ArticlePDF Available

Competing in Groups

September 2004
Managerial and Decision Economics 25(6-7):453-471

September 2004
25(6-7):453-471

DOI:10.1002/mde.1201

Source
RePEc

Authors:

Tim Rowley

University of Toronto

Joel A. C. Baum

University of Toronto

Andrew V. Shipilov

INSEAD

Henrich Greve

INSEAD

Show all 5 authorsHide

In this study, we examine how the characteristics of clique structures affect the performance of firms embedded within the cliques. Although it is generally accepted in organization theory and strategic management that firms are embedded within ego and overarching industry networks that each affect their behavior and performance, there is little evidence on whether cliques are stable features of industry networks or affect firm behavior or performance. We theorize that the value of a clique to its members depends on (1) the network centrality of the clique and (2) the internal structure and organization (heterogeneity and inequality) of the clique. Our analysis of the Canadian investment banking industry from 1952 to 1990 provides empirical evidence of stable cliques, and indicates that while a clique's internal structure and organization materially affect firm-level benefits of clique membership, its positional embeddedness within the industry network does not. Copyright © 2004 John Wiley & Sons, Ltd.

Content uploaded by Joel A. C. Baum

Content may be subject to copyright.

MANAGERIAL AND DECISION ECONOMICS

Manage. Decis. Econ. 25: 453–471 (2004)

Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/mde.1201

Competing in Groups

Tim J. Rowley

, Joel A. C. Baum

*, Andrew V. Shipilov

Henrich R. Greve

and Hayagreeva Rao

Rotman School of Management, University of Toronto, Canada

Norwegian School of Management, Norway

Kellogg Graduate School of Business, Northwestern University, USA

In this study, we examine how the characteristics of clique structures aﬀect the performance of

ﬁrms embedded within the cliques. Although it is generally accepted in organization theory

and strategic management that ﬁrms are embedded within ego and overarching industry

networks that each aﬀect their behavior and performance, there is little evidence on whether

cliques are stable features of industry networks or aﬀect ﬁrm behavior or performance. We

theorize that the value of a clique to its members depends on (1) the network centrality of the

clique and (2) the internal structure and organization (heterogeneity and inequality) of

the clique. Our analysis of the Canadian investment banking industry from 1952 to 1990

provides empirical evidence of stable cliques, and indicates that while a clique’s internal

structure and organization materially aﬀect ﬁrm-level beneﬁts of clique membership,

its positional embeddedness within the industry network does not. Copyright #2004 John

Wiley & Sons, Ltd.

INTRODUCTION

Observation contradicts the traditional economic

view of ﬁrms as distinct and autonomous units of

action. Instead, ﬁrms are embedded in networks

comprised of close, robust and multidimensional

ties that blur organizational boundaries (Grano-

vetter, 1994; Powell and Smith-Doerr, 1994;

Gulati and Gargiulo, 1999). Interﬁrm networks

are collections of ﬁrms joined by ties that vary in

formality, but are stable and signiﬁcant enough to

create reasonably persistent interﬁrm structures.

Sparked by recognition of the increasing preva-

lence and economic signiﬁcance of interﬁrm net-

works, a thriving literature in organization theory

and strategic management now stresses the need to

understand how the social context in which ﬁrms

are embedded inﬂuences their behavior and

performance.

Early work focused on identifying general

motivations for interﬁrm collaboration including

reducing uncertainty that results from resource

requirements (Pfeﬀer and Salancik, 1978), acces-

sing new knowledge and complementary assets

(Kogut, 1988; Teece, 1992), and solving market

failures that emerge under conditions of bounded

rationality (Williamson, 1985, 1991). This work

locates the interﬁrm network within broader

justiﬁcations for institutions, building on Coase’s

(1937, 1988) transaction cost and March and

Simon’s (1958) and Cyert and March’s (1963)

behavioral theory of the ﬁrm. From this perspec-

tive, interﬁrm networks are institutional arrange-

ments for managing interdependence among

organizations. Under conditions of bounded

rationality (limited cognitive capacity and costly

information), exchange is inhibited by the inability

to foresee contingencies and observe all actions of

exchange partners (Coase, 1937, 1988; Simon,

1957, Williamson, 1975). Like the ﬁrm, the

interﬁrm network smoothes exchange by provid-

ing rules and understandings that limit costs and

uncertainty.

*Correspondence to: Rotman School of Management,

University of Toronto, 105 St. George Street, Toronto, Ontario

M5S 3E6, Canada. E-mail: Baum@rotman.utoronto.ca

Interdependence in the face of transaction costs

provides a useful explanation of factors inﬂuen-

cing the propensity of ﬁrms to form ties in the ﬁrst

place, but does not explain the choice of partners

or the kinds of network structures that may result.

Although interﬁrm relations can be a valuable

means of managing environmental uncertainty

and interdependence, there is also considerable

uncertainty associated with entering cooperative

ties. Given that imperfect information about

potential partners raises search costs and the risk

of exposure to opportunistic behavior, how do

ﬁrms choose partners? Recent work suggests that

ﬁrms’ ‘relational embeddedness’ in an immediate

circle of direct and indirect relations with other

ﬁrms and ‘positional embeddedness’ in the overall

network systematically inﬂuence their decisions

about cooperative ties. Work on relational em-

beddedness suggests that information about the

capabilities and trustworthiness of potential part-

ners obtained through previous direct ties (Gulati,

1995; Chung et al., 2000) and ties with potential

partners’ partners (Burt and Knez, 1995) shapes

ﬁrms’ search for partners. Work on positional

embeddedness indicates that the signaling property

of central network positions is important to ﬁrms’

partnering decisions because it introduces systema-

tic status diﬀerences among ﬁrms that travel

beyond their immediate circle of direct and

indirect ties (Podolny, 1993).

The signiﬁcance of interﬁrm networks as a form

of organization for carrying out economic activity

has been highlighted in several accounts of the

dependence of economic performance of indivi-

dual ﬁrms on their embeddedness within a net-

work (Burt, 1992; Podolny, 1994; Powell et al.,

1996). One line of work emphasizes the inﬂuence

of relational embeddedness within the direct and

indirect ties comprising a ﬁrm’s ego network on its

performance. This research examines how access

to complementary assets and resources and the

quality and composition of a ﬁrm’s ﬁrst- and

second-order ties inﬂuences the ﬁrm’s economic

performance, growth and survival (Uzzi, 1996,

1997; Baum et al., 2000). A second stream of

research focuses on the inﬂuence of a ﬁrm’s

positional embeddedness in an industry-wide net-

work on the ﬁrm’s performance. This work shows

how ﬁrms’ centrality in industry networks raises

their innovative (Powell et al., 1996) and economic

performance (Podolny, 1994; Rowley et al., 2000).

Taken together, these studies show how the direct

and indirect ties comprising a ﬁrm’s ego network,

and its position within the broader industry

network, shape the ﬁrm’s performance in several

important ways.

While the performance consequences of em-

beddedness in ﬁrm-level ego networks and indus-

try-level interﬁrm networks are increasingly well

understood, we know considerably less about

the role of intermediate network substructures

}‘cliques’}that lie between the ﬁrm and industry

levels (Dorian, 1992). Cliques are relatively stable

groups of ﬁrms, more densely interconnected to

one another than to other ﬁrms in the industry

network, and reproduced over time by repeated

interactions among a set of ﬁrms (Wasserman and

Faust, 1994). In a clique, the outcomes for ﬁrms

are fundamentally intertwined with those of other

ﬁrms that belong to the same clique. Clique

structures appear central to the frequently ob-

served industry networks composed of regions that

are more or less densely connected by relationships

(Nohria and Garcia-Pont, 1991; Gulati and

Gargiulo, 1999). Such industry networks are

typically decentralized, with no one ﬁrm connected

to most other ﬁrms, sparsely connected, with each

ﬁrm having few ties relative to the number of ﬁrms

in the industry, and locally clustered, in which

partners of partners are also frequently partners

(Walker et al., 1997; Watts, 1999; Kogut and

Walker, 2001). Such characteristics are consistent

with the simultaneous operation of ﬁrm and

industry level embeddedness processes (Baum

and Ingram, 2002).

When an industry contains cliques of intercon-

nected ﬁrms, competition is no longer solely ﬁrm

against ﬁrm, but instead a mixture of competition

among cliques and among ﬁrms within each clique

(Gomes-Casseres, 1994). Competition among busi-

ness networks such as the Keiretsu is an example

of such competition within systems of ﬁrms

combined with contests within the group for the

value generated by the group (Odagiri, 1994). The

organization of cliques may explain both their

competitive strength relative to other cliques and

their internal distribution of value.

Cliques thus resemble strategic groups, which

also subdivide an industry into groups of ﬁrms

that compete against other groups and internally

(Porter, 1979). Cliques have a stronger element of

coordination than strategic groups because they

are generated by network ties, whereas strategic

groups are generated by product-market and

T.J. ROWLEY ET AL.454

resource deployment similarities (e.g., Cool and

Schendel, 1987; Clark and Montgomery, 1999).

Cliques may, or may not, be comprised of ﬁrms

with similar product and resource portfolios.

Strategic groups may also contain elements of

coordination, however, making this diﬀerence less

important in practice than it seems at ﬁrst glance

(Peteraf and Shanley, 1997; Dranove et al., 1998).

Despite the evidence that interﬁrm networks are

often comprised of relatively stable cliques of ﬁrms

whose fates are shared, it remains unclear how

ﬁrms’ clique membership inﬂuences their perfor-

mance in general, and how clique characteristics

enhance or depress their members’ performance in

particular. This is the question we address in our

analysis. Below, we develop our theoretical frame-

work, in which the value of clique membership

depends on (1) the clique’s positional embedded-

ness within the industry network and (2) the

internal structure and organization of the clique.

We test our ideas in a study of Canadian

investment brokerage ﬁrms, which, as we show

below, competed in relatively stable groups, or

cliques, as they formed syndicates to underwrite

securities oﬀerings over time between 1952 and

1990.

THE VALUE OF CLIQUES

Cliques that can create value for their members are

likely to become stable. The value of clique

membership is a function of (1) the value created

by the clique and (2) the distribution of the value

created among clique members. These, in turn,

depend on (1) the networks of clique members,

and (2) the internal structure and organization of

the clique.

Clique Positional Embeddedness: Centrality in the

Industry Network

A clique’s ability to create value for its members is

high when the clique occupies a central position in

the industry network and one that places it in a

brokerage position vis- "

aa-vis the ﬁrms whose ex-

changes it mediates. Centrality refers to the extent

to which an actor is involved in relationships

within a network (Freeman, 1977; Bonacich, 1987;

Borgatti and Everett, 1992). The most central ﬁrm

in an industry network is the ﬁrm on the shortest

path between all pairs of other ﬁrms in the

network; the ﬁrm that is the fewest steps away

from reaching all ﬁrms in the network. At the

clique-level, centrality in the industry network

requires either each member of a clique to occupy

a central position in the industry network, or that,

taken together, the clique members’ relationships

produce a central position.

Central cliques have greater access to resources

and market opportunities (Podolny, 1993), and

more detailed information about ﬁrms’ capabil-

ities, needs and reliability (Powell et al., 1996).

