A Competition-based Explanation of Collaborative Invention With the Firm

Strategic Management JournalStrat. Mgmt. J., 34: 1186–1208 (2013)

Published online EarlyView 22 March 2013 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/smj.2059

Received 13 November 2010 ; Final revision received 14 June 2012

A COMPETITION-BASED EXPLANATION OFCOLLABORATIVE INVENTION WITHIN THE FIRM

PUAY KHOON TOH1* and FRANCISCO POLIDORO2

1 Carlson School of Management, University of Minnesota, Minneapolis, Minnesota,U.S.A.2 McCombs School of Business, The University of Texas at Austin, Austin, Texas,U.S.A.

Existing literature shows that collaborative invention within the firm enhances innovativenessby facilitating knowledge recombination. Despite such benefit, firms vary in their use ofcollaborative invention when drawing on their individual inventors’ knowledge. In addressingthis puzzle, we argue that competition from rival products building on similar knowledge compelsfirms to favor search depth over exploratory search and respond expeditiously, thus reducing afirm’s inclination toward collaborative invention. In contrast with prior research’s focus onhow upstream resources influence a firm’s position in downstream markets, this study showsthat downstream competition drives heterogeneity across firms in their utilization of upstreamresources. Copyright 2013 John Wiley & Sons, Ltd.

INTRODUCTION

Collaboration has become increasingly prominentin the organization of knowledge-based activ-ities (Wuchty, Jones, and Uzzi, 2007). Col-laboration between inventors allows a firm torecombine knowledge residing across these inven-tors (Almeida and Kogut, 1999; Corredoira andRosenkopf, 2010; Fleming and Sorenson, 2004). Itexpands a firm’s combinatorial opportunities andenhances the potential for breakthrough inventions(Fleming, Mingo, and Chen, 2007; Rosenkopfand Almeida, 2003). Collaborative invention thusconstitutes a useful way for a firm to utilizeits inventors’ knowledge and create a competi-tive advantage (Felin and Hesterly, 2007; Grant,1996; Liebeskind, 1996). However, despite these

Keywords: collaboration; competition; innovation;knowledge-based resources; exploration; pharmaceuticals*Correspondence to: Puay Khoon Toh, Carlson School ofManagement, University of Minnesota, 321 Nineteenth AvenueSouth, RM 3-364, Minneapolis, MN 55455, U.S.A.E-mail: [email protected]

Copyright 2013 John Wiley & Sons, Ltd.

known benefits of collaborative invention, theliterature has stopped short of explaining whyits incidence varies across firms. This leaves anunanswered question—why are some firms moreinclined toward collaborative invention while otherfirms have more of their inventors inventing indi-vidually?

To address this question, a helpful start is tonote that collaborative invention likely involves adifferent type of search (Dosi, 1982; Katila andAhuja, 2002) than inventing with individual inven-tors. It also entails more communication and coor-dination difficulties between individuals (Beckerand Murphy, 1992; Kretschmer and Puranam,2008). Consequently, collaborative invention maynot always be the most appropriate way for a firmto expend its R&D efforts and coordinate its inven-tors. It is a strategic choice rather than a superiorpractice that firms will uniformly adopt. Accord-ingly, a theory of collaborative invention can beone that identifies situations requiring the type ofsearch that collaboration involves, or justifying thedifficulties that collaboration entails.

A Competition-Based Explanation of Collaborative Invention 1187

In this paper, we take an initial step towardestablishing such a theory by examining the com-petitive environment as a driver of a firm’s col-laborative invention. Competition has pervasiveeffects on a firm’s inventive process. Competi-tive threats can arise from rivals’ products, pro-duction capacity, market power, or intellectualproperty (IP) rights (Demsetz, 1973; Dixit, 1980;Gilbert and Newbery, 1982). They influence howmuch inventive effort a firm invests in particu-lar markets, as evidenced by the extensive studieson R&D races (Doraszelski, 2003; Lerner, 1995;Reinganum, 1985). Competition also influenceswhich technological space a firm explores (Clark-son and Toh, 2010) and what products a firminvests in (Martin and Mitchell, 1998). What isless known thus far is how, for a given amountof inventive effort within a particular technologi-cal space, competition also affects the way a firmorganizes its inventors, specifically, the extent towhich it engages in collaborative invention.

Competition in a product market comes in twoforms: products building on similar knowledgeand products building on alternative knowledgebases (Dosi, 1982; Nelson and Winter, 1982; Poli-doro and Toh, 2011). Prior research focusing onthe former has shown that it induces a firm todefend its market position by refining existingproducts (Anderson and Tushman, 1990; Mar-tin and Mitchell, 1998). We extend this line ofresearch by examining how such form of compe-tition, with its inherent cumulative improvementsin product features (Green and Scotchmer, 1995;Murray and O’Mahoney, 2007), induces a firm toorganize its inventors for greater search depth withless collaborative invention, rather than for greaterexploratory search via more collaborative inven-tion. In addition, as such competition increases, afirm likely has less slack time to deal with col-laboration difficulties, and having inventors workindividually instead may be preferable. Accord-ingly, we propose that as a firm faces greatercompetition from rival products building on sim-ilar knowledge, it engages in less collaborativeinvention for a given level of inventive effort.1

1 We do not expect our propositions to generalize to competitionfrom products building on alternative knowledge bases because,as prior studies have shown, this form of competition exertsconflicting influences, sometimes compelling a firm to refine itsown technology (Anderson and Tushman, 1990: 611), while atother times inducing a firm to branch away from its technology

Further, we examine a firm’s commercial and tech-nical abilities as boundary conditions for this mainproposition, which help explain why firms differ intheir use of collaborative invention even when theyface the same level of competition.

To test our propositions, we focus on pharma-ceutical firms producing antihypertensives (drugsthat lower blood pressure) from 1980 to 2004.Three features of this setting are particularly rele-vant to our empirical tests. First, by tracing thepharmacological mechanism underlying a firm’santihypertensive, we can ascertain the knowledgeunderlying that drug (Reuben and Wittcoff, 1989)and precisely identify rival drugs building on sim-ilar knowledge. This fine-grained measure of com-petition enables a sharper test of our proposi-tions, by identifying competition from only rivalproducts (antihypertensives) building on similarknowledge (mechanism) as a firm’s product, ratherthan including all competing products in the mar-ket. Second, by identifying specific patents pro-tecting antihypertensives, we can map rival prod-ucts building on a particular knowledge to therelevant subset of a firm’s subsequent inventiveactivities involving such knowledge. Third, byexploiting a flux period in the mideighties inducedby mergers and acquisitions within the pharma-ceutical industry, we can instrument the variablecapturing competition from similar antihyperten-sives. Such instrumental variable approach helpsaddress potential endogeneity issues, specifically,that firms engaging in collaborative invention maytend to operate in areas with less competition, orthat the nature of a certain technological area maydraw more competition and concurrently requiregreater depth in search through individual inven-tors. We elaborate on the merits of these featuresin a later section.

Through our propositions, we introduce acompetition-based explanation for why, despitethe well-established benefits of combining knowl-edge across individuals, firms vary in their use ofcollaborative invention. In the innovation process,whether a firm utilizes its individual inventorsmore as a set of collaborative inventors or more asa collection of solo inventors depends on the com-petition it faces. This paper, by highlighting suchrole of competition in firms’ resource utilization,advances theory on the sources of competitive

and explore alternatives instead (Martin and Mitchell, 1998:756).

Copyright 2013 John Wiley & Sons, Ltd. Strat. Mgmt. J., 34: 1186–1208 (2013)DOI: 10.1002/smj

1188 P. K. Toh and F. Polidoro

advantage. A common theme in strategy researchis that firms perform differently because theypossess heterogeneous resources (Barney, 1991;Wernerfelt, 1984), and resource heterogeneity,in turn, arises when firms accumulate resourcesin dissimilar ways (Ahuja and Katila, 2004;Helfat, 1997). This paper, in contrast, focusesmore on firms’ resource utilization rather thanaccumulation, and shows that, even for an activ-ity such as collaborative invention that seemspurely upstream and unrelated to downstreammarkets, product competition in fact still plays acrucial role. Simply put, although firms’ upstreamresources shape their downstream performances,in some circumstances it is precisely downstreamcompetition that drives the way firms utilizetheir upstream resources. In the final section, weelaborate on these theoretical implications.

THEORY AND HYPOTHESES

Collaborative invention allows a firm to draw on agreater number and diversity of knowledge com-ponents from its inventors, creating greater recom-binant potential (Fleming, 2001). Moreover, col-laborative invention entails social interactions thatenhance creativity and enable inventors to com-bine previously unconnected ideas, technologies,and processes (Fleming et al., 2007), which in turnincreases the incidence of impactful, breakthroughinnovations (Singh and Fleming, 2010). Althoughcollaborative invention provides these benefits, itspractice is not uniform across firms. Some firmstend to have more collaboration between inven-tors, whereas others are more inclined to let theirinventors work individually (Dahlin, Taylor, andFichman, 2004; Singh and Fleming, 2010). Toexplain this divergent choice, we first highlight twoattributes of collaborative invention—the type ofsearch it facilitates and the difficulties it entails. Wethen examine how competition from rival productsbuilding on similar knowledge diminishes a firm’sinclination to engage in collaborative invention.

Collaborative invention: type of search andcollaboration difficulties

Collaborative invention is conducive to distantand exploratory search (Levinthal, 1997; March,1991), as pooling multiple individuals’ knowl-edge expands the combinatorial space (Singh and

Fleming, 2010). When multiple inventors cometogether with different knowledge, as long as a cri-terion is to include knowledge components fromeach inventor, the resultant combination will onaverage incorporate more distant components andaccordingly be more radical. In contrast, whenan inventor operates individually, search tendsto occur in a narrower space and often withgreater depth as the inventor engages in more useand reuse of her existing knowledge (Katila andAhuja, 2002). This deepens individual inventors’understanding of concepts and scientific principlesunderlying their inventions, and reduces potentialerrors, leading to more reliable and predictablesearch (Eisenhardt and Tabrizi, 1995). The inven-tive outcomes are consequently more incrementaland cumulative over prior inventions (Tushmanand Anderson, 1986). Hence, firms tend to choosecollaborative invention when they wish to engagein exploratory search and, conversely, tend to haveits inventors inventing individually in situationsrequiring search depth. Indeed, firms often reallo-cate inventors, for instance, from more exploratoryprojects to ones more focused on the inventors’specialized expertise, so as to increase their effi-ciency in product innovation activities (Edmond-son and Nembhard, 2009).

The second attribute of collaborative inventionconcerns the difficulties it entails. Collaborationrequires coordination, which can be problem-atic when collaborating individuals are special-ists in different fields (Becker and Murphy,1992; Kretschmer and Puranam, 2008). It alsorequires individuals to synchronize objectives,agree on the sequence of activities, and plan forconnecting interfaces between different parts ofprojects (Lawrence and Lorsch, 1967; Sanchezand Mahoney, 1996). Moreover, when inventorscollaborate, a firm cannot easily monitor eachindividual’s inputs to the joint production ofknowledge, or even identify what each individualcan potentially contribute (Eisenberg, 2001). Whenindividuals anticipate such problems of imper-fect attribution in jointly obtained outcomes, theirincentives to contribute optimal effort toward thejoint activity in the first place further reduces(Aghion and Tirole, 1994), which can render afirm vulnerable to shirking, free-riding, and outputexpropriation by opportunistic individuals (Pisano,1990; Williamson, 1975). Additional obstacles canarise from the interpersonal context surroundingcollaboration. The very diversity of knowledge



and perspectives within teams that spurs innova-tions can also give way to disagreements aboutdelegation of duties and resources, which eventu-ally prevent the team from realizing those poten-tial benefits (Jehn, Northcraft, and Neale, 1999).Although these collaboration difficulties are notinsuperable, reducing frictions in coordination,designing suitable reward structures, and mitigat-ing interpersonal conflicts take effort and time.Thus, when deciding whether or not to engagein collaborative invention, a firm likely assesseswhether the situation calls for a greater need ofexploratory search through collaborative inventionso as to justify the effort and time required to mit-igate the associated collaboration difficulties.

