Vous êtes sur la page 1sur 23

Strategic Management Journal

Strat. Mgmt. J., 29: 639661 (2008)


Published online in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/smj.679
Received 31 May 2002; Final revision received 2 January 2008

DESIGNING ALLIANCE NETWORKS: THE INFLUENCE


OF NETWORK POSITION, ENVIRONMENTAL
CHANGE, AND STRATEGY ON FIRM PERFORMANCE
BALAJI R. KOKA1 and JOHN E. PRESCOTT2 *
1
W. P. Carey School of Business, Arizona State University, Tempe, Arizona, U.S.A.
2
Katz Graduate School of Business, University of Pittsburgh, Pittsburgh,
Pennsylvania, U.S.A.

Alliance networks are strategic decisions involving trade-offs between two stylized structural
design choices: prominent and entrepreneurial. Prominent alliance networks emphasize benefits
arising out of multiple access and affiliation to other prominent firms in the network. An
entrepreneurial position, on the other hand, emphasizes brokerage and diversity benefits arising
out of access to nonredundant and diverse information. We demonstrate that the performance
benefits of each type of alliance network are contingent on environmental change and strategy,
and are thus time dependent. Following an environmental change event in the steel industry,
alliance networks that were more entrepreneurial performed better, while those that were more
prominent suffered performance decline. However, when the change was radical, both types of
alliance networks were negatively related to performance. We suggest that following a radical
change, industry alliance networks may not have the requisite information necessary for quick
and effective strategic responses. Firms pursuing an analyzer strategy performed better when
emphasizing a prominent, and to a lesser extent, entrepreneurial alliance network. However, firms
that develop an alliance network high on both prominent and entrepreneurial structural positions
had lower relative performance. Our results indicate the need for managers to assess their
alliance portfolio over time and redesign it based on environmental and strategic contingencies.
Copyright 2008 John Wiley & Sons, Ltd.

INTRODUCTION networks affect performance has important man-


agerial consequences. Research has emphasized
If and how alliance network positions influence two distinct yet interrelated approaches to the
firm performance has become a dominant theme structural design of alliance networks. In the first
of strategy and organizational research. Such inter- approach, firms pursue a network strategy that
est reflects the widespread belief that the struc- enables them to attain a position of prominence in
ture of alliance networks influences firm perfor- the network. Benefits to network prominence arise
mance and such influence is more often than not out of access to critical and key information in
positive. Since the structural design of alliance the network due to the formation of multiple ties
networks has significant strategic consequences, with many partners. In addition, such a position
establishing the conditions under which alliance enables the firm to influence its partners in ways
that enable it to pursue and establish its own strate-
gic agenda, thus enhancing its own performance. In
Keywords: alliance networks; network position; entrepre- the second approach, firms pursue a network strat-
neurial network position; prominent network position; egy that enables them to occupy an entrepreneurial
network structure and performance; social networks position in the network. Benefits from such an
*Correspondence to: John E. Prescott, Katz Graduate School of
Business, University of Pittsburgh, Pittsburgh, PA 15 260, U.S.A. entrepreneurial position occur because of nonre-
E-mail: prescott@ katz.pitt.edu dundant information, information diversity, and

Copyright 2008 John Wiley & Sons, Ltd.


640 B. R. Koka and J. E. Prescott

control of information flow, which in turn enhance time, managers need to assess their alliance port-
firm performance. folios and make structural adjustments based on
However there is little consistent empirical evi- environmental and strategic contingencies. Third,
dence for the hypothesis that either of these firms pursuing an analyzer strategy perform bet-
two different structural approaches to designing ter when they develop network positions that are a
alliance networks positively affect firm perfor- combination of prominent and, to a lesser extent,
mance. For instance, being central (a proxy for entrepreneurial positions. Fourth, firms that have
prominence) in the network was associated with high levels on both network positions have lower
more innovation in the biotech industry (Powell, relative performance. Our findings cumulatively
Koput, and Smith-Doerr, 1996) but poorer per- reinforce the need for researchers to identify and
formance in the steel industry (Madhavan, 1996). explicate the conditions under which the struc-
Similarly, multiple structural holes (a proxy for ture of network positions are likely to enhance or
an entrepreneurial position) in the network were detract from firm performance.
associated with higher innovativeness (Zaheer and
Bell, 2005) but fewer patents (Ahuja, 2000a).
We argue that such inconsistencies in the evi- THEORY DEVELOPMENT
dence are a result of a theoretical approach that
has typically not examined the contextually-based In recent years, strategic alliances have become an
performance trade-offs between the two network integral component of a firms strategy to access
positions over time. Network researchers have and acquire resources from external sources. View-
only recently begun to examine contingencies that ed as access relationships, alliances act as conduits
impact the design of a firms alliance portfolio for the flow of hitherto unavailable resources and
(Burt, 2000; Rowley, Behrens, and Krackhardt, capabilities. Information, knowledge, and other
2000; Koka and Prescott, 2002; Podolny, Stu- resources diffuse through the network created by
art, and Hannan, 1996). Building on these stud- the individual alliances and interfirm linkages that
ies, we make the argument that these two net- bind all the firms in the network. In essence, such a
work positions provide different benefits and con- network is the reservoir of all the benefits residing
sequently will have different effects on firm perfor- within the industry, and the firms network position
mance conditioned by environmental change and is an indicator of the benefits the firm can access
firm strategy. Instead of asking if network posi- through its alliance relationships. However, there
tions affect firm performance, we thus focus on has been considerable debate regarding how differ-
a more important and nuanced question: under ent types of structural network positions influence
what conditions are the two different types of the flow of benefits and consequently firm perfor-
network positions effective in enhancing firm per- mance.
formance? We study this question by focusing Two different yet interrelated structural appro-
on two important contingenciesenvironmental aches have largely driven research (Podolny,
change and firm strategythat we argue interact 2001). In essence, these two approaches reflect
with the two different network positions to effect alternative network strategies employed by firms
firm performance. in the pursuit of competitive advantage. The first
Our findings uniquely contribute to the network approach emphasizes the benefits arising to firms
literature in four ways. First, when environmen- because of their prominence in the network. The
tal change occurs in an industry, firms that have second approach draws on Burts (1992) structural
developed an entrepreneurial position perform bet- holes construct to emphasize entrepreneurial ben-
ter while firms that were prominent in the network efits arising to firms. Debates on the importance
suffer performance decline. However, when there of dense versus sparse networks (Coleman, 1988;
is more rather than less radical change, both net- Burt, 1992); direct versus indirect ties (Ahuja,
work positions are negatively related to perfor- 2000b); information volume versus information
mance. Second, our methods allowed us to ask diversity (Koka and Prescott, 2002); centrality
if the benefits of a network position are stable (Powell, et al. 1996) versus structural holes (Burt,
over time. Our results indicate that the effect of 1992); seem to naturally align both explicitly and
particular network positions on firm performance implicitly with these two structural positions. For
changes over time. This finding suggests that over instance, structural holes, indirect ties, and sparse
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
Designing Alliance Networks 641

