bridging the mutual knowledge gap: coordination and … · 2017. 5. 26. · bridging the mutual...

BRIDGING THE MUTUAL KNOWLEDGE GAP:COORDINATION AND THE COMMERCIALIZATION

OF UNIVERSITY SCIENCE

REDDI KOTHASingapore Management University

GERARD GEORGEImperial College London

KANNAN SRIKANTHIndian School of Business

We examine why commercialization of interdisciplinary research, especially fromdistant scientific domains, is different from commercialization of inventions fromspecialized or proximate domains. We argue that anticipated coordination costs aris-ing from the need to transfer technology to licensee firms and from the need for aninventor team’s members to work together to further develop a technology significantlyimpact commercialization outcomes. We use a sample of 3,776 university inventiondisclosures to test whether variation in the types of experience of the scientists on ateam influences the likelihood that an invention will be licensed. We proffer evidenceto support our hypotheses that anticipated coordination costs influence whether aninvention is licensed and that specific forms of team experience attenuate such coor-dination costs. The implications of these findings for theories of coordination, inno-vation, and entrepreneurship are discussed.

Organizations exist to achieve joint action, de-fined as individuals working together to achieve a

desired outcome. Coordination is the problem ofaligning actions so they are synchronized toachieve this objective. The ability of interdepen-dent actors to coordinate actions stems from ade-quate mutual knowledge that enables individualsto act as if they can predict others’ actions (Cam-erer, 2003; March & Simon, 1958; Puranam,Raveendran, & Knudsen, 2012). The traditionalview is that coordination problems are strictly aconsequence of misaligned incentives and that,when appropriate incentives are provided, suchproblems should disappear (Holmstrom & Mil-grom, 1994). This incentive-based approach con-flicts with findings that coordination problems per-sist even when incentives are aligned (Camerer,2003; Faems, Janssens, Madhok, & Van Looy, 2008;Gulati, Lawrence, & Puranam, 2005). Specifically,scholars from the knowledge-based view (KBV) ar-gue that a significant proportion of joint productionfailures in organizations occur because of inabilityto synchronize joint efforts (Grant, 1996; Heath &Staudenmayer, 2000; Postrel, 2009), either becauseof inadequate mutual knowledge or difficulty increating such knowledge. In the context of commer-cializing discoveries, we examine how different

An earlier version of this article received the Distin-guished Paper Award of the Business Policy and StrategyDivision of the Academy of Management in 2010. Wethank the staff of the Wisconsin Alumni Research Foun-dation (WARF) for their generous contribution of timeand their willingness to share data, ideas, expertise, andorganizational memory. In particular, we thank Carl Gul-brandsen, Michael Falk, Howard Bremer, Ken Lutz, andKristi Sullivan. Our manuscript benefited from the com-ments of Olav Sorenson, Ed Zajac, and seminar partici-pants at Copenhagen Business School, Emory University,Erasmus University, IESE Barcelona, London School ofEconomics, and Singapore Management University.Reddi Kotha acknowledges the support of SingaporeManagement University research grant for part fundingof this study (07-C207-SMU-030). Gerard George ac-knowledges the support of the Professorial Fellowship ofthe UK’s Economic and Social Research Council (RES-051-27-0321) and the Rajiv Gandhi Centre for Innovationand Entrepreneurship at Imperial College London. Kan-nan Srikanth acknowledges the support of the StrategicOrganization Design Unit, University of SouthernDenmark.

498

� Academy of Management Journal2013, Vol. 56, No. 2, 498–524.http://dx.doi.org/10.5465/amj.2010.0948

Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s expresswritten permission. Users may print, download, or email articles for individual use only.

kinds of inventor experience help mitigate coordi-nation problems under varying levels of mutualknowledge.

The scientific discovery process has increasinglybecome a team production function, with a gradualshift in the locus of invention from solo inventorsto teams and networks of inventors (Paruchuri,2010). Using a comprehensive sample of 19 millionpublications and 2 million patents spanning fivedecades, Wuchty, Jones, and Uzzi (2007) found thatteams generate far more innovations than solo in-ventors, that large teams are commonplace, andthat teams generate more heavily cited (more im-pactful) research. From an examination of patentrecords, Singh and Fleming (2010) confirmed thatteams are more likely to produce breakthrough in-ventions than solo inventors. Given a large scien-tific base, longer educational training in progres-sively narrower specializations is necessary topush the knowledge frontier (Jones, 2009). Conse-quently, the proliferation of team-based inventionsis driven largely by two trends: first, the need forspecialization to push the scientific frontier andsecond, the need for interdisciplinary work to solvechallenging scientific and practical problems.Whereas teams facilitate and focus use of collectiveknowledge to solve a problem, concomitant coor-dination costs incurred when teams perform thesecomplex tasks hinder team activity (Cummings &Kiesler, 2005; Guimerà, Uzzi, Spiro, & Nunes Ama-ral, 2005). Though solving complex scientific prob-lems can enhance social benefit (Basalla, 1988), theproductivity of such efforts is necessarily dimin-ished by their attendant coordination losses(Phelps, 2010; Weitzman, 1998). Therefore, how togainfully combine different knowledge domains isa core problem in both the innovation and organi-zation literatures (Vural, Dahlander, & George,2013; Zahra & George, 2002).

Drawing from the knowledge-based view, we ap-proach technology commercialization as a jointproduction effort involving multiple specialistswho are required to coordinate effort to performcomplex tasks (Grant, 1996; Kogut & Zander, 1992;Lawrence & Lorsch, 1967). Though knowledge in-tegration can be difficult and costly (Kapoor & Lim,2007), it is essential for the commercialization ofimpactful technology (George, Kotha, & Zheng,2008; Kotha, Zheng, & George, 2011), although thisfacet is often neglected. Studies that draw on evo-lutionary theories argue that the inventions thatbuild on the same knowledge domain have limitedupside potential, whereas those that build on dif-

ferent, complementary knowledge domains couldpotentially be valuable (Fleming, 2001, 2004; Flem-ing & Sorenson, 2004; Makri, Hitt, & Lane, 2010).For example, combining knowledge domains suchas molecular biology and photolithography canyield fundamental breakthroughs such as DNA ar-rays that transform genetic sequencing and map-ping capabilities.

Scholars from organization theory, social psy-chology, and sociology have pointed out that it isextremely difficult for specialists from differentknowledge domains to work together fruitfully.This is because specialists have different knowl-edge, beliefs, languages, and norms of working, sothat it becomes difficult for them to understandeach other and effectively complement each others’knowledge (Cronin & Weingart, 2007; Dougherty,1992; Heath & Staudenmayer, 2000; Latour, 1986).This suggests that the most desirable recombina-tions may also be difficult to fructify (Fleming,2004; Phelps, 2010; Vural et al., 2013). The desir-ability-difficulty trade-off makes it challenging, exante, to understand how the likelihood of commer-cialization varies with the underlying science dis-tance. Science distance in an invention is the dis-tance between multiple knowledge domains onwhich the invention is based. We argue that antic-ipated coordination costs in the joint productiontask of technology commercialization help explainwhy commercialization of interdisciplinary re-search, especially from distant domains, differsfrom the commercialization process for inventionsfrom specialized or proximate domains.

Using insights from the KBV on coordinationcosts and from individual/group learning theorieson the role of experience in reducing coordinationcosts, we add to the literatures on the productionand commercialization of science as well as tobroader discussions of coordination in teams. Wecontribute to the innovation literature by distin-guishing between two sources of coordination costsin technology commercialization: Coordinationcosts arise both from the need to coordinate acrossinventor-licensee teams and from the need to coor-dinate within an inventor team for further refine-ment and modification of its invention to facilitatecommercialization. Whereas prior work on coordi-nation in teams and in interfirm technology rela-tionships has generally referred to the importanceof prior joint work experience, we identify twospecific types of experience—licensing experienceand collaboration experience—as reducing the an-

2013 499Kotha, George, and Srikanth

ticipated coordination costs1 for inventions charac-terized by differing levels of mutual knowledge.

We tested our predictions using event historymodels that estimate the licensing hazard of 3,776unique inventions from a large US research univer-sity disclosed between 1981 and 2007. This studyis among the first large-sample, longitudinal, em-pirical tests of coordination among scientific teamsand its implications for commercialization. We adda complementary perspective to prior work byhighlighting the coordination-related aspects ofcombining distant knowledge domains in pro-ducing the type of inventions that are valuable(Fleming, 2001; Kapoor & Lim, 2007). Unlike priorwork on technology commercialization, which haslargely ignored nonpatented inventions (e.g.,Shane, 2002; Vural et al., 2013), this study ad-dresses the commercialization of both patented andnot-patented inventions. This comprehensive sam-ple raises the quality of theoretical and practicaldiscourse beyond inventions for which intellectualproperty rights have been secured through patents,and this extension has important implications forthe commercialization of scientific knowledge andthe practice of technology transfer.

THEORY AND HYPOTHESES

The theory development is structured as follows.First, we observe that the commercialization of in-ventions comprises two tasks: (1) effective knowl-edge transfer and coordination between an inventorteam and subsequent (downstream) parts of thesame or different organization and (2) effectiveknowledge transfer and coordination within theinventor team to further develop and modify theinvention as required for successful commercialapplication. We hypothesize that these coordina-tion costs and the value of an invention vary sys-tematically with science distance and affect thelikelihood of licensing. Next, we propose thatlearning from prior licensing experience helps at-tenuate coordination costs between inventor teamand licensee firm (henceforth, firm-team coordina-tion). Finally, we posit that learning from priorcollaboration experience helps mitigate coordina-

tion costs between members of the inventor team(henceforth, within-team coordination).

Commercialization of University Research

Scholars following an evolutionary economicstradition have argued that inventions are created byrecombining existing knowledge (Basalla, 1988;Fleming, 2001; Schumpeter, 1934). Though inven-tors can combine any two pieces of knowledge,what actually gets combined is constrained by thelocalness of search and the social construction ofwhat pieces can be gainfully combined (Basalla,1988; George et al., 2008; Kotha et al., 2011). Stud-ies have argued that breakthrough inventions aretypically the result of combining distant knowledgedomains and are rarely produced by recombiningproximate knowledge domains (Fleming, 2001,2004; Fleming & Sorenson, 2001; Vural et al., 2013).Our focus, however, is not the discovery of, butrather the commercialization of, such inventions.Inventions that recombine elements from the sametechnological domain are likely to be incremental,and they are therefore less valuable to a firm seek-ing commercial benefits than are inventions thatcombine relatively more distant domains that offerthe potential for breakthroughs. Therefore, from acommercial value perspective, inventions thatcombine proximate knowledge domains are lesslikely to be licensed than those that combine moredistant knowledge domains.

Increasing science distance, however, creates aconcomitant coordination problem. For a typicaluniversity invention, licensing is only the firststage in a long process of commercialization. Fromon a survey of 62 American universities, Jensen andThursby (2001) concluded that most universitytechnology is embryonic and requires significantfurther work for successful commercialization.They report that about 75 percent of inventionswere no more than a “proof of concept” at the timeof licensing; 48 percent did not have a prototype;and manufacturing feasibility was known for only 8percent of inventions. The authors concluded that“at the time of license, most university inventionsare at such an early stage of development that noone knows if they will eventually result in a com-mercially successful innovation or not. Moreover,they are so embryonic that further developmentwith the active involvement by the inventor isrequired for any chance of commercialization”(Jensen & Thursby, 2001: 240; emphasis added).Examples abound of the time and effort required to

1 We use “coordination costs” as a convenient short-hand to denote both the cost of generating adequate pre-dictive knowledge (Puranam et al., 2012) and the cost ofthe “wasted effort” that arises from misaligned actions(Postrel, 2009).

500 AprilAcademy of Management Journal

move inventions to commercial viability (Clarysse,Wright, Lockett, Van de Velde, & Vohora, 2005;George, Zahra, & Wood, 2002; Jain, George, & Mal-tarich, 2009). As a professor at St. Mary’s Hospital,part of Imperial College London, Alexander Flem-ing discovered that a mold inhibited the growth ofbacteria in 1928. Although the significance of thisdiscovery was immediately felt, attempts to treathuman subjects failed owing to insufficient quan-tities of the drug being produced. Commercial pro-duction of penicillin was achieved only in the early1940s, when its chemical structure was decoded.

