ArticlePDF Available

Technological exploration : a longitudinal study of the role of recombinatory search and social capital in alliance networks /

January 2003

January 2003

Authors:

University of Oklahoma

Typescript. Thesis (Ph. D.)--New York University, Graduate School of Business Administration, 2003. Includes bibliographical references (leaves 232-269).

Motorola's 1992 Ego Network Structure of Technology Alliances a

…

Figures - uploaded by Corey C. Phelps

Content may be subject to copyright.

Content uploaded by Corey C. Phelps

Content may be subject to copyright.

A LONGITUDINAL STUDY OF THE INFLUENCE OF ALLIANCE

NETWORK STRUCTURE AND COMPOSITION ON FIRM

EXPLORATORY INNOVATION

COREY C. PHELPS

HEC Paris

This study examines the influence of the structure and composition of a firm’s alliance

network on its exploratory innovation—innovation embodying knowledge that is

novel relative to the firm’s extant knowledge. A longitudinal investigation of 77

telecommunications equipment manufacturers indicated that the technological diver-

sity of a firm’s alliance partners increases its exploratory innovation. Further, network

density among a firm’s alliance partners strengthens the influence of diversity. These

results suggest the benefits of network “closure” (wherein a firm’s partners are also

partners) and access to diverse information can coexist in an alliance network and that

these combined benefits increase exploratory innovation.

A core area of research on strategic alliances con-

cerns their influence on firm performance (Gulati,

1998). Within this domain of inquiry, researchers

have often characterized alliances as wellsprings of

innovation and new capabilities (e.g., Hamel, 1991;

Leonard-Barton, 1995). Many studies have shown

the alliance networks in which firms are embedded

can enhance firm learning and innovation (e.g.,

Ahuja, 2000; Shan, Walker, & Kogut, 1994; Smith-

Doerr et al., 1999; Soh, 2003). Despite this evi-

dence, substantial opportunity exists to expand un-

derstanding of how and under what conditions

alliance networks influence firm innovation. A re-

view of nearly 40 years of research published in 12

leading management and social science journals

(Phelps, Heidl, & Wadhwa, 2010) showed the liter-

ature on alliances and firm innovation is limited in

at least four important respects.

First, although some research has examined the

influence of alliance network structure on firm in-

novation, the composition of firms in these net-

works has received little attention. Network struc-

ture refers to the pattern of relationships that exist

among a set of actors, and network composition

refers to the types of actors in a network character-

ized in terms of their stable traits, features, or re-

source endowments (Wasserman & Faust, 1994).

Recent research has recognized that alliance net-

work studies have largely ignored network composi-

tion and has called for more attention in network

research to the heterogeneity of the resources of firms

in networks (Lavie, 2006; Maurer & Ebers, 2006). The

few studies that have examined both structure and

composition have focused on the depth of partner

technological resources and found that they improve

a firm’s innovation performance (Baum, Calabrese, &

Silverman, 2000; Stuart, 2000). Although dyad-level

research has examined the influence of technological

differences between partners on firm innovation

(Sampson, 2007), research has largely overlooked the

influence of network-level technological diversity—

the technological differences between a firm and its

partners and among the partners. Such compositional

diversity is relevant to a current debate in the social

network and alliance literatures.

Second, research has yielded conflicting results

about the influence of alliance network structure

on firm innovation. Research that examines the

influence of social networks on creativity and in-

novation has stressed the benefits actors derive

from network structure and explored how these

benefits, or “structural social capital” (Nahapiet &

This article is based on my doctoral dissertation, which

received the Academy of Management’s TIM Division’s

Best Dissertation Award and the State Farm Companies

Foundation Dissertation Award. I thank my dissertation

committee members, Raghu Garud, Theresa Lant, and J.

Myles Shaver, for their advice and guidance. I am grateful

for comments on drafts of this article by Sanjay Jain, Mel-

issa Schilling, J. Myles Shaver, Kevin Steensma, Kate

Stovel, Anu Wadhwa, and Mina Yoo. I thank Simon Rodan

for his help with computing the diversity measure used

here. I also thank Associate Editor Chet Miller for his me-

ticulous feedback and thank the three anonymous review-

ers for their comments, all of which greatly contributed to

the improvement of the article. I acknowledge financial

support from the State Farm Companies Foundation and

the Berkley Center for Entrepreneurship at the Stern School

of Business, NYU. Finally, I am grateful for the extraordi-

nary database development and programming skills of

Ralph Heidl and Tim Nali. All errors are my responsibility.

娀Academy of Management Journal

2010, Vol. 53, No. 4, 890–913.

890

written permission. Users may print, download or email articles for individual use only.

Ghoshal, 1998), influence knowledge creation. In

particular, the configuration of an actor’s set of

direct ties (i.e., the actor’s “egocentric network

structure”) has received some attention. This re-

search has focused on triadic closure (i.e., whether

an actor’s partners are partners), but two competing

perspectives exist, each with different causal mech-

anisms linking network structure to innovation.

The argument of one view is that disconnected

networks increase creativity and innovation be-

cause they provide actors with timely access to

diverse information (Burt, 1992, 2004).

An alter-

native view suggests dense networks, in which tri-

ads are closed and “structural holes” (unconnected

partners) are absent, provide social capital because

such structures generate trust, reciprocity norms,

and a shared identity, which increase cooperation

and knowledge sharing (Coleman, 1988; Portes,

1998). Research has found support for both views,

yielding conflicting results. Although studies have

found structural holes in a firm’s network enhance

its knowledge creation (Hargadon & Sutton, 1997;

McEvily & Zaheer, 1999), other research has sug-

gested that network closure improves knowledge

transfer and innovation (Ahuja, 2000; Dyer & No-

beoka, 2000; Schilling & Phelps, 2007).

One plausible reason for these conflicting results

is that most studies have examined the influence of

network structure and largely overlooked network

composition. An examination of network composi-

tion may help resolve these conflicting results and

lead to a better understanding of how alliance net-

works influence firm innovation. Another possible

explanation is that different studies have examined

different types of ties, different institutional con-

texts, and different outcome variables. It is unlikely

a particular network structure is universally bene-

ficial (see Adler & Kwon, 2002). Research has sug-

gested that the value of open versus closed net-

works for innovation and creativity is contingent

on the type of task (Hansen, 1999), type of tie

(Ahuja, 2000), and particular institutional environ-

ment (Owen-Smith & Powell, 2004). In contrast,

network structure may act as a contingency vari-

able and moderate the influence of network com-

position on firm innovation. Moreover, this effect

may depend on the type of learning and innovation

actors pursue. Both of these contingencies have

been largely unexplored in prior research and are

examined in this study.

A third limitation of research on alliances and

firm innovation concerns an often-used, yet largely

unexamined, assumption about the benefits of

structural holes. Although a principal benefit at-

tributed to structural holes is timely access to di-

verse information (Burt, 1992), structural holes are

neither a necessary nor a sufficient condition for

such access (Reagans, Zuckerman, & McEvily,

2004). The informational benefits contacts provide

can be directly observed by examining the extent to

which they specialize in different domains of

knowledge (Reagans & McEvily, 2003; Rodan &

Galunic, 2004). Observing differences in competen-

cies also allows a finer-grained measure of diversity

than simply counting structural holes. Because

competencies are stable and durable properties of

firms (Patel & Pavitt, 1997), they are a composi-

tional variable. Ties to partners with dissimilar

knowledge stocks provide a firm with access to

diverse information and know-how, independent

of the structure of its local network. Thus, the so-

cial control benefits of network closure and access

to diverse information and know-how can coexist.

Research on interfirm alliances (Ahuja, 2000) and

interpersonal networks (Rodan & Galunic, 2004)

has shown network density and knowledge diver-

sity are empirically distinct. However, alliance re-

search has not examined the independent and in-

teractive effects of network structure and network

knowledge diversity on firm innovation.

A final limitation of research on alliance net-

works and firm innovation is that it largely ignores

the novelty of the knowledge created and embodied

in the innovations measured. Instead, studies have

focused on the amount of innovation indicated by

survey items or counts of new products and pat-

ents. This approach implicitly rests on the as-

sumption that innovations are similar in their

knowledge content. Although research has sug-

gested firms typically search for innovative solu-

tions to problems in the domains of their existing

expertise (“local search”) and produce “exploi-

tive” innovations that represent incremental im-

provements to their prior innovative efforts (e.g.,

Dosi, 1988; Martin & Mitchell, 1998), some re-

search has shown firms vary in the scope of their

search and the exploratory content of their inno-

vations (Ahuja & Lampert, 2001; Rosenkopf &

Nerkar, 2001). A few studies have examined how

organizational design decisions influence ex-

ploratory knowledge creation (Jansen, Van Den

Bosch, & Volberda, 2006; Sigelkow & Rivkin,

2005). However, with the exception of some qual-

itative case study research (Dittrich, Duysters, &

Burt (1992) also argued structural holes allow actors

freedom from the normative expectations of others in a

network, yet research into the influence of network struc-

ture on innovation and creativity has stressed informational

benefits, rather than control benefits, as the primary causal

motor (e.g., Ahuja, 2000; Burt, 2004; Obstfeld, 2005).

2010 891Phelps

de Man, 2007; Gilsing & Noteboom, 2006), re-

search has generally ignored the effects of alli-

ance network structure and network diversity on

exploratory knowledge creation.

The purpose of this study is to address these

limitations. I do so by examining the influence of

the structure and composition of a firm’s network

of horizontal technology alliances on its explor-

atory innovation. I focus on horizontal technology

alliances for theoretical clarity. Exploratory inno-

vation is the creation of technological knowledge

that is novel relative to a firm’s extant knowledge

stock. Research has often portrayed exploration as a

process (March, 1991), yet the manifestation of this

process can be observed by examining the explor-

atory content of a firm’s innovations (Benner &

Tushman, 2002; Rosenkopf & Nerkar, 2001). Ex-

ploratory innovations embody knowledge that dif-

fers from knowledge used by the firm in prior in-

novations and shows the firm has broadened its

technical competence (Greve, 2007; Rosenkopf &

Nerkar, 2001).

Understanding the origins of exploratory innova-

tion is an important endeavor. Because the results

of exploration (versus exploitation) typically take

longer to realize, are more variable, and produce

lower average returns, organizations generally pur-

sue exploitative innovation at the expense of ex-

plorative innovation (March, 1991). They face a

fundamental challenge: although exploitation im-

proves an organization’s short-term performance,

exploration increases its long-term adaptability and

survival (Levinthal & March, 1993). This formula-

tion does not suggest that exploratory innovation is

preferred over incremental innovation, only that a

balance is necessary (March, 1991). The strong in-

centives to pursue exploitation at the expense of

exploration raise the question of how and when

firms are able to explore effectively. Research has

documented the propensity of firms to pursue local

search and exploitative innovation (e.g., Dosi,

1988; Helfat, 1994), but much less is known about

how and when firms overcome this predisposition

and develop exploratory innovations. Explaining

the production of exploratory innovations should

provide a better understanding of how organiza-

tions are able to thrive and survive.

I derive two predictions about the effect of hori-

zontal technology alliances on firm exploratory in-

novation. First, in highlighting the role of network

composition, I draw on the recombinatory search

literature (e.g., Fleming, 2001) and examine the

benefits and costs of increasing network technolog-

ical diversity for exploratory innovation. I predict

network diversity has an inverted U-shaped effect

on firm exploratory innovation performance. Sec-

ond, building on interfirm learning and network

research, I argue that the extent to which a firm’s

partners are densely interconnected generates trust

and reciprocity, which enhance the benefits of net-

work diversity and mitigate some of its costs. I

predict the density of a firm’s alliance network

positively moderates the effect of network diversity

on the firm’s exploratory innovation performance.

In the empirical work reported here, I tested

these predictions on a panel of 77 leading commu-

nications equipment manufacturers during 1987–

97 and found partial, yet robust, support for both

hypotheses. A positive linear effect of network di-

versity emerged, rather than a curvilinear effect,

and a positive linear interaction between diversity

and density, rather than a curvilinear interaction.

This study contributes to the alliance and innova-

tion literatures by addressing significant gaps in

research on the influence of alliance networks on

firm innovation. This is the first study of which I

am aware that investigates the influence of alliance

network structure and composition on firm explor-

atory innovation. The results show the technologi-

cal diversity in a firm’s alliance network and the

density of the network increase exploratory inno-

vation, independently and in combination. The re-

sults also suggest the presence of structural holes in

a firm’s network is not a necessary condition for

providing the firm with access to diverse informa-

tion. The extent to which an actor’s network is

composed of alters with diverse knowledge bases

provides it access to diverse information, indepen-

dent of network structure. The benefits of network

closure and access to diverse information and

know-how can coexist in a firm’s alliance network,

and combining the two increases the firm’s explor-

atory innovation. Because I find network diversity

begets diverse innovations (Kauffman, 1995), the

results suggest that dense networks populated by

diverse actors generate more, rather than less, di-

verse knowledge.

THEORY AND HYPOTHESES

To understand when alliances influence a firm’s

exploratory innovation, I build on two complemen-

tary theoretical bases: recombinatory search and

As He and Wong stated, “Exploration versus exploi-

tation should be used with reference to a firm itself and

its existing capabilities, resources and processes, not to a

competitor or at the industry level” (2004: 485). Thus,

“one can only view acts of exploration or exploitation

relative to a particular actor’s vantage point” (Adner &

Levinthal, 2008: 49).

892 AugustAcademy of Management Journal

social capital. The recombinatory search literature

casts innovation as a problem-solving process in

which solutions to valuable problems are discov-

ered via search (Dosi, 1988). Search processes lead-

ing to the creation of new knowledge typically in-

volve the novel recombination of existing elements

of knowledge, problems, or solutions (Fleming,

2001; Nelson & Winter, 1982) or reconfiguring the

ways knowledge elements are linked (Henderson &

Clark, 1990). Search is uncertain, costly, and guided

by prior experience (Dosi, 1988). Over time, feedback

from past search efforts becomes embodied in organ-

izational routines, which efficiently guide current

search efforts (Nelson & Winter, 1982).