networks all focus on the entrepreneurial potential the direct contact that they have with multiple part-
for brokerage and diversity as the basis for net- ners. Information is available at a shorter distance
work benefits. Researchers have, however, tended (because of direct relations or fewer intermedi-
to explore the implications of these similar con- aries) to well-connected firms, and is therefore
structs separately. There has been little attempt available early. Direct access also reduces search
to develop a broad theory of network structural costs (Cyert and March, 1963), which enhances
position that integrates these similar constructs efficiency of search and transfer process.
into a broad umbrella position by focusing on Apart from quantity, network prominence ben-
the nature of the underlying benefits that accrue efits also stem from the quality of access. Mul-
to the firm. Similarly, arguments underlying the tiple ties between partners engender interdepen-
importance of multiple direct ties with many part- dence between the partners (Rowley et al., 2000).
ners, and/or centrality in the design of alliance Interdependence is directly related to the level of
networks, all highlight the benefits accruing to resource commitment to the relationship. High lev-
firms because of multiple accesses to valuable els of resource commitment increase frequency of
information in the network. Here again, rarely interaction between the partners, which enhances
have network researchers examined the degree to the quality of the exchange relationships. Such
which these different constructs are theoretically ties enable firms to know more about each others
and empirically related. Given the above frag- operations, facilitating the transfer of tacit knowl-
mented nature of this literature, it is therefore not edge across organizational boundaries (Hagg et al.,
surprising that there is little consistent evidence 1982; Larson 1992) that is geared toward problem
for the structural effect of alliance networks on solving (Uzzi, 1996). Transfer is also facilitated
firm performance (Koka and Prescott, 2002). We by the institution of similar processes and routines
draw on these themes to explicate the two differ- between the exchange partners. Similarity in pro-
ent approaches to designing alliance networks, to cesses and routines in turn facilitate joint and coor-
review their underlying theoretical logic and deter- dinated action by the partners resulting in enhanced
mine how researchers have operationalized the two effectiveness and consequent performance benefits.
positions, and to identify the sources and benefits Network prominence benefits go beyond access
of each position. to include those based on affiliation. As Podolny
(2001) points out, alliances are also signaling
devices. Attributions based on the characteristics
Network prominence of the partner firms are significant influences on
firms eventual performance because of legitimacy
Firms pursuing prominence in the alliance network and reputation effects. In other words, affiliations
seek to take advantage of network benefits arising to prominent firms by way of alliances enhance
out of two distinct mechanisms. First, prominence the reputation of the partner firm leading to per-
provides significantly more access to key and valu- formance benefits. At one level, legitimacy con-
able information relative to those less prominent cerns ensure that prominent firms seek to partner
in the network. Second, affiliation-based benefits with other firms that have similar status. Designing
provide opportunities for the prominent firm to alliance networks following this strategy ensures
establish its competitive and strategic agenda as that prominent firms continue to be prominent
the defining norm in the industry/network. Theo- in their network. Additionally, formation of ties
retically, both lead to competitive advantage and with similar others leads to alignment of inter-
superior performance. ests. It thus facilitates mobilization and pooling of
Access arguments primarily focus on the quan- resources aimed at legitimizing and establishing
tity of reliable information that the prominent firm dominance in the industry. Such joint and coordi-
can acquire through its network. Benefits here is a nated action facilitates the exploitation of existing
function of the number of partners with whom the technologies and organizational practices result-
firm has alliances, as well as the number of ties it ing in competitive advantage. Network prominence
has with each partner. Closely related to the idea of also depends on forming relationships with less
multiple accesses is the notion of centrality. Cen- prominent firms. For the less prominent firms, rela-
tral firms possess more access to key information tionships with prominent firms provide legitimacy
that is locally residing in the network because of benefits. Prominent firms on the other hand benefit
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
642 B. R. Koka and J. E. Prescott

by way of favorable contracts and benefits aris- diffusion between different and often unconnected
ing from their stronger bargaining position. Such groups in the network (Burt, 2000). In addi-
a bargaining position translates into an opportu- tion, firms in an entrepreneurial position benefit
nity for the prominent firm to establish standards from information diversity because of their focus
and norms with their weaker partners that favor its on partner diversity in designing their alliance
own bases of competitive advantage, thus further- network.
ing its strategic agenda. Thus prominence by virtue Such a position draws on the key insight that
of its many ties with multiple partners enables network density is rarely uniform throughout a
the widespread transmission of processes and rou- network. While parts of a network are likely
tines that revolve around particular technologies to be dense, other areas are more likely to be
and strategies favored by the prominent firm. Not sparsely populated with ties. It thus provides
only do these alliances enable the prominent firm entrepreneurial opportunities to firms that can act
to control and exploit existing opportunities in the as bridges between the different parts of the net-
industry, they also enable it to control the trajec- work. Such entrepreneurial opportunities exist only
tory and direction of any new knowledge created in so far as the actors contacts are otherwise discon-
the network by controlling and directing resources nected, such as in a brokering role. The more the
only to those projects that favor its own competi- actors partners know one another, the more likely
tive position. it is that the actors ability to bridge the different
Finally, being prominent in the network also groups is diminished.
enables firms to reinforce their structural position. A key construct that underlies the entrepre-
Multiple experiences at managing collaborative neurial position is the structural holes construct
activities across firm boundaries create opportu- that emphasizes the social capital benefits aris-
nities to leverage alliance capabilities with other ing out of control over the flow of information
partners in other contexts. Further, initial collab- and the access to diverse information. Since social
orations lead to reputation effects that facilitate actors with many structural holes are likely to
the creation of more ties, not only with each possess sparsely connected networks (because of
other, but also with other partners, which increases the lack of ties between the firms partners), they
their prominence (Powell et al., 1996). Since such benefit from brokering the connection between
partners are more likely to be those that already the groups that are otherwise unconnected (Burt,
have ties with the firms existing partners, net- 2000). They are more likely to have informa-
work prominence may also result in dense cohe- tion benefits that are additive because their con-
sive networks that further facilitate transfer of tacts are nonredundant. They also have access to
valuable fine-grained information. Since informa- diversity of information because they bridge dif-
tion acquisition, exploitation, and dissemination is ferent groups that have different information con-
time-consuming and costly, firms that are highly tent. Such diversity of information benefits also
prominent in the network are likely to realize sig- occurs because of the differences in the character-
nificant benefits necessary to maintain competitive istics of the partners (Koka and Prescott, 2002).
advantage. A limitation of such a position is that Consistent with the firm heterogeneity tenet, firms
new and different information not part of the local in a network possess differential information con-
network may not be received, may be received tent. Since differentially endowed firms linked in a
more slowly, or may be subject to attention and network operate in different market segments, uti-
blind-spot biases. Thus, when change occurs in an lize different technologies, and sometimes belong
industry that requires access to new and/or differ- to different industries, they are sources of differ-
ent types of information, firms that are prominent ent information content. Between two firms hav-
in the existing network are at a considerable dis- ing the same number of alliances, the firm that
advantage. has more diverse contacts (say ties distributed
among firms in different strategic groups) has
access to more nonredundant and diverse infor-
Entrepreneurial position
mation.
Firms pursuing an entrepreneurial position in the From an alliance network point of view, such a
network seek to take advantage of the benefits aris- network position primarily provides opportunities
ing out of both control and participation in resource for the firm to recombine the diverse information
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
Designing Alliance Networks 643

