modes of cooperative r&d commercialization by …8].pdfthe context (dyer, kale, and singh,...

Strategic Management JournalStrat. Mgmt. J., 30: 835–864 (2009)

Published online in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/smj.765

Received 9 July 2007; Final revision received 11 February 2009

MODES OF COOPERATIVE R&DCOMMERCIALIZATION BY START-UPS

VIKAS A. AGGARWAL and DAVID H. HSU*The Wharton School, University of Pennsylvania, Philadelphia, Pennsylvania, U.S.A.

This study empirically examines the determinants of heterogeneous firm-level cooperative R&Dcommercialization strategies. While the volume of interfirm collaboration has increased dra-matically in recent decades, the determinants of firm-level choices among alternate modes ofsuch cooperative activity remain relatively understudied. We develop a conceptual model of fac-tors determining collaborative mode choice at the organizational portfolio level. These factorsinclude the firm-level appropriation environment, in which deal-level choices have portfolio-level spillover implications, as well as governance capabilities developed by the firm over time.Using a random sample of innovating biotechnology start-ups, we assemble a firm-year paneldataset that aggregates transaction-level collaboration data to the firm-year level, allowing usto characterize firms’ portfolios of collaborative deals. We find broad empirical support for ourmodel, suggesting that a firm’s appropriation environment and governance capabilities stronglyinfluence portfolio-level collaboration mode choices. In addition, we explore the implications ofgovernance capability development, finding that experience with particular modes, as well asdeviations from existing capabilities, impact firm valuation. Copyright 2008 John Wiley &Sons, Ltd.

INTRODUCTION

Collaborative research and development (R&D)commercialization activity has increased signifi-cantly in recent decades (Hagedoorn, 2002),prompting a flurry of research seeking to explainthe causes and consequences of this phenomenon.Increasing technological complexity and fasterproduct cycle times are among the factors that haveled to greater value chain specialization and madeinterfirm collaboration a critical component offirm strategy in many technology-based industries

Keywords: strategic alliances; technology commercializa-tion; entrepreneurship; biotechnology*Correspondence to: David H. Hsu, The Wharton School, Uni-versity of Pennsylvania, 2000 Steinberg-Dietrich Hall, Philadel-phia, PA 19104, U.S.A. E-mail: [email protected]

(Cukier, 2005). While the organizational perfor-mance benefits of such cooperation have been doc-umented in a wide range of settings (Shan, Walker,and Kogut, 1994; Das, Sen, and Sengupta, 1998;Dyer, 2000; Rothaermel, 2001; Gomes-Casseres,Jaffe, and Hagedoorn, 2006), cooperative strate-gies are particularly important for start-up firms,as the option to forward-integrate into downstreamproduction activities is often not feasible due toresource constraints. The division of labor betweenupstream innovation by start-ups and downstreamcommercialization by established firms with com-plementary assets allows for specialization in theeconomy and can enable firm-level productive effi-ciency gains (Gans, Hsu, and Stern, 2002).

Firms have a range of options in structuring col-laborative relationships, and a growing body ofwork has begun to address the issue of governance

Copyright 2008 John Wiley & Sons, Ltd.

836 V. A. Aggarwal and D. H. Hsu

choice, focusing on how and why organizationschoose particular forms of cooperative activity(e.g., product licensing vs. equity alliance). Ana-lyzing the organization of collaboration is impor-tant, as a given mode of cooperation may be moretargeted and appropriate than another mode giventhe context (Dyer, Kale, and Singh, 2004). Argu-ments based on both contractual and organizationalconsiderations have addressed the question of howto optimally structure these dyadic relationships,and an emerging stream of literature has begunto compare transaction cost and organizationalresource-based perspectives (Leiblein and Miller,2003; Sampson, 2004b; Villalonga and McGahan,2005; Mayer and Salomon, 2006) in explaining thegovernance structure of individual transactions.

Much of this prior work examining governancechoice, however, has focused on outcomes at thetransaction level, examining decisions made forindividual collaborative events without an explicitfocus on the firm’s overall portfolio of collabora-tive activities. In this study we suggest that tak-ing a portfolio-level approach helps fill an impor-tant gap in our understanding of the drivers offirms’ cooperative activities. Our primary theoret-ical goals are, thus, not only to conceptualize thedifferences among cooperative governance modes,but more importantly to develop a model of thedrivers of cooperative mode choice at the organi-zational portfolio level of analysis. In developingthis model, we follow in the spirit of an emergingbody of work that uses multiple theoretical lensesto address the governance choice issue. We departfrom this literature, however, by taking a portfolio-level perspective. In doing so, we outline in detailthe mechanisms that cause deal-level choices tohave portfolio-level implications.

Approaching R&D commercialization using anorganizational lens is particularly important forstart-up innovators, for whom important tradeoffsmust be made in allocating and deploying criti-cal resources against a portfolio of external col-laborative opportunities. Consider a firm such asQualcomm, the wireless communications semicon-ductor company, which has chosen to special-ize in technology licensing, electing to earn themajority of its returns from innovation througharm’s-length licensing of its technological devel-opments. Such an organization will tend to allocateits resources to develop a significant capability inproducing patentable inventions through its tech-nical staff, perhaps at the expense of developing

other organizational areas such as manufacturingor sales/marketing functions. Thus, developing adistinctive capability in one area is a strategicchoice with the associated trade-off of not being atthe leading edge of another organizational domain.

Start-ups’ decisions to pursue one collaborativeform over another can also have implications forindustry incumbents, as recipient firms of technol-ogy transfer will have to devote varying levelsof managerial attention, oversight, and resourcesin governing different types of interorganizationalrelationships. Consider Cisco Systems, the com-munications network equipment supplier, with areputation for acquiring and integrating innovativestart-ups. While acquisitions are an extreme formof innovator cooperation, the economic effect oftransferring innovation to the counterparty is alsoachieved in this form of collaborative commer-cialization. Organizations that frequently acquirestart-up innovators likely invest relatively moreresources in acquiring evaluation and integrationcapabilities (Dyer et al., 2004).

In this study, we take the perspective of thestart-up innovator, seeking to explain the determi-nants and consequences of their cooperative modechoice at the portfolio level. Focusing on start-upsis particularly beneficial for the purposes of thisstudy for two reasons. First, organizational capa-bility explanations are an important component ofour theoretical model (more on this below), and theability to track firms from the time of their incep-tion can thus mitigate the concern that unobservedroutines and/or prior experiences might be influ-encing our observed firm-level outcomes. Second,start-ups provide a setting where resource con-straints are especially binding, forcing tradeoffsacross different modes of organizing cooperationportfolios, making our research question and con-ceptual model particularly salient. It is important tonote that the governance of collaborative commer-cialization decisions is inherently bilateral (involv-ing both the innovator and its counterparty), and anequilibrium analysis based on observed outcomesthus reflects outcomes that take into account thesebilateral choices. As a result, the innovator’s over-all bargaining power, including the choice of coop-erative arrangement, depends on a host of factorsincluding technology uniqueness and the externalfinancing environment (Lerner and Merges, 1998),and we adopt an empirical design aimed at takinginto account such circumstances.

Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 30: 835–864 (2009)DOI: 10.1002/smj

Modes of Cooperative R&D Commercialization by Start-Ups 837

To address the primary research question ofwhat drives organizational portfolio-level varia-tion in start-ups’ collaborative commercializationmodes, we construct a conceptual model built onexplanations rooted in firms’ appropriation envi-ronment and organizational governance capabili-ties, an approach consistent with the multitheo-retical view toward mode choice taken in recentliterature (e.g., Villalonga and McGahan, 2005).1

As previously noted, we depart from prior workby explicitly discussing mode choice determinantsat the organizational portfolio level. Expropria-tion hazards, for example, have typically beendiscussed at the deal level; a theoretical contri-bution of this study is, thus, to explicitly detail themechanisms of action that lead to the firm-levelappropriation environment. In addition, we expli-cate mechanisms leading to organizational gover-nance capability development.

In our empirical analysis we focus on collabo-rative R&D mode choice drivers, and also addressthe interlinked question of the consequences ofpursuing collaborative commercialization modesthat differ from existing governance capabilities.We select a research setting where we can abstractaway from the self-commercialization option andfocus solely on choices among various modes ofcommercialization. Biotechnology is an industrywhere cooperation is an overwhelmingly commonstrategy for innovators, and where there is signif-icant variation among collaboration modes; thus,it is an ideal setting for our study.2 We assem-ble a dataset of biotechnology start-ups, which wetrack over time beginning with their year of found-ing, and find broad support for both appropriation

1 Our research is perhaps closest in spirit to this study, in whichthe authors conduct a comprehensive assessment of firm scopedecisions (and their drivers) varying from divestiture to acqui-sition, spanning intermediate alliance forms. While we do notexamine the range of firm boundary choices they do, we ana-lyze variation among different alliance forms (which Villalongaand McGahan [2005] do not treat) due to the organizational andstrategic rationale previously discussed. As they do, we con-dition our analysis on firms’ decisions to undertake a changein their firm boundary, thereby studying variation among thevarious organizational collaboration modes.2 Due to the combination of a strong appropriability environmentand high costs of assembling complementary assets, the biotech-nology industry exhibits a great deal of commercialization viastart-up cooperation with incumbent firms (Gans et al., 2002). Inaddition, while there is variation in the number of cooperativeactivities that comprise a focal start-up’s portfolio, innovatingfirms in this industry typically commercialize multiple technolo-gies, making a portfolio interpretation relevant, particularly aswe observe the evolution of the firm’s cooperation decisionsover time.

environment and especially governance capabilityexplanations for collaborative mode choice. Usingfirm valuation as the measure of the consequencesof deviating from existing governance capabilities,we find that a higher intensity of licensing deal userelative to firms’ historic use of such activities ispositively correlated with valuation, but that thiseffect is moderated by ‘hot’ capital markets.

In the following section, we discuss the litera-ture and hypotheses underlying our analysis, anddiscuss our conceptual model of cooperative modechoice drivers. We then turn to a discussion ofthe research design, followed by a description anddiscussion of our empirical results. A final sectionconcludes by discussing the managerial and theo-retical implications of this study.

LITERATURE AND HYPOTHESES

This section begins by outlining our rationale forequity use and arm’s-length licensing as focalmodes of cooperation. We then discuss our use ofa firm-level of analysis, contrasting our approachwith the prior literature. The second half of thesection derives empirical predictions of organi-zation-level collaborative mode choice and therelated firm valuation consequences.

Modes of organizing collaborativecommercialization

Collaborative commercialization modes can takemany forms, spanning the dimensions of contrac-tual formality, level of interorganizational inte-gration, and equity use (Mathews, 2006; Robin-son and Stuart, 2007). In our analysis, we focuson start-up firms’ aggregate use of two contrac-tual features. The first, equity use, is an indi-cator of organizational hierarchy (among otherthings),3 while the second, arm’s-length licensing,allows for the division of labor between innovatorsand commercializers (Arora, Fosfuri, and Gam-bardella, 2001; Gans et al., 2002). Organizationalspecialization in developing collaborative portfo-lios that use one mode more intensively than theother has implications for investment and resource

3 As Gulati and Singh (1998) argue, alliance coordination costs,which do not necessarily coincide with equity use, are also animportant factor in alliance governance design. Due to measure-ment issues in our empirical context, we abstract away from thisdistinction.



allocation within the firm (e.g., recruiting certaintypes of technical and legal personnel) and fororganizational design more broadly (e.g., settingincentive schemes). We elaborate on these twomodes of cooperative activity before turning tothe broader discussion of their organizational-levelimplications.

Equity use

The prior literature has highlighted a number ofsituations in which structuring transactions withan equity component may be beneficial. Gulati(1995) suggests that the use of equity alliancesrelative to non-equity alliances is driven by fac-tors including transaction cost and relational con-siderations. In general, equity is tied to greaterformal control in interorganizational governance,which may be particularly important in mitigatingthe threat of transactional hazards (Pisano, 1989;Oxley, 1997). In the context of strategic alliances,equity use has also been conceptualized as help-ful in capturing upside gain through option value(Santoro and McGill, 2005), monitoring and infor-mation acquisition (Filson and Morales, 2006), andin enabling learning effects (Anand and Khanna,2000).

In addition, equity use may facilitate strate-gic goals for start-ups. In their discussion of‘judo strategy,’ Yoffie and Kwak (2002 : 8) sug-gest that an essential component of successfulstrategy for smaller players is the ability to exertcontrol over the larger player. They note thatone way of doing this is to ‘give potential com-petitors a stake in your success through part-nerships, joint ventures, or equity deals’ (Yoffieand Kwak, 2002 : 11). Mathews (2006) formal-izes this logic, suggesting that because a coop-erative strategy requires negotiating details thatin itself may expose the start-up to an expro-priation threat by the incumbent, including anequity provision may help diffuse the competi-tive threat inherent in the collaborative processshould bargaining break down. More generally,entrepreneurial equity assignment is an impor-tant means by which corporate control rights(board seats) are allocated (Robinson and Stuart,2007). We therefore focus on equity use as animportant distinguishing feature of collaborativerelationships.

Arm’s-length licensing

In considering non-equity modes of organizing col-laborative relationships, there are many such alter-native forms, including supply arrangements andother non-R&D activities. We are interested inthose forms of collaboration where a focal firmis seeking to commercialize an innovation, andso contractual arrangements that lead to technol-ogy sharing or that enable the creation of marketsfor intellectual property are of particular interest.Prior work has discussed the use of arm’s-lengthlicensing relationships in this context, noting thatlicensing enables firms to utilize markets for tech-nological assets while avoiding more integratedand hierarchical contractual forms (Teece, 1986;Arora et al., 2001; Mathews, 2006).