Central cliques are also more visible, and more

likely to be in brokerage positions where they can

mediate exchanges among ﬁrms who lack direct

connections to each other (Burt, 1992). Central

cliques are located at the ‘crossroads’ of their

industry networks and are positioned to dispro-

portionately (and most quickly) amass informa-

tion circulated in the network and to inﬂuence the

network by gatekeeping this information. Cliques

occupying central positions are thus able to

arbitrage information and resources gained from

one part of the network to another, while less

central cliques experience greater constraint and

poorer performance.

The investment banking industry, our empirical

setting, is relationship-oriented because banks

must partner with one another to facilitate under-

writing deals (Eccles and Crane, 1988). Investment

banks are ﬁnancial intermediaries linking issuers

(ﬁrms) wishing to raise funds on capital markets to

investors. The underwriting process begins with

the issuer choosing a bank as lead manager to

oversee the underwriting responsibilities. In many

cases, the lead manager invites additional invest-

ment banks to participate as co-lead managers in

underwriting the issue as a means of spreading risk

and reaching a wider range of investors. Invest-

ment banks thus make two types of partnering

decisions that, in the aggregate, constitute invest-

ment banking cliques and the broader industry

network at any given time. As lead manager of an

issue, a bank must select co-lead managers for an

underwriting syndicate, and as a potential co-lead

manager, it must decide whether or not to

accept an oﬀer to participate in an underwriting

syndicate.

With underwriting fees generally constant

across the industry, improved performance is

largely a function of increasing market share

(transaction volume), which is achieved by

COMPETING IN GROUPS 455

obtaining greater access to underwriting opportu-

nities}either directly (as lead manager) through a

bank’s network of issuing organizations or indir-

ectly (as co-lead manager) through a bank’s

network of partners. By occupying central posi-

tions within the industry network, investment

banks are able to maximize the number of under-

writing deals to which they have access and

control. Information regarding potential under-

writing deals, other banks’ relationships and

investor interests is essential to successful invest-

ment banking, and receiving high quality and

diverse information sooner rather than later is a

vital source of competitive advantage.

Thus, consistent with their alternate moniker,

‘brokerage houses,’ we expect investment banks to

derive network advantages from occupying central

network positions (Rowley and Baum, 2002,

2004). Consequently, we expect investment bank

cliques situated in central positions to outperform

those that are less central. These arguments

suggest that, controlling for individual clique

members’ centrality:

Hypothesis 1 (H1):

The more central a clique is to its industry network,

the better the performance of its members.

Clique Structure and Organization:

Diversity and Inequality

The internal organization of a clique can be

characterized as a multidimensional space of

structural parameters (e.g., organizational size,

performance, or network position) along which

clique members are distributed and which reﬂects

and aﬀects members’ roles and associations with

one another (Blau, 1977, p. 3). The parameters

describe diﬀerences among clique members, and

two types of diﬀerences can be distinguished.

Diﬀerences on ranked characteristics such as

power or assets give rise to inequality}or vertical

diﬀerentiation}of the cliques. Diﬀerences on

unranked characteristics such as functional spe-

cialization create diversity}or horizontal diﬀer-

entiation}of the clique. The distinction between

diversity and inequality is important. Diversity is

related to greater variation in capabilities, infor-

mation, and ideas. Inequality, in contrast, is

related to diﬀerences in power and status that

allows more powerful members to extract a greater

share of the value.

Diversity. Diversity can have two distinct eﬀects of

the value of a clique. Diversity that reﬂects

diﬀerentiation in members’ resource needs and

capabilities lowers the potential for competition

and conﬂict among them, while at the same time

creating the potential for cooperation based on

complementary functional diﬀerences. Diversity can,

however, make it diﬃcult to eﬀectively distribute the

rewards of cooperation, particularly when the

diversity reﬂects diﬀerences in clique members’

contributions to the value created by the clique.

Diversity and complementarity. Organizations

diﬀerent in size and specializing in diﬀerent roles

typically compete in diﬀerent ways, for diﬀerent

resources, and using diﬀerent strategic and oper-

ating routines (Aldrich, 1979; Hannan and Free-

man, 1977; McKelvey, 1982). Such diﬀerentiation

reduces the likelihood for conﬂict in the group by

creating a beneﬁcial mutual dependence. Conver-

sely, cliques comprised of similarly sized ﬁrms,

specialized in similar roles, and thus require

similar resources to thrive (e.g., raw materials,

labor, ﬁnancial support, and customers) are likely

to experience intense within-clique competition

that undermines any potential value the clique

may create. The potential for intragroup conﬂict

also depends on how many clique members

perform similar functions or take on duplicate

roles. To a point, internal conﬂict can increase

ﬂexibility and foster innovation, but it can also

fragment the network as partners’ competing

interests pull in diﬀerent directions and appro-

priation concerns derail cooperative eﬀorts

(Gomes-Casseres, 1994). Such competition and

conﬂict pushes ﬁrms to seek out distinct functions

in which they hold a competitive advantage

(Hawley, 1950), and fosters creation of subgroups

in which ﬁrms fulﬁll complementary roles in which

they depend on, but do not compete directly with,

each other, lowering the potential for within-group

competition by reducing the number of direct

competitors in the clique. Increasing heterogeneity

thus increases the probability of interﬁrm relations

(Blau, 1977).

In investment banking, this suggests that

‘functionally integrated’ cliques comprised of a

mix of banks}some large, some small; some

specializing in the lead role, some in the co-lead

role}will outperform more homogeneous cliques

(Hawley, 1950). So, for example, a clique com-

prised of large investment banks specialized in the

lead manager role should perform poorly because

T.J. ROWLEY ET AL.456

competition among its members for syndicates will

undermine the value it creates for them. A clique

comprised of small, co-lead manager specialists

should also perform poorly, although less as a

result of competition among them than because of

their limited capabilities to attract deals and form

syndicates. These observations suggest that role

specialization and coordination may be funda-

mental to understanding the value created by a

clique. A bank specialized in the lead manager role

will improve its lot by belonging to a clique with

others specialized in the co-lead manager ro-

le}and vice versa. Therefore we hypothesize:

Hypothesis 2 (H2):

The more diverse a clique is in member size and

role specialization, the better the performance of

its members.

Diversity and distribution. To remain viable,

cliques must distribute the value they create so that

their members do not receive less value as a

member of the clique than they would by leaving

the clique (either for another clique, or to act

independently) (Fujiwara-Greve and Greve, 2000).

The distribution of a clique’s value among its

members is the result of negotiation among clique

members, and each member’s contribution to the

value created by the clique serves as a resource that

can be used to demand a higher share of the

created value (Pfeﬀer and Salancik, 1978). Because

central network positions are advantageous in

creating valuable brokerage opportunities across

structural holes (Burt, 1992), centrality represents

a resource that clique members can use to claim a

greater share of the value created by their clique.

Equitable distribution of the clique’s value can

occur only when multiple members occupy central

positions, not when the network of a single

member is central. Such a member would likely

leave the clique since it would unlikely be able to

capture the entire value creation for itself. Clique

members must thus need each other to some extent

for the clique to create value for its members.

However, as member diversity in network positions

increases, negotiations become more complex, and

it becomes increasingly diﬃcult to distribute

rewards satisfactorily, prompting conﬂicts among

clique members that can undermine their stability.

Hypothesis 3 (H3):

The more diverse a clique is in member centrality,

the poorer the performance of its members.

Inequality. Inequality, too, may have two distinct

eﬀects of the value of a clique. Inequality that

reﬂects diﬀerentiation in clique members’ resource

needs and capabilities can weaken the social

cohesion that supports value-creating interﬁrm

cooperative behavior within the clique (Coleman,

1988). Inequality in network position can, how-

ever, facilitate value creation in a clique as

diﬀerences in power and status facilitate clique

coordination and control by high-ranking ﬁrms,

and motivate lower ranking ﬁrms to seek endor-

sing aﬃliations with them.

Inequality and cohesion. While diversity that

reﬂects horizontal diﬀerentiation in members’ size

and role specialization can facilitate clique perfor-

mance by lowering the potential for intra-clique

competition and conﬂict, inequality}vertical dif-

ferentiation}on these dimensions can weaken the

social cohesion that supports interﬁrm coopera-

tion within the clique.

Homophily theory predicts preferential interac-

tion based on similarity. Lazarsfeld and Merton

(1954) distinguished two types of homophily:

status homophily, in which similarity is based on

demographic features that stratify, or diﬀerentiate

vertically, and value homophily, which is based on

values, attitudes and beliefs. Our focus here is on

status homophily, which implies that vertical

diﬀerentiation of ﬁrms translates into network

distance.

There are three reasons to expect status homo-

phily among ﬁrms. First, ﬁrms of similar rank are

more likely to collaborate because of the signaling

role of their hierarchical position: when a ﬁrm’s

quality is uncertain, the status of other ﬁrms with

which a ﬁrm interacts is used to gauge its quality

(Podolny, 1993). Indeed, interacting with low-

ranking ﬁrms may damage a high-ranking ﬁrm’s

own attractiveness (Podolny, 1993; Gulati and

Garigiulo, 1999). Thus, consistent with the ten-

dency toward homophily under conditions of

uncertainty (Podolny, 1994), in a vertically diﬀer-

entiated network, the likelihood of partnering

increases with similarity in status (Chung et al.,

2000).

Second, ﬁrms of similar size and specialization

typically compete in similar ways, for similar

resources, and using similar strategic and operat-

ing routines (Hannan and Freeman, 1977).

Although such ﬁrms are potentially direct compe-

titors, the similarity of their capabilities makes it

possible for them to cooperate more eﬀectively

COMPETING IN GROUPS 457

with one another (Lorange and Roos, 1992).

Third, dissimilarity in rank is likely to discourage

ﬁrms from committing to partnerships. The higher

ranked ﬁrm commits resources of the same quality

as the lower-ranked ﬁrm}which is less than it is

capable of committing}while the lower ranked

ﬁrm anticipates commitment of the higher ranked

ﬁrms capabilities. This commitment gap can

undermine any potential relationship (Chung

et al., 2000).

Inequality can also increase the cost of negotia-

tion and bargaining, severely limiting possible

bargaining outcomes available to clique members.

Johnson and Libecap (1982), for example, for-

mulate a model based on their observation of the

South Texas shrimp ﬁshery, in which ﬁshers are

diﬀerentiated vertically by their productivity. They

ﬁnd that neither ﬁsher-speciﬁc quotas nor bilateral

payments among ﬁshers (which amount to the

same thing) are practical to administer because

such schemes are too diﬃcult to monitor and

enforce. The only viable option, they conclude, is a

system of uniform quotas. The more productive

the ﬁsher, however, the larger the restriction

implied by this regime, and so more productive

ﬁshers will tend to oppose it.

In sum, higher-ranking clique members will

commit less to and beneﬁt less from group

cooperation. Consequently, they are more likely

to leave the clique since they would be unlikely to

capture a suﬃciently high share of its beneﬁts.

Large investment banks, for example, are likely to

extract proportionally less value from clique

membership than smaller banks, and so are less

willing to commit their substantial resources to the

clique. Similarly, high-ranking lead managers (i.e.,

banks that form many syndicates) are likely to

extract less value from or commit fewer high-

quality resources to a clique than banks that are

low in the lead manager rankings. Greater inequal-

ity on these dimensions exacerbates such pro-

blems, making the emergence of stable value-

creating cliques still less likely. Therefore, we

hypothesize:

Hypothesis 4 (H4):

The greater the inequality in a clique in member

size and role specialization, the poorer the

performance of its members.

