OrganizationScienceArticles in Advance, pp. 1–19ISSN 1047-7039 (print) � ISSN 1526-5455 (online) http://dx.doi.org/10.1287/orsc.1110.0722

© 2012 INFORMS

Firm-Specific, Industry-Specific, and OccupationalHuman Capital and the Sourcing of Knowledge Work

Kyle J. MayerDepartment of Management and Organization, Marshall School of Business, University of Southern California,

Los Angeles, California 90089, kmayer@marshall.usc.edu

Deepak SomayaCollege of Business, University of Illinois at Urbana–Champaign, Champaign, Illinois 61820, dsomaya@illinois.edu

Ian O. WilliamsonMelbourne Business School, Melbourne, Victoria 3053, Australia, i.williamson@mbs.edu

Whereas capability differences are known to impact governance decisions, what drives heterogeneity in firm capabil-ities? We propose that capability differences may arise from governance choices related to the focal activity and

study how firms accumulate capabilities in the firm-specific, industry-specific, and occupational human capital necessaryto perform knowledge work. We theorize that prior outsourcing decisions influence the development of firm- and industry-specific human capital and that buyer–supplier differences in the management of skilled employees can produce systematicdifferences in capabilities based on occupational human capital. Additionally, we explore some contingencies in the devel-opment of these types of human capital and their impacts on outsourcing knowledge work. These propositions are testedwith a unique data set on the outsourcing of legal work involved in filing patents (i.e., patent prosecution).

Key words : human capital; organizational capabilities; knowledge-based view; outsourcing; knowledge work;resource-based view; transaction cost economics

History : Published online in Articles in Advance.

IntroductionKnowledge-based business activities have become anincreasingly important component of firm performance(Grant 1996, Itami 1997, Kogut and Zander 1992), andsimultaneously, the market for outsourced knowledgework has also been growing rapidly (e.g., in fieldssuch as accounting, consulting, information technology(IT), law, marketing, product design, and research anddevelopment (R&D)). Statistics from the U.S. Censusindicate that this economic sector’s GDP value addednearly doubled in the decade 1998–2008, whereas man-ufacturing grew by a mere 24%.1 Many firms performsome of the same knowledge-based functions internallywhile simultaneously outsourcing others to external sup-pliers. Therefore, understanding how firms decide tosource knowledge-based projects and how these deci-sions are influenced by the firm’s development of valu-able knowledge-based competencies is important formanagement research. Sourcing decisions have beenaddressed by several theories that highlight differentbut potentially complementary motivations, but in priorwork, the origins of firm capabilities that subsequentlyaffect boundary choices are typically either unexplainedor based on serendipitous experience (e.g., Argyres1996, Leiblein and Miller 2003, Mayer and Solomon2006). In this paper, we examine how these differential

firm capabilities may arise and in turn influence out-sourcing decisions, especially in the context of knowl-edge work.

We develop and examine the proposition thatknowledge-based capabilities may have “governance”origins; specifically, capabilities may arise from priormake-or-buy choices and buyer–supplier differences inthe management of skilled employees. Because it iswidely acknowledged that firms’ knowledge and capa-bilities are primarily situated in their human capital (HC)(Grant 1996, Felin and Hesterly 2007), we draw on priorresearch to develop a deeper understanding of the dif-ferent types of HC required for skilled knowledge work(e.g., Becker 1964; Castanias and Helfat 1991, 2001;Gibbons and Waldman 2004; Kambourov and Manovskii2009). We advance our ideas by focusing on the firm-specific, industry-specific, and occupational HC rele-vant to performing knowledge projects and examiningthe extent to which firms may have built capabilitiesin each of these types of human capital as a result ofthe governance-based logics noted above. In turn, wehypothesize and test how these presumed differences inhuman capital-based capabilities will affect the outsourc-ing decision for individual knowledge projects.

Our research makes four main contributions to thecapabilities and outsourcing literatures. First, we drawon and apply existing typologies of human capital to

1

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Published online ahead of print June 15, 2012

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge Work2 Organization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS

develop a logic of capability development at the func-tional level within firms. Although prior research haslargely employed human capital typology to understandfirm capabilities derived from top-level management(e.g., Castanias and Helfat 1991), Castanias and Helfat(2001) also note the potential to extend the analysis tofunctional HC, which we hereby realize.

Second, and most importantly, we develop two cen-tral logics of capability development within an HCframework, which in turn impact outsourcing decisions.The first logic relates to the impact of prior gover-nance (make-or-buy) choices on the firm’s human capi-tal. In a recent theory paper, Argyres and Zenger (2012)propose that current capabilities may ultimately havetheir roots in the firm’s prior boundary choices. Build-ing on their work, we advance our understanding ofthe dynamic relationship between prior sourcing deci-sions, their influence on the accumulation of humancapital through learning by doing (specifically in rel-evant firm-specific and industry-specific HC), and theimpact on subsequent outsourcing decisions. The sec-ond logic, which draws on ideas relating to diminishedincentives and selective intervention within firms rela-tive to markets (Williamson 1985, Foss 2003), highlightsthe differences between (client) firms making sourcingdecisions and specialized suppliers in the way they gov-ern their employees. In particular, suppliers specializedin a particular occupation (e.g., law, accounting, IT) maybe better able than buyer (client) firms to incentivizeand enable employees to develop very high levels ofoccupational HC. Taken together, these logics suggestthat managers walk a fine line with their firm bound-ary choices—excessive outsourcing can “hollow out”the firm’s knowledge (Reitzig and Wagner 2010) andincrease dependence on suppliers (Ring and Van de Ven1994, Langlois and Robertson 1995), whereas exces-sive integration can hurt the firm’s competitive positionif superior human capital (and associated capabilities)exists in the market.

Third, recognizing that the relevance of human cap-ital for knowledge work does not exist in a vacuum,we examine a number of contingencies that impacthow different types of HC are accumulated and linkedto outsourcing decisions. To begin with, we examinehow firm-specific HC and industry-specific HC rele-vant to a particular knowledge project moderate eachother’s impact on outsourcing. Furthermore, we studyhow projects with certain attributes may rely more heav-ily on particular types of human capital; for example,a high level of occupational HC may be essential forprojects in highly contested areas. Finally, we examinethe potential for firms to attenuate the theorized supplieradvantage in occupational HC by investing in the devel-opment of large internal functions within which occupa-tional expertise can be better developed and maintained.

Fourth, we employ detailed project-level data frompatent prosecution work—that is, the work of drafting,filing, and obtaining patents—to test the effects of dif-ferent types of human capital on outsourcing decisions.Patent data provide precise measures of our key humancapital variables and facilitate robust controls for alter-native explanations. Thus, our study lays the ground-work for future research that seeks to test microanalytictheories of firm capabilities based on human capital.We complement our large sample analyses with unstruc-tured interviews of managers, which enable us to furthercorroborate our findings and discuss their theoretical andmanagerial implications.

Theory and HypothesesHistorically, the implications of transactional governance(informed by transaction cost economics (TCE)) andcapabilities (informed by resource-based view (RBV))for firm boundaries were often characterized as beingat odds (e.g., Conner and Prahalad 1996, Ghoshaland Moran 1996, Foss 1996, Kogut and Zander 1996,Williamson 1996); however, their central tenets appearto be more compatible than antithetical, and subse-quent research has sought to realize these complemen-tarities in various ways. One approach has sought toexamine how capabilities—both productive and gover-nance related—help to explain the effective governanceof transactions within and across firm boundaries (Mayerand Salomon 2006, Argyres and Mayer 2007). Anotherapproach has acknowledged that both capability differ-ences and transaction costs are necessary for a theoryof firm boundaries, and has argued that transaction costsare often endogenously determined by capability dif-ferences and resulting firm specialization (Langlois andRobertson 1995, Jacobides and Winter 2005). In the cur-rent paper, we seek to advance the increasing integrationof governance and capabilities views in a slightly differ-ent way. Echoing Argyres and Zenger (2012), we pro-pose and test a theory that focuses on how governanceleads to the development of capabilities. In so doing,we also develop a potential response to a foundationalbut challenging question in RBV research (Barney 1986,Dierickx and Cool 1989): What are the main sources ofinterfirm differences in capabilities?

We begin with the human capital framework weadopt to develop our theory of capability developmentin knowledge work. Knowledge work essentially con-sists of activities, tasks, or projects (henceforth, projects)that require the application of knowledge to solve busi-ness problems. For example, a market may need to beresearched, a business plan developed, a legal documentprepared, or a product designed. Accordingly, we focuson the expertise needed to competently solve businessproblems and execute knowledge projects (March 1991,Simon 1991), and such expertise is ultimately rooted in

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge WorkOrganization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS 3

human capital. Thus, although there are many epistemo-logical debates on what essentially constitutes knowl-edge (Spender 1996), our view of firm knowledge isthat it is composed of the expertise and skills embed-ded within the human capital of the firm’s employees.Therefore, exploring the implications of different typesof human capital may be helpful for explaining capabil-ity development.

Building on Becker’s (1964) pioneering work,Castanias and Helfat (1991, 2001) advance three maintypes of managerial human capital that underlie firm cap-abilities—firm specific, industry specific, and general.2

The three types are conceptually nested, with firm-specific human capital being the most narrowly applica-ble and general human capital being the broadest. In thispaper, we employ and extend the Castanias and Helfat(1991) typology by shifting its focus from top managersto skilled employees at the functional level within theorganization who are responsible for knowledge-basedactivities. Within the broad Castanias and Helfat (1991)category of general human capital, we focus on occu-pational HC, a type of general HC that is highly rele-vant for knowledge workers who are members of rela-tively well-defined occupations or professions (e.g., law,accounting, engineering) (Kambourov and Manovskii2009, Shaw 1984). Occupational expertise can be a sig-nificant driver of performance in knowledge work (e.g.,an expert IT engineer or lawyer will likely deliver a farsuperior work product), and by highlighting this cate-gory, we hope to direct more research attention to theimportance of this type of general HC. As a practicalmatter, many dimensions of general HC are manifestedin occupational HC—e.g., more intelligent attorneys arealso likely “better lawyers”—and to that extent they areincluded in our approach. We do not mean to imply thatother nonoccupational aspects of general HC are not rel-evant, and indeed, we cannot empirically rule out theirimpact on capability development and sourcing deci-sions. However, our theorizing about capability develop-ment in general HC relies in part on suppliers who arespecialized along occupational lines, and that logic doesnot translate fully to other types of general HC (outsideoccupational HC).

