OrganizationScienceVol. 19, No. 1, January–February 2008, pp. 3–24issn 1047-7039 �eissn 1526-5455 �08 �1901 �0003

informs ®

doi 10.1287/orsc.1070.0311©2008 INFORMS

Founding the Future: Path Dependence in the Evolution ofTop Management Teams from Founding to IPO

Christine M. BeckmanPaul Merage School of Business, University of California, Irvine, Irvine, California 92697, cbeckman@uci.edu

M. Diane BurtonSloan School of Management, Massachusetts Institute of Technology, Cambridge, Massachusetts 02142, burton@mit.edu

We contrast life-cycle and path-dependent views of entrepreneurial firms by examining the evolution of top managementteams. We show how initial conditions constrain subsequent outcomes by demonstrating that the founding team’s

prior functional experiences and initial organizational functional structures predict subsequent top manager backgroundsand later functional structures. We find that narrowly experienced teams have trouble adding functional expertise not alreadyembodied in the team. We also find that firms beginning with a limited range of functional positions are less likely todevelop complete functional structures. Importantly, we do not find functional structure and functional experience to beinterchangeable. We find that firms beginning with more complete functional structures are likely to go public faster, andfirms beginning with broadly experienced team members obtain venture capital more quickly regardless of the experienceand structural composition of the top management team in place at the time of these outcomes. Further, broadly experiencedfounding teams that build an early team with a full complement of functional positions achieve important milestonesfaster than firms that start with neither experience nor structure. This suggests that creating positions as “placeholders” innew ventures, where positions are created and filled with the intent of bringing individuals with more relevant experienceonboard later, is not obviously a path by which to succeed. By examining the origins of top management team experienceand functional structures, we illustrate the lasting imprint of founders on top management team composition and firmoutcomes.

Key words : entrepreneurship; organization and management theory; organizational evolution and change; organizationaldemography

IntroductionThe popular press often portrays successful high-tech-nology firms being launched by specialized techno-logical geniuses: Brilliant scientists found a company(frequently in a garage), then attract more broadly expe-rienced executives and venture capital to bring the firmto the next level (Audia and Rider 2005). A similar viewis taken by entrepreneurship scholars who take a life-cycle perspective (e.g., Greiner 1972): Executives arereplaced as the firm outgrows their capabilities in a pro-cess of “professionalization” (Hellman and Puri 2002,Boeker and Karachalil 2002). Our research examines theextent to which these images reflect reality. There isbroad consensus that successful firms are led by sea-soned professional managers who bring a full rangeof skills to the venture (Roberts 1991, Cooper et al.1994, Burton et al. 2002) and create an organizationof functional structure with clear roles and accountabil-ity (Ancona and Caldwell 1992, Keck 1997, Roure andKeeley 1990, Jensen and Zajac 2004, Sine et al. 2006).Yet we know relatively little about how teams and firmsevolve over time.Although many pundits explicitly recommend that

entrepreneurial firms replace founders with professionalmanagers (Charan et al. 1980, Flamholtz and Randle2000, Wasserman 2003), it is less clear how a new

firm founded by narrowly experienced technologistsin an undifferentiated functional structure is able toattract broadly experienced team members and developa broad portfolio of functional positions. In fact, muchof the extant organizational literature that emphasizeshomophily (McPherson et al. 2001, Ruef et al. 2003),imprinting (Boeker 1988, Burton and Beckman 2007),and inertia (Hannan et al. 1996, Phillips 2005) suggeststhat this kind of professionalization and organizationalevolution would be both difficult and unlikely.In this paper we explore how the breadth of founder

prior experiences and early decisions about functionalstructures influence the types of executives who areattracted and retained and the types of structures that aresubsequently put into place. We contrast the life-cycleperspective with a path-dependent view of firm develop-ment. We argue that, contrary to the images of the pop-ular press and life-cycle theories of entrepreneurial firmdevelopment, homophily and imprinting operate suchthat subsequent executives and structures bear a strongresemblance to founding executives and structures. Ourresearch extends the idea that founders bring importantexperiences and make critical choices early in a firm’shistory that leave a lasting organizational imprint.Building on established traditions in top management

team demography and upper echelon literatures, we

3

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management Teams4 Organization Science 19(1), pp. 3–24, © 2008 INFORMS

examine two facets of the team: the breadth and depthof the position assignments in the new venture and theprior experiences that team members bring to the firm(Bunderson and Sutcliffe 2002). We conceptualize priorfunctional experience as a measure of the team’s humancapital (whether someone has prior sales or engineeringexperience, for example) and functional assignments asa measure of functional structure distinct from the skillsand qualifications of any specific incumbent (whether afirm has a vice president of sales or engineering, forexample).Interestingly, although the extant literature recognizes

both structural and human capital differences amongnascent ventures, they are rarely considered simultane-ously. Instead, scholars interested in functional struc-ture tend to ignore the characteristics of the individualswho occupy structural positions (e.g., Roure and Keeley1990, Sine et al. 2006), and scholars interested in teamsand individuals tend to ignore structure (e.g., Cooperet al. 1994). Indeed, in much of the top managementteam (TMT) demography research, functional structure,and experiences are considered interchangeable (e.g.,Ancona and Caldwell 1992, Keck 1997).Treating structure and experience as conceptual and

empirical equivalents may be reasonable in establishedorganizations, where we expect to see team memberswith relevant experience in a position (e.g., a person witha sales background in a sales job). In entrepreneurialventures, however, there may or may not be a matchbetween prior functional experience and current func-tional assignment. Consider the technologist with nomanagement experience who takes on the role of pres-ident, the sales person who is responsible for humanresource management or customer service, and the recentbusiness school graduate with prior engineering expe-rience who fulfills the business development or mar-keting function. This potential for a mismatch betweenprior experience and structural assignment offers acompelling arena in which to compare life-cycle andpath-dependent perspectives. Whereas life-cycle mod-els advocate that firms create positions as “placehold-ers” until executives with the relevant experience canbe hired and the firm can be professionalized (Hell-mann and Puri 2002, Boeker and Wiltbank 2005), path-dependent models would be more cautionary, given thepotential dangers of functional structures filled by in-dividuals with atypical experience (Burton and Beckman2007). In sum, we argue that functional experience andfunctional structure are distinct, and we explore theirinterrelationships and their effects on firm outcomes.

Theory and HypothesesPrior experience of the top manager and the functionalstructure of a firm have long been of interest in the TMTdemography (Pfeffer 1983) and upper echelons litera-tures (Hambrick and Mason 1984). A relatively large

body of empirical research confirms the intriguing andsensible possibility that the demographic composition ofTMTs has consequences for organizational strategy andperformance (see Finkelstein and Hambrick 1996 andWilliams and O’Reilly 1998 for reviews).A large number of studies have focused on functional

diversity and demonstrated a relationship among topmanager functional diversity and a variety of outcomesincluding firm performance, reorientation, and exter-nal communication (e.g., Eisenhardt and Schoonhoven1990, Ancona and Caldwell 1992, Lant et al. 1992).When theorists posit a positive main effect of functionaldiversity, they generally argue that diversity enhancesorganizational performance because broad functionalrepresentation ensures that the TMT has the full rangeof capabilities needed to manage the organization (e.g.,Keck 1997, Randel and Jaussi 2003). Although the main-stream demography literature has advanced to identify avariety of process variables that account for the relation-ship between diversity and outcomes (e.g., Smith et al.1994), as well as a host of factors that moderate therelationship between diversity and outcomes (e.g., Jehnet al. 1999), an important conceptual distinction betweenfunctional experience and functional structure has beenlargely ignored. The literature has generally used thesame logic to account for diverse functional structuresand diverse functional experiences, treating these dimen-sions as interchangeable and considering the choice ofmeasuring structure or experience a methodological detail(but see Bunderson and Sutcliffe 2002 for an exception).Functional structure refers to the existence of func-

tional roles or positions irrespective of any person whomight occupy the role, but functional experience refersto the human capital characteristics of the individualincumbents. This distinction is theoretically interesting,as it allows us to consider team evolution from botha structural perspective and a human capital/social psy-chological perspective. If structure and experience aredistinct and separable constructs, each should have aunique effect on firm outcomes. Structure and experi-ence may also evolve through different paths, or onemay be antecedent to the other. Our research considersthese possibilities and extends the existing empirical andtheoretical literatures by distinguishing functional expe-rience from functional structure. Our work contributesto the small but growing number of studies that are lon-gitudinal by design (Boone et al. 2004, Sørensen 2004),and so can also respond to calls to explore the sourcesof demographic distributions in organizations (Mittman1992, Lawrence 1997, Carroll and Harrison 1998) andaddress why TMTs “look the way they do” (Hambrick2007, p. 338).We address Hambrick’s call by examining the role of

the founding team in creating the experiences and struc-ture of the subsequent TMT. If path dependence occurs,then initial functional structures will predict a deepening

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management TeamsOrganization Science 19(1), pp. 3–24, © 2008 INFORMS 5

of these structures over time. If homophily occurs, theninitial functional experiences will predict a deepening ofthese experiences over time. If functional structure andfunctional experience are interchangeable, then havingeither will enable a firm to broaden as it evolves. Further-more, we examine whether these initial characteristics ofthe team and firm have lasting consequences. If, as life-cycle theories predict, initial functional structures andexperiences are irrelevant to subsequent changes, thenthere should be little relationship between the foundingteam experiences and structures and subsequent experi-ences and structures, particularly as founders are replacedand firms mature. Thus, in stark contrast to a path-dependent perspective, where origins are a source of sub-sequent constraint, the life-cycle perspective emphasizesopportunity and adaptation.In summary, we consider five interrelated questions:

(1) whether the functional experience of the foundersshapes initial decisions about functional structure,(2) whether the initial functional structure constrainsfuture functional structure, (3) whether the initial func-tional structure shapes the functional experience of exec-utives hired by the firm, (4) whether the functionalexperience of the founders influences the experiences ofexecutives who join subsequently; and (5) whether theinitial functional structure and experiences shape firmoutcomes.To examine these questions, it is important to first dis-

tinguish between the founding team and TMT. Foundingteams are made up of people who create the firm, irre-spective of whether they hold executive titles. For exam-ple, a founder may hold a position of “director of engi-neering” rather than vice president. TMTs are made upof people who hold executive titles, regardless of whenthey join the firm. Although there is overlap betweenthese teams at founding if all founders hold executivetitles, the teams are conceptually and empirically dis-tinct. Founders are not added as the firm grows, and,as TMT members join the firm and founders leave, theteams continue to diverge.Our first question about the relationship between prior

functional background experience and initial functionalstructural decisions draws on two ideas—that func-tional background experiences shape a person’s worldview and that people carry their prior experiences withthem across organizational settings. We rely on earlystatements and recent empirical evidence suggestingthat functional training conditions individual cognitions(Dearborn and Simon 1958, Sutcliffe 1994, Tripsas andGavetti 2000) and an established empirical traditiondemonstrating that when employees leave an organiza-tion, they carry that organization’s routines and ideaswith them (Baty et al. 1971, Boeker 1997, Sørensen1999, Phillips 2002). For example, the functional expe-rience and strategy of the founder’s prior firm influ-ences the strategy of the new firm (Boeker 1988), and

founders’ experiences in a prior firm shape the routinesand practices they put in place in a new firm (Phillips2005, Beckman 2006). In addition, there is growing evi-dence that the functional background experiences of aTMT influence how it defines and enacts organizationalpositions (Fligstein 1990, Phillips 2005). Extending thisidea to initial decisions about the allocation of positionsand responsibilities, we hypothesize that founders willput functional structures in place that mirror their ownexperiences (Schein 1992, Burton and Beckman 2007).For example, a founding team with a member from asales and business planning background may decide tocombine the sales and business planning functions undera single executive and build an organization where thesefunctions are closely interrelated and externally (cus-tomer) oriented. The background of the initial marketingexecutive—business development, classical marketing,or sales—leads to different choices about what the mar-keting function does and with which other functions itis aligned. Thus we hypothesize:

Hypothesis 1 (H1). Founding team functional expe-rience will shape initial functional structure.

Once the initial functional structures are put intoplace in an organization, they are likely to be main-tained and strengthened over time. Support for this argu-ment comes from several theoretical traditions. Socialtheorists have long argued that institutions are self-reinforcing (Michels 1915, Downs 1967). In addition,theories of organizational imprinting suggest that initialdecisions about the allocation of responsibility amongteam members have lasting consequences (Stinchcombe1965, Baron et al. 1996, Hannan et al. 1996, Burtonet al. 2002). For example, Baron et al. (1999) findthat founding organizational blueprints, around whichemployment relations are managed, predict subsequentlyadopted decisions and structures (including adding newfunctional positions; creating formal documentation andorganizational charts). Finally, arguments of structuralinertia suggest that initial structures will be maintainedover time (Hannan and Freeman 1984). Thus, a rangeof work suggests that initial functional structures predictlater functional structures. We therefore hypothesize:

Hypothesis 2 (H2). Founding functional structurewill shape the breadth and depth of subsequent func-tional structure.

