»DO FIRM BOUNDARIES EXPLAIN THE NATURE OF PATENT
TRANSACTIONS? INVESTIGATING TRADING PATTERNS BETWEEN
AND WITHIN CORPORATE GROUPS«
MASTER THESIS
to obtain the degree of
MSc Business Administration – Strategic Innovation Management University of Groningen JANUARY 18, 2021 _____________________ Ruben Weinmiller S4146123 r.weinmiller@student.rug.nl Supervisor Pere Arque-Castells Co-Assessor dr. Wim Biemans _____________________ Total Words 12,346
Abstract
Table of Content
Abstract ... 2
1. Introduction ... 4
2. Theoretical Background and Hypotheses ... 6
2.1. Theoretical Background ... 6
Patents and Markets for Technology ... 6
Corporate Groups and Ambidexterity ... 8
2.2. Hypotheses Development ... 9
2.2.1. The Effect of Group Boundaries on the Nature of Patent Transactions ... 9
A. Incremental Patent Transactions Within Corporate Groups ... 10
B. Radical Patent Transactions Across Corporate Groups ... 11
2.2.2. The Moderating Effect of Technological Proximity ... 12
3. Data and Sample ... 14
3.1. Data Collection ... 14
3.2. Variable Construction ... 15
3.3. Descriptive Statistics ... 17
4. Method of Analysis ... 20
5. Results ... 23
5.1. Regression Results and Hypotheses Testing ... 23
5.2. Additional Analyses and Robustness Checks ... 24
6. Discussion ... 25
6.1. Theoretical Implications ... 25
6.2. Managerial Implications ... 26
6.3. Limitations and Future Research ... 27
7. Conclusion ... 28
8. References ... 29
1.
Introduction
The Schumpeterian notion of innovation as a product of a sole inventor is being increasingly replaced by the conception that innovation is an interplay involving various actors that engage in an iterative process with the goal of commercializing an invention (Gulati, 1998; Pisano, 1990). Special emphasis in these more recent models has been paid to the network character of innovation, involving many interactions of a variety of players along the way (Gulati, Nohria, & Zaheer, 2000). This context sets the framework for the emergence of markets for technology in which patents take on a fundamental role by virtue of their immanent characteristics that facilitate the transfer of Intellectual Property (IP) between entities (Arora, Fosfuri, & Gambardella, 2001; Spulber, 2015). In light of todays’ open innovation economy, a declining number of patents is filed at the locus that is optimal for commercialization. Therefore, markets for technology promote allocative efficiencies by facilitating technology exchanges (Bloom, Jones, van Reenen, & Webb, 2017; Jones, 2009; Monk, 2009).
Especially the increased technological complexity and specialization in an open innovation economy imply an imperative to participate in markets for technology. Due to the innovative aptitude of corporations, this imperative is augmented in the context of corporate groups (CGs) (Belenzon & Berkovitz, 2010). The importance of conducting further research in this regard is based on the aforementioned innovative capabilities of corporations that account for large proportions of R&D activity in an economy (Eurostat, 2019; OECD, 2019). On the other hand, the relevance builds on the growth that pertains to markets for technology and the increase in patenting activities on a worldwide scale (Kulatilaka & Lin, 2006). An abundance of research has examined particularities of markets for technology and the behavior of CGs in those markets in terms of their ability to pursue incremental and radical innovation (Arora et al., 2001; Ayfer & Cockburn, 2016; Caviggioli & Ughetto, 2013; Ceccagnoli, Graham, Higginsy, & Leez, 2010). However, notwithstanding the relevance of this topic, there is no previous research on the impact of firm boundaries on the nature of patent transactions in the context of CGs.
Therefore, this paper forms an addition to the theory on markets for technology while extending previous research concerning the behavior of CGs in those markets (Ayfer & Cockburn, 2016; Caviggioli & Ughetto, 2013; Ceccagnoli et al., 2010). Specifically, this analysis will combine insights from the knowledge-based view and transaction cost theory to in order to contribute to research on markets for technology, CGs and the role of technological proximity in that context. In the following, this study develops several hypotheses on transactions between and within CGs and how the nature of those transactions is determined by the boundaries of the firm and accompanying transaction cost considerations1. Further, it
examines how this relationship is shaped by the technological proximity of the transactors. Therefore, the following research question is formulated:
“How do corporate group boundaries affect the nature of conducted patent trades and how is this relationship shaped by the technological proximity of the transactors?”
To address this research question, three hypotheses are developed. The first hypothesis suggests that CGs provide a transaction-facilitating infrastructure that allows the internal transfer of patents that could not be realized between separate corporations due to exacerbated transaction costs. In particular, the transfer of incremental assets that confer comparatively little technological advancement is hypothesized to dominantly come to fruition within group boundaries. On the other hand, transactions across group boundaries supposedly favor involving radical patents, compared to those within. These, due to their pioneering merits, are likelier to surpass heightened transaction costs in trades that cross firm boundaries.
This hypothesized mechanism is based on expected transaction cost asymmetries between and within CGs. Under the influence of a transaction-cost aligning instrument, namely, the technological proximity of the transactors, this effect is hypothesized to attenuate, so that transactions between and within groups are more evenly distributed. To test the impact of firm boundaries on the nature of patent transactions, a binary logistic regression is conducted based on a sample of 51,558 patent transactions of firms rooted under the same corporate tree and firms associated with a different corporate parent. The observed time period ranges from transactions executed between 2005 and 2010.
The findings provide partial support for the hypotheses. That is, the locus of a transaction was found to affect the nature of patent trades only upon consideration of technological proximity in the statistical model. Only then, firm boundaries were found to potentially enable the transfer of incremental assets that would not materialize in external transactions due to heightened transaction costs. In turn, technological proximity as a transaction-cost reducing instrument has shown to exert an attenuating influence on the main mechanism only at high levels of proximity, thus negatively moderating the main mechanism.
2.
Theoretical Background and Hypotheses
This section reviews relevant theory on patents and their role in emerging markets for technology. Moreover, it elaborates on CGs and their ambidextrous ability to pursue exploration and exploitation resulting in the development of three hypotheses.
2.1. Theoretical Background
Patents and Markets for Technology
As a formal knowledge protection mechanism, a patent grants the owner temporary monopoly rights over an invention in return for its disclosure (Hall, Jaffe, & Trajtenberg, 2001; James, Leiblein, & Lu, 2013). Thus, by granting appropriability while disclosing the underlying technology, patents’ inherent characteristics confer the potential to stimulate transactions of intellectual property (IP) within markets for technology (James et al., 2013; Spulber, 2015). Disclosure and transferability promote transaction efficiencies, while the codification and standardization of inventions through patents nurture transparency prior to a transaction. Therefore, patents, by their immanent virtues, lay the foundation for the functioning of markets for technology (Akerlof, 1970; Spulber, 2015). Evidence of the functioning of markets for technology is positively related to the appropriability regime in host countries (Branstetter, Fisman, & Foley, 2006). Relatedly, Arora & Ceccagnoli, (2006), link an increase in the effectiveness of patent protection to a thicker market of technology that features a wider range of potential participants. Another underlying factor behind the growth of markets for technology pertains to the increased grounding of technology in science and computer technology which facilitates the codification and manipulation of physical occurrences, respectively. In turn, those factors cumulatively promote the emergence of IP as a commercial asset within markets for technology (Arora & Gambardella, 2010).