In summary, we focus on three types of human cap-ital—firm-specific, industry-specific, and occupational.To further clarify our framework, we define these HCtypes and explain how they apply in the empirical con-text of our study, namely, patent legal work. Firm-specific HC refers to knowledge and skills that areunique to a firm, such as knowledge about specific strate-gies, processes, and technologies of the firm (Castaniasand Helfat 1991). For example, effective patent legalwork may require firm-specific knowledge about therelationship of the focal patent to the relevant productline or technological trajectory of the firm. Industry-specific HC refers to knowledge about the industry set-ting or domain in which a project is situated, and thus

it is redeployable across the (limited) set of firms withprojects in the same industry domain (Castanias andHelfat 1991). Because patents are filed on inventionslocated in different technological areas, knowledge aboutthe technological domain of a patent is an importanttype of industry-specific HC needed to file and prose-cute patents. Finally, occupational HC consists of theknowledge and skills required to perform work withina professional or functional area, and as a type of gen-eral HC, it is most easily transferred across industry andfirm settings (Kambourov and Manovskii 2009, Shaw1984). Because industries often develop around an occu-pation, we should be careful to distinguish betweenoccupational and industry-specific HC, whereby the for-mer refers to knowledge and skills in the “task domain”of the focal project and the latter refers to the “appli-cation domain” of the project. For example, legal workon a biotechnology patent requires knowledge aboutboth patent law and biotechnology, and in this con-text, occupational HC refers to knowledge about patentlaw and industry-specific HC refers to knowledge aboutbiotechnology.

Before moving forward with our theorizing, we noteone final wrinkle in applying human capital typologiesto understand capabilities and outsourcing in knowledgework. Whereas all firm-specific and industry-specificHC may matter to performance at the top managementlevel (Castanias and Helfat 1991, 2001), the humancapital relevant to project-level performance may bea much narrower subset of these broad groupings.3

Top managers need to know a great deal about theentire firm and the whole industry to make the bestdecisions. When looking at functional-level knowledgework, however, a narrower range of human capital abouta specific firm context (e.g., a specific product line)or a specific industry domain (e.g., a specific appli-cation area such as digital signal processing) may bemore relevant. Recent labor economics research has alsobegun to highlight the value of focusing on such nar-row “task-specific” categories of human capital (Gibbonsand Waldman 2004, Gathmann and Schönberg 2010).To ensure clarity and precision, we refer henceforth tothe narrower subsets of human capital that matter specif-ically to a focal project by labeling them “relevant”(firm- or industry-specific) HC. Drawing on the threetypes of human capital—firm specific, industry specific,and occupational—we now turn to developing our the-oretical arguments about how transactional governanceaffects the firm’s capability development and in turndrives outsourcing decisions on the focal project.

Firm-Specific Human CapitalWhen knowledge projects are closely tied to a uniquetechnology, organizational unit, or strategic area of thefirm, effective problem solving for the completion of

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge Work4 Organization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS

the project requires significant firm-specific human cap-ital. For example, when a patent relates to a well-established product of the company, understanding themarket position, complementary technologies, and com-petitive background of the product would be invaluablefor writing and obtaining a legally meaningful patent.External suppliers face distinct disadvantages, relativeto internal departments, in the development of firm-specific HC because they may not share a common lan-guage, relationships, or a sense of identification thatexists among departments within a firm (Grant 1996;Kogut and Zander 1992, 1996). Suppliers may alsobe reluctant to make investments in firm-specific HC(Williamson 1975) and will have to be compensated todo so. Moreover, because of time compression disec-onomies (Dierickx and Cool 1989, Knott et al. 2003),suppliers may be unable to quickly develop the samedepth of firm-specific HC as internal providers. Becausefirm-specific HC tends to be associated with strategicallyimportant areas of the firm, outsourcing to suppliers canalso raise appropriability and monitoring issues (Mayerand Nickerson 2005).

Despite these traditional rationales for the internal-ization of projects needing substantial firm-specific HC,buyers and suppliers generally have longer histories thana single transaction, and the supplier may have com-pleted related projects for the client in the past. Argyresand Liebeskind (1999) refer to governance inseparabili-ties whereby outsourcing choices made for one transac-tion materially influence the choice sets or relative costsfor outsourcing future transactions. Thus, suppliers mayhave already developed relevant firm-specific HC fromtheir experience in executing closely related projects forthe firm. The outsourcing of prior related projects mayalso improve performance in the focal project becausethe parties learn how to effectively work and con-tract with each other (Langlois and Robertson 1995,Mayer and Argyres 2004). Additionally, prior experi-ence may mitigate appropriability concerns because ofthe deployment of relational governance mechanismsand the development of mutual trust and understandingbetween the firms (Gulati 1995, Uzzi 1996). By con-trast, if closely related projects were internalized in thepast, the relevant firm-specific HC would be built upwithin the firm instead of suppliers, and the develop-ment of relational governance with suppliers would alsobe curtailed. Consequently, it is not simply the need forfirm-specific HC but the extent to which the relevantfirm-specific HC has been retained internally that shouldbe associated with a higher probability of internalizationon the focal project.

Hypothesis 1 (H1). Firms are less likely to out-source knowledge projects the greater their relevantfirm-specific human capital developed by performingprior related projects internally.

Industry-Specific Human CapitalThe successful execution of a knowledge-based projectalso requires an understanding of the industry setting inwhich the activity will occur. Knowledge work generallyinvolves trade-offs between different task dimensions,and a good understanding of the application domain isnecessary to make such trade-offs and exercise creativity.When (client) firms diversify into new areas, they mayinitially lack the in-house industry-specific HC requiredto effectively perform in these areas. For example, evena firm with a strong IT department may not feel con-fident in developing a new type of software applica-tion in-house for a new industry context (e.g., a bankdiversifying into online trading). As such, organizationaldecision makers face an important strategic choice: theymay decide to rely on the industry-specific HC of exter-nal suppliers, or they may choose to develop it inter-nally and rely on external sources only in limited situ-ations (Parmigiani 2007). Although the firm may makea sourcing decision based on relevant current consider-ations (e.g., workload, staffing budgets, time pressures,transaction costs), this decision can have lasting impactson its future development of industry-specific HC, thusshaping the industry domain capabilities of the firm andinfluencing subsequent outsourcing decisions.

Firms may attempt to learn by working closelywith external suppliers and thus seek to develop theirindustry-specific HC (Rothaermel et al. 2006, Parmigianiand Mitchell 2009). However, several factors may limitthe effectiveness of this strategy. First, suppliers maynot want to share their knowledge with the client—theywill provide the good or service, but not necessarily helptheir clients learn it so well that they could do it them-selves. Second, industry-specific HC may be difficult tolearn because of tacitness, causal ambiguity, proprietarytechnology, or other appropriability barriers. Therefore itis likely to be more difficult to transfer industry-specificHC across firms than within organizational boundaries(Darr et al. 1995, Argote and Ingram 2000). Finally, tothe extent that industry-specific HC is built up throughlearning by doing in a given domain (Hatch and Dyer2004) and requires constant improvement and updating,reliance on suppliers creates and repeatedly replenishessupplier capabilities, leaving client firms to continuallyplay catch-up. As suppliers develop greater familiaritywith the firm and its technologies (and the two learn towork together), the firm’s lack of capabilities in industry-specific HC can become an entrenched driver of furtheroutsourcing.

Alternatively, organizational decision makers maydecide to develop industry-specific HC within the firm.Although firms can hire new employees and/or investin training to partially address their industry-specificHC needs, it will ultimately be necessary to “learn bydoing” knowledge projects internally (Hatch and Dyer2004). In dynamic, fast-moving domains in particular,

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge WorkOrganization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS 5

the challenges of learning from suppliers may be diffi-cult to overcome without actually engaging in the rel-evant knowledge work. Overall, we conclude that theoutsourcing of knowledge work will be less likely whenthe firm has developed relevant industry-specific HCthrough the internalization of prior projects in the sameindustry domain.

Hypothesis 2 (H2). Firms are less likely to out-source knowledge projects the greater the relevantindustry-specific human capital developed by previouslyinternalizing knowledge projects in the same domain.

Interaction Between Firm-Specificand Industry-Specific Human CapitalWhen performing knowledge work, firm-specific andindustry-specific human capital may both be relevant fora particular activity. For example, prosecuting a patentsimultaneously requires knowledge about developmentsin the technological domain of the patent (industry-specific HC) and the role played by the patent within thefirm (firm-specific HC). The issue we seek to addresshere is the interaction of firm- and industry-specific HCin predicting outsourcing decisions, which is a valu-able addition to our understanding of how firms sourceknowledge projects. If the relevant firm- and industry-specific HC held by the firm are both low (e.g., becauseof prior outsourcing), there will be little internal capa-bility to perform the project, and the tendency to out-source will likely be strong. However, when the firmdevelops a high level of one type of HC (firm specificor industry specific), we propose that this is likely tosignificantly reduce the tendency to outsource for tworeasons: (1) an increased capability of the firm to exe-cute the knowledge project and (2) coordination benefitsthe firm realizes from utilizing internal human capitalinstead of working with suppliers. The latter followsfrom the knowledge-based view (KBV), which arguesthat advantages in coordinating knowledge-based activi-ties are the foundation of firms (Kogut and Zander 1992,1996; Conner and Prahalad 1996; Nickerson and Zenger2004). These coordination advantages stem from com-plex routines, communication channels, and a sense ofidentification within the firm, which make it more eco-nomical to transfer knowledge and cooperate on knowl-edge projects.

Therefore, because of the additional boost to inter-nalization from the ease of intrafirm coordination ofknowledge projects, the marginal effect of either firm-specific or industry-specific HC on internalization willbe highest when the other HC is low and decrease asthe magnitude of the other HC increases. In essence,once a firm is already high on one type of HC, higherlevels of the second type increase firm capability toperform knowledge work but do not add much in addi-tional coordination benefits. We expect that simultane-ous high levels of firm-specific and industry-specific HC

will lead to the lowest probability of outsourcing; how-ever, from a marginal impact perspective, the greatestmarginal impact of firm-specific or industry-specific HCon outsourcing will occur when the firm is weak in theother type of human capital.4

Hypothesis 3 (H3). The marginal impact of higherfirm-specific or industry-specific human capital on thetendency of firms to internalize knowledge work is high-est when the other type of human capital ( firm specificor industry specific) is low.

Occupational Human CapitalAs noted earlier, occupational HC refers to knowledgeand skills relating to a profession or field of knowl-edge (e.g., law, accounting, software engineering) andis the most widely applicable of our three types of HCbecause it applies across a variety of industry settings.Two key issues must be addressed to understand howoccupational HC impacts the sourcing decision for agiven knowledge project: (1) whether the (client) firm islikely to have a higher level of occupational HC versusspecialized suppliers and (2) the extent to which occu-pational HC is needed to complete a particular task.

