LEVERAGING INTERNAL AND EXTERNAL · PDF file1 ESSEC Business School, Cergy Pontoise, ......

Strategic Management JournalStrat. Mgmt. J., 31: 734–758 (2010)

Published online EarlyView in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/smj.834

Received 30 October 2007; Final revision received 25 November 2009

LEVERAGING INTERNAL AND EXTERNALEXPERIENCE: EXPLORATION, EXPLOITATION,AND R&D PROJECT PERFORMANCE

HA HOANG1* and FRANK T. ROTHAERMEL2

1 ESSEC Business School, Cergy Pontoise, France2 College of Management, Georgia Institute of Technology, Atlanta, Georgia, U.S.A.

Although one tenet in the alliance literature is that firms learn from prior experience, we posit thatany potential learning effects depend on the type of experience. In particular, we hypothesizethat alliance exploitation experience has positive effects on R&D project performance, whilealliance exploration experience has negative effects. We further posit that an internal explorationcompetence allows firms to leverage their external exploitation experience more fully. In contrast,when firms combine internal exploitation experience with external exploration experience, thenegative effects on R&D project performance become more pronounced. To test this integrativemodel of organizational learning, we leverage a unique and detailed dataset of 412 R&D projectsin biotechnology conducted by large pharmaceutical companies between 1980 and 2000. Usinga competing risk event history model predicting successful product approval versus projecttermination, we find support for our theoretical model. Copyright 2010 John Wiley & Sons,Ltd.

INTRODUCTION

A firm’s ability to adapt to shifting knowledgeenvironments is a key dynamic capability to ensurecontinued survival and competitiveness (Eisen-hardt and Martin, 2000; Teece, Pisano, and Shuen,1997). To answer the question of how incum-bent firms adapt to and even capitalize on radi-cal technological change, one stream of researchhighlights the role of interfirm research and devel-opment (R&D) collaborations (Arora and Gam-bardella, 1990; Hill and Rothaermel, 2003; Teece,1992; Tripsas, 1997b). Collaborative R&D as aresponse to shifting knowledge environments hasbeen linked to a variety of positive outcomes

Keywords: exploration and exploitation; ambidexterity;dynamic capabilities; new product development; biotech-nology; pharmaceutical industries*Correspondence to: Ha Hoang, ESSEC Business School,Avenue Bernard Hirsch, BP 50105 Cergy, 95021 Cergy PontoiseCedex, France. E-mail: [email protected]

including greater firm innovativeness and perfor-mance (Nicholls-Nixon and Woo, 2003; Rothaer-mel, 2001). A second stream of research on howincumbent firms might adapt to radical technolog-ical change has emphasized the need for firms topossess sufficient absorptive capacity (Cohen andLevinthal, 1990; Zahra and George, 2002). A num-ber of empirical studies have found support for thenotion that the capacity to recognize, value, assimi-late, and apply new external knowledge is a signif-icant predictor of successful organizational trans-formation (Arora and Gambardella, 1994; Helfat,1997; Kaplan, Murray, and Henderson, 2003; Trip-sas, 1997a).

An important consequence of sustained involve-ment in collaborative R&D is that firms withgreater alliance or external experience can extractmore benefits than firms with less experience(Rothaermel and Deeds, 2006; Sampson, 2005).This is because repeated engagements in the focalactivity allow firms to learn from their experience

Copyright 2010 John Wiley & Sons, Ltd.

Leveraging Internal and External Experience 735

(learning-by-doing), and to store and retrieve theinferred learning for future engagements in thefocal activity (Levitt and March, 1988). Consistentwith the learning curve hypothesis (Yelle, 1979),external experience has been found to be positivelyrelated to subjective and objective outcome mea-sures of individual alliances as well as overall firmperformance (Anand and Khanna, 2000; Chang,2003; Hoang and Rothaermel, 2005; Kale, Dyer,and Singh, 2002; Zollo, Reuer, and Singh, 2002).

The link between external experience and per-formance, however, has been based on somewhatcourse-grained analyses. With few notable excep-tions (e.g., Baum, Calabrese, and Silverman, 2000;Dussauge, Garrette, and Mitchell, 2000; Lavieand Rosenkopf, 2006; Park, Chen, and Gallagher,2002; Rothaermel, 2001; Rothaermel and Deeds,2004), prior studies have generally not systemati-cally distinguished between alliances at differentfoci of the value chain and are thus at risk ofaggregation bias. Ignoring the potential variancethat exists across R&D alliances initiated underdifferent motivations can lead to spurious results.Moreover, prior alliance studies generally focuson outcomes at the firm level of analysis, whichare more theoretically distant from performanceat the collaboration level. As a consequence, weargue that a more subtle understanding of learn-ing within the alliance context and its relationshipto the development of internal capabilities necessi-tates a more fine-grained analysis. We accomplishthis by analyzing different types of R&D alliancesand different types of internal experience com-bined with a focus on performance effects at theproject level of analysis.

We leverage Koza and Lewin’s (1998) typol-ogy of alliance activity to examine the impactof external exploration and exploitation on sub-sequent R&D project performance. This typologyrecognizes that firms emphasize external activi-ties in different components of the R&D activ-ity chain: some alliances are formed to explorenew competencies to identify new opportunities,while others are used to exploit existing compe-tencies in order to leverage known opportunities.By and large, efforts to link alliance activity tooutcomes aggregate these different alliances and,as a result, ignore important variation across themin the different types of alliance partners and thenovelty and ambiguity of their knowledge con-tent (exceptions Baum et al., 2000; Lavie andRosenkopf, 2006; Rothaermel and Deeds, 2006).

A focus on the project level of analysis brings tothe fore the challenges that firms face to lever-age these different alliance experiences appropri-ately to enhance subsequent project outcomes. Weuse time to drug approval of biotechnology-basedR&D projects undertaken by established pharma-ceutical firms to assess the performance impact ofexternal exploratory and exploitative activity. Weposit that a firm’s experience in external explo-ration versus external exploitation has opposingdirect effects: external exploration experience hasa negative effect on subsequent R&D project per-formance, while the effect of external exploitationexperience is positive.

We then introduce a contingency perspective byexamining how internal R&D experience moder-ates external exploration and exploitation expe-rience. Internal R&D experience is critical to afirm’s successful adaptation to radical technologi-cal innovation, because it builds the foundation ofa firm’s absorptive capacity (Cohen and Levinthal,1990). We build on the theoretical insight that thekey to understanding a firm’s dynamic capabili-ties lies in how a critical process such as newproduct development interacts with a firm’s abilityto leverage external partnerships (Eisenhardt andMartin, 2000; Teece et al., 1997). By focusing onthe project level of analysis and identifying howexploration and exploitation activities are interre-lated in the new product development process, weseek to shed greater light on how firms tap existinginternal experience to leverage externally driveninnovation.

Understanding the dynamics of R&D collabora-tions in new product development is particularlyrelevant in the pharmaceutical industry, which hasseen a significant increase in the opportunitiesto partner with new entrants who focus on drugdiscovery and development projects that lever-age scientific advances in biotechnology. It is acontext where project timeliness is also critical:when a firm is the first to introduce an inno-vative product, it is often able to extract tem-porary monopoly rents based on patent-protectedintellectual property (Lieberman and Montgomery,1988; Macher and Boerner, 2006). In the phar-maceutical industry, being fast to market andproduct sales are highly correlated (Grabowskiand Vernon, 1990; Roberts, 1999). To empiri-cally test our theoretical model, we draw on anunusually fine-grained dataset documenting 412biotechnology-based drug development projects

Copyright 2010 John Wiley & Sons, Ltd. Strat. Mgmt. J., 31: 734–758 (2010)DOI: 10.1002/smj

736 H. Hoang and F. T. Rothaermel

undertaken by established pharmaceutical compa-nies under a variety of governance arrangementsover the 21-year time period between 1980 and2000.

THEORY AND HYPOTHESES

External exploration and exploitation

We build herein on March’s (1991) exploration-exploitation framework of organizational learning,and the work of Koza and Lewin (1998) who dis-tinguished between alliance activity that is moti-vated by the need to explore for new opportu-nities and alliances that are formed to exploitknown opportunities. Applying an exploration-exploitation lens to strategic alliances is well estab-lished in the literature (e.g., Lavie and Rosenkopf,2006; Park et al. 2002; Rothaermel, 2001; Rothaer-mel and Deeds, 2004). In addition to capturing thetensions inherent in different types of knowledge(Levinthal and March, 1993), the exploration andexploitation framework of organizational learn-ing also maps quite well onto the research con-text of this study. Following the prior litera-ture, we employed the exploration versus exploita-tion distinction to alliances because characteriz-ing alliances as exploratory is highly consistentwith the pharmaceutical drug discovery and earlystage development process. Similarly, exploitationalliances map well onto activities that occur inlater stages of the value chain that tap a firm’sexisting knowledge including clinical testing, reg-ulatory affairs, distribution, and marketing/sales.

As with the broader notion of exploratory search,exploration alliances are characterized by longtime horizons and unpredictable, high variancereturns. Exploration alliances are conduits fororganizational learning. In some cases, alliancesinvolve the cocreation of new knowledge by part-ners (Lubatkin, Florin, and Lane, 2001). In othercases, each partner in an exploration allianceattempts to identify, transfer, and absorb part orall of the partner’s valuable knowledge assets.In contrast, exploitation alliances involve shortertime horizons between the point of learning andthe realization of predictable, low variance bene-fits. Exploitation alliances are motivated by (tem-porary) access to the partner’s knowledge assetsto leverage complementarities across different andunique competencies along the value chain

(Bresser, Heuskel, and Nixon, 2000), while eachalliance partner maintains its comparative knowl-edge advantage (Grant and Baden-Fuller, 2004).

Differentiating between exploration and exploi-tation alliances is especially salient in the con-text of the R&D process. In exploration alliances,partners are motivated to discover something new,frequently advancing the boundaries of basic sci-ence (Rosenkopf and Nerkar, 2001; Rothaermeland Deeds, 2004). For established pharmaceuticalfirms, such collaborations typically involve uni-versities and research intensive start-ups as theirpartners (Arora and Gambardella, 1990; Chang,2003; Shan, Walker, and Kogut, 1994). In exploita-tion alliances, firms seek to leverage their existingcapabilities in areas such as clinical testing, regula-tory affairs, distribution, and sales and marketing.A number of studies highlight that distinguish-ing between exploration and exploitation alliancesallows for a more fine-grained understanding ofhow alliance activity can lead to differential out-comes. For example, Rothaermel (2001) docu-ments how large pharmaceutical firms that focuson exploitation rather than exploration alliancesin their network strategy exhibit higher perfor-mance when adapting to biotechnology. Rothaer-mel and Deeds (2004) find that biotechnologystart-ups that orchestrate an integrative alliancesystem that leverages exploration and exploitationalliances in a sequential fashion achieve supe-rior new product development performance. Morerecently, Lavie and Rosenkopf (2006) also notethat the exploration-exploitation distinction revealsimportant nuances regarding how firms balancethese different activities when forming alliances.