Licensing activity can impact firms’ strategy byinfluencing organizational scope and value appro-priation decisions. Arora et al. (2001) suggestthat markets for technology, which are facilitatedby the presence of licensing contracts, can haveimportant implications for industry competition,writing that such markets ‘lower the barriers ofentry into the industry, increase competition, andcompress product life cycles’ (Arora et al., 2001:224). These industry changes further affect cor-porate strategies, particularly for start-ups. Theygo on to state that start-ups need not necessarilyinvest in downstream assets, as they ‘can profitfrom their research even if they lack the comple-mentary assets, or if the markets for such assets areunderdeveloped’ (Arora et al., 2001: 251)—whichhas the effect of enlarging the commercializationstrategy space.

Licensing has therefore been treated as a dis-tinct alliance form in much of the prior litera-ture. For example, Oxley (1997) uses a hierarchyof alliance types, with bilateral contracting repre-senting a licensing-type relationship, while Anandand Khanna (2000) explore the degree to whichlearning effects exist for licensees. While licens-ing relationships sometimes come bundled with avariety of other features such as co-developmentor co-marketing, in this study we focus on ‘pure’technology licensing in order to capture arm’s-length innovation transfer effects.

Levels of analysis

We turn now to our motivation for examiningmode choice outcomes at the firm-level. Early



work examining collaborative mode choice drewlargely from the transaction cost economics (TCE)paradigm. This theory lends itself naturally tounderstanding heterogeneous deal-level gover-nance by examining contractual hazards, expro-priation risk, and asset specificity (Williamson,1975; Klein, Crawford, and Alchian, 1978). TCEtheories have been applied to a broad range offirm boundary issues, including vertical integration(Monteverde and Teece, 1982), equity use in col-laborative R&D relationships (Pisano, 1989), andorganizational governance (Oxley, 1997).

While understanding individual deal-level out-comes is certainly important, organization-leveloutcomes are equally important, since managerslikely make decisions taking into account thestrategic implications across a portfolio of deals(Hoffmann, 2007). Recent work building on thecapabilities and competence views of the firm(Nelson and Winter, 1982; Wernerfelt, 1984; Bar-ney, 1991) suggests that firms often make cor-porate-level choices such as the development of adedicated alliance function (Kale, Dyer, and Singh,2002), which entail significant resource allocationtrade-offs. Consequently, individual deals cannotbe considered in isolation, but should instead beexamined within the larger context of the firm.In this spirit, Williamson (1999) examines theinterplay between the governance (TCE) and com-petence views of the firm, concluding that thereare important complementarities between the twoperspectives. In a related vein, Madhok (2002)discusses the implications of firm resources fortransaction-level governance.

Figure 1 provides a conceptual diagram of thedifferent levels of analysis at which we can exam-ine cooperative mode choice. On each side of

the diagram we illustrate a firm, represented asa bundle of individual deals. We can conceptual-ize the determinants of mode choice at the deal,portfolio, and firm levels. Mode choice outcomescan similarly be examined at these same threelevels. The level of analysis of the inputs andoutputs depends on the level of analysis of theassociated theory. For example, while much ofthe TCE literature (Pisano, 1989; Oxley 1997) hasexamined deal-level explanations and effects (i.e.,[A]→[X] in Figure 1), a conceptualization of firm-level resources determining deal-level governancechoice (Mayer and Salomon, 2006), would suggesta different model ([A + C] → [X] in Figure 1).4

Portfolio-level explanations and outcomes be-come especially relevant when there are interac-tions and spillovers among individual deals thatmake the portfolio more than a bundle of trans-actions. For example, Vassolo, Anand, and Folta(2004) discuss the role of portfolio effects inthe context of firms with multiple explorationalliances, showing that correlation among inputsassociated with individual projects can increaseoverall portfolio value, while correlation amongoutputs reduces portfolio value. Portfolio charac-teristics are thus potentially important as an input(category [B] in Figure 1), and can also potentiallyaffect choices made at the portfolio level (category[Y] in Figure 1).5

4 We distinguish between portfolio- and firm-level characteris-tics, with firm-level characteristics referring to attributes that arenot directly related to the firm’s collaborative portfolio, suchas firm age and size. Similarly, firm-level outcomes refer tocharacteristics such as firm valuation.5 The recent literature has begun to address some of the moregeneral implications of alliance portfolio configuration for firmstrategy and performance (Hoffmann, 2007; Lavie, 2007).

Dealcharacteristics

Portfoliocharacteristics

Firmcharacteristics

Deal-leveloutcomes

Portfolio-leveloutcomes

Firm-leveloutcomes

Explanatory factors Mode choice outcomes

(A)

(B)

(C)

(X)

(Y)

(Z)

Figure 1. Levels of analysis



Examining mode choice at the portfolio level isthus most informative when governance decisionsare made across a number of deals and when com-mon organizational inputs are necessary to struc-ture cooperative activity. The start-up innovatorcontext fits these specifications well, as organi-zational resource inputs are typically highly con-strained for these firms. Our focus in this study is,therefore, on the innovator’s portfolio-level use ofequity and licensing, with explanatory factors atthe portfolio and firm levels.6 In order to justifyany particular set of factors as driving portfolio-level outcomes, it is necessary to describe howinteractions among governance decisions cause aportfolio to be more than just the sum of individualdeals. We turn to this issue in the next section.

Drivers of mode choice

Our analysis considers two primary explanatorycategories for portfolio-level cooperative modeoutcomes: the firm-level appropriation environ-ment and governance capabilities. We first discussthe rationale for focusing on these two factors. Wethen explore each factor in turn, describing themechanisms that make a portfolio-level of anal-ysis appropriate before delving into the associatedpredictions.

As previously discussed, an influential streamof the alliance governance literature has emergedfrom the TCE paradigm. Applied to the generalissue of appropriating returns from innovation, themain TCE insight is that the organization of inno-vation (through internal production or via con-tractual relations with an external entity) cruciallydepends on the appropriation environment (Teece,1986; Gans and Stern, 2003). When transactioncosts are high, such as when innovation appro-priation is threatened through interorganizationalbargaining, firms are more likely to commercial-ize via internal production. Even when cooper-ative innovation commercialization via allianceactivity is pursued, these same issues can sur-face, as Gulati and Singh suggest: ‘appropriationconcerns. . . originate from the pervasive presenceof behavioral uncertainty, combined with the dif-ficulties of specifying intellectual property rights,

6 We focus primarily on portfolio-level outcomes ([B] + [C] →[Y] in Figure 1) while also examining implications for deal-level mode choice ([B] + [C] → [X]) and firm performance([B] + [C] → [Z]).

and by the challenges of contractual monitoringand enforcement’ (Gulati and Singh, 1998: 788).

The literature on interorganizational trust andreputation highlights a distinct mechanism bywhich firms may be able to achieve a strongerappropriation environment (Gulati, 1998 and ref-erences therein). This literature tends to empha-size the social mechanisms by which organiza-tional behavior is conditioned by the extent ofsocial relations, which can lead to interorganiza-tional trust and/or reputation. While pure trust-based mechanisms of appropriation do not high-light the threat of future punishment in the sameway that reputation-based ones do, the result ofboth is an improved means by which firms canreap the returns from their innovations. Appropri-ation via interorganizational trust and reputation onthe one hand, and minimizing transaction costs onthe other, can be two sides of the same coin. Poppoand Zenger (2002), for example, conclude that for-mal contracting and relational governance can bereinforcing complements as opposed to dividingsubstitutes. We thus focus on appropriability as akey concern in start-up alliance settings, particu-larly in the context of innovators facing expropri-ation hazards from collaborative activity.

Another stream of literature focuses on alliancegovernance stemming from the resource-basedview of the firm (Wernerfelt, 1984; Barney, 1991),in which the primary concern is developing andusing organizational capabilities. These studiesalso address the determinants of alliance gover-nance, but differ from the approaches discussedabove by stressing characteristics internal to thefirm as opposed to governing interorganizationalrelationships. The relational view (Dyer and Singh,1998) extends capabilities concepts to the dyadiclevel; however, the focus of this theory and relatedstudies such as Kale et al. (2002) and Zollo, Reuer,and Singh (2002) is still on the role of resourcesand capabilities in the formation, governance, andperformance of collaborative relationships.

We therefore focus on appropriation and capa-bilities considerations in our analysis of collab-orative mode choice. These two theories arisefrom the transaction cost and resource-based viewsof the firm, and represent central themes in thestrategy literature, combining economic reason-ing with organization theory (Williamson, 1999).While other theories (e.g., real options) offercomplementary perspectives on governance choice



Deal-levelfactors

Portfolio/firmcharacteristics

Expropriationthreats

Deal-levelgovernance

• Patents• Relational• Equity

Stocks of IP andrelational factors

• Patent portfolio• Aggregate partner-

specific experience

Spillover effects,Portfolio/firm levels

• Crossover patentprotection thickets

• Reputation effects and(implicit) threat ofwithholding future deals

(Initial)governance

choices

Deal-specificcapabilities

• Procedural knowledge• Human resources• Legal know-how

Organizationalgovernance capabilities

• Patterning of activities• Knowledge repositories• Routines/org. memory

Influences ongovernance choice

• Organizational inertia • Risk aversion• Rational choice• Project selection• Governance inseparability

Appropriation environment Governance capabilities

Mechanismsdriving mode

choice

Portfolio/firm-levelgovernance outcomes

Figure 2. Determinants of cooperative mode

(Villalonga and McGahan, 2005)7, the transac-tion cost and resource-based perspectives consti-tute two of the primary paradigms employed inprior work, with recent studies marked by anapproach examining these theories simultaneously(Colombo, 2003; Leiblein and Miller, 2003; Samp-son, 2004b; Mayer and Salomon, 2006).8 Whereasthis stream of research has generally examined

7 While these authors consider other theories in their analysis ofgovernance choice, they too highlight the dominant roles of theresource-based and transaction cost views (see Villalonga andMcGahan, 2005: 1185).8 These theories offer distinct mechanisms in the way they shapecooperative R&D mode choice, and offer non-transitive channelsof operation. Specifically, the appropriation mechanisms are con-ceptually distinct from governance capability, and possessing agovernance capability is more inclusive than the ability to appro-priate the returns from collaborative commercialization. Further-more, resource-based theories are distinguished from appropri-ation theories in that the former primarily stress characteristicsinternal to the firm (such as knowledge and capability in struc-turing and governing interorganizational collaboration), whileappropriation theories stress interorganizational and legal-basedmechanisms, respectively.

transaction-level outcomes, our analysis differsimportantly in its focus on firm-level mechanismsand implications. Figure 2 summarizes our argu-ments for firm-level appropriation and capabilitiesas drivers of mode choice. We turn next to expli-cating these two mechanisms.

Firm-level appropriation environment

Although transaction cost theory can explain therole of expropriation threats and the associatedmechanisms employed by a firm at the deal levelto appropriate returns from collaborative activity,examining firm-level governance choices requiresdeveloping a theory that can describe the condi-tions under which deal-level characteristics haveimplications for outcomes at a more aggregatelevel. We would like to understand the conditionsunder which a deal-level analysis fails to explainvariance in observed portfolio- and firm-level gov-ernance organization. Such a superadditive situ-ation might occur, for example, when there are



individual deal-level spillovers that have implica-tions for the governance of other deals in the firm’scollaborative portfolio, as per our prior level ofanalysis discussion, which builds on the Vasalloet al. (2004) logic.

Figure 2 outlines our theoretical framework,with the left side of the figure summarizing ourappropriation arguments. Expropriation threatscause firms to structure deal-level governance inways to mitigate such threats. These mechanismsinclude intellectual property protection, relationalcapital, and equity use (more on these below).These deal-level choices create a set of portfolio-and firm-level characteristics (the firm’s patentportfolio and aggregate partner-specific experi-ence), which then result in spillover effects thatimpact choices made at the portfolio level. Wediscuss two such spillover effects in this section:crossover patent protection and trust/reputation-based effects arising from repeated interactionswith existing partners.

Before describing the two firm-level spillovermechanisms in detail, we briefly explain their deal-level antecedents. Firms employ a variety of strate-gies to overcome deal-level expropriation threats.Intellectual property protection (via patents), forexample, is often cited as an important mecha-nism by which innovators can overcome threatsthat arise through collaborative activity (Teece,1986; Gans and Stern, 2003). A second strategyfor managing threats of expropriation at the deallevel is through repeated interactions among part-ners (Gulati, 1995; Baker, Gibbons, and Murphy,2002; Poppo and Zenger, 2002; Robinson andStuart, 2007), as this channel creates a ‘shadowof the future’ effect in which firms can threatento withhold future deals from contracting partiesdepending on their behavior in the focal transac-tion.9 Finally, a third strategy to assuage deal-levelexpropriation threats is through contracting, and inparticular through the use of equity-based relation-ships (Pisano, 1989; Oxley, 1997).

Two key firm-level attributes emerge as a resultof these strategies for managing appropriation con-cerns at the deal level: the firm’s patent portfolio

9 This strategy is particularly important when there is not a com-petitive supply of (future) complementary innovations accessibleto the party insourcing the technology, or when the innovatorholds assets outside the focal object of transfer that will addunique future value to the recipient (tacit knowledge or non-technical, specialized complementary assets).

and aggregate partner-specific experience.10 Whileeach of these characteristics represents the aggre-gate implications of deal-level governance choices,each construct also has unique implications at thefirm level because of its spillover effects onto otherdeals (see Nickerson, Hamilton, and Wada [2001]for a parallel argument on constellations of trans-actions). A firm’s patent portfolio can enable thefirm to mitigate appropriation concerns that ariseacross multiple deals. This occurs when patentsare not specific to an individual deal, but ratherapply across multiple technology commercializa-tion projects. Such an intellectual property port-folio can thus act as a ‘patent thicket’ (Shapiro,2000), making it more difficult for collaborativepartners to expropriate the innovating firm’s tech-nology.11 The degree of protection afforded bysuch a portfolio will, of course, necessarily bedependent on the degree to which patents are rel-evant across multiple commercialization projects.We return to this issue in our discussion of theempirical context.