and knowledge into novel combinations that enable and entrepreneurial position with environmental
it to provide unique offerings to the market. In that change and firm strategy influences their effect on
sense it enables the exploration of new markets firm performance.
providing first mover advantages to the firm. Such
first mover advantages are also facilitated by the
fact that the firms in the entrepreneurial position Environmental level of analysis contingency:
get new and novel information earlier than others environmental change
in the network. Advantages to such a position also
include the nonredundancy benefits of the network The influence of environments on firm perfor-
structure since it provides for an efficient use of the mance has been one of the central themes in strat-
firms alliance capabilities. In addition, due to their egy (Porter, 1980). More directly related to our
strategic location at the crossroads of information research question is the contingency approach that
between different groups, such firms also con- has focused on explaining firm performance as a
trol the flow and diffusion of information within function of the fit between organizational struc-
the network by providing these groups opportuni- ture and environmental characteristics (Lawrence
ties to form consortia and networks wherein they and Lorsch, 1967). Extending this argument to
can orchestrate the actions of multiple firms with the interorganizational level, researchers (Rowley
diverse capabilities in order to create new mar- et al., 2000; Burt, 2000) have argued that the effect
kets and products. Thus, apart from getting access of network positions on firm performance is mod-
to diverse information earlier than others, such erated by environmental context. Underlying this
firms more than others are likely to control the argument is the fact that changes in the environ-
pace and rate of change and innovation in the net- ment affect the nature of skills and capabilities
work thereby enhancing their competitive position important for competitive advantage. Since net-
in the network. A limitation of such a position work positions are a critical mechanism for firms
is that the ability to develop strong cohesive ties to acquire these capabilities and since different net-
is attenuated and thus the transfer of tacit knowl- work positions provide differential benefits, firms
edge is limited. Such a position is also less con- need to strategically design their networks if they
ducive to the quick mobilization and coordination are to acquire the appropriate skills and capabil-
of resources because the lack of alignment of inter- ities in the changed environment. Any mismatch
ests between the partners ultimately has an impact between the needs of the environment and the
on efficiency. firms alliance position could thus have serious
performance implications.
We build on Burt (2000) and Rowley et al.
HYPOTHESES DEVELOPMENT (2000) studies to examine this contingency. Both
argued that turbulent environments require a net-
An appreciation for the prominent and entrepre- work position that facilitates exploration, while
neurial approaches to the structural design of stable environments emphasize the importance of
alliance networks is important because they focus network positions that enable exploitation. How-
our attention on the different sources and types ever, Rowley et al. (2000) found limited support
of benefits that are available to a firm from for their proposition that strong ties are valuable
its network of interfirm relations. Understanding in environments that require exploitation, while
the different sources and types of benefits that weak ties would be more appropriate for explo-
accrue to a firm as a result of these two struc- ration. Apart from the narrow focus on only tie-
tural positions provide an opportunity for man- strength to operationalize network position, their
agers to design an alliance network that fits weak results could be attributed to an implicit
their strategic context. Thus, our logic leads us assumption in their analysis that environmental
to propose a contingency element in specify- characteristics and competitive dynamics are stable
ing the relationship of the two types of net- over time. Such an assumption may not be valid.
work positions and firm performance. We exam- Even in the steel industry, which is the context of
ined two contingencies at the environment and their study, the implementation of Compact strip
firm levels of analysis to explore the proposi- casting technology changed competitive dynam-
tion that the interaction of network prominence ics. In other words, while exploitation might be
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
644 B. R. Koka and J. E. Prescott

valid prior to the introduction of a new technol- their network with a view to reinforcing their
ogy, network positions that enable exploration may competitive advantage. By definition, however,
be more appropriate in the new environment. We environment-changing events change the existing
examine this proposition by focusing on environ- bases of competition. Firms that were prominent
mental change as a key contingency that affects the prior to the environment-changing event are likely
effect of network position on firm performance. to be adversely affected by this shift in the environ-
We follow an event-based logic to operational- ment. Much of the information from their existing
ize environmental change. Evidence suggests that ties may lose its relevance in the new environ-
firms take advantage of environmental change trig- ment. Because it is built around existing tech-
gered by industry events as occasions to structure nologies and routines, such a position limits the
their networks (Barley, 1986; Madhavan, Koka and extent to which the firm can change its direc-
Prescott, 1998) since such changes alter the bases tion and strategy. Instead of forming relationships
of competition. Such changes thus create pressures with new and diverse partners, prominent firms
for different capabilities requiring firms to mod- are likely to respond to this uncertainty by rein-
ify the criteria for selecting partners and managing forcing their existing network by forming more
alliances. A network position that was effective in ties within their existing network partners, which
the previous environment may be inappropriate in further negatively affects their competitive posi-
the new environment following the industry event. tion. Uzzis (1996) study highlights this paradox
Firms need to maneuver to discard old alliances of such a position that on the one hand provides
and seek new partners who can provide benefits fine-grained information for exploitation, but on
necessary for competitive advantage in the new the other hand could result in adverse effects if all
environment. Following such events, the entire net- the ties in the firms network comprise similar ties
work is in considerable flux as firms strive to that do not provide diverse information. Recent
attain network positions that will be effective in criticisms of the dysfunctional effects of social
the changed environment. capital (Locke, 1999) have similarly focused on
Focusing on industry events as occasions for the negative consequences of firms getting locked
environmental change as a key driver affecting net- into their existing relationships.
work change also highlights the need for firms Firms that are entrepreneurially positioned are
to achieve a balance between exploration and ideally positioned to take advantage of the shift
exploitation in order to be consistently success- in the bases of competitive advantage. Not only
ful. As Levinthal and March (1993) suggest, an are their networks designed to secure nonredun-
overemphasis on either one could lead to dete- dant information, they are explicitly designed to
rioration in competitive advantage. Their logic take advantage of diversity within their network,
suggests that following an environment-changing which enables, as we noted earlier, novel combi-
event, firms would be better off following an nations and recombinations of knowledge in the
exploration orientation. As the change stabilizes, network. Such explorations into new technology,
firms should exploit the new learning acquired routines, new offerings, and markets enable firms
when the environment was in flux. More impor- in an entrepreneurial position to take advantage of
tantly, such an approach leads to the perspective the opportunities provided by the changed environ-
that industry environments differ not in terms of ment. It thus enables them to more effectively com-
whether they emphasize exploration or exploita- pete in the new environment resulting in enhanced
tion types of learning, but in terms of when either performance. Thus:
type of learning is more appropriate. It thus places
a premium on firms to (re)design their network Hypothesis 1: Following an environment-chang-
positions to fit the new environmental context if ing event, firms that are more prominently posi-
they are to enhance performance. tioned in the interfirm network will exhibit lower
As we observed earlier, firms that are promi- relative performance.
nently positioned in the current network are un-
likely to go beyond their existing ties to seek Hypothesis 2: Following an environment-chang-
out novel information even when the environment ing event, firms that are more entrepreneurially
changes, which might be inappropriate given that positioned in the interfirm network will exhibit
firms pursuing prominence in the network design higher relative performance.
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
Designing Alliance Networks 645

