podolny baron 1997

RESOURCES AND RELATIONSfflPS:SOCIAL NETWORKS AND MOBILITY IN THE WORKPLACE*

Joel M. PodolnyStanford University

James N. BaronStanford University

We examine how the structure and content of individuals' networks in theworkplace affect intraorganizational mobility. Consistent with prior re-search, we find that an individual's mobility is enhanced by having a large,sparse network of informal ties for acquiring information and resources.However, in contrast to previous work, we emphasize the importance of con-sistent role expectations for performance and mobility. We find evidence thatwell-defined performance expectations are more likely to arise from a small,dense network of individuals. We develop a typology of network contents anddocument the interaction between network structure and content in analysesof mobility among employees of a high-technology firm. We also examinehow the effects of tie duration on mobility vary by tie content. We discuss theimplications of our results for theory and research on networks and organi-zational mobility.

I nterest in understanding how careers un-fold within organizations has recently in-

creased. However, these efforts have focusedalmost exclusively on features of formal or-ganizations that influence mobility, over-looking one of the most important organiza-tional influences on careers and work satis-faction—informal networks in the workplace(Gartrell 1987; Simpson 1989; Baron andPfeffer 1994). Even in the most bureaucraticsettings, informal social relations provide animportant source of task advice (Blau 1955;Dalton 1959; Kanter 1977), can affect the

* Direct all correspondence to Joel Podolny,Graduate School of Business, Stanford Univer-sity, Stanford, CA 94305 (podolnyjoel@gsb,Stanford,edu). The authors acknowledge generousresearch support from the Women's EducationalFund of the Community Foundation of SantaClara County and from the Graduate School ofBusiness at Stanford University, and financialsupport to Podolny from the Fletcher Jones Fac-ulty Fellowship, We also gratefully acknowledgehelpful comments from Ron Burt, Roberto Fer-nandez, Mike Hannan, Herminia Ibarra, TedKastelic, Ed Lazear, Charles O'Reilly, KathyPriebe, Jeroen Weesie, and three anonymous ASRreviewers. Finally, we thank Katrina Jaggears,Megan Miller, Douglas Park, and Joni Podolskifor their assistance, and the employees of thecompany in which this research was conductedfor their cooperation.

content and quality of decision-making (Cro-zier 1964; Hickson et al, 1971), and oftenbecome valued personal relationships formany workers (Roethlisberger and Dickson1946; Warner and Low 1947),

Researchers examining how networksshape mobility usually have focused on howsocial ties affect status or income attainmentin the broad labor market (Lin, Ensel, andVaughn 1981; Lin 1982; Campbell, Marsden,and Hurlbert 1986; Lin and Duman 1986;Boxman, De Graaf, and Flap 1991), entryinto organizations (Granovetter 1974;Bridges and Villemez 1986; Montgomery1992), or perceived potential for advance-ment (Ibarra 1995), Less attention has fo-cused on how networks affect the outcomesof promotion contests within the firm. Per-haps the most systematic exploration of net-work effects on career advancement withinthe firm is Ronald Burt's analysis of intra-organizational mobility in his work Struc-tural Holes (1992). Burt documents how thesame configuration of network ties that cre-ates opportunities for brokering and entre-preneurialism in relations among firms—anetwork full of structural holes (i,e,, beingconnected to many actors who are them-selves unconnected)—also enhances careeropportunities for actors competing for pro-motions within an organization. While ad-

American Sociological Review, 1997, Vol, 62 (October:673-693) 673

674 AMERICAN SOCIOLOGICAL REVIEW

miring Burt's work, we argue that his studyignores the interaction between networkstructure and content. Burt's structuralistconception—developed to understand ex-change relations among firms and industriesin a competitive market context—reflects ahighly instrumental view of networks that isless appropriate for studying mobility withinorganizations.

Burt conceptualizes network ties in termsof information and resources that actors canaccess in competitive situations, includingmobility contests. In a network containingmany "structural holes," it is easier to assimi-late diverse resources and information and toplay people off against one another. We em-phasize a different conception of social net-works, drawing on Coleman's (1990) notionof "social capital" and the extensive litera-ture on community ties (Wellman, Carring-ton, and Hall 1988; Wellman and Wortley1990). Both literatures see informal networkties as the primary bases of social identity,conveying a sense of personal belongingwithin a collectivity and clear normative ex-pectations associated with one's role. Yet aclear social identity is facilitated by smallernetworks that display high closure and cohe-siveness, not large networks full of structuralholes. We suggest that when competing forcareer opportunities, a coherent and well-de-fined organizational identity—includingclear and consistent expectations about one'srole—may be no less important than know-ing whom to consult for information and aid.Thus, the network structure most conduciveto maximizing access to information, re-sources, and "brokerage" opportunities is notthe structure most conducive to a clear socialidentity, and vice versa. This suggests theimportance of examining how network struc-ture and network content interact in shapingcareers.

We develop this alternative conception ofworkplace networks and show how it illumi-nates our understanding of how network tiesshape career mobility within organizations.We document empirically that the networkstructure most conducive to organizationaladvancement depends significantly on thecontent of the social tie involved. We con-sider the same informal ties analyzed byBurt: task advice, strategic information,"buy-in" (those who have "fate control" over

one's initiatives at work), social support, andmentorship. But rather than aggregatingacross these different types of ties, we de-velop hypotheses regarding how each type ofnetwork tie affects upward mobility.

THE INTERACTION OF NETWORKSTRUCTURE AND CONTENT

Granovetter's (1973) pioneering work on the"strength of weak ties" showed the signifi-cance of social networks for mobility. Unlikethe strong ties that bind cliques of individu-als and primarily convey redundant, within-group information, weak ties are sources ofnew information because they bridge localcliques. Thus, the more weak ties an indi-vidual (hereafter "ego") has in his/her net-work, the more valuable the network is as asource of information.

Burt (1992) argues that it is more usefulanalytically to focus on the pattern of rela-tionships among the people (hereafter "al-ters") to whom ego is tied. He argues thatweak ties are a correlate, rather than a deter-minant, of the amount of unique informationprovided by ego's network. Ego derivesunique information from a particular alter tothe extent that alter is disconnected from oth-ers in ego's network. Burt uses the termstructural hole to indicate the absence ofconnections among those in the network, ar-guing that the more structural holes sur-rounding ego, the more conducive ego's net-work is to mobility. We use the term directties to refer to ego's connections to those inhis or her network and indirect ties to referto ties among the alters in ego's network. Thenumber of direct ties should have a positiveeffect on upward mobility within the organi-zation, whereas, for a given size, the numberof indirect ties should have a negative effect.That is, larger and less redundant networksshould aid career advancement.

Although Granovetter (1973) emphasizesthe information benefits that derive from alarge, nonredundant network, Burt (1992) ar-gues that structural holes yield control ben-efits as well, basing his argument onSimmel's (1955) conception of the tertiusgaudens ("third who benefits"). When twoindividuals are suppliers or buyers of thesame resource, a third individual can exploitthe competitive relation between the other

SOCIAL NETWORKS AND MOBILITY 675

two to play them off against one another.Even if two individuals are not vying for orproffering the same resource, the third canexploit the lack of connection by inducing acompetition for his or her time. In both cases,the third's bargaining power is strongly im-paired if the two network members arestrongly interconnected. This proposition isechoed in the network literature on broker-age, although this literature tends to focus oncontexts in which no formal authority rela-tions link actors (e.g., Marsden 1982; Fer-nandez and Gould 1994).

Burt applies the same reasoning to indi-viduals and career mobility within organiza-tions. He argues that a subordinate can influ-ence superiors if there is some uncertaintyabout who has ultimate decision-making au-thority, and this uncertainty is enhanced bythere being many unconnected decision-makers. The greater the uncertainty aboutwho has ultimate authority and whose pref-erences should be taken into account, theeasier it is for a subordinate to play differingpreferences off against each other and estab-lish his or her own agenda as a solution to a"swirl" of conflicting demands. Thus, tomaximize information and control, egoshould maximize the size and nonredun-dancy of his or her network.

A strong market-based conception loomsin the background of this theory. Ego's con-trol over others is a function of the extent towhich ego can play partners off against oneanother. Indeed, Burt (1980) initially appliedthis theory of control to firms and industriesin the market context. There are importantadvantages to transforming a market-basedtheory of control into a general networktheory applicable in diverse contexts. Thehigh level of abstraction allows the re-searcher to focus on common network struc-tures, processes, methods, and measuresacross disparate levels of analysis and socialphenomena, such as interorganizational rela-tions within corporate networks, competitionacross industrial sectors, influence patternswithin political institutions, power strugglesamong organizational subunits, and citationpatterns and career advancement within sci-entific communities. This type of formula-tion has advanced network theory and hasaided the development of standard networkmeasures and methods. Indeed, network ana-

lysts interested in careers often measure anddiscuss social networks in the same way,whether they are studying career processesand network ties within a particular organi-zation or analyzing a representative sampleof the labor force.

