Convergence in Technology Space and Strategic Alliance Activity 1

Transcription

1 Convergence in Technology Space and Strategic Alliance Activity 1 Anindya Ghosh The Wharton School of the University of Pennsylvania 2062 Steinberg Hall - Dietrich Hall Philadelphia, PA (215) anindya@wharton.upenn.edu Johannes Pennings The Wharton School of the University of Pennsylvania 2107 Steinberg Hall - Dietrich Hall Philadelphia, PA (215) pennings@wharton.upenn.edu 3 May 2008 Paper prepared for INSEAD Network Conference. 1 Support from the Mack Center at the Wharton School and the Wharton-SMU Research Center in Singapore is gratefully acknowledged. We also appreciate the comments from Gino Cattani, Mike Halperin, Michael Jacobides, Daniel Phelps, Sheen Levine, Evan Rawley, Xavier Martin, Gordon Walker, Filippo Wezel, Matthew White and Jason Woodard.

2 Convergence in Technology Space and Strategic Alliance Activity ABSTRACT This paper investigates the patenting activities of firms in the imaging sector by constructing its so called "technology space" as platform of bundled knowledge that presents incentives for strategic collaboration. We identify the relative centrality of firms to make inferences about their organizational status to account for their propensity to engage in strategic alliances. The status system is derived from five year windows of organizational networks through mutual citation of intellectual capital (i.e. patents) that we construe as an evolving technology space, and allows a determination of the flow (as distinct from stock) of status to predict their propensity to engage in strategic alliances. Higher status firms, afflicted with loss aversion exhibit lower while firms with established knowledge legacy exhibit higher alliance propensity. We also consider the firms degree of knowledge diversity, their inclusion in the technology space relative to overall R&D output as complements to status, spatially defined. The implications of the findings are reviewed through the lens of technological evolution and dominant design. Keywords: Technology space; convergence; alliance; imaging; technology evolution, status

3 INTRODUCTION Established or mature industries often exhibit strategic renewal through R&D efforts, strategic alliances or joint ventures, mergers & acquisitions and new product introductions. The renewal might wreak havoc in industries when discontinuities are so radical that their arrival renders obsolete many of the current skills or other assets and result in the elimination of several if not all market incumbents, even if certain skills continue to persist and maintain strategic value. In this paper, we examine the aggregate output of firms' R&D, which comprises their collective "technological space" and creates opportunities for collaborative relationships among them. The technology space comprises clusters of technological knowledge through connections among firms whose collective intellectual capital reflect convergence and substitution. By tracking the mutual citations of firms in the imaging sector, technological networks become exposed furnishing insights in the process of convergence, and thus potentially triggering shifts in status stratification among its established and new incumbents. Status is deemed critical when firms select strategic partners in monetizing their R&D output and we will furnish two hypotheses to spell out the alliance propensity as a function of inferred status. The evolutionary clustering among firms motivates them to engage in joint ventures and other forms of strategic pooling of strategically relevant technology subject to their aversion of status leakage and motivation to recycle complementary assets. Our objective is therefore two fold: identify the firms location in the imaging sector, here framed as evolutionary technology space and account for their collaborative strategic behaviors such as alliances due to shifts in status differentiation that emanates from that evolving space.

4 It has been somewhat conventional to position innovative conduct along a continuum of pure research and development to product-market, in that firms participate in the accumulation of technology producing or dissolving clusters of proprietary intellectual capital. Such collective gales of creative destruction are discernible from emergence of bundles of patents, some of which become converted into components and architectures of new or improved product-market applications. Yet, a clear one-to-one mapping of firms from their basic knowledge to IPR and then to product-markets is tenuous, if not elusive and problematic (compare Ahuja & Kattila, 2002). This paper attempts to make such a connection by tracing strategic alliance activity to a firm s topography in the technology space conceived of as a status system (Podolny 2005) Specifically, we claim that transformations in the photographic imaging sector, evolving from mechanical-chemical to electronic platforms provide an impetus for the formation of joint ventures and other alliances. We ask whether imaging firms engage in alliances as they encounter growing density and clustering of domains of technology and the boundaries of their space become fluid and shifting. The technology space comprises the collective intellectual capital of firms around pin hole imaging and will be mapped through the connections that surface by their mutual citations. Firms self-accredit as imaging firms when they join the space by dint of citing or becoming cited by other firms who participate in imaging relevant technology domains. It would be tempting to invoke conventionally available categories such as SIC codes (e.g. Cool & Schendel, 1980, Henderson & Clark, 1989), ecological communities (Ruef, 2004) or populations (e.g., Hannan et al, 1995) and regions like Silicon Valley (Phillips, 2005) Detroit (Klepper, 2007) or Reggio Emilia (e.g., McEnright, 1999). Such categories are deficient in capturing the strategically relevant technology space. Instead, in view of its spatial ambiguity, we should delimit a firm's domain as a self-referential platform onto which it enters through lasting

