how do technology clusters emerge and become sustainable ...keck graduate institute of applied life...

Research Policy 36 (2007) 438–455

How do technology clusters emerge and become sustainable?Social network formation and inter-firm mobility

within the San Diego biotechnology cluster

Steven Casper !

Keck Graduate Institute of Applied Life Sciences, 535 Watson Drive, Claremont, CA 91711, United States

Available online 9 April 2007

Abstract

Regional technology clusters are an important source of economic development, yet in biotechnology few successful clustersexist. Previous research links successful clusters to heightened innovation capacity achieved through the existence of social tieslinking individuals across companies. Less understood are the mechanisms by which such networks emerge. The article uses socialnetwork analysis to examine the emergence of social networks linking senior managers employed in biotechnology firms in SanDiego, California. Labor mobility within the region has forged a large network linking managers and firms, while ties linkingmanagers of an early company, Hybritech, formed a network backbone anchoring growth in the region.© 2007 Elsevier B.V. All rights reserved.

Keywords: Biotechnology; Social network analysis; Careers; Technology clusters; Innovation

1. Introduction

Clusters of high-technology firms have become animportant source of economic development across theadvanced industrial economies, and a central focus oftechnology policy. Studies of technology clusters haveyielded persuasive accounts of how successful clusterswork, when they work. A central explanation focuses onsocial networks and labor market mobility within tech-nology clusters. In an influential study of the success ofSilicon Valley, Saxenian (1994) argues that a culture ofdecentralized social ties linking scientists and engineersacross local companies helps diffuse innovation, whilefrom the point of view of skilled individuals, manage thecareer risks of working in failure-prone companies. Thiscreates a strong justification for approaches stressing

! Tel.: +1 909 607 0132; fax: +1 909 607 0119.E-mail address: steven [email protected].

the social embeddedness of economic action, as com-panies embedded within regions with a decentralizedculture of high mobility and knowledge diffusion willhave a “regional advantage” over companies that are not(Saxenian, 1994; Herrigel, 1993; Sabel, 1992; Storper,1997).

While providing a persuasive explanation for the suc-cess of some clusters over others, a problem with theinter-firm mobility hypothesis is that it only makes senseonce a large agglomeration of companies coupled withnorms and social networks facilitating mobility exist.Left unexplored are the mechanisms by which regionsmove from a starting position in which neither theagglomeration of companies or social networks under-pinning mobility exist to one in which they do. Howdo regional technology clusters, and the social networksunderpinning them, emerge?

This article examines the emergence and sustain-ability of social networks linking senior managersof biotechnology firms in San Diego, California. It

0048-7333/$ – see front matter © 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.respol.2007.02.018

mailto:[email protected]

dx.doi.org/10.1016/j.respol.2007.02.018

S. Casper / Research Policy 36 (2007) 438–455 439

investigates how a region went from having virtually nopresence in commercial biotechnology to developingone of the world’s most vibrant biotechnology clusters.Drawing on career histories of over 900 managersemployed in San Diego biotechnology firms, the paperuses social network analysis tools to trace the develop-ment career affiliation networks linking senior managerswho worked together at one or more San Diego biotech-nology firms during their careers. Over a 27-year period,a tiny network of less than a dozen people working in avery small number of firms incrementally expanded intoa large, efficient, and sustainable social network linkingseveral hundred senior managers working in an agglom-eration of over 120 high-risk biotechnology companies.

The study is motivated by two analytic goals. A firstis to address the debate as to how technology clustersbecome sustainable. The career history data allows thegeneration of career affiliation networks linking seniormanagers on a year by year basis over the history ofthe cluster. This facilitates a year by year analysis of thecluster’s emergence. Did the cluster emerge and becomesustainable slowly, or quickly? The paper draws on anumber of simple measures developed by social networktheorists to explore sustainability. These include year byyear measures of the usefulness of the network in termsof its size and connectivity, its efficiency in develop-ing ties linking senior managers to firms in the region,and its durability or robustness given company failures.This analysis provides strong support for the notion thatsustainable clusters are linked to the existence of densesocial networks across key personnel supporting careermobility. However, the San Diego case indicates that sus-tainable social networks emerge relatively slowly; it tookabout 15 years for this cluster to become sustainable, atleast in terms of social network organization.

The second objective is to examine the mechanismsby which social networks linked to career mobilityemerge. How does a regional economy develop a socialstructure favoring job mobility? An important literature,which has roots in studies of Silicon Valley’s engineeringindustries, emphasizes the role of companies in promot-ing job mobility. The company focused explanation isof central importance to the development of San Diego’sbiotechnology cluster. Managers linked to one phenome-nally successful early start-up company called Hybritechplayed a dominant role in seeding the development oftwo generations of successor companies in San Diego,and while doing so helped form the backbone of socialnetworks linking managers during the early history of thecluster. Without this cohort of experienced biotechnol-ogy executives willing to embrace and commercializelocal university technologies (see Higgins, 2005), it is

likely that San Diego’s biotechnology cluster would nothave developed as successfully as it did.

2. Labor market mobility and the sustainabilityof high-technology clusters

The social structure and career mobility explanationdraws on a large literature linking environmental fac-tors to the innovative performance of firms within newtechnology industries such as semiconductors, software,or biotechnology. Many of the core arguments underly-ing approach were developed by Saxenian (1994) in hercomparison of the Silicon Valley and Route 128/Bostonregional semiconductor industries. Saxenian argues thatSilicon Valley’s success is linked to the developmentof a social structure encouraging the development ofnumerous informal links across the region’s scien-tists, engineers, and managers. These links raised theinnovative capacity of Silicon Valley’s firms through dif-fusing technological and market intelligence. Drawingon Granovetter’s (1973) research on referral networkswithin labor markets, Saxenian argues that social net-works within Silicon Valley increased labor mobilityacross firms and by doing so created an additional mech-anism of knowledge diffusion. The declining fortunesof Route 128’s computer and semiconductor industry,on the other hand, is influenced by autarkic practicesof long-term employment within its companies that hin-dered the creation of flexible labor markets, coupled withvery limited informal sharing across firms through socialnetworks. Almeida and Kogut (1999) followed-up Sax-enian’s research with a quantitative study using patentdata from 12 US semiconductor clusters. In their study,patent data were used to gather information on levelsof inter-firm mobility of inventors within each clusterand as an indicator of aggregate innovativeness. Theirstudy supported Saxenian’s argument, showing that onlySilicon Valley had both high levels of job mobility andmarkedly higher levels of patenting.

The strength of the labor market mobility researchstream is its ability to connect career mobility to height-ened innovative capacity of start-up firms specializingwithin new technology industries, while also estab-lishing a mechanism by which presumably risk-averseskilled employees commit to failure prone jobs. Moststart-up companies within new technology industriessuch as biotechnology, software, or semiconductorsbegin life as relatively simple project-based firmsdeveloped to recruit and incentivize teams of talented sci-entists and engineers to work on well defined technologydevelopment goals (Whitley, 2004; Baron and Hannan,2002). Their success is in part determined by their ability

440 S. Casper / Research Policy 36 (2007) 438–455

to entice skilled managers and employees to leave lucra-tive and often ‘safe’ jobs in established companies oruniversity labs to join a new venture. Common patternsof financing and organizing start-ups enhance the attrac-tiveness of working in a start-up. A founding team ofmanagers and employees usually share ownership of theteam with venture capitalists and other investors. Directownership stakes provide extremely high performanceincentives for employees of early stage start-up com-panies. Should the company succeed and “go public”through a stock offering or be acquired at a favorablevaluation early employees can earn vast payouts. More-over, though demanding, careers at start-ups are oftendescribed as exhilarating due to their fast pace and broadrange of activities.

The potential benefits of working within a start-upare countered by a high likelihood that employmenttenures within start-ups will be short due to dismissalsor outright failure. Most start-ups fail to reach a lucra-tive exit, be it an initial public offering or acquisition bya larger firm at a favorable valuation. Start-up compa-nies are usually funded by venture capitalists through aseries of financing rounds as the firm passes through aseries of technical and market milestones developed byits board (Kenney and Florida, 1998). Venture capital-ists often decide to halt investments in companies thatfail to meet key milestones. Dismissals of top manage-ment are often a common response by VC-led boards tofirms that have failed to meet development milestones.Managers and employees within start-ups also find them-selves at risk of dismissal due to strategic decisionsto change the competency structure of the firm. More-over, as a condition for funding many venture capitalistsinsists on the right to replace early technical foundersof companies with professional managers as a companydevelops.

