Tc

SK

ARRAA

KARTBS

1

veWc1uwnJtapcsg

W

0h

Research Policy 42 (2013) 1313– 1324

Contents lists available at SciVerse ScienceDirect

Research Policy

jou rn al hom epage: www.elsev ier .com/ locate / respol

he spill-over theory reversed: The impact of regional economies on theommercialization of university science

teven Casper ∗

eck Graduate Institute of Applied Life Sciences, USA

a r t i c l e i n f o

rticle history:eceived 24 January 2012eceived in revised form 25 April 2013ccepted 29 April 2013vailable online 28 June 2013

eywords:cademic commercializationegional economiesechnology transferiotechnologyocial network analysis

a b s t r a c t

The concept of regional technology spill-overs created by university research is one of the most enduringtheories within the economic geography and innovation management fields. This article introduces analternative perspective on academic commercialization, arguing that the quality of a university’s regionalenvironment can significantly impact a university’s success in commercializing science. Recent researchon university technology transfer stresses the importance of personal contacts between academic andindustry scientists in driving commercialization. The social structure of the regional economy in which auniversity is embedded will strongly influence the density of contacts linking university scientists withindividuals in industry, and through doing so, impact the density of networks through which universityknowledge can be commercialized. Social network analysis is used to examine the quality of social tieslinking industry and university scientists within the San Francisco and Los Angeles California biotechnol-ogy industries over the 1980–2005 period. Results support the theory that the existence of strong social

networks linking inventors heightens university commercialization output. Despite similar universityresearch endowments, universities in San Francisco have dramatically commercialization outputs thanSan Francisco, which is correlated with the existence of cohesive inventor networks linking industryand university scientists in this region, but not Los Angeles. Moreover, longitudinal analysis shows thatthe commercialization output of San Francisco universities increased substantially starting in the early1990s, the time period in which cohesive inventor networks emerged in the region.

. Introduction

The concept of regional technology spill-overs created by uni-ersity research is one of the most enduring theories within theconomic geography and science and technology study fields.hile scholars had long viewed basic research as a public good that

ould spill-over into society (Arrow, 1962; Rosenberg and Nelson,994, 1996), the recognition that knowledge being developed inniversities is often tacit or “sticky” (Von Hippel, 1994) lead to aave of research viewing universities as anchors of regional eco-omic development (see e.g., Audretsch and Feldman, 1996, 2003;

affe et al., 1993). Firms have an incentive to locate near universi-ies, as proximity to universities reduces the cost of accessing andbsorbing knowledge spill-overs (Audretsch and Lehmann, 2005,. 1115). The process by which university knowledge leads to the

reation of regional spin-off companies has also been intenselytudied, both in terms of the importance of such firms to economicrowth (see e.g. Florida and Choen, 1999) but also the creation of

∗ Correspondence address: Keck Graduate Institute of Applied Life Sciences, 535atson Drive, Claremont, CA 91711, USA. Tel.: +1 909 607 0132.

E-mail addresses: scasper@kgi.edu, steven casper@kgi.edu

048-7333/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.respol.2013.04.005

© 2013 Elsevier B.V. All rights reserved.

regional technology clusters (Braunerhjelm and Feldman, 2006;Audretsch and Lehmann, 2005). Universities have been heraldedas “engines of growth” (Florax, 1992). Economic development hasbecome a “third mission” of universities (Etzkowitz, 2002), com-plementing education and basic research, and is also frequentlyused to justify public investments in university research (see Hage,2011; National Academies, 2007).

The management of university knowledge spill-overs hasbecome a central issue within the field of technology transfer stud-ies. The enactment in the United States of the 1980 Bayh–DoleAct legitimated the idea that university research should be treatedas intellectual property that can be commercialized, and in mostcases transferred ownership of federally funded research to theuniversity in which it was conducted (see Mowery et al., 2004).Though controversial (see Kenney and Patton, 2009), the Bayh–Doleframework has become emulated in the UK, Japan, Germany, andmany other nations around the world (Mowery and Sampat, 2005).Much research on academic commercialization has focused on vari-ation in the ability of universities to successfully commercialize

technology. Scholars have identified a range of factors internal touniversities linked to effective technology transfer. These includethe quality of a university’s basic research endowments (Powersand McDougall, 2005), the university’s prestige (Sine et al., 2003),

1 olicy 4

at

crcrtabceeaTuttw

genrw(RonheuwSrnobs

swuswdtslhdtscltii

itt2ltn

314 S. Casper / Research P

nd organizational practices and funding surrounding technologyransfer (Siegel et al., 2003; O’Shea et al., 2005).

This article introduces an alternative perspective on academicommercialization, arguing that the quality of a university’segional environment can significantly impact a university’s suc-ess in commercializing science. While the idea that universityesearch can usefully flow into regional economies is not contested,he mechanisms structuring this flow are. While most research oncademic commercialization emphasizes “push” variables – seeny the predominance of internal factors in explaining success inommercialization – this article argues that “pull” factors can bequally important. Recent research within the technology literaturemphasizes the importance of personal contacts between academicnd industry scientists in driving commercialization (Thursby andhursby, 2004). The quality of the regional economy in which aniversity is embedded will strongly influence the density of con-acts linking university scientists with individuals in industry, andhrough doing so, structure the existence of networks throughhich university knowledge can be commercialized.

There is a long tradition of economic geographers and sociolo-ists comparing the social and industrial organization of regionalconomies (see e.g. Storper, 1997; Piore and Sabel, 1984). Eco-omic sociologists argue that the fabric of social ties withinegions strongly impacts the pattern by which individuals workingithin companies, universities, and other organizations interact

Saxenian, 1994; Herrigel, 1993). Regional factors can influence the&D strategies of companies, and through doing so structure therganization of inter-organizational ties linking scientists and engi-eers within regional economies. Walter Powell and collaboratorsave made particularly important contributions to this approach,mphasizing the importance of social networks at both the individ-al and organizational level in impacting the performance of firmsithin biotechnology and other science based industries (Owen-

mith and Powell, 2004; Powell et al., 1996, 2002, 2005a,b). Powell’sesearch was the first to employ social network methods to mapetworks across both firms and individuals within the biotechnol-gy field, generating direct evidence that the structure of social tiesoth varies regionally and impacts the performance of firms withincience-based industry.

This article contributes to Powell’s general strategy of mappingocial networks linking scientists, engineers, and entrepreneursithin regional biotechnology clusters, applying the approach toniversity commercialization. The article argues that universitycientists are likely to develop pervasive networks of contactsith industry scientists when they work at a university embed-ed within a regional economy in which local industry adoptsechnology strategies that encourage collaboration with univer-ity scientists. Universities embedded within regions that eitherack a significant industry presence or one in which companiesave developed primarily inward looking R&D strategies will notevelop extensive networks of ties linking academics with indus-ry, and as a result will have lower commercialization output. In thisense, the perspective offered here reverses the causal mechanismommonly associated with spill-over theory: technology is moreikely to flow from universities to a regional environment whenhe ‘plumbing’ – or network of contacts between university andndustry scientists – is in place to pull knowledge from universitiesnto the regional economy.

The article uses social network analysis to examine the qual-ty of social ties linking industry and university scientists withinhe San Francisco and Los Angeles California biotechnology indus-ries over the 1980–2005 period. Drawing on data from close to

0,000 biotechnology patents, it maps the emergence of networks

inking inventors in San Francisco and Los Angeles. This data ishen used to explore whether the existence of cohesive inventoretworks that connect university and industry scientists within a

2 (2013) 1313– 1324

region is correlated with increased commercialization output fromlocal universities.