Nevertheless, in our view, a market-basedtheory of control obscures several fundamen-tal features of informal networks in work or-ganizations. In Burt's theory of control, rela-tions are understood primarily as conduits ofinformation and resources exchanged by ac-tors in pursuit of instrumental objectives.While many ties in the organizational con-text serve as conduits for some resource, par-ticularly access and information, informalties also are crucial sources of organizationalidentity and social support, often valued fortheir own sake (Ibarra and Smith-Lovin1997). By referring to ties as determinants ofidentity, we mean that informal social rela-tions not only forge a sense of personal be-longing, but also create and sustain a clearnormative framework within which individu-als can rationally determine which courses ofaction are in their interest.

The idea that social ties convey contentother than material resource flows has re-ceived considerable attention in the literatureon neighborhood and community ties (e.g.,Wellman and Wortley 1990). However, re-searchers interested in network effects withinorganizations have neglected the implica-tions of the diverse contents transmittedthrough informal ties at work. Obviously,many organizational ties, such as relations tosupervisors and mentors, involve access toresources and conferral of social identity, in-sofar as normative expectations of superiorsand mentors have strong implications forhow an individual perceives his or her iden-tity in the organization.' Indeed, Grano-vetter's (1985) notion of the "social embed-dedness" of economic exchange seems toimply a significant overlap of individualsamong networks that convey resource-based

' Recent work in the sociology of markets con-ceptualizes market networks as determinants ofidentity. For example, Podolny (1993) argues thatmarket relations are important determinants of anorganization's status. Here, however, we are pri-marily contrasting the organizational context withthe vision of the market that underlies Burt'stheory of control.


and identity-based content. Some types ofinformation exchanged through a specificnetwork tie, such as the performance feed-back an employee may receive from and giveto others, lie at the boundary between re-source-based and identity-based content.

Although the boundary between resource-based and identity-based content is admit-tedly imprecise, the conceptual distinction isimportant because it illuminates how andwhy the effects of structural holes on careermobility within organizations should varyacross different types of network ties. To theextent that ties convey information or re-sources, Burt's theory is generally correct:Large, sparse networks are more advanta-geous in mobility contests. However, to theextent that ties convey normative expecta-tions and social identity, there are clear ad-vantages to a more cohesive network. Con-ceptualizing networks in terms of socialcapital rather than in terms of autonomy, de-pendency, and brokering opportunities, Cole-man (1990) discusses the advantages of so-cial closure in an individual's network. A co-hesive network conveys a clear normative or-der within which the individual can optimizeperformance, whereas a diverse, discon-nected network exposes the individual toconflicting preferences and allegianceswithin which it is much harder to optimize.Burt (1992) acknowledges this same point:"Leisure and domestic clusters are a conge-nial environment of low-maintenance, redun-dant contacts. Efficiency mixes poorly withfriendship" (p. 24). What Burt does not seemto acknowledge, however, is that in the orga-nization, a dense, redundant network of tiesis often a precondition for: (1) internalizinga clear and consistent set of expectations andvalues in order to be effective in one's role;and (2) developing the trust and support fromothers that is necessary to access certain cru-cial resources (political aid, sensitive infor-mation, etc.) and to implement strategic ini-tiatives.

Coleman's (1990) argument finds supportin the organizational literature on "boundaryspanning." Boundary spanners are particu-larly susceptible to role conflict arising fromdiffering and inconsistent expectationsamong multiple constituencies (Whyte 1949;Kahn et al. 1964; Spekman 1979; Van Sell,Brief, and Schuler 1981). Faced with contra-

dictory expectations, the boundary spannerexperiences considerable stress, and eachconstituency grows increasingly suspiciousthat its needs are receiving less attentionfrom the boundary spanner than someoneelse's needs. Role conflict is associated withindividual-level maladies, such as less effi-cient and effective task performance andlower organizational commitment (Kahn etal. 1964:49).

Consider the implications of this line ofreasoning for an individual's "buy-in" net-work—a network composed of those indi-viduals whose support an actor needs in or-der to pursue initiatives successfully withinthe organization. Members of the buy-in net-work have normative expectations for howego should allocate time and what goals egoshould pursue. These expectations define theresponsibilities of ego's position. If there arefew alters in the buy-in network and they arecohesive, ego is likely to face a well-definedand consistent normative milieu withinwhich to pursue his or her interests. In con-trast, when ego is confronted with a diversearray of unconnected individuals with con-flicting preferences and is accountable tothose individuals, ego is less able to satisfythese multiple constituencies simultaneouslyand develop a coherent identity. Thus, wepredict that a buy-in network replete withstructural holes adversely affects ego's jobperformance, and hence mobility, as well asego's well-being at work.

A second difference between Burt's visionof ties in the market and the social ties thatarise within organizations concerns the dis-tinction between actors and the formal posi-tions they occupy within a social structure.In Burt's conception of the market, there areno formal positions independent of actorsand their relations. Within organizations,however, ties and networks exist among for-mal positions, as well as among individuals.For example, organizations often character-ize network connections among formal posi-tions through work flow diagrams or organi-zational charts. These ties are independent ofthe individuals occupying particular posi-tions. Other types of informal ties reflect in-terdependencies among positions but are col-ored by the personal relations between indi-viduals. Consider, for instance, the task-re-lated information network of a faculty mem-


ber who becomes a dean. Were we to dia-gram his or her network in terms of flows ofelectronic mail, we would certainly expect tosee major differences in the size, composi-tion, and structure of that network before andafter assuming the administrative post. Putdifferently, because of task interdependen-cies, resource flows, and opportunities forcontact associated with organizational roles,individuals to some degree "inherit" net-works by virtue of their formal organiza-tional positions.

Formal position is especially likely to cir-cumscribe and structure network ties thattransmit task-related information and re-sources. After all, the task-related informa-tion that ego can provide and that ego needsis largely determined by formal position, andthe alters best situated to provide ego withtask advice also depend on ego's formal role.In contrast, informal ties of friendship andsocial support are likely to be discretionary,reflecting interpersonal attraction and trust,and therefore are less rigidly circumscribedby formal position. (Although formal posi-tion determines opportunities for contactwith friends or sources of gossip that in turninfluence what ties are actually formed, thecontent of such relationships depends less onthe formal positions that ego and alter oc-cupy than does the content of job-related ties,like task-advice ties.)

We hasten to emphasize that this distinc-tion between ties that are "position-centered"or induced by organizational structure versusthose that are "person-centered" or inducedby interpersonal attraction and trust shouldnot be overstated, as it is more a matter ofdegree than of kind. For example, some typesof strategic information conveyed in theworkplace depend, almost by definition, onego's and alter's formal positions, whereasother types of (more diffuse) organizational"gossip" and political intelligence may bedecoupled from formal organizational rolesand depend on interpersonal attraction andtrust.

Notwithstanding these empirical ambigu-ities, the underlying distinction is importantbecause it has implications for how the valueof ties changes over time. Ties based prima-rily on formal position are unlikely to bemaintained after the individual shifts posi-tions or are likely to be of limited value if

they are maintained. One tie that is usuallynot maintained when occupants of a positionmove is a formal reporting relation. If theoccupant shifts horizontally within the orga-nization, ego usually cannot maintain a re-porting relation to the supervisor. The facultymember who becomes a dean illustrates a tiethat can be maintained but loses its value af-ter a change in organizational roles. Thetask-advice ties that were valuable sources ofinformation for research or teaching are notlikely to be contacts that can offer advice inthe new administrative role. Thus, either be-cause position-centered ties cannot be main-tained after mobility or because their valuedeclines after mobility, task-advice ties havelittle portability. In contrast, person-centeredties, such as those that convey gossip, socialsupport, and deference, are more portable—they continue to provide value after ego hasshifted positions because their content is notso closely linked to ego's job. Whereas a ma-jor change in role is likely to disrupt or severposition-centered ties, it is less likely to in-fluence person-centered ties, which are influ-enced more by changes in ego's or alter's lifesituations, values, temperaments, and otherfactors that influence interpersonal attrac-tion.

Figure 1 summarizes these arguments. Theaxes of the figure are two dimensions alongwhich network contents are distinguished.The vertical axis distinguishes ties that pri-marily link positions (reflecting job interde-pendencies) from those that primarily linkpersons (reflecting interpersonal attractionand trust). Positional ties tend to be updatednaturally or to decline in value over time. Forexample, the value of a strong tie to some-one with fate control over ego's initiatives islikely to diminish sharply if ego takes on anew position, whereas person-to-person tiesare unlikely to diminish because they are notclosely linked to ego's formal position. Thehorizontal axis distinguishes ties that conveyresources from ties that convey identity ornormative expectations. We agree with Burtthat structural holes are beneficial for net-works that are conduits of resources. How-ever, for ties that are conduits of normativeexpectations or identity, structural holes arenot beneficial to mobility. Indeed, if facingconsistent expectations benefits job perfor-mance, then structural holes in identity-based


CONTENT CONVEYED THROUGH TIE

Q.