Founding functional structure will also likely shapesubsequent management functional experience. Exten-sive literature on job analysis—which in modern humanresource management is the foundation of job descrip-tions, job design and redesign, performance appraisal,selection, and training—defines jobs as combinations oftasks that require particular knowledge, skills, and abili-ties that can be generically specified (cf. Fine and Cron-shaw 1999). Indeed, the extensive literature on organi-zational design, formalization, and even internal labor

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management Teams6 Organization Science 19(1), pp. 3–24, © 2008 INFORMS

markets has a vision of clearly defined and hierarchi-cally ordered positions or roles (see Osterman and Bur-ton 2004 for a recent review). Jobs exist in organiza-tions as “placeholders” against which individuals in thelabor market can be matched (e.g., Wanous 1992; seeMiner 1987 for a review of this literature and an alter-native view). Similarly, these characterizations of roles,jobs, and positions imply that the organizational func-tional structure shapes the characteristics of subsequentincumbents.Consider again the firm in which sales and business

planning positions are closely aligned. When a newexecutive is recruited to take over the combined position,the structure of the position will influence the choiceof executive for the position. A different founding teammight decide to create stand-alone sales and businessplanning functions, which would lead to different futurecandidates. Thus, the structural choices made by thefounding team have implications for the characteristicsof the subsequent executives who would be capable offilling the defined positions, attracted to the opportu-nity, and attractive to the incumbent team members. Ourarguments imply that structural decisions made by firmfounders constrain the pool of people who might be will-ing or able to join the firm. Thus we hypothesize:

Hypothesis 3 (H3). Founding functional structurewill shape the breadth and depth of subsequent func-tional experience.

Finally, there are reasons to expect an associa-tion between the prior functional experiences of thefounding team and subsequent TMT experience. Sev-eral mechanisms (which may operate simultaneously)could generate this path dependence. For instance, theattraction-selection-attrition (ASA) cycle described bySchneider (1987) would predict that founding and futureTMTs will share similar characteristics. Managers seekorganizations where existing personnel have similar char-acteristics, founders select managers like themselves, andmanagers who do not fit will leave. For example, Booneet al. (2004) find that even when environmental conditionsshifted dramatically, Dutch newspaper executives tendedto hire new executives who were demographically similarto those already in place. Thus, founders and managersalike should be attracted to one another when they sharecommon experiences and knowledge.A large literature on homophily (Rogers and Bhowmik

1971, Ruef et al. 2003) also suggests this relationship.Although research on homophily has generally focusedon categories such as race and gender, there is grow-ing evidence that the same processes apply to occupa-tions (McPherson et al. 2001). Founders may privilegeand recruit executives with functional experiencessimilar to their own rather than executives with func-tional experience that may be more relevant for the posi-tion being filled. For example, engineers might prefera CEO with a technical background over one with a

finance background. It is important to note that these ten-dencies toward functional homophily within a positionmay result in team-level functional heterogeneity. Teamswill be functionally diverse if, for instance, individ-ual preferences for homophily exist on founding teamswhere members come from a variety of functional back-grounds. Alternately, founders with diverse functionalbackgrounds (functional generalists) may value that vari-ety and seek TMT members who themselves have diverseexperiences. Thus, at the team level, homophily andsimilarity-attraction arguments imply similarity betweenthe experiences of the founding team and the experiencesof the TMT. We therefore hypothesize:

Hypothesis 4 (H4). Founding team functional expe-rience will shape the breadth and depth of subsequentTMT functional experience.

Following a path-dependent logic, our hypotheses sug-gest a tight linkage between initial team experiences/structures and subsequent TMT experiences/structures.Importantly, this implies there will be relatively littlechange in functional structure or human capital experi-ence from the founding team to later teams. Thus, insharp contrast to the dominant image of narrowly experi-enced engineers in an unstructured nascent venture whoare replaced by professionals who evolve the firm into aprofessional bureaucracy, our hypotheses imply that ini-tial founding team characteristics and structural choicesshould be a powerful predictor of later experiences andstructures—even controlling for compositional changes.Contrary to a life-cycle perspective that would be agnos-tic about initial conditions, our theoretical argumentsimply that initial conditions restrict subsequent firm andteam options.To further understand the lasting consequences of

these initial conditions, we also directly link found-ing experiences and structures to firm outcomes. Boththe entrepreneurship and TMT literatures offer ampleevidence linking human capital characteristics such asthe type and amount of prior experience to firm suc-cess (Aldrich and Zimmer 1986, Roberts 1991, Cooperet al. 1994). Other work finds that a range of priorexperiences as well as shared experiences benefit thefirm (Eisenhardt and Schoonhoven 1990, Beckman et al.2007). Still other work links the functional structure ofthe firm to firm outcomes (Ancona and Caldwell 1992,Keck 1997, Roure and Keeley 1990). More recently,Sine et al. (2006) demonstrate that formalized functionalpositions and functional specialization is positively asso-ciated with firm success in a sample of Internet start-ups.Together these studies document that a broad set of bothteam functional experiences and functional structures arepredictive of firm outcomes. Thus, we hypothesize:

Hypothesis 5 (H5). The breadth of founding teamfunctional experiences and functional structures are pos-itively associated with firm success.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management TeamsOrganization Science 19(1), pp. 3–24, © 2008 INFORMS 7

Interestingly, the existing literature is silent as towhether structure might be more important to firm suc-cess than experience or vice versa. If the structure andexperience are distinct conceptual constructs, then theyshould have independent effects. If, as the extant demog-raphy literature has implied, structure and experience areinterchangeable representations of the same underlyingconstruct, then having either broad functional experiencerepresented among the founding team members or a dif-ferentiated functional structure should be sufficient tofuel the development of a firm with a broadly experiencedTMT and a differentiated functional structure. However,if structure and experience are not interchangeable, thenan increase on one dimension cannot compensate for aweakness on another dimension. Our analyses explorethese alternatives.If, as we posit, the founding conditions set teams and

firms on a particular path, then cross-sectional studiesof TMTs overattribute outcomes to the current team andoverestimate the extent to which teams can be reformed.We pursue a research strategy that begins with the found-ing team and then tracks changes in personnel—boththrough entrances and exits—and in functional compo-sition to explore exactly how the founding team andinitial structures shape the subsequent TMT and laterstructures.

Data and MethodsThis paper begins with a sample of entrepreneurial high-technology firms in California’s Silicon Valley that werestudied as part of the Stanford Project on EmergingCompanies (SPEC) (see Baron et al. 1999 for a detaileddescription) but includes additional data on individualexecutives. The focus on firms in a single region allowsus to hold constant key labor market and environmentalconditions. Within the region, the SPEC study focuseson the high-technology industries of computer hardwareand/or software, telecommunications, medical devicesand biotechnology, semiconductors, manufacturing, andresearch. The sampling frame explicitly oversamplesyoung and small firms. The firms in the sample haveat least 10 employees and are no more than 10 years oldat the time data collection begins in 1994–1996. Abouthalf of the firms are founded before 1989, and thus themedian age of the firms in 1994 is five years. Our sampleis not representative of all start-up firms in that we havefewer solo founders and more teams of three or morefounders, but this is typical of high-technology start-ups(Eisenhardt and Schoonhoven 1990, Ruef et al. 2003).Our sampling frame has the disadvantage of biasing thesample toward firms that have survived several years andthus are more likely to be (or become) successful. Thatsaid, we sample firms that both do and do not receiveventure funding and go public. We see no a priori rea-son to expect that our hypothesized effects, where TMTsare constrained and shaped by founding teams, will bedifferent for a broader sample of start-up firms.

The key independent and dependent variables for thisstudy are constructed from career histories. We constructa monthly database of every founder and every execu-tive that ever worked for one of the sampled firms fromfounding through December 2000 or the time of initialpublic offering (IPO), acquisition, or failure (see below).The founding team was identified in an interview withthe founder (the mean founding team has 2.82 mem-bers), and all subsequent TMT members are identified asthose individuals ranked as vice president or higher (e.g.,senior vice president, CTO, CIO, COO). It is impor-tant to note that founders do not always hold TMTtitles (as evidenced by an average TMT smaller than thefounding team at time zero). Our data sources includeinterviews (conducted in 1994–1995 and 1996–1997),internal company documents (business plans and pro-motional documents), Lexis/Nexis news searches, DowJones Interactive, Edgar Archives (useful for firms aboutto go public and for top managers who have beeninvolved with public companies), The San Jose MercuryNews (the local paper has a regular column on promo-tions, movements, and resignations in the Silicon Val-ley), and extensive Web searches. Over a six-year periodwe completed at least four complete searches for eachperson and spent thousands of person-hours collectingcareer data on team members. We confirmed, via inter-view or telephone, nearly 50% of the career histories col-lected through 1996–1997 with the person designated bythe CEO or the human resources person. This increasesthe reliability of the earliest team data, the most difficultperiod to gather consistent data through archival sources.Despite extensive research, we do not obtain complete

career histories on all team members. Often the chronol-ogy of careers is correct but the dates unclear. This dataproblem precludes us from constructing duration vari-ables such as years of experience in a particular function,a method commonly pursued by demography scholars.Because some of our team data may be incomplete, wemake a point to control for variables that may impact thecompleteness of the data (i.e., firm size) as well as theaverage amount of person data collected by firm.Our final database contains information on 1,485

executives in 167 firms holding 1,940 positions in oursampled firms. We collect a median of two past posi-tions for each person, including employer identity andjob title, with a maximum of 19 positions for a sin-gle person, although we limit our analysis to the priorthree employers. This includes data for executives (oftenfounders) who we confirm had no prior work experience(at least 38 founders join one of our firms directly fromschool). For Hypotheses 1–4 and our analysis of teamevolution we rely on our full sample (167 firms). Forour analysis of outcomes, firms that are not independent(e.g., wholly owned subsidiaries) or at risk of going pub-lic (e.g., nonprofit research centers) are excluded (result-ing in a sample of 158 firms for our outcome analysis).

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management Teams8 Organization Science 19(1), pp. 3–24, © 2008 INFORMS

Dependent VariablesIn this research we examine three types of dependentvariables: (1) functional organizational structure, (2) teammember functional experience, and (3) firm out-comes. The difference between functional structureand team functional experience mirrors the differencebetween functional assignment diversity and func-tional background diversity in the demography literature(Bunderson and Sutcliffe 2002).Functional organizational structure is measured as

whether the firm has defined executive-level posi-tions in each of the following six functional areas:sales and marketing, general administration (includ-ing human resources), science/R&D/engineering, oper-ations, business development/strategic planning, andfinance/accounting. For each firm, we calculate whethera firm has a given functional position, when it is firstestablished, and how many executive-level positions ina given function simultaneously exist in a team for eachmonth of the firm’s life. For example, a technology-intensive firm may be founded with a CTO and a vicepresident of engineering. The vice president of engineer-ing job is later transformed into the position of seniorvice president of engineering, and a vice president ofhardware and a vice president of software are hired. Thisfirm clearly has a science/R&D/engineering function. Itis established at age 0 and there are ultimately fourexecutive-level positions in this function on the TMT.We also create a count measure ranging from 0 to 6

that represents the number of functional categories cov-ered each month to predict the breadth of functionalstructure. Firms with no executive positions are coded 0for structure; firms with executive positions for all sixfunctions are coded six.Prior functional experience of TMT members is gath-

ered from career histories. We code up to three priorpositions for every individual into one of the six func-tional areas (see above). From team-level career historieswe calculate a variable, updated monthly, indicating howmany team members have prior experience in a functionto capture the depth of functional experience on a team.We also create a count measure ranging from 0 to 6

that represents the range of prior functional experiencebrought by the team in each month and that thus cap-tures the breadth of functional experience. A six indi-cates that all possible functions are represented in theprior experiences of the team.We have two firm outcome measures: time to receive

venture capital (VC) and time to IPO, and we con-duct event-history analyses to examine these outcomes.These firm outcomes represent the most significant mile-stones in the life of a young start-up (Shane and Stuart2002), particularly during the time period and in theregion that we study. We focus on time to first VC fund-ing rather than cumulative rounds or total funds raised,

because future rounds are based on more direct knowl-edge about the firm than the first VC financing, andamount is a firm and industry-specific choice (Gompersand Lerner 1999). By modeling time to first VC fund-ing, we examine characteristics of the team that allowedthe firm to obtain funding more quickly than other firms,as well as whether they receive VC backing. By choos-ing IPO as our second dependent variable, we can com-pare the performance of entrepreneurial firms acrossmultiple industries—a task that is quite difficult usingaccounting-based measures of profitability. Recent stud-ies have examined IPO as an outcome variable indicatingfirm success (Stuart et al. 1999, Certo et al. 2001, Gulatiand Higgins 2003, Shane and Stuart 2002, Hannan et al.2006). This IPO measure also allows us to examine afirm outcome that occurs over a longer time horizonthan initial VC funding. VC financing and IPO data arecollected via public and proprietary databases (such asVenture One and Venture Economics), government fil-ings, annual reports, internal company documents, and asurvey instrument sent to the most senior finance exec-utive in the firm (see Hellmann and Puri 2002).Of our 158 firms at risk of IPO or VC, 26 exit the

risk set through IPO, acquisition, merger, or failure with-out securing VC, and another 14 do not receive VCby the end of 2000. Thus, 118 firms (71%) obtain VCfunding and 87 (51%) go public during our samplingperiod. At first glance these numbers seem unusuallyhigh; however, the end of our time period (late 1990s)witnessed an explosion of VC investments and IPOs inentrepreneurial firms in the United States.