Firm level of analysis contingency: firm prospector on either end of the continuum with
strategy analyzers in the middle (Miles and Snow, 1978;
Doty et al., 1993). During the time period of our
The alliance strategy of a firm constitutes an study, firms in the steel industry were consid-
important component of a firms strategy. The type ered to be relatively conservative (Hogan, 1991).
of alliances, the criteria to choose partners, the tim-
As such, prospector-like activity was limited and
ing of the alliances, and the relationship process
as we will demonstrate below, most firms pur-
affecting the alliances so formed are driven by the
sued either an analyzer or defender strategy. We
firms objectives and strategies. In the steel indus-
thus frame our strategy hypotheses on the ana-
try for instance, alliance activity clearly reflected
lyzer strategy, implicitly recognizing that defender
firm goals and objectives. Thus control of raw
strategies are represented by our operationalization
materials drove the mining ventures of the fifties,
of firm strategy.
technology needs led to a rash of technical assis-
Analyzers have been defined as a unique com-
tance agreements through the early eighties and
bination of the [p]rospector and [d]efender types
market access needs drove the United States-Japan
(Miles and Snow, 1978: 68). Analyzers typically
alliances thereafter in the late eighties. If a firm
service a firms stable base of traditional products
is to enhance competitive advantage, the type of
network position that managers design should be and customers while simultaneously locating new
in consonance with its overall strategy. Any mis- product and market opportunities that are closely
match may have a detrimental effect on firm per- related to its existing technological base. In this
formance. A firms strategy is thus an important respect, analyzers are like defenders in that a sig-
contingency factor. nificant portion of their revenues is based on serv-
To specify the directional hypotheses for the ing a stable product market. However, they operate
interacting effect of firm strategy, we classified in a few more product markets than the defenders,
firm strategy in terms of the Miles and Snow resulting in a broader scope than defenders (Doty
typology. Miles and Snow (1978) distinguished et al., 1993). In addition to their core technology,
between three types of firm strategies: Analyzer, they have other technologies that enable them to
Prospector, and Defender. The effectiveness of serve additional products and markets. Thus, ana-
these strategies depends on internal consistency lyzers are likely to be more diversified than defend-
and fit between the different structural, contex- ers. However, they are unlike prospectors in that
tual, and strategic factors of the organization they rarely are the first movers into these markets
(Doty, Glick, and Huber, 1993). Several stud- nor are they likely innovators of new technologies.
ies have adopted this contingency argument to As Miller (1988) pointed out, they are more likely
study the effectiveness of these strategies. Ham- to enter these markets or adopt new technologies
brick (1983), for instance, examined the effective- only after they are well established. Examples of
ness of these strategies under differing environ- analyzers are Japanese integrated steel firms. They
ments and concluded that defenders were more appear to be more analyzer-like because apart from
likely to be successful in stable environments. producing traditional steel products and structural
Miller (1988: 239) developed what he termed steel using continuous casting technology, they
technology-strategy gestalts as a way of exam- have expanded into related services that include
ining the effect of fit between technology and steel consulting in the emerging markets. In addi-
strategy on firm performance. Snow and Hrebiniak tion to servicing their traditional domestic mar-
(1980) found that only firms whose strategies were kets, they have moved internationally to service
accompanied by appropriate competences exhib- Japanese auto transplants in the United States and
ited higher performance. In the context of this the European Union.
study, adopting the Miles and Snow typology to Firms following an analyzer strategy require
test for the interacting effect of the network posi- frequent and, more importantly, reliable industry-
tion and strategy on firm performance is thus con- specific information flow that results from being
sistent with the approach adopted by prior research prominent in their network. Prominence in the
studies. network enables them to exploit their existing
Most studies have viewed these three strate- technological base, resulting in significant opera-
gies as forming a continuum with defender and tional efficiencies because competitive advantage
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
646 B. R. Koka and J. E. Prescott

partially lies in the heavy exploitation of exist- Hypothesis 3: Firms pursuing an analyzer strat-
ing technology. Such firms require specific, reli- egy and an alliance structure characterized by
able information from others whose competitive network prominence will exhibit higher rela-
advantage is also based on similar technology in tive performance than firms pursuing a defender
order to drive down operating costs (Von Hippel, strategy.
1988). Critical for the flow of such information is
the need for multiple durable relationships char- Emphasizing the dysfunctional effect of an
acterized by strong interdependence between the entrepreneurial position in the network leads us
partners. to argue that firms that pursue a defender strat-
Analyzers also expand into new products and egy may exhibit lower relative performance if
markets that arise out of leveraging opportuni- they design alliance structures that provide them
ties created by their existing technological base. with a large amount of unique and diverse infor-
Given their reluctance to move into markets and/or mation. Being highly entrepreneurially positioned
technologies as a first mover, being more promi- provides firms the advantage of getting unique
nent provides them with reliable information and
information from different parts of the network
knowledge that reduces uncertainty. More impor-
that are more suitable for firms that are prospectors
tantly, such a position enables them to actively
and the prospect-like areas of the analyzer strat-
shape the creation of new knowledge that opens
egy (Miles and Snow, 1978). As Miles and Snow
up markets and products for their new partners.
(1978) argue, prospectors operate in a wide range
For instance, in the steel industry, integrated firms
of environmental conditions. They seek to avoid
partner with other integrated firms in order to
commitment to a single technological process and
develop lighter steel materials that could com-
attain growth by exploring new product and market
pete with plastics and aluminum, thereby expand-
opportunities. Their strategy involves exploration-
ing their markets with the automotive and appli-
based learning in their pursuit of growth and prof-
ance industry. Such a position enables firms to
itability. As Rowley et al. (2000) argue, firms with
offer broader product/market offerings, which in
sparse networks are ideally suited for exploration-
turn enables additional relationships with cur-
related activities because they are able to bridge
rent and new partners and establishes a more
information from diverse parts of the network.
prominent position that results in higher perfor-
Firms that are entrepreneurially positioned are
mance.
thus ideally positioned to pursue a strategy that
Establishing a prominent position however may
enables them to take advantage of the diverse
not be appropriate for a defender as compared to
information such a position provides. However,
an analyzer. Defenders focus on a narrow por-
as noted earlier, the steel industry does not con-
tion of the market. They typically have a sin-
tain many firms pursuing a prospector strategy.
gle core technology that they pursue in order
In terms of framing the hypothesis for analyzers,
to achieve competitive advantage. American min-
imills with electric arc furnace-based technology the question then becomes: how much diversity
and integrated steel mills with batch-processing do analyzers need if they are to enhance perfor-
technologywith their focus on structural steel mance? Since analyzers also operate in a broader
products and the American marketcan thus range of markets, technologies, and products than
be classified as pursuers of a defender strategy. defenders, this argument can be extended to tenta-
While a few close ties leading to exploitation tively argue that analyzers that are entrepreneuri-
of existing technology is important for such a ally positioned get the benefits of diversity of infor-
strategy, defenders in comparison with analyz- mation that enable them to access new markets
ers need fewer ties because their focus is on a and technologies, resulting in higher performance.
smaller segment of the market. Efficiency of oper- Thus,
ations includes efficiency in terms of the number
of collaborative ties necessary to enhance perfor- Hypothesis 4: Firms pursuing an analyzer strat-
mance. Thus, firms that are pursuing a defender egy and an alliance structure characterized by
strategy are likely to be adversely affected if an entrepreneurial position will exhibit higher
they design their networks in pursuit of network relative performance than firms pursuing a
prominence. defender strategy.
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
Designing Alliance Networks 647