The second spillover mechanism we proposerelates to a firm’s prior experience with collabora-tive partners, which can act to foster trust betweenthe parties and/or build interorganizational reputa-tion. A large body of prior literature has examinedthe reputations that firms can develop as a resultof their repeated interactions (e.g., Milgrom andRoberts, 1982; Kreps and Wilson, 1982). Suchtrust and/or reputation can be particularly impor-tant for start-up innovators, as their resources forpursuing legal recourse are likely limited (Park,Chen, and Gallagher, 2002; Huyghebaert and Vande Gucht, 2004). At the same time, potential col-laborative partners have an interest in retainingstart-up partners who have accumulated relationalcapital to mitigate search costs for future collabora-tion partners. Such future activity would hold value

10 We focus on these two constructs because they representcharacteristics that influence governance choice through appro-priation-based mechanisms. Aggregate equity use at the firmlevel is also a relevant characteristic, as we discuss in thesubsection entitled ‘Firm governance capabilities.’11 While the prior literature has discussed patent thickets inthe context of fragmented markets for intellectual property, weconceptualize an analogous situation with the innovator holdinga stock of interrelated patents that deters potential expropriationof a range of related products, a situation especially relevantfor start-up innovators, as most such firms confine their productdevelopment to a single or small number of related products forresource reasons.



for the collaborative partner that might be jeopar-dized should the partner expropriate the innova-tor’s technology. Thus, while prior partner expe-rience can serve as a governance mechanism forindividual transactions (Baker et al., 2002; Robin-son and Stuart, 2007), it can also have portfolio-level implications to the degree that it createsspillover effects onto other deals. These spillovereffects can thus serve to mitigate appropriationconcerns for both current and new partners, dueto the innovator’s reputation and the associatedpromise of future value.

These spillover effects thus shape the innova-tor’s appropriation environment at the firm level,with the ability to manage potential threats ofexpropriation through a patent portfolio and aggre-gate prior partner experience having implicationsfor how the innovator structures its broader portfo-lio of collaborative activity. Both spillover effectsare particularly important in a start-up setting,where resource constraints make it relatively costlyto invest in alternate means of mitigating appropri-ation. The start-up innovator’s ability to mitigateappropriation concerns by using one or both ofthese mechanisms would thus make it less likelyto structure its collaborative portfolio using morehierarchical governance structures such as equity.We thus propose:

Hypothesis 1a: Collaborative mode choice at theorganizational portfolio level will involve lowerlevels of equity use when the firm has a largerintellectual property portfolio.

Hypothesis 1b: Collaborative mode choice at theorganizational portfolio level will involve lowerlevels of equity use when the firm has higherlevels of repeated collaborative experience withexisting partners.

Firm governance capabilities

While the literature has begun to allude to thepossibility that firms’ ability to govern their col-laborative alliances might be a strategic organiza-tional capability, we aim to develop a more com-plete picture in this study. Villalonga and McGa-han introduce the term ‘governance specialization,’which they write ‘measures the degree to which afirm has repeatedly engaged in deals of the cur-rent type’ (Villalonga and McGahan, 2005: 1189).

However, they do not elaborate on the conceptexcept to show empirical evidence of its impor-tance in explaining firms’ choice among acquisi-tions, alliances, and divestitures when altering theirorganizational boundaries. Mayer and Salomon(2006) argue that a firm’s technological compe-tence may indirectly enhance its governance capa-bilities (through enhancing parties’ observability oftechnological milestones), which can in turn miti-gate contractual hazards. To better understand thestrategic fundamentals of governance capability,we address two questions in this section: what isan organizational governance capability, and howcan such capabilities persist in use over time?

The general requirements for a strategic capabil-ity have been well explored in the literature, withthe resource-based view conceptualizing a firm’scompetitive advantage as accruing from its uniqueset of valuable, rare, and inimitable resources andcapabilities (Barney, 1991). For governance struc-turing to be a firm capability, it must be difficultto replicate and not be an asset accessible on theopen marketplace.12 In conceptualizing governancecapabilities, we focus on firms’ prior governancestructuring experience (building on Villalonga andMcGahan [2005]),13 and suggest that a firm candevelop governance mode capabilities (e.g., equityalliances and arms-length licensing deals). Theability to structure and manage a particular typeof relationship can involve a combination of tacitand codified organizational knowledge, along withan associated set of organizational routines, pro-cedures, structures, and personnel. A governancecapability is therefore a firm’s aggregate collectionof knowledge, routines, and organizational struc-tures associated with a particular collaborativemode.

In many ways, governance capabilities representa ‘patterning of activities’ as suggested by Winter,where he defines an organizational capability as ‘ahigh-level routine (or collection of routines) that,together with its implementing input flows, confers

12 We neither make claims nor evaluate the optimality of gover-nance capability and use at this stage, as such a study requires aperformance evaluation of aligned and misaligned governancesituations relative to a benchmark (see Sampson, 2004a andMayer and Nickerson, 2005 for examples in other contexts).13 In complementary efforts, the prior literature has shown thatfirms can build capabilities through collaboration (Kale, Singh,and Perlmutter, 2000), and that prior general collaborative-,partner specific-, and technology specific-experience positivelyimpact the performance of their interorganizational relationships(Zollo et al., 2002).



upon an organization’s management a set of deci-sion options for producing significant outputs of aparticular type’ (Winter, 2003: 991, italics in orig-inal). We conceptualize a governance capability asa capability that enables a firm to structure col-laborative relationships in a particular way. Struc-turing licensing deals, for example, can involvea combination of legal and business developmenthuman resources, along with supporting procedu-ral knowledge that can be both tacit and codified.Over time, this set of (interacting) knowledge androutines can become increasingly embedded andintegrated within the firm’s larger set of activitiesand business model. The capability of structur-ing downstream collaborative activities, for exam-ple, can become more integrated with the firm’supstream innovation activities over time as incen-tives and organizational structures associated withinnovative activities become increasingly gearedtoward utilizing the firm’s capabilities in structur-ing particular cooperative forms. Our conceptual-ization of a governance capability thus relates toa firm’s activity system (Porter, 1996; Porter andSiggelkow, 2008), in which unique (and difficultto imitate) configurations of activities can often bea source of competitive advantage.

As particular governance activities and routinesare refined over time via managerial search pro-cesses (Nelson and Winter, 1982; March, 1991),and as knowledge of particular governance formsbecomes increasingly codified and embeddedwithin the organization, the firm’s stock of a par-ticular governance capability can increase. Firmswith greater experience structuring a particulartype of collaboration are therefore more likely tohave higher levels of capabilities associated withthe governance form, though the existence of acapability need not necessarily impel the firm tostructure current collaborative deals in a particu-lar way.

Two factors link prior collaborative governanceform experience with current choices. First,because selecting forms where a capability doesnot exist or is smaller in magnitude can entailbearing greater risk due to the lower level ofknowledge associated with such forms, we mightexpect firms’ capability stocks to influence currentchoice. Second, governance inseparability, wherecurrent choices are constrained (contractually orotherwise) by governance choices made in priorperiods (Argyres and Liebeskind, 1999), can alsoplay a role. We elaborate on these factors and on

the link between the stock of governance capa-bilities and current period firm-level governancechoice in the remainder of this section.

We begin with the concept of path dependence,which suggests that a firm’s set of prior expe-riences plays an important role in conditioningboth the set of choices available in the present,and even (in some accounts) what firms selectfrom their current menu of possibilities. Considera start-up’s initial choices regarding governancemode, which can be the result of anything fromserendipity to rational planning and investmentchoice. From these origins, through investmentsand learning processes, firms enhance their abilityto govern using particular styles. Heterogeneousgovernance capabilities can then result from orga-nizational learning processes (Levitt and March,1988; Walsh and Ungson, 1991), reinforcing theeffect of firms’ different levels of experience withvarious alliance governance modes.

Ongoing organizational and managerial invest-ments can affect a firm’s organizational memoryand impact the development of governance capa-bilities. For example, while the literature has pro-posed different loci of alliance knowledge repos-itories within the firm, such as dedicated alliancedepartments (Kale et al., 2002) and formal con-tracts (Mayer and Argyres, 2004), explicit invest-ments in building repositories of knowledge (andtheir ability to effectively retrieve and extend suchknowledge) are important determinants of orga-nizational learning. In addition, there may be ahost of other managerial investments necessaryto develop a competence in governing differ-ent modes of collaboration, such as identifying,recruiting, and retaining partners willing to engagein collaborations of a particular type, and structur-ing transactions appropriate for firm and partnercapabilities (Dyer et al., 2004).

We would expect a combination of risk aversion(to investing in new and unknown capabilities),organizational inertia, and rational choice to causefirms to be more likely to structure new collabora-tions in such a way as to be consistent with theirexisting stock of governance capabilities, particu-larly when those capabilities have been developedin prior transactions through forward-looking, cog-nitive processes (Gavetti and Levinthal, 2000) thattake into account the nature of future deals. Theseeffects are more likely to be salient for start-upfirms, where resources constraints are a primaryconcern (e.g., Park et al., 2002), making start-ups



less likely than established firms to take on thepotentially risky (and costly) proposition of invest-ing in new capabilities. A related effect involvesproject selection, wherein a firm would be morelikely to select projects based on its existing gover-nance capabilities. This effect, combined with thebehavioral implications of existing organizationalstructures and resources, suggests that governancecapabilities can influence a firm’s portfolio of col-laborative choices.

A second explanation for the persistence in useof governance capabilities is rooted in contrac-tual origins, as developed in Argyres and Liebe-skind’s (1999) article on governance insepara-bility. Using path dependence arguments, theseauthors argue that a firm’s ability to choose gov-ernance structures may be shaped, and indeed lim-ited, by governance use in prior time periods.Relatively long-lasting organizational contractualcommitments may perpetuate governance persis-tence. Particularly for start-ups, offering longerterm contracts and/or contracts with broader scopemay be necessary to induce would-be collabora-tors to make relatively specific investments, thusexposing themselves to opportunism hazards. Con-sequently, governance adjustments may be difficultto execute in the face of contractual commitments.An additional mechanism leading to governanceinseparability is the pressure to preserve contrac-tual uniformity across collaborative partners. Thismight be important, even if such contracts spantechnological or other domains, as parties to atransaction may care not only about the terms oftheir contract, but also how their terms compare toothers contracting with the same innovating start-up. Mayer and Bercovitz (2008) provide empiricalevidence for these effects in their sample of inter-firm contracts.

Our arguments in this section are summarized onthe right side of Figure 2. For both organizationalgovernance capability and governance inseparabil-ity reasons, we propose:

Hypothesis 2: Firms’ governance capabilitiesresult in path dependencies within their portfo-lios of cooperative governance mode use.

Valuation implications

We end this section by examining the valua-tion effects associated with firm investments ingovernance capability development. What are the

implications of pursuing paths that diverge fromexisting capabilities? In particular, as firms con-sider the development of new capabilities, is therea penalty associated with lower use of a given col-laborative governance mode capability?

Collaborative capability development might benecessary in the long run, but requires short-runinvestments in order to develop these competen-cies. We are interested in understanding the degreeto which the interplay between capability invest-ment by managers and recognition of this invest-ment by investors is reflected in firms’ valuations.An example of this is Abgenix, a biotechnol-ogy innovator that had a history of licensing its‘XenoMouse’ technology to pharmaceutical firmsin an arm’s-length ‘hand-off’ strategy. In order todevelop downstream commercialization capabili-ties, the firm needed to pursue a shift in its strat-egy to include co-development, which in the shortrun forced the firm to trade off a steady royaltystream from arm’s-length licensing to develop itscommercialization capabilities. The company’s topmanagers worried that this shift in business modelwould receive a negative shareholder reaction, andweighed that factor when considering the form ofcooperative activity decision (Dolan, 2001).

The prior literature has discussed the valua-tion impact of cooperative events generally (Chanet al., 1997; Das et al., 1998; Baum, Calabrese,and Silverman, 2000), suggesting that collabora-tive activity has a positive impact on firm per-formance, and that heterogeneity in cooperativemodes can be correlated with differential valua-tion implications. Additionally, Sampson (2004a)suggests that there are potentially important costsassociated with ‘misaligned’ governance decisionsthat can reduce the positive benefits of collabo-ration. However, the valuation impact of gover-nance capability development, and in particular theimpact of deviating from an existing stock of gov-ernance capabilities, has not, to our knowledge,been addressed.14

We would expect that the choice to pursuea particular mode entails a trade-off between

14 We focus our theoretical development on deviations in gov-ernance capability and not on changes in appropriability envi-ronment. Since we only develop predictions relating the appro-priability mechanisms to firms’ equity-use intensity (and nottheir licensing intensity—see our discussion in the ‘EmpiricalResults’ section), we decided to focus on governance capabili-ties, as our theoretical development suggests that such capabili-ties apply symmetrically across collaborative R&D modes.



two competing effects: first, the extant capabilitystock associated with a given mode influences thepropensity of the firm to use that mode; second,investments in new capabilities may be beneficialin the future, but costly today. Since investmentsare inherently risky (in that they might not payoff), we might moreover expect that the capitalmarkets may not immediately take into account thefull value associated with the choice to invest ina new capability. This may occur either becausemarkets are myopic, or because they hold a dif-ferent perception of the risk associated with firmchoices.15 Thus, we would expect firms that devi-ate from their existing capability stock in order topursue investments in new governance capabilitieswould be punished by the capital markets throughlower firm valuations.