5 R&D and other commitments in technology and products. We selected the imaging sector because its technology constitutes on of the prime contemporary exemplars of convergence most notably between its chemical, optical, electronic and other classes of knowledge--even if some of its product-market features (e.g., optics) exhibit continuity. Its incumbents and new entrants face interesting and daunting prospects in accommodating legacy and emergent technologies as their technology space evolves and new product-market components and architectures reach the market. We believe that their space can be represented as a "coordinate system" whose evolutionary clustering and contours uncover the shifting location of firms and their attendant propensity to engage in strategic networking. We go even further and suggest that firms engage in strategic alliances on the basis of their status in the technology space perhaps to capitalize on the value of their unique or shared intellectual capital which when pooled with that of other firms grows in value and permits them to acquire higher returns form their R&D investments (compare Arora, Fosfuri and Gambardella, 2003). Recursively, joint ventures might also accelerate and even alter the technological trajectory when firms merge their technologies and consolidate its propagation (compare Schilling and Phelps, 2007) but such considerations are beyond the scope of this paper. Rather we attempt to predict a firm's propensity to selectively engage in strategic alliances when faced with R&D output of other firms which are attractive to the focal firm in the technology space, the evolving coordinates and internal differentiation we attempt to identify. If faced with a strategic partner, enjoying high status, the focal firm might derive higher rents form its intellectual capital if pooled accordingly. Status is inferred form the firms evolving centrality as determined by their network embeddedness.

6 In the following parts of this paper we further belabor the notions of technology space exhibiting technological convergence. We provide a theoretical outline for framing the evolutionary structure of firms within the imaging space and its defining properties such as clustering and status differentiation. We present an empirical analysis of technological networking but also consider the firms legacy, diversification and their actual inclusion in that technology space over the period as a precursor for the level of alliance formation. We find ourselves then in a position to account for their re-combinative conduct at more tangible levels for example as strategic partners in a conventionally defined sector where intangible, proprietary assets become bundled. THEORY Technology Space Further Defined The theoretical framework presented here starts from the premise that firms are embedded in a network of knowledge flows, often illustrated and quantified by patent citations or mobility of scientists and engineers (e.g., Fleming, 2006). Such networks have a fractal-like structure in which vertices, (or agents, firms or actors) cluster together into groups that then join to form groups of groups (Clauset, Moore and Newman, 2006). Hierarchies of such clusters invite us to locate or describe communities through an evolutionary perspective (compare Chakarbarti, Kumar and Tomkins, 2006) and might be interpreted as enabling or inhibiting alliance activities among actors. For example, firms in a given cluster might view their intellectual capital more complementary with members of different and divergent clusters or communities. Networking has become a prominent concept in organization theory and strategic management providing conceptual inputs in framing firms in their competitive context as well as

7 a set of methodological tools for measuring and estimating the impact of connections among firms. In the present study we explore networking for a theory about organizational environments in terms of communities" (e.g., Clouset, Moiore and Newman, 2006) with status differentiation (e.g., Podolny 2005). We do so by also leveraging new developments in network analysis, most notably data reduction tools for very large data sets. Network "theory" as a study of organizational environments has surfaced in diverse inquiries ranging from the study of innovation and performance along the value chain (Burt, 2004, Pontikes, 2008), development and creative renewal of artistic standards (e.g., Uzzi and Spiro, 2005) or organizational learning and evolution (e.g., Watts, 2004). Network methodology has expanded to handling large data sets covering both egocentric and network-centric metrics (Podolny and Stewart, 1995), and moving from cross sectional snapshots to longitudinal movies (e.g., Chakrabarti, Kumar and Tomkins 2006; Moody, McFarland and Bender 2005; Clauset, Newman and Moore, 2004). A technology space informed theory entails the delineation of its boundaries and internal variations in density to spell out status differentiation among its incumbents. Firms place themselves into clusters through what we call acts of self-accreditation in which they legitimize themselves through application and granting of patents with attendant sanctioning through the acknowledgment of peer firms. The mere act of patenting self-accredits the firm and actually can be interpreted as a contributing to what others have called institutionalization (e.g., DiMaggio and Powell, 1983). The firm as incumbent in technology space can no longer be treated as a self-sufficient actor if such an assumption was ever tenable, even if framed with an anti-commons perspective (Murray & Stern, 2002: Bessen & Meurer, 2008). The firm as knowledge repository requires a

8 new perspective in which it becomes treated as being embedded in a wider context, such as technology space. The well known Teece (1995) discussion on appropriability and "isolating" mechanisms suggests that firms should retreat to those activities they excel in, or in which knowledge leakage is not jeopardized, while placing all other skills and knowledge beyond their legal boundaries. We assume that the firms' technology space amounts to some ordering of firms in terms of deference relationships and commensurate status differences. The connections among firms in terms of pools of intellectual capital, --e.g., through current or accumulated patents or employment of scientists and engineers and linked through citations and inter-firm mobility respectively, places them in the core or the periphery. Their topography reveals relative prominence or simply status in this technology space. Central firms are attractive partner candidates in the formation of joint ventures, or they might be more attractive targets for employment, especially for scientist and engineers but are disproportionately restrained due to their aversion of status leakage. Peripheral firms, not facing such downside risks, are located in the lower strata of the technology space and face higher thresholds in attracting alliance partners, for example. The two pronged perspective on environment-firm interface through patent citations and strategic alliances can be observed cross-sectionally but more informatively, longitudinally. Firms and their technological environments evolve over time and aggregate knowledge flows exhibit convergence in which strands of knowledge become more proximate or distant. As we elaborate below, in the research setting of this study, convergence is paramount with a putative change form chemical to electronic technology platforms. That very convergence produces changes in the status system and will push certain firms away from the center, while and other