Effect of competition from similar products oncollaborative invention

Having highlighted the type of search that col-laborative invention facilitates and the difficultiesit entails, we now explain how competition fromproducts building on similar knowledge, by shap-ing the salience of these two attributes, affects afirm’s collaborative invention. Competition fromproducts building on similar knowledge createspressure for a firm to deal with rivals’ incre-mental product improvements (Green and Scotch-mer, 1995). These rival products often do notencompass major changes in such knowledge.They constitute minor improvements upon prod-uct features relevant to users. For example, inthe cell phone industry, competing smart phonesbuild on existing platforms with minor improve-ments in physical features and usability. Simi-larly, rivalry in the pharmaceutical industry oftenarises from competing versions of similar drugsimproving on some dimensions of efficacy orsafety (Danzon, 2000). For instance, differentangiotensin-converting enzyme inhibitors, suchas Parke-Davis’s quinapril (Accupril) and Ciba-Geigy’s benazepril (Lotensin), achieve a similareffect in the treatment of hypertension (Scriabine,1999: 188). Rival products building on similarknowledge may also compete for new applicationsthrough incremental modifications. For example,with small tweaks to polytetrafluoroethylene (alsoknown as Teflon), the technology’s applicationshifted from nonstick materials in cooking pansto slide-inducing materials in gears, bearings,uranium-holding pipes, and roof coatings. Like-wise, firms have used antihypertensives building

on the beta-blockers mechanism to treat arrhyth-mia (Scriabine, 1999).

Such competition exists even in the presence ofpatent protection. The economics literature showsthat patent protection does not fully prevent rivalsfrom inventing around and creating imitative prod-ucts (Dasgupta, 1988: 74; Gilbert and Newbery,1982: 519). As Mansfield, Schwarz, and Wagnernote, “Contrary to popular opinion, patent pro-tection does not make entry impossible, or evenunlikely”2 Mansfield, Schwarz, and Wagner (1981:913). Although pharmaceutical firms actively usepatents to deter imitation (Cohen, Nelson, andWalsh, 2000), patents do not award exclusion-ary rights over the scientific principles underly-ing drugs. A patent protecting a drug does notprevent rivals from drawing on the mechanismof action underlying that drug to create simi-lar “me-too” drugs (Danzon, 2000; Higgins andRodriguez, 2006). For instance, following the suc-cess of chlorothiazide, a diuretic antihypertensive,many firms invented around Merck’s patent, result-ing in more than twenty related diuretics in themarket (Scriabine, 1999: 196).

Competition from rival products building onsimilar knowledge influences a firm’s collaborativeinvention in two ways. First, it shifts a firm’s pri-ority away from exploratory search toward greatersearch depth. With the threat of losing marketshare to these rival products, a firm tends to con-centrate on improving its existing products viaminor variations based on that same knowledge,so as to defend its market position (Anderson andTushman, 1990; Martin and Mitchell, 1998). Forexample, increasing competition among calciumchannel antagonists in the treatment of hyperten-sion induced firms to focus on minor improve-ments relative to similar drugs, such as longerduration of action in the case of Pfizer’s amlodip-ine, or availability of a water-soluble derivative forintravenous administration in the case of Syntex’snicardipine (Scriabine, 1999: 193). Also, as rivalproducts cumulatively improve within a domain,the problems they aim to solve become increas-ingly nuanced and specific to the domain. To cre-ate the next technically superior product withinthe same domain, a firm needs greater domain-specific experience and deeper understanding of

2 These authors showed that 60 percent of the patented successfulinnovations in their sample were imitated within four years oftheir introduction.



the underlying knowledge. These may come inthe form of knowing how to decompose prob-lems, what specific knowledge combinations aremeaningful, or where to look for feasible solu-tions within the domain (Eisenhardt and Tabrizi,1995; Katila and Ahuja, 2002). As knowledgegains depth, it often becomes more tacit and harderto communicate across individuals. This is akin tothe idea that mastery of skills is hard to codifyand typically requires prolonged apprenticeships.In these instances, having inventors individuallyburrowing for greater depth is likely more effec-tive. Moreover, exploratory search inherent in col-laborative invention tends to be risky (March,1991), as it involves experimenting with relativelyuntested knowledge combinations. Bearing suchrisk requires a firm to secure at least a minimumbase of financial performance in the interim. Com-petition, by threatening the market share of a firm’sproducts, challenges such base. Thus, in the face ofcompetition, exploratory search through collabora-tive invention tends not to be a high firm priority.

Second, competition from rival products build-ing on similar knowledge typically requires swiftresponses, which diminishes the time that a firmcan afford to incur for dealing with collaborationdifficulties. As such rival products typically targetusers with similar preferences and introduce refine-ments over relevant functional dimensions, theycould quickly erode a firm’s market share (Chris-tensen, 1997; Martin and Mitchell, 1998). Further-more, a firm’s expected returns from investing ina particular underlying knowledge arise from notonly its current product building on this knowl-edge but also the whole stream of potential prod-ucts in the future exploiting this knowledge. Rivalproducts cumulatively exhaust these future oppor-tunities. To avoid losing current market share andmissing future exploitative opportunities, a firmneeds to rapidly introduce new products build-ing on its current technology. This time pres-sure is aggravated by the fact that cumulativerival products, by advancing on various technicaldimensions, raise technical hurdles for subsequentproducts, and a firm needs time to overcome thesehurdles. This time urgency induces a firm to directits inventive effort away from collaborative inven-tion, as collaborative invention requires time forovercoming its inherent difficulties, toward havingmore inventors working individually instead.

In summary, holding other influences con-stant, competition from rival products building

on similar knowledge, by shifting a firm’s prior-ity away from exploratory search toward greatersearch depth and by increasing the pressure for afirm to swiftly refine its products, pushes that firmaway from collaborative invention toward a greaterdegree of individual invention.

Hypothesis 1 (H1): The more rival productsbuild on similar knowledge, the less a firmsubsequently engages in collaborative inventionwhen using this knowledge.

H1 accounts for firm variation within the sameproduct market at a point in time. Depending onthe knowledge they use, firms face different levelsof similar product competition and, accordingly,exhibit different propensities toward collaborativeinvention. However, variation in collaborativeinvention across firms using the same knowledgeand hence facing the same competition remainsto be explained. To explain this variation, weexamine firm attributes rendering a firm especiallysusceptible to the main effect in H1. Lookingclosely at the earlier mechanisms, we identifytwo such attributes—commercial abilities andtechnical abilities.

The first contingency is a firm’s commercialabilities. Abilities to commercialize inventionsreside not only in tangible downstream assets, suchas production and distribution systems (Cohen andKlepper, 1996), but also in intangible skills, suchas the craft of managing the trial-and-error processin development phases, awareness of appropriateexperimental settings for testing prototypes’ feasi-bility, sense of the right market for product launch(Nerkar and Roberts, 2004), and relationshipswith distribution networks (Mitchell, 1989).

While similar rival products steer a firm towardgreater search depth (less collaborative invention)to advance its product’s technical attributes, itscommercial abilities dampen such effect. With suf-ficient commercial abilities, a firm can still push itsproduct to commercial success even if it is techni-cally inferior. For example, in the pharmaceuticalindustry where physicians cannot fully observe theintrinsic quality of drugs (Bodewitz, Buurma, andde Vries, 1987), a firm can rely on its commercialabilities to shape physicians’ prescribing behavior(Avorn, Chen, and Hartley, 1982; Azoulay, 2002).A firm with strong commercial abilities is lessthreatened by similar rival products and better able



to secure a minimum base of financial performancedespite rivals’ technical advances. This allows afirm to still engage in exploratory search for poten-tial breakthroughs via collaborative invention inspite of competition.

In contrast, a firm with weaker commercialabilities is more susceptible to the pressure ofcompetition. With fewer downstream assets tobolster its product’s performance, a commerciallyweaker firm is more vulnerable to the technicalchallenges stemming from similar rival products.If this firm decides to remain a competitor inthis market, its weaker commercial abilities createsa greater need for it to keep up with technicalimprovements in its products, so as to avoid losingmarket share. These incremental improvementscan be best achieved via greater search depth (lesscollaborative invention). This leads to our firstcontingency effect:

Hypothesis 2 (H2): Rival products building onsimilar knowledge reduce a firm’s subsequentcollaborative invention to a greater degree whenthat firm has lower commercial abilities .

The second contingency is a firm’s techni-cal abilities. This refers to a firm’s abilities torapidly create new products based on a partic-ular knowledge. It includes how quickly a firmcan figure out which and how knowledge com-ponents can be combined (Fleming and Sorenson,2004). For example, in new drug creations, it refersto how quickly a firm can decipher what com-binations of compounds are scientifically soundand will result in feasible drugs, and can knowhow a core set of drug components can be mod-ified to improve efficacy or safety (Thomke andKuemmerle, 2002). Related components exhibitvarying degrees of interdependence (Thompson,1967); learning which existing couplings to breakand which new ones to establish instead takestime (Yayavaram and Ahuja, 2008). These abil-ities are sticky and accumulate with time andexperience (Nerkar and Roberts, 2004; Szulanski,1996). When new inventive opportunities arise, forinstance, upon scientific discovery of new knowl-edge, a firm with technical abilities is quicker tocapitalize on them through the creation of newtechnically feasible inventions.

Technical abilities attenuate competition’s effecton eroding a firm’s time for dealing with collab-oration difficulties. With greater technical abilities

giving faster inventive speed, a firm has more timeto resolve collaboration difficulties without beingtoo slow to market. Accordingly, such firm has lessneed to reduce its collaborative invention whensimilar rival products threaten its market share.Also, it can still afford to use collaborative inven-tion for capturing further inventive opportunitiesfrom the same knowledge, despite rivals erodingslack time and jostling to expropriate these oppor-tunities. Moreover, this firm requires less time toovercome technical hurdles raised by rivals’ cumu-lative products, which we described earlier.

Conversely, a technically weaker firm facesmore time pressure to overcome the hurdles arisingfrom competing rival products. As we discussedearlier, similar rival products, with improvementsalong narrow functional dimensions, create pres-sure on a firm to defend its market positionand respond with incremental improvements. Forexample, when Akrimax’s propranolol (a beta-blocker antihypertensive) demonstrated a nega-tive feature of producing vivid dreams in somepatients, its rival, Bristol, improved on its beta-blocker nadolol by inhibiting that feature (Scri-abine, 1999: 184). Clearly this improvementrendered the rival drug (nadolol) more attractive,especially to patients vulnerable to the side effectof the firm’s drug (propranolol). Indeed, nadololbecame one of the elite 100 best-selling drugs inthe early 1980s according to the Pharmacy Times .For a firm that decides to continue competing inthe beta-blocker market, the immediate pressureis not about exploring new combinations basedon beta-blocker knowledge. Rather, the immediatepressure that such firm faces is to swiftly mitigateits drug’s side effect to avoid losing market share.This time pressure is more acute when a firm lackstechnical abilities and cannot count on its inventivespeed to quickly counter the side effects. Hence, atechnically weaker firm has even less time to dealwith collaboration difficulties, and would tend tohave more inventors individually addressing thecumulative technical challenges instead.3

3 We stress again that the issue here is not whether a technicallyweak firm will exit the area, or reduce investments in this areaand explore some other new areas instead. Rather, it is aboutwhether a technically weak firm chooses collaborative invention,within its given amount of inventive effort, to deal withincremental challenges brought on by competition. Moreover,in the context of antihypertensives, firms are unlikely to simplygive up the market in the face of competition from similar drugs,since antihypertensives are typically commercially successful



In sum, competition’s effect of reducing collab-orative invention is more pronounced when a firmhas lower technical abilities. This leads to our sec-ond contingency effect.