There are important differences between intrafirmfunctional departments of companies (e.g., the legaldepartment within a semiconductor company) and sup-pliers who specialize in an activity (e.g., an intellectualproperty law firm) in how they manage and incentivizetalented individuals to accumulate human capital. Spe-cialized suppliers may have a particular advantage in(and motivation for) developing superior occupationalHC by virtue of their differential ability to attract, moti-vate, and retain knowledge workers. The suppliers’ labormarket advantage is driven in part by their ability toprovide a highly incentivized work environment, whichattracts talented employees and rewards them for devel-oping high levels of occupational HC. Because superioroccupational HC can in turn be employed to servicethe needs of multiple clients, it becomes a fundamentalvalue proposition for some supplier firms, who in turncreate highly targeted incentives and a strong selectionenvironment for occupational experts within their orga-nizations. Talented individuals are also drawn to thesesuppliers for professional development, which stemsfrom the variety of occupationally interesting and chal-lenging work they offer, and opportunities to work andlearn alongside occupational experts. Although suppli-ers must bear the costs of coordination with client firms,these costs can be offset by the premium they receivefor their expertise, some of which is passed on throughincentivized pay and professional development to theiremployees.

Intrafirm departments, in contrast, trade on theirknowledge about the processes, technologies, and strate-gies specific to the firm (firm-specific HC), and con-sequently, they tend to engage in a narrower range of

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge Work6 Organization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS

activities than external suppliers. As a result, intrafirmdepartments are unable to reproduce supplier incen-tive structures for the development of occupational HC.Client firms can try to attract and motivate talentedemployees by promising them interesting work as well ashigh pay, but they may be unable to fulfill such commit-ments because they face inherent difficulties in creatinga market-like environment for their employees, in partbecause unlike specialized suppliers they do not have tocompete for work from external clients. In other words,focal firms are at least partly handicapped in their devel-opment of occupational HC because of the well-knownchallenges posed by selective intervention within firms(Williamson 1985, Foss 2003). Thus, the two core issuesare that the firm cannot replicate supplier incentives foremployees to be experts in their occupation and thatthe range of projects (and occupational HC developmentopportunities) within any one firm is inherently limitedby the scope of the firm.

An example from our empirical context is patent attor-neys. Internal patent attorneys are likely to stay cur-rent only in narrow aspects of the law that affect theiremployers’ technologies (and may not have time evenfor that, as they deal with various firm-specific issues),whereas patent attorneys for a top law firm will be cur-rent in many aspects of patent law because of the learn-ing opportunities they have and the incentives they faceto win business and service clients. Specialized law firmsmay be disadvantaged relative to internal patent depart-ments in their knowledge of the firm (firm-specific HC);however, they can compete with internal departmentsbased on superior occupational (and potentially industry-specific) HC, which in turn can be developed throughand applied in services provided to multiple client firms.

Thus, some specialized suppliers are likely to accu-mulate superior occupational HC and attract customerslooking for such expertise, which leads to more expertisedevelopment. It is important to recognize that suppli-ers may not develop superior occupational HC under allconditions of knowledge work. Nor is it the case that allsuppliers will be more expert than intrafirm functions.For example, many suppliers may get business (andfirms may outsource to them) because of lower costsor scale economies, because of industry-specific HC, orto smooth out internal firm workloads. Moreover, evenwhen suppliers have superior occupational HC, clientfirms may not always outsource because expert service isinherently expensive (as we note above) and potentiallycostly to coordinate across firm boundaries (Grant 1996;Kogut and Zander 1992, 1996; Somaya et al. 2007).

Acknowledging that (client) firms can access supe-rior occupational HC through the market, the ques-tion then shifts to when such expertise is most likelyto be required. We propose that the incentives toaccess external occupational HC are strongest when thenegative consequences associated with errors in using

occupational HC are large or, equivalently, when thereare significant perceived gains from using expert occupa-tional human capital and therefore a willingness to payfor accessing this expertise.

One factor in particular that may increase the cost oferrors (and thus the value of occupational HC) is theextent to which the project is in an arena with exten-sive competitive actions by rivals (what we refer toas a “highly contested” area). Competitive actions are“specific and detectable competitive move[s]…initiatedby a firm to improve or defend its relative competi-tive position” (Chen et al. 1992, p. 440) and can takemany forms including price cuts, new product introduc-tions, and legal actions (Ferrier et al. 1999, Ketchen etal. 2004). Actions by rivals often threaten the competi-tive position of a firm by increasing the complexity ofthe environment in which the firm must execute keybusiness-related tasks (Porter 1980). Thus, it is morechallenging to execute knowledge-based activities in anenvironment that is highly contested because actions insuch an area typically attract close scrutiny and strongreactions from active rivals.

For example, if a firm wishes to file a patent in anarea that is highly contested—that is, the area has seenmultiple patent lawsuits as rivals fight over their intel-lectual property claims—then it should expect the focalpatent to also be challenged by competitors in multipleways. Rivals are likely to seek to overturn the patent,either proactively or in response to the patenting firm’sefforts to enforce the patent, reflecting the competitiveback-and-forth between rival firms. Therefore, the occu-pational HC required to obtain such a patent is likely tobe more demanding because the negative consequencesassociated with errors in execution (which can surfacelater in litigation and undermine the firm’s intellectualproperty position) are likely to be very significant.

Therefore, in highly contested areas, firms will de-mand high levels of occupational HC, and specializedsuppliers who possess such expertise may be especiallyvaluable. So we predict that firms will be more likely tooutsource in an area that is highly contested.

Hypothesis 4 (H4). Firms are more likely to out-source knowledge projects situated in areas that arehighly contested (and thus rely heavily on occupationalhuman capital).

The reliance on external suppliers for knowledge workin areas that are highly contested (and therefore requiresuperior occupational HC) may be impacted by theextent to which the firm has developed its own stock ofoccupational HC. One key factor that may impact thefirm’s development of superior occupational HC is thesize of its internal staff with training and credentials inthe focal occupation. A larger internal staff will clearlylead to a lower baseline level of outsourcing. The firmmust utilize its staff, and, controlling for other factors,

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge WorkOrganization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS 7

it would outsource less. The main issue we address hereis how that larger internal staff will differentially affectoutsourcing when occupational HC is particularly impor-tant (e.g., in contested areas). We make the case that alarge internal staff may enable the focal firm to offsetsome of the incentive and learning advantages of suppli-ers in developing occupational HC.

One potential advantage of a large internal staff isthat the firm can organize its internal function moreeffectively as a knowledge hierarchy (Garicano 2000,Garicano and Rossi-Hansberg 2006), where some indi-viduals specialize as occupational experts and leveragethe talents of others for more mundane work. Althoughsuppliers may be similarly organized, firms with largerinternal staffs can also use knowledge hierarchies tonarrow the occupational expertise gap with suppliers.Knowledge hierarchies have a number of incentive fea-tures that can help the firm attract and motivate talent.First, the firm’s staff may value opportunities for pro-motion up the hierarchy, and experts at the top of thehierarchy may enjoy compensation that is more “marketlinked” than at other firms. Second, inexperienced talentmay value the chance to work with and learn from theseexperts. Third, the experts themselves may be able tochoose more interesting and challenging projects, whichthey intrinsically value. Finally, a large internal staff mayalso act as a signal to talent that the occupational areais important to the firm and will receive resources andtraining investments in the future.

Additionally, firms with large internal staffs (that areorganized as knowledge hierarchies) may have a num-ber of advantages for learning and expertise develop-ment within the firm. In addition to the within-firmlearning imparted by occupational experts, the existenceof significant occupational HC within the firm mayincrease its absorptive capacity for knowledge from itsenvironment (Cohen and Levinthal 1990). For exam-ple, the firm’s professionals may be better connectedwith outside experts, receive more information aboutnew developments, and be better able to understand andassimilate this information. Such firms are also likely tohave a wider scope of occupational activities, thus pro-viding their staff with a richer variety of opportunitiesto learn and develop their occupational HC.

For all of these reasons, a firm with a large inter-nal staff may be more able to develop occupational HC,which allows it to better handle the complexities asso-ciated with knowledge work in contested areas, thusdecreasing the firm’s proclivity to outsource these typesof projects.

Hypothesis 5 (H5). The larger the firm’s internalstaff (in the focal occupational area), the weaker the(positive) effect of a project being in a highly contestedarea on the probability of outsourcing (i.e., the size ofthe firm’s internal staff negatively moderates the effectof being in a highly contested area on outsourcing).

Data and VariablesSampleWe tested our hypotheses by examining the patent out-sourcing decisions of a sample of Fortune 500 firmsin technology-based industries, in which firms weremost likely to file and obtain patents, over the period1990–1995. Our sample included all (129) public U.S.firms in the 1989 Fortune 500 survey (published in1990) from the following five industries: chemicals(39 firms), computer manufacturing (22 firms), electron-ics (40 firms), pharmaceuticals (12 firms), and scien-tific and photographic equipment (16 firms). Using theDirectory of Corporate Affiliations, we gathered infor-mation about the entire corporate family (all subsidiariesand divisions) for each firm in each year of our dataset. We then matched these various corporate businessunits with the unique patent assignee codes used bythe U.S. Patent and Trademark Office (USPTO) andobtained all patents filed by (or issued to) these firmsduring the period 1985–1995. In essence, we reproducedthe approach adopted for the National Bureau of Eco-nomic Research patent data set (Hall et al. 2001), albeitat a more detailed year-by-year level for a smaller sam-ple of large firms. Based on this approach, we devel-oped a core sample of 80,129 patents filed over six years(1990–1995), although the complete data set (includingprior years used to code some variables) has a little over132,000 patents.

Our interest lies in studying the outsourcing of patentprosecution, which consists of activities related to fil-ing and obtaining a patent from the USPTO. To identifywhether or not a patent was outsourced, we used datafrom the front page of each patent in the “Attorney orAgent” field (see below). In addition, we employed mul-tiple years of data on the universe of registered patentattorneys (from the Office of Enrollment and Disci-pline (OED) of the USPTO) to code the affiliation ofpatent attorneys (or agents) listed on the patent. Patentattorneys are highly specialized professionals who arerequired to have a technical background and pass thepatent bar. Only registered patent attorneys may practicebefore the USPTO, and maintaining a current addresswith the OED is a requirement for registration. Addi-tional data on each patent were collected and coded fromvarious data sets made available by the USPTO.

VariablesThe dependent variable in our analyses is an indicatorvariable, outsourced, which is coded as 1 if the “Attor-ney or Agent” field on the patent lists a law firm orexternal attorney and as 0 otherwise. In many patents,the focal company’s attorneys are listed in this field, asidentified from the OED attorney roster, so these patentswere clearly not outsourced. In most outsourced patents,the “Attorney or Agent” field lists the name of a specific

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge Work8 Organization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS

law firm, usually without the names of attorneys whoworked on the patent.5 In less than 10% of outsourcedpatents, the field lists attorneys employed by a law firmor in private practice, which we identified from the OEDattorney roster. In approximately 12.7% of patents, thefield is blank. We report results by dropping these latterpatents from our sample, but we also estimated modelsby including these patents and coding them as not out-sourced and obtained similar results.6 All coding of theoutsourced variable was done by hand because it wasstraightforward to categorize entries in the “Attorney orAgent” field by visual inspection (but difficult to antici-pate variations in names and spellings beforehand).