From the point of view of individuals, there is astrong rationale for choosing to work only within start-up companies embedded within an agglomeration inwhich social ties promoting mobility are strong. Doingso can dramatically lower the career risk for foundingteams and R&D staffs by creating numerous alternateemployment options should a given venture fail, undergomanagerial shakeups at the behest of investors, or needto change its competency structure due to technologicalvolatility. According to this logic, successful technologyclusters develop what Bahrami and Evans (1999) call“recycling mechanisms” to help preserve the value ofassets committed to failed enterprises. To quote Saxe-nian, “Moving from job to job in Silicon Valley was notas disruptive of personal, social, or professional ties asit could be elsewhere” (Saxenian, 1994: 35). This helps

explain why successful and presumably risk adverse sci-entists and managers would give up prestigious careers inestablished companies or university labs to work withinlucrative but highly risky start-ups: within successfulclusters the embeddedness of individuals within socialnetworks makes it safe to do so.

In addition to creating a regional recruiting advan-tage, social structures facilitating mobility may providecompetitive advantage for firms operating in market seg-ments in which technological cumulativeness is low.During the early phase of new industries, technolog-ical paradigms are still being established (Utterback,1996). Companies compete to validate technologicalapproaches and secure property rights over the approachor, at times, develop a dominant design (Teece, 1986). Togive an example from biotechnology, Penan (1996) con-ducted a bibliometric survey of approaches being used todevelop therapies for Alzheimer’s disease and found over20 distinct technological approaches being pursued bycompeting teams of biotechnology firms, basic researchlabs, and large pharmaceutical companies. Within highlyuncertain technological environments companies mayneed to routinely adjust their portfolio of approaches.If embedded within a regional economy with high labormarket mobility, firms can more easily use “hire and fire”practices to alter research and development strategies.This returns to Saxenian’s core argument: in additionto providing a recruiting advantage, inter-firm mobilityhelps create informal social ties across a region’s firms.Informal ties may provide market or technological intel-ligence, allowing companies to make superior decisionsas to which technologies to adopt or, at times discontinue.Firms may be able to react to market developments fasterthan competitors.

While focusing attention on explaining successfulcases, and especially Silicon Valley, the career mobil-ity approach also contains an explanation of why mostregional economies fail. Most clusters, even if they reachsufficient size, do not develop the social networks ornorms of high labor market flexibility needed to createthe ‘regional advantage’ associated with Silicon Valley.Lacking a safety net provided by career affiliation net-works, leaving a safe job to work within a failure pronestart-up is truly a risky proposition, one that most ratio-nal and risk-averse individuals will resist. Individualsdo however have an incentive to relocate themselves,abandoning disadvantaged regions to join firms locatedin a more successful cluster. Agglomeration or scaleeffects may develop from such human capital migrationand, following the Marshall tradition (see, e.g. Krugman,1991), provide “first mover” type advantages to success-ful regional clusters. However, the labor market mobility


theory provides an important complement to the scaleeffect explanation, as high labor market mobility withina region, and a social structure of dense inter-firm tiessupporting it, may be a necessary condition for the regionto sustain high levels of innovation by a sizeable numberof companies.

The mobility explanation is in core aspects similar toa game theoretic equilibrium model. It links the rationalbehavior of talented individuals to different labor mar-ket equilibriums which are then linked to the generationof different innovative capacities for companies. Inter-firm mobility helps diffuse technology across companiesand, from the point of view of skilled personnel, gener-ates the formation of social networks that can be used tooffset the career risk of leaving a ‘safe’ job to work in ahigh-risk start-up. However, most localities have neitherthe agglomeration of firms and people nor the social tiesneeded to sustain high-risk firms. In such regions keypersonnel are unable to reduce the risk of participatingin a high-risk firm. This leads to a second, much morecommon equilibrium, that of failed cluster development.A structure promoting extensive career mobility does notexist. From the career perspective, leaving a safe job inan established company or university to join a start-uptruly is a high-risk proposition that most will choosenot to do. It becomes easier to understand why mostlocalities fail to develop successful technology clusters:talented individuals might populate a region, but theyface a collective action problem. They lack the appro-priate social ties needed to reduce the risk of workingwithin a high-risk venture.

The failure equilibrium is the most common occur-rence. Much comparative institutional research haspinpointed the United States as a country with an idealinstitutional infrastructure to support clusters of technol-ogy firms (Hall and Soskice, 2001; Casper and Whitley,2004). Yet even in the United States most regions, eventhose housing leading universities, are not home to suc-cessful technology clusters. Almeida and Kogut (1999)semiconductor study found only 1 cluster out of a sam-ple of 12, Silicon Valley, that had both exceptional levelsof innovation and career mobility. Systematic compar-ative studies are lacking in the biotechnology area, butindustry surveys suggest that there are only three largeclusters that analysts universally agree are performingwell: those in San Francisco, San Diego, and Boston(see DeVol et al., 2005). In Europe, there are at best twowell-performing biotechnology clusters, located in Cam-bridge, UK, and Munich, Germany (Casper and Murray,2004); both are much smaller than the successful Amer-ican clusters in terms of number of independent R&Dintensive companies.

The emergence of successful biotechnology clustersis a problem of social coordination that may be diffi-cult to resolve. The decentralized social infrastructurecharacterizing successful technology clusters is in someways analogous to a collective or public good: its bene-fits accrue to most if not all individuals and companieswithin the regional economy (though perhaps dispro-portionately depending on position within the network,see Owen-Smith and Powell, 2004). However, unliketraditional public goods (roadways, the air), social infras-tructures supporting technology clusters may be difficultto orchestrate or maintain in a systematic fashion. Rather,it is an emergent property, a product of the collectivebehavior of individuals and firms within a regional econ-omy. As such, it is unlikely that individuals or firms cansingle-handedly develop the necessary mesh of socialties needed to sustain a highly innovative cluster. A rel-atively large number of individuals must develop andmobilize social ties in order to develop a density ofties sufficient to generate useful networks. What are themechanisms by which regions move from a starting posi-tion in which neither the agglomeration of companiesor social networks underpinning mobility exist to onein which they do? If their development has collectiveaction problems, how do social ties develop into usefuland sustainable networks?

A plausible scenario is that social networks developslowly or incrementally. Early entrants to a cluster mightbe particularly risk acceptant individuals. Over time, theycould plausibly seed a nucleus of companies and estab-lish social ties between them. As these ties expand theybecome a so-called backbone to a social infrastructurethat other entrants, both individuals and new companies,can draw upon. It is possible that, after reaching a cer-tain size and rate of mobility, a tipping point could bereached whereby the cluster becomes sustainable andregional innovation effects begin to accrue. Once sustain-able, agglomeration effects might become established asjobs within the cluster become attractive to more riskadverse individuals.

A difficulty with this explanation is that social net-work effects may only be pronounced once a largenumber of individuals participate in the network; benefitsmay only develop as social networks become relativelylarge and efficiently organized. If so, early pioneerswithin a cluster may be particularly failure prone. Earlyfailures are likely to be much more costly, in terms oftheir effects on network growth, than later failures. If so,nascent technology clusters might never reach the crit-ical mass to become sustainable. This could lead to theoutright collapse of a cluster or the decision by individ-uals and companies to abandon “radically innovative”


strategies in order to pursue safer and more incrementalinnovation strategies. Companies would then adopt moreinternally focused research and development strategiesthat draw less on external ties and accommodate saferjobs with relatively long employment expectations.

The rarity of well-performing clusters suggests thatthe emergence of appropriate social infrastructures is adifficult problem, perhaps one rarely solved. A key issuethen becomes one of emergence. How do individualswithin a region develop the social infrastructure neededto sustain agglomeration of high-risk firms? This issue,of moving essentially from “nothing” to the generation ofa decentralized social infrastructure capable of diffusinginnovation and facilitating career management, has beenlargely ignored in studies of high-technology clusters. Itmotivates the present study.