Consequently, reports estimate worldwide patenting revenues to have risen from $15 billion in 1990 to $100 billion in 2000 (Kulatilaka & Lin, 2006). US firms in particular have shown to increase their revenues following technology licensing activities, stimulating a rise in both patenting activities and transaction numbers (Arora et al., 2001; Marco, Graham, & Apple, 2015; Serrano, 2010). Apart from monetary reasons, scholars have identified several non-pecuniary motives to partake in the market for technology such as entry deterrence, attaining strategic leeway to operate and the enhancement of a firms’ bargaining power (Caviggioli & Ughetto, 2013; Grindley & Teece, 1997). Other motives can be rooted in superfluous patents that may emerge following a shift in a firms’ strategy. Since around 17% of patents are found to be “sleeping patents”, markets for technology have the potential to foster allocative efficiencies through gains from trade (Arora & Gambardella, 2010; Giuri et al., 2007).
on the degree of technological development in its area (Alcacer & Gittelman, 2005). That is, the more prior knowledge a patent comprises, the more incremental it is considered to be. In turn, patents that incur few prior art are deemed more radical (Ayfer & Cockburn, 2016). The increasing difficulty of innovating without expanding on prior knowledge is reflective of the cumulative nature of innovation, mirrored particularly in complex technological areas where large numbers of patents are imperative to innovate. Consequently, the technological landscape is characterized by an increasing fragmentation of patent rights, often denying one firm to make a claim on all patents that comprise an innovation (Noel & Schankerman, 2013).
Said factors have increasingly complicated the in-house cultivation of all innovation facets and have led to the replacement of the 20th century view that put vertically integrated but
isolated firms at the centerstage of R&D (Chandler, 1990). In the meantime, markets for technology have taken on an essential role in the more recent concept of open innovation (Graham, Marco, & Myers, 2018, Chesbrough, 2006). Open innovation emphasizes the “purposive inflows and outflows of knowledge to accelerate internal innovation and expand the markets for external use of innovation respectively" (Chesbrough, 2006, p.1). In the context of open innovation, facilitating the trade of patents becomes essential to enable aforementioned knowledge flows (Hagui & Yoffie, 2013; Yanagisawa & Guellec, 2009).
Hence, markets for technology isolate a firms’ inventive capabilities and allow for the evaluation of technology apart from the downstream complementary assets it possesses (Arora et al., 2001; Teece, 1986a). Consequently, scholars have identified a disaggregation of competition between the product market and the upstream market for technology (Arque-Castells & Spulber, 2019; Gans & Stern, 2003). This implies enlarged strategic options for firms and calls for a careful and proactive management of IP (Arora et al., 2001). Thus, a firms’ competitive advantage will no longer be rooted solely in its internal R&D capabilities, but rather in its dynamic capability to identify, assimilate, transform and utilize knowledge from multiple domains. In turn, this capacity to absorb knowledge has shown to be positively related to a firms’ existing knowledge stock (Cohen & Levinthal, 1990; Zahra & George, 2002). Transactions in the market for technology can take a variety of forms ranging from mergers and acquisitions (M&As), R&D joint ventures or alliances to licensing and cross-licensing agreements (Caviggioli & Ughetto, 2013). However, in the following, the focus will be solely on transfers and sales of patent ownership from one entity to another in the context on CGs. Those entities may be grouped under the same corporate umbrella or belong to different CGs. By taking on this perspective, a new dimension is added to open innovation: the dual concept of internal knowledge development and external sourcing omits the possibility that knowledge can be transferred internally, and that way contribute to the optimal internal
transfers of incremental patents that confer comparatively little technological advancement could be materialized within CGs due to efficiency benefits around transactions. In an external setting, those trades could be dominated by overshadowing transaction costs. This speaks to the relevance of researching the role of CGs in the context of markets for technology and adds new insights to the discussion on the boundaries of the firm and the optimal organization of innovation.
Corporate Groups and Ambidexterity
Among the worlds’ 2000 R&D investors, 70% of investments are performed by the top-250 firms, indicating that the distribution of R&D activity is highly skewed in favor of corporations (OECD, 2019; Eurostat, 2019). Andersson & Ejermo, (2005), find that the larger accumulation of R&D resources within a corporation has a positive effect on innovation performance. CGs have evident advantages in R&D rooted in their size and availability of capital. Those advantages pertain to the exploitation of economies of scale and scope, out of which economies of scope were found to have a greater impact on innovation performance (Cockburn & Henderson, 2001). This finding relates to synergies that can be derived from conducting research throughout a diverse environment that is integral to CGs (Cockburn & Henderson, 2001; Miller, Fern, & Cardinal, 2007; Zhou & Li, 2012).
Relatedly, Belenzon & Berkovitz, (2010), argue that CG affiliation is largely indicative of innovation performance. This finding pertains to the efficiency benefits of internal capital markets, low information asymmetries and reduced transaction costs within CGs. This relates to scholars who attribute the dominant role of CGs in innovation to the rationale of corporate headquarters (CHQs) and transaction efficiencies that arise from grouping firms under the parenting mandate of the CHQ (Foss, 1997; Pisano, 1990). By internalizing transactions, CGs are able to reduce transaction costs stemming from coordination problems, that is, “mitigate the effects of incentive-conflicts” (Foss, 1997, 1999, p. 2; Williamson, 1983). The hierarchical relationships between business units under the same corporate tree largely circumvent morally hazardous behavior through shared interests, alleviating costly monitoring and enforcement of contractual relationships (Liebeskind, 1996; Teece, 1986b).
contingencies (Foss, 1996). Relatedly, flexibility captures a CHQ’s ability to alter the configuration of the CG to promote speed in knowledge acquisition and reactiveness to externalities. This function can be executed more effectively within a hierarchy than between independent firms due to non-contractible coordinative capabilities (Foss, 1996, 1997). More specifically, flexibility refers to the addition, divesture or alteration of business units grouped under the same corporate umbrella (Foss, 1996, 1997).
It is those characteristics of CGs that lay the foundation for the positive rationale of CGs and facilitate the simultaneous pursuit of exploration and exploitation as argued in the context of ambidexterity (Gibson & Birkinshaw, 2004). Exploitation refers to the refinement of existing competencies and is more closely associated with incremental innovation. As such, exploitation is rooted in harnessing efficiency benefits derived from combining extant knowledge elements that foster improvements to the existent (Atuahene-Gima, 2005). Paradoxically, exploration entails leveraging and experimenting with diverse, novel or distant knowledge elements in the pursuit of radical innovation (Ahuja & Lampert, 2001). In the context of CGs, the exploitative aspect is mainly accommodated by knowledge direction and the efficiency benefits associated with it. Flexibility, on the other hand, largely pertains to the allocation of resources in a way that is conducive to exploration (Foss, 1997). The configuration of CGs that enables the pursuit of exploration and exploitation in tandem speaks to the ability of CGs to find a balance between those two mechanisms (Belenzon & Berkovitz, 2010; Tushman & O’Reilly, 1996). The overriding echo in previous literature is that an unspecified equilibrium between exploration and exploitation, that is, ambidexterity, drives innovation performance (Andriopoulos & Lewis, 2009; Atuahene-Gima, 2005; He & Wong, 2004; Tushman & O’Reilly, 1996).