Firms create knowledge by engaging in local and

distant search (March, 1991). Local search, which

is synonymous with exploitation, produces recom-

binations of familiar and well-known knowledge

elements and is often the preferred mode of search

(March, 1991; Stuart & Podolny, 1996). In contrast,

distant search, or exploration, involves recombina-

tions of novel, unfamiliar knowledge and involves

higher costs and uncertainty (March, 1991). Although

distant search can be less efficient and less certain

than local search, it increases the variance of search

and the potential for highly novel recombinations

(Fleming, 2001; Levinthal & March, 1981).

Innovation search research has primarily focused

on where firms search for solutions (i.e., local ver-

sus distant); the interfirm learning literature, on the

other hand, has emphasized how firms search. Ac-

cording to this research, interfirm relationships are

a mechanism for search and a medium of knowl-

edge transfer (Ingram, 2002). Because knowledge

is widely and heterogeneously distributed (von

Hayek, 1945), the exchange of knowledge is neces-

sary for recombination (Nahapiet & Ghoshal, 1998).

Yet the nature of knowledge involved in innovation

poses challenges to exchange. Technical innova-

tion involves tacit and socially embedded knowl-

edge (Dosi, 1988). Technology is knowledge em-

bedded in communities of practitioners (Layton,

1974) who develop tacit understandings of how to

solve problems related to its use and reproduction

(von Hippel, 1988). Such knowledge is also stored

in organizational routines (Nelson & Winter, 1982).

The specialized, tacit, and embedded nature of

technical knowledge makes market trading for it

subject to severe exchange problems (Teece, 1992).

Firms that can identify potentially useful elements

of technological knowledge, conceive of how these

elements can be fruitfully combined, and effectively

access and assimilate this knowledge increase their

potential for knowledge creation (Galunic & Rodan,

1998). Strategic alliances are important in each of

these aspects of successful recombination.

Strategic alliances are a means of accessing

knowledge a firm does not have and can be an

effective medium of knowledge transfer and inte-

gration (Hamel, 1991). Alliances provide a firm

with direct and repeatable access to its partners’

organizational routines, which reduces its ambigu-

ity about a partner’s knowledge and increases the

efficacy of its transfer and assimilation (Jensen &

Szulanski, 2007). Because of the increased social

interaction and enhanced incentive alignment and

monitoring features they provide, alliances are in-

stitutions better suited than market transactions for

the repeated exchange of tacit, routine-embedded

knowledge (Teece, 1992).

Although alliances provide access to external

knowledge, they do not guarantee its effective de-

tection, transfer, and assimilation. These processes,

and thus the odds of successful recombination, are

influenced by the incentives partners have to coop-

erate and share knowledge with each other (Hamel,

1991). Because the risk of opportunism is pro-

nounced in horizontal technology alliances, effec-

tive cooperation and knowledge sharing are diffi-

cult to achieve (Gulati & Singh, 1998). Alliance

research has typically emphasized the role of for-

mal governance mechanisms—such as detailed

contracts, the use of equity as a “hostage,” and joint

venture structures—in curbing opportunism and

increasing cooperation (e.g., Kogut, 1988; Mowery,

Oxley, & Silverman, 1996; Sampson, 2007). Other

research has suggested that mutual trust and reci-

procity norms between partners provide effective

and efficient informal governance (Dyer & Singh,

1998; Kale, Singh, & Perlmutter, 2000).

Trust and reciprocity serve as social control

mechanisms that mitigate opportunism and safe-

guard exchange in alliances (Dyer & Singh, 1998).

As such, they are forms of social capital, because

they represent resources that are instrumentally

valuable for, and appropriable by, partners in a

social exchange relationship (Coleman, 1988). The

extent to which social capital exists in a firm’s

network of alliances can increase the firm’s access

to its partners’ knowledge, the motivation of its

partners to transfer knowledge, and the efficiency

of knowledge exchange and transfer (Inkpen &

Tsang, 2005), resulting in more successful recom-

binations (Galunic & Rodan, 1998). In the next two

sections, I build on the recombinatory search lit-

erature and research on interfirm networks and

social capital to develop predictions about how

and when network technological diversity and

network density influence a firm’s exploratory

innovation performance.

2010 893Phelps

Network Technological Diversity

Diversity refers to the extent to which a system

consists of uniquely different elements, the fre-

quency distribution of these elements, and the de-

gree of difference among the elements (Stirling,

2007). Thus, I define alliance network technologi-

cal diversity as the extent to which the technologies

pursued by a firm’s alliance partners are different

from one another and from those of the focal firm.

Although network diversity provides benefits for a

firm’s exploratory innovation efforts, it also poses

significant costs. Diversity affects the relative nov-

elty of knowledge available in a network and the

ease with which a firm can recognize, assimilate,

and utilize this knowledge.

Increasing network diversity increases the rela-

tive novelty of the knowledge a firm can access.

Because exploratory innovations embody relatively

novel knowledge, a necessary condition for firm

exploratory innovation is access to dissimilar

knowledge (Greve, 2007; Jansen et al., 2006). Diver-

sity increases the number and variety of possible

combinations and the potential for highly novel

solutions (Fleming, 2001). The “value of variance”

(Mezias & Glynn, 1993) in distant search is that

though it increases failures, it also increases the

number of highly novel solutions (Levinthal &

March, 1981). In contrast, individuals and organi-

zations that exploit established competences in

their innovative problem-solving efforts typically

experience more certain and immediate returns,

but produce mostly incrementally innovative solu-

tions (Audia & Goncalo, 2007; Dosi, 1988). Search-

ing diverse knowledge domains challenges existing

cognitive structures, including premises and be-

liefs about cause-effect relationships (Duncker,

1945), which can promote new associations and

lead to highly novel insights and solutions (Simon-

ton, 1999). By searching diverse and novel do-

mains, firms can develop multiple conceptualiza-

tions of problems and solutions and apply

solutions from one domain to problems in another

(Hargadon & Sutton, 1997). Diverse knowledge

sources also provide firms with access to diverse

problem-solving heuristics (Page, 2007), which can

increase the exploratory content of new combina-

tions of knowledge (Audia & Goncalo, 2007). Fi-

nally, searching diverse, nonredundant knowledge

can stimulate intensive experimentation with new

combinations, leading to highly novel innovations

(Ahuja & Lampert, 2001).

Network diversity also influences a firm’s relative

absorptive capacity. As the technological distance be-

tween partners increases, their ability to recognize,

assimilate, and apply each other’s knowledge de-

clines (Lane & Lubatkin, 1998), increasing the costs

of recombinatory innovation (Weitzman, 1998). A

firm must expend greater effort and resources to

understand and integrate dissimilar knowledge

(Cohen & Levinthal, 1990). This can manifest in

costly, excessive, and inconclusive experimenta-

tion (Ahuja & Lampert, 2001). A firm’s cognitive

capacity constraints and its relative inexperience

with dissimilar knowledge components will limit

its ability to comprehend increasingly complex in-

teractions among these components (Fleming &

Sorenson, 2001). Moreover, integrating novel

knowledge from dissimilar sources often requires

changing existing patterns of communication and

social exchange, which is difficult in established

organizations (Kogut & Zander, 1992). Attempting

to assimilate and integrate highly diverse knowl-

edge components can lead to information overload,

confusion, and diseconomies of scale in innovation

efforts (Ahuja & Lampert, 2001). Thus, as a firm’s

network diversity increases, its costs of absorbing

and utilizing this knowledge greatly increase.

Given these benefits and costs of network diver-

sity, I expect it to exhibit a curvilinear effect on a

firm’s exploratory innovation. At low levels of di-

versity, a firm has a high degree of relative absorp-

tive capacity in its portfolio of partners, but the

knowledge to which it has access provides little

novelty. At high levels of network diversity, ab-

sorptive capacity costs are likely to outweigh the

benefits of highly novel knowledge. Although in-

creasing diversity exponentially increases opportu-

nities for novel recombinations (Fleming, 2001), an

organization is greatly constrained in its ability to

process an abundance of potentially novel recom-

binations into usable innovations (Weitzman,

1998). Research has shown that as knowledge com-

ponents become more diverse, the chance of their

recombination into useful innovations declines,

with excessive diversity reducing innovation

(Fleming & Sorenson, 2001). In contrast, at a mod-

erate level of network diversity a firm’s exploratory

innovation efforts benefit from a balance of access

to a moderate degree of novel knowledge and mod-

erately efficient relative absorptive capacity. Thus,

some degree of diversity is valuable for exploratory

innovation; too much can be detrimental.

Hypothesis 1. The technological diversity in

a firm’s alliance network has an inverted U-

shaped relationship with the firm’s subsequent

degree of exploratory innovation.

Network Density

Although an alliance provides access to a part-

ner’s knowledge, it does not guarantee the effective

894 AugustAcademy of Management Journal

detection, transfer, and assimilation of this knowl-

edge (Hamel, 1991). The tacit and embedded nature

of technological knowledge makes it difficult for

partners to detect, transfer, and assimilate (Teece,

1992), reducing its potential for successful recom-

bination (Galunic & Rodan, 1998). Increasing net-

work diversity worsens this problem, since a firm’s

absorptive capacity in relation to its partners will

decline (Lane & Lubatkin, 1998). Greater diversity

reduces the odds partners share a common under-

standing of technical issues, a language for discuss-

ing them, and an approach to codifying knowledge

(Cohen & Levinthal, 1990). The exchange hazards

in horizontal technology alliances compound these

problems. Because partners have incentives to

compete, the risk of opportunism is elevated. Such

alliances are also inherently uncertain and pose

large measurement and monitoring problems

(Pisano, 1989). Partners are at risk of involuntary

knowledge leakage, the withholding of effort and

resources needed to achieve alliance goals, misrep-

resentation of newly discovered knowledge, and

challenges in transferring tacit knowledge devel-

oped during the relationships (Gulati & Singh,

1998). Network diversity also compounds these

problems. Increasing diversity increases the rela-

tive novelty of knowledge and the variety of tacit

knowledge, thereby increasing the amount of

unique tacit knowledge. High novelty and tacitness

increase partner uncertainty and contractual haz-

ards (Pisano, 1989). Technological diversity in-

creases coordination problems and the potential for

costly contractual renegotiations (Sampson, 2004).

These exchange hazards can reduce cooperation

and knowledge sharing, hindering a firm’s recom-

bination efforts.

The extent to which a firm’s alliance partners are

densely interconnected mitigates some of the costs

and amplifies some of the benefits of increasing

network diversity, thus positively moderating its

effect on exploratory innovation. Dense networks

facilitate the production of trust and reciprocity

among networked firms, which decrease exchange

hazards in alliances, increase cooperation among

partners, and mitigate absorptive capacity prob-

lems. These problems become more challenging,

and thus more important to resolve, as network

diversity grows.

Network density promotes trust and reciprocity

between partners because they share common

third-party partners. Dense networks allow firms to

learn about current and prospective partners

through common third parties, reducing informa-

tion asymmetries among firms and increasing their

“knowledge-based trust” in one another (Gulati,

Nohria, & Zaheer, 2000). Network closure also pro-

motes trust by increasing the costs of opportunism

(Coleman, 1988). Because a firm’s behavior is more

visible in a dense network, an act of opportunism

can damage its reputation, jeopardizing its existing

alliances and reducing future alliance opportuni-

ties (Gulati, 1998). Because the costs of opportun-

ism can outweigh the benefits, firms will refrain

from such behavior. Thus, dense networks also

generate “enforceable” or “deterrence-based” trust

(Kreps, 1990; Raubb & Weesie, 1990). Research has

provided empirical support for these arguments

(Gulati & Sytch, 2008; Holm, Eriksson, & Johanson,

1999; Husted, 1994; Robinson & Stuart, 2007;

Rooks, Raub, Selten, & Tazelaar, 2000; Uzzi, 1996).

Network density also generates reciprocity exchanges

in which partners share privileged resources because

they expect recipients will repay them with some-

thing of equivalent value (Coleman, 1988). A firm

can encourage reciprocity between two of its part-

ners by transferring reciprocal obligations one part-

ner owes to the firm to the other partner (Uzzi,

1997). Dense networks also promote reciprocity by

protecting relationships from opportunism, in-

creasing actors’ confidence that obligations for re-

payment will eventually be met (Coleman, 1988).

The trust and reciprocity benefits of dense net-

works can mitigate some of the exchange hazards

and challenges to effective interfirm cooperation

associated with greater network diversity. Trust

and reciprocity generated by network density act as

informal safeguards of dyadic exchange, supple-

menting formal alliance governance mechanisms

(Powell, 1990). Given the challenges of formal gov-

ernance in horizontal technology alliances among

technologically diverse firms, informal governance

becomes more important in mitigating opportun-

ism and promoting cooperation as diversity in-

creases. Informal governance reduces the threat of

opportunism and increases each partner’s motiva-

tion to cooperate and share resources (Dyer &

Singh, 1998). Trust reduces the extent to which

alliance partners protect knowledge, increases their

willingness to share knowledge, and increases in-

terfirm learning and knowledge creation (Kale et

al., 2000; Larson, 1992). Reciprocity norms rein-

force this motivation to share, since firms can be

confident partners will reciprocate (Dyer & No-

beoka, 2000). As a result, the information and know-

how shared will be less distorted, richer, and of

higher quality (Dyer & Nobeoka, 2000; Uzzi, 1997).

Research has suggested dense interfirm networks are

better for transferring and integrating complex and

tacit knowledge than networks with structural holes

(Dyer & Nobeoka, 2000; Kogut, 2000).

Alliance network density also reduces absorptive

capacity problems related to growing network di-

2010 895Phelps

versity. Network closure promotes intense social

interaction, experimentation, joint problem solv-

ing, and triangulation, which enhance a firm’s abil-

ity to absorb and apply increasingly diverse partner

knowledge. The trust and reciprocity benefits of

network closure promote intense interaction

among personnel from partnered firms (Larson,

1992), which improves the detection and transfer of

tacit and embedded knowledge (Zander & Kogut,

1995). Intense interaction can also lead to the cre-

ation of partner-specific knowledge-sharing rou-

tines that facilitate knowledge transfer (Lane & Lu-

batkin, 1998). The social capital produced in dense

alliance networks encourages such relation-spe-

cific investments (Walker, Kogut, & Shan, 1997).