3. Research design

The study traces the emergence and sustainabilityof social networks linking senior managers employedover the history of a successful biotechnology clus-ter located in San Diego, California. It examines yearby year employment histories of several hundred topmanagers from the cluster’s early formation, starting in1978, to 2005. This data gathering strategy allows usto trace the growth of San Diego biotechnology firmsand an analysis of both agglomeration patterns (peoplemoving to San Diego) and mobility across firms. More-over, career histories allow us to construct and studya relatively complete set of social ties formed betweenmanagers through joint employment in the same biotech-nology firms.

In terms of social networks, the study examinesthe emergence of career affiliation networks formedbetween senior managers on the basis of ties betweenindividuals that are formed through joint employment atthe same organization (see Casper and Murray, 2004).A focus on career affiliation ties is warranted dueto the study’s emphasis on mobility patterns. More-over, we can plausibly assume that through servingon senior management teams, individuals form durablesocial ties with one another and have obtained rela-tively full information about one another. These tiesshould be particularly useful when used for job refer-rals.

The study explores the emergence of career affiliationnetworks linking senior managers. Within the biotech-nology industry, senior management usually includesa company’s chief executive, chief scientific officer,chief finance officer, and a number of vice presi-dents and senior personnel involved in research and

development, business development, and, within somecompanies, human resources and legal affairs. Seniormanagers must define a firm’s strategy and mobilizethe necessary resources to implement it. Recruitingtalented senior management is strongly linked to the suc-cess of biotechnology companies (Higgins and Gulati,2003). In this respect, an emphasis on top managementagain links directly to the emphasis on career mobil-ity.

The biotechnology industry was chosen for studydue to its status as a high technology industry con-taining high technological volatility (see Henderson etal., 1999). Large, multi-billion dollar markets exist fordrug and diagnostics products meeting unmet medicalneeds, and intellectual regimes surrounding new treat-ments are strong. While a few successful firms havegenerated enormous profits, failure rates are high in thebiotechnology industry. Most companies will fail or becheaply acquired and integrated into competitors or largepharmaceutical companies. An Internet database locatedwithin Biotech Career Center (2005) lists several hun-dred failed companies. Finally, a key theme emergingfrom research on the biotechnology industry is the decen-tralization of knowledge within the industry and the needfor companies to develop and tap into a variety of exter-nal networks if they are to succeed (Powell, 1996; Powellet al., 1996; Shan et al., 1994).

The study’s research design selects a successful case,San Diego. Given that social networks linked to mobilityhave been identified in several other cluster studies, thisresearch design is biased towards a finding that strongcareer affiliation networks do exist (as is indeed the out-come). However, the goal of the study is to rigorouslyexplore the emergence of social networks within hightechnology industry, an area that has only very recentlybegun to be investigated (see, e.g. Uzzi and Spiro, 2005;Fleming et al., 2006) and were unknown with regardsto the San Diego biotechnology cluster. Moreover, giventhat Silicon Valley is the only success case identifiedin the literature linking social networks and mobility tocluster performance, an exploration of San Diego willfacilitate broader comparative analysis. If social networkdevelopment within clusters is rare, carefully document-ing mechanisms of emergence across the few successfulclusters is an important step in designing comparativeresearch capable of yielding generalizations applicableacross clusters.

San Diego is an excellent laboratory to study clus-ter development. The region went from having virtuallyno presence in commercial biotechnology at the start ofthe 1980s to developing one of the world’s most vibrantbiotechnology clusters by the late 1990s (DeVol et al.,


2005). While San Diego has recently developed a clus-ter of wireless telecom companies to complement itsbiotechnology presence (see Simard, 2004), the regiondid not have a presence in high technology industry dur-ing the late 1970s, and was primarily known for its largenaval base and defense contractors. This suggests thatits biotechnology was the first high-technology indus-try to develop in the region, with the implication thatearly companies could not draw on previously estab-lished local venture capitalists, labor market pools, orother resources. The San Diego biotechnology industrywas founded in 1978 and gained critical mass in the late1980s. While this was early in the history of biotech-nology (Genentech was founded in 1976), San Diegocompanies did benefit from the demonstration effect ofGenentech’s success as well as the ability of some earlycompanies to establish links with San Francisco venturecapitalists.

San Diego has long been home to several worldclass biomedical research institutes, such as the ScrippsResearch Institute and the Salk Institute, while the Uni-versity of California, San Diego (UCSD) has developeda medical school and strong departments in chemistry,biology, and other fields with links to biotechnology.Through the early 2000s, these institutes were col-lectively receiving over $500 million federal fundingfor biomedical research from the US National ScienceFoundation and National Institute of Health (BrookingsInstitute, 2002). Considerable research has demonstratedthat the performance of biotechnology companies isimproved by the existence of ties to leading academicscientists (Zucker et al., 1998; Murray, 2004). The impor-tance of university-firm ties has led to the establishmentof most biotechnology clusters in close proximity toleading universities. Without a supply of world class sci-ence, it is unlikely that a biotechnology cluster wouldexist in San Diego.

San Diego is also one of very few successful biotech-nology clusters within the United States (BrookingsInstitute, 2002). Although San Diego has very favor-able conditions for development, there exist many moreworld class universities than biotechnology clusters, andseveral US regions are home to world-class biomedi-cal research bases, such as Los Angeles, Chicago, orNew York, but have not developed sizeable biotechnol-ogy clusters. Thus, while San Diego might be considereda “most favorable” place to develop a biotechnologycluster, we should keep in mind that many other favor-able locations within the United States have failed tomatch the region’s success. How was this area able togrow a viable biotechnology industry, while others havenot?

4. Methods: data gathering and networkconstruction

The research process for this study had three steps:locating firms, gathering career histories, and then usingsocial network analysis tools and related descriptivestatistics to gather and analyze results.

4.1. Locating firms

High failure rates within the biotechnology industrymake the identification of firms over the history of aregional cluster difficult. A two-step approach was usedto identify firms. Industry directories were used to locatebiotechnology companies active in the San Diego regionduring the years 2004 and 2005 (AlexanderX, 2004,2005). The San Diego region was defined as San Diego,La Jolla, Carlsbad, and other communities located withinSan Diego, but excluded San Juan Capistrano and othercoastal communities located approximately 25 milesnorth within Orange County. Industry guides list hun-dreds of companies, including consultancies, contractresearch organizations, equipment manufacturers, andsubsidies of companies with central headquarters locatedelsewhere. Only research intensive and independentbiotechnology companies were included in the database.As a criteria to aid company selection, only companiesthat had published at least one scientific article wereincluded (Casper and Murray, 2003). Searches on theWeb of Science and Pubmed on-line databases wereused to search for company publications. This searchstrategy yielded 125 active companies for 2004 and 121companies for 2005.

Due to the high failure rate of biotechnology compa-nies, it was important to locate as many companies thatfailed or lost independence as possible. A primary sourceof failed companies were career histories of senior man-agers working in active firms, which often listed jobs infailed local biotechnology companies that were addedto our database. While a useful strategy of identifyingfirms, the exclusive use of career histories to locate failedfirms could create biases in the social network data. Aswill be discussed below, social network ties will be cre-ated through mobility across firms. If the only sourceof locating failed companies was career histories, thiscould create a potential bias in the network data towardsincreasing network connectivity, as chains of ties willexist between all failed firms and all on-going firms.To minimize this potential bias, it is important to locateas many failed firms as possible, ideally from sourcesother than career histories. Two useful lists of San Diegobiotechnology companies active at earlier points in time


were used to locate additional firms. The first source wasa newspaper article from 1993 that contained a list of 106San Diego biotechnology companies active during thatyear, many of which had failed by 2005. Second, a reporton the commercialization of academic science within theSan Diego region (Lee and Walshok, 2000) listed over120 companies spun-out of UCSD, the Scripps ResearchInstitute, and the Salk Institute prior to 2002, a majorityof which were biotechnology related. While informa-tion on many of these companies could not be located,presumably because they failed very quickly due to aninability to secure finance or other reasons, informa-tion was found for a small number of companies locatedthrough this list.

In sum, 193 San Diego biotechnology companieswere identified for inclusion in the study, 70 of whichhad failed. This implies a failure rate of 36%. This figureis probably lower than the actual failure rate within SanDiego biotechnology, for two reasons. First, many firmsin the study are relatively young, having been foundedin the post-2000 period, and may not have had suffi-cient opportunity to fail. Second, only firms for whichinformation on senior managers could be found wereincluded in the database. This creates a biased towardsthe inclusion of firms that have secured significant ven-ture capital financing needed to recruit sophisticatedsenior management teams The inability to find infor-mation on managers for many of the companies listed inthe study of university spin-outs suggests that numeroussmall companies were founded and quickly failed due toan inability to secure capital.