Hypothesis 3 (H3): Rival products building onsimilar knowledge reduce a firm’s subsequentcollaborative invention to a greater degree whenthat firm has lower technical abilities .

METHODS

We test our propositions in the setting of firms pro-ducing antihypertensives in the United States from1980 to 2004. Collaborative invention is salient inthis setting, as firms’ knowledge of chemical com-pounds and experience in recombining them arecrucial in creating feasible drugs (Henderson andCockburn, 1994; Thomke and Kuemmerle, 2002).Moreover, commercial and technical abilities arerelevant firm attributes in this setting. Firms needto demonstrate their drugs’ merits to the U.S. Foodand Drug Administration (FDA), which reviewsnew drugs before market launch, and to physi-cians, who prescribe drugs (Polidoro and Theeke,2012; Schweitzer, 1997). Also, given the complextechnical hurdles involved in creating efficaciousand safe drugs, firms’ technical abilities are clearlycritical.

Importantly, in this setting we are able to iden-tify rival products building on knowledge similarto that underlying a firm’s product. Antihyper-tensives lower blood pressure through differentmechanisms of action.4 These mechanisms cap-ture the key knowledge on human physiology andpharmacology that drugs build on to achieve theintended therapeutic effect (Reuben and Wittcoff,1989). For instance, angiotensin receptor blockersrelax blood vessels, whereas beta-blockers obstructreceptors that otherwise stimulate cardiac output.Table 1 illustrates mechanisms underlying antihy-pertensives. By tracing the mechanisms underly-ing antihypertensives, we can go beyond broadly

drugs. Indeed, the Pharmacy Times listed 44 antihypertensivedrugs among the 200 best-selling drugs in the United States in1999.4 The therapeutic class of antihypertensives contains bothlong-standing mechanisms, such as diuretics, and mechanismsintroduced more recently, such as angiotensin receptor blockers.Moreover, this class accounts for a significant number ofdrugs—70 distinct antihypertensives were launched by 2004.

observing rival products with comparable func-tionalities, to narrowly identify those that build onsimilar knowledge. This fine-tuned identificationfacilitates a sharper test, as the arguments behindour propositions involve not just any product mar-ket competition, but rather the type of competitionthat exhausts combinatorial opportunities for par-ticular knowledge. Despite pharmaceutical firms’high patenting propensities (Cohen et al., 2000),patent protection does not fully prevent rivals fromcreating similar drugs (Danzon, 2000; Higgins andRodriguez, 2006). Examples of imitative drugsamong antihypertensives include “chlorothiazide(Diuril), the first antihypertensive diuretic (1958),with 15 imitations; propranolol (Inderal), the firstantihypertensives β-blocker (1964), with 24 imita-tions” (Achilladelis, 1999: 3). The success of thebeta-blocker propranolol “led almost every largepharmaceutical company to mount an intensivesearch” for similar drugs (Scriabine, 1999: 183).Likewise, the “success of the first three calciumchannel antagonists—verapamil, nifedipine, anddiltiazem—led almost every major pharmaceuti-cal company to initiate research projects in thisarea” (Scriabine, 1999: 193).

Further, this setting allows us to map the imple-mented form of an innovation in the product mar-ket (drug) to the inventive activities in R&D thatgenerated the innovation (patents). Following the1984 Hatch-Waxman Act, the FDA started to com-pile information about patents protecting pharma-ceutical drugs, which enables this mapping. Thisenables us to go beyond studying collaborativeinvention between scientists as reflected in patents(Fleming et al., 2007; Singh and Fleming, 2010) toexamine how competing drugs affect such collab-orative invention. Further, it enables us to narrowdown from a firm’s collaborative invention in all ofits inventive activities to that in the relevant subsetinvolving a particular knowledge underlying drugs.This in turn allows a sharper identification of theeffect of competition we propose.

Data

The FDA was the source of data on antihyper-tensives. For each drug, we collected data on theactive ingredient, the innovating firm, its reviewstatus,5 and approval date. We also gathered data

5 The FDA assigns priority review status to drugs that representpromising therapeutic advance.



Table 1. Examples of mechanisms of action among antihypertensive drugs

Mechanism of action Examples of drugs building on mechanism

Alpha blockers—reduce blood pressure by blocking receptorsthat reduce blood vessels diameter and increase resistance toblood flow.

Methyldopa (Merck)Doxazosin (Pfizer)

Angiotensin-converting enzyme (ACE) inhibitors—inhibit theenzyme associated with the system of the body preventingrapid blood pressure loss.

Benazepril (Ciba-Geigy)Quinapril (Parke-Davis)

Angiotensin receptor blockers—block angiotensin receptors,thus preventing the enzyme from contracting blood vessels.

Losartan (Merck)Irbesartan (Sanofi)

Beta-blockers—decrease blood pressure by blocking thereceptors that otherwise stimulate cardiac output.

Metoprolol (Novartis)Nadolol (Bristol)

Calcium channel blockers—lower blood pressure by limitingthe amount of calcium that enters heart cells that generatesignals for heart contractions.

Nifedipine (Pfizer)Nicardipine (Syntex)

Diuretics—lead to an increase in urine production, whichresults in a decrease of the blood volume.

Amiloride (Merck)Indapamide (Sanofi-aventis)

on antihypertensives approved by the FDA priorto 1980 to construct some measures that we detaillater. To do so, we searched for all antihyper-tensives listed in the pharmacological databasesMicromedex , Mosby’s Drug Consult , and DrugFacts and Comparisons . We then looked for therespective FDA-approval dates in the FDA’s list ofApproved Drug Products with Therapeutic Equiv-alence Evaluation , also known as the OrangeBook . The pharmacological databases also fur-nished information on the mechanism underlyingeach drug. We then subjected this identificationof mechanisms to validation by an external expertin pharmacology and medicinal chemistry. Fur-ther, we used the Orange Book to identify thepatents protecting the drugs in the sample, captur-ing the multiple patents that a firm oftentimes usesto protect different components of a drug. We thencollected detailed data on these patents, includingassigned technology classes, inventors, and theirlocations, from the Cassis database managed bythe U.S. Patent and Trademark Office (USPTO).Further, we used the annual lists of the PharmacyTimes to identify best-selling antihypertensives.Finally, to construct a control variable we detaillater, we used the Web of Science to collect data onscientific publications about the drugs in the study.

To build our sample, we started with all FDA-approved antihypertensives between 1980 and2004. We traced the associated firms, observedthe mechanism underlying each firm’s drug, andidentified all competing drugs based on the samemechanism in each year. Next, we traced eachfirm’s inventive activities that potentially involved

antihypertensive knowledge, so as to measure afirm’s collaborative pattern within these activities.Pharmaceutical firms typically operate in multipletherapeutic classes, some of which may not relateto antihypertensive knowledge. We identified therelevant subset of a firm’s inventive activities inthree steps. First, we compiled all patents for afirm’s antihypertensives. Second, we identifiedthe inventors listed in these patents, by mappingtheir last names and initials of first names withina firm.6 Finally, we compiled all patents filed sub-sequently by that firm involving those inventors.The resulting subset of patents contains inven-tions created by a firm’s inventors who possessknowledge about that firm’s antihypertensives.7

The unit of analysis for the eventual panel datasetis the firm-year, and the dataset contains 292observations in the full sample with 27 firms.

Dependent variable

Collaboration ratio

We use patent coauthorships to measure theextent to which a firm engages in collaborative

6 Note that by keeping this mapping within a firm, we minimizeerrors arising from different inventors across firms having thesame last names and initials of first names. To further ensurethe accuracy of this mapping, we randomly selected 100 pairsof equivalent last names and initials within firms, and manuallychecked their full first names and middle names. We found noinstances where a pair included different inventors.7 53.4 percent of patents in this subset refer to subsequent com-binations involving firms’ antihypertensive knowledge gearedtoward other therapeutic applications.



invention. We trace, for each year, a firm’s patentsthat include inventors with knowledge on antihy-pertensives, and count the ones listing more thanone inventor. As this count may be confoundedwith a firm’s overall inventive effort, we measurethe dependent variable as a ratio of such countto a firm’s total patents that include inventorswith knowledge on antihypertensives in the year.This variable thus indicates a firm’s extent ofcollaborative invention (with the alternative beingindividual invention) for a given amount ofinventive effort.

Patent coauthorships are appropriate proxiesfor collaborative invention. The U.S. patent lawsrequire that, when an invention results from con-tributions of several individuals, the “joint inven-tors must apply for a patent jointly and eachmust make the required oath or declaration; nei-ther of them alone, nor less than the entirenumber, can apply for a patent for an inven-tion invented by them jointly” (USPTO, 2012).Accordingly, firms have strong incentives to matchpatent coauthorship with the underlying collabora-tion, so as to minimize risk of subsequent patentinvalidation. Indeed, past research has demon-strated through field studies that patent coauthor-ships reflect intensive collaboration and workingrelationships between coinventors over nontriv-ial time periods (Fleming et al., 2007) and hasestablished that they capture collaborative inven-tion (e.g., Cockburn and Henderson, 1998; Singhand Fleming, 2010). We also consulted two patentlawyers with vast experience in patent prosecu-tion and strategic patent portfolio management onthe possibility that a firm may strategically listonly one inventor when an invention indeed resultsfrom a larger collaborative effort. They promptlyconsidered this improbable, stressing that a firm’sidentification of a patent’s coinventors is a legal,not a strategic or political, matter. These justifi-cations notwithstanding, we further conducted aseries of tests to examine the appropriateness ofpatent coauthorships as proxies for collaborativeinvention, which we explain in a later section.

Another potential issue with patent-based mea-sures is that not all firms’ inventions are patented.Firms’ motivations and patenting propensitiesmay differ significantly across industries (Cohenet al., 2000), such that the collaboration mea-sure may not encompass all relevant inventiveactivities within a firm. By restricting our sampleto antihypertensives, we minimize this problem,

as patenting propensities are likely stable withina product class (Griliches, 1990). Also, whilepatents may not be perfect measures of inventiveoutputs, they are relatively reliable indicators ofinventive efforts (Hausman, Hall, and Griliches,1984). Moreover, our focus is on collaborativetendencies as indicated in patents, rather thancounts of patents, and a priori it is not clear thatcollaboration tendencies would systematically dif-fer between patented and nonpatented inventionsfor a firm in a particular year.

Independent variables

Competition from similar antihypertensives ( H1)

To capture rival products building on similarknowledge, we create a variable with the num-ber of rival drugs building on the same mecha-nism of action underlying a firm’s drug availablein the antihypertensive market in the year pre-ceding the observation year. 8 By tracing thesecommercialized rival drugs rather than measuringrivals’ knowledge similarity based on patents (e.g.,Ahuja and Katila, 2001; Mowery, Oxley, and Sil-verman, 1998), this variable captures the actualcompetitive threats rather than early outcomes ofrivals’ experimentation with similar knowledge asreflected in patents. This is more appropriate forour purpose since patents may only reflect uncer-tain signs of potential competition, most of whichmay not materialize.