The trust and reciprocity benefits of network clo-

sure also increase partners’ joint problem-solving

efforts and stimulate experimentation with differ-

ent knowledge combinations, improving knowl-

edge detection and transfer from diverse partners

(Dyer & Nobeoka, 2000; Uzzi, 1997). Trust and rec-

iprocity can also increase a partner’s motivation to

“teach” (Szulanski, 1996), which is more important

for student firms as partner diversity increases,

since they should find it easier to learn unaided

from similar partners (Szulanski, 1996). Alliance

partners also provide alternative interpretations of

technical problems and solutions, allowing a firm

to compare, contrast, and triangulate these perspec-

tives (Nonaka, 1994). Alternative perspectives dif-

fuse rapidly in dense networks (Smith-Doerr &

Powell, 2005) and are more valuable when partners

are diverse, since a variety of perspectives in-

creases the chances some will be useful in a firm’s

recombination efforts (Nonaka, 1994). Finally, the

rapid flow of information in dense networks pro-

vides firms with more opportunities to share and

expand their understanding of technical issues and

can help establish a shared mode of discourse

(Smith-Doerr & Powell, 2005), allowing diverse

partners to more efficiently communicate with and

learn from one another (Kogut & Zander, 1996).

In sum, increasing network density improves a

firm’s ability to absorb and utilize knowledge from

more diverse partners. I expect these benefits of

network density to moderate the curvilinear effect

of network diversity on firm exploratory innova-

tion in four distinct ways. First, increasing density

will increase the slope (i.e., strength) of the positive

relationship between diversity and exploratory in-

novation (i.e., the positive slope to the left of the

peak of the curve). Second, increasing density will

increase the amplitude of the curvilinear effect of

diversity. That is, as density increases, the maxi-

mum value of exploratory innovation achieved will

increase. Third, the value of diversity that maxi-

mizes exploratory innovation will increase as den-

sity increases, shifting the peak of the curve to

higher values of diversity. Finally, increasing den-

sity will reduce the slope of the negative relation-

ship between diversity and exploratory innovation.

That is, after the effect of diversity turns negative,

increasing density will dampen the negative effect

of diversity on exploratory innovation.

Hypothesis 2. The density of a firm’s alliance

network moderates the curvilinear relationship

between network diversity and exploratory in-

novation in such a fashion that increasing den-

sity will: (a) increase the slope of the positive

effect of diversity, (b) increase the amplitude of

the effect of diversity, (c) increase the value

of diversity that maximizes exploratory inno-

vation, and (d) reduce the negative effect of

diversity.

METHODOLOGY

Sample and Data

The research setting for this study was the global

telecommunications equipment industry (SIC 366).

Firms in this industry produce and market hard-

ware and software that enable the transmission,

switching, and reception of voice, images, and data

over both short and long distances using digital,

analog, wire line, and wireless technology. I chose

this setting for two reasons. First, during the 1980s

and 1990s this industry experienced significant

changes in technology and competition, resulting

in a growing use of technology alliances by incum-

bents (Amesse, Latour, Rebolledo, & Se´guin-Du-

lude, 2004). Second, since I used patent data, I

chose to study an industry in which firms routinely

and systematically patent their inventions (Hage-

doorn & Cloodt, 2003; Levin, Klevorick, Nelson, &

Winter, 1987).

To minimize survivor bias and right censoring, I

limited the study period to 1987–97. I limited the

sample frame to public companies to ensure the

availability and reliability of financial data. I lim-

ited the sample to the firms in the industry with the

largest sales because complete and accurate alli-

ance data are more available for industry leaders

than for smaller firms (Gulati, 1995). To minimize

survivor bias, I identified the top-selling firms in

the industry at the beginning of the study period

rather than the end because numerous mergers, re-

structurings, and failures occurred during the study

period (Amesse et al., 2004). To minimize the influ-

ence of right censoring, I ended the study period in

1997 to allow sufficient time for the (non)approval of

patent applications that sample firms made during

896 AugustAcademy of Management Journal

the period (also see footnote 4). Following prescrip-

tions for establishing network boundaries in empiri-

cal research (Laumann, Marsden, & Prensky, 1983), I

restricted the network to both firms and alliances that

focused on the telecommunications equipment in-

dustry. Recent alliance network research has used

similar network construction criteria (Rowley, Beh-

rens, & Krackhardt, 2000; Schilling & Phelps, 2007).

These sampling criteria resulted in a sample of 77

firms headquartered in 13 countries.

I used patent data to measure technological

knowledge because patents are valid and robust

indicators of knowledge creation (Trajtenberg,

1987). Knowledge is instantiated in inventions, and

patents are measures of novel inventions externally

validated through the patent examination process

(Griliches, 1990). A patent application represents a

positive expectation by an inventor of the eco-

nomic significance of his or her invention, since

getting such protection is costly (Griliches, 1990).

Patents measure a codifiable portion of a firm’s

technical knowledge, yet they correlate with mea-

sures that incorporate tacit knowledge (Brouwer &

Kleinknecht, 1999). For these various reasons, pat-

ents are a reliable and valid measure of innovation

in the telecom equipment industry (Hagedoorn &

Cloodt, 2003).

Information on U.S. patents was obtained from

Delphion. Using patents from a single country

maintains consistency, reliability, and comparabil-

ity across firms (Griliches, 1990). U.S. patents are a

good data source because of the rigor and proce-

dural fairness used in granting them, the large in-

centives firms have to obtain patent protection in

the world’s largest market for high-tech products,

the high quality of services provided by the U.S.

Patent and Trademark Office (USPTO), and the rep-

utation of the United States for providing effective

intellectual property protection (Pavitt, 1988; Riv-

ette, 1993). I used the application date to assign a

granted patent to a firm because this date closely

captures the timing of knowledge creation (Grili-

ches, 1990). Because patents are often assigned to

subsidiaries, I carefully aggregated patents to the

firm level.

The collaboration data were obtained from mul-

tiple sources. I initially collected alliance data from

the SDC Alliance Database. Although this database

provided substantial content, it had many limita-

tions. I overcame these limitations through system-

atic archival research using annual reports, 10K

and 20F filings, Moody’s Manuals, Factiva, Lexis-

Nexis, and Dialog. These last three databases index

the historical full texts of hundreds of business

publications from all regions of the world and in-

clude articles translated to English from their orig-

inal languages, and non-English publications. I

conducted broad keyword searches to identify all

instances of interfirm cooperation involving the

sample firms. Individuals fluent in the respective

language read non-English articles and reports,

identified instances of interfirm cooperation, and

translated the documents into English. I recorded

only collaborations that could be confirmed in mul-

tiple sources. Around 1,200 annual reports and Se-

curities and Exchange Commission (SEC) filings

and over 180,000 electronic articles were exam-

ined, and over 8,500 relevant news stories were

printed out. Overall, the data set from which this

study draws includes 7,904 alliances and 1,967

acquisitions initiated during 1980-96. I reviewed

every record from the SDC data and corrected du-

plicate entries and other errors and omissions using

secondary sources.

Firm attribute data were collected from Compus-

tat, annual reports, SEC filings, the Japan Company

Handbook, Worldscope, and Global Vantage.

Measurement: Dependent Variable

Exploratory innovation. Exploratory innovation

is the creation of technological knowledge by a firm

that is novel relative to its existing knowledge stock

(Benner & Tushman, 2002; Rosenkopf & Nerkar,

2001). Following prior research (Benner & Tush-

man, 2002, Katila & Ahuja, 2002; Rosenkopf &

Nerkar, 2001), I measured exploratory innovation

using patent citations. I began with the list of U.S.

patent classes that corresponded to the telecommu-

nications equipment industry at the beginning of

the sample period (see Table 1). I assessed the

exploratory innovation of firm iin year tby classi-

fying and tabulating all citations in the firm’s tele-

communications equipment patents applied for in

year t(and eventually granted). I traced each cita-

tion to determine if the firm had used the same

citation or if the citation was to a patent developed

by the firm during the seven years before the focal

year. I used a seven-year window because organi-

zational memory in high-tech firms is imperfect,

causing the value of knowledge to depreciate rap-

idly over time (Argote, 1999) and creating signifi-

I identified all divisions, subsidiaries, and joint ven-

tures of each sample firm (using Who Owns Whom and

the Directory of Corporate Affiliations) as of 1980. I then

traced each firm’s history to account for name changes,

division names, divestments, acquisitions, and joint ven-

tures and obtained information on the timing of these

events. This procedure yielded a master list of entities

that I used to identify all patents belonging to sample

firms for the period of study.

2010 897Phelps

cant problems for intertemporal knowledge transfer

(Nerkar, 2003). Although prior research (e.g., Katila

& Ahuja, 2002) has used a 5-year window to assess

exploration, I chose a 7-year window because the

median age of cited patents in telecom technologies

is about 6.5 years (Hicks, Breitzman, Olivastro, &

Hamilton, 2001). Using this window, I classified

each citation as “new” or “used.” I computed the

variable as the result of dividing new citations by

total citations (exploratory innovations

new cita-

tions

/total citations

.) Because this formula mea-

sures a share of new citations, rather than their full

count, it captures a firm’s propensity to produce ex-

ploratory innovations, independent of firm scale.

The extent to which a firm draws on elements of

knowledge (e.g., patent citations) it has previously

used reflects its practice of local search and exploi-

tation of its extant knowledge stock. The extent to

which it uses citations with which it has no expe-

rience is indicative of distant search and explor-

atory innovation (Benner & Tushman, 2002). This

measure ranges from pure exploitation (no explo-

ration) at the low end to pure exploration (no ex-

ploitation) at the high end. It is consistent with

research that has conceptualized and measured ex-

ploitation and exploration, or local and distant

search, as the ends of a continuum

(Benner &

Two aspects of the patent data used to construct this

measure merit discussion. First, during the period of

study, the USPTO did not publish patent applications. A

patent application date was only observable when a

patent was granted. Because I observed patents using

their date of application and because there is a delay

between the date of application for a patent and its even-

tual granting, I may not have observed all patents applied

for in a particular year and eventually granted, because

the USPTO had not rendered a decision by the time I

collected my patent data. The influence of such a right-

censoring bias, caused by the delay between patent ap-

plication and issuance, is likely to be negligible in this

study. Around 99 percent of all applications are re-

viewed within five years of application (Hall et al., 2001),

which is the period between the end of the sample (1997)

and the last year of patent data collection (2002). Second,

patent examiners often add citations to patent applications

(Alcacer & Gittelman, 2006), which suggests applicant firms

are not necessarily aware of all cited patents. Third-party

citations often manifest as noise in the measurement of

patent-based variables (Jaffe, Trajtenberg, & Fogarty, 2002).

Noise in the measurement of a dependent variable in-

creases standard errors and reduces the likelihood of find-

ing statistically significant effects (Gujarati, 1995).

Though I focus on one domain of search (i.e., tech-

nological knowledge), firms search multiple domains,

such as customer and geographic space (Gupta, Smith, &

Shalley, 2006; Sidhu et al., 2007). Portraying exploitation

TABLE 1

Primary U.S. Patent Classes Used to Represent

Telecommunications Equipment

Class Number Title

178 Telegraphy

179 (discontinued) Telephony

329 Demodulators

332 Modulators

333 Wave transmission lines and networks

334 Tuners

340 Communications: electrical

341 Coded data generation or conversion

342 Communications: directive radio wave

systems & devices

343 Communications: radio wave antennas

348 Television

358 Facsimile and static presentation

processing

359 Optics: systems (including

communication) and elements

367 Communications, electrical: acoustic

wave systems and devices

370 Multiplex communications

375 Pulse or digital communications

379 Telephonic communications

381 Electrical audio signal processing

systems and devices

382 Image analysis

385 Optical waveguides

455 Telecommunications

725 Interactive video distribution systems

Because patents are classified by technological and func-

tional principles, they do not map easily to product-based in-

dustrial definitions such as SIC codes (Griliches, 1990). That is,

there is not a one-to-one mapping between primary patent

classes and industries. Multiple patent classes are used in a

single industry, and a single patent class can be used in multiple

industries. Consequently, to identify the areas of technology that

constitute telecommunications equipment, I needed to develop

a concordance between primary patent classes and the three-

digit SIC code 366, “communications equipment.” To do so, I

utilized both Silverman’s (1996) concordance method and con-

cordances provided by experts. I used the concordance for com-

munications equipment developed by scholars at Science and

Technology Policy Research (SPRU), a unit of Sussex University

in the United Kingdom, and the concordance developed by the

Community of Science Inc., an internet company that provides

collaborative tools and services for research scientists and engi-

neers. I identified the primary patent classes common to both of

these expert-based concordances as a baseline and then com-

pared this list of classes with a rank-ordered list delineating the

degree to which specific international patent classes (IPCs) were

associated with SIC 366 as the industry of manufacture as of

1988. To make this comparison, I used the USPTO’s USPC-IPC

concordance. The primary classes listed in the baseline concor-

dance were associated with the highest ranked IPC classes as-

sociated with U.S. SIC 366 (except for class 725, which did not

exist in the late 1980s). This indicated that the 22 primary

classes used in this study to represent communications equip-

ment technology in this table are most frequently associated

with SIC 366.

898 AugustAcademy of Management Journal

Tushman, 2002; Greve, 2007; Sidhu, Commandeur,

& Volberda, 2007).

As a robustness check, I applied an alternative

measure of exploratory innovation from prior re-

search (e.g., Ahuja & Lampert, 2001; McGrath &

Nerkar, 2004). I computed this measure as the num-

ber of new three-digit technology classes in which

firm ipatented in year t, classifying a technology

class as new if the firm had not patented in that

class in the past seven years. The USPTO assigns

patents to about 450 technology classes, with each

class demarcating an area of technology. The extent

to which a firm enters new technological domains

is indicative of exploration (Ahuja & Lampert,

2001; McGrath & Nerkar, 2004). This measure was

broader than the citation-based measure since it

took into account all technology classes in which a

firm might patent.