Inequality, status enhancement, and coordina-

tion and control. As we noted earlier, network

positions can also be represented as a status order

in which a ﬁrm’s hierarchical rank is determined

by the centrality of its position in the industry

network and structural holes it spans (Bonacich,

1987; Borgatti and Everett, 1992; Burt, 1992;

Podolny, 1993). Central ﬁrms have greater access

to ﬁne-grained information about competencies,

needs, and reliability of potential partners and as a

better comprehension of the existing network

(Krackhardt, 1990; Powell et al., 1996). Firms

occupying central positions are gatekeepers in the

industry network, able to arbitrage information

and resources gained from one part of the network

in another. Firms in such positions are also more

visible, earn greater rewards in the market (e.g.,

higher revenues, margins, market share) and are

thus more attractive partners.

Status ﬂows through interconnections between

ﬁrms (Podolny, 1993). Ties to higher-status ﬁrms

enhance a ﬁrm’s own status, while ties to lower-

status ﬁrms detract from it (Faulkner, 1983; Stuart

and Podolny, 1996; Stuart et al., 1999). If a ﬁrm is

uncertain of the quality of the available partners,

or is unable to bear the search costs of investigat-

ing all the diﬀerent alternatives, then the regard

other market participants display for a given ﬁrm

provides a strong indication of the quality of the

ﬁrm.

These material and signaling beneﬁts encourage

low-status ﬁrms to try to enhance their own

position by engaging in strategic behavior, includ-

ing bargaining, exchange and coercion (Brager and

Holloway, 1978; Willer and Anderson, 1981), or

co-sponsorship and cooptation (Sharpe, 1985), to

establish and maintain interactions with higher

status ﬁrms (Leik, 1992), which may not have any

incentive to accept their advances. Thus, while

homophily theory predicts greater interaction

based on similarity, theories of status organizing

processes suggest that many interactions will be

directed upward in status. Notably, because status

distributions are nearly always positively skewed,

higher-status ﬁrms will generally have more

extensive downward relations than lower-status

ﬁrms will have upward relations. Thus, notwith-

standing the possibility that interacting with low-

ranking ﬁrms may damage a high-ranking ﬁrm’s

own attractiveness, except for the lowest ranking

ﬁrms, the probability of ﬁrms associating with

others below their status is typically greater

than the probability of their associating with

others equidistant above them. Higher inequality

T.J. ROWLEY ET AL.458

increases the likelihood of status-distant associa-

tion despite the tendency for homophily.

In addition to the foregoing status organizing

processes, inequality in network positions within a

clique also creates a basis for coordination and

control. Both centrality and structural holes are

associated with power. Firms with high network

centrality exhibit both higher levels of reputed

power and greater degrees of success in political

conﬂicts than less central organizations (e.g.,

Laumann and Knoke, 1987). Central ﬁrms gain

power from their access to resources and informa-

tion as well as choice of alternatives. They also

derive power from their information and control

advantages over other ﬁrms in the industry

network, to which they can strategically control

the ﬂow of information. As network position

inequality within a clique increases, the emergence

of a small number of powerful ﬁrms may thus

facilitate coordination and control of clique

members, and increasing the value created by the

clique.

Hypothesis 5 (H5):

The greater the inequality in a clique in member

centrality, the better the performance of its

members.

NETWORK DATA AND EMPIRICAL

SETTING

To study the relationship between alliance moves

and network positions over time in the Canadian

investment banking industry, we collected data

comprising all securities oﬀerings on all Canadian

stock exchanges}Vancouver, Alberta, Winnipeg,

Montreal, and Toronto}from 1952 to 1990,

inclusively. The sample covers all banks}domes-

tic and foreign}that participated in one or more

underwriting syndicates for a securities oﬀering on

a Canadian exchange. The Record of New Issues,

published yearly since 1952 by Financial Data

Group in Toronto, was the main data source and

provided the following information: issuer’s name,

date of issue, name of lead issuing investment bank

and co-lead investment banks, security type (debt,

bond, preferred share, common share, right and

unit), value of issue proceeds, and issue price and

volume. In 1952, 19 investment banks participated

in 27 underwriting syndicates in Canada. The life

histories of this subset of banks, which represents

5% of the 403 banks in the sample, are left-

censored. Over the observation period, the number

of syndicates grew rapidly, and by 1989, 83 banks

participated in 422 syndicates. Thus, although the

life histories for a small number of banks are left-

censored, our data cover the period during which

raising equity in the capital markets superceded

bank debt as the dominant mode of corporate

ﬁnancing (Davis and Mizruchi, 1999).

The history of partnerships among lead and co-

lead manager banks in underwriting syndicates is a

meaningful setting for our analysis of clique value

creation. Investment banks add value to corpora-

tions raising capital in primary markets by

eﬀectively pricing and placing their issues (Po-

dolny, 1993). The industry is characterized as

‘relationship-oriented’, because banks often colla-

borate in underwriting deals in order to reach a

wider market of investors and spread risk.

Relationships are not only a common practice

but also a vital resource, as ties are conduits to

underwriting opportunities and contribute to

ﬁrms’ reputations. At the network level, our

comprehensive network data and longitudinal

research design permit us to measure interorgani-

zational networks comprehensively, avoiding the

common problem of boundary-setting in network

analysis (Doreian and Woodard, 1992), and model

network dynamics over a period of time suﬃcient

to yield meaningful variation in network structure

and composition. At the ﬁrm level, our design

avoids the common sample selection problem of

over-representing currently successful organiza-

tions or particular points in time that can under-

mine inferences about factors producing

organizational success (Berk, 1983).

Network Deﬁnition

Industry networks were deﬁned based on member-

ships in underwriting syndicates. Networks were

constructed from adjacency matrices capturing the

number of times each bank participated in a

syndicate with each other bank for two-year

moving periods (i.e., 1952–1953, 1953–1954,

1954–1955, etc.). We used two-year periods

because syndicates can remain intact up to six

months or more prior to the date of the oﬀering.

Consequently, constructing the network based on

one-year periods would represent the network

inaccurately because syndicates that conclude in a

COMPETING IN GROUPS 459

given year may have been established in the prior

year, but not measured in that year. Moreover,

syndicate ties represent the visible manifestation of

relationships among banks. Banks participating in

syndicates together in a given year are also likely

to interact with each other in years proximate to

the syndicate. Our two-year moving period

approach to the network’s construction allows

for this possibility. These networks were used to

compute data on the characteristics of each bank’s

network position in the second year of each two-

year period. Thus, the 1952–1953 network was

used to measure banks’ network positions for

1953, the 1953–1954 network for positions in 1954,

etc.

Clique Deﬁnition

Cliques were also constructed based on syndicate

ties. We transformed each two-year adjacency

matrix into a similarity matrix, which we then

analyzed using hierarchical clustering techniques

to form dendograms from which to classify banks

into cliques. We deﬁned cliques at rescaled

combining distance of 10 (the rescaled maximum

was 25). To improve the reliability of the cliques,

we then constructed new adjacency matrices

indicating how many years within a given ﬁve-

year period (e.g., 1952–1957, 1953–1958, and so

on) each bank belonged to the same clique as each

other bank. We then repeated the process using the

ﬁve-year clique matrices to identify cliques. Final-

ly, we constructed a genealogy of the ﬁve-year

cliques, linking them over time. A clique was

considered ongoing from period tto period t+1 if

it retained more than 50% of its membership from

the previous period; in most cases the percentage

was far higher. Thus, if a clique identiﬁed from the

1952–1957 clique adjacency matrix shared more

than 50% of its members with a clique identiﬁed

from the 1953–1958 clique adjacency matrix, we

considered the 1953–1958 clique to be a continua-

tion of the 1952–1957 clique.

Using this procedure, we identiﬁed seven

cliques, with an average span of 15 years

(range 8–23). The cliques averaged 10 members

(range 3–21), of which, on average, 70% remained

the same from one year to the next. Between 1952

and 1990, there were 19 years in which there were

two bank cliques, 14 years in which there were

three, and six years in which there were four. The

average density of syndicate ties within these

cliques is 0.51; between them it is 0.02. This large

and signiﬁcant (p50.001) gap for within and

between clique syndicate ties corroborates the

meaningfulness of the cliques, and also indicates

that the investment banking industry network is

sparsely connected and locally clustered, which is

consistent with many previously studied industry

networks (e.g., Nohria and Garcia-Pont, 1991;

Walker et al., 1997; Gulati and Gargiulo, 1999).

Variables computed for these cliques were

assigned to banks in the ﬁnal year of each ﬁve-

year period, and based on their clique membership

in that year. Thus, characteristics of the 1952–1957

cliques were assigned to banks based on the cliques

to which they belonged in 1957, the 1953–1958

cliques for 1958, and so on. To avoid losing data

on the early observation years, for 1953–1956, we

measure clique characteristics based on the 1952-

1957 cliques.

Dependent Variable and Model Speciﬁcation

For the analysis, we pool the yearly data and

estimate a single model on the pooled cross-

sections using time series regression models. Each

bank is represented in the sample for the years in

which it is a member of a clique.

Pooling repeated

observations on the same organizations is likely to

violate the assumption of independence from

observation to observation and result in the

model’s residuals being autocorrelated. In parti-

cular, unmeasured diﬀerences in the strategy or

capability of banks lead to bank eﬀects on the

performance. Therefore, we estimated random-

eﬀects GLS models, which correct for autocorrela-

tion of disturbances due to constant ﬁrm-speciﬁc

eﬀects and has better small sample properties than

ﬁxed-eﬀects models, which are the main alternative

to random eﬀects (Tuma and Hannan, 1984).

A further complication arises from the fact that

our sample only includes observations for invest-

ment banks that participated in syndicated deals

and were assigned to cliques, which creates a

potential source of sample selection bias. We

corrected for this potential bias using Lee’s

(1983) generalization of Heckman’s (1979) two-

stage sample selection estimation procedure. This

entailed estimating a logistic regression model

predicting clique membership to calculate l, each

banks probability of being in a clique, which was

then included as a regressor in the market share

T.J. ROWLEY ET AL.460

models to estimate of the selectivity bias and

improve the consistency of coeﬃcient estimates.

We measured the dependent variable, ﬁrm

performance, as a bank’s market share at time t.

Investment bank performance is based primarily

on the ability to provide issuing ﬁrms and

investors with market advice, which is contingent

on being involved in many deals with many

investors and issuers (Podolny, 1993). As such,

high market share is an important objective in

investment banking. Also, because underwriting

fees and margins are relatively constant across the

industry (a standard underwriting fee of seven

percent is the norm), proﬁt increases are derived

primarily from increased market share/volume. To

measure market share at time t, we allocated the

dollar value (inﬂation adjusted) of each oﬀering

made during a year among the members of the

syndicate that underwrote the deal. For deals

involving a lead manager only (i.e., with no

syndicate), the bank was assigned 100% of the

oﬀering’s dollar value. For deals involving multi-

ple bank syndicates, we assigned the lead bank

50% of the underwriting value, and split the

remaining value among the co-lead banks equally.

We logged banks’ market shares to correct for the

skewness in its distribution.

In all our models, we include lagged market

share (logged) to predict the current year’s market

share. This helps to account for the possibility

that, even though well speciﬁed, our empirical

models likely still suﬀer from speciﬁcation bias due

to unobserved heterogeneity (Jacobson, 1990), and

permits us greater conﬁdence when inferring

causal relationships between independent and

dependent variables. In particular, if banks’

market shares are themselves a result of unob-

served factors aﬀecting performance, controlling

for the lagged dependent variable should eliminate

spurious eﬀects resulting from such endogeneity.