External exploration/exploitation experienceand R&D project performance

Distinguishing between an exploration versusexploitation focus in collaborative activity canhave important implications for the ease of learn-ing and, hence, the degree to which firms can buildand leverage external experience for greater perfor-mance in subsequent R&D projects. We suggestthat, in contrast to the challenges of leveragingexploratory alliance experience, firms can morereadily leverage their experience from exploita-tive alliances. Exploitation alliances are typicallyfocused on incremental improvements to existing



routines, and are readily codified and embod-ied in refinements to current products and pro-cesses, because they leverage existing complemen-tarities between partners (Benner and Tushman,2003; Teece, 1992). In the collaborative R&Dcontext, the large pharmaceutical firms contributemore explicit and codifiable knowledge, includingtheir manufacturing capabilities, regulatory know-how, and significant sales force management anddeployment knowledge. This downstream knowl-edge has been honed over decades through cumu-lative experience, and as a consequence, pharma-ceutical companies have developed highly efficientorganizational routines and standard operating pro-cedures to effectively manage the production anddistribution process (Pisano, 1996). Biotechnologyfirms, in turn, contribute a potential new drug thathas undergone significant preclinical development,thus considerably reducing the ambiguity and tac-itness of the knowledge to be managed in thealliance. In short, exploitation alliances are basedon a division of labor through a matching of spe-cialized complementary resources and skills.

Due in part to these knowledge characteristics,firms can more readily accumulate and leverageexternal exploitation experience through repeatedengagements in the focal activity or learning-by-doing (Pisano, 1994). Although entering, manag-ing, and exiting alliances creates nontrivial coordi-nation costs, an emphasis on exploitative alliancesreduces these costs significantly, because the needfor extensive and deep communication with part-ners is generally lower at later stages of the prod-uct development process (Rothaermel and Deeds,2006). Lower effort and costs, especially in themanagerial attention required to coordinate andleverage external exploitation, translates intohigher learning benefits. Pharmaceutical firms, forexample, may learn which problems can arise asa biotech drug candidate moves from small-scaledevelopment to large-scale production, a signifi-cant challenge that can derail successful and timelyproject completion (Pisano, 1996). The speed ofproject completion is a critical performance met-ric in this industry, because being fast to marketand product sales are endogenous to a large extent(Grabowski and Vernon, 1990; Roberts, 1999).

The learning task is also greatly simplified in anexploitative alliance, because each firm focuses onits area of specialization. Such a focus is likelyto benefit project completion speed, because it

allows each partner to focus on its distinctive com-petence, thus leveraging comparative advantagesacross firms (Azoulay, 2004; Mowery, Oxley, andSilverman, 1996). Rather than sending staff intoeach other’s laboratories to further push out thefrontier of basic science, as is frequently thecase in exploration alliances, firms in exploita-tive alliances can focus on managing the allianceinterface as a particular drug moves from preclini-cal to clinical development. This situation is oftendescribed as a ‘hand-off’ from the biotechnol-ogy venture to a pharmaceutical company (Pisanoand Mang, 1993). Leveraging their prior experi-ence in their respective area of expertise requiresless intensive managerial attention and resources,because firms can more readily identify sources ofmisalignment and rectify them before they under-mine alliance performance. With a clear focus, firminvestments in an alliance management capability,including an alliance management office, are likelyto be more successful, prompting further invest-ment in this capability (Kale et al., 2002).

In the biotechnology R&D context, exploitationalliances are well suited to be managed throughsuch a formal business process, because exploita-tion alliances involve the downstream capabili-ties of the established pharmaceutical companies,including manufacturing, legal expertise, and sales,distribution, and marketing. For example, the phar-maceutical company Lilly established an alliancemanagement process that is well designed to cap-ture benefits from exploitation alliances (Sims,Harrison, and Gueth, 2001). Each alliance is man-aged through a three-person team, including asenior manager for high-level oversight and sup-port, an alliance leader responsible for the day-to-day management of the alliance, and an alliancemanager from the corporate office of alliancemanagement, who serves as business integratorbetween the two alliance partners. Through inter-views with industry experts, we learned that explo-ration alliances, in contrast, tend to initially remain‘under the radar’ of management, because they aregenerally formed by the partner firms’ scientists,and only later do they receive managerial attentionshould promising results emerge that fit the firm’sstrategy. Moreover, because of the tacit knowledgethat is being developed in exploration alliances,they are largely incompatible with formal man-agement processes (Benner and Tushman, 2003).Because the ease of learning from alliance experi-ence is greater when the focus is on exploitation,



thus allowing incumbent pharmaceutical firms tomove down the learning curve of how to success-fully complete R&D projects in a timely fashion,we posit that:

Hypothesis 1: External exploitation experiencehas a positive effect on R&D project perfor-mance.

Exploration typically involves the developmentof new knowledge that is tacit and of uncertainvalue. To be of strategic value, however, this newknowledge must then be integrated into broaderorganizational capabilities that allow for the exe-cution of key tasks. In comparison to externalexploitation, exploratory alliances expose a focalfirm to new, cutting-edge knowledge. This, in turn,commensurately raises the challenge of learningand integrating new capabilities from partners. Dif-ficulties arise when there are differences betweenfirms, particularly in their dominant logics, knowl-edge bases, and organizational structures (Lane andLubatkin, 1998).

In the biotechnology context, firms must oftenwork with partners that operate under different cul-tures, incentive systems, and norms. Large pharma-ceutical companies engage in exploratory allianceswith research universities and biotechnology firms.Research universities are generally large, bureau-cratic structures whose primary goal is the cre-ation and widespread dissemination of cutting-edge, basic knowledge. Biotechnology firms arefor-profit entities that place far greater emphasison developing and leveraging applied, proprietaryknowledge. In contrast to the large pharmaceuticalfirms, they are also able to provide high-poweredincentives for breakthrough innovation with theuse of stock options for their scientific and man-agerial staff, especially if the new venture is pre-initial public offering (IPO). Because knowledgeis embedded in particular social contexts (Kogutand Zander, 1992), dissimilarity between partnersincreases the difficulty of knowledge transfer andlearning within alliances, such that overcomingthese barriers to knowledge transfer requires sig-nificant managerial resources and attention (Mora-Valentin, Montoro-Sanchez, and Guerras-Martin,2004; Rothaermel and Deeds, 2006; Simonin,1997, 1999).

Learning from partners and the ability to lever-age prior experience is not only affected by thetypes of partners but also by the knowledge that

is being transferred within an exploratory alliance.The primary goal of an exploration alliance withuniversities and entrepreneurial start-ups involvesintegrating leading-edge scientific discoveries intoa new product or process. This process generallyinvolves the transfer of complex and tacit knowl-edge between partners. Pisano (1996) insightfullyhighlights the differences in the type of knowl-edge between exploratory search in the discoveryphase and exploitative activities in the commer-cial development phase: ‘In the discovery phase ofpharmaceutical R&D projects, research scientistsdevelop crude processes for synthesizing relativelysmall amounts of the molecule under investigation.These laboratory methods of production, however,are almost always completely unsuitable for man-ufacturing the compound in commercial volumesat required cost and quality levels’ (Pisano, 1996:1104). Because the knowledge base is basic andemergent in exploratory alliances, there is con-siderable uncertainty about the potential applica-tions of new knowledge and its contribution toan organization’s capabilities. Moreover, partnershave little or no prior experience with advanc-ing this type of knowledge, and unlike relativelyhomogeneous contexts that allow for experimenta-tion and the accumulation of systematic feedback,novel conditions increase the difficulty of buildingand leveraging prior experience.

Indeed, repeated exposure to novel learning con-texts can lead to negative knowledge transfer, aconcept that originates in cognitive psychology,and which has frequently been demonstrated at theindividual level (Gick and Holyoak, 1987). Nega-tive knowledge transfer describes a situation whereexperience gained in a prior activity is transferredto a new activity that appears to be similar onthe surface, but is, in fact, fundamentally differ-ent. This, in turn, implies that prior experience canactually hurt rather than help future performance.For example, Cohen and Bacdayan (1994) demon-strate how individuals who accumulated experi-ence through repeated engagements in a card gameplayed under specific rules were outperformed bynovice, untrained card players when the rules of anew game differed slightly from the game in whichthe prior experience was accumulated.

Within the context of biotechnology R&Dprojects, in particular, Pisano (1997: 216) detailshow attempts to leverage biotechnology projectexperience can lead to negative knowledge trans-fer, and thus reduce project performance, because



of the immature knowledge base upon which newproducts and processes are developed. He docu-ments project delays that occurred because a firmtried to apply basic process technology developedin an earlier project to a subsequent project. Afterfollowing codified knowledge embodied in proto-cols developed from prior experience, the firm wasunable to replicate the success of the prior projectin the subsequent project. In this case, the methodused led to a biologically inactive molecule; as aconsequence it was of no therapeutic value. Thesignificant challenges of learning between dissim-ilar organizations partnering under conditions ofsignificant knowledge novelty and ambiguity thusleads us to posit that prior exploratory allianceexperience will have a negative impact on thesuccessful completion of an established pharma-ceutical firm’s subsequent R&D projects.

Hypothesis 2: External exploration experiencehas a negative effect on R&D project perfor-mance.

Moderating effects of internalexploration/exploitation experience

Scholars have long held that the ability to lever-age external activities depends on the extent towhich external knowledge is related and assimi-lated with a firm’s own knowledge base (Cohenand Levinthal, 1989). The extent of a firm’sabsorptive capacity—understood as ‘the ability ofa firm to recognize the value of new, external infor-mation, assimilate it, and apply it to commercialends’ (Cohen and Levinthal 1990: 128)—can, inturn, influence its perceived returns on subsequentinvestments to develop new knowledge or exploitits existing knowledge. Relevant internal capabil-ities are thus likely to play an important role indetermining whether firms are able to fully lever-age their external experiences.

The successful completion of the research com-ponent of the R&D process (exploration) requiressubsequent exploitative activities including regu-latory expertise, manufacturing, sales, and distri-bution (Rothaermel and Deeds, 2004). We thusconsider how complementary internal experiencemoderates a firm’s external activities, where thedefinition of complementarity is based on thesequence of exploration-exploitation that char-acterizes the value chain of the pharmaceuticalresearch and development process. These combi-nations can contribute to ambidexterity, defined as

the ‘ability of a firm to simultaneously exploreand exploit’ (O’Reilly and Tushman 2008: 185).We therefore consider: 1) whether a firm’s inter-nal exploration experience moderates the effectsof its external exploitation experience, and 2)whether internal exploitation experience moder-ates the effects of its external explorationexperience.

External exploitation and internal exploration

We suggest that the benefits to external exploita-tion experience on subsequent R&D project perfor-mance are enhanced when combined with internalexploration experience. In their development ofthe concept of absorptive capacity, scholars haverepeatedly emphasized the importance of transfor-mative skills and routines in order for firms tobenefit and adjust to rapid technological change(Cohen and Levinthal, 1990; Garud and Nay-yar, 1994; Lane and Lubatkin, 1998; Zahra andGeorge, 2002). A firm’s absorptive capacity is builtthrough continuous engagements in basic researchover time (Cohen and Levinthal, 1989, 1990), and,thus, through repeated engagements in exploratoryactivities. A firm’s internal exploration experi-ence, in turn, leads to firm-specific knowledge thatenables a firm to monitor, screen, evaluate, andleverage externally generated knowledge (Helfat,1997; Mowery, 1983).

