features of this paper regards the development and empirical testing of a theoretical approach that integrates the extended resource based view of the firm into contract theory. The findings suggest that the extent to which the licensing deal involves core technologies (to the licensee and the licensor) and the uncertainty related to its future trajectory are significant predictors for this use of this clause in licensing deals 21 1.1.4 Essay 4: Understanding the Rent Dissipation Effect in Technology Licensing Contracts The fourth paper introduces and empirically tests a framework to understand the profit dissipation experienced by firms that license out their technologies. While the generation of revenues is an important incentive for firms to license out, granting other firms access to relevant technologies can also produce negative implications for the licensor’s competitiveness (Choi, 2002). Along this line, rent dissipation is a phenomenon related to the increasing competition that licensors with downstream assets in the product market might experience in the periods subsequent to the licensing deal (Fosfuri, 2006). The fact that the licensee can use the acquired technology to improve its own internal capabilities and become an aggressive competitor has been repeatedly suggested in conceptual papers dealing with this issue. Accordingly, the importance of this phenomenon lies of the fact that in several sectors the market for inventions might remain underdeveloped, given the licensor’s concerns about undermining its competitive advantages. This paper aims at explaining the dissipation effect experienced by licensors using a perspective that incorporates three important dimensions of the markets for technology: 1) whether the licensors possess downstream assets, 2) licensee size, and 3) technological overlap between the licensor and the licensee. The results lend empirical support to the idea that licensing out core technologies is negatively related to subsequent changes in the licensor’s market share. I also find that as the licensee size increases the dissipation effect is strengthened. However, if the licensee and the licensor operate in different technological areas (i.e., they are technologically fragmented) the negative effect caused by licensing core technologies becomes weaker. Those findings add to the current literature by discussing the importance of also taking into account the licensee’s characteristics as a way to understand more comprehensively the dissipation effect. 22 1.2 Contributions and Implications This dissertation as a whole contributes to the innovation and licensing literatures in different respects. Generally speaking, it provides empirical and theoretical insights into how firms use technology licensing to feed their demands for external knowledge. This dissertation also provides an overview of licensing practices related to the way that contractual clauses can be used to shape the incentives that those involved have to enter into licensing deals. More than that, it also looks at the motives that firms on both sides of markets for technology have to buy and sell technologies through licensing contracts. In fact, by mainly focusing on the demand side of markets for technology, the papers in this dissertation join a growing body of literature suggesting that so far little is known about the determinants of firms’ decisions to license in (e.g., Laursen et al., 2010; Leone and Reichstein, 2012; Ceccagnoli & Jiang, 2013). Accordingly, in all of the four papers in this dissertation the licensee perspective is incorporated into the conceptual and empirical analysis. 23 1.3 References Anand, B. N. & Khanna, T. 2000. Do firms learn to create value? The case of alliances. Strategic Management Journal, 21(3): 295-315. Arora, A. & Gambardella, A. 1990. Complementarity and External Linkages: The Strategies of the Large Firms in Biotechnology. Journal of Industrial Economics, 38(4): 361-379. Arora, A. 1995. Licensing tacit knowledge: intellectual property rights and the market for know-how. Economics of Innovation and New Technology, 4(1): 41-60. Arora, A., Fosfuri, A., & Gambardella, A. 2001. Markets for Technology and their Implications for Corporate Strategy. Industrial & Corporate Change, 10(2): 419-451. Arora, A. & Gambardella, A. 2010. Ideas for rent: an overview of markets for technology. Industrial & Corporate Change, 19(3): 775-803. Atuahene-Gima, K. 1993. Determinants of inward technology licensing intentions: An empirical analysis of Australian engineering firms. Journal of Product Innovation Management, 10(3): 230-240. Audretsch, D. & Feldman, M. 2003. Small-Firm Strategic Research Partnerships: The Case of Biotechnology. Technology Analysis & Strategic Management, 15(2): 273-288. Blundell, R., Griffith, R., & Reenen, J. V. 1999. Market Share, Market Value and Innovation in a Panel of British Manufacturing Firms. The Review of Economic Studies, 66(3): 529-554. Cassiman, B. & Veugelers, R. 2006. In Search of Complementarity in Innovation Strategy: Internal R&D and External Knowledge Acquisition. Management Science, 52(1): 68-82. Ceccagnoli, M. & Jiang, L. 2013. The cost of integrating external technologies: Supply and demand drivers of value creation in the markets for technology. Strategic Management Journal, 34(4): 404-425. Cefis, E. & Marsili, O. 2005. A matter of life and death: innovation and firm survival. Industrial and Corporate Change, 14(6): 1167-1192. Chesbrough, H. W. 2003. Open innovation : the new imperative for creating and profiting from technology. Boston, Mass.: Harvard Business School Press. Choi, J. P. 2002. A dynamic analysis of licensing: The "boomerang" effect and grant-back clauses. International Economic Review, 43(3): 803-829. Cohen, W. M. & Levinthal, D. A. 1990. Absorptive Capacity: A New Perspective on Learning and Innovation. Administrative Science Quarterly, 35(1): 128-152. Fosfuri, A. 2006. The licensing dilemma: understanding the determinants of the rate of technology licensing. Strategic Management Journal, 27(12): 1141-1158. 30 position and first-to-market successes (Kessler & Chakrabathi, 1996). We also add a complementary perspective to prior work on social networks as the locus of recombination (Carnabuci & Operti, 2013; Guler & Nerkar, 2012; Nerkar & Paruchuri, 2005; Phelps, Heidl, & Wadhwa, 2012) which has mostly examined internal knowledge recombination. Our study highlights the function that intrafirm networks serve in recombining external knowledge. Finally, we add to research on the role of intraorganizational social networks for overall firm innovation outcomes (Kleinbaum & Tushman, 2007). 2.2 Theory and Hypotheses An invention is the outcome of a search process that involves problem-solving by inventors and eventually, recombination of existing knowledge components in a novel manner (Fleming, 2001; Hargadon & Sutton, 1997; Schumpeter, 1934). The invention process has shifted from taking place solely within the firm to a more open model in which firms acquire knowledge from a variety of sources (Chesbrough, Vanhaverbeke, & West, 2006; Laursen & Salter, 2006). Acquisition of external knowledge facilitates firm invention due to the complementarity between externally and internally generated knowledge components (Cassiman & Veugelers, 2006). Firms do not have all relevant knowledge in-house and therefore engage in alliances, licensing, and hiring to update their R&D process (Arora & Gambardella, 1990; Levin et al., 1987). The process of knowledge recombination thus increasingly relies on the recombination of both internal and external knowledge components. In this respect, Cohen & Levinthal (1990) argue that firms vary in the ability to draw on external knowledge. The absorptive capacity of firms refers to the ability to recognize, assimilate, and exploit external knowledge and “is largely a function of the level of prior related knowledge” (Cohen & Levinthal, 1990: 128). 31 According to the knowledge-based theory of the firm, knowledge is collectively stored among employees and firms can be seen as social communities (Kogut & Zander, 1996; Matusik & Heeley, 2005). Social communities are the origin of knowledge creation and knowledge transfer within the firm (Tsai, 2000, 2001). In a similar manner, the literature on organizational learning asserts that learning involves knowledge transfer among individuals and business units within the firm (Argote, Mcevily, & Reagans, 2003; Huber, 1991). Organizations can thus be understood as network arrangements (Brass, Galaskiewicz, Greve, & Tsai, 2004; Reinholt, Pedersen, & Foss, 2011; Tsai, 2001). Networks among employees, and especially those individuals that are active in a firm’s R&D process, inventors, influence the extent to which knowledge is diffused and generated within a firm (Guler & Nerkar, 2012; Nerkar & Paruchuri, 2005). Intrafirm social networks can be seen as an antecedent of a firm’s absorptive capacity (Volberda et al., 2010) because intrafirm networks shape knowledge flows among individuals and determine the efficiency of communication between them. Relevant knowledge for problem-solving is distributed among individuals within the firm (Lenox & King, 2004) and can be detected and shared through networks (Brass et al., 2004; Turner & Makhija, 2012). To illustrate this, Nerkar & Paruchuri (2005:773) argue that “bounded rational inventors search across the internal knowledge network on the basis of incomplete information about which knowledge should be recombined”. Networks among inventors also constitute communication patterns. The efficiency of communication (Cohen & Levinthal, 1990) refers to inward-looking absorptive capacity and determines the effectiveness of internal sharing of external knowledge (Volberda et al., 2010). In this sense, intrafirm inventor networks influence firm innovation through sharing, development, and recombination of external knowledge. As a consequence, interpersonal networks can be seen as an antecedent of a firm’s capacity to deal with external 32 knowledge, constituting the micro-foundations of a firm’s inventive capabilities (Allen & Cohen, 1969; Brown & Duguid, 2001; Tushman & Scanlan, 1981). The use of external knowledge in a firm’s R&D process may shorten the time of the invention process (Kessler & Chakrabathi, 1996; Leone & Reichstein, 2012). Speeding up the invention process is crucial to consolidate the competitive position of firms. Yet, the effect of external knowledge acquisition on subsequent invention speed depends on the channel through which external knowledge is acquired (Lee & Allen, 1982; Tzabbar, Aharonson, & Amburgey, 2012; Vasudeva & Anand, 2011) and a firm’s absorptive capacity (Cohen & Levinthal, 1989, 1990). In this paper we examine the influence of specific intrafirm network configurations of inventors on the speed with which a firm integrates and recombines externally acquired knowledge. We define external knowledge recombination speed as the time it takes a firm to recombine externally acquired knowledge into the firm’s own invention. In the next paragraphs we develop hypotheses on how structural and compositional features of intrafirm networks among inventors affect the recombination speed of external knowledge. Technological distance and recombination speed. Firms acquire external knowledge to complement their own technological knowledge base. In fact, in order to fill in the gaps related to the lack of specific knowledge components, firms tend to reach out for technologically distant knowledge (Rosenkopf & Almeida, 2003). Yet, we argue here that even though firms are prone to engage in distant knowledge sourcing, this comes at a cost with regard to recombination speed. The ease with which firms recombine external knowledge hinges upon having related prior experience with the acquired knowledge (Cohen & Levinthal, 1990; Zahra & George, 2002). Prior experience becomes the natural starting point for subsequent searches for new knowledge, and a firm’s knowledge stock, which is accumulated over the years, is used as a lens through which the firm makes sense of knowledge from the environment (Rosenkopf & 33 Almeida, 2003). The technological development of a firm over time thus affects the technological distance between a firm’s knowledge base and external knowledge. Assimilation of external knowledge requires a common base of understanding, or overlap in the knowledge base, in order to achieve successful application of this piece of knowledge (Cohen & Levinthal, 1990). As a result, when the technological distance between the firm’s knowledge base and acquired external knowledge increases, the absorptive capacity of a firm declines (Gilsing, Nooteboom, Vanhaverbeke, Duysters, & Vandenoord, 2008; Lane & Lubatkin, 1998). This means that the cost and effort to recombine external knowledge increases with distance (Leone & Reichstein, 2012; Weitzman, 1998). To illustrate this, integration of distant external knowledge will require more effort and time as inventors in the firm are likely to encounter problems when they deal with unfamiliar knowledge. The solution generation process will subsequently prolong the time it takes for the firm to recombine distant external knowledge into an invention. Consequently, a firm requires more time to understand distant knowledge and may need more time to invest in its absorption, and this will slow down the process of external knowledge recombination. Our baseline hypothesis therefore states: Hypothesis 1. The larger the distance between the externally acquired knowledge and the firm’s knowledge base, the longer it takes the firm to recombine external knowledge Intrafirm network density and the recombination speed of distant external knowledge. Dense networks (also called cohesive or closed networks) are networks in which the members are wellconnected with each other. From an innovation perspective, previous studies have indicated that network density may either be beneficial or harmful for firm innovation (Burt, 1992; Coleman, 1988). On the one hand, network density leads to knowledge-sharing among members of the network and fosters information flow through the network (Gargiulo, Ertug, & Galunic, 2009; Obstfeld, 2005; Reagans & McEvily, 2003). Furthermore, dense networks are likely to have 34 effective norms, promote trust (Coleman, 1988), and facilitate the exchange of tacit and complex knowledge (Hansen, 1999; Hansen, Podolny, & Pfeffer, 2001; Uzzi, 1997). On the other hand, the opposite of a dense network, a sparse network, may also be effective for firm innovation (Burt, 2004). A sparse network, which features structural holes between clusters or sub-networks, enhances firm innovation through the likelihood that such a network structure exhibits diverse information and fosters creativity. Although sparse networks have been shown to be associated with high levels of heterogeneity, which facilitate the creation of new knowledge, the absence of connections between the network members reduces the speed with which individuals can share knowledge and access information (Singh et al., 2010). In fact, even though knowledge heterogeneity is important for inventors to deal with unfamiliarity, existing ties are necessary to provide individuals the right channels to tap into each other’s experience and knowledge. This is particularly true for intrafirm networks, given that relevant knowledge might exist within the firm boundaries and still remain unutilized if network configurations do not favor its detection and dissemination (Hansen, 1999). Therefore, we claim that intrafirm network density is particularly relevant to firms’ ability to quickly recombine and eventually integrate distant external knowledge. Intrafirm inventor network density shortens the time it takes to recombine distant external knowledge for at least three reasons. First, dense networks ease the search for and detection of relevant knowledge available in the network of inventors. Through their ties, inventors may hear about and observe potentially relevant inventors with the knowledge and skills needed to recombine distant external knowledge. Thus, dense networks tend to speed up the search time for relevant information within the network (Zaheer & Bell, 2005). Second, dense inventor networks tend to encourage knowledge sharing and the willingness to devote time and effort to support peers (Reagans & McEvily, 2003). Such cooperative behavior is likely to create 35 cooperative norms and fosters knowledge transfer between inventors in the firm. For this reason, one may expect that the prolonged recombination time inherent to distant knowledge tends to be shorter in dense networks as a result of a mutually supportive environment. Third, network density promotes the formation of norms, which, in turn, enhances mutual understanding between inventors and lowers the possibility of misinterpretation and loss of relevant information (Reagans & McEvily, 2003; Zaheer & Bell, 2005). Inventors in dense networks thus tend to save time due to the formation of successful communication routines. In line with our predictions, we claim that firms with a dense intrafirm co-invention network experience a shorter recombination time for distant external knowledge. Our second hypothesis thus states the following: Hypothesis 2. Firms with an intrafirm inventor network that has a high level of network density recombine distant knowledge faster than firms with an intrafirm inventor network that has a low level of network density Intrafirm average tie strength and recombination speed of distant external knowledge. Tie strength refers to the intensity of interaction between two members of the network and is “a combination of the amount of time, the emotional intensity, the intimacy (mutual confounding) and the reciprocal services which characterize the tie” (Granovetter, 1973: 1361). Tie strength characteristics tend to increase with increasing frequency of collaboration between inventors. Tie strength promotes trust and facilitates knowledge transfer, especially knowledge that is complex and tacit (Hansen, 1999; Levin, Walter, & Murnighan, 2010; McFadyen et al., 2009). While weak ties help in the search of useful knowledge it also impedes individuals to exchange complex information, limiting the extent to which complex knowledge flows within the network (Hansen, 1999). In fact, Hansen (1999) points out that, particularly in the case of innovation, useful knowledge may fail to be appropriately shared among individuals even though 36 information regarding the whereabouts of the knowledge is disseminated across the network. This argument emphasizes the need of strong ties in order to individuals’ knowledge and expertise to move from one point to another in the network. Strong ties among inventors within a firm are likely to mitigate disadvantages related to integrating distant external knowledge according to two main arguments. First, trust and knowledge-sharing among inventors increases with recurring interaction (Hansen, 1999; Reagans & McEvily, 2003). This, in turn, increases the willingness of inventors to spend more time and effort on supporting each other (Rost, 2010; Seibert, Kraimer, & Liden, 2001; Sosa, 2010), for example in problem-solving related to the integration of unfamiliar pieces of knowledge. Second, knowledge that is tacit and highly complex is better transferred through strong ties (Hansen, 1999; Phelps et al., 2012). Distant knowledge is likely to be a complex matter for inventors within the firm, and therefore, tie strength increases the likelihood that such complexity is shared throughout the firm, which accelerates the integration process (Hansen, 1999). Taken together, we expect that high average tie strength will shorten the recombination process of distant knowledge and we therefore posit the following hypothesis: Hypothesis 3. Firms with an intrafirm inventor network that has high average tie strength recombine distant knowledge faster than firms with an intrafirm inventor network that has low average tie strength Intrafirm network diversity and recombination speed of distant external knowledge. Network diversity refers to the diversity of resources available in the network. Or, in other words, the extent to which network connections span boundaries (Reagans & McEvily, 2003). In the context of this paper, network diversity refers to variety in technological experience among the collaborating inventors inside the firm (Harrison & Klein, 2007) or the extent to which inventor ties span technological boundaries. Network diversity or range increases knowledge sharing among members of the network (Reagans & McEvily, 2003) and promotes the problem-solving 37 ability of members through access to diverse resources available in the network (Phelps, 2010). An intrafirm network composed of a diverse group of inventors will accelerate the time it takes to recombine distant external knowledge for at least three reasons. First, due to the inherent uncertainty of knowledge recombination, inventors benefit from having diverse partners in their intrafirm network. Diverse connections provide a single inventor with access to a diverse set of problem-solving heuristics (Page, 2007) and support the accomplishment of complex tasks related to recombining distant knowledge (Mors, 2010; Rodan & Galunic, 2004). Thus, the collective problem-solving ability of inventors increases with diversity and shortens the time it takes to recombine complex distant knowledge acquired from outside the boundaries of the firm. Second, when inventor with different technological backgrounds collaborate they expand their ability to convey knowledge across distinct bodies of meta-knowledge (Reagans & McEvily, 2003; Tortoriello, Reagans, & McEvily, 2012). Over time, building experience in interacting with dissimilar colleagues increases inventors’ capability to efficiently and successfully frame their communication with other inventors, which, in turn, may accelerate the recombination of distant knowledge based on future interactions among heterogeneous inventors. Third, diversity within the intrafirm network increases the likelihood of overlap between the acquired external knowledge component and available relevant knowledge already existent in the intrafirm coinventor network (Cohen & Levinthal, 1990). Diversity among collaborating inventors thus eases the comprehensibility of distant external knowledge and leads to shorter recombination time. Our final hypothesis therefore states: Hypothesis 4. Firms with an intrafirm inventor network that has a high level of network diversity recombine distant knowledge faster than firms with an intrafirm inventor network that has a low level of network diversity 38 In short, we posit that while technological distance prolongs the time it takes to recombine external knowledge into own invention, network density, average tie strength and diversity shorten the recombination process of distant knowledge pieces1. 