Independent Variables

To test H1, we computed each clique’s network

centrality in each year as the average betweenness

centrality of its members in that year. Betweenness

centrality measures the extent to which a bank falls

‘between’ any two other banks that are not

connected themselves (Freeman, 1979). The most

central bank in the industry network is the bank

on the shortest path between all pairs of other

banks in the network, and is thus the fewest steps

away from reaching all other banks in the

network. More formally, if g

is the number of

geodesic paths from bank ito bank jand g

ikj

is the

number of paths from bank ito bank jthat pass

through bank k, then betweenness centrality is the

proportion of geodesic paths from bank ito bank j

that pass through bank k:

B¼X

gikj

gij



Betweenness centrality was computed for each

bank for each of the two-year moving windows

used to construct the industry networks. In the

computations, values for each two-year window

were assigned to the second year of the period (i.e.,

the 1952–1953 count was used for 1953, and so

on...). Clique centrality was computed as the

mean centrality of its most central members,

speciﬁcally members in the top one-third of the

clique’s centrality distribution. We use this ap-

proach to avoid confounding the eﬀects of cliques

being composed of a relatively small number of

highly central banks, which better represent the

clique’s position in the industry network than the

larger number of less central clique members

that beneﬁt from the membership of the more

central banks. Values for clique centrality were

assigned to banks on the basis of their clique

memberships in each observation year, and

lagged one year for model estimation to avoid

simultaneity problems.

To test H2 and H3, we used the coeﬃcient of

variation, which is the standard measure of

diversity for continuous-valued variables. The

coeﬃcient of variation equals the standard error

of the structural variable divided by the mean:

C¼1=ðn1Þ½Pjxi%

xxðÞ

21=2

ð1=nÞPixi

;

where xis the structural variable and nis the

number of ﬁrms in the clique. It is larger for

greater diversity and bounded at the lower end by

zero.

We computed three coeﬃcients of variation for

each clique. The ﬁrst captured variation in clique

member size, measured as the dollar value of deals

each bank completed in a year. To measure bank

size at time t, we allocated the dollar value

(inﬂation adjusted) of each oﬀering in which a

bank participated according to the procedure

outlined above for market share. The second

COMPETING IN GROUPS 461

reﬂected variation in role specialization among a

clique’s members. For this purpose, we con-

structed a variable that indicated the extent to

which a bank specialized in the lead manager or

co-lead manager role. Operationally, the variable

was deﬁned as a bank’s number of lead manager

roles divided by the total number of syndicates in

which it participated in each year. Banks scoring

one (the maximum) on this measure specialized in

the lead role, forming all the syndicates in which

they participated; those scoring zero (the mini-

mum), specialized in the co-lead role, acting as co-

lead manager in all the syndicates in which they

participated. The third coeﬃcient captured varia-

tion in a clique’s members’ network centrality, as

deﬁned above. As before, these variables were

computed using the two-year moving window

industry networks, with the values for each two-

year period assigned to the second year of the

period, values assigned to banks on the basis of

their clique memberships in each observation year,

and then lagged one year to avoid simultaneity

problems.

To test H4 and H5, we used the Gini index,

which Blau (1977) recommended as a measure of

inequality for continuous structural parameters.

The Gini index operationalizes inequality as the

average gain in the parameter ﬁrms could get if

given the choice of trading places with a randomly

drawn other ﬁrm. This logic yields the following

equation:

G¼ð1=2n2ÞPiPjjxixjj

ð1=nÞPixi

Here, xis the parameter along which clique

members are ranked, and the expression simply

averages the diﬀerences in xfor all pairs of banks i

and j, scaled by the average x. It is larger for

greater inequality, and bounded by zero and one.

We computed measures of inequality among the

members of each clique in each year for the same

three variables as we computed coeﬃcients of

variation}bank size, manager role specialization,

and network centrality. Also, as before, the

variables were computed for the two-year moving

windows used to construct the interﬁrm networks,

with the count values for each two-year period

being assigned to the second year of the period

(i.e., the 1952–1953 count was used for 1953, and

so on...). Values of the variables were assigned to

banks on the basis of their clique memberships in

each observation year, and lagged one year for

model estimation to avoid simultaneity problems.

Control Variables

Many other factors may inﬂuence the performance

and network positions of investment banks.

Accordingly, our analysis controls for a baseline

model that includes a range of additional bank

characteristics and industry-speciﬁc environmental

factors.

Firm-level. We controlled for lagged eﬀects of

each bank’s size (logged) and role specialization.

We also controlled for ﬁrm-level network central-

ity, which prior theory and research on network

position eﬀects indicates can materially aﬀect ﬁrm

performance (e.g., Burt, 1992; Podolny, 1993).

Controlling for these ﬁrm-level variables, which we

used to construct the clique-level variables testing

H1–H5, helps ensure that eﬀects of clique level

speciﬁcations for these variables do not result

spuriously from uncontrolled eﬀects at the ﬁrm

level.

In addition to a bank’s position in the industry

network, the composition of a bank’s ego network

has also been shown to impact ﬁrm performance

(e.g., Uzzi, 1996, 1999). We therefore controlled

for each bank’s the ﬁrst-order (direct) and second-

order (indirect) network coupling. Each bank’s

ﬁrst-order network coupling (FONC) was com-

puted as

FONCi¼X

ðPij Wij Þ2;

where the N

is the number of banks tied to bank

i,P

is the percentage of the syndicates that bank i

participated in that also involved bank j, and W

the percentage of the total dollar volume of the

syndicates that bank iparticipated in that also

involved bank j. As FONC increases toward one,

bank i’s syndicates are increasingly formed with

small number of relationships, rather than dis-

persed among a larger number of partners. Our

variable diﬀers from Uzzi’s in its incorporation of

information on both the extent to which banks’

ties (P

) and resource ﬂows (W

) were concen-

trated individual players, rather than only the

former.

Second order network coupling (SONC) was

computed as the average ﬁrst-order network

T.J. ROWLEY ET AL.462

coupling of banks with which bank ihad direct

ties:

SONCi¼PjFONCj

When SONC is low, the network of banks with

which bank icooperates is spread out over a large

number of other banks, each of which accounts for

a small proportion of the syndicated deals. When

SONC is high, in contrast, the network to which

bank iis tied is comprised of a small number of

banks among which its syndicate deal are con-

centrated (Uzzi, 1996). A limitation of these

measures is that they are sensitive to the size of a

bank’s network. Therefore, we also included

counts of each bank’s number of lead and co-lead

manager roles in each year to controls for

individual bank’s network size (Uzzi, 1999).

Clique-level. At the clique level, in addition to the

variables testing H1–H5, we controlled for each

clique’s age in years and size in number of

members. We also controlled for each clique’s

within-clique density of ties. Finally, we included a

set of seven clique dummy variables, coded 1 for

the clique of which a bank was a member, and zero

otherwise. We suppressed the constant and in-

cluded dummies for all seven cliques. Values for

the clique-level control variables were assigned to

banks on the basis of their clique memberships in

each year.

Environmental factors

. Environmental controls

emphasized economic factors likely to inﬂuence

the issuer behavior and/or the underwriting

process. These included the prime interest rate,

the volume of shares traded on the Toronto Stock

Exchange, logged to normalize its distribution,

and the value of the Toronto Stock Exchange 300

market index, the largest and most inﬂuential

exchange and index in Canada. These variables

were measured annually, and lagged one year in

the analysis.

We also included a control for the logged total

value of all public oﬀerings, which may aﬀect the

number and type of partnering opportunities

available to banks, and in turn their performance.

This variable was computed based on the two-year

moving windows used to construct the industry

networks, and, as before, variable values for each

two-year period assigned to the second year of the

period, and then lagged one year for the analysis.

Lastly, we created dummy variables to control for

whether a particular deal took place in the 1950s,

1960s, 1970s or 1980s to account for possible addi-

tional period-speciﬁc shifts in bank performance.

Descriptive statistics for the independent and

control variables are given in Table 1. Correlations

among the variables are low to moderate, with few

larger than 0.5 (25% shared variance), and

unlikely to pose an estimation problem.

RESULTS

Table 2 presents estimates for random-eﬀects GLS

models of investment banks’ market share, and

gives w

statistics to compare the ﬁt of models.

Model 1 presents a baseline model, and Model 2

tests our hypotheses. The baseline model includes

lagged bank- and clique-level controls (including

clique ﬁxed eﬀects, which not reported for

simplicity), as well as the environmental controls.

Among the control variables, a bank’s lagged

market share and betweenness centrality, and its

clique’s size (number of members) are the main

predictors of its market share, with each positively

related to current market share.

Model 2, which adds the theoretical variables to

test for predicted eﬀects of clique centrality,

heterogeneity and inequality on banks’ market

share performance, provides a signiﬁcant improve-

ment over Model 1 (Dw

=34.4, 7 df, p50.001).

The coeﬃcient estimate for clique centrality is

not signiﬁcant in Model 2. We explored several

diﬀerent speciﬁcations of this variable (e.g.,

average centrality of top two-thirds, average

centrality of all members), but none was signiﬁ-

cant. One explanation for this non-ﬁnding is that

the beneﬁt of the clique having a core of highly

central banks is not distributed equally among the

clique’s members. For example, it is possible that

the core of central banks alone secure the beneﬁt

of their association with each other and the clique.

Another possibility is that competition among

clique’s most central banks for deals and network

position lowers their individual performance,

while less central members of the clique beneﬁt

from their rivalry (Rowley and Baum, 2002). To

examine these possibilities, we performed a sup-

plementary analysis in which we added to the

model a dummy variable coded one for the top

one-third centrality banks and zero otherwise, and

COMPETING IN GROUPS 463

Table 1. Descriptive Statistics

MeanStDev12345678910111213141516171819202122232425262728

1 Interest rate 7.55 2.90

2 Value shares traded

on TSE (logged)

27.37 1.38 0.80

3 TSE 300 market

index

1519.8 977.11 0.70 0.96

4 Total PO value

(logged)

23.11 1.93 0.68 0.91 0.91

5 Clique age

(number of years)

7.52 4.92 0.35 0.39 0.34 0.31

6 Clique size

(number of banks)

12.06 4.10 0.50 0.55 0.49 0.54 0.51

7 Clique density of

ties within

1.16 0.83 0.53 0.70 0.68 0.74 0.17 0.41

8 Clique 1 0.28 0.45 0.50 0.51 0.46 0.49 0.10 0.09 0.34

9 Clique 2 0.07 0.25 0.35 0.33 0.22 0.24 0.30 0.38 0.23 0.17

10 Clique 3 0.13 0.34 0.25 0.25 0.28 0.27 0.02 0.30 0.36 0.24 0.10

11 Clique 4 0.01 0.11 0.01 0.03 0.05 0.07 0.13 0.23 0.10 0.07 0.03 0.04

12 Clique 5 0.36 0.48 0.54 0.43 0.31 0.38 0.32 0.51 0.39 0.46 0.20 0.29 0.08

13 Clique 6 0.02 0.15 0.14 0.06 0.01 0.06 0.17 0.19 0.05 0.09 0.04 0.06 0.02 0.11

14 Clique 7 0.13 0.34 0.35 0.54 0.57 0.55 0.27 0.15 0.48 0.24 0.11 0.15 0.04 0.29 0.06

15 Bank market

share (logged)