The issue of further development effort requiredof university inventions continues to remain a chal-lenge today and gives rise to firm-team and within-team coordination costs. The firm-team coordina-tion cost arises from the need for adequateknowledge transfer and synchronization betweeninventor team and licensee firm to help the firmunderstand the value of the invention and workjointly toward commercialization. The within-teamcoordination cost arises from the need for themembers of the inventor team to work together tointegrate their knowledge to further develop theinvention to create working prototypes and estab-lish manufacturing scale. Below, we argue howjointly considering the likely value from a recom-bination and the two types of coordination costsinvolved in its commercialization help us under-stand the likelihood of licensing for university-gen-erated technologies.

Science Distance and CoordinationCosts in Commercialization

Several studies have argued that combining moredistant knowledge domains is more risky than com-bining more proximate knowledge domains, but afew of these distant combinations result in funda-mental breakthroughs of immense value (Ahuja &Lampert, 2001; Fleming, 2001; Fleming & Soren-son, 2001; Foster, 1986; Katila & Ahuja, 2002;Utterback, 1994). Inventions that recombine distantknowledge domains can be valuable because re-combining different disciplines can introduce plu-ralism in mental models and facilitate problemsolving (Amabile, 1988). Increasing variety in prob-lem solving also concomitantly increases the like-lihood that solutions can be found for challengingproblems (Fleming, 2007; George et al., 2008; Singh& Fleming, 2010). However, many of these recom-binations have little value because inventors havelimited knowledge of the nature of interdepen-

dence between elements from two distant domainsand ex ante have inaccurate expectations regardingtheir potential value. In contrast, inventors typi-cally have some idea of the nature of interactionsbetween two moderately distant domains and canmake informed choices regarding which experi-ments are more likely to succeed (Basalla, 1988).For example, as Fleming and Sorenson (2004)pointed out, inventors typically use science as a“map” to indicate regions of potentially valuablecombinations in their search attempts. Since thereis little interaction between distant sciences, scien-tific research is likely to provide a crude map. Incontrast, the search space for recombinations in-volving proximate domains, though well under-stood, tends to be saturated, since a lot of experi-mentation has already occurred (Ahuja & Katila,2001; Fleming, 2001; Rosenkopf & Almeida, 2003;Rosenkopf & Nerkar, 2001; Vural et al., 2013).Therefore, any recombination involving proximatedomains is likely to yield incremental improvements.

Fleming and coauthors examined how the valueof an invention covaries with the knowledge dis-tance between the sciences on which it is founded.Drawing on patent data analysis, Fleming (2004)argued that inventions that arise from proximateknowledge domains have limited upside potential.In contrast, when distant domains are combined,the average value of such inventions is lower, butthe variance increases, with a small number of in-ventions becoming extremely valuable. Fleming(2001) found that patents belonging to a single pat-ent class have significantly fewer forward citations.Fleming and Sorenson (2001) similarly showedthat interdependence between the knowledge basesthat underlie a patent positively impacts patentvalue, but the square of interdependence negativelyimpacts that value; an inverted U-shaped effect isshown. They measured interdependence as thedensity of citations between two knowledge do-mains; the higher the cross-citation, the closer theknowledge domains and the more familiar special-ists in one domain are with advances in the otherdomain. Kotha et al. (2011) also showed that theknowledge distance of a firm, calculated on thebasis of its patenting profile, has a positive impacton both innovative output and innovation quality,but the squared term for distance has a negativeimpact on both innovative output and innovationquality. Taken together, the evidence suggests thatif potential licensee firms were concerned onlyabout the value of inventions, researchers would onaverage see an inverse U-shaped relationship be-


tween an invention’s science distance and its haz-ard (likelihood) of licensing.

Next, we argue that this potential relationshipbetween science distance and licensing of an in-vention is reinforced when the coordination costsinvolved in commercializing academic inventionsare considered. Since most university inventionslead their industry by several years (Bayus & Agar-wal, 2007), a technology transfer organization andan inventor team need to educate a potential li-censee firm regarding the prospective value of in-ventions. Apart from being the target of an effort toconvey invention value, the licensee needs to bereassured that the inventor team will fully engagewith the commercialization process, is committedto generating further add-on inventions, and willwork with the firm’s employees to transfer ade-quate knowledge to make the required modifica-tions for scale-up and manufacturing (Dechenaux,Thursby, & Thursby, 2009; Jensen & Thursby,2001). Our premise is that if potential coordinationcosts are perceived to be high for an inventionrelative to its perceived value, firms are unlikely tolicense such inventions.

When an invention combines proximate do-mains, a firm is more likely to possess the requisiteabsorptive capacity to evaluate its commercial po-tential (Cohen & Levinthal, 1990). It is also morelikely to have scientists and engineers who possessthe requisite knowledge to work with inventors, tounderstand the nuances of the invention, and tofurther develop it for commercial application.From a within-team coordination cost perspective,inventors from the same or proximate disciplinesare likely to have shared knowledge and commontraining that will allow them to easily work to-gether. Though both types of coordination costs arelower when science distance is low relative towhen science distance is high, they may not betrivial. There are multiple demands on academicscientists’ time (Mowery, Nelson, Sampat, & Ziedo-nis, 2004), and given the considerable distance be-tween typical industry and academic knowledgefrontiers (Bayus & Agarwal, 2007), considerable ef-fort may be needed in firm-team coordination. Yetsuch proximate inventions may not have muchvalue for firms because they may lie adjacent toexisting technology portfolios or be perceived asmarginal improvements. Firms are thus more selec-tive in commercializing proximate inventions(Rosenkopf & Nerkar, 2001). Consequently, jointlyconsidering the potential value of an invention andthe likely coordination costs for its commercializa-

tion, it is likely that low science distance inven-tions have a lower likelihood of licensing than dis-tant inventions.

In an invention that combines distant scientificknowledge, coordination costs are likely to bemuch higher (Lockett, Siegel, Wright, & Ensley,2005; Schilling & Phelps, 2007; Siegel, Waldman,Atwater, & Link, 2003). First, from a firm-team co-ordination cost perspective, it becomes more diffi-cult to educate a potential licensee firm regardingthe benefits of the invention, because the averagelicensee firm may not possess the requisite absorp-tive capacity in all the different knowledge do-mains. This is especially difficult since, as notedabove, the knowledge underlying academic inven-tions on average leads the industry by several years.Whereas firms may recognize the potential ofbreakthrough inventions, they may not have theability to evaluate their commercial potential in themedium term. This is a nontrivial problem, sinceeven breakthrough technologies have valuable andless valuable patents. This difficulty is com-pounded by managerial coalitions and dominantparadigms of “what works” in their industry,which are likely to hinder the adoption of theseinventions (Kaplan, Murray, & Henderson, 2003).Second, when distant domains are combined, thelack of absorptive capacity makes it more difficultto transfer the underlying tacit knowledge fromthese domains to licensee scientists and engineers(Lane & Lubatkin, 1998). Given these high costs andthe large effort needed to develop an invention forcommercialization, a firm may not choose to li-cense it unless the value of the invention outweighsthe cost of this effort.

In addition, since the average licensee firm typi-cally does not possess adequate knowledge to fur-ther develop such an invention to make it commer-cially viable, this burden usually falls on theinventor team (Dechenaux et al., 2009). When dis-tant knowledge domains are combined, intrateamcoordination costs are also likely to be higher be-cause inventors from multiple domains are un-likely to have high levels of shared knowledge(Cronin & Weingart, 2007). First, scientists fromdistant domains are less likely to share a commonlanguage that allows them to effectively communi-cate their ideas between themselves. Second, dif-ferent disciplines vary in their dominant logics re-garding the nature of science and the mostprofitable avenues for further explorations to solveproblems. For example, Knez and Camerer (1994)discussed how experiments are conducted for dif-


fering reasons using varied methods by psycholo-gists and economists. Finally, disciplinary trainingand the scientific communities in different disci-plines also typically adopt different approaches re-garding acceptable trade-offs in research methods.Since much of this knowledge is tacit and taken forgranted, it becomes cumbersome for specialistsfrom multiple knowledge domains to coordinateamong themselves in further developing theirinventions.

Even though the members of an innovating teamhave coordinated among themselves to create a fo-cal invention, the low level of mutual knowledgeamong specialists remains a coordination challengein the commercialization process. Team membersneed to carry forward their ideas from proof-of-concept stage to creating a prototype, then on tomanufacturing scale (Jensen & Thursby, 2001).Typically, at this stage, interdependence betweenvarious elements that are combined becomestighter as solutions have to be optimized for scaleefficiencies. For example, there was a time lag ofabout a decade after the discovery of human em-bryonic stem cells for it to show commercial prom-ise. A significant roadblock was the fragility of thecell lines themselves and the difficulty in propagat-ing adequate quantities of the cells for commercialapplications. The inventor team generated furtherinventions in the manufacturing processes to culti-vate these cells to be more robust for commercialapplications (Jain & George, 2007). The inventorteams needed to invest significantly to learn andsolve these additional problems. Therefore within-team coordination is likely to remain a severe chal-lenge for diverse teams even though they have al-ready worked together to create the focal invention.For a firm, even though a distant recombination isvaluable, it may be reluctant to license because ofhigh anticipated within-team coordination costs. Insum, the lower average value of inventions thatcombine distant domains when they are comparedwith those that combine moderately distant do-mains, and the high anticipated coordination costs,together reduce the hazard of licensing for a high-science-distance invention.

A recombination of moderately distant knowl-edge domains is likely to be more valuable thanthose that recombine proximate or very distant sci-ences (Fleming & Sorenson, 2001; Kotha et al.,2011; Vural et al., 2013). These moderately distantinventions are also less likely to suffer from veryhigh coordination costs because there are overlapsin the knowledge base, and scientists need to invest

less effort to communicate the benefits of their in-ventions. Adequate absorptive capacity may alsohelp a firm’s scientists to understand the improve-ment necessary for commercialization and signifi-cantly reduce the knowledge transfer effort com-pared to that needed for distant inventions (Lane &Lubatkin, 1998; Phelps, 2010). Firms are morelikely to see moderately distant recombinant inven-tions as extensions of their product lines and capa-bilities (Katila & Ahuja, 2002). Dominant para-digms and managerial coalitions are therefore lesslikely to interfere with adoption of these new tech-nologies. These effects on average reduce the firm-team coordination costs for inventions combiningmoderately distant sciences as compared with in-ventions combining distant sciences.

Within-team coordination costs are also corre-spondingly lower in moderately distant inventions.When knowledge domains are only moderately dis-tant, different specialists in an innovating team aremore likely to have some degree of shared knowl-edge. For instance, they are more likely to read eachothers’ journals and are conversant with scientificnorms and paths of progress in the differentfields—that is, they share a common language thatfacilitates exchanging ideas and integrating knowl-edge. Therefore, potential licensee firms are likelyto conclude that within-team coordination costs aremanageable in moderately distant inventions andthat further development for commercialization isfeasible. In sum, the higher average value of mod-erately distant combinations and the lower antici-pated coordination costs relative to the inventionvalue together, on average, increase the hazard oflicensing for an invention that combines moder-ately distant knowledge domains over that of oneinvolving very proximate or very distant combina-tions. Drawing on these arguments, we posit:

Hypothesis 1. The relationship between inven-tor science distance and licensing likelihoodis curvilinear and is such that the slope ispositive from low to moderate science distanceand negative from moderate to high sciencedistance.

Prior Licensing Experience andFirm-Team Coordination Costs

Because new technology can be difficult to de-scribe and evaluate, it takes significant effort toconvince a potential licensee of the value of aninvention. Successful commercialization also re-


quires significant further development, often to cre-ate prototypes, experimenting to ensure stability,robustness, and refinement to ensure manufactur-ability to sufficient scale; these modifications likelywill draw on the commercialization expertise of thelicensee firm (Jensen & Thursby, 2001). From theperspective of the inventor team, we argue belowthat the firm-team coordination costs are lower fora team with prior experience in commercialization.

Inventor teams with prior licensing experienceare likely more aware of firm-team coordinationchallenges and the need to invest significant timeand resources in commercializing their inventions.Licensing experience likely provides inventorswith an understanding of the state of the art in anindustry and the primary concerns that drive in-dustrial R&D (Owen-Smith & Powell, 2003). Inother words, licensing experience provides higherlevels of shared knowledge between an inventorteam and a licensee firm, enabling them to moreclearly explain the technology and tailor theirknowledge transfer efforts to the needs of the li-censee firm (George, 2005). Prior licensing experi-ence is also likely to provide the inventor teamwith the knowledge and routines needed to betterinteract with their industry partners. An inventorteam that has limited commercialization experi-ence may not anticipate the effort involved in trans-ferring knowledge and coordinating actions with alicensee and may be unable to communicate thevalue of an invention persuasively to counter dom-inant logics. These teams are also unlikely to haveroutines in place to effectively transfer tacit knowl-edge to licensee firm scientists and engineers.These coordination challenges may lead the li-censee firm to take a skeptical view of the team’scapacity to assist with commercialization. Thissuggests that inventor teams with high licensingexperience are likely to have greater odds of licens-ing than teams with limited experience.