2.3 Data and Methods We test the aforementioned hypotheses in the context of the global pharmaceutical industry. Firms in this industry develop and commercialize drugs, chemical components, and biological products. The focus on pharmaceutical firms provides a good research context for at least four reasons. First, the pharmaceutical industry is characterized as technology driven and R&D intensive, which makes technological knowledge a critical component to develop and sustain competitive advantages (Roberts, 1999). Second, firms in this industry routinely and systematically protect and document their inventions (Hagedoorn & Cloodt, 2003). In particular, patenting is an important and common mechanism used in this industry (Levin et al., 1987). Since patents provide reliable documentation of a firm’s innovative activities we rely on patent information to identify the technological profile of the firms in our sample (Roberts, 1999; Adegbesan and Higgins, 2010; Hoang and Rothaermel, 2010). Third, R&D collaboration with other firms and universities represents an important driver of technology development (Arora & Gambardella, 1990). Indeed, firms in this industry actively engage in external knowledge or technology acquisition to foster their own inventive activity. Finally, the pharmaceutical industry has proven to be a valuable context to identify and measure the effect of inventor networks on innovative output (Paruchuri, 2009). 1 We acknowledge the fact that prior work has identified costs related to excessive network density and diversity in particular (Phelps, 2010). We address this issue empirically in the section on robustness checks and theoretically in the discussion section. 39 The data used in this study derive from four data sources. First, we used detailed information on licensing agreements from the Deloitte Recap Database, which covers licensing deals in the global pharmaceutical industry for the period 1983 – 2008. This database is one of the most accurate sources of information regarding partnerships and technology exchange in the pharmaceutical industry (Audretsch & Feldman, 2003; Schilling, 2009). More specifically, this database allowed us to access the original licensing contracts, from which it was possible to extract precise information regarding the date of the licensing event, characteristics of the licensed technologies, contractual specifications, and information related to the identification of licensees and licensors (e.g. firm name, address and operating segment). Second, we drew on the NBER patent project to merge the specific patent numbers connected to the traded technologies from the Deloitte Recap Database with patents registered at the United States Patent and Trademark Office (USPTO). Furthermore, the information retrieved from the NBER project was used to identify the technological profile of the firms that acquire technologies through licensing (i.e. licensees), and the firms that sell the technologies (i.e. licensors). Therefore, we were able to include in the analysis variables capturing the characteristics of firms on both sides of the licensing contract, allowing us to disentangle potentially confounding firm effects from the variables of interest. Third, we relied on the Harvard Patent Network Dataverse, which provided us with the disambiguated inventor names and inventor identification numbers. This allowed us to construct intrafirm inventor networks based on co-invention as well as to derive inventorlevel information. Prior research has used qualitative evidence (i.e. interviews) to validate copatenting ties as a measure of collaboration among inventors (Carnabuci & Operti, 2013; Fleming, King III, & Juda, 2007). Finally, we utilized the WRDS Compustat database mainly for control variables. The final sample consists of 113 firms involved in the acquisition of 708 USPTO patents using licensing contracts. Given that the information regarding inventors’ patenting 46 references listed in the backward citations of the licensed technology as a way to capture crosstechnology differences in terms of the development stage. The final set of control variables is related to the licensor’s characteristics. First, we control for the number of successfully applied patents that the licensor filed in the seven years prior to the licensing contract as the licensor’s size and technological capabilities may also affect the licensee’s willingness to quickly invent using the licensed technology. Second, in order to control for differences between firms and universities as licensors we added a dummy variable to identify the contracts in which the licensor is a university. Finally, following the convention in this literature, we added sector dummies indicating the segment within the pharmaceutical firm in which the licensee operates and year dummies. 2.4 Model Specification and Estimation Given that the hypotheses refer to the time it takes to recombine knowledge, we generated the dependent variable following an event history analysis structure. This type of model is conventionally used to examine the conditional probability that an event occurs in a particular time interval (t) (Blossfeld, Golsch, & Rohwer, 2007; Yu & Cannella, 2007). In this respect, we apply event history analysis to model the time taken, T, between the licensing date and the first time the licensing technology is cited by the licensee in a new patent. The use of event history analysis to investigate the effect of the explanatory variables on the time it takes to recombine knowledge offers at least two major advantages. First, it makes it possible to directly model time as the dependent variable without the need to transform it into a discrete outcome (Pennings & Wezel, 2009). Second, this technique also allows for modeling the observations that do not experience the transition during the time frame covered by the data by dealing with issues emerging from right-censoring as a non-random process (Blossfeld et al., 2007). Compared to alternative model specifications (e.g. logit or OLS), employing event history analysis allows us 47 to include the observations for which we only have partial information, which covers the time they enter the sample (the licensing date) until the last date that patent data for backward citation are available. In order to decide among the possible models within even history analysis we considered the underlying mechanisms driving the hazard to knowledge recombination. We expect that firms that license-in technologies with low distance will be able to recombine the new knowledge with existing components at a rapid pace, which increases the hazard to knowledge recombination as the time increases. However, as the time elapses, the technologies with lower distance exit the sample, leaving in the sample technologies that take more time to be recombined. This effect is expected to become dominant and lowers the hazard rate until a point at which the hazard function starts to decline. Accordingly, we decided to employ a log-logistic model as a way to accommodate the expected process of an initial increase followed by a decreasing rate (Mills, 2011). Alternatively, we also employed a log-normal specification as a robustness check and, as expected, both models produced comparable results. Considering that the capacity to deal with distant knowledge is likely to be also determined by firm characteristics that are not captured by the explanatory variables used in the econometric model, we correct for potential endogeneity issues originating from the presence of unobserved heterogeneity across the firms. Prior studies using a similar setting to the one presented in this paper have dealt with unobserved firm-level differences affecting duration dependence by employing frailty estimators (e.g. Hoang & Rothaermel, 2010; Pennings & Wezel, 2009; Polidoro, Ahuja, & Mitchell, 2011). Following the recommendation by Blossfeld et al. (2007), we model the unobserved heterogeneity using a shared gamma mixture specification associated with the log-logistic model. The alternative to the use of a gamma mixture model would be the inverse Gaussian frailty model, but as demonstrated by Jenkins (2005), it is straightforward to assume a gamma or normal distribution for the frailty of log- 48 logistic models. The inclusion of a gamma mixture refers to the incorporation of an “error term” in the model that relates multiplicatively to the hazard rate for each firm in the analysis (Blossfeld et al., 2007; Hougaard, 1986). Additionally, the use of shared frailty also offers the possibility to model intragroup correlation, which in the case of our sample is created from repeated group observations (Gutierrez, 2002). 2.5 Descriptive Statistics and Correlations Table 1 reports the means, standard deviations, and Pearson correlation coefficients of the variables used in the analysis. The results raised no concerns regarding collinear variables, except for the correlations between Average path length with Network density and Clustering with Average Tie Strength. The moderate correlations between those variables are in line with theoretical expectations, but in order to check for potential bias we entered the variables in a stepwise manner and the results for the main explanatory variables do not change as the variables enter the model. Additionally, the maximum variance inflation factor (VIF) associated with any of the independent variables was 4.34 (mean VIF = 2.15), which is well below the rule- of-thumb value of ten (Gujarati, 1995). In order to identify potential model estimation issues regarding the stability of the coefficients and standard error we also added the main explanatory variables one at a time. Finally, the likelihood ratio comparison test at the bottom of Table 2 indicates that models II – V provide significant improvement relative to the baseline model. Looking specifically into the likelihood ratio comparison for model V (likelihood ratio: 35, df: 4, p<0.001) we observe a substantial improvement compared to the restricted model. [Insert Table 1 around here] We were able to track the patenting behavior of the firms in our sample until December 2006; therefore, our analysis is censored at the latest dates available in the patent citation data. 49 Looking into the knowledge recombination speed, the longest time to transition for the firms in our sample was 168 months. Out of 708 firm-technology observations, a total of 116 firms cited the licensed technology in a new patent (made the transition) during the time frame of our analysis. For the observations that experienced the transition, the average time for knowledge recombination was 25 months. In contrast, the average time of at-risk months for all firms in the sample (including censored observations) was 74 months. Considering the average time for knowledge recombination between the uncensored observations with high versus low technological distance (using mean values), small distance technologies are, on average, cited within 24 months, while large distance technologies are cited within 83 months. Among the 592 firm-technology observations that did not experience the transition during the time window of our analysis, 129 observations exit the sample earlier than December 2006. These observations were subject to a different type of right-censoring. In the empirical setting used in this paper these observations exit the sample earlier because their latest records on COMPUSTAT ended earlier than the latest information available in the patent data. We modeled those observations differently by setting the exit time at the date of the latest Compustat record, implying that although these observations exit the sample, they do not experience the transition. The fact that the financial records for a given firm are discontinued is likely to be due to bankruptcy or an M&A process, which eliminates the possibility of a firm being observed in the patent citation data5. To supplement, we plot the cumulative hazard function after the estimation of the log-logistic model to visualize the patterns of the hazard function regarding the non-monotonic shape. Indeed, the results (see figure 1) indicate an initial increase followed by a decrease in the hazard rate for the observations in our sample, suggesting the suitability of the log-logistic 5 If we consider those firms exiting the sample earlier, approximately 20% of the observations experience the transition within the time frame of the event history analysis 50 model specification. Additionally, in order to visualize the shape of the hazard rate for observations with high and low levels of technological distance we generated two groups on the basis of the mean values of distance. As expected, the visualization of the cumulative hazards indicates that the observations that present lower levels of distance exhibit a higher hazard rate compared to those with higher levels of distance, with the curves for the two groups exhibiting a similar non-monotonic pattern. This result offers initial support for our hypothesis regarding the effect of distance on the firm’s capacity to recombine external knowledge. As suggested in the graph, the firms dealing with lower levels of distance have a higher probability of experiencing a transition earlier compared to those dealing with high distance levels. [Insert Figure 1 around here] 2.6 Results Table 2 reports the results for the log-logistic model with the shared gamma mixture specification. The dependent variable across the six models reported in this table reflects the time gap between the licensing date and the first time the licensed technology was cited in a new patent (for the non-censored observations). Model I reports the estimators for controls and the main effects of the interaction terms. Additionally, we included year dummies to control for period effects, such as overall differences in patenting behavior in the pharmaceutical industry. In models II – VI the interaction terms capturing the relationships described in the hypotheses were entered one-by-one along with all the controls. For the sake of simplicity we will focus the discussion of the results on the full model in column VI. [Insert Table 2 around here] 51 Hypothesis 1 predicted that the larger the distance between the externally acquired knowledge and the firm’s knowledge base, the longer it takes the firm to recombine external knowledge. The coefficient for the technological distance variable is positive and significant at the 1% level when all controls are included in the equation, providing strong evidences in favor of our first hypothesis. The result lends support to the fundamental idea developed in this paper that distance (unfamiliarity) is an important predictor of a firm’s capacity to recombine external knowledge at a faster pace. This finding is similar to the results obtained by Leone & Reichstein (2012) regarding the joint effect of unfamiliarity and contractual specifications (the use of grantback clause) on the time a licensee takes to produce its first invention after a licensing contract. Hypothesis 2 stated that firms with an intrafirm inventor network that has a high level of network density recombine distant knowledge faster than firms with an intrafirm inventor network that has a low level of network density. Accordingly, the interaction term between technological distance and network density exhibits a negative and significant coefficient, indicating that the positive effect of distance on the time it takes to recombine knowledge becomes less positive (or more negative) when interacted with network density. This result supports the expected effect described in hypothesis 2. Thus, the negative and significant interaction term indicate that firms with a densely connected intrafirm inventor network are better able to deal with technological distance in a faster way. Hypothesis 3 did not find support in the results. We predicted that firms with an intrafirm inventor network that has high average tie strength recombine distant knowledge faster than firms with an intrafirm inventor network that has low average tie strength. The interaction between technological distance and tie strength did not produce significant coefficients at the conventional level. Hence, the insignificant coefficient for this interaction term indicates that distance is positively related to knowledge recombination regardless of the tie strength among the inventors within the firm. In other words, we do not find evidence of a 52 significant moderating effect of tie strength on the relationship between distance and the dependent variable. Finally, the results offered support for the moderation effect predicted in hypothesis 4 regarding the fact that firms with an intrafirm inventor network that has a high level of network diversity recombine distant knowledge faster than firms with an intrafirm inventor network that has a low level of network diversity. Accordingly, the interaction between technological distance and network diversity produced a significant and negative coefficient. This finding supports the idea that network diversity negatively moderates the relationship between distance and the time it takes to recombine knowledge and thus accelerates the recombination of distant knowledge. 