0.01 0.02 0.14 0.23 0.26 0.28 0.04 0.09 0.15 0.09 0.11 0.01 0.07 0.05 0.08 0.22

16 Bank role specialization 0.47 0.35 0.09 0.05 0.03 0.02 0.05 0.10 0.00 0.01 0.07 0.01 0.05 0.05 0.02 0.00 0.19

17 Bank network centrality 2.68 3.93 0.06 0.07 0.07 0.09 0.03 0.09 0.09 0.01 0.02 0.17 0.06 0.20 0.05 0.05 0.43 0.26

18 Bank value of

deals (logged)

8.03 1.08 0.33 0.29 0.23 0.29 0.24 0.29 0.29 0.21 0.16 0.19 0.10 0.52 0.01 0.13 0.34 0.17 0.41

19 Bank number of

lead manager roles

13.69 33.37 0.21 0.22 0.19 0.21 0 0.09 0.18 0.1 0.1 0.1 00.10 0.34 0.18 0.33 0.49 0.28

20 Bank number of

co-lead manager roles

10.83 19.02 0.24 0.3 0.26 0.3 0.03 0.16 0.29 0.2 0.1 0.2 0.1 0.18 0.1 0.31 0.11 0.2 0.31 0.16 0.07

21 Bank First order

network coupling

0.24 0.32 0.26 0.22 0.17 0.21 0.11 0.26 0.23 0.11 0.21 0.13 0.18 0.32 0.02 0.03 0.17 0.23 0.32 0.2 0.2 0.51

22 Bank Secondorder

network coupling

0.17 0.16 0.22 0.16 0.09 0.15 0.13 0.16 0.16 0.13 0.13 0.13 0.06 0.33 0.03 0.07 0.06 0.24 0.15 0.03 0.1 0.14 0.04

23 Clique average centrality

(top 1/3)

5.59 2.41 0.18 0.15 0.22 0.27 0.13 0.33 0.39 0.07 0.05 0.56 0.17 0.57 0.13 0.20 0.04 0.03 0.25 0.11 0.16 0.40 0.23 0.21

24 Clique size heterogeneity 0.44 0.15 0.10 0.10 0.18 0.16 0.18 0.25 0.04 0.14 0.26 0.13 0.10 0.48 0.02 0.50 0.15 0.08 0.18 000.30 0.23 0.19 0.51

25 Clique role specialization

heterogeneity

0.18 0.09 0.34 0.22 0.09 0.06 0.29 0.35 0.13 0.07 0.43 0.03 0.22 0.36 0.47 0.00 0.01 0.03 0.04 0.1 0.1 0.16 0.13 0.11 0.09 0.41

26 Clique network centrality

heterogeneity

0.32 0.10 0.30 0.30 0.26 0.26 0.51 0.55 0.21 0.14 0.34 0.08 0.36 0.15 0.01 0.06 0.02 0.10 0.01 000.16 0.18 0.05 0.15 0.26 0.09

27 Clique size inequality 0.81 0.12 0.49 0.45 0.33 0.28 0.30 0.32 0.29 0.24 0.27 0.10 0.11 0.22 0.15 0.19 0.09 0.01 0.05 0.07 0.1 0.12 0.07 0.14 0.12 0.11 0.18 0.21

28 Clique role specialization

inequality

0.48 0.16 0.03 0.03 0.03 0.03 0.10 0.24 0.21 0.08 0.08 0.19 0.29 0.42 0.37 0.21 0.15 0.02 0.08 0.02 00.20 0.17 0.13 0.38 0.29 0.34 0.10 0.28

29 Clique: network centrality

inequality

0.55 0.10 0.25 0.25 0.27 0.25 0.28 0.52 0.32 0.18 0.21 0.37 0.34 0.29 0.27 0.11 0.04 0.10 0.08 0.1 0.1 0.20 0.25 0.10 0.42 0.32 0.17 0.64 0.06 0.56

T.J. ROWLEY ET AL.464

an interaction between this dummy variable and

the average centrality of the top one-third banks.

Neither the dummy variable nor the interaction

term was signiﬁcant, failing to support either of

the scenarios, and the estimates for the theoretical

variables remained unchanged. Thus, we conclude

that H1 is not supported}controlling for banks’

own network centrality, the centrality of the clique

to which the banks belong does not inﬂuence their

market share performance.

Table 2. Random Eﬀect GLS Models of Banks’Market Share

Variable Model 1 Model 2

Interest rate 0.0003 0.0003

0.0003 0.0003

Value of shares traded on TSE (logged) 0.0011 0.0011

0.0016 0.0017

TSE 300 marked index 0.0000 0.0000

0.0000 0.0000

Total PO value (logged) 0.0010 0.0010

0.0007 0.0008

Clique age (number of years) 0.0003 0.0001

0.0002 0.0003

Clique size (number of banks) 0.0005



0.0004



0.0002 0.0002

Clique density of ties within 0.0002 0.0013

0.0009 0.0010

Bank market share (logged) 0.6794



0.6685



0.0291 0.0295

Bank role specialization 0.0009 0.0013

0.0015 0.0014

Bank network centrality 0.0003



0.0003



0.0001 0.0001

Bank value of deals (logged) 0.0004 0.0005

0.0007 0.0007

Bank number lead manager roles/10 0.0000 0.0000

0.0002 0.0002

Bank number of co-lead manager roles/10 0.0002 0.0002

0.0003 0.0003

Bank ﬁrst order network coupling (FONC) 0.0045 0.0046

0.0072 0.0073

FONC squared 0.0016 0.0016

0.0068 0.0068

Bank second order coupling (SONC) 0.0019 0.0045

0.0072 0.0073

SONC squared 0.0056 0.0088

0.0092 0.0093

H1(+): Clique average network centrality (top 1/3) 0.0001

0.0004

H2(+): Clique size heterogeneity 0.0129



0.0064

H2(+): Clique role specialization heterogeneity 0.0285



0.0112

H3(): Clique network centrality heterogeneity 0.0130

0.0088

H4(): Clique size inequality 0.0062

0.0062

H4(): Clique: role specialization inequality 0.0171



0.0056

H5(+): Clique network centrality inequality 0.0256



0.0094

Heckman correction 0.0102



0.0104



0.0029 0.0030

Degrees of freedom 25 32

Chi-square 2369.95 2404.35

p50.10,



p50.05,



p50.01,



p50.001. Standard errors below coeﬃcients.

COMPETING IN GROUPS 465

H2, which predicted that ‘functionally in-

tegrated’ cliques comprised of a mix of banks}

some large, some small; some specializing in the

lead role, some in the co-lead role}would out-

perform more homogeneous cliques is partially

supported. The signiﬁcant positive coeﬃcient for

clique role specialization heterogeneity supports

the prediction; the signiﬁcant negative coeﬃcient

for clique size heterogeneity does not. One possible

explanation for these divergent estimates is that

while role specialization produces the kinds of

complementary diﬀerences among banks that we

anticipated would facilitate a clique’s perfor-

mance, such complementarities do not arise as a

result of diﬀerences in size. Instead, diﬀerences in

size foster distributional conﬂicts such as those we

predicted would result from heterogeneity in

network positions.

H3 is supported by the signiﬁcant (p50.10)

negative coeﬃcient for clique centrality hetero-

geneity. Thus, as we predicted, diversity in

members’ network positions can undermine the

value of the clique to its members by making it

diﬃcult to distribute rewards satisfactorily and

prompting internal conﬂicts. The signiﬁcant posi-

tive coeﬃcient for bank network centrality indi-

cates that more central banks earned higher

market shares, and thus, consistent with our

interpretation, diﬀerentially from the clique.

Turning to the inequality predictions, H4

receives partial support. Although the coeﬃcient

for clique size inequality is not signiﬁcant, the

coeﬃcient for clique role specialization inequality

is signiﬁcant, and negative as predicted. This

supports the idea that greater inequality in

manager roles creates commitment problems for

high-ranking lead managers (i.e., banks specialize

in forming syndicates), undermining the clique’s

ability to creating value for its members.

The signiﬁcant positive coeﬃcient for clique

centrality inequality supports H5, which predicted

that inequality in clique members’ network posi-

tions would facilitate coordination and control of

clique members by a small number of powerful

ﬁrms, increasing the value created by the clique.

In sum, while we ﬁnd no eﬀect of clique

centrality on clique member market share perfor-

mance (H1), the heterogeneity (H2–H3) and

inequality (H4–H5) predictions are supported for

role specialization and network centrality, but not

for bank size. As noted above, contradictory eﬀect

of clique size heterogeneity may have resulted

because diﬀerences in bank size create conﬂicts

rather than complementary diﬀerences among

banks.

Support for the opposing eﬀects of role specia-

lization heterogeneity and inequality and network

centrality heterogeneity and inequality point to an

interesting tradeoﬀ between heterogeneity and

inequality. Figure 1 illustrates these graphically,

both to aid in interpreting of our results and to

demonstrate the magnitude of eﬀects. The panels

in Figure 1 plot estimated market share as a

0.2

0.4

0.6 0.8 1

0.2

0.4

0.6

0.8

Role Heterogeneity

Role Inequality

Heterogeneity-Inequality Tradeoff

0.2

0.4

0.6 0.8 1

0.2

0.4

0.6

0.8

% Market Share % Market Share

Centrality Heterogeneity

Centrality Inequality

Heterogenity-Inequality Tradeoff

(b)

(a)

Figure 1. The heterogeneity-inequality tradeoﬀ.

T.J. ROWLEY ET AL.466

function of inequality and heterogeneity in role

specialization (Panel a) and network centrality

(Panel b) based on coeﬃcients in Model 2.

Although the magnitude of the market share

eﬀects may appear small, it is large relative to

the mean (standard deviation) of market shares

among banks in our sample of 2.33% (4.71%).

Figure 1(a) shows how increasing role hetero-

geneity within a clique leads to performance gains

for its members on the one hand (H2), and how

increasing role inequality lowers market share on

the other hand (H4). Notably, the slope of the

surface is steeper on the role heterogeneity axis

than on the role inequality axis, illustrating the

larger magnitude eﬀect of role heterogeneity on

clique members’ market shares. Figure 1(b) plots

the eﬀects for centrality inequality and hetero-

geneity. The ﬁgure shows the negative eﬀect of

centrality heterogeneity on clique members’ mar-

ket share (H3), and the positive eﬀect of centrality

inequality (H5). For centrality inequality, the slope

is steeper on the inequality axis, indicating a

stronger eﬀect of clique inequality than hetero-

geneity on clique members’ market shares. Taken

together, the panels in Figure 1 indicate that, for

the investment banks in our sample, complemen-

tary diﬀerences based on role heterogeneity and

coordination and control beneﬁts of network

centrality inequality are essential to increasing

the value created by the clique.

DISCUSSION AND CONCLUSION

It is generally accepted in organization theory and

strategy management that ﬁrms are ‘relationally’

and ‘positionally’ embedded in interﬁrm networks

that aﬀect their behavior and performance (e.g.,

Podolny, 1993; Gulati, 1995; Chung et al., 2000).