4.2. Gathering career histories

Several sources were used to gather career histories.The most important are SEC company filings. Filings areavailable from 1994 onwards for most firms in the on-lineEdgar database. Annual reports (10k forms) and prospec-tus documents filed when selling shares usually providecareer histories for senior managers, whom are requiredto include information on the past 5 years employ-ment but, in practice, almost always include completecareer histories, including graduate and postgraduatetraining.

The availability of SEC company documents overseveral years allows direct tracing of top managementpersonnel changes within companies. As a result, thedatabase contains both founding senior managementteams and subsequent senior managers who were hired asfirms expanded or went through periods of managementturnover. Additional sources of information used werecompany web pages, which usually contain career his-

tories for current top managers as well as press releasesdocumenting personnel changes. Finally, in some cases,general Google searches were used to search for missingdata.

Career histories were constructed for 923 senior man-agers of San Diego biotechnology firms employed in atleast one firm listed in our database between 1978 and2005. This includes all types of senior management posi-tions, both those in scientific and general managementpositions. Sixty two percent (573) had a graduate sciencedegree (PhD) or, in a few cases, were medical doctors.The remaining individuals had no graduate science train-ing and were presumably general managers. Only 61 hadreceived PhDs or held postdoctoral or professor positionsfrom UCSD, Scripps or the Salk Institute. Well over 90%of the scientists (512) had their last place of scientificemployment (graduate training, postdoctoral student, orin rare cases, professor) outside of San Diego, and moregenerally most non-scientists also moved to their firstSan Diego biotechnology job from outside the region.This suggests that ties forged through joint employmentat San Diego biotechnology firms will be an importantresource for both scientists and non-scientists.

While this search strategy yielded career histories fora large number of senior managers within the region,there are important sources of missing data that couldbias the results. A first issue is incomplete data on compa-nies obtained through SEC searches. Companies are onlyrequired to make public SEC filings once they reach cer-tain financial and ownership distribution thresholds, andhence do not make filings during their early years of oper-ation. For companies founded during the period of 2000onwards that had not filed SEC documents, extensiveinformation on senior management teams was availablethrough company websites and press releases. How-ever, for companies founded less recently data availablethrough Internet searches was less available, creatinga source of missing data on senior managers that leftcompanies during their early history. If such managersobtained a job within another San Diego biotechnologyfirm, however, missing data would be recovered throughcareer histories of that individual identified through sub-sequent employment in the region.

A second problem with the SEC filing data is thatfiles are only easily accessible on-line for the post-1994period. This leads to the possibility of missing informa-tion on the composition of top management teams for allcompanies active prior to 1994. This means that the mainsource of data for the pre-1994 data are the career his-tory information of managers working within the clusterduring the post-1994 period. Managers that worked ina San Diego biotechnology firm during the 1980s early


1990s and left the region prior to 1994 are likely to beexcluded from the database. How frequent are exits fromthe region? From the 923 individuals that were located,106 left the region for jobs outside of San Diego. Thissuggests that while exits do occur, close to 90% of indi-viduals have not exited the region during the time periodsampled. If the exit rate is assumed to be constant overthe cluster’s history, this would imply that only a modestnumber of individuals exited prior to 1994 and have beenexcluded from the database.

The issue of missing data during the early history ofthe cluster is significant, as a major goal of this study istrace mechanisms of social network emergence duringthose early years. Missing data could result in impor-tant ties linking senior managers from early companiesbeing excluded from the data, suggesting that less con-nectivity exists within the network than is actually thecase. An important finding surrounding the emergenceof the network, discussed at length in Section 5, is that agroup of senior managers sharing joint affiliations withan early company named Hybritech were responsible forcreating connectivity within the network during the keyperiod between 1986 and the early 1990s. It is unlikelythat this result has been impacted by missing data. Mul-tiple sources of evidence for this result will be presentedfrom the social network data and will also be supportedwith evidence from secondary sources surrounding thegrowth of San Diego biotechnology.

A final problem concerns the consistency of data.Firms have different definitions of what positions con-stitute senior management. A few firms only includethe chief executive officer, chief financial officer, andchief science officer. Most list vice presidents, and a fewlist directors or “key personnel.” All individuals listedas senior management in SEC filings or web-ages wereincluded in the project database. The result is that somefirms have more senior managers contained in the net-work in any given year than others or similar size. Thiscould impact the calculation of some social networkstatistics in which aggregate count measures of ties (ordegrees) are important. For example, some network cen-trality measures use counts of ties as an indicator; thesemeasures could be biased. Certain individuals or com-panies may be assigned a more central position in thenetwork due to their association with a company whichprovides exhaustive information on the composition oftheir senior management team. While we are unableto control for this effect, we do not believe it stronglyimpacts more general measures of overall network con-nectivity or density, which will be used to examine thegeneral level and sustainability of social ties within theregion.

4.3. Constructing career affiliation networks

All career histories were entered into a database list-ing each senior manager and the name and years workedat each organization during their career. Ties betweenindividuals are created through joint employment withinthe same organization. Under this rule of tie formation,ties linking individuals across organizations are onlyformed through mobility. Upon changing jobs a man-ager maintains ties with members of the old organization,while creating new ties at the new place of employment.Networks were created for each year between 1978 and2005. The yearly network data allows detailed processtracing as to the formation of the network.

An important issue surrounding the construction ofnetworks is how long ties should be assumed to lastonce an individual leaves an organization. Once anindividual moves jobs there is a probability that tieswill decay, or weaken over time as people lose con-tact with one another. From a theoretical perspective,if ties are assumed to last indefinitely, dense social net-works become much easier to produce and the problemof sustaining the network drops away. By creating amodel where ties decay, new ties must be continuouslygenerated in order for a network to become sustain-able. As ties linking organizations are only producedthrough mobility, this assumption generates a systemin which relatively high levels of labor market mobilitywill be needed to maintain useful networks. Followingan approach implemented in similar network emergencestudies by Uzzi and Spiro (2005) and Fleming et al.(2006), the study assumes that ties linking an individ-ual to others within an organization cease to exist 5years after an individual changes jobs, unless renewed bysubsequent joint employment at the same organization.The 5-year decay rule is a strict assumption, allowing anempirical analysis of the emergence of social networkswhen it is hard to generate and sustain connectivity.1

1 As one of the anonymous reviewers to the article pointed out, indi-viduals commonly maintain ties long after joint employment throughmeeting at industry conferences or through informal contact. Thismight warrant using a longer, more generous decay model. To testthe robustness of the 5-year decay rule, models using decay rules of 3and 10 years were also tested. Reporting on the final, 2004 networksand referring to results reported later in the paper in Table 1, the longerdecay model increases both connectivity and density within the result-ing network, but only marginally. Using a 10-year decay rule increasesconnectivity, defined as the number of individual’s linked together inthe network’s giant or main component from 95% using the 5 year ruleto 99%, while the density of ties increases from 2 to 2.8%. Adoptinga stricter decay model of 3 years, however, has a more pronouncedimpact on connectivity within the network, which decreases to 51%,


Table 1San Diego biotechnology company statistics, 1978–2005

Year Total firms Growth rate (%) Entrants Exits Public firms

1978 1 100 1 0 01979 1 0 0 0 01980 1 0 0 0 01981 2 100 1 0 01982 5 150 3 0 01983 7 40 2 0 01984 11 57 4 0 01985 11 0 1 1 01986 15 36 4 0 01987 22 47 7 0 21988 30 36 9 1 01989 38 27 8 1 01990 38 0 3 3 51991 46 21 8 0 71992 55 20 10 1 91993 61 11 8 1 151994 64 5 7 5 171995 69 8 6 1 191996 77 12 10 3 241997 95 23 16 6 301998 103 8 20 6 361999 108 5 11 3 392000 124 15 25 10 502001 127 2 15 9 522002 131 3 7 4 562003 127 "3 5 7 572004 125 "2 2 3 632005 121 "3 0 5 64

5. Results

Findings from the study are reported in two parts.First, a variety of indicators are used to investigate thesustainability of career affiliation networks among SanDiego senior managers. Several indicators are assessed,including the general connectivity and density of socialties within the network, a measure of the usefulnessof these ties for individuals in securing referrals toother firms, and an examination of the robustness of thenetwork to failure. The second section examines mech-anisms of emergence. How was connectivity within thenetwork established?