In this regard, Foss’s (1997) seminal paper on the role of the CHQ in knowledge direction and flexibility as antecedents of exploitation and exploration provides a theoretical foundation to this study. Furthermore, this paper adds to the theory of markets for technology and expands on previous research regarding the behavior of CGs in those markets (Ayfer & Cockburn, 2016; Caviggioli & Ughetto, 2013; Ceccagnoli et al., 2010). Combining insights from the knowledge-based view and transaction cost theory, this paper will expand existing research on markets for technology, CGs and the role of technological proximity in that context. In the following, this study develops several hypotheses on different types of patent transactions between and within CGs and how the nature of transactions is influenced by the boundaries of the firm and concomitant transaction cost considerations. Further, it examines how this relationship is shaped by the technological proximity of the transactors.
2.2. Hypotheses Development
2.2.1. The Effect of Group Boundaries on the Nature of Patent Transactions
A. Incremental Patent Transactions Within Corporate Groups
In theory, transactions only materialize when the gains from trade outweigh the costs of spillovers (Arque-Castells & Spulber, 2019). Therefore, the consensus in literature is that particularly those patents that are inherently valuable are most likely to be transferred. Evidence for that is given by the mean value of a patent which has been found to be three times higher for a traded patent compared to an untraded one (Figueroa & Serrano, 2019; Serrano, 2018). Thus, from a patent level perspective, it has been argued that the likelihood of a transaction increases when a patent exhibits certain criteria related to the number of citations received by a given age, prior transactions or its generality (Serrano, 2010). That is, an asset has to outweigh the transaction costs that arise prior to, during and after its transaction in order to qualify for a transfer (Arque-Castells & Spulber, 2020; Serrano, 2018). Those costs are particularly related to searching, evaluating, negotiating and specifying the transactional mode as well as monitoring and enforcement of the transaction (Pisano, 1990; Teece, 1986b).
However, the aforementioned patent-level research largely neglects the possibility that patent characteristics which qualify an asset for a transaction as well as related costs are contingent on the locus of the transaction. Within CGs, costs related to negotiating and executing a transaction can be largely avoided due to the coordinative abilities of the corporate infrastructure (Foss, 1997). These virtues of CGs could facilitate incremental-type transactions within their boundaries that could not be realized between separate entities. This can be due to the incremental nature of the patent that confers comparatively little additional value which is dwarfed by the concomitant transaction costs when transferred externally. Exemplarily, previous findings indicate that particularly incremental patents are likely to be subject to litigation which has shown to be a hindering factor in trade (Lanjouw & Schankerman, 2001). Conversely, transacting those assets internally circumvents part of the risks associated with incremental patents and litigation suits. Therefore, the smaller gains from trade that arise from incremental patent transfers are compensable by reduced transaction costs within a CG ecosystem. In particular, reducing transaction costs by 50% were found to bolster realized gains from trade by an additional 8.7% (Serrano, 2018). This speaks to the efficiency benefits that arise from internalizing part of the patent transactions within CGs (Griffith, Lee, & Straathof, 2017; Pisano, 1990; Teece, 1986b).
particularly those patents that feature incremental improvements to existing technologies within firm boundaries (Caviggioli & Ughetto, 2013).
Extending this line of thought, spillovers in internal transactions might even be encouraged by the transactors in order to benefit from complementary knowledge while promoting the ability to utilize the technology (Arora & Gambardella, 1994; Yamin & Otto, 2004). Relatedly, Jensen, Palangkaraya, & Webster, (2015), demonstrate that familiarity and trust lubricate transactions in the market for technology. This underscores the ability of CGs to materialize exchanges within their boundaries that would not induce gains from trade between separate organizations due to the outweighing potential of spillovers and transaction costs (Arque-Castells & Spulber, 2020). Given the efficiency benefits that arise from internalizing transactions within a CG and the ability to avoid expropriation that is augmented in the case of incremental-type patents, CGs are expected to provide conditions that are particularly favorable to transact patents that are incremental in their nature. That is, patent transactions within CGs will be more incremental in nature, compared to patent transactions across CGs.
B. Radical Patent Transactions Across Corporate Groups
By definition, patents acquired across firm boundaries will be radical to the acquiring firm due to their novelty to the organization (Ahuja & Lampert, 2001; Katila, 2000). In light of the enlarged strategy space that arises from markets for technology, particularly patents that incur greater deviation from the existing in-house pool of knowledge promise to be a significant add-on to the acquiring firms’ knowledge base (Arora et al., 2001; Figueroa & Serrano, 2019). However, the following line of argumentation regards radicality as the degree of technological radicalness of the transacted patent measured by the amount of knowledge it builds on (Ahuja & Lampert, 2001; Katila, 2000). That is, the extent of prior research as indicated by patent citations will be inversely related to the radicalness of the patent (Ayfer & Cockburn, 2016; Hall et al., 2001).
The benefits that arise from transacting radical patents between separate CGs are twofold and relate to value creation and value protection aspects. Change-theoretical argumentation found external resource acquisitions to be a channel to break path-dependencies that could arise from routinely executed internal knowledge transactions. Therefore, the potential to exploit path-breaking opportunities is expected to be larger for patents that build on little previous research and are thus deemed radical (Karim & Mitchell, 2000; Williamson, 1999). In turn, utilizing those patents requires a great extent of absorptive capacity (ACAP) from the acquirer in order to identify the value of the technology and thereupon enable its exploitation (Akerlof, 1970; Cohen & Levinthal, 1990; Zahra & George, 2002). By virtue of their size and organizational setup, CGs feature a large knowledge base that yields enhanced potential to absorb external knowledge (Argyres, 1996; Cohen & Levinthal, 1990). Those characteristics augment the ability to identify, assimilate, transform and exploit external knowledge that is more radical in its nature (Ahuja & Katila, 2001; Zahra & George, 2002).
(Griffith et al., 2017). Thus, congruent to the previous argumentation, heightened transaction costs in transfers across CGs must be outweighed in order to realize gains from trade (Arque-Castells & Spulber, 2020; Serrano, 2018). Patents that are more radical are more likely to amortize heightened transaction costs between CGs because of the greater potential value they bestow (Ahuja & Lampert, 2001). This comes down to a “selection effect”, by virtue of which patents that promise higher returns augment the likelihood to be traded (Serrano, 2018). In spite of that, radical patents do not promise immediate returns since they incur a significant deviation from established technology (Katila, 2000). Consequently, they also build less closely on the firms’ existing knowledge base (Ahuja & Lampert, 2001). Situated peripheral to the technology-providing firms’ core knowledge base, radical-type patents confer a smaller probability of unwanted spillovers. Since radical patents build on less related knowledge that can be inferred from, they pose a reduced threat of expropriation for the providing firm (Ahuja & Lampert, 2001). Thus, from a knowledge protection perspective, transacting patents that build on fewer existing technologies are less likely to cause unwanted spillovers that could jeopardize realized gains from trade between CGs (Arque-Castells & Spulber, 2020). Consequently, it is expected that patents transacted across CGs are more radical in nature compared to patents that are transferred within CGs.
Hypothesis 1a: Patents transacted within groups are more incremental in nature than patents transacted between groups.