Hypothesis 2a. The greater the prior licensingexperience of an inventor team, the greater thelicensing likelihood for an invention, ceterisparibus.

Next, we argue that the prior licensing experienceof an inventor team moderates the relationship be-tween science distance and the hazard of licensingby differentially changing anticipated firm-teamcoordination costs for inventions at low, medium,and high science distance. At medium science dis-tance, firm-team coordination costs are neither lownor high, as they are in the corresponding low and

high science distance inventions. As discussed ear-lier, the greater a team’s prior licensing experience,the lower the likely firm-team coordination costs.

However, inventions that recombine moderatelydistant knowledge domains are on average morevaluable (Fleming, 2004; Fleming & Sorenson,2001). Furthermore, firms may perceive these mod-erate distance inventions as logical extensions totheir current work and therefore perceive these in-ventions to be more valuable. To realize this per-ceived value, licensee firms may be more willing toundertake the additional firm-team coordinationcosts when working with an inventor team withlow prior licensing experience. Therefore, taking afirm-team coordination cost perspective, we arguethat teams with high licensing experience are un-likely to have a significant advantage over teamswith average licensing experience; similarly, teamswith low licensing experience are unlikely to havea significant disadvantage when compared to teamswith average prior licensing experience. In otherwords, for medium science distance inventions,higher licensing experience is unlikely to be asso-ciated with significantly higher odds of licensing.

High science distance inventions are character-ized by a significant need for knowledge transferand coordination in multiple knowledge domainsbetween an inventor team and a licensee firm. Thisis likely to be difficult because the internal organ-ization architecture of firms reflects prevalent com-binations and links between fields rather thannovel and infrequent combinations (Henderson &Clark, 1990). Inventor teams with high levels ofprior licensing experience are likely to have lowerfirm-team coordination costs. These teams can relyon routines and processes developed from previousinteractions to more effectively convey theirknowledge to an industry’s participants. Theirprior experience interacting with the industry isalso likely to enable them to better understand andact upon the concerns of licensee firms about im-proving inventions for commercial use. Therefore,at high science distance, increase in prior licensingexperience from average to high is likely to beassociated with a significant increase in the hazardof licensing.

On the other hand, a team with low levels oflicensing experience is likely to lack the tools, rou-tines, and processes that enable an invention teamto relate to and address the concerns of a licenseefirm and therefore is likely to face significant firm-team coordination costs. The typical licensee firmunderstands the crucial role played by an inventor


team in transferring knowledge and developing aninvention, and it therefore, may look for a positivetrack record among the inventors for licensing(Mowery & Rosenberg, 1998). Therefore, the licens-ing likelihood for an invention at high science dis-tance is likely to significantly decrease as licensingexperience decreases from mean to low.

Firm-team coordination costs for low science dis-tance inventions, though lower than at mediumand high science distance, are still likely to bechallenging because academic inventions typicallylead industrial research by a significant margin.Since low science distance inventions are on aver-age incremental improvements that have lowervalue than moderate science distance inventions,licensee firms are likely to license only those in-ventions that they believe can be commercializedwith very little additional cost. Since teams withhigh prior licensing experience have low firm-teamcoordination costs, their low science distance in-ventions are likely to have a higher hazard of li-censing when compared with the inventions ofteams with average experience. Similarly, sinceteams with low prior licensing experience on aver-age have high firm-team coordination costs, theirlow science distance inventions are likely to have alower hazard of licensing than those of teams withaverage experience. Putting the above argumentstogether:

Hypothesis 2b. Prior licensing experience mod-erates the curvilinear relationship between in-ventor science distance and licensing likeli-hood: the inverted U-shaped pattern is amplifiedwhen experience is low and neutralized whenexperience is high.

Prior Collaboration Experience andWithin-Team Coordination Costs

Within-team coordination costs arise from theneed for members of an inventor team to jointlywork to further develop their inventions for suc-cessful commercialization. Commercialization re-quirements impose significant coordination needson inventors to configure subsystems for proto-types and achieving efficient scale (Jensen &Thursby, 2001). These concerns may differ fromtypical early stage invention concerns such as dem-onstrating proof of concept. For example, in drugsresearch and development, early concerns as to theefficacy of the active ingredient in a medicine aresupplanted later on by concerns about finding for-

mulations that help the body absorb the medicinewith few side effects (Henderson & Cockburn,1996). Therefore, even teams that have worked to-gether in developing a focal invention need to in-vest in significant learning to solve these commer-cialization problems; this gives rise to within-teamcoordination costs (Hoang & Rothaermel, 2005).

Coordination problems that arise when special-ists from different disciplines work together havebeen studied by scholars in diverse streams of re-search, including organization theory, social psy-chology, and sociology. Lawrence and Lorsch(1967) argued that organizations need to straddlethe differentiation-integration trade-off and pro-posed that mechanisms such as cross-functionaltraining and experience working in multifunctionalteams can help resolve the most difficult coordina-tion problems. In our context, studies of coordina-tion problems that arise when multiple specialistswork together in scientific labs have shown thatthese teams incur significant coordination prob-lems that are mitigated by specialists’ working to-gether and developing degrees of shared knowledge(Knorr-Cetina, 1999).

Social psychologists have also extensively stud-ied this issue, comparing the information gainsfrom diversity, especially functional diversity, andthe concomitant process losses that arise from thedifficulty diverse individuals have in working to-gether (Van Knippenberg & Schippers, 2007). In oursetting, within-team coordination requires that sci-entists relate to each other’s needs and require-ments and understand the trade-offs they have tomake and the consequences of such trade-offs.These inventors need to develop shared knowledgeregarding the challenges in further development,decide on future directions for development, andagree on a division of labor based on who is bestequipped to solve specific challenges. They mayneed to understand specific tools and methods pro-posed by members and dovetail their own effort tomatch those of their colleagues, all of which in-crease within-team coordination costs (Cronin &Weingart, 2007; Heath & Staudenmayer, 2000).

Prior collaboration experience alleviates within-team coordination costs by enabling participants togenerate shared mental models (Klimoski & Mo-hammed, 1994; Mohammed & Dumville, 2001).Prior experience increases both aspects of sharedmental models—that is, it provides both a sharedrepresentation of a task, called a task mental model,and shared understanding of the distribution ofexpertise in a team, called a team mental model


or transactive memory system (Liang et al., 1995;Mathieu, Heffner, Goodwin, Salas, & Cannon-Bowers, 2000). Recent work has suggested that co-ordination utility of a transactive memory systemis not limited to the task on which it was obtainedbut is transferrable to distinct but related tasks(Lewis, Belliveau, Herndon, & Keller, 2007; Lewis,Lange, & Gillis, 2005). Collaboration experiencealso allows a team to develop a shared languageand routines that facilitate coordinating futurework. Finally, teams with prior collaboration expe-rience also share tacit knowledge on tools andtechniques, and their uses and constraints, thusenabling inventors with prior collaboration experi-ence to substantially reduce intrateam coordina-tion costs.

Our theoretical concern is the reduction in coor-dination costs that accompanies an increase in mu-tual knowledge among interdependent individuals.Our specific argument is that when more membersof an inventor team have worked together before,the greater the development among them of sharedmental models, shared language, and shared rou-tines for achieving coordination. The density ofprior collaborative ties among an inventor teamcaptures our core theoretical construct. In contrast,most prior work on the effects of prior experiencehas measured intensity of prior collaboration. In-tensity of collaboration measures the number ofprior interactions, but this need not be an accurateindicator of increase in mutual knowledge in ateam. For example, in a five-member team, if twomembers have worked together ten times before,and none of the others have worked together, thesetwo members are the only ones who likely havecommon ground. In contrast, if all five team mem-bers have worked together before, but only twicefor each dyad, this team’s prior experience count isstill 10; however, a team with the latter experiencepattern is likely to have much greater commonground that enables them to reduce coordinationcosts more effectively. In other words, we expectthat the density of prior collaborations among ainventor team leads to increase in common groundand a corresponding decrease in coordinationcosts, even if we control for the intensity of priorexperience. Drawing on the arguments above, wehypothesize:

Hypothesis 3a. The greater the prior collabora-tion experience within an inventor team, thegreater the licensing likelihood for an inven-tion, ceteris paribus.

Next, we posit that collaboration experiencemoderates the relationship between science dis-tance and licensing by differentially changing an-ticipated within-team coordination costs for inven-tions at low, medium, and high science distance.Parallel to our arguments for licensing experience,first take the case of a moderate-science-distanceinvention for which prior collaboration experienceincreases from average to high. At moderate sciencedistance, within-team coordination costs are signif-icant, but manageable. Inventors are likely to sharesome common ground regarding underlying knowl-edge domains and may be able to build on thispreexisting knowledge to modify their inventionfor commercialization. This is especially likely tobe true if the inventors have some level of priorcollaboration experience. Teams with low priorcollaboration experience undoubtedly face higherwithin-team coordination costs than teams withhigh experience. However, inventions that recom-bine moderately distant knowledge domains are onaverage more valuable (Fleming & Sorenson, 2001)and, as argued above, are also likely to be perceivedto be more valuable. Given this high value estima-tion, firms may be willing to incur the additionalcoordination costs associated with commercializ-ing an invention from a team with low experience.Since the knowledge domains are marginally dis-tant, licensee firms may also believe that these ad-ditional costs may be low. Therefore, taking a co-ordination costs perspective, we argue that teams atmoderate science distance, with high collaborationexperience, are unlikely to have a large advantageover teams with average experience; and also,teams with low collaboration experience are un-likely to have a significant disadvantage.

At high science distance, within-team coordina-tion costs are likely to be extremely high (Makri etal., 2010; Schilling & Phelps, 2007; Siegel et al.,2003). Unlike the moderate-distance case, the high-distance case is likely to involve very little com-mon ground between scientists from very differentdisciplines (Cronin & Weingart, 2007; Dougherty,1992). It is likely to be quite difficult for inventorsto understand the nuances of sciences with verylittle overlap and build on them substantially tocreate new knowledge (Heath & Staudenmayer,2000; Huber & Lewis, 2010). As argued above,teams with high level of collaboration experienceare likely to have developed requisite commonground among all experts, reducing anticipatedwithin-team coordination costs. Therefore, at highscience distance, increasing collaboration experi-


ence from average to high is likely to have signifi-cant positive impact on the hazard of licensing.Conversely, decreasing collaboration experiencefrom average to low is likely to exacerbate within-team coordination costs. Therefore, at high sciencedistance, decreasing collaboration experience fromaverage to low is likely to be associated with alower hazard of licensing.

When an invention has low science distance,teams with high collaboration experience are, forsimilar coordination cost reasons, likely to be veryefficient in working together in further developingit. Because low science distance inventions, on av-erage, have lower value than moderate science dis-tance inventions, licensee firms are likely to licenseonly those inventions that they believe can be com-mercialized with little additional cost. Therefore,for low–science-distance inventions, teams withhigh collaboration experience are likely to have ahigher hazard of licensing than teams with averagelevels of experience.

Similarly, given an invention at low science dis-tance, even though inventors are fairly familiarwith each other’s knowledge domains, they stillneed to invest considerable time and effort in fur-ther developing the invention for successful com-mercialization (Jensen & Thursby, 2001). Inventorswith no prior experience are unlikely to have de-veloped shared routines and have limited knowl-edge of each other’s skills, expertise, and idiosyn-cratic preferences, likely resulting in increasedcosts of further collaboration. Given that inventionsat low science distance, on average, have lowervalue than inventions at moderate science distance(Fleming & Sorenson, 2001), and that academic in-ventors have competing demands on their time(Mowery et al., 2004), a licensee firm may believethat an inventor team will not invest the requisiteeffort in collaboration to further develop an inven-tion for commercialization. Therefore, reducingcollaboration experience from average to low islikely to reduce the licensing hazard for low sci-ence distance inventions. Given the above argu-ments, we predict that:

Hypothesis 3b. Prior collaboration experiencemoderates the curvilinear relationship be-tween inventor science distance and licensinglikelihood: The inverted U-shaped pattern isamplified when collaboration experience islow and neutralized when collaboration expe-rience is high.