The interaction of the two different positions RESEARCH METHODOLOGY


Our final hypothesis examines the interaction bet- We tested our hypotheses using longitudinal allian-
ween the two network positions.1 While these ce data in the global steel industry. Since the eight-
two network positions are analytically distinct in ies, the global steel industry has witnessed sub-
terms of the mechanisms that provide benefits, stantial alliance activity. Licensing and technical
both positions draw on the same network structure agreements enabled the transfer of technology and
to provide these different benefits. Consequently, best practices. Joint ventures facilitated reductions
every firm has some level of both prominent and in capacity by consolidating operations and plants.
entrepreneurial benefits in the network. In other In addition, alliances enabled firms to overcome
words, these two positions are not mutually exclu- the barriers imposed by cross-regulatory borders.
sive with firms varying in terms of the degree to Alliances thus emerged as a significant component
which they pursue either or both simultaneously. of steel firms strategy to manage the compet-
However, given that different processes are neces- itive dynamics (Madhavan et al., 1998), making
sary to access the benefits that accrue to the firm this industry an appropriate arena in which to test
for each position, it is important to examine the our hypotheses.
consequences to firms if they maintain high levels
on both positions simultaneously.
Data
The logic for each position was developed ear-
lier and need not be repeated here. However, given Our analysis focuses on a population of strate-
the difference in the nature of skills and capabil- gic alliances formed in the steel industry between
ities necessary to benefit from these positions, it 19801994.2 Consistent with the definition in the
is unlikely that firms can possess high levels on literature, we identified alliance as any durable
both of these capabilities simultaneously. As pro- collaborative activity that involved exchange of
ponents of the resource-based view and learning resources between firms (Ahuja, 2000b). All steel
theory (March, 1999) argue, it is difficult for firms industry firms that had formed at least one alliance
to gain expertise at both of these different skills during the time period of this study19801994
and capabilities. More importantly, pursuing the were included in our database. The American
different skills and capabilities may actually lead Metal Market (AMM ), which is the major trade
to a dilution of their efforts resulting in poor per- journal of the steel industry, was our primary data
formance. Thus, source. We examined daily issues of the AMM
spanning all 15 years for our data collection pur-
poses.3 We collected information on 706 alliances
Hypothesis 5: Firms that are simultaneously formed between 422 firms from 48 countries; 162
both more prominent and entrepreneurially posi- of which operated in the steel industry. Apart
tioned in the network relative to other firms will from information on the partners, we also col-
exhibit lower relative performance. lected data on the start dates of the alliances,
their end dates (if they were terminated), purpose,
and alliance structure. We used this data to cre-
The distinction between Hypothesis 5 and the
ate alliance networks and matrices for each year,
previous hypotheses related to analyzers is impor-
which were then utilized to operationalize the net-
tant because it highlights that context matters.
work constructs. Since we had end dates for sev-
Hypotheses 3 and 4 propose that firms that put
eral alliances, our network matrices for every year
emphasis either position will experience higher
performance only if they are pursuing an analyzer
2
strategy. Hypothesis 5, on the other hand, proposes We draw on Koka and Prescott (2002) in our description of the
data and methodology.
that putting emphasis on both positions will lead 3
We used multiple data sources in order to ascertain the accuracy
to lower performance if it is achieved irrespective of our database. These include the Metal Bulletin, Dow Jones
of context. Retrieval Index, Annual Reports of various firms, several Steel
Industry Monitoring Reports published by the United States
International Trade Commission. In addition we drew on two
books published on the steel industry (Mangum, Kim, and
1
We thank an anonymous reviewer for encouraging us to Tallman, 1996; Hogan, 1991) to ascertain the accuracy of our
develop this hypothesis. data.

Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
648 B. R. Koka and J. E. Prescott

included only alliances that were active in that position. Stronger ties reflect the potential for inter-
year. Forming year-wise matrices was necessary dependence and capability for coordinated action
to create a panel dataset that enabled a year-wise between the partners. Finally, Bonacich centrality
test of the hypotheses. Such a strategy allowed us is calculated as the weighted sum of the central-
to test whether our theory holds across time, and ities of the firms partners. It captures the extent
was particularly useful in testing the environmental to which the firm is connected to the other cen-
change hypotheses. tral firms, enabling us to evaluate the influence of
We followed Nohria and Garcia-Pont (1991) in the firm to establish the nature of the competitive
assigning a numerical score to each alliance that dynamics in the industry, while simultaneously
reflected the intensity of cooperation and there- providing us with the extent of access that the firm
fore the interdependence of partners on a 10-point has to the industrys information and resources. We
scale. Higher scores indicate higher levels of inten- used the network adjacency matrix to calculate this
sity of cooperation and interdependence, indicating measure using the UCINET V software program
a potential for a greater amount and a higher qual- (Borgatti, Everett, and Freeman, 1999). In addi-
ity of information flow than alliances with lower tion we operationalized network prominence using
scores. In addition, higher interdependence reflects a count measure of the number of partners (PART-
the potential for the partners to engage in coordi- NER) a firm had in each year. This measure is a
nated action that involves considerable mobiliza- simple indicator of the size of the firms network
tion of resources, which as we pointed out earlier (Uzzi and Gillespie, 2002) and measures the num-
is a significant aspect of the strategy of the promi- ber of other firms in the network with whom the
nent firms in the network. Where pairs of partners firm has direct ties. Finally, we used a count mea-
had multiple alliances with each other in a year, sure of the number of ties (TIES) that the firm had
we added the individual scores of each alliance to in each year. PARTNER is the number of partners
form a composite index. Coding alliances along with whom the firm had at least one tie. TIES on
these lines enabled us to incorporate interdepen- the other hand, is the number of direct ties that
dence between the partners into our analysis. a firm had with all its partners, and is designed
to capture the multiple ties that a firm may enjoy
with certain of its partners. Apart from acting as
Operationalization of constructs
indicators for the potential for resource flows to
In line with our argument that network positions the focal firm, all three of these measures provide
should be operationalized using multiple indica- a means of determining the extent of influence that
tors, and in order to capture the different aspects the firm has in the overall network.
of the network constructs, we developed several
measures to arrive at a composite variable for each
Entrepreneurial position (ENTREP)
construct.
Entrepreneurial position was operationalized using
the network measurestructural holes (Burt,
Network prominence (PROMIN)
1992). As noted in the theory section, firm net-
We identified three measures to operationalize net- works that have many structural holes indicate that
work prominence. Theoretically we argued that the partners are relatively unconnected to others
firms that pursued a prominent position in the net- in the network. We calculated this as a constraint
work tended to occupy a central position therein. measure (HOLES) (Burt, 2000) using UCINET V.
Of the many network measures that have been High constraint indicates that the firms partners
utilized to capture the notion of centrality, the are densely connected to one another with high
eigenvector-based measure of centrality (CEN- redundancy in information flow. A low constraint
TRAL) is particularly apposite (Bonacich, 1972). measure on the other hand indicates a sparsely
First, it is a status measure of the firm (Podony connected network. We also developed two other
et al., 1996). As we argued, affiliation with more measures to reflect the diversity in the structure
prominent firms is desired because of legitimacy of the alliance network: the first measured the
benefits. Second, this measure of centrality takes diversity of the partners in terms of their national-
into account the strength of ties between the part- ities; the second reflected the diversity of the part-
ners, which is an integral aspect of a prominent ners in terms of six technology groupsIntegrated
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
Designing Alliance Networks 649