Within the context of the pharmaceutical indus-try’s adaptation to biotechnology, Rothaermel andHill (2005) document that a pharmaceutical firm’sinternal R&D capability provides it with a superiorability to understand and value new biotechnologyknowledge. This, in turn, enables the pharmaceuti-cal firm to select the most promising alliance part-ners among the swarm of new entrants (Schum-peter, 1942), with over 2,000 new biotechnologyfirms vying for reputable alliance partners amonga relatively small number of incumbent pharma-ceutical companies (Stuart, Hoang, and Hybels,1999). The ability to select the most promisingalliance partners is a valuable competence, becausepharmaceutical firms manage multiple new prod-uct development projects with different partners inseveral market domains simultaneously (Rothaer-mel and Deeds, 2006; Vassolo, Anand, and Folta,2004). More generally, pursuing a larger number ofexternal knowledge sources simultaneously with-out the requisite internal absorptive capacity hasbeen linked to reduced innovative performance in



a recent study of innovation by U.K. manufactur-ing firms (Laursen and Salter, 2006).

A firm’s internal exploration experience buildsthe foundation of successfully completing R&Dprojects in a timely fashion, because of the sequen-tial nature of the drug discovery and develop-ment process, where, at the level of each individ-ual project, exploration activities precede exploita-tion activities (Rothaermel and Deeds, 2004). Afirm’s internal exploration experience also enablesthe firm to more fully assimilate and transformthe learning benefits obtained from its externalexploitation experience, because it allows for moreeffective interfirm capability transfer (Hamel,1991; Simonin, 1997, 1999). As an example, thepharmaceutical firm Lilly was able to success-fully transfer important biotech process capabilitiesthrough an exploitation alliance with Genentech,which Lilly applied successfully to subsequentR&D projects conducted in-house (Fisher, 1995;McDaniel, 1994). Altogether, we argue that afirm’s internal exploration experience has positiveimplications for the firm’s ability to successfullycomplete R&D projects in timely fashion whencoupled with external exploitation experience.

Hypothesis 3: The positive effect of externalexploitation experience on R&D project perfor-mance is enhanced in the presence of internalexploration experience.

External exploration and internal exploitation

We further propose that the negative effect ofexternal exploration experience on R&D projectperformance is intensified when combined withinternal exploitation experience. At first glance,past and current drug research and developmentprojects share broad features that would seem tofacilitate the application of internal exploitationexperience. The biotechnology knowledge base,however, is still in its infancy relative to the tradi-tional drug discovery process, and thus representsa new knowledge paradigm that undermines thevalue of pharmaceutical firms’ prior knowledge(Rothaermel, 2001). Moreover, the pharmaceuti-cal firms’ dominant logics (Prahalad and Bettis,1986) were developed and refined through decadesof competing on drug discovery and developmentwithin the traditional paradigm of chemical syn-thesis. Established pharmaceutical firms, therefore,can be prone to misapplying learning from external

exploration activities, because of the lack of theo-retical and practical guidance to aid in subsequentprojects due to the newness of knowledge explored(Pisano, 1996). This effect is likely to be more pro-nounced in the presence of strong internal exploita-tion experience that implies high levels of successwithin the traditional paradigm, potentially leadingto a competency trap (Levitt and March, 1988).

Firms that rely heavily on external knowledgesources for basic R&D are at a relative dis-advantage when it comes to building the skillsand resources needed to respond to technologicalchange (Bettis, Bradley, and Hamel, 1992; Lei andHitt, 1995). An internal focus on building exploita-tion experience and complementary assets reducesa firm’s ability to select the most promisingexploratory research projects. Moreover, the exten-sive use of external sources for exploratory knowl-edge can invite opportunism on the part of thealliance partners. Concerning biotechnology R&Dknow-how, Pisano (1997) found empirical supportfor a lemons hypothesis in the market for collab-orative drug development projects. Due to infor-mation asymmetry combined with quality uncer-tainty, the biotechnology ventures have a tendencyto offer inferior projects for collaboration, whilemaintaining the more promising projects in-housefor solo development and commercialization. Thisproblem is accentuated for the large pharma-ceutical companies when they focus on internalexploitation, because they lack the requisite inter-nal exploratory experience to evaluate the qualityof the R&D projects offered for collaboration bythe biotechnology ventures. Moreover, firms thatattempt to integrate external exploration with inter-nal exploitation within the context of an R&Dcollaboration face additional complexity in theknowledge transfer process, because tacit knowl-edge developed within an exploratory project isdifficult to transfer and apply without an adequatelevel of internal knowledge (Simonin, 1997, 1999).

Given the fundamental role that internal explo-rative activities play in building a firm’s absorptivecapacity, we suggest that the effect of negativeknowledge transfer is stronger when a firm focuseson external exploration combined with internalexploitation.

Hypothesis 4: The negative effect of externalexploration experience on R&D project perfor-mance is enhanced in the presence of internalexploitation experience.



METHODOLOGY

Research setting

We assess the new drug discovery and develop-ment project performance of established pharma-ceutical companies in biotechnology. Establishedpharmaceutical companies like Pfizer or GlaxoWellcome are the firms that were in existence priorto the emergence of biotechnology. We concentrateon pharmaceutical companies that are engaged inthe discovery, development, and commercializa-tion of biotechnology-based drugs that are placedinside the human body (in vivo). Focusing on thein-vivo biotechnology segment ensures a homo-geneous sample, and thus controls for varianceacross different industry segments. In addition, thepharmaceutical firms in the in-vivo biotechnologysegment are exposed to extensive and strict regu-latory oversight (i.e., Food and Drug Administra-tion [FDA] in the United States and the EuropeanMedicines Evaluation Agency [EMEA]), whichmandates that these firms disclose detailed data onnew drug development projects.

Underlying biotechnology are important scien-tific breakthroughs in genetic engineering (recom-binant DNA, 1973) and hybridoma technology(monoclonal antibodies, 1975), among others. Sub-sequently, the first new biotechnology drugsreached the market in the 1980s. The emergence ofbiotechnology, therefore, constitutes a radical pro-cess innovation in the drug discovery and develop-ment process for established pharmaceutical firms(Stuart et al., 1999). Responding to new technolog-ical developments has become critical as pharma-ceutical firms face tremendous pressures to inno-vate, as illustrated by the following trends (Hig-gins and Rodriguez, 2006): total R&D expendi-tures have grown from $6.8 billion in 1990 to$21.3 billion in 2000 (17% of sales); the aver-age new drug development costs have increasedfrom $231 million to $802 million between 1990and 2000, and average sales per patented drug hasfallen from $457 million in 1990 to $337 millionin 2001. Although the scientific expertise of incum-bent pharmaceutical firms rests on chemistry andchemical engineering, they nevertheless retainextensive knowledge of specific therapeutic areasand hold critical skills in clinical trial manage-ment and drug marketing and sales that have ledto extensive alliance formation with new biotech-nology firms (Rothaermel and Boeker, 2008).

Sample and data

To overcome a potential survivor bias, we iden-tified all pharmaceutical firms active in biotech-nology as of 1980 through a detailed study ofannual Standard Industrial Classification listingsand a comprehensive set of industry databases andpublications. Through this process, we identified43 global pharmaceutical companies, which wethen tracked forward over the 21-year study period,1980–2000. The number of firms is consistent withthe oligopolistic industry structure of the globalpharmaceutical industry, which is dominated by afew large companies that are active in proprietarydrug discovery and development.

The pharmaceutical industry has become moreconsolidated over the 21-year time period studiedthrough horizontal mergers among large pharma-ceutical companies. To account for this, we con-structed a detailed ‘family tree’ for each of these43 firms for the 1980–2000 time period. We usedmultiple industry publications to construct the fam-ily tree from 1980 onward, including Dun andBradstreet’s Who Owns Whom? and Standard &Poor’s Industry Surveys.1 Through this method, weidentified 13 horizontal mergers among the sam-ple firms. When a horizontal merger took place,we combined the past data of the two mergingfirms, and tracked the combined entity forward.We created an indicator variable for a firm thathad merged with or acquired another firm in thesample. This variable was not significant, however,in explaining time to drug approval or project ter-mination.

While the scientific breakthroughs underlyingbiotechnology were accomplished in the mid-1970s, we chose our study period to begin in 1980,because this year marks the start of commercializ-ing biotechnology. This can partly be explained byfour important events that occurred in 1980 (Stuartet al., 1999): (1) the successful IPO of Genentech,the first public biotechnology firm; (2) the passageof the Bayh-Dole act, which provides incentivesfor university patenting of inventions that resultedfrom federally funded research programs; (3) thedecision of the Supreme Court that life forms can

1 Dun & Bradstreet publish Who Owns Whom? annual worldwidedirectories that link companies to their corporate families andprovide key information regarding the corporate family tree.Standard & Poor’s Industry Surveys are published by McGraw-Hill, New York.



be patented; and (4) the Cohen-Boyer patent, dis-closing the recombinant DNA technology under-lying genetic engineering, was granted to StanfordUniversity (U.S. Patent 4,237,224), which licensedthis breakthrough technology widely for a nomi-nal fee.

The underlying data for analysis are at theproject level, and capture drug discovery devel-opment projects undertaken by pharmaceuticalcompanies in biotechnology and in their tradi-tional domains. These data were obtained fromLifecycle, a proprietary database maintained byIMS Health, an industry research firm specializ-ing in the pharmaceutical industry. Lifecycle iscommercially available and provides fine-graineddata on R&D projects covering a large numberof pharmaceutical firms globally. To obtain thesedata, IMS Health associates collect informationfrom governmental agencies, attend industry con-ferences, scan issued patents and scientific publi-cations, and maintain contacts with scientists andmanagers within the focal firms.

Lifecycle allows researchers to identifyprojects that are based on biotechnology. To ensurethe accuracy of these data and to prepare them forstatistical analysis, however, these data were codedby a researcher on our team holding a Doctor ofMedicine degree. Taken together, we were ableto collect data on 415 new biotechnology-baseddrug discovery and development projects com-menced by 43 pharmaceutical companies between1980 and 1998, while the observation of the out-comes of these projects ended in 2000. Missingdata for one firm regarding its portfolio of tradi-tional R&D projects led us to drop three records,which reduced our sample to 412 projects. Theseprojects were organized as follows: 122 (30%)were conducted alone by the pharmaceutical firms,235 (57%) were conducted in cooperation with abiotechnology firm, and 55 (13%) were initiatedby biotechnology firms after they were acquiredby the pharmaceutical firm.

The full set of 412 biotechnology projects hadcomplete data on project governance, but not allprojects had a project start or termination date. Wefilled in missing project dates through a detailedanalysis of data obtained from BioScan, Recap,and PharmaProjects. In addition, we trackedprojects through a fine-grained analysis of articlesavailable on Nexis-Lexis and company Securitiesand Exchange Commission (SEC) 10-K reports,

among other sources. This also allowed us to tri-angulate the accuracy of the start and termina-tion dates reported in the IMS data. In total, 385projects had accurate project start dates necessaryfor the event-history analysis (93% of initial sam-ple). Although the product development cycle inthis industry tends to be lengthy, the extended timeperiod covered by these data allows us to observeclear success or failure outcomes in 51 percent ofthese cases. It is important to note, however, thatthe utilized hazard rate estimation enables us totake advantage of all available information in thedata, including projects still ongoing at the end ofthe study period (Greene, 2003).