But, in addition to relational embeddedness in an

immediate circle of direct and indirect relations

with other ﬁrms and positional embeddedness

within overarching industry networks, ﬁrms are

also commonly embedded in intermediate network

substructures or ‘cliques’ involving cohesive sub-

groups of ﬁrms that are more densely intercon-

nected to one another than to other ﬁrms in the

industry network. Despite the apparent frequency

with which interﬁrm networks are observed to be

sparsely connected and locally clustered (e.g.,

Uzzi, 1996; Walker et al., 1997; Watts, 1999;

Kogut and Walker, 2001), there is little evidence

addressing whether such clique structures aﬀect

ﬁrm behavior or performance.

In this study, we examined how the character-

istics of clique structures, which fall between ﬁrm

and industry network levels of analysis, aﬀect the

performance of ﬁrms embedded within the cliques.

We conceptualized the value of a clique to its

members as a function of its positional embedded-

ness within the industry network and its internal

structure and organization.

First, we established the existence of clique

substructures within the broader industry net-

work. Using an iterative clustering procedure, we

identiﬁed seven long-lived and statistically distinct

cliques, with, on average, 70% of each clique’s

members remaining the same from year to year.

The high within clique and low between clique

densities of syndicate ties (0.51 and 0.02, respec-

tively) that characterized these cliques indicates

that the investment banking industry network is

decentralized, sparsely connected, and locally

clustered. This is consistent with previously

studied industry networks (e.g., Nohria and

Garcia-Pont, 1991; Walker et al., 1997; Gulati

and Gargiulo, 1999), and with the simultaneous

operation of ﬁrm and industry level embeddedness

processes (Baum and Ingram, 2002).

We predicted that a clique’s ability to create

value for its members is high when it occupies a

central position in the industry network (H1).

Although ﬁrm-level network centrality was sig-

niﬁcantly and positively related to bank perfor-

mance in our analysis, clique-level network

centrality was not (H1). We estimated several

diﬀerent speciﬁcations of the clique centrality

variable, none of which was signiﬁcant. We also

performed a supplementary analysis to examine

the possibility that the beneﬁts of clique centrality

were not equally distributed among clique mem-

bers, which yielded no support for the idea. Thus

we found no evidence to support the hypothesis

that the centrality of the clique in the industry

network creates value for its members.

A clique’s internal structure and organization,

however, did signiﬁcantly and materially aﬀect the

performance of its members. Following Blau

(1977), we conceptualized the internal organiza-

tion of a clique as a multidimensional space of

structural parameters representing the heterogene-

ity and inequality in role specialization, size, and

network centrality of the clique’s members.

COMPETING IN GROUPS 467

Heterogeneity in lead and co-lead manager role

specialization increased the value of a clique to its

members. This supports our prediction (H2) that a

clique creates value for its members when com-

prised of ﬁrms with complementary diﬀerences

that lead them to rely on one another rather than

compete. Contrary to H2, however, size hetero-

geneity among clique members did not appear to

contribute to their functional integration, but

rather decreased the value of the clique to its

members. Conﬂict among diﬀerently sized banks

over distribution of the value created by their

clique may account for this ﬁnding. Such a dispute

would be similar to the conﬂict we predicted, and

found evidence of, arising from heterogeneity in

network positions (H3).

While role specialization heterogeneity en-

hanced a clique’s ability to create value, role

specialization inequality reduced it signiﬁcantly

(H4). This ﬁnding is consistent with the prediction

that inequality impedes the development of social

cohesion among tightly linked social actors,

promoting cooperation and coordination. Inequal-

ity in network centrality, in contrast, increases the

performance of a clique’s members signiﬁcantly

(H5). This ﬁnding is consistent with the prediction

that inequality in centrality creates power diﬀer-

ences, allowing more central banks to control less

central banks, which acquiesce in order to gain

access to the better-positioned ﬁrms’ networks and

resources. Thus, while centrality heterogeneity

creates problems for determining how value is

distributed among clique members, centrality

inequality creates a pseudo-governance mechan-

ism for coordinating a clique’s syndication activ-

ities and distributing their value.

Taken together our theoretical analysis and

empirical ﬁndings point to a notable tension

between heterogeneity and inequality in role

specialization within a clique. On the one hand,

status homophily promotes equality in role spe-

cialization, fostering cohesion among clique mem-

bers, and enhancing cooperation. On the other

hand, heterogeneity in role specialization enhances

the value of a clique to its members by promoting

functional integration based on members’ com-

plementary diﬀerences. Thus, the value of a clique

to its members is optimized when the distribution

of members’ role specializations is at once homo-

philous and heterogeneous. A bimodal or ‘hour

glass’ distribution of clique members’ role specia-

lizations would appear to provide such a joint

optimization of homophily and heterogeneity by

creating two specialized subgroups, equal within

and heterogeneous between, i.e., a clique com-

prised of specialist banks, some specialized in the

lead manager role, others specialized in the co-lead

manager role.

However, as illustrated earlier in

Figure 1(a), for our sample banks, the positive

eﬀect of role specialization heterogeneity on clique

members’ market shares is larger than the negative

eﬀect of inequality; thus the optimal distribution

favors heterogeneity over homophily.

There is an analogous, but inverted, tension

between heterogeneity and inequality in clique

members’ network centrality. Inequality in clique

member centrality increases the ability of the

clique’s most central ﬁrms to coordinate and

control their subordinate partners. Conversely,

heterogeneity in clique member centrality leads to

instability and ineﬃciency because positional

heterogeneity increases the diversity of members’

contribution to the clique, which makes it diﬃcult

to distribute the value created equitably among the

clique’s members (Johnson and Libecap, 1982).

Again, a bimodal distribution of clique members’

network centralities jointly optimizes the eﬀects of

inequality and homogeneity by creating two

unequal, but homogeneous, subgroups within the

clique. As illustrated earlier in Figure 1(b),

however, for our sample banks, the positive eﬀect

of network centrality inequality on clique mem-

bers’ market shares is larger than the negative

eﬀect of heterogeneity, indicating that clique

structures favoring the control and coordination

beneﬁts of centrality inequality will outperform

those emphasizing the negotiation beneﬁts of

centrality homogeneity.

Our study suggests that ﬁrms can beneﬁt from

membership in cliques, but only if the clique’s have

particular structural and organizational features.

Our implicit assumption is that cliques that

provide greater value to their members would be

more stable. Future research examining clique

stability (i.e., entry and exits of members) is thus

an important next step in the study of mid-range

network phenomena. Additionally, our focus here

on how clique network position and internal

organization aﬀects member performance should

be supplemented by consideration of the compe-

titive dynamics among the cliques. To the extent

that clique members’ success depends on the

competitive strength that they build collectively,

competitive dynamics will likely occur at a

T.J. ROWLEY ET AL.468

clique-versus-clique level as well as ﬁrm-versus-

ﬁrm level (Gomes-Casseres, 1994). Such clique-vs-

clique competition alters the nature of competition

in a way that increases the competitive signiﬁcance

of a ﬁrm’s clique membership, and may (or may

not) lessen the importance of ﬁrm-level competi-

tion. This points to the importance of incorporat-

ing clique-level competition into models of clique

value creation, as well as to the need for measures

capturing the collective competitive strength of

cliques. Our ﬁndings suggest that measures of

cliques’ internal structure and organization may

serve as useful proxies for cliques’ competitive

strength, with more eﬀectively structured cliques

outcompeting those lacking functional integration

and coordination.

As with all single industry studies, we must

consider the generalizability of our ﬁndings.

Consistent with Burt’s (1992) argument, ﬁrms in

the investment banking industry acquire network

advantage by occupying central positions between

unconnected partners. In other industries, how-

ever, positions densely interconnected with part-

ners rather than positions between them provide

network advantage (Ahuja, 2000; Rowley et al.,

2000). It is likely that heterogeneity and inequality

in role specialization and network centrality will

both have diﬀerent eﬀects on ﬁrm performance in

these distinct competitive environments. This

contingency view provides an opportunity to

extend and reﬁne the implications derived from

our study.

Overall our ﬁndings indicate that, for Canadian

investment banks, a clique’s position does not

aﬀect its ability to create value for its members.

What does aﬀect a bank clique’s ability to create

value for its members is its internal structure and

organization. Clique advantage is realized through

heterogeneity in role specialization that creates

complementary functional diﬀerences and equality

in role specialization that facilities cohesion.

Clique advantage also derives from homogeneity

in banks’ network centrality and inequality in

network centrality that creates a basis for co-

ordination and control. Although there is an

inherent tension among these advantageous struc-

tural features, as noted, it is possible for cliques to

beneﬁt from both heterogeneity and inequality in

role specialization as well as network centrality.

For Canadian investment banks, however, this

tension is muted in part by the dominance of role

specialization heterogeneity over equality and

network centrality inequality over homogeneity

in creating value for clique members. Banks should

thus not join cliques with the goal of improving

their network advantage by virtue of the network

positions of the clique’s core ﬁrms; beneﬁts of

clique membership do not derive from greater

centrality of the clique to the industry network.

Rather, they derive from the clique’s internal

structure and organization, and in particular, the

functional integration of members’ specialized

roles and the coordination and control eﬀorts of

a few powerful ﬁrms.

Despite the apparent frequency of cliquey

industry networks, no previous study has exam-

ined how the characteristics of clique structures

falling between ﬁrm and industry network levels

aﬀect the performance of ﬁrms embedded within

them. Our study shows that they do, and identiﬁes

Blau’s (1977) work on heterogeneity and inequal-

ity as a conceptual foundation for characterizing

their internal structure and organization. We hope

our ﬁndings spark more ‘meso’ and ‘multilevel’

analyses of ﬁrm and industry networks.

Acknowledgements

This research was supported by a grant from the Social Sciences

and Humanities Research Council of Canada. For assistance

with data collection and coding we are grateful to Stan Li and

Danny Tzabbar.

NOTES

1. In a supplementary analysis, we dropped the

1953–1956 observations from the analysis. The

results are not substantively diﬀerent from the

ﬁndings we report below.

2. Several banks, typically with low network centrality,

participated in syndicates but did not belong system-

atically to any particular clique, and so were removed

from the data.

3. While our empirical approach may not resolve the

endogeneity problem entirely, our focus on clique-

level predictors of ﬁrm performance mitigates such

concerns. It is unlikely that clique-level characteristics

are endogenously determined in a substantive way by

ﬁrm-level capabilities. Nevertheless, as discussed

below, we control for ﬁrm-level equivalents for all

the clique-level variables as a further check against

spuriousness.

4. The Canadian Socioeconomic Information and

Management Database was our source for economic

data.

5. We conducted a series of simulations, varying the

distribution of role specialization that conﬁrmed this

intuition.

COMPETING IN GROUPS 469

REFERENCES

Ahuja G. 2000. Collaboration networks, structural

holes, and innovation: a longitudinal study. Adminis-

trative Science Quarterly 45(3): 425–455.

Aldrich HE. 1979. Organizations and Environments.

Prentice-Hall: Englewood Cliﬀs, NJ.

Allison PD. 1978. Measures of inequality. American

Sociological Review 43: 865–880.

Baum JAC, Calabrese T, Silverman B. 2000. Don’t go it

alone: alliance network composition and startups

performance in Canadian biotechnology. Strategic

Management Journal 21(Special issue): 267–294.

Baum JAC, Ingram P. 2002. Toward a behavioral

theory of the interﬁrm. In Augier M, March JG (eds).

The Economics of Choice,Change,and Organization:

Essays in the Memory of Richard M. Cyert. Edward

Elgar: Cheltenham, UK; 191–218.

Berk RA. 1983. An introduction to sample selection bias

in sociological data. American Sociological Review 48:

386–398.