Independent and Control VariablesOur key independent variables are the founding team’sfunctional experiences and initial functional structures.These measures are created like the dependent vari-ables but focus on the team and functional struc-tures at founding. That is, we code the breadth anddepth of initial functional structure and founding teamfunctional experience in each of the following six func-tional areas: sales and marketing, general administration(including human resources), science/R&D/engineering,operations, business development/strategic planning, andfinance/accounting. At founding the mean number ofunique functional positions, our measure of structure,is 1.8, which typically represents a firm with a generaladministrative position such as president or CEO and avice president of science or engineering. The mean num-ber of prior functional experiences at founding is 1.5.

Firm-Level Controls. Firm size is the number of em-ployees at the start of a calendar year, lagged by oneyear. Because of the skewed distribution of this mea-sure, we use a logged measure in our analyses. Wecontrol for whether the firm had an innovation strat-egy, because strategy has been linked to VC financingin prior research (Burton et al. 2002). We also control

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management TeamsOrganization Science 19(1), pp. 3–24, © 2008 INFORMS 9

for whether the firm sought to differentiate itself throughits sales, marketing, or service, as this strategy impliesa greater need for the sales and marketing function thanother strategies. Initial strategy is coded from interviewswith the founder and has been empirically validated (seeHellmann and Puri 2000). For example, firms with inno-vation strategies are those that described themselves as“first movers.” In addition, we control for the date thefirm shipped its first product, because product shipmentmay affect the skills and structures required as well asfirm outcomes. We also include the cumulative num-ber of rounds of VC funding that the firm had obtainedbecause such firms are more likely to go public (Gom-pers and Lerner 1999), but we obtain similar resultswhen we use a simple dummy variable (e.g., equal to1 when the firm obtains VC). Finally, we control forthe amount of firm-level team data collected by includ-ing the average number of prior positions collected foreach person in the firm. This allows us to control forthe possibility that we have more data on individuals insuccessful firms.

Team-Level Controls. We include founding team sizeand then include measures for the cumulative entrancesand exits to the TMT as controls. These variables cap-ture both change and growth in the teams (Tushmanand Rosenkopf 1996, Beckman et al. 2007). A life-cycleperspective suggests that growth is the critical determi-nant of success, and these entrances and exits wouldbe expected to wipe out any effects of origin. Further,these controls ensure that our effects for experience andstructure are net effects of turnover. The proportion ofTMT positions held by founders is included to controlfor the extent to which founders with executive titles arecounted as part of both teams. All team variables areupdated monthly.

Industry-Level Controls. Medical devices and the bio-technology industry (combined as medically related) isthe only significantly different industry for our out-comes; thus, we include a dummy variable for this indus-try in all models. The number of IPOs in each industryby year controls for industry-specific variation in ratesof IPO, and the number of VC deals by year controlsfor available financial resources.

ResultsDescriptive statistics and a correlation matrix for thefull sample are presented in the appendix. To exam-ine the influence of the founding team’s prior func-tional experience on the initial functional structure (H1),we conduct an event history analysis predicting timeto the first executive-level position in a functionalarea using Cox proportional hazards models with theHuber/White/sandwich estimator of error and clusteredby firm to correct for heteroskedasticity and serial corre-lations. We regress the time to the first position for five

different functions on the prior functional experience ofthe founding team. The key independent variable is howmany members of the founding team had prior experi-ence in the function. (We exclude general administrationbecause almost all firms start with a president or CEO.)We report these results in Table 1. For each func-

tional area we generate a baseline model that includescontrol variables only and then a test model that addsa variable indicating the amount of relevant prior func-tional experience on the founding team. The number ofexits indicates the number of firms that ever create anexecutive-level position in the function reported. Duringour observation period, 75% of the firms create a posi-tion in the science/engineering function (125 of 167);only half ever create an executive position responsiblefor manufacturing/operations (82 of 167), and fewer thanone-fourth create an executive position in business devel-opment or strategic planning (42 of 167). Firms thathave a function at founding—in other words, a memberof the founding team creates the position—are censoredimmediately. Our models therefore capture the rate offunctional position creation for those firms that do andthat do not have the relevant prior functional experience.Table 1 reports hazard ratios and reveals that larger

firms with larger founding teams, and those that obtainVC financing, create functional positions at a higher rate.For two of our five functional areas, science/engineeringand sales/marketing, we find that having functional expe-rience represented on the founding team significantlyincreases the rate at which the structural position isadded to the TMT. Founding teams with sales experi-ence are 69% more likely to add a sales position. Found-ing teams with science experience are 44% more likelyto add a science position. The coefficients are in thepredicted direction for both finance and strategic plan-ning/business development, although they do not achievestatistical significance. Based on these results, we findmodest support for H1.To examine the relationships between initial func-

tional experience and initial functional structure on sub-sequent team functional experience and later functionalstructures (H2, H3, and H4), we perform yearly panel-poisson regression analyses predicting counts of expe-rience and structure for each function. We specify anAR1 error structure to account for serial autocorrelationand again report robust standard errors. These results arepresented in Table 2.Table 2 is organized into two panels with yearly team-

level observations. The first panel presents our results forTMT functional structure. The second panel presents ourresults for TMT prior experience. The first six modelsin each panel represent each of the functional areas ofinterest and are counts of the depth or number of func-tional positions or experience in each area. The finalmodel in each panel is a cumulative count variable that

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management Teams10 Organization Science 19(1), pp. 3–24, © 2008 INFORMS

Table 1 Event History Analysis Predicting Time to First Functional Position

(1) (2) (3) (4) (5)Science/ Sales/ Mfg./ Finance/ Strategy/

Engineering Marketing Operations Acct. Bus. Dev.

Log firm size 1�20∗∗ 1�29∗∗ 1�25∗ 1�08 1�49∗∗∗�0�09� �0�13� �0�14� �0�11� �0�23�

Medical-related industry 0�98 0�28∗∗∗ 0�9 1�21 2�05∗�0�21� �0�11� �0�32� �0�30� �0�85�

Innovation strategy 1�26 1�14 1�23 0�97 1�46�0�22� �0�27� �0�38� �0�22� �0�70�

Marketing strategy 0�8 0�84 1�81∗ 0�9 1�59�0�23� �0�24� �0�62� �0�24� �0�76�

Cumulative rounds of VC 1�49∗∗∗ 1�29∗∗∗ 1�26∗∗∗ 1�21∗∗∗ 1�19∗�0�14� �0�10� �0�08� �0�08� �0�12�

Prop. founders on TMT 0�92 0�88 0�82 0�59∗∗ 0�95�0�07� �0�12� �0�19� �0�13� �0�28�

Product developed 1�01 0�77 0�50∗∗ 1�18 0�55∗�0�23� �0�17� �0�15� �0�27� �0�20�

Average number of prior person positions 1�06 0�91 0�97 0�88 1�2�0�08� �0�09� �0�09� �0�09� �0�16�

Founding team (FT) size 1�17∗∗∗ 1�16∗∗ 0�99 1�18∗ 1�1�0�06� �0�09� �0�09� �0�10� �0�15�

Cumulative entrances to the TMT 1�12 1�25∗∗ 1�1 1�02 1�05�0�13� �0�14� �0�11� �0�09� �0�17�

Cumulative exits from the TMT 0�82 0�87 0�9 0�79 0�85�0�23� �0�20� �0�19� �0�23� �0�17�

FT prior science/engineering exp. 1�44∗∗∗�0�11�

FT prior sales/marketing exp. 1�69∗∗∗�0�24�

FT prior mfg/operations exp. 0�76�0�31�

FT prior finance/accounting exp. 1�4�0�38�

FT prior strategy/business dev. exp. 1�63�0�97�

Observations 6,054 7,343 9,762 8,756 12,562Exits 125 121 82 114 42Firms 167 167 167 167 167Log-pseudolikelihood −558�62 −515�27 −367�73 −498�34 −187�38Note. Robust standard errors.∗Significant at 10%; ∗∗significant at 5%; ∗∗∗significant at 1%.

ranges from 0 to 6 to capture the breadth of functionalpositions or experiences.We test our second hypothesis, that founding func-

tional structure will be associated with subsequentfunctional structure, in the first panel. Across all speci-fications, there is a strong positive relationship betweenthe founding structure and the subsequent structure, evencontrolling for whether the founding team has experi-ence in a given functional area. Model 7 in Panel 1presents the count variable predicting the number offunctional positions eventually covered by the firm.Consistent with the results for each function, firms thatbegin with more initial structure are likely to cover morefunctional areas over time. Thus, while the foundingteam prior experiences shape initial structures (in mod-est support of H1 in Table 1), it is these initial structures

that shape subsequent structures (in strong support of H2in Panel 1 of Table 2).We test the impact of initial structure and founding

team prior experience on later TMT prior experiencein Panel 2 of Table 2 (H3 and H4). Contrary to H3,initial functional structures do not serve as placehold-ers to facilitate bringing in later executives with rele-vant functional experience. Indeed, the only statisticallysignificant finding (for sales/marketing positions) is theopposite of this prediction.However, the pattern of results revealed in Panel 2 of

Table 2 shows support for H4. There is a strong rela-tionship between the functional background experienceof the founding team and that of the later TMT priorexperience, even controlling for changes in team compo-sition. The relationship is consistently revealed for eachof the functional areas (Models 1–6) and in the cumu-

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management TeamsOrganization Science 19(1), pp. 3–24, © 2008 INFORMS 11

Table

2Pan

elPoisso

nwithRobu

stStandardErrors,withAR1ErrorStructure,PredictingTMTExp

erience

andStructure

Panel1:T

MTfunctionalstru

cture

Panel2:T

MTpriorexpe

rience

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Science/

Sales/

Mfg./

Finance/

Adm

in/

Strategy/

Num

.of

Science/

Sales/

Mfg./

Finance/

Adm

in/

Strategy/

Num

.of

Engin.

Mktg.

Ops.

Acct.

HR

Bus.d

ev.

functions

Engin.

Mktg.

Ops.

Acct.

HR

Bus.d

ev.Priorexp.

FTpriorscienceexp.

0�07

0�43

∗∗∗

�0�06�

�0�06�

FTsciencepo

sition

0�33

∗∗∗

−0�04

�0�06�

�0�06�

FTpriorsalesexp.

0�29

∗∗∗

0�54

∗∗∗

�0�10�

�0�09�

FTsales/mkt.p

osition

0�15

∗−0

�17∗

∗

�0�07�

�0�08�

FTpriormfg./ops.e

xp.

−0�05

1�49

∗∗∗

�0�23�

�0�21�

FTmfg./ops.p

osition

0�96

∗∗∗

0�17

�0�15�

�0�20�

FTpriorfin./acct.exp.

−0�03

0�89

∗∗∗

�0�21�

�0�24�

FTfin./acct.po

sition

0�87

∗∗∗

0�18

�0�15�

�0�21�

FTpriorad

min./H

Rexp.

0�05

0�47

∗∗∗

�0�06�

�0�13�

FTad

min./H

Rpo

sition

0�22

∗∗∗

−0�06

�0�03�

�0�08�

FTpriorbu

siness

dev.exp.

−0�11

1�24

∗∗∗

�0�51�

�0�31�

FTbu

siness

dev.po

sition

0�71

∗∗0�05

�0�28�

�0�57�

FTpriorexpe

rience

−0�00

0�14

∗∗∗

count

�0�02�

�0�03�

FTstructure

0�11

∗∗∗

0�01

count

�0�02�

�0�03�

Log-firm

size

0�07

∗∗∗

0�07

∗∗∗

0�11

∗∗0�05

∗0�03

∗∗0�15

∗0�03

∗∗∗

0�01

−0�01

−0�01

0�09

∗0

0�03

0�01

�0�02�

�0�03�

�0�04�

�0�03�

�0�01�

�0�09�

�0�01�

�0�02�

�0�03�

�0�05�

�0�05�

�0�02�

�0�07�

�0�01�

Med

ical-related

0�53

∗∗∗

−0�37

0�35

0�45

∗∗∗

0�22

∗∗1�17

∗∗∗

0�14

∗∗0�19

−0�46∗

0�42

0�27

−0�03

0�56

∗0�11

∗

indu

stry

�0�14�

�0�34�

�0�25�

�0�15�

�0�10�

�0�39�

�0�07�

�0�14�

�0�27�

�0�26�

�0�24�

�0�24�

�0�33�

�0�07�

Innovationstrategy

0�1

−0�24∗

0�07

−0�25∗

0�08

0�26

−0�03

0�1

−0�1

0�42

∗0

0�19

0�22

0�06

�0�13�

�0�13�

�0�19�

�0�14�

�0�07�

�0�41�

�0�05�

�0�11�

�0�13�

�0�24�

�0�18�

�0�17�

�0�37�

�0�06�

Marketingstrategy

0�1

−0�17

0�31

−0�03

0�20

∗∗0�44

−0�02

0�17

0�09

0�39

0�19

0�13

0�12

0�08

�0�16�

�0�17�

�0�20�

�0�16�

�0�09�

�0�45�

�0�05�

�0�13�

�0�15�

�0�31�

�0�22�

�0�20�

�0�48�

�0�07�

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management Teams12 Organization Science 19(1), pp. 3–24, © 2008 INFORMS

Table

2(cont’d

.)