We measured relevant firm-specific human capital(Hypothesis 1) for a given patent prosecution projectby the number of “backward” citations to patents ofthe same company that were also processed in-house.Because we expect diminishing returns in the accumula-tion of firm-specific HC, and because this variable (likeothers below) is highly skewed, we logged it. Reflect-ing our theoretical discussion, we seek to differentiatebetween situations in which relevant firm-specific HC islocated within the firm’s internal staff through its priorwork experience versus situations in which prior out-sourcing (measured separately as a control variable; seebelow) may have prevented such HC development. Wemeasure this variable over the five years prior to patentfiling (see Endnote 4 regarding robustness to alternativemeasures).

We measured relevant industry-specific human capi-tal (Hypothesis 2) by the company’s previous in-houseexperience in the technical domain of the focal patent.Because a patent is essentially a property claim in anarea of technology, effective patent legal work requiresknowledge about the patent’s technical domain, which isin effect a narrow area of industry-specific HC. Specifi-cally, we measured the (logged) number of prior patentsprocessed internally by the firm in the same primaryseven-digit International Patent Classification (IPC) asthe focal patent. The IPC has several advantages overU.S. Patent Classification (USPC), the most notable ofwhich are the unambiguous technological basis for IPCclasses and the nested nature of the IPC system (seeLerner 1995). The IPC ensures that each patent classmaps to a narrow but coherent technological subfield(e.g., A61K 038/xx: Medicinal preparations containingpeptides). We measured industry-specific human cap-ital with patents applied for or issued over the pre-vious five years (from the focal patent’s filing date)to account for the obsolescence of human capital overtime and economize on the resource demands of coding.We reproduced our results using a four-digit level of IPCaggregation and a three-year capture period, giving ussubstantial confidence that our findings are not sensitiveto these definitions.7

In Hypothesis 4, we propose that knowledge work sit-uated in highly contested areas is likely to be outsourcedto access the superior occupational HC of market sup-pliers. The ultimate goal of patent prosecution is toobtain a watertight intellectual property right with eco-nomically valuable scope, and both validity and scopeare inevitably challenged if a patent ends up in litiga-tion. At the time of patenting, patents that are in a con-tested domain marked by previous litigation are them-selves likely to be viewed as being at risk of litigation.Among patents issued between 1990 and 2000, patentsciting litigated patents are approximately 6.5 times aslikely to be litigated as the average patent.8 Therefore,patents in such contested areas are particularly likely tobe perceived as needing occupational expertise, whichis corroborated by our interviews with company gen-eral counsels (see the Discussion section). Accordingly,we test H4 by using the dummy variable highly con-tested area (coded as 1 if the focal patent cited a patentthat was litigated in the past). We collected data onpatent litigation in U.S. district courts from the LitAlertdatabase compiled by the USPTO and combined thiswith patent cases from the Section 337 forum of theInternational Trade Commission, which is an alternativeforum for (import-related) patent disputes in the UnitedStates. We test Hypothesis 5 by interacting highly con-tested area with internal staff size, which measures thenumber of patent attorneys employed by the focal firm(coded from the OED roster).

In addition to these independent variables, we alsocoded a set of patent-level control variables. We focuson controls at the patent level because both interfirmcross-sectional variation and quarter-to-quarter intrafirmvariation is already accounted for by our use of firm-quarter fixed effects (see below). Naturally, attributes ofthe patent play the biggest role in outsourcing differ-ences among patents applied for by the same firm in thesame quarter. For this reason, we also do not (indeed,cannot) include the main effect of internal staff size,which has no variation among patents filed within thesame firm-quarter. We use prior outsourcing in industrydomain to control for the (logged) number of same-IPCpatents outsourced (over the previous five years) by thecompany. Note that this is the analog of our industry-specific human capital variable, except that these patentswere outsourced rather than being processed internally.We also control for prior related outsourcing—the ana-log of our firm-specific human capital variable—whichmeasures the (logged) number of self-cited patents out-sourced to external service providers (over the previousfive years).

Furthermore, we include the control variables foreigninventor (if the first listed inventor is based outside theUnited States), number of patent claims (the numberof distinct patent claims listed on the patent), and for-ward citations (a logged count of citations to the focal

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge WorkOrganization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS 9

patent from subsequent patents) in our analyses. Becauseforward citations are truncated by the number of yearsof data available after the patent is issued, we face apotential bias in measuring this variable. Put simply,patents issued earlier will on average have more cita-tions merely because they have had more time in whichto be cited. We therefore obtain an unbiased estimateof forward citations (for 15 years from the issue dateof each patent) by applying a correction factor from thecitation function (versus time) for the average patent ineach technology class (for more detail on this, see Hallet al. 2001). Finally, we code a set of six technology-grouping dummy variables using a USPC-based classifi-cation scheme widely used in prior research (Jaffe et al.1998). Specifically, the groupings we use are drugs andmedical, chemicals, electrical and electronics, comput-ers and communications, and other technologies (withmechanical as the reference category).9

Analyses and ResultsEach observation in our data is at the level of a patent,for which we need to model the dichotomous firm deci-sion whether or not to outsource the legal patent pros-ecution work to an external supplier. Although our dataset is longitudinal, spanning several years for each firm,it is not a panel data set of firm-year observations.Rather, because our companies typically filed for mul-tiple patents in even a short time period (e.g., a three-month quarter), this provides a unique opportunity toaccount for all company-level factors (even ones thatmay vary from quarter to quarter) and focus only onthe variation between patents filed by the same firmwithin the same quarter. Accordingly, we adopt a fixedeffects model with firm-quarter fixed effects. Althoughboth logit and probit approaches are well developed tomodel binary outcomes, the conditional fixed effects pro-bit estimator is biased. Therefore, we employ the con-ditional fixed effects logit estimator to analyze our data,with robust standard errors clustered at the firm-quarterlevel (which ensures consistency in the standard errorestimates).

The firm-quarter fixed effects model renders a veryconservative test of our hypotheses because it accountsfor both interfirm and within-firm (temporal) heterogene-ity, and thus it rules out many alternative explanations.For example, any firm-level factor—such as financing,staffing, or strategy—that may affect outsourcing fromquarter to quarter is accounted for, and the model essen-tially leverages the differences between patents filed byfirms within the same quarter to estimate coefficients.10

This automatically means that we may lose some obser-vations. For example, if a firm does not outsource anypatents (or outsources all) in a given quarter, we can-not use data from this firm-quarter because idiosyncraticfirm-level drivers of outsourcing in that quarter cannot

be separated from patent-level drivers. Consequently, anadditional 13% of our observations drop out of the sam-ple, leaving us with a final set of 59,590 observations.11

Table 1 provides descriptive statistics of our data, splitup by the subsamples of patents that are outsourcedand internalized. Approximately 27% of patents in oursample are outsourced, and there is considerable varia-tion, ranging from 0% in some firm-quarters to 100% inothers. The 25th, 50th, and 75th percentiles of patentsoutsourced are, respectively, 4.6%, 28.6%, and 98.7%across all firm-quarters. The differences between thetwo subsamples in the means of our key independentvariables are consistent with our hypotheses. The dataalso show substantial within unit (firm-quarter) varia-tion, indicating that the patent-level differences are large.We report a correlation table (Table 2), which shows lim-ited correlation across our key variables. In particular,the low correlation between firm- and industry-specificHC is reassuring, indicating that the empirical overlapbetween these two constructs is quite low.

Table 3 presents our main results from the firm-quarterfixed effects logit model. For ease of interpretation, theestimates are reported as odds ratios, with a null valueof 1 (rather than 0). Model 1 in Table 3 shows thatthe control variable measuring the focal patent’s for-eign origin is significantly associated with internaliza-tion rather than outsourcing, consistent with a rationalethat the associated patent work is more efficiently coor-dinated internally. The main effects of firm-specific (H1)and industry-specific (H2) HC are both supported at the1% level of significance in Model 2. Hypothesis 3 pre-dicted that firm-specific and industry-specific HC wouldhave marginally diminishing interactive effects on firms’decisions not to outsource. The positive and significantinteraction reported in Model 3 is consistent with thepredictions of Hypothesis 3. Furthermore, our predictionin Hypothesis 4 that activities in highly contested areasare more likely to be outsourced is supported in Model 4,and Model 5 corroborates that this effect is negativelymoderated by the size of the internal occupational staffwithin the firm, both at the 1% level of statistical signif-icance. Finally, in the full model (Model 6), all of theaforementioned findings are again corroborated. Likeli-hood ratio tests also confirm large statistically significantimprovements in model fit with each additional variable.

Interestingly, the relationships we find are not onlystatistically significant but the magnitudes are also quitelarge, indicating that they underpin economically mean-ingful findings. For example, based on Model 6, a onestandard deviation increase in relevant firm-specific HCdecreases the odds of outsourcing by 21.9% (Hypoth-esis 1). Similarly, a one standard deviation increase inindustry-specific HC is associated with 40.7% lowerodds of outsourcing (Hypothesis 2). Moreover, wefind that prior related outsourcing and prior outsourc-ing in industry domain, the analogous variables that

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge Work10 Organization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS

Table 1 Descriptive Statistics of Key Variables

Not outsourced Outsourced

Within WithinVariables Measures Mean Std. dev. std. dev. Mean Std. dev. std. dev.

Firm-specific humancapital (H1)

(Log) No. of firm’s owncited patents that wereinternalized

00538 00643 00612 00296 00519 00446

Industry-specific humancapital (H2)

(Log) No. of internalizedpatents in same class

20801 10819 10584 20077 10853 10293

Highly contested area(H4)

Dummy = 1 if patent citesa litigated patent

00126 00332 00320 00151 00358 00337

Internal staff size(moderator for H5)

No. of patent attorneysemployed by firm

62.95 42.56 (n/a) 55.96 59.09 (n/a)

Forward citations (Log) No. of citations tofocal patent (est.)