2.7 Alternative Explanations and Robustness Checks Despite the large number of prior studies indicating that technology licensing leads to knowledge transfer (Arora, 1996; Ceccagnoli & Jiang, 2012; Laursen et al., 2010), we acknowledge that the link between licensing-in and patent citations has not yet been established in the literature. Therefore, we performed a robustness check to evaluate the number of citations received by a technology after and before the licensing date using a conditional difference-in- differences design (Singh & Agrawal, 2011). By doing so, we expect to strengthen the confidence in the main results by focusing on two important aspects. First, it could be argued that the licensing firm is more likely to cite a technology of relatively higher quality or relevance regardless of whether or not it licenses the technology. Accordingly, technologies with such characteristics may also be more likely to be commercialized in the markets for technology, which creates a selection problem in which backward citations do not reflect the true effect of licensing. Second, a licensee may be more likely to license a technology in a domain in which the firm is intending to expand its technological activities. Therefore, it is 53 likely that the licensing efforts would also be associated with other measures aiming to improve a firm’s access to a specific technological area. To perform the difference-in-differences we followed the steps described in the study by Singh & Agrawal (2011). First, each licensed technology in our sample was matched on the basis of propensity scores using the application year, patent class, and subclass to the closest technology in the entire technological space (USPTO patents). Second, we certified that no observation in the control group was in fact licensed by the focal firm in the sample. Third, we computed the total number of citations that the focal firm made to both groups of technologies (the treatment and control) after and before the licensing date. There were only eight observations in which the licensed technology had been cited by the licensee before the licensing contract; those observations were removed from the event history analysis but were used to estimate the difference-in-differences model. On the basis of this matching sample between licensed and non-licensed technologies sharing similar characteristics, we evaluated the change in the number of citations. The results indicate (see Table 3) a significant and substantial increase in the number of citations received by a licensed technology when the number of citations received by the technologies in the control group is taken into account. Considering the baseline period, it is observed that the patents in the control group received an average number of citations of 0.055, while the licensed technologies had on average 0.031 citations. However, considering the years after the licensing date it is possible to observe that the average number of citations for the licensed technologies increases to 1.541 while the control group remains the same. [Insert Table 3 around here] 54 Further robustness checks are not reported here because of space limitation. First, the literature on network analysis has also pointed out to limitations in the extent that increasing levels of network density and diversity can benefit knowledge sharing and diffusion within networks. This claim naturally leads to the idea that density and diversity curvilinearly moderate the effect of distance on time to knowledge recombination. We empirically investigated if that is the case by including in the log-logistic model interaction terms between Technological Distance and the squared version of our measures for Network Density and Network Diversity, the results were statistically insignificant. Second, an alternative explanation for the effect of distance on time to knowledge recombination is related to the fact that the distant technologies may not be licensed with the intention of applying them in a new invention. Therefore, it could also be suggested that our results regarding the effect of technological distance on time to knowledge recombination comes from the censored observations, for which we have only partial information. To address this concern and check the plausibility of this argument we conducted a t-test comparing the level of distance between those observations that experience the transition and those that do not during the time window of our analysis. We found no evidence of statistical significance between the two groups. 2.8 Discussion and Conclusion The present study was motivated by the fact that the absorptive capacity literature has overlooked the actions and interactions of individuals within the organization in the process of external knowledge integration. In addition, research on absorptive capacity has not paid enough attention to how quickly firms can recombine knowledge from the external environment with internal knowledge. The ability to speed-up the process of external knowledge recombination is a competitive advantage, especially in fast-paced industries. In this paper we address these 55 shortcomings and examine the influence of intrafirm inventor networks on firms’ ability to recombine external knowledge with internal knowledge into own invention. We specifically investigated how network structure and network composition within the firm affect the absorption speed of distant external knowledge. We made the argument that firms often engage in distant knowledge acquisition, yet distant knowledge requires substantial time to be devoted to recombination due to inventors’ lack of familiarity with it. By drawing on social network theory and literature on search within organizations we subsequently claimed that network density, average tie strength and network diversity shorten the time to recombine distant external knowledge with internal knowledge pieces. The empirical results indeed showed that technologically distant external knowledge prolongs the time of external knowledge recombination compared to close knowledge. More importantly, the results showed that intrafirm network density and diversity shorten the time in which firms assimilate distant external knowledge. This is in line with our predictions. Yet, our results did not support our prediction that tie strength moderates the relationship between technological distance and the speed of external knowledge recombination. We discuss our results in light of previous research on absorptive capacity and external knowledge sources. Our finding that strong average intrafirm ties among inventors do not accelerate the recombination of distant knowledge is in contrast to what we expected on the basis of the literature on knowledge-sharing within firms (e.g. Hansen, 1999; McFadyen & Cannella, 2004). Two explanations can be put forward for why this is the case. First, in addition to its benefits, tie strength can also impair the inventors’ ability to develop distant external knowledge. Recurring interaction between a pair of inventors may lead to a trustworthy relationship characterized by supportive behavior (Granovetter, 1973). Yet, inventors with a limited number of partners with whom they collaborate can become myopic and focus on a limited set of colleagues. As a result, 62 Griliches, Z. 1990. Patent Statistics as Economic Indicators: A Survey. Journal of Economic Literature, 28(4): 1661–1707. Gujarati, D. N. 1995. Basic Econometrics. New York: McGraw-Hill. Guler, I., & Nerkar, A. 2012. The Impact of Global and Local Cohesion on Innovation in the Pharmaceutical Industry. Strategic Management Journal, 33: 535–549. Gutierrez, R. G. 2002. Parametric frailty and shared frailty survival models. Stata Journal, 2(1): 22–44. Hagedoorn, J., & Cloodt, M. 2003. Measuring Innovative Performance: Is There an Advantage in Using Multiple Indicators ? Research Policy, 32: 1365–1379. Hall, B. H., Jaffe, A. B., & Trajtenberg, M. 2001. The NBER Patent Citations Data File: Lessons, Insights and Methodological Tools. Working paper no. 8498, National Bureau of Economic Research, Cambridge, MA, 1–74. Hansen, M., Podolny, J. M., & Pfeffer, J. 2001. So Many Ties, So Little Time: A Task Contingency Perspective on Corporate Social Capital in Organizations. In S. M. Gabbay & R. T. A. J. Leenders (Eds.), Social Capital of Organizations (Research in the Sociology of Organizations, Volume 18). Emerald Group Publishing Limited. Hansen, M. T. 1999. The Search-Transfer Problem: The Role of Weak Ties in Sharing Knowledge across Organization Subunits. Administrative Science Quarterly, 44(1): 82– 111. Hargadon, A., & Sutton, R. 1997. Technology Brokering and Innovation in a Product Development Firm. Administrative Science Quarterly, 42(4): 716–749. Harrison, D. A., & Klein, K. J. 2007. What’s the Difference? Diversity Constructs as Separation, Variety, or Disparity in Organizations. Academy of Management Review, 32(4): 1199– 1228. Hoang, H. A., & Rothaermel, F. T. 2010. Leveraging Internal and External Experience: Exploration, Exploitation, and R&D Project Performance. Strategic Management Journal, 31: 734–758. Hougaard, P. 1986. Survival Models for Heterogeneous Populations Derived from Stable Distributions. Biometrika, 73(2): 387–396. Huber, G. P. 1991. Organizational Learning: The Contributing Processes and the Literatures. Organization Science, 2(1): 88–115. Jenkins, S. 2005. Survival analysis. Unpublished manuscript, Institute for Social and Economic Research, University of Essex, Colchester, Essex, UK. 63 Kessler, E. H., & Chakrabathi, A. K. 1996. Innovation Speed: A Conceptual Model of Context, Antecedents, and Outcomes. The Academy of Management Review, 21(4): 1143–1191. Kleinbaum, A. M., & Tushman, M. L. 2007. Building Bridges: The Social Structure of Interdependent Innovation. Strategic Entrepreneurship Journal, 1: 103–122. Kogut, B., & Zander, U. 1992. Knowledge of the Firm, Combinative Capabilities, and the Replication of Technology. Organization Science, 3(3): 383–397. Kogut, B., & Zander, U. 1996. What Firms Do? Coordination, Identity, and Learning. Organization Science, 7(5): 502–518. Lane, P. J., & Lubatkin, M. 1998. Relative Absorptive Capacity and Interorganizational Learning. Strategic Management Journal, 19: 461–477. Laursen, & Salter, A. 2006. Open for Innovation: The Role of Openness in Explaining Innovation Performance among U.K. Manufacturing Firms. Strategic Management Journal, 27(2): 131–150. Lee, D. M. ., & Allen, T. J. 1982. Integrating New Technical Staff: Implications for Acquiring New Technology. Management Science, 28(12): 1405–1421. Lenox, M., & King, A. 2004. Prospects for Developing Absorptive Capacity through Internal Information Provision. Strategic Management Journal, 25: 331–345. Leone, M. I., & Reichstein, T. 2012. Licensing-In Fosters Rapid Invention! The Effect of the Grant-Back Clause and Technological Unfamiliarity. Strategic Management Journal, 33(8): 965–985. Levin, D. Z., Walter, J., & Murnighan, J. K. 2010. Dormant Ties: The Value Of Reconnecting. Organization Science, 22(4): 923–939. Levin, R. C., Klevorick, A. K., Nelson, R. R., Winter, S. G., Gilbert, R., & Griliches, Z. 1987. Appropriating the Returns from Industrial Research and Development. Brookings Papers on Economic Activity, (3): 783–831. Markman, G. D., Gianiodis, P. T., Phan, P. H., & Balkin, D. B. 2005. Innovation Speed: Transferring University Technology to Market. Research Policy, 34(7): 1058–1075. Marsden, P. V. 1990. Network Data and Measurement. Annual Review of Sociology, 16: 435– 463. Matusik, S. F., & Heeley, M. B. 2005. Absorptive Capacity in the Software Industry: Identifying Dimensions That Affect Knowledge and Knowledge Creation Activities. Journal of Management, 31(4): 549–572. 64 McFadyen, M. A., & Cannella, A. A. 2004. Social Capital and Knowledge Creation: Diminishing Returns of the Number and Strength of Exchange. The Academy of Management Journal, 47(5): 735–746. McFadyen, M. A., Semadeni, M., & Cannella, A. A. 2009. Value of Strong Ties to Disconnected Others: Examining Knowledge Creation in Biomedicine. Organization Science, 20(3): 552–564. Mors, M. L. 2010. Innovation in a Global Consulting Firm: When the Problem is Too Much Diversity. Strategic Management Journal. Nerkar, A., & Paruchuri, S. 2005. Evolution of R&D Capabilities: The Role of Knowledge Networks within a Firm. Management Science, 51(5): 771–785. Obstfeld, D. 2005. Social Networks, the Tertius lungens Orientation, and Involvement in Innovation. Administrative Science Quarterly, 50(1): 100–130. Page, S. E. 2007. The Difference: How the Power of Diversity Creates Better Groups, Firms, Schools, and Societies. Princeton, NJ: Princeton University Press. Paruchuri, S. 2009. Intraorganizational Networks, Interorganizational Networks, and the Impact of Central Inventors: A Longitudinal Study of Pharmaceutical Firms. Organization Science, 21(1): 63–80. Paruchuri, S., Nerkar, A., & Hambrick, D. C. 2006. Acquisition Integration and Productivity Losses in the Technical Core: Disruption of Inventors in Acquired Companies. Organization Science, 17(5): 545–562. Pennings, J. M., & Wezel, F. C. 2009. Faraway, Yet So Close: Organizations in Demographic Flux. Organization Science, 21(2): 451–468. Perry-Smith, J. E. 2006. Social Yet Creative: The role of Social Relationships in Facilitating Individual Creativity. Academy of Management Journal, 49(1): 85–101. Phelps, C. C. 2010. A Longitudinal Study of the Influence of Alliance Network Structure and Composition on Firm Exploratory Innovation. The Academy of Management Journal, 53(4): 890–913. Phelps, C. C., Heidl, R., & Wadhwa, A. 2012. Knowledge, Networks, and Knowledge Networks: A Review and Research Agenda. Journal of Management, 38(4): 1115–1166. Polidoro, F., Ahuja, G., & Mitchell, W. 2011. When the Social Structure Overshadows Competitive Incentives: The Effects of Network Embeddedness on Joint Venture Dissolution. The Academy of Management Journal, 54(1): 203–223. Powell, W. W., Koput, K. W., & Smith-Doerr, L. 1996. Interorganizational Collaboration and the Locus of Innovation: Networks of Learning in Biotechnology. Administrative Science Quarterly, 41(1): 116–145. 65 Reagans, R., & McEvily, B. 2003. Network Structure and Knowledge Transfer: The Effects of Cohesion and Range. Administrative Science Quarterly, 48(2): 240–267. Reinholt, M., Pedersen, T., & Foss, N. J. 2011. Why a Central Network Position Isn’t Enough: The Role of Motivation and Ability for Knowledge Sharing in Employee Networks. Academy of Management Journal, 54(6): 1277–1297. Roberts, P. W. 1999. Product Innovation, Product-Market Competition and Persistent Profitability in the U.S. Pharmaceutical Industry. Strategic Management Journal, 20: 655–670. Rodan, S., & Galunic, C. 2004. More than Network Structure: How Knowledge Heterogeneity Influences Managerial Performance and Innovativeness. Strategic Management Journal, 25(6): 541–562. Rosenkopf, L., & Almeida, P. 2003. Overcoming Local Search through Alliances and Mobility. Management Science, 49(6): 751–766. Rost, K. 2010. The Strength of Strong Ties in the Creation of Innovation. Research Policy, 40(4): 588–604. Rothaermel, F. T., & Alexandre, M. T. 2008. Ambidexterity in Technology Sourcing: The Moderating Role of Absorptive Capacity. Organization Science, 20(4): 759–780. Sampson, R. C. 2007. R&D Alliances and Firm Performance: The Impact of Technological Diversity and Alliance Organization on Innovation. Academy of Management Journal, 50(2): 364–386. Schilling, M. A. 2009. Understanding the Alliance Data. Strategic Management Journal, 30: 233–260. Schumpeter, J. 1934. The Theory of Economic Development: An Inquiry into Profits, Capital, Credit, Interest and the Business Cycle. Cambridge, MA: Harvard University Press. Seibert, S. E., Kraimer, M. L., & Liden, R. C. 2001. A Social Capital Theory of Career Success. Academy of Management Journal, 44(2): 219–237. Singh, J. 2005. Collaborative Networks as Determinants of Knowledge Diffusion Patterns. Management Science, 51(5): 756–770. Singh, J., & Agrawal, A. 2011. Recruiting for Ideas: How Firms Exploit the Prior Inventions of New Hires. Management Science, 57(1): 129–150. Singh, J., & Fleming, L. 2009. Lone Inventors as Sources of Breakthroughs: Myth or Reality? Management Science, 56(1): 41–56. 66 Singh, J., Hansen, M. T., & Podolny, J. M. 2010. The World Is Not Small for Everyone: Inequity in Searching for Knowledge in Organizations. Management Science, 56(9): 1415–1438. Sosa, M. E. 2010. Where Do Creative Interactions Come From? The Role of Tie Content and Social Networks. Organization Science, 22(1): 1–21. Tortoriello, M., & Krackhardt, D. 2010. Activating Cross-Boundary Knowledge: The Role of Simmelian Ties in the Generation of Innovations. Academy of Management Journal, 53(1): 167–181. Tortoriello, M., Reagans, R., & McEvily, B. 2012. Bridging the Knowledge Gap: The Influence of Strong Ties, Network Cohesion, and Network Range on the Transfer of Knowledge Between Organizational Units. Organization Science, 23(4): 1024–1039. Tsai, W. 2000. Social Capital, Strategic Relatedness and the Formation of Intraorganizational Linkages. Strategic Management Journal, 21(9): 925–939. Tsai, W. 2001. Knowledge Transfer in Intraorganizational Networks: Effects of Network Position and Absorptive Capacity on Business Unit Innovation and Performance. The Academy of Management Journal, 44(5): 996–1004. Tsai, W., & Ghoshal, S. 1998. Social Capital and Value Creation: The Role of Intrafirm Networks. The Academy of Management Journal, 41(4): 464–476. Turner, K. L., & Makhija, M. V. 2012. The Role of Individuals in the Information Processing Perspective. Strategic Management Journal, 680: 661–680. Tushman, M. L. 1977. Special Boundary Roles in the Innovation Process. Administrative Science Quarterly, 22(4): 587–605. Tushman, M. L., & Scanlan, T. J. 1981. Boundary Spanning Individuals: Their Role in Information Transfer and Their Antecedents. The Academy of Management Journal, 24(2): 289–305. Tzabbar, D., Aharonson, B. S., & Amburgey, T. L. 2012. When Does Tapping External Sources of Knowledge Result in Knowledge Integration? Research Policy, (Advance Access). Uzzi, B. 1997. Social Structure and Competition in Interfirm Networks: The Paradox of Embeddedness. Administrative Science Quarterly, 42(1): 35–67. Vasudeva, G., & Anand, J. 2011. Unpacking Absorptive Capacity: A Study of Knowledge Utilization from Alliance Portfolios. Academy of Management Journal, 54(3): 611–623. Volberda, H. W., Foss, N. J., & Lyles, M. A. 2010. Absorbing the Concept of Absorptive Capacity: How to Realize Its Potential in the Organization Field. Organization Science, 21(4): 931–951. 67 Weitzman, M. L. 1998. Recombinant Growth. Quarterly Journal of Economics, 113(2). Yang, H., Phelps, C. C., & Steensma, H. K. 2010. Learning from What Others Have Learned from You: The Effects of Knowledge Spillovers on Originating Firms. Academy of Management Journal, 53(2): 371–389. Yu, T., & Cannella, A. A. 2007. Rivalry between Multinational Enterprises: An Event History Approach. Academy of Management Journal, 50(3): 665–686. Zaheer, A., & Bell, G. G. 2005. Benefiting from Network Position: Firm Capabilities, Structural Holes, and Performance. Strategic Management Journal, 26(9): 809–825. Zahra, S. A., & George, G. 2002. Absorptive Capacity: A Review, Reconceptualization, and Extension. The Academy of Management Review, 27(2): 185–203. Ziedonis, A. A. 2007. Real Options in Technology Licensing. Management Science, 53(10): 1618–1633. 68 2.10 Appendix Figure 1. Estimated Hazard Functions of Small versus Large Distance Licensed Technologies 0.1 . 025 .05 . 075 .1 020 40 60 80 100 Analysis Time Low Distance High Distance Loglogistic regression 69 Table 1. Descriptive Statistics and Correlations Coefficients (N = 708) Variable Mean S.D. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [1] Technological Distance 0.824 0.280 1.00 [2] Network Density 0.259 0.285 -0.19 1.00 [3] Average Tie Strength 1.581 1.007 -0.01 -0.21 1.00 [4] Network Diversity 0.622 0.228 0.15 -0.57 0.13 1.00 [5] Clustering 2.246 0.854 -0.07 -0.05 0.77 0.09 1.00 [6] Average Path Length 2.557 1.632 0.20 -0.64 0.25 0.43 0.10 1.00 [7] Same Sector 0.357 0.479 0.01 0.07 0.10 0.06 -0.07 -0.18 1.00 [8] Co Patent 0.892 0.310 -0.18 -0.08 0.05 0.07 0.07 0.04 -0.04 1.00 [9] Prior Citations 1.261 7.847 -0.04 0.12 -0.06 0.01 0.03 -0.12 -0.05 0.04 1.00 [10] Scientific References 30.381 52.714 -0.01 -0.06 0.11 0.02 0.10 0.07 0.05 0.07 -0.01 1.00 [11] Technology Value 60.671 155.548 -0.11 0.14 -0.03 0.00 0.02 -0.08 -0.01 0.07 0.07 0.00 1.00 [12] Technological Furnishing 0.559 0.497 0.08 -0.29 -0.01 0.15 -0.06 0.26 0.15 0.05 -0.10 -0.05 -0.11 [13] Grant-back Clause 0.242 0.429 0.02 -0.03 -0.05 0.04 -0.13 0.08 0.17 -0.08 -0.07 -0.07 0.00 [14] Milestone 0.613 0.487 -0.00 -0.05 0.12 -0.15 0.13 0.07 -0.01 -0.03 0.08 -0.04 -0.08 [15] R&D Intensity 124.633 132.409 -0.21 0.26 -0.02 -0.20 0.05 -0.32 0.13 0.00 -0.01 0.01 -0.10 [16] Licensor University 0.177 0.381 -0.12 0.21 -0.11 -0.34 -0.04 -0.27 -0.35 -0.18 0.03 -0.07 -0.06 [17] Licensor Number of Patents 334.927 1.451.046 -0.05 0.16 -0.07 -0.11 0.00 -0.17 -0.11 0.00 0.77 -0.04 0.03 [18] US Firm 0.898 0.303 0.01 0.12 -0.00 -0.06 -0.05 -0.09 0.08 -0.09 0.05 -0.06 -0.04 [19] Log(Number Employees) 7.129 2.844 0.24 -0.52 0.16 0.32 0.08 0.68 -0.25 0.25 -0.08 0.10 0.06 [20] Average Patenting Time 4.091 5.165 -0.05 0.39 -0.24 -0.38 -0.31 -0.51 0.11 -0.39 -0.07 -0.13 -0.04 [21] Previous Year Patent 0.766 0.423 0.09 -0.27 0.13 0.34 0.13 0.33 -0.01 0.63 0.03 0.09 0.06 [22] Slack 165.471 149.181 0.09 -0.37 0.35 0.28 0.21 0.42 -0.01 0.24 -0.11 0.14 -0.09 Variable Mean S.D. [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [12] Technological Furnishing 0.559 0.497 1.00 [13] Grant-back Clause 0.242 0.429 0.18 1.00 [14] Milestone 0.613 0.487 0.17 0.04 1.00 [15] R&D Intensity 124.633 132.409 -0.06 -0.07 0.15 1.00 [16] Licensor University 0.177 0.381 -0.32 -0.23 0.08 0.28 1.00 [17] Licensor Number of Patents 334.927 1.451.046 -0.13 -0.02 0.12 0.21 0.06 1.00 69 Table 1. Descriptive Statistics and Correlations Coefficients (N = 708) Variable Mean S.D. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [1] Technological Distance 0.824 0.280 1.00 [2] Network Density 0.259 0.285 -0.19 1.00 [3] Average Tie Strength 1.581 1.007 -0.01 -0.21 1.00 [4] Network Diversity 0.622 0.228 0.15 -0.57 0.13 1.00 [5] Clustering 2.246 0.854 -0.07 -0.05 0.77 0.09 1.00 [6] Average Path Length 2.557 1.632 0.20 -0.64 0.25 0.43 0.10 1.00 [7] Same Sector 0.357 0.479 0.01 0.07 0.10 0.06 -0.07 -0.18 1.00 [8] Co Patent 0.892 0.310 -0.18 -0.08 0.05 0.07 0.07 0.04 -0.04 1.00 [9] Prior Citations 1.261 7.847 -0.04 0.12 -0.06 0.01 0.03 -0.12 -0.05 0.04 1.00 [10] Scientific References 30.381 52.714 -0.01 -0.06 0.11 0.02 0.10 0.07 0.05 0.07 -0.01 1.00 [11] Technology Value 60.671 155.548 -0.11 0.14 -0.03 0.00 0.02 -0.08 -0.01 0.07 0.07 0.00 1.00 [12] Technological Furnishing 0.559 0.497 0.08 -0.29 -0.01 0.15 -0.06 0.26 0.15 0.05 -0.10 -0.05 -0.11 [13] Grant-back Clause 0.242 0.429 0.02 -0.03 -0.05 0.04 -0.13 0.08 0.17 -0.08 -0.07 -0.07 0.00 [14] Milestone 0.613 0.487 -0.00 -0.05 0.12 -0.15 0.13 0.07 -0.01 -0.03 0.08 -0.04 -0.08 [15] R&D Intensity 124.633 132.409 -0.21 0.26 -0.02 -0.20 0.05 -0.32 0.13 0.00 -0.01 0.01 -0.10 [16] Licensor University 0.177 0.381 -0.12 0.21 -0.11 -0.34 -0.04 -0.27 -0.35 -0.18 0.03 -0.07 -0.06 [17] Licensor Number of Patents 334.927 1.451.046 -0.05 0.16 -0.07 -0.11 0.00 -0.17 -0.11 0.00 0.77 -0.04 0.03 [18] US Firm 0.898 0.303 0.01 0.12 -0.00 -0.06 -0.05 -0.09 0.08 -0.09 0.05 -0.06 -0.04 [19] Log(Number Employees) 7.129 2.844 0.24 -0.52 0.16 0.32 0.08 0.68 -0.25 0.25 -0.08 0.10 0.06 [20] Average Patenting Time 4.091 5.165 -0.05 0.39 -0.24 -0.38 -0.31 -0.51 0.11 -0.39 -0.07 -0.13 -0.04 [21] Previous Year Patent 0.766 0.423 0.09 -0.27 0.13 0.34 0.13 0.33 -0.01 0.63 0.03 0.09 0.06 [22] Slack 165.471 149.181 0.09 -0.37 0.35 0.28 0.21 0.42 -0.01 0.24 -0.11 0.14 -0.09 Variable Mean S.D. [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [12] Technological Furnishing 0.559 0.497 1.00 [13] Grant-back Clause 0.242 0.429 0.18 1.00 [14] Milestone 0.613 0.487 0.17 0.04 1.00 [15] R&D Intensity 124.633 132.409 -0.06 -0.07 0.15 1.00 [16] Licensor University 0.177 0.381 -0.32 -0.23 0.08 0.28 1.00 [17] Licensor Number of Patents 334.927 1.451.046 -0.13 -0.02 0.12 0.21 0.06 1.00 70 [18] US Firm 0.898 0.303 -0.14 -0.05 -0.10 0.08 0.09 0.07 1.00 [19] Log(Number Employees) 7.129 2.844 0.07 0.08 -0.01 -0.60 -0.37 -0.22 -0.20 1.00 [20] Average Patenting Time 4.091 5.165 -0.07 -0.02 -0.02 0.13 0.22 0.03 0.11 -0.49 1.00 [21] Previous Year Patent 0.766 0.423 0.08 0.03 -0.02 -0.08 -0.36 0.01 -0.09 0.40 -0.55 1.00 [22] Slack 165.471 149.181 0.14 -0.03 0.12 -0.25 -0.32 -0.16 -0.08 0.54 -0.26 0.25 1.00 70 [18] US Firm 0.898 0.303 -0.14 -0.05 -0.10 0.08 0.09 0.07 1.00 [19] Log(Number Employees) 7.129 2.844 0.07 0.08 -0.01 -0.60 -0.37 -0.22 -0.20 1.00 [20] Average Patenting Time 4.091 5.165 -0.07 -0.02 -0.02 0.13 0.22 0.03 0.11 -0.49 1.00 [21] Previous Year Patent 0.766 0.423 0.08 0.03 -0.02 -0.08 -0.36 0.01 -0.09 0.40 -0.55 1.00 [22] Slack 165.471 149.181 0.14 -0.03 0.12 -0.25 -0.32 -0.16 -0.08 0.54 -0.26 0.25 1.00 71 Table 2. Results of Log-Logistic Hazard Models with Gamma Frailty Predicting the Time to Knowledge Recombination Variable Model I Model II Model III Model IV Model V Model VI Technological Distance 2.439*** 2.339*** 2.446*** 2.019*** 1.915*** (0.614) (0.555) (0.605) (0.530) (0.460) Technological Distance x Network Density -6.418** -12.107*** (2.191) (3.433) Technological Distance x Avg. Tie Strength 1.368 0.959 (1.629) (0.972) Technological Distance x Network Diversity -9.815** -10.860** (3.234) (4.008) Network Density -2.561+ -1.309 -1.160 -1.014 -1.238 -1.097 (1.484) (2.132) (1.325) (2.075) (1.779) (1.219) Average Tie Strength 0.184 0.374 0.504+ 0.424 0.535+ 0.617+ (0.312) (0.395) (0.295) (0.419) (0.296) (0.349) Network Diversity -1.083 0.332 -0.282 0.123 1.828 0.227 (1.255) (1.684) (1.070) (1.838) (1.404) (1.330) Clustering -0.428 -0.677 -0.933** -0.649 -0.930** -0.867* (0.371) (0.526) (0.326) (0.569) (0.326) (0.380) Average Path Length -0.499* -0.633*** -0.688*** -0.584** -0.745*** -0.713*** (0.197) (0.177) (0.168) (0.178) (0.163) (0.156) Same Sector 0.235 -0.190 -0.103 -0.238 -0.020 -0.198 (0.457) (0.546) (0.449) (0.581) (0.398) (0.394) Co Patent 7.060** 6.449* 3.483 8.458+ 13.465*** 6.821 (2.401) (3.045) (2.291) (4.779) (3.323) (5.064) Prior Citations 0.060* 0.038 0.024 0.032 0.040+ 0.020 (0.029) (0.037) (0.037) (0.040) (0.023) (0.023) Scientific References 0.003 0.005 0.006 0.006 0.004 0.006 (0.004) (0.005) (0.005) (0.005) (0.005) (0.004) Technology Value -0.011*** -0.010** -0.011*** -0.010** -0.008* -0.010*** (0.003) (0.003) (0.002) (0.003) (0.003) (0.003) Technological Furnishing -0.078 -0.360 -0.154 -0.331 -0.323 -0.374 (0.454) (0.412) (0.383) (0.419) (0.453) (0.397) Grant-back Clause -1.186** -1.116** -1.225*** -1.152** -0.823* -1.020** (0.429) (0.385) (0.353) (0.394) (0.362) (0.341) 71 Table 2. Results of Log-Logistic Hazard Models with Gamma Frailty Predicting the Time to Knowledge Recombination Variable Model I Model II Model III Model IV Model V Model VI Technological Distance 2.439*** 2.339*** 2.446*** 2.019*** 1.915*** (0.614) (0.555) (0.605) (0.530) (0.460) Technological Distance x Network Density -6.418** -12.107*** (2.191) (3.433) Technological Distance x Avg. Tie Strength 1.368 0.959 (1.629) (0.972) Technological Distance x Network Diversity -9.815** -10.860** (3.234) (4.008) Network Density -2.561+ -1.309 -1.160 -1.014 -1.238 -1.097 (1.484) (2.132) (1.325) (2.075) (1.779) (1.219) Average Tie Strength 0.184 0.374 0.504+ 0.424 0.535+ 0.617+ (0.312) (0.395) (0.295) (0.419) (0.296) (0.349) Network Diversity -1.083 0.332 -0.282 0.123 1.828 0.227 (1.255) (1.684) (1.070) (1.838) (1.404) (1.330) Clustering -0.428 -0.677 -0.933** -0.649 -0.930** -0.867* (0.371) (0.526) (0.326) (0.569) (0.326) (0.380) Average Path Length -0.499* -0.633*** -0.688*** -0.584** -0.745*** -0.713*** (0.197) (0.177) (0.168) (0.178) (0.163) (0.156) Same Sector 0.235 -0.190 -0.103 -0.238 -0.020 -0.198 (0.457) (0.546) (0.449) (0.581) (0.398) (0.394) Co Patent 7.060** 6.449* 3.483 8.458+ 13.465*** 6.821 (2.401) (3.045) (2.291) (4.779) (3.323) (5.064) Prior Citations 0.060* 0.038 0.024 0.032 0.040+ 0.020 (0.029) (0.037) (0.037) (0.040) (0.023) (0.023) Scientific References 0.003 0.005 0.006 0.006 0.004 0.006 (0.004) (0.005) (0.005) (0.005) (0.005) (0.004) Technology Value -0.011*** -0.010** -0.011*** -0.010** -0.008* -0.010*** (0.003) (0.003) (0.002) (0.003) (0.003) (0.003) Technological Furnishing -0.078 -0.360 -0.154 -0.331 -0.323 -0.374 (0.454) (0.412) (0.383) (0.419) (0.453) (0.397) Grant-back Clause -1.186** -1.116** -1.225*** -1.152** -0.823* -1.020** (0.429) (0.385) (0.353) (0.394) (0.362) (0.341) 78 3.2 Theoretical Background As the traditional integrated R&D models have been replaced for more open and collaborative forms of partnerships (Chesbrough, 2003; Laursen and Salter, 2006), the importance of markets for technology has dramatically increased (Ceccagnoli & Jiang, 2012). In a context of knowledge transactions, licensing contracts are one of the main mechanisms used by firms to trade know-how and technologies (Arora & Gambardella, 2010). Accordingly, licensing can be described as an arm’s length contractual deal through which firms can trade know-how and intellectual property (IP) rights (Arora, 1995). Arora & Gambardella (2010) proposes that a distinguishing feature of licensing in contrast to other mechanisms for knowledge acquisition (i.e., joint ventures, M&A, and the mobility of human capital) regards the contractual nature of the knowledge. While ex-ante contracts such as research alliances involve higher uncertainty about potential outcomes, in ex-post contracts such as licensing, the traded technology can be more easily defined. Furthermore, in most licensing contracts between firms, as in the two examples mentioned earlier, the deals involve already developed and proven technologies (Atuahene-Gima, 1993; Leone & Reichstein, 2012). As the exchange of knowledge is at the core of technology licensing, it is a mechanism that is closely related to the notion of organizational learning. Indeed, the idea of organizational learning as “a change in an organization’s capacity for doing something new” (Tannenbaum, 1997, p. 438) can be closely connected with the licensee’s position in a licensing deal. In fact, existing literature focusing on understating the motives for firms to license-in indicates that the acquisition of new technologies through licensing-in positively affects distinct dimensions of firm innovation. For example, Laursen et al. (2010) suggested that firms can explore more distantly from their current technological trajectory using licensed technologies. Leone and Reichstein (2012) found that licensees are able to innovate faster than non-licensees 79 due to the possibility to build on technologies that are already developed. However, although previous studies have documented the importance of technology licensing, little is known about how the licensee’s organizational characteristics affect its capacity to innovate based on the licensed technologies. Along this line, it is well established in the organizational literature that one of the main determinants of a firm’s capacity to assimilate and integrate external knowledge regards its level of absorptive capacity. Accordingly, recent studies on technology licensing have shown that firms with high levels of absorptive capacity are better able to deal with technologies acquired through a licensing contract (e.g., Laursen et al., 2010; Ceccagnoli & Jiang, 2012). Despite the fact that absorptive capacity is one of the main determinants of a firm’s ability to benefit from licensing-in, it is not the only one. Considering the existing literature on technology licensing, it is possible to identify other relevant contingencies that are expected to affect the licensee’s capacity to draw on and learn from licensing-in. For example, Ceccagnoli & Jiang (2012) argue that integration costs caused by a high degree of cospecialization between R&D and downstream activities can limit the licensee’s capacity to deal with licensed-in technologies. In another example, Arora & Gambardella (2010) call attention to the importance of understanding how internal resistance to external knowledge affects firms operating on the demand side of markets for technology. Following the idea of integration costs proposed by Ceccagnoli & Jiang (2012), it is expected that the successful assimilation and application of licensed-in technologies will require the licensee to allocate substantial resources in the exploitation of the newly acquired technology. Consequently, unabsorbed resources at the time that a technology has been acquired can be necessary for the licensee to be able to deal with integration challenges. Furthermore, previous studies have shown that certain levels of organizational slack are not only relevant in 80 terms of the resource availability but also to induce experimentation and risk taking and to absorb the uncertainty related to the development and exploitation of new technologies (Cyert and March, 1963; Nohria and Gulati, 1996). The second point, regarding the internal resistance that licensed-in technologies can experience, is also at the core of the integration process that licensees have to go through in order to benefit from licensing-in. Thus, even if a firm spends large amounts to license-in a technology, internal resistance from individuals or groups can significantly reduce or prevent the newly acquired knowledge from being assimilated and disseminated within the organization. This type of resistance against external knowledge can be fostered by lack of familiarity or skills (Cohen & Levinthal, 1990), path-dependency (Dosi 1982), and the concern that individuals’ or groups’ own ideas may be disregarded within the organization (Katz & Allen, 1982; Gupta & Govindarajan, 2000). Considering the impact that internal resistance may have on firm capacity to benefit from external knowledge acquisition, I operationalize the concept of myopia using the licensee perspective. To summarize, previous studies have analyzed whether firms can use technology licensing-in as a mechanism to acquire external knowledge, but little is yet known about firmspecific characteristics that affect the process through which licensed technologies can foster organizational learning. In this paper I focus on how the knowledge acquired through licensing affects the number of innovations produced by the licensee and also on the moderating effect that organizational slack and myopia have on this main relationship. 81 3.3 Theory and Hypothesis 3.3.1 Technology Licensing and Firm Innovation Firms can use licensing contracts as a main mechanism to acquire externally developed technologies and feed their needs for inventive knowledge (Laursen, et al., 2010). Indeed, through licensing contracts the acquiring firm can improve its innovation performance by assimilating and adapting new knowledge connected to the licensed technologies. In this sense, the reliance on technologies that have already been developed and proven not only saves efforts that would otherwise have been spent on creating a totally new technology, but also provides the acquiring firm with a larger set of technological opportunities (Klevorick, Levin, Nelson, & Winter, 1995). Furthermore, the fact that the licensed technology was developed by a different organization that naturally possesses a different set of capabilities and skills (Teece, Pisano, & Shuen, 1997) opens up new learning opportunities to the licensee. Considering the link between licensing and organizational learning, previous studies have proposed the term “learning-by-licensing” (Johnson, 2002) to indicate the learning possibilities that firms can access when engaging in licensing agreements. According to this perspective, the acquisition of new knowledge results in organizational learning through an interactive combinatorial process in which new and existing elements are linked together through a continuous process of experimentation (Pisano, 1996). In this context, a licensed technology can be understood as an input that increases the size and diversity of the firm’s knowledge base. Indeed, following the knowledge-based theory of the firm, in order to develop and sustain innovation, firms must be able to manage previously accumulated knowledge as well as the inflow of knowledge generated outside their boundaries (Deeds & Decarolis, 1999). Accordingly, licensing-in is expected to have a positive impact on the number of innovations 82 produced by the licensee, and also provide the grounds for the adoption of new search trajectories (Laursen et al., 2010). There are at least two distinct mechanisms through which licensing-in leads to organizational learning that can be subsequently converted into innovations. First, licensing agreements can be considered as an alternative for firms to directly acquiring/buying the IP rights to exploit a specific technology (Arora & Ceccagnoli, 2011; Ziedonis, 2004). In this case, the acquiring firm can experience learning benefits from gaining access to state-of-the-art methods and process, which can increase the R&D efficiency and the way that innovations are internally produced and managed (Gallini & Winter, 1985). Second, the other possibility for organizational learning relates to the acquisition of technologies associated with knowledge that the firm knows very little or nothing about. The unfamiliarity of the licensee with the licensed technology regards the degree to which the knowledge about the technology is not present within the firm. In this situation, the newly acquired knowledge provides the licensee with a context for novel combinations between existing and new elements (Grant, 1996; Nooteboom, Van Haverbeke, Duysters, Gilsing, & van den Oord, 2007). Accordingly, previous research indicates that licensing is an efficient mechanism for firms to fill internal gaps in their knowledge bases, complement internal capabilities, and create the potential for new knowledge combinations. Therefore, my baseline hypothesis is the following: Hypothesis 1: Engaging in technology licensing will be positively related to a firm’s subsequent capacity to produce innovations However, despite the positive effect that licensing-in is expected to have on firms’ capacity to produce innovations, I propose that integrating a new technology can also be challenging for the licensee. Indeed, the process of knowledge transfer and integration is directly dependent on the 83 organizational capabilities and resources that the acquiring firm possesses to tap into external knowledge sources (Grant, 1996; Van Den Bosch, Volberda, & De Boer, 1999). Therefore, I also focus on organizational factors related to resource availability and firms’ capacity to drawn on external knowledge to explain cross-firm differences in benefiting from licensing-in. Following this baseline hypothesis, in the next sections I will focus on the moderating effect that recoverable slack and organizational myopia have on the main relationship between licensing-in and innovation. 3.3.2 Recoverable Slack and Knowledge Integration Previous studies have pointed out that despite the fundamental role that external knowledge acquisition has for firms’ innovation performance, the assimilation and integration of knowledge generated outside the firms’ boundaries can pose several difficulties and risks (Grant, 1996). Specifically examining the case of licensing-in, there is the possibility that the acquiring firm will face significant challenges in understanding and successfully applying the new knowledge (Ceccagnoli & Jiang, 2012). Furthermore, in several cases, firms will have limited comprehension of how the new technology can be further developed, becoming dependent on the licensor to support the process of knowledge integration (Arora, 1995). Consequently, given that external technologies can be difficult to assimilate and integrate, it is not uncommon that the acquiring firm needs to invest significant efforts and resources in order to benefit from it (Ceccagnoli & Jiang, 2012; Kotha et al., 2013).In fact, successful integration requires the licensed-in technology to go through a process of experimentation, refinement and, especially in the case of licensing, adaptation to the specific needs of the acquiring firm (Jensen & Thursby, 2001). The licensee’s difficulties in integrating a newly licensed-in technology increase if the licensor and the licensee operate in different industries or at different stages of the value 84 chain (Wilcox King & Zeithaml, 2003). This can be attributed to the fact that in most cases, a technology has been developed to be applied and to meet the needs of a firm in a specific context (Gambardella & Giarratana, 2012). In fact, it is not uncommon that licensors choose to license their technologies only to firms operating in different industries in order to avoid an increase in the number of direct competitors with similar technological capabilities (Arora, Fosfuri, & Gambardella, 2001; Fosfuri, 2006). This fact suggests that the licensee needs to possess high levels of absorptive capacity relative to the licensed technology (Lane & Lubatkin, 1998) or the necessary resources to deal with the integration challenges in a timely manner (Ceccagnoli & Jiang, 2012). Given those challenges, the integration process of licensed-in technologies can be associated with high uncertainty regarding its potential outcomes. Indeed, several studies have pointed out that the process of external knowledge acquisition and development is associated with high failure rates (Das & Teng, 2000; Park & Russo, 1996). In this context, Laursen et al., (2010) argue that licensing-in can also be considered a form of exploratory search that firms use to reach distant (unfamiliar) technological domains. In fact, if we refer back to the seminal work of March (1991), exploratory search is described as being directly associated with high risk and experimentation. Departing from this perspective, I argue that the level of recoverable slack within the acquiring firm is critical for the process of assimilating and integrating licensed-in technologies. Organizational literature has conceptualized recoverable slack as “the pool of resources in an organization that is in excess of the minimum necessary to produce a given level of organizational output” (Nohria & Gulati, 1996). Indeed, slack resources can be observed in the form of excess inputs such as unused capacity, unnecessary capital expenditures, and unexploited opportunities to increase outputs (Cheng & Kesner, 1997; George, 2005; Nohria & Gulati, 1996; Singh, 1986). High levels of slack are particularly useful for firms when dealing 85 with environmental jolts (Meyer, 1982) and engaging in high-risk R&D activities (George, 2005). Hence, I expect slack to directly affect firm capacity to benefit from licensing-in. Therefore, I propose that the positive effect of technology licensing-in on a firm’s capacity to produce innovations will be stronger under conditions of high recoverable slack. As the level of unabsorbed slack available to the licensee increases, the amount of resources that can be redeployed for the successful integration of the newly acquired technology grows. Indeed, considering that before being able to benefit from a licensed-in technology a firm must first be able