Hypothesis 1b: Patents transacted between groups are more radical in nature than patents transacted within groups
2.2.2. The Moderating Effect of Technological Proximity
Technological proximity captures the degree of overlap in technology between providers and adopters (Arque-Castells & Spulber, 2020; Jaffe, 1986). Following literature on ACAP, proximate technology bases between firms ease the transferability of assets due to facilitated identification and understanding of relevant knowledge components (Aldieri, 2011; Bloom, Schankerman, & Van Reenen, 2013; Cohen & Levinthal, 1990). In the context of markets for technology, this finding relates to an increased ability to identify and assess the value of a technology due to proximate links between the transactors (De Marco et al., 2017). Therefore, transferring patents will be favored by technological proximity through the establishment of mutual knowledge between parties prior to a transaction (Ayfer & Cockburn, 2016). Concurrently, knowledge flows largely pertain to firms located in the same technological field because of enhanced capabilities to generate, transfer and absorb knowledge within close technological proximity (Arque-Castells & Spulber, 2017; Bloom et al., 2013).
promote efficiencies (Griffith et al., 2017). As previously argued, in internal transactions, the preexisting CG ecosystem lubricates transfers and eases the identification of assets that confer incremental technological advancement. Lacking the coordinative parenting mandate of a common CHQ, technological proximity should exert a substituting influence on the role of a CG ecosystem and therefore increase the likelihood of incremental patents being transferred across group boundaries (Foss, 1997, 1999). That is, the identification and utilization of incremental assets that suit the current technological development stage will be enhanced by the operational depth of technologically proximate transactions (Arora & Gambardella, 1994; Ayfer & Cockburn, 2016).
While costs of discovery and utilization are increased by 30% for transactions that cross firm boundaries, transaction costs may be reduced if the ability of the buyer to identify the value of the technology is enhanced by transfers between groups that operate in close technological proximity (Akerlof, 1970; Griffith et al., 2017). This mechanism should be particularly pervasive for incremental transactions, which, as previously argued, provide comparatively little advancement in technology and hence call for low transaction costs (Serrano, 2018; Tushman & O’Reilly, 1996). In particular, technological proximity is likely to reduce insecurities associated with the applicability of incremental patents across a CG-ecosystem. This is because incremental assets require a substantial amount of previous knowledge in a similar field in order to materialize post transaction (Atuahene-Gima, 2005; Ayfer & Cockburn, 2016).
By the same token, the increase in transactions of incremental patents across CGs under conditions of technological proximity is expected to cause an internalization effect of radical patents within group boundaries, ceteris paribus (Belenzon & Berkovitz, 2010; Figueroa & Serrano, 2019; Zhou & Li, 2012). The driving forces behind this effect are twofold: on the one hand, high levels of internal knowledge overlap make the internalization of radical patents particularly favorable. Knowledge sharing in close technological proximity highlights the “kaleidoscopic” ability of a CG ecosystem to evoke explorative resource allocations across a broad internal knowledge base by means of the CHQs’ flexibility function (Foss, 1997, 1999; Kanter, 1988; Zhou & Li, 2012). Especially the transfer of complementary tacit knowledge which assists the assimilation and utilization of patents that confer little previous research has shown to be lubricated by similar technology bases (Arora & Gambardella, 1994).
On the other hand, the alignment of transaction costs within and across groups at high levels of technological proximity causes the formerly described displacement effect in patent transactions, so that transfers within and across groups are more evenly distributed. Namely, the lower proximity-induced transaction cost threshold that allows incremental transactions to materialize in external transactions should in turn increase the likelihood of radical assets being traded within group boundaries (Andriopoulos & Lewis, 2009)
transactors supposedly exerts a substitution effect on the mechanism described in H1a and
H1b. That is, technological proximity will partly offset the CG ecosystem that favors incremental transactions within its boundaries and radical transfers across, so that transactions within and across CGs are more evenly distributed. This leads to the final hypothesis:
Hypothesis 2: Technological proximity will attenuate the effect of group boundaries on the nature of the transacted asset, so that patent transactions within and across groups are more evenly distributed.
3.
Data and Sample
This section outlines the data sources this analysis draws from before describing the variable construction. Subsequently, after providing descriptive statistics and a Chi-Square test, the correlations among variables will be analyzed in order to exclude multicollinearity.
3.1. Data Collection
This research largely draws from datasets on interactions in the market for technology constructed by Arque-Castells & Spulber, (2017). The main transaction-level dataset is derived from matching assignees and assignors listed in the USPTO Patent Assignment Dataset (PAD) with a Dynamic Corporate Tree Dataset through unique firm identifiers (GVKEYS). The latter dataset was constructed by integrating data from Compustat, SDC Platinum Osiris and the NBER PDP DYNASS file. Since standalone firms were neglected, the final dataset only contains information on patent transactions within and between CGs and their subsidiaries (IDDRS) at the transaction record level. A subset of this dataset that is limited to transactions executed between 2005 – 2010 constitutes the base for this research and contains 22,142 observations.
The second obtained dataset covers patent-level information on 8,715,825 transferred patents, while the third contains patents and their original assignment to the respective corporate parent identifier (GVKEY), resulting in 1,362,904 entries.
on a limited time period allows for a more fine-grained analysis of patent transfers, while keeping a substantial number of observations.
The patent citation dataset uspatentcitation contains information on citations made by US granted patents. This dataset is constructed dyadically, that is, each backward citation of a given patent A to a cited patent B equals one entry, adding up to a total of 111,026,440 observations. Moreover, Compustat was used to obtain data on R&D expenses of the assignee parent firm. To execute the desired patent-level analysis, relevant observations of all six datasets were sequentially merged into one, resulting in 51,558 observations at the patent-level.
3.2. Variable Construction
The following section will describe in detail the process of constructing the variables in this analysis.
Dependent Variable.
Radical: As previously stated, patents are required to cite all prior technological developments they build on (Hall et al., 2001; Trajtenberg, Henderson, & Jaffe, 1992). The dataset used to construct this variable, uspatentcitation, differentiates between citations that are added by the applicant, the examiner and other non-disclosed citations for entries prior to 2002.Since the applicant may not be aware of all technological antecedents of the applied patent and especially the examiner is considered an expert in the respective field, their added citations are kept in the data. While there is an often-reported noise associated with citations added by patent examiners, said noise mainly accrues to forward citations (Alcacer & Gittelman, 2005).This conservative approach is followed to capture the maximum paper trail of knowledge that underlies an invention (Hall et al., 2001; Trajtenberg et al., 1992). Previous research has used multidimensional measures of radicalness that capture both the originality and impact of a patent (Ahuja & Lampert, 2001; Ayfer & Cockburn, 2016). In this study, the binary dependent variable is constructed following the measure of a pioneering patent based on backward citation counts adopted from Ahuja & Lampert, (2001). According to this measure, a patent is deemed pioneering if it does not build on any previous patents. This research therefore uses the term “radical” synonymously to what the aforementioned authors measure as a pioneering patent. To construct this variable, the dyadic dataset on patent citations was collapsed at the citing patent level in order to obtain the respective citation counts. Consequently, for each logarithmic value of patent citations that equals zero, the patent is deemed radical, and incremental otherwise. The logarithmic values of zero were adopted because one citation was the minimum number of backward citations indicated in the uspatentcitation dataset.
Independent Variable.
are associated with the same CG, and zero if a transaction was executed between two distinct CGs.
Moderating Variable.
Tech_proximity: Following Jaffe, (1986), this variable is constructed by integrating patent-level datawith firm identifiers and the main technology class of the respective patents. Based on the USPTO allocation of patents in one of 426 technological classes, a vector Ti=(Ti1,Ti2,…,Ti426) is constructed, according to which Tij is the number of patents a firm i holds
in technology class j. The vector with the share of patents in each technology class for each firm reflects the distribution of patents across technology classes. This vector identifies the positioning of firms in a technological space, measured as the uncentered correlation coefficient between corresponding technology vectors (Aldieri, 2011). After computing the vector Tifor each firm, technological proximity according to Jaffe (1986) for each unique
assignor/assignee pair was constructed as follows:
!"#$
!"=
!