To preview the article’s findings, the San Francisco and LosAngeles cases support the theory that the existence of strongsocial networks linking inventors heightens university commer-cialization output. San Francisco has developed a large, cohesivenetwork linking thousands of biotechnology inventors, into whichhundreds of university scientists have become connected. Thisnetwork developed incrementally between 1980 and 2005 andonly became highly interconnected in the post-1990 period. Theacademic commercialization output of universities located withinthe San Francisco region increased substantially in the post-1990period, after a cohesive social network linking inventors emerged.Moreover, two universities whose scientists developed particularlystrong linkages within regional industry inventor networks, Stan-ford and the University of California San Francisco, saw the largestincreases in commercialization output. In the Los Angeles region,in contrast, the biotechnology industry has not developed a large orcohesive inventor network, with the consequence that fewer tieslinking university and industry scientists have developed. Univer-sities in the region, despite having comparable resources to thosein the San Francisco region, have a much lower commercializationoutput.

2. The quality of regional environments and academiccommercialization

This article argues that organization of inter-organizationalsocial ties among scientists within a university’s regional econ-omy can influence commercialization patterns. This argument hastwo elements. First, drawing on recent research within the technol-ogy transfer studies field (Thursby and Thursby, 2004; Kenney andPatton, 2009; Lam, 2007), it is argued that personnel ties betweenuniversity and company personnel are an important driver of aca-demic commercialization. Second, focusing on contributions fromeconomic sociology (Saxenian, 1994; Piore and Sabel, 1984; Powellet al., 1996, 2005a,b), it is claimed that the social organization ofregional economies shape the development of inter-organizationalnetworks linking scientists and engineers. Regions that developdense social networks across inventors will foster more numer-ous personnel ties linking local companies and university scientistscompared to regions that do not, increasing the potential for com-mercialization processes to develop within the region.

Recent research within the technology transfer literature hasemphasized the importance of personal contacts linking industryand academic scientists in driving forward university commer-cialization. Thursby and Thursby (2004), in discussing the originsof university technology licensing agreements, emphasize the“extreme importance of personal contacts between the firm’s R&Dstaff and university personnel.” Their survey of over a hundredcompanies that had initiated technology licensing arrangementswith universities found that 45.7% of companies cited personal con-tacts as “extremely important” and a further 31.4% of respondentscited contacts as “important” (Thursby and Thursby, 2004, p. 169).Kenney and Patton (2009) also emphasize the key role of universityscientists in driving technology commercialization. Compared tolicensing officers, faculty have superior knowledge of the technol-ogy and broader research field, and often have pre-existing contactsto companies that can be crucial in easing concerns on the part ofthe company that a university technology, once licensed, can be

successfully transferred (Kenney and Patton, 2009, p. 1411). Lam(2007), in a study of several university–industry relationships inthe United Kingdom, has emphasized the role of “linked” universityscientists, defined both in terms of translational research interests

olicy 4

ar

ncdwleMatpptta

sblspM1iectamiwcplSmtsfioc1

pir1tco(an(iolluotp

e

S. Casper / Research P

nd contacts with industry, in driving successful industry collabo-ations.

While both university and industry scientists are likely to haveetworks of contacts of international reach, the existence of localontacts may be particularly important in pushing forward aca-emic commercialization (Zucker et al., 1998). As emphasizedithin research on university spillovers into regional economies,

ocal ties may facilitate the creation and transfer of tacit knowl-dge across inventors (Liebeskind et al., 1996; Zucker et al., 2002).ore generally, the co-location of universities and agglomer-

tions of technology oriented companies may conduce towardhe creation of social networks linking university and industryersonnel as such individuals become acquainted through localrofessional and social activities. From this perspective, universi-ies located within sizeable technology clusters may be expectedo commercialize more technology compared to universities thatre not.

While such agglomeration effects are no doubt important, atronger, less haphazard explanation locates the origin of local tiesetween universities and companies as a result of strategic action of

ocal actors. Sociologists, economic geographers, and other regionaltudies experts have long argued that regions can develop uniqueatterns of social organization (see e.g. Piore and Sabel, 1984;arkusen, 1987; Herrigel, 1993; Storper, 1997; Scott, 1998; Locke,

999). Saxenian’s (1994) well-known research on the computerndustry in Silicon Valley and the Route 128 region of Boston helpsxplain how social structure can shape the behavior of a region’sompanies (see also Almeida and Kogut, 1999). Saxenian argueshat that in Silicon Valley dense networks linking area managersnd scientists were created by high patterns of inter-firm laborarket mobility and sustained by the existence of norms legitimiz-

ng frequent contact between scientists, engineers, and managersorking within area companies. Firms in the Boston region, by

ontrast, had developed more insular or autarchic human resourceractices that, coupled with low labor market mobility across firms,

imited the development of strong social networks across firms.axenian argues that social networks can rapidly diffuse infor-ation, allowing firms plugged into such networks to react to

echnological developments faster than competitors. Moreover,ocial ties coupled with high labor market mobility can increase arm’s human capital flexibility in reacting to market and technol-gy changes, while diminishing career risks for talented individualsontemplating careers in a high-risk firm (see Bahrami and Evans,999; Casper, 2007).

Saxenian’s research on Silicon Valley and Route 128/Boston isart of a long tradition of research suggesting that the compet-

tiveness of local economies is tied to the quality of inter-firmelationships within a region (Becattini, 1978; Piore and Sabel,984; Sabel and Zeitlin, 1985; Herrigel, 1993; Locke, 1999). As men-ioned earlier, Powell and collaborators are particularly importantontributors to this research tradition, emphasizing the importancef network forms of organization within science based industryPowell, 1996; Powell et al., 1996, 2005a,b) and using networknalysis to compare the organization of social ties linking orga-izations, scientists, entrepreneurs and financiers cross-regionallyOwen-Smith and Powell, 2004). A key finding from this researchs that social structures conducive toward the formation of inter-rganizational ties are likely to pull university scientists intoocal technical communities. The corporate activities that areinked to regional social network formation across companies andniversities – high labor market mobility and extensive inter-rganizational collaboration – are likely to facilitate also facilitate

he formation of ties drawing universities into commercializationrocesses.

A common mechanism by which university–industry tiesmerge is through employment relationships. A significant

2 (2013) 1313– 1324 1315

number of PhD students and postdoctoral fellows move into indus-try careers (NSF, 2006; Smith-Doer, 2005). University scientistsmoving into local firms will presumably preserve ties with theirold laboratory. Employment ties are likely to accumulate morerapidly in regions with high labor market mobility, as firms willneed to hire more frequently. Moreover, as former university sci-entists move across a series of jobs within a regional economy, theygenerate referral networks across those companies and the univer-sity. Regions that develop high job mobility cross organizations, asSaxenian and other have argued exists in Silicon Valley, will developdenser social networks. It may be easier for scientists within uni-versity labs to become embedded within such referral networkswhen they are located in regions with rapid job circulation.

The decision of a company to embrace a particular form ofindustrial organization can be influenced by the broader socialorganization a region’s economy. Saxenian’s comparison of Sili-con Valley and Route 128 echoes earlier accounts of industrialdistricts, such as Herrigel’s (1993) research on decentralized sup-plier networks in the Baden Wuerttenberg machine tool industry orPiore and Sabel’s (1984) discussion of flexible specialization withinNorthern Italy’s textile sector. Within each of these accounts mostcompanies in a region participate in a dominant form of indus-trial organization. To the extent that most organizations in a regiondevelop similar organizational practices, their activities will pro-duce a distinct social structure within a region.

3. Research design

This study investigates the organization and impact of socialnetworks linking scientists in the biotechnology field on aca-demic commercialization within two regions of California, the SanFrancisco Bay Area, and the greater Los Angeles area, includingOrange County but not San Diego. To assess commercialization out-put within each region, data on biotechnology related universitypatenting from major universities in each region from the 1980 to2005 time period is used. Social network analysis is used to exam-ine the size, connectedness, and composition of inventor networkslinking biotechnology scientists across the same time period, allow-ing correlations to be made comparing the relative strength of socialnetworks to university commercialization outputs.