to assimilate it, the avaiability of resources can be particularly critical to the integration process. Under conditions of resource constraint, firms are more likely to prioritize R&D projects that have more certain outcomes and a low risk of failure (Nohria & Gulati, 1996). Consequently, licensed technologies may remain underutilized in terms of their potential as inputs in the generation of innovations under conditions of limited resources. On the other hand, when the amount of resources available to be redeployed in the assimilation and application of licensed-in technologies is high, firms are likely to increase the effects of licensing-in on innovation. Therefore, I argue that when the knowledge inputs provided by technology licensing-in are held constant, the capacity that the licensee will have to assimilate and to produce novel combinations from the licensed technologies will increase under conditions of a high level of recoverable slack. Consequently, the positive effect of licensing-in on the number of innovations produced by a firm tends to increase with the total amount of unobserved resources at the time of the licensing agreement. Based on those arguments I propose the following moderating effect for recoverable slack: Hypothesis 2: The effect of technology licensing on innovation will be moderated by the firm’s level of recoverable slack in such a fashion that increasing slack will increase the positive effect of licensing on the firm’s subsequent capacity to produce innovations 86 3.3.3 Organizational Myopia and Licensing My first hypothesis argues that licensing-in increases firms’ subsequent capacity to produce innovations. However, this effect may be also contingent upon the openness of the acquiring firm to build upon and learn from external knowledge. Katz and Allen (1982) proposed the term “Not Invented Here (NIH) Syndrome” to describe the propensity that stable research teams have to systematically overlook and reject the ideas generated by outsiders. As suggested by the NIH literature, especially among individuals affiliated with highly cohesive groups within organizations, the knowledge possessed by insiders is often seem as superior to the knowledge that lies outside the firm’s boundaries (Katz and Allen, 1982). Indeed, Katz and Allen define the NIH syndrome as a the fact that “…stable project teams become increasingly cohesive over time and begin to separate themselves from external sources of technical information and influence by communicating less frequently with professional colleagues outside their teams” (1982, p. 7). In this context, it is well known that organizations often draw disproportionally on internally developed technologies and do not pay enough attention to external sources (Agrawal, Cockburn, & Rosell, 2009; Arora & Gambardella, 2010). This is especially true for large firms, where inventors are more likely to find a larger set of technological alternatives generated inhouse (Agrawal et al., 2009). Relying on the NIH syndrome definition, Agrawal et al. (2009) describe as myopic organizations firms in which inventors disproportionately build upon their own innovations7. The presence of persistent myopic behavior in the process of generating innovations is, at least partially, a consequence of the lower incentives that single individuals or groups have to use external knowledge as compared to their own (Katz & Allen, 1982). For 7 In their paper, Agrawal et al. propose five different types of myopia: 1) Self-citation to Own Prior Inventions; 2) Use of New Knowledge; 3) Technological Myopia; 4) Locational Myopia; and 5) Temporal Myopia. Because I am interested in firms’ resistance to external knowledge, I focus in my analysis on the first concept of myopia. 87 example, apart from the fact that inventors are likely to find it simpler to continuously build on their own inventions and accumulated knowledge, individuals tend to be rewarded if their ideas are recurrently used and applied within the organization (Rotemberg & Saloner, 1994). In fact, putting aside the implications related to the lack of access to external knowledge sources, it is less costly and simpler to access and use internal knowledge relative to external knowledge. However, as the NIH syndrome emerges, organizations’ capacity to absorb and use external knowledge becomes jeopardized (Agrawal et al., 2009). Furthermore, given that firms need to continuously tap into knowledge sources to learn and absorb new technological knowledge, internal resistance can harm their capacity to stay competitive. I expect the deleterious bias created by the NIH syndrome to be particularly prominent in the case of licensed-in technologies. Given that licensing agreements often involve technologies that are already substantially developed, individuals may not find space to incorporate their own previous innovations into the core of the licensed-in technology. Consequently, in comparison with strategic alliances, where the output is usually the fruit of joint development efforts between inventors in two or more firms (Sampson, 2007; Schilling & Phelps, 2007), licensed technologies are more prone to suffer internal resistance. Individuals and teams might see their position and expertise threatened if a technology whose development they had relatively little, or no, influence on is adopted instead of their own. In other words, they have stronger incentives to invest more effort and time into developing and incorporating knowledge generated by themselves or by other members of their group. Consequently, licensed-in technologies may be overlooked in favor of internally developed ones. Another explanation of why individuals might resist a licensed-in technology is related to the path-dependent nature of organizational routines and capabilities (Dosi & Marengo, 2007; Pentland, Feldman, Becker, & Liu, 2012). While developing and establishing routines that allow individuals to work in teams and facilitate collaboration between employees 94 where  5 represents the total number citations that firm i made to its own patents, and , the total number of citations, regardless of the ownership of the cited patent. Following previous studies I applied a seven year moving window to allow for enough patents to be produced and to account for knowledge depreciation over time (Phelps, 2010). - Control Variables Presample Patents. To control for unobserved heterogeneity concerning firms’ patenting propensity, I use the presample average innovation count proposed by (Blundell, Griffith, & Reenen, 1995). To operationalize this variable I calculate the cumulative number of patents obtained by a firm in the five years prior to its entry in the sample. Industry Competition. Given that both the decision to license in and the decision to innovate might be triggered by competitive pressures, I include as a control variable the total number of firms listed in the same four-digit SIC code of firm i in COMPUSTAT in year t. Technological Collaborator. One way that firms have to deal with problem solving related to the assimilation of new technologies is the use of external partners (Powell, Koput, & Smith- Doerr, 1996). Therefore, I account for the number of co-patents applied for by firm i in the five years prior to the licensing-in deal as a proxy for propensity to collaborate with external partners. Patenting Experience. I control for technological experience using the number of years that elapsed between the first time the firm applied for a patent and year t. Firms with more experience in innovating might also be better at managing licensed technologies. Furthermore, I also expect this variable to be significantly correlated with firm age, which has been shown to affect innovation (e.g.,Owen-Smith & Powell, 2004; Sørensen & Stuart, 2000). 95 Technological Complexity. I control for firms’ ability to handle technological complexity by computing the average number of claims on patents applied for in the seven years preceding year t. Experience in dealing with complex bodies of knowledge makes it easier for firms to integrate the acquired technology into their own knowledge bases (Leone & Reichstein, 2012). Evaluation Capacity. Firms differ significantly in their capacity to evaluate external knowledge. On the basis of Arora and Gambardella (1994), I calculate the Evaluation Capacity variable using the average number of scientific references in the backward citations of patents accumulated in the seven years preceding year t. This measure aims to capture firms’ in-house scientific capabilities, which reflect both the extent to which a firm’s inventors search the frontier of technological space and also their ability to deal with scientific knowledge15. Technology Diversity. Higher levels of diversity in a firm’s knowledge base increase the likelihood that a licensed technology will be easily absorbed (Laursen et al., 2010). I measure firm technology diversity in year t using the Herfindahl index applied to the technological classes of the patents that firm i produced before year t: Technology diversity =1 − /0! !2. 3, Where Nit is the total number of patents that a firm accumulated in the previous seven years and Njit represents the number of patents in technology class j. The final measure is obtained by subtracting 1 from the value reflecting the concentration of patent classes across the different technological domains. 15 In their paper, Arora and Gambardella measure Evaluation Capacity using the number of scientific publications produced by a firm’s inventors. I do not have information on the scientific outputs of inventors; nevertheless, I expect that the extent to which firms rely on scientific references to produce innovations is also a good indicator of inventors’ search behavior and their capacity to deal with scientific knowledge. 96 Licensing Experience. Accumulated experience with licensing-in might enhance firms’ overall innovation capabilities by establishing internal routines related to the way that licensed-in technologies are managed. Furthermore, as discussed earlier, the effect of licensing-in on innovation might extend over more than one period. I control for the number of licensing deals that firm i has accumulated in the three years before year t. Firm R&D Intensity. A firm’s expenditure in R&D activities is one of the main determinants of its capacity to absorb external knowledge (Cohen & Levinthal, 1990). So, in order to control for firm differences in terms of absorptive capacity, I measure R&D intensity by dividing a firm’s R&D expenses by its sales in year t. Firm Size. To control for firm size, I used the natural log of the number of employees (thousands) for firm i in year t. U.S./Canada Firm. I used a region dummy that takes the value 1 if the firms’ headquarter is located in U.S. or Canada. 3.6 Statistical Method Given that I measure firm innovation performance using citation-weighted patents, the model used to conduct the empirical analysis had to appropriately accommodate non-negative integer count values. Furthermore, previous studies have shown that modeling patent count requires using a regression approach that deals with many zeros (Sampson, 2007; Ziedonis, 2004). I first considered using a Poisson model as it is one of the simplest alternatives to deal with count data (Hausman, Hall, & Griliches, 1984). However, the Poisson distribution has the strong assumption that the variance is proportional to the mean, #(6)=7(6)=8. If this assumption is violated, one of the implications concerns the fact that the coefficients will be estimated 97 consistently, but underestimated standard errors might be reflected in spurious significance levels (Cameron & Trivedi, 1986; Gourieroux, Monfort, & Trognon, 1984). The test for overdispersion provided evidence against using a Poisson model and in favor of a model that allows the variance of the dependent variable to exceed its mean. The usual alternative to the pure Poisson model is the conditional Negative Binomial specification (Hausman et al., 1984), which is a generalization of the Poisson that is appropriate under conditions of overdispersion. The estimation of a Negative Binomial model in a longitudinal setting allows implementing either fixed or random effects as an alternative to deal with unobserved characteristics regarding the subjects in the sample. I chose to use the fixed effects specification as it allows for arbitrary correlation between unobserved timeconstant factors and the explanatory variables (Wooldridge, 2012), providing more conservative estimators. In the case of the present study, I expect that not all relevant time invariant attributes simultaneously affecting firm capacity to deal with licensed-in technologies and to innovate can be successfully incorporated into the model. Furthermore, although the fixed effects estimators remove both desirable and undesirable variation across subjects (Angrist & Pischke, 2009), failing to control for unobserved heterogeneity might result in significant specification errors (Heckman, 1979). Despite the large number of studies applying the conditional Negative Binomial fixed effects model, this model has been criticized for not providing “true” fixed effects estimators as it does not control for all time-invariant attributes (Allison & Waterman, 2002). Schilling & Phelps (2007) suggest that incorporating into the model firm attributes to account for the residual unobserved heterogeneity specifically associated with firm patenting behavior helps to deal with this problem. Following Schilling and Phelps’ steps, in addition to firm fixed effects, I also used the presample information approach of Blundell et al. (1995) as an 98 explanatory variable. In their approach, Blundell et al. argued that, in innovation models, one of the main sources of firm heterogeneity comes from differences in the knowledge stock that firms enter the sample. They propose that using the “entry stock” of patents can adequately control for fixed effects and virtually eliminate persistent serial correlation in count models (i.e., Poisson and Negative Binomial). This approach has also been applied in several studies related to innovation performance as a way to minimize the failure of negative binomial fixed effects to capture all firm-specific effects (e.g., Ahuja & Katila, 2001; Keil, Maula, Schildt, & Zahra, 2008; Schilling & Phelps, 2007). Accordingly, I included in the Negative Binomial model the Presample Patents as an explanatory variable. Finally, I also used year and industry dummies to control for unobserved industry and period effects not captured by the firm fixed effects. 3.7 Results Table 1 reports descriptive statistics and simple correlations between the variables used in the regression analysis. The reported dependent variable corresponds to the citation-weighted count of patents following a one year lag (Patents t+1)16. Results of the pairwise correlation raised no concerns regarding multicollinearity. Particularly, the explanatory variables concerning the hypothesized effects do not present any strong correlations among themselves or with the control variables. Additionally, the maximum variance inflation factor (VIF) associated with any of the independent variables was 2.17 (mean VIF = 1.31), which is well below the rule-of- thumb value of ten (Gujarati, 2003). On average, 18% of the firm year observations in the sample involve a Technology Licensing deal, with larger firms being more likely to license in. In line with my expectations, the presample estimator presented a moderate correlation with the dependent variable (r=0.29), suggesting that firms’ pre-entry patent stock is positively associated with firms’ future capacity to produce innovations. 16 There are no substantial changes in the correlation between the dependent variable and the main explanatory variables, except to the fact that some of the correlations decrease in magnitude. 99 [Insert Table 1 around here] Table 2 reports the negative binomial panel with fixed effects for the three dependent variables (Patents t+1; Patent t+2; Patents t+3). The results for the different year lags are reported in different models. Models 1, 2, 3, and 4 report the results for a one-year lag between the explanatory variables and the citation-weighted patent count (Patents t+1). Models 5, 6, 7, and 8 report the results using a two-year lag (Patents t+2). Finally, models 9, 10, 11, and 12 report the results using a three-year lag (Patents t+3). In order to check the stability of the coefficients and standard errors, I added the main explanatory variables one at a time. For each of the three dependent variables, the first models (1, 5, and 9) include the control variables and the main explanatory variable Technology Licensing. In the second models (2, 6, and 10), I included the interaction Technology Licensing X Organizational Myopia. In the third group of models (3, 7, and 11), the interaction term Technology Licensing X Recoverable Slack also enters the regression without the previous interaction term. Finally, the final models (4, 8, and 12) include the main explanatory variable and the two interaction terms simultaneously. I will focus on the interpretation of the coefficients estimated in the final models. To conserve space, the coefficients regarding industry and time period effects, while estimated in all models, are not reported. [Insert Table 2 around here] The results provided support for Hypothesis 1, which stated that engaging in technology licensing will be positively related to a firm’s subsequent capacity to produce innovation. While the reported coefficients for Technology Licensing are positive and statistically significant (p<0.001) across all the models, it is possible to observe a decrease in their magnitude from Model 4 to Model 8 and Model 12. I estimated the marginal effects for the coefficients regarding 100 the Technology Licensing variable in order to compare their magnitudes in absolute terms17. The marginal effects estimated on the basis of the full models indicated that licensing-in is associated with an increase of 7.2% in firm patenting after a one-year lag, 5.8% after a two-year lag, and 5.2% after a three-year lag. Hypothesis 2 proposed that the effect of technology licensing on innovation will be moderated by the firm’s level of recoverable slack in such a fashion that increasing slack will reinforce the positive effect of licensing on the firm’s subsequent capacity to produce innovations. The coefficient for the interaction term between Technology Licensing and Recoverable Slack is positive and significant, exhibiting different levels of significance, across the models. This result suggests that the positive effect of licensing-in on firm patenting is augmented in conditions of high levels of recoverable slack, supporting the effect described in hypothesis 2. Additionally, I used a Wald test to verify whether the combined effect of this interaction term and Technology Licensing are simultaneously equal to zero, which would suggest that removing the interaction term would not significantly reduce the model fit. The results for the three dependent variables rejected the null hypothesis that both terms are simultaneously equal to zero (for a one-year lag: chi2 (2) = 49.47, p<0.001). Finally, the results supported the moderation effect predicted in Hypothesis 3 regarding the fact that the effect of technology licensing on innovation will be moderated by the firm’s level of organizational myopia in such a fashion that increasing myopia will weaken the positive effect of licensing on the firm’s subsequent capacity to produce innovations. Accordingly, the interaction term between technology licensing and organizational myopia produced statistically significant and negative coefficients (p<0.001). This finding supports the idea that organizational myopia negatively moderates the relationship between technology 17 The other independent variables were set to equal the sample means. 101 licensing and firm innovation. I also used the Wald test to check whether the joint effect of Technology Licensing X Organizational Myopia and the main variable is statistically different from zero. The results also indicated that the inclusion of the interaction term creates a statistically significant improvement in the fit of the model for the three dependent variables (for a one year lag: chi2 (2) = 63.84, p<0.001). In sum, the overall results support the idea that recoverable slack and organizational myopia are important moderators for firms’ capacity to produce innovations out of technology licensing-in. To illustrate the magnitude of the interaction effect, Figure 2 presents the plot of the two interactions regarding Recoverable Slack and Organizational Myopia with the variable Technology Licensing. The graphs are calculated on the basis of the effect of one standard deviation above and below the mean. The graphic representation of the interaction effects is consistent with the results in model 4 reported on table 2, with increasing levels of slack leading to an increase in the positive effect of licensing on patenting. On the other hand, the graph shows increasing levels of myopia leading to a decrease in the effect of licensing-in on firm patenting. [Insert Figure 2 around here] The examination of the control variables reveals that, as expected, Size and R&D Intensity had significant and positive effects on firm patenting. In addition to that, Licensing Experience also displays a positive and significant coefficient. This result could be attributed to either the residual effect of previous licensing deals on firm patenting or an improvement on firm capacity to deal with licensing through accumulated experience. The fact that the coefficients for Presample Patents remained significant across all the models reinforces the importance of controlling for firm heterogeneity in terms of “pre-entry” knowledge stock. 