9 firms migrate from the periphery including start-ups which might travel to its more central areas. Convergence alters the incentives to engage in alliances, and creates opportunities for bundling intangible assets among various communities of firms, subject to the earlier mentioned constraints of status differentiation. Status in Technology Space and Alliance Formation The social network literature has identified two main sources of value that may accrue to a firm due to its network position. First, its network position may provide the firm with opportunities for brokerage in the flow of information between clusters of firms otherwise disconnected from each other (Burt, 2005; Podolny, 2005). Social networks, however, are not only "pipes" (Podolny, 2001) or conduits that enable access to information. Relationships confer status, and act as a signal of the firm's quality or of its superior standing among peers (Podolny, 2001). As we suggested above status in terms of centrality is proxied by the extent to which a firm is linked to others. Podolny (2005) argues that status signals quality. An investment bank or California winery commands a premium if its location in an exchange network is central. Podolny and Stuart (1995) claim that an actor's status affects the attention that it receives in a community of innovators as revealed by citations around technological niches, even if the attention is misplaced, for example when an innovation fails. Citations indicate ties among firms R&D technological networking and suggest a stratification depending on how central they are in their network. For example, technological output from central, and derivatively high-status actors are more likely to be given attention and benefit from further injections compared to "lesser" technologies originating form peripheral firms. Following their logic, we would expect that high

10 status firms draw are more prominent as targets for alliance formation, but face the classic aversion loss (Thaler, 1984), while low status firms express stronger propensities to ally and need not contend with such loss. Centrality is such a status indicator with the shorter a firm's pathway in the knowledge space the more attractive is its candidacy for other firms who seek to monetize their complementary assets, or legitimize their identity as member of the sector (compare Tripsas, 2008). Podolny (2005) however, also claims that high-status actors are highly selective and restrict their availability to other firms seeking to engage in exchange relationships. He implies that prestige is susceptible to spillovers with the onerous implication that status might leak away if an alliance in the knowledge space involves two firms with unequal status. Conversely, a lower status firm might gain in prestige when bundling its assets with a higher status firm. The implication is that under conditions of convergence where uncertainty regarding partners is paramount, high status firms, residing in the center of the set of core firms, shun most of the other firms, while firms in the periphery aggressively pursue intellectually prominent firms to achieve greater legitimacy and other pay-offs in the knowledge space. Our position is therefore inconsistent with Stewart (1998) who argued that high status firms are more at risk in forming alliances. Stewart assumes that status confers a favorable bargaining position, permit firms to extract higher IPR rents, but disregards the considerable risk of leakage among high status firms: the higher the status the greater the disincentive to engage other firms in partnering since the potential loss weighs much more heavily compared to those with lower status (compare Thaler, 1984). For example Kodak would rather form a joint venture with a high status firm such as IBM or Fuji than with a newer or smaller, and by implication a more marginal firm whose legitimacy as digital player is (still) questionable. Its marginal status renders it more ambiguous as provider

11 of economically valuable assets that when bundled with a high status firm's intellectual capital improves the rents for the former. Note also that the status of a major photography player like Kodak or Polaroid might decline as their knowledge legacy decreases in value as proxied by diminished centrality. We argue therefore that low status firms seek high status partners to better exploit their proprietary capital. The quality of that capital is unobserved and cannot be specified, but the status as captured through knowledge flows inferred centrality can be specified and measured, and serves as a proxy for that capital's quality. From an evolutionary standpoint status or "prestige" is susceptible to spillovers, and might leak away towards a lower status firm. Lower status firms in a converging technology space are strongly induced to stimulate such leakage as it would elevate theirs. We rely on prospect theory of in that gains have a comparatively lower convexity than losses, consistent with the recent advancements in behavioral economics that stress the presence of loss aversion (compare Kahneman and Tversky, 1979; Thaler, 1984). Organizations are much more prone to avoid status leakage compared to their peers who are motivated to embellish their status. Hence we hypothesize: H1 The higher a firm's status in the imaging knowledge space, the lower its rate of alliance formation with peer firms Technology Capability and Alliance Formation A converging sector demands new skills and capabilities. Strategic alliances and joint ventures are an effective vehicle for accessing emerging or discontinuous knowledge (Dyer et al., 1998), especially when firms face obsolescence. Even if facing obsolescence, firms endowed