Since a negative reaction of the capital marketsto a new capability investment is dependent uponthe markets perceiving a greater risk associatedwith this investment than the firm managers them-selves, we might further believe that in situationswhere markets are less risk averse, this differentialbetween the market’s risk perception and that ofmanagers will be reduced. We propose that sucha situation occurs in ‘hot’ markets, where marketsare generally less risk averse and more forgivingof managerial decisions in new capability invest-ments. Based on this discussion, we offer the fol-lowing hypotheses:

Hypothesis 3a: Selecting cooperative modes thatdeviate from existing governance capabilitieswill be associated with lower firm valuations.

Hypothesis 3b: The effect of governance capa-bility deviations on firm valuation will be lowerwhen external financial conditions are morefavorable to the firm’s industry.

15 At least two other mechanisms have been proposed in the lit-erature for altered firm valuations resulting from changing orga-nizational strategies. Stein (1989) argues that myopic managerialbehavior (in the context of corporate earnings manipulation)can be perpetuated, even when capital markets are operatingefficiently as a result of a prisoner’s dilemma effect. Anothermechanism involves managers and firms updating their beliefsabout their own skills and capabilities over time through sequen-tial options-based investments (Bernardo and Chowdhry, 2002).In this second mechanism, it is the volatility of capability uncer-tainty (rather than the resources themselves) that leads to differ-ences in valuation.

RESEARCH DESIGN

To address the correlates and consequences of het-erogeneous modes of collaborative R&D commer-cialization, we choose an empirical setting thatsatisfies two requirements: the empirical contextshould exhibit variation among cooperative formsof organization, and there should be nonexistent orinfrequent self-commercialization.16 The biotech-nology industry is close, if not ideal, in fulfillingthese criteria, as the economics of the industryare such that collaborative commercialization withlarger pharmaceutical firms is often the dominantchoice, yet there is considerable variation in theway in which such transactions are organized andgoverned.

We examine organizational governance capabil-ity and appropriation-based explanations for het-erogeneous forms of collaboration intensity andconduct analyses at both the firm-year level in apanel dataset and at the transaction level in anunderlying, associated dataset. While the formeranalysis accords with the study’s conceptual devel-opment on organizational portfolios of collabora-tive commercialization, the latter analysis allowsus to compare our results with the prior litera-ture that has sought to examine transaction-levelgovernance.

Throughout our firm-level analyses, we employstart-up firm fixed effects ordinary least squares(OLS) specifications. Our estimates are, there-fore, derived from within-firm, across-time vari-ation, which mitigates the risk that unobserveddifferences across firms in the sample are driv-ing the results. Using this method, unmeasuredtime invariant factors do not bias our estimates.The transaction-level analysis employs multino-mial logit regressions (with standard errors clus-tered by firm), thereby modeling alternativechoices of organizing a deal as a function offirm capability and transaction cost measures. Thissection discusses the data construction and pro-vides a description of the variables used in theempirical analysis.

Data sample

To construct a firm-year panel, we begin by ran-domly sampling a set of biotechnology firms. We

16 For an analysis of the determinants of commercialization strat-egy between self commercialization and cooperative develop-ment, see Gans et al. (2002).



use the Thomson One Banker database to create alist of all firms in Standard Industrial Classifica-tion (SIC) codes 2833–2836 located in the UnitedStates that have conducted an initial public offer-ing (IPO)17 in the past 25 years. From this list of468 firms, we randomly select 117 for our sample.By using the CorpTech and Hoover’s databases,combined with Lexis-Nexis news article searches,we determine the founding and dissolution datesfor each firm. Dissolution is defined as any situ-ation where the firm ceases to exist because it isacquired or because it goes bankrupt or dissolvesfor other reasons. We obtain yearly data for oursample of firms from the year of founding to theyear of dissolution (or 2004 if the firm is still inexistence).

The primary source for our cooperation vari-ables is the Recombinant Capital (ReCap) RDNAAlliances database.18 For each firm in ReCap, thereis a full listing of all cooperation events (alliancesand acquisitions) the firm has been involved withthroughout its history. For each cooperation event,we download and code a relevant set of informa-tion associated with that event. In particular, wenote whether the firm was a ‘client’ or ‘R&D’partner in the alliance, the name of the partner,the stage at signing, a subject description of thealliance, and a variety of contractual characteris-tics, including whether there is an equity or licens-ing component associated with alliance. Our sam-ple is composed of randomly selected biotechnol-ogy firms drawn from the set of R&D innovators,which allows us to make inferences about R&Dcommercialization from the innovator’s perspec-tive. The cooperation data are organized both at thetransaction level as well as aggregated to the firm-year level for the empirical analysis. We obtaincooperation data for 1,200 collaborative transac-tions across 91 firms, which we then aggregate tothe firm-year level.

17 While we select firms that have conducted an IPO in order toensure sufficient availability of firm-level data, in constructingthe firm-year panel we obtain data for all years, both public andprivate.18 Another possible source of alliance data, which has been usedin prior literature (e.g., Anand and Khanna, 2000) is SecurityData Company’s (SDC) Joint Ventures and Alliances database.We compared the list of transactions for our focal firms fromReCap with transactions from the SDC database, and foundthe ReCap is significantly more comprehensive in reportingtransactions and transaction details. ReCap is specific to thebiotechnology and pharmaceutical industries, and we attributeits increased comprehensiveness to this specialization.

Next, we code the particular therapeutic classesassociated with our sample of alliances and firmyears. For each alliance, we use the alliance sub-ject information to allocate the alliance to one ormore of 24 therapeutic classes. Virtually all of thealliances are coded as belonging to only a singleclass. We then analyze Securities and ExchangeCommission (SEC) documents (S-1s and 10-Ks) todetermine the therapeutic classes in which the firmitself is involved during each observation year.This is generally done by looking through the‘Company Overview’ or ‘Our Business’ sectionsof the documents, isolating keywords, and cross-referencing with drug pipeline descriptions if avail-able. The appendix provides a detailed discussionof our classification scheme.

Finally, we obtain patent, venture capital (VC),and firm-level data for our panel by utilizing avariety of sources. For the patent data, we usethe United States Patent and Trademark Office(USPTO) Web site to identify all patents grantedto our sample of firms. We then use the NationalBureau of Economic Research (NBER) patentdatabase (Hall, Jaffe, and Trajtenberg, 2001),which contains patent information through 1999,combined with a proprietary patent database atthe National University of Singapore that con-tains fields similar to the NBER data, but updatedthrough 2004, to generate a variety of patent-basedmeasures for our sample. Our primary source forthe VC data is VentureXpert. We retrieve informa-tion on all VC funding events such as valuation andfunding amounts for each of the firm years in oursample. Finally, we obtain firm-level informationsuch as firm size and market capitalization througha combination of CorpTech, Hoover’s, SEC docu-ments, and Compustat.

Variables

Addressing our research question calls for con-structing measures of outputs (collaborative orga-nizational form and firm valuation) as well asinputs (organizational governance capabilities,firms’ appropriation environment, and controls).We describe the variables used in our analysis inthis section. Definitions, means and standard devi-ations of these variables are provided in Table 1,and a correlation matrix of the independent vari-ables is provided in Table 2.



Table 1. Summary statistics and variable definitions∗

Variable Definition Mean Std. Dev.

Dependent variablesEquity-use intensity Percentage of all R&D deals for the firm in the

year where there is an equity component0.17 0.34

License-only intensity Percentage of all R&D deals for the firm inthat year where licensing is the only dealcomponent

0.21 0.35

Firm valuation Most recent VC valuation for private years;market capitalization for public years (in$M)

373.99 1219.76

Independent variablesGovernance capability measures1) Equity-use deal stock Total number of equity R&D deals by the focal

firm to the current year0.88 1.42

2) License-only deal stock Total number of pure licensing R&D deals bythe focal firm to the current year

1.25 2.75

3) Equity-use negative deviation Equity-use activity deviation in year t from theaverage equity-use activity normalized bytotal alliances

−0.02 0.29

4) License-only negative deviation Licensing-only activity deviation in year tfrom the average licensing-only activitynormalized by total alliances

0.07 0.49

Appropriation environment measures5) Patent stock Total stock of patents held by the firm up to

the current year9.73 25.17

6) Prior partner experience Number of alliance events by a focal firm inthe current year with firms partnered with inthe past

0.11 0.45

Firm-level controls7) Total deal experience Total number of alliances conducted by the

focal firm up to the current year8.24 16.24

8) Average deal scope Number of business categories spanned by theaverage deal in the year excludingequity-use and licensing-only

0.99 1.02

9) Firm age Age of firm in years (since founding) 8.24 6.3210) Number of employees Number of employees in the current year 552.42 1806.6311) Public firm dummy Dummy = 1 if the focal firm is publicly traded 0.61 0.4912) Discovery stage projects Number of alliances during a current year that

are at the discovery stage0.30 0.98

13) Therapeutic scope Number of distinct therapeutic segments thefirm is involved in during the current year

2.27 1.83

14) New therapeutic area dummy (t-2) Dummy = 1 if the firm entered a newtherapeutic area via an alliance two-yearsprior to time t

0.17 0.37

15) VC amount invested Total amount of VC equity invested in the firmby all VC firms up to the current year

11046.40 24847.06

16) Lerner index level Average level of the Lerner BiotechnologyIndex for the current year

4.81 2.51

∗ The natural logarithm of a variable, X, will be denoted L X.

Dependent variables

As previously discussed, we are primarily con-cerned with analyzing collaborative mode choiceintensities at the firm level since top managers are

concerned with both the organization of particulartransactions and the governance of their firm’scollaborative portfolio (we describe our trans-action-level analysis shortly). Our primary col-laborative mode dependent variables at the



Table 2. Pair-wise correlation matrix of independent variables

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16)

(1) 1.00(2) 0.12 1.00(3) −0.03 0.05 1.00(4) 0.10 0.14 −0.04 1.00(5) 0.23 0.19 0.10 −0.10 1.00(6) 0.10 0.37 −0.00 −0.02 0.10 1.00(7) 0.01 0.62 0.04 0.37 0.10 0.37 1.00(8) 0.14 −0.23 −0.25 0.21 −0.06 −0.17 −0.30 1.00(9) 0.27 0.30 0.04 −0.06 0.55 0.06 0.18 −0.16 1.00(10) −0.02 −0.01 −0.07 −0.12 0.10 −0.05 −0.02 −0.17 0.11 1.00(11) 0.06 0.03 −0.05 0.05 0.06 0.05 0.07 −0.19 0.12 0.03 1.00(12) −0.07 0.39 0.05 −0.03 −0.04 0.48 0.39 −0.21 −0.11 −0.01 −0.03 1.00(13) 0.09 0.28 0.10 0.08 0.27 0.25 0.44 −0.08 0.04 −0.14 0.03 0.30 1.00(14) −0.01 −0.10 −0.11 0.06 −0.15 −0.02 −0.03 0.13 −0.24 0.01 −0.09 0.01 −0.03 1.00(15) 0.02 0.00 0.06 −0.08 0.42 0.03 −0.09 0.07 −0.01 −0.10 0.02 0.02 0.18 0.01 1.00(16) 0.11 0.17 0.12 −0.13 0.20 0.03 0.19 −0.17 0.19 −0.16 0.02 0.20 0.29 −0.17 0.29 1.00

Note: Independent variable numbering corresponds to Table 1 numbering.

firm-year level of analysis are equity-use inten-sity and license-only intensity. The denominatorfor each of these variables is the total number ofalliance events in a given firm year where the focalfirm is the R&D partner. We focus on these alliancetypes at the portfolio level in order to capture twoends of the commercialization form spectrum, asother alliance types are intermediate in form (suchas co-development without equity, joint market-ing, and the like). The numerator of the equity-useintensity dependent variable is the total numberof alliances during the given firm year in whichequity is a component of the deal. The numeratorof the license-only intensity dependent variable isthe number of pure licensing alliances during thegiven firm year.19 These two variables operational-ize the concept of managers’ portfolio choice ofcollaboration governance in a fundamentally dif-ferent way than much of the existing literature onthe topic, which has sought to predict the gover-nance of a particular transaction.

In order to compare our analysis at the firmportfolio level to a transaction-level analysis, weemploy a dependent variable that accommodatesmultiple discrete outcomes in the context of multi-nomial logit regressions at the transaction levelof analysis. The dependent variable = 1 if a deal

19 These two dependent variables are not simply the mirror imageof each other, as the numerator of equity-use intensity excludesany non-equity form of R&D collaboration, while the numeratorof license-only intensity excludes deals with any componentother than a pure licensing agreement.

involves equity use and = 2 if a deal employsarm’s-length licensing only (and = 0 otherwise).

In addition to analyzing the incidence of equity-and license-based collaboration modes, we exam-ine their performance impact as measured by firmvaluation (as well as the associated impact offirms’ deviations from their historic experience instructuring collaborative transactions, their gov-ernance capabilities). We therefore construct firmvaluation as a dependent variable. For firms’ pri-vate years, we use the valuation at the most recentround of venture funding, while for public yearswe use market capitalization. The variable is meantto capture the overall value of the firm as assessedby its shareholders. We transform firm valuationby taking the natural log to address its skeweddistribution.