Measurement: Explanatory Variables

Following prior research (e.g., Ahuja, 2000; Stu-

art, 2000), I sampled alliances involving technology

development or exchange because my phenome-

non of interest and theory concerned the transfer

and creation of technological knowledge. I ex-

cluded unilateral licensing deals and alliances

formed for the sole purpose of marketing, distribu-

tion, or manufacturing.

Network technological diversity. To measure

network technological diversity, I employed Rodan

and Galunic’s (2004) measure of knowledge hetero-

geneity. This measure incorporates information about

the knowledge distance between a focal actor and

each of its partners and the distances among the part-

ners. I began at the dyad level and measured the

technological distance between pairs of firms using

Jaffe’s (1986) index. For each firm-year, I measured

the distribution of a firm’s patents across primary

patent classes. Following Sampson (2007), I used a

moving four-year window to establish a firm’s patent-

ing profile. This distribution located a firm in a mul-

tidimensional technology space, captured by a K-di-

mensional vector (f

[f

...f

], where f

represents

the fraction of firm i’s patents that are in patent

class k). This approach rests on an assumption that

the distribution of a firm’s patents across classes

reflects the distribution of its technical knowledge

(Jaffe, 1986). The technological distance, d, be-

tween firms iand jin year twas calculated as:

dijt ⫽1⫺

冋

冘

k1

fikfjk

冒

冉

冘

k1

fik2

冊

1/ 2

冉

冘

k1

fjk2

冊

1/ 2

册

This measure was bounded between 0 (complete

similarity) and 1 (maximum diversity) and sym-

metric for the two firms. I used these pairwise

distance values to construct annual distance matri-

ces, D

, which reflected the technological distances

between all possible pairs of sample firms.

Next, I computed the uniqueness of the knowl-

edge of each partner jin firm i’s alliance network in

year t. The uniqueness of firm jis a function of the

uniqueness of its partners, k, and firm j’s distance

from them. Following Rodan and Galunic (2004), I

defined the uniqueness of firm j,u

, as:

␭

uj⫽

冘

djk ⫻uk.

The uniqueness of each firm is found in the solu-

tion of the eigen equation(

␭

UDU), where Uis an

eigenvector of Dand

␭

is its associated eigenvalue.

The elements of Uare the uniqueness values for

each firm, and Dis the matrix of pairwise techno-

logical distances. I measured the technological di-

versity available to firm iin its (ego) network of

alliance partners in year tas:

Network technological diversityit ⫽1

冘

j1

dij

␭

uj,

where d

is partner j’s distance from iand

␭

is j’s

uniqueness score computed for i’s Npartners. The

1/Nterm compensates for the fact that lambda in-

creases linearly with network size. This measure

increases linearly with the distances among iand

its partners (Rodan & Galunic, 2004).

Network density. To measure ego network den-

sity, I constructed annual adjacency matrices for

the period 1987–96 that indicated the presence of a

technology alliance, in existence at the end of a

focal year, between all possible undirected pair-

wise combinations of sample firms. An alliance

with more than 2 firms entered the adjacency ma-

trix as separate dyadic combinations of all firms in

the alliance. Of all sample alliances, 89 percent

involved only 2 firms, and the average alliance had

2.38 firms. Because alliances often endure longer

than one year, constructing adjacency matrices us-

ing only alliances formed in a focal year would

have understated the true connectivity of the net-

work. Consequently, I collected alliance data for

each firm beginning in 1980 and researched each

and exploration as ends of a continuum in one domain of

search does not preclude the possibility that firms can

simultaneously achieve high levels of both exploitation

and exploration in multiple domains (Gupta et al., 2006).

A universal argument about the mutual exclusivity or

independence of exploitation and exploration may be

impossible (Gupta et al., 2006).

2010 899Phelps

alliance to identify its date of dissolution or con-

tinuance through the last sample year.

Ego network density was the percentage of all

possible ties among an ego’s alters that had been

formed (Scott, 1991). Ego networks in which a

firm’s alliance partners are themselves allied imply

higher values of density. To test the robustness of

the effect of density, I substituted Burt’s (1992)

measures of efficiency and then constraint into al-

ternative specifications. The Appendix presents

these specifications. Both efficiency and constraint

are measures of triadic closure (see Borgatti [1997]

for a comparison). Figure 1 presents an example of

a sample firm’s ego network, specifically, Motoro-

la’s network of technology alliances at the end of

1992, and lists the values for the density, effi-

ciency, and constraint of this network. Algebraic

explanations of each measure are also shown.

Control Variables

To minimize alternative explanations and isolate

the marginal effects of the explanatory variables, I

controlled for several firm- and alliance-level vari-

ables whose influence on exploratory innovation

might be confounded with the explanatory vari-

ables. Given the firm-level analysis used in this

study, I aggregated alliance-level observations to

the firm level. I used multiple-year moving win-

dows of differing lengths to compute five control

variables. These window lengths ranged from four

to seven years and differed by control variable. I

based the choice of window length for each control

variable on prior research. Using alternative win-

dow lengths (1 year) for these control variables

did not substantively change the results of the ex-

planatory variables presented in Table 2.

Network size. More alliance partners may pro-

vide a firm with access to greater technical diver-

sity. Moreover, measures of ego network density

are sensitive to network size, making network size

an important control variable (Friedkin, 1981). I

computed network size as the natural logarithm of

the number of telecom technology alliance partners

maintained by firm iin year t.

Alliance duration. Alliance longevity can lead to

greater interfirm trust (Gulati, 1995), stronger reci-

procity norms (Larson, 1992) and relation-specific

routines (Levinthal & Fichman, 1988), increasing

interfirm learning (Simontin, 1999). I measured al-

liance duration as the average number of years firm

ihad participated in its existing telecom technol-

ogy alliances at the end of year t(see footnote 6).

Repeated ties. Prior ties between firms can in-

crease interfirm trust (Gulati, 1995), the develop-

ment of relation-specific learning heuristics, and

interfirm learning (Lane & Lubatkin, 1998). Follow-

ing Gulati and Gargiulo (1999), I calculated re-

peated ties as the average number of alliances firm

ihad formed with its current group of alliance

partners in the five years prior to year t.

Joint venture. Research has suggested equity

joint ventures are superior governance mechanisms

for interfirm learning and knowledge transfer

(Kogut, 1988; Mowery et al., 1996). I computed the

variable as the proportion of firm i’s telecom tech-

nology alliances governed by equity joint ventures

in year t.

International alliance. International alliances

provide access to diverse knowledge (Rosenkopf &

Almeida, 2003), but they experience greater coor-

dination and communication problems and cul-

tural conflicts than domestic alliances, and this

experience diminishes interfirm learning (Lyles &

Salk, 1996). I measured this variable as the fraction

of firm i’s telecom technology alliances in year t

involving foreign firms.

Partners’ market overlap. Because partners

tend to protect their knowledge when they are

product-market competitors, overlaps in partners’

markets can impede interfirm knowledge transfer

(Dutta & Weiss, 1997). I computed market overlap

as the proportion of firm i’s portfolio of telecom

technology alliances in year thaving partners with

the same primary four-digit SIC code as firm i.

Firm sales. Firm size can have both negative and

positive effects on firm innovation (Teece, 1992). I

controlled for firm size using the natural log of sales

(in millions of U.S. dollars) for firm iin year t.

I researched each alliance using the sources de-

scribed previously. I also contacted company personnel

to identify dissolution dates, which proved very useful in

identifying the termination or ongoing status of joint

ventures (JVs). For nearly all JVs, I was able to identify

the months they were ended or their ongoing status at the

end of the sample period. For each remaining JV, I as-

sumed it existed until the end of the last year in which it

was documented or until the end of the year after the year

it was founded, whichever was later. For non-JV alli-

ances, I recorded termination on the basis of specified

tenure, if mentioned in the archival sources, or an-

nouncement of dissolution (either from archival sources

or company contact). In cases in which I could not es-

tablish precise dissolution, I followed Ahuja (2000) and

presumed an alliance to exist until the end of the last

year in which it was documented or until the end of the

year after the year it was founded, whichever was later. I

performed a t-test of the difference in mean duration

between alliances with formal dissolution announce-

ments and those with assumed dissolution dates and

found no significant difference.

900 AugustAcademy of Management Journal

FIGURE 1

Motorola’s 1992 Ego Network Structure of Technology Alliances

In the figure, Motorola is the focal actor, or ego. Below are the values of ego network density, efficiency and constraint for Motorola’s

1992 technology alliance network and an explanation of each measure. Burt (1992) provides a detailed explanation of the measures of

efficiency and constraint and Borgatti (1997) provides a comparison of the three measures.

The values for the density, efficiency, and constraint of this network and their algebraic computation are as follows:

Density 26.67%

Ego network densityi

冋冉

冘

qxjq

冊

冒

冉再

冉

N1

冊冎

冒

冊册

100, jq,

where x

represents the relative strength of the tie between alter jand alter q, and Nrepresents the number of alters to which ego iis connected.

Because I treated alliances as either present or absent (i.e., they do not vary in terms of strength), all values of x

were set to 1 if a relationship

existed and 0 otherwise. The term [N(N1)] was divided by 2 to reflect that alliances are undirected ties. Variable range, 0–100%.

Efficiency 0.75

Ego network efficiencyi

冋

冘

冉

1

冘

qpiqmiq

冊册

冒

N,jq,

where p

is the proportion of i’s ties invested in the relationship with q, m

is the marginal strength of the relationship between alter jand alter

q(as I used binary data, all values of m

were set to 1 if a tie existed and 0 otherwise), and Nrepresented the number of alliance partners to which

focal firm was connected. This measure could vary from 0 to 1, with higher values indicative of greater efficiency (i.e., structural holes).

Constraint 0.15

Ego network constrainti

冘

冋

pij 

冘

qpiqpqj

册

,qi,j,

where p

is the proportion of i’s ties invested in the relationship with j, p

is the proportion of i’s ties invested in the relationship with

q, and p

is the proportional strength of alter q’s relationship with alter j. This measure can vary from of 0 to 1, with higher values

indicative of greater constraint (i.e., fewer structural holes).

2010 901Phelps

TABLE 2

Descriptive Statistics and Correlations

Variables Mean s.d. Min. Max. 123456789101112131415161718

1. Exploratory innovation 0.72 0.19 0 1 4

2. Network technological

diversity

0.22 0.33 0 1.71 .19

3. Network density 33.88 32.99 0 100 .18 .27

4. Network size

1.33 0.99 0 3.56 .25 .79 .38

5. Alliance duration 3.10 1.87 0 14 .08 .14 .14 .20

6. Repeated ties 0.38 0.46 0 4 .19 .49 .17 .43 .13

7. Joint venture 0.18 0.19 0 1 .02 .39 .18 .38 .26 .20

8. International alliance 0.30 0.25 0 1 .06 .16 .14 .18 .02 .08 .28

9. Partners’ market

overlap

0.33 0.24 0 1 .02 .04 .08 .07 .02 .09 .12 .14

10. Sales

6.47 2.66 2.99 11.34 .17 .64 .20 .61 .39 .55 .40 .17 .05

11. Firm current ratio 2.32 1.58 0.003 23.46 .13 .21 .09 .22 .19 .26 .29 .17 .06 .37

12. Firm R&D intensity 0.13 0.76 0 20.25 .03 .04 .07 .01 .27 .03 .18 .08 .02 .20 .09

13. Firm patent stock 582.06 1,267.74 1 6,875 .14 .50 .21 .44 .21 .67 .29 .15 .11 .63 .25 .04

14. Firm age 45.39 36.03 2 150 .13 .58 .18 .52 .38 .37 .50 .25 .04 .74 .30 .09 .11

15. Firm alliance

experience

0.14 0.75 0.001 19.25 .02 .09 .14 .18 .08 .03 .15 .07 .08 .29 .05 .36 .52 .16

16. Firm technological

diversity

0.76 0.31 0 1 .11 .34 .19 .32 .32 .25 .40 .11 .09 .59 .38 .02 .07 .53 .18

17. Firm acquisitions 0.96 1.71 0 16 .06 .34 .08 .32 .03 .07 .24 .06 .34 .11 .04 .06 .38 .09 .20

18. U.S.-Canada 0.65 0.48 0 1 .1 .50 .14 .31 .30 .39 .52 .29 .11 .65 .29 .06 .09 .61 .13 .43 .19

19. Europe 0.19 0.39 0 1 .04 .51 .10 .39 .14 .06 .50 .21 .12 .18 .18 .03 .06 .55 .08 .30 .40 .66

n(firms) 77; n(observations) 707. All correlations greater than 兩.07兩are significant at p.05.

Logarithm.

Firm current ratio. The availability of slack

resources can increase exploratory search (Singh,

1986) and lead to greater innovative performance

(Nohria & Gulati, 1996). I controlled for the un-

absorbed slack resources of firm iin year tusing

its current ratio (current assets/current liabilities)

(Singh, 1986).

Firm R&D intensity. A firm’s R&D expenditures

are investments in knowledge creation (Griliches,

1990) and contribute to its ability to absorb extra-

mural knowledge (Cohen & Levinthal, 1990). I mea-

sured R&D intensity by dividing firm i’s R&D ex-

penses by its sales in year t.

Firm patent stock. The more patents a firm has,

the more patents and references it can cite; a large

patent stock could thus negatively affect the pri-

mary measure of exploratory innovation here. A

firm’s patent stock also reflects the depth of its

technological resources and absorptive capacity

(Silverman, 1999). I controlled for the number of

firm i’s patents obtained in the four years prior to

the end of year t.

Firm age. As firms age, they tend to exploit their

existing technological competencies rather than ex-

plore new and unfamiliar technologies (Sorensen &

Stuart, 2000). I operationalized firm age as the

number of years from the date of founding of firm i

to year t.

Firm alliance experience. Alliance experience

enhances the collaborative capability of a firm,

which facilitates interfirm knowledge transfer

(Sampson, 2005). I controlled for the number of all

types of alliances formed by firm iin the seven

years before year t, divided by its sales in year t.