Commercial abilities ( H2)

When measuring a firm’s commercial abilities,we focus on abilities a firm needs to advance newdrugs through the commercialization stage. Evenafter a firm establishes that a new drug is effi-cacious and safe, it requires commercial abilitiesto successfully distribute and promote the drug’suse (Schweitzer, 1997). Such abilities are not justrepresented by a firm’s physical distributions ormarketing assets, but also reside in intangibleassets such as relationships and knowledge ofdealing with physicians. A comprehensive mea-sure should thus focus less on a drug’s inventionin early stages and more on its commercial

8 In cases where a firm had antihypertensives based on differentmechanisms, we compute the average count of rival drugs acrossthese mechanisms. Results are robust when we retain only onemechanism for each firm.



performance, and encompass more than traceablephysical inputs. We construct a variable with thenumber of versions of a firm’s antihypertensivesthat made it into the top 100 drugs generatingthe most sales in dollars in the U.S. prescriptionmarket in the year preceding the observation year.To ensure that the variable does not overstatecommercial abilities by including duplicate ver-sions with minimal technical advances, we onlyconsider a firm’s distinct molecular entities.

Technical abilities (H3)

A key element in H3 is that for a firm withgreater inventive speed, its collaborative inventionwill be relatively unaffected by competition. Thisspeed is manifested in how quickly a firm, uponthe discovery of a new mechanism of actionto treat hypertension, can create a new drugbased on this mechanism. New mechanisms arediscovered either in scientific breakthroughs inresearch institutions or through a private firm’sdrug-creation process (Scriabine, 1999). Uponsuch discovery, rival firms would typically vieto create new drugs based on this mechanism(Danzon, 2000). As explained earlier, this rivalryexists in the presence of patent protection, sincemechanisms of action per se cannot be patented.A firm with greater inventive speed tends to morequickly figure out which knowledge combinations(to create new drugs) are feasible within thismechanism, and hence be an earlier producerof drugs for this mechanism. Accordingly, tocreate the measure of technical abilities, we tracethe order of a firm’s antihypertensive within itsrespective mechanism; that is, whether its drugwas the first, or second, or third (and so on) drugto appear in the market upon the introduction ofthis mechanism. We then calculate the inverse ofthis order (i.e., reciprocal, 1/n), such that highervalues of this measure reflect greater technicalabilities.9

9 When computing this measure for a firm that had antihyper-tensives based on different mechanisms, we used the averagevalue for its drugs across mechanisms. Results are robust whenwe retain only one mechanism for these firms. Even thoughthis measure is constructed as being relative to rivals, one mayargue that it does not reflect the firm’s contemporaneous (i.e.,specific to a point in time) relative abilities. We separately usean alternative measure of technical abilities calculated as the“firm’s total patents relative to the average rival’s total patents”in the year. Subsequent findings remain fully robust with thisalternative measure.

We acknowledge that our measures of commer-cial and technical abilities do not directly captureabilities but rather capture the performance out-comes of abilities, despite our attempts to closelyconnect them to the underpinnings of H2 and H3.Observing a firm’s abilities is a challenging task(Henderson and Cockburn, 1994). When examin-ing whether these abilities contribute to firm per-formance, measuring abilities using performanceoutcomes clearly entails a problem of tautology(Priem and Butler, 2001). However, this tautologyis not salient in our study, since we are not exam-ining whether these abilities contribute to firmperformance. In our propositions, we do not pre-sume collaborative invention to imply, or lead to,better firm performance than solo-inventor inven-tions. In fact, for our purpose, the more closelythese measures based on performance outcomes(e.g., inverse order of a firm’s drug within themechanism) correlate with the unobserved latentabilities (e.g., technical abilities), the more accu-rate these measures will be. Moreover, we are notaware of how a potential gap between a firm’sabilities and the respective outcomes would varysystematically with both that firm’s collaborativeinvention (dependent variable) and the competitionit faces (independent variable). Hence, we have noreason to suspect that this gap would create a sys-tematic bias in our findings.

Control variables

We control for factors that may correlate with thecompetition from similar products a firm faces andthe extent to which it engages in collaborativeinvention. To account for the potential influence ofcompeting products building on alternative knowl-edge bases, we include the number of antihyperten-sives building on mechanisms different from thatof a firm’s drug. We control for a firm’s inven-tiveness with the count of a firm’s patents involv-ing inventors working on antihypertensives in theyear. Firms may file for more patents (with soloinventors) for each given invention when facedwith greater competition, so as to strengthen itsprotection against rivals over a particular techno-logical area. We control for a firm’s total num-ber of patents filed in the year (spanning beyondpatents protecting antihypertensive drugs), whichalso helps account for firm size. To address thepossibility that competition may induce firms todiversify geographically in search of lower costs,



which in turn may hinder collaboration, we controlfor a firm’s number of distinct geographic loca-tions (states or foreign countries) based on infor-mation in patents about where inventors reside. Toaccount for effects associated with technologicalprogression, models include the number of yearselapsed since the start of the analysis period.

Further, to address the concern that the dissemi-nation of scientific knowledge may spur incidencesof imitative drugs (Danzon, 2000; Scriabine, 1999)and reduce a firm’s need to explore via collabora-tive invention, we add a variable with the numberof published scientific papers regarding drugs inthe focal mechanism of action in the relevant topjournals. We define top journals as those account-ing for more than two-thirds of the total impactfactor in general and experimental medicine, aswell as pharmacology, according to the Insti-tute for Scientific Information.10 A firm may faceincreasing competition and become more “desper-ate” as its patents approach expiration (Higginsand Rodriguez, 2006), turning to solo-inventorresearch to quickly exploit remaining patent lives.We control for such desperation effect with a countof a firm’s antihypertensive-related patents near-ing expiration, that is, past 15 years since applica-tion.11 More alternative mechanisms of action maymean fewer rivals building on a firm’s mechanismand increased risk of obsolescence for a firm’sknowledge, which accordingly prompt that firm toengage in collaboration to mitigate such risk. Tocontrol for this possibility, we include the numberof antihypertensive mechanisms. Rivals in com-petitive environments frequently engage in litiga-tions, and this may induce a firm to invest in morenarrow-scope (solo inventor) patents with tighterlegal boundaries. We control for such rivals’ liti-giousness with a count of litigations filed by afirm’s rivals that operate in the same mechanismof action. We also add year dummies to control for

10 It is possible that availability of scientific knowledge doesnot facilitate a firm’s exploratory search in instances whererivals’ patent protection is strong, such that this control variableis unnecessary. Additional models dropping this variable showfully robust results.11 We use 15 years since application as a proxy for “nearingexpiration” to account for the Uruguay Round Agreements Actof 1994, which extended patent lengths from 17 years sincegrant date to 20 years since application date. To ensure thatthis control is not affected by how we define patents nearingexpiration, we also use an alternative measure of the average ageof a firm’s antihypertensive-related patents (since application)in a robustness test. Results remain robust with this alternativemeasure.

temporal heterogeneities. We lag all time-varyingindependent and control variables by one year.

Empirical model

Arguably, the controls above may not fullyeliminate omitted-variable bias, as there may stillbe unobserved features of a firm’s technologicalarea that both draw greater competition and neces-sitate less collaboration. Moreover, the controlsdo not fully address the possibility of reversecausality; that is, that by engaging in collaborativeinvention a firm may venture into new areaswhere competition from similar drugs is low.Both concerns may result in biased estimates dueto nonrandomized assignments of observationsto levels of the independent variable of interest(Holland, 1986). To address these concerns,we use the instrumental variable approach intwo-stage least squares (2SLS) estimation (Berryand Waldfogel, 2001; Wooldridge, 2002).

To instrument competition from similar drugsin the first stage, we use the period in the mid-1980s when there was a spike in merger andacquisition (M&A) activities in the pharmaceu-tical industry. Unlike M&As in the mid-1990s,which were largely motivated by firms accessingexternal technologies and accelerating pipelines ofnew drugs in the midst of impending patent expi-ration (Higgins and Rodriguez, 2006), M&As inthe mid-1980s mostly followed a rationale of cor-porate growth via best-selling drugs and furtherresearch (Achilladelis, 1999).12 Acquisitions in themid-1980s occurred at breakneck speed due tocorporate quest for greener pastures, momentumin restructuring programs, reassessments of assetvalues, and strategies of going private (ChemicalWeek , 1986). During this time, rumors of prospec-tive M&A activities generated huge swings instock prices (The New York Times , 1985), and phar-maceutical firms sought to quickly expand theirproduct lines through M&A activities13 and pur-sued potential acquisitions to make themselves less

12 This difference in M&A rationale is starkly represented inour sample: over our M&A period (1985–1987), there are nofirms in our sample that have antihypertensive-related patentsnear expiration, specifically, beyond 15 years since applicationdate. Hence, firm desperation due to patent expiration is likelynot an influence here.13 For instance, by acquiring Revlon’s USV Pharmaceutical andArmour Pharmaceutical operations, Rorer added the antihyper-tensives Hygroton, Regroton, and Lozol to its product line (TheNew York Times , 1985).



vulnerable to takeovers (Financial Times , 1985).This widespread M&A prospect created uncer-tainty that induced firms to hold back on marketlaunch of new drugs, especially in areas with majortherapeutic advances, and to focus instead onminor modifications of their drugs.14 This M&A-induced slowdown in the launch of major thera-peutic advances serves as a useful instrument toestimate differential changes in the levels of com-petition from similar antihypertensives. We use adummy variable taking the value of 1 for years1985–1987 (inclusive), and 0 otherwise, to cap-ture the period of spike in M&A activities.15 Wethen use a variable with the number of a firm’santihypertensives that received priority review sta-tus from the FDA, to indicate the extent to whichthe firm was working in areas that constitute poten-tial major breakthroughs or significant advances inhypertensive therapy.16 Finally, we multiply thesetwo variables, and the resulting interaction termserves to estimate differential changes in com-petition across firms over the M&A-induced fluxperiod.

In the first stage of the 2SLS estimation,we predict changes in Competition from similarantihypertensivesit with the instruments and con-trol variables:

Competition from similar antihypertensivesit = β0 + β1 M &A spike periodit

+ β2 Priority drugsit + β3 M &A spike periodit X Priority drugsit + βhControls + εijt (1)

In the second stage, we lag the independentvariables and estimate the main model:

Collaboration ratioit = δ0 + δ1 predicted (Competition from similar antihypertensivesit−1)

+ δh Controls + ξijt (2)

14 We appreciate valuable insights from industry practitioners onthis matter.15 To address the possibility that the M&A spike period mayhave had longer term effect on slowing down firms’ marketlaunch of major antihypertensives, we extend that period to four(1985–1988) and five (1985–1989) years in additional analyses.Findings remain fully robust.16 In sensitivity analysis, we dichotomized this variable (Cardand Krueger, 1994), capturing whether or not a firm had priority-status antihypertensives. Findings in both stages remain fullyrobust.

By using the 2SLS command in Stata, whichperforms the standard variance adjustments forthe coefficient of δ1 in the second stage, weobtain a consistent and efficient estimate for δ1(Wooldridge, 2002), which we use to test H1.To test the contingency effects in H2 and H3,the conventional approach is to multiply the mainindependent variable, competition (x1), with thecontingency variables, commercial abilities (x2)and technical abilities (x3), separately. However,in 2SLS estimations, this approach creates compli-cations in the variance adjustments. For instance,in the estimated coefficient on the interactionterm between competition and commercial abili-ties (x1 × x2), only the portion of variance arisingfrom predicted competition (predicted x1) requiresadjustments, while the portion arising from com-mercial abilities (x2) does not. To circumventthese complications, we use split-sample analy-ses (Penner-Hahn and Shaver, 2005). To test H2,we split the sample by the median of commer-cial abilities and obtained δ1

LC and δ1HC, capturing

respectively the effects of competition from sim-ilar drugs in observations with low (below themedian) and high (equal or above the median)levels of commercial abilities. We then performa t-test17 for the difference in these coefficients to

examine if the effect of competition differs acrossdifferent (low versus high) levels of the contin-gency variable. We follow a similar procedure totest H3.