102 Finally, the result related to the direct effect of Recoverable Slack also merits more detailed discussion. The main term for Recoverable Slack remained negative and highly significant across all the models (p<0.001), which suggests that this type of slack has a negative effect on firm patenting. This result is similar to the one found by Geiger & Makri (2006), which reported that for firms with high levels of R&D intensity (which is the case for pharmaceutical firms) the effect of Recoverable Slack on the firms’ number of new patents is negative. Furthermore, Geiger & Makri also found that Recoverable Slack is particularly important for firms’ engagement in exploratory innovation18, which is also in line with my findings. Scholars have attributed the negative effect of slack on firm innovation to inefficiencies associated with resource allocation and adaptation processes (Nohria & Gulati, 1996). 3.8 Robustness Checks I performed a number of additional robustness checks to verify whether the main results are sensitive to alternative specifications. First, I analyzed the data using a Poisson model with fixed effects. I chose to re-estimate the analysis using a Poisson model because although it is likely to suffer from overdispersion, the estimators can be regarded as true fixed effects providing robust results related to unobservable (stable) firm attributes (Allison & Waterman, 2002). Table 3 reports the results for a one-year lag19 without the time invariant explanatory variables. The Poisson model produced similar estimators to the Negative Binomial model with fixed effects, suggesting that the specification used in the main regression analysis adequately accounts for the simultaneous effect of firms’ stable attributes on the dependent and independent variables. I also assessed the marginal effects of the Technology Licensing variable on firm patenting using the 18Laursen et al., (2010) show that licensing-in is an effective mechanism for firms to engage in exploratory search. 19 I choose to report a one-year lag in the robustness check because that is the period in which the effect of licensing-in on patenting is more pronounced, but the results for the other two dependent variables (Patent t+2; Patents t+3) are qualitatively similar to the Negative Binomial estimators. 103 Poisson estimator; the result indicates a reduction of 3% compared to the Negative Binomial model. [Insert Table 3 around here] A source of potential concern regards the presence of significant heterogeneity across the technologies in the sample. The empirical strategy that I chose to test the hypotheses did not allow me to incorporate variables that relate only to the licensing contracts and the licensors. Nevertheless, I also investigated potential bias related to the origin of the licensed technology, if it comes from a university or from another firm. In fact, Jensen & Thursby (2001) concluded that in most cases, the technologies commercialized by universities are at early stages, requiring substantial further work to reach a stage that would allow them to be commercially exploited by firms. In my sample, 12% of the licensing contracts have a university as the licensor. While this is not a large number, I still decided to investigate whether there are difference between those contracts and pure inter-firm agreements. Although I cannot assess precisely the stage of development related to the technologies in the sample, I have information regarding the age of those technologies20. Considering the full sample, the average age of the licensed technologies is 6.3 years. However, the results of a t-test indicate that when accounting only for the contracts coming from universities, this value is reduced to 4.8 years. Considering this fact, I also estimated the econometric models without university contracts and the results remained the same in terms of significance and direction. In addition, I also calculated the marginal effect for the variable technology licensing on the dependent variable using a one-year lag; the results indicate an increase in the order of 1.8% in the sample without university contracts. As expected, if the licensed technology is more developed, the effect on firm 20 While age is not an ideal proxy for the stage of development of a certain technology, I expect that the higher the age, the longer the time the technology had to be refined and further developed. 110 Atuahene-Gima, K. 1993. Determinants of inward technology licensing intentions: An empirical analysis of Australian engineering firms. Journal of Product Innovation Management, 10(3): 230-240. Blundell, R., Griffith, R., & Reenen, J. V. 1995. Dynamic Count Data Models of Technological Innovation. Economic Journal, 105(429): 333-344. Bourgeois, L. J. 1981. On the Measurement of Organizational Slack. Academy of Management Review, 6(1): 29-39. Bromiley, P. 1991. Testing a Causal Model of Corporate Risk Taking and Performance. The Academy of Management Journal, 34(1): 37-59. Cameron, A. C. & Trivedi, P. K. 1986. Econometric Models Based on Count Data: Comparisons and Applications of Some Estimators and Tests. Journal of Applied Econometrics, 1(1): 29-53. Cassiman, B. & Veugelers, R. 2006. In Search of Complementarity in Innovation Strategy: Internal R&D and External Knowledge Acquisition. Management Science, 52(1): 68-82. Ceccagnoli, M. & Hicks, D. 2012. Complementary Assets and the Choice of Organizational Governance: Empirical Evidence From a Large Sample of U.S. Technology-Based Firms. IEEE Trans. Eng. Manage. IEEE Transactions on Engineering Management: 1-14. Ceccagnoli, M. & Jiang, L. 2012. The cost of integrating external technologies: Supply and demand drivers of value creation in the markets for technology. Strategic Management Journal(forthcoming ). Ceccagnoli, M. & Jiang, L. 2013. The cost of integrating external technologies: Supply and demand drivers of value creation in the markets for technology. Strategic Management Journal, 34(4): 404-425. Cheng, J. L. C. & Kesner, I. F. 1997. Organizational Slack and Response to Environmental Shifts: The Impact of Resource Allocation Patterns. Journal of Management, 23(1): 1. Chesbrough, H. W. 2003. Open innovation : the new imperative for creating and profiting from technology. Boston, Mass.: Harvard Business School Press. Choi, J. P. 2002. A dynamic analysis of licensing: The "boomerang" effect and grant-back clauses. International Economic Review, 43(3): 803-829. Cohen, W. M. & Levinthal, D. A. 1990. Absorptive Capacity: A New Perspective on Learning and Innovation. Administrative Science Quarterly, 35(1): 128-152. Conti, R., Gambardella, A., & Novelli, E. 2013. Research on Markets for Inventions and Implications for R&D Allocation Strategies. The Academy of Management Annals, 7(1): 717-774. 111 Das, T. K. & Teng, B.-S. 2000. Instabilities of Strategic Alliances: An Internal Tensions Perspective. Organization Science, 11(1): 77-101. David, P. A. & Olsen, T. E. 1992. Technology adoption, learning spillovers, and the optimal duration of patent-based monopolies. International Journal of Industrial Organization, 10(4): 517-543. Deeds, D. L. & Decarolis, D. M. 1999. The impact of stocks and flows of organizational knowledge on firm performance: an empirical investigation of the biotechnology industry. Strategic Management Journal. Dosi, G. & Marengo, L. 2007. On the Evolutionary and Behavioral Theories of Organizations: A Tentative Roadmap. Organization Science, 18(3): 491-502. Fleming, L. 2001. Recombinant uncertainty in technological search. Management Science, 47(1): 117-132. Fosfuri, A. 2006. The licensing dilemma: understanding the determinants of the rate of technology licensing. Strategic Management Journal, 27(12): 1141-1158. Gallini, N. T. & Winter, R. A. 1985. Licensing in the theory of innovation. RAND Journal of Economics (RAND Journal of Economics), 16(2): 237-252. Gambardella, A. 1992. Competitive advantages from in-house scientific research: The US pharmaceutical industry in the 1980s. Research Policy, 21(5): 391-407. Gambardella, A. & Giarratana, M. S. 2012. General technological capabilities, product market fragmentation, and markets for technology. Research Policy(forthcoming). Gans, J. S. & Stern, S. 2003. The product market and the market for “ideas”: commercialization strategies for technology entrepreneurs. Research Policy, 32(2): 333-350. Gans, J. S. & Stern, S. 2010. Is there a market for ideas? Industrial and Corporate Change, 19(3): 805-837. Geiger, S. W. & Makri, M. 2006. Exploration and exploitation innovation processes: The role of organizational slack in R & D intensive firms. The Journal of High Technology Management Research, 17(1): 97-108. George, G. 2005. Slack Resources and the Performance of Privately Held Firms. The Academy of Management Journal, 48(4): 661-676. Gourieroux, C., Monfort, A., & Trognon, A. 1984. Pseudo Maximum Likelihood Methods: Applications to Poisson Models. Econometrica, 52(3): 701-720. Grant, R. M. 1996. Prospering in Dynamically-Competitive Environments: Organizational Capability as Knowledge Integration. Organization Science, 7(4): 375-387. 112 Griliches, Z. 1990. Patent Statistics as Economic Indicators - a Survey. Journal of Economic Literature, 28(4): 1661-1707. Grindley, P. C. & Teece, D. J. 1997. Managing Intellectual Capital: Licensing and Cross- Licensing in Semiconductors and Electronics. California Management Review, 39(2): 8- 41. Gujarati, D. N. 2003. Basic Econometrics. 4th: New York: McGraw-Hill. Hausman, J., Hall, B. H., & Griliches, Z. 1984. Econometric Models for Count Data with an Application to the Patents-R&D Relationship. Econometrica, 52(4): 909-938. Heckman, J. J. 1979. Sample Selection Bias as a Specification Error. Econometrica, 47(1): 153- 161. Hornsby, J. S., Kuratko, D. F., & Zahra, S. A. 2002. Middle managers' perception of the internal environment for corporate entrepreneurship: assessing a measurement scale. Journal of Business Venturing, 17(3): 253-273. Jensen, R. & Thursby, M. 2001. Proofs and Prototypes for Sale: The Licensing of University Inventions. The American Economic Review, 91(1): 240-259. Johnson, D. K. N. 2002. "Learning-by-Licensing": R&D and Technology Licensing in Brazilian Invention. Economics of Innovation and New Technology, 11(3): 163-177. Katila, R. & Ahuja, G. 2002. Something Old, Something New: A Longitudinal Study of Search Behavior and New Product Introduction. The Academy of Management Journal, 45(6): 1183-1194. Katila, R. & Shane, S. 2005. When Does Lack of Resources Make New Firms Innovative? The Academy of Management Journal, 48(5): 814-829. Katz, R. & Allen, T. J. 1982. Investigating the Not Invented Here (NIH) syndrome: A look at the performance, tenure, and communication patterns of 50 R & D Project Groups. R&D Management, 12(1): 7-20. Keil, T., Maula, M., Schildt, H., & Zahra, S. A. 2008. The effect of governance modes and relatedness of external business development activities on innovative performance. Strategic Management Journal, 29(8): 895-907. Klevorick, A. K., Levin, R. C., Nelson, R. R., & Winter, S. G. 1995. On the sources and significance of interindustry differences in technological opportunities. Research Policy, 24(2): 185-205. Kotha, R., George, G., & Srikanth, K. 2013. Bridging the mutual knowledge gap: coordination and the commercialization of university science. Academy of Management Journal, 56(2): 498-524. 113 Lane, P. J. & Lubatkin, M. 1998. Relative absorptive capacity and interorganizational learning. Strategic Management Journal, 19(5): 461. Laursen, K. & Salter, A. 2006. Open for innovation: The role of openness in explaining innovation performance among U.K. manufacturing firms. Strategic Management Journal, 27(2): 131-150. Laursen, K., Leone, M. I., & Torrisi, S. 2010. Technological exploration through licensing: new insights from the licensee's point of view. Industrial and Corporate Change, 19(3): 871- 897. Leone, M. I. & Reichstein, T. 2012. Licensing-in fosters rapid invention! the effect of the grantback clause and technological unfamiliarity. Strategic Management Journal, 33(8): 965- 985. Levin, R. C., Klevorick, A. K., Nelson, R. R., & Winter, S. G. 1987. Appropriating the Returns from Industrial Research and Development. Brookings Papers on Economic Activity(3): 783. Levinthal, D. A. & March, J. G. 1993. The myopia of learning. Strategic Management Journal, 14: 95-112. March, J. G. 1991. Exploration and Exploitation in Organizational Learning. Organization Science, 2(1): 71-87. McGahan, A. M. & Silverman, B. S. 2006. Profiting from technological innovation by others: The effect of competitor patenting on firm value. Research Policy, 35(8): 1222-1242. Meyer, A. D. 1982. Adapting to Environmental Jolts. Administrative Science Quarterly, 27(4): 515-537. Mezias, S. J. & Glynn, M. A. 1993. The three faces of corporate renewal: institutional, revolution, and evolution. Strategic Management Journal, 14(2): 77-101. Miller, K. D. & Leiblein, M. J. 1996. Corporate Risk-Return Relations: Returns Variability versus Downside Risk. The Academy of Management Journal, 39(1): 91-122. Mulotte, L., Dussauge, P., & Mitchell, W. 2013. Does pre-entry licensing undermine the performance of subsequent independent activities? Evidence from the global aerospace industry, 1944–2000. Strategic Management Journal, 34(3): 358-372. Nohria, N. & Gulati, R. 1996. Is slack good or bad for innovation? Academy of Management Journal, 39(5): 1245-1264. Nooteboom, B., Van Haverbeke, W., Duysters, G., Gilsing, V., & van den Oord, A. 2007. Optimal cognitive distance and absorptive capacity. Research Policy, 36(7): 1016-1034. 114 Owen-Smith, J. & Powell, W. W. 2004. Knowledge Networks as Channels and Conduits: The Effects of Spillovers in the Boston Biotechnology Community. Organization Science, 15(1): 5-21. Oxley, J. E., Sampson, R. C., & Silverman, B. S. 2009. Arms race or détente? How interfirm alliance announcements change the stock market valuation of rivals. Management Science, 55(8): 1321-1337. Park, S. H. & Russo, M. V. 1996. When Competition Eclipses Cooperation: An Event History Analysis of Joint Venture Failure. Management Science, 42(6): 875-890. Paruchuri, S. 2009. Intraorganizational Networks, Interorganizational Networks, and the Impact of Central Inventors: A Longitudinal Study of Pharmaceutical Firms. Organization Science, 21(1): 63-80. Paruchuri, S. & Eisenman, M. 2012. Microfoundations of Firm R&D Capabilities: A Study of Inventor Networks in a Merger. Journal of Management Studies, 49(8): 1509-1535. Pentland, B. T., Feldman, M. S., Becker, M. C., & Liu, P. 2012. Dynamics of Organizational Routines: A Generative Model. Journal of Management Studies, 49(8): 1484-1508. Phelps, C. C. 2010. longitudinal study of the influence of alliance network structure and composition on firm exploratory innovation. Academy of Management Journal, 53(4): 890-913. Phene, A., Fladmoe-Lindquist, K., & Marsh, L. 2006. Breakthrough innovations in the U.S. biotechnology industry: the effects of technological space and geographic origin. Strategic Management Journal, 27(4): 369-388. Powell, W. W. 1990. Neither market nor hierarchy: Network forms of organization. Research in Organizational Behavior, 12: 295. Powell, W. W., Koput, K. W., & Smith-Doerr, L. 1996. Interorganizational Collaboration and the Locus of Innovation: Networks of Learning in Biotechnology. Administrative Science Quarterly, 41(1): 116-145. Radjou, N. P. J. C. A. S.; Jugaad innovation think frugal, be flexible, generate breakthrough growth. Rivette, K. G. 1993. Corporate patent portfolios. Technology Forecast. Price Waterhouse: New York., Price Waterhouse: New York. Roberts, P. W. 1999. Product innovation, product-market competition and persistent profitability in the U.S. Strategic Management Journal, 20(7): 655-670. Rosenkopf, L. & Nerkar, A. 2001. Beyond Local Search: Boundary-Spanning, Exploration, and Impact in the Optical Disk. Strategic Management Journal, 22(4): 287. 115 Rosenkopf, L. & Almeida, P. 2003. Overcoming local search through alliances and mobility. Management Science, 49(6): 751-766. Rotemberg, J. J. & Saloner, G. 1994. Benefits of Narrow Business Strategies. The American Economic Review, 84(5): 1330-1349. Sampson, R. C. 2007. R&D Alliances and Firm Performance: The Impact of Technological Diversity and Alliance Organization on Innovation. Academy of Management Journal, 50(2): 364-386. Saner, R., Yiu, L., & Søndergaard, M. 2000. Business diplomacy management: A core competency for global companies. Academy of Management Executive, 14(1): 80-92. Schilling, M. A. & Phelps, C. C. 2007. Interfirm Collaboration Networks: The Impact of Large- Scale Network Structure on Firm Innovation. Management Science, 53(7): 1113-1126. Schilling, M. A. 2009. Understanding the alliance data. Strategic Management Journal, 30(3): 233-260. Schreyögg, G. & Kliesch-Eberl, M. 2007. How dynamic can organizational capabilities be? Towards a dual-process model of capability dynamization. Strategic Management Journal, 28(9): 913-933. Singh, J. V. 1986. Performance, slack, and risk taking in organizational decision making. Academy of Management Journal, 29(3): 562-585. Sørensen, J. B. & Stuart, T. E. 2000. Aging, Obsolescence, and Organizational Innovation. Administrative Science Quarterly, 45(1): 81-112. Teece, D. J. 1986. Profiting from technological innovation: Implications for integration, collaboration, licensing and public policy. Research Policy, 15(6): 285-305. Teece, D. J., Pisano, G., & Shuen, A. 1997. Dynamic capabilities and strategic management. Strategic Management Journal, 18(7): 509-533. Trajtenberg, M. 1987. Patents, Citations and Innovations: Tracing the Links: National Bureau of Economic Research, Inc, NBER Working Papers: 2457. Van Den Bosch, F. A., Volberda, H. W., & De Boer, M. 1999. Coevolution of firm absorptive capacity and knowledge environment: Organizational forms and combinative capabilities. Organization Science, 10(5): 551-568. Verbeek, M. & Nijman, T. 1992. Testing for Selectivity Bias in Panel Data Models. International Economic Review, 33(3): 681-703. Wilcox King, A. & Zeithaml, C. P. 2003. Measuring organizational knowledge: a conceptual and methodological framework. Strategic Management Journal, 24(8): 763-772. 116 Wooldridge, J. M. 2012. Introductory econometrics: a modern approach: Cengage Learning. Wuyts, S., Colombo, M. G., Dutta, S., & Nooteboom, B. 2005. Empirical tests of optimal cognitive distance. Journal of Economic Behavior & Organization, 58(2): 277-302. Ziedonis, R. H. 2004. Don't Fence Me In: Fragmented Markets for Technology and the Patent Acquisition Strategies of Firms. Management Science, 50(6): 804-820. 