To obtain information on pharmaceutical firms’areas of focus using their traditional chemical-based method for lead compound identification anddevelopment, we obtained additional data on over3,500 projects primarily from IMS Lifecycle. Wecollected data pertaining to the lead company foreach project, project start dates, and therapeuticarea targets to provide information on pharma-ceutical firms’ traditional areas of R&D expertise.When relevant, IMS identifies patents that under-pin a particular project, allowing us to use patentownership to assign projects back to the originat-ing firm in the case of firms that had experienceda merger. We used project data from a seconddatabase, PharmaProjects, for six firms for whichIMS project records were sparse. Similar to IMSLifecycle, PharmaProjects is a publicly avail-able database containing detailed information onnew product development projects of pharmaceu-tical companies based on company questionnaires,annual reports, SEC and FDA filings, journals,investment reports, press releases, industry confer-ences, among others.

Furthermore, we collected historical pharmaceu-tical sales data for the firms in the sample. Bygenerating a sample of retail pharmacies and drugstores, obtaining their sales data and extrapolat-ing these data, IMS is able to report global salesfigures for leading pharmaceutical firms (Nerkarand Roberts, 2004). Sales data are broken downby the firm’s top 10–15 therapeutic categories,which allowed us to relate them to our project-leveldata, thereby providing a fine-grained measure ofa firm’s ability to exploit existing knowledge andcapabilities.

To obtain data on alliance experience by thepharmaceutical companies, we linked the sample



firms to alliance information obtained from vari-ous volumes of BioScan and from Recap, likelyto be the two most comprehensive publicly avail-able data sources documenting alliance activityin the biotechnology industry. Both sources arefairly consistent and accurate in reporting alliances(intersource reliability >0.90). These sources cata-loged alliance activity over the time period of ourstudy and also included alliances initiated in the1970s that allowed us to create lagged externalexperience measures.

Finally, we obtained patent data assigned bythe U.S. Patent and Trademark Office from 1975onward. We focused on patents obtained in theUnited States, because it is the largest market forbiotechnology worldwide, and thus it is almostcompulsory for firms to first patent in the UnitedStates. In addition, firms active in biotechnologyhave a strong incentive to patent, because intellec-tual property protection has been held up consis-tently in court and patenting is thus considered tobe a necessary activity to protect critical intellec-tual property (Albert et al., 1991).

Dependent variables

Drug approval. The primary dependent variableof this study is the hazard rate for drug approval.The hazard rate incorporates information onwhether the event occurred and project duration,proxied by the number of months from initiation ofthe project to market approval by either the FDA inthe United States, or the EMEA in Europe. About80 percent of the marketed drugs were introducedin the United States before or simultaneously withtheir introduction in Europe.

Project termination. We also model the hazardrate for project termination, that is, when a projectis discontinued. Indeed, project termination is amore common occurrence in this context andprovides important complementary insights intoproject outcomes. In this sample, 139 projects wereterminated before the end of the study period. Afocus on termination in addition to project suc-cess enables us to apply a competing hazard ratemodel that leverages all the available data. We thusovercome a sample selection bias that would beintroduced by only including successful projects.

Independent variables

External experience. We proxied external explo-ration experience by the percentage of R&Dalliances entered into by the pharmaceutical firmin the biotechnology field up to, but not includ-ing, the year of the focal project. We measuredexternal exploitation experience by the percentageof licensing and manufacturing agreements withinthe biotechnology field. Both external experiencemeasures controlled for the cumulative total num-ber of alliances that had been formed prior to theyear of the focal project. These proxies can alsobe viewed as the firms’ strategic orientation towardexploration and exploitation.

It is important to note that the external expe-rience data are based on all such alliances thatthe large pharmaceutical firms and their biotech-nology subsidiaries have entered into within thefield of biotechnology, and, thus, the experiencevariables are not limited to our sample of collabo-rative R&D projects. The average pharmaceuticalfirm had entered approximately 18 exploitation and27 exploration alliances prior to engaging in thefocal R&D project.

Internal experience. We developed measures of afirm’s internal exploration and exploitation expe-rience based on a firm’s past R&D efforts (explo-ration) and product sales (exploitation). Followingprior research (Macher and Boerner, 2006; Nerkarand Roberts, 2004), we characterized a firm’sexperience at the therapeutic class level. Akin toproduct market segments, we used 15 therapeuticclasses in accordance with the ‘Anatomical Ther-apeutic Classification’ (ATC) employed by IMSHealth and maintained by the World Health Orga-nization’s Collaborating Centre for Drug StatisticsMethodology. A firm’s internal exploration experi-ence is captured by the lagged, cumulative numberof R&D projects that a firm had initiated in a par-ticular therapeutic domain. Projects contribute to afirm’s internal exploration experience if the focalfirm is listed as the lead firm for the project. Toavoid double counting with solo biotech experi-ence, a control variable, the project must not beincluded in the focal sample of biotechnology-based R&D projects for this study. The firms inour sample accumulated experience by engagingon average in 10 internal R&D projects within agiven therapeutic area.

To capture internal exploitation experience, wecollected historical pharmaceutical sales data for



the firms in the sample, as detailed above. Weused the cumulative percentage of sales derivedfrom that therapeutic class as a proxy for internalexploitation experience. This measure captures theextent of a firm’s regulatory, marketing and salesexpertise, which are key downstream competenciesthat may be leveraged in a biotechnology-basedR&D project.

Control variables

Therapeutic area fixed effects. Because unob-served characteristics of specific therapeutic areasmay lead to differential project outcomes (Dan-zon, Nicholson, and Pereira, 2005; Macher andBoerner, 2006), we created four broad thera-peutic area dummies, which capture the primarytherapeutic area targeted by the focal project.These dummies covered 63 percent of the sam-ple, while the reference category represented the11 other therapeutic areas targeted by the remain-ing projects.

Medical indications. We proxied for the numberof medical indications or diseases that a projectcould potentially target. If the drug developmentproject concerns several indications, scientists areable to leverage more readily accessible knowl-edge, because multiple indications share underly-ing biological processes or target molecules thatare common to those indications. Thus, the sci-entists can draw on a greater number of researchmodels for testing and allow for greater knowledgetransfer across the different indications, therebyincreasing the odds of completing a project suc-cessfully. As such, 31 percent of projects in oursample targeted more than one indication.

Project year. To control for year effects and forright truncation, we included the year the projectwas initiated, with the expectation that projectsinitiated later in the study period are less likelyto lead to successful (or unsuccessful) outcomes.Some projects did not have project start dates sowe included the earliest date provided indicatingthe start of a particular stage of the developmentprocess. Such projects contributed to the first-stageselection model, but not to the second-stage event-history analysis.

Project patent protected. Patent protected projectsare more novel and, therefore, may be expected

to be more successful. Such projects are likelyto attract more resources and managerial atten-tion due to their expected, positive effect on firmperformance. We controlled for whether the under-lying project was protected by a U.S. and/or Euro-pean patent (dummy coded ‘1’). In addition tonovelty, whether a project is patent protected or notis also inversely related to the age of the project.When a project is young, potential patent claimstend to be less specifiable. At the time of our anal-ysis, every second project was patent protected.

Firm patents. To control for the pharmaceuticalfirms’ overall competence in biotechnology, weincluded firms’ patent data. Since many pharma-ceutical companies tend to patent in a wide rangeof areas, we attempted to eliminate unnecessarynoise in the patent data by focusing on techno-logical areas in which biotechnology patents wereemerging, such as U.S. patent class 435, Chem-istry: Molecular Biology and Microbiology. Weweighted each patent obtained by a pharmaceu-tical company in the relevant patent classes by itsforward citations to capture the quality of a firm’spatent portfolio (Trajtenberg, 1990). Thus, we cal-culated a cumulative variable for each pharmaceu-tical firm by summing the annual citation-weightedpatent counts up to the year before the initiationof the focal project.

Organization of new drug development. We codedfor the governance mode of each project, with‘1’ for solo development (solo project). We alsocoded for whether the project was initiated by abiotechnology firm after it had been acquired bya pharmaceutical firm in our sample (1 = biotechsubsidiary). The reference category consisted ofprojects that were conducted collaboratively.

Past solo experience. Prior direct engagement inthe focal activity may be an additional source oflearning that can affect project outcomes, and,thus, we control for this effect. The averagesample firm had initiated approximately six solobiotechnology-based projects before engaging inthe focal project.

Collaboration stage. Every successful projectgoes through the entire value chain from discov-ery to development, but collaborations with part-ners may be formed in either stage. Using this



distinction, all solo projects are, by definition, ini-tiated in the exploration or discovery phase, 60percent of the collaborative projects started withdrug discovery, while the remaining 40 percent ofcollaborations began in a later (clinical) develop-ment phase. We would expect that collaborativeprojects initiated in the preclinical stage (dummycoded ‘1’) would take longer to complete thanprojects where a partner joins during the clini-cal trial stage (dummy coded ‘0’). In contrast toprojects begun in the preclinical stage, later stagealliances are based on explicit and less ambiguousknowledge and skills, allowing for easier coordi-nation between partners that may increase projectspeed.

Inverse Mills ratio. Because the initial projectgovernance decision may be determined by unob-served firm competences, past choices, or unob-served characteristics of the project (Argyres andLiebeskind, 2002), we employed a first-stagemodel that allowed us to create a selection termthat corrects for endogeneity in the subsequentsecond-stage event history analysis (Heckman,1979). Unobserved factors that influence both thegovernance of the drug development project andits subsequent performance could otherwise leadto biased or spurious results (Hamilton and Nick-erson, 2003).

In the first stage, we applied a multinomial logitmodel to estimate the probability that a firm willchoose to either pursue a project solo, undertakeit through a biotechnology firm that has beenacquired, or pursue it collaboratively. The first-stage model consisted of our independent, controlvariables, and an instrument capturing the degreeof competitive intensity. This measure was equal tothe number of other pharmaceutical firms that hadinitiated projects in the same two-digit therapeuticclass up to the year prior to the start of thefocal project. The first-stage model returned anadjustment term, the inverse Mills ratio, whichwe then inserted in the second-stage event-historymodels to explicitly correct for self-selection.

Estimation procedure

Because the dependent variable combines the prob-ability of and the time to a focal event, we employevent history analysis. Drug development projectsare at risk to be successfully completed or ter-minated. Since these two outcomes are mutually

exclusive, we model them as a competing risk(Allison, 1984; Greene, 2003). We apply a modelthat estimates the competing hazard rates of eachproject making the transition to either success-ful completion or termination. This transition iscaptured by the instantaneous transition rate, r ,defined as

rk(t) = lim�t→0

Pr (t ≤ T < (t + �t), D = k|T ≥ t)

�t,

where k refers to one of the two mutually exclusiveoutcomes in D, describing the possible terminaloutcomes. The variable T measures the time spentat risk of making one of the two possible transi-tions, and the probability Pr describes the likeli-hood of experiencing a terminal transition duringthe time interval from t to (t + �t), conditionalon the project being at risk of making a transitionat time t (Tuma and Hannan, 1984). We specifyeach rate using the Cox (1972) proportional hazardmodel:

rk(t) = r0(t) exp(bX),

where X is a vector of covariates, assumed tohave a multiplicative effect on the baseline haz-ard, and b are the parameters to be estimated. Weestimated the Cox model with a robust specifica-tion, which adjusts the standard errors to allow forthe possibility of nonindependence across projectsinitiated by the same firm. A positive (negative)coefficient sign indicates a greater (lower) haz-ard of the focal event occurring (drug approvalor project termination, respectively), and thus canbe interpreted to mean that the variable of interestleads to a faster (slower) occurrence of the focalevent. Higher (lower) hazard rates, in turn, suggesta larger (smaller) number of such events within agiven time period.