Resources Organizational identity

Workflow inputs/outputsFinancial support

Task Advice

StrategicInformation

Organizational gossip

11111111

1 1L

Authority

Performance feedback

Buy

• •II

-, iVientorship ] —

III 111111111

IV

-In

Organizational rulesFormal socialization

Social SupportFriendshipDeference

Figure 1. A Typology of the Content of Network Ties in Organizations

networks could inhibit performance, andhence inhibit mobility as well.

Each quadrant in Figure 1 contains ex-amples of the content that might flowthrough ties. The position of each contentwithin a quadrant reflects our informal as-sessment of how purely the content conformsto the characteristics of the quadrant. Thecloser a content is to the border of multiplequadrants, the more that we believe the con-tent is an "intermediate type," conveyingsome combination of identity and resources,or person- and position-centered contents.

The quadrants differ in terms of how sig-nificant the network contents are for job per-formance and mobility. Resource flows pri-marily linked to position (e.g., task advice )are clearly of instrumental significance to jobperformance. In contrast, flows of identityand expectations based on interpersonal at-traction have the least instrumental signifi-cance. Although these ties may be importantto work-related outcomes, such as satisfac-tion or turnover intentions, it is not clear whysocial support or friendship ties (absent anyinformational or resource content) should be

relevant to task performance. Quadrants IIand III (person-centered resource flows andidentity-based positional flows) are less clearcut. Some contents in these off-diagonalquadrants—relations to those with authorityand access to certain kinds of strategic infor-mation, such as who is likely to be fired orreceive more resources—could affect ego'sjob performance. Other flows, such as ge-neric company gossip, are probably less re-lated to job performance.

Our analysis explores predictions impliedby this typology, particularly how the ef-fects of structural holes on mobility varyacross types of network ties. We focus onfive types of social relations that illustratethe diversity of informal ties observed inwork organizations: task advice, strategicinformation, "buy-in" or "fate control," so-cial support, and mentorship. These net-works were included in our network surveyto maintain compatibility with Burt (1992)and other network research, rather than tocapture maximal variation along the dimen-sions portrayed in Figure 1. Nonetheless,each of these networks of social relations—


with the exception of mentorship—repre-sents one of the four quadrants in Figure 1.This lets us explore the usefulness of theconceptual distinctions that underlie Figure1. Task advice and social support representquadrants I and IV, respectively. Ties thatinvolve strategic information (quadrant III)and "buy-in" (quadrant II) are somewhatmore ambiguous. Clearly, flows of sometypes of strategic information within infor-mal networks may be a result of the formalpositions of ego and alter. However, in ournetwork survey, strategic information wasdefined more like organizational gossip or"scuttlebutt," which is based more on inter-personal attraction than on job interdepen-dencies. We locate "buy-in" relations (tiesto individuals with fate control over ego'sjob-related initiatives) in quadrant II, be-cause they derive from organizational posi-tion and transmit role expectations, organi-zational identity, and support. (Obviously,individuals with fate control may also beimportant sources of material resources, butthis need not be the case—a coworker whois important socially or politically withinthe organization might be central in the"buy-in" network, even if that person is nota source of important material resources.)

The final tie, mentorship, could fit intoseveral quadrants in Figure I. An extensiveliterature examines how mentor relationshipsinfluence advancement, particularly forwomen and minorities (Marcus and House1973; Kanter 1977; Burke and McKeen1990; Ragins and McF^rlin 1990; Thomas1993). To date, this literature has producedlittle consensus, perhaps because mentorshipinvolves multiple ties of disparate content. Ifmentorship is a source of resources, then themost valuable mentor tie would be to an al-ter who is not tied to ego's other contacts(Burt 1992). However, mentors also providefriendship and social support and convey ex-pectations about role performance and theprogress of an individual's career, thus rep-resenting a crucial determinant of organiza-tional identity and belonging. How thementor's expectations correspond to the ex-pectations of others who have authority orfate control over ego must be considered.Therefore, we distinguish between mentorswho are in ego's buy-in network and men-tors who are not. A view of mentorship that

emphasizes the identity-based content ofmentor-prot^g^ ties rather than their task-re-lated content suggests the opposite predic-tion from Burt's structural holes argument:A mentor not in ego's buy-in network repre-sents a different constituency whose norma-tive expectations must be appeased, thus un-dercutting ego's ability to forge a clear orga-nizational identity and signifying another po-tential land mine in the pursuit of upwardmobility.

DATA, MEASUREMENT, ANDMETHODS

The Setting

To test these predictions, we examine the ef-fects of different network contents on careermobility among exempt (i.e., salaried) per-sonnel in a high-technology engineering andmanufacturing corporation with approxi-mately 25,000 employees worldwide in1994. (Roughly speaking, exempt workersare those in professional and managerial po-sitions.) The company, which grew consider-ably in the years prior to our study, had an-nual revenues of approximately 10 billiondollars in 1993. Though headquartered inCalifornia, the firm also has large establish-ments in several western states and has pro-duction facilities overseas. It is divided intothree main operating divisions and severalstaff functions. Human resource representa-tives in the organization provided us with arandom sample of 658 names of exempt em-ployees drawn from two of the three maindivisions. All employees in the sample werelocated in the United States at the time thesample was drawn.

A computer-administered questionnaire,approximately 30 to 45 minutes in length,was placed on computer diskette. We senteach of the 658 individuals a packet that in-cluded: (1) a letter, co-signed by a companyhuman resources official, requesting theindividual's participation in the survey, as-suring confidentiality, and providing instruc-tions on how to start the survey; (2) a dis-kette containing the survey; and (3) a post-age-paid mailer for returning the disketteupon completing the survey. After threeweeks, we issued a follow-up letter remind-ing individuals of the survey, again request-


ing their participation if they had not yetcompleted the survey.

The survey was divided into three generalsections. First, the respondent was presentedwith a set of name-generating questions thatasked for the names or initials of the impor-tant individuals within his or her network.^In general, we used one name-generatingquestion to define each type of network. Ap-pendix A lists the five name-generating itemsrelevant here.

To maximize comparability across studies,we drew extensively on the name-generatingquestions used by Burt (1992). Respondentscould list up to five names in response toeach name-generator, except that no morethan two mentors could be listed. Respon-dents could list the same individual in re-sponse to the different name-generatingitems. Respondents who provided networkinformation on five alters for a particular net-work were asked to estimate the number ofadditional alters who would meet the criteriaimplied by the name-generating question.This enabled us to gauge the percentage ofrespondents for whom we obtained reason-ably complete network information (thoughwe could not include these additional altersin the analysis because we did not obtaincomplete information on them).

The work force we studied, exempt em-ployees of a high-technology firm, resemblesthe one analyzed by Burt (1992). The majordifferences are that our company is somewhatsmaller than the company Burt analyzed, andour sample includes a broader spectrum ofgrade levels than Burt's did. Our sample isdrawn from 10 grade levels, and only the twoor three highest grades (approximately 10percent of our sample) are comparable to themanagers in Burt's sample. Nevertheless, webelieve the similarities across the twosamples outweigh the differences. By using asimilar set of network measures in a similarorganizational context, we are fairly confi-dent that any differences in results reflect dif-

^ We anticipated that some employees mighthave concerns about the confidentiality of theirsurvey responses, so the computerized question-naire informed respondents that they could useconsistent first names or initials in lieu of fullnames of people in their networks. In this way,the program could prompt the respondent for ad-ditional information on network members.

ferences in how we have specified networkeffects, rather than differences in samples,survey techniques, or name-generating items.

The second section of our survey asked re-spondents about the people they identified asbeing in their networks (e.g., gender, formalposition, etc.) and the nature of each relation-ship (duration, closeness, frequency of con-tact). Respondents were also asked to iden-tify strong ties among alters in their network,where strong ties were defined as relation-ships characterized by closeness and fre-quent contact (Marsden and Campbell 1984).

The third section of the survey asked re-spondents about their satisfaction and orga-nizational commitment. Appropriate re-sponse categories and wording for questionsin these sections were determined throughextensive consultation with members of thefirm's Human Resource department and bypretesting the survey instrument in the com-pany.

We also were given access to companypersonnel information for all 658 persons inthe target sample. These computer recordsincluded demographic information, careerand performance appraisal history, and cur-rent salary. Using this information, wetested for selectivity bias in responses to oursurvey. Of the 658 surveys issued, we ob-tained usable responses from 236 individu-als (36 percent).-' According to company of-ficials, this response rate is comparable towhat the firm achieves from its in-housepencil-and-paper surveys done on companytime. This response rate was achieved de-spite the fact that our survey was more timeconsuming and was administered in an unfa-miliar (computerized) format. Respondentswere given the option of requesting a reportof the research results and a printout com-paring their personal networks to those ofsimilarly situated employees at the com-pany; 99 percent of the respondents re-quested this information.