20185 10122 10034 20414 10105 00980

Prior outsourcing inindustry domain

(Log) No. outsourcedpatents in same class

00918 10281 00927 10938 10637 10250

Prior related outsourcing (Log) No. of own citedpatents outsourced

00049 00208 00192 00227 00437 00403

No. of patent claims (Log) No. of claims infocal patent

10786 00985 00952 10691 00899 00846

Foreign inventor Dummy = 1 if first inventoris outside the UnitedStates

00091 00287 00274 00063 00243 00212

Drugs and medical Technology group dummyvariable

00125 00330 00231 00066 00248 00153

Chemicals Technology group dummyvariable

00212 00408 00362 00120 00325 00263

Electrical and electronics Technology group dummyvariable

00273 00445 00399 00294 00456 00404

Computers andcommunications

Technology group dummyvariable

00211 00408 00331 00350 00477 00378

Other technologies Technology group dummyvariable

00023 00149 00143 00016 00127 00112

No. of observations 43,301 16,289No. of (firm-quarter)

groups1,242 1,242

Table 2 Correlations Between Variables

Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14

1 Firm-specific 1human capital

2 Industry-specific 0029 1human capital

3 Highly contested 0002 −0001 1area

4 Internal staff size 0038 0015 −0002 15 Forward citations 0006 001 0009 0011 16 Prior outsourcing 0004 0049 0003 0025 002 1

in industry domain7 Prior related 0002 0 0006 0001 0011 0032 1

outsourcing8 No. of patent 0002 −0007 0002 −0002 0005 −0011 −0002 1

claims9 Foreign inventor −0005 0005 −0001 0006 −0005 0002 −0003 −0001 1

10 Drugs and medical −0001 0008 001 −0018 −0006 −0007 −0006 0014 0005 111 Chemicals 0005 −0014 −0003 −0007 −0018 −0018 −0006 0003 0002 −0017 112 Electrical and −0004 0007 −0008 0001 0003 0001 −0001 −0008 −0003 −0022 −003 1

electronics13 Computers and −0001 0016 0006 0024 0024 0034 0012 −0009 0 −002 −0028 −0036 1

communications14 Other technologies 0001 −0004 0 −0002 −0004 −0009 −0003 0003 −0001 −0006 −0008 −001 −0009 1

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge WorkOrganization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS 11

Table 3 (Firm-Quarter) Fixed Effects Logit Models for the Outsourcing of Patent Legal Work

Dependent variable: Probability of outsourcing

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Firm-specific human capital 00720∗∗ 00711∗∗ 00712∗∗

4000115 4000125 4000125Industry-specific human 00726∗∗ 00655∗∗ 00654∗∗

capital 4000195 4000335 4000325Firm-specific HC ∗ 10033∗∗ 10034∗∗

Industry-specific HC 4000135 4000135Highly contested area 10182∗∗ 10383∗∗ 10313∗∗

4000525 4000855 4000805Highly contested area ∗ 00998∗∗ 00998∗

Internal staff size 4000015 4000015Forward citations 00986 00993 00993 00982 00981 00990

4000155 4000155 4000155 4000145 4000145 4000155Prior outsourcing in 10092∗∗ 10519∗∗ 10515∗∗ 10092∗∗ 10092∗∗ 10513∗∗

industry domain 4000175 4000345 4000335 4000175 4000175 4000335Prior related outsourcing 20491∗∗ 20378∗∗ 20374∗∗ 20480∗∗ 20479∗∗ 20365∗∗

4001775 4001545 4001545 4001755 4001765 4001545No. of patent claims 00926∗ 00939∗ 00942+ 00927∗ 00926∗ 00942+

4000295 4000295 4000295 4000295 4000295 4000295Foreign inventor 00742∗∗ 00718∗∗ 00720∗∗ 00746∗∗ 00746∗∗ 00722∗∗

4000755 4000715 4000715 4000755 4000755 4000725Drugs and medical 10097 10301∗∗ 10306∗∗ 10085 10083 10288∗∗

4001085 4001165 4001175 4001065 4001065 4001155Chemicals 00779∗∗ 00801∗∗ 00802∗∗ 00779∗∗ 00778∗∗ 00800∗∗

4000455 4000445 4000445 4000455 4000455 4000445Electrical and electronics 10002 10104+ 10106+ 10002 10001 10104+

4000565 4000615 4000615 4000565 4000565 4000615Computers and 00828∗ 00966 00965 00819∗∗ 00824∗∗ 00958

communications 4000615 4000665 4000665 4000605 4000605 4000655Other technologies 00772∗ 00821+ 00818∗ 00768∗ 00766∗ 00813∗

4000825 4000845 4000845 4000825 4000825 4000835No. of observations 59,590 59,590 59,590 59,590 59,590 59,590No. of (firm-quarter) groups 1,242 1,242 1,242 1,242 1,242 1,242Log likelihood −201899015 −201337008 −201332099 −201888003 −201881068 −201320056

Notes. Robust clustered (by firm-quarter) standard errors are in parentheses. Estimates reported are exponentiated coefficients (oddsratios); thus, values above (below) 1 indicate a higher (lower) probability of outsourcing.

+Significant at the 10% level; ∗ significant at the 5% level; ∗∗significant at the 1% level.

measure prior outsourcing rather than internalization,were associated with statistically significant (1% level)increases—as opposed to decreases—in the odds of out-sourcing. In other words, it is not just the firm’s priorprojects within the same firm-specific context or industrydomain that matters but the extent to which the rele-vant human capital was developed internally by keepingthe prior projects in-house. The impact of patents beingin highly contested areas is also quite large; Model 6indicates that this increases the odds of outsourcing by31.3% (for a firm with an internal staff size of zeropatent attorneys). At a more reasonable internal staffsize of 14 (mean minus one standard deviation for thisvariable), a patent in a highly contested area still hasapproximately 28.2% higher odds of outsourcing.

We now turn our attention to interpreting the inter-action effects that we estimate. Interpreting the mag-

nitude and direction of change implied by interactionterms can be tricky in choice models because of thenonlinear character of these models (Huang and Shields2000). Therefore, we follow recommended best practice(Hoetker 2007) and evaluate the joint impact of interac-tion variables at representative high and low values fromour data (see Figures 1 and 2). The chosen values of thevariables are mentioned in the notes accompanying thefigures. In each figure, there are two sets of graphs, onerepresenting the estimated impact in terms of odds ratiosand the other in terms of probabilities.12 Consistent withHypothesis 3, Figure 1 shows that the marginal impactof an increase in firm- or industry-specific HC on out-sourcing is greatest when the level of the other type ofHC (industry specific or firm specific) is low. For exam-ple, an increase in firm-specific HC from low to highlevels is associated with a 58.6% drop in the odds of

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge Work12 Organization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS

Figure 1 Interactive Effects of Relevant Firm-Specific andIndustry-Specific Human Capital on Outsourcing

0

0.1

0.2

0.3

0.4

0.5

0.6

Low industry-specifichumancapital

High industry-specifichuman capital

Pro

babi

ty o

f our

sour

cing

Combined effect of relevant firm-specific andindustry-specific human capital on P (outsourcing)

(baseline average probability = 0.3)

00.10.20.30.40.50.60.70.80.9

Low industry-specifichuman capital

High industry-specifichuman capital

Odd

s ra

tio im

pact

Combined effect of relevant firm-specific andindustry-specific human capital on the odds of outsourcing

Low firm-specific human capitalHigh firm-specific human capital

Note. For both firm-specific human capital and industry-specifichuman capital, we use the 25th and 75th percentiles of the vari-ables as the low and high values, respectively.

outsourcing when industry-specific HC is low but only a47.9% when it is high. Figure 2 shows that the positiveimpact of highly contested areas on the probability ofoutsourcing is diminished in firms with larger internalstaff. Based on the coefficient estimates, we infer thatwhen a firm has a staff comprising 161 patent attorneys,the tendency to outsource patent prosecution in highlycontested areas is completely nullified. In our sample,the largest internal staff size is 159 attorneys (for IBM),which indicates that the impact of highly contested areasis essentially zero for the largest firms in our sample.

DiscussionIn this paper, we have developed a human capital-centered framework of capability development andsourcing decisions in knowledge work. We highlightthe value of examining the governance of related (butdistinctly separate) transactions of the firm and howthey influence the focal sourcing decision through theirimplications for capability development. To this end, weextend Castanias and Helfat’s (1991, 2001) frameworkof firm-specific, industry-specific, and general humancapital by focusing on occupational HC as a type of gen-eral HC that is particularly relevant to the performance

Figure 2 Combined Effects of Highly Contested Area andInternal Staff Size on Outsourcing

0.75

0.85

0.95

1.05

1.15

1.25

1.35

Small internal staff size (20) Large internal staff size (87)

Odd

s ra

tio im

pact

Combined effect of contested area andinternal staff size on the odds of outsourcing

0.250.260.270.280.290.300.310.320.330.340.35

Small internal staff size (20) Large internal staff size (87)

Pro

babi

lity

of o

utso

urci

ng

Combined effect of contested area and internal staff sizeon P (outsourcing) (baseline average probability = 0.3)

Uncontested areaContested area

Note. For internal staff size, we use the 25th and 75th percentilesof the variable as the low and high values, respectively.

of knowledge work. Although occupational HC is spe-cific to a particular profession (e.g., law, accounting), itis characteristic of general HC because it still crossesfirm and (applied) industry boundaries.

Employing the framework of firm-specific, industry-specific, and occupational HC, we examine the impactsof two main governance choices on HC-based capa-bility development, which in turn influence the focalsourcing decision. First, with respect to firm-specificand industry-specific human capital, we advance a path-dependent logic whereby prior sourcing decisions affectHC development (either at the firm or its suppliers)and thus influence future sourcing decisions. In sodoing, we build on Argyres and Liebeskind’s (1999)ideas about intertemporal “governance inseparabilities”between transactions and explore how the origins offirm capabilities—which affect current outsourcing—are rooted in the governance choices of prior relatedtransactions (Argyres and Zenger 2012). Inter alia, ourresearch underscores prior research findings that out-sourcing can decrease organizational learning (Reitzigand Wagner 2010) and create excessive dependenceon suppliers (Ring and Van de Ven 1994, Langloisand Robertson 1995). Put differently, we take adynamic approach to argue that past sourcing deci-sions may have lasting effects on firm learning andcapabilities, which then impacts future sourcing deci-sions. In this way, our research responds directly to

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge WorkOrganization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS 13

Williamson’s (1999, p. 1104) observation that “a predic-tive theory of economic organization will be enriched bymaking more prominent provision for the many ways inwhich learning influences the intertemporal governancechoice calculus.”

Second, we argue that there are systematic differencesbetween (client) firms and suppliers in how they gov-ern the employment relationship with skilled employ-ees, and these differences impact the development ofcertain types of HC-based capabilities. Whereas clientfirms have an advantage in developing firm-specific HC,suppliers have an advantage in developing occupationalHC because of their employees’ incentive structures andscope of work. Drawing on the TCE logic regardingselective intervention (Foss 2003, Williamson 1985), wemake the case that it is difficult for client firms torecreate the incentive structures and learning opportu-nities present in supplier firms with respect to occupa-tional HC. The resulting differences in firm capabilitiescan set in motion a virtuous cycle of outsourcing to sup-pliers and development of their occupational expertise.