Firm technological diversity. Technologically

diverse firms may be more innovative because of

diverse internal knowledge flows (Garcia-Vega,

2006), and they may be more able to absorb external

knowledge (Cohen & Levinthal, 1990). I measured

firm i’s diversity in year tusing a modified Herfin-

dahl index (Hall, 2002):

Technological diversityit ⫽

冋

1⫺

冘

j1

冉

Njit

Nit

冊

册

⫻Nit

Nit ⫺1,

where N

is the number of patents obtained by firm

iin the past four years. N

jit

is the number of patents

in technology class jin firm i’s four-year patent

stock. This variable could range from 0 to 1 (max-

imum diversity).

Firm acquisitions. Acquisitions can enhance ac-

quirer innovation (Ahuja & Katila, 2001). Telecom

equipment firms often use both acquisitions and

alliances to source knowledge (Amesse et al.,

2004). I controlled for the number of telecom equip-

ment acquisitions (i.e., those in which the target

company’s primary SIC code was 366) made by

firm iduring the four years prior to and including

year t.

U.S.-Canada/Europe/Asia. I used dummies de-

noting the regional origin of a firm to control for

regional effects. “U.S.-Canada” was coded 1 if a

firm was headquartered in the United States or

Canada. “Europe” was coded 1 if the firm was

headquartered in Europe. Asia was the omitted

category.

Model Specification and Estimation

The dependent variable was a proportion and

presented several challenges to linear regression

(Gujarati, 1995). Thus, I used three alternative mod-

eling approaches. First, I estimated the models with

exploratory innovation as the dependent variable

using panel linear regression and robust standard

errors. Following common econometric practice

(Greene, 1997), I also estimated models with a log-

odds transformation of exploratory innovation.

Fi-

nally, I estimated models using a generalized esti-

mating equation (GEE) approach in which I

specified a probit link function and an exchange-

able correlation matrix and computed robust errors

(Papke & Wooldridge, 2005). As a robustness check,

I compared the results from these alternative spec-

ifications. I included year dummies to control for

period effects, such as differences in macroeco-

nomic conditions or industry technological oppor-

tunity. Either firm-specific fixed or random effects

can be used to control for unobserved firm hetero-

geneity (Greene, 1997), such as differences in mo-

tivations to pursue, and abilities to develop, explor-

atory innovations. Because the use of random

effects relies on an assumption that errors and re-

gressors are uncorrelated, I used a Hausman (1978)

test to choose between fixed and random effects. I

also checked for first-order serial autocorrelation in

the errors. I lagged all independent variables one

year, which reduced concerns of reverse causality

and avoided simultaneity.

The transformed variable is as follows: ln(explor-

atory innovation/1–exploratory innovation). Because

the transformation is undefined when exploratory inno-

vation is equal to 0 or 1, I recoded these values as follows:

00.0001 and 1 0.9999.

2010 903Phelps

RESULTS

Table 2 reports descriptive statistics and correla-

tions. The panel was unbalanced and consisted of

77 firms and 707 firm-year observations. Table 3

presents the results of the panel regression analysis

used to test the hypotheses. I report the results for

untransformed exploratory innovation for ease of

interpretation. The results using a logit transforma-

tion and those from GEE estimation are consistent

with those reported in Table 3. I estimated models

1–7 using firm random effects for three reasons: (1)

significant unobserved heterogeneity was present,

(2) Hausman specification tests were not signifi-

cant, supporting the use of random effects, and (3)

significant serial correlation was not present. Hu-

ber-White (or “sandwich”) robust standard errors

are reported, and all significance levels are for two-

tailed tests. Multicollinearity does not seem to have

unduly influenced the regression results because

the average variance inflation factor (VIF) for each

model and the VIFs for all variables were below the

rule-of-thumb value of ten (Gujarati, 1995).

Hypothesis 1 predicts an inverted U-shaped ef-

fect of network technological diversity on firm ex-

ploratory innovation. Models 2–6 in Table 3 pro-

vide partial support for this hypothesis. In each of

these models, network technological diversity

it 1

exhibited a positive and significant effect on ex-

ploratory innovation. However, the squared term

was not significant in any model in which it was

entered. Thus, although I found evidence of a pos-

itive linear effect of network diversity, I did not

TABLE 3

Results of Random-Effects Panel Linear Regression Analysis Predicting Firm Exploratory Innovation

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

Constant 0.85** (0.06) 0.89** (0.07) 0.89** (0.07) 0.89** (0.07) 0.86** (0.07) 0.85** (0.07)

Network size

0.03** (0.01) 0.05** (0.01) 0.05** (0.01) 0.04** (0.01) 0.04** (0.01) 0.04** (0.01)

Alliance duration 0.01 (0.01) 0.01 (0.01) 0.01 (0.01) 0.01 (0.01) 0.01 (0.01) 0.01 (0.01)

Repeated ties 0.01 (0.02) 0.01 (0.02) 0.02 (0.02) 0.01 (0.02) 0.02 (0.02) 0.01 (0.02)

Joint venture 0.08 (0.06) 0.08 (0.06) 0.08 (0.06) 0.07 (0.06) 0.05 (0.06) 0.05 (0.06)

International alliance 0.09* (0.04) 0.08* (0.04) 0.08* (0.04) 0.08* (0.04) 0.07* (0.04) 0.08* (0.04)

Partners’ market overlap 0.04 (0.04) 0.03 (0.04) 0.03 (0.04) 0.02 (0.04) 0.02 (0.04) 0.01 (0.04)

Firm sales

0.02* (0.01) 0.03* (0.01) 0.03* (0.01) 0.02* (0.01) 0.00 (0.01) 0.00 (0.01)

Firm current ratio 0.00 (0.01) 0.00 (0.01) 0.00 (0.01) 0.00 (0.01) 0.00 (0.01) 0.00 (0.01)

Firm R&D intensity 0.08 (0.08) 0.08 (0.08) 0.08 (0.08) 0.08 (0.08) 0.07 (0.08) 0.06 (0.08)

Firm patent stock/1,000 0.004** (0.00) 0.002** (0.00) 0.002** (0.00) 0.002** (0.00) 0.003** (0.00) 0.003** (0.00)

Firm age 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00)

Firm alliance

experience

0.01 (0.03) 0.02 (0.03) 0.02 (0.03) 0.02 (0.03) 0.02 (0.03) 0.02 (0.03)

Firm technological

diversity

0.05 (0.04) 0.05 (0.04) 0.05 (0.04) 0.05 (0.04) 0.05 (0.04) 0.05 (0.04)

Firm acquisitions 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00)

U.S.-Canada 0.01 (0.02) 0.01 (0.02) 0.01 (0.02) 0.01 (0.02) 0.01 (0.02) 0.01 (0.02)

Europe 0.01 (0.03) 0.00 (0.03) 0.00 (0.03) 0.00 (0.03) 0.01 (0.03) 0.00 (0.03)

Network technological

diversity

0.06* (0.03) 0.06* (0.03) 0.06

†

(0.03) 0.07* (0.04) 0.08* (0.04)

Network technological

diversity squared

0.00 (0.04) 0.00 (0.04) 0.05 (0.04) 0.06 (0.09)

Network density 0.05* (0.02) 0.10* (0.04) 0.07 (0.05)

Network technological

diversity density

0.46** (0.13) 0.53** (0.14)

Network technological

diversity squared 

density

0.48 (0.36)

Year dummies included Yes Yes Yes Yes Yes Yes

0.13 0.14 0.14 0.16 0.17 0.18

Wald

␹

(df)4.67** (1) 4.66* (2) 6.18* (3) 9.94** (4) 11.71** (5)

n(firms) 77; n(observations) 707. Huber-White robust standard errors are in parentheses.

Logarithm.

†

p.10

*p.05

** p.01

Two-tailed tests.

904 AugustAcademy of Management Journal

find evidence of a curvilinear effect. Hypothesis 2

predicts network density strengthens the effect of

network technological diversity on exploratory in-

novation. Models 5–6 show the interaction had a

significant, positive effect on exploratory innova-

tion, supporting Hypothesis 2. The interaction of

network diversity squared and network density

was not significant (model 6). Although not pre-

dicted, network density had a positive and signifi-

cant effect on exploratory innovation, independent

of diversity (models 4–5). The Wald statistics at the

bottom of Table 3 indicate models 2–6 provide

significant improvement in fit relative to model 1. I

constructed each test for incremental improve-

ment in fit relative to the baseline model, because

making it relative to the previous model would

have provided the same information as the sig-

nificance level of the newly entered variable,

since each explanatory variable was entered

alone (Gujarati, 1995). The Appendix contains an

assessment of the robustness of the results and

alternative explanations.

DISCUSSION

This study was motivated by important limita-

tions of research on alliance networks and firm

innovation. This literature has largely ignored the

potential influence of network composition, partic-

ularly the technological diversity of a firm’s part-

ners. This research also draws on seemingly incom-

patible theoretical arguments and has produced

conflicting empirical results regarding the influ-

ence of network structure. These conflicts stem

from an assumption that a firm’s access to diverse

information and the innovation benefits of network

closure are mutually exclusive. In part because of

this assumption, potential complementarities be-

tween network structure and composition have

been largely unexamined. Finally, research on alli-

ance networks and firm innovation has focused on

the volume of firm innovation, with little consid-

eration of its exploratory content.

This study addressed these limitations by exam-

ining the influence of the composition and struc-

ture of a firm’s network of horizontal technology

alliances on its degree of exploratory innovation.

The theoretical framework suggested network com-

position and structure play different, yet comple-

mentary, roles in exploratory innovation. Regard-

ing network composition, I drew on research on

recombinatory search to predict that the technolog-

ical diversity in a firm’s alliance network has an

inverted U-shaped relationship with its exploratory

innovation. Although increasing diversity in-

creases the number, variety, and novelty of poten-

tial innovative combinations, excessive diversity

impairs a firm’s ability to recognize and utilize

knowledge components in its network, reducing its

ability to produce exploratory innovations. Regard-

ing network structure, I built on research on inter-

firm networks and interfirm learning and argued

the density of a firm’s horizontal alliance network

increases its ability to access, mobilize, and inte-

grate its partners’ knowledge, thus increasing its

ability to benefit from technologically diverse part-

ners. In so doing, this study moved beyond the

dyadic perspective typically used in interfirm

learning research (cf. Tiwana, 2008).

The results are mostly consistent with the pre-

dictions of the theoretical framework. I predicted a

curvilinear effect of network diversity, yet I found

evidence of a positive linear effect on exploratory

innovation. I speculate on this result below. I also

found the density of a firm’s network of horizontal

technology alliances strengthened the effect of di-

versity. These results do not seem to be biased by

endogeneity and are robust to the use of many firm-

and alliance-level controls, alternative specifica-

tions and estimation routines, firm fixed and ran-

dom effects, and the use of alternative measures.

Although I predicted an inverted U-shaped effect

of network technological diversity, I found a posi-

tive, linear effect. There are at least three possible

explanations for this result. First, sample firms may

have avoided alliances with excessively diverse

partners. Indeed, Mowery et al. (1998) found that

firms typically avoid forming alliances with highly

dissimilar partners. Without a sufficient number of

excessively diverse networks, only a linear rela-

tionship can be observed. Although this argument

suggests the parameter estimates for network diver-

sity and its square might be biased by sample self-

selection, I tested for such endogeneity and found

none (see the Appendix). Second, Rosenkopf and

Almeida (2003) found that once a firm had formed

an alliance, it was just as likely to learn from tech-

nologically dissimilar firms as from similar firms.

They theorized firms typically make the necessary

investments in interfirm learning mechanisms to

learn effectively from highly diverse partners. If

sample firms typically made such investments,

then they would have been able to mitigate, to some

extent, the absorptive capacity problems associated

with increasingly diverse partners. Finally, increas-

ing technological distance among a firm’s partners

may have increased their willingness to share

knowledge with the focal firm because they were

less concerned their knowledge would leak to ri-

vals via a common partner. When a firm’s network

consists of partners with similar and thus substi-

tutable knowledge stocks, competitive concerns

2010 905Phelps

can lead them to withhold information and knowl-

edge from a common partner to prevent its leakage

to rivals via this common intermediary (Khanna,

Gulati, & Nohria, 1998).

This study has important implications for re-

search and practice. First, this study contributes to

a debate in the literature concerning the network

structure of social capital by suggesting that re-

search has overemphasized the informational ben-

efits of structural holes for firm innovation. The

prior research assumption has been that structural

holes increase an actor’s timely access to diverse

information. Because structural holes and network

closure are inversely related, this argument implies

the informational benefits of structural holes must

come at the expense of the benefits of network

closure, and vice versa. Prior conflicting findings

about the effect of structural holes on firm innova-

tion may be influenced by a confounding of the

structural holes effect with an unobserved compo-

sitional effect of partner knowledge diversity. This

study suggests the extent to which an actor’s net-

work is composed of alters with diverse knowledge

bases will provide it access to informational diver-

sity, independent of network structure. The bene-

fits of network closure and access to diverse infor-

mation and know-how can coexist in a firm’s

alliance network, and the combination of the two

enhances its exploratory innovation. This finding

coincides with the results of a recent longitudinal

qualitative study of interfirm networks. In their

examination of six biotechnology firms, Maurer

and Ebers (2006) found firms with dense networks

of partners with diverse resources experienced

greater growth and development.

Second, this study contributes to the innovation

search literature. Much of this literature stresses

the proclivity of firms to practice local search. Lit-

tle research explores how firms are able to over-

come the inertial tendencies of local search. The re-

sults of this study suggest having access to diverse

knowledge is important. This finding reinforces and

complements the results of recent alliance-level re-

search, which shows partner technological diversity

affects the rate of firm innovation (Sampson, 2007).