We tested for selection bias on the basis ofgrade, occupation, division, race, gender,length of employment with the firm, past rateof mobility, and performance ratings. Wefound no evidence of selection bias on thebasis of grade, occupation, division, gender.

^ We received 240 surveys, but four disketteswere unreadable.


or length of employment with the firm. Em-ployees with higher rates of mobility prior tothe survey were slightly more likely to re-spond to the survey, while non-Whites wereless likely than Whites to respond. Followingthe standard procedure for exploring sampleselection bias (Berk 1983), we used logisticregression to compute the predicted probabil-ity of responding to the survey. We then testedwhether inclusion of the predicted probabil-ity in our models affected our results. In nocase was the effect of the selection variablesignificant, nor did its inclusion affect themagnitude or significance of any other coef-ficients. Therefore, although there is slightevidence of nonrandom responses to the sur-vey, they have no discernible effects on theanalyses reported here, and we excluded theselection instrument from the final analyses.

Dependent Variable: Grade Advancement

The survey was administered and the archi-val information collected beginning in Janu-ary 1994; the bulk of the responses were re-ceived by March 1994. To analyze mobility,we model the probability of an exemptemployee's experiencing a grade shift be-tween March 1, 1993 and March 1, 1994, es-sentially the year prior to the administrationof the survey. In this company, grade shiftsdo not necessarily imply job shifts. We focuson grade mobility rather than job mobilitybecause it is extremely difficult to differenti-ate horizontal shifts from vertical job shiftsin this firm independent of grade changes,whereas grade changes clearly signify up-ward advancement."* Indeed, in this company,grade shifts invariably imply an increase insalary because each grade has a clearly de-fined salary range. Of the 229 individuals inour mobility analysis, 57 (25 percent) expe-rienced at least one grade shift over the one-year period we analyzed.^

•* Several human resources personnel could notname their past or current job titles but knew theirgrade history. The fact that human resource spe-cialists were unfamiliar with their formal jobtitles helped persuade us that grade-level changeswere the meaningful career transitions to modelin this organization.

' We excluded from the mobility analyses 7 ofthe 236 employees who were nonexempt as ofMarch 1993, because by definition these indi-

Given this measure of mobility, we ex-cluded all ties in an individual's network thathad a duration of one year or less. By exclud-ing ties of duration shorter than the promo-tion window, we ensure that the direction ofcausality flows from the network to the pro-motion event rather than the reverse (cf. Burt1992:173-80).*

Our approach requires two assumptions:(1) that no important ties in a respondent'snetwork were severed during the one-yearwindow, because alters would not have beennamed or described in the survey; and (2)each alter's function in ego's network re-mained reasonably stable over the durationof the window. For example, if an altershifted from being a source of only task ad-vice to a source of only strategic information,this change would not be reflected in ourdata. We believe these assumptions are notproblematic for two reasons. First, the one-year window makes major changes in net-work composition unlikely.^ Second, and

viduals must have experienced a move (into ex-empt status) to have been in the sample. In effect,inclusion of these 7 individuals would be tanta-mount to selecting on the dependent variable forthis subgroup.

* We conducted supplementary analyses thatincluded these short-duration ties formed after thestart of the promotion window. In general, net-work effects on mobility declined when theseshort-duration ties were included. This suggeststhat the relationship between networks and mo-bility is properly specified. If network effects onmobility outcomes simply reflected some unob-served third factor (e.g., underlying abilities) andthus were spurious, then the observed relationshipbetween network characteristics and mobilitywould not decrease with the inclusion of theshort-duration ties. On the other hand, if mobilityoutcomes are a result of network effects, as wecontend, then including these short-duration tiesshould weaken the observed effect because tiesformed after a promotion obviously cannot causethat promotion.

' To investigate the validity of this claim, weanalyzed retrospective data we collected aboutrespondents' networks in their first six months atthe firm. In our sample, 23 individuals were hiredin 1992. Comparing their initial (1992) informa-tion networks with current ones (January-March1994) yields an estimate of how much change oc-curred in alters and in tie contents for a givenego-alter tie. These 23 individuals named 33 task-advice ties and 32 strategic-information ties in


even more important, to the extent that theseassumptions do affect the results, they arelikely to bias the results against our hypoth-eses by adding additional noise to the data,^

Independent Variables: Network Ties

Task-advice network. We examined threeproperties of ego's task-advice network—itssize (number of direct ties), density (numberof indirect ties among alters), and the dura-tion of ego's ties to alters in the network. Sizecan range between 0 to 5, and the number of

indirect ties ranges from 0 to -, where

N is the size of the task-advice network. Be-cause task advice is a resource flow, we pre-dict that a large, less-dense task advice net-

their initial networks. Of the 33 initial task-ad-vice ties, 27 (82 percent) remained in ego's cur-rent task-advice network. Of the 32 initial strate-gic-information ties, 22 (69 percent) were still in-tact. Moreover, of the 33 initial task advice ties,just four were "transformed" (i,e,, becamesources of only strategic information in the indi-vidual's current network). Of the 32 initial strate-gic-information ties, only 8 became sources oftask-advice in the current network. Because re-call bias is a concern, we regard this informationas a rough guide to turnover and changes in con-tent of ties. We also emphasize that these esti-mates are based on individuals who had been withthe firm for no more than two years. Fewerchanges in network composition and structuremight be expected among employees who havebeen with the firm longer because their work as-signments and social relations would be morestable, (Employees in our sample had an averagetenure of 8,4 years,) Accordingly, the assumptionthat ties remained stable over a one-year periodseems reasonable,

* Note that our dependent measure differs sig-nificantly from Burt's, Burt (1992) analyzes twooutcomes: "fast promotion," which measures theextent to which an individual is promoted to hisor her current rank at a younger-than-normal age,and "early promotion," which measures whetheran individual has been at the current rank longerthan expected. Values for these mobility mea-sures are affected by events that could occur priorto the formation of ties that constitute ego's cur-rent network. The network that Burt measuresthus antedates much of the mobility captured inhis dependent variables, making it difficult to seehow the network could have caused the advance-ment that Burt analyzes. This is why we adopt thealternative approach just described.

work facilitates advancement. However, wehave argued that task-advice networks aremore position-centered than are strategic in-formation networks—the task advice that egoneeds and the task advice that an alter canprovide are closely linked to their formalroles. Accordingly, we predict that task ad-vice ties formed before ego's most recentgrade shift (prior to our one-year promotionwindow) will be less valuable than those task-advice ties formed after ego's most recentgrade shift. To test this hypothesis, we mea-sured duration in two ways. First, we con-structed one variable that is the average dura-tion of ego's task-advice ties. Second, we di-vided network size into two components: thenumber of direct ties formed prior to ego'slast grade shift before the one-year promo-tion window, and the number formed afterthat grade shift. Ties formed after the mostrecent grade shift should facilitate mobility,but those formed prior to this shift shouldhave no effect (or possibly even a negativeeffect).

Strategic-information network. To gener-ate strategic-information networks we askedrespondents to name individuals "on whom[they] have relied for general information onthe 'goings on' at [COMPANY]—people whohave given you special insight into the goalsand strategies of important individuals, divi-sions, or perhaps even the firm as a whole,"Thus, this network taps flows of general or-ganizational gossip. Because strategic infor-mation is a resource, we expect the value ofthe strategic-information network to increasewith network size and sparseness (structuralholes), as we predicted for the task-advicenetwork. However, because we argue that or-ganizational gossip networks involve pre-dominantly person-to-person ties, the valueof this network for mobility should not de-cline with average tie duration. This con-trasts with our hypothesis for the task-advicenetwork.

As with task advice, size of the strategic-information network can range from 0 to 5,and the number of indirect ties ranges from 0

N ( N - l ) , XT • u •

to —^—'-, where N is the size of the net-work. We also include the average durationof strategic-information ties to contrast theeffects of tie duration in the strategic infor-mation and task advice networks.


Buy-in (fate-control) network. We arguethat, net of the other ties we are examining,buy-in or fate-control ties convey organiza-tional identity and normative expectationsrather than constitute tangible resources. Ac-cordingly, we predict that a small, dense fate-control network is more conducive to mobil-ity than is a large, sparse network containingmany unconnected constituents. As with thestrategic-information and task-advice net-works, size can range from 0 to 5, and thenumber of indirect ties in a network of size

N can range from 0 to —'- . We do not

include a variable for average duration of thebuy-in network ties because these ties areless discretionary than those in the informa-tion networks. The longevity of the buy-innetwork ties is a function of how long egohas been in a particular formal position in theorganization (which our models do controlfor). These ties will be updated when an em-ployee acquires a new formal position.