METHODS

The research site for testing our hypotheses wasthe technology transfer office (TTO) of a large USuniversity. Established in 1925, this TTO receivesover $50 million in licensing and other annualincome. The members in our research team visitedthe TTO frequently and conducted extensive inter-views with managers from the four major func-tional areas: intellectual property, licensing, legal,and general administration. Almost all universityemployees disclose their inventions to the TTO.The intellectual property right managers (IPMs) li-aise with the inventors and educate them on theneed for, and the process of disclosure. This TTOprides itself in its outreach efforts to achieve com-prehensive disclosures, even of risky inventions.When a disclosure is made, the TTO records thenames of all the inventors, and the principal inves-tigator is typically assigned the role of the firstinventor. The IPMs then interview the inventorsand write a report that evaluates the commercialpotential of the invention. A decision about pursu-ing patent protection and commercialization istaken on the basis of this disclosure report at amonthly review meeting. The TTO records all dis-closures and decisions including whether to pat-ent, how widely to patent, and whether to seeklicensing without patent protection. By document-ing archived paper records and computerized data-bases, we constructed this data set and gatheredadditional data as described next.

Sample

The data for this study, collected in 2006 and2007, include disclosures made to this universityTTO between 1981 and 2006. We started at 1981 toreflect a dramatic change in institutional environ-ment with the passing of the Bayh-Dole Act. Thisact enabled university faculty to benefit from theintellectual property (IP) created by government-sponsored research and obliged universities to en-sure that such output be transferred as goods andservices back into the economy. Although we re-cord all disclosures from 1981 onward, we includeall relevant information regarding inventors andTTO, such as experience, since 1960. From 1981 to2006, inventors in teams of two or more disclosed3,776 unique inventions to the TTO. Of the 3,776disclosures, 755 (20%) were licensed; of these, 431inventions were licensed once, and the rest werelicensed multiple times. Of these 3,776 disclosures,


874 were patented, 339 (or 38%) of which werelicensed. Of the remaining 2,902 disclosures thatwere not patented, 416 (14%) were licensed.

Estimation Strategy and Dependent Variable

To estimate the impact of inventor characteristicson technology licensing, we used an event historymethodology. For all disclosures, we collected dis-closure date and licensing dates. About 80 percentof all disclosures were not licensed at the end ofthe study period. Since survival models control forright censoring, they were the most appropriatemodel to use in this study. They also control formultiple failure events, which was important forour estimation, since in our data 43 percent oflicensed disclosures were licensed multiple times.Our dependent variable is the licensing odds for agiven invention.

The estimation strategy was based on the stepsrecommended by Cleves, Gould, Gutierrez, andMarchenko (2010) and Kleinbaum and Klein(2012). First we focused on the underlying theoret-ical rationale—that is, the drivers of the distribu-tional form of licensing hazard. An invention dis-closed to the TTO is unlikely to be licensedinstantly; hence with time the hazard of licensingmay increase. In our data, the average licensing timeis about 7.8 years, and the modal licensing time is4 years. However, after a given time, inventionsthat have not been licensed have a lower hazard oflicensing. We checked to see if the underlying dis-tribution of licensing hazard was indeed initiallyincreasing and then decreasing, and we found thisto be the case. This distributional form meant wehad to use parametric estimations that fit non-monotonic hazard functions. We used a log-logisticmodel to estimate the survival odds that best fit thishazard function. We conducted tests to see if thelog-logistic form was appropriate and found this tobe the case. Survival odds are the inverse of licens-ing odds that we use to state our hypothesis and aresimilar in theory and interpretation to (the inverseof) the hazard of licensing (Kleinbaum & Klein,2012). That is, coefficients that are negative andsignificant imply that an increase in the indepen-dent variable leads to an increased hazard of licens-ing. We tested to see whether our results weresensitive to the specific parametric assumption andfound that they were robust to multiple distribu-tional specifications and also to nonparametricestimations.

Explanatory Variables

Science distance. Our theory suggests that themore distant the knowledge domains that are com-bined in an invention, the more difficult it is tolicense that invention. We captured the breadth ofknowledge domains that are combined by consid-ering how many different specialists contributed toan invention. We assigned its inventors to special-izations on the basis of the schools and depart-ments in which these scientists were employed.2

A simple construction of science distance is tocode an invention as “proximate” if its inventorsare employed in the same department/school, andas “distant” if they are employed in differentschools. This simplicity may overstate science dis-tance; for example, it masks the fact that scientistsin biology and medicine may share a larger overlapin mutual knowledge than scientists in physics andbiology. Therefore, we replaced the coarse categoryclassification with a fine-grained measure of dis-tance between sciences based on the prevalence ofcross-disciplinary research between any given pairof sciences.

Sociology of science has long focused on study-ing the relationship between sciences. One tooloften used for this purpose is citation analysis. Themore field A draws from field B, the lower theknowledge distance between fields A and B. Tomap the knowledge distances between sciences, wefollowed Leydesdorff and Rafols (2009). To under-stand cross-disciplinary citation patterns, they firstclassified 6,164 journals into 172 subject categoriesused by the Institute of Science Information (ISI).3

2 Recent work to capture the knowledge distance be-tween scientists has focused on the patent subcategoriesin which their discoveries are classified (Fleming & So-renson, 2001; Tzabbar, 2009), the intuition being that iftwo scientists file in the same category, their knowledgebases overlap. The distance between the categories ismeasured using the citation proportion between all pat-ents in one subcategory with the other patent subcate-gory. We do not use the patent-based measure but rely onthe intuition behind these measures. The reasons wedo not use the patent-based measure are (i) about 77percent of the inventions in our data are not patent pro-tected and (ii) recent evidence (Alcácer & Gittelman,2004) suggests that assignment of a patent to subclasses isdone by patent appraisers, and the inventors may not seetheir work as being in a particular category.

3 Interdisciplinary journals complicate the story, asthey are hard to assign to a unique science category.Scholars have followed one of two approaches: either


They did not conduct this analysis at the journallevel because the matrix was sparse; out of thenearly 38 million possible dyads (6,164 � 6,164),only 3 percent had values greater than zero. Theauthors used the 172 subject categories listed by ISIin 20064 as the starting unit of analysis. Journals areassigned to a subject category by ISI staff based ona list of factors that include the journal’s coverageand main audience. The average number of jour-nals per category is 56.3, and the range by categoryvaries from a low of 5 to a high of 262. Even withthe reduced set of 172 subject categories, the re-vised matrix of citation patterns between the 172 �172 categories results in nearly 30,000 cells, mostof which are zero. The authors collapsed the 172subject categories into 14 broad categories via clus-ter analysis and recommended that analysis ofthe citation pattern between the 172 subject cate-gories identified by ISI and the 14 broad categoriesyielded the best results for measuring knowledgedistance between sciences.

The values in the cells in this 172�14 matrixrange from 0 (which indicates no citations betweenjournals in category i to the journals in category j) toa high of .91 (which indicates that 90 percent of thecitations in category i are to journals in category j).We calculated science distance as one minus theproximity value calculated using the science prox-imity matrix described above, so that high valuesindicate high science distance representing limitedcross-citation between two fields.

To use this scientific distance matrix, we had twograduate students with engineering or science un-dergraduate degrees classify all inventors in ourdatabase into the 172 scientific specializations, us-

ing the department and school in which each in-ventor worked. In the few cases in which the codersdisagreed, we consulted experts to resolve the dif-ference. We measured science distance at the levelof inventor team. We aggregated the dyadic distanceof every pair of inventors and averaged it across alldyads to get a team-level measure of science dis-tance. Science distance is bounded by 0 and 1, with0 indicating that all inventors are in domains withhigh knowledge overlap and 1 indicating littleoverlap between inventors’ knowledge domains. Inour sample, the average science distance for a teamis 0.3, with a standard deviation of 0.4.

Prior licensing experience. Prior licensing expe-rience for a team was measured as the total numberof prior licenses held by its members from 1960until the time of licensing or censoring, as appro-priate, divided by team size. The variable has amean of 14.5 licenses and standard deviation of27.9. An alternative measure of prior licensing ex-perience, the cumulative count of the licensing expe-rience of all inventors in a team without standardiz-ing, yielded the same results in our estimations.

Prior collaboration experience. Inventors withlarge stocks of common ground are more able tocoordinate with each other in further improving aninvention. We used prior joint work experience as aproxy for preexisting stocks of common ground thatcan be leveraged in this coordination situation. Forgood intrateam coordination, it is best if all themembers of a team have adequate shared knowl-edge in the form of a common language, rules androutines, and common understanding of decisionpremises. For this purpose, we were interested inthe completeness of shared experience (Latour,1986) rather than in a simple count of prior sharedexperience, as we argued in the hypothesis devel-opment section. We measured prior collaborationexperience as the proportion of dyads in an inven-tor team with prior collaboration experience. Spe-cifically, we counted the dyads who had workedtogether prior to licensing date or censoring date (asappropriate) and divided this count by the totalpossible number of dyads in an inventor team. Thisvariable ranges from 0 to 1, where 0 implies noinventors have worked together before and 1 impliesthat all the inventors have worked together before.Its mean is 0.6 and its standard deviation is 0.4.

Control Variables

Following prior research, we constructed a hostof control variables derived from the patents that

ignore interdisciplinary journals altogether and concen-trate on the top few journals in each science or includethe interdisciplinary journals in only one sciencecategory.

4 We use a static measure of science distance instead ofa truly time-varying measure from 1981 through 2006. ISIhas published data classifying journals into subject cat-egories only since 1997; therefore it is empirically diffi-cult to obtain this information before 2006. However,Rafols, Porter, and Leydesdorff (2010) suggested that athigher levels of aggregation such as ours, science dis-tance measures are very robust. Also, since we used datafrom the end of our sample period in 2006, we are morelikely to understate than overstate science distance. Thisis more likely to downward bias our coefficients andwork against our finding any results in support of ourhypotheses.


underlay the inventions (Elfenbien, 2007; Shane,2002). We also used a number of non-patent-basedmeasures for all the inventions, including whethereach focal invention was patented. For inventionsthat were not patented, we set the control variablesderived from patents to be zero. We also used analternative approach to derive the mean value forthese variables based on observed values of allother variables used in the estimation. The resultsfor the theory variables are similar under both theseapproaches.

Quality of invention. To understand why somedisclosures are licensed, we needed to control forthe quality of each studied invention. Ziedonis(2007) suggested that the number of citations re-ceived by a patent is a good indicator of economicpotential. We counted the total citations a patenthad received, excluding self-citations, up to 2007.

Stage of development of invention. Most uni-versity disclosures are at an early stage in the tech-nology commercialization cycle and include nocommercial prototype or field data (Thursby &Thursby, 2002). Ziedonis (2007) used the numberof other patents that a focal patent cited as a mea-sure of whether the patent was radical or was builton existing technologies and further along the com-mercialization cycle. Similarly, we counted thenumber of backward citations of a focal patent as ameasure of its stage of development.

Inventor academic prestige. Elfenbein (2007) re-ported that inventors’ academic prestige was posi-tively linked to licensing outcomes. To control forthe quality of inventors, we collected data on theacademic publications of the members of each in-ventor team.

Government-funded intellectual property. Weused an indicator variable to determine whetherintellectual property (IP) associated with or consti-tuting a sampled invention was funded by a gov-ernment agency. Government-sponsored scientificprojects may likely push the frontiers of science,which may have a material effect on licensinghazard.

Privately funded intellectual property. We usedan indicator variable to determine whether the IPfor an invention was funded by a private organiza-tion. Funded projects are likely to be of greatercommercial interest to a sponsoring firm and,hence, are likely to get licensed sooner.

Intellectual property not protected. Given thatour sample includes nonpatented intellectual prop-erty such as biological materials, tools, and softwarethat can also be licensed, we used a binary variable

(1 � “not patented”) to control for the absence ofpatent protection at the time of licensing.

Collaboration intensity. In our theory develop-ment, we argued that completeness of prior jointwork experience among a team’s members is im-portant for achieving coordinated outcomes. Wealso controlled for prior collaboration intensityamong team members, measured as the average ofthe count of the number of prior collaborationsbetween each dyad in a team until the time oflicensing or censoring.