117 3.11 Figures Figure 1: Distribution of Licensing Events and Number of Firms (1983-2004) Figure 2: Graph of Interactions of Fixed Effects Models (Patents t+1) 0 50 100 150 200 250 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 Frequency Year Number of Licensing Events Number of Firms 0 0,2 0,4 0,6 0,8 1 1,2 1,4 Licensing (0) Licensing (1) Technology Licensing Low Myopia High Myopia 118 0 0,5 1 1,5 2 2,5 3 3,5 4 4,5 Licensing (0) Licensing (1) Technology Licensing Low Slack High Slack 119 Table 1. Descriptive Statistics and Correlation Coefficients (N = 2.509) Mean S.D. (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (1) Patents it+1 264.138 598.055 1.00 (2) Technology Licensing 0.180 0.384 0.11 1.00 (3) Recoverable Slack 1.077 1.114 0.02 -0.04 1.00 (4) Organizational Myopia 0.335 0.293 -0.03 -0.00 0.03 1.00 (5) Firm Size 6.548 2.558 0.50 0.08 0.19 0.05 1.00 (6) Firm R&D Intensity 149.692 286.873 -0.17 0.01 -0.16 -0.07 -0.38 1.00 (7) Licensing Experience 0.496 0.734 0.10 0.14 -0.01 -0.03 0.08 0.01 1.00 (8) Technology Diversity 0.612 0.303 0.23 0.03 0.01 -0.27 0.36 -0.09 0.01 1.00 (9) Evaluation Capacity 18.239 27.767 -0.18 0.04 -0.21 -0.13 -0.28 0.23 0.13 -0.24 1.00 (10) Technological Complexity 15.511 11.160 -0.11 -0.03 0.02 0.08 -0.16 0.04 0.02 -0.42 0.35 1.00 (11) Patenting Experience 9.279 7.102 0.23 0.02 0.22 -0.03 0.54 -0.26 0.15 0.15 -0.10 0.09 1.00 (12) Technological Collaborator 2.747 9.474 0.09 0.05 0.07 -0.04 0.26 -0.05 0.11 0.13 0.06 0.02 0.30 1.00 (13) Industry Competition 146.007 84.928 -0.10 0.06 -0.01 -0.11 0.03 0.11 0.23 -0.11 0.15 0.04 0.15 0.13 1.00 (14) U.S./Canada Firm 0.959 0.198 0.07 -0.04 -0.10 -0.07 -0.18 0.02 0.10 -0.10 0.09 0.08 -0.07 -0.14 -0.01 1.00 (15) Presample Patents 72.675 258.379 0.29 0.09 0.10 0.01 0.44 -0.13 0.02 0.23 -0.13 -0.07 0.35 0.57 -0.05 -0.30 1.00 119 Table 1. Descriptive Statistics and Correlation Coefficients (N = 2.509) Mean S.D. (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (1) Patents it+1 264.138 598.055 1.00 (2) Technology Licensing 0.180 0.384 0.11 1.00 (3) Recoverable Slack 1.077 1.114 0.02 -0.04 1.00 (4) Organizational Myopia 0.335 0.293 -0.03 -0.00 0.03 1.00 (5) Firm Size 6.548 2.558 0.50 0.08 0.19 0.05 1.00 (6) Firm R&D Intensity 149.692 286.873 -0.17 0.01 -0.16 -0.07 -0.38 1.00 (7) Licensing Experience 0.496 0.734 0.10 0.14 -0.01 -0.03 0.08 0.01 1.00 (8) Technology Diversity 0.612 0.303 0.23 0.03 0.01 -0.27 0.36 -0.09 0.01 1.00 (9) Evaluation Capacity 18.239 27.767 -0.18 0.04 -0.21 -0.13 -0.28 0.23 0.13 -0.24 1.00 (10) Technological Complexity 15.511 11.160 -0.11 -0.03 0.02 0.08 -0.16 0.04 0.02 -0.42 0.35 1.00 (11) Patenting Experience 9.279 7.102 0.23 0.02 0.22 -0.03 0.54 -0.26 0.15 0.15 -0.10 0.09 1.00 (12) Technological Collaborator 2.747 9.474 0.09 0.05 0.07 -0.04 0.26 -0.05 0.11 0.13 0.06 0.02 0.30 1.00 (13) Industry Competition 146.007 84.928 -0.10 0.06 -0.01 -0.11 0.03 0.11 0.23 -0.11 0.15 0.04 0.15 0.13 1.00 (14) U.S./Canada Firm 0.959 0.198 0.07 -0.04 -0.10 -0.07 -0.18 0.02 0.10 -0.10 0.09 0.08 -0.07 -0.14 -0.01 1.00 (15) Presample Patents 72.675 258.379 0.29 0.09 0.10 0.01 0.44 -0.13 0.02 0.23 -0.13 -0.07 0.35 0.57 -0.05 -0.30 1.00 126 The grant-back clause “requires the potential licensee to agree to grant back to the patentee [i.e., the licensor] rights to improvement patents developed by the licensee that relate to the original patent as partial consideration for the license right” (Schmalbeck 1975: 733). In this context, Leone and Reichstein (2012) show that generally licensees achieve more rapid innovation than comparable non-licensees but that this effect is negated by the inclusion in the contract of a grant-back clause. It is argued that this is because the grant-back clause reduces the licensee’s incentive to develop the licensed technology further. This paper aims to increase our rather limited knowledge on the contingencies that determine when the contracting parties agree to the inclusion of a grant-back clause, by focusing on the factors that might explain its inclusion in a technology licensing agreement. In this case, it is a result of negotiation between licensor and licensee. To our knowledge, this attempt to try systematically to explain this phenomenon is unique. We consider especially the characteristics of the licensed technology in terms of the similarities between licensor’s and licensee’s knowledge bases, and technical uncertainties about the scope, level, and quality of follow-on innovations. We combine insights from the resource-based view (RBV) of the firm and contract economics to build theoretical arguments regarding the inclusion of grant-back clauses in technology licensing contracts. Our fundamental theoretical argument centers on the contracting firm’s need to balance protection of its technological resources with learning through internal and external processes. We argue that licensing agreements are increasingly more likely to include a grant-back clause, the closer is the licensed technology to the core of the licensor’s patent portfolio. On the licensee’s side, we conjecture that such agreements are decreasingly likely to include a grant-back clause, the closer the licensed technology to the core of the licensee’s patent portfolio. 127 In addition to those effects described above, we also hypothesize that a licensing agreement will be increasingly likely to include a grant-back clause if the licensed technology is associated with high uncertainty. We expect that the contracts negotiating uncertain technologies are more likely to include the grant-back clause as licensors have incentives to reduce future threats that uncertain technologies may pose. However, these variables interact in important ways: We would argue that technology licensing agreements involving technologies that are core to the licensor and are at the same time uncertain will further increase the probability of inclusion of a grant back-clause. We argue also that the lower likelihood of a grant-back clause in technology licensing agreements involving technologies that are related to the licensee’s core knowledge will be reversed if the licensed technology is uncertain. We test the proposed hypotheses using a sample of 404 licensed technologies extracted from Recombinant Capital’s Biotech Alliance (Recap) Database for the period 1984- 2004. We merge the information retrieved for licensing deals with patent data from the NBER project and with firm information from the COMPUSTAT database. We employ a hierarchical nested decision model to account for the fact that the inclusion in a licensing contract of a grantback clause is nested in the decision about which technologies to out-license. Accordingly, the empirical model comprises a two-level asymmetric nested tree in which the option of a grantback clause is available only if the licensor decides to out-license a technology. To implement this technique we use the unique USPTO (United States Patent and Trademark Office) patent number assigned to each technology in the licensing database to estimate the likelihood that a specific technology will be licensed. Then we estimate the extent to which the licensed technology represents a core technological activity for licensor and licensee, and the level of technological uncertainty, to estimate the likelihood that a grant-back clause will be included in the contract. We find overall empirical support for our theoretical arguments. 128 4.2 Theoretical Background The RBV and the competence-based view of the firm highlight that sustained competitive advantage can be achieved through ownership of valuable resources that are imperfectly mobile and imperfectly imitable (Barney 1991, Peteraf 1993, Wernerfelt 1984). We draw on these perspectives and the extension proposed by Lavie (2006) who relaxes the fundamental assumption of the original RBV that firms must own or at least fully control the resources that confer competitive advantage. Relaxation of this assumption is central to the extended RBV which claims that firms achieve sustained competitive advantage through collaboration with other firms. The conventional RBV frame considers only internal rents. Internal rents are a combination of Ricardian rents and quasi-rents derived from the focal firm’s internal resources (Lavie 2006, Peteraf 1993). Lavie (2006) also considers three additional types of rents that are related to the focal firm’s external relations. Appropriated relational rents refer to the joint benefit that accrues to collaboration partners through the combination, exchange, and codevelopment of unique resources, and are the rents considered in the relational view of the firm (Dyer and Singh 1998). Inbound spillover rents are the private benefits that are derived by the focal firm exclusively from external resources subject to unintended leakages of knowledge from collaboration partners related to the partners’ shared and non-shared resources.22 Outbound spillover rents refer to the opposite situation where the unintended leakage of the resources of the focal firm produces private benefits for the collaborating partners, thereby reducing the focal firm’s competitive advantage. In our hypotheses we apply these distinctions to the types of rents relating to the benefits/rents that can be gained or lost in interactions with collaboration partners. 22 The two-sided nature of spillovers is acknowledged and exploited analytically in Myles, Shaver and Flyer (2000). The notion of inbound and outbound spillovers corresponds to Cassiman and Veugelers’s (2002) concepts of incoming and outgoing spillovers, which are applied for instance, in Alcácer and Chung (2007). 129 In the present paper, we focus on technological resources. In many industries, technological resources are vital for sustained competitive advantage (Silverman 1999). To understand the inclusion of a grant-back clause in licensing contracts we combine insights from the RBV of the firm with the logic of incomplete contract theory in contract economics. Under the assumption of information asymmetry between trading parties, incomplete contract theory studies whether either of the contracting parties has an incentive to act opportunistically, and relatedly, whether either party has an incentive to invest effort in the transaction (Bolton and Dewatripont 2005). In this context, Choi (2002) is a particularly important contribution which is based on incomplete contract theory and addresses the already mentioned boomerang effect. The boomerang effect implies that granting licensees the right to use the licensors’ intellectual property “may enable them to develop new products, which make the licensed technology obsolete and leave the licensor in backwater of technology” (Choi 2002: 804). This possibility indicates that the risk of increasing future competition can distort the licensing relationship, preventing deals to take place. Choi (2002) shows that a quantitydependent royalty payment can act as a “hostage” to help facilitate the transfer of the cuttingedge technology. However, even though the use of royalty payments can deal with the licensor’s concerns regarding monetary compensations, the use of a royalty rate might incur in high costs for the licensee. Accordingly, if the royalty payments are too high, then the best technology will not be traded. As an alternative, a grant-back clause can reduce the costs imposed by a quantitydependent royalty payment, thereby facilitating trade by providing both parts the incentives to trade cutting-edge technologies. However, depending on the type of technology and its importance in the trading parties’ technological portfolios—related to their ability to maintain competitive advantage—a grant-back clause will be more or less likely to be included in the licensing contract. 130 4.3 Hypotheses 4.3.1 Licensors’ core technologies A given firm possessing technological resources will have a portfolio of core and non-core technologies related to its activities (Granstrand, Patel and Pavitt 1997). The firm’s core technologies will be underpinned by a set of in-house core competencies (Prahalad and Hamel 1990). The licensing literature generally does not consider the licensing out of core technologies, because of the potential for loss of competitive advantage (see e.g., Caves, et al. 1983). However, firms do license-out core technologies if they can avoid creating direct competitors (for an overview, see Leone and Laursen 2011). In these cases, licensing contracts are more likely to include a grant-back clause for three main reasons. First, the potential boomerang effect will be more severe for the licensor if core technology and related core competencies are involved, because of the competitive advantage and internal rents they provide. If these advantages are eliminated through the boomerang effect and resulting obsolescence of the core technology, the consequences for the licensor may be dire: There will be a strong likelihood that competition in the market will increase and new competitors will emerge. However, a grant-back clause reduces these risks. From the licensee’s perspective, although the inclusion of a grant-back may reduce the potential advantages of investing in the licensing deal, it still assures access to a technology with a high potential, which subsequently may result in a successful commercial exploitation. Second, assigning more residual rights of control to the principal (licensor) shifts the incentives for opportunistic and distorting behavior as a result of the lower ex-post returns to the agent (licensee). As a result, the licensee will commit fewer resources to the development of the in-licensed technology since the potential for achieving competitive advantage based on 131 development of the technology will be reduced. Van Dijk (2000: 1433) states that “future exchange clauses obviously weaken (licensee’s) incentives to improve current technology.” For licensees, the potential competitive advantage deriving from technology improvements is reduced because the advances achieved have to be transferred to the licensor. For the licensor, the risk of being overtaken by the licensee in a core technology is reduced as the result of a grant-back clause which increases the chances of maintaining the internal rents related to the core technology. Third, the inclusion of a grant-back clause provides an incentive for the licensor to assist the licensee in developing the technology further and forging a collaborative learningrelated arrangement based on the license (Leone and Reichstein 2012). A collaborative arrangement will increase the possibility of outbound spillover rents from the point of view of the licensor. This may be particularly dangerous if the collaboration is related to core technologies and the associated core competencies. However, the inclusion of a grant-back clause means that the potential outbound spillover rents become realized relational rents. In sum, we propose: Hypothesis 1: Technology license agreements, ceteris paribus, are increasingly likely to include a grant-back clause the closer the licensed technology is to the core of the licensor’s patent portfolio. 4.3.2 Licensees’ core technologies There are many reasons why firms choose to in-license technologies. In the standard licensing literature, potential licensees are attracted by rapid access to technologies that have been developed and proven by their licensors in other competitive arenas (Atuahene-Gima 1992, Atuahene-Gima 1993). In this literature, licensing-in is seen as a tactical response to a shortfall in internal R&D capabilities (Lowe and Taylor 1998). However, in-licensing is also considered 132 a learning mechanism allowing combinations of complementary pieces of internal and external knowledge (Choi 2002, Johnson 2002, Laursen, Leone and Torrisi 2010, Leone and Reichstein 2012, Lowe and Taylor 1998). According to Choi (2002: 807): “licensing of a new technology serves as a stepping stone for further developments of the licensed technology.” We hypothesize that a licensing agreement related to a core technology in the licensee’s patent portfolio is very unlikely to contain a grant-back clause for two main reasons. The first is the level of absorptive capacity needed to integrate the in-licensed technology (Laursen, et al. 2010). Cohen and Levinthal (1990: 128) argue that a firms’ absorptive capacity is “largely a function of the level of prior related knowledge.” This related prior knowledge allows “the firm to better understand and therefore evaluate the import of intermediate technological advances that provide signals to the eventual merit of a new technological development” (Cohen and Levinthal 1990: 136). If the licensee’s core technology and related core competencies are technologically proximate to the in-licensed technology, it is likely that the licensee firm will have identified the appropriate patent and will be able to assimilate and exploit the knowledge contained in the in-licensed technology (for a similar logic applied to R&D-related strategic alliances, see Mowery, Oxley and Silverman 1996): Based on its prior knowledge, the licensee will probably understand the externally acquired technology. However, in the case of an unfamiliar technology, it will be difficult for the licensee to integrate the licensed technology into its activities. In this case, a grant-back clause will create an incentive for the licensor to help the licensee to integrate and develop this new technology (Leone and Reichstein 2012). This is unnecessary for a technology that is close to the licensee’s core technologies. The second and closely related reason is that licensing agreements entail the risk of involuntary bidirectional spillovers. They hold the potential for outbound spillover rents seen 133 from the point of view of either party. As already mentioned, grant-back clauses are often included to facilitate technological cooperation, and provide the licensor with an incentive to help the licensee understand, integrate, and develop the in-licensed knowledge. The licensor might be encouraged to invest extra time and resources to provide supplementary knowledge if there is some potential for future benefits from grant-backs (Parr and Smith 2005). However, when the licensed in technology is close to the licensee’s core technology (involving a high probability of absorption without direct technological collaboration with the licensor) then the licensee will not be keen to work cooperatively. It will be more interested in avoiding leaks of knowledge about core technologies and related core competencies to the licensor (inbound spillover rents for the licensor/outbound spillover rents for the licensee). Indeed, if the licensee is legally committed to transfer back to the licensor any technological improvement related its core technologies, there is always the risk that the licensee might lose its competitive advantages. On this basis, we posit the following: Hypothesis 2: Technology license agreements, ceteris paribus, are decreasingly likely to include grant-back clauses the closer the licensed technology is to the core of the licensee’s patent portfolio 4.3.3 Licensors and uncertain technology Another aspect affecting the structure of contracts is the level of uncertainty of the licensed technology. According to Ziedonis (2007: 2624), technological uncertainty is related to “the commercial potential of the patent [...] and is likely to be higher for technologies that are more ‘basic’ or more ‘distant’ from commercialization.” It is more difficult to forecast the technical performance and feasibility of these types of technologies since they have not been commercialized. In other words, technological uncertainty refers to the uncertain future payoffs from investment in the new technology (Ziedonis 2007). In line with this definition, our measure 134 of uncertainty is intended to reflect the level of uncertainty related to the future development of the licensed technology. If the licensor cannot predict the future trajectory of the licensed technology it will be less inclined to license it out. A licensee’s attempt to improve on the original technology may fail totally, may result in an incremental improvement to the original technology, or may materialize as a radical innovation that challenges the licensor’s competitive advantage. Therefore, the licensor cannot predict what might happen but will hedge against the worst possible outcome, which could result in it being overtaken by the licensee. As we pointed out earlier, the inclusion of a grant-back clause provides a safeguard to the licensor by ensuring it access to the results of developments undertaken by the licensee before the latter can exploit them in the market (Schmalbeck 1975). Hence, the potential utility of a grant-back clause increase with the uncertainty of the technology since such a clause may facilitate technology transfer that otherwise would not have happened due to the licensor’s fear of being overtaken by the licensee. In the case of uncertainty about the future of a technology, joint learning may attribute more competencies to the development of the intellectual asset than possessed by either individual firm, as well as introduce risk sharing. This may hence be a feasible way to resolve development problems (e.g., Mariti and Smiley 1983, Sampson 2007). A grant-back clause provides the licensor with an incentive to collaborate with the licensee over the technology, which can benefit both contracting parties. Based on these considerations, we conjecture that: Hypothesis 3: Technology license agreements are, ceteris paribus, increasingly likely to include a grant-back clause with increasing levels of uncertainty. 