Unlike the task-advice or strategic-infor-mation networks, ego could cite an individualas someone whose buy-in is critical withoutthe named alter necessarily being amongego's network of direct personal relations orfriends. Accordingly, we include a variablethat represents the average closeness to thosein ego's fate-control network. Closeness ismeasured on a four-point scale: distant, notclose, close, extremely close,' We assignedequal-interval scores between 0 and 1 to theseresponses and calculated the mean closeness.We expect that closeness increases promotionchances. However, causality could run in theopposite direction: individuals might makeinferences about their closeness to key deci-sion-makers based on whether they had beenpromoted, assuming that once promoted theyare "close" to those with authority or influ-ence. Thus, causality for this variable is diffi-cult to determine from our analysis.

We also included variables representingthe size and number of indirect ties amongthose in the buy-in network. These variablesallow us to adjudicate between Burt's (1992)structural holes argument and the oppositepredictions derived from the research onboundary spanners and the literature empha-

* Respondents were given a one-sentence defi-nition of each of these categories.

sizing the identity-enhancing effects ofsmall, dense networks.

Mentor relations. Our questionnaire lim-ited the number of mentor ties to two. Wedistinguished between mentors inside andoutside ego's buy-in network. If mentor tiesare primarily conduits of resources, it shouldbe beneficial to have a mentor who is out-side of ego's immediate work setting andbuy-in network because, as Burt (1992)ar-gues, such a mentor provides nonredundantaccess. However, if mentor ties are primarilyconduits for conveying normative expecta-tions and organizational identity, then itshould not be beneficial to have a mentoroutside of ego's buy-in network because sucha mentor is another constituency that egomust appease.

Friendship or social support relations.Quadrant IV in Figure 1 contains identity-based person-to-person ties that are unlikelyto aid job performance directly. We did notpredict any net effect of the social supportnetwork on mobility, and preliminary analy-ses confirmed this suspicion. However, in-cluding these variables inflated the standarderrors of variables characterizing the strate-gic-information network because of a corre-lation between the features of the social sup-port network and the strategic informationnetwork, (A moderately strong correlationbetween these two networks is not surprisinggiven that both are person-driven rather thanposition-determined,) We therefore omit thesocial support network measures from ouranalyses.

Control Variables

Our analyses include additional controls, in-cluding three demographic attributes of re-spondents: race (non-White =1), gender (fe-male = 1), and age (in years),'° We includelinear and quadratic terms for the amount oftime an individual has been in the currentgrade (in years), (The mean duration in cur-rent grade is 2,61 years with a standard de-viation of 2,46,) In alternative specifications,

'" Separate analyses for women and minoritiesfound no systematic differences in network ef-fects on mobility by gender or race. However, oursample included few women (47) and non-Whites(37), which may have limited our ability to detectsuch differences.


we also control for tenure in the organiza-tion, but when age and grade are included ascontrols, the effect of tenure is nonsignifi-cant, (Most people enter this organization atapproximately the same age and grade,)

We control for prior mobility in two ways.First, we calculate ego's rate of prior promo-tions by dividing the total number of gradepromotions achieved before the beginning ofthe one-year promotion window by ego's ten-ure in the organization (in years) as of thatdate, (The mean number of grade shifts peryear is ,26 with a standard deviation of ,21,)Second, we include a dummy variable indi-cating whether ego had any prior promotionswithin the company. This permits mobilitychances for employees who have never beenpromoted in the past to differ discontinu-ously. These two measures capture the pro-pensity for ego to have been promoted be-fore our one-year promotion window,thereby controlling for any unobserved char-acteristics that could affect observed mobil-ity. This makes for a strict test for networkeffects on mobility.

We also control for several aspects of ego'sformal position in the company: grade, divi-sion, and occupation. The effect of grade isreported in the tables," Controls for division(dummy variables) are included in the analy-sis, but their effects are not reported becausethey have little significance outside this par-ticular organization. In other analyses (notshown), we also included controls for occu-pation, none of which approached statisticalsignificance; they were therefore excludedfrom the final analyses, '^

Methods

We modeled the probability of a grade ad-vancement over the one-year promotion win-dow using logistic regression. The outcomevariable was coded 1 if the respondent re-

" Specifications incorporating a dummy vari-able for each grade did not improve on the re-ported (linear) specification,

'̂ In analyses not reported here, we controlledfor ego's task-advice relations, strategic-informa-tion relations, and mentor relations during thefirst six months at the company. Although respon-dents may not have recalled such informationwith perfect accuracy, we included these vari-ables to control for ego's unobserved propensity

ceived a promotion over the one-year win-dow, 0 otherwise,'^

RESULTS

Descriptive statistics and bivariate correla-tions for the independent variables of inter-est are reported in Tables la and lb and inTable 2, respectively. Table la indicates thatmost individuals reported content-specificnetworks that were no larger than five. Al-though a substantial minority (30 percent)named more than five network alters in theirtask-advice network (Table lb), these resultssuggest that we obtained rather complete in-formation on the networks of interest,'*Table 2 indicates that correlations among thesize of the task-advice, strategic-information,and buy-in networks range from ,34 to ,49,Although statistically significant, these cor-relations are sufficiently low that each net-work can be considered reasonably distinct.

Table 3 reports the effects of social net-works on grade mobility. For those coeffi-cients for which there are contradictory hy-potheses regarding their direction, we em-ploy two-tailed tests of significance. For allothers, we employ one-tailed tests. We donot report significance levels for controlvariables.

Consistent with Burt's (1992) structuralhole hypothesis, the size of the strategic-in-formation network in the full model (Model

to form particular types of network ties. If indi-viduals of different ability levels form networksof different types, then observed network effectsmay reflect these ability differences. Our controlfor the rate of prior promotion reflects some ofthese unobserved differences, but including theseearlier network variables provided an additionalcontrol for such differences. These early networkvariables had no significant net effects on mobil-ity and therefore were excluded from the finalanalyses,

'̂ We conducted the analyses using theLOGIST procedure in SAS (version 6,07),

''' Concerned that there might be something dis-tinctive about employees who have networkslarger than five, we conducted supplementaryanalyses including dummy variables to indicate ifa respondent would have named more than fivealters for any of the name-generating items in thesurvey. None of these dummy variables had a sta-tistically significant effect on mobility. Accord-ingly, they were excluded from the final analyses.


Table la.

Size ofNetwork

0

1

2

3

4

5

Descriptive Statistics fornology Corporation, 1994

Task-Advice

MeanIndirect Ties

Number amongof Cases Alters

15 0(0)

34 0(,00)

46 ,24(,43)

43 ,44(,76)

28 1,26(1,22)

69 1,37(1,45)

Types of Networks, bji' Networli

Type of Network

Strategic-Information

MeanIndirect Ties


34 0(,00)

36 0(•00)

44 ,18(,34)

53 ,47(,80)

34 1,02(1,10)

35 1,22(1,23)

Size: Employees at a

Buy-In

MeanIndirect Ties


51

40

46

38

19

42

0(,00)

0(,00)

,13(•34)

,44(,72)

,79(1,03)

1,44(1,50)

Social

High-Tech-

Support

MeanIndirect Ties


40

46

71

37

25

17

0(,00)

0(.00)

,45(,50)

1,22(,91)

1,60(1,19)

1,52(1,32)

Note: Numbers in parentheses are standard deviations; N = 236,

Table lb. Additional Characteristicsogy Corporation, 1994

Characteristic Task-Advice

of Networks by Network Type:

Type of Network

Strategic-Information Buy-In

Employees

SocialSupport

at a High-Technol-

Mentor

Mean size of network

Mean duration ofdirect ties

Number who would namemore than five sources

Average closeness

Number who named atleast i mentor

Number who would namemore than 2 mentors

3,04(1,63)

3,19(2,58)

58

—

—

—

2,52(1,62)

1,35(1,85)

34

—

—

—

2,25(1,76)

—

35

,43(,28)

—

—

2,05(1,45)

—

17

—

—

—

144

49

Note: Numbers in parentheses are standard deviations; N = 236,

4, Table 3) has a positive and statistically sig-nificant effect on the odds of grade promo-tion, whereas the number of indirect ties inthis network has a significant negative effect.When an individual's rate of prior mobilityis excluded (Model 3), the average durationof strategic-information ties has a positive,statistically significant effect on mobility.