Focal independent variables

The independent variables of interest are proxiesfor two categories of factors: the firm’s gover-nance capabilities and the firm-level appropriationenvironment. The first set of variables measuresfirms’ experience in collaborating and structuringcollaborative deals. For each firm year, equity-use deal stock and license-only deal stock rep-resent the total number of equity or license-onlydeals the focal firm has been involved with (up tothe given year) since the firm’s inception. Suchexperience likely captures a set of governance



capability components ranging from recruiting col-laborative partners to negotiating and structuringagreements of a particular type with them. As aresult, such experience proxies for the interlockingactivity system of organizational knowledge, rou-tines, and design that collectively comprise gover-nance capability.20

To explore how reinforcing or deviating from afirm’s prior collaborative mode experience affectsits market valuation, we introduce two additionalvariables: equity-use negative deviation andlicense-only negative deviation. For a particularmode, the deviation variable is constructed bytaking the difference between the percentage ofall prior cooperation events utilizing a particu-lar mode (based on the stock of such experience),and the percentage of all current year coopera-tion events utilizing that mode. Positive values ofthese deviation variables suggest that historic useof a given governance mode exceeds current usage.Because the deviation terms are potentially nonlin-ear, we also include the square of these terms inour valuation regressions.

Another set of independent variables measuresthe firm’s appropriation environment. In line withour conceptual model, we construct the variable,patent stock, which counts the total stock of patentsheld by the firm up to a given firm year. We usethis measure as a proxy for the degree of firm-levelspillovers that exist as a result of a firm’s deal-level choices to manage the threat of expropriation.The conventional wisdom is that the appropriabil-ity regime for innovators via formal intellectualproperty rights is strongest in the health scienceand chemical sectors (Levin et al., 1987). Yet evenwithin the biotechnology sector, there has been asubstantial rise in patenting rates, even outstrip-ping the increase in overall patenting rates overthe past 20 years (Kortum and Lerner, 1999), sug-gesting that patent awards within relatively strongintellectual property regimes are meaningful. Animplicit assumption underlying our use of this con-struct is that the more patents held by the firm, themore likely the firm’s R&D deals are to includeits own patents. Particularly given the importance

20 Governance capabilities likely do not progress in a linearfashion, so learning is greater in the first phase of experiencerelative to later experience (i.e., there may be a nonlinear‘learning curve’). This learning process, together with the right-skewed nature of our data, guides our decision to specify atranslog function form (the dependent and independent variablesare specified in logs, with the exception of dummy variables).

of patenting in the biotechnology industry to theR&D context as discussed above, we feel that thisis a reasonable assumption.

Another spillover mechanism we discuss in ourconceptual model relates to prior collaborativepartner experience. Consistent with this mecha-nism, we include a second variable character-izing the appropriation environment at the firmlevel, prior partner experience. This variable isdefined as the number of alliance events by afocal firm in the current year with firms part-nered with in the past. Prior partner experienceproxies for the extent to which past bilateral col-laborations may lower expropriation hazards due toreputation and/or trust effects and the potential forfuture deals in enforcing discipline on opportunismthreats between parties in the current period.

Control variables

We include a number of control variables in theanalysis. The first, total deal experience, is the totalnumber of alliances conducted by the focal firmup to the current year, and is included to accountfor the possibility that collaborative mode choicemight depend on overall collaborative experience.The second, average deal scope, is the number ofbusiness categories spanned by an average deal in agiven firm year excluding equity-use and licensing-only deals (counting those categories would arti-ficially increase the average deal scope correla-tion with the main dependent variables, equity-useintensity and licensing-only intensity). The aver-age deal scope variable aims to capture more com-plex and/or ambitious collaborative transactions,which in turn shape cooperative structure and/orvaluation. Such deals may require more commit-ment by the commercialization partners.

The next control, therapeutic scope, is a countof the number of therapeutic areas the firm isinvolved in during the given firm year. Highervalues of therapeutic scope may proxy for ele-vated administrative structure necessary to gov-ern firms’ existing R&D and commercializationactivities. While there may be a minimum orga-nizational administrative structure regardless of afirm’s span of activities across therapeutic areas,bureaucratic structure may still increase with firms’therapeutic scope due to the need for program-specific governance. A learning interpretation oftherapeutic scope is also possible, as organiza-tional adjustment and learning associated with



firms’ changing therapeutic scope may also besubstantial.

Another variable, new therapeutic area dummy,indicates whether a firm undertook an alliance dur-ing a given year involving a therapeutic area inwhich it had not previously conducted an alliance(the therapeutic areas are based on our previouslydescribed 24-category classification scheme). Thisvariable is lagged two years, so that we can exam-ine the effect of entering a new area via an alliance,and aims to controls for the possibility that firmswill systematically choose one governance formover another when entering into a new therapeu-tic area. For example, with no established experi-ence in a given domain, firms may have to offeran equity-stake inducement in order to interest apotential collaboration partner.

The next control variable aims to capture firms’access to external resources via affiliation withventure capitalists. We construct the variable VCamount invested, which is the total amount ofequity invested in the firm up to a given firmyear. The VC literature has stressed the importanceof network access for start-up resource acquisi-tion (Hellmann and Puri, 2002; Hsu, 2004), andso higher levels of VC amount invested may becorrelated with firms’ external access to organiza-tional resources. The literature has also highlightedthe organizational and strategic influence of ven-ture capitalists (Gompers and Lerner, 1999), whichmay impact start-ups’ collaborative commercial-ization governance decisions.

A final set of variables control for firm age, stageof business development, and business environ-ment effects. These include: firm age, number ofemployees, public (a dummy for publicly tradedfirms), discovery stage projects (the number ofalliances in which the firm is in the ‘discovery’stage), and Lerner index level (an index track-ing the munificence of external funding from VCs,corporations, and the public markets available tobiotechnology firms).21 Due to the skewed distri-bution of the independent variables (except for theLerner index level), we add one to all continuous

21 The Lerner index is available through 1999 (see Lerner,1994 for a description of the index). For 2000–2005 we useestimates of the index that we construct by regressing theLerner index on the Nasdaq Biotechnology Index (NBI) for1993–1999 (the available overlapping years), then using theestimated coefficients of the regression, combined with actualvalues of the NBI, to generate the predicted values.

variables and take a natural log transformation ofthe underlying variable.

EMPIRICAL RESULTS

The empirical analyses are organized to examineorganizational governance capability and appro-priation-based explanations for observed variationin interorganizational collaborative mode, as wellas firm valuation implications of collaborative gov-ernance choice. We first examine correlates ofequity-use intensity and licensing-only intensity infirms’ portfolio of collaborative activities at thefirm-year level of analysis. We then compare theseresults to a transaction-level analysis in whichthe decision to organize a given deal is mod-eled as a choice among equity-use, licensing-only,and other forms of cooperative commercialization.Comparing the firm-portfolio and transaction-levelresults explicitly informs the potential importanceof firm-portfolio level spillover effects. Finally,returning to the firm level of analysis, we exploreshareholder reaction (via firm valuations) to cur-rent period deviations from historic capabilities oforganizing interorganizational collaborations. Eachanalysis is discussed in turn.

Drivers of R&D collaboration mode

We begin the empirical analysis by examining therelationship between measures of firms’ appropri-ation environment and the incidence of equity useamong firms’ collaborative alliances. The depen-dent variable we analyze in Table 3 is the fractionof firms’ deals in a given firm year involving anequity component via the variable equity-use inten-sity. Each of the specifications here and in all thefirm-level analyses throughout the article containfirm fixed effects, which mitigates the possibil-ity that unobserved firm differences are drivingthe results by basing the estimates on within-firm, across-time data variation. The predictions ofHypotheses 1a and 1b are that larger patent port-folios and repeated collaborative experience areeach negatively related to equity use. While bothpatent portfolios and repeated collaborative experi-ence are measures of firms’ appropriation environ-ment, they are distinct mechanisms with differentspillover processes (as detailed previously).

The first specification of Table 3 includes botha full set of control variables and measures of the



Table 3. Equity-use intensity fixed effects OLS regres-sions (firm-year level analysis)

Dependent variable =equity-use intensity

(3-1) (3-2) (3-3)

Independent variablesL equity-use 0.390∗∗∗ 0.400∗∗∗

deal stock (0.086) (0.086)L license-only −0.021 0.064

deal stock (0.086) (0.089)L patent stock −0.090∗∗ −0.112∗∗∗

(0.041) (0.041)L prior partner 0.082∗ 0.046

experience (0.049) (0.047)

Control variablesL total deal −0.110 −0.148∗ −0.146∗

experience (0.075) (0.080) (0.079)L average deal 0.184∗∗∗ 0.152∗∗∗ 0.174∗∗∗

scope (0.044) (0.042) (0.042)L therapeutic −0.073 −0.172∗∗∗ −0.123∗

scope (0.066) (0.063) (0.064)New 0.102∗∗ 0.065 0.063

therapeutic (0.047) (0.046) (0.045)area dummy(t-2)

L VC amount 0.024 0.017 0.017invested (0.016) (0.016) (0.015)

L firm age 0.344∗∗ 0.019 0.215(0.139) (0.118) (0.134)

L number of 0.028 −0.026 −0.012employees (0.036) (0.037) (0.036)

Public firm 0.145 −0.020 −0.072dummy (0.233) (0.227) (0.223)

L discovery −0.012 −0.002 −0.014stage (0.038) (0.036) (0.036)projects

Lerner index 0.003 −0.001 −0.001level (0.008) (0.008) (0.008)

Constant −0.681∗∗ 0.283 −0.114(0.313) (0.293) (0.323)

R-squared 0.45 0.50 0.53Number of

observations203 203 203

∗ , ∗∗ , or ∗∗∗ indicate statistical significance at the 10%, 5%, or1% level, respectively.

two appropriation environment variables, patentstock and prior partner experience. There are threesignificant control variables, average deal scope,new therapeutic area dummy, and firm age. Whilethe last two of these variables are positive and sig-nificantly related to equity-use intensity, they areno longer significant once governance capabilityis introduced into the specifications. Average deal

scope is positive and significant across the spec-ifications, suggesting that firms are more likelyto organize their portfolio of collaborative R&Defforts using an equity component the broaderand/or more complex their average deals are ina given firm year, as more complex collabora-tive transactions may involve more intensive (andcostly) governance arrangements. A higher aver-age deal scope could also proxy for more valuabledeals, in which case there is a larger threat of valueexpropriation.22

The main variables of interest are patent stockand prior partner experience. Patent stock is esti-mated with a negative and significant coefficient(at the 1% level), suggesting that crossover patentprotection serves as a substitute for organizationalequity-use intensity. The prior partner experiencevariable, however, is not significant. Equity rela-tionships may be more intricate and complex rel-ative to other modes on average (in ways whichaverage deal scope does not control for), whichcan be enabled by having prior relationships withgiven counterparties. This effect works in theopposite direction of the main effect we predict inthe theory development, and provides a possibleexplanation for why we find a net zero statisti-cal effect between prior partner experience andequity-use intensity in Table 3.

We examine the incidence of firms’ share ofdeals involving arm’s-length contracting (licen-sing-only intensity) in Table 4, as this representsanother well-defined cooperative R&D commer-cialization mode. As previously discussed, arm’s-length technology licensing is an independentlyinteresting organizational governance mode sinceit entails the choice of foregoing collaborative co-development and equity use. We again take anorganizational portfolio of collaborative mode per-spective using firm fixed effects to examine thecorrelates of firms’ licensing-only intensity.

We do not predict parallel appropriation envi-ronment effects in the other major governancemode we examine, license-only intensity, for thefollowing reasons. First, Gans, Hsu, and Stern(2008) find that there can be both pre-patentallowance and post-patent allowance technologylicensing between start-up innovators and moreestablished players in equilibrium. This results

22 While we are not able to test the effect of technological scopeand overlap between collaboration partners using patent charac-teristic data (as in Sampson, 2004a), it would be interesting todo so in future work.



Table 4. License-only intensity firm fixed effects OLSregressions (firm-year level analysis)

Dependent variable= license-only intensity

(4-1) (4-2)

Independent variablesL equity-use deal −0.076

stock (0.096)L license-only deal 0.616∗∗∗

stock (0.098)

Control variablesL patent stock 0.034 −0.050

(0.049) (0.046)L prior partner 0.092 0.105∗∗

experience (0.058) (0.052)L total deal −0.112 −0.354∗∗∗

experience (0.089) (0.088)L therapeutic 0.021 0.096

scope (0.079) (0.071)New therapeutic −0.041 −0.010

area dummy (0.056) (0.050)(t-2)

L VC amount 0.022 −0.014invested (0.019) (0.017)

L firm age 0.034 0.149(0.166) (0.150)

L number of −0.060 0.031employees (0.043) (0.041)

Public firm dummy −0.503∗ −0.339(0.275) (0.249)

L discovery stage −0.012 0.030projects (0.045) (0.041)

Lerner index level 0.019∗∗ 0.011(0.009) (0.008)

Constant 0.828∗∗ 0.330(0.371) (0.359)

R-squared 0.34 0.50Number of

observations203 203

∗ , ∗∗ , or ∗∗∗ indicate statistical significance at the 10%, 5%, or1% level, respectively.

from innovation- and strategic environment-levelheterogeneity in the benefits and costs to early(pre-patent allowance) licensing, though the for-mal allowance of patents is important in raising thehazard rate of technology licensing. Consequently,there may not be a systematic relationship betweenlicensing intensity and firms’ patent stock.