12 with important complementary assets might survive the onslaught of discontinuous technologies important to the actors endowed with the new technology (Teece, 1986; Tripsas, 1997). For example, the expertise of a high status firm like Kodak in distribution and marketing, its brand and imaging-knowledge and not just its chemical knowledge might be attractive to other firms in the imaging technology space. Thus, there is a double motivation for engaging in alliances for actors with a legacy in the entrenched technology. This scenario is well illustrated by two alliances in the imaging sector. Eastman Kodak and Canon have long collaborated to bring out novel products. They forged an alliance in 1985 whereby Kodak exclusively sold copiers made by Canon in the USA. Canon relied on Kodak's distribution channel for the marketing and sales of its copier product. The joint venture was further expanded in 1992 to ten other peripheral products. More recently the partners have extended their collaboration into the digital printing arena. A second joint venture example involves Fuji Photo Film and Microsoft. In 2000 the firms announced an alliance to make digital imaging more accessible for consumers. As part of the agreement, Fujifilm incorporated Microsoft Picture It! Express 2000 photo-editing software into every new Fujicolor CD. The Fujicolor CD was intended to make it easier and more affordable for consumers to download their photos into their PC and experiment with the editing of digital imaging. Conversely, Microsoft sought to leverage the consumer base of Fuji to promote its proprietary digital imaging software. In general we would assume that in the technology space, firms with a potentially more obsolete knowledge legacy, to have a greater incentive to partner with firms in the knowledge space whose technology platform is fresh or novel, such as to stretch the life of their intangible assets that might rapidly decline economically as the sector exhibits technological convergence.

13 Age might convey status, but as we will elaborate, we capture status by dint of a firm's betweeness centrality in its technology space, but apart from spillovers due to status, the legacy in a technology space undergoing a dramatic "paradigm shift," should matter as well..we should expect firms with a stronger chemical legacy and associated receding status, to bundle their cumulative knowledge with firms not historically confined by such obsolete technology in order to preserve their continuityand to exploit what Tripsas (1994) refers to as recycling proprietary bundles of technology as they enter a new era in their sector. Based on the above discussion it can be hypothesized: H2. The higher the firm's patenting in a technology category threatened by obsolescence, the greater its rate of alliance formation. EMPIRICAL SETTING Brief History of Imaging In 2003, sales of digital cameras exceeded sales of silver halide-based or analog cameras. Electronically based imaging had gradually replaced chemically based imaging. Imaging had been traditionally looked at as encompassing two domain, image capture and image display. The former category included the actual camera hardware that consisted of the optical system, mechanical and electrical system like auto focus and red eye reduction and the latter category consisted of sensitized materials to capture the image, including the negative and the photographic paper on which the images were developed. The key element of the image capture process and the traditional photographic industry is silver halide which when interacting with

14 light waves becomes transformed into an image and can be transmitted from film to paper, or projected onto some other medium like a screen. Digital or electronic imaging and its underlying intellectual capital expanded the domains of imaging further into image capture, image storage, image manipulation and image display. It entailed taking pictures and developing those pictures using electrons instead of film and then transmit, store, and process these images electronically, as if they were files of data, unlike silver halide based imaging where the film and paper is covered by a layer of silver embedded substrates. Many of the challenges especially related to the optical part were similar. In fact, early digital cameras looked like their analog counterpart and were units with the film subsystem replaced by a CCD (Charge Coupled Device) unit. However, many of the same problems, the red eye reduction technology for example, required new solutions completely different from the chemical based methods used in analog cameras. The basic image capture technology is based on the CCD sensor, which serves the function of converting light energy into a digital data file. The CCD technology remained virtually dominant until recently, when CMOS (Complementary Metal Oxide Semiconductor) based technology began to replace it; sensors using CMOS sensors are about ten times as energy efficient as CCD's, and cost substantially less. The earliest versions of digital cameras did not have any storage device, thus severely constraining their portability as it was meant to be a computer peripheral. Today, two major competing formats for the storage of digital photo files exists; removable storage cards and micro drives. These are removable media, which effectively function like a roll of film. The file format in which digital images are transferred to a computer, and then further undergo manipulation constitute still another critical aspect of the digital imaging concept. Competing alternatives are currently available for the format in which digital

15 imaging files can be stored, as well as for the software needed to manipulate and use them. Finally, a microprocessor chip, which controls the operation of the camera, completes its product architecture. Its key metrics are speed and size. In addition, most present day digital cameras have an LCD display, and a lithium battery to meet the power requirements. Related components of the camera's "eco-system" are printers, computers and other visual display tools such as MP4 and Blue-Ray. Digital imaging technology emerged in the early 1970's; this technology is intricately linked to computer technology, benefited from the associated so called direct network" (Katz & Shapiro, 1985) effects and as costs of computer processing fell, that technology rapidly diffused into other areas. From the realm of consumer electronics, the development of video cameras had an impact on the way initial digital cameras were configured. Video technology in pioneering cameras like the Mavica by Sony had already shown that it was possible to dispense with film, though that video industry remained firmly rooted in analog technology till the late eighties. Before 1990, the usage of digital photography was largely restricted to a few scientific (medicine and satellite imaging) and commercial (publishing and real estate marketing) applications. Its primary advantage included the ability to manipulate and edit pictures on computers and the ease and speed of development, storage, recall and transmission. With decreasing costs and increasing functionality among component digital technologies, particularly semiconductors, computer hardware and software, the digital camera has since been making steady inroads into conventional, silver halide based film, but has also been spawning new products and services such as optics, photolithography and the internet. The seeds of this imminent convergence through clustering and fragmentation of disparate knowledge bundles were already discernable in the late 1970s, early 1980s as revealed by the clustering of firms in this technology space.