4.3.4 Licensors’ core and uncertain technologies We have argued that licensing agreements are more likely to include a grant-back clause if they represent a core technology of the licensor, or if there is uncertainty about the future 135 development of the technology. When both conditions hold, this should further increase the likelihood of a grant-back clause. We argue that if the given development of the technology is relatively predictable and “safe”, then even if it is a core technology of the licensor, the license may not include a grant-back clause because the risk of a critical boomerang effect will be small. If the technology is uncertain and non-core, the licensor firm might consider not including a grant-back clause because the damage caused by a boomerang effect—and the consequential reduction in the internal rents—will likely be small in the case of a non-core technology. However, if the technology is core to the licensor and also uncertain regarding future opportunities, the potential damage to the competitive advantage of the licensor firm could be huge. Therefore, we posit that the relationship between licensors’ core technologies and the likelihood of a using the grantback clause will be positively moderated by the level of uncertainty related to the licensed technology. In other words, we expect that increasing levels of uncertainty will reinforce the positive effect of licensor’s core technology on the likelihood of including the grant-back clause in a contract. Consequently, we suggest: Hypothesis 4: Ceteris paribus, the increasing likelihood of a grant-back clause appearing in technology licensing agreements involving technologies that are core to the licensor increases further when the uncertainty of the licensed technology is high. 4.3.5 Licensees, and core and uncertain technologies We have argued that the inclusion of a grant back-clause will be less likely if the licensed technology is close to the licensee’s core technologies. However, we would argue also that this effect will be reversed if the technology is also uncertain in relation to its application and development. The theory is that, as discussed above, the effect of uncertainty on its own typically leads the licensor to require the inclusion of a grant-back clause to reduce the 142 (Somaya, et al. 2011). Therefore, the use of this clause has implication for both parts involved in the licensing deal. While for licensee it provides the right incentives to commit more resources and actively engage in the exploitation of the licensed technology, for the licensor it is a relevant contractual mechanism to ensure the commercial success of the deal. Accordingly, given the value creation possibilities associated with this clause, we expect that the grant-back clause is more likely to be used in exclusive licensing contracts. Technological overlap: The decision to include the grant-back clause in a licensing contract might also be affected by the extent to which licensor and licensee build on the same technological fields. In order to capture technological overlap we use the measure proposed by Jaffe (1986) which indicates the technological position of firm A relative to firm B in terms of the technological classes in which both firms have patented. In order to construct our measure we generated the technological profile of licensors and licensees by computing the distribution of accumulated patents across different classes in the five years previous to the licensing contract. Accordingly, we obtained a multidimensional vector F = F,…FG where F5 represents the number of patents assigned to firm i in the patent class s. The final measure is computed as follows: Technological overlap: HIHJ L(HIHN) HPHJ This measure takes the value 0 for firms that have orthogonal vectors, value 1 for firms with identical vectors, and a value between 0 and 1 for the cases in which there is an intermediate degree of orthogonality between the firms. Patent value: Following the convention in patent studies (Lahiri 2010, Trajtenberg 1990, Yang, Phelps and Steensma 2010, Ziedonis 2007), we proxy the economic value of a technology as a 143 time invariant measure of the total number of forward citations received by a patent from its date of publication to 2006.27 Technology radicalness: Radical technologies have higher potential to produce significant changes in the way that economic activities are organized (Shane, 2001). Therefore, we expect licensors to be more likely to request the use of the grant-back clause as the level of radicalness of a certain technology increases. The radicalness of the technology is measured following Rosenkopf and Nerkar (1999). They use the number of different three-digit level International Patent Classification (IPC) categories related to the patents cited by the patent for the focal technology, excluding the class of the focal patent. The fact that a patent’s backward citations refer to different classes (from its own) indicates that the invention builds on several different technological fields (Shane 2001). Technology scope: Technology scope is an indication of its applicability which may be a sign of the potential for further development and may increase the licensor’s incentive to include a grant-back clause in a licensing contract with another firm. We follow the measure for scope proposed in Lerner (1994), which considers the number of IPCs that USPTO assigns to a patent as an indication of the breadth of its technology base and intellectual property protection. Technology age: The age of a technology can influence the licensor’s decision to commercialize it by licensing it out or exploiting it in-house. Several studies suggest that licensors are less likely to license out technologies that might undermine their competitive position in the industry (see e.g. Leone and Reichstein 2012). Firms will therefore be less likely to commercialize more recent inventions, given that these technologies supposedly are at the technological frontier of their inventive activities. 27 The latest year available in the NBER patent database. 144 Backward citations: It has been claimed that the total number of backward citations in a patent is a good indicator of the size of the technological space and scope of the intellectual property rights of a given technology (Harhoff and Reitzig 2004, Reitzig, Henkel and Schneider 2010). Technologies with a large number of backward citations may be more likely to be licensed since they may overlap more technological actors and be attractive to more agents on the demand side of the market for technology. R&D intensity: The firm’s relative R&D expenditure may affect its decisions regarding technology licensing. Firms that are R&D intensive are likely to be less dependent on specific technologies, while firms with low levels of in-house R&D are likely to have fewer technological opportunities (Dosi, Marengo and Pasquali 2006). It is likely also that firms that invest hugely in R&D are pursuing purely technology driven strategies which do not include traditional commercialization, and whose profits lie in exchanges of intellectual property. In addition to that, firms that exhibit high levels of R&D intensity are also more likely to have inhouse capabilities to internally develop and exploit technologies. This may introduce heterogeneity in the decision to enter the markets for technology. R&D intensity is measured as firm i’s total amount of R&D investment divided by its sales in year t. Licensor technological specialization: The firm’s level of technological specialization is likely to affect the way it operates in the markets for technology: narrower technological scope renders the firm more susceptible to rent dissipation when licensing core technologies. Therefore, we include a measure of technological specialization by calculating a Herfindahl index for the total number of patents in the firm j’s patent portfolio accumulated during in the seven years before the license agreement. We operationalize this measure as follows: Licensor technological specialization: ∑+IP +I. 3, 145 Firm slack. The availability of slack resources can affect the novelty of innovations (Nohria and Gulati 1996). Given that a firm’s ability to introduce innovations characterized by a high degree of novelty may affect the firm’s licensing decision, we control for firm i's slack, using the ratio current assets/current liabilities in year t. Firm size. We control for firm size using the logarithm of total number of employees in a given year. Licensor market diversification: We control for the number of different markets in which the licensor operates by counting the total number of different SIC codes reported in the COMPUSTAT database at year t. This may spread the risks for the licensor, which might influence the inclusion or not of a grant-back clause. Technological fragmentation. The degree of fragmentation of ownership in the firm’s patent portfolio has been shown to affect patenting behavior and the strategic decisions related to exploiting the market to commercialize new technologies (Ziedonis 2004). We control for fragmentation of ownership rights of firm j’s patents produced at year t are using the fragmentation index proposed in Ziedonis (2004): Technological fragmentation = 1−∑+QRSUVGIP +QRSUVGI.,≠Y 3, , where j refers to the unique entities cited by the patents granted to firm i in a given year. Based on this idea, !Z[7\ concerns to the aggregate(total) number of backward citations present in firms’ i patents within year t, and !Z[7\ to the number of unique entities listed in those backward citations28. 28 In line with how the measure was calculated originally we exclude from the backward citations references to the firm’s own patents, to expired patents, and to scientific references. 146 Sales change. Licensors that experience a decrease in their sales may be under pressure to generate short-term revenue by licensing their more valuable technologies (Katz and Shapiro 1986). Percentage change in licensor’s sales between the licensing years t and t-1 is used to control for a licensing decision motivated by financial pressures. Finally, patenting propensity varies across years and industry segments, resulting in the need to protect an invention differing across and within the firms in our sample. To account for these effects we include dummy variables for biotech firms and medical firms (in the pharmaceutical industry) and year of the licensing contract. 4.5 Econometric analysis and model choice With a categorical multinomial dependent variable, the first modeling choice is a multinomial logit. However, the grant-back clause cannot be considered independently of the likelihood that a technology is licensed-out. The likelihood of a technology being licensed may have an impact on the inclusion of a grant-back clause, and the grant-back clause may be subject to bias depending on the possibility of its inclusion. Hence, the different outcomes for the dependent variable may not be considered independent irrelevant alternatives (IIA), as assumed by the multinomial logit. We investigate also whether the IIA problem persists only theoretically or is an empirical challenge as well. Using a Brant test, we find strong evidence of a violation of the IIA assumption when applying multinomial logit estimation. This paper applies a hierarchical nested logit specification to model the likelihood that a grant-back clause will be included in the licensing contract. These specifications split the categorical values into nests representing mutually dependent decisions (Manski and McFadden 1981). The nested logit therefore, is congruent with the decisions over licensing the technology and including a grant-back clause being interlinked. This model choice enables joint estimation 147 of the impact of firm and technology characteristics on the licensing decision and inclusion of a grant-back clause. The applied specification is a two-level nested logit model with random utility maximization (RUM) and full information maximum-likelihood estimation. This setting allows separation between use of a grant-back clause and the licensing decision while preserving the correlation between these two outcomes (Ziedonis 2007). [Insert Figure 1 around here] Figure 1 shows that the nest splits the sample across the three levels of the dependent categorical variable creating an asymmetric tree structure. The first nest utilizes all the USPTO patents granted to the licensor in the same year as the licensed technology on the assumption that they are all included in the portfolio of technologies that potentially could be licensed out. We identified a total of 7416 technologies of which 7012 patents were not included in the Recap dataset and we assume they were never licensed out, leaving 404 patents which we classified as being licensed out.29 A potential limitation of this setting is that firms might also license non-patented inventions, which are not included in this empirical setting. However, previous studies (Arora and Ceccagnoli 2006: 294) show that there is a connection between patenting behavior and licensing activity, suggesting “the presence of a patent is almost essential for licensing.” These authors show that less than 10 percent of licensors do not patent.30 Additionally, using only patented inventions to compare licensed versus non-licensed technologies ensures analytical consistency. Another potential issue related to our setting is that a certain technology may have been licensed but not reported in the Recap database. However, we have no reason to suspect 29 We excluded 239 technologies produced in the same year as the licensed technologies because they had two or more different assignees, indicating that the property rights for those patents were shared among firms. 30 In the Recap database, after dropping the firms for which there was no publicly available information, we had no cases of licensors that did not patent. 148 that, were this the case the technologies not reported in the licensing database would be systematically correlated with the likelihood of being licensed under a grant-back clause. This study uses interaction terms to estimate the determinants of the likelihood of a grant-back clause in a licensing contract. This modeling technique requires some shaping of the data (Drucker and Puri 2005) so that observations are classified as: (1) non licensed technology, (2) licensed technology without the grant-back clause, or (3) licensed technology with the grantback clause. This increases the number of observations threefold, transforming the 7416 observations into 22,248 and generating three evenly-distributed dummy variables, with 7416 positive outcomes for the three possible outcomes. Because there is no within-case variability in the second nest, following Drucker and Puri (2005) we created pseudo alternative specific outcomes for the explanatory variables by interacting them, in this nest, with the outcome variable (grant-back clause). To assess the magnitudes of the effects of a marginal change in the explanatory variables on the probability of observing a grant-back clause in a technology license contract, we estimate marginal effects. This requires partially differentiating the probability of the grantback clause with respect to the explanatory variables. This is problematic given the equations underlying the nested logit. There is no standardized method for capturing marginal effects for the nested logit. We follow Cameron and Trevedi’s (2009) recommendations and estimate only the marginal effects not their significances. These are estimates at mean values. 4.6 Results Table 1 reports the descriptive statistics for the variables considered in the analysis and their Pearson correlation coefficients (N=22 347). None of the correlations suggest any multicollinearity problems in the regression analysis. This is confirmed by variance inflation 149 factor (VIF) analysis. The maximum VIF associated with any of the independent variables is 2.78 (mean VIF = 1.40). Where it is possible to use the entire dataset for those variables, the statistics are consistent. [Insert Table 1 around here] Table 2 summarizes the results of the regression analysis. Model I reports the results considering only the controls, while models II-V introduce the explanatory variables and their interactions gradually. Table 2 provides support for Hypothesis 1, suggesting technology license agreements are increasingly likely to contain a grant-back clause the closer the licensed technology is to the licensor’s core technology. Hypothesis 2 is also supported: Technology license agreements are decreasingly likely to contain a grant-back clause the closer is the licensed technology to the licensee’s core technology. The parameter estimates for licensor’s/licensee’s core technology are statistically positively/negatively significant in all the models in which they are considered (II-V). The evidence slightly favors Hypothesis 1 compared to Hypothesis 2. However, both hypotheses are supported at a minimum 5 percent level of significance. [Insert Table 2 around here] Table 2 provides also supports Hypothesis 3 that the more uncertain the licensed technology, the more likely the technology license agreement will include a grant-back clause. The parameter estimates for uncertainty are significantly positive in all the models. Model II provides weak support at the 10 percent level of significance; the significance level is higher in models III, IV and V. The data do not support Hypothesis 4 regarding the likelihood of a grant-back clause in technology licensing agreements involving technologies that are core to the licensor increasing if the licensed technology is uncertain. None of the interaction parameter estimates 150 between uncertainty and licensor core technology are significant. Accordingly, we find no evidence to suggest that the parameter estimate of the interaction between the licensor’s core technology and technological uncertainty will be greater in absolute terms than the estimate associated with the licensor’s core technology. This is confirmed by a Wald test. We find statistical support for Hypothesis 5 that the decreasing likelihood of a grant-back clause in a technology licensing agreement involving technology that is core to the licensee will become an increasing likelihood if the licensed technology is uncertain. The Wald test suggests that the parameter associated with the interaction between licensee’s core technology and uncertainty is significantly greater than the absolute value of the parameter estimate for licensee’s core technology. Model V shows that (1.8+4.7=0) is statistically greater than zero. Among the controls we find that contracts specifying higher royalty rates are more likely to include grant-back clauses, and that contracts between parties where the licensor is technological superior tend not to contain grant-back clauses. The evidence suggests that higher value, more radical, and older technologies are more likely to be licensed. We find evidence also that technologies that are broader in scope typically are not licensed. Among firm characteristics, the empirical results suggest that R&D intensive and more technologically specialized licensors tend to engage in licensing activity. In line with previous studies, we find that larger firms, and firms characterized by a higher level of technological fragmentation tend not to engage in technology licensing. [Insert Table 3 around here] In order to test if the inclusion of the main explanatory variables provides significant improvement in the model fit, we used a log likelihood test to compare unrestricted 151 models against the restricted ones (reported at the bottom of table 2). Given that the Model 1 represents a baseline, there are no statistics reported for this model. The comparison for Model 2 - Model 1 indicates that adding the two main explanatory variables regarding Licensor and Licensee core technology significantly increased the overall model fit. To access Models III – V we used Model II as a baseline, with the results for all comparisons indicating statistically significant improvement in the overall model fit. Table 3 reports the marginal effects corresponding to the estimates of Model V in Table 2. These estimates basically confirm the direction indicated in the standard nested logit parameters. In addition, the marginal effects reveal that a one unit increase in how core the technology is for the licensor results in a 0.03 increase in the probability of a grant-back clause for the average observation. The corresponding number for the licensee is a 0.035 decrease in the probability of a grant-back clause. The comparative effect of the interaction between licensee core technology and technological uncertainty is more than double in absolute terms exhibiting a marginal effect of 0.076. 4.7 Sensitivity analysis We conducted several additional analyses to ensure that our results were not a by-product of our empirical choices. First, we considered those variables where we chose a particular time window, and varied the time dimensions (plus/minus 2 years). We found no evidence that our choice had any impact on the overall results of the model. We considered the fact that some firms appear more than once in the dataset since they had licensed-out more than one technology. This means that not all observations are independent of one another, which potentially could introduce some bias in our estimators because it is standard practice in some firms and not a by-product of a general tendency in a