TTO capabilities. A TTO’s degree of experiencelicensing different technologies may impact thelikelihood of licensing for a given invention. Wecontrolled for TTO capabilities by counting thenumber of prior inventions a TTO had handled inthe technological domain of the focal invention.

Year effects and time since sample start date.We included a year indicator variable to control forunobserved year fixed effects. In addition, weadded a control for the number of days since Janu-ary 1, 1981, and a disclosure date. This variablecaptures any time trends that may lead to changesin the dependent variable. Furthermore, this vari-able accounts for the possibility that disclosures inrecent periods may be treated differently from thedisclosures in prior periods. Omitting the timesince start of sample variable did not change theresults for the theory variables.

RESULTS

We report descriptive statistics and the correla-tions between variables in Table 1. All correlationsabove .02 are significant at the .05 level. In Table 2,we report the results of survival models with robuststandard errors, clustered for multiple disclosuresby the same inventor team. The dependent variableis the survival odds of an invention—that is, theodds that an invention survives in the observationperiod without a licensing event. Consequently,positive values of the coefficient imply a negativerelationship between the variable and licensinghazard. In model 1, we report the baseline modelcontaining only the control variables. In model 2,we included the main effects of theory variables:science distance, science distance squared, priorlicensing experience, and prior collaboration expe-rience. In model 3, we add the moderations be-tween science distance and the experiencevariables.

Hypothesis 1 predicts a curvilinear relationshipbetween science distance and survival odds that is


such that moderate distance inventions shouldhave the lowest survival odds. In line with thisprediction, in model 2 the main effect for sciencedistance is negative and significant (� � �1.54, p �.05) and the squared term is positive and significant(� � 1.86, p � .01). To test if moderate distanceinventions have the lowest survival odds, we plot-ted the estimated licensing odds across the ob-served range of science distance in our sample. SeeFigure 1 for this plot. Note that the model predictssurvival odds; we calculated the licensing odds asone minus predicted survival odds. As hypothe-sized, we find that moderate distance inventionshave the highest licensing odds. The highest valuelicensing odds reaches is 0.30 when science dis-tance is 0.42. Whereas when science distance is atthe mean minus one standard deviation, the licens-ing odds are 0.23; and when science distance is atthe mean plus one standard deviation, the licensingodds are 0.26. Hypothesis 1 is supported. Hypoth-esis 2a predicts that licensing odds will be higheras licensing experience increases. In model 2, thereis a negative and significant relationship betweenlicensing experience and survival odds (� ��0.009, p � .01), supporting our prediction in Hy-pothesis 2a. Hypothesis 3a predicts a positive rela-tionship between prior collaboration experienceand licensing odds for an invention. The coefficientfor prior collaboration experience (� � �0.0149)is not significant in model 2. Hypothesis 3a is notsupported.

We tested the moderation effects using the re-sults from model 3 in Table 1. Hypothesis 2b pre-dicts positive moderation by licensing experience

of the relationship between science distance andlicensing odds that is such that the curvilinear re-lationship is neutralized as prior licensing experi-ence increases. The interaction term between sci-ence distance and licensing experience is negativeand significant (� � �0.081, p � .01), and theinteraction term between science distance squaredand licensing experience is positive and significant(� � 0.095, p � .01). To facilitate interpretation,we plotted the licensing odds at different levels ofscience distance and different levels of licensingexperience, as shown in Figure 2. We plotted thesefigures at the median time it takes to license aninvention in our sample (approximately eight years).5

When prior experience is low (i.e., one standarddeviation below mean prior licensing experience),the hazard of licensing an invention is 0.24, 0.40,and 0.23 at low, moderate, and high science dis-tance, respectively. When prior experience is at themean, the hazard of licensing an invention is 0.27,0.50, and 0.28 at low, moderate, and high sciencedistance, respectively. When prior experience ishigh (i.e., one standard deviation above the mean),the hazard of licensing and invention is 0.29, 0.69,and 0.29 at low, moderate, and high science dis-tance, respectively (Figure 2). This suggests thatcurvilinearity is more pronounced as experienceincreases, contrary to Hypothesis 2b. Though not

5 We checked for robustness using different years: 3, 5,10, and 12 years, as well as at the mean (7.8 years) andmode (4 years). Our results are similar to those presentedin the text.

TABLE 1Descriptive Statistics and Correlations of the Variables in the Studya

Variables Mean s.d. 1 2 3 4 5 6 7 8 9 10 11 12 13

1. Time (years) 8.0 5.42. Government funded 0.2 0.4 �.073. Privately funded 0.6 0.5 �.08 �.514. Quality of invention 4.7 12.5 .04 .05 �.065. Stage of development 1.9 5.3 �.01 .03 .01 .486. Inventor academic prestige 6.7 1.7 �.15 �.05 .03 �.04 .007. Intellectual property not protected 0.7 0.5 �.06 .00 .00 �.58 �.56 �.048. Time from start 8.6 5.9 �.62 .09 .13 �.34 �.12 .03 .339. TTO capabilities 6.3 1.6 �.41 .02 .15 �.14 �.01 .33 .06 .45

10. Collaboration intensity 1.2 1.5 .16 �.01 �.15 .25 .09 .32 �.25 �.55 �.0211. Science distance 0.3 0.4 .18 .04 �.12 .09 �.02 �.12 .01 �.18 �.24 �.0712. Science distance squared 0.2 0.4 .17 .05 �.10 .08 �.02 �.17 .03 �.12 �.25 �.13 .9613. Licensing experience 14.5 27.9 .05 �.02 �.08 .06 .05 .55 �.15 �.36 .11 .61 �.06 �.1114. Prior collaboration experience 0.6 0.4 .05 �.06 �.02 .15 .11 .28 �.18 �.22 .09 .51 �.12 �.13 .34

a n � 3,776 disclosures.


presented here, a plot of the entire data rangeshows that at extreme values of prior licensing ex-perience, in line with our hypothesis, the relation-ship between science distance and hazard of licens-ing is essentially flat except at the ends, very lowand very high science distance. However, withinthe two standard deviation range, Hypothesis 2b

is not supported. Hypothesis 3b predicts that thecurvilinear relationship between science distanceand licensing hazard is more pronounced at lowvalues of prior collaboration and is neutralized athigh values of prior collaboration. We tested themoderation effects using the results from model 3of Table 1. The interaction term between science

TABLE 2Predicting the Survival Odds of an Inventiona

Dependent Variable: Survival Odds Model 1 Model 2 Model 3

Government funded �0.214 �0.252 �0.227(0.169) (0.169) (0.167)

Privately funded �0.393** �0.408** �0.421**(0.144) (0.143) (0.142)

Quality of invention �0.0333** �0.0354** �0.0364**(0.005) (0.005) (0.005)

Stage of development 0.0213* 0.0217* 0.0221*(0.010) (0.010) (0.010)

Inventor academic prestige �0.0544 0.0420 0.0408(0.038) (0.044) (0.044)

Intellectual property not protected 0.505** 0.471** 0.499**(0.144) (0.143) (0.141)

Time from start �0.000302** �0.000318** �0.000336**(0.000) (0.000) (0.000)

TTO capabilities �0.261** �0.214** �0.216**(0.049) (0.047) (0.047)

Collaboration intensity �0.269** �0.196** �0.181**(0.043) (0.058) (0.057)

Theory VariablesScience distance �1.542** �1.404*

(0.551) (0.555)Science distance squared 1.858** 1.788**

(0.559) (0.560)Licensing experience �0.00882** �0.00907**

(0.003) (0.003)Prior collaboration experience �0.0149 0.0501

(0.170) (0.168)Licensing � science distance �0.0809**

(0.017)Licensing � science distance squared 0.0950**

(0.017)Collaboration � science distance 4.329**

(1.317)Collaboration � science distance squared �4.497**

(1.336)Constant 8.866** 5.960** 6.063**

(1.001) (0.854) (0.854)Year fixed effects Yes Yes Yes

Chi-square 507.8 546.8 545.6Log-likelihood �4,374.3 �4,336.0 �4,297.1Observations 5,010 5,010 5,010

a Survival means an invention was not licensed in a given observation period. Hence negative and significant coefficients mean that thehazard of licensing increases with an increase in the explanatory variable.

* p � .05** p � .01

Robust clustered standard errors are in parentheses.


distance and collaboration experience is positiveand significant (� � 4.329, p � .01), and the inter-action between science distance squared and li-censing experience is negative and significant (� ��4.497, p � .01). To interpret the moderation hy-potheses, we plotted Figure 3, which shows therelationship between science distance and licens-ing hazard at various levels of prior collaborationexperience. When prior collaboration experience isat the mean, the hazard of licensing an invention is0.29, 0.33, and 0.24 at low, moderate, and highscience distance, respectively. When prior collabo-ration experience is low (mean – 1 s.d.), the valueof the hazard of licensing and invention is 0.25,0.34, and 0.20 at low, moderate, and mean science

distance, respectively. This suggests that the curvi-linear relationship of science distance and licens-ing hazard is more curvilinear (inverted U-shaped)at low than at moderate distance, as per Hypothesis3b. When prior collaboration is high (mean � 1s.d.), the hazard of licensing an invention is 0.32,0.30, and 0.28 at low, moderate, and high sciencedistance, indicating that the curvilinear relation-ship is now flatter, supporting Hypothesis 3b.

Robustness Checks

We conducted several robustness checks to testthe sensitivity of our results to alternative specifi-cations, different measures, and the influence of

FIGURE 1Hazard of Licensing at Different Levels of Science Distance

0.2 0.4 0.6 0.8 1.0Science Distance

0.18

0.20

0.22

0.24

0.26

0.28

0.30

Licensing Odds

FIGURE 2Science Distance and Hazard of Licensing Moderated by Licensing Experience

0.0 0.2 0.4 0.6 0.8 1.0Science Distance

0.2

0.3

0.4

0.5

0.6

0.7

Licensing Odds

Low Experience

Mean Experience

High Experience


outliers. First, our sample combines disclosuresthat are patented and those that are not patented.We checked to see if our theory holds across thesesubsamples. We report the estimations by patented(model 3) and not patented (model 2) subsamplesin Table 3. The results for the theory variables arebroadly similar in these two subsamples. The maindifference is that the moderation terms for priorlicensing experience are not significant in the pat-ented subsample. One explanation for this result isthat the patenting process potentially codifies alarge part of the tacit knowledge that underlies aninvention, which may reduce firm-team coordina-tion costs.

We ran several other estimations that are notreported in this article. We tested results’ sensitiv-ity to our assumption of the parametric form in thesurvival models. Our results were qualitatively un-changed for several different parametric assump-tions as well as in nonparametric estimations. Wetested the robustness of our results to multiple li-censing events of the same disclosure. We ran sur-vival models with only a single failure event (li-censing or censoring) per disclosure and comparedit to survival models with multiple failure events(multiple licenses for the same disclosure), clus-tered by disclosure and inventor team. Our resultswere qualitatively identical for these estimations.Some licensee firms may have licensed more thanone invention from a TTO and had routines inplace that reduced the coordination cost in licens-ing. Since we could not explicitly use a controlvariable capturing the history of licensing between

a licensee firm and TTO, as this record is not de-fined for inventions not licensed, we reran ourestimations clustering for the repeated licensing bysome firms. Similar results were obtained.

In addition to controls for demand conditions,we included variables from the Yale Survey onindustrial research and development, on the rela-tionship between sciences and industry, that mayinfluence certain types of science to be commer-cialized faster (Klevorick, Levin, Nelson, & Winter,1995). These variables are not significant in thehazard of licensing and hence are not included inour estimations. Furthermore, we ran estimationsdropping outliers, and our results remain un-changed. We also tried to check if unobserved fac-tors regarding team formation impacted licensing.We followed a two-stage method to predict thelikelihood that a given team would form and thenused the hazard ratio from this prediction as anadditional control in predicting licensing. Our re-sults remain robust.

DISCUSSION

We examined the coordination problems thatstem from lack of mutual knowledge when innova-tive teams attempt joint action. By contrasting thecoordination of experts from distant versus proxi-mate scientific domains, we were able to highlightthe different coordination challenges in commer-cializing inventions, specifically, the challenges oftaking university research to industry application.Our findings suggest that mutual knowledge among

FIGURE 3Science Distance and Hazard of Licensing Moderated by Collaboration Experience

0.2 0.4 0.6 0.8 1.0Science Distance

0.20

0.25

0.30

0.35

Licensing OddsLow Collaboration

Mean Collaboration

High Collaboration


inventors and routines that transfers knowledge tolicensee firms is important to overcome coordina-tion challenges. We find that prior collaborativeand prior licensing experience respectively helpmitigate these two types of coordination costs. Theimplications for theories of coordination in organ-izational design and the production and commer-cialization of discoveries are discussed next.