After controlling for the rate of prior mobil-ity, however, the effect is reduced consider-ably and is no longer significant (Model 4).Note that long-lived strategic informationties not only benefit ego's current mobilityprospects, but also presumably play a role inego's mobility. Therefore, it is not surprisingthat controlling for the rate of past mobility


II

zOS

I2oU

,6BO

Xa

a.

n

•cn

"B01•e

B

o

nenBO-.13

o

00 C— O

o o Tt

cs — —en t O

oo >o Oen — O

00 O

>o o

I I

I

.01

cs

SO"0

1-

.12

.07

.54

Tt

-.01

q

.10

-.01

20"

.06

.61

enen

q

o

q

-.05

.23

.18

.18

OS

.11

.01

1

.08

-.02

.10

20"

.40

oen

-.02

enq

q

.02

r.1

2.5

1

enen

q

ooq

q

-.15

CM

-.01

.08

en

.07

.14

-.04

-.01

.10

.25

ocs

q

02"

-.02

.03

.12

.37

CMCS

.10

Tt

q

.10

.01

-.02

s

.43

es

.03

o

i"

q

.00

.06

to-

.27

T)

.12

r-or

.02

-.02

.07

-.13

.21

VOcs

-.01

o

q

-.18

.11

.09

.67

§

-.01

80"0

1-

.03

.14

1.00

1

o

-.27

-.12

.01

1.00

o1

-.26

.06

1.00

1

1

1

g1

to-

1.00

1

cs

1.00

-30

§—

Tten

.10

Tten

.13

o

.24

en cs

20"

o

.02

o

-.03

enTl

-.06

o

.36

.49

o

.61

.32

VOCS

.02

.18

00

.28

.39

en

.30

.19

OS

-.01

.02

VO

o

.32

69"

oo

.55

1.00

1

cs — oo ~^ or r "

.56

oo

I I I

o

Iu

s

I.1

u

u

s•S

oa

dui

ubiau:>

? 5~ in £•

11

3

I.«

00

c

XI

o

' ^ w 5 5 - S < u - Z O Q

o

- .2.2 cs en

c§


lessens the effect of tie duration. In effect,the rate of prior mobility is a proxy for howbeneficial ego's long-lived strategic informa-tion ties have been in the past; once thisproxy is included in the analysis, the directeffect of duration diminishes. Clearly, thereis little evidence that the effects of strategic-information ties decline with duration.Rather, the value of this person-to-person tieappears to increase slightly with duration.This positive relationship may suggest theimportance of trust—which presumably canonly be established over time—for person-to-person ties.

Models 2 and 3 present alternative specifi-cations for the effects of task-advice networkson grade mobility. Model 2 includes size (di-rect ties), indirect ties, and average durationof ties in the task-advice network. Size has apositive coefficient, although it is not statisti-cally significant, whereas the number of indi-rect ties has a significant negative effect. Av-erage duration has a negative coefficient thatfails to attain statistical significance.

These results mask important differences,however, in the effects of long-lived versusshort-lived task-advice ties. Model 3 dividesthe task-advice network into two compo-nents: ties formed before the last grade shiftpreceding the one-year promotion windowand ties formed while in the respondent'scurrent grade (as of the beginning of the pro-motion window). As we argued above, task-advice networks are likely to require signifi-cant updating after promotion, given thechanges in duties and interdependencies as-sociated with a higher position in the organi-zation. Consistent with our argument, task-advice ties formed since the respondent'smost recent grade shift have a positive andsignificant effect on subsequent mobility,whereas ties predating the most recent gradeshift have a negative (but not statistically sig-nificant) coefficient during the one-year pro-motion window. Therefore, unlike person-centered strategic-information ties, position-centered task-advice ties show clear evidenceof obsolescence following mobility. Both thepositive effect of size and the negative effectof indirect ties are statistically significant inthe fully specified model (Model 4). Basedon Model 4, each additional task-advice tieformed after the most recent grade changemore than doubles the odds of promotion

over the one-year window (e-^-' = 2.07),whereas each additional indirect tie reducespromotion odds by a little more than one-half(e"^"* = .43). Given that the unconditionalodds of an individual's being promoted overthe one-year window are 1:3 (25 percent ofthe sample was promoted), these effects onthe odds are substantial.

Consistent with the literature on boundary-spanning roles, the size of the buy-in net-work has a negative, statistically significanteffect on mobility (Model 4), whereas thenumber of indirect ties among those withinthis fate-control network has a significantpositive effect. Apparently, a lack of cohe-siveness among those with fate control im-pedes advancement, in direct contrast to theprediction of Burt's tertius gaudens argu-ment. Average closeness to those with fatecontrol has a strong positive effect on mobil-ity. We reiterate that this effect should be in-terpreted cautiously; a respondent could in-fer closeness to those in the buy-in networkbased on his or her own advancement.

The effects of mentor relations provide fur-ther evidence that structural holes amongthose with fate control diminish mobility.Model 1, which includes only the mentor tievariables, shows that a tie to a mentor withinthe buy-in network has a significant positiveimpact on the odds of grade mobility. Thereis no statistically significant difference ingrade mobility between individuals with amentor outside the buy-in network and thosewho did not name a mentor. Model 2 revealsthat the positive effect of having a mentor inthe buy-in network is largely spurious—when other network ties are included in theanalysis, the effect of mentors becomes non-significant. Additional analyses (not reportedhere) revealed that the spuriousness reflectsthe strength of ego's connections to those inthe buy-in network.'^ In other words, havinga tie to a mentor with fate control facilitatesmobility, but no more so than having a rela-tionship to any individual whose buy-in iscritical. This spurious relationship impliesthat mentor effects on career mobility prima-

'^When ego's average closeness to those in thebuy-in network is included in Model 1, the effectof having a mentor in the buy-in network is re-duced by approximately one-half and becomesnonsignificant.


Table 3. Logistic CoefTidents from the Regressioi1 of Grade MobilityEmployees at a High-Technology Corporation, 1994

Independent Variables

Intercept

Task-Advice Network

Size

Number of ties formedafter most recent grade shift

Number of ties formedbefore most recent grade shift

Indirect ties

Average duration

Strategic-Information Network

Size

Indirect ties

Average duration

Buy-In Network

Average closeness

Size

Indirect ties

Mentor RelationsMentor in the buy-in network

Mentor not in the buy-in network

Control Variables

Grade

Duration in grade

(Duration in grade)^

Non-White

Female

Age

Rate of promotionsprior to one-year window

Any promotionsprior to one-year window

Chi-square

Degrees of freedomPercent concordant pairs

Number of cases"

Model 1

Coef.

.34

.65*

-.07

-.15

1.47

-.16

-.69

.39

-.09

S.E.

(1.17)

—

—

—

—

—

—

—

—

—

—

(.33)

(.27)

(.12)

(.36)

(.05)

(.46)

(.44)

(.03)

—

—

46.00"10

78.4229

Model 2

Coef.

.69

.19

-.56+

-.09

.64+-

-.50

.16+

1.71+

-.27

.92*

-.22

-.40

-.17

1.68

-.18

-.71

.44

-.10

S.E.

(1.32)

(.19)

—

_

(.34)

(.08)

' (.22)

(.34)

(.09)

' (.79)

(.22)

(.38)

(.48)

(.33)

(.13)

(.41)

(.05)

(.51)

(.52)

(.40)

—

—

74.80**

19

85.7

229

on Measures

Model 3

Coef.

.31

.53++

-.13

-.70+

.66++

-.60+

.15+

1.79++

-.30

1.08**

-.09

-.41

-.17

1.34

-.16

-.89

.50

-.13

S.E.

(1.42)

—

(.24)

(.22)

(.36)

—

(.21)

(0.36)

(.09)

" (.81)

(.23)

(.40)

(.49)

(.33)

(.13)

(.42)

(.05)

(.53)

(.52)

(.04)

—

—

82.00**

1987.1

229

of Network Type:

Model 4

Coef.

-.70

.73++

-.03

-.84++

.60++

-.74++

.08

1.68+

-.50*

1.00*

.38

-.70

-.18

1.61

-.17

-.15

.32

-.12

2.11

2.13

104

S.E.

(1.78)

—

(.28)

(.24)

(.40)

(.23)

(.36)

(.08)

(.87)

(.25)

(.41)

(.56)

(.37)

(.15)

(.47)

(.06)

(.60)

(.56)

(.04)

(.69)

(1.16)

.86**21

90.7

229

" Number of cases is 229 because 7 individuals were nonexempt employees (see note 5).+p < .05 ++P < .01 (one-tailed tests)*p < .05 "p < .01 (two-tailed tests)


rily reflect the mentor's authority or influ-ence over decisions and initiatives in ego'sdepartment or office; there is little tangiblebenefit of a mentor on grade promotion in-dependent of this ability. Put differently, it isbeing close to those with fate control, ratherthan having a mentor per se, that helpspeople get ahead.

Although having a mentor in the buy-innetwork has no independent effect on mobil-ity chances, the fully specified model (Model4) shows that having a mentor who is out-side of the buy-in network actually has anegative effect on mobility chances thatclosely approaches significance {p = .06,two-tailed test). This result provides furtherevidence that dispersion of fate control im-pedes mobility chances. By trying to impressor placate a mentor who is disconnected fromthose who control one's fate, while also ac-commodating the interests and concerns ofthose in the buy-in network, an employeemay perform in ways that displease both con-stituencies, thereby diminishing his or herfuture prospects.