Our empirical analyses of patent outsourcing deci-sions produced several key findings, which we discussbelow. To reinforce and complement these findings, wedraw on unstructured field interviews with retired gen-eral counsels from six different firms in our sample. Theinterviewees were asked to explain when and why theyoutsourced patent prosecution work in their companies.In two cases, the general counsels referred us to theirchief patent counsels, and for those companies, our inter-view responses below reflect primarily the opinion of thelatter. Six of these executives had built their careers aspatent law professionals, and all eight had direct expe-rience in administering and supervising the patent lawfunction at their companies during the 1990s. We con-ducted these interviews at the beginning of our researchand revisited the transcripts after conducting the empir-ical analyses.

Firm-Specific Human CapitalOne important finding of our research is that a knowl-edge project is less likely to be outsourced when itrequires higher levels of firm-specific HC, and thishuman capital has been developed within the firmthrough the internalization of prior related projects. BothTCE and KBV would argue that the need for firm-specific HC should decrease the probability of outsourc-ing; however, we develop the logic and empirical teststo show that this is only true when prior related projectsof the firm are also internalized. We find in our anal-yses that firms are actually more likely to outsourcethe focal activity if prior related projects were out-sourced, thus giving suppliers the opportunity to buildfirm-specific HC instead of the focal (client) firm. Ourinterviewees noted a number of potential obstacles to

outsourcing projects that are more firm specific, includ-ing the transaction costs of sharing such knowledge,the communication barriers across organizational bound-aries, and potential appropriability concerns. Althoughnone of our interviewees (labeled anonymously A–Ffor each company/interviewee) specifically highlightedthe need for internal firm-specific HC, their commentsappear to assume that such human capital would beavailable in-house:

[T]o educate them from the beginning the concept ofwhat you do and the prior art and all the rest of the kindsof things 0 0 0 to get somebody to come in and do that fromthe outside 0 0 0 it’s very expensive, to just come in andlearn without having to 0 0 0you know 0 0 0produce. (A)

If they have to contract out then there is another relation-ship that has to be worked out, and that’s between insideattorneys and outside attorneys, and that tends to breedsome barrier 0 0 0you don’t want your clients [firm R&D]to be talking to these outside guys and not talking withyou 0 0 0 and so on. (A)

[O]ur [internal] patent attorneys are more efficientbecause they know the client and they know the product,and they don’t have to learn things on a new job. (E)

In the high-tech business 0 0 0 if it is [the company]’s busi-ness, for example, it’s also going to be somebody else’sbusiness 0 0 0because you know who your competitors are[acknowledges potential for knowledge leakage and con-flict of interest] 0 0 0 0 Usually, we would try to keep themost important patent applications in-house. Becausethey often called for a lot of back-and-forth betweenthe engineers and the patent lawyers 0 0 0 and the businesspeople 0 0 0 0 (F)

Industry-Specific Human CapitalOur results also show that outsourcing decisions areinfluenced by how firms develop their HC in vari-ous industry-specific domains. In particular, internal-izing prior patent prosecution projects in an industry(technology) domain contributes to the firm’s industry-specific human capital and thus reduces the probabilityof outsourcing future projects in the same domain. Con-versely, our results also show that if a firm outsourcesprior patent prosecution projects in an industry domain,it may suffer from an “incompetence trap,” wherebysuppliers develop and maintain an industry-specific HCadvantage relative to the firm. Thus, initial decisionsto outsource (or not) may spawn trajectories over timethat create heterogeneous capabilities between firms andtheir suppliers. Consistent with the resource-based view,these heterogeneous capabilities would not be sustain-able without effective barriers to imitation (Barney 1991,Wernerfelt 1984), which in this case may be providedby the costs of learning by doing and time compressiondiseconomies in catching up with the knowledge fron-tier (Dierickx and Cool 1989). Two of our intervieweesin particular highlighted the role played by industry-specific HC in outsourcing decisions:

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge Work14 Organization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS

[We outsourced] 0 0 0 if it was in an area where we felt weneeded the [industry domain] expertise that we did nothave in house. Initially, we didn’t have patent attorneyswith a biotechnology background—it was a new area.We recognized very rapidly that we needed someone [apatent attorney] that had that biotech background to workin that area. [Question: So you recruited someone?] Werecruited someone, but initially we went outside to dothe work. (B)

So you may find in a large company that half the patentapplication work is being sent to outside counsel, whoin turn are people who are specialists in their techno-logical fields and they can write patent applications effi-ciently because 0 0 0presumably, if they really worked inthe same area for a long time, they really know the sub-ject matter 0 0 0 0 (F)

Oftentimes, if there is [industry domain] expertise in aparticular area in a law firm it will likely spread from oneattorney to another and so on. So there might be a groupof folks who are pretty good in a certain technology. Sothat might come about if, say, business increases from aparticular corporate client. (F)

Taken together, we believe our findings provide com-pelling initial evidence that the outsourcing of knowl-edge work is not a static process but is instead shapedby the dynamic path-dependent development of firm-and industry-specific HC resulting from prior outsourc-ing decisions (Argyres and Zenger 2012). Furthermore,consistent with KBV theories (Kogut and Zander 1992,1996; Conner and Prahalad 1996; Nickerson and Zenger2004), our results show that firms have an organizationaladvantage in coordinating knowledge activities withintheir boundaries. The effectiveness of knowledge coor-dination within the firm has the effect of amplifying theadvantages internal departments enjoy relative to exter-nal suppliers when they have strengths in at least onetype of HC. Accordingly, we find that the marginal effectof either firm-specific or industry-specific human capi-tal on sourcing decisions is greatest when the other islow. Although we did not ask our interviewees directlyabout these differences in coordination, one interviewee(B) noted that sometimes there “wasn’t any time savingat all” (for internal staff) from outsourcing because ofall the additional coordination involved. Thus, one crit-ical factor diminishing the attractiveness of outsourcingin knowledge work is the amount of firm–supplier coor-dination that is sometimes needed.

Occupational Human CapitalWe also found significant support for our hypothe-sis that firms prefer to outsource knowledge work inhighly contested areas to access the potentially supe-rior occupational HC of external suppliers. We theorizedthat suppliers accumulate superior occupational HCon account of their ability to incentivize occupationalexperts and provide occupational learning opportunities,both of which attract and motivate talented employees.

Outsourcing to external suppliers is of course costly,both in the organizational costs of coordinating withthem and in the pecuniary costs of paying for their expertservice. Although our interviewees noted these costs,they also acknowledged the value of law firm occupa-tional expertise. Consistent with our empirical results,a number of our interviewees explained that this exper-tise is an important driver of outsourcing decisions forpatents that are likely to be litigated:

[You outsource] 0 0 0 if you have a special situation 0 0 0oh,I don’t know, something related to litigated matters orsomething related to you are in litigation, and you are indevelopment within the scope of litigation 0 0 0 in general,I think it works much better inside [the company]. (A)

The reason you don’t go out to an outside patent firmunless you absolutely have to is because it’s too expen-sive. The patent lawyers, I mean they’re just voracious!Have you talked to patent lawyers? They’re absolutelyhorrible. They’re so expensive. There are few of them,their skills are much in demand, and the good firms arereally very good, and then they price accordingly 0 0 0 0 Youdo as much in house as you can. It’s cost effective. (B)

[I]t’s all based on the speculation that you may be suedor may wish to sue. And in that circumstance, why, ofcourse you go out, you hire the best you could and youpay the fees, and you’re happy that those people are avail-able to work for you, and are willing to. (B)

Well, that is a way to do it, is to have an outside law firm,but that’s a more expensive way to do it 0 0 0 0 In termsof long-range factors that is not the most economic wayto go. (C)

[G]enerally, patent attorneys in-house can be retained ata lower cost per task 0 0 0we’re not carrying the burden ofmarketing and some of the productivity and cost effec-tiveness shortfalls the law firms have, and we can producein-house cheaper because of that 0 0 0 0 (E)

That having been said, you go outside 0 0 0 for expertise 0 0 0 0There are certain things that are at a high enough levelof importance that you want to pay the extra buck to getthe best there is 0 0 0 [because] you lose a lot of other rev-enue if you get beat [in litigation], you may go outsidejust to cover yourself, to only get the best 0 0 0 0 It would belegal expertise, it would be someone who had been therebefore. (E)

Thus our empirical findings and qualitative interviewsboth suggest that, because of the need for superior occu-pational HC, outsourcing is likely when the focal activ-ity is in a highly contested area. Drawing on researchin knowledge hierarchies (Garicano 2000, Garicano andRossi-Hansberg 2006) and absorptive capacity (Cohenand Levinthal 1990), we hypothesized that firms withlarge internal staffs may be able to narrow the occu-pational HC gap with leading suppliers. Our empiricalfindings corroborate that the higher probability of out-sourcing in highly contested areas is negatively moder-ated by internal staff size.

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge WorkOrganization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS 15

Limitations and Future ResearchAlthough our research provides valuable insights intooutsourcing in knowledge-based activities, it is not with-out limitations, which may also provide valuable oppor-tunities for future research. Let us begin by noting somecharacteristics of patent legal work, which may limit thegeneralizability of our empirical findings to other typesof knowledge work. First, patents are usually prosecutedby individuals or small (two- to three-person) teams,and the final work product falls wholly within a singleprofession (patent law), so there are few challenges ofmanaging large teams of individuals, especially cross-functional teams. Second, patent work is an importantsupport activity and perhaps a valuable complementarycapability for technology firms, but it is different frommore strategic knowledge-based activities such as, say,R&D or marketing. Third, there is a vibrant market forpatent legal services arising from demand outside thefocal firms in our study, which is also the case with manyother types of knowledge work (e.g., IT, human resourcemanagement), but not always so. Moreover, our studyfocuses only on large Fortune 500 firms, and small orstart-up organizations may face very different resourceconstraints and competitive pressures, which may in turnimpact their outsourcing decisions. Thus, future researchcan build on our work by examining the extent to whichour findings apply to a range of other empirical set-tings. Finally, our research adopts a simple dichotomousapproach by treating any project as outsourced if a sup-plier was involved in it and internalized if not. In theory,a number of intermediate solutions are also feasible thatinvolve various degrees of buyer–supplier coproduction(but are less discernibly prevalent in our context; seeEndnote 5), which is also a potential topic of interest forfuture research.