The other mechanism of appropriation, priorpartner experience, invokes trust and/or reputation-based means of safeguarding expropriation risksassociated with collaborative R&D contracting. Tothe extent that pure trust and or pure reputation

mechanisms are operating, the need for formallicensing contracts is unnecessary either becausethe parties have trust in each other (Macaulay,1963) or because the ‘shadow of the future’effects are so powerful so as to make contract-ing less useful (Baker et al., 2002). This is thetraditional logic for why there might be a neg-ative relationship between licensing use intensityand proxies for trust/reputation such as prior part-ner experience. A more recent literature, how-ever, argues and provides evidence for the oppo-site—the view that prior relations is associatedwith creating more detailed and customized con-tracts with more enforcement mechanisms (Poppoand Zenger, 2002; Mayer and Argyres, 2004; Ryalland Sampson, 2009), suggesting that contractuallicensing and prior relationships are complementsrather than substitutes (the positive coefficient onprior partner experience in Table 4 is consistentwith this view, though a credible case for thesubstitutability between licensing and prior rela-tionships prevented us from making an ex anteprediction about the overall relationship). In sum-mary, because there was no clear ex ante predictionof the empirical relationship between patent andrelational appropriation mechanisms on the onehand and firm license-only intensity on the other,we treat the appropriation variables as controls inTable 4.23 We next turn our attention to examiningour second hypothesis, that governance capabilityshapes the current choice of collaborative modeorganization. We do this by examining the rela-tionship between equity-use deal stock and equity-use intensity, and between license-only deal stockand license-only intensity. The second column ofTable 3 shows that equity-use stock is positivelyrelated to equity-use intensity (at the 1% statisti-cal significance level), a result that persists afteralso including the appropriation environment vari-ables (Table 3, Column 3). The second columnof Table 4 shows that license-only deal stock ispositively related to license-only intensity (also atthe 1% level of statistical significance). Together,these results confirm the second hypothesis, andare consistent with the importance of governance

23 Throughout the analysis in Table 4, we exclude the variableaverage deal scope because, by definition, the dependent vari-able license-only intensity is based on deals with scope equalto one.



capabilities in structuring and influencing the orga-nization of a firm’s current portfolio of collabora-tive transactions.24

Transaction-level analysis of collaborative mode

The analysis thus far has examined firms’ relativeuse of equity and arm’s-length licensing in orga-nizing their collaborative relationships at the firmlevel of analysis. We now present a parallel anal-ysis at the transaction level of analysis, where anobservation is a deal. By using multinomial logitregressions, we examine three alternative modes oforganizing a collaborative transaction: equity-use,licensing-only, and other collaborative modes (thelast of which represents the baseline category inthe multinomial logit). The results, in which thereported standard errors are clustered by firm inthe analysis, are presented in Table 5.

In a first specification, we examine governancecapability measures together with the set of firm-level controls and deal-level controls. While theformer controls are the same as those used in theprior tables, the latter variables include a set of fourdeal stage dummies, and a set of five therapeuticarea dummies. All estimates in this table shouldbe interpreted relative to the baseline collaborativemode, non-equity co-development. For example,the negative and significant coefficient on Lernerindex level associated with equity-use (but notlicense-only) suggests that relative to non-equityco-development, equity is used more sparingly inless munificent external funding conditions. Themain result of the first specification is that equity-use deal stock is positively correlated with transac-tion equity-use (but not with license-only use), andlicense-only deal stock is positive and significantlycorrelated with deal-level license-only use (but not

24 An anonymous referee suggested that we conceptualize gov-ernance capabilities as a depreciating stock rather than sim-ply accumulated experience. We therefore constructed depreci-ated knowledge stocks consistent with the literature on orga-nizational ‘forgetting’ (Nelson and Winter, 1982). FollowingArgote, Beckman, and Epple (1990), among other studies, weconstructed exponential depreciation parameters in computingknowledge/capability stocks to capture the conceptual possi-bility of organizational forgetting. Since there is little strongtheory to motivate the rate of forgetting, we experimented withdepreciation rates ranging from 0–50 percent. We found thathigher discount rates are associated with larger economic esti-mates across the spectrum of discount rates we tested (and are ofsimilarly strong statistical significance). We chose to report theconservative estimates associated with an undepreciated knowl-edge stock in the empirical tables.

equity-use —and indeed the correlation is negativeand significant) after controlling for various firmand deal variables.

The second specification retains the same set ofcontrol variables, but instead focuses on appro-priation variables.25 The only notable result isthat patent stock is weakly correlated with equity-use (though even this result is no longer statis-tically significant in the fully specified model).The final specification of Table 5 puts all the vari-ables together, and the picture that emerges rein-forces the importance of organizational governancecapability in explaining the variation in organiz-ing transaction-level collaborative relationships. A‘non-result’ of an appropriation variable relativeto the firm portfolio analysis in the final speci-fication of the multinomial logit is important. Inthe transaction-level analysis, there is no signifi-cant relationship between patent stock and equity-use, which contrasts with the negative relationshipbetween these variables in the firm portfolio-levelanalysis. This disparity, we believe, is due to thelack of organization-level cross-patent spilloversassociated with the transaction-level analysis.26

In unreported regressions (available on request),we checked result robustness using: (a) probits ofgoverning a transaction via equity-use and license-only (separately), (b) negative binomial and zero-inflated negative binomial specifications predictingthe number of equity-use and license-only deals(separately) by firms treating the data as a pooledcross-section, and (c) ordered probits of increas-ing hierarchical structure.27 In all these cases, the

25 Note that our transaction cost variables are constructed at thefirm level. We experimented with a deal breadth measure that isanalogous to the average deal scope variable. We would haveto modify this variable for the transaction level multinomiallogits because one of the outcomes is license-only, which bydefinition has a deal scope of one. To preserve consistency in theempirical tables, we therefore used only firm-level transactioncost variables.26 Similarly, while we do not hypothesize a relationship betweenprior partner experience and license-only, the positive relation-ship between these two variables present in the firm portfolioanalysis disappears in the deal-level analysis, probably due tothe lack of organization spillover effects.27 We thank an anonymous referee for suggesting that we imple-ment an ordered conceptualization of governance hierarchywhich Oxley (1997) and others have presented in the prior litera-ture. The dependent variable in the analysis is a measure of morehierarchical cooperative mode, which takes the value of one forpure arm’s length licensing, two for other non-equity licens-ing, and three for equity use. The baseline category is neitherlicensing nor equity use. The results of ordered probit modelswith robust standard errors (clustered by firm) are available onrequest, and show results consistent with those we have reported



Table 5. Cooperative mode multinomial logits (transaction-level analysis)

Dependent variable = Equity-use or License-only[default = other], robust std errors reported (clustered by firm)

(5-1) (5-2) (5-3)

Equity-use License-only Equity-use License-only Equity-use License-only

Independent variablesL equity-use deal stock 2.977∗∗∗ −0.257∗ 2.889∗∗∗ −0.268∗∗

(0.611) (0.148) (0.580) (0.133)L license-only deal stock −0.066 1.744∗∗∗ −0.060 1.762∗∗∗

(0.304) (0.190) (0.326) (0.184)L patent stock 0.475∗∗ 0.203 0.248 0.251∗∗

(0.187) (0.130) (0.186) (0.101)L prior partner experience 0.092 0.374 0.086 0.136

(0.449) (0.237) (0.522) (0.161)

Control variablesL total deal experience −1.252∗∗∗ −1.409∗∗∗ −0.961∗∗∗ −0.250 −1.336∗∗∗ −1.469∗∗∗

(0.249) (0.154) (0.338) (0.201) (0.264) (0.163)L therapeutic scope 0.010 0.513∗∗ 0.294 0.578∗∗∗ −0.065 0.435∗∗

(0.478) (0.223) (0.639) (0.218) (0.416) (0.178)New therapeutic area dummy (t-2) −0.495∗ −0.055 −0.519∗ −0.142 −0.510∗ −0.064

(0.288) (0.262) (0.308) (0.245) (0.288) (0.263)L VC amount invested −0.042 −0.037 0.017 0.015 −0.060 −0.073∗∗∗

(0.047) (0.024) (0.058) (0.036) (0.045) (0.025)L firm age −0.798∗∗ −0.108 −0.529 0.237 −1.105∗∗ −0.545∗∗

(0.361) (0.253) (0.415) (0.347) (0.441) (0.244)L number of employees 0.045 0.128 0.025 −0.153 0.013 0.106

(0.155) (0.080) (0.106) (0.173) (0.168) (0.090)Lerner index level −0.188∗∗ 0.022 −0.182∗∗ −0.007 −0.205∗∗∗ 0.010

(0.077) (0.033) (0.085) (0.055) (0.074) (0.033)Deal stage dummies Yes (4) Yes (4) Yes (4) Yes (4) Yes (4) Yes (4)Therapeutic area dummies Yes (5) Yes (5) Yes (5) Yes (5) Yes (5) Yes (5)Constant 0.843 −0.513 1.149 −1.978∗∗ 1.774 0.548

(1.005) (0.632) (0.972) (0.858) (1.087) (0.599)Log likelihood/num obs LL = −435.76/N = 866 LL = −491.67/N = 866 LL = −433.67/N = 866

∗ , ∗∗ , or ∗∗∗ indicates statistical significance at the 10%, 5%, and 1% levels, respectively

results are qualitatively the same as those reportedin the multinomial logits.

Firm valuation and cooperative mode

We conclude the empirical analysis by examiningvaluation correlates of both organizational gover-nance capabilities and firm deviations from theirhistoric capabilities. We do this by returning to thefirm level of analysis in Table 6. The valuationmeasure spans both private valuations (assignedby VCs) and valuations when the firm is public,as measured by market capitalization (the prod-uct of publicly traded share price and number of

in the main text—namely the important role of governance modecapability.

shares outstanding). While we pool the privateand public valuation observations in the reportedregressions, the results are broadly consistent withdisaggregated results separating private and publicvaluations.

We first examine the impact of equity-use dealstock and license-only deal stock, together withnegative deviations from historic average levels ofequity and arm’s-length license use in the first col-umn. Equity-use deal stock and license-only dealstock are each positively associated with firm val-uation, suggesting that collaborative deal volumeis valuable.

The two key independent variables are license-only negative deviation and equity-use negativedeviation. Each variable is the normalized devia-



tion of current period governance mode use rel-ative to the firm’s historic levels. Higher valuesof these variables suggest lower current use of thegovernance mode relative to past usage. The neg-ative coefficient on license-only negative deviationin regression (6-1) suggests that controlling forexisting levels of licensing-only governance capa-bility (as measured by license-only deal stock ), lessuse of that capability on average correlates with alower firm valuation.28 The included squared termof license-only negative deviation is also negative,implying that the license-only negative deviationeffect is linear (reinforcing the main effect). Theanalogous equity-use deviation variables are notsignificant.

One important issue in interpreting these resultsis that observed deviations of cooperative gov-ernance mode capability can arise for a multi-tude of reasons, including reduced (actual or per-ceived) efficacy of a given governance mode.Another reason might be a result of proactivemanagers at some firms, with corporate planningdepartments that are forward looking in varyingdegrees. Unfortunately we are not able to directlyobserve or measure the reason for an observedshift in cooperative governance. We therefore relyon a control variable, VC amount invested, toproxy for governance efficacy. The venture cap-ital literature (Gompers and Lerner, 1999) pro-vides detail on the mechanisms by which VCsmay be involved in new venture governance andstrategy. VCs will invest less in ventures withpoor or misaligned governance as compared tothe strategic environment, and more when theopposite holds (in addition, a higher VC amountinvested also likely comes with corporate controland governance implications—see, for example,Hsu, 2004—and so degree of VC involvement canhave direct implications for collaborative commer-cialization governance mode decisions). The VCvariable, together with the full slate of control vari-ables, is included in specification (6-2) of Table 6.The main result on license-only negative deviationis preserved.

28 We also experimented with breaking up these deviation vari-ables into the positive and negative directions. Doing so yieldsthe same qualitative results, but we preferred using our contin-uous measure, as these measures can take on negative values,which would mean that the focal firm is using more of the givenmode relative to their own past use. The directionality of devia-tion is therefore preserved in the continuous measures, with theadditional benefit that the overall fit of the regressions modelswas better using the continuous measures.

The final specification of Table 6 adds in theLerner index level, along with interaction termsof this measure with the license and equity-usenegative deviation measures. A positive value ofthe Lerner index level suggests, as expected, thatmore munificent funding environments are associ-ated with higher valuations. The Lerner index levelinteraction term with licensing-only negative devi-ation suggests that in ‘hot’ biotechnology finan-cial markets, the negative impact of conductingfewer licensing deals in the present as comparedto historical norms is attenuated, consistent withHypothesis 3b.

Taken as a whole, because we cannot directlymeasure the underlying reason for shifting use ofgovernance modes, we rely on an indirect con-trol for governance efficacy (via venture capitalinvolvement, as described above). Moreover, whilewe are not able to conclude a strict causal rela-tionship from this evidence, we are not aware oftheory that would suggest reverse causality, andthe results are robust to an analysis of a one-and two-year lag structure of the key indepen-dent variables (available from the authors). There-fore, controlling for existing levels of governancecapability, our results are consistent with the pre-diction that firms attain lower valuations, ceterisparibus, if they do not exercise their license-only governance capability, though this effectis moderated by ‘hot’ capital markets.29 We donot find corresponding evidence for equity-usecapability.

DISCUSSION AND CONCLUSIONS

This study examined the determinants of start-upfirms’ strategic choices in organizing their inter-firm collaborative R&D relationships. We focused

29 While we can observe the valuation impact of divergence incollaborative mode choice, we cannot disentangle the impact ofmanagerial choice from market recognition of that choice. A neg-ative valuation associated with capability divergence can resultbecause managers have made a choice that is not fully rationaland markets (efficiently) recognize this choice, or from managersmaking a fully rational choice and markets (inefficiently) failingto recognize this. Confirmation of Hypothesis 3a, however, doesallow us to reject the hypothesis that managers make optimaldecisions and at the same time these decisions are efficientlyreflected in market valuations. As well, regarding the possibilityof reverse causality (changes in governance form in response tochanges in firm valuation), our panel data techniques in whichannual data are most prevalent (for firms’ public valuations)mitigates this concern.