Research on Silicon Valley has documented the existence ofhigh labor market mobility and pervasive social networks linkingscientists and mangers within the region’s electronics and com-puter industries (Fleming et al., 2007; Almeida and Kogut, 1999;Saxenian, 1994). Given the existence of a large and vibrant biotech-nology cluster in the San Francisco region (Romanelli and Feldman,2006), it is likely that similar social networks link inventors work-ing within the life sciences. Such a finding would allow subsequentanalysis of whether such ties are linked to increased incidence ofacademic commercialization. The Los Angeles region is home totwo prominent large biotechnology companies, Amgen and Aller-gan, but otherwise is not known to be a region with a vibrantcluster of entrepreneurial biotechnology companies (see Casper,2009). From a research design perspective Los Angeles was chosenbecause it is likely to be a negative case, one where strong socialnetworks do not connect academic and industry scientists. If so,this would allow Los Angeles to serve as a baseline case for com-parison to the San Francisco region. The expected empirical findingof the study is that dense social networks linking regional biotech-nology scientists from both industry and academia exist in the SanFrancisco area but not in the Los Angeles area, and that, possibly

as a result, academic commercialization is higher in San Franciscoregional universities, compared again to the Los Angeles region.

The Los Angeles and San Francisco regions share several simi-larities that help control for variables that may drive differences in

1316 S. Casper / Research Policy 42 (2013) 1313– 1324

Table 1University capabilities across the greater Los Angeles and San Francisco regions.

Size of faculty Nobel prizes(total/prizes inmedicine)

Members of nationalacademies(total/membersnational institute ofmedicine)a

NationalInstitute ofHealth Funding(2009)($million)

TotalSponsoredResearchFunding (2009)($million)

Los Angeles regionCalifornia Institute of Technology 416 31/9 109/6 47 260University of Southern California 3200 3/0 57/13 175 484UC Los Angeles 2654 6/1 119/37 384 750UC Irvine 1685 3/0 40/5 113 218UC Riverside 549 0/0 7/0 14 110

San Francisco regionStanford University 1910 16/1 351/66 307 1380UC Berkeley 2047 22/0 234/13 114 649UC San Francisco 1934 5/5 111/74 464 933UC Davis 1503 0/0 34/11 173 622

Los Angeles total 9589 43/10 332/61 733 1822San Francisco total 7394 43/6 730/164 1058 3584

S pectivg, and

avr2rU(ALeIChSFlmbib

vbda2trh4beNaisA

bNooft

ource: NIH funding from the NIH reports website. All other data gathered from resa National Academies include the US Academy of Science, Academy of Engineerin

cademic commercialization rates. An important predictor of uni-ersity commercialization output is the size and quality of basicesearch endowments within a university (Powers and McDougall,005). Both Los Angeles and San Francisco have a number of largeesearch oriented universities. Each region is home to three majorniversity of California campuses: Berkeley (UCB), San Francisco

UCSF), and Davis (UCD) in the broader San Francisco area and Losngeles (UCLA), Irvine (UCI), and Riverside (UCR) in the greateros Angeles region. Large private research universities also exist inach region: Stanford in the San Francisco area and the Californianstitute of Technology (Caltech) and the University of Southernalifornia (USC) in the greater Los Angeles region. Each region alsoas significant medical schools attached to universities. Within thean Francisco area Stanford and the University of California at Sanrancisco each have medical schools, while in the greater Los Ange-es area USC, UCLA, and UCI have medical schools. The presence of

edical schools is an important driver of commercialization in theiomedical field due to the importance of clinical studies in validat-

ng efficacy and safety in humans of molecules discovered throughasic laboratory research (see e.g. Jong, 2006; Kenney, 1986).

While both Los Angeles and San Francisco have a number of uni-ersities, it is useful to compare more closely various indicators ofasic science endowments. Table 1 presents a range of summaryata on the size and performance of major research universitiescross the greater Los Angeles and San Francisco regions as of010. Both regions have very large university research infrastruc-ures, with about 7400 total faculty employed in the San Franciscoegion and about 9600 in and around Los Angeles. Both regions areome to numerous prominent faculty: faculty have been awarded3 Nobel Prizes in each region, though faculty in Los Angeles haveeen awarded more prizes in medicine (10 compared to 6). How-ver, the San Francisco region is home to more members of the USational Institute of Medicine (133 to 64). The San Francisco regiont an aggregate level conducts more university research: for 2009ts four major research universities generated over $3.5 billion inponsored research funding, compared to $1.8 billion in the Losngeles region.

A more direct of biomedical research that could potentiallye commercialized is peer reviewed funding secured from theational Institutes of Health (NIH). Data is available from the NIH

n the number of projects funded to each university from 1989nwards, and on aggregate funding awarded to most universitiesrom 2000 onwards. With only minor variation on a yearly basis,his data shows that Los Angeles area universities, as a whole, were

e university websites in June 2010. the National Institute of Medicine.

awarded about 70% as many projects as San Francisco region uni-versities, and also about 70% as much funding. For example, in1989 San Francisco area universities were awarded 1821 projects,compared to 1293 in the Los Angeles area (71% as many projects).In 2005 these figures were 1911 compared 2603 (73% as manyprojects for Los Angeles). Thus, if the level of NIH funding wereused as a simple proxy for the size of basic research endowmentsin the biosciences, one would expect that the San Francisco regionwould have somewhat stronger commercialization rates, at about30% higher than the Los Angeles region.

The comparison of the Los Angeles and San Francisco regionsalso helps to partially control for variation in the organization andpolicies surrounding university technology transfer offices. Six ofthe nine major universities across the two regions are University ofCalifornia (UC) campuses, which operate under a centralized tech-nology transfer system, with common rules and procedures acrossall campuses and a centralized budgeting procedure (Mowery et al.,1999). While particular UC campuses might develop unique capa-bilities due to the presence, for example, of exceptionally strong(or poor) licensing officers or administrative staff, it is reasonable toassume that in the case of UC campuses variation in academic com-mercialization output will not be strongly driven by differences inthe financial resources or rules and procedures governing licens-ing. That being said, the study does not control for variation inthe effectiveness of licensing offices within the private universi-ties across the two regions. It is also worth noting, however, thatStanford and the California Institute of Technology were ranked 3rdand 4th nationally in a study of the effectiveness of university tech-nology licensing offices across 135 American research universitiesduring the 2000–2004 period (Devol and Bedroussian, 2006). TheUniversity of California System as a whole was ranked 2nd in thisstudy, and the University of Southern California was ranked 15th.These findings suggest that effective technology transfer organiza-tions existed across each of the nine universities located across SanFrancisco and Los Angeles.

4. Methodology

Information from a database of all biotechnology related USpatents applied for between 1980 and 2005 is used to structure the

analysis. Two types of evidence were gathered from patents: dataon the academic commercialization outputs of universities and dataused to examine social networks linking biotechnology inventorsin each region.

olicy 4

4

tagdnoiiosaifioiltAvslT

tllustnUytvtloKbci

4

octesakpmdtasbtci

S. Casper / Research P

.1. Academic commercialization data

University patenting is the primary indicator of commercializa-ion output used. Patent data was chosen for three reasons. Firstly,ll patents include the application year, allowing count data to beenerated on a year by year basis and then correlated with evi-ence on social ties linking university scientists to broader inventoretworks for that year. Secondly, all patents are assigned a technol-gy class, which allows relatively simple grouping of patents intondustry fields, in this case biotechnology. Thirdly, patenting datas publicly available, allowing the creation of a complete datasetf biotechnology related patenting for all nine of the universitiestudied. There are at least two weaknesses presented by patentss an indicator of commercialization success. One potential issues that a large number of patents developed by universities fail tond commercial licensees. A study of the Massachusetts Institutef Technology’s licensing office, widely regarded as a leading officen the United States, found that only 51% of patents were undericense at any given time (Shane, 2002), and across all universitieshe percentage of patents under license may be as low as 30% (seeUTM, 2010). A second potential problem is that there is a wideariation in the value of particular patents, even when licensed. Amall number of licensed patents generate tens of millions of dol-ars in yearly income, while most receive much less (AUTM, 2010;hursby and Thursby, 2007).