Panel1:T

MTfunctionalstru

cture

Panel2:T

MTpriorexpe

rience

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Science/

Sales/

Mfg./

Finance/

Adm

in/

Strategy/

Num

.of

Science/

Sales/

Mfg./

Finance/

Adm

in/

Strategy/

Num

.of

Engin.

Mktg.

Ops.

Acct.

HR

Bus.d

ev.

functions

Engin.

Mktg.

Ops.

Acct.

HR

Bus.d

ev.

Priorexp.

Cum

ulativeVC

0�01

0�04

∗∗0�03

0�06

∗∗0�01

0�04

0�02

∗∗0�02

0�07

∗∗∗

0�08

∗∗0�08

∗∗∗

0�04

0�01

0�04

∗∗∗

rounds

�0�02�

�0�02�

�0�02�

�0�02�

�0�01�

�0�05�

�0�01�

�0�01�

�0�02�

�0�04�

�0�03�

�0�03�

�0�05�

�0�01�

Prop

.found

ers

0�17

−0�72∗

∗∗−0

�64∗

∗−1

�31∗

∗∗0�31

∗∗∗

−0�88

−0�37∗

∗∗−0

�13

−0�39

−0�91∗

∗∗−1

�49∗

∗∗−0

�66∗

∗∗−0

�49

−0�44∗

∗∗

onTM

T�0�23�

�0�20�

�0�27�

�0�30�

�0�09�

�0�61�

�0�08�

�0�18�

�0�25�

�0�33�

�0�36�

�0�25�

�0�49�

�0�11�

Avg

num.p

rior

0�02

−0�25∗

∗∗−0

�1−0

�01

−0�02

−0�02

−0�03

0�17

∗∗∗

0�20

∗∗∗

0�28

∗∗∗

0�18

∗∗0�08

0�31

∗∗0�14

∗∗∗

pers.p

ositions

�0�07�

�0�06�

�0�09�

�0�07�

�0�03�

�0�16�

�0�02�

�0�04�

�0�06�

�0�09�

�0�08�

�0�06�

�0�12�

�0�03�

Founding

team

size

0�08

∗∗0�06

−0�09

0�06

0�01

0�04

0�02

∗−0

�03

0�08

∗0�08

0�1

0�1

0�02

0�03

�0�04�

�0�05�

�0�06�

�0�05�

�0�02�

�0�09�

�0�01�

�0�04�

�0�04�

�0�09�

�0�07�

�0�06�

�0�16�

�0�02�

Cum

ulativeTM

T0�16

∗∗∗

0�19

∗∗∗

0�17

∗∗∗

0�13

∗∗∗

0�09

∗∗∗

0�20

∗∗∗

0�09

∗∗∗

0�15

∗∗∗

0�24

∗∗∗

0�21

∗∗∗

0�16

∗∗∗

0�15

∗∗∗

0�25

∗∗∗

0�13

∗∗∗

entra

nces

�0�03�

�0�03�

�0�03�

�0�03�

�0�01�

�0�06�

�0�01�

�0�02�

�0�03�

�0�03�

�0�04�

�0�03�

�0�05�

�0�02�

Cum

ulativeTM

T−0

�22∗

∗∗−0

�22∗

∗∗−0

�26∗

∗∗−0

�20∗

∗∗−0

�09∗

∗∗−0

�35∗

∗∗−0

�15∗

∗∗−0

�16∗

∗∗−0

�22∗

∗∗−0

�17∗

∗∗−0

�27∗

∗∗−0

�13∗

∗∗−0

�29∗

∗∗−0

�15∗

∗∗

exits

�0�03�

�0�03�

�0�05�

�0�04�

�0�02�

�0�10�

�0�02�

�0�03�

�0�05�

�0�04�

�0�05�

�0�05�

�0�08�

�0�02�

Prod

uctd

evelop

ed−0

�29∗

−0�26

−0�35∗

∗−0

�19

0�08

−0�46

−0�10∗

−0�33∗

∗−0

�11

0�17

−0�53∗

∗0�17

0�04

−0�03

�0�16�

�0�17�

�0�17�

�0�15�

�0�07�

�0�48�

�0�06�

�0�14�

�0�14�

�0�27�

�0�25�

�0�16�

�0�35�

�0�06�

Constant

−1�65∗

∗∗−0

�37

−1�42∗

∗∗−1

�01∗

∗∗−0

�50∗

∗∗−3

�15∗

∗∗0�86

∗∗∗

−1�26∗

∗∗−1

�97∗

∗∗−3

�36∗

∗∗−2

�12∗

∗∗−1

�51∗

∗∗−3

�62∗

∗∗0�00

�0�33�

�0�26�

�0�36�

�0�33�

�0�14�

�0�70�

�0�10�

�0�23�

�0�29�

�0�51�

�0�38�

�0�31�

�0�76�

�0�15�

Observations

1,256

1,256

1,256

1,256

1,256

1,256

1,256

1,256

1,256

1,256

1,256

1,256

1,256

1,256

Num

berof

firms

167

167

167

167

167

167

167

167

167

167

167

167

167

167

Waldchi-squ

are

359.44

437.37

290.98

170.94

380.4

193.19

502.59

288.2

352.42

220.31

157.88

152

129.33

507.41

Difference

inchi-squ

are

120.22

12.4

106.78

−18�7

89.59

74.33

23.99

123.7

75.42

103.99

5.61

−3�87

6.97

111.08

from

baseline

Note.

Rob

uststand

arderrors

inbrackets.

∗ Significanta

t10%

;∗∗ significanta

t5%;∗

∗∗sign

ificant

at1%

.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management TeamsOrganization Science 19(1), pp. 3–24, © 2008 INFORMS 13

lative count measure (Model 7). It is important to notethat only the parallel structure or experience is predic-tive. In supplementary analyses we confirmed there wereno effects for the noncorresponding functions. In otherwords, sales experience does not predict the establish-ment of a science position; it is only science experiencethat is significantly associated with science functionalstructures and subsequent science experience. We presentthe more parsimonious models in the paper for purposesof clarity.Across most specifications we see that large firms,

and firms where teams are growing, have more broadlyexperienced TMTs and more functional positions. Thissuggests that firms are growing and becoming moreexperienced over time, which is consistent with a life-cycle perspective. Further, firms with VC funding oftengain TMT experience and functional structures. How-ever, our findings imply that structures and experiencesaccrue more to those firms that are already advantaged.Overall, our findings for structural precedent (we findsupport for H2 but not for H3) are weaker than our find-ings for cognitive framing (H1) and homophily (H4). Yetwe see that initial functional structures matter (H2).Finally, to test H5 regarding firm outcomes, in Table 3

we conduct event history analyses on monthly observa-tions and report Cox proportional hazards models usingmaximum likelihood estimation and robust estimates ofstandard error (Lin and Wei 1989). The founding yearis represented as age= 0 with all the initial conditionsrepresented as covariates that are updated where appro-priate. Firms remain in the sample until they achieve theoutcome of interest; until they cease to exist as indepen-dent entities through failure, merger, or acquisition; oruntil the end of the sample period, at which time theyare censored.We present our results in two panels in Table 3 to

demonstrate the influence of founding team on the tim-ing of two different outcomes: obtaining VC and com-pleting an IPO. Model 1 is the baseline model thatincludes controls only. Note that team compositionalchanges are accounted for by the inclusion of measuresof both entrances to and exits from the TMT. Model 2examines founding team variables and includes countmeasures for both the amount of initial functional struc-ture in the firm and the amount of prior functional expe-rience possessed by the founders. In Model 3 we alsoinclude the interaction of these terms. The main effectsare centered prior to calculating the interaction (as sug-gested by Cronbach 1987) to reduce collinearity betweenthe main effects and the interaction term. Model 4 addsthe same count variables and their interaction for theTMT team to assess whether the effects of the foundingteam are eliminated when the subsequent team is con-sidered. It is in this model that we most strongly assesspath dependence against life cycles in determining therelationship between team and firm characteristics and

outcomes. As in Table 1, we report hazard ratios, sonumbers larger than 1 indicate an increase in the rate,and numbers smaller than 1 indicate a decrease in rateof achieving the firm outcomes.In the first panel, the VC analysis, we see that firm

size and the average amount of data collected for teammembers are both positively associated with attainingVC financing. In addition, TMT departures are posi-tively related to obtaining financing. This may indicatethat firms replace executives to build a team attractiveto venture capitalists. In Model 2 we find that eachadditional functional area represented in the prefound-ing career history increases the likelihood of attractingVC financing by 14%. The effect of functional struc-ture is also positive but not statistically significant. Inter-estingly, in Model 3, the interaction of structure andexperience is negative and statistically significant. Theinteraction effect should be interpreted as the effect ofexperience on the outcome when structure is 0. Theinfluence of founding functional experience on the timeto VC financing is weaker for firms that begin with nofunctional structure; they have a 13% lower likelihoodof receiving VC. Not surprisingly, the founding teameffects are attenuated when we include characteristics ofthe subsequent team in Model 4 (although the interactionremains significant) because, as we know from Table 2,founding team experience also predicts TMT experience.The strong correlation between the founding and TMTcomposition in the early years makes Model 4 difficultto interpret; however, the pattern of results suggest thatthe range of prior experience held by the founding teamis an important correlate of VC financing, and firms thatstart with both experience and structure reach this mile-stone faster than other firms.In Panel 2 of Table 3 we present a similar set of anal-

yses for time to IPO. The control variables have largereffects on the rates of IPO. Firms in medical-relatedindustries go public at least five times more quickly thanfirms in other industries, and firms in industries with ahigh number of IPOs go public more quickly than firmsin other industries. In addition, we see that the effectof turnover has two countervailing effects on rates ofIPO: firms benefit from entrances and are hindered byexits. These effects, while consistent with the life-cycleperspective, do not wipe out the founder effects, as seenin Model 2. Thus, origins do matter for outcomes. Incontrast to our VC findings, Model 2 in Panel 2 revealsthat functional structure is associated with an increasedlikelihood of IPO, and prior experience is not. In Model3 we see that the interaction of structure and experienceis negative and statistically significant, suggesting thatstructure cannot substitute for a lack of experience. Theinfluence of founding functional structure on the time toIPO is weaker for firms that begin with no prior func-tional experience; they have a 23% lower likelihood of

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management Teams14 Organization Science 19(1), pp. 3–24, © 2008 INFORMS

Table 3 Event History Analysis: Effect of Founding Team on Firm Outcomes

Panel 1: VC Panel 2: IPO

(1) (2) (3) (4) (1) (2) (3) (4)

Firm size 1�65∗∗∗ 1�65∗∗∗ 1�70∗∗∗ 1�61∗∗∗ 1�29 1�16 1�19 1�18�0�16� �0�16� �0�17� �0�17� �0�22� �0�21� �0�21� �0�23�

Medical-related industry 1�14 1�3 1�37 1�38 5�46∗∗∗ 6�97∗∗∗ 8�63∗∗∗ 7�10∗∗∗

�0�27� �0�31� �0�33� �0�36� �1�86� �2�52� �2�92� �2�65�

Innovation strategy 1�29 1�33 1�36 1�38 0�74 0�7 0�68 0�77�0�30� �0�31� �0�32� �0�34� �0�20� �0�19� �0�19� �0�23�

Marketing strategy 0�8 0�82 0�78 0�79 1�11 1�03 0�93 0�9�0�23� �0�24� �0�23� �0�24� �0�39� �0�35� �0�31� �0�30�

VC deals/IPOs in industry 1 1 1 1 1�18∗∗∗ 1�19∗∗∗ 1�18∗∗∗ 1�18∗∗∗

�0�00� �0�00� �0�00� �0�00� �0�03� �0�03� �0�03� �0�03�

Cumulative VC 1�12∗∗ 1�14∗∗∗ 1�14∗∗∗ 1�08�0�06� �0�06� �0�05� �0�06�

Avg. person positions 1�19∗∗ 1�19∗∗ 1�18∗∗ 1�19∗∗ 2�21∗∗∗ 2�30∗∗∗ 2�46∗∗∗ 2�58∗∗∗

�0�09� �0�09� �0�09� �0�09� �0�31� �0�33� �0�35� �0�40�

Founding team (FT) size 1�02 0�99 1 1�01 1�13 1�02 1�09 1�03�0�05� �0�05� �0�05� �0�05� �0�09� �0�09� �0�10� �0�10�

Product developed 0�91 0�94 0�89 0�89 0�66 0�69 0�71 0�60∗

�0�20� �0�21� �0�20� �0�21� �0�18� �0�19� �0�20� �0�18�

Prop. founders on TMT 0�95 0�97 0�94 0�97 0�54 0�59 0�53 0�77�0�10� �0�11� �0�12� �0�12� �0�22� �0�25� �0�24� �0�33�

Cumulative TMT entrances 0�86 0�81 0�83 0�66∗∗ 1�63∗∗∗ 1�66∗∗∗ 1�68∗∗∗ 1�43∗∗

�0�13� �0�13� �0�13� �0�12� �0�16� �0�17� �0�18� �0�21�

Cumulative TMT exits 1�71∗ 1�63∗ 1�80∗∗ 2�75∗∗∗ 0�66∗∗∗ 0�65∗∗∗ 0�63∗∗∗ 0�75∗

�0�48� �0�46� �0�50� �1�00� �0�06� �0�07� �0�07� �0�11�

FT structure count 1�07 1�11 0�87 1�41∗∗∗ 1�50∗∗∗ 1�45∗∗∗

�0�13� �0�13� �0�18� �0�21� �0�21� �0�20�

FT experience count 1�14∗ 1�19∗ 0�95 0�94 1�05 1�0�11� �0�11� �0�13� �0�12� �0�15� �0�15�

FT structure ∗experience 0�87∗ 0�79∗∗ 0�77∗∗∗ 0�78∗∗∗

�0�07� �0�09� �0�07� �0�07�

TMT structure count 1�4 1�78∗∗∗

�0�29� �0�33�

TMT experience count 1�53∗∗∗ 1�33�0�24� �0�24�

TMT structure ∗experience 1�05 0�89�0�12� �0�08�

Observations 6,107 6,107 6,107 6,107 13,085 13,085 13,085 13,085Log-pseudolikelihood −516�91 −515�43 −513�86 −506�96 −290�02 −287�6 −284�12 −276�69Exits 118 118 118 118 86 86 86 86Firms 158 158 158 158 158 158 158 158

Note. Robust standard errors.∗Significant at 10%; ∗∗significant at 5%; ∗∗∗significant at 1%.

going public. The effects of the founding structure per-sist through the inclusion of contemporary structure (seeModel 4). Structure may be an important signal to publicinvestors.In general, our findings suggest that venture capitalists

are more concerned with the prior experience of execu-tives in firms that they fund, whereas the public marketsevaluate functional structure. Both milestones are moreeasily achieved by the firms that start with both structureand experience. A key conclusion from the analyses in

Table 3 is that although the impact of the TMT tends tobe stronger than the founding team on these outcomes,founding team effects persist, controlling for changes tothe team.