!!
"′
((!
!!
!$)(!
"
!
"′))
% &Ranging from zero to one, the value encompasses a spectrum between no technological commonalities and complete overlap. It therefore captures the inferable ability of the adopter to utilize the technology following a transaction (Arque-Castells & Spulber, 2020; Cohen & Levinthal, 1990; Jaffe, 1986).
Control Variables.
Several variables are included to control for factors that may affect the previously described dependent and independent variables and thus allow more definitive conclusions. Control variables are included at the firm- and patent-level, and to account for varying industry effects across different technology groups.
XRD: R&D expenses of the adopter parent firm was chosen as a control variable, measured as the logarithmic mean value of R&D expenses over the examined time period. The logarithmic value was chosen to account for the skewness of the distribution within the sample that ranged from values of less than a million up to more than 8 billion USD. The importance of controlling for this variable is reflected in suspected differences of technology-adoption patterns of firms according to their R&D expenditures. Varying R&D expenses of adopter parent firms may explain differences in the characteristics of traded assets. Namely, firms featuring higher R&D expenditures may be less susceptible to transaction cost considerations that underlie the hypotheses development.
measured as the execution year minus the year of application. This measure nonetheless resulted in some negative values caused by fragmentary inaccuracies in given execution dates.
Repeated_Trade: Controlling for repeatedly traded patents is done via a dummy variable that takes the value one if a patent occurs more than once in the dataset and zero otherwise. Repeatedly traded patents may hold different characteristics than patents that are traded once in the examined time period (Serrano, 2010).
Npatents: With varying sizes of transaction bundles, controlling for the number of patents per transaction is commendable. Since transaction costs of patent transfers are spread over a higher number of patents in larger transactions, the transaction size could affect the characteristics of traded patents. Therefore, Npatents is measured as the number of patents that correspond to one transaction-identifier, rf_id.
Technology_Groups: Based on the technology class of each patent and the USPC-system, the transacted patents were manually allocated in one of three technology groups as indicated by the USPTO: Technology Group 1: Chemical & related Arts, Technology Group 2: Communication, Radiant Energy, Weapons, Electrical, and Computer Arts, Technology Group 3: Body Treatment and Care, Heating and Cooling, Material Handling and Treatment, Mechanical Manufacturing, Mechanical Power, Static, and Related Arts (USPTO, 2012). A fourth category, Industrial Designs, did not correspond to any of the traded patents in the observed time period. The three resulting technology groups are measured as three dummy variables that take the value one if a patent belongs to the respective category, and zero otherwise. Sporadically, technology classes appear in multiple technology groups. This measurement is undertaken to control for industry effects that may explain differences in the characteristics of traded assets (De Rassenfosse, Palangkaraya, & Webster, 2016).
3.3. Descriptive Statistics
Table 1: Descriptive Statistics
In order to further differentiate transactions between and within CGs, descriptive statistics are additionally subcategorized. The mean value of “Radical” is shown to be slightly higher in transactions within CGs (M=0.056), compared to transactions between groups (M=0.049), indicating that slightly more radical patents are transacted internally. Appendix B indicates histograms of technological proximity for internal and external transactions. By mean comparison, technological proximity is found to be higher for transactions within groups (M=0.59), compared to those across (M=0.50). What is more, the average transaction bundle is larger across CGs (M=10.09) than within CGs (M=4.18).
The conducted Chi-Square Test (Appendix C) determines whether the differences in the observations provided in the descriptive statistics are also statistically significant. This is to affirm a statistical difference among the categorical dependent and explanatory variables if the Null-Hypothesis can be rejected. Generally, a Chi-Square Test is conducted to determine whether expected occurrences of observations between categorical variables differ significantly from observed observations. The Chi-Square Test indeed highlights a statistically significant difference for radical and incremental patents regarding the locus of a transaction (p<0.01). However, it does not indicate whether the locus of a transaction influences the likelihood of a radical or incremental transaction.
In Table 2, correlations amongthe variables included in the statistical model are highlighted. With two exceptions, one can observe that the variables signify a low or moderate correlation amongst each other (r<0.6). However, two correlations occur among dummy variables between Technology Group 1 and 2 and Technology Group 2 and 3. This could be due to the observation that some patent technology classes pertain to multiple Technology Groups. However, due to the dummy nature of the correlated variables, multicollinearity in the present model can be excluded.
Variables Obs Mean Std. Dev. Min Max
Radical 51,558 0.0538811 0.225785 0 1 Locus (Within) 51,558 0.6191086 0.4856107 0 1 Tech Proximity 51,558 0.55642 0.3108593 0 1 R&D Expenses 51,558 1631.058 1886.549 0.3558 8654.833 Patent Age 51,558 5.888126 4.688031 -6 38 Repeated Trades 51,558 0.2587183 0.4379348 0 1
Patents per Transaction 51,558 5.387461 39.21834 1 1274
Technology Group 1 51,558 0.1502192 0.3572896 0 1
Technology Group 2 51,558 0.7234183 0.4473121 0 1
Table 2: Correlation Matrix Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (1) Radical 1 (2) Locus 0.0136 1 (3) Technological Proximity 0.0088 0.1393 1 (4) R&D Expenses 0.0753 0.1368 -0.0764 1 (5) Patent Age -0.0468 -0.0209 -0.1116 -0.0091 1 (6) Repeated Trades -0.0062 0.0106 -0.0720 -0.0872 0.0437 1
(7) Patents per Transaction -0.0087 0.0375 -0.0240 -0.0146 0.1410 0.1222 1
(8) Technology Group 1 0.0596 -0.0292 -0.0139 0.0510 0.0041 -0.0461 -0.1400 1
(9) Technology Group 2 -0.0036 -0.0266 -0.0478 -0.0130 -0.0268 0.0759 0.1834 -0.6602 1
20
4.
Method of Analysis
This section outlines the methodological proceeding of the conducted analysis. A binary logistic regression will test the hypotheses, before additional analyses will determine if differences in results occur at marginal levels of technological proximity. Robustness checks will amplify the validity and reliability of the obtained results.
Since the dependent variable “radical” can only take on two values, one for a radical patent and zero for an incremental one, the outcome is binary and thus opposes the two polar groups. This construct calls for a binary logistic regression2 that, applied to the present
analysis, yields the following form (Hosmer, Lemeshow, & Sturdivant, 2013):
E(Y|x) = $!+ $"'()*+ + $#,-)ℎ/(0(12)30 45(627289 + $$'()*+ ∗
,-)ℎ/(0(12)30 45(627289 + ;(/85(0 <3523=0-+
Rather than measuring effects via beta-coefficients in a linear model, a binary logistic regression “approximates how much more likely (or unlikely) it is for the outcome to be present among those with x = 1 than among those with x = 0“ Hosmer et al. (2013, p. 50). Applied to the present analysis, the sign of the logit coefficient highlights whether transactions within (across) groups are more incremental (radical). By nature of logistic regression, this at the same time can be interpreted as the likelihood of an incremental (radical) transfer to occur within (between) groups.
H1a and H1b will be tested by examining the sign of the logit coefficient $" of the independent
variable “locus” which is expected to take on a negative value, indicating that internal transactions are more incremental, compared to transactions across groups. By virtue of the binary independent variable, this would at the same time indicate that transactions observed across group boundaries are more radical in nature. That is, the sign of the coefficient will be inverted for transactions across CGs. If this observation turns out to be significant, it would support H1.