Other potential indicators of commercialization output includehe number of invention disclosures filed by faculty, number oficensing deals negotiated between a university and third parties,icensing revenue, and number of new ventures launched usingniversity technology (often referred to as “spin-offs”). Given thistudy’s core claim that social ties between universities and scien-ists are linked to commercialization output, the number of dealsegotiated per year would be a particularly good measure to use.nfortunately, published data on this measure does not exist on aearly basis for many of the universities examined, and when it doeshis data is not broken down by industry field. Some studies of uni-ersity technology transfer focus on new venture creation linkedo university technology, as this activity is most closely linked toocal economic development and, moreover, ownership in spin-uts can be financially lucrative for universities (see Breznitz, 2011;enney and Patton, 2009). While information on the number ofiotechnology spin-out is not available on a yearly basis, data on theumulative number of biotech spin-outs for California universitiess available up until 2002, and is included as a secondary indicator.

.2. Inventor networks

Social network analysis can help explore the existence andrganization of ties linking scientists working in universities andompanies within a given region. Personal contacts between scien-ists may result from any number of activities, such as shared socialxperiences or receiving graduate training in the same univer-ity. Social ties developed through collaborative scientific activitiesre particularly salient as they demonstrate common scientificnowledge. Such ties are frequently documented through scientificublication and patent applications. The analysis here draws on aethodology first employed by Fleming et al. (2007) to use patent

ata to capture networks forged among scientists listed as inven-ors on patents. This data is used to construct inventor networks on

yearly basis, which can then be used to examine whether cohe-ive networks have evolved within each of the three core California

iotechnology regions. Simple network statistics and visualiza-ions will be used to compare the relative size of each region’so-inventor networks, as well as the composition of organizationsnhabiting the network.

2 (2013) 1313– 1324 1317

Patent data for the project was drawn from a large databasesummarizing all US patents issued between 1963 and 2009 whichwas purchased from the US Patent Office. The database used is sim-ilar in form to the National Bureau of Economic Research PatentProject database (Hall et al., 2001) but had the advantage of contain-ing data on issued patents up to 2009. The US Patent Office routinelytakes several years for a patent application to be processed. As aresult, data from patents could only reliably be examined up untilthe year 2005, and it is likely that additional patents applied forwithin the 2002–2005 period have been granted after 2009 and arethus not included in the database used for this project. Using indus-try classifications from the NBER Patent project, data was collectedon all biotechnology related patents in which at least one inven-tor resided in a Los Angeles or San Francisco region city between1980 and 2005. This search yielded 19,441 patents, of which 15,388were assigned to one or more San Francisco inventors, and 4053 toinventors in the Los Angeles region. The year 1980 was chosen for astarting date because the Bayh–Dole Act, which formally assignedownership of federally funded research to universities, came intoeffect that year. Moreover this year coincides with the early historyof the biotechnology industry, commonly ascribed to the foundingof Genentech in 1976.

Patent applications list names and the city of residence of allinventors, as well as the organization that originally filed for thepatent, called the assignee. The address information of inventorswas used to sort patents into one of the two California biotechnol-ogy regions selected for study. Once sorted into regional files, thenames of inventors was manually cleaned to insure uniformity ofspellings within the social network analysis, as slight variation innames sometimes occurs when individuals move to jobs in differ-ent organizations. All patents were also classified by organizationaltype based on the assignee, allowing the identification of patentsfiled by universities and biotechnology companies.

Co-inventor networks are formed when patents are filed withmultiple inventors: all inventors listed on a patent are assumed tohave network ties to one another. Ties linking individuals acrosspatents are created as inventors filed subsequent patents with dif-ferent co-inventors. This commonly occurs through career moves tonew organizations, inter-organizational collaborations, or throughnew collaborations with scientists within an organization. Thus,if patent A listed Cindy and Doug as inventors, and patent B listedDoug and Melissa as inventors, a network linking Cindy and Melissawould be formed through their common tie to Doug. Followingmethods used in previous studies social network studies, inventorswere assumed to be active within regional co-inventor networksuntil five years after their most recent patent; after this timeinventors were removed from the network (Uzzi and Spiro, 2005;Fleming et al., 2007).

5. The growth of co-inventor networks and academiccommercialization

The results of the analysis are presented in three parts. Firstly,descriptive statistics describing commercialization outputs acrossthe San Francisco and Los Angeles universities are reviewed. Sec-ondly, data on the size and organization of inventor networks overthe 1980–2005 period is analyzed for both regions. Finally, morespecific data on the embeddedness of university scientists withininventor networks is examined.

5.1. Descriptive statistics on academic commercialization

Table 2 presents data on commercialization outputs across SanFrancisco and Los Angeles universities over the 1980–2005 period.The patent counts refer to the application year of biotechnology

1318 S. Casper / Research Policy 42 (2013) 1313– 1324

Table 2Number of biotechnology patents issued to SF and LA universities, 1980–2005.

Stanford UCSF UC Berkeley UC Davis Caltech UCLA USC UCI UCR SF Bay Area LA Area

1980 1 4 2 2 1 3 0 0 2 9 61981 2 4 0 0 0 3 1 0 3 6 71982 4 8 1 4 0 4 1 1 1 17 71983 8 6 3 3 2 4 2 0 1 20 91984 11 6 4 5 0 7 3 4 0 26 141985 10 3 3 2 2 3 0 3 1 18 91986 4 7 3 4 1 2 2 1 0 18 61987 6 5 1 3 1 5 1 3 0 15 101988 6 4 1 2 3 3 1 1 0 13 81989 7 4 8 3 1 3 2 3 2 22 111990 5 3 9 5 0 3 2 3 2 22 101991 9 11 8 5 4 2 7 3 1 33 171992 10 7 7 3 3 5 2 2 3 27 151993 18 23 11 5 5 12 5 5 0 57 271994 30 34 14 13 11 17 9 8 2 91 481995 32 90 28 22 17 24 12 10 2 172 651996 19 37 16 9 18 14 8 8 2 81 501997 28 49 28 9 11 21 7 11 3 114 531998 28 58 31 21 13 24 6 8 3 138 541999 30 47 40 14 27 18 18 11 2 131 762000 28 50 30 11 20 21 11 8 5 119 652001 40 58 29 13 33 21 14 22 3 140 932002 29 31 25 15 14 24 11 10 3 100 622003 24 37 13 5 11 14 5 12 1 79 432004 13 26 10 6 5 3 6 9 1 55 232005 9 9 6 1 6 6 0 1 0 25 13

Total patents 411 621 331 185 209 266 136 147 26 1548 801

S Inform

pifnugh

tiTv(itpprn

tial differences exist in commercialization output across the tworegions. San Francisco universities have been granted close to twiceas many patents as LA universities, 1548 to 801, and have launchedalmost three times as many companies. Referring to Fig. 2, it is

Total biotech spin-outs 117 79 87 26

ource: Data on university patenting collected from the United States Patent Office.

atents that were eventually issued by the US Patent Office. Datas presented for each university, as well as aggregate patent totalsor each region as a whole. Table 2 also includes data on the totalumber of spin-out biotechnology companies from each universityp until 2002. To help better visualize the aggregate results, Fig. 1raphically illustrates the total number of biotechnology patentseld by universities in each region on a yearly basis.

Before reviewing variation in university patenting, it is usefulo mention three general points. A first trend is the steady increasen university patenting across all universities from 1980 onward.his finding is consistent with studies documenting increased uni-ersity patenting since the enactment of the Bayh–Dole Act in 1980Mowery et al., 2004). A second observation is the decline in patent-ng from 2002 to 2005. This is caused by the lengthy review periodaken by the US Patent Office before granting patents. Many of the

atents applied for by universities in the 2002–2005 period hadresumably not finished review prior to 2009, the year of the mostecent US Patent Office database was released. Finally, there is aoticeable spike in patenting in 1995. This was caused by a change

0

20

40

60

80

100

120

140

160

180

200

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

San Francisco

Los Angeles

San Francisco

Fig. 1. Total academic biotechnology patents, LA and SF universities, 1980–2005.