While Sampson (2007) also found that the use of

equity joint venture, a formal alliance governance

mechanism, positively moderated the influence of

partner dissimilarity on firm innovation, my results

suggest informal governance provided by network

closure positively moderated the influence of net-

work-level diversity on firm exploratory innova-

tion performance. Research has shown alliances

enhance firm innovation performance, but it is dif-

ficult to establish from these past studies whether

firms expanded their technical competencies in the

process. The findings of this study suggest alliances

can spur exploratory innovation when they provide

access to technologically diverse partners that are

densely connected to one another.

Finally, the results of this study have managerial

implications. The findings confirm alliances can

improve a firm’s development of exploratory inno-

vations. The theory and results point to the benefits

of forming alliances with technologically diverse

partners in densely connected networks. Thus,

managers should attend to the structure of the alli-

ance networks in which their firms are embedded,

because these structures have implications for firm

performance. Although technology alliance part-

ners are often selected based on their technological

capabilities (Stuart, 1998), the results of this study

suggest a firm’s ability to learn from technologi-

cally diverse partners depends on the degree of

network closure around these relationships. Man-

agers should evaluate how their choices about

forming new alliances and ending existing relation-

ships will affect the structure of their networks.

Moving from the dyad level of analysis to the net-

work level can sensitize managers to the impor-

tance of understanding how social structure influ-

ences firm performance (Gulati, 1998).

The results and contributions of this study

should be considered in light of its limitations.

First, although I emphasized the benefits of dense

and diverse alliance networks for firm innovation, I

did not consider their long-term costs. Research

suggests network density reduces the diversity of

information available in a network over time (Lazer

& Friedman, 2007). Dense links provide redundant

paths to the same information sources. Soon every-

one in the network comes to have the same infor-

mation (Burt, 1992). Over time, this homogeneity

would harm innovation. This argument implies

that the diversity of information in a network is

fixed and results from the diversity of information

possessed by actors when the network was formed.

Thus, the only way to inject novel information into

a network is to add connections to new actors who,

as a function of their ties to others outside the focal

network, can provide such novelty (Burt, 1992;

Granovetter, 1973). Access to diverse information

is determined solely by the connective structure of

ties among actors (Obstfeld, 2005).

These are unrealistic assumptions. Not only does

this argument assume actors are equally and easily

able to absorb or imitate the information they do

not initially possess, it also rules out the possibility

of recombinant innovation. Given some degree of

heterogeneity among actors in the information and

knowledge they possess, the sharing and diffusion

of these resources provides the potential for their

906 AugustAcademy of Management Journal

novel recombination into new knowledge that did

not previously exist (Fleming, 2001). If innovation

is a process of the recombination of existing knowl-

edge, then innovations actually increase the poten-

tial for subsequent innovations. In short, recombi-

nations beget more recombinations (Fleming,

2001). From this perspective, a network established

with some degree of diversity in the information

and knowledge actors possess will facilitate the

development of even more diverse information and

knowledge. Thus, diversity begets diversity (Kauff-

man, 1995: 291). Moreover, as the results of this

study suggest, network density can facilitate this

process. Rather than driving out diversity, dense

networks that begin with specialized and therefore

diverse actors may generate more, rather than less,

diversity. Although a detailed investigation of this

issue was beyond the scope of this study, it repre-

sents an important topic for future research.

Next, because I used patents to assess exploratory

innovation, the measure may not capture all of a

firm’s exploratory innovations. If firms systemati-

cally patent explorative knowledge for unobserved

reasons, parameter estimates may be biased. I at-

tempted to control for this potential source of bias

using control variables and firm effects. Addition-

ally, firms may patent knowledge in anticipation of

entering alliances because of concerns about future

leakage of this knowledge to partners (Brouwer &

Kleinknecht, 1999). Exploratory inventions tend to

have a greater impact on subsequent technological

development (Rosenkopf & Nerkar, 2001) and may

therefore be of greater economic value (Narin,

Noma, & Perry, 1987). Thus, firms may patent ex-

ploratory inventions before entering alliances to

appropriate their greater economic value. The use

of a one-year lag between collaboration and patent-

ing and the use of firm effects reduces the likeli-

hood of such a bias.

Another possible limitation is that an alliance

survivor bias may have influenced the results. If

sample firms formed alliances with the intent of

exploratory learning and if successful alliances sur-

vived, then observed alliances will be those that

yielded the greatest exploratory benefit. Such a

self-selection bias is unlikely in this study. First,

because I have time-varying data on alliances and I

observe alliance formation and dissolution, my

data include both successful and unsuccessful alli-

ances. Second, research shows firms often exit al-

liances before they yield knowledge transfer bene-

fits (Deeds & Rothaermel, 2003) and often maintain

alliances that negatively affect interfirm knowl-

edge transfer (Gomes-Casseres, Hagedoorn, &

Jaffe, 2006). Third, firms enter technology alli-

ances for reasons other than technological explora-

tion (Hagedoorn, 1993). Finally, if alliances that are

beneficial for exploration tend to survive, I would

expect a positive effect of the number of alliances

maintained by a firm on its exploratory innovation.

I do not observe such an effect.

Finally, the archival data used in this study can-

not provide direct evidence of the causal processes

and mechanisms that I hypothesized. Although my

hypothesis concerning network density relied on

an established and empirically validated argument

that density promotes trust and reciprocity, my

data did not allow me to observe trust and reciproc-

ity among personnel involved in the sample alli-

ances. The results are consistent with theoretical

expectations, yet a better understanding of the mi-

crosociological foundations that underlie the ob-

served effects of alliance network structure and

composition is needed to validate the causal infer-

ences of this study. In particular, longitudinal quali-

tative research should explore how interorganization-

al and interpersonal networks interact to produce

social capital and how this social capital influences

knowledge transfer and innovation.

Conclusion

Because firms have strong incentives to pursue

exploitation at the expense of exploration, the

question of how and when firms are able to explore

effectively is fundamental to understanding how

organizations adapt, thrive, and survive. As Moran

and Ghoshal concluded, “An organization that is

not adequately enabling and motivating new possi-

bilities is more likely to witness its own decline”

(1999: 410). The results of this study reinforce the

“relational view” of firm resource creation and ad-

vantage (Dyer & Singh, 1998) by helping to identify

the conditions under which alliances enable a firm

to create exploratory technological innovations that

can provide it with the technological foundations

for new commercial possibilities. The results sug-

gest the benefits of network closure and access to

diverse information can coexist in a firm’s alliance

network and the combination of the two increases

exploratory innovation.

REFERENCES

Adler, P. S., & Kwon, S. K. 2002. Social capital: Prospects

for a new concept. Academy of Management Re-

view, 27: 17–40.

Adner, R., & Levinthal D. 2008. Doing versus seeing: acts

of exploitation and observations of exploration.

Strategic Entrepreneurship Journal, 2: 43–52.

Ahuja, G. 2000. Collaboration networks, structural holes

2010 907Phelps

and innovation: A longitudinal study. Administra-

tive Science Quarterly, 45: 425–455.

Ahuja, G., & Katila, R. 2001. Technological acquisitions

and the innovation performance of acquiring firms:

A longitudinal study. Strategic Management Jour-

nal, 22: 197–220.

Ahuja, G., & Lampert, C. M. 2001. Entrepreneurship in

the large corporation: A longitudinal study of how

established firms create breakthrough inventions.

Strategic Management Journal, 22: 521–543.

Alcacer, J., & Gittelman, M. 2006. How do I know what

you know? Patent examiners and the generation of

patent citations. Review of Economics and Statis-

tics, 88: 774–779.

Amesse, F., Latour, R., Rebolledo, C., & Se´ guin-Dulude,

L. 2004. The telecommunications equipment indus-

try in the 1990s: From alliances to mergers and ac-

quisitions. Technovation, 24: 885–897.

Argote, L. 1999. Organizational learning: Creating, re-

taining, and transferring knowledge. Boston: Klu-

wer Academic.

Audia, P. G., & Goncalo, J. A. 2007. Past success and

creativity over time: A study of inventors in the hard

disk drive industry. Management Science, 53: 1–15.

Baum, J. A. C., Calabrese, T., & Silverman, B. S. 2000.

Don’t go it alone: Alliance network composition and

startups’ performance in Canadian biotechnology.

Strategic Management Journal, 21: 267–294.

Benner, M. J., & Tushman, M. 2002. Process management

and technological innovation: A longitudinal study

of the paint and photography industries. Adminis-

trative Science Quarterly, 47: 676– 698.

Borgatti, S. P. 1997. Structural holes: Unpacking Burt’s

redundancy measures. Connections, 20(1): 35–38.

Brouwer, E., & Kleinknecht, A. 1999. Innovative output

and a firm’s propensity to patent: An exploration of

CIS micro data. Research Policy, 28: 615–624.

Burt, R. S. 1992. Structural holes. Cambridge, MA: Har-

vard University Press.

Burt, R. S. 2004. Structural holes and good ideas. Amer-

ican Journal of Sociology, 110: 349–399.

Cohen, W. M., & Levinthal, D. A. 1990. Absorptive ca-

pacity: A new perspective on learning and innova-

tion. Administrative Science Quarterly, 35: 128–

152.

Coleman, J. S. 1988. Social capital in the creation of

human capital. American Journal of Sociology,

94(supplement): S95–S120.

Davidson, R., & MacKinnon, J. G. 1993. Estimation and

inference in econometrics. New York: Oxford Uni-

versity Press.

Deeds, D. L., & Rothaermel, F. T. 2003. Honeymoons and

liabilities: The relationship between age and perfor-

mance in research and development alliances. Jour-

nal of Product Innovation Management, 20: 468–

484.

Dittrich, K., Duysters, G., & de Man, A.-P. 2007. Strategic

repositioning by means of alliance networks: The

case of IBM. Research Policy, 36: 1496–1511.

Dosi, G. 1988. Sources, procedures, and microeconomic

effects of innovation. Journal of Economic Litera-

ture, 26: 1120–1171.

Duncker, K. 1945. On problem solving. Psychological

Monographs, 58(5): 270.

Dutta, S., & Weiss, A. M. 1997. The relationship between

a firm’s level of technological innovativeness and its

pattern of partnership agreements. Management

Science, 43: 343–356.

Dyer, J. H., & Nobeoka, K. 2000. Creating and managing a

high-performance knowledge-sharing network: The

Toyota case. Strategic Management Journal, 21:

345–367.

Dyer, J. H., & Singh, H. 1998. The relational view: Coop-

erative strategy and sources of interorganizational

competitive advantage. Academy of Management

Review, 23: 660– 679.

Fleming, L. 2001. Recombinant uncertainty in technolog-

ical search. Management Science, 47: 117–132.

Fleming, L., & Sorenson, O. 2001. Technology as a com-

plex adaptive system: Evidence from patent data.

Research Policy, 30: 1019–1039.

Friedkin, N. E. 1981. The development of structure in

random networks: An analysis of the effects of in-

creasing network density on five measures of struc-

ture. Social Networks, 3: 41–52.

Galunic, D. C., & Rodan, S. 1998. Resource recombina-

tions in the firm: Knowledge structures and the po-

tential for Schumpeterian innovation. Strategic

Management Journal, 19: 1193–1201.

Garcia-Vega, M. 2006. Does technological diversification

promote innovation? Research Policy, 35: 230–246.

Gilsing, V., & Noteboom, B. 2006. Exploration and ex-

ploitation in innovation systems: The case of phar-

maceutical biotechnology. Research Policy, 35:

1–23.

Gomes-Casseres, B., Hagedoorn. J., & Jaffe, A. 2006. Do

alliances promote knowledge flows? Journal of Fi-

nancial Economics, 80: 5–33.

Granovetter, M. S. 1973. The strength of weak ties. Amer-

ican Journal of Sociology, 78: 1360–1380.

Greene, W. H. 1997. Econometric analysis (3rd ed.).

Upper Saddle River, NJ: Prentice Hall.

Greve, H. R. 2007. Exploration and exploitation in prod-

uct innovation. Industrial and Corporate Change,

16: 945–975.

Griliches, Z. 1990. Patent statistics as economic indica-

tors: A survey. Journal of Economics Literature, 28:

1661–1707.

908 AugustAcademy of Management Journal

Gujarati, D. N. 1995. Basic econometrics (3rd ed.). New

York: McGraw-Hill.

Gulati, R. 1995. Does familiarity breed trust? The impli-

cations of repeated ties for contractual choice in

alliances. Academy of Management Journal, 38:

85–112.

Gulati, R. 1998. Alliances and networks. Strategic Man-

agement Journal, 19: 293–317.

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

tional networks come from? American Journal of

Sociology, 104: 1439–1493.

Gulati, R., Nohria, N., & Zaheer, A. 2000. Strategic

networks. Strategic Management Journal, 21: 203–

215.

Gulati, R., & Singh, H. 1998. The architecture of cooper-

ation: Managing coordination costs and appropria-

tion concerns in strategic alliances. Administrative

Science Quarterly, 43: 781–814.

Gulati, R., & Sytch, M. 2008. Does familiarity breed trust?

Revisiting the antecedents of trust. Managerial and

Decision Economics, 29: 165–190.

Gupta, A. K., Smith, K. G., & Shalley, C. E. 2006. The

interplay between exploration and exploitation.

Academy of Management Journal, 49: 693–706.

Hagedoorn, J. 1993. Understanding the rationale of stra-

tegic technology partnering: Interorganizational

modes of cooperation and sectoral differences. Stra-

tegic Management Journal, 14: 371–385.

Hagedoorn, J., & Cloodt, M. 2003. Measuring innovative

performance: Is there an advantage in using multiple

indicators? Research Policy, 32: 1365–1379.

Hall, B. H. 2002. A note on the bias in the Herfindahl

based on count data. In A. Jaffe & M. Trajtenberg

(Eds.), Patents, citations, and innovation: 149–156.

Cambridge, MA: MIT Press.

Hamel, G. 1991. Competition for competence and inter-

partner learning within international strategic alli-

ances. Strategic Management Journal, 12(summer

special issue): 83–103.

Hansen, M. T. 1999. The search-transfer problem: The

role of weak ties in sharing knowledge across orga-

nization subunits. Administrative Science Quar-

terly, 44: 82–111.

Hargadon, A., & Sutton, R. I. 1997. Technology brokering

and innovation in a product development firm. Ad-

ministrative Science Quarterly, 42: 716–749.