Supplementary AnalysesThe above findings begin to establish organizationalfunctional structure and incumbent prior functionalexperience as distinct constructs and demonstrate sup-port for our main theoretical argument that initial con-

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management TeamsOrganization Science 19(1), pp. 3–24, © 2008 INFORMS 15

Table 4 Descriptive Statistics for Founding Teams Initial Types

(1)Simple structure (2)(narrow range of Simple structure, (3)

functional positions), broadly Complete structurenarrowly experienced team (broad range of (4)

experienced team (broad range of functional Complete (4) (4) (3) (2)Full (limited range of prior prior functional positions), narrowly structure, broadly vs. vs. vs. vs.

Row Variable sample functional experiences) experiences) experienced team experienced team (1) (3) (1) (1)

1 Number of firms 167 50 14 44 592 Mean founding 2�82 2�16 2�71 2�82 3�41 17�57∗∗∗ 3�65∗∗ 4�23∗∗ n.s.

team size3 Cumulative entrances 5�50 4�30 5�64 5�61 6�39 10�32∗∗∗ n.s. 3�53∗ n.s.4 Cumulative exits 1�68 1�48 1�93 1�73 1�76 n.s. n.s. n.s. n.s.5 Proportion of founders 0�55 0�60 0�50 0�56 0�52 n.s. n.s. n.s. n.s.

on TMT6 Mean TMT size 4�42 3�24 4�43 4�32 5�49 20�02∗∗∗ 5�06∗∗ 3�97∗∗ n.s.7 TMT experience 2�78 2�12 3�36 2�61 3�32 14�56∗∗∗ 4�71∗∗ n.s. 6�23∗∗∗8 TMT functions 3�27 2�56 3�64 3�55 3�58 9�21∗∗∗ n.s. 7�49∗∗∗ 4�22∗∗9 Proportion VC backed 0�71 0�58 0�71 0�68 0�83 8�42∗∗∗ 2�76∗ n.s. n.s.10 Proportion that 0�51 0�42 0�43 0�52 0�61 3�93∗∗ n.s. n.s. n.s.

achieve IPO11 Firm age at end of 7�00 9�00 7�50 7�00 6�00 16�58∗∗∗ n.s. 5�58∗∗∗ 2�94∗

observation period

Notes. Two-tailed tests for F -statistic. We do not include comparisons between cells 2 and 3 and cells 2 and 4 because there are notsignificant differences. ∗p < 0�10; ∗∗p < 0�05; ∗∗∗p < 0�01.

ditions constrain subsequent evolution. However, ourhypothesis tests do not allow us to fully understand theinterrelationships between structure and experience overtime. The strong support for the effects of initial expe-rience on subsequent experience and for the effects ofinitial structure on subsequent structure contrasted withthe weaker or nonexistent support for the cross effects.This suggests that structure and experience are not sub-stitutable. Having one does not appear to compensate forthe other.To explore the interrelationships of structure, expe-

rience, and outcomes, we first descriptively character-ize the team and firm evolutionary patterns revealed inour data (see Table 4). The first column (Full sample)presents overall means for all firms. We then divideour sample, using mean splits on counts of foundingteam functional experience and counts of initial struc-tural positions, into four subsamples representing eachof the archetype cells described in Figure 1. Cell 1 mostclosely captures the mythical “engineers in a garage”as it describes a firm with a narrow range of functionalpositions and functional expertise. In contrast, Cell 4most closely captures the professional team in a func-tional organization. The off-diagonals, Cells 2 and 3,represent situations where the founding team has eitherstructure or experience, but not both.Several points are noteworthy. First, Figure 1 and

Table 4 demonstrate that experience and structure areclearly distinct concepts, and structure and prior experi-ence do not always go hand-in-hand. The least commonstarting state is represented by Cell 2—simple struc-ture, but broadly experienced executives—which char-acterizes 14 of 167 firms; however, firms are relatively

dispersed across the other cells. Firms with neither struc-ture nor experience at founding (50 of 167 firms repre-sented in Cell 1) appear to be disadvantaged on manyfronts: They start with smaller teams, do not grow asmuch, and are less likely to receive VC or to go publicthan firms with the most broadly experienced teams withcomplete structures. (See Table 4, Rows 2, 6, 9, and 10.)Table 4 also lends additional support for H5, that thecombination of initial background experience and func-tional structure is important for firm success. Firms inCell 4 seem to achieve critical milestones more quickly

Figure 1 Archetype Founding Team Types

Cell 2 Cell 4

Cell 1 Cell 3

Organizational functionalstructure

Low

Low

High

High

Priorfunctionalexperience

Complete structureand

broadly experiencedteam

Simple structureand

broadly experiencedteam

Complete structureand

narrowly experiencedteam

Simple structureand

narrowly experiencedteam

General partners

Engineers ina garage

Business plancompetition entrants

Functionalorganization

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management Teams16 Organization Science 19(1), pp. 3–24, © 2008 INFORMS

than firms in Cell 1. In addition, there is suggestive evi-dence that narrowly experienced teams are not able toaccrue needed additional functional expertise as easily asunstructured teams acquire structure. That is, comparingnarrowly and broadly experienced teams at each levelof structure (Cells 1 and 2; Cells 3 and 4), it appearsthat, regardless of the starting structure, narrowly experi-enced founding teams have significantly less broad TMTexperience at the end of the sample period (see Table 4,Row 7). In contrast, when firms have broadly experi-enced teams, the firms that lack initial functional struc-tures (Cell 2) have the same level of structure at the endof the time period as those that started with more com-plete functional structures (Cell 4) (see Table 4, Row 8).Thus, those teams with early functional experience candevelop functional structure, but those teams with earlyfunctional structure do not develop functional experi-ence. Finally, it is worth noting that executives exit alltypes of firms at the same rate (see “Cumulative exits,”Row 4 in Table 4). There are no significant differencesacross the four starting states, suggesting there is not aclear signal about “winning” firms.In Table 5 we present monthly panel regression mod-

els, again correcting for autocorrelation and with robuststandard errors, where we attempt to discern the relation-ship between origin state and ultimate destination. Giventhe evidence that successful firms have broadly expe-rienced TMTs and complete functional structures, weare particularly interested in understanding the path bywhich firms end up in this state. Recall our count mea-sures for breadth of experience and structure that rangefrom 0 to 6. If we consider these as interval representa-tions along two continua, then the product of experienceand structure, an interaction variable with a theoreticalrange of 0 to 36, is a linear representation of the com-bined amount of structure and experience. The largerthe product of TMT structure and TMT experience, thecloser the firm is to having a complete structure andall functions represented in the prior experiences of theTMT (i.e., Cell 4). We treat this product as our depen-dent variable in a series of regression models wherethe key independent variables are initial structure, initialexperience, and the interaction of the two.In Table 5, we report a baseline model, Model 1, with

controls only. Models 2 and 3 add founder functionalexperience and initial functional structure, respectively,to the model. Model 4 includes both key independentvariables. Model 5 includes the interaction of these twocounts to assess the relative importance of each. Thisallows us to examine the extent to which experienceand structure can serve as substitutes and whether oneis more important than the other in influencing firmand team evolution. Consistent with life-cycle predic-tions of professionalization, firm size, VC, and entrancesto the TMT are positively associated with having abroadly experienced team and complete structure, and

TMT entrances is one of the largest effects in the model.Yet net of all of these effects, we see that foundingteam experience and initial structure are important corre-lates of the final destination state. Comparing Models 2and 3, we see that founder functional experience aloneimproves model fit more than initial functional structurealone. Thus, it appears that initial experience is a moresubstantial predictor than initial structure. But Model 4demonstrates a significant joint effect, and Model 5reveals a slight positive interaction. This is evidence ofthe advantages associated with starting with both struc-ture and experience (Cell 4) at founding; however, thestrength of the experience main effect and the weak sta-tistical significance attached to the interaction imply thatCell 2 is a better origin state than Cell 3.We explored several alternative explanations of our

findings. One additional explanation is that a matchbetween initial and subsequent functional experience iscreated through preferences for experienced executives.Although all firms may seek to hire broadly experiencedexecutives, it is only firms with broadly experiencedfounding team members that are able to attract the exec-utive talent necessary for future success. Although sim-ilar to homophily, it would be a result of only broadlyexperienced founders being successful at recruitment.Although it is difficult to empirically disentangle thesemechanisms, it is worth noting that all varieties of found-ing team experiences influence TMT experience, regard-less of whether the experience is particularly relevantfor the firm’s strategy. This gives more credence to thehomophily explanation.Another alternative explanation is that “good” firms

attract both broadly experienced founding teams andTMTs rather than founding teams predicting TMTs. Thisconcern of unobserved heterogeneity is a common issuein empirical work, and we have done what we canto explore this possibility—that people are attracted tofirms and not to other people. This would imply thatrather than initial experience and structure increasing theability of the firm to reach important milestones, expe-rienced TMT members may be drawn to firms that looklike winners. Thus, the causality could go the other way:Good firms attract good people rather than good peoplecreating good firms. Although we cannot rule out thispossibility, we do control for several measures of firmquality. In addition to controlling for size and strategy,we are also including whether the firm had a product asa control. Also, recall that the exit rates and proportionof founders on the TMT across our four cells in Table 4are not significantly different. On average, 1.7 executivesleave the firm and founders account for more than 50%of the TMT. This suggests that firm quality is difficultto observe; otherwise, executives and founders wouldleave low-quality firms at higher rates. Thus, it seemslikely that prospective TMT members have an easiertime assessing the quality of the existing team than the

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management TeamsOrganization Science 19(1), pp. 3–24, © 2008 INFORMS 17

Table 5 Population-Averaged Panel Regression Analysis Predicting TMT Structure ∗Experience, with RobustStandard Errors and AR1 Error Structure

(1) (2) (3) (4) (5)

Log firm size 0�33∗∗∗ 0�34∗∗∗ 0�33∗∗∗ 0�34∗∗∗ 0�34∗∗∗

�0�06� �0�06� �0�06� �0�06� �0�06�

Medical-related industry −1�03 −0�35 0�14 0�18 −0�03�1�11� �1�05� �1�10� �1�07� �1�03�

Innovation strategy 1�33 1�40 1�44 1�45 1�44�0�87� �0�81� �0�83� �0�80� �0�79�

Marketing strategy 0�39 0�49 0�12 0�32 0�59�1�05� �0�94� �1�02� �0�94� �0�93�

Cumulative rounds of venture capital 0�35∗∗∗ 0�37∗∗∗ 0�36∗∗∗ 0�37∗∗∗ 0�38∗∗∗

�0�06� �0�06� �0�06� �0�06� �0�06�

Prop. founders on TMT −0�09∗∗∗ −0�09∗∗ −0�09∗∗ −0�09∗∗∗ −0�09∗∗∗

�0�03� �0�03� �0�03� �0�03� �0�03�

Product developed 0�21 0�23 0�22 0�23 0�23�0�12� �0�12� �0�12� �0�12� �0�12�

Average number of prior person positions 0�12 0�12 0�13 0�12 0�12�0�08� �0�08� �0�08� �0�08� �0�08�

Founding team (FT) size 0�66∗∗ 0�39∗ 0�28 0�23 0�17�0�27� �0�19� �0�23� �0�19� �0�18�

Cumulative entrances to the TMT 0�16∗∗∗ 0�17∗∗∗ 0�16∗∗∗ 0�17∗∗∗ 0�17∗∗∗

�0�05� �0�05� �0�05� �0�05� �0�05�

Cumulative exits from the TMT 0�08 0�07 0�08 0�07 0�06�0�12� �0�12� �0�12� �0�13� �0�13�

FT prior experience count 1�90∗∗∗ 1�52∗∗∗ 1�29∗∗∗

�0�37� �0�42� �0�43�

FT structure count 1�80∗∗∗ 1�02∗∗ 0�86∗

�0�45� �0�48� �0�46�

FT exp. ∗ structure 0�70∗

�0�38�

Constant 2�16∗∗ 2�61∗∗∗ 3�02∗∗∗ 3�00∗∗∗ 2�82∗∗∗

�0�85� �0�71� �0�81� �0�71� �0�69�

Observations 14,269 14,269 14,269 14,269 14,269Chi-square 131.58 183.36 160.08 192.54 192.27

Note. Robust standard errors.∗Significant at 10%; ∗∗significant at 5%; ∗∗∗significant at 1%; two-tailed tests.

firm itself because the quality of the firm (independentof people) is hard to assess. However, there may also bean “escalation of commitment” (Staw 1976), wherebysubsequent team members remain with their firm despitelower potential returns. Of course, the entrance rates aresignificantly different across these cells. The foundingteams with both experience and structure grow morerapidly than all other firms, but it appears these changesin composition are driven by founding team composi-tion. Still, understanding how people evaluate prospec-tive employment opportunities in entrepreneurial firmsis an important area for future research.Because we were able to observe firms for a rel-

atively short period of time, we are limited in ourability to assess if and when founder imprints mightdecay. In supplementary analyses we explore the extentto which our results are driven by having relativelyyoung firms. We first replicated all our time-series anal-yses, including yearly dummy variables to control for

the main effects of firm age and found results consis-tent with those reported. We also replicated the analysesin Table 2, Model 7, predicting both TMT functionalstructure and TMT experience, including interactionswith initial experience and structure and firm age. Wefind statistically significant age interactions; however,the magnitude of the effects is trivial. We thus havesome confidence that our results hold for many yearspost-founding.We were concerned about several additional factors

that could have influenced our results. First, our effectsmay be skewed by the bubble market of the late 1990s.In supplementary analyses, we examine the possibilitythat our results arise from period effects rather than ourtheorized mechanisms. Our results are robust to restrict-ing our VC financing events prior to 1993 (before themarket began to heat up) and our IPO events prior to1996 (the first year that IPO rates jump significantly).

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management Teams18 Organization Science 19(1), pp. 3–24, © 2008 INFORMS

Firms may add positions to the TMT just prior toreceiving external financing or going public to signalreadiness to potential investors. We control for thispotential “window dressing” by reporting team entrancesand exits, and proportion founders on the TMT, lagged12 months in all outcome analyses. Thus, changes madewithin a year of the event were not considered. We foundsubstantially the same results with shorter and longerlags, thus we feel comfortable using a conservative one-year window. However, because we have announcementdates and not dates when these discussions commenced,we cannot completely rule out reverse causality.Finally, we were concerned that the time-series data

may overemphasize the within-firm linkage between thefounding team and TMT. In addition to using robuststandard errors and correcting for autocorrelation in thereported models, we also ran a series of panel regres-sions. In these models we used founding team expe-rience and structure to predict TMT experience andstructure (as in Table 2) at a given firm age (e.g., at year4 and at year 8), and we found similar results.

LimitationsWe have taken a first step in understanding TMT evo-lution with rich longitudinal data. Of course, our studyhas several limitations. First, as noted above, we observefirms for a relatively short period of time and are thuslimited in our ability to assess the durability of foundereffects over long periods. A second weakness is that ourdata for both executives and firms are somewhat incom-plete. We do not have the kind of detailed performanceoutcomes that are available for established public com-panies. In addition, we may have been unable to find dataon TMT members who were not successful during theircareer. However, ours is the first study to our knowledgeto attempt such a detailed look at the career histories ofexecutives in small firms over time, and such an examina-tion almost by definition involves problems with missingdata. Furthermore, we have controlled for the potentialproblem to the extent possible in all analysis.A third limitation is that our sample has a success

bias because the firms survived on average five yearsbefore we contacted them. We believe, however, that indemonstrating an empirical phenomenon in an interest-ing empirical setting and establishing a new model oflongitudinal research, the benefits of the sample far out-weigh the concerns. Finally, there are questions of unob-served heterogeneity that cannot be ruled out, althoughwe explore them to the extent possible in the above sup-plementary analyses.

DiscussionOur research demonstrates that founding teams stronglyinfluence the TMT through path dependence. Consis-tent with homophily expectations, founding teams thatbegin with broadly experienced team members are more

likely to attract broadly experienced executives. Consis-tent with ecological research, firms that begin with arange of functional structures are more likely to developmore complete functional structures. Thus, path depen-dence, where the founding team shapes the subsequentTMT, occurs through both homophily and imprinting.We see cumulative advantage, as high-quality found-

ing teams become high-quality TMTs and less well-endowed founding teams never catch up. In particular,we see that 26% of our teams begin with broadfunctional structures but narrowly experienced people(Table 4), and they never catch up in terms of attracting abroad range of functional experience. Firms founded bynarrowly experienced founders have difficulty attractingbroadly experienced executives. It does not seem to bethe case that narrowly experienced founders simply failto recognize the importance of other types of functionalexpertise. Instead, even when the firm has created anexecutive level position for a given function—a strongstatement that it needs and values that expertise—thefirm is limited in its ability to fill the position with therelevant expertise if it does not already have an executivein place with at least some experience in that function.This suggests that structure is a poor substitute for expe-rience. Thus, our results stand in sharp contrast to thedominant image of engineers in an unstructured nascentventure who are replaced by broadly experienced profes-sionals who then formalize the structures and evolve thefirm into a professional bureaucracy. Our results implythat the narrowly experienced “garage” entrepreneur isnot likely to succeed (Audia and Rider 2005).Despite the significant findings, two of our hypothe-

ses receive modest or no support. We find only modestsupport that initial functional experience predicts theinitial functional structure. Given the wealth of otherresearch documenting this transfer (Boeker 1988; Burtonand Beckman 2007; Phillips 2002, 2005), it is possiblethat our measures of functional structure do not ade-quately capture the nuance of how executives enact orga-nizational positions. In addition, path dependence mayoperate more strongly through homophily in the cases ofrace or gender (see Phillips 2005). This would be a use-ful area for future research, as initial functional structureis an important decision that founders make.Our lack of support for functional structures predict-

ing experience (H3) is quite relevant for human resourcemanagement. At least in entrepreneurial firms, the struc-ture of the position does not predict the experiencepeople bring to the firm, once controlling for initialexperience. Instead, it appears that the ASA frame-work (Schneider 1987) and structural inertia (Hannanand Freeman 1984) together account for the relation-ship between early structure and experience and laterstructure and experience. As noted above, establishingpositions and filling them with people who are “place-holders” appears to be detrimental to the firm’s abilityto ultimately attract broadly experienced executives.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management TeamsOrganization Science 19(1), pp. 3–24, © 2008 INFORMS 19

Little research has examined changes in teams overtime, much less from founding, and a major contributionof this paper is the detailing of the relationship betweenthe founding team and TMT. We demonstrate that thefounding team exerts an influence on the firm not onlythrough directly influencing firm outcomes, but also byshaping the very nature of the organization. An ahistoricalaccount of the TMT will overstate the ability of the TMTto change; in fact, future teams may be best understoodby a detailed examination of the teams that have comebefore. This finding limits the role of agency and suggestssharp deviations from the initial path are unlikely (andprobably risky, see Hannan et al. 2006). Our research thusfurther highlights the benefits of an evolutionary perspec-tive on firms and teams (see Aldrich 1999).We clearly see evidence of path dependence in our

analysis, and these effects are net of changes and growthamong the firms. The significant influence of teamentrances in many of our analyses, however, also offerssupport for a life-cycle model of entrepreneurial firmdevelopment. Growth does allow firms to add both func-tional experience and structure, and these effects are sub-stantive in our models. However, our intent here is todemonstrate that the initial conditions do matter and notall firms easily professionalize and grow. Particularly inlight of our lack of support for H3—initial structure doesnot predict attracting that functional experience to thefirm—firms should be cautious in assumptions of addingexperience later. In contrast, our findings suggest deliber-ate planning into the future is usefully done at founding.Our results also add to the organizational demogra-

phy and upper echelon research, which has not to datefocused on developing dynamic models or explainingthe source of demographic distributions in organizations(Lawrence 1997, Hambrick 2007). Our work begins toaddress this weakness in the demographic approach byexploring the path-dependent process that results in con-tinued heterogeneity or homogeneity among teams. Thisis not to say that organizational demography research hasnot moved in important directions. To the contrary, criti-cal work examines the team processes generated by diver-sity (Knight et al. 1999, Pelled et al. 1999, Reagans andZuckerman 2001) and the difficulties of using indirectmeasures of demographic composition rather than directmeasures of social networks to predict firm performance(Reagans et al. 2004). Still, scholars rarely conduct longi-tudinal studies (see Boone et al. 2004 and Sørensen 2004for recent exceptions). More importantly, scholars havedevoted little attention to the sources of demographicdiversity. In addition, we point to a distinction betweenfunctional experience and functional structure that hasreceived little attention in this literature; thus, our paperfills several important gaps in the literature.For the entrepreneurial literature, our research sug-

gests that the relevance of founding teams is more signif-icant than has been acknowledged. Not only do founding

teams directly impact firm outcomes, but, through a pro-cess of path dependence, the founding team shapes theTMT. Entrepreneurship research often focuses on theindividual entrepreneur, but understanding the teams thatcome together and develop over time is essential tounderstanding the performance of entrepreneurial firms.We advocate more studies with longitudinal data as wellas a focus on functional structure in addition to teamhuman capital. Also important, but largely unexploredin this paper, is the impact of VC in shaping the TMT.Our findings offer suggestive evidence that VC-backedfirms are better able to add structure and experience thannon-VC-backed firms; however, the precise mechanismsare worthy of additional research.Much more remains to be understood about entrepre-

neurial teams. How does a team attract founders withvaried experiences when we know diversity is atypical(Ruef et al. 2003)? Given the long-term effects of theseinitial choices, such an exploration would be very useful.We know founding is not truly the beginning, becauseentrepreneurs bring experience and networks with them(Burton et al. 2002, Shane and Stuart 2002). A qualitativeassessment of how people, ideas, experiences, and struc-tures come together to create a firm would be an impor-tant contribution of future work. In addition, delving intoindividual executive positions in the organization wouldhelp us understand in more detail the mechanisms thatresult in path dependence (Burton and Beckman 2007).Despite the remaining questions, we are encouraged

by the consistency of our results, how they contribute toand reflect current theories, and by the potential rewardsof examining teams over time in this rich research set-ting of entrepreneurial firms. We demonstrate that found-ing teams matter—both directly and indirectly—largelythrough a process of path dependence. This sociologi-cal and evolutionary approach demonstrates how initialteams have a lasting impact on the firm.

AcknowledgmentsBoth authors contributed equally to this manuscript. Theauthors thank Greg Northcraft for his insights and contribu-tion to this manuscript. The authors also thank Roberto Fer-nandez, Mauro Guillen, Nandini Rajagopalan, Jesper Sørensen,Ezra Zuckerman, seminar participants at University of Califor-nia, Berkeley, Harvard Business School, Massachusetts Insti-tute of Technology, University of California, Los Angeles, Uni-versity of California Irvine, Wharton, and attendees at the WestCoast Entrepreneurial Research Conference for constructivecomments on earlier drafts of this paper. An earlier version ofthis paper was presented at the Academy of Management Meet-ing in 2006. This research was supported by Stanford Project onEmerging Companies (SPEC) at the Graduate School of Busi-ness, the Harvard Business School Division of Research, UCIrvine, and the MIT Entrepreneurship Center. The authors alsothank Jon Reuter, Candice Tulberg, and Stephanie Woerner forhelpful research assistance. All errors are the responsibility ofthe authors.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management Teams20 Organization Science 19(1), pp. 3–24, © 2008 INFORMS

Appendix

Table A Descriptive Statistics �N = 167�

Variable Mean Std. dev. Min. Max.