28 40 25 16 3 299 112

ation on biotechnology spin-outs from the California Healthcare Institute (2004).

in patent rules governing patent life brought about by the enact-ment of the Uruguay Round of international trade negotiations inmid-1995. Under the new rules, all patents would be granted a 20-year life from the date of filing, as opposed to the previous systemwhich granted a 17-year patent life from the date of issue. Becausethe US Patent Office commonly took more than 3 years to evaluatebiotechnology patents, universities and other organizations pushedforward new patent applications in 1995 in order to take advantageof the pre-Uruguay Round rules.

Viewing the data from a comparative perspective, substan-

Fig. 2. Main component of Los Angeles inventor network, 2002.

olicy 4

imbwimufCma

sLrhfpsotm

5

twhvPusSdio

TG

S. Casper / Research P

nteresting to note that while SF universities have always filedore biotechnology patents, the difference in commercialization

ecomes particularly pronounced in the post-1990 period. It isorthwhile to take note of this periodization, as it will become

mportant when analyzing the inventor network results. Moreover,uch of the regional variation in patenting is driven by two SF area

niversities, Stanford and UCSF. Stanford, for example, has success-ully applied for close to twice as many biotechnology patents asaltech (411 to 209), and UCSF has collected more than twice asany patents as UCLA (621 to 266). On the other hand, UC Berkeley

nd UCLA have broadly comparable totals (331 to 266).The review of university endowments did show an overall

uperiority of San Francisco regional universities compared to theos Angeles region, but of an incremental nature. The NIH dataeviewed earlier suggests that San Francisco universities in totalad about 30% more individuals projects and overall research

unding. The finding that these universities have dramatically out-erformed Los Angeles area universities in commercialization,uccessfully applying for nearly twice as many patents and spinningff close to three times as many companies, supports a conclusionhat other influences, such as the quality of the external environ-

ent, may help explain this variation in commercialization output.

.2. Inventor networks

Social network analysis can help investigate the premise thathe existence social ties linking academic and industry scientistsill increase academic commercialization. A variety of indicatorsave been developed to link the organization of networks to inno-ative performance (see e.g. Powell et al., 1996; Owen-Smith andowell, 2004; Fleming et al., 2007; Burt, 1992). The analysis hereses simple indicators to explore the structure and composition ofocial ties linking scientists in the biotechnology field within the

an Francisco and Los Angeles regions. Structural data will includeescriptive statistics on the overall size and degree of connectiv-

ty within inventor networks and will be examined for each regionn a yearly basis from 1980 to 2005. Data on the organizational

able 3rowth of biotechnology inventor networks in San Francisco and Los Angeles, 1980–2005

Number of total inventors innetwork

Nunet

SF LA SF

1980 407 181 41981 455 190 31982 541 204 31983 638 226 31984 750 266 51985 859 290 111986 947 309 151987 1092 335 211988 1188 378 241989 1302 413 261990 1442 461 301991 1581 539 481992 1741 605 501993 2008 696 751994 2507 867 1001995 3121 1044 1411996 3433 1129 1651997 3892 1257 1911998 4357 1362 2271999 4645 1484 2512000 4705 1573 2502001 5184 1715 2892002 6142 1777 3502003 5328 1761 3122004 5083 1621 2992005 4497 1423 241

2 (2013) 1313– 1324 1319

affiliation of inventors within the network will be used to ascertainwhether university scientists have established ties with industryscientists.

Social networks differ in their level of fragmentation. The higherthe level of connectivity within the network, measured by the num-ber of individuals one can conceivably reach through either directcontact or through intermediaries, the more useful a network mightbe to its members. A simple measure of connectivity used fre-quently within social network analysis is the size of the largestgrouping or component of individuals in which everyone is con-nected (Faust and Wasserman, 1994). This cluster of individuals istypically called the “main component” within network studies. Ina similar study exploring social network structure and innovationwithin regional economies, Fleming et al. (2007) demonstrated thatthe innovative output of firms increases when the firm is a memberof a sufficiently large main component of inventors. This finding isconsistent with the research by Saxenian (1994) and others linkingthe development of cohesive social networks to increased innova-tive capacity of firms in Silicon Valley.

Table 3 presents descriptive statistics on the general size andconnectivity within the San Francisco and Los Angeles regionalinventor networks. While a sizeable network of inventors hasdeveloped in both regions, San Francisco has a much larger generalagglomeration of biotechnology scientists who have filed patents.At its peak year, in 2002, over 6000 scientists were active in the SanFrancisco regional network, compared to about 1750 in Los Ange-les. While some of this variation may be explained by higher levelsof university patenting, the primary difference is the existence ofa much larger biotechnology industry in San Francisco comparedto Los Angeles. During the 1980–2005 time period over three hun-dred biotechnology firms were founded in the San Francisco region,compared to less than 50 in Los Angeles. Another general findingfrom Table 3 is that inventor networks grew in size gradually over

time. Between 1980 and 2002 the San Francisco region grew at anaverage of 13% a year, compared to 9% in Los Angeles.

While San Francisco developed a much larger community ofbiotechnology inventors, the major difference across the regions is

.

mber of inventors inwork main component

Percent of inventors innetwork main component

LA SF LA

2 4 10% 2%9 5 9% 3%4 6 6% 3%6 6 6% 3%0 8 7% 3%9 9 14% 3%5 9 16% 3%7 10 20% 3%9 11 21% 3%0 10 20% 2%4 11 21% 2%1 13 30% 2%8 16 29% 3%1 21 37% 3%8 25 40% 3%9 33 45% 3%9 32 48% 3%8 39 49% 3%7 47 52% 3%9 44 54% 3%7 56 53% 4%2 67 56% 4%7 71 57% 4%4 101 59% 6%1 127 59% 8%7 122 54% 9%

1320 S. Casper / Research Policy 42 (2013) 1313– 1324

Fran

tofppaGyctcowtosaa

stnirscwj

aAoaionUi

Fig. 3. Main component of San

he level of connectivity within each network. Table 3 contains datan the number of individuals within the network main componentor each year and the percent of individuals within the main com-onent. The Los Angeles region has little network connectivity. Theercentage of inventors active within the main component hoverst 3–4% over most of the region’s history, and peaks at 9% in 2005.iven the relatively small size of the inventor community over mostears, the number of individuals linked to one another in the mainomponent is small, at less than 10 individuals in the 1980s, lesshan 50 through the 1990s, and peaking at 127 in 2004. Networkonnectivity in San Francisco, on the other hand, increased steadilyver the 25-year history of the network. While only 10% of inventorsere members of the main network component in the early 1998s,

his level rose to 20% by the late 1980s, 30% in the early 1990s, andver 50% from 1997 onwards. Because the general network grewubstantially over time, the number of inventors connected to onenother in the network main component exceeded 1000 in 1994nd grew to over 3500 by 2002.

A graphical visualization of the inventor networks helps empha-ize the dramatic difference in the structure of social ties acrosshe two regions. Figs. 2 and 3 provide network visualizations of theetwork main component for the San Francisco and Los Angeles

nventor networks for 2002, the year of peak membership in bothegions. Within these charts the dots, or nodes, represent individualcientists, while the connecting lines represent ties formed througho-inventorships on patents. The San Francisco co-inventor net-ork main component for 2002 links 3507 scientists, compared to

ust 77 for Los AngelesThe quality of the regional environment facing universities

ppears to be much stronger in San Francisco compared to Losngeles, particularly from the early 1990s onwards. Biotechnol-gy scientists located in San Francisco frequently collaborate with

variety of scientists over their careers. As a result, large, cohesivenventor networks developed within the San Francisco biotechnol-

gy community, but not in Los Angeles. Referring back to Fig. 1,ote that San Francisco universities, and particularly Stanford andCSF, established a sizeable advantage in filing patents starting

n the 1990s. A correlation exists between the establishment of

cisco inventor network, 2002.

cohesive regional inventor networks and increased university com-mercialization outputs. To provide stronger evidence of causality, itis useful to examine the extent to which university scientists haveestablished ties with industry scientists.