Conversely, H2 will be tested by analyzing the total effect of locus. Thus, the logit coefficient of technological proximity $$, upon interacting with the independent variable, is expected to
take on the opposite coefficient compared to $". That is, a positive $$ would indicate a
weakening interaction effect that causes transactions within and across groups to be more
2A binary logistic regression takes the basic form as follows (Hosmer et al., 2013): !(#|x) = $!+ $"'
21
evenly distributed, as hypothesized in H2. This is because the interaction term attenuates the negative effect that $" is said to comprise, so that the overall effect of locus ($"+ $$) is close
22
Table 3: Binary Logistic Regression
Dependent Variable: Radical Model 1 Model 2 Model 3 Model 4 Model 5
Independent Variable Locus (Within) -0.010 -0.040 -0.162* -0.189** (0.045) (0.045) (0.092) (0.082) Moderator Variable Tech Proximity 0.270*** 0.115 (0.070) (0.122) Medium 0.118 (0.0904) High 0.041 (0.084) Interaction Effect
Locus X Tech Proximity 0.229
(0.149)
Locus X Medium Tech Proximity 0.042
(0.118)
Locus X High Tech Proximity 0.353***
(0.109) Control Variables R&D Expenses 0.241*** 0.242*** 0.250*** 0.252*** 0.256*** (0.016) (0.016) (0.016) (0.016) (0.016) Patent Age -0.049*** -0.050*** -0.048*** -0.049*** -0.047*** (0.005) (0.005) (0.005) (0.005) (0.005)
Patents per Transaction 0.000 0.000 0.000 0.000 0.000
(0.000) (0.000) (0.000) (0.000) (0.000) Repeated Trade 0.010** 0.010** 0.123** 0.115** 0.119** (0.050) (0.050) (0.051) (0.051) (0.051) Technology Group 1 0.875*** 0.874*** 0.889*** 0.892*** 0.895*** (0.117) (0.117) (0.117) (0.117) (0.118) Technology Group 2 0.413*** 0.412*** 0.430*** 0.433*** 0.441*** (0.113) (0.113) (0.114) (0.114) (0.114) Technology Group 3 -0.691*** -0.690*** -0.686*** -0.691*** -0.688*** (0.105) (0.106) (0.106) (0.106) (0.107)
Year Dummies Yes Yes Yes Yes Yes
Observations 51,558 51,558 51,558 51,558 51,558
LR Chi2 775.28 775.34 790.54 792.90 811.66
Prob > Chi2 0.0000 0.0000 0.0000 0.0000 0.0000
Pseudo R2 0.0358 0.0358 0.0365 0.0367 0.0375
Log Likelihood -10428.608 -10428.58 -10420.981 -10419.799 -10410.422
23
5.
Results
5.1. Regression Results and Hypotheses Testing
Table 3 presents the results of the conducted binary logistic regression and the corresponding logit coefficients. The first model only includes control variables that inspect explanations for the dependent variable other than the hypothesized ones, followed by the addition of the independent variable in Model 2. Subsequently, the moderator variable and finally the interaction term were added in Models 3 and 4. Model 5 includes a categorization in the measurement of technological proximity in order to examine marginal effects at low, medium and high levels of proximity, leaning on Huang & Shields, (2000). This sequential approach highlights the increasing explanatory power of the model through the addition of variables, indicated by the steady increase of the overall likelihood-ratio chi-square from 775.28 to 811.66 throughout the models.
Considering the binary dependent and explanatory variables, H1a suggests a negative effect of internal transactions on the nature of transacted patents to be radical, while H1b makes the flipside argument. Although the coefficient is negative as expected, adding the independent variable “locus” in Model 2 does not yield significant results (b=-0.01, p>0.1). Without considering the subsequent addition of the interaction effect of locus and technological proximity in the model, this would lead to the rejection of H1a and H1b.
Conversely, the addition of technological proximity as a moderator in Model 3 proves to exert a highly significant, positive impact on the likelihood of a radical patent transaction, regardless of CG affiliation (b=0.27, p<0.01).
Testing the interaction term of locus and technological proximity in Model 4 according to Hypotheses 2 does not yield significance (b=0.229, p>0.1). Normally, this would lead to the rejection of H2. However, before rejecting the hypotheses, it is commendable to inspect the effect of technological proximity at low, medium and high levels, in particular since the hypotheses is formulated for high levels of technological proximity (Huang & Shields, 2000). According to the mentioned authors, marginal effects may differ across observations. Therefore, a fifth Model includes an interaction term at the aforementioned levels of proximity.
24
transactions within and across CGs to be more evenly distributed, hence confirming H2 under these conditions.
Several additional insights pertain to the addition of technological proximity as an interaction term in Model 4. Namely, upon interacting the moderator variable with the locus of the transaction, the negative logit coefficient of locus indeed highlights a negative impact of internal transactions on the nature of the trade to be radical as hypothesized. However, this impact is only weakly significant (b=-0.162, p<0.1). Therefore, internal transactions are indeed likelier to feature incremental patents, compared to transactions across groups. Conversely, transactions across groups will likelier be radical, confirming H1a and H1b upon interacting technological proximity with locus in the model. Moreover, the categorization of the variables in Model 5 further increases the significance of the independent variable as hypothesized (b=-0.189, p<0.05). That is, H1a and H1b is supported upon consideration of the interaction effect of technological proximity in the model.
Regarding the influence of control variables in Model 1, the logit coefficient of R&D Expenses indicates a highly significant impact on the likelihood of transacting a radical patent (b=0.241 p<0.01). Conversely, Patent Age exerts a highly significant, negative influence on the likelihood of a radical transaction (b=-0.049, p<0.01). Repeatedly traded patents indicate a positively significant likelihood to be radical (b=0.01, p<0.05). Patents that belong to Technology Groups 1 (b=0.875, p<0.01) and 2 (b=0.413, p<0.01) positively affect the likelihood of a radical transaction, respectively. In contrast, patents assigned to Technology Group 3 are likelier to be incremental (b=-0.691, p<0.01). Those observations hold a constant significance throughout all Models.
5.2. Additional Analyses and Robustness Checks
25
6.
Discussion
This section discusses theoretical and managerial implications of the obtained results before addressing limitations and providing avenues for future research.
6.1. Theoretical Implications
While an abundance of literature has examined the peculiarities of CGs and patent transactions (Arora et al., 2001; Arque-Castells & Spulber, 2020; Belenzon & Berkovitz, 2010; Foss, 1997) the nature of traded patents in the realm of transactions within and across CGs remained uninvestigated. Therefore, this study aimed to combine patent-level research with transactions in the market of technology in the context of CGs (Arque-Castells & Spulber, 2020; Ayfer & Cockburn, 2016; Serrano, 2010). This identified literature gap was filled by examining how CG boundaries affect the likelihood of radical and incremental trades respectively, and how this relationship is shaped by the technological proximity of transactors.
This study illuminates several interesting findings that complement previous literature. It confirms the negative relationship of a radical asset and its age by the time it is traded, indicating that pioneering patents tend to be transferred earlier in their lifespan (Serrano, 2010). What is more, differences in transaction patterns across technology groups point to the importance of distinguishing between industries and varying patent trading flows across them (Arora & Ceccagnoli, 2006; James et al., 2013).