5.1. Network properties and sustainability

Table 1 displays descriptive statistics surroundingthe growth of companies in the San Diego region,

while density decreased to 1.3%. Overall, the 5-year decay rule appearsto provide robust results while adopting a relatively strict assumptionabout tie maintenance needed to provide a strong test of the socialnetworks and mobility explanation for cluster development.

while Table 2 displays information on the organizationof career affiliation networks linking senior managers.Turning first to companies, the size of the cluster grewincrementally from an initial start-up launched in 1978,Hybritech, to a relatively large agglomeration of over125 companies from the year 2000 onwards. The patternof growth within the region is incremental. On average,the growth of rate for the number of firms in the regionis about 15% a year. A significant percentage of SanDiego biotechnology firms were able to attain fundingfrom stock markets, an important indicator of success.While few firms completed initial public offerings dur-ing the 1980s, during the 1990s and early 200s over 60companies completed IPOs.

Turning to Table 2, the region also experienceddramatic growth in the number of senior managersemployed. Starting in 1978 with the two initial foundersof the region’s first company, Hybritech, the number ofindividuals grew incrementally to over 100 by 1988, 500by 1998, and 867 by 2005. The growth rate of senior man-agers within the area was also relatively incremental, atabout 20% per year. The higher growth rate for indi-viduals suggests that, in addition to new start-ups, firmswere consistently expanding their senior managementteams. Note also that the average number of individ-uals working per company increases over time. Thisis in part due to the ability of over 60 companies tocomplete initial public offerings during the mid 1990s;cash infusions from going public are usually used togrow companies. However, as mentioned in the method-ology discussion, it is also possible that there is lessmissing data for the post-1994 network history, whichwould be reflected in larger numbers of individuals percompany.

The impressive growth rate of both firms and seniormanagers attests to the success of the San Diego biotech-nology industry. But did this growth occur in conjunctionwith the development of social networks linking thesenior managers? An important measure related to thedevelopment of sustainable networks is connectivity.Connectivity relates to the usefulness of the networkin terms of the number of individuals linked together.Most networks initially consist of several fragmentedclusters of individuals with ties to one another, callednetwork components. As ties continue to form, theseclusters begin to coalesce, eventually forming one giantor main component. Table 2 displays the size of themain component linking individuals, as well as the per-cent of people in the main component and the totalnumber of components in the network, all on a yearlybasis. Three distinct periods of development occurred.During the early 1980s, the network was fragmented,


Table 2Descriptive statistics, San Diego career affiliation networks, 1978–2005

Year Totalindividuals

Avg. peopleper firm

Size of maincomponent

Percent in maincomponent

Average pathlength

Networkdensity

1978 2 2.0 2 1 1.0 1.001979 4 4.0 3 75% 1.0 0.501980 7 7.0 4 57% 1.0 0.331981 9 4.5 4 44% 1.0 0.191982 19 3.8 9 47% 1.0 0.251983 27 3.9 15 56% 1.0 0.331984 39 3.5 17 44% 1.1 0.211985 47 4.3 24 51% 1.2 0.251986 59 3.9 35 59% 1.8 0.191987 78 3.5 57 73% 2.8 0.131988 107 3.6 81 76% 3.1 0.101989 132 3.5 103 78% 3.3 0.071990 165 4.3 135 82% 3.9 0.061991 188 4.1 151 80% 3.7 0.051992 232 4.2 204 88% 3.8 0.051993 273 4.5 243 89% 4.1 0.041994 317 5.0 290 92% 4.0 0.041995 342 5.0 300 88% 3.6 0.041996 397 5.2 347 87% 3.5 0.041997 452 4.8 409 91% 3.6 0.031998 503 4.9 466 93% 3.6 0.031999 547 5.1 498 91% 3.6 0.032000 624 5.0 559 90% 3.8 0.022001 702 5.5 648 92% 3.8 0.022002 771 5.9 719 93% 3.8 0.022003 817 6.4 760 93% 3.8 0.022004 852 6.8 806 95% 3.9 0.022005 867 7.2 824 95% 4.2 0.02

with no more than roughly one-half of a relatively smallnetwork of senior managers linked into the main com-ponent. During the late 1980s and early 1990s, thenetwork begins to gain coherence, as over two-thirdsof the members of a larger network consisting of morethan 100 people are members of the main component.From the post-1993 period onwards, the network con-tinues to grow but has gained coherence; over 95% ofits members are connected to the main component by2005.

The very high level of connectivity points to theexistence of a potentially vibrant network, at least interms of the availability of most senior managers in SanDiego to contact peers through career affiliation ties. Itis important to note that connectivity within the networkremained high through the mid 1990s onwards despite adramatic growth in the overall size of the network; over700 individuals joined the network from 1990 onwards.One measure of the impact of movement of new individ-uals is a decline in network density over time. Networkdensity is defined as the ratio of actual ties within the net-work to possible ties. Table 2 indicates that declined from

a relatively high ratio of 20% or more during the earlyhistory of the network to less than 10% by 1988, to 5% by1992, and about 2% from the year 2000 onwards. Thesedata differs from findings by Powell et al. (2004) on thegeneral increase of network density in alliance activitywithin the biotechnology industry over time. The maindriver of declining density within the San Diego networkis the 5-year decay rule of ties. In essence, new individ-uals are joining the network at a faster rate than new tiesare forming through mobility.

A large pool of senior managers are connected intoSan Diego career affiliation networks, but how useful isthe network to its members? The declining density of net-work suggests that the efficiency of the network, in termsof the ability of individuals to use the network effectively,might be low. While this study does not measure the inno-vative capacity of companies, it can examine the degreeto which the network becomes useful to its membersfrom the point of view of career mobility. It does throughexamining whether structure of the network, as it evolvesover time, becomes efficient in developing ties betweensenior managers and other companies. The more ties cre-


ated between senior managers at different companies,the lower the career risk of working within a firm con-nected within this network becomes. If the labor marketmobility hypothesis is correct, the social network beganto link a large population of companies and lowered therisk of individuals joining high-risk technology start-upsand simultaneously increased the innovative capacity ofits companies.

How easy is it, on average, for members of the net-work to develop ties to other individuals and firms inany given year? Referrals are often developed by “work-ing the network” or asking acquaintances for contactsthat may know at target companies. A common statis-tic to measure indirect ties is average path length, or“degrees of separation,” between individuals in the net-work. Table 2 displays path length between individualslocated within the network main component on a yearlybasis. During the early history of the region, path lengthdata is low, at less than 2 ties, due to the small size ofthe network. However, from 1990 onwards, the aver-age path length averages at about 3.5–4. Given thatthe size of the network increased seven-fold duringthis period, the relatively constant path length statisticsis impressive. Despite the increasing size and sparse-ness of the network over time, inter-firm mobility wassufficient to generate a sufficient number of new ties link-ing individuals across companies to maintain networkefficiency.

While useful, the path length statistic only measurestie structures linking individuals, not between individu-als and firms. It is also biased by the clustering of directcontacts between people working within the same firm.A more measure of network usefulness would measurethe number of companies accessible to individuals at var-ious degrees of separation, i.e. how many direct contacts,how many contacts linked through one individual, twoindividuals, and so forth. If a significant number of com-panies were accessible to individuals at a low numberof intermediaries, this would be a strong indicator thatsocial ties have are linked to career mobility within thenetwork.

To develop an indicator of links between individualsand companies, path length information for each indi-vidual was recalculated on a yearly basis to examine theaverage number of contacts each senior manager hadat other companies at each degree of separation. Thiswas calculated by obtaining the minimum path lengthbetween each individual and each company on a year byyear basis, and then calculating the average number ofcompanies a manager could contact at each degree offreedom. All senior managers that were members of thenetwork main component for each year starting in 1985

Fig. 1. Average network distance of San Diego managers to companiesat given degrees of freedom, 1985–2004.