5.3. Academics within inventor networks

Is there evidence that university scientists have established tieswith industry scientists, particularly within the cohesive inven-tor networks that developed in San Francisco? Table 4 presentsdescriptive data on the activity of university scientists withinregional inventor networks during the 1980–2005 period. At a gen-eral level, the number of academic inventors within each networkagain grows incrementally over time, as universities increase thecommercialization output. San Francisco universities consistentlyout patented Los Angeles universities, and as a result have morescientists within regional networks, 893 in a peak year, 2001, com-pared to 334 in the Los Angeles area.

The columns comparing the number and percent of local aca-demics within the network main component are particularlyrelevant. Academic scientists that are members of the networkmain component, particularly at times when the main componentis large in size, have numerous direct or indirect ties with industryscientists. During the 1980s there is no difference across the tworegions, as only a handful of academic scientists were members ofthe network main component in either region. In Los Angeles thisis partly driven by the lack of a sizeable main component acrossthe broader network. In San Francisco a network main componentwas growing during the 1980s, but university scientists only playeda minor role. These findings are consistent with an interpretationthat universities in both regions were essentially “on their own”in commercializing science during the 1980s, with no or little pulleffect created by the embeddedness of university scientists withinregional inventor networks. This state of affairs remains constant

for Los Angeles area universities through the 19990s up until 2005.Only small numbers of university scientists are members of themost connected part of the Los Angeles regional inventor commu-nity. Referring back to Table 3, this outcome is primarily driven

S. Casper / Research Policy 42 (2013) 1313– 1324 1321

Table 4Involvement of university scientists within regional inventor networks.

Number of local academicinventors in network

Number of local academics innetwork main component

Percent of academic inventorsin network main component

SF LA SF LA SF LA

1980 27 18 0 4 0% 22%1981 35 22 0 5 0% 23%1982 65 37 0 6 0% 16%1983 96 52 0 6 0% 12%1984 126 77 0 4 0% 5%1985 148 79 5 4 3% 5%1986 159 81 4 4 3% 5%1987 174 78 5 5 3% 6%1988 161 77 5 4 3% 5%1989 150 72 6 5 4% 7%1990 150 73 8 4 5% 5%1991 174 108 16 4 9% 4%1992 201 122 21 4 10% 3%1993 266 140 37 3 14% 2%1994 363 180 57 5 16% 3%1995 520 235 144 0 28% 0%1996 622 251 153 0 25% 0%1997 707 281 196 0 28% 0%1998 809 309 288 0 36% 0%1999 835 347 294 1 35% 0%2000 822 340 357 1 43% 0%2001 893 376 334 4 37% 1%2002 885 415 346 4 39% 1%2003 871 483 324 8 37% 2%

332200

bA

int3moeanwim

mia

Fa

2004 812 375

2005 685 346

y the general level of fragmentation across LA inventor networks.cademic inventors in Los Angeles lack a cohesive network to join.

In San Francisco, however, a different dynamic unfolds. Start-ng in the early 1990s the number of scientists belonging to theetwork main component increases, jumping to about 150 duringhe 1995–1996 period and peaking at over 350 in 2000. Between0% and 40% of academic scientists within the broader network areembers of the main component from 1997 onwards. Referring

nce more to Table 3, there is a strong correlation between the gen-ral growth of connectivity within San Francisco inventor networksnd the increased involvement of university scientists within theetwork main component. When the quality of the regional net-ork increased, several hundred scientists were able to become

nvolved in local inventor networks. During this same period, com-ercialization output by local universities increased.

Fig. 4 graphically displays the variation in main component

embership by academics across the two regions and also includesnformation from individual universities in San Francisco. Oncegain, note that from the mid-1990s onwards all San Francisco

0

50

100

150

200

250

300

350

400

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

UCSF

Stanford

UC Berkeley

UC DavisAll Los Angeles

All San Francisco

ig. 4. Number of academics within the network main component by universitynd region.

8 41% 2% 5 29% 1%

regional universities saw an increase in the number of scientiststhat were members of the network main component. Stanford andUCSF, however, had particularly strong participation within thisnetwork. In the year 2000, for example, about 170 UCSF scientistswere members of the regional biotechnology network main compo-nent, as were close to 100 Stanford scientists. This data is consistentwith the observation, noted earlier, that much of the increasein patenting by San Francisco universities during the mid-1990sonwards was driven by Stanford and UCSF. Scientists from thesetwo universities had the strongest network of contacts within themore general industry network, and possibly as a result, engagedin markedly more commercialization activities.

A final issue to investigate is the number of university scientistswithin each region that have direct ties with industry scientists. It islikely that many university scientists have only indirect links withcompanies. An academic scientist that co-invented a technologywith a colleague that has a direct tie to an industry scientist could,for example, be a member of the network main component throughthis indirect tie. While still plugged into a viable inventor network,such scientists would need to ask colleagues for referrals in orderto contact company scientists.

How many academic scientists have direct network ties toindustry inventors within each region? Such ties are commonlyformed in two ways. Firstly, a university scientist may move toa company, establishing ties between the firm and universitythrough patenting with scientists at both organizations. Such tiesoften develop as PhD students or postdoctoral fellows move intoindustry, but sometimes develop when more senior professors takesabbaticals at companies or, rarely, decide to move more perma-nently to companies. Secondly, many professors accept sponsoredresearch grants from companies. Patents developed from suchresearch are usually assigned to the company, creating networklinkages as professors co-invent new technology with scientists

from the company.

Fig. 5 presents the number of inventors from each regionaluniversity who are inventors of at least one patent at a local uni-versity and a company. At a regional level, there are more than

1322 S. Casper / Research Policy 4

0 20 40 60 80 100 120 140

UC Riverside

UC Irvine

USC

Caltech

UCLA

UC Davis

UC Berkeley

UCSF

Stanford

F1

tintwossalltwuc

6

ecmilWcSamrAlitnc

oeSc1mSotwt

university laboratories within corporate research activities might

ig. 5. Number of “linked” inventors, San Francisco and Los Angeles universities,980–2005.

hree times as many scientists in the San Francisco area with directndustry contacts than in Los Angeles: 329 to 100. It is also worthoting that Stanford and UCSF, the two universities with excep-ionally high patent output, each have dramatically more scientistsith direct industry contacts, 130 and 119 respectively, than the

ther universities in either region (see Kenney and Goe, 2004 for aimilar comparison focused on engineering). This evidence is con-istent with findings from the technology transfer literature linkingcademic commercialization outputs to the establishment of tiesinking faculty with industry. San Francisco academics are moreikely to be both embedded in broad inventor networks with indus-ry scientists, and have dramatically more direct co-inventor tiesith industry. From this perspective, it is not surprising that theniversities housing these scientists are more effective in commer-ializing science.

. Concluding discussion

The embeddedness of university scientists within a regionalconomy can influence the ability of a university to effectivelyommercialize science. The findings here, linking academic com-ercialization output to the size and cohesiveness of regional

nventor networks within the San Francisco and Los Ange-es biotechnology industry, support this theoretical approach.

hile basic resource endowments are an important predictor ofommercialization output, the data presented here shows thatan Francisco regional universities obtained more than twices many biotechnology patents and spun-out three times asany firms, though only attracting about 30% more sponsored

esearch funding within its universities compared to the Losngeles area. This stronger commercialization output is corre-

ated with the existence of dramatically larger and more cohesivenventor networks in the San Francisco region. Though facili-ating contacts between academic and industry networks, theseetworks helped facilitate the commercialization of university dis-overies into industry.