Hausman, J. 1978. Specification tests in econometrics.

Econometrica, 46: 1251–1271.

He, Z. L., & Wong, P. K. 2004. Exploration versus exploi-

tation: An empirical test of the ambidexterity hy-

pothesis. Organization Science, 15: 481–494.

Helfat, C. E. 1994. Evolutionary trajectories in petroleum

firm R&D. Management Science, 40: 1720–1747.

Henderson, R. M., & Clark, K. B. 1990. Architectural

innovation: The reconfiguration of existing product

technologies and the failure of established firms.

Administrative Science Quarterly, 35: 9–30.

Hicks, D., Breitzman, T., Olivastro, D., & Hamilton, K.

2001. The changing composition of innovative activ-

ity in the US—A portrait based on patent analysis.

Research Policy, 30: 681–703.

Holm, D. B., Eriksson, K., & Johanson, J. 1999. Creating

value through mutual commitment to business net-

work relationships. Strategic Management Journal,

20: 467–486.

Husted, B. W. 1994. Transaction costs, norms, and social

networks. Business and Society, 33(1): 30– 44.

Ingram, P. 2002. Interorganizational learning. In J. A. C.

Baum (Ed.), Companion to organizations: 642–663.

New York: Blackwell.

Inkpen, A. C., & Tsang, E. W. K. 2005. Social capital,

networks, and knowledge transfer. Academy of

Management Review, 30: 146–165.

Jaffe, A. 1986. Technological opportunity and spillovers

of R&D: Evidence from firms’ patents, profits and

market value. American Economic Review, 76:

984–1001.

Jaffe, A. B., Trajtenberg, M., & Fogarty, M. S. 2002. The

meaning of patent citations: Report on the NBER/

Case-Western Reserve survey of patentees. In A. B.

Jaffe & M. Trajtenberg (Eds.), Patents, citations and

innovations: A window on the knowledge econ-

omy: 379– 401. Cambridge, MA: MIT Press.

Jansen, J. J. P., Van Den Bosch, F. A. J., & Volberda, H. W.

2006. Exploratory innovation, exploitative innova-

tion, and performance: Effects of organizational an-

tecedents and environmental moderators. Manage-

ment Science, 52: 1661–1674.

Jensen, R. J., & Szulanski, G. 2007. Template use and the

effectiveness of knowledge transfer. Management

Science, 53: 1716–1730.

Kale, P., Singh, H., & Perlmutter, H. 2000. Learning and

protection of proprietary assets in strategic alliances:

Building relational capital. Strategic Management

Journal, 21: 217–238.

Katila, R., & Ahuja, G. 2002. Something old, something

new: A longitudinal study of search behavior and

new product introduction. Academy of Manage-

ment Journal, 45: 1183–1194.

Kauffman, S. 1995. At home in the universe: The search

for the laws of self-organization and complexity.

New York: Oxford University Press.

Khanna, T., Gulati, R., & Nohria, N. 1998. The dynamics

of learning alliances: Competition, cooperation, and

relative scope. Strategic Management Journal, 19:

193–210.

Kogut, B. 1988. Joint ventures: Theoretical and empirical

perspectives. Strategic Management Journal, 9:

319–332.

2010 909Phelps

Kogut, B. 2000. The network as knowledge. Strategic

Management Journal, 21: 405–425.

Kogut, B., & Zander, U. 1992. Knowledge of the firm,

combinative capabilities, and the replication of tech-

nology. Organization Science, 3: 383–397.

Kogut, B., & Zander, U. 1996. What firms do? Coordina-

tion, identity, and learning. Organization Science,

7: 502–518.

Kreps, D. 1990. Corporate culture and economic theory.

In J. Alt & K. Shepsie (Eds.), Perspectives on posi-

tive political economy: 90–143. New York: Cam-

bridge University Press.

Lane, P. J., & Lubatkin, M. 1998. Relative absorptive

capacity and interorganizational learning. Strategic

Management Journal, 19: 461–477.

Larson, A. 1992. Network dyads in entrepreneurial set-

tings: A study of the governance of exchange rela-

tionships. Administrative Science Quarterly, 37:

76–104.

Laumann, E. O., Marsden, P. V., & Prensky, D. 1983. The

boundary specification problem in network analysis.

In R. S. Burt & M. J. Minor (Eds.), Applied network

analysis: 18–34. Beverly Hills, CA: Sage.

Lavie, D. 2006. The competitive advantage of intercon-

nected firms: An extension of the resource-based

view. Academy of Management Review, 21: 825–

860.

Layton, E. 1974. Technology as knowledge. Technology

and Culture, 15: 31–41.

Lazer, D., & Friedman, A. 2007. The network structure of

exploration and exploitation. Administrative Sci-

ence Quarterly, 52: 667–694.

Leonard-Barton, D. 1995. Wellsprings of knowledge:

Building and sustaining the sources of innovation.

Boston: Harvard Business School Press.

Levin, R. C., Klevorick, A. K., Nelson, R. H., & Winter, S.

1987. Appropriating the returns from industrial re-

search and development. Brookings Papers on Eco-

nomic Activity, 3: 783–831.

Levinthal, D., & March, J. G. 1981. A model of adaptive

organizational search. Journal of Economic Behav-

ior and Organization, 2: 307–333.

Levinthal, D. A., & March, J. G. 1993. The myopia of

learning. Strategic Management Journal, 14: 95–

112.

Levinthal, D. A., & Fichman, M. 1988. Dynamics of interor-

ganizational attachments: Auditor-client relation-

ships. Administrative Science Quarterly, 33: 345–

369.

Lyles, M. A., & Salk, J. E. 1996. Knowledge acquisition

from foreign parents in international joint ventures:

An empirical examination in the Hungarian context.

Journal of International Business Studies, 27: 877–

903.

March, J. G. 1991. Exploration and exploitation in organ-

izational learning. Organization Science, 2: 71–87.

Martin, X., & Mitchell, W. 1998. The influence of local

search and performance heuristics on new design

introduction in a new product market. Research

Policy, 26: 753–771.

Maurer, I., & Ebers, M. 2006. Dynamics of social capital

and their performance implications: Lessons from

biotechnology start-ups. Administrative Science

Quarterly, 51: 262–292.

McEvily, B., & Zaheer, A. 1999. Bridging ties: A source of

firm heterogeneity in competitive capabilities. Stra-

tegic Management Journal, 20: 1133–1156.

McGrath, R. G., & Nerkar, A. 2004. Real options reasoning

and a new look at the R&D investment strategies of

pharmaceutical firms. Strategic Management Jour-

nal, 25: 1–21.

Mezias, S. J., & Glynn, M. A. 1993. The three faces of

corporate renewal: Institution, revolution, and evo-

lution. Strategic Management Journal, 14: 77–101.

Moran, P., & Ghoshal, S. 1999. Markets, firms, and the

process of economic development. Academy of

Management Review, 24: 390– 412.

Mowery, D. C., Oxley, J. E., & Silverman, B. S. 1996.

Strategic alliances and interfirm knowledge transfer.

Strategic Management Journal, 17(winter special

issue): 77–91.

Mowery, D. C., Oxley, J. E., & Silverman, B. S. 1998.

Technological overlap and interfirm cooperation:

implications for the resource-based view of the firm.

Research Policy, 27: 507–523.

Nahapiet, J., & Ghoshal, S. 1998. Social capital, intellec-

tual capital, and the organizational advantage. Acad-

emy of Management Review, 23: 242–266.

Narin, F., Noma, E., & Perry, R. 1987. Patents as indica-

tors of corporate technological strength. Research

Policy, 16: 143–155.

Nelson, R. R., & Winter, S. 1982. An evolutionary theory

of economic change. Cambridge, MA: Harvard Uni-

versity Press.

Nerkar, A. 2003. Old is gold? The value of temporal

exploration in the creation of new knowledge. Man-

agement Science, 49: 211–229.

Nohria, N., & Gulati, R. 1996. Is slack good or bad for

performance? Academy of Management Journal,

39: 1245–1264.

Nonaka, I. 1994. A dynamic theory of organizational

knowledge creation. Organization Science, 5:

14–37.

Obstfeld, D. 2005. Social networks, the tertius iungens

orientation, and involvement in innovation. Admin-

istrative Science Quarterly, 50: 100–130.

Owen-Smith, J., & Powell, W. W. 2004. Knowledge net-

works as channels and conduits: The effects of spill-

910 AugustAcademy of Management Journal

overs in the Boston biotechnology community. Or-

ganization Science, 15: 5–21.

Page, S. E. 2007. The difference: How the power of

diversity creates better groups, firms, schools, and

societies. Princeton, NJ: Princeton University Press.

Papke, L. W., & Wooldridge, J. M. 2005. Panel data meth-

ods for fractional response variables with an appli-

cation to test pass rates. Journal of Econometrics,

145: 212–233.

Patel, P., & Pavitt, K. 1997. The technological competen-

cies of the world’s largest firms: Complex and path-

dependent, but not much variety. Research Policy,

26: 141–156.

Pavitt, K. 1988. Uses and abuses of patent statistics. In

A. F. J. van Raan (Ed.), Handbook of quantitative

studies of science and technology: 509–536. Am-

sterdam: Elsevier.

Phelps, C. C., Heidl, R., & Wadhwa, A. 2010. Networks,

knowledge, and knowledge networks: A literature

review. Working paper, HEC Paris, Jouy-en-Josas,

France.

Pisano, G. P. 1989. Using equity participation to support

exchange: Evidence from the biotechnology indus-

try. Journal of Law, Economics and Organization,

5: 109–126.

Portes, A. 1998. Social capital: Its origins and applica-

tions in modern sociology. In J. Hagan & K. S. Cook

(Eds.), Annual review of sociology, vol. 24: 1–24.

Palo Alto, CA: Annual Reviews.

Powell, W. W. 1990. Neither market nor hierarchy: Net-

work forms of organization. In B. M. Staw & L. L.

Cummings (Eds.), Research in organizational be-

havior, vol. 12: 295–336. Greenwich, CT: JAI Press.

Raubb, W., & Weesie, J. 1990. Reputation and efficiency

in social interactions: An example of network ef-

fects. American Journal of Sociology, 96: 626– 654.

Reagans, R. W., & McEvily, B. 2003. Network structure

and knowledge transfer. Administrative Science

Quarterly, 48: 240–267.

Reagans, R. W., Zuckerman, E. W., & McEvily, B. 2004.

Two holes in one? Information and control in the

analysis of structural advantage. Working paper,

Carnegie-Mellon University, Pittsburgh.

Rivette, K. G. 1993. Corporate patent portfolios. Technol-

ogy Forecast (New York: Price Waterhouse).

Robinson, D. T., & Stuart, T. E. 2007. Network effects in

the governance of strategic alliances. Journal of

Law, Economics, and Organization, 23: 242–273.

Rodan, S., & Galunic, C. 2004. More than network struc-

ture: How knowledge heterogeneity influences man-

agerial performance and innovativeness. Strategic

Management Journal, 25: 541–562.

Rooks, G., Raub, W., Selten, R., & Tazelaar, F. 2000. How

inter-firm co-operation depends on social embed-

dedness: A vignette study. Acta Sociologica, 43:

123–137.

Rosenkopf, L., & Almeida, P. 2003. Overcoming local

search through alliances and mobility. Management

Science, 49: 751–766.

Rosenkopf, L., & Nerkar, A. 2001. Beyond local search:

Boundary spanning, exploration and impact in the

optical disc industry. Strategic Management Jour-

nal, 22: 287–306.

Rowley, T., Behrens, D., & Krackhardt, D. 2000. Redun-

dant governance structures: An analysis of structural

and relational embeddedness in the steel and semi-

conductor industries. Strategic Management Jour-

nal, 21: 369–386.

Sampson, R. C. 2004. Organizational choice in R&D alli-

ances: Knowledge-based and transaction cost per-

spectives. Managerial and Decision Economics, 25:

421–436.

Sampson, R. C. 2005. Experience effects and collabora-

tive returns in R&D alliances. Strategic Manage-

ment Journal, 26: 1009–1031.

Sampson, R. C. 2007. R&D alliances and firm perfor-

mance: The impact of technological diversity and

alliance organization on innovation. Academy of

Management Journal, 50: 364–386.

Schilling, M., & Phelps, C. 2007. Interfirm collaboration

networks and knowledge creation: The impact of

large scale network structure on firm innovation.

Management Science, 53: 1113–1126.

Scott, J. 1991. Social network analysis: A handbook.

London: Sage.

Shan, W., Walker, G., & Kogut, B. 1994. Interfirm coop-

eration and startup innovation in the biotechnology

industry. Strategic Management Journal, 15: 387–

394.

Sidhu, J. S., Commandeur, H. R., & Volberda, H. W. 2007.

The multifaceted nature of exploration and exploi-

tation: Value of supply, demand, and spatial search

for innovation. Organization Science, 18: 20–38.

Siggelkow, N., & Rivkin, J. W. 2005. Speed and search:

Designing organizations for turbulence and com-

plexity. Organization Science, 16: 101–122.

Simonton, D. K. 1999. Origins of genius. New York:

Oxford University Press.

Silverman, B. S. 1999. Technological resources and the

direction of corporate diversification: Toward an in-

tegration of transaction cost economics and the re-

source-based view. Management Science, 44:

1109–1124.

Singh, J. V. 1986. Performance, slack, and risk taking in

organizational decision making. Academy of Man-

agement Journal, 29: 562–585.

Smith-Doerr, L., Owen-Smith, J., Koput, K. W., & Powell,

W. W. 1999. Networks and knowledge produc-

tion: Collaboration and patenting in biotechnology.

2010 911Phelps

In R. Th. A. J. Leenders & S. Gabbay (Eds.), Corpo-

rate social capital: 331–350. Norwell, MA: Kluwer

Academic.

Smith-Doerr, L., & Powell, W. W. 2005. Network and

economic life. In N. Smelser & R. Swedberg (Eds.),

The handbook of economic sociology (2nd ed.):

379– 402. Princeton, NJ: Russell Sage Foundation/

Princeton University Press.