Steel, Specialty Steel, Minimill, Downstream (to are significantly correlated in the same direction
the steel industry), Upstream, and Others (Koka indicating that they are all indicators of a similar
and Prescott, 2002; Madhavan et. al., 1998). We construct. There is thus considerable face validity
then operationalized both diversity measures using for these measures.
Blaus heterogeneity index. A high heterogeneity We conducted a principal components factor
index indicates high diversity in terms of nation- analysis with varimax rotation to develop a com-
ality/technology of the partners. posite measure for each of the position constructs.
As expected, the six measures loaded onto two
Blaus heterogeneity index 1 !Pj 2 ; factors. The two-factor model was consistently
supported on a variety of criteria such as eigen-
where Pj is the proportion of partners in the j values (>than 1), variance explained, and factor
category. loadings.6 We utilized the factor scores developed
Thus, we had two heterogeneity index mea- from the above analysis as a composite measure for
suresnationality heterogeneity index (COUN- each of the network position constructs. In addi-
TRY) and technology heterogeneity index (TECH) tion, these measures were then utilized to develop
to indicate the diversity in the structure of the the interaction variables.
alliance network.
The descriptive statistics (Table 1) reveal signif- Performance
icant correlations between the measures.
As we argued, both types of network positions Based on discussions with several steel industry
draw on some aspect of network structure and is analysts at World Steel Dynamics and Prudential
therefore not very surprising. The magnitude and Securities, we operationalized our dependent vari-
the direction of the correlations are as anticipated. able firm performance as Sales per employee in
Thus, network prominence (PROMIN) measures constant US dollars (PRODUCTIVITY). We used
are all correlated with each other as expected. this measure for several reasons. First it is consis-
The count number of ties (TIES) has a high tent with the way productivity has been measured
correlation to centrality (CENTRAL) (0.89; p < at the macroeconomic level (output/man hour).
0.01) and count measure of partners (PARTNER) Second, prior research on the steel industry (Nair
(0.96; p < 0.01). Similarly, PARTNER and CEN- and Kotha, 2001; Koka and Prescott, 2002) has
TRAL are also correlated (0.78; p < 0.01), provid- used sales per employee as a measure of produc-
ing evidence that they measured the same under- tivity. It also provides a measure of the efficiency
lying component indicating a convergent valid- of the firms operation7 which has been the driving
ity. Further support for the convergent validity concern of many firms in the steel industry (Ghe-
can also be discerned from the correlations for mawat, 1993). Finally, differences in accounting
the entrepreneurial position (ENTREP) measures. practices prevent us from utilizing accounting mea-
sures of performance given that our data includes
Technology heterogeneity (TECH) and country
firms from over 40 countries. Thomson Financials
heterogeneity (COUNTRY) are, as expected, high-
Worldscope database was our main source for the
ly correlated with the HOLES measure (0.78,
p < 0.01; 0.78, p < 0.01 respectively)4 and with
each other (0.77; p < 0.01). Additionally, there is number of ties, and centrality are interrelated. The high cor-
relation between HOLES and PARTNER can be explained by
discriminant validity. TECH and COUNTRY are examining the operationalization. HOLES calculates if the part-
less correlated with CENTRAL (0.38, p < 0.01; ners of a firm are themselves connected to one another. Thus,
0.28, p < 0.01 respectively) and the other network a firm can have many partners who are unconnected among
themselves leading to a high entrepreneurial position with high
prominence measures.5 Finally all the measures nonredundancy and therefore low constraint to the firm. A lower
correlation with CENTRAL can be attributed to the fact that
CENTRAL, on the other hand, depends on relations between
4
The negative correlations reflect the way we operationalized the the partners because it is a based on partner centrality.
variables. A high TECH and COUNTRY measure indicates high 6
The results of all these analyses are available from the first
entrepreneurial position just as a low HOLES measure indicates author.
high entrepreneurial position. 7
Ideal performance measures for the steel industry are typically
5
An issue here is the correlation between the constraint mea- based on steel tonnage. Thus, a better productivity measure
sure for structural holes (HOLES) and the prominence position would be tonnage per employee. However, tonnage details were
measures CENTRAL (0.46; p < 0.01) and PARTNER (0.70; not available for all the firms for all the years on a consistent
p < 0.01). The correlations indicate that ego network density, basis.

Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
Copyright 2008 John Wiley & Sons, Ltd.

650
B. R. Koka and J. E. Prescott
Table 1. Descriptive statistics and correlations

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

1 Eigen centrality 8.50 16.20 1.00 0.78 0.89 0.46 0.38 0.28 0.94 0.13 0.01 0.002 0.095 0.29 0.29 0.34
2 Partners 8.34 9.3 1.00 0.96 0.70 0.56 0.5 0.87 0.41 0.09 0.015 0.05 0.175 0.455 0.28
3 Ties 12.83 17.88 1.00 0.6 0.5 0.42 0.95 0.3 0.08 0.015 0.06 0.25 0.413 0.32
4 Structural holes 0.49 0.33 1.00 0.78 0.78 0.39 0.85 0.09 0.08 0.08 0.09 0.585 0.26
5 Tech. heterogeneity 0.374 0.272 1.00 0.77 0.26 0.88 0.01 0.09 0.23 0.06 0.644 0.37
6 Country heterogeneity 0.42 0.287 1.00 0.14 0.92 0.015 0.1 0.20 0.14 0.601 0.43
7 Network prominence 0 1 1.00 0 0.048 0.02 .151 0.32 0.23 0.50
8 Entrepreneurial position 0 1 1.00 0.03 0.11 0.24 0.209 0.623 0.54
9 Legislative changes 0.25 0.43 1.00 0.38 0.027 0.19 0.08 0.034
10 Technology changes 0.3 0.46 1.00 0.014 .122 0.16 0.066
11 Strategy 0.05 0.11 1.00 0.231 0.384 0.32
12 Performanceproductivity 0.28 0.21 1.00 0.416 0.188
13 Firm sizeemployees 9.09 1.288 1.00 0.242
14 Interaction of prominence and 0.00 0.462 1.00
entrepreneurial


- p < 0.01; p < 0.05; n = 585
Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
Designing Alliance Networks 651