17 The t-statistic was manually calculated using the formula:t = (δ1 − δ2)/{[(σ 1

2/n1) + (σ 22/n2)]0.5}, where δ is the coefficient

of estimated competition, σ is the standard deviation, and n isthe sample size for each of the subsamples.



RESULTS

Table 2 reports summary statistics and pairwisecorrelations. As this table shows, on average, about83 percent of a firm’s antihypertensive-relatedpatents involved collaborative invention. Firms inthe sample faced about 10 competing similar anti-hypertensives in a given year. The pairwise corre-lations between firms’ counts of antihypertensive-related patents and states (0.76), and between thetime elapsed and the number of mechanisms ofaction (0.84) are rather high. To ensure that multi-collinearity does not affect findings, we run addi-tional models separately dropping states and mech-anisms of action. Findings remain robust.

Our premise that firms readjust their collabo-rative invention level as a strategic response tocompetition implies that such level is somewhatfluid, as firms change it over time. Figure 1 plotsthe average of yearly within-firm changes in thecollaboration ratio and shows that there is clearlyvariance in such within-firm changes. In otherwords, firms do change the extent of collabora-tion on a yearly basis, sometimes as drastically as100 percent in either direction. Interestingly, formost years, both negative and positive changesoccur, suggesting that firms react heterogeneouslyto external impetus for change.

Figure 2 illustrates changes in competitionfrom similar antihypertensives over the M&Aspike period. As mentioned, we expect firmsoperating in areas with major therapeutic advancesto face lower levels of competition from similardrugs over this period in the mid-1980s. Asthis figure shows, although the total number ofantihypertensives on the market increases over theperiod of analysis, the breakdown of competitionfor firms with and without priority drugs revealspatterns consistent with our expectation. Overthe spike period, these two groups of firmsfaced notably divergent levels of competition fromsimilar drugs, suggesting that firms did indeedhold back from launching new drugs in areasmarked by major therapeutic advances, as ourearlier qualitative evidence suggests. This findinglends confidence to the appropriateness of theinstruments used to predict competition.

Table 3 shows results of the first stage of the2SLS estimation. Model 1 contains control vari-ables only. Models 2 and 3 add the instrumen-tal variables predicting competition from similardrugs, without and with robust errors, respectively. Ta

ble

2.D

escr

iptiv

est

atis

tics

and

pair

wis

eco

rrel

atio

ns

Var

iabl

eM

ean

S.D

.1

23

45

67

89

1011

12

1.C

olla

bora

tion

ratio

0.83

0.12

1.00

2.C

ompe

titio

nfr

omsi

mila

r10

.03

3.73

−0.0

61.

003.

Com

mer

cial

abili

ties

0.34

0.63

0.00

−0.4

31.

004.

Tech

nica

lab

ilitie

s0.

180.

190.

05−0

.59

0.32

1.00

5.A

ntih

yper

tens

ives

with

diff

eren

t41

.20

11.9

00.

44−0

.16

−0.1

50.

161.

006.

Ant

ihyp

erte

nsiv

es-r

elat

ed21

1.97

273.

92−0

.07

−0.1

30.

46−0

.03

−0.4

11.

007.

Tota

lpa

tent

s68

6.39

3400

.71

0.02

0.02

−0.0

6−0

.03

−0.0

30.

011.

008.

Stat

es11

.70

7.20

−0.0

4−0

.04

0.37

−0.1

5−0

.26

0.76

0.03

1.00

9.T

ime

elap

sed

13.6

85.

850.

640.

030.

05−0

.09

0.64

−0.0

20.

010.

071.

0010

.Pu

blic

atio

nsin

top

med

ical

1309

.72

404.

060.

260.

010.

07−0

.03

0.37

0.10

0.03

0.24

0.39

1.00

11.

Pate

nts

near

170.

4627

5.29

0.27

0.04

0.31

−0.1

2−0

.05

0.43

0.01

0.39

0.44

0.27

1.00

12.

Mec

hani

sms

of5.

470.

500.

530.

040.

12−0

.01

0.57

0.04

0.01

0.12

0.84

0.25

0.42

1.00

13.

Liti

gatio

ns15

.08

24.5

50.

350.

09−0

.16

−0.1

20.

46−0

.18

−0.0

1−0

.13

0.64

−0.0

30.

070.

54



-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002

With

in-f

irm

cha

nges

in c

olla

bora

tion

ratio

Year

Max

Min

Average

Figure 1. Within-firm changes in collaborative invention over time

0

10

20

30

40

50

60

70

0

5

10

15

20

25

30

1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003

Year

Com

petit

ion

from

sim

ilar

antih

hype

rten

sive

s

Tot

al n

umbe

r of

ant

ihyp

erte

nsiv

es

Total number of antihypertensives

Competition from similar antihypertensivesfor firms without priority drugs

Competition from similar antihypertensivesfor firms with priority drugs

M&A spike period

Figure 2. Competition from similar antihypertensives over time

The coefficient on the M&A spike period is gen-erally positive in Models 2 and 3, though not sig-nificant. More importantly, the coefficient on theinteraction term is significantly negative in Mod-els 2 and 3, indicating that the M&A spike perioddifferentially affected competition for firms operat-ing in areas with major therapeutic advances; thatis, they faced a slower rate of increase in com-petition during that period. This is in line withFigure 2 and consistent with our earlier ratio-nale that the prospect of M&A induced firms toslow down on market launch of antihypertensivesin areas with major therapeutic advances and tofocus instead on incremental drugs launches.18

18 In additional analysis, we apply a more stringent criterion,keeping only firms that existed both during and outside ofthe flux period, which further reduces sample size. First-stage findings are fully robust. Second-stage results show a

We use Model 2 to predict competition for thesecond stage.19

Table 4 reports second-stage estimations ofinfluences on firms’ collaboration ratio. Model 1comprises the full sample and tests H1. Models 2and 3 refer to the split-sample analysis that testsH2; similarly, Models 4 and 5 test H3. Models6–10, in turn, reproduce the analyses with firm-fixed effects,20 showing that the results reportedbelow are robust.

significantly negative effect of competition (p < 0.10; z = −1.83)and robust split-sample findings.19 We also run analysis using Model 3 to predict competitionfrom similar drugs. Results show a significantly negativeinfluence of competition (p < 0.10; z = −1.90). Findings for thesplit-sample remain fully robust.20 This approach is highly stringent given that the M&A shockvariables already account for potential omitted firm attributeswhen obtaining exogenous changes in competition. Moreover,the difference-in-difference estimator in the first stage based on



Table 3. First-stage regressions predicting competitionwith instrumental variables (N = 247)

Model 1 Model 2 Model 3a

M&A spike period 1.212 1.212(1.80) (2.33)

Priority drugs 0.0264 0.0264(0.024) (0.021)

M&A spikeperiod × prioritydrugs

−0.194***(0.074)

−0.194***(0.049)

Commercial abilities −1.531*** −1.616*** −1.616***−0.32 (0.33) (0.27)

Technical abilities −11.23*** −10.74*** −10.74***(1.61) (1.61) (3.08)

Antihypertensives withdifferent mechanisms

−0.183*** −0.176*** −0.176***(0.029) (0.029) (0.036)

Antihypertensives-relatedpatents

−0.0023**(0.0010)

−0.0024**(0.0010)

−0.0024**−0.00009

Total patents −0.000005 −0.000002 −0.000002(0.000047) (0.000047) (0.000035)

States −0.0533 −0.0451 −0.0451(0.043) (0.042) (0.040)

Time elapsed −0.0599 −0.0576 −0.0576(0.11) (0.11) (0.12)

Publications in topmedical journals

0.0038*** 0.0037** 0.0037**(0.0015) (0.0014) (0.0017)

Patents near expiration −0.0002 0.0001 0.0001(0.0009) (0.0009) (0.0007)

Mechanisms of action 2.243** 2.187** 2.187**(1.05) (1.03) (0.89)

Litigations 0.0387** 0.0386** 0.0386***(0.016) (0.016) (0.010)

Constant 4.187 3.991 3.991(6.14) (6.07) (6.36)

Year dummies Included Included Included

Two-tailed tests for all variables; standard errors in parentheses.*p < 0.10; **p < 0.05; ***p < 0.01.a Model 3 allows for robust errors.

H1 predicts that competition from rival productsbuilding on similar knowledge reduces a firm’ssubsequent collaborative invention involving thisknowledge. In support of H1, the coefficient oncompetition from antihypertensives building on thesame mechanism as a firm’s drug (−0.035) issignificantly negative in Model 1. Its economicmagnitude is substantial—one standard deviationin the number of rival drugs (3.73) reduces thedependent variable by about 13 percent, which is

firm attributes (whether a firm operates in areas of potentiallymajor breakthroughs or significant advances in hypertensivetherapy) further restricts such variance required for within-firmanalyses.

more than one standard deviation of the dependentvariable21 (12 percent).

H2 predicts that the negative effect of compe-tition on collaborative invention is greater whena firm has lower commercial abilities. Models 2and 3 in Table 4 show the influence of competi-tion from similar drugs for firms with low and highvalues of commercial abilities, respectively.22 Thecoefficient on competition is significantly negativein Model 2 but not in Model 3, suggesting thatcompetition reduces collaboration more notablyfor firms with lower commercial abilities. In sup-port of H2, a t-test of difference in coefficients ofcompetition across Models 2 and 3 reveals that theeffect of competition is indeed significantly morenegative for firms with lower commercial abilities(t = −32.39).23 H3, in turn, predicts that the neg-ative effect of competition on collaborative inven-tion is accentuated when a firm has lower technicalabilities. Models 4 and 5 similarly test this contin-gency effect. The coefficient on competition fromsimilar drugs is significantly negative for firmswith lower technical abilities (Model 4) but not forthose with high technical abilities (Model 5). Fur-ther, the t-test shows that such difference is indeedsignificant (t = −35.05).24 This result supports H3.

It is possible that firms take longer than ayear to change their collaboration in response to

21 Standard deviations for number of similar drugs and collabo-ration ratio are reported in Table 2.22 Due to reduced sample sizes, split-sample models droppedyear dummies. A concern may be that year-specific occurrencesare driving our results. However, these occurrences likely affectAntihypertensives with different mechanismsit−1 as well, whichwe control for. Moreover, our 2SLS model, by restrictingestimation to changes in Competitionit−1 over the flux period,further mitigates this concern. Alternatively, we try retainingthe year dummies and drop Mechanisms of actiont−1 instead.Findings remain fully robust.23 A potential concern is that the measure for commercialabilities may incorporate a firm’s technical abilities as well, sincea drug’s technical superiority also contributes to its eventualmarket success. We mitigate this problem by controlling fortechnical abilities in models examining the contingency effectsof commercial abilities.24 In Models 4 and 5, the significant coefficients for “Technicalabilities” with different signs across the split-samples hintthat this variable may have a curvilinear effect on thefirm’s collaborative invention. To ensure that our findingsare not affected by omission of the second-order effect of“Technical abilities,” we add its square term to all models inTable 4 in a robustness test. The square term is significantlypositive, indicating a U-shape for technical abilities’ effect.Importantly, our findings remain fully robust with this inclusion.Alternatively, we remove “Technical abilities” from the split-sample analyses to check if its presence was driving our findings.Again, findings remain fully robust.



Tabl

e4.