Firm size (logged) 4�07 0�99 0�69 7�47Medical-related industry 0�14 0�35 0 1Innovation strategy 0�48 0�50 0 1Sales/marketing/service strategy 0�23 0�42 0 1Cumulative rounds of VC 3�10 2�82 0 11Proportion of founders on TMT 0�67 0�68 0 5Product developed 0�77 0�42 0 1Average number of prior 2�12 0�85 0 5person positions (by firm)

Founding team (FT) size 2�82 1�61 1 10Cumulative entrances to TMT 5�50 3�46 0 18Cumulative exits from TMT 1�68 2�22 0 15

FT prior experienceScience/engineering 0�95 0�96 0 3Sales/marketing 0�31 0�55 0 3Manufacturing/operations 0�10 0�30 0 1Finance/accounting 0�11 0�31 0 1Administration/HR 0�38 0�58 0 3Business development/ 0�07 0�25 0 1strategic planning

Initial structureScience/engineering 0�58 0�80 0 3Sales/marketing 0�23 0�56 0 3Manufacturing/operations 0�11 0�35 0 2Finance/accounting 0�14 0�35 0 1Administration/HR 1�53 1�11 0 6Business development/ 0�04 0�22 0 2strategic planning

TMT prior experienceScience/engineering 1�34 1�24 0 5Sales/marketing 1�03 1�23 0 5Manufacturing/operations 0�40 0�64 0 2Finance/accounting 0�53 0�62 0 2Administration/HR 0�80 0�99 0 5Business development/ 0�28 0�55 0 3strategic planning

TMT structureScience/engineering 1�23 1�29 0 7Sales/marketing 1�28 1�51 0 6Manufacturing/operations 0�56 0�77 0 3Finance/accounting 0�68 0�67 0 3Administration/HR 2�32 1�68 0 8Business development/ 0�24 0�54 0 2strategic planning

FT experience count 1�47 1�03 0 4TMT experience count 2�78 1�71 0 6FT structure count 1�77 0�96 0 4TMT structure count 3�38 1�63 1 6IPOs in industry 4�51 4�49 0 18VC deals in industry 1502�22 1000�60 567 3367Ever VC 0�71 0�46 0 1Ever IPO 0�51 0�50 0 1

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management TeamsOrganization Science 19(1), pp. 3–24, © 2008 INFORMS 21

Table

BCorrelationMatrix

12

34

56

78

910

1112

1314

1516

1718

1920

21

1Firm

size

(log)

1�00

2Med

icalindu

stry

−0�15

1�00

3Innovationstrategy

−0�08

0�26

1�00

4Marketingstrategy

0�08

−0�10

−0�52

1�00

5Venturecapital

0�20

0�06

0�20

−0�19

1�00

6Prop

.found

erson

TMT

−0�18

0�11

−0�07

0�07

−0�18

1�00

7Prod

uctd

evelop

ment

0�17

−0�27

0�01

0�16

0�13

−0�06

1�00

8Average

priorpo

sitions

0�03

0�04

0�07

−0�06

0�07

−0�17

−0�03

1�00

9Founding

team

size

0�13

0�11

0�08

0�02

0�03

0�40

0�01

−0�17

1�00

10Cum

.TMTentra

nces

0�37

−0�16

0�24

−0�10

0�31

−0�37

0�22

−0�04

0�24

1�00

11Cum

.TMTexits

0�21

−0�18

0�12

−0�04

0�14

−0�03

0�19

−0�21

0�18

0�63

1�00

12FT

exp.

science/eng.

0�06

0�08

0�20

−0�18

0�06

−0�14

−0�03

0�01

0�22

0�20

−0�04

1�00

13FT

exp.

sales/mktg.

0�01

−0�17

−0�14

0�19

−0�02

−0�08

0�07

0�11

0�17

0�05

−0�02

0�15

1�00

14FT

exp.

mfg./ops.

−0�02

−0�03

−0�01

0�01

−0�03

0�05

0�04

−0�08

−0�05

−0�01

0�06

0�04

−0�08

1�00

15FT

exp.

fin./acct.

−0�02

−0�03

−0�02

0�00

−0�06

−0�02

−0�04

−0�02

0�23

0�10

0�11

0�08

0�23

0�01

1�00

16FT

exp.

admin./H

R0�04

−0�09

−0�11

−0�01

−0�05

−0�12

−0�02

0�00

0�14

0�17

0�11

−0�16

0�09

−0�08

0�04

1�00

17FT

exp.

bus.de

v.−0

�05

0�10

0�04

0�09

0�14

−0�07

0�09

0�08

−0�03

0�01

−0�06

−0�04

0�07

0�15

−0�01

−0�05

1�00

18FT

struc.

science/eng.

0�07

−0�04

0�17

−0�13

0�19

−0�04

0�05

0�05

0�31

0�23

0�08

0�60

0�20

0�03

0�18

0�06

−0�07

1�00

19FT

struc.

sales/mktg.

0�23

−0�17

−0�08

0�03

−0�08

−0�03

0�10

−0�03

0�29

0�21

0�11

0�11

0�48

0�04

0�20

0�23

0�06

0�22

1�00

20FT

struc.

mfg./ops.

0�18

−0�08

−0�17

0�19

0�01

−0�04

0�05

−0�17

0�13

0�04

−0�06

0�12

0�16

0�00

−0�06

0�23

−0�09

0�04

0�26

1�00

21FT

struc.

fin./acct.

0�07

−0�12

−0�05

0�06

−0�20

−0�05

−0�02

−0�24

0�00

0�03

0�00

0�15

0�05

−0�03

0�19

−0�12

−0�11

0�11

0�07

0�16

1�00

22FT

struc.

admin./H

R−0

�01

−0�01

0�03

−0�02

−0�19

0�26

−0�05

−0�14

0�52

0�12

0�14

0�00

0�03

−0�05

0�22

0�19

−0�04

−0�09

0�15

0�01

−0�14

23FT

struc.

bus.de

v.−0

�04

0�17

0�06

0�04

−0�06

0�08

−0�11

0�04

0�31

0�06

−0�04

0�07

0�06

−0�06

0�03

0�18

−0�04

0�09

0�08

0�10

−0�07

24TM

Texp.

science/eng.

0�20

0�18

0�25

−0�09

0�22

−0�28

0�08

0�17

0�17

0�44

−0�11

0�48

0�00

−0�04

−0�06

−0�08

−0�03

0�34

0�10

0�06

−0�08

25TM

Texp.

sales/mktg.

0�19

−0�16

0�02

0�01

0�30

−0�30

0�15

0�18

0�11

0�52

0�03

0�13

0�25

−0�09

−0�07

0�07

0�09

0�15

0�12

0�12

0�07

26TM

Texp.

mfg./ops.

0�15

0�04

0�11

0�02

0�23

−0�18

0�21

0�07

0�06

0�42

0�09

0�04

−0�09

0�25

−0�13

−0�09

0�10

0�08

−0�06

0�06

−0�12

27TM

Texp.

fin./acct.

0�12

−0�05

0�03

0�06

0�13

−0�28

0�05

0�23

0�10

0�35

−0�10

0�24

0�19

−0�03

0�26

0�04

0�08

0�22

0�16

0�11

0�03

28TM

Texp.

admin./H

R0�22

0�00

0�14

−0�07

0�12

−0�27

0�01

0�07

0�12

0�47

0�14

0�08

0�07

−0�11

0�15

0�30

−0�07

0�12

0�18

−0�04

0�03

29TM

Texp.

bus.de

v.0�05

0�17

0�18

−0�06

0�13

−0�17

−0�04

0�11

0�00

0�24

−0�04

0�05

0�00

−0�06

−0�03

0�03

0�27

0�00

0�04

−0�13

−0�11

30TM

Tstruc.

science/eng.

0�20

0�26

0�27

−0�11

0�23

−0�16

−0�03

0�08

0�35

0�38

−0�09

0�48

0�03

−0�06

0�09

0�10

−0�01

0�46

0�11

0�10

−0�01

31TM

Tstruc.

sales/mktg.

0�34

−0�20

0�03

−0�05

0�22

−0�32

0�21

−0�02

0�19

0�59

0�08

0�15

0�20

−0�04

0�05

0�14

0�05

0�10

0�39

0�20

0�10

32TM

Tstruc.

mfg./ops.

0�29

0�01

0�05

0�07

0�16

−0�27

0�14

0�05

−0�03

0�30

−0�09

0�07

0�12

−0�04

−0�15

−0�02

0�06

−0�04

0�11

0�29

−0�01

33TM

Tstruc.

fin./acct.

0�20

0�02

0�06

0�00

0�14

−0�33

0�11

0�13

0�05

0�44

−0�02

0�11

0�02

−0�11

0�08

0�02

0�02

0�12

0�10

0�03

0�17

34TM

Tstruc.

admin./H

R0�10

0�13

0�16

0�04

0�02

0�19

0�05

−0�04

0�55

0�39

0�10

0�09

−0�05

−0�09

0�12

0�12

−0�05

0�06

0�09

−0�03

−0�09

35TM

Tstruc.

bus.de

v.0�06

0�23

0�11

0�00

0�06

−0�11

−0�08

0�04

0�20

0�20

−0�06

0�14

0�06

−0�08

0�06

0�13

0�06

0�09

0�09

−0�02

−0�06

36FT

exp.

count

0�00

−0�10

−0�03

−0�01

0�04

−0�20

0�04

0�04

0�22

0�19

0�05

0�43

0�58

0�27

0�46

0�37

0�28

0�34

0�36

0�17

0�10

37TM

Texp.

count

0�26

0�05

0�19

−0�06

0�35

−0�43

0�14

0�29

0�12

0�61

−0�01

0�24

0�14

−0�06

0�06

0�12

0�12

0�22

0�14

0�06

−0�05

38FT

structurecount

0�19

−0�22

−0�10

0�10

−0�09

−0�06

0�06

−0�12

0�33

0�22

0�08

0�44

0�38

0�04

0�21

0�15

−0�04

0�56

0�56

0�52

0�48

39TM

Tstructurecount

0�30

0�05

0�13

−0�07

0�30

−0�40

0�13

0�14

0�12

0�54

−0�07

0�20

0�06

−0�12

0�03

0�04

0�09

0�12

0�17

0�08

0�06

40IPOsinind.

0�07

0�13

0�08

0�00

0�08

−0�05

0�02

0�05

0�14

0�17

0�05

−0�03

0�06

0�13

0�00

0�05

0�03

0�07

0�04

0�08

−0�05

41VC

dealsinind.

−0�06

−0�19

−0�14

0�07

−0�17

0�18

−0�11

−0�19

−0�07

−0�30

0�13

−0�11

0�01

0�03

0�00

0�03

−0�07

−0�04

−0�02

−0�02

−0�05

42VC

0�19

0�15

0�22

−0�18

0�71

−0�15

0�15

0�22

0�09

0�28

0�01

0�12

0�07

0�00

−0�03

−0�06

0�12

0�17

−0�01

0�02

−0�15

43IPO

0�14

0�19

0�16

−0�10

0�27

−0�29

−0�04

0�31

0�12

0�39

−0�17

0�22

0�04

−0�11

−0�09

0�05

−0�03

0�27

0�06

0�04

−0�01

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management Teams22 Organization Science 19(1), pp. 3–24, © 2008 INFORMS

Table

B(cont’d

.)

2223

2425

2627

2829

3031

3233

3435

3637

3839

4041

42

22FT

struc.

admin./H

R1�00

23FT

struc.

bus.de

v.0�20

1�00

24TM

Texp.

science/eng.

−0�06

0�13

1�00

25TM

Texp.

sales/mktg.

−0�03

−0�09

0�36

1�00

26TM

Texp.

mfg./ops.

−0�02

0�11

0�44

0�34

1�00

27TM

Texp.

fin./acct.

0�02

−0�01

0�32

0�40

0�32

1�00

28TM

Texp.

admin./H

R0�03

0�29

0�27

0�16

0�05

0�14

1�00

29TM

Texp.

bus.de

v.0�01

0�12

0�22

0�27

0�04

0�13

0�16

1�00

30TM

Tstruc.

sci./eng.

0�10

0�27

0�69

0�28

0�18

0�37

0�35

0�13

1�00

31TM

Tstruc.

sales/mktg.

0�05

0�01

0�34

0�68

0�29

0�44

0�36

0�28

0�29

1�00

32TM

Tstruc.

mfg./ops.

−0�18

−0�01

0�36

0�38

0�44

0�35

0�11

0�09

0�18

0�38

1�00

33TM

Tstruc.

fin./acct.

0�00

0�04

0�31

0�38

0�34

0�67

0�25

0�21

0�30

0�45

0�36

1�00

34TM

Tstruc.

admin./H

R0�40

0�12

0�33

0�22

0�24

0�19

0�24

0�17

0�30

0�24

0�12

0�19

1�00

35TM

Tstruc.

bus.de

v./st.pl.

0�16

0�34

0�31

0�21

0�07

0�19

0�18

0�33

0�42

0�18

0�09

0�15

0�12

1�00

36FT

exp.

count

0�11

0�14

0�10

0�19

−0�05

0�31

0�18

0�05

0�23

0�24

0�02

0�09

0�02

0�13

1�00

37TM

Texp.

count

−0�09

0�12

0�61

0�65

0�54

0�63

0�43

0�44

0�49

0�58

0�48

0�61

0�30

0�33

0�22

1�00

38FT

structurecount

0�08

0�22

0�15

0�22

0�02

0�24

0�15

−0�07

0�30

0�31

0�11

0�18

0�04

0�06

0�48

0�13

1�00

39TM

Tstructurecount

−0�03

0�06

0�54

0�55

0�44

0�61

0�32

0�30

0�47

0�62

0�64

0�71

0�25

0�42

0�13

0�80

0�15

1�00

40IPOsinind.

0�09

0�11

0�12

0�13

0�17

0�07

0�15

0�17

0�21

0�18

−0�02

0�08

0�10

0�18

0�08

0�15

0�07

0�13

1�00

41VC

dealsinind.

0�10

−0�10

−0�39

−0�32

−0�37

−0�44

−0�32

−0�15

−0�38

−0�38

−0�31

−0�50

−0�20

−0�17

−0�05

−0�57

−0�06

−0�54

−0�26

1�00

42VC

−0�09

0�05

0�24

0�35

0�30

0�26

0�11

0�16

0�29

0�26

0�25

0�30

0�03

0�07

0�09

0�42

0�01

0�39

0�21

−0�35

1�00

43IPO

−0�07

0�05

0�59

0�46

0�44

0�53

0�28

0�14

0�53

0�39

0�48

0�56

0�28

0�21

0�03

0�66

0�13

0�67

0�27

−0�52

0�45

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management TeamsOrganization Science 19(1), pp. 3–24, © 2008 INFORMS 23

ReferencesAldrich, H. E. 1999. Organizations Evolving. Sage, London.