Our findings help explain how capabilities maydevelop as a result of prior sourcing choices that givefirms the opportunity to develop firm- and industry-specific HC that improves their relative capabilitiesfor future transactions. However, plausible alternativeexplanations may arise if these prior human capitaldevelopment opportunities are also correlated with trans-action costs or other drivers of internalization in therelated future projects. Moreover, we need to understandmore deeply the mechanisms by which firms maximizethe learning (and thus capability development) poten-tial of these opportunities to accumulate certain types ofhuman capital. Furthermore, although our study focuseson how prior sourcing choices influence current sourc-ing decisions, the decision to outsource or not may alsoanticipate future needs in certain areas (e.g., in specificindustry domains) and include the firm’s choice to com-mit resources into these areas. We should note that clientfirms may sometimes be amenable to having supplierswho develop and maintain the needed firm- and industry-specific HC and thus provide them with a high level of

service. Our interviews also highlight a potential role forstrategic firm actions to break out of incompetence traps(e.g., the hiring of patent attorneys) or supplier actions tolure firms into one. Thus, we have highlighted a numberof areas in which our understanding of capability devel-opment through prior governance choices is limited, anda broader investigation of these questions awaits futureresearch.

With respect to occupational HC development, webuilt our arguments by relying on the differencesbetween (client) firms and suppliers in their governanceof employees and in the learning experiences they pro-vide. However, there may be limits to these mutuallyreinforcing mechanisms, and understanding how andwhen such limits operate can contribute significantly toour understanding of the evolution of vertical bound-aries in industries (Jacobides and Winter 2005). Further-more, occupational HC may be important for knowledgeprojects in many other contexts besides highly contestedareas, thus requiring outsourcing to expert suppliers. Forexample, the activity in question may involve a verycomplex problem (e.g., a difficult strategic dilemma isoften referred to a management consulting firm), or thecost of even small mistakes may be very high (e.g., amission-critical IT system is often outsourced to a lead-ing IT services firm). Future work can examine otherkinds of business problems that cannot be solved withstraightforward applications of occupational HC and sorequire the expertise of specialized suppliers. Futureresearch can also study the extent to which outsourcingto a particular supplier may influence future outcomesthrough signaling or other mechanisms. For example,does outsourcing to a particular attorney or law firm actas a signal that affects future patent citation or litigation?

Given that knowledge is substantially embeddedwithin the human capital of individuals (Felin andHesterly 2007), another useful extension of our researchis to understand how the mobility of knowledgeworkers influences the ability of firms to develop,utilize, and retain the human capital needed for knowl-edge work. Firms experiencing high employee mobil-ity may find it difficult to retain critical HC, andthey may seek to develop routines and structures toshare and embed knowledge organizationally. Con-versely, employee mobility may allow firms to acquireindividuals with certain expertise and reinforce meri-tocratic selection mechanisms that lead to HC devel-opment. We conducted a preliminary investigation ofthis issue by estimating our models with subsamples ofour data that had high (above median) and low (belowmedian) mobility, but we found no statistical differ-ences. Mobility from (all) suppliers to client firms israre in our sample, averaging at approximately 0.5 attor-neys per firm-year, which may explain why our resultshold despite such mobility. In general, understanding

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge Work16 Organization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS

the boundary conditions created for knowledge acquisi-tion and retention by the mobility of talented knowledgeworkers across firm boundaries is a promising area forfuture work.

ConclusionOverall, our findings suggest that organizations shouldcarefully consider the long-term effects of their out-sourcing decisions on the development of humancapital—learning will accrue to either the firm or a sup-plier, which will influence the firm’s future capabilitiesrelative to suppliers for similar types of projects. Theselonger-term implications for human capital developmentneed to be considered in conjunction with the short-term cost–benefit analysis for the project to fully addressthe implications for sustainable competitive advantage.At the same time, we find a distinct organizational dis-advantage for firms in developing occupational humancapital relative to their specialized suppliers. The disad-vantage stems from differences in governing and incen-tivizing skilled employees, which can be partially offsetby a large internal occupational staff that seeks to repli-cate some governance features of supplier organizations.

In conclusion, we have highlighted the role of afirm’s governance choices as not only a transaction-cost-minimizing effort but a building block for thefirm’s stock of human capital and, ultimately, capabili-ties. We built on extant typologies of human capital tofocus on three types of human capital that are centralto knowledge work—firm specific, industry specific, andoccupational—each of which poses different challengesfor firms’ capability development and sourcing choices.As our economy transitions into one that increasinglyrelies on knowledge, this emerging understanding of thenexus between capability and governance issues entailedin knowledge work will likely have important implica-tions for firm strategy, the careers of knowledge workers,the evolution of industry boundaries, and the theory ofthe firm.

AcknowledgmentsThe authors gratefully acknowledge comments from threeanonymous referees, the editors of this special issue,Benjamin Campbell, Tim Folta, Nicolai Foss, Gerry George,Royston Greenwood, Rahul Kapoor, Joe Mahoney, HughO’Neill, Jackson Nickerson, Torben Pedersen, and CoreyPhelps. They also received valuable feedback from presen-tations at the Academy of Management Annual Confer-ence (Anaheim 2008), Annual International Business ResearchForum (Philadelphia 2008), Atlanta Competitive AdvantageConference (Emory University 2008), Clifford Chance Confer-ence on Professional Service Firms (Oxford University 2008),Copenhagen Business School, Fisher School of Business (OhioState University), Imperial College Business School, Kenan-Flagler School of Business (University of North Carolina),Krannert School of Management (Purdue University), MooreSchool of Business (University of South Carolina), Richard

Ivey School of Business (University of Western Ontario),Robinson School of Business (Georgia State University), SloanIndustry Studies Conference (Boston 2008), SIEPR Semi-nar (Stanford University), and Strategic Management Soci-ety Conference (Cologne 2008). This project was supportedby research grants from the Intellectual Property ResearchInstitute of Australia and the Robert H. Smith School ofBusiness (University of Maryland). Ry D’Antonio, CristianLopez, Caitlin Mulligan, and Kiran Sheshadri provided valu-able research assistance.

Endnotes1Assuming a conservative definition of knowledge-based ser-vices as “NAICS 54: Professional, scientific, and technical ser-vices,” this sector was about two-thirds (66.4%) the size ofmanufacturing in 2008 in value added (see http://www.bea.gov/industry/gdpbyind_data.htm; accessed June 4, 2010).2These HC types have been slightly elaborated in subse-quent work to include “related industry” HC (Castanias andHelfat 2001); however, we base our approach on the originalthree types.3We focus on the breadth of firm- and industry-specific HC, butarguably, the occupational HC needed for a project may alsosometimes be narrow (e.g., JAVA coding as opposed to soft-ware engineering). However, our empirical context of patentlaw is already quite narrow in occupational terms, so we glossover this nuance.4To illustrate our logic, consider the following example. Whena firm is low in both firm- and industry-specific HC, the prob-ability of outsourcing may be 80%, but if the firm develops ahigh level of firm-specific or industry-specific HC, the proba-bility of outsourcing may drop to 35%. Then if the firm movesfrom having a high level of just one type of HC (firm specific orindustry specific) to having high levels of both types of HC, theprobability of outsourcing may drop further to 10%. The keyinsight is that the biggest marginal effect occurs when goingfrom being low in both types of HC to having high levels ofone type of HC.5We validated these data by comparing them with the firm orattorney responsible for 100 outsourced patents according tothe Patent Application Information Retrieval (PAIR) databaseof the USPTO. The PAIR database records the attorney(s) orfirm/office that currently holds the legal power of attorney torepresent the inventor at the patent office. We found that thePAIR data match the “Attorney or Agent” data in 90 of 100patents we examined. In nine other cases, the differences in thedata plausibly result from a new power of attorney or change inpatent ownership post grant. The discrepancy with one patentremains unexplained. Thus, the law firm or attorney(s) listedin the “Attorney or Agent” field appear to be responsible forpatent prosecution in most patents. Our interviews suggest thatthe firm’s own attorneys are also typically involved as bound-ary spanners, but we do not have a reliable way to measure anyadditional involvement. However, because the PAIR data indi-cate that a law firm listed on the patent is typically responsiblefor it, we conclude that the law firm is likely doing much ofthe heavy lifting on the patent.6We investigated a sample of 100 such “blank” patents in theUSPTO’s PAIR database and found that these patents wereabout as likely to be outsourced (27%) as the average patent inour sample. Because not outsourcing is the higher-probability

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge WorkOrganization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS 17

event in our sample, we conducted the sensitivity analysis bycoding “blank” patents in that category. Additionally, approx-imately 4.5% of patents have both external and internal attor-neys listed on them. Our reported results reflect our judgmentthat the internal attorneys likely acted as exchange managers,but our results are also robust to the recoding of these patents asnot outsourced (similar magnitudes, same significance levels).7The five-year and three-year capture periods produce similarresults for all variables, and the additional effects on outsourc-ing from human capital accumulated in the fourth and fifthyears was insignificant. Therefore it appears that there may bediminishing returns to older knowledge. We also found that theadditional marginal effect of domain knowledge in the broaderfour-digit IPC class (in addition to the narrower seven-digit IPCclass) was quite small, albeit significant. Therefore, the levelsof aggregation we have chosen appear to be reasonable, bothin terms of patent classes and prior years of experience, andmoreover, they are robust to changes in specification.8Although there is some variation by technology sector—from4.5 times to 8.5 times—the overall pattern is consistent acrossall technologies.9In unreported models, we included the (logged) number ofclaims in the patent, the (logged) number of backward citations,and dummy variables for patents in technological domains(seven-digit IPC classes) with no prior patenting by the firmand for patents that contained no self-citations. Our reportedresults are robust to the inclusion of these additional variables.10Our findings are robust to the use of firm fixed effects withtime-varying firm-level controls and year dummies. However,statistical support for Hypothesis 5 slips below conventionallevels in these models, which is not entirely surprising becausethe comparison group for an outsourced patent is now changedto all the patents filed by the firm in any time frame (and thussubject to much greater variation in unobserved firm-level het-erogeneity).11The 13% of the sample that was dropped either had all patentsoutsourced or all patents processed internally for that particularfirm-quarter. A very large number of firms (111) had at leastone quarter with either all patents outsourced or all internal-ized. If these quarters coincided with (unobservable) firm-levelchanges in strategy that simply happen to be correlated withour key variables, our methods essentially prevented them fromconflating our findings. Therefore, our firm-quarter fixed effectsare particularly good at controlling for time-varying within-firmheterogeneities, which would be a primary concern with firm-level fixed effects models. These dropped data are spread outmore or less evenly over the six years of our sample, with 15%–20% of dropped observations appearing in any given year.12In logit models, the effect of a change in a variable on theprobability (but not odds) of choosing an alternative dependson the values of the other variables, or equivalently, on thebaseline probability estimated from the other variables. In thegraphs that show the joint impact of variables on the probabil-ity of outsourcing, we choose the baseline probability such thatthe average probability across the combination of the interactedvalues is 0.3 (which is about that of the sample on average).It should also be noted that, unlike in regression models (or forodds ratios), the estimated probabilities at intermediate valuesof the variables do not simply lie on the straight lines con-necting the high and low levels in the bar charts. Rather, therelationship is curvilinear.

ReferencesArgote, L., P. Ingram. 2000. Knowledge transfer in organizations:

A basis for competitive advantage in firms. Organ. Behav.Human Decision Processes 82(1) 150–169.