Theories of OrganizationalDesign and Coordination

Recent approaches to organization design draw-ing on the knowledge-based view have emphasizedthe role of achieving coordination in joint produc-tion tasks and the role of organizations as vehiclesfor generating adequate mutual knowledge or com-mon ground that enables interdependent agents to

TABLE 3Robustness Subsample Estimation of the Survival Odds of Inventionsa

Dependent Variable: Survival Odds Full Sample Not Patented Patented

Government funded �0.227 �0.204 �0.259(0.167) (0.291) (0.243)

Privately funded �0.421** �0.845** �0.0625(0.142) (0.231) (0.202)

Quality of invention �0.0364** �0.0298**(0.005) (0.005)

Stage of development 0.0221* 0.0179†

(0.010) (0.010)Inventor academic prestige 0.0408 0.0328 0.0489

(0.044) (0.067) (0.063)Intellectual property not protected 0.499**

(0.141)Time from start �0.000336** �0.000277* �0.000452**

(0.000) (0.000) (0.000)TTO capabilities �0.216** �0.216** �0.254**

(0.047) (0.066) (0.087)Collaboration intensity �0.181** �0.230* �0.140†

(0.057) (0.093) (0.082)Theory VariablesScience distance �1.404* �1.460

†�2.218**

(0.555) (0.900) (0.773)Science distance squared 1.788** 2.390** 1.835*

(0.560) (0.926) (0.786)Licensing experience �0.00907** �0.0148** �0.0064†

(0.003) (0.004) (0.0038)Prior collaboration experience 0.0501 0.0838 0.241

(0.168) (0.257) (0.214)Licensing � science distance �0.0809** �0.118** �0.0307

(0.017) (0.024) (0.023)Licensing � science distance squared 0.0950** 0.140** 0.0304

(0.017) (0.025) (0.028)Collaboration � science distance 4.329** 5.879** 3.667†

(1.317) (2.005) (2.000)Collaboration � science distance squared �4.497** �5.433** �4.317*

(1.336) (2.046) (1.995)Constant 6.063** 3.814** 6.587**

(0.854) (0.761) (0.772)Year fixed effects Yes Yes YesChi-square 545.6 476.4 237.9Log-likelihood �4,297.1 �2,734.2 �1,430.2Observations 5,010 3,510 1,500

a Robust clustered standard errors are in parentheses.† p � .10* p � .05

** p � .01


coordinate their activities (Grant, 1996; Kogut &Zander, 1992, 1996; Puranam et al., 2012). Theseapproaches articulate a view of organizations ascoordination systems, as opposed to the traditionalview of organizations as incentive systems (Holm-strom & Milgrom, 1994; Williamson, 1991). We ex-plicitly adopted a coordination lens to study thejoint production problem, specifically by focusingon the variation in the level of mutual knowledgein an inventor team, holding other factors constant.We found support for our arguments that antici-pated coordination costs help explain licensingpatterns of university technology.

The main effect of prior collaboration experienceon licensing was not significant; instead, prior col-laborative experience became pertinent only as amoderator of science distance. We found that low-experience teams perform much more poorly athigh science distance that is due to lack of mutualknowledge and its attendant coordination costs.Joint work experience is likely to reduce coordina-tion costs under all conditions; however, our find-ing suggests that in the invention commercializa-tion process, such efficiencies gain importanceonly under some conditions. Specifically, within-team coordination costs become important onlywhen the cost of coordination is expected to belarge relative to the expected value of an invention.Further, theoretically, we differentiate firm-teamcoordination costs from within-team coordinationcosts and argue that these costs are respectivelymitigated by different kinds of experience—that is,licensing experience and collaboration experience,respectively. Though we do not empirically distin-guish between these two kinds of coordinationcosts, our moderation results strongly suggest thatthese different costs do drive the invention com-mercialization process.

It is intuitive that prior joint working experienceis helpful in collaboration efforts by specialistsfrom multiple domains. Specifically, we contributeto the literature on the impact of prior experienceon coordination in specialist teams in two ways.First, prior research on joint work has typicallyused dichotomous measures for knowledge dis-tance, conceptualizing distance as within-domainor across-domains. By measuring knowledge dis-tance as a continuous variable, we contribute tounderstanding how prior experience likely impactscoordination costs depending on variation in thelevel of mutual knowledge available in a team. It isinteresting to note that though levels of mutualknowledge likely vary monotonically with knowl-

edge distance, prior experience that bridges thismutual knowledge gap has a nonmonotonic impacton the likelihood of commercialization. Second,studies have typically measured prior collaborationexperience by intensity of joint work, using metricssuch as the amount of time partners teams haveworked together (e.g., Latour, 1986; Vural et al.,2013). However, intensity in addition to mutualknowledge creation can capture other effects, suchas socialization. This confound prevents pinpoint-ing whether observed effects are indeed a functionof generating mutual knowledge. For instance, inour data, though collaboration intensity has a pos-itive main effect on licensing odds, the interactionswith science distance are not significant. This find-ing would be in line with our theory if collabora-tion intensity were actually capturing socializationeffects rather than mutual knowledge effects. Incontrast, we argue that the density or the closure ofprior experience (i.e., the proportion of team mem-bers that have worked together before) is also animportant consideration in reducing coordinationcosts. This is a more accurate measure of the exis-tence of mutual knowledge, especially in large in-novative teams, which are becoming more com-mon. We show that the density of prior experiencehas an impact on coordination costs (and the like-lihood of forming technology commercializationrelationships) even after the intensity of prior ex-perience is controlled for.

Our findings reinforce the crucial role played bymutual knowledge and coordination costs in organ-izational design and joint action. Future extensionsof this study could take a more dynamic approachto the distances between science fields and con-sider how incentives and coordination costs jointlyand independently influence outcomes of team ac-tion. Nevertheless, this is among the first large-scale longitudinal settings for a test of coordinationwhen mutual knowledge is low in joint team pro-duction of inventions.

The Commercialization of Scientific Discoveries

Our study proffers evidence for the coordinationcosts perspective in technology commercialization.In so doing, this study speaks to the heart of com-mercialization of scientific discoveries and to thecore challenges in integrating knowledge from dis-parate technical niches. The trend toward teamproduction of scientific discoveries (Paruchuri,2010; Wuchty et al., 2007) and the shift towardinterdisciplinary science that recombines distant


rather than insular domains push forward the fron-tiers of human knowledge (Basalla, 1988; George etal., 2008). In that context, our empirical finding oflicensing likelihood at moderate science distancepresents an interesting pattern. Figures 2 and 3show that at moderate science distance, teams witha high level of collaboration experience do not per-form better than teams with moderate experience.Though average experience seems to be beneficialwhen compared to low experience, the magnitudeof the impact of experience seems to be lower thanat the low or high ends of science distance. It islikely that moderate-distance inventions occupy a“sweet spot” in the commercialization process—they tend to be more valuable than inventions re-sulting from adjacent or very distant combinations,and potential licensee firms are likely to recognizetheir value and partner with these inventions’ aca-demic inventors for commercialization. Therefore,potential licensees may be more willing to licensesuch inventions and partner with teams that havesome (rather than low) experience. Future researchshould explore these processes further. However, itis interesting to note that prior licensing experienceis valued highly by potential licensee firms even atmoderate science distance, and TTOs have highersuccess licensing inventions by teams with highprior licensing experience.

An explanation complementary to our coordina-tion costs logic is that inventor teams with highlicensing experience are working in more appliedareas that are more readily commercialized, andteams with prior collaboration experience may beworking on more incremental improvements totheir prior work that could also have commercialapplications. We attempted to control for these ef-fects. First, we controlled for the academic publi-cation record of inventors; presumably, more aca-demic publications point to more basic research.Second, in the patented subsample, we controlledfor the relative weight of academic citations toother (patent) citations. Arguably, the more aca-demic work cited by a patent, the more basic it is;and the more applied work cited by the patent, themore incremental the invention. Our results for ourcoordination cost variables were unchanged inthese robustness tests. Future work should morecarefully delineate the impact of these differentmechanisms.

The findings also add to the substantial reposi-tory of empirical work on the value of patents. Thecitations that a patent receives have been used as aproxy for the quality of the patent. In our study, the

results suggest that the likelihood of licensing in-creases by 2 percent with every additional patentcitation. The innovation literature has predomi-nantly relied on patents as an indicator to explainboth the process of creating knowledge and theperformance of firms, which has drawn some crit-icism as scholars have questioned the value of pat-ents, given the explosion of patenting activity byfirms (Jaffe & Lerner, 2004). Furthermore, othershave questioned the link between empirical mea-sures and the reality of the patenting process,wherein some critical variables such as citations toprior art are frequently added by patent examiners;this suggests that either inventors may have notused the prior art cited or may have been unawareof it (e.g., Alcácer & Gittelman, 2004). There is apaucity of empirical work that examines commer-cialization beyond patenting.

In this study, we include all inventions disclosedat a university that could potentially be licensed.Of the 3,776 disclosures, only 23 percent were pat-ented. Of the 2,902 disclosures that were not pat-ented, 416 were licensed. The number of licensesfrom nonpatented inventions is 1.23 times greaterthan the number of licenses from patented inven-tions. This suggests that it is important to gobeyond patenting to understand discovery andcommercialization processes. For example, Shane(2002) found that nearly 60 percent of inventions atMIT were patented, out of which 52 percent werelicensed. In that study and most others in this lit-erature, inventions that could be licensed butwere not patented are ignored. By including allinventions whether patented or not, we provide thefirst large-sample investigation of licensing of non-patented inventions.

Our research also has implications for more tra-ditional licensing of technology between firmsand, more broadly, for interorganizational technol-ogy transfer or development. In the past decade,an increasing number of studies on technology-sourcing relationships, including buyer-supplierrelationships (Becker & Zirpoli, 2011; Srikanth &Puranam, 2011), technology alliances, joint ven-tures (Gulati & Singh, 2009; Hoetker & Mellewigt,2009), and acquisition integration–acquisition per-formance (Ahuja & Katila, 2001; Makri et al., 2010;Puranam & Srikanth, 2007) have begun to empha-size the role of knowledge integration in the suc-cess of interorganizational relationships. We ex-tend this coordination costs logic to the formationof technology-licensing relationships characterized


by need for close joint work between the licenseeand licensor.

Our setting is particularly interesting because weobserve the formation of technology licensing rela-tionships where the traditional make-or-buy con-siderations do not apply. The university does notcommercialize its own inventions. Therefore, anyinventions that are not licensed essentially havezero net present value, which provides an oppor-tunity to observe the role of coordination costs inthe formation of a licensing relationship. We drawfrom theories of experience, especially theories oflearning by doing at the individual and group lev-els of analysis, to inform our understanding of theinfluence of coordination problems on the forma-tion of interorganizational licensing relationships.Considering the role of coordination costs in tech-nology commercialization relationships is an inter-esting extension to prior research on role of coor-dination costs in make-or-buy decisions (Hoetker,2005; Monteverde, 1995).

The literature on open innovation suggests thatfirms should take advantage of innovations pro-duced by specialists outside their boundaries(Chesborough, Vanhaverbeke, & West, 2006). A sig-nificant and growing portion of these licenses arefrom universities and scientific research laborato-ries (Mowery & Rosenberg, 1998), where the respec-tive partners have complementary capabilities:basic research and commercialization of innova-tions. Our study suggests that the licensing of in-terdisciplinary inventions could involve signifi-cant coordination costs that may not occur inlicensing specialized inventions. Since managingthese coordination costs is likely to hold the key tocommercial success, firms should keep in mindsuch costs when making decisions regarding theexternal procurement versus internal developmentof innovations.

Incentives and Coordination

In this study, we focused on the anticipated co-ordination cost as an explanation of why someinventions are licensed and others are not. Presum-ably incentives offered to license intra- and inter-disciplinary inventions could be heterogeneous.Whereas a wealth of theoretical work focuses onthe optimal contract structure between licenseeand licensor, there is little empirical work on thistopic. Future work could consider the variations inincentive contracts with differences in types of in-ventions and how these contracts change with the

experience of licensee and licensor capabilities.Such research efforts would throw light on the roleof incentive structure on successful commercializa-tion of a licensed invention as a product or service.