Seco

nd-s

tage

estim

ates

ofef

fect

ofco

mpe

titio

non

firm

’sco

llabo

rativ

ein

vent

ion

Rob

ustn

ess

mod

els

with

firm

-fixe

def

fect

s

Com

mer

cial

abili

ties

Tech

nica

lab

ilitie

sC

omm

erci

alab

ilitie

sTe

chni

cal

abili

ties

Full

sam

ple

Low

Hig

hL

owH

igh

Full

sam

ple

Low

Hig

hL

owH

igh

Mod

el1

Mod

el2

Mod

el3

Mod

el4

Mod

el5

Mod

el6

Mod

el7

Mod

el8

Mod

el9

Mod

el10

Com

petit

ion

from

sim

ilar

antih

yper

tens

ives

−0.0

35**

−0.0

46**

0.00

2−0

.038

***

−0.0

02−0

.025

**−0

.069

−0.0

12−0

.039

**−0

.004

(0.0

16)

(0.0

19)

(0.0

03)

(0.0

11)

(0.0

04)

(0.0

12)

(0.0

53)

(0.0

08)

(0.0

19)

(0.0

27)

t-te

stof

diff

eren

ceac

ross

mod

els

−32.

39−3

5.05

−13.

82−1

1.62

Com

mer

cial

abili

ties

−0.0

61**

−0.0

00−0

.05*

**−0

.085

*−0

.007

−0.0

17−0

.000

−0.0

525*

**0.

062*

−0.0

50**

*(0

.029

)(0

.000

)(0

.010

)(0

.047

)(0

.010

)(0

.014

)(0

.000

)(0

.018

)(0

.034

)(0

.014

)Te

chni

cal

abili

ties

−0.2

72−1

.886

**0.

061

−2.8

00**

*0.

119*

*−0

.234

−6.6

80−0

.121

−7.7

63**

0.16

2*(0

.20)

(0.8

8)(0

.042

)(1

.06)

(0.0

54)

(0.1

9)(4

.72)

(0.1

3)(3

.89)

(0.0

96)

Ant

ihyp

erte

nsiv

esw

ithdi

ffer

ent

mec

hani

sms

−0.0

07**

−0.0

06**

0.00

03−0

.004

**0.

002

−0.0

0012

0.00

6−0

.000

10.

009*

*0.

003

(0.0

03)

(0.0

02)

(0.0

01)

(0.0

02)

(0.0

01)

(0.0

016)

(0.0

06)

(0.0

03)

(0.0

04)

(0.0

03)

Ant

ihyp

erte

nsiv

es-r

elat

edpa

tent

s−0

.000

1*−0

.000

05−0

.000

04*

0.00

010.

0001

**−0

.000

1**

−0.0

004

−0.0

001*

*0.

0000

2−0

.000

1(0

.000

06)

(0.0

001)

(0.0

0002

)(0

.000

1)(0

.000

04)

(0.0

0005

6)(0

.000

3)(0

.000

05)

(0.0

001)

(0.0

0005

)To

tal

pate

nts

−0.0

0000

1−0

.000

001

0.00

0002

−0.0

0000

3−0

.000

002

−0.0

0000

2*−0

.000

003

−0.0

0000

01−0

.000

001

−0.0

0000

2(0

.000

002)

(0.0

0000

2)(0

.000

003)

(0.0

0000

2)(0

.000

002)

(0.0

0000

13)

(0.0

0000

2)(0

.000

003)

(0.0

0000

1)(0

.000

002)

Stat

es−0

.002

0.00

1−0

.001

−0.0

02−0

.005

***

0.00

020.

007

−0.0

0002

−0.0

02−0

.001

(0.0

02)

(0.0

03)

(0.0

01)

(0.0

02)

(0.0

02)

(0.0

02)

(0.0

07)

(0.0

02)

(0.0

02)

(0.0

02)

Tim

eel

apse

d0.

0132

***

0.01

9***

0.01

3***

0.01

6***

0.01

7***

0.01

4***

0.01

8***

0.00

9**

0.00

60.

014

(0.0

048)

(0.0

05)

(0.0

02)

(0.0

05)

(0.0

03)

(0.0

02)

(0.0

07)

(0.0

04)

(0.0

04)

(0.0

10)

Publ

icat

ions

into

pm

edic

aljo

urna

ls0.

0001

050.

0000

50.

0000

040.

0000

10.

0000

20.

0000

10.

0000

070.

0000

1−0

.000

04*

0.00

002

(0.0

0008

9)(0

.000

04)

(0.0

0002

)(0

.000

03)

(0.0

0002

)(0

.000

02)

(0.0

0003

)(0

.000

02)

(0.0

0002

)(0

.000

02)

Pate

nts

near

expi

ratio

n0.

0000

0518

−0.0

0004

0.00

004

0.00

001

−0.0

0001

−0.0

0003

−0.0

002

0.00

01−0

.000

10.

0000

2(0

.000

037)

(0.0

0005

)(0

.000

03)

(0.0

0005

)(0

.000

03)

(0.0

0003

)(0

.000

2)(0

.000

1)(0

.000

1)(0

.000

05)

Mec

hani

sms

ofac

tiona

0.06

370.

004

−0.0

01−0

.016

−0.0

49*

(0.0

58)

(0.0

40)

(0.0

23)

(0.0

35)

(0.0

29)

Liti

gatio

ns0.

001

0.00

01−0

.001

0.00

050.

0003

0.00

03−0

.000

3−0

.000

30.

0002

0.00

02(0

.000

9)(0

.000

5)(0

.000

6)(0

.000

4)(0

.000

6)(0

.000

3)(0

.000

6)(0

.000

6)(0

.000

3)(0

.000

5)C

onst

ant

0.92

5***

1.48

9***

0.72

1***

1.61

5***

0.79

3***

0.96

8***

1.93

9**

0.92

***

1.69

6***

0.55

3***

(0.2

7)(0

.32)

(0.0

93)

(0.2

70)

(0.1

20)

(0.1

3)(0

.90)

(0.1

0)(0

.50)

(0.0

65)

Yea

rdu

mm

ies

Incl

uded

Not

incl

uded

Not

incl

uded

Not

incl

uded

Not

incl

uded

Not

incl

uded

Not

incl

uded

Not

incl

uded

Not

incl

uded

Not

incl

uded

Obs

erva

tions

246

175

7113

111

524

617

571

131

115

*p<

0.10

,**

p<

0.05

,**

*p<

0.01

;tw

o-ta

iled

test

sfo

ral

lva

riab

les;

stan

dard

erro

rsin

pare

nthe

ses.

aM

odel

s6

–10

dono

tco

ntai

nva

riab

le“m

echa

nism

sof

actio

n”du

eto

redu

ced

with

in-fi

rmva

riat

ion.



competition. As robustness checks, we add two-year lag, and separately two- and three-year lags,of competition to all models in Table 4, and testfor joint significance of all lags in each model.In the split-sample, we test for differences in thejoint lags across models (Model 2 versus 3, andModel 4 versus 5). All results remain robust. It isalso possible that different mechanisms of actionthat firms use come with different availability of IPprotection and scientific knowledge, in ways thatconfound our findings but are not captured by ourcontrol variables (rivals’ litigations and scientificpublications). Additional analyses adding dummyvariables for mechanisms of action showed robustresults.

Validity of patent coauthorships as a measureof collaborative invention

Notwithstanding our earlier justifications of patentcoauthorship as a valid measure of collaborativeinvention, we conducted a variety of supplementalanalyses to further examine such validity. Specifi-cally, we examine the possibilities that firms mayreduce patent coauthorships for strategic reasons,that coauthorships may be picking up firms’ over-all inventive effort, or that other structural factorsmay exist that influence coauthorships.

First, one may conjecture that a firm lists on apatent only the inventor who most contributed toan invention, in fear of rivals’ attempts to inval-idate the patent by challenging the contributionsof coinventors. Such possibility should be morepronounced when rivals are litigious, and our anal-yses control for rivals’ litigiousness. Additionally,we examine the relationship between the num-ber of inventors listed on patents and the natureof inventions underlying patents. Our premise isthat inventions involving more inventors typicallyentail greater search scope. This alternative conjec-ture would suggest that such a relationship wouldnot be observed, since collaboration may haveactually occurred even in what one observes as asolo-inventor patent. We reconstruct the databaseat the patent level, and trace, for each patent, thepatents that it cites and identify their USPTO-assigned technology classes. Next, we run a seriesof regressions to test the effect of the numberof inventors listed in a patent on that patent’ssearch scope (i.e., unique technology classes itcited). Findings in Table 5 show a persistent pos-itive effect across models; that is, the greater the

number of inventors listed on patents, the widerthe search scope. This finding provides additionalevidence that inventors listed on patents do corre-spond with collaboration underlying the inventionprocess, contrary to the conjecture that firms arestrategically listing fewer inventors on patents.

We further examine this conjecture by checkingauthorships on scientific publications associatedwith patents. The logic is that firms are less likelyto strategically manipulate authorships on scien-tific publications, since these do not face the samerisk of subsequent invalidation that patents face.The patent experts we consulted raised a potentialissue that scientific publications about an inven-tion may very well comprise authors who are notcoinventors of the respective patent. For example,an individual who collaborates in other phases ofan invention beyond its conception may qualify asa coauthor in a publication about the invention,but does not qualify as a patent’s coinventor.25

This issue notwithstanding, evidence that scien-tific publications associated with a solo-inventortend to have a limited number of coauthors shouldhelp mitigate the concern. We identify 49 solo-inventor patents in our sample and search formatching scientific publications for each of thesepatents, following the patent-paper matching pro-cedure adopted in prior work26 (Murray, 2002). Wefind matching scientific publications to only foursolo-inventor patents. This is not surprising, giventhat inventions with solo-inventors tend to be nar-rower in scope and thus less likely to warrant anovel scientific contribution. Importantly, in onlyone of these cases did the inventor of a patent haveanother collaborator from the same department ascoauthor of the associated publication. This evi-dence bolsters confidence that solo-author patentsare not associated with larger collaborative efforts.

A related issue is that firms may split the out-comes of a large collaborative project into multiplesolo-inventor patents, despite the abovementionedlegal considerations. This interpretation suggests

25 An individual who implements an inventive idea or partici-pates in its reduction to practice may quality as a coauthor in ascientific publication, but not as a coinventor of the respectivepatent (USPTO, 2012).26 Murray (2002) proposed a single patent-pair match; that is,she linked patent 5,041,138 (neomorphogenesis of cartilage invivo from cell culture) to a publication by Vacanti et al. (1991).Our procedure involved identifying all scientific publications(co)authored by the inventor of each patent, followed by analysisof patent and publication abstracts to ascertain whether theircontents overlapped.



Table 5. Patent-level effect of number of inventors on search scope

Model specification

Negativebinomialwith yeardummies

Negativebinomial withyear dummies,robust errors

Negativebinomial with

year and technologyclass dummies

Poissonwith yeardummies

Poissonwith year and

technology classdummies

Random-effectsPoisson withyear dummies

Number ofinventors onpatent

0.0153*** 0.0153*** 0.0149*** 0.0153*** 0.0149*** 0.0153***

(0.0033) (0.0035) (0.0033) (0.0033) (0.0033) (0.0033)

Two-tailed tests for all variables; standard errors in parentheses. *p < 0.10; **p < 0.05; ***p < 0.01.

these firms will have more patents, all else equal.Our analyses address this issue by controllingfor a firm’s number of antihypertensive-relatedpatents, as well as a firm’s total patents (span-ning beyond antihypertensive therapy). Further, weconsider that, if our proposed effect of competi-tion on reducing collaboration is only driven byfirms splitting collaborative projects into multiplesolo-inventor patents, then this effect will likelydisappear in a subset of firms with a higher numberof patents. We split the sample along the medianof firms’ antihypertensive patents in a previousyear, retained the subsample of firms with morepatents, and rerun the analyses.27 Findings remainfully robust, further assuaging this concern.