26
In particular, high levels of technological proximity have shown to increase the buyers’ potential to discern the fewer amount of prior art conferred in radical patents, while also augmenting the ability to identify and utilize incremental patents regardless of shared cognitive structures of a common CHQ (Ahuja & Lampert, 2001; Foss, 1997, 1999; Zahra & George, 2002). Moreover, technological proximity was found to encourage knowledge flows and interaction, thus promoting market thickness and the potential for gains from trade by reducing transaction costs in both types of transfers (Aldieri, 2011; Arque-Castells & Spulber, 2020; Roth, 2008).
This finding echoes the mechanism that underlies the hypotheses building which is based on expected transaction cost differences across CG boundaries and technological proximity as a transaction cost reducing instrument that attenuates those differences. Therefore, the findings imply that, absent of transaction costs, one should observe no differences in trading patterns within and across CGs. However, previous research on other transaction cost-induced idiosyncrasies in technology transfers indicate that transaction costs exist and need to be considered in this context (Teece, 1986b). What is more, the simultaneous occurrence of significance in the final model speaks to the complementarity of firm boundaries and technological proximity as transaction-cost reducing instruments. Thus, the findings further complement previous literature on other mechanisms to align transaction costs in technology transfers such as the level of integration in transaction modes, the transaction size, trust, geographical or market proximity (Aldieri, 2011; Arque-Castells & Spulber, 2017; Caviggioli & Ughetto, 2013; Figueroa & Serrano, 2019; Jensen et al., 2015).
However, the obtained results have to be interpreted with caution, partially due to the weak significance of the logit coefficient in Model 4 and the conditional nature of the significance, as well as small effects that call for further investigation. Moreover, the findings do not infer causality.
6.2. Managerial Implications
27
transaction-cost reducing instruments if the full potential of the increased strategy space of technology transfers is to be realized (Arora et al., 2001; Griffith et al., 2017). Therefore, this study assists managers to make decisions on patent internalizations in a CG and thus adds to the discussion on boundaries of the firm, indicating when and where to internalize patent transactions (Teece, 1986b). Taking those insights into account will assist managers to augment benefits from the increased strategy space that arises from markets for technology, thereby further boosting the innovative radiance of corporations.
6.3. Limitations and Future Research
This study suffers from several assumptions that potentially confine the generalizability of the obtained results. While the focus on CGs is not a limitation per se, neglecting standalones and smaller firms in the data potentially limits the obtained insights and forbids generalizing implications onto the entire market for technology, something that could be addressed by future research.
Moreover, the study focuses exclusively on the US patent market. While this market has shown to be the dominant market for technology worldwide, obtained insights could differ in the setting of a different market, for instance by virtue of weaker appropriability regimes, that, as previously argued, affect trading patterns in the market of technology and the realization of patent transfers as such (Branstetter et al., 2006). Namely, the research is conducted in an environment that provides particularly favorable “laboratory” conditions for patent transfers that may not be applicable to other market settings. Comparing the obtained results with other markets could yield interesting insights for future research
Thirdly, the obtained results largely hinge on the measurement of radicalness of the transacted patents. While the adopted measure of a radical patent based on citation counts has been applied by prominent literature in the past, an important limitation of this study pertains the one-dimensional proxy of radicalness that does not encompass aspects such as generality and impact but merely focuses on the pioneering element a radical patent is said to comprise (Ahuja & Lampert, 2001; Ayfer & Cockburn, 2016; Henderson & Clark, 1990; Zhou & Li, 2012). This is problematic because not all patents that are deemed radical in this study are paths to technological or commercial success. Some may be technological deadlocks and would thus question the transaction-cost bridging potential ascribed to them in this study (Ahuja & Lampert, 2001). However, also alternative measures have shown to suffer from inconclusiveness. For instance, measuring radicalness via citations received by a given age confers problems of truncation, that is, missing data due to potential future citations received by the time the patent is transferred (Hall et al., 2001). Overall, previous research has not been conclusive of a transcendent measure of patent radicalness. This is an important topic to be addressed by future research.
28
for more compelling conclusions, representative of the increasing dynamics that pertain to the market for technology.
7.
Conclusion
This study adds to the theory of CGs and expands on former research regarding the behavior of CGs in markets for technology (Ayfer & Cockburn, 2016; Caviggioli & Ughetto, 2013; Ceccagnoli et al., 2010). Therefore, it fills the gap in previous literature regarding the influence of firm boundaries on the nature of patent transactions in the context of CGs. In particular, this research examined transactions of radical and incremental patents and hypothesized how those are influenced by the boundaries of the firm and accompanying transaction cost considerations. Further, it has analyzed how this relationship is shaped by the technological proximity of the transactors in an attempt to answer the following research question: “How do corporate group boundaries affect the nature of conducted patent trades and how is this relationship shaped by the technological proximity of the transactors?”
29
8.
References
Ahuja, G., & Katila, R. (2001). Technological acquisitions and the innovation performance of acquiring firms: A longitudinal study. Strategic Management Journal, 22(3), 197–220. https://doi.org/10.1002/smj.157
Ahuja, G., & Lampert, C. M. (2001). Entrepreneurship in the large corporation: A longitudinal study of how established firms create breakthrough inventions. Strategic Management Journal, 22(6–7), 521–543. https://doi.org/10.1002/smj.176 Akerlof, G. A. (1970). The market for “lemons”:
Quality uncertainty and the market mechanism. Decision Science, 84(3), 488–500.
https://doi.org/10.1017/cbo97805116093 81.002
Alcacer, J., & Gittelman, M. (2005). How do I Know what you Know? Patent Examiners and the Generation of Patent Citations. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.548003 Aldieri, L. (2011). Technological and
geographical proximity effects on knowledge spillovers: Evidence from the US patent citations. Economics of Innovation and New Technology, 20(6), 597–607.
https://doi.org/10.1080/10438599.2011.5 54632
Andersson, M., & Ejermo, O. (2005). How does accessibility to knowledge sources affect the innovativeness of corporations? - Evidence from Sweden. Annals of Regional Science, 39(4), 741–765. https://doi.org/10.1007/s00168-005-0025-7
Andriopoulos, C., & Lewis, M. W. (2009). Exploitation-exploration tensions and organizational ambidexterity: Managing paradoxes of innovation. Organization Science, 20(4), 696–717.
https://doi.org/10.1287/orsc.1080.0406 Argyres, N. (1996). Capabilities, Technological
Diversification and Divisionalization. Strategic Management Journal, 17(5), 395–410. Retrieved from http://dx.doi.org/10.1002/(SICI)1097- 0266(199605)17:5%3C395::AID-SMJ826%3E3.0.CO;2-E
Arora, A., & Ceccagnoli, M. (2006). Patent protection, complementary assets, and firms’ incentives for technology licensing. Management Science, 52(2), 293–308. https://doi.org/10.1287/mnsc.1050.0437 Arora, A., Fosfuri, A., & Gambardella, A. (2001).
Markets for technology and their implications for corporate strategy. Industrial and Corporate Change, 10(2), 419–452.
https://doi.org/10.1093/icc/10.2.419 Arora, A., & Gambardella, A. (1994). Evaluating
technological information and utilizing it. Scientific knowledge, technological capability, and external linkages in biotechnology. Journal of Economic Behavior and Organization, 24(1), 91– 114. https://doi.org/10.1016/0167-2681(94)90055-8
Arora, A., & Gambardella, A. (2010). Ideas for rent: An overview of markets for technology. Industrial and Corporate Change, 19(3), 775–803. https://doi.org/10.1093/icc/dtq022 Arque-Castells, P., & Spulber, D. F. (2017). The
Market for Technology: Harnessing Creative Destruction. SSRN Electronic
Journal, (November).
https://doi.org/10.2139/ssrn.3041558 Arque-Castells, P., & Spulber, D. F. (2019).