Future research could also be enriched by exam-ining the role of a central coordinating agent in ateam. It is plausible that, in lieu of dyadic tiesbetween inventors working across distant scientificdomains, one central coordinating agent, who hasworked with all the inventors on a team and caninterpret results of the work of all members, may bejust as effective. Because this coordinating teammember will expend more effort in orchestratingthe actions of others, whether the central coordina-tor should be rewarded with a higher share of theincome is an important question. More fundamen-tally, does the bearer of the coordination cost alsocommensurately benefit from its success? The lim-ited empirical evidence on incentive sharing isseen in the literature on entrepreneurs distributingequity to their helpers. A substantial majority offounders of nascent ventures tend to share incen-tives equally with their helpers (Ruef, 2009),whereas some entrepreneurs with specific humancapital are able to deviate away from the equaldistribution of incentives (Kotha & George, 2012).Consequently, the effects of specific and generalhuman capital of the inventors on incentive distri-bution and its implications for team performancebecome interesting and economically meaningfulquestions for future research.

Opportunity Recognition in Entrepreneurship

The central debate in the literature on entrepre-neurship concerns the nature of opportunities (Al-varez & Barney, 2010; Dimov, 2011). Some scholarshave taken the view that opportunities for bringingnew products and services to market are objectiveand that entrepreneurs identify these opportunitiesdrawing on their prior experience (Shane, 2012).Other scholars have viewed entrepreneurs as con-structing opportunities where it is not possible tospecify an opportunity ex ante unless it has beenenacted (Venkataraman, Sarasvathy, Dew, & For-ster, 2012). Our study considers inventions as po-tential opportunities that might be commercializedand that vary in value on the basis of science dis-tance and anticipated coordination costs. We findsupport for the likelihood that moderately distantinventions have a higher probability of being suc-cessfully licensed. These results are consistentwith the treatment of opportunities as objective and


the view that inventor experience improves theodds of success, which is consistent with an oppor-tunity-individual nexus view of entrepreneurship(Shane, 2012). By starting with all inventions thatcould potentially be licensed, we avoided the fal-lacy of defining opportunities on the basis of theirsuccessful outcome. On the issue of whether oppor-tunities objective or subjective, we proffer evidenceto support the view that the value of opportunities,on average, may be objectively assessed by the sci-entific distance between the inventors in a team.Individual factors further improve the likelihood oflicensing in distant domains, in keeping with thenotion of opportunities being objective, as is con-sistent with the literature’s treatment of individualexperience effects in entrepreneurship.

TTO Role and Managerial Implications

We argued that the variation in the level of mu-tual knowledge of inventors in a team influencesthe perceived value of a technology and its antici-pated coordination cost, thereby influencing whichinventions gets licensed. The central actors in oursetting are inventors, TTOs, and licensing firms.TTOs are tasked with protecting and licensing in-ventions. It is TTOs that communicate with poten-tial licensee firms, negotiate and monitor contracts,and protect the inventions assigned to them againstinfringements. Given the importance of TTOs, it isnot surprising that they are able to license moreinventions in domains in which they have highlevel of capabilities—that is, have handled a highernumber of inventions. We found that TTO exper-tise is positively and significantly related to thehazard of licensing an invention. It would be inter-esting to consider the joint variation in TTO capa-bilities and mutual knowledge of inventors as fur-ther steps in future studies.

This study also holds implications for managingtechnology transfer organizations, especially in auniversity setting. From initial interviews, wefound that TTO intellectual property managers andlicensing managers realize how helpful and avail-able specific inventors have been in prior deals.TTO managers use this information to assuageconcerns from potential licensee firms regardingknowledge transfer and industry engagementissues—the effects of prior licensing experience un-derscore the importance of this strategy. The addedimplication is that TTO managers should considerthe mutual knowledge in teams and how comfort-able inventors are as a team, especially if they are

from different scientific domains. The findings re-veal that prior collaboration experience makes asubstantive difference to inventions combininghighly proximate and distant domains—highlight-ing that the “chemistry” between inventors andtheir ability to work together in commercializationis especially important.

Limitations

This study is not without its limitations. First, werelied on a single institution for data. There may bea risk of idiosyncratic institutional practices thatinfluence our results. However, a single institutionnaturally controls for between-institution differ-ences that can confound results, especially withregard to coordination costs that arise from differ-ences in administrative bureaucracies. Further-more, we follow the tradition of prior studies thathave argued that with the advent of technologytransfer associations and journals, best practicesare widely spread (Clarysse et al., 2005; Nerkar &Shane, 2007; Shane, 2000; Siegel et al., 2003). An-other issue that stems from our setting is that therelationship between academic inventors and li-censee firms is mediated by universities’ TTOs,organizations dedicated to pursuing commercial-ization opportunities stemming from academic set-tings. Findings from this context may not be di-rectly applicable to commercialization of R&D fromcorporate labs. However, the rise of contract re-search organizations and the increasing establish-ment of units dedicated to licensing technologiesarising from corporate labs that a firm chooses notto commercialize itself may be appropriate avenuesfor generalizing the insights from this study.

Second, we used a time-invariant measure of sci-ence distance, measuring it at the end of our sampleperiod in 2006. Since sciences are growing closer,we are more likely to be understating, rather thanoverstating, science distance. If there is any bias, itis more likely to suggest that sciences are closerthan they actually were at the time of the invention.Since our moderation hypotheses specify differenteffects at high science distance, this bias shouldwork against our finding any results. Finally, sci-entists may elect to pursue projects taking into ac-count their likelihood of commercialization, andplausibly, our study may suffer from an endogene-ity problem. Theoretically, whereas endogeneitycan confound the interpretation of main effects, itis hard to think of a compelling story that explainsthe moderator effects. It is possible that only inven-


tors who believe their coordination costs are lowengage in inventions combining distant sciences. Ifthis were the case, then we should not find evi-dence for the moderation by prior collaborationexperience. The fact that we still find significantresults suggests that ours is a more conservative testfor the impact of coordination costs on licensing.

Conclusion

This study adds to the literatures on coordinationof experts from distant domains to commercializenew inventions. We capture disclosure and licens-ing data over a 26-year-long window that startswith the uniform change in legislation (Bayh DoleAct) for all US universities. Our sample includesboth patented and nonpatented inventions, and ourestimation is robust for single and multiple licens-ing events. We find support for our predictions thatprior licensing and collaboration experience in-creases the hazard of licensing an invention. Thepattern of our results suggests a nuanced mecha-nism of how prior collaboration within a team in-fluences intrateam coordination costs and the haz-ard of licensing an invention. Specifically, we findthat prior collaboration experience increases thehazard of licensing more for inventions that com-bine very proximate or very distant sciences than itdoes for inventions that combine moderately dis-tant sciences. In doing so, this study expandsempirical investigation to the practice of commer-cialization of innovation and to theories of coordi-nation and organizational design.

REFERENCES

Ahuja, G., & Katila, R. 2001. Technological acquisitionsand the innovation performance of acquiring firms:A longitudinal study. Strategic Management Jour-nal, 22: 197–220.

Ahuja, G., & Lampert, C. M. 2001. Entrepreneurship inthe large corporation: A longitudinal study of howestablished firms create breakthrough inventions.Strategic Management Journal, 22: 521–543.

Alcácer, M., & Gittelman, M. 2004. Patent citations as ameasure of knowledge flows: The influence of exam-iner citations. Review of Economics and Statistics,88: 774–779.

Alvarez, S., & Barney, J. 2010. Entrepreneurship andepistemology: The philosophical underpinnings ofthe study of entrepreneurial opportunities. In A. P.Brief & J. P. Walsh (Eds.), Academy of Managementannals, vol. 4: 557–583. Essex, UK: Routledge.

Amabile, T. M. 1988. A model of creativity and innova-tion in organizations. In B. M. Staw & L. L. Cum-mings (Eds.), Research in organizational behavior,vol. 10. Greenwich, CT: JAI.

Basalla, G. 1988. The evolution of technology. Cam-bridge, UK: Cambridge University Press.

Bayus, B. L., & Agarwal, R. 2007. The role of pre-entryexperience, entry timing, and product technologystrategies in explaining firm survival. ManagementScience, 53: 1887–1902.

Becker, M. C., & Zirpoli, F. 2011. What happens whenyou outsource too much? MIT Sloan ManagementReview, 52(2): 59–67.

Camerer, C. 2003. Behavioral game theory: Experi-ments in strategic interaction. Princeton, NJ:Princeton University Press.

Chesborough, H., Vanhaverbeke, W., & West, J. 2006.Open innovation: Researching a new paradigm.Oxford, UK: Oxford University Press.

Clarysse, B., Wright, M., Lockett, A., Van de Velde, E., &Vohora, A. 2005. Spinning out new ventures: A ty-pology of incubation strategies from European re-search institutions. Journal of Business Venturing,20: 183–216.

Cleves, M., Gould, W., Gutierrez, R., & Marchenko, Y.2010. An introduction to survival analysis usingStata. College Station, TX: Stata.

Cohen, W. M., & Levinthal, D. A. 1990. Absorptive-ca-pacity: A new perspective on learning and innovation.Administrative Science Quarterly, 35: 128–152.

Cronin, M. A., & Weingart, L. R. 2007. Representationgaps, information processing and conflict in func-tionally diverse terms. Academy of ManagementReview, 32: 761–773.

Cummings, J. N., & Kiesler, S. 2005. Collaborative re-search across disciplinary and organizational bound-aries. Social Studies of Science, 35: 703–722.

Dechenaux, E., Thursby, M. C., & Thursby, J. G. 2009.Shirking, sharing risk, and shelving: The role of uni-versity license contracts. International Journal ofIndustrial Organization, 27: 80–91.

Dimov, D. 2011. Grappling with the unbearable elusive-ness of entrepreneurial opportunity. Entrepreneur-ship: Theory and Practice, 35: 57–81.

Dougherty, D. 1992. Interpretive barriers to successfulproduct innovation in large firms. Organization Sci-ence, 3: 179–203.

Elfenbein, D. W. 2007. Publications, patents and the mar-ket for university inventions. Journal of EconomicBehavior and Organization, 63: 688–715.


Faems, D., Janssens, M., Madhok, A., & Van Looy, B.2008. Toward an integrative perspective on alliancegovernance: Connecting contract design, contract ap-plication, and trust dynamics. Academy of Manage-ment Journal, 51: 1053–1078.

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

Fleming, L. 2004. Perfecting cross-pollination. HarvardBusiness Review, 82(9): 22–24.

Fleming, L. 2007. Breakthroughs and the “long tail” ofinnovation. MIT Sloan Management Review, 49(1):69–74.

Fleming, L., & Sorenson, O. 2001. Technology as a com-plex adaptive system: Evidence from patent data.Research Policy, 30: 1019–1039.

Fleming, L., & Sorenson, O. 2004. Science as a map intechnological search. Strategic Management Jour-nal, 25: 909–928.

Foster, R. 1986. Innovation: The attackers advantage.New York: Summit.

George, G. 2005. Learning to be capable: Patenting and li-censing at the Wisconsin Alumni Research Founda-tion 1925–2002. Industrial and Corporate Change,14: 119–151.

George, G., Kotha, R., & Zheng, Y. 2008. Entry into insu-lar domains: A longitudinal study of knowledgestructuration and innovation in biotechnology firms.Journal of Management Studies, 45: 1448–1474.

George, G., Zahra, S., & Wood, D. 2002. The effects ofbusiness-university alliances on innovative outputand financial performance: A study of publiclytraded biotechnology companies. Journal of Busi-ness Venturing, 17: 577–609.

Grant, R. M. 1996. Toward a knowledge-based theory ofthe firm. Strategic Management Journal, 17: 109–122.

Guimerà, R., Uzzi, B., Spiro, J., & Nunes Amaral, L. A.2005. Team assembly mechanisms determine collab-oration network structure and team performance.Science, 308: 697–702.

Gulati, R., Lavie, D., & Singh, H. 2009. The nature ofpartnering experience and the gains from alliances.Strategic Management Journal, 30: 1213–1233.

Gulati, R., Lawrence, P., & Puranam, P. 2005. Adaptationin vertical relationships: Beyond incentive conflict.Strategic Management Journal, 26: 415–440.

Heath, C., & Staudenmayer, N. 2000. Coordination ne-glect: How lay theories of organizing complicate co-ordination in organizations. In R. I. Sutton & B. M.