RESULTS

Table 1 depicts the descriptive statistics of the vari-ables and the bivariate correlation matrix. A totalof 57 projects (10 conducted alone, 47 conductedin collaboration) were successful. When outcomesare averaged over the total number of spells (mea-sured in project months), the average proportionalhazard rate for a successful product development



Tabl

e1.

Des

crip

tive

stat

istic

san

dbi

vari

ate

corr

elat

ion

mat

rix

Mea

nSD

Min

Max

12

34

56

78

910

1112

13

1.M

edic

alin

dica

tion

s1.

480.

851.

005.

001.

002.

Pro

ject

year

1992

3.74

1980

2000

−0.1

21.

003.

Pro

ject

pate

ntpr

otec

ted

0.51

0.50

0.00

1.00

0.15

−0.0

31.

004.

Fir

mpa

tent

s18

10.2

918

84.8

50.

0081

63.0

0−0

.03

0.14

0.06

1.00

5.B

iote

chsu

bsid

iary

0.13

0.34

0.00

1.00

0.06

0.15

0.04

0.46

1.00

6.So

lopr

ojec

t0.

340.

470.

001.

00−0

.07

0.11

−0.3

10.

04−0

.01

1.00

7.P

ast

solo

expe

rien

ce6.

379.

450.

0042

.00

−0.0

70.

470.

070.

480.

490.

151.

008.

Col

labo

rati

onst

age

0.73

0.44

0.00

1.00

−0.1

30.

38−0

.23

−0.0

5−0

.04

0.43

0.21

1.00

9.In

vers

eM

ills

rati

o0.

710.

460.

082.

97−0

.06

0.12

−0.1

80.

100.

190.

350.

110.

201.

0010

.In

tern

alex

plor

atio

nex

peri

ence

10.3

010

.90

0.00

59.0

0−0

.07

0.33

0.14

0.07

−0.0

80.

000.

130.

05−0

.09

1.00

11.

Inte

rnal

expl

oita

tion

expe

rien

ce49

.00

93.7

90.

0063

1.10

0.02

0.09

0.09

0.14

0.04

−0.0

30.

150.

08−0

.08

0.28

1.00

12.

Ext

erna

lex

ploi

tati

onex

peri

ence

0.34

0.16

0.00

1.00

−0.0

40.

11−0

.10

0.17

0.24

−0.1

40.

21−0

.06

−0.0

1−0

.15

0.00

1.00

13.

Ext

erna

lex

plor

atio

nex

peri

ence

0.51

0.18

0.00

1.00

−0.0

30.

010.

05−0

.10

−0.1

50.

11−0

.04

0.06

−0.0

50.

130.

00−0

.56

1.00

N=4

12.

Not

e:Po

sitiv

ean

dne

gativ

eco

rrel

atio

nsgr

eate

rth

an0.

09ar

esi

gnifi

cant

atp<

0.01

leve

l.



outcome is 0.001 and the average time to drugapproval for successful projects is 100 months.The proportional hazard rate for project termina-tion is 0.002 and the average time to terminationfor failed projects is 59 months.

We present the results predicting new drugapproval and project termination in Table 2. Mod-els 1 and 2 are the respective baseline estimationscontaining fine-grained project-level and firm-levelcontrols, as well as the inverse Mills ratio, whichwas obtained in the first-stage Heckman selec-tion model described above. In Models 3 and4, we entered the internal and external experi-ence variables, distinguished by the exploration-exploitation dimension, in order to test the directeffect hypotheses (Hypotheses 1 and 2). In Models5 and 6, we entered the interaction effects to testthe contingency hypotheses (Hypotheses 3 and 4).

We proposed opposing direct effects of exter-nal experience on R&D project performance: apositive effect for external exploitation experience(Hypothesis 1) and a negative one for externalexploration experience (Hypothesis 2). We foundsupport for Hypothesis 1 in the analysis of projectsuccess, because the effect of external exploita-tion experience is positive and significant (Model3, p < 0.05). The results obtained in Model 3also provide support for Hypothesis 2, because theeffect of external exploration experience is nega-tive and significant (p < 0.05).

Because project terminations are an importantcomplementary outcome to project success andallow for a more complete understanding of R&Dperformance, an analysis of the factors that affectproject termination rates can provide additionalinsights. A comparison of Models 3 and 4 revealsthat the factors that increase the rate of project suc-cess do not necessarily decrease the rate of projecttermination. Instead, subtle but important differ-ences appear. Namely, greater external exploitationexperience increases the hazard rate of project ter-mination by 22 percent with a one standard devi-ation increase in external experience (Model 4,p < 0.05). It appears that firms may be able toleverage their exploitation experience to improveproduct approval rates, but at the cost of increasingproject termination rates.

We further hypothesized that the positive effectof external exploitation experience on R&D projectperformance is enhanced in the presence of inter-nal exploration experience (Hypothesis 3), whilethe negative effect of external exploration on R&D

project performance is accentuated when combinedwith internal exploitation experience (Hypothe-sis 4). We present the results for the interac-tion hypotheses in Models 5 and 6. We find thatinternal exploration experience positively moder-ates the impact of external exploitation experience,because the interaction term is positive and signifi-cant (Model 5, p < 0.01). This indicates that firmswith greater internal exploration experience in thefocal therapeutic area are better able to leveragethe benefits of external exploitation experience.

The results presented in Model 5 also providesupport for Hypothesis 4 (p < 0.001), indicatingthat as internal exploitation experience grows, theimpact of external exploration experience on thehazard rate for product approval grows increas-ingly negative. This implies that firms with greaterinternal exploitation experience are at a distinctdisadvantage in mitigating the difficulties associ-ated with leveraging external exploration experi-ence. This finding seems to reflect the challengesof organizational learning in a collaborative con-text when knowledge is highly novel and tacit.

The analysis of Model 6 examines the compet-ing risk of project termination. We found that theinteraction effects for project termination were inthe opposite direction to their effects for productapproval. Firms with greater internal explorationexperience are able to leverage external exploita-tion experience to decrease the rate of project ter-mination (p < 0.05). On the other hand, increas-ing internal exploitation experience coupled withgreater external exploration experience increasesthe rate of project termination (p < 0.10). Thispattern of results reinforces the notion that R&Dperformance is enhanced in the case of internalexploration experience while it is dampened withincreasing internal exploitation experience.

To gain an intuitive understanding of the resultsobtained, we graphically display them in Figures 1and 2. In Figure 1, we depict the results forHypotheses 1 and 3. The bottom line of the graphprovides a baseline estimation for the cumulativehazard rate of product approval derived from theindependent variables evaluated at their mean. Themiddle hazard rate line indicates that a one stan-dard deviation increase in external exploitationexperience results in a 36 percent increase in thehazard rate for time to drug approval comparedto firms with an average level of such experience.The hazard rate for drug approval improves evenfurther when firms combine internal exploration



Tabl

e2.

Pred

ictin

gne

wdr

ugap

prov

alan

dpr

ojec

tte

rmin

atio

n

Mod

el1

Dru

gap

prov

al

Mod

el2

Proj

ect

term

inat

ion

Mod

el3

Dru

gap

prov

al

Mod

el4

Proj

ect

term

inat

ion

Mod

el5

Dru

gap

prov

al

Mod

el6

Proj

ect

term

inat

ion

The

rape

utic

area

fixed

effe

cts

incl

uded

incl

uded

incl

uded

incl

uded

incl

uded

incl

uded

Med

ical

indi

cati

ons

0.51

23∗∗

∗−0

.158

3∗0.

4854

∗∗∗

−0.1

507∗

0.47

73∗∗

∗−0

.146

5∗

(0.1

484)

(0.0

913)

(0.1

491)

(0.0

851)

(0.1

430)

(0.0

847)

Pro

ject

year

−0.0

510

0.05

40†

−0.0

778

0.03

23−0

.080

30.

0429

(0.0

869)

(0.0

383)

(0.0

896)

(0.0

429)

(0.0

887)

(0.0

464)

Pro

ject

pate

ntpr

otec

ted

1.95

79∗∗

∗−2

.124

1∗∗∗

2.16

52∗∗

∗−2

.150

1∗∗∗

2.05

88∗∗

∗−2

.107

3∗∗∗

(0.4

274)

(0.2

740)

(0.4

502)

(0.2

883)

(0.4

522)

(0.2

880)

Fir

mpa

tent

s0.

0001

−1.6

7E-0

50.

0001

−1.6

1E-0

50.

0001

−1.1

2E-0

6(0

.000

1)(0

.000

1)(0

.000

1)(0

.000

1)(0

.000

1)(0

.000

1)B

iote

chsu

bsid

iary

−1.0

837∗∗

0.02

93−1

.412

6∗∗∗

0.00

16−1

.424

9∗∗∗

−0.0

628

(0.3

808)

(0.4

565)

(0.3

845)

(0.4

315)

(0.4

424)

(0.4

384)

Solo

proj

ect

1.06

68†

0.36

51†

1.25

10†

0.40

72†

0.84

830.

4155

†(0

.873

1)(0

.256

7)(0

.832

3)(0

.281

2)(0

.910

0)(0

.298

6)P

ast

solo

expe

rien

ce0.

1043

−0.1

470

0.18

25−0

.171

9†0.

2310

−0.1

347

(0.2

645)

(0.1

440)

(0.2

878)

(0.1

365)

(0.2

788)

(0.1

371)

Col

labo

rati

onst

age

−1.3

455∗∗

0.10

07−1

.497

9∗∗0.

1631

−1.6

518∗∗

0.11

63(0

.571

9)(0

.286

0)(0

.539

6)(0

.285

1)(0

.533

5)(0

.302

4)In

vers

eM

ills

rati

o0.

0634

0.10

290.

1678

0.11

600.

1125

0.15

52(0

.349

3)(0

.243

8)(0

.315

1)(0

.236

6)(0

.343

1)(0

.236

9)



Tabl

e2.

(Con

tinu

ed)

Mod

el1

Dru

gap

prov

al

Mod

el2

Proj

ect

term

inat

ion

Mod

el3

Dru

gap

prov

al

Mod

el4

Proj

ect

term

inat

ion

Mod

el5

Dru

gap

prov

al

Mod

el6

Proj

ect

term

inat

ion

Inte

rnal

expl

orat

ion

−0.1

187

0.13

71†

−0.1

647

0.01

39ex

peri

ence

(0.2

562)

(0.0

940)

(0.2

730)

(0.1

175)

Inte

rnal

expl

oita

tion

−0.0

139

0.08

52−0

.003

10.

0899

expe

rien

ce(0

.144

8)(0

.117

3)(0

.134

0)(0

.128

9)E

xter

nal

expl

oita

tion

0.31

19∗

0.20

24∗

0.56

54∗∗

0.09

67ex

peri

ence

(0.1

624)

(0.1

187)

(0.2

159)

(0.0

963)

Ext

erna

lex

plor

atio

n−0

.264

3∗0.