CONCLUSION

We have documented how informal socialties affect advancement within a high tech-nology corporation. In certain respects, ourfindings are consistent with those reported inBurt's (1992) pioneering study of a similarorganization. Like Burt, we found that largeinformation networks that lack indirect ties(i.e., are full of structural holes) promote up-ward mobility within the company. Ouranalyses also substantiate claims by Burt andother network structuralists (e.g.. White,Boorman, and Breiger 1976; Mayhew 1980)that the pattern or structure of social rela-tions is a meaningful determinant of anindividual's fate, including intraorganiza-tional advancement.

However, by disaggregating social ties intospecific types, we have shown that Burt's pre-dictions apply only to a restricted class ofnetwork contents. Among position-centeredresource flows, it is necessary to be sensitiveto the duration of ties—simply adding up thenumber of structural holes in the networkleads to a misspecification of the network ef-fect on mobility because the value of some ofthese holes clearly diminishes when ego

changes positions. More significantly, withinbuy-in networks, structural holes have a nega-tive effect on mobility. Supplementary multi-variate analyses (not reported here) revealedthat job satisfaction was lowest, all else be-ing equal, among those whose buy-in net-works were the largest and displayed the moststructural holes, as would be predicted by theboundary-spanning literature and the iden-tity-based approach to networks. Perhaps itcan be said that all structural holes are not ofthe same color; some are "white holes," pro-pelling the individual upward through the or-ganization and providing socioemotionalbenefits, while others are clearly "blackholes," holding individuals at a particularrank in the organization and causing negativepsychological consequences.

Given our finding that the effects of struc-tural holes on promotion are positive for tiesthat convey resources and information andnegative for ties that transmit identity andexpectations, the standard practice in net-work research of aggregating disparate kindsof ties when relating network structure tomobility outcomes seems ill-conceived.'* In-deed, Burt (1995) reanalyzed his data de-composing his aggregate network into two

'* In supplementary analyses, we adoptedBurt's approach and calculated aggregate mea-sures of network size and structure. There was nonet effect of overall network size or structure onmobility. Why did Burt (1992) find an effect ofstructural holes for his aggregated network mea-sure and we did not? Perhaps the difference re-flects differences between the two organizationsand samples studied. Burt's sample consisted ofsenior managers for whom issues of organiza-tional identity and belonging may no longer besalient for career advancement, whereas oursample was more heterogeneous. Another expla-nation may rest in methodological differences be-tween the two studies. Whereas we examine mo-bility events that occur after the formation of anindividual's network, Burt examines the relation-ship between an individual's current aggregatenetwork and that individual's total mobility sinceentering the firm. Since much of an individual'smobility may occur prior to the formation of thenetwork, Burt's specification of the relationshipopens up the possibility of reverse causality. Thatis, a high rate of mobility may determine the for-mation of subsequent structural holes. Such re-verse causality would increase the possibility ofa statistically significant relationship between theaggregate network and an individual's mobility.


disaggregated networks—a network pertain-ing to resource fiows and a network pertain-ing to authority relations. With the overallnetwork disaggregated in this fashion, struc-tural holes in the resource-based network hada positive effect on mobility, but structuralholes in the authority network did not.

We suggested that when applying the struc-tural hole hypothesis to organizational con-texts (including mobility contests), the dis-parate contents conveyed through informalsocial ties must be situated within a broaderconceptual framework. We emphasize twodimensions along which social ties in workorganizations vary: (1) whether the tie is prin-cipally a conduit of task-related informationand resources rather than a means of trans-mitting role expectations, an organizationalidentity, and a clear sense of belonging; and(2) whether the tie is primarily a link amongpositions, reflecting task interdependenciesand the exigencies of the division of labor, oramong individuals, reflecting processes ofinterpersonal attraction and trust. Althoughthe hypotheses we tested were broadly con-sistent with our findings, the specific types ofnetwork ties we studied were selected to par-allel previous studies in this field, not to pro-vide an optimal representation of the concep-tual space portrayed in Figure 1. Future theo-retical and empirical work on networks canvalidate and refine our typology and improveour understanding of the contours and conse-quences of different types of content that flowthrough informal social ties.

Indeed, we believe our typology casts lighton some disconfirming evidence regardingthe structural holes hypothesis that Burt un-earthed in his own study. Burt (1992) re-ported that mobility among female managersand entry-level male managers was enhancedby building hierarchical networks around astrategic partner and by building dense tiesto the immediate work group, rather than bymaximizing structural holes and brokerageopportunities. Burt writes that this "speaks totheir more defensive positions in the firm"(1992:157). Stated another way, it is espe-cially important for women in the seniormanagement ranks of a technology company,as well as entry-level male managers, toforge clear organizational identities and tointernalize a coherent set of normative ex-pectations about their organizational roles.

For these organizational "neophytes," resolv-ing identity concerns may be of greater mo-ment than maximizing access to information,resources, and brokerage opportunities.

Extending this line of argument, one couldtest a set of hypotheses about how the ben-efits of structural holes versus dense cohe-sive networks vary not only with the contentof ties but also cross-culturally (particularlybetween individualistic and collectivist cul-tures), across organizational contexts (e.g., intraditional bureaucratic firms, where struc-tural holes may be most beneficial, versusstrong culture organizations, where a senseof belonging and a clear organizational iden-tity may be crucial), among different typesof occupations (e.g., as a function of demo-graphic composition and degree of technicalinterdependence), and in response to ascrip-tive characteristics, tenure, and other indi-vidual attributes. Such investigations into theboundary conditions around structural holeeffects represent a promising direction forfuture research.

More broadly, our approach to networkscalls into question the instrumentalist, stra-tegic conception of social networks implicitin most research relating social networks tomobility. Many important informal ties in or-ganizations are a result of ego's and alter'spositions within the organizational divisionof labor and are therefore only minimallycontrollable by the actors involved. More-over, individuals seek not only resources andinformation through social networks, butalso a sense of belonging and an understand-ing of what is expected of them, and some-times the very same tie (e.g., to a mentor orsupervisor) can be a source of both resource-based and identity-based flows. Conse-quently, individuals are highly constrained intheir ability to form a network that is maxi-mally efficient with respect to some property,such as the number of structural holes.

The sources of constraint within organiza-tions are numerous. One source of constraintis the opportunity to form new ties. For in-stance, it may not be possible to form newmentor relations to individuals with buy-in;therefore, ego may retain a relationship witha prior mentor that continues to providesocioemotional benefits even if the mentor'svalue for future mobility is limited (or evennegative).


A second source of constraint follows fromthe first. To the extent that ego has limitedopportunities to form ties, he or she may relyon the same ties for multiple contents (e.g.,receiving task advice, strategic information,and social support). Such "multiplex" tiesmay constrain the individual's ability tomaximize his or her network with respect toany particular content. If an individual re-ceives multiple contents from the same indi-vidual, it may be difficult to drop one contentwithout dropping the other. For example, ifego goes to a given alter for strategic infor-mation, social support, and buy-in, it willobviously be quite difficult to withdraw thestrategic information component of the rela-tionship without risking the loss of the tie al-together. Yet as we have seen, structural holesin the strategic information network facilitateego's mobility, whereas structural holes inego's buy-in network have the opposite ef-fect. Thus, if the two networks overlap sub-stantially, ego will by definition have to livewith suboptimal networks in both domains.

A third source of constraint is that indi-viduals may experience negative reputationalconsequences in an organization by droppingperson-to-person ties that are no longer valu-able. As Burt (1992) colorfully writes,"Judging friends on the basis of efficiency isan interpersonal flatulence from whichfriends will fiee" (pp. 24-25). If ego is per-ceived as dropping ties and reneging on im-plicit obligations when a relation is no longervalued, others may be less willing to formties with ego. Put another way, the individualmay need to preserve ties that are no longerinstrumentally valuable because of normsagainst the breaking of ties.

The presence of such constraints calls intoquestion the value of conceptualizing work-place networks in highly strategic and vol-untaristic terms. A promising direction forfuture research involves examining stabilityand change in various types of social ties inorder to understand how individuals adapt tothese constraints and manage the trade-offsthey imply. What structural or demographicfactors affect an individual's opportunities toform new ties? What factors account for therate of change in the number, structure, andcomposition of different types of informalties? Are some types of relations (e.g.,friendship and strategic information) more

persistent than others? How and why do em-ployees vary in the extent to which theirworkplace networks are "multiplex" versusbeing compartmentalized across specific do-mains or types of relations? Why do somegroups of employees have more diverse andexpansive networks than others? Researchalong these lines should illuminate the rela-tive importance of organizational constraintsversus employees' strategic choices in shap-ing social ties, and should provide additionalinsight into how informal social relations af-fect mobility and satisfaction in the work-place.

Joet M. Podotny is Associate Professor of Stra-tegic Management and Organizational Behaviorat the Graduate School of Business, StanfordUniversity. His areas of interest include eco-nomic sociology and complex organizations. Hiscurrent projects involve examining how organi-zational status impacts on organizational learn-ing and investigating the causes and conse-quences of cross-national differences in work-place networks.