Staw (Eds.), Research in organizational behavior,vol. 22: 153–191. Greenwich, CT: JAI.

Henderson, R., & Cockburn, I. 1996. Scale, scope andspillovers: The determinants of research productiv-ity in drug discovery. RAND Journal of Economics,27: 32–59.

Hoang, H., & Rothaermel, F. 2005. The effect of generaland partner-specific alliance experience on jointR&D project performance. Academy of Manage-ment Journal, 48: 332–345.

Hoetker, G. 2005. How much you know vs. how well Iknow you: Selecting a supplier for a technically in-novative component. Strategic Management Jour-nal, 26: 75–96.

Hoetker, G., & Mellewigt, T. 2009. Choice and perfor-mance of governance mechanisms: Matching alli-ance governance to asset type. Strategic Manage-ment Journal, 30: 1025–1044.

Holmstrom, B. & Milgrom, P. 1994. The firm as an incen-tive system. American Economic Review, 84: 972–991.

Huber, G. P., & Lewis, K. 2010. Cross-understanding:Implications for group cognition and performance.Academy of Management Review, 35: 6–26.

Jaffe, A., & Lerner, J. 2004. Innovation and its discon-tents: How our broken patent system is endanger-ing innovation and its progress, and what to doabout it. Princeton, NJ: Princeton University Press.

Jain, S., & George, G. 2007. Technology transfer offices asinstitutional entrepreneurs: The case of WisconsinAlumni Research Foundation and human embryonicstem cells. Industrial and Corporate Change, 16:535–567.

Jain, S., George, G., & Maltarich, M. 2009. Academics orentrepreneurs? Investigating role identity modifica-tion of university scientists involved in commercial-ization activity. Research Policy, 38: 922–935.

Jensen, R., & Thursby, M. 2001. Proofs and prototypes forsale: The licensing of university inventions. Ameri-can Economic Review, 91: 240–259.

Jones, B. F. 2009. The burden of knowledge and the“death of the renaissance man”: Is innovation gettingharder? Review of Economic Studies, 76: 283–317.

Kaplan, S., Murray, F., & Henderson, R. 2003. Disconti-nuities and senior management: Assessing the role ofrecognition in pharmaceutical firm response to bio-technology. Industrial and Corporate Change, 12:185–233.

Kapoor, R., & Lim, K. 2007. The impact of acquisitions onthe productivity of inventors at semiconductor


firms: A synthesis of knowledge-based and incen-tive-based perspectives. Academy of ManagementJournal, 50: 1133–1155.

Katila, R., & Ahuja, G. 2002. Something old, somethingnew: A longitudinal study of search behavior andnew product introduction. Academy of Manage-ment Journal, 45: 1183–1194.

Kleinbaum, D., & Klein, M. 2012. Survival analysis aself learning text. New York: Springer Science andMedia.

Klevorick, A. K., Levin, R. C., Nelson, R. R., & Winter,S. G. 1995. On the sources and significance of inter-industry differences in technological opportunities.Research Policy, 24: 185–205.

Klimoski, R., & Mohammed, S. 1994. Team mental mod-el: Construct or metaphor? Journal of Management,20: 403–437.

Knez, M., & Camerer, C. 1994. Creating expectationalassets in the laboratory: Coordination in “weakest-link” games. Strategic Management Journal, 15:101–119.

Knorr-Cetina, K. 1999. Epistemic cultures: How the sci-ences make knowledge. Cambridge, MA: HarvardUniversity Press.

Kogut, B., & Zander, U. 1992. Knowledge of the firm,combinative capabilities, and the replication of tech-nology. Organization Science, 3: 383–397.

Kogut, B., & Zander, U. 1996. What firms do? Coordina-tion, identity and learning. Organization Science, 7:502–518.

Kotha, R., & George, G. 2012. Friends, family, or fools:Entrepreneur experience and its implications forequity distribution and resource mobilization. Jour-nal of Business Venturing, 27: 525–543.

Kotha, R., Zheng, Y., & George, G. 2011. Entry into newniches: The effects of firm age and the expansion oftechnological capabilities on innovative output andimpact. Strategic Management Journal, 32: 1011–1024.

Lane, P. J., & Lubatkin, M. H. 1998. Relative absorptivecapacity and interorganizational learning. StrategicManagement Journal, 19: 461–477.

Latour, B. 1986. Laboratory life: The construction ofscientific facts. Princeton, NJ: Princeton UniversityPress.

Lawrence, P. R., & Lorsch, J. W. 1967. Organization andenvironment: Managing differentiation and inte-gration. Boston: Harvard University Press.

Lewis, K., Belliveau, M., Herndon, B., & Keller, J. 2007.Group cognition, membership change, and perfor-

mance: Investigating the benefits and detriments ofcollective knowledge. Organization Behavior andHuman Decision Processes, 103: 159–178.

Lewis, K., Lange, D., & Gillis, L. 2005. Transactive mem-ory systems, learning, and learning transfer. Organi-zation Science, 16: 581.

Leydesdorff, L., & Rafols, I. 2009. A global map of sciencebased on the ISI subject categories. Journal of theAmerican Society for Information Science andTechnology, 60: 348–362.

Liang, D. W., Moreland, R., & Argote, L. 1995. Groupversus individual training and group performance:The mediating role of transactive memory. Person-ality and Social Psychology Bulletin, 21: 384–393.

Lockett, A., Siegel, D., Wright, M., & Ensley, M. 2005. Thecreation of university spin-off firms at public re-search institutions: Managerial and policy implica-tions. Research Policy, 34: 981–993.

Makri, M., Hitt, M. A, Lane, P. J. 2010. Complementarytechnologies, knowledge relatedness, and inventionoutcomes in high technology mergers and acquisi-tions. Strategic Management Journal, 31: 602–628.

March, J. G. & Simon, H. A. 1958. Organizations. NewYork: Wiley.

Mathieu, J. E., Heffner, T. S., Goodwin, G. F., Salas, E., &Cannon-Bowers, J. A. 2000. The influence of sharedmental models on team process and performance.Journal of Applied Psychology, 85: 273–283.

Mohammed, S., & Dumville, B. C. 2001. Team mentalmodels in a team knowledge framework: Expandingtheory and measurement across disciplinary bound-aries. Journal of Organization Behavior, 22: 89–106.

Monteverde, K. 1995. Technical dialog as an incentivefor vertical integration in the semi-conductor indus-try. Management Science, 41: 1624–1638.

Mowery, D. C., Nelson, R. R., Sampat, B. N., & Ziedonis,A. A. 2004. Ivory tower and industrial innovation:University-industry technology transfer beforeand after the Bayh-Dole Act. Palo Alto, CA: Stan-ford University Press.

Mowery, D., & Rosenberg, N. 1998. Paths to innovation.Cambridge, UK: Cambridge University Press.

Nerkar, A., & Shane, S. 2007. Determinants of inventioncommercialization: An empirical examination of ac-ademically sourced inventions. Strategic Manage-ment Journal 28: 1155–1166.

Owen-Smith, J., & Powell, W. 2003. The expanding roleof university patenting in the life-sciences: Assess-


ing the importance of experience and connectivity.Research Policy, 32: 1695–1711.

Paruchuri, S. 2010. Intraorganizational networks, interor-ganizational networks, and the impact of central in-ventors: A longitudinal study of pharmaceuticalfirms. Organization Science, 21: 63–80.

Phelps, C. C. 2010. A longitudinal study of the influenceof alliance network structure and composition onfirm exploratory innovation. Academy of Manage-ment Journal, 53: 890–913.

Postrel, S. 2009. Multitasking teams with variable com-plementarity: Challenges for capability management.Academy of Management Review, 34: 273–296.

Puranam, P., Raveendran, M., & Knudsen, T. 2012. Orga-nization design: The epistemic interdependence per-spective. Academy of Management Review, 37:419–440.

Puranam, P., & Srikanth, K. 2007. What they know vs.what they do: How acquirers leverage technologyacquisitions. Strategic Management Journal, 28:805–825.

Rafols, I., Porter, A. L., & Leydesdorff, L. 2010. Scienceoverlay maps: A new tool for research policy andlibrary management. Journal of the American Soci-ety for Information Science and Technology, 61:1871–1887.

Rosenkopf, L., & Almeida, P. 2003. Overcoming localsearch through alliances and mobility. ManagementScience, 49: 751–766.

Rosenkopf, L., & Nerkar, A. 2001. Beyond local search:Boundary spanning, exploration and impact in theoptical-disk industry. Strategic Management Jour-nal, 22: 287–306.

Schilling, M., & Phelps, C. 2007. Interfirm collaborationnetworks: The impact of large-scale network struc-ture on firm innovation performance. ManagementScience, 53: 1113–1126.

Schumpeter, J. A. 1934. The theory of economic devel-opment. New York: Springer.

Shane, S. 2000. Prior knowledge and the discovery ofentrepreneurial opportunities. Organization Sci-ence, 11: 448–469.

Shane, S. 2002. Selling university technology: Patternsfrom MIT. Management Science, 48: 122–137.

Shane, S. 2012. Delivering on the promise of entrepre-neurship as a field of research. Academy of Man-agement Review, 37: 10–20.

Siegel, D., Waldman, D., Atwater, L., & Link, A. 2003.Commercial knowledge transfers from universitiesto firms: Improving the effectiveness of university-industry collaboration. Journal of High TechnologyManagement Research, 14: 111–133.

Singh, J., & Fleming, L. 2010. Lone inventors as sourcesof breakthroughs: Myth or reality? Management Sci-ence, 56: 41–56.

Srikanth, K., & Puranam, P. 2011. Integrating distributedwork: Comparing task design, communication andtacit coordination mechanisms. Strategic Manage-ment Journal, 32: 849–875.

Thursby, J. G., & Thursby, M. C. 2002. Who is selling theivory tower? Sources of growth in university licens-ing. Management Science, 48: 90–104.

Tzabbar, D. 2009. When does scientist recruitment affecttechnology reposition? Academy of ManagementJournal, 52: 873–896.

Utterback, J. M. 1994. Mastering the dynamics of inno-vation. Boston: Harvard Business School Press.

Van Knippenberg, D., & Schippers, M. C. 2007. Workgroup diversity. In M. I. Posner & M. K. Rothbart(Eds.), Annual review of psychology, vol. 58: 515–541. Palo Alto, CA: Annual Reviews.

Venkataraman, S., Sarasvathy, S. D., Dew, N., & Forster,W. R. 2012. Reflections on the 2010 AMR decadeaward: Whither the promise? Moving forward withentrepreneurship as a science of the artificial. Acad-emy of Management Review, 37: 21–33.

Vural, O., Dahlander, L., & George, G. 2013. Collaborativebenefits and coordination costs: Learning and capa-bility development in science. Strategic Entrepre-neurship Journal, 7: In press.

Weitzman, M. L. 1998. Recombinant growth. QuarterlyJournal of Economics, 113: 331–360.

Williamson, O. 1991. Comparative economic organiza-tion: The analysis of discrete structural alternatives.Administrative Science Quarterly, 36: 269–296.

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

Zahra, S. A., & George, G. 2002. Absorptive capacity: Areview, reconceptualization and extension. Acad-emy of Management Review, 27: 185–203.

Ziedonis, A. A. 2007. Real options in technology licens-ing. Management Science, 53: 1618–1633.


Reddi Kotha ([email protected]) is an assistantprofessor in the Strategy and Organization Group atthe Lee Kong Chian School of Business, Singapore Man-agement University. Reddi received his Ph.D. in entre-preneurship from London Business School. He studiesorganizational structures and processes used by entrepre-neurs and firms to bring new products and services tomarket.

Gerry George ([email protected]) is the deputyprincipal for faculty and programs and professor of in-novation and entrepreneurship at Imperial College Lon-don. Gerry also serves as the director of the Rajiv Gandhi

Center at Imperial College London. He received his Ph.D.from Virginia Commonwealth University. His recentwork examines strategic and organizational factors, suchas business models and organizational design, and theirimpact on innovation and performance in technologyventures and emerging markets.

Kannan Srikanth ([email protected]) is an assis-tant professor of strategy in the Indian School of Busi-ness. He received his Ph.D. from the London BusinessSchool. His research focuses on learning and coordina-tion aspects of organizational adaptation and innovation.


bridging the mutual knowledge gap: coordination and … · 2017. 5. 26. · bridging the mutual...

Documents