performance data. We supplemented this informa- given a further impetus by the introduction of CSC
tion by drawing on annual reports of the various technology. By lowering the capital costs for set-
firms.8 We lagged the performance variable by ting up a new plant, CSC technology changed the
one year in our analysis, which is consistent with industry structure because it brought new compe-
the use of a productivity measure since efficiency tition to the integrated steel mills in the form of
improvements are likely to have an immediate minimills (Ghemawat, 1993). These events thus
impact. We also conducted the analysis with a two- engendered disequilibrium in the environment in
year lag that produced consistent, but not as strong, the eighties making them a good context to test
results. our hypotheses over time.
We developed two variables to separately test
Contextual and control variables for the effects of each of the two events in order
to test for the contingency effect of environmen-
Environmental Change tal change. In line with our argument that these
Consistent with our theoretical arguments, we drew events change competitive dynamics, we exam-
on an events-based classification scheme to oper- ined the effect of network position on firm per-
ationalize environmental change. Industry events formance in the period immediately following the
change the competitive dynamics, which forces events. For the legislative event, we created a
firms to jockey for position in the network in order dummy variable (SRYEAR) that took the value
to acquire capabilities necessary to compete. Dur- of 1 for the years 1984, 1985, and 1986 and 0
ing the eighties the steel industry was buffeted otherwise. Extending the effect of the legislative
by two key industry events. First were a series acts for a period of three years was a judgment
of legislative acts in 1984, such as the Voluntary based on prior studies (Madhavan et al., 1998).
Restraint Agreements (VRAs) between the United For the technology event, we created a dummy
States and the steel industries of the European variable (SLYEAR) that took the value of 1 for
Union and Japan, and the relaxation of antitrust the years 1987, 1988, and 1989 and 0 otherwise.
laws, which enabled mergers between domestic Extending the effect of this event for a period
steel companies such as the LTV-Republic steel of three years was again based on prior research,
merger and set the stage for new competitive industry knowledge, and the desire for consistency
actions in the industry (Madhavan et al., 1998). A with the SRYEAR variable. We developed four
second key event was the technology breakthrough interaction variablesPROM8486, PROM8789,
spearheaded by Nucor in 1987 (Madhavan et al., ENTR8486, and ENTR8789as a product of
1998; Ghemawat, 1993). This event introduced these two variables, and two network position
the Compact Strip Casting (CSC) technology that variablesPROMIN and ENTREPto test for
enabled Nucor and other minimills to enter the flat the interaction effect of environmental change
steel segment of the market. and network position on firm performance. Thus,
Both of these events spurred several strategic Hypotheses 1 and 2 were tested twice; once with
actions by incumbent firms as firms jockeyed to the legislative event and a second time with the
attain favorable positions. The VRAs forced for- technology event. Testing our hypotheses over two
eign firms to form alliances with U.S. firms so they change events not only produced a more robust
could supply the United States-based Japanese auto test of our hypotheses but also helped us explore
transplants. The relaxation of the antitrust laws the question of whether the benefits of network
enabled U.S. firms to form technology alliances positions are stable over time.
with firms from Europe and Asia in order to access Since the dummy variables created above were
continuous casting technology. It also inspired a judgment calls based on our understanding of the
series of consolidations and restructuring both in industry and prior research, we conducted addi-
the United States and Europe. Such dynamics were tional validity checks. We tested for the interaction
effect of each of the years in the data and the net-
8
We adjusted for currency differences. We converted all the work position measure and examined their effect
figures into U.S. dollars using the exchange rate between the on firm performance. The results are consistent
U.S. dollar and the foreign currency prevailing as of 30 June
each year. We also adjusted for inflation by developing constant with the other operationalization, which we discuss
price measures with the year 1991 as a base. in the results section.
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
652 B. R. Koka and J. E. Prescott

Firm strategy did this analysis for two years (1986 and 1991)
to examine both the stability of memberships in
We operationalized an analyzer strategy in two dif- the clusters as well as the reliability of the clus-
ferent waysfirst, as a categorizing variable and ters. Memberships within the clusters as well as
second as a continuous measure. We ran different the means remained largely similar for both years,
analyses using the two different measures to test providing more validity for the measures.
our hypotheses in order to test for the validity and Utilizing a dichotomous approach to testing the
robustness of our results. hypotheses has two issues. First, a dichotomous
approach makes the implicit assumption that all
Categorizing variable as an indicator for strategy firms are similar in terms of how they are pursu-
ing a particular strategy. Differences between firms
Prior research studies have typically operational- in terms of their distance from an ideal type are
ized the Miles and Snow typology by cluster masked by taking a dichotomous approach. This is
analyzing firms on a number of organizational an important issue given that our theoretical argu-
attributes. We adopted a similar approach by clus- ments are based on degrees rather than an absolute
ter analyzing firms on six organizational measures. level. Second, utilizing a time invariant variable to
Two measuresproportion of sales in the steel characterize a firms strategy leads to collinearity
industry and proportion of international sales problems with the use of the firm as a dummy
measured the breadth of an organizations oper- variable (to control for unobserved heterogeneity),
ations (Doty et al., 1993). We also developed two which is another time invariant variable resulting
measures that addressed diversification: the num- in two constants in the panel regression methodol-
ber of product segments in the steel industry that ogy. We accordingly developed a continuous mea-
the firm operated in and the number of two-digit sure for testing the hypotheses in this study that is
SIC codes that it operated in. Since efficiency a subset of the above measures, and demonstrated
is another attribute that distinguishes between the that the two operationalizations are highly related.
strategies, we classified firms in terms of whether
they used continuous casting technology. Finally,
we used the size of the firms as an indicator Continuous strategy measure (STRG)
of administrative complexity (Doty et al., 1993). We had argued that analyzers would have a mix
Given the conservative nature of the industry, as of business with large exposure to one product
discussed earlier, we accordingly tested for a two- market domain but with significant presence in
cluster solution. The means for the first cluster for other domains. That is, an analyzers focus is on
proportion of steel sales was 0.76 as opposed to a broader product market scope. We thus opera-
0.93; for the second cluster, proportion of inter- tionalized firm strategy in this context as a com-
national sales was 0.30 as opposed to 0.17. Firms posite measure of two ratios. The first measures
in the first cluster had adopted continuous cast- the proportion of sales in the steel industry to the
ing technology, they also operated in three steel total sales for each firm. Firms pursuing an ana-
segments as opposed to one steel segment for the lyzer strategy were likely to service markets other
second cluster. Firms in the first cluster operated than steel. As indicated earlier, most steel firms
in three industries (at the two-digit SIC level), seemed to sell products that were related to the
while those in the second were limited to one steel industry. Ranging from 0.28 to 1, the mean
industry. Finally, the firms in the first cluster were of this measure for the dataset was 0.85. The sec-
larger with an average employee size of approxi- ond measure is the proportion of domestic sales to
mately 20,000 as opposed to 2,400 for firms in the total sales for each firm. Much of the sales in this
second cluster. Thus, there is considerable valid- industry seem to be directed to the domestic mar-
ity to argue that firms in the first cluster were ket. Though this measure ranged from 0.0648 to 1,
following an analyzer strategy given their broad the mean of this measure was 0.78. We reversed
product market scope, some diversification, focus the two measures and then multiplied the two ratios
on efficiency, and larger size. The second clus- to arrive at a composite measure of firm strategy
ter seemed to comprise firms following a defender (STRG) that ranged between 0 and 0.67. The mean
strategy given their focus on narrow product mar- for this measure was 0.05. Firms with scores closer
ket scope and relatively little diversification. We to 1 indicate that they are pursuing an analyzer
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
Designing Alliance Networks 653