16 Figure 2, revealing fast "greedy community" detection (Newman and Girvan, 2004) shows the clustering of the firms that comprised the technology space in the period , derived from their patent citation based between ness (Limitations in visual display of such quantitative information for later periods is computational impossible). The imaging arena now contains participants from multiple industries including many of its progenitors- the chemically based, silver halide photographic technologies with major players such as Kodak, Agfa, Polaroid and Fuji, for whom digital imaging technology represented a competence destroying discontinuity. The knowledge that would support future products in imaging like the celebrated CD Photo and APS endured a truncated life perhaps because the convergence between old and new techno logy was not as promising as it seemed at he onset of the microelectronic sweep (compare Braun & Macdonald, 1982). One group of new entrants into imaging came from the consumer electronics industry (e.g. Panasonic), and attempted to leverage their experience with video cameras into digital imaging. Yet another group of firms originated from the graphic arts and printing industry, which had pioneered the use of electronic scanning. Finally, many entrants entered from computer hardware, software and semiconductor industries (e.g. Intel, Hewlett Packard, Adobe) as digital cameras began to be accepted as computer peripherals. Imaging today draws on technological competencies from the semiconductor and electronics industries, computer hardware and software industries, and conventional film based imaging industries. Figure 1 provides a graphic display of the evolutionary trajectory of imaging as a "technology space" in which gradually the chemical knowledge becomes replaced by electronic knowledge. The diagram is based on the firms' R&D output as captured by the granting of patents.

17 Figure 1 and 2 about here Imaging "Technology Space" The brief history of the imaging sector over three decade reveals digital technologies to increasingly dominate a sector once heavily entrenched in chemical knowledge. It also shows the importance of complementary domains such as optics, mechanical controls and electrical circuitry. To capture the evolution of all these areas individually or jointly is a challenging task. However, one newer tradition in the management of technology and innovation has been the growing interests in patenting and the resulting body of intellectual property. Patent data have been used extensively to map knowledge flows and to create a topography of technology space (Fleming & Sorenson, 2001, 2004; Jaffe, Trajtenberg, & Henderson, 2002; Podolny & Stuart, 1995; Podolny, Stuart, & Hannan, 1996; Stuart, 1998; Stuart & Podolny, 1996). Their limitations and even empirical shortcomings notwithstanding, patents and their citations amount to a trail of knowledge flows (Jaffe et al., 2002) and enable the identification of technological trajectories and the uncovering of new and emergent sectors. Furthermore, patents provide detailed information about the underlying invention and the domain of knowledge to which it belongs. Several authors have tried to categorize the technology space and the European Patent Office even provides an IP "web-guide" by country, scientific field and other classes. Hall, Jaffe and Trajtenberg (2001) classified US patents into six technological categories: Chemical (excluding Drugs); Computers and Communications (C&C); Drugs and Medical (D&M); Electrical and Electronics (E&E); Mechanical; and Others (see Appendix 1 of, Hall, Jaffe and Trajtenberg (2001)). They correctly indicate the arbitrariness in devising such an aggregation system and in assigning the patent classes into various technological categories. However, their

18 classification scheme could serve us well in our efforts to create a topography of the photographic knowledge space. The two of their broad categories, Chemical and Computer & Communication (C&C) crudely capture knowledge in Chemical and Digital domains,.by contrast, the Electrical & Electronics category captures only a small portion of that knowledge which is related to semiconductor patents and should not affect the overall characterization of the knowledge space. In this paper we construct the knowledge space by dint of a time evolving network of firms from 1976 to 2002 whose accumulation of intellectual property becomes spatially tied to that of other firms through their backward patent citations. In other words, backward citations define the topological location of firms in the evolving field of photographic imaging. As we describe below, in the Method section, 23 strings of networks using a five-year sliding window were obtained to document the evolution of that knowledge spaces as a status system followed by an effort to predict the firms' propensity to establish strategic alliances that should emanate from organizational status as a precursor for alliance formation. We believe that presumption to be justified as convergence creates incentives for firms to monetize their intellectual complementarities with lower status firms having a stronger proclivity to bundle their intellectual capital with that of high status partners. The Imaging "Technology Space" Empirically Defined The delineation of sector boundaries is often carried out by conveniently available categories such as patent class or some SIC based scheme. While SIC codes have a technological product focus, patents are strictly based on claims, not on industry, function or effect. We chose to construct a spatial unit by consulting industry experts in the imaging sector to provide a set of "focal patents" to define the imaging knowledge space. Scientists at Eastman Kodak provided us