Table 6. Firm valuation firm fixed effects OLS regressions (firm-year level analysis)

Dependent variable = L firm valuation

(6-1) (6-2) (6-3)

Independent variablesL equity-use deal stock 1.783∗∗∗ 0.670∗ 0.521∗

(0.180) (0.349) (0.307)L license-only deal stock 0.424∗∗∗ 0.818∗∗ 0.648∗∗

(0.160) (0.362) (0.317)L license-only negative deviation −0.465∗∗∗ −0.393∗ −0.777∗∗

(0.148) (0.203) (0.320)L license-only negative deviation squared −0.215∗∗∗ −0.358∗∗∗ −0.302∗∗∗

(0.075) (0.113) (0.098)L equity-use negative deviation 0.258 −0.403 −0.644

(0.218) (0.311) (0.512)L equity-use negative deviation squared 0.188 −0.151 −0.099

(0.136) (0.218) (0.207)Lerner index level ∗ 0.077∗∗

license-only negative deviation (0.038)Lerner index level ∗ 0.056equity-use negative deviation (0.063)

Control variablesL patent stock −0.062 −0.031

(0.168) (0.148)L prior partner experience −0.237 −0.036

(0.180) (0.160)L total deal experience 0.184 0.200

(0.349) (0.303)L average deal scope −0.232 −0.136

(0.211) (0.186)L therapeutic scope 0.279 −0.071

(0.248) (0.223)New therapeutic area dummy (t-2) 0.116 0.195

(0.176) (0.154)L VC amount invested −0.135∗∗ −0.072

(0.060) (0.052)L firm age −0.347 −0.801∗

(0.532) (0.473)L number of employees 0.574∗∗∗ 0.477∗∗∗

(0.189) (0.168)L discovery stage projects 0.348∗∗ 0.241∗

(0.143) (0.126)Lerner index level 0.176∗∗∗

(0.027)Constant 3.021∗∗∗ 2.130∗ 2.616∗∗∗

(0.184) (1.147) (0.997)R-squared 0.72 0.83 0.88Number of observations 407 193 193

∗ , ∗∗ , or ∗∗∗ indicate statistical significance at the 10%, 5%, or 1% level, respectively.

our analysis at the organizational level,highlighting the role of the firm’s appropriationenvironment and governance capabilities inexplaining variation in collaborative mode choice.By jointly examining appropriation and capabil-ities considerations, we follow in the spirit of anemerging body of literature that has recognized the

importance of taking a multitheoretical approachto understanding the governance of interorgani-zational collaboration. At the same time, how-ever, we depart from the transaction-level focusof this prior work by developing a theoreticalframework that underscores the importance of fac-tors that transcend the individual deal level. Our



framework conceptualizes a set of appropriationand capabilities-based mechanisms that transformportfolio- and firm-level characteristics (resultingfrom deal-level choices) into portfolio-level gov-ernance outcomes. We test our conceptual modelusing a sample of biotechnology start-ups, findingbroad empirical support for our predictions.

The empirical results confirm that governancecapabilities have a strong influence on firms’ coop-erative mode choice, supporting resource-basedtheories of capability development that draw onideas of path dependence and organizational learn-ing through prior experience. The results also pointto market valuation effects associated with thedevelopment of these capabilities. For licensingcapabilities specifically, deviations from prior lev-els of capability use have a negative effect onvaluation, indicating that investments in new gov-ernance capabilities require firms to trade off thepotential benefits associated with accessing newmodes with the learning arising from prior col-laborative experiences. In addition to the gover-nance capabilities results, we find that the firm’sappropriation environment plays an important rolein shaping portfolio-level mode choice outcomes.Our results show that firms’ patent portfolios andaggregate partner-specific experience significantlyinfluence equity mode choice at the portfolio level,outcomes consistent with our conceptual model inwhich spillover effects arising from aggregate lev-els of deal-level governance choices serve as animportant mechanism driving mode choice.

Before discussing the broader implications ofour results, we mention two sample selectionissues that may affect the interpretation of ourresults. The first relates to selecting start-up firmsthat (eventually) choose downstream cooperationas a means of technology commercialization. Onthis issue, we purposefully selected our empiricalsetting to examine variation among cooperativemodes rather than variation between cooperativeand noncooperative commercialization strategies(Gans et al., 2002 study the latter issue). While weunderstand that self-commercialization does takeplace even within the biotechnology industry, webelieve that this industry setting represents thedimension of empirical variation we desire (amongcooperative modes) while allowing us to abstractaway from the self-commercialization strategy.Consequently, while the rate of collaborative com-mercialization between biotechnology and pharma-ceutical firms is high, there is substantial variation

in the governance modes employed. Due to ourselection based on cooperative events, however,the results should appropriately be interpreted asconditional on cooperation.

A second selection issue concerns conditioningthe sample on firms that (ex post) conducted anIPO, even though we collect information aboutfirms’ pre- and post-IPO activity.30 We do thisin order to ensure availability of adequate data(a motivating factor for most of the studies inthis literature). Our sample is therefore likely tocontain higher performing firms relative to theunderlying population of biotechnology firms. Thequestions then are whether higher quality firmsemploy a different mix of cooperative forms rela-tive to the biotechnology population, and whetherthe explanatory variable estimates apply similarlyto the underlying distribution of firms. While weacknowledge the firm quality issue as a caveat tothe study, we leave the importance of this selec-tion issue to future research. We note, however,that a benefit of sampling companies that even-tually went public is that the interpretation of agovernance portfolio is strengthened. As previ-ously mentioned, a desirable quality of samplingstart-ups for the purposes of this study is the abil-ity to trace the initial conditions and evolution ofexpropriation hazards and governance capabilitiesover time. The concept of a governance portfolioespecially among this group of firms is likely tobe an important consideration. As well, couplingthe sampling scheme with a start-up fixed-effectsmethod that estimates results based on within-firmacross-time changes allows us to rule out timeinvariant unobserved heterogeneity as a source ofomitted variable bias.

With these selection caveats in mind, we nowturn to some of the theoretical and practical impli-cations of this study. A primary goal of thisresearch has been to explain the underlying driversof variation among cooperative modes by shift-ing the locus of analysis from individual deal-level

30 A related limitation of our sample is that the data are lesscomplete for the private years than for the public years. Privatefirm-year data is generally difficult to obtain, and our data col-lection efforts have been fairly exhaustive given the constraintsof the research design and available data sources. The use of afirm-year level of analysis, combined with controls for publicyears, however, likely mitigates some of the issues associatedwith the reduced number of private-year observations; more-over, the effect of this skew is thus likely to be a reductionin the statistical power of our results rather than a bias in thecoefficients.



determinants of mode choice to factors that tran-scend the deal level. This has enabled us to moresystematically understand the effects of organiza-tional considerations, and in particular to evaluatethe role that appropriation and capabilities-basedmechanisms play when individual deals are con-sidered not only in isolation, but also within thebroader context of the firm. We contribute to theliterature on appropriation considerations by dis-cussing explicitly how firm-level factors can arisefrom deal-level origins. This allows us to placeboth appropriation concerns and capabilities-basedfactors on the same level of analysis, in contrast toprior work where the levels of analysis have oftenbeen mixed.

Our conceptual framework and results have anumber of implications for theory, to which wemake two broad contributions. First, we developthe concept of collaborative governance capabil-ity and detail the underlying mechanisms con-tributing to the capability, path dependence, andorganizational learning. Second, by arguing thathazards of appropriation can operate at the firmlevel, we highlight the importance of examiningTCE-oriented considerations at levels of analy-sis beyond the transaction. Furthermore, in thiscontext, we outline the particular role that inter-dependency across activities plays in generatingfirm-level effects. These results follow much in thespirit of Nickerson et al. (2001), where the authorssuggest that firm strategy derives from a combi-nation of market position, resource profile, andorganizational form. They show that choices alongthese dimensions reinforce each other in such away that TCE factors cannot be considered in iso-lation, and so a ‘constellation of activities’ maybe an important unit of analysis. Focusing on bun-dles of transactions, and the associated effects thatoccur across multiple activities, is thus importantto understanding the implications of TCE. In ourstudy, we place the spotlight on the portfolio as theunit of analysis, developing the concept of a firm-level appropriation environment that we comparewith governance capabilities. In doing so, we alsocontribute to a body of work that links TCE withthe resource-based view of the firm. Silverman(1999), for example, suggests that in the contextof diversification, TCE considerations can play arole in influencing diversification outcomes. Ourpoint of departure from this literature, however, isthat we do not focus on appropriation considera-tions at the industry level (e.g., by examining the

conditions that facilitate licensing or the protectionof intellectual property), but rather on factors thatresult from firm choices and influence firm-leveloutcomes.

The idea of an appropriation environment (andthus, exchange hazards) at the firm level raisesa number of questions for future research. Forexample, to what degree is the firm-level appropri-ation environment a firm capability in the resource-based sense? Does the aggregation of transactions,and the associated interdependencies among thesetransactions, imply the creation of a resource that isany different from the firm’s other resources andcapabilities? More generally, at the firm level ofanalysis, can a TCE perspective on appropriationspillovers together with an organizational capa-bilities perspective offer an adequate theoreticalunification? Furthermore, it would be interesting toexplore how factors such as the relative focus ofthe firm’s patent portfolio and the degree to whichintellectual property is a necessary component ofexpropriation deterrence for the firm’s industryshape the role that patent portfolios play in influ-encing mode choice. A broadening of portfolioscope and an increase in expropriation deterrencevalue arising from patent protection, for example,could increase the impact of portfolio-level patentspillovers. Similarly, for repeat collaboration, set-tings where reputation considerations are of greaterimportance might increase the role of repeat col-laboration as a substitute for equity governancemodes.

The firm-level appropriation mechanisms dis-cussed in this article also have possible impli-cations for the start-up’s organizational develop-ment. If patent spillovers serve to mitigate possibleexpropriation of future products, managers mighthave an incentive to constrain the degree of explo-ration associated with subsequent product intro-ductions, since a more homogenous product lineis more likely to reap the benefits of intellectualproperty spillovers. When the incentive to avoidthreats of future expropriation is sufficiently high,this may affect the balance between explorationand exploitation at the product level. In a relatedmanner, as reputation effects become increasinglysalient, particularly with respect to increasing theimplicit threat of withholding future deals, firmsmay be compelled to conduct a higher proportionof their collaborative activity with repeat partners.Thus, the balance between selecting new partners



and developing deeper ongoing relationships withexisting partners would likewise be affected.

The governance capability results also have anumber of implications. Our results suggest thateven beyond time-invariant firm-specific factors,early choices made in structuring collaborativetransactions can have a lasting impact on the firm’sfuture governance choices and, as a consequence,on an innovator’s overall business model. Thishighlights the importance to managers of under-standing the evolutionary process inherent in astart-up’s development trajectory, along with theimpact that path dependencies ultimately have ongovernance capability development, use, and valu-ation.31 There are also implications for the bodyof literature examining the locus of knowledge(in this case governance knowledge) inside a firm.To the degree that knowledge can be codified incontracts and internal documents, personnel canserve as substitutes for another. On the other hand,if governance knowledge is primarily held withinindividuals, there are different implications for theorganizational and individual changes that need totake place when seeking to develop capabilities tostructure collaborative activity in different ways.While the present study does not settle this debate,it does highlight the importance of understand-ing the trade-offs associated with different typesof governance capability investment, as the roleplayed by these capabilities significantly shapesthe organization of firms’ collaborative portfoliosas well as their subsequent valuation.

Other avenues for future research include explor-ing in more detail the conditions that facilitatethe development, use, and recognition of gov-ernance capabilities. In the start-up developmentphase, for example, founding team and venturecapitalist experience may play an important rolein influencing investments in capability develop-ment. Alternatively, another set of factors influ-encing governance capability development mightrelate to interactions with the firm’s appropriationenvironment. Another area for research concernsthe link between governance capabilities and valu-ation. While markets may penalize firms for devi-ating from existing governance capabilities, we

31 In this regard, a benefit of utilizing a research design that tracksfirms from their year of founding (as opposed to a more staticstudy) is that we can observe the origins of path dependence andunderstand the endogenous evolution of corporate developmentprograms.

might ask whether these effects imply that mar-kets are efficiently recognizing inefficient choices,or whether markets are being myopic and not rec-ognizing the full value associated with investing innew capabilities.

To conclude, this study has sought to understandthe drivers of variation in the firm-level choice ofcooperative commercialization mode, an issue witha host of strategic and resource allocation impli-cations. We address a theoretical gap in the liter-ature by developing a model of cooperative modechoice at the portfolio level. The empirical resultssuggest that patent-based and relational spillovers,along with governance capabilities, drive variationin mode choice; additionally, investments in anddeviations from licensing-related governance capa-bilities influence firm valuation. These results haveimplications for how managers conceptualize thedeterminants and broader consequences of collabo-rative activity. More generally, this study suggestsdirections for future research on the role of a firm’sappropriation environment and governance capa-bilities, and on the associated organization-levelimplications of cooperative mode choice.

ACKNOWLEDGEMENTS

We thank Nick Argyres, Michael Leiblein, DanLevinthal, Kyle Mayer, Nicolaj Siggelkow, Har-bir Singh, two anonymous reviewers, and con-ference participants at London Business School,the Atlanta Competitive Advantage Conference,the Strategic Management Society conference, theAcademy of Management meetings, and theSchumpeter Society for thoughtful comments. Wethank Mark Edwards and Storn White for grant-ing us access to the Recombinant Capital database,and Tong Zhao for excellent research assistance.We gratefully acknowledge Wharton’s Mack Cen-ter for Technological Innovation and the AlfredP. Sloan Foundation for funding this project.

REFERENCES

Anand BN, Khanna T. 2000. Do firms learn to createvalue? The case of alliances. Strategic ManagementJournal , March Special Issue 21: 295–315.

Argote L, Beckman SL, Epple D. 1990. The persistenceand transfer of learning in industrial settings.Management Science 36(2): 140–154.



Argyres NS, Liebeskind JP. 1999. Contractual commit-ments, bargaining power, and governance inseparabil-ity: incorporating history into transaction cost theory.Academy of Management Review 24(1): 49–63.