(the first year, the main component was of significantsize) up to 2004 were included.2

Fig. 1 displays the results of this analysis. The figureshows the number of companies available to the averageindividual within the network at each degree of freedomon a yearly basis. For example, throughout most of thehistory of the network, the average individual has directcontacts to three San Diego biotech companies, which bydefinition were forged through prior employment at thesefirms. As the network grew, however, progressively morecompanies become accessible to San Diego senior man-agers. It is open to debate as to how many people a givenmanager can easily use as intermediaries when “workingthe network” to gain access to a given firm. It is reason-able to suggest, however, that companies requiring twoor less intermediaries to contact are readily accessibleto most senior managers. In this respect, an interest-ing result from this figure is that the usefulness of thenetwork is strongly dependent upon size. For example,during the early years of the network, fewer than 10 com-panies were reachable to the average manager with twoor fewer contacts. By the late 1990s onwards, this figureincreases to about 30 companies for most years. Theseresults are strongly driven by the 5-year decay rule. Ifties were allowed to persist, then the percentage of com-panies within close contact would accumulate over time.However, the demonstration that relatively high numbersof companies are at close reach to managers during muchof its history under conservative modeling assumptions

2 Data from an earlier version of the network analysis including602 individuals and 142 San Diego firms are used in this calculation.As UCINET and other network analysis software does not supportthis type of calculation, the analysis was performed manually usingMicrosoft Excel and was very labor intensive. While I could not repli-cate the calculation for larger network used elsewhere in this paper, theresults generally confirm the usefulness of San Diego career affiliationnetworks in linking individuals to firms.


supports the argument that social ties are supportive ofcareer mobility under risky employment conditions thattypify the biotechnology industry.

A final indicator of sustainability is the network’srobustness to failure. What happens to connectivitywithin the network when biotechnology firms fail, asthey are prone to do? If managers exit the network afterfailure, presumably through retirement or an inabilityto find a job at another local biotechnology firm, thenconnectivity within the network will be lost after tiesdecay. Seventy San Diego biotechnology firms failedor were acquired during the 1978–2005 time period.Fifty-seven of these failures, or about 80%, occurred inthe post-1995 period, during which network connectiv-ity regularly achieved 90% or more. During this sameperiod, over 500 senior managers entered the network,and 117 new firms appeared. The robustness of connec-tivity within this career affiliation network, given theturbulence created by large numbers of company fail-ures and start-ups, should be considered strong evidenceof extensive job mobility within the region. The careerrisk of joining a biotechnology firm that ultimately losesits independence appears to be relatively low.

How robust is the network to more catastrophic fail-ure? Many of the companies that failed were smallfirms that failed to gain critical mass. Managers workingwithin these firms tend to have peripheral positions in thenetwork. What happens to network connectivity if largerfirms whose managers occupy central position in careeraffiliation networks fail? To examine this possibility, themost central firm was identified for every year using ameasured called the betweeness centrality, a commonlyused indicator of brokerage within social networks (seeWassermann and Faust, 1994). To investigate the impactof more catastrophic failure, this company and all tieslinking individuals to it were removed from the networkfor that year. Network statistics for the new network werecalculated and then compared to the size of the maincomponent of the original network.

Fig. 2 displays the percentage of individuals in theoriginal main component and the main component afterremoving the most central company for each year. Dur-ing the early 1980s, the removal of the most central firmhas a pronounced effect on network coherence. Seniormanagers from the early entrant Hybritech dominatedthe early formation of social networks within San Diegobiotechnology. Hybritech was the network’s most centralfirm between 1978 and 1989. During this period, removalof Hybritech firm from the network results in a loss ofbetween 30 and 40% of connectivity. Beginning in 1991,however, the robustness of the network to catastrophicfailure increases significantly. Hybritech was acquired

Fig. 2. Effect on size of network main component of removing mostcentral company.

in 1986 with the result that most of its managers left tojoin a wave of start-ups formed during the late 1980s. Aseries of companies founded by former Hybritech alums,Idec, Gensia, Ligand, and Amylin, occupied the positionof most central company between 1990 and 1997. By1991, a cohesive network linking these and other firmsin the region exists to the extent that the removal of thelargest firm has a relatively minor impact on networkconnectivity. Less than 10% of connectivity is lost in theearly 1990s, and from 1995 onwards the loss is between1 and 3%. This result complements the evidence on net-work usefulness, suggesting that from 1995 onwards SanDiego career affiliation networks were both extremelywell connected and durable. In general, the sustainabilityof San Diego career affiliation networks is high.

5.2. Mechanisms of emergence: the role ofHybritech managers in creating a network backbone

Given the organic pattern of growth in San Diego,a key issue to investigate is whether a mechanismdeveloped to overcome collective action problems sur-rounding the early growth of flexible labor markets.Through coupling network analysis with a closer analy-sis of history of the cluster’s key firms it is possible toexamine the mechanisms by which the network emerged.An important catalyst of a network’s development is theemergence of what network theorists call a “backbone”or group of initial ties that later entrants to a network canlatch on to, stimulating the growth of a cohesive network(see Powell et al., 2004). An interesting finding in SanDiego is that a network backbone did develop, and canbe attributed almost entirely to the career strategies of aset senior managers with ties to Hybritech, a prominentearly San Diego biotech company.

While a small number of biotechnology companiesexisted in San Diego by the early 1980s, only Hybri-tech was launched by a world class team of venturecapitalists, scientific founders, and general managers.


Hybritech,was founded in late 1978. The com-pany commercialized molecular diagnostics technologydeveloped at UCSD by Ivor Royston and Howard Birn-dorf. Hybritech received immediate credibility due toits ability to attract funding from the by the same teamof Silicon Valley venture capitalists at Kleiner, Perkins,and Byers that launched Genentech in San Franciscoa few years earlier. The Kleiner Perkins venture capi-talists were able to recruit a strong management team,lead by Howard Greene, one of several up and com-ing young general managers who left the medical devicefirm Baxter to accept leadership positions within thefirst generation of US biotechnology start-ups (Higgins,2005).

In 1986, Hybritech was acquired by the large phar-maceutical firm Lilly for $300 million, a princely sumgiven the time period and the fact that few biotechnol-ogy firms had at this point successfully commercialized aproduct (Crabtree, 2003). This acquisition had the imme-diate effect of transforming Hybritech’s top managementteam, all of whom owned shares in the company, intoextremely wealthy individuals. As part of the acquisition,the top management team was encouraged to remain,but Hybritech became a subsidiary of a large Indianabased pharmaceutical company with a relatively con-servative managerial ethos. Hybritech had developed afree-flowing, informal corporate culture typical of tech-nology start-ups. This created immediate clashes with

the Lilly managers. Tina Nova, one of the senior sci-entists at Hybritech, reflects that “It was like ‘AnimalHouse’ meets ‘The Waltons’ (Fikes, 1999). Lilly alsobegan a practice of rotating its managers into and outof the Hybritech facility at frequent intervals, making itdifficult for the original Hybritech managerial crew todevelop working relationships with the Lilly managers.Lilly was ultimately unable to integrate Hybritech’s man-agement and scientific team into its corporate culture,and in the years immediately following the acquisitionmost of the former Hybritech senior managers, includ-ing all managers located in this study’s database, left.Hybritech is now regularly referred to as a failed acqui-sition.

The cadre of former Hybritech managers are nowwidely credited within San Diego for “seeding” the SanDiego biotechnology industry. This group of managerscould serve as a reliable and trusted referral networkto one another. Their credibility as successful biotechentrepreneurs was also important in recruiting highlyskilled individuals to join San Diego start-ups to whichthe Hybritech managers were linked. These managershad the financial resources, managerial experience, anda reputation for developing one of the biotechnologyindustry’s early and rare success stories. A litany ofimportant San Diego companies were founded by formerHybritech managers, including Amylin, IDEC, Gen-sia, Gen-probe, Ligand, Nanogen, Immune Response,

Fig. 3. San Diego career affiliation network 1984.



Biosite, and many others. One recent study found over40 biotechnology companies in San Diego employinga senior manager or board advisor linked to Hybritech(Fikes, 1999). Moreover, Greene and Royston, two of the

original three founders of Hybritech, eventually becameimportant venture capitalists within the region. Birn-dorf, the third founder, developed a reputation with localventure capitalists as an excellent CEO of early stage




biotechnology companies, and became a short-term CEOof several companies.