Measuring the Private and Social Returns to R&D: Unintended Spillovers versus Technology Markets. SSRN Electronic
Journal, (February).
30 Arque-Castells, P., & Spulber, D. F. (2020).
Firm Matching in the Market for Technology: Harnessing Creative Destruction. Academy of Management Proceedings, 2020(1), 19790. https://doi.org/10.5465/ambpp.2020.197 90abstract
Atuahene-Gima, K. (2005). Resolving the Capability – Rigidity Paradox in New Product Innovation. Journal of Marketing, 69(October), 61–83.
Ayfer, A., & Cockburn, I. (2016). Buyer Behavior in Markets for Technology: Technology Proximity Between Firm Portfolio and In-Licensed Patents. Mimeo. https://doi.org/10.1017/CBO9781107415 324.004
Belenzon, S., & Berkovitz, T. (2010). Innovation in Business Groups. Management Science, (56), 519-535.
Bloom, N., Schankerman, M., & Van Reenen, J. (2013). Identifying Technology Spillovers and Product Market Rivalry. Econometrica, 81(4), 1347–1393. https://doi.org/10.1017/CBO9781107415 324.004
Bloom, Nicholas, Jones, C. I., van Reenen, J., & Webb, M. (2017). Are ideas getting harder to find?†. American Economic Review, 110(4), 1104–1144. https://doi.org/10.1257/aer.20180338 Branstetter, L., Fisman, R., & Foley, C. F.
(2006). Stronger Intellectual Property Rights Increase International Technology Transfer? Empirical Evidence from U.S. Firm-Level Data. NBER WORKING PAPER SERIES.
Caviggioli, F., & Ughetto, E. (2013). The drivers of patent transactions: Corporate views on the market for patents. R and D Management, 43(4), 318–332. https://doi.org/10.1111/radm.12016 Ceccagnoli, M., Graham, S. J. H., Higginsy, M.
J., & Leez, J. (2010). Productivity and the role of complementary assets in firms’ demand for technology innovations.
Industrial and Corporate Change, 19(3), 839–869.
https://doi.org/10.1093/icc/dtq033 Chandler, A. D. (1990). Strategy and Structure:
Chapters in the History of the Industrial Enterprise. MIT Press, 120.
Chesbrough. (2006). Open Innovation: A New Paradigm for Understanding Industrial Innovation. Oxford University Press, 75– 80. https://doi.org/citeulike-article-id:5207447
Cockburn, I. M., & Henderson, R. M. (2001). Scale and scope in drug development: Unpacking the advantages of size in pharmaceutical research. Journal of Health Economics, 20(6), 1033–1057.
https://doi.org/10.1016/S0167-6296(01)00108-4
Cohen, W. M., & Levinthal, D. A. (1990). Absorptive Capacity: A New Perspective on Learning and Innovation. Administrative Science Quarterly, 35(1), 128. https://doi.org/10.2307/2393553 De Marco, Scellato, G., Ughetto, E., &
Caviggioli, F. (2017). Global markets for technology: evidence from patent transactions. Research Policy. https://doi.org/10.1016/j.respol.2017.07.0 15
De Rassenfosse, G., Palangkaraya, A., & Webster, E. (2016). Why do patents facilitate trade in technology? Testing the disclosure and appropriation effects.
Research Policy.
https://doi.org/10.1016/j.respol.2016.03.0 17
Eurostat. (2019). Intramural R&D Expenditure (GERD) by Sectors of Performance. Retrieved October 23, 2020, from http://appsso.eurostat.ec.europa.eu/nui/s how.do?wai=true&dataset=rd_e_gerdtot Figueroa, N., & Serrano, C. J. (2019). Patent
trading flows of small and large firms.
Research Policy.
31 Fosfuri, A. (2006). The Licensing Dilemma:
Understanding the Determinants of the Rate of Technology Licensing. Strategic Management Journal, 27(12), 1141– 1158. https://doi.org/10.1002/smj
Foss, N. J. (1996). Firms, Incomplete Contracts and Organizational Learning. Human Systems Management, 15, 17–26. Foss, N. J. (1997). On the Rationales of
Corporate Headquarters. Industrial and Corporate Change, 6(2), 313–338.
Retrieved from
https://academic.oup.com/icc/article-abstract/6/2/313/794297
Foss, N. J. (1999). The Use of Knowledge in Firms. Journal of Institutional and Theoretical Economics (Jite)/Zeitschrift Für Die Gesamte Staatswissenschaft, 270–284.
https://doi.org/10.1142/9789812701275_ 0025
Gans, J. S., & Stern, S. (2003). The product market and the market for “ideas”: Commercialization strategies for technology entrepreneurs. Research Policy. https://doi.org/10.1016/S0048-7333(02)00103-8
Gibson, C. B., & Birkinshaw, J. (2004). The antecedents, consequences, and mediating role of organizational ambidexterity. Academy of Management Journal, 47(2), 209–226. https://doi.org/10.2307/20159573
Giuri, P., Mariani, M., Brusoni, S., Crespi, G., Francoz, D., Gambardella, A., … Verspagen, B. (2007). Inventors and invention processes in Europe: Results from the PatVal-EU survey. Research Policy, 36(8), 1107–1127. https://doi.org/10.1016/j.respol.2007.07.0 08
Graham, S. J. H., Marco, A. C., & Myers, A. F. (2018). Patent transactions in the marketplace: Lessons from the USPTO Patent Assignment Dataset. Journal of Economics and Management Strategy,
27(3), 343–371.
https://doi.org/10.1111/jems.12262 Griffith, R., Lee, S., & Straathof, B. (2017).
Recombinant innovation and the boundaries of the firm. International Journal of Industrial Organization, 50, 34– 56.
https://doi.org/10.1016/j.ijindorg.2016.10. 005
Grindley, P. C., & Teece, D. J. (1997). Managing Intellectual Capital: Licensing and Cross-Lincensing in Semiconductors and Electronics. Research Technology Management.
https://doi.org/10.1080/08956308.2005.1 1657346
Gulati, R. (1998). Alliances and networks. Strategic Management Journal, 19(4), 293–317.
https://doi.org/10.1002/(SICI)1097- 0266(199804)19:4<293::AID-SMJ982>3.0.CO;2-M
Gulati, R., Nohria, N., & Zaheer, A. (2000). Strategic networks. Strategic Management Journal, 21(3), 203–215. https://doi.org/10.1002/(SICI)1097- 0266(200003)21:3<203::AID-SMJ102>3.0.CO;2-K
Hagui, A., & Yoffie, B. D. (2013). The New Patent Intermediaries: Platforms, Defensive Aggregators, and Super-Aggregators. Electrical Engineering,
27(1), 45–66.
https://doi.org/10.1109/ee.1953.6438261 Hall, B. H., Jaffe, A. B., & Trajtenberg, M.
(2001). The NBER Patent Citation Data File: Lessons, Insights and Methodological Tools. Working Paper 8498.
Hamel & Prahalad, C. K. (1990). The core competence of the corporation. Harvard Business Review, 29, 147–158.