19 a search algorithm through Micropatent, a US patent search provider, and collected a set of thirty-five thousand odd focal patents filed in the USPTO that are deemed central in the imaging sector and which is used by Eastman Kodak for conducting intellectual property intelligence. Currently, we have adopted a triangulated search algorithm to validate the choice of these focal patents, but the map which we furnish here is based on the Kodak furnished domain of intellectual property. This Kodak defined knowledge space furnishes a unique window into the evolutionary trajectory in this dynamic and converging sector. Once we obtained the 35,473 focal patents we also collected all the patents that these focal patents cited and were cited by over the period from the USPTO. We used the NBER patent dataset (Hall, Jaffe, & Trajtenberg, 2001) to obtain information on the citations and the pertinent technology categories. To map the convergence of technologies we plotted the patenting activity by the six main technology categories as defined above. Figure 1 shows the overall patenting in the six categories from 1976 to Method Data Sources The data from this study came from a variety of sources. The patent data are obtained from the USPTO stored in readable format by the NBER and compiled by Hall, Jaffe and Trajtenberg (2001). For the alliance data SDC platinum was sourced. The populations of interest for this study are the incumbents of what we have called the imaging technology space. The delineation of sector boundaries is often carried out by conveniently available categories such as patent class or some SIC based scheme. While SIC codes have a technological product focus, patents are strictly based on claims, not on industry,

20 function or effect. We constructed a spatial unit by consulting industry experts in the imaging sector to provide us with a set of "focal patents" to define the imaging space. Scientists at Eastman Kodak shared with us a search algorithm through Micropatent, a US patent search provider, and collected a set of approximately thirty-five thousand focal patents filed in the USPTO that are deemed central in the imaging sector and which is used by Eastman Kodak for conducting intellectual property intelligence. This Kodak R&D defined knowledge space furnishes a unique window into the evolutionary trajectory in the dynamic and converging sector and helps us set the population of firms in the sector. The 3039 firms with patents in this focal set is our population of interest for alliance incidence analysis. Once we obtained the 35,473 focal patents we also collected all the patents that these focal patents cited and were cited by over the period from the USPTO. We used the NBER patent dataset (Hall, Jaffe, & Trajtenberg, 2001) to obtain information on the citations and the pertinent technology categories. Next the inter-firm network in the knowledge space through backward patent citations was extracted. The focal patents yields a universe of 178,795 patents from through backward and forward citations which are assigned (conferred legal ownership) to 18,371 unique proprietors including firms, individuals, government agencies, universities and hospitals. We were interested only in firms as assignees, so we discarded all other ownership classes reducing them to a set of 17,400 firms. We further checked for duplicates and joint ventures by using a name matching algorithm. This laborious process reduced the number of firms to 16,475. This forms the population of firms in the knowledge space of the imaging sector and is a superset of our main population of interest. This population is mainly used to calculate the status of the firms in the knowledge space.

21 Next we searched SDC platinum for data on alliance activity of the 3039 firms in our population of interest. We were able to collect alliance data on 249 firms from SDC for the period 1984 and Since the data before 1988 in SDC are not reliable we selected a window of 10 years from 1989 to 1998 to predict alliance formation among these firms. This set of 249 firms is our main sample of interest for analyzing alliance formation. The sample is a convenient sample of firms that have at least one alliance reported in SDC. Therefore our results are not generalizable to firms in the entire population. Variables The main variables constructed are those that characterize the firm in the "technology space" and are obtained from the patent data by constructing networks using backward citations. Since the USPTO provides patents of firms as well as joint ventures of the firms, we assigned to the parents of each joint venture the patents in proportion of the number of firms in the joint venture. Thus patent counts in our sample are not always an integer. We then created 23 networks for the 16,475 firms using a sliding window of five years from 1976 using Pajek, a tool for creating and analyzing large networks (Batagelj & Mrvar, 1998). Next the betweenness network centrality of each firm, that is, the number of times the shortest path between two other firms in the network contain the firm of interest, was calculated. Additionally, a technology diversity measure of patenting of each firm using the six categories of technologies defined by Hall, Jaffe and Trajtenberg (2001) was constructed. The six relevant categories are Chemical (excluding Drugs & Medical), Computer & Communication (which we take as a proxy for Digital technologies), Mechanical, Drugs & Medical, Electrical and Electronics and Other. The

22 diversity measure is a Herfindahl Index, Hi, of the patents in each of the six categories for each firm H i: r X p j, p J is the proportion of patents in category j. j= 1 The variables that we extracted from the NBER and SDC archives are described in table 3a and Table 1 b furnishes descriptive statistics and product moment correlations. Tables 1a and 1 b about here Research Design & Method In the ideal experiment to test the question above, a set of randomly chosen firm from the population of firms in the sector would receive the treatment of being assigned status (network centrality). The alliance activity of this group would then be compared to a control group. However, this is not possible as we have observational data on a subset of firm that undertake alliance. Therefore causality will be assumed using a response schedule given below. Confounding issues, the model and method of analysis is described below. Confounding & Inferential Issues The major confounding in this case is that status and alliance activity are interdependent event and therefore reverse causality could potentially lead us to wrong inferences. Also unobserved firm level time invariant fixed effects can potentially explain alliance formation. To overcome this we use fixed effects and panel data and lag the alliance count, that is, the status at time t will predict the alliance activity in time t+1. Furthermore the status is calculated for a network of patent citations from t-4 to t. This leading five year window and given that firms have