Network visualization can help tracing the role ofthe former Hybritech managers in forging connectiv-ity within the San Diego biotechnology community.Figs. 3–6 display career affiliations for several yearsbetween 1984 and 1995. Within these figures, the dots,or nodes, represent senior managers while the lines con-necting dots represent ties. Companies are representedby groups of individuals sharing ties to one another. Tosimplify the network figures, individuals with no ties toother people within the network (so-called isolates) wereremoved from the analysis. Managers with career affil-iations to Hybritech are colored black, while all otherindividuals are shaded gray.

Fig. 3 displays San Diego career affiliation networksfor 1984. At this time, the network comprised a fewdistinct clusters of individuals, each representing com-panies founded during the early formation of the region’sbiotechnology industry. While Hybritech has over adozen senior managers represented in the network in1984, most of the other companies are smaller. Duringthe 1979–1984 period, there was very little labor mobil-ity across firms. The only instance of mobility acrossfirms in the network surrounded the formation of Gen-Probe in 1983, in which three Hybritech managers joineda new entrant to the network to found the company.

Fig. 4 displays career affiliation networks in 1987,a year after the acquisition of Hybritech by Lilly. Thisnetwork diagram shows the formation of new two com-panies formed by Hybritech alumni, Gensia and Idec,as well as the general growth of the network main com-ponent as Quidel and Strategene established initial tiesto the growing cluster of individuals linked throughHybritech. By 1990 (Fig. 5), the main component hadgrown significantly, in part through the formation of sev-eral more companies founded by Hybritech alumni. Atthis point, it appears a network backbone had emerged,though it is fragile in some places. For example, whileconnectivity linking individuals is high surroundingthe core group of Hybritech alumni, senior managersworking at Agouron are weakly connected to the maincomponent through two individual ties. The 1990 net-work also shows that, due to the 5-year decay model,direct ties linking some former Hybritech managers toone another had ended, though these individuals stilloccupied central positions within the network. By 1995(Fig. 6), a robust network has formed linking a largenumber of companies. All ties to Hybritech had decayedfrom the network, and, while most former Hybritechmanagers were still active within the biotechnologycommunity, their central role in holding the networktogether appears to have declined. Labor market mobil-ity within the region was sufficient to create sustainable


Fig. 7. Average degree centrality of Hybritech alumni and non-Hybritech alumni, 1978–2005.

career affiliation networks linking most firms in theregion.

To help document more systematically, the role ofHybritech managers in shaping robust career affilia-tion networks in San Diego. Fig. 7 displays the averagenumber of ties held by Hybritech alumni through the1978–1995 period as opposed to all other alumni. Whilethis evidence shows that the Hybritech alumni hold moreties, on average, than non-Hybritech alumni through allyears, the evidence for the years 1987–1992 is particu-larly striking. These are the years in which the Hybritechmanagers left to found new companies. During theseyears, the Hybritech alumni held between 23 and 26 tieson average, about 5 times the average connectedness ofthe non-Hybritech alumni. This evidence is consistentwith the explanation that the Hybritech alumni, whenfounding companies during the 1987–1992 period, cre-ated a network backbone of social ties while doing so.

As discussed in Section 4, a problem surrounding theearly formation of social networks in San Diego biotech-nology are possible biases created by the exclusion ofindividuals due to the reliance of career history dataonly to locate senior managers working in San Diegobefore 1994. It is reasonable to conclude that missingdata has not biased the evidence surrounding the keyrole of Hybritech managers in forming a viable networkbackbone. The role of Hybritech managers in seedingthe network is substantiated by both the network visu-alization and network centrality results. Moreover, theHybritech story generally well known within the SanDiego area and reported in local newspaper articlesand academic reports (see Fikes, 1999; Crabtree, 2003).Given that most of the key firms launched between 1986and the early 2000s were founded by former Hybritechmanagers, it also seems unlikely that missing data hashidden additional mechanisms by which the networkemerged.

In sum, the mechanism of network emergence sur-rounding the failed Hybritech acquisition helps justifythe claim that dynamics surrounding the formation ofan appropriate social structure within the region wereimportant in explaining the success of the region in devel-oping a large cluster of companies. Through both seedinga generation of follow-on companies to Hybritech and,through mobility to and from these firms, creating aweb of social ties across the new firms, a crediblenetwork backbone emerged. Referring back to the ear-lier game theoretic analogy, career expectations acrosssenior managers within San Diego “flipped” from outof the low commitment equilibrium into one wherethe career risk of taking jobs within regional biotech-nology firms was reduced by the existence of large,vibrant social networks linking managers within theregion.

6. Conclusion

This study helped identify mechanisms by whichsocial networks linked to career mobility emerged andbecame sustainable within the San Diego biotechnol-ogy cluster. How does a regional economy develop asocial structure favoring job mobility? Process tracingthrough a year by year analysis of the network revealsthat viable social networks linking San Diego seniormanagers were initially created by a small cadre of for-mer Hybritech managers that created a number of newcompanies in the region. These new firms initially werelinked through their former ties to Hybritech, but soondeveloped a shared labor market pool that helped con-solidate and then expand a viable network backbonefor the San Diego biotechnology industry. During themid-1990s onwards, patterns of career mobility acrossSan Diego managers became sufficient to generate andsustain social networks.

While this study does not attempt to relate social net-works to the innovative performance of area firms, itdid develop a number of measures to evaluate the use-fulness of networks on a yearly basis. Social networkslinking San Diego networks are large, have maintainedtheir efficiency despite a doubling of the network overthe last several years, and are robust to failure. More-over, as the network grew most managers developednumerous ties to other companies through direct careermobility, and could gain access to two dozen or morecompanies through only one or two intermediaries. Thisevidence supports the core claims of the labor marketmobility research stream: San Diego networks facilitatecareer mobility, reducing the risk of talented individu-als working within a start-up company and supporting


innovation strategies within technologically volatileindustries.

While the social network results for San Diego sug-gest that cohesive ties exist senior managers in theregion, one important area not investigated in this study iswhether the organization of networks differ across dif-ferent types of managers. For example, once enteringcommercial biotechnology, do scientists display a biastowards the formation of ties with other scientists, per-haps on the basis of shared technical knowledge, or areshared experience within companies the primary basisby which ties are formed? The findings on Hybritechalso demonstrate the possibility that networks linkingcompany founders may be particularly important. Dofounder networks exhibit different characteristics thanthe more general career affiliation networks examined inthis research? Future research might usefully investigatethese and other issues surrounding the micro-dynamicsof social network formation.

How do these results fit into the larger debatesurrounding the economic development of technologyclusters? One interesting puzzle is why there is somuch variation in the success of technology clusterswithin given economies. Much research has empha-sized the attractiveness of the United States in generatinghigh-technology technology clusters (see, e.g. Hall andSoskice, 2001). Reinforcing this theme is a more spe-cialized literature linking the American success incommercializing science to favorable regulatory frame-works, such as the Bayh–Dole Act (Mowery et al., 2004).Yet there is tremendous variation in the success and fail-ure of clusters even within the United States. A keyassertion emerging from the social network approach isthat more than attractive national frameworks are neededfor clusters to develop. In the San Diego case, effec-tive social ties were able to coalesce around a groupof managers linked to the failed Hybritech acquisi-tion. Combined with access to world class technologyfrom local universities, a small hub of activity was ablecoalesce into a world class biotechnology cluster. Oneinteresting question for future research is whether socialnetworks, especially those promoting mobility, can spurcluster formation in business systems that have beenidentified as disadvantageous for cluster formation. Theappearance of wireless telecommunication clusters inScandinavia, or biotechnology in Southern Germany, areattractive cases to explore.

If cluster development depends on the formation ofsocial networks that have primary origins through sharedcareer experiences, what is the role for governments?Governments across the world are spending large sumsof money in attempts to orchestrate the development of

technology clusters. In the field of biotechnology, thesepolicies usually link subsidized venture capital with poli-cies to encourage and hasten the commercialization ofuniversity science. One finding of this study is that sus-tainable social networks emerge relatively slowly; it tookabout 15 years for this cluster to become sustainable,at least in terms of social network organization. Whilethis might seem like a relatively short time period, itis a much longer time-span than that assumed in manypolitical projects scattered across the world that seek torapidly create new biotechnology clusters through tech-nology policies, but within the confines of short electoralcycles.