Blau P. 1977. Inequality and Heterogeneity. The Free

Press: New York.

Bonacich P. 1987. Power and centrality: the family

of measures. American Journal of Sociology 92:

1170–1183.

Borgatti SP, Everett MG. 1992. Notions of position in

social network analysis. Sociological Methodology 22:

1–35.

Brager G, Holloway S. 1978. Changing Human Service

Organizations:Politics and Practice. Free Press: New

York, NY.

Burt R. 1992. Structural Holes:The Social Structure of

Competition. Harvard University Press: Cambridge,

MA.

Burt R, Knez M. 1995. Kinds of third-party eﬀects on

trust. Rationality and Society 7: 225–292.

Chung SA, Singh H, Lee K. 2000. Complementarity,

status similarity and social capital as drivers of

alliance formation. Strategic Management Journal

21(1): 1–22.

Clark BH, Montgomery DB. 1999. Managerial identi-

ﬁcation of competitors. Journal of Marketing 63

(July), 67–83.

Coase RH (ed.). [1937] 1988. The Firm,the Market and

the Law. University of Chicago Press: Chicago.

Coleman JS. 1988. Social capital in the creation of

human capital. American Journal of Sociology 94:

S95–S120.

Cool KO, Schendel DE. 1987. Strategic group forma-

tion and performance: the case of the U.S. pharma-

ceuticals industry, 1963–1982. Management Science

1102–1124.

Cyert R, March J. 1963. A Behavorial Theory of the

Firm. Prentice-Hall: Englewood Cliﬀs, NJ.

Davis GF, Mizruchi MS. 1999. The money center

cannot hold: commercial banks in the U.S. system of

corporate governance. Administrative Science Quar-

terly 44: 215–239.

Doreian P, Woodard KL. 1992. Fixed list versus

snowball selection of social networks. Social Science

Research 21: 216–233.

Dranove D, Peteraf M, Shanley M. 1998. Do strategic

groups exist? An economic framework for analysis.

Strategic Management Journal 19: 1029–1044.

Eccles RG, Crane DB. 1988. Doing Deals:Investment

Banks at Work. Harvard Business School Press:

Boston, MA.

Freeman LC. 1977. A set of measures of centrality based

on betweeness. Sociometry 40: 35–41.

Fujiwara-Greve T, Greve HR. 2000. Organizational

ecology and job mobility. Social Forces 79: 547–568.

Gomes-Casseres B. 1994. Group versus group: how

alliance networks compete. Harvard Business Review

62–74.

Granovetter M. 1994. Business groups. In Handbook

of Economic Sociology, Smelser NJ, Swedberg R

(eds). Princeton University Press: Princeton, NJ;

453–475.

Granovetter MS. 1985. Economic action and social

structure: the problem of embeddedness. American

Journal of Sociology 91: 481–510.

Gulati R. 1995. Social structure and alliance formation

patterns: a longitudinal analysis. Administrative

Science Quarterly 40: 619–652.

Gulati R, Gargiulo M. 1999. Where do interorganiza-

tional networks come from? American Journal of

Sociology 104(5): 1439–1494.

Hannan MT, Freeman JH. 1977. The population

ecology of organization. American Journal of Sociol-

ogy 82: 929–965.

Hawley A. 1950. Human Ecology. Ronald Press: New

York.

Heckman JJ. 1979. Sample selection bias as speciﬁcation

error. Econometrica 47: 153–161.

Jacobson R. 1990. Unobservable eﬀects and business

performance. Marketing Science 9: 74–85.

Johnson R, Libecap G. 1982. Contracting problems and

regulation: the case of the ﬁshery. American Economic

Review 72: 1005–1021.

Kennedy P. 1992. A Guide to Econometrics (3rd ed.).

MIT Press: Cambridge, MA.

Kogut B. 1988. Joint ventures: theoretical and empirical

perspectives. Strategic Management Journal 9:

319–332.

Kogut B, Walker G. 2001. The small world of Germany

and the durability of national networks. American

Sociological Review 66: 317–335.

Krackhardt D. 1990. Assessing the political landscape:

structure, cognition, and power in organizations.

Administrative Science Quarterly 35: 342–369.

Laumann EO, Knoke D. 1987. The Organizational

State:Social Choice in National Policy Domains.

University of Wisconsin Press: Madison, WI.

Lazarsfeld PF, Merton RK. 1954. Friendship as social

process: a substantive and methodological analysis,

pp. 298–348 in The Varied Sociology of Paul F.

Lazarsfeld, edited by P.L. Kendall. NY: Columbia

University Press.

Lee LF. 1983. Generalized economic models with

selectivity. Econometrica 51: 507–512.

Leik RK. 1992. New directions for network exchange

theory: strategic manipulation of network linkages.

Social Networks 14: 309–323.

T.J. ROWLEY ET AL.470

Lorange P, Roos J. 1992. Strategic Alliances. Blackwell:

Cambridge, MA.

Madhavan R. 2001. Does network structure really

matter? A meta-analytic synthesis of the evidence.

Working Paper, Katz Graduate School of Business,

University of Pittsburgh.

Madhavan R, Koka B, Prescott J. 1998. Networks in

transition: how industry events (re)shape interﬁrm

relationships. Strategic Management Journal 19:

439–459.

March JG, Simon HA. 1958. Organizations. New York:

Wiley.

McKelvey B. 1982. Organizational Systematics:Taxon-

omy,evolution,classiﬁcation. University of California

Press: Berkeley, CA.

Nohria N, Garcia-Pont C. 1991. Global strategic

linkages and industry structure. Strategic Management

Journal 12: 105–124.

Odagiri H. 1994. Growth Through Competition,Compe-

tition Through Growth:Strategic Management and the

Economy in Japan. Oxford University Press: Oxford,

UK.

Peteraf M, Shanley M. 1997. Getting to know you:

a theory of strategic group identity. Strategic

Management Journal 18 (Summer Special Issue):

165–186.

Pfeﬀer J, Salancik GR. 1978. The External Control of

Organizations. Harper and Row: New York.

Podolny JM. 1994. Market uncertainty and the social

character of economic exchange. Administrative

Science Quarterly 39(3): 458–470.

Podolny JM. 1993. A status-based model of market

competition. American Journal of Sociology 98(4):

829–872.

Porter ME. 1979. The structure within industries and

companies’ performance. Review of Economics and

Statistics 214–227.

Powell WW, Koput KW, Smith-Doerr L. 1996. Inter-

organizational collaboration and the locus of innova-

tion: networks of learning in Biotechnology.

Administrative Science Quarterly 41: 116–145.

Powell WW, Smith-Doerr L. 1994. Networks and

economic life. In The Handbook of Economic Sociol-

ogy Smelser NJ, Swedberg R (eds). Princeton Uni-

versity Press: Princeton, NJ; 368–402.

Rowley T, Behrens D, Krackhardt D. 2000. Redundant

governance structures: an analysis of structural and

relational embeddedness in the steel and semiconduc-

tor industry. Strategic Management Journal 21:

369–386.

Rowley T, Baum JAC. 2002. The dynamics of network

positions and strategies. Paper presented at the

Academy of Management meetings, Denver, CO,

August.

Rowley TJ, Baum JAC. 2004. Sophistication of inter-

ﬁrm network strategies in the Canadian investment

banking industry. Scandinavian Journal of Manage-

ment. Forthcoming.

Sharpe, LJ. 1985. Interorganizational policy-making:

the limits of subnational autonomy. In Guidance,

Control and Evaluation in the Public Sector:The

Bielefeld Interdisciplinary Project, Kaufmann FX,

Majone G, Ostrom V (eds). De Gruyter: Berlin;

159–181.

Simon HA. 1957. Models of Man. New York: Wiley.

Stuart TE, Podolny JM. 1996. Local search and the

evolution of technological capabilities. Strategic

Management Journal 17 (Summer Special Issue):

21–38.

Stuart TE, Hoang H, Hybels RC. 1999. Interorganiza-

tional endorsements and the performance of entrepre-

neurial ventures. Administrative Science Quarterly

44(2): 315–349.

Teece DJ. 1992. Competition, cooperation, and innova-

tion: organizational arrangements for regimes of rapid

technological progress. Journal of Economic Behavior

and Organization 18: 1–25.

Teece DJ, Pisano G, Shuen A. 1997. Dynamic capabil-

ities and strategic management. Strategic Management

Journal 18(7): 509–533.

Tuma NB, Hannan MT. 1984. Social Dynamics:Models

and Methods. New York: Academic Press.

Uzzi B. 1996. The sources and consequences of

embeddedness for the economic performance of

organizations: the network eﬀect. American Socio-

logical Review 61: 674–698.

Uzzi B. 1997. Social structure and competition in

interﬁrm networks: the paradox of embeddedness.

Administrative Science Quarterly 42: 35–67.

Uzzi B. 1999. Embeddedness in the making of ﬁnancial

capital: how social relations and networks beneﬁt

ﬁrms seeking ﬁnancing. American Sociological Review

64: 481–505.

Walker G, Kogut B, Shan W. 1997. Social capital,

structural holes and the formation of an industry

network. Organization Science 8(2): 109–125.

Wasserman S, Faust K. 1994. Social Network Analysis:

Methods and Applications. Cambridge University

Press: Cambridge.

Watts D. 1999. Networks, dynamics and the small world

phenomenon. American Journal of Sociology 105:

493–527.

Willer D, Anderson B (eds). 1981. Network, Exchange

and Coercion. Elsevier: New York, NY.

Williamson OE. 1975. Markets and Hierarchies. New

York: Free Press.

Williamson OE. 1985. The Economic Institutions of

Capitalism:Firms,Markets,Relational Contracting.

Free Press: New York.

Williamson OE. 1991. Comparative economic organiza-

tion: The analysis of discrete structural alternatives.

Administrative Science Quarterly 36: 269–296.

COMPETING IN GROUPS 471

How do different types of interorganizational ties matter in technological exploration?

Article

Full-text available

May 2019
MANAGE DECIS

Wim Vanhaverbeke

Purpose-Boundary-spanning exploration through establishing alliances is an effective strategy to explore technologies beyond local search in innovating firms. The purpose of this paper is to argue that it is useful to make a distinction in boundary-spanning exploration between what a firm learns from its alliance partners (explorative learning from partners (ELP)) and what it learns from other organisations (explorative learning from non-partners (ELN)). Design/methodology/approach-The authors contend that alliances play a role in both types of exploration. More specifically, the authors discern three types of alliances (inside ties, clique-spanning ties and outside ties) based on their role vis-à-vis existing alliance cliques. Clique members are highly embedded, and breaking out of the cliques through clique-spanning and outside alliances is crucial to improving explorative learning. Thereafter, the authors claim that clique-spanning ties and outside ties have a different effect on ELN and ELP. Findings-The empirical analysis of the "application specific integrated circuits" industry indicates that inside ties have negligible effects on both types of explorative learning. Clique-spanning ties have a positive effect on ELP, but not on ELN. The reverse is true for outside ties. The results show that research on explorative learning should devote greater attention to the various roles alliance partners and types of alliances play in advancing technological exploration. Originality/value-The literature only emphasises the learning from partners, focussing mainly on accessing their technology. In sum, alliance partners play different roles in exploration, and their network position influences the role they are able to play.