Aldrich, H. E., C. Zimmer. 1986. Entrepreneurship through social net-works. D. L. Sexton, R. W. Smilor, eds. The Art and Science ofEntrepreneurship. Ballinger, Cambridge, MA, 3–23.

Ancona, D., D. F. Caldwell. 1992. Demography and design: Predictorsof new product team performance. Organ. Sci. 3 321–341.

Audia, P. G., C. I. Rider. 2005. A garage and an idea: What more doesan entrepreneur need? California Management Rev. 48 6–28.

Baron, J. N., M. D. Burton, M. T. Hannan. 1996. The road taken:Origins and evolution of employment systems in emerging com-panies. Indust. Corporate Change 5 239–275.

Baron, J. N., M. D. Burton, M. T. Hannan. 1999. Engineering bureau-cracy: The genesis of formal policies, positions, and structures inhigh-technology firms. J. Law, Econom. Organ. 15 1–41.

Baty, G. B., W. M. Evan, T. W. Rothermel. 1971. Personnel flows asinterorganizational relations. Admin. Sci. Quart. 16 430–443.

Beckman, C. M. 2006. The influence of founding team company affil-iations on firm behavior. Acad. Management J. 49(4) 741–758.

Beckman, C. M., M. D. Burton, C. O’Reilly III. 2007. Early teams:The impact of team demography on VC financing and going pub-lic. J. Bus. Venturing 22(2) 147–173.

Boeker, W. 1988. Organizational origins: Entrepreneurial and environ-mental imprinting at time of founding. G. R. Carroll, ed. Ecolog-ical Models of Organizations. Ballinger, Cambridge, MA, 33–51.

Boeker, W. 1997. Executive migration and strategic change: The effectof top manager movement on product entry. Admin. Sci. Quart.42 213–236.

Boeker, W., R. Karichalil. 2002. Entrepreneurial transitions: Fac-tors influencing founder departure. Acad. Management J. 45818–826.

Boeker, W., R. Wiltbank. 2005. New venture evolution and managerialcapabilities. Organ. Sci. 16 123–133.

Boone, C., W. van Olffen, A. van Witteloostuijn, B. De Brabander.2004. The genesis of top management team diversity: Selectiveturnover among top management teams in Dutch newspaper pub-lishing, 1970–94. Acad. Management J. 47 633–656.

Bunderson, J. S., K. M. Sutcliffe. 2002. Comparing alternative concep-tualizations of functional diversity in management teams: Processand performance effects. Acad. Management J. 45 875–893.

Burton, M. D., C. M. Beckman. 2007. Leaving a legacy: Positionimprints and successor turnover in young firms. Amer. Sociol.Rev. 72 239–266.

Burton, M. D., J. B. Sørensen, C. Beckman. 2002. Coming fromgood stock: Career histories and new venture formation.M. Lounsbury, M. Ventresca, eds. Research in the Sociology ofOrganizations, Vol. 19. JAI Press, Inc., Greenwich, CT, 229–262.

Carroll, G. R., J. R. Harrison. 1998. Organizational demography andculture: Insights from a formal model and simulation. Admin. Sci.Quart. 43 637–667.

Certo, S. T., J. G. Covin, C. M. Daily, D. R. Dalton. 2001. Wealthand the effects of founder management among IPO-stage newventures. Strategic Management J. 22(6/7) 641–658.

Charan, R., C. W. Hofer, J. F. Mahon. 1980. From entrepreneurialto professional management: A set of guidelines. J. Small Bus.Management 18 1–10.

Cooper, A. C., F. J. Gimeno-Gascon, C. Y. Woo. 1994. Initial humanand financial capital as predictors of new venture performance.J. Bus. Venturing 9 371–395.

Cronbach, L. 1987. Statistical tests for moderator variables: Flaws inanalysis recently proposed. Psych. Bull. 102 414–417.

Dearborn, D. W. C., H. A. Simon. 1958. Selective perception: Anote on the departmental identification of executives. Sociometry21(2) 140–144.

Downs, A. 1967. Inside Bureaucracy. Little Brown, Inc., Boston, MA.

Eisenhardt, K. M., C. B. Schoonhoven. 1990. Organizational growth:Linking founding teams, strategy, environment and growthamong U.S. semi-conductor ventures. Admin. Sci. Quart. 28274–291.

Fine, S. A., S. F. Cronshaw. 1999. Functional Job Analysis: A Foun-dation for Human Resources Management. Lawrence ErlbaumAssociates, Florence, KY.

Finkelstein, S., D. C. Hambrick. 1996. Strategic Leadership: Top Exec-utives and Their Effects on Organizations. West Publishing Com-pany, Minneapolis/St. Paul, MN.

Flamholtz, E., Y. Randle. 2000. Growing Pains. JAI Press, Greenwich,CT.

Fligstein, N. 1990. The Transformation of Corporate Control. HarvardUniversity Press, Cambridge, MA.

Gompers, P. A., J. Lerner. 1999. The Venture Capital Cycle. MIT Press,Cambridge, MA.

Greiner, L. E. 1972. Evolution and revolution as organizations grow.Harvard Bus. Rev. 50(4) 37–46.

Gulati, R., M. C. Higgins. 2003. Which ties matter when? The contin-gent effects of interorganizational partnerships on IPO success.Strategic Management J. 24(2) 127–144.

Hambrick, D. C. 2007. Upper echelons theory: An update. Acad. Man-agement Rev. 32 334–343.

Hambrick, D. C., P. Mason. 1984. Upper echelons: The organizationas a reflection of its top managers. Acad. Management Rev. 9193–206.

Hannan, M. T., J. Freeman. 1984. Structural inertia and organizationalchange. Amer. Sociol. Rev. 49 149–164.

Hannan, M. T., M. D. Burton, J. N. Baron. 1996. Inertia and change inthe early years: Employment relations in young, high technologyfirms. Indust. Corporate Change 5 503–536.

Hannan, M. T., J. N. Baron, G. Hsu, O. Kocak. 2006. Organizationalidentities and the hazard of change. Indust. Corporate Change15 755–784.

Hellmann, T., M. Puri. 2000. The interaction between product marketand financing strategy: The role of venture capital. Rev. FinancialStud. 13(4) 959–984.

Hellmann, T., M. Puri. 2002. Venture capital and the profession-alization of start-up firms: Empirical evidence. J. Finance 57169–197.

Jehn, K. A., G. B. Northcraft, M. A. Neale. 1999. Why differencesmake a difference: A field study of diversity, conflict, and perfor-mance in workgroups. Admin. Sci. Quart. 44(4) 741–763.

Jensen, M., E. J. Zajac. 2004. Corporate elites and corporate strategy:How demographic preferences and structural position shape thescope of the firm. Strategic Management J. 25(6) 507–524.

Keck, S. 1997. Top management team structure: Differential effects byenvironmental context. Organ. Sci. 8 143–156.

Knight, D., C. L. Pearce, K. G. Smith, J. D. Olian, H. Sims, K. A.Smith, P. Flood. 1999. Top management team diversity, groupprocess, and strategic consensus. Strategic Management J. 20445–465.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.

Beckman and Burton: Path Dependence in the Evolution of Top Management Teams24 Organization Science 19(1), pp. 3–24, © 2008 INFORMS

Lant, T. K., F. J. Milliken, B. Batra. 1992. The role of managerial learn-ing and interpretation in strategic persistence and reorientation:An empirical exploration. Strategic Management J. 13 585–608.

Lawrence, B. 1997. The black box of organizational demography.Organ. Sci. 8 1–22.

Lin, D. Y., L. J. Wei. 1989. The robust inference for the Cox propor-tional hazards model. J. Amer. Statist. Assoc. 84 1074–1078.

McPherson, J. M., L. Smith-Lovin, J. M. Cook. 2001. Birds of afeather: Homophily in social networks. Annual Rev. Sociol. 27415–444.

Michels, R. 1915. Political Parties: A Sociological Study of the Oli-garchical Tendencies of Modern Democracy. The Free Press,New York.

Miner, A. S. 1987. Idiosyncratic jobs in formalized organizations.Admin. Sci. Quart. 32 327–351.

Mittman, B. S. 1992. Theoretical and methodological issues in thestudy of organizational demography and demographic change.P. Tolbert, S. Bacharach, eds. Research in the Sociology of Orga-nizations, Vol. 10. JAI Press, Greenwich, CT, 3–53.

Osterman, P., M. D. Burton. 2004. Ports and ladders: The natureand relevance of internal labor markets in a changing world.P. Tolbert, R. Batt, eds. Oxford Handbook on Work and Organi-zation. Oxford University Press, Oxford, UK.

Pelled, L. H., K. M. Eisenhardt, K. R. Xin. 1999. Exploring the blackbox: An analysis of work group diversity, conflict, and perfor-mance. Admin. Sci. Quart. 44 1–28.

Pfeffer, J. 1983. Organizational demography. B. Staw, L. Cummings,eds. Res. in Organ. Behavior, Vol. 5. JAI Press, Greenwich, CT,299–357.

Phillips, D. J. 2002. A genealogical approach to organizational lifechances: The parent-progeny transfer among Silicon Valley lawfirms, 1946–1996. Admin. Sci. Quart. 47 474–506.

Phillips, D. J. 2005. Organizational genealogies and the persistence ofgender hierarchies: The case of Silicon Valley law firms. Admin.Sci. Quart. 50 440–472.

Randel, A. E., K. S. Jaussi. 2003. Functional background identity,diversity, and individual performance in cross-functional teams.Acad. Management J. 46 763–774.

Reagans, R., E. W. Zuckerman. 2001. Networks, diversity, and produc-tivity: The social capital of corporate R&D teams. Organ. Sci. 12502–517.

Reagans, R., E. W. Zuckerman, B. McEvily. 2004. How to make theteam: Social networks vs. demograhy as criteria for designingeffective teams. Admin. Sci. Quart. 49 101–133.

Roberts, E. B. 1991. Entrepreneurs in High Technology: Lessons fromMIT and Beyond. Oxford University Press, New York.

Rogers, E. M., D. K. Bhowmik. 1971. Homophily-heterophily: Rela-tional concepts for communication research. Public OpinionQuart. 34 523–538.

Roure, J., R. Keeley. 1990. Predictors of success in new technologybased ventures. J. Bus. Venturing 5 201–220.

Ruef, M., H. E. Aldrich, N. M. Carter. 2003. The structure of found-ing teams: Homophily, strong ties, and isolation among U.S.entrepreneurs. Amer. Sociol. Rev. 68(2) 195–222.

Schein, E. H. 1992. Organizational Culture and Leadership, 2nd ed.Jossey Bass, San Francisco, CA.

Schneider, B. 1987. The people make the place. Personnel Psych. 40437–453.

Shane, S., T. Stuart. 2002. Organizational endowments and the perfor-mance of university start-ups. Management Sci. 48(1) 154–170.

Sine, W. D., H. Mitsuhashi, D. A. Kirsch. 2006. Revisiting burns andstalker: Formal structure and new venture performance in emerg-ing economic sectors. Acad. Management J. 49(1) 121–132.

Smith, K., K. Smith, J. Olian, H. Sims, D. O’Bannon, J. Scully. 1994.Top management team demography and process: The role ofsocial integration and communication. Admin. Sci. Quart. 39412–438.

Sørensen, J. B. 1999. Executive migration and interorganizationalcompetition. Soc. Sci. Res. 28 289–315.

Sørensen, J. B. 2004. The organizational demography of racialemployment segregation. Amer. J. Sociol. 110 626–671.

Staw, B. M. 1976. Knee-deep in the big muddy—A study of escalatingcommitment to a chosen course of action. Organ. Behav. HumanPerformance 16(1) 27–44.

Stinchcombe, A. L. 1965. Social structure and organizations. J. G.March, ed. Handbook of Organizations. Rand McNally, Chicago,IL, 142–193.

Stuart, T. E., H. Hoang, R. C. Hybels. 1999. Interorganizationalendorsements and the performance of entrepreneurial ventures.Admin. Sci. Quart. 44 315–349.

Sutcliffe, K. M. 1994. What executives notice: Accurate percep-tions in top management teams. Acad. Management J. 37(5)1360–1378.

Tripsas, M., G. Gavetti. 2000. Capabilities, cognition and inertia: Evi-dence from digital imaging. Strategic Management J. 21(10/11)1147–1161.

Tushman, M. L., L. Rosenkopf. 1996. Executive succession, strategicreorientation and performance growth: A longitudinal study inthe U.S. cement industry. Management Sci. 42 939–953.

Wanous, J. P. 1992. Organizational Entry: Recruitment, Selection,Orientation, and Socialization of Newcomers. Addison-Wesley,Reading, MA.

Wasserman, N. 2003. Founder-CEO succession and the paradox ofentrepreneurial success. Organ. Sci. 14(2) 149–172.

Williams, K., C. O’Reilly. 1998. Demography and diversity in organi-zations: A review of 40 years of research. Res. Organ. Behavior20 77–140.

INFORMS

holds

copyrightto

this

article

and

distrib

uted

this

copy

asa

courtesy

tothe

author(s).

Add

ition

alinform

ation,

includ

ingrig

htsan

dpe

rmission

policies,

isav

ailableat

http://journa

ls.in

form

s.org/.