Longitudinal data provides additional support for the the-ry that the structure of social networks within regionalconomies promote academic commercialization. Universities inan Francisco saw a marked increase in patenting after large, well-onnected inventor networks developed in the region in the early990s. Moreover, the two universities whose scientists becameost involved in these networks from the early 1990s onwards,

tanford and UCSF, saw the largest increases in commercializationutput. Stanford and UCSF also had the most academic scien-

ists with direct ties to industry scientists. Los Angeles, a regionith several strong universities but a weak biotechnology indus-

ry, never established well-connected inventor networks, and its

2 (2013) 1313– 1324

universities fell behind those in San Francisco in commercializationoutput during the 1990s.

The study was designed to put forward evidence that wouldhelp support the theory that the organization of social tieswithin regional economies can impact university commercializa-tion. These findings contribute to a small but growing literatureexamining the impact of environmental factors on university com-mercialization output (see Powell et al., 1996; Mowery et al., 2004;Siegel et al., 2003). Aside from the brief comparison of universityendowments, it did not systematically investigate a variety of inter-nal factors, such as university licensing policies or variation in theorganization or financing of university technology transfer offices.Which strong variation in internal factors can influence commer-cialization output, it is also important to note the convergenceacross universities, particularly within the United States, in licens-ing policies and organization (see generally Breznitz and Feldman,2012). Future research might introduce more sophisticated statis-tical modeling techniques in order to investigation of the relativeweight of internal versus external variables in influencing univer-sity commercialization.

From a policy perspective, an implication of this study is thatexpectations as to a university’s commercialization output shouldfactor in the quality of its local regional economy. Universities sit-uated in high quality regions, in terms of the existence of highlycollaborative local industry networks, should have higher commer-cialization outputs compared to similar to universities in lowerquality regional economies. To provide one example, this studyshowed that UC Berkeley and UCLA have broadly similar basicresearch endowments and that UC Berkeley had a slight lead inbiotechnology patents developed from 1980 to 2005 (333 to 266).UCLA, however, has been criticized within Southern Californiafor underperforming in its commercialization efforts (see e.g.Silverman, 2007; Casper, 2009). This criticism could very well beunfounded. Given the low quality regional biotechnology environ-ment and paucity of local networking in the Los Angeles region, thecommercialization output of UCLA might actually be strong. On theother hand, given the vibrancy of San Francisco inventor networks,UC Berkeley might very well be underperforming. Why have notscientists at UC Berkeley failed to become strongly involved inregional inventor networks to the extent that scientists at UCSFor Stanford have? It is here that internal factors might very wellbe important. But in any case, an implication of this study is thatresource endowments should not serve as the only benchmarksused to compare university commercialization output. The poten-tial of a university to effectively commercialize science is influencedby the quality of its regional environment.

The study has implications for thinking about the general roleof universities in economic development. The university spill-overperspective assumes a linear flow of knowledge from universitiesto the external environment. The results from this study do notdirectly challenge the idea that knowledge flows from universi-ties toward industry, but suggests that the quality of a university’senvironment can strongly impact whether the networks throughwhich such knowledge flows develop. A more controversial impli-cation of the study is that the linear or waterfall image of universityknowledge flows might be incorrect. The involvement of univer-sity scientists within regional knowledge networks might shapethe flow of knowledge more directly, creating more of a circularflow of knowledge to and from universities. The dense inventornetworks found in the San Francisco region may in part be a productof highly collaborative “open science” strategies embraced by manyarea companies (Rhoten and Powell, 2007). The involvement of

shape the research capabilities and agenda of university scientists.The present study presented evidence that, in the cases of Stanfordand UCSF, hundreds of academic scientists can become involved in

olicy 4

renio

A

gtc

R

A

A

A

A

A

A

B

B

B

B

B

B

C

C

C

D

EF

F

F

F

HH

H

J

J

K

K

K

S. Casper / Research P

egional commercial innovation networks. Future research couldxamine more carefully the collaborations underpinning theseetworks and examine the extent to which the involvement of

ndustry collaborators shapes the research capabilities or agendaf university professors (see e.g. Murray, 2004).

cknowledgements

The research reported on in this article was supported by arant from the Joseph Rudolph Haynes Foundation. I would likeo thank Martin Kenney and three anonymous reviewers for usefulomments that helped improve the article.

eferences

lmeida, P., Kogut, B., 1999. Localization of knowledge and the mobility of engineersin regional networks. Management Science 45, 905–917.

rrow, K., 1962. Economic welfare and the allocation of resources for invention.In: Nelson, R. (Ed.), The Rate and Direction of Inventive Activity: Economic andSocial Factors. Princeton University Press, Princeton, NJ, pp. 609–625.

ssociation of University Technology Managers (AUTM), 2010. The AUTM LicensingSurvey, FY 2010. AUTM, Norwalk, CT.

udretsch, D., Feldman, M.P., 2003. Knowledge spillovers and the geography of inno-vation. In: Vernon Henderson, J., Jacque Thisse (Eds.), Handbook of Urban andRegional Economics, vol. 4. North Holland Publishing, Amsterdam.

udretsch, D.B., Feldman, M.P., 1996. R&D spillovers and the geography of innovationand production. American Economic Review 86, 630–640.

udretsch, D.B., Lehmann, E., 2005. Does the knowledge spillover theory ofentrepreneurship hold for regions? Research Policy 34, 1191–1202.

ecattini, G., 1978. The development of light industry in Tuscany: an interpretation.Economic notes. Monte dei Paschi di Siena 7, 107–123.

raunerhjelm, P., Feldman, M.P. (Eds.), 2006. Cluster Genesis Technology-basedIndustrial Development. Oxford University Press, Oxford.

ahrami, H., Evans, S., 1999. Flexible re-cycling and high-technology entrepreneur-ship. California Management Review 37, 62–88.

reznitz, S., Feldman, M.P., 2012. The engaged university. Journal of TechnologyTransfer 37, 139–157.

reznitz, S., 2011. Improving or impairing? Following technology transfer changesat the University of Cambridge. Regional Studies 45, 463–478.

urt, R., 1992. Structural Holes: The Social Structure of Competition:. Harvard Uni-versity Press, Cambridge.

alifornia Healthcare Institute, 2004. California’s Biomedical Industry, 2004.California Healthcare Institute, La Jolla.

asper, S., 2009. The Marketplace for Ideas: Can Los Angeles Build a SuccessfulBiotechnology Cluster? Report Written for the John Randolph Haynes Founda-tion, Claremont.

asper, S., 2007. How do technology clusters emerge and become sustainable? Socialnetwork formation and inter-firm mobility within the San Diego biotechnologycluster. Research Policy 36, 438–455.

evol, R., Bedroussian, A., 2006. Mind to Market: A Global Analysis of UniversityBiotechnology Transfer and Commercialization. Milken Institute, Los Angeles.

tzkowitz, H., 2002. MIT and the Rise of Entrepreneurial Science. Routledge, London.aust, K., Wasserman, S., 1994. Social Network Analysis: Methods and Applications.