Status quo and challenges in air transport management research

Article

Full-text available

Jun 2024

Air transport management research, concerned with all facets of aviation operations, policies, and strategies, is an essential element of making our aviation system more sustainable and preparing it for the challenges inherent to the present and future. Based on a data-driven categorization of almost 2,000 papers published on the subject, we discuss the status quo in air transport management research. Through our data-driven categorization we have identified 15 broad topics. For each topic, we provide a description of the state of the art and propose 2-3 challenges , respectively. Overall, our study provides a set of 35 challenges to the research community. Accordingly, we hope and believe that our study makes a valuable contribution, mainly by guiding the air transport management research community towards a delineated work plan on the research landscape of air transport as well as the present challenges, ultimately helping to improve the global air transport system.

United we stand? Organizational groups and spinoff mortality in the context of academic entrepreneurship

Article

Full-text available

Jan 2024
J BUS VENTURING

We study failures between 1993 and 2017 in the complete population of 1731 English and Scottish university spinoffs founded since 1977. We borrow and expand the concept of density dependence from organizational ecology to theorize that a spinoff's propensity to fail is affected by the number of spinoffs active not only in the aggregate population but also within its parent university's portfolio. We contribute to organizational theory, demonstrating the importance of organizational groups that form within larger populations on individual organizations' propensity to fail. We contribute to literature on academic entrepreneurship showing that, for most universities, spinoff portfolio growth can lower associated spinoffs' failure rates, but that such effects need to be juxtaposed to the aggregate population's finite capacity to support an expanding number of spinoffs.

The role of network community characteristics for firms' rapid business scaling

Article

Sep 2023
TECHNOL FORECAST SOC

Which factors lead to firms' massive rapid business scaling? The prior work has predominantly focused on factors inside the firm, assuming that the possession of internally controlled idiosyncratic resources and capabilities largely explains the phenomenon. We extend this work by arguing that the alliance network community in which a scaling firm is embedded represents a distinct valuable channel through which it can access and orchestrate resources needed for its rapid growth. Network communities are structural groups of firms that are densely connected internally and sparsely linked to firms in other groups. We focus on three key characteristics of network communities: size, density, and stability. We theorize that the access to and orchestration of diverse and complementary network community resources influence firm rapid scaling through the expansion, replication, and synchronization of resources and business practices. Using a sample of firms publicly traded on the North American stock exchanges and observed over 1990–2020, we find that these firms' business scaling positively relates to the size and density of their network communities, yet the relationship between firm scaling and network community stability is negative.

Unpacking the intellectual structure of ecosystem research in innovation studies

Article

Full-text available

Jul 2023
RES POLICY

In light of the scattered nature of extant ecosystem literature, we seek to (1) identify the intellectual structure and theoretical roots of ecosystem literature within the broader field of innovation, and (2) take stock of how ecosystem research enriches and extends the innovation literature. Based on a bibliometric analysis of innovation and management journals over the past 30 years, we identify three lineages of ecosystem research within the "umbrella" of innovation studies. The first lineage finds its theoretical roots in the literature of product and architecture innovation. The second follows the lineage of open innovation and innovation strategy. The third adopts the traditions of regional innovation and entrepreneurship. Further, we offer an extended-boundary model that features two sets of synopses, explicating ecosystems' contributions to the field of innovation. The first set demonstrates how the ecosystem concept extends the three lineages with its unique attributes from a meta-organisational perspective, i.e., consumption-side synergies, boundary spanning and self-organisation. The second indicates how the concept may be repositioned on the foundational economic theories, i.e., Schumpe-terian Innovation and Transaction Cost Economics, and advance how it may integrate the two schools and extend their research implications vis-à-vis profiting from, organising for, and outcomes of innovation. We contribute by organising the fragmented ecosystem literature into an integrative framework, forging stronger links with traditional innovation theories, and offering a cumulative theoretical basis for this promising concept to further flourish in the field of innovation studies.

Behavioral antecedents of firm's ego-network competitiveness: The case of the global pharmaceuticals

Article

Full-text available

Mar 2023
LONG RANGE PLANN

Intra-industry alliance networks provide a firm with both collaborative opportunities and competitive challenges. When forming alliance networks within a particular industry, firms need to consider to which extent they compete in the same markets with their alliance network partners, who are also their industry peers. However, previous literature has not exhaustively addressed the antecedents of a firm's competitive behavior with their alliance network peers – i.e., a phenomenon we label ego-network competitiveness. This study draws on extensive panel data from the top global pharmaceuticals to examine this question. Combining behavioral and network perspectives, we test two competing hypotheses on how ego-network competitiveness varies relative to performance feedback and whether structural prominence moderates this relationship. Our results show that performance above and below aspirations increases ego-network competitiveness through high-intensity responses (i.e., problemistic and slack search). We also find that performance above aspirations increases ego-network competitiveness for firms with high structural prominence.

How much does the firm's alliance network matter?

Article

Jan 2022

Research summary Extant empirical work partitioning the variance in firm (business segment) profitability has identified industry, corporate parent, business segment and time as key sources. However, this variance decomposition research stream has treated firms as atomistic, autonomous entities. We employ a fast-unfolding community-detection algorithm to detect firms' network memberships and use the Shapley Value method to isolate the effect of the firm's alliance network, in addition to industry, corporate parent, business segment and year effects, on the variance in business unit performance. Our findings demonstrate that the effect of the firm's alliance network explains 11% of the variance in firm ROA among 16,381 business segments from 1979 through 1996. We also extend the time period through 2018 and find that our results broadly hold. Managerial summary In the search for superior firm performance, managers typically focus their attention externally on profitable industries in which to operate, as well as internally on their firms’ idiosyncratic and valuable resources and capabilities. In addition to these profitability sources, our work suggests another important, but heretofore overlooked, factor in the managerial quest for competitive advantage: the value-creating potential of alliance networks. We employ a machine learning algorithm to detect firms' network memberships. Our findings indicate that as much as 11% of the variance in firm profitability (ROA) is explained by the network of alliances of which the firm is a part. Our study also implies that the emphasis on networks continues to be relevant in a technology age in which the industry boundaries are blurring. This article is protected by copyright. All rights reserved.

The synergy of inventor cooperative network dual embeddedness and firm innovation: the mediating role of ambidextrous learning

Article

Jan 2022

Dual embeddedness of inventors’ cooperative networks influences firm innovation. This paper explores the mechanism of synergy between network structural embeddedness (network centrality and structural hole) and relational embeddedness (relational strength) on firm innovation using data from Chinese listed automobile companies. This paper further explores the mediating role of exploitative learning and exploratory learning between the synergy of cooperative network dual embeddedness and innovation. The results show that the synergy of network centrality and relational strength hinders innovation, while the synergy of the structural hole and relational strength promotes innovation. The synergy of centrality and relational strength hinders exploitative learning, while the synergy of the structural hole and relational strength promotes exploitative learning. Exploitative learning plays a mediating role between synergy of structure and relational embeddedness and innovation.

Structural Susceptibility to Corruption in FIFA: a Social Network Analysis

Article

Oct 2020

Voting and decision making in FIFA was in 2011 shifted from the executive committee to the FIFA congress with the aim to reduce risk of corrupt vote mobilisation in these elections. A key rationale behind this decision was that it would not be financially feasible to bribe a majority of the congress. However, the network structure with asymmetries of influence can still be susceptible to corruption. The aim of this study is therefore to analyse the potential susceptibility of the organisation structure to a grand vote-rigging corruption scheme of FIFA World Cup host elections by means of a social network analysis approach. Based on a theoretical notion of vote mobilisation and inter-organisational networks, the organisational structure is evaluated to identify potential susceptibility to corruption. Two possible regimes were analysed, the centrally controlled and horizontally controlled, based on structural cohesion, embeddedness, network centralisation and correlation between actor centrality distributions. Findings do not suggest a susceptibility to a centrally controlled, but to a horizontally controlled vote-rigging corruption scheme. After the social network analysis was conducted, the holders of five out of eleven of the identified vulnerable positions in the network have been charged with allegations of corruption. This is a strong indicator of the validity of the method to assess the structural susceptibility to corruption.

A foot in two camps or your undivided attention? The impact of intra- and inter-community collaboration on firm innovation performance

Article

Jan 2020

Many inter-organisational networks exhibit a community structure. However, the level of understanding of the effect of firms’ embeddedness in a network community on their innovation performance is limited. This study attempts to fill this gap by investigating the impact of intra- and inter-community collaboration strength on firms’ innovation performance, with patent data pertaining to the new-energy vehicle industry during the period from 2000 to 2017. The results show that, first, firms with moderate levels of intra-community collaboration strength achieve the greatest innovation performance, as these firms obtain rapid and efficient access to knowledge within their community. Second, firms with moderate levels of inter-community collaboration strength achieve the greatest innovation performance, as they acquire heterogeneous knowledge from outside the community. Third, the interaction of intra- and inter-community collaboration strength has a significant negative effect on firms’ innovation performance. This finding suggests that simultaneously building collaborations inside and outside the community is not an ideal network strategy for firms to employ.

Organizational Systematics: Taxonomy, Evolution, Classification

Book

Dec 1982

Bill McKelvey

Growth through Competition, Competition through Growth

Book

May 1994

Hiroyuki Odagiri

Redundant governance structures: an analysis of structural and relational embeddedness in the steel and semiconductor industries

Article

Mar 2000
STRATEGIC MANAGE J

Network researchers have argued that both relational embeddedness—characteristics of relationships—and structural embeddedness—characteristics of the relational structure—influence firm behavior and performance. Using strategic alliance networks in the semiconductor and steel industries, we build on past embeddedness research by examining the interaction of these factors. We argue that the roles relational and structural embeddedness play in firm performance can only be understood with reference to the other. Moreover, we argue that the influence of these factors on firm performance is contingent on industry context. More specifically, our empirical analysis suggests that strong ties in a highly interconnected strategic alliance network negatively impact firm performance. This network configuration is especially suboptimal for firms in the semiconductor industry. Furthermore, strong and weak ties are positively related to firm performance in the steel and semiconductor industries, respectively. Copyright © 2000 John Wiley & Sons, Ltd.

Don't go it alone: alliance network composition and startups' performance in Canadian biotechnology

Article

Mar 2000
STRATEGIC MANAGE J

We combine theory and research on alliance networks and on new firms to investigate the impact of variation in startups’ alliance network composition on their early performance. We hypothesize that startups can enhance their early performance by 1) establishing alliances, 2) configuring them into an efficient network that provides access to diverse information and capabilities with minimum costs of redundancy, conflict, and complexity, and 3) judiciously allying with potential rivals that provide more opportunity for learning and less risk of intra‐alliance rivalry. An analysis of Canadian biotech startups’ performance provides broad support for our hypotheses, especially as they relate to innovative performance. Overall, our findings show how variation in the alliance networks startups configure at the time of their founding produces significant differences in their early performance, contributing directly to an explanation of how and why firm age and size affect firm performance. We discuss some clear, but challenging, implications for managers of startups. Copyright © 2000 John Wiley & Sons, Ltd.

COLLABORATION NETWORKS, STRUCTURAL HOLES AND INNOVATION: A LONGITUDINAL STUDY.

Article

Aug 1998

Gautam Ahuja

I find that a firm's innovation output increases with the number of collaborative linkages maintained by it, the number of structural holes it spans, and the number of partners of its partners. However, innovation is negatively related to the interaction between spanning many structural holes and having partners with many partners.

Group versus Group: How Alliance Networks Compete

Article