A potential concern is that the measure for col-laborative invention might simply reflect a firm’sinventive effort. For instance, a firm might listinventors on the same patent even when the inven-tors separately create the same invention withoutcollaboration. This is more likely to happen whena firm’s overall inventive effort increases, whichstresses the need to control for inventive effort. Aswe clarified earlier, our analyses control for both afirm’s antihypertensive-related patents, as well astotal patents. Also, we find that the pairwise cor-relation between the collaboration ratio measureand a firm’s number of antihypertensive-relatedpatents is 0.0047 and nonsignificant (p > 0.10).This provides reasonable assurance that a firm’sinventive effort is likely not confounding ourdependent variable. Furthermore, if our findingsare only driven by inventive effort, then ourfindings should diminish in the subset of firmsthat are unconstrained, that is, have a large number

27 Due to the reduced sample size, the year dummies are droppedfrom the second-stage estimation.

of total patents. However, we have shown that ourfindings remain robust in the subset of firms witha higher number of patents (see earlier issue).We further entertain the possibility that increasedR&D expense may increase a firm’s numberof inventors, which increases the likelihood ofinventors coming up with the same idea and hencethe incidences of patent coauthorship withoutactual collaboration. We control for a firm’snumber of inventors involved in antihypertensivesin a robustness test. Findings remain fully robust.

Another concern is that, despite the legalrequirements for inventors listed on patents, coau-thorships might in a few cases reflect a hierarchicalstructure where managers’ names appear in patentseven when they did not substantively contribute tothe inventions. This scenario likely does not con-found our findings, as it is not clear why this inclu-sion would systematically vary with the relativechange in competition between different types offirms over the M&A flux period. Further address-ing this concern, we note that if coauthorshipsindeed reflect managers’ inclusion in patents, thena firm operating with more mechanisms of actionwithin antihypertensives and hence needing moremanagers to coordinate across them should exhibitgreater coauthorships. We include a firm’s numberof mechanisms of action in the second-stage esti-mation (Table 4) in a robustness test, and find that,in the full model, the coefficient on such variableis in fact significantly negative (z = −2.20) ratherthan positive. All earlier findings remain robust,providing additional assurance that this conjecturedoes not affect the findings.

Finally, although our research question aboutcollaboration versus solo inventors is best testedwith a dependent variable capturing the proportionof collaborative inventions in a firm’s patents,



the mechanisms underlying our propositions canarguably be extended to explain team size incollaborative invention as well. Specifically, thegreater need for search depth and invention speedthat competition spurs may induce a firm tochoose fewer inventors in collaborative projects.As a robustness check, we replace the dependentvariable with the average number of inventorson a firm’s relevant patents in the observationyear. This dependent variable has a mean andstandard deviation of 2.87 and 0.91, respectively,and ranges between 1 and 9.4. The coefficient ofcompetition remains negative (p < 0.1), and thesplit-sample tests remain fully robust (t = −25.17,p < 0.01 for H2 and t = −25.11, p < 0.01 for H3).

DISCUSSION AND CONCLUSION

Our central proposition is that competition fromsimilar products diminishes a firm’s propensityto engage in collaborative invention. Greatercollaborative invention corresponds with moreexploratory search and requires more time for afirm to overcome the attendant collaboration dif-ficulties. Competition from similar products com-pels a firm to favor search depth over exploratorysearch and to respond expeditiously, thereby push-ing it away from collaborative invention towardhaving its inventors work more individually. Fur-thermore, we argue that this main effect is moreprominent when a firm is weaker in commercialor technical abilities, because such weaknessesdiminish the attractiveness of exploratory searchand aggravate a firm’s time pressure to respond tocompetition. In support of our central proposition,we find that the greater the number of rival anti-hypertensives building on the same mechanism ofaction as a firm’s antihypertensive, the less thatfirm engages in collaborative invention involvingknowledge of that mechanism. Findings also showthat this effect of rival drugs is accentuated whenthat firm lacks commercial or technical abilities.

Implications for management practice

This study furthers our understanding of chal-lenges that managers face when organizinginventive effort. From a managerial standpoint,prior research stressing the benefits of collabora-tive invention may have been taken as a messagefor managers to always foster collaboration

between inventors. We stress in contrast thatcollaborative invention is not a universallysuperior practice. Instead, managers can exerciseconsiderable discretion over how to organize theirinventors so as to balance the tradeoff betweenthe benefits that collaboration produces and theadditional difficulties it entails. Importantly, afirm’s idiosyncratic situation affects such tradeoff.As our findings reveal, the level of competitionthat a firm faces plays a substantial role inexplaining variations in the level of collaborativeinvention. Specifically, with intense competition,managers move away from collaboration towardsolo inventions. While this forgoes benefits ofcollaborative invention, it ensures timely andpointed responses to the high technical bars thathave been raised as rivals become increasinglyentrenched in a particular technology. Conversely,when a firm faces less intense competition,managers are in a better position to take oncollaboration difficulties in exchange for a broaderarray of potential knowledge combinations.

Theoretical contributions and opportunitiesfor future research

This study has significant implications for thefirm heterogeneity puzzle at the core of strategyresearch. In examining why firms in the sameproduct market perform differently, scholars pro-posed that downstream performance differentialsarise from a nonuniform distribution of upstreamfirm-specific resources (Barney, 1991, Wernerfelt,1984). To explain this distribution, prior studieshave examined how firms search for resourcesdifferently (Ahuja and Katila, 2004) and howthey organize their upstream inventive activitiesto facilitate this search (Nickerson and Zenger,2004). The implicit approach along this vein is toattribute resource differences to some other form oforganizational heterogeneity upstream, ultimatelytracing upstream heterogeneity to some form ofinitial endowment differences at firms’ inceptions.This study, in contrast, reverses the focus frompreexisting upstream heterogeneity to competitionin downstream markets as a driver of resourceheterogeneity across firms. Such reversal is theo-retically meaningful because it potentially circum-vents the need for researchers to perpetually searchupstream for the seemingly elusive source of het-erogeneity. Rather, it suggests that perhaps firmheterogeneity is really shaped by how downstream



competition unfolds at various points in time foreach firm.

Besides offering a different approach to explainhow firm heterogeneity arises, this study also raisesquestions about why firm heterogeneity persists.Equally puzzling as the origin of firm differencesis the persistence of such differences. Dissimi-lar firms within an environment do not alwaysconverge subsequently toward homogeneity. Con-ceptually, for convergence to occur, there musteither be resource transfer across firms or firmsmust independently create similar resources sub-sequently. To explain why resource transfer orimitation can be difficult, prior studies have arguedthat resources are “sticky,” and organizational fea-tures such as complexity further impede resourcetransfer (Liebeskind, 1996; Rivkin, 2000). Withrespect to why firms do not subsequently createsimilar resources, predominant explanations cen-ter on firms’ tendencies toward path dependenceand local search (Cyert and March, 1963), due tobounded rationality, inertia, and routines (Nelsonand Winter, 1982). This study, in turn, hints atan alternative approach to explain this persistence,by showing how a firm’s commercial or technicalabilities affect its response to competition. Whencompetition causes a firm with weaker abilities toreduce more of its exploratory search, this firmmay be compromising its inventive potential andsubsequent abilities in the long run, which iron-ically renders it more vulnerable to competitionin the future. Consequently, as competition pushesfirms with differential abilities further apart, smallinitial differences in abilities may spiral into sig-nificant heterogeneity over time.

This study prompts further questions about howelse firms may engage in exploratory search. Ourmain proposition is that competition reduces firms’priority and feasibility of exploratory search andshifts their focus away from collaborative inven-tion toward greater search depth. Yet, it is possiblethat a firm facing competition, while diminishingits exploratory search via collaborative invention,may turn to other avenues of exploratory search,such as alliances, licensing, or tie-ups with uni-versities, in ways that entail less commitment ofresources.

Another avenue for future research is to exam-ine the extent that our propositions are applicableacross industries. Our propositions presume afirm’s ability to identify rival products building onsimilar knowledge. In many industries, competing

platforms, paradigms, or fundamental technologiesare indeed clearly identifiable. Yet, there are plau-sibly other instances, such as during early stagesof a technology, where the boundaries of platformsare not clearly defined. In these instances, firmsmight be unable to observe if rival products areeroding market opportunities and requiring themto engage in search depth as we theorize. Further,our study examines a setting where patent pro-tection is relatively strong. Arguably, the effectsof competition that we theorize are even strongerin contexts where firms cannot rely on patents todeter the emergence of similar products and arethus exposed to greater levels of competition. Weencourage future research to explore firms’ collab-oration propensities in these instances.

Finally, this study points to opportunities forfuture research on how competition, a centralconstruct in strategy research, affects firms. Priorresearch has highlighted competition’s impacton firms’ decisions regarding what markets toenter and how much inventive effort to investin these markets. However, researchers havelargely neglected the next step of examining howcompetition affects a firm’s use of its inventors’existing knowledge to create new knowledge.Addressing this gap may be crucial as such processdetermines a firm’s resource accumulation andconsequently a firm’s growth (Penrose, 1959).Through this paper, we hope to kick start thisexamination. A potential point of departure fromexisting research is that competition does not drivea firm out of a market or force it to give upits existing knowledge within this market. Rather,it induces a firm to reorient its approach towardexploiting this knowledge.

In sum, by showing that competition influenceswhether a firm utilizes its individual inventorsmore as a collection of solo inventors or closer to aset of collaborative inventors, this study indicatesfecund research opportunities on firms’ utilizationof knowledge and their resulting processes ofresource accumulation and growth. We believethat by reversing prior research’s focus on howupstream resources influence a firm’s positionin downstream markets and also by consideringhow downstream competition shapes a firm’supstream resource utilization, we can attain amore comprehensive picture of how firms developand accumulate resources. We hope that thepossibilities we delineated above will stimulatesuch stream reversal in future studies, leading to



better understanding of the origin and persistenceof firm heterogeneity.

ACKNOWLEDGEMENTS

We acknowledge research funding from the Carl-son School of Management at the University ofMinnesota, and the McCombs School of Busi-ness at the University of Texas at Austin. Weappreciate helpful comments from Nick Argyres,Victor Bennett, Susan Cohen, Martin Ganco,Aseem Kaul, Alfie Marcus, Will Mitchell, Clau-dia Schoonhoven, Myles Shaver, Andy Van deVen, and two anonymous reviewers. All remainingerrors are our own.

REFERENCES

Achilladelis B. 1999. Innovation in the pharmaceuticalindustry. In Pharmaceutical Innovation: Revolutioniz-ing Human Health , Landau R, Achilladelis B, Scri-abine A (eds). Chemical Heritage Press: Philadelphia,PA; 1–147.

Aghion P, Tirole J. 1994. The management of innovation.Quarterly Journal of Economics 109(4): 1185–1209.

Ahuja G, Katila R. 2001. Technological acquisitionsand the innovation performance of acquiring firms: alongitudinal study. Strategic Management Journal 22:197–220.

Ahuja G, Katila R. 2004. Where do resources comefrom? The role of idiosyncratic situations. StrategicManagement Journal 25: 887–907.

Almeida P, Kogut B. 1999. Localization of knowledgeand the mobility of engineers in regional networks.Management Science 45(7): 905–917.

Anderson P, Tushman ML. 1990. Technological discon-tinuities and dominant designs: a cyclical model oftechnological change. Administrative Science Quar-terly 35: 604–633.

Avorn J, Chen M, Hartley R. 1982. Scientific versuscommercial sources of influence on the prescribingbehavior of physicians. American Journal of Medicine73(1): 4–8.