Moreover, there is little empirical research suggestingthat the social networks underpinning the developmentof clusters of high-risk entrepreneurial firm can quicklybe orchestrated solely by non-market activities. Whilethe commercialization of university of science may bean effective instrument to catalyze regional cluster cre-ation, the San Diego case shows that critical ties linkingsenior managers within the region were formed primar-ily through shared market experience. In this respect,a disturbing implication of this study is that the eventcatalyzing tie formation in San Diego—a failed acquisi-tion of a highly successful company-may be rare andin any case helps accumulate little knowledge usefulto individuals or governments interested in develop-ing successful clusters. The creation of social structurescapable of supporting high labor market mobility withina region may be outside the purview of direct govern-ment policy. In this respect, research investigating howpolicy may shape the formation of shared market expe-riences by experienced scientists and managers couldhelp reveal alternative mechanisms by which decentral-ized social networks within emerging high-technologyclusters could be forged.

Acknowledgements

For valuable research assistance, I’d like to thankTiffany Sun, a student supported by a National ScienceFoundation REU Summer Research Assistance Grant.For helpful comments on an earlier draft, I’d like to thankthe editors of this special issue, Mark Ebers and WoodyPowell, as well as three anonymous reviewers. I’d alsolike to thank Pepper Culpepper, David Finegold, Hen-rik Glimsted, Fiona Murray, Gunnar Trumbull, RichardWhitley and participants of Woody Powell’s ‘lab’ semi-nar at Stanford, the “Corporate Governance in the PacificRim” workshop at UCSD, and the “Exploring Biotech-nology” theme group at the 21st EGOS Colloquium,Berlin. All errors of course remain my own.


References

AlexanderX, 2004. AlexanderX Guide to San Diego Bioscience.AlexanderX, San Diego.

AlexanderX, 2005. AlexanderX Guide to California Bioscience.AlexanderX, San Diego.

Almeida, P., Kogut, B., 1999. Localization of knowledge and the mobil-ity of engineers in regional networks. Management Science 45,905–917.

Baron, J.N., Hannan, M.T., 2002. Organizational blueprints for suc-cess in high-tech start-ups: lessons from the Stanford project onemerging companies. California Management Review 44, 8–36.

Bahrami, H., Evans, S., 1999. Flexible re-cycling and high-technologyentrepreneurship. California Management Review 37, 62–88.

Biotech Career Center, 2005. Biotech Graveyard URL: http://www.biotechcareercenter.com/Graveyard index.html.

Brookings Institute, 2002. Profile of Biomedical Research and Biotech-nology Commercialization: San Diego Metropolitan StatisticalArea. Available on the internet at: http://www.brookings.edu/es/urban/publications/biotechsandiego.pdf.

Casper, S., Murray, F., 2003. Examining the marketplace for ideas:how local are European biotechnology clusters. In: McKelvey,Maureen, Jens Hageman (Eds.), Industrial Dynamics in EuropeanBiotechnology. Elgar, London.

Casper, S., Murray, F., 2004. Careers and clusters: analyzing career net-work dynamics of biotechnology clusters. Journal of Engineeringand Technology Management 22, 51–74.

Casper, S., Whitley, R., 2004. Managing competences inentrepreneurial technology firms: a comparative institu-tional analysis of Germany, Sweden and the UK. Research Policy33, 89–106.

Crabtree, P., 2003. A Magical Place. San Diego Union Tribune,September 14.

DeVol, P., Ki, J., Bedroussian, A., Koepp, R., 2005. America’s Biotechand Life Science Clusters: San Diego’s Position and EconomicContributions. Milken Institute, Los Angeles.

Fikes, B., 1999. Why San Diego Has Biotech, San Diego Metropolitan.http://www.sandiegometro.com/1999/apr/biotech.html.

Fleming, L., King, C., Juda, A., 2006. Small Worlds and RegionalInnovation Working Paper, Harvard Business School, March 21.

Granovetter, M., 1973. The strength of weak ties. American Journal ofSociology 78 (May).

Hall, P.A., Soskice, D. (Eds.), 2001. Varieties of Capitalism: TheInstitutional Foundations of Comparative Advantage. Oxford Uni-versity Press, Oxford.

Henderson, R., Orsenigo, L., Pisano, G., 1999. The pharmaceuti-cal industry and the revolution in molecular biology: interactionsamong scientific, institutional and organizational change. In: Mow-ery, D., Nelson, R. (Eds.), Sources of Industrial LeadershipCambridge. Cambridge University Press, pp. 267–311.

Herrigel, G., 1993. Power and the redefinition of industrial districts: thecase of Baden-Wurttemberg. In: Grabher, G. (Ed.), The EmbeddedFirm. Routledge, London.

Higgins, M., 2005. Career Imprints: Creating Leaders Across an Indus-try. Wiley and Sons, New York.

Higgins, M., Gulati, R., 2003. Getting off to a good start: the effectsof upper echelon affiliations on interorganizational endorsements.Organization Science 14, 244–263.

Kenney, M., Florida, R., 1998. Venture-capital financed innovationand technological change in the USA. Research Policy 17, 119–139.

Krugman, P., 1991. Geography and Trade. MIT Press, Cambridge.Lee, C., Walshok, M., 2000. Making Connections: The Evolution

of Links Between UCSD Researchers and San Diego’s BiotechIndustry.

Mowery, D., Nelson, R., Sampat, B., Ziedonis, A., 2004. Ivory Towerand Industrial Innovation: University–Industry Technology Trans-fer Before and After the Bayh–Dole Act in the United States.Stanford University Press, Palo Alto.

Murray, F., 2004. The role of inventors in knowledge transfer: sharingin the laboratory life. Research Policy 33, 643–659.

Owen-Smith, J., Powell, W., 2004. Knowledge networks as channelsand conduits: the effects of spillovers in the Boston biotechnologycommunity. Organization Science 51, 5–21.

Penan, H., 1996. R&D strategy in a techno-economic network:Alzheimer’s disease therapeutic strategies. Research Policy 25,337–358.

Powell, W., 1996. Inter-organizational collaboration in the biotechnol-ogy industry. Journal of Institutional and Theoretical Economics152 (1), 197–215.

Powell, W., Koput, K., Smith-Doerr, L., 1996. Inter-organizationalcollaboration and the locus of innovation: networks of learn-ing in biotechnology. Administrative Science Quarterly 41, 116–145.

Powell, W., White, D., Koput, K., Owen-Smith, J., 2004. Networkdynamics and field evolution: the growth of inter-organizationalcollaboration in the life sciences. American Journal of Sociology110, 1132–1205.

Sabel, C., 1992. Studied trust: building new forms of cooperationin a volatile economy. In: Swedberg, R. (Ed.), Explorations inEconomic Sociology. Russell Sage Foundation, New York, pp.104–144.

Saxenian, A., 1994. Regional Advantage: Culture and Competitionin Silicon Valley and Route 128. Harvard University Press, Cam-bridge, MA.

Simard, C., 2004. From Weapons to Cell Phones: Knowledge Networksin San Diego’s Wireless Valley. Dissertation. Stanford University.

Shan, W., Walker, G., Kogut, B., 1994. Interfirm cooperation andstartup innovation in the biotechnology industry. Strategic Man-agement Journal 155, 387–394.

Storper, M., 1997. The Regional World. The Guilford Press, New York.Teece, D., 1986. Profiting from technological innovation: implica-

tions for integration, collaboration, licensing, and public policy.Research Policy 15, 285–305.

Utterback, J., 1996. Mastering the Dynamics of Innovation. HarvardUniversity Press, Cambridge.

Uzzi, B., Spiro, J., 2005. Collaboration and creativity: the small worldproblem. American Journal of Sociology 11, 447–504.

Wassermann, S., Faust, K., 1994. Social Network Analysis Cambridge:Cambridge University Press.

Whitley, R., 2004. Project-based firms: new organisational form orvariations on a theme? Working Paper, Manchester BusinessSchool.

Zucker, L., Darby, M., Brewer, M., 1998. Intellectual human capital andthe birth of U.S. biotechnology enterprises. American EconomicReview 88, 290–306.

http://www.biotechcareercenter.com/Graveyard_index.html

http://www.biotechcareercenter.com/Graveyard_index.html

http://www.brookings.edu/es/urban/publications/biotechsandiego.pdf

http://www.brookings.edu/es/urban/publications/biotechsandiego.pdf

http://www.sandiegometro.com/1999/apr/biotech.html

how do technology clusters emerge and become sustainable ...keck graduate institute of applied life...

Documents