1049

−0.2

786∗

0.05

41ex

peri

ence

(0.1

565)

(0.1

011)

(0.1

597)

(0.0

850)

Inte

rnal

expl

orat

ion

×0.

7995

∗∗−0

.248

2∗

exte

rnal

expl

oita

tion

(0.2

768)

(0.1

268)

Inte

rnal

expl

oita

tion

×−0

.275

7∗∗∗

0.10

62†

exte

rnal

expl

orat

ion

(0.0

822)

(0.0

671)

Spel

ls62

,268

62,2

6862

,268

62,2

6862

,268

62,2

68L

og-l

ikel

ihoo

d−2

63.4

0−7

59.2

2−2

59.4

6−7

56.8

5−2

54.8

3−7

53.3

2C

hisq

uare

281.

46∗∗

∗21

4.19

∗∗∗

489.

65∗∗

∗20

9.25

∗∗∗

864.

22∗∗

∗48

5.41

∗∗∗

Rob

ust

stan

dard

erro

rsar

ein

pare

nthe

ses.

†p

<.1

0;∗p

<.0

5;∗∗

p<

.01;

∗∗∗p

<.0

01.



0

0 50 100 150

Baseline

0.5

0.4

0.3

0.2

0.1

Analysis time

Cox Proportional Hazard Rates: product approvalInt exploration x extexploitation

+1 SD ext exploitationexperience

Figure 1. Effect of external exploitation and internal exploration on cumulative hazard for product approval. Thisfigure is available in color online at www.interscience.wiley.com/journal/smj

0

0.15

Cum

ulat

ive

haza

rd

0 50 100 150

Baseline

0.1

0.05

Analysis time

Cox Proportional Hazard Rates: product approval

+ 1 SD ext explorationexperience

Int exploitation x extexploration

Figure 2. Effect of external exploration and internal exploitation on cumulative hazard for product approval. Thisfigure is available in color online at www.interscience.wiley.com/journal/smj

experience with external exploitation experience(top line). A one standard deviation increase inboth internal exploration experience and externalexploitation experience more than doubles the haz-ard rate for time to drug approval compared to thebaseline model.

Figure 2 displays the results for Hypotheses2 and 4. With increasing external explorationexperience, firms experience a significant decreasein the hazard rate for drug approval: a one standarddeviation increase in external exploration experi-ence leads to a 23 percent decline in the hazard



rate for drug approval (middle line). The haz-ard rate for drug approval declines even furtherwhen firms combine internal exploitation expe-rience with external exploration experience (bot-tom line). A one standard deviation increase inboth internal exploitation experience and externalexploration experience results in an additional 24percent decrease in the rate of drug approvals com-pared to the baseline.

The results also reveal that a number of project-level controls are significant predictors of productapproval and project termination rates. The resultsobtained in Models 5 and 6 indicate that projectswith more indications and those with patent pro-tection have significantly higher hazard rates forapproval and a corresponding lower hazard rate forproject termination. Both controls capture aspectsof project quality, which implies that these projectsare likely to attract more scientific, managerial,and financial resources that, in turn, aid in speed-ing these high-return projects to completion andcommensurately decrease their potential for termi-nation. The results also indicate that collaborativeprojects undertaken at a later stage of develop-ment are faster to market than projects initiatedin the preclinical stage. Because the pharmaceuti-cal firm can choose to collaborate after a projecthas proven to be viable by entering into clinicaltrials, accounting for collaboration stage appearsto capture underlying project quality. Projects thatwere begun by biotechnology firms that had beenacquired by large pharmaceutical firms were asso-ciated with decreased project approval rates, whileconducting a project alone increased the hazardrate for project termination.

Robustness checks

To confirm the robustness of our results, wereanalyzed the data using the cumulative num-ber of exploratory and exploitative alliances inplace of the ratio-based experienced measures(McNamara and Baden-Fuller, 2007; Rothaermel,2001; Rothaermel and Deeds, 2004). Firms tendto engage in both exploratory and exploitativealliances as indicated by their high intercorrela-tion when applying simple count measures (r =0.84). However, because we track firms longitu-dinally, this correlation is also capturing the factthat exploratory and exploitative experience growsover time. While the high correlation is not sur-prising, it does introduce a limitation when using

cumulative counts to assess alliance experience.A high correlation between independent variablesincreases standard errors and results in a greaterlikelihood of making a type II error or accept-ing a null hypothesis of no effect when it shouldbe rejected. High correlations between our inde-pendent variables may also make it more diffi-cult to interpret the interaction terms that are cre-ated from the predictors (exploration and exploita-tion alliance activity) and our moderator variables(internal exploration and exploitation ability).

With these concerns in mind, we continue tofind support for Hypothesis 1 (positive impact ofalliance exploitation experience on project out-comes) and Hypothesis 2 (negative impact ofalliance exploration experience) in the model pre-dicting project approvals. Regarding our interac-tion effects, all the signs are in the expected direc-tion for both the models of project approval andtermination. The interaction effects are significantwith respect to Hypothesis 3 (combining internalexploration and external exploitation), but not forHypothesis 4. Overall, we found that our resultsare robust to a different definition of our predic-tor variables. The model we present in the paperhas the advantage of having lower correlationsbetween exploration and exploitation thereby facil-itating interpretation of our results.

We included potentially reinforcing interactionsbetween internal and external experience in theexploration and exploitation domains, respectively.Neither interactions between internal and externalexploration experience nor between internal andexternal exploitation experience were statisticallysignificant.

We used data from company annual reports,Lexis-Nexis, and Recombinant Capital, to recon-struct sales figures for 47 percent (27/57) of ourproduct approvals beginning in 1986. With thiscaveat in mind, we found that 33 percent (9/27)of the approved drugs for which we have salesdata did reach the milestone of generating over$1 billion in sales in a given year (this is usuallythe cutoff used to define a blockbuster drug). Thosewith the highest levels of sales in their therapeuticcategory were among the first to market. In ourdata, the first entry had a probability of 66 percentto be the leading drug in terms of sales. No druglater than being third to market was able to capturethe top sales spot. Time to product approval andproduct sales tend to be endogenous.



Because there is no standard treatment for bothproblems of endogeneity and unobserved hetero-geneity in hazard rate models, we undertook anumber of additional analyses that provide anindication of the robustness of our results. Fol-lowing Dolton, Makepeace, and Treble (1994),we reran our analyses using an accelerated fail-ure time model with treatment effects. The resultsof the analyses were consistent with the hazardrate results reported in Model 5. For the modelpredicting project termination, the results differedfrom Model 6 to the extent that the interactioneffect between internal exploration and externalexploitation experience became marginally signif-icant (p < 0.10).

We also ran hazard rate models akin to random-effects models in linear regression to capture unob-served heterogeneity. Using a shared frailty modelin which an additional estimated parameter witha known (gamma) distribution enters multiplica-tively on the hazard rate for each firm, we foundthe same pattern of results for our main indepen-dent and interaction variables in models of bothproject approval and termination. Since the param-eter did not reach statistical significance in eithermodel, we have further confidence that unobservedfirm-level heterogeneity did not materially influ-ence the results.

DISCUSSION

The emergence of biotechnology represents a rad-ical exogenous change in terms of drug discoveryand development for pharmaceutical companies.A focus on these incumbents, in turn, providesa natural laboratory for researchers to investigatewhether experience gained in collaboration is aneffective means by which to build a new set ofproduct development capabilities. Thus far, accu-mulating empirical research has generally linkedalliances to positive firm-level outcomes; however,the few studies that examine knowledge transfermore directly indicate more equivocal outcomes(Hoang and Rothaermel, 2005; Reuer and Zollo,2005; Zollo et al., 2002).

To better understand the role of alliance expe-rience, our study identified R&D alliances in dif-ferent parts of the value chain and assessed theimpact of different types of external experienceon R&D project-level performance. Recognizingexploration and exploitation alliances explicitly is

critical, because it highlights how leveraging exter-nal experience is systematically related to charac-teristics of the knowledge exchanged, demands forknowledge integration, and differences in organi-zational contexts between the partners. Our find-ing that external exploration experience leads topoorer R&D project outcomes suggests that theseimpediments may overwhelm firm attempts tolearn. In contrast, when firms engage in exploita-tion alliances, which involve the transfer of lessambiguous knowledge and require lower knowl-edge integration between partners, they are betterable to leverage external experience in order toimprove R&D project performance.

Our findings indicate support for the notion thatnew product development alliances are best aimedat gaining access to partners’ resources and capa-bilities to reap gains from specialization (Grantand Baden-Fuller, 2004; Teece, 1992). Exploitationalliances seek to leverage existing complementari-ties between partners. In this context, pharmaceu-tical firms hold important downstream capabilitiesthat their partners often lack, including manufac-turing, legal expertise, and sales, distribution, andmarketing. Leveraging such specialized capabil-ities in alliances has been an important strate-gic response to technological innovations intro-duced by numerous, smaller biotechnology-basedentrants, and is facilitated by the highly struc-tured and sequential nature of the pharmaceuti-cal new product development process that allowsfor manageable ‘hand-offs’ between biotechnologyand pharmaceutical firms.

In addition to the benefits of exploitationalliances, our results also suggest that certainkinds of internal capabilities play an importantrole in fully leveraging external activities. By tak-ing a detailed look at the internal exploration andexploitation activities of our focal firms, we iden-tified an important, additional source of comple-mentarities. Specifically, a firm’s internal explo-ration experience, as indicated by greater R&Dactivity in a given therapeutic area, played akey role in enhancing the benefits that accruedfrom engaging in exploitation alliances. The roleof internal exploration experience likely increasesa firm’s absorptive capacity, thereby allowing itto not only initiate more promising new R&Dprojects but also to recognize and exploit the exter-nal opportunities afforded by new technologicaldevelopments.



This interpretation of our results is strengthenedby the findings presented in an article pertain-ing to innovation in the Spanish manufacturingsector (Beneito, 2006), where significant innova-tions were initiated in-house (through exploration),while contractual (exploitation) alliances providedaccess to knowledge that was more incremental innature. Our findings at the project level of analysisalso resonate with recent research on antecedentsto innovation at the firm level of analysis, whichfound a positive reinforcing effect of firms’ R&Dexpenditures and strategic alliances when predict-ing the number of new patents assigned in biotech-nology to pharmaceutical firms (Rothaermel andHess, 2007). However, we differentiate betweendifferent types of R&D alliances and their effectson project-level outcomes, which allowed for moresubtle results to emerge pertaining to the interac-tion between internal and external capabilities.

Internal exploitation experience appears to offerno complementary benefits to a firm’s externalexploratory experience. Indeed, the negative inter-action effect indicates that the challenges of lever-aging this type of external experience are evengreater for firms that have gained more of theirhistorical sales revenues from the focal therapeu-tic area. Firms with greater internal exploitationexperience who simultaneously pursue explorationalliances may lack the internal knowledge neces-sary to recognize the most viable early-stage part-nership options. This leads to a scenario wherenegative learning may occur. Their extensive prod-uct market experience in a particular therapeuticdomain may also come with greater complemen-tary assets and the potential for scale benefits;in order to better exploit both factors, firms maybe less stringent in their selection of collabora-tive projects and thus face higher failure projectrates. Taken together, this may be indicative ofa situation where core competencies in internalexploitation can turn to core rigidities (Leonard-Barton, 1992) when combining them with externalexploration experience.