Arora A, Fosfuri A, Gambardella A. 2001. Markets forTechnology: Economics of Innovation and CorporateStrategy. MIT Press: Cambridge, MA.

Baker G, Gibbons R Murphy KJ. 2002. Relationalcontracts and the theory of the firm. Quarterly Journalof Economics 117(1): 39–84.

Barney J. 1991. Firm resources and sustained competitiveadvantage. Journal of Management 17(1): 99–120.

Baum JAC, Calabrese T, Silverman BS. 2000. Don’t goit alone: alliance network composition and startups’performance in Canadian biotechnology. StrategicManagement Journal , March Special Issue 21(3):267–294.

Bernardo AE, Chowdhry B. 2002. Resources, realoptions, and corporate strategy. Journal of FinancialEconomics 63(2): 211–234.

Chan SH, Kensinger JW, Keown AJ, Martin JD. 1997.Do strategic alliances create value? Journal ofFinancial Economics 46(2): 199–221.

Colombo MG. 2003. Alliance form: a test of thecontractual and competence perspectives. StrategicManagement Journal 24(12): 1209–1229.

Cukier K. 2005. A market for ideas: a survey ofpatents and technology. The Economist (22 October)377(8449): 3–20.

Das S, Sen PK, Sengupta S. 1998. Impact of strategicalliances on firm valuation. Academy of ManagementJournal 41(1): 27–41.

Dolan RJ. 2001. Abgenix and the xenomouse. HarvardBusiness School case 9-501-061. Harvard BusinessSchool: Boston, MA.

Dyer JH. 2000. Collaborative Advantage: Winningthrough Extended Enterprise Supplier Networks .Oxford Press: New York.

Dyer JH, Kale P, Singh H. 2004. When to ally andwhen to acquire. Harvard Business Review 82(7–8):108–115.

Dyer JH, Singh H. 1998. The relational view: cooperativestrategy and sources of interorganizational competitiveadvantage. Academy of Management Review 23(4):660–679.

Filson D, Morales R. 2006. Equity links and informationacquisition in biotechnology alliances. Journal ofEconomic Behavior & Organization 59(1): 1–28.

Gans JS, Hsu DH, Stern S. 2002. When does start-upinnovation spur the gale of creative destruction?RAND Journal of Economics 33(4): 571–586.

Gans JS, Hsu DH, Stern S. 2008. The impact of uncertainintellectual property rights on the market for ideas:evidence from patent grant delays. ManagementScience 54: 982–997.

Gans JS, Stern S. 2003. The product market and themarket for ‘ideas’: commercialization strategies fortechnology entrepreneurs. Research Policy 32(2):333–350.

Gavetti G, Levinthal DA. 2000. Looking forward andlooking backward: cognitive and experiential search.Administrative Science Quarterly 45(1): 113–137.

Gomes-Casseres B, Jaffe AB, Hagedoorn J. 2006. Doalliances promote knowledge flows? Journal ofFinancial Economics 80(1): 5–33.

Gompers P, Lerner J. 1999. The Venture Capital Cycle.MIT Press: Cambridge, MA.

Gulati R. 1995. Does familiarity breed trust? Theimplications of repeated ties for contractual choicein alliances. Academy of Management Journal 38(1):85–112.

Gulati R. 1998. Alliances and networks. StrategicManagement Journal , April Special Issue 19:293–317.

Gulati R, Singh H. 1998. The architecture of cooperation:managing coordination costs and appropriationconcerns in strategic alliances. Administrative ScienceQuarterly 43(4): 781–814.

Hagedoorn J. 2002. Inter-firm R&D partnerships: anoverview of major trends and patterns since 1960.Research Policy 31(4): 477–492.

Hall BH, Jaffe AB, Trajtenberg M. 2001. The NBERpatent citations data file: lessons, insights andmethodological tools. NBER working paper 8498,National Bureau of Economic Research, Cambridge,MA.

Hellmann T, Puri M. 2002. Venture capital and the pro-fessionalization of start-up firms: empirical evidence.Journal of Finance 57(1): 169–197.

Hoffmann WH. 2007. Strategies for managing a portfolioof alliances. Strategic Management Journal 28(8):827–856.

Hsu DH. 2004. What do entrepreneurs pay forventure capital affiliation? Journal of Finance 59(4):1805–1844.

Huyghebaert N, Van de Gucht LM. 2004. Incumbentstrategic behavior in financial markets and the exitof entrepreneurial start-ups. Strategic ManagementJournal 25(7): 669–688.

Kale P, Dyer JH, Singh H. 2002. Alliance capability,stock market response, and long-term alliance success:the role of the alliance function. Strategic ManagementJournal 23(8): 747–767.

Kale P, Singh H, Perlmutter H. 2000. Learning andprotection of proprietary assets in strategic alliances:building relational capital. Strategic ManagementJournal , March Special Issue 21: 217–237.

Klein B, Crawford RG, Alchian AA. 1978. Verticalintegration, appropriable rents, and the competitivecontracting process. Journal of Law and Economics21(2): 297–326.

Kortum S, Lerner J. 1999. What is behind the recentsurge in patenting? Research Policy 28(1): 1–22.

Kreps DM, Wilson R. 1982. Reputation and imperfectinformation. Journal of Economic Theory 27(2):253–279.

Lavie D. 2007. Alliance portfolios and firm performance:a study of value creation and appropriation in theU.S. software industry. Strategic Management Journal28(12): 1187–1212.

Leiblein MJ, Miller DJ. 2003. An empirical examinationof transaction- and firm-level influences on the verticalboundaries of the firm. Strategic Management Journal24(9): 839–859.



Lerner J. 1994. Venture capitalists and the decision togo public. Journal of Financial Economics 35(3):293–316.

Lerner J, Merges RP. 1998. The control of technologyalliances: an empirical analysis of the biotechnologyindustry. Journal of Industrial Economics 46(2):125–156.

Levin RA, Klevorick A, Nelson RR, Winter SG. 1987.Appropriating the returns from industrial R&D. Brook-ings Papers on Economic Activity–Microeconomics .Brookings Institution: Washington, D.C.

Levitt B, March JG. 1988. Organizational learning. InAnnual Review of Sociology (Vol. 14), Scott WR,Short JF (eds). Annual Reviews: Palo Alto, CA;319–340.

Macaulay S. 1963. Non-contractual relations in business:a preliminary study. American Sociological Review28: 55–69.

Madhok A. 2002. Reassessing the fundamentals andbeyond: Ronald Coase, the transaction cost andresource-based theories of the firm and the institutionalstructure of production. Strategic Management Journal23(6): 535–550.

March JG. 1991. Exploration and exploitation inorganizational learning. Organization Science 2(1):71–87.

Mathews RD. 2006. Strategic alliances, equity stakes,and entry deterrence. Journal of Financial Economics80(1): 35–79.

Mayer KJ, Argyres NS. 2004. Learning to contract:evidence from the personal computer industry.Organization Science 15(4): 394–410.

Mayer KJ, Bercovitz J. 2008. The influence of intertia oncontract design: contingency planning in informationtechnology service contracts. Managerial and Deci-sion Economics 29(2–3): 149–164.

Mayer KJ, Nickerson JA. 2005. Antecedents and per-formance implications of contracting for knowledgeworkers: evidence from information technology ser-vices. Organization Science 16(3): 225–242.

Mayer KJ, Salomon RM. 2006. Capabilities, contractualhazards, and governance: integrating resource-basedand transaction cost perspectives. Academy ofManagement Journal 49(5): 942–959.

Milgrom P, Roberts J. 1982. Predation, reputation, andentry deterrence. Journal of Economic Theory 27(2):280–312.

Monteverde K, Teece DJ. 1982. Supplier switching costsand vertical integration in the automobile industry.Bell Journal of Economics 13(1): 206–213.

Nelson RR, Winter SG. 1982. An Evolutionary Theory ofEconomic Change. Belknap Press: Cambridge, MA.

Nickerson JA, Hamilton BH, Wada T. 2001. Marketposition, resource profile, and governance: linkingPorter and Williamson in the context of internationalcourier and small package services in Japan. StrategicManagement Journal 22(3): 251–273.

Oxley JE. 1997. Appropriability hazards and governancein strategic alliances: a transaction cost approach.Journal of Law, Economics, & Organization 13(2):387–409.

Park SH, Chen R, Gallagher S. 2002. Firm resourcesas moderators of the relationship between marketgrowth and strategic alliances in semiconductorstart-ups. Academy of Management Journal 45(3):527–545.

Pisano GP. 1989. Using equity participation to supportexchange: evidence from the biotechnology industry.Journal of Law, Economics, & Organization 5(1):109–126.

Poppo L, Zenger T. 2002. Do formal contracts andrelational governance function as substitutes orcomplements? Strategic Management Journal 23(8):707–725.

Porter ME. 1996. What is strategy? Harvard BusinessReview 74(6): 61–78.

Porter ME, Siggelkow N. 2008. Contextual interactionswithin activity systems and sustainability of competi-tive advantage. Academy of Management Perspectives22(2): 34–56.

Robinson DT, Stuart TE. 2007. Financial contractingin biotech strategic alliances. Journal of Law andEconomics 50(3): 559–595.

Rothaermel FT. 2001. Incumbent’s advantage throughexploiting complementary assets via interfirm cooper-ation. Strategic Management Journal , June–July Spe-cial Issue 22: 687–699.

Ryall MD, Sampson RC. 2009. Formal contracts inthe presence of relational enforcement mechanisms:evidence from technology development projects.Management Science 55: (forthcoming). Available at:http://works.bepress.com/michael ryall/11.

Sampson RC. 2004a. The cost of misaligned governancein R&D alliances. Journal of Law, Economics andOrganization 20(2): 484–526.

Sampson RC. 2004b. Organizational choice in R&Dalliances: knowledge-based and transaction costperspectives. Managerial and Decision Economics25(6–7): 421–436.

Santoro MD, McGill JP. 2005. The effect of uncer-tainty and asset co-specialization on governance inbiotechnology alliances. Strategic Management Jour-nal 26(13): 1261–1269.

Shan W, Walker G, Kogut B. 1994. Interfirm cooperationand startup innovation in the biotechnology industry.Strategic Management Journal 15(5): 387–394.

Shapiro C. 2000. Navigating the patent thicket: cross-licenses, patent pools and standard setting. Workingpaper, Competition Policy Center, University ofCalifornia at Berkeley.

Silverman BS. 1999. Technological resources andthe direction of corporate diversification: towardan integration of the resource-based view andtransaction cost economics. Management Science45(8): 1109–1124.

Stein JC. 1989. Efficient capital markets, inefficient firms:a model of myopic corporate behavior. QuarterlyJournal of Economics 104(4): 655–669.

Teece DJ. 1986. Profiting from technological innovation:implications for integration, collaboration, licensingand public policy. Research Policy 15(6): 285–305.



Vassolo RS, Anand J, Folta TB. 2004. Non-additivity inportfolios of exploration activities: a real options-based analysis of equity alliances in biotechnology.Strategic Management Journal 25(11): 1045–1061.

Villalonga B, McGahan AM. 2005. The choice amongacquisitions, alliances, and divestitures. StrategicManagement Journal 26(13): 1183–1208.

Walsh JP, Ungson GR. 1991. Organizational memory.Academy of Management Review 16(1): 57–91.

Wernerfelt B. 1984. A resource-based view of the firm.Strategic Management Journal 5(2): 171–180.

Williamson OE. 1975. Market and Hierarchies: Analysisand Antitrust Implications . The Free Press: New York.

Williamson OE. 1999. Strategy research: governanceand competence perspectives. Strategic MangementJournal 20(12): 1087–1108.

Winter SG. 2003. Understanding dynamic capabilities.Strategic Management Journal , October Special Issue24: 991–995.

Yoffie DB, Kwak M. 2002. Mastering balance: howto meet and beat a stronger opponent. CaliforniaManagement Review 44(2): 8–24.

Zollo M, Reuer JJ, Singh H. 2002. Interorganizationalroutines and performance in strategic alliances.Organization Science 13(6): 701–713.



APPENDIX: ALLIANCE ANDFIRM-LEVEL SEGMENTCLASSIFICATIONS

Our 24 category therapeutic classification schemeis a derivative of the Anatomical Therapeutic Clas-sification (ATC) system published by the WorldHealth Organization’s Collaborating Centre forDrug Statistics Methodology. The ATC systemcontains a five-level classification system for drugswith increasing specificity at lower levels. The top-level contains 14 categories; we retain nine of theseoriginal categories, split three of the categories intoeither two or three subcategories, and add eightadditional categories to accommodate firms andalliances with process or delivery-focused tech-nologies, resulting in a 24-category classificationscheme. In classifying the alliances and firm-years,the original ATC therapeutic categories take prece-dence over the new categories.

Therapeutic categoriesAllergic (x)Autoimmune (x)Bioinformatics/diagnostics/drug discovery (*)Musculoskeletal (o)Cancer (x)

Cardiovascular (o)Nervous system (o)Drug delivery (*)Endocrinology and metabolic (x)Gastrointestinal (x)Genitourinary and gynecological (o)Hematologic (o)Infectious—bacterial/fungal (x)Infectious—viral (x)Ophthalmology (o)Orthopedics (*)Respiratory (o)Wound healing (o)Generics/supplements (*)Devices (*)Veterinary (o)RNAi (*)Supplies/reagents/processing/manufacturing (*)Agricultural/industrial biotech (*)

(o) indicates an original ATC category or a deriva-tive of an original ATC category

(x) indicates a category that has been split from anoriginal ATC category

(*) indicates a new category


modes of cooperative r&d commercialization by …8].pdfthe context (dyer, kale, and singh,...

Documents