23 relatively less flexibility when it comes to citing other firms' R&D output compared to for example academic citation with their "commons" and "public goods" identity, and moreover often incur licensing fees not to mention the presence of a third party, i.e., a patent examiner, who might impose additional citations, all this reduces the confounding issues due to reverse causality. In short it is rather counterfactual to claim that alliance activity influences status as we have operationalized. A related confounding issue involves time variant firm attributes which is correlated with status. Our regression attempts to account for such attributes as far as possible. Sample selection that could limit the generalization of our results is beyond the scope of our research objective and will be further discussed below. Model, Method and Analysis The research design models alliance formation over time as a Poisson process and approximates a quasi-experiment in which rival hypotheses are ruled out. However, the data show over-dispersion with variance much larger than the mean mandating instead a negative binomial model for the regressions. We use a conditional fixed effect Negative Binomial for panel data as used in most of the patent literature for over-dispersed count data (Hausman, Hall & Griliches, 1984). The negative binomial model is a generalized form of a Poisson model where an individual, unobserved effect is introduced in the conditional mean (Greene, 2000): E (Alliance it+1 I X it ) = exp (P'X it + a*allianceexp.+ v i +8 t +^'F it +Si t ) Allianceit+i is the dependent variable, the number of alliance by firm I and time t+1. X it is a set of variables that characterizes the firms in the "technology space" and is based on the patent data as described above and listed in Table 1a and 1b:

24 Network centrality (Betweenness) captures firm status Patenting in the six technology categories to capture the importance of chemical or digital legacy of the firm (Chem IP, C&C IP, EEE IP, Drug IP, Mech IP and Other IP). Hi (IP Specialization) the diversity of patenting in various categories to control for firms that are diversified across technologies against those who are specialized. IP Flow (total patenting by a firm in a given year), Technology-Space Participation (percentage of patents of total in the imaging knowledge space) and Technology- Space Entry (dummy specifying the time period when a firm entered the knowledge space). The stock of strategic partnering (AlliancExp) has been shown to be a predictor of alliance formation and therefore is included as control variable through the firm's cumulative count till time t-1 (Gulati, 1999). y i is a set of firm dummies to capture firm fixed effects and 8 t would be year dummies to capture time trends. Fit is a set of financial variables to account for the size of the firm, profitability and resources in case they affect alliance activity. s it is everything else that influences alliance formation not captured by the technology, financial and experience variables and the firm and year dummies. We assume a one-year lag between our dependent variable and our regressors. We use the xtnbreg command in STATA with the fe option to fit our data to the conditional fixed effect negative binomial model. Allison and Waterman (2002) point out that the conditional fixed-effects negative binomial model is not a true fixed-effects model since it fails to control for all of its predictors. Therefore to verify our results we also fit a Poisson fixed effects model although it does not handle over-dispersion. An unconditional negative binomial

25 model is not used as it is not recommended when there are more than 20 panels (Hilbe, 2007). We also use Tim Simcoe's xtpqml STATA code for robust standard errors in the Fixed Effects Poisson Model based on Wooldridge algorithm (1999) which shows that the fixed effects Poisson estimator produces consistent estimates of the parameters in an unobserved components multiplicative panel data model under very general conditions. In fact, all that is required is an assumption about the conditional mean of the dependent variable and quite useful for two reasons. First, it implies that fixed effects Poisson estimation is appropriate for any non-negative dependent variable not just count that follows a Poisson distribution. Second, the estimator is robust to arbitrary patterns of serial data correlation. The final model we report here does have any financial information as we found that they don't have any effect on alliance formation as predicted by prior literature (Gulati, 1999). Results Table 2 show sour result. We only report the results from the conditional fixed effect negative binomial regression. We find that status (Betweenness Centrality) predicts alliance formation with a p-value of We also find that IP Specialization (p-value 0.015), Chem IP (number of patents in chemical technologies, p-value 0.03) and Technology Space Entry (dummy specifying the time period when a firm entered the space, p-value 0.006) are significant. All year dummies expect for 1990 are highly significant. We got consistent results with conditional Poisson fixed effects model and the fixed effects model with xtpqml. Table 2 about here

26 Interpretation and Discussion of Analysis Our results confirm our supposition that status of a firm predicts alliance formation in the future for our sample of firms. To answer our research question, the higher the status the lower the alliance rate for the firm. However, the effect size is small. For example, an increase of status by 100,000 results in change in expected count of 0.98, i.e., a 2% decrease in expected alliance count. The implication is that firms need to make substantial R&D investments and patent in large numbers over several years to increase their status before their addition to status propagates their future alliance formation. An alternative interpretation suggests that they need to patent innovations which build on number of innovations from different firms and which themselves are sufficiently important to be leveraged as status enhancement. This is a challenging task which adds to our argument of ruling out the reverse causality argument. Below we discuss some caveats of our analysis. Caveats Our analysis is contingent upon various limitations. First, as we pointed out before we use a convenient sample from our population. Ideally we would like to develop a sample selfselection model. Hilbe (2007) points out that Heckman and similar approaches are not appropriate for use with count response model. Recent work by Greene (2006) and Terza (1998) along with a selection model of alliance formation may help us develop a sample self-selection model. Right now our results hold for those forms that have at least one alliance from 1989 onwards involving firms with higher patenting activity. Second, we fit our model using a procedure that is known to have problems. Although we have used multiple methods to ensure the robustness of our results, we need to address this issue further and if needed develop code for our purposes.