Recently, the pursuit of cross exploration-exploitation activities has attracted significantattention in research on ambidexterity (O’Reillyand Tushman, 2008; Raisch and Birkinshaw,2008). This is because pursuing exploration andexploitation simultaneously creates a significanttension due to divergent goals and different timehorizons (Levinthal and March, 1993; March

1991). Yet firms that are able to balance and rec-oncile this tension through technology sourcingappear to achieve greater innovative and finan-cial performance based on a study of the U.S.manufacturing sector (Rothaermel and Alexandre,2009). When applying an ambidexterity lens toR&D projects in the biopharmaceutical industry,however, more nuances emerge because we areable to study each individual project rather thanoutcomes at the more aggregate and theoreticallydistant firm level. We find that ambidexterity isbeneficial when firms focus on internal explorationcombined with external exploitation. In contrast,ambidexterity can have negative performance con-sequences when internal exploitation is combinedwith external exploration.

Managerial implications

The overall pattern of results suggests that theleveraging of internal and external experience ischallenging. It may even have deleterious effectsin part because the partnering process and the valuechain activities on which it is overlaid operate withdistinctive logics and time horizons that can onlybe effectively combined in certain ways. In newproduct development, internal and external expe-riences appear to be reinforcing if the focal firmhas extensive internal exploration experience andexploitative collaborative experience. Given theshorter time horizons of strategic alliances com-pared to the R&D timeline, opportunities to learnfrom partners may be best achieved in relativelyshort-term, codifiable exploitation alliances.

In contrast, the highly uncertain and tacit natureof early stage pharmaceutical R&D may favorinternal organization rather than collaborativestructures. Given the difficulty of learning in highlyuncertain exploratory alliances, internal experiencedoes not appear to be synergistic when it is basedon expertise in downstream capabilities that maycontribute little to a firm’s ability to identify andleverage external opportunities. To the extent thatfirms have invested to enhance learning and knowl-edge transfer, they typically focus on transferringknowledge within a particular stage of the productdevelopment process or across contiguous stages.The current use of structures such as multidisci-plinary project teams may be insufficient to fos-tering complex learning across project and firmboundaries. Managers may need to make more sig-nificant investments in terms of resource allocation



and organizational structure to build internal explo-ration competences before attempting to leverageexternal exploitation experience.

Limitations and future research

Our study contains a number of limitations thatopen the door for future research. Although clearlyimportant, performance metrics such as time toproduct approval and project duration may notfully reflect how firms leverage their externalexploration experience. Knowledge conversionmay be reflected in firms’ publication and sub-sequent patent output, which are important pre-cursors to products in the market. This is aninteresting proposition that should be taken up infuture research. From a strategy perspective, how-ever, bringing new products to market in a timelyfashion is imperative to gaining and sustainingcompetitive advantage in this industry (Grabowskiand Vernon, 1990; Graham and Higgins, 2006;Roberts, 1999). A future study that links internaland external experience to economic performance,such as sales revenue generated from each productintroduction (in the spirit of Nerkar and Roberts,2004), would be welcome since many pharmaceu-tical companies compete on the introduction ofblockbuster drugs.

In addition, although time to project comple-tion is clearly an important performance metric,especially for new drug development in the phar-maceutical industry, we do need to caution thatspeed represents only one dimension of perfor-mance. One can argue that an excessive focuson speed to secure a patent-protected first moveradvantage could lead to decisions that might com-promise the safety of a product. For example,additional safety and side effect studies are notconducted once a minimum acceptable standard isreached for FDA approval, or critical data are notinvestigated in sufficient depth. These are possibleexplanations for the dramatic drug recalls, such asthat of Merck’s Vioxx, which we have witnessed inthe recent past. A future study, therefore, is clearlyneeded to illuminate the trade-offs between qualityand speed in new product development.

We are also in need of a deeper understand-ing of the factors that determine project termi-nation. While Reuer and Zollo (2005) sought todetermine whether terminated alliances had beensuccessful or had failed, we were not able to dis-criminate between appropriate and inappropriate

project terminations. Indeed, project terminationsmight be considered a success if the firm was ableto stop committing resources to a low-potentialproject. Or, managers may find a failed projectvaluable if the project was undertaken initially asan option on an emerging technological area (Zolloet al., 2002). Unsuccessful terminations, in con-trast, would involve prematurely ending a poten-tially viable project, and thus committing a typeII error. To allow for a wider array of outcomes,future research might usefully combine subjectiveand objective assessments of project outcomes.

CONCLUSION

Our empirical findings provide some initial bound-ary conditions for the dynamic capabilities per-spective when applied to an industry setting thatexperienced radical technological change. Thepharmaceutical industry seems particularly wellsuited to test theoretical conjectures derived fromthis perspective due to the importance of strate-gic alliances and new product development in thisindustry setting. Indeed, Eisenhardt and Martin(2000) view those organizational activities to be atthe core of the dynamic capabilities perspective:‘Dynamic capabilities include well-known orga-nizational and strategic processes like alliancingand product development whose strategic valuelies in their ability to manipulate resources intovalue-creating strategies’ (Eisenhardt and Martin,2000: 118). Moreover, one key theoretical tenetof the dynamic capabilities perspective is that theastute combination of internal and external com-petencies can enable firms not only to adapt, butalso to take advantage of rapidly changing envi-ronments: ‘We refer to this as the “dynamic capa-bilities” approach in order to stress exploitingexisting internal and external firm-specific compe-tences to address changing environments’ (Teeceet al., 1997: 510). Combining internal and exter-nal competencies allows for a dynamic strategicfit, which has been shown to improve firm perfor-mance (Zajac, Kraatz, and Bresser, 2000).

While the combination of external and internalcompetencies is critical to address rapidly chang-ing environments, it is only certain combinationsthat are beneficial to performance, while others canactually be harmful. In particular, we find that acombination of internal exploration with externalexploitation improved R&D project performance,



while a combination of internal exploitation andexternal exploration reduced R&D project perfor-mance. It appears that internal exploration com-petencies lay the necessary foundation to leverageexternal experience.

ACKNOWLEDGEMENTS

We thank Associate Editor Rudi Bresser, theanonymous reviewers, Jeffrey Edwards, AntonGueth (former director, Office of Alliance Man-agement, Eli Lilly), Cheryl Gaimon, Stuart Gra-ham, Andrew Hess, Matthew Higgins, SteliosKavadias, Christopher Liu, Christoph Loch, DavidLuvison (of Alliance Vista), Fiona Murray, AtulNerkar, Jackson Nickerson, Joanne Oxley, GulruOzkan, Samuel Ransbotham, Sancha Salgueiro,Scott Shane, Scott Stern, Toby Stuart, and DanielTzabbar for helpful comments and suggestions.We thank Faisal Aftab, Shanti Agung, MeganMenkveld, Padmaja Jasti, Deboleena Sengupta,and Sreten Simac, M.D., for research assistancein data collection and data coding. We thank Car-olyn Hughes, David Ewbank, Ben Manning, andSarah Rickwood (all of IMS Health), and MarkEdwards of Deloitte Recap LLC for their inputand for making their various databases available tous. We thank Lois Gast for effective managementof this manuscript, and Deborah Gray for expertcopyediting.

Prior versions of this paper were presented atthe Academy of Management Meetings, AtlantaCompetitive Advantage Conference (ACAC), Lon-don Business School SIM Seminar Series, Univer-sity of Maryland Entrepreneurship Conference, theStrategic Management Society Conference, and theToronto Alliance Edge Conference.

Rothaermel acknowledges support for thisresearch from the National Science Foundation(NSF SES 0545544) and the Sloan Foundation.Rothaermel is an affiliate of the Sloan Biotechnol-ogy Industry Center at the University of Maryland.All opinions expressed as well as all errors andomissions are our own

REFERENCES

Albert MB, Avery D, Narin F, McAllister P. 1991.Direct validation of citation counts as indicators ofindustrially important patents. Research Policy 20(3):251–259.

Allison PD. 1984. Event History Analysis. Regression forLongitudinal Data . Sage Publications: Newbury Park,CA.

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

Argyres NS, Liebeskind JP. 2002. Governance insepara-bility and the evolution of US biotechnology indus-try. Journal of Economic Behavior and Organization47(2): 197–219.

Arora A, Gambardella A. 1990. Complementarity andexternal linkages: the strategies of the large firms inbiotechnology. Journal of Industrial Economics 38(4):361–379.

Arora A, Gambardella A. 1994. The changing technologyof technological change: general and abstractknowledge and the division of innovative labor.Research Policy 23(5): 523–532.

Azoulay P. 2004. Capturing knowledge within and acrossfirm boundaries: evidence from clinical development.American Economic Review 94(5): 1591–1612.

Baum JAC. Calabrese T, Silverman BS. 2000. Don’tgo it alone: alliance network composition andstartups’ performance in Canadian biotechnology.Strategic Management Journal , March Special Issue21: 267–294.

Beneito P. 2006. The innovative performance of in-houseand contracted R&D in terms of patents and utilitymodels. Research Policy 35(4): 502–517.

Benner MJ, Tushman ML. 2003. Exploitation, explo-ration, and process management: the productivitydilemma revisited. Academy of Management Review28(2): 238–256.

Bettis RA, Bradley SP, Hamel G. 1992. Outsourcing andindustrial decline. Academy of Management Executive6(1): 7–22.

Bresser RKF, Heuskel D, Nixon RD. 2000. The decon-struction of value chains: practical and conceptualchallenges. In Winning Strategies in a Deconstruct-ing World (Wiley Strategic Management Series),Bresser RKF, Hitt MA, Nixon RD, Heuskel D (eds).Wiley: Chicester, UK.; 1–21.

Chang YC. 2003. Benefits of co-operation on innovativeperformance: evidence from integrated circuits andbiotechnology firms in the UK and Taiwan. R&DManagement 33(4): 425–437.

Cohen MD, Bacdayan P. 1994. Organizational routinesare stored as procedural memory: evidence from alaboratory study. Organization Science 5(4): 554–568.

Cohen WM, Levinthal DA. 1989. Innovation and learn-ing: the two faces of R&D. Economic Journal99(397): 569–596.

Cohen WM, Levinthal DA. 1990. Absorptive capacity:a new perspective on learning and innovation.Administrative Science Quarterly 35(1): 128–152.

Cox DR. 1972. Regression models and life tables.Journal of the Royal Statistical Society Series B 34:187–202.

Danzon PM, Nicholson S, Pereira NS. 2005. Productivityin pharmaceutical-biotechnology R&D: the role ofexperience and alliances. Journal of Health Economics24(2): 317–339.



Dolton PJ, Makepeace GH, Treble JG. 1994. The youthtraining scheme and the school–to–work transi-tion. Oxford Economic Papers–New Series 46(4):629–657.

Dussauge P, Garrette B, Mitchell W. 2000. Learningfrom competing partners: outcomes and durations ofscale and link alliances in Europe, North America andAsia. Strategic Management Journal 21(2): 99–126.