Soh, P.-H. 2003. The role of networking alliances in

information acquisition and its implication for new

product performance. Journal of Business Ventur-

ing, 18: 727–744.

Sorensen, J. B., & Stuart, T. E. 2000. Aging, obsolescence,

and organizational innovation. Administrative Sci-

ence Quarterly, 45: 81–112.

Stirling, A. 2007. A general framework for analysing di-

versity in science, technology and society. Journal

of the Royal Society Interface, 4: 707–719.

Stuart, T. E. 1998. Network positions and propensities to

collaborate: An investigation of strategic alliance for-

mation in a high-technology industry. Administra-

tive Science Quarterly, 43: 668– 698.

Stuart, T. E. 2000. Interorganizational alliances and the

performance of firms: A study of growth and inno-

vation rates in a high-technology industry. Strategic

Management Journal, 21: 791–811.

Stuart, T., & Podolny, J. 1996. Local search and the evo-

lution of technological capabilities. Strategic Man-

agement Journal, 17(summer special issue): 21–38.

Szulanski, G. 1996. Exploring internal stickiness: Imped-

iments to the transfer of best practice within the

firm. Strategic Management Journal, 17(winter

special issue): 27–43.

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

tion: Organizational arrangements for regimes of

rapid technological progress. Journal of Economic

Behavior and Organization, 18: 1–25.

Tiwana, A. 2008. Do bridging ties complement strong

ties? An empirical examination of alliance ambidex-

terity. Strategic Management Journal, 29: 251–272.

Trajtenberg, M. 1987. Patents, citations, and innova-

tions: Tracing the links. Working paper no. 2457,

National Bureau of Economic Research. Washington,

DC.

Uzzi, B. 1996. The sources and consequences of embed-

dedness for the economic performance of organiza-

tions: The network effect. American Sociological

Review, 61: 674– 698.

Uzzi, B. 1997. Social structure and competition in inter-

firm networks: The paradox of embeddedness. Ad-

ministrative Science Quarterly, 42: 35–67.

von Hayek, F. 1945. The use of knowledge in society.

American Economic Review, 35: 519–530.

von Hippel, E. 1988. The sources of innovation. New

York: Oxford University Press.

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

structural holes and the formation of an industry

network. Organization Science, 8: 109–125.

Wasserman, S., & Faust, K. 1994. Social network analy-

sis: Methods and applications. Cambridge, U.K.:

Cambridge University Press.

Zander, U., & Kogut, B. 1995. Knowledge and the speed

of transfer and imitation of organizational capabili-

ties: An empirical test. Organization Science, 6:

76–92.

APPENDIX

Alternative Explanations and Robustness Checks

I considered several alternative explanations and as-

sessed the robustness of the results. First, I removed the

time-invariant variables and used firm fixed-effects. The

results were similar to those obtained using random ef-

fects, which is consistent with the insignificant Hausman

tests (1978) mentioned in the main text. Next, I consid-

ered the potential endogeneity of network structure and

network diversity. The formation and dissolution of alli-

ances reflect choices made by firms. These choices may

be based on expectations of the exploration-enhancing

benefits of alliances. This introduces the possibility of an

unobserved sample self-selection process causing an en-

dogeneity bias. Network structure may, however, be ex-

ogenous for a few reasons. Firms form technology alli-

ances for reasons other than exploratory innovation

(Hagedoorn, 1993) and do not easily or quickly alter their

alliances to optimize their networks for particular objec-

tives (Maurer & Ebers, 2006). Thus, at any point in time,

alliance networks are not necessarily structured to max-

imize exploratory innovation and are, at least weakly,

exogenous. Last, the structure of a firm’s alliance net-

work is beyond the sole control and influence of any one

firm in the network and is therefore not a firm choice

variable. Although network diversity may change slowly

because of inertia in a firm’s alliance relationships and

thus may not be optimized for exploratory innovation at

a given point in time, the level of diversity in a firm’s ego

network is largely under its control.

Because endogeneity is an empirical question, I tested

for the presence of deleterious endogeneity related to

both network density and network diversity. I used Da-

vidson and MacKinnon’s test (1993), as implemented by

the “dmexogxt” procedure in Stata 10. This test com-

pares the estimated coefficient for the assumed endoge-

nous regressor (e.g., density or network diversity) ob-

tained from ordinary least square (OLS) fixed-effects

regression with the estimate obtained from a two-stage

instrumental variables fixed-effects regression. The null

hypothesis is that OLS fixed effects yields a consistent

parameter estimate. This procedure requires a valid in-

strumental variable for the two-stage estimator so that the

second-stage estimates can be identified. I used firm

technological diversity to instrument for network density

and network diversity in separate regressions because it

was not significantly correlated with exploratory innova-

912 AugustAcademy of Management Journal

tion but was correlated with density and network diver-

sity. Neither the endogeneity test associated with net-

work density nor that associated with network diversity

was significant. Thus, the parameter estimates for these

variables in Table 3 do not appear to be unduly influ-

enced by endogeneity.

I performed additional unreported analyses to assess

the robustness of my findings. First, I experimented with

alternative specifications by removing insignificant vari-

ables and then removing all control variables. The results

related to the three explanatory variables were robust to

these alternative specifications. Second, I estimated the

full model using a GEE approach in which I specified a

probit link function and an exchangeable working corre-

lation matrix and computed robust standard errors

(Papke & Wooldridge, 2005). Results from this analysis

for the three explanatory variables were consistent with

those reported in Table 3. Third, I substituted Burt’s

(1992) measures of network efficiency and constraint for

the density measure discussed above. The results ob-

tained using these alternative measures of ego network

closure were statistically stronger but otherwise consis-

tent with those reported in Table 3. Finally, I used the

alternative measure of exploratory innovation discussed

in the main text. Because this variable was a count and

took on only nonnegative integer values, I estimated the

full model with negative binomial panel regression, us-

ing year dummies and firm random effects (Greene,

1997). The results were consistent with those reported in

Table 3. Overall, the results of the various robustness

analyses converged and provided added support for both

hypotheses.

Corey Phelps (phelps@hec.fr) is an associate professor of

strategy and business policy at HEC Paris. He completed

his Ph.D. in strategic management at the Stern School of

Business, New York University. Dr. Phelps’s current re-

search is focused on understanding how companies learn

from extramural sources of knowledge.

2010 913Phelps

Interfirm Collaboration Networks:: The Impact of Network Structure on Rates of Innovation

Chapter

Full-text available

Mar 2021

A jack-of-all-trades or a master of none: the performance effects of balancing exploration and exploitation within vs across alliance portfolio domains

Article

Sep 2018
J Knowl Manag

Purpose Theory predicts that balancing exploratory and exploitative learning (i.e., ambidexterity) across alliance portfolio domains (e.g. value chain function, governance modes) increases firm performance, whereas balance within domains decreases performance. Prior empirical work, however, only assessed balance/imbalance within and across two domains. The purpose of this study is to determine if theory generalizes beyond specific domain combinations. The authors investigated across multiple domains to determine whether alliance portfolios should be imbalanced toward exploration or exploitation within domains or balanced across domains. The authors also extended prior research by exploring whether the direction of imbalance matters. Current theory only advises managers to accept imbalance without helping with the choice between exploration and exploitation. Design/methodology/approach Hypotheses are tested using fixed-effects generalized least squares (GLS) regression analysis of a large 13-year panel sample of Fortune 500 firms from 1996 to 2008. Findings With respect to the balance between exploration and exploitation within each of the five domains investigated, imbalanced alliance portfolios had higher firm performance. No evidence was found that balance across domains relates to performance. Instead, for four of the five domains, imbalance toward exploration related positively to firm performance. Originality/value An alliance portfolio that allows for exploration in some domains and exploitation in other domains appears more difficult to implement than prior theory suggests. Firms benefit mostly from using the alliance portfolio for exploratory learning.

Country-governance quality and signals in the alliance formation process: a meta-analytic approach

Article

Jan 2022

Country-governance quality and signals in the alliance formation process: a meta-analytic approach

Article

Full-text available

Jan 2022

Going green: the dynamics of green technological alliances

Article

Aug 2020

This paper investigates the impact of strategic technological alliances and environmental regulation, and their interaction, on the generation of green technologies. The empirical analysis is carried out on a newly constructed dataset of European firms over the period 2005–2011 and it is articulated in two steps. Firstly, we test the existence of a relationship between the environmental regulation, as measured by the OECD Environmental Policy Stringency index, and GTs, proxied by patent applications. We then employ a dynamic network analysis model to explore the dual role of GTs both as the determinant of the collaboration network and as the outcome of firms' collaboration strategies. We find that, even though there exists a strong and positive relationship, the regulatory framework has not a direct effect on GTs but rather it stimulates firms to search for new and qualified collaborations. Then, it is the nature and the structure of these collaborations that encourages firms to generate new green technological knowledge.

A Transaction Cost Perspective of Alliance Portfolio Diversity

Article

Jun 2019

Alliance portfolio diversity (APD) helps firms access diverse capabilities and knowledge. APD can also increase transaction costs, but it is unknown whether and how transaction cost theory's (TCT's) insights about hierarchical integration operate at the portfolio level. We adapt TCT to the portfolio level to suggest that the transaction costs from APD encourage integration into alliance partners' industries, and we introduce the concept of shared‐specific investments to pinpoint one source of transaction costs within portfolios and predict which industries will be integrated. Using data from 1996‐2013 on S&P 500 firms, we find evidence in support of our theorising. Juxtaposing results with other theoretical perspectives suggests that TCT offers complementary insights about which activities to perform in the firm versus the alliance portfolio. This article is protected by copyright. All rights reserved.

Frontiers of Strategic Alliance Research

Article

Full-text available

Mar 2019

Cambridge Core - Strategic Management - Frontiers of Strategic Alliance Research - edited by Farok J. Contractor

Surviving Alliance Network Evolution during Industry Convergence: Negotiating, Structuring and Governing Partnerships

Chapter

Feb 2019

Frontiers of Strategic Alliance Research - edited by Farok J. Contractor March 2019

The rise and fall of R and D networks

Article

Aug 2017

Drawing on a large database of publicly announced R &D alliances, we empirically investigate the evolution of R & D networks and the process of alliance formation in several manufacturing sectors over a 24-year period (1986-2009). Our goal is to empirically evaluate the temporal and sectoral robustness of a large set of network indicators, thus providing a more complete description of R &D networks with respect to the existing literature. We find that most network properties are not only invariant across sectors, but also independent of the scale of aggregation at which they are observed, and we highlight the presence of core-periphery architectures in explaining some properties emphasized in previous empirical studies (e.g. asymmetric degree distributions and small worlds). In addition, we show that many properties of R & D networks are characterized by a rise-and-fall dynamics with a peak in the mid-nineties. We find that such dynamics is driven by mechanisms of accumulative advantage, structural homophily, and multiconnectivity. In particular, the change from the "rise" to the "fall" phase is associated to a structural break in the importance of multiconnectivity.

Research on value-creation efficiency of regional generic culture overflow and diffusion in SEM innovation network

Conference Paper

Aug 2014

Xuan Wang

Overflow and diffusion of regional generic culture can produce great value in SEM(small and medium enterprise) innovation network. The spread of this culture affect the action value of the enterprise and the product" value of their own in addition to affect the person's psychology to create value. Based on the mechanism of each type of value creation, we explore the value creation channel of the regional generic cultural diffusion in theoretical level whether it has the reality guiding sense, and use confirmatory factor analysis tested the hypothesis. Verification results show that the CFA measurement model has good distinction validity. Using this model, we analysis the efficiency of node enterprise in SEM innovation network node. Study found that psychological value, action value and “products” value can effectively distinguish the efficiency type and have high percentage. This research is an important measure innovation method about cultural value-creation efficiency, and has become an important standard of dividing clusters in regional network, which has tremendous implications for the development of regional economy.

Something Old, Something New: A Longitudinal Study of Search Behavior and New Product Introduction

Article

Full-text available

Dec 2002

Does Familiarity Breed Trust? The Implications of Repeated Ties for Contractual Choice in Alliances

Article

Feb 1995

Ranjay Gulati

Performance, Slack, and Risk Taking in Organizational Decision Making

Article

Sep 1986

Jitendra V. Singh

Is Slack Good or Bad for Innovation?

Article

Oct 1996

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.

Interfirm cooperation and startup innovation in the biotechnology industry

Article

Aug 1993

This paper examines the association between interfirm cooperation and the innovation output of startup firms in the biotechnology industry. A reciprocal association is hypothesized. The results, however, show only that cooperation affects innovation, not the reverse. Several control variables are related to cooperation and innovation, especially startup size and the startup's position in the cooperative network.

Appropriating the Returns from Industrial Research and Development

Article

Jan 1987

Knowledge in the network and the network as knowledge

Article

Jan 1993

Networks and alliances

Article

Apr 1998
STRATEGIC MANAGE J

Ranjay Gulati

This paper introduces a social network perspective to the study of strategic alliances. It extends prior research, which has primarily considered alliances as dyadic exchanges and paid less attention to the fact that key precursors, processes, and outcomes associated with alliances can be defined and shaped in important ways by the social networks within which most firms are embedded. It identifies five key issues for the study of alliances: (1) the formation of alliances, (2) the choice of governance structure, (3) the dynamic evolution of alliances, (4) the performance of alliances, and (5) the performance consequences for firms entering alliances. For each of these issues, this paper outlines some of the current research and debates at the firm and dyad level and then discusses some of the new and important insights that result from introducing a network perspective. It highlights current network research on alliances and suggests an agenda for future research.© 1998 John Wiley & Sons, Ltd.

Social Capital in the Creation of Human Capital

Article

Jan 2000

J.S. Coleman

Technological exploration : a longitudinal study of the role of recombinatory search and social capital in alliance networks /

Abstract and Figures

Recommended publications

Essays on investor heterogeneity /

An exploration of information technology use for inter-unit coordination [microform] /

Construction through deconstruction : a compositional approach to the development of brand names /

Managing the globalization process : an exploratory study : a thesis /