27 Finally, we could not refine our analysis for different categories of alliances as we had foreseen in the proposal. The categorization of alliances into Chemical and Digital is not an easy process and we were not able to finish it in time. This is something we will do going forward. DISCUSSION This paper sought to expose a connection between convergent trends in a comparatively well-bounded status system and the subsequent strategic collaboration among corporate owners of proprietary knowledge. The status system under study comprises the various technologies associated with what we have called the (pinhole based) imaging sector -- still and moving imaging, the firms of which can be represented topographically in a technology space derived from a network of patent citations. In order to expose such a connection, we theorized that patent citations by owners of intellectual property proxies' complementarities and substitution and motivates them to combine their knowledge in order to exploit it commercially. We investigated the citation derived networking among firms constructing an evolving status system to obtain insights about its boundaries and internal stratification. Thus we could construct an evolutionary trajectory of their status system with convergence and substitution. Like any other community of people or organizations, the imaging sector manifest differences in status resulting in deference relationships among them. Patenting produces variations among the owners of intellectual property in terms of their position in a network of relationships based on mutual citation densities. Low status firms have a stronger proclivity to complement their intellectual property with high status partners than vice versa compared to high status firms exhibiting avoidance behavior towards those in the subordinate strata. Such differences in convexity, common in research on decision-making and judgment (e.g.,

28 Kahneman and Tversky 1979) are also presumed to exist in the effects of status on alliance propensity. We followed a two step investigation of the status system. First we ascertained a rolling 5 year representation of the relative centrality among the firms during the period , the era during which imaging evolved form mechanical/chemical imaging to electronic imaging. For example the over firms belonging to imaging resulting in each five year window in a betweeness centrality network permitted the identification of central, or core and peripheral firms in terms of their status, the presumption being that firms between other firms in the network enjoy higher status and are therefore less inclined to engage in strategic alliances compared to firms residing in the periphery, all under the presumption that high status firms avoid status leakage while low status firms seek status gains by partnering with high status firms and thus securing higher rents from their intangible assets, all this while controlling for the focal firm' s alliance experience. These results were complemented with other interesting findings regarding firms and their technology space for example the degree to which firms participate in that space or disperse and connect their R&D productivity over one or more other sectors. A related and equally interesting finding suggests that incumbents might also leverage their proprietary complementary asset legacies and offer attractive targets for corporate development activities among other players in the sector. Incumbent firms tend to have lower status, and their legacy knowledge might eventually decay as the sector migrates to substations in the present case electronic technologies. This is borne out by our finding that patenting in the Chemical field is associated with heightened alliance activity. This result could be interpreted as still another manifestation of the Tripsas effect: obsolete technologies might be recycled during a process of technological convergence when appropriately bundled with new capabilities.

29 (Tripsas, 1997). Empirical issues involve size of the status system, which renders computational methods challenging. Thanks to recent work in computing science (e.g., Chakrabarti, Kumar and Tomkins, et al. (2006) and Newman and Girvan (2004) the threshold for conducting network analysis on large samples is lowering and becomes it feasible to map firms into a network with clusters varying in centrality. A second issue is the empirical justification for the inclusiondecision rule, especially that involving what has become known as sample selection issues. The prediction of strategic alliances was carried out on a set of firms which were classified as central to the status system. Peripheral firms were excluded. In fact only 242 firms out of an aggregate of over firms informed about the propensity to expand organizational boundaries, thus restricting the generalizability of our results. Future research should address the differences and similarities among firms to be included and excluded in samples which are deemed part of a community of practice or knowledge, a market, industry or sector. Such issues need to be resolved if we want to understand how and where joint ventures (and other strategic conduct such as mergers and acquisitions, licensing agreements, competitive intelligence) become implemented. If successful such data files can also inform us on the recursive effects of joint ventures on the changing configuration of the status system (Oxley and Sampson, 2006). Thus far our study provides limited but useful insight both for research and practice of the evolving imaging sector. It opens up a variety of interesting issues to for researchers in strategic management, technology management and networking. Much of the conventional research on patenting has centered around innovation, property rights and assets defining the strategic options. The concept of status system is instructive for circumscribing status as a basis on which

30 firms compete and collaborate and should complement the research on reputation and branding as attributes which have been left out.

31 Figure 1. Chemical and Digital patent evolution from

32 Figure 2, Clustering in technology Space during initial period ( )