James N. Baron is the Walter Kenneth KilpatrickProfessor of Organizational Behavior and Hu-man Resources at the Graduate School of Busi-ness, Stanford University. He is spending the1997-1998 academic year as a Marvin BowerFellow at the Harvard Business School. His re-search interests include economic sociology,complex organizations, careers, and socioeco-nomic inequality. With Michael T. Hannan andM. Diane Burton he is conducting a large-scalestudy of the evolution of organizational structuresand employment practices in high-technologystart-up companies.

Appendix A. Name-Generating Questions Usedin Computerized Questionnaire

The following name-generating questions, whichclosely follow Burt (1992), were used to obtain in-formation on ego's current network:

Task advice Over the last six months, are there anywork-related contacts from whom youregularly sought information and ad-vice to enhance your effectiveness onthe job?

Buy-in Suppose you were moving to a newjob and wanted to leave behind thebest network advice that you could forthe person moving into your currentjob. Are there any individuals whomyou would name to your replacement


whose "buy-in" is essential for initia-tives coming out of your office or de-partment?

Strategic Thinking back over the past sixinformation months, are there any individuals on

whom you have relied as sources forgeneral information on the "goings-on" at [COMPANY NAME]—people who

have given you special insight into thegoals and strategies of important indi-viduals, divisions, or perhaps even thefirm as a whole?

Mentor Are there any individuals whom youregard as a mentor—that is, someonewho has taken a strong interest inyour professional development overthe last six months by providing youwith opportunities and/or access to fa-cilitate your career advancement?

Social Is there anyone in your work environ-support ment over the last six months whom

you regard as a source of social sup-port—that is, someone with whomyou are comfortable discussing sensi-tive matters?

REFERENCES

Baron, James N. and Jeffrey Pfeffer. 1994. "TheSocial Psychology of Organizations and In-equality." Social Psychology Quarterly 57:190-209.

Berk, Richard A. 1983. "An Introduction toSample Selection Bias in Sociological Data."American Sociological Review 48: 386-98.

Blau, Peter M. 1955. The Dynamics of Bureau-cracy: A Study of Interpersonal Relationshipsin Two Government Agencies. Chicago, IL:University of Chicago Press.

Boxman, Ed A. W., Paul M. De Graaf, andHendrick D. Flap. 1991. "The Impact of Socialand Human Capital on the Income Attainmentof Dutch Managers." Social Networks 13:51-73.

Bridges, William P. and Wayne J. Villemez.1986. "Informal Hiring and Income in the La-bor Market." American Sociological Review51:574-82.

Burke, Ronald J. and Carol A. McKeen. 1990."Mentoring in Organizations: Implications forWomen." Journal of Business Ethics 9:317-32.

Burt, Ronald S. 1980. "Autonomy in a Social Ty-pology." American Journal of Sociology 85:892-925.

. 1992. Structural Holes: The Social Struc-ture of Competition. Cambridge, MA: HarvardUniversity Press.

. 1995. "The Contingent Value of SocialCapital." Graduate School of Business, Univer-sity of Chicago, Chicago, IL. Unpublishedmanuscript.

Campbell, Karen E., Peter V. Marsden, andJeanne S. Hurlbert. 1986. "Social Resourcesand Socioeconomic Status." Social Networks8:97-117.

Coleman, James S. 1988. "Social Capital in theCreation of Human Capital." American Jour-nal of Sociology 94(Supp.):S95-S120.

Crozier, Michel. 1964. The Bureaucratic Phe-nomenon. Chicago, IL: University of ChicagoPress.

Dalton, Melville. 1959. Men Who Manage: Fu-sions of Feeling and Theory in Administration.New York: John Wiley.

Fernandez, Roberto M. and Roger V. Gould.1994. "A Dilemma of State Power: Brokerageand Influence in the National Health PolicyDomain." American Journal of Sociology 99:1455-91.

Friedman, Raymond A. and Joel Podolny. 1992."Differentiation of Boundary Spanning Roles:Labor Negotiations and Implications for RoleConflict." Administrative Science Quarterly37:28-47.

Gartrell, C. David. 1987. "Network Approachesto Social Evaluation." Pp. 49-66 in Annual Re-view of Sociology, vol. 13, edited by W. R.Scott and J. F. Short, Jr. Palo Alto, CA: An-nual Reviews Inc.

Granovetter, Mark S. 1974. Getting a Job. Cam-bridge, MA: Harvard University.

Hickson, David. J., C. R. Hinings, C. A. Lee,R. E. Schneck and J. M. Pennings. 1971. "AStrategic Contingencies Theory of Intraorgani-zational Power." Administrative Science Quar-terly 14:378-97.

Ibarra, Herminia. 1995. "Race, Opportunity, andDiversity of Social Circles in Managerial Net-works." Academy of Management Journal 38:673-701.

Ibarra, Herminia and Lynn Smith-Lovin. 1997."New Directions in Social Network Researchon Gender and Organizational Careers." InHandbook of Organizational Behavior, editedby S. Jackson and C. Cooper. Sussex, England:J. Wiley.

Kahn, Robert H., Donald M. Wolfe, RobertQuinn, and J. Diedrick Snoek. 1964. Organi-zational Stress: Studies in Role Conflict andAmbiguity. New York: John Wiley.

Kanter, Rosabeth M. 1977. Men and Women ofthe Corporation. New York: Basic Books.

Lin, Nan. 1982. "Social Resources and Instru-mental Action." Pp. 131-45 in Social Structureand Network Analysis, edited by P. V. Marsdenand N. Lin. Beverly Hills, CA: Sage.

Lin, Nan and Mary Dumin. 1986. "Access to Oc-


cupations Through Social Ties." Social Net-works 8: 365-85.

Lin, Nan, Walter M. Ensel, and John C. Vaughn.1981. "Social Resources and the Strength ofTies." American Sociological Review 46: 393-405.

Marcus, Philip M. and James S. House. 1973."Exchange between Superiors and Subordi-nates in Large Organizations." AdministrativeScience Quarterly 18:209-22.

Marsden, Peter V. 1982. "Brokerage Behavior inRestricted Exchange Networks." Pp. 201-19 inSocial Structure and Network Analysis, editedby P. V. Marsden and N. Lin. Beverly Hills:Sage.

Marsden, Peter V. and Karen E. Campbell. 1984."Measuring Tie Strength." Social Forces 63:482-501.

Mayhew, Bruce. 1980. "Structuralism versus In-dividualism: Shadowboxing in the Dark." So-cial Forces 59: 335-75.

Montgomery, James D. 1992. "Job Search andNetwork Composition: Implications of theStrength-of-Weak-Ties Hypothesis." AmericanSociological Review 57:586-96.

Podolny, Joel M. 1993. "A Status-Based Modelof Market Competition." American Journal ofSociology 98: 829-72.

Ragins, Belle R. and Dean B. McFarlin. 1990."Perceptions of Mentor Roles in Cross-CenderMentoring Relationships." Journal of Voca-tional Behavior 37: 321-39.

Roethlisberger, F. J. and William J. Dickson.1946. Management and the Worker. Cam-bridge, MA: Harvard University Press.

Simmel, Georg. [1922] 1955. Conflict and Webof Group Affiliations. Translated by K. Wolff

and R. Bendix. New York: Free Press.Simpson, Ida Harper. 1989. "The Sociology of

Work: Where Have the Workers Gone?" SocialForces 61\ 563-81.

Spekman, Robert E. 1979. "Influence and Infor-mation: An Exploratory Investigation of theBoundary Role Person's Basis of Power."Academy of Management Journal 22: 104-17.

Thomas, David A. 1993. "Racial Dynamics inCross-Race Developmental Relationships." Ad-ministrative Science Quarterly. 38: 169-94.

Van Sell, Mary, Arthur P. Brief, and Randall S.Schuler. 1981. "Role Conflict and Role Ambi-guity: Integration of the Literature and Direc-tions for Future Research." Human Relations34:43-71.

Warner, W. Lloyd and J. O. Low. 1947. The So-cial System of the Modern Factory. New Ha-ven, CT: Yale University Press.

Wellman, Barry, Peter J. Carrington, and AlanHall. 1988. "Networks as Personal Communi-ties." Pp. 130-84 in Social Structures: A Net-work Approach, edited by B. Wellman and S.D. Berkowitz. New York: Cambridge Univer-sity Press.

Wellman, Barry and Scott Wortley. 1990. "Dif-ferent Strokes from Different Folks: Commu-nity Ties and Social Support." American Jour-nal of Sociology 96: 558-88.

White, Harrison C , Scott Boorman, and RonaldL. Breiger. 1976. "Social Structure from Mul-tiple Networks, I: Blockmodels of Roles andPositions." American Journal of Sociology81:730-80.

Whyte, William Foote. 1949. "The Social Struc-ture of the Restaurant." American Journal of^ / 54:302-10.

podolny baron 1997

Documents