Azoulay P. 2002. Do pharmaceutical sales respondto scientific evidence? Journal of Economics andManagement Strategy 11(4): 551–594.

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

Becker G, Murphy K. 1992. The division of labor,coordination costs, and knowledge. Quarterly Journalof Economics 107(4): 1137–1160.

Berry S, Waldfogel J. 2001. Do mergers increase productvariety? Evidence from radio broadcasting. QuarterlyJournal of Economics 116(3): 1009–1025.

Bodewitz HJHW, Buurma H, de Vries GH. 1987. Regu-latory science and the social management of trust in

medicine. In The Social Construction of TechnologicalSystems: New Directions in the Sociology and Historyof Technology , Bijker WE, Hughes TP, Pinch TJ (eds).MIT Press: Cambridge, MA; 243–260.

Card D, Krueger A. 1994. Minimum wages and employ-ment: a case study of the fast-food industry in NewJersey and Pennsylvania. American Economic Review84(4): 772–793.

Chemical Week . 1986. Who owns whom: the mergers of’85. Chemical Week 1 January: Management 50.

Christensen CM. 1997. The Innovator’s Dilemma . Har-vard Business School Press: Boston, MA.

Clarkson G, Toh PK. 2010. ‘Keep out’ signs: the role ofdeterrence in the competition for resources. StrategicManagement Journal 31(11): 1202–1225.

Cockburn IM, Henderson RM. 1998. Absorptive capacity,coauthoring behavior and the organization of researchin drug discovery. Journal of Industrial Economics46(2): 157–182.

Cohen W, Klepper S. 1996. Firm size and the natureof innovation within industries: the case of processand product R&D. Review of Economics and Statistics78(2): 232–243.

Cohen WM, Nelson R, Walsh J. 2000. Protecting theirintellectual assets: appropriability conditions and whyUS manufacturing firms patent (or not)? NBERWorking paper 7552, National Bureau of EconomicResearch, Cambridge, MA.

Corredoira R, Rosenkopf L. 2010. Should auld acquain-tance be forgot? The reverse transfer of knowledgethrough mobility ties. Strategic Management Journal31: 159–181.

Cyert RM, March J. 1963. A Behavioral Theory of theFirm . Prentice-Hall: Englewood Cliffs, NJ.

Dahlin K, Taylor M, Fichman M. 2004. Today’s Edisonsor weekend hobbyists: technical merit and success ofinventions by independent inventors. Research Policy33(8): 1167–1183.

Danzon PM. 2000. The pharmaceutical industry. InThe Encyclopedia of Law and Economics , Bouck-aert B, Geest G (eds). Edward Elgar: Cheltenham;1055–1091.

Dasgupta P. 1988. Patents, priority and imitation, or theeconomics of races and waiting games. EconomicJournal 98(389): 66–80.

Demsetz H. 1973. Industry structure, market rivalry, andpublic policy. Journal of Law and Economics 16:1–10.

Dixit A. 1980. The role of investment in entry deterrence.Economic Journal 90(357): 95–106.

Doraszelski U. 2003. An R&D race with knowledgeaccumulation. RAND Journal of Economics 34(1):20–42.

Dosi G. 1982. Technological paradigms and technologicaltrajectories. Research Policy 11: 147–162.

Edmondson A, Nembhard IM. 2009. Product develop-ment and learning in project teams: the challenges arethe benefits. Journal of Product Innovation Manage-ment 26: 123–138.

Eisenberg R. 2001. Bargaining over the transfer of pro-prietary research tools: is this market failing or emerg-ing?. In Expanding the Boundaries of Intellectual



Property: Innovation Policy for the Knowledge Soci-ety , Dreyfuss RC, Zimmerman DL, First H (eds).Oxford University Press: Oxford; 223–250.

Eisenhardt K, Tabrizi B. 1995. Accelerating adaptiveprocesses: product innovation in the global com-puter industry. Administrative Science Quarterly 40:84–110.

Felin T, Hesterly W. 2007. The knowledge-based view,nested heterogeneity, and new value creation: philo-sophical considerations on the locus of knowledge.Academy of Management Review 32(1): 195–218.

Financial Times. 1985. Rorer to buy Revlon drugs unit.3 December: International companies and finance: 20.

Fleming L. 2001. Recombinant uncertainty in technolog-ical search. Management Science 47(1): 117–132.

Fleming L, Mingo S, Chen D. 2007. Collaborativebrokerage, generative creativity, and creative success.Administrative Science Quarterly 52(3): 443–475.

Fleming L, Sorenson O. 2004. Science as a map intechnological search. Strategic Management Journal25: 909–928.

Gilbert R, Newbery D. 1982. Preemptive patenting andthe persistence of monopoly. American EconomicReview 72: 514–526.

Grant RM. 1996. Toward a knowledge-based theory ofthe firm. Strategic Management JournalSpecial Issue17: 109–122.

Green J, Scotchmer S. 1995. On the division of profitin sequential innovation. Rand Journal of Economics26(1): 20–33.

Griliches Z. 1990. Patent statistics as economic indica-tors: a survey. Journal of Economic Literature 28:1661–1707.

Hausman J, Hall B, Griliches Z. 1984. Econometric mod-els for count data with an application to the patents-R&D relationship. Econometrica 52: 909–938.

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

Henderson R, Cockburn I. 1994. Measuring competence?Exploring firm effects in pharmaceutical research.Strategic Management Journal 15: 63–84.

Higgins MJ, Rodriguez D. 2006. The outsourcing of R&Dthrough acquisitions in the pharmaceutical industry.Journal of Financial Economics 80: 351–383.

Holland P. 1986. Statistics and causal inference (in theoryand methods). Journal of the American StatisticalAssociation 81(296): 945–960.

Jehn KA, Northcraft GB, Neale MA. 1999. Whydifferences make a difference: a field study ofdiversity, conflict, and performance in workgroups.Administrative Science Quarterly 44: 741–763.

Katila R, Ahuja G. 2002. Something old, somethingnew: a longitudinal study of search behavior andnew product introduction. Academy of ManagementJournal 45(6): 1183–1194.

Kretschmer T, Puranam P. 2008. Integration throughincentives within differentiated organizations. Orga-nization Science 19(6): 860–875.

Lawrence PR, Lorsch JW. 1967. Differentiation andintegration in complex organizations. AdministrativeScience Quarterly 12(1): 1–47.

Lerner J. 1995. Patenting in the shadow of competitors.Journal of Law and Economics 38(2): 463–495.

Levinthal D. 1997. Adaptation on rugged landscapes.Management Science 43(7): 934–950.

Liebeskind JP. 1996. Knowledge, strategy, and the theoryof the firm. Strategic Management JournalSpecialIssue 17: 93–107.

Mansfield E, Schwartz M, Wagner S. 1981. Imitationcosts and patents: an empirical study. EconomicJournal 91(364): 907–918.

March J. 1991. Exploration and exploitation in organiza-tional learning. Organization Science 2(1): 71–87.

Martin X, Mitchell W. 1998. The influence of localsearch and performance heuristics on new designintroduction in a new product market. Research Policy26: 753–772.

Mitchell W. 1989. Whether and when? Probability andtiming of incumbents’ entry into emerging industrialsubfields. Administrative Science Quarterly 34(2):208–230.

Mowery DC, Oxley JE, Silverman BS. 1998. Strategicalliances and interfirm knowledge transfer. StrategicManagement Journal, Winter Special Issue 17: 77–91.

Murray F. 2002. Innovation as co-evolution of scientificand technological networks: exploring tissue engineer-ing. Research Policy 31: 1389–1403.

Murray F, O’Mahoney S. 2007. Exploring the foundationsof cumulative innovation: implications for organiza-tion science. Organization Science 18: 1006–1021.

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

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

Nickerson JA, Zenger TR. 2004. A knowledge-basedtheory of the firm – the problem-solving perspective.Organization Science 15(6): 617–632.

Penner-Hahn P, Shaver JM. 2005. Does internationalresearch and development increase patent output? Ananalysis of Japanese pharmaceutical firms. StrategicManagement Journal 26: 121–140.

Penrose ET. 1959. The Theory of the Growth of the Firm .Wiley: New York.

Pisano GP. 1990. The R&D boundaries of the firm: anempirical analysis. Administrative Science Quarterly35(1): 153–176.

Polidoro F, Theeke M. 2012. Getting competition downto a science: the effects of technological competitionon firms’ scientific publications. Organization Science23(4): 1135–1153.

Polidoro F, Toh PK. 2011. Letting rivals come close orwarding them off? The effects of substitution threat onimitation deterrence. Academy of Management Journal54(2): 369–392.

Priem RL, Butler JE. 2001. Is the resource-based view auseful perspective for strategic management research?Academy of Management Review 26(1): 22–40.

Reinganum J. 1985. Innovation and industry evolution.Quarterly Journal of Economics 100(1): 81–99.

Reuben BG, Wittcoff HA. 1989. Pharmaceutical Chemi-cals in Perspective. Wiley: New York.



Rivkin J. 2000. Imitation of complex strategies. Manage-ment Science 46(6): 824–844.

Rosenkopf L, Almeida P. 2003. Overcoming local searchthrough alliances and mobility. Management Science49(6): 751–766.

Sanchez R, Mahoney JT. 1996. Modularity, flexibility,and knowledge management in product and organi-zation design. Strategic Management JournalSpecialIssue 17: 63–76.

Schweitzer SO. 1997. Pharmaceutical Economics andPolicy . Oxford University Press: New York.

Scriabine A. 1999. Discovery and development ofmajor drugs currently in use. In PharmaceuticalInnovation: Revolutionizing Human Health , Landau R,Achilladelis B, Scriabine A (eds). Chemical HeritagePress: Philadelphia, PA.

Singh J, Fleming L. 2010. Lone inventors as sources ofbreakthroughs: myth or reality? Management Science56(1): 41–56.

Szulanski G. 1996. Exploring internal stickiness: imped-iments to the transfer of best practice within the firm.Strategic Management Journal 17: 27–43.

The New York Times . 1985. Rorer buys drug unit ofRevlon. The New York Times 30 November: FinancialDesk 29.

Thomke S, Kuemmerle W. 2002. Asset accumulation,interdependence and technological change: evidencefrom pharmaceutical drug discovery. Strategic Man-agement Journal 23(7): 619–635.

Thompson JD. 1967. Organizations in Action . Mcgraw-Hill: New York.

Tushman M, Anderson P. 1986. Technological disconti-nuities and organizational environments. Administra-tive Science Quarterly 31: 436–465.

USPTO. 2012. Patent Laws United States Code Title35 – Patents. Available at: http://www.uspto.gov/web/offices/dcom/olia/aipa/PatLaws1214.pdf (accessed 04March 2013).

Vacanti CA, Langer R, Schloo B, Vacanti JP. 1991.Synthetic polymers seeded with chondrocytes providea template for new cartilage formation. Plastic andReconstructive Surgery 88(5): 753–759.

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

Williamson O. 1975. Markets and Hierarchies . PrenticeHall: Englewood Cliffs, NJ.

Wooldridge JM. 2002. Econometric Analysis of CrossSection and Panel Data . The MIT Press: London,UK.

Wuchty S, Jones BF, Uzzi B. 2007. The increasingdominance of teams in production of knowledge.Science 316: 1036–1039.

Yayavaram S, Ahuja G. 2008. Decomposability in knowl-edge structures and its impact on usefulness of inven-tions and knowledge-base malleability. AdministrativeScience Quarterly 53(2): 333–362.


Date post:	19-Feb-2016
Category:	Documents
Upload:	mirianalbertpires
View:	3 times
Download:	0 times

A Competition-based Explanation of Collaborative Invention With the Firm

Documents