Eisenhardt KM, Martin JA. 2000. Dynamic capabilities:what are they? Strategic Management Journal ,October–November Special Issue 21: 1105–1121.

Fisher L. 1995. Drug makers will settle patent fight. NewYork Times , 6 January.

Garud R, Nayyar PR. 1994. Transformative capac-ity: continual structuring by intertemporal technol-ogy transfer. Strategic Management Journal 15(5):365–385.

Gick ML, Holyoak KJ. 1987. The cognitive basisof knowledge transfer. In Transfer of Learning ,Cormier SM, Hagman JD (eds). Academic Press: NewYork; 9–46.

Grabowski H, Vernon J. 1990. A new look at the returnsand risks in pharmaceutical R&D. ManagementScience 36(7): 804–821.

Graham JH, Higgins MJ. 2006. The impact of patentingon new product introductions in the pharmaceuticalindustry. Paper presented at 6th Annual Roundtablefor Engineering Entrepreneurship Research (REER),Georgia Institute of Technology.

Grant RM, Baden-Fuller C. 2004. A knowledge access-ing theory of strategic alliances. Journal of Manage-ment Studies 41(1): 61–84.

Greene WH. 2003. Econometric Analysis (5th edn).Prentice Hall: Upper Saddle River, NJ.

Hamel G. 1991. Competition for competence andinter-partner learning within international strategicalliances. Strategic Management Journal , SummerSpecial Issue 12: 83–103.

Hamilton BH, Nickerson JA. 2003. Correcting forendogeneity in strategic management research.Strategic Organization 1(1): 51–78.

Heckman JJ. 1979. Sample selection bias as aspecification error. Econometrica 47(1): 153–161.

Helfat CE. 1997. Know-how and asset complementarityand dynamic capability accumulation: the case ofR&D. Strategic Management Journal 18(5): 339–360.

Higgins MJ, Rodriguez D. 2006. The outsourcing ofR&D through acquisitions in the pharmaceuticalindustry. Journal of Financial Economics 80(2):351–383.

Hill CWL, Rothaermel FT. 2003. The performance ofincumbent firms in the face of radical technologicalinnovation. Academy of Management Review 28(2):257–274.

Hoang H, Rothaermel FT. 2005. The effect of generaland partner-specific alliance experience on joint R&Dproject performance. Academy of Management Journal48(2): 332–345.

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.

Kaplan S, Murray F, Henderson R. 2003. Discontinuitiesand senior management: assessing the role ofrecognition in pharmaceutical firm response tobiotechnology. Industrial and Corporate Change12(2): 203–233.

Kogut B, Zander U. 1992. Knowledge of the firm,combinative capabilities, and the replication oftechnology. Organization Science 3(3): 383–397.

Koza MP, Lewin AY. 1998. The co-evolution of strategicalliances. Organization Science 9(3): 255–264.

Lane PJ, Lubatkin MH. 1998. Relative absorptivecapacity and interorganizational learning. StrategicManagement Journal 19(5): 461–477.

Laursen K, Salter A. 2006. Open for innovation:the role of openness in explaining innovationperformance among U.K. manufacturing firms.Strategic Management Journal 27(2): 131–150.

Lavie D, Rosenkopf L. 2006. Balancing explorationand exploitation in alliance formation. Academy ofManagement Journal 49(4): 797–818.

Lei D, Hitt MA. 1995. Strategic restructuring andoutsourcing: the effect of mergers and acquisitions andLBOs on building firm skills and capabilities. Journalof Management 21(5): 835–859.

Leonard–Barton D. 1992. Core capabilities and corerigidities: a paradox in managing new product devel-opment. Strategic Management Journal , Summer Spe-cial Issue 13: 111–125.

Levinthal DA, March JG. 1993. The myopia of learning.Strategic Management Journal , Winter Special Issue14: 95–112.

Levitt B, March JG. 1988. Organizational learning.Annual Review of Sociology 14: 319–340.

Lieberman MB, Montgomery DB. 1988. First moveradvantages. Strategic Management Journal , SummerSpecial Issue 9: 41–58.

Lubatkin M, Florin J, Lane P. 2001. Learning togetherand apart: a model of reciprocal interfirm learning.Human Relations 54(10): 1353–1382.

Macher JT, Boerner CS. 2006. Experience and scaleand scope economies: trade-offs and performance indevelopment. Strategic Management Journal 27(9):845–865.

March JG. 1991. Exploration and exploitation inorganizational learning. Organization Science 2(1):71–88.

McDaniel CS. 1994. Biotechnology trade secrets, uni-versity research and international perspective. Whitepaper, Conley Rose International Property Law. Avail-able at: http://conleyrose.com/TRADESEC.pdf.

McNamara P, Baden-Fuller C. 2007. Shareholder returnsand the exploration-exploitation dilemma: R&Dannouncements by biotechnology firms. ResearchPolicy 36(4): 548–565.

Mora-Valentin EM, Montoro-Sanchez A, Guerras-MartinLA. 2004. Determining factors in the success of R&Dcooperative agreements between firms and researchorganizations. Research Policy 33(1): 17–40.

Mowery DC. 1983. Innovation, market structure, andgovernment policy in the American semiconductorelectronics industry: a survey. Research Policy 12(4):183–197.



Mowery DC, Oxley JE, Silverman BS. 1996. Strategicalliances and interfirm knowledge transfer. StrategicManagement Journal , Winter Special Issue 17:77–91.

Nerkar A, Roberts PW. 2004. Technological and product-market experience and the success of new productintroductions in the pharmaceutical industry. StrategicManagement Journal , August–September SpecialIssue 25: 779–799.

Nicholls–Nixon CL, Woo CY. 2003. Technology sourc-ing and output of established firms in a regime ofencompassing technological change. Strategic Man-agement Journal 24(7): 651–666.

O’Reilly CA III, Tushman ML. 2008. Ambidexterityas dynamic capability: resolving the innovator’sdilemma. Research in Organizational Behavior 28:185–206.

Park SH, Chen R, Gallagher S. 2002. Firm resources asmoderators of the relationship between market growthand strategic alliances in semiconductor start-ups.Academy of Management Journal 45(3): 527–545.

Pisano GP. 1994. Knowledge, integration, and thelocus of learning: an empirical analysis of processdevelopment. Strategic Management Journal , WinterSpecial Issue 15: 85–100.

Pisano GP. 1996. Learning-before-doing in the develop-ment of new process technology. Research Policy 25:1097–1119.

Pisano GP. 1997. The Development Factory. Unlockingthe potential of process innovation . Harvard BusinessSchool Press. Boston, MA.

Pisano GP, Mang P. 1993. Collaborative product devel-opment and the market for know-how: strategies andstructures in the biotechnology industry. In Researchon Technological Innovation, Management, and Pol-icy . Rosenbloom R, Burgelman R (eds). JAI. Press:Greenwich, CT; 109–136.

Prahalad CK, Bettis RA. 1986. The dominant logic:a new linkage between diversity and performance.Strategic Management Journal 7(6): 485–501.

Raisch S, Birkinshaw J. 2008. Organizational ambidex-terity: antecedents, outcomes, and moderators. Journalof Management 34(3): 375–409.

Reuer JJ, Zollo M. 2005. Termination outcomes ofresearch alliances. Research Policy 34(1): 101–115.

Roberts P. 1999. Product innovation, product-marketcompetition and persistent profitability in theU.S. pharmaceutical industry. Strategic ManagementJournal 20(7): 655–670.

Rosenkopf L, Nerkar A. 2001. Beyond local search:boundary-spanning, exploration, and impact in theoptical disk industry. Strategic Management Journal22(4): 287–306.

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

Rothaermel FT, Alexandre MT. 2009. Ambidexterity intechnology sourcing: the moderating role of absorptivecapacity. Organization Science 20(4): 759–780.

Rothaermel FT, Boeker W. 2008. Old technology meetsnew technology: complementarities, similarities, and

alliance formation. Strategic Management Journal29(1): 47–77.

Rothaermel FT, Deeds DL. 2004. Exploration andexploitation alliances in biotechnology: a systemof new product development. Strategic ManagementJournal 25(3): 201–221.

Rothaermel FT, Deeds DL. 2006. Alliance type, allianceexperience, and alliance management capabilityin high-technology ventures. Journal of BusinessVenturing 21(4): 429–460.

Rothaermel FT, Hess AM. 2007. Building dynamiccapabilities: innovation driven by individual- firm-,and network-level effects. Organization Science 18(6):898–921.

Rothaermel FT, Hill CWL. 2005. Technological disconti-nuities and complementary assets: a longitudinal studyof industry and firm performance. Organization Sci-ence 16(1): 52–70.

Sampson RC. 2005. Experience effects and collaborativereturns in R&D alliances. Strategic ManagementJournal 26(11): 1009–1031.

Schumpeter JA. 1942. Capitalism, Socialism, and Democ-racy . Harper & Row: New York.

Shan WJ, Walker G, Kogut B. 1994. Interfirm coopera-tion and startup innovation in the biotechnology indus-try. Strategic Management Journal 15(5): 387–394.

Simonin BL. 1997. The importance of collaborativeknow-how: an empirical test of the learningorganization. Academy of Management Journal 40(5):1150–1174.

Simonin BL. 1999. Ambiguity and the process ofknowledge transfer in strategic alliances. StrategicManagement Journal 20(7): 595–623.

Sims N, Harrison R, Gueth A. 2001. Managing alliancesat Lilly. In Vivo. Business & Medicine Report 19(6):1–9.

Stuart TE, Hoang H, Hybels RC. 1999. Interorgani-zational endorsements and the performance ofentrepreneurial ventures. Administrative Science Quar-terly 44(2): 315–349.

Teece DJ. 1992. Competition, cooperation, and innova-tion: organizational arrangements for regimes of rapidtechnological progress. Journal of Economic Behaviorand Organization 18(1): 1–25.

Teece DJ, Pisano G, Shuen A. 1997. Dynamic capabili-ties and strategic management. Strategic ManagementJournal 18(7): 509–533.

Trajtenberg M. 1990. A penny for your quotes: patentcitations and the value of innovations. Rand Journalof Economics 21(1): 172–187.

Tripsas M. 1997a. Surviving radical technological changethrough dynamic capability: evidence from thetypesetter industry. Industrial and Corporate Change6(2): 341–377.

Tripsas M. 1997b. Unraveling the process of creativedestruction: complementary assets and incumbent sur-vival in the typesetter industry. Strategic ManagementJournal , Summer Special Issue 18: 119–142.

Tuma NB, Hannan MT. 1984. Social Dynamics: Modelsand Methods . Academic Press: Orlando, FL.



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.

Yelle LE. 1979. The learning curve: historical reviewand comprehensive survey. Decision Science 10(2):302–328.

Zahra SA, George G. 2002. Absorptive capacity: areview, reconceptualization, and extension. Academyof Management Review 27(2): 185–203.

Zajac EJ, Kraatz MS, Bresser RKF. 2000. Modelingthe dynamics of strategic fit: a normative approachto strategic change. Strategic Management Journal21(4): 429–453.

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


LEVERAGING INTERNAL AND EXTERNAL · PDF file1 ESSEC Business School, Cergy Pontoise, ......

Documents

Transcript of LEVERAGING INTERNAL AND EXTERNAL · PDF file1 ESSEC Business School, Cergy Pontoise, ......