Cambridge University Press, Cambridge.leming, L., King, C., Juda, A., 2007. Small worlds and innovation. Organization Sci-

ence 18, 938–954.lorax, R., 1992. The University: A Regional Booster? Economic Impacts of Academic

Knowledge Infrastructure. Aldershot, Avenbury.lorida, R., Choen, W.M., 1999. Engine or infrastructure? The university role in

economic development. In: Brandscome, L.M., Kodama, F., Florida, R. (Eds.),Industrializing Knowledge: University–Industry Linkages in Japan and theUnited States. MIT Press, Cambridge, pp. 589–610.

age, G., 2011. Restoring the Innovative Edge. Stanford University Press, Stanford.all, B.H., Jaffe, A.B., Trajtenberg, M., 2001. The NBER Patent Citation Data File:

Lessons, Insights and Methodological Tools. NBER Working Paper 8498.errigel, G., 1993. Large firms, small firms, and the governance of flexible special-

ization: the case of Baden Württenberg and sozialized risk. In: Kogut, B. (Ed.),Country Competitiveness. Oxford University Press, New York.

affe, A., Trajtenberg, M., Henderson, R., 1993. Geographic localization of knowledgespillovers, as evidence by patent citations. Quarterly Journal of Economics 108,863–911.

ong, S., 2006. How organizational structures in science shape spin-off firms: thebiochemistry departments of Berkeley, Stanford, and UCSF and the birth of thebiotech industry. Industrial and Corporate Change 15, 251–283.

enney, M., 1986. Biotechnology: The University–Industrial Complex. Yale Univer-sity Press, New Haven.

enney, M., Patton, D., 2009. Reconsidering the Bayh–Dole Act and the currentuniversity invention ownership model. Research Policy 38, 1407–1422.

enney, M., Goe, W.R., 2004. The role of social embeddedness in professorialentrepreneurship: a comparison of electrical engineering and computer scienceat UC Berkeley and Stanford. Research Policy 33, 691–707.

2 (2013) 1313– 1324 1323

Lam, A., 2007. Knowledge networks and careers: academic scientists inindustry–university links. Journal of Management Studies 44, 993–1016.

Liebeskind, J., Oliver, A., Zucker, L., Brewer, M., 1996. Social networks, learning, andflexibility: sourcing scientific knowledge in new biotechnology firms. Organiza-tion Science 7, 428–443.

Locke, R., 1999. Remaking the Italian Economy. Cornell University Press, Ithaca.Markusen, A., 1987. Regions, the Economies and Politics of Territory. Rowman &

Littlefield, New York.Mowery, D., Nelson, R., Sampat, B., Ziedonis, A., 2004. Ivory Tower and Indus-

trial Innovation: University–Industry Technology Transfer Before and After theBayh–Dole Act in the United States. Stanford University Press, Stanford.

Mowery, D., Nelson, R., Sampat, B., Ziedonis, A., 1999. The effects of the Bayh–DoleAct on U.S. University Research and Technology Transfer: an analysis of datafrom Columbia University, the University of California, and Stanford University.In: Branscomb, L., Florida, R., Ishii, M. (Eds.), Industrializing Knowledge. MITPress, Cambridge, Mass.

Mowery, D., Sampat, B., 2005. The Bayh–Dole Act of 1980 and university–industrytechnology transfer: a model for other OECD governments? Journal of Technol-ogy Transfer 30, 115–127.

Murray, F., 2004. The role of inventors in knowledge transfer: sharing in the labora-tory life. Research Policy 33, 643–659.

National Academies, 2007. Rising Above the Gathering Storm: Energizing andEmploying America for a Brighter Economic Future. The National AcademiesPress, Washington, DC.

National Institutes of Health, 2010. NIH Reporter Website.http://projectreporter.nih.gov/reporter.cfm (accessed 21.05.10).

National Science Foundation, 2006. Survey of Doctoral Recipients. National ScienceFoundation, Arlington.

O’Shea, R., Allen, T., Chevalier, A., Roche, F., 2005. Entrepreneurial orientation, tech-nology transfer and spin off performance of U.S. universities. Research Policy 34,994–1109.

Piore, M., Sabel, C., 1984. The Second Industrial Divide. Basic Books, New York.Powers, J.B., McDougall, P.P., 2005. University start-up formation and technol-

ogy licensing with firms that go public: a resource-based view of academicentrepreneurship. Journal of Business Venturing 20, 291–311.

Powell, W., 1996. Inter-organizational collaboration in the biotechnology industry.Journal of Institutional and Theoretical Economics 120, 197–215.

Powell, W., Koput, W., Smith-Doerr, L., 1996. Interorganizational collaboration andthe locus of innovation: networks of learning in biotechnology. AdministrativeScience Quarterly 41, 116–145.

Powell, W., White, D., Koput, K., Owen-Smith, J., 2005a. Network dynamics and fieldevolution: the growth of inter-organizational collaboration in the life sciences.American Journal of Sociology 1104, 1132–1205.

Powell, W., Porter, K., Bunker, K., 2005b. The institutional embeddedness ofhigh-tech regions. In: Breschi, S., Malerba, F. (Eds.), Clusters, Networks, andInnovation. Oxford University Press, Oxford.

Powell, W., Koput, K., Bowie, J., Smith-Doerr, L., 2002. The spatial clustering of scienceand capital: accounting for biotech firm–venture capital relationships. RegionalStudies 36, 291–305.

Owen-Smith, J., Powell, W., 2004. Knowledge networks as channels and conduits:the effects of spillovers in the Boston Biotechnology Community. OrganizationScience 51, 5–21.

Rhoten, D., Powell, W., 2007. The frontiers of intellectual property: expanded pro-tection vs. new models of open science. Annual Review of Law and Social Science3, 345–373.

Romanelli, E., Feldman, M.P., 2006. Anatomy of cluster development: the case of U.S.human biotherapeutics, 1976–2003. In: Braunerhjelm, P., Feldman, M. (Eds.),Cluster Genesis: The Origins and Emergence of Technology-based EconomicDevelopment. Oxford University Press, Oxford.

Rosenberg, N., Nelson, R., 1996. The roles of universities in the advance of industrialtechnology. In: Rosenbloom, R.S., Spencer, W.J. (Eds.), Engines of Innovation:U.S. Industrial Research a the End of an Era. Harvard Business School Press,Cambridge, MA, pp. 87–109.

Rosenberg, N., Nelson, R., 1994. American universities and technical advance inindustry. Research Policy 23, 323–348.

Sabel, C., Zeitlin, J., 1985. Historical alternatives to mass production: politics, marketsand technology in nineteenth century industrialization. Past and Present 108,133–176.

Saxenian, A., 1994. Regional Advantage: Culture and Competition in Silicon Valleyand Route 128. Harvard University Press, Cambridge, Mass.

Scott, A., 1998. Regions and the World Economy. Oxford University Press, Oxford.Shane, S., 2002. Selling university technology: patterns from MIT. Management Sci-

ence 48, 122–137.Siegel, D., Waldman, D., Link, A., 2003. Assessing the impact of organizational prac-

tices on the relative productivity of university technology transfer offices: anexploratory study. Research Policy 32, 27–48.

Silverman, E., 2007. The trouble with tech transfer. In: The Scientist, January 1.Sine, S., Shane, S., Di Gregario, D., 2003. The halo effect and technology licensing:

the influence of institutional prestige on the licensing of university inventions.Management Science 49, 478–496.

Smith-Doer, L., 2005. Institutionalizing the network form: how life scientists legiti-

mate work in the biotechnology industry. Sociological Forum 20, 271–299.

Storper, M., 1997. The Regional World: Territorial Development in a Global Economy.Guilford Press, London.

Thursby, J., Thursby, M., 2007. University licensing. Oxford Review of EconomicPolicy 23, 620–639.

1 olicy 4

T

U

V

Zucker, L., Darby, M., Brewer, M., 1998. Intellectual human capital and the birth of

324 S. Casper / Research P

hursby, J., Thursby, M., 2004. Are faculty critical? Their role in university–industry

licensing. Contemporary Economic Policy 22, 162–178.

zzi, B., Spiro, J., 2005. Collaboration and creativity: the small world problem. Amer-ican Journal of Sociology 111, 447–504.

on Hippel, E., 1994. Sticky information and the locus of problem solving: implica-tions for innovation. Management Science 404, 429–439.

2 (2013) 1313– 1324

U.S. biotechnology enterprises. American Economic Review 88, 290–306.Zucker, L., Darby, M., Armstrong, J., 2002. Commercializing knowledge: university

science, knowledge capture, and firm performance in biotechnology. Manage-ment Science 48, 138–153.