Argyres, N. 1996. Evidence on the role of firm capabilities in verticalintegration decisions. Strategic Management J. 17(2) 129–150.

Argyres, N. S., J. P. Liebeskind. 1999. Contractual commitments, bar-gaining power and governance inseparability: Incorporating his-tory into transaction cost theory. Acad. Management Rev. 24(1)49–63.

Argyres, N., K. J. Mayer. 2007. Contract design as a firm capability:An integration of learning and transaction cost perspectives.Acad. Management Rev. 32(4) 1060–1077.

Argyres, N., T. Zenger. 2012. Capabilities, transaction costs, and firmboundaries: An integrative theory. Organ. Sci. Forthcoming.

Barney, J. 1986. Strategic factor markets: Expectations, luck, and busi-ness strategy. Management Sci. 32(10) 1231–2144.

Barney, J. 1991. Firm resources and sustained competitive advantage.J. Management 17(1) 99–121.

Becker, G. S. 1964. Human Capital. Columbia University Press,New York.

Castanias, R. P., C. E. Helfat. 1991. Managerial resources and rents.J. Management 17(1) 155–171.

Castanias, R. P., C. E. Helfat. 2001. The managerial rents model:Theory and empirical analysis. J. Management 27(6) 661–678.

Chen, M.-J., K. G. Smith, C. M. Grimm. 1992. Action characteristicsas predictors of competitive responses. Management Sci. 38(3)439–455.

Cohen, W., D. Levinthal. 1990. Absorptive capacity: A new per-spective on learning and innovation. Admin. Sci. Quart. 35(1)128–152.

Conner, K. R., C. K. Prahalad. 1996. A resource-based theory of thefirm: Knowledge versus opportunism. Organ. Sci. 7(5) 477–501.

Darr, E. D., L. Argote, D. Epple. 1995. The acquisition, transfer, anddepreciation of knowledge in service organizations: Productivityin franchises. Management Sci. 41(11) 1750–1762.

Dierickx, I., K. Cool. 1989. Asset stock accumulation and sus-tainability of competitive advantage. Management Sci. 35(12)1504–1513.

Felin, T., W. S. Hesterly. 2007. The knowledge-based view, nestedheterogeneity, and new value creation: Philosophical considera-tions on the locus of knowledge. Acad. Management Rev. 32(1)195–218.

Ferrier, W., K. Smith, C. Grimm. 1999. The role of competitive actionin market share erosion and industry dethronement: A studyof industry leaders and challenger. Acad. Management J. 42(4)372–388.

Foss, N. J. 1996. Knowledge-based approaches to the theory of thefirm: Some critical comments. Organ. Sci. 7(5) 470–476.

Foss, N. J. 2003. Selective intervention and internal hybrids: Inter-preting and learning from the rise and decline of the Oticonspaghetti organization. Organ. Sci. 14(3) 331–349.

Garicano, L. 2000. Hierarchies and the organization of knowledge inproduction. J. Political Econom. 108(5) 874–904.

Garicano, L., E. Rossi-Hansberg. 2006. The knowledge economy atthe turn of the twentieth century: The emergence of hierarchies.J. Eur. Econom. Assoc. 4(2–3) 396–403.

Gathmann, C., U. Schönberg. 2010. How general is human capital?A task-based approach. J. Labor Econom. 28(1) 1–49.

Gibbons, R., M. Waldman. 2004. Task-specific human capital. Amer.Econom. Rev. 94(2) 203–207.

Ghoshal, S., P. Moran. 1996. Bad for practice: A critique of the trans-action cost theory. Acad. Management Rev. 21(1) 13–47.

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge Work18 Organization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS

Grant, R. M. 1996. Prospering in dynamically-competitive envi-ronments: Organizational capability as knowledge integration.Organ. Sci. 7(4) 375–387.

Gulati, R. 1995. Does familiarity breed trust? The implications ofrepeated ties on contractual choice in alliances. Acad. Manage-ment J. 38(1) 85–112.

Hall, B. H., A. B. Jaffe, M. Trajtenberg. 2001. The NBER patentcitation data file: Lessons, insights and methodological tools.Working Paper 8498, National Bureau of Economic Research,Cambridge, MA.

Hatch, N., J. H. Dyer. 2004. Human capital and learning by doing asa source of sustainable competitive advantage. Strategic Man-agement J. 25(12) 1155–1178.

Hoetker, G. 2007. The use of logit and probit models in strategicmanagement research: Critical issues. Strategic Management J.28(2) 331–343.

Huang, C., T. G. Shields. 2000. Interpretation of interaction effectsin logit and probit analyses: Reconsidering the relationshipbetween registration laws, education, and voter turnout. Amer.Politics Quart. 28(1) 80–95.

Itami, H. 1997. Mobilizing Invisible Assets. Harvard University Press,Cambridge, MA.

Jacobides, M. G., S. G. Winter. 2005. The co-evolution of capabilitiesand transaction costs: Explaining the institutional structure ofproduction. Strategic Management J. 26(5) 395–413.

Jaffe, A. B., M. S. Fogarty, B. A. Banks. 1998. Evidence from patentsand patent citations on the impact of NASA and other fed-eral labs on commercial innovation. J. Indust. Econom. 46(2)183–205.

Kambourov, G., I. Manovskii. 2009. Occupational specificity ofhuman capital. Internat. Econom. Rev. 50(1) 63–115.

Ketchen D. J., Jr., C. C. Snow, V. L. Hoover. 2004. Research oncompetitive dynamics: Recent accomplishments and future chal-lenges. J. Management 30(6) 779–804.

Knott, A., D. Bryce, H. Posen. 2003. On strategic accumulation ofintangible assets. Organ. Sci. 14(2) 192–207.

Kogut, B., U. Zander. 1992. Knowledge of the firm, combinativecapabilities, and the replication of technology. Organ. Sci. 3(3)383–397.

Kogut, B., U. Zander. 1996. What firms do? Coordination, identity,and learning. Organ. Sci. 7(5) 502–518.

Langlois, R. N., P. L. Robertson. 1995. Firms, Markets and EconomicChange: A Dynamic Theory of Business Institutions. Routledge,London.

Leiblein, M. J., D. J. Miller. 2003. An empirical examination oftransaction- and firm-level influences on the vertical boundariesof the firm. Strategic Management J. 24(9) 839–859.

Lerner, J. 1995. Patenting in the shadow of competitors. J. LawEconom. 38(2) 463–495.

March, J. G. 1991. Exploration and exploitation in organizationallearning. Organ. Sci. 2(1) 71–87.

Mayer, K. J., N. S. Argyres. 2004. Learning to contract: Evi-dence from the personal computer industry. Organ. Sci. 15(4)394–410.

Mayer, K. J., J. A. Nickerson. 2005. Antecedents and perfor-mance consequences of contracting for knowledge workers:Evidence from information technology services. Organ. Sci.16(3) 225–242.

Mayer, K. J., R. Salomon. 2006. Capabilities, contractual hazards,and governance: Integrating resource-based and transaction costperspectives. Acad. Management J. 49(5) 942–959.

Nickerson, J. A., T. R. Zenger. 2004. A knowledge-based theory ofthe firm: The problem-solving perspective. Organ. Sci. 15(6)617–632.

Parmigiani, A. 2007. Why do firms both make and buy? An inves-tigation of concurrent sourcing. Strategic Management J. 28(3)285–311.

Parmigiani, A., W. Mitchell. 2009. Complementarity, capabilities, andthe boundaries of the firm: The impact of within-firm and inter-firm expertise on concurrent sourcing of complementary com-ponents. Strategic Management J. 30(10) 1065–1091.

Porter, M. E. 1980. Competitive Strategy: Techniques for AnalyzingIndustries and Competitors. Free Press, New York.

Reitzig, M., S. Wagner. 2010. The hidden costs of outsourcing:Evidence from patent data. Strategic Management J. 31(11)1183–1201.

Ring, P. S., A. H. Van de Ven. 1994. Developmental processes ofcooperative interorganizational relationships. Acad. Manage-ment Rev. 19(1) 90–118.

Rothaermel, F. T., M. A. Hitt, L. A. Jobe. 2006. Balancing verticalintegration and strategic outsourcing: Effects on product port-folios, new product success, and firm performance. StrategicManagement J. 27(11) 1033–1056.

Shaw, K. L. 1984. A formulation of the earnings function using theconcept of occupational investment. J. Human Resources 19(3)319–340.

Simon, H. A. 1991. Bounded rationality and organizational learning.Organ. Sci. 2(1) 125–134.

Somaya, D., I. O. Williamson, X. Zhang. 2007. Combining patentlaw expertise with R&D for patenting performance. Organ. Sci.18(6) 922–937.

Spender, J. C. 1996. Making knowledge the basis for a dynamic theoryof the firm. Strategic Management J. 17(Winter special issue)45–62.

Uzzi, B. 1996. The sources and consequences of embeddedness forthe economic performance of organizations: The network effect.Amer. Sociol. Rev. 61(4) 674–698.

Wernerfelt, B. 1984. A resource-based view of the firm. StrategicManagement J. 5(2) 171–181.

Williamson, O. E. 1975. The Markets and Hierarchies: Analysis andAntitrust Implications. Free Press, New York.

Williamson, O. E. 1985. The Economic Institutions of Capitalism.Free Press, New York.

Williamson, O. E. 1996. Economic organization: The case for candor.Acad. Management Rev. 21(1) 48–57.

Williamson, O. E. 1999. Strategy research: Governance and compe-tence perspectives. Strategic Management J. 20(12) 1087–1108.

Kyle J. Mayer is an associate professor of strategyand a member of the faculty at the Marshall School ofBusiness, University of Southern California. He receivedhis Ph.D. from the University of California, Berkeley.His research interests include contracting, interfirm rela-tionships, alliances, and vertical integration, combiningan economic orientation with attention to the role ofcognition.

Deepak Somaya is an assistant professor of strategyand entrepreneurship at the College of Business of theUniversity of Illinois at Urbana–Champaign. He receivedhis Ph.D. from the Walter A. Haas School of Business,University of California, Berkeley. In his research, he

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Mayer, Somaya, and Williamson: Human Capital and the Sourcing of Knowledge WorkOrganization Science, Articles in Advance, pp. 1–19, © 2012 INFORMS 19

examines how companies strategize about and derivecompetitive advantage from their knowledge assets, inparticular from their knowledge workers and intellectualproperty.

Ian O. Williamson is the Associate Dean of Inter-national Relations at the Melbourne Business School

(Australia). He currently serves as the director of theAsia Pacific Social Impact Leadership Centre and is alsoa research fellow of the Intellectual Property ResearchInstitute of Australia. His research focuses on how “talentpipelines” impact community social issues and organiza-tional outcomes.

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.