strategy. It is pertinent to point out that a score natural logarithm of the total number of employees
around the mean of 0.05 indicates a firm is fairly (LNEMP).
well diversified because it means that on the aver- All the above measures were calculated for each
age, at least one-fourth of the total sales of the firm firm for every year they were in the network to
is directed to markets outside of the steel industry enable the longitudinal testing of our hypotheses.
and/or the domestic market. We developed a com-
posite measure using factor analysis of the two
Analyses
different measures as the analysis indicated that
there was only one factor. The correlation of that To test the hypotheses, we used a two-way fixed
factor score with STRG was 0.96. We conducted effects panel regression model. A fixed effects
the analysis for each of the measures separately. panel model enabled us to test the validity of
The results were stronger and consistent for the our theory over time. It also addresses the issue
proportion of sales in the steel industry measure of serially correlated errors that could otherwise
than for the geographic measure. affect the results validity. In addition, fixed effects
We compared this measure with the category controls for unobserved heterogeneity within firms
variable. First, the nonparametric correlation was by introducing a dummy variable for each firm
0.55 (p < 0.001), indicating that the two measures in the regression equation (Greene, 1993). The
were largely similar. Second, we calculated the dif- dummy variable controls for all other unobserved
ference in means between the two clusters on the and unknown time-invariant and firm-specific fac-
continuous strategy measure (STRG). The mean tors. It is quite likely that using the firm as a
for STRG for cluster one was 0.08 and 0.001 control variable may lead to inconclusive results
for cluster two (difference of means significant for the main hypotheses of this study because it
at p < 0.001), supporting our argument that firms may capture all of the variance in performance. In
that were higher on STRG were following an ana- that sense, using a fixed effects model may be a
lyzer strategy. Third, we tested for the strategy conservative test of the hypotheses. To the extent
hypotheses using both the continuous and categor- that the hypothesized variables are significant, we
ical measure and found similar results (which we argue that it provides strong support for the validity
report in the results section). All these reasons give of the results as well as the theoretical logic under-
us confidence in terms of the reliability and validity lying it. In addition to controlling for firm effects,
of the measures that we used for operationalizing we also controlled for time effects by introduc-
firm strategy. ing each year as a dummy variable. This variable
We report the results based on STRG. We cal- captures any variation in performance that is time-
culated each measure using data from World- specific that again provides for a conservative test
scope. We used information from Annual Reports of the hypotheses. We conducted the fixed effects
if the information was not available from World- test using the TSCSREG procedure in SAS.
scope. We again developed two interaction vari- We controlled for industry effects by restricting
ablesPROMSTRG and ENTRSTRGas a pro- the analysis to only firms from the steel indus-
duct of the strategy variable STRG and the position try. From the 162 steel firms in the dataset, we
variables PROMIN and ENTREP. conducted the analysis on a sample comprising 70
steel firms with 585 observations over 11 years
Interaction of network prominence and the (19831993). Even though we collected alliance
entrepreneurial position data from 19801994, we conducted the analysis
only for the 11-year period 19831993 in order
We calculated this variable (PROMENTR) as a to address left-censoring issues. The choice of the
product of the two position variables PROMIN and 70 steel firms was based on three reasons. First, of
ENTREP. the 162 steel firms in the dataset, only 153 were
actively engaged in alliances between 19831993;
data availability issues, particularly for firms oper-
Firm size
ating outside the Triad (Europe, North America,
We included firm size as a control measure to and Japan), further restricted the number of firms
test the alternative hypotheses that size affects we could include in the sample; and the final
productivity. Firm size was operationalized as the choice of 70 firms included the top 50 firms in the
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
654 B. R. Koka and J. E. Prescott

steel industry from the chosen countries in terms effects of both position measures on firm perfor-
of sales turnover. Second, the choice of these 70 mance. On the other hand, the estimates for the
firms was also consistent with the current method- interaction variables are more consistent in signif-
ology in network studies that restrict their analysis icance, magnitude, and direction across the differ-
to a selected group of firms (Gulati, 1995; Ahuja, ent models, providing face validity for our argu-
2000a). Third, competitive dynamics among firms ment. More importantly, the incremental change
from the Triad were more interdependent than with in explained variance is significant for each of the
firms from the other countries (Mangum et al., models, providing support for the importance of
1996), indicating that much of the alliance activity the interaction variables in explaining firm perfor-
was concentrated among the firms in the sample. mance.
To ensure that there was no bias in the validity Hypothesis 1 predicted that following an envi-
of the results, we conducted random effects model ronmental change, network prominence would be
analysis to test for the assumption that the 70 firms negatively related to performance, while Hypoth-
were a random sample of the population. We found esis 2 predicted that firms pursuing an entrepre-
results that were consistent with the fixed effects neurial position would exhibit higher performance.
model reported below, reinforcing our confidence As noted in our operationalization section, we
in the robustness of the results. examined the interacting effects of two types of
environmental changethe legislative change of
1984 and the technology change of 1987on firm
RESULTS performance to test these hypotheses. We discuss
each separately.
Table 2 provides the panel regression results. To
ensure clarity, we dropped the 69 firm dummy
Legislative change
variables and the 10 year dummy variables and
show only the parameter estimates, standard errors, We find support for both Hypotheses 1 and 2 for
and significance of the coefficients of the main and the interacting effect of environmental change trig-
interaction variables. Model 1 tested for the main gered by legislative events on firm performance
effects of the two network position measures, the (Model 2 and Model 5). The interaction variable
contingency variables and size as a control vari- (PROM8486) of network prominence (PROMIN)
able. Models 25 included the interaction effects and the period dummy variable (SR8486) is neg-
separately for the different contingency hypothe- ative and significant. Similarly, the interaction
ses. Model 6 examined the effect of the complete variable (ENTR8486) of entrepreneurial position
model with all the interaction effects in one model. (ENTREP) and SR8486 is significant and posi-
Model 7 gives the results for the complete model tive. Thus, following an environmental change,
and is similar to Model 6 except that the strategy entrepreneurial position is positively related to firm
variable is a dichotomous variable based on the performance. Both these relationships continue to
cluster analysis described in the operationalization be supported in the full model (Model 6), provid-
section. Model 8 in Table 3 tested for the com- ing strong support for the contingency effect of the
plete model similar to Model 6, but replaced the environmental context on firm performance.
dummy environmental change variables with the
interaction variables for each year and the position
Technology change
measures.
Since our interest is on the interactive effects The interacting variable for network prominence
of the position and contingency variables, we (PROM8789) following a technology event is in
focus our discussion on them. The main effects the expected direction, but not significant in Model
depicted in Model 1 for comparison purposes 3 (Hypothesis 1). However when we examine
provide preliminary support for the validity of the full Model 6, Hypothesis 1 is strongly sup-
a contingency perspective for the effect of net- ported. Following the technology change, firms
work position on firm performance. As can be that were prominent in the network experienced
seen across the different models in Table 2, there adverse effects on firm performance. However, the
is little consistency in terms of the magnitude interacting variable for the entrepreneurial position
of the estimates and the direction for the main (ENTR8789) is significant and negative (Model 3).
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
Designing Alliance Networks 655
Table 2. Panel analysis results for productivity as the dependent variable

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7

Main Hypotheses Hypotheses Hypotheses Hypothesis Full Full Model


Effects 1&2- 1&2 3&4 5 Interaction Model (Strategy as
Legislative Technology Strategy of the two dichotomous
Change Change positions variable)@

Intercept 0.904 0.858 0.892


0.76 0.8 0.899 0.866
(0.11) (0.105) (0.108)
(0.102) (0.106) (0.107) (0.10)
Firm size (LNEMP) 0.06 0.053 0.059
0.042 0.048 0.059 0.063
(0.014) (0.014) (0.014)
(0.013) (0.014) (0.014) (0.012)
Strategy (STRG) 0.613 0.403 0.560.61 0.99 0.496 0.034
(0.2) (0.19) (0.2) (0.2) (0.204) (0.2) (0.05)
Network prominence 0.053 0.04 0.058
0.013 0.006 0.068 0.023
(PROMIN) (0.011) (0.01) (0.01)(0.015) (0.014) (0.014) (0.029)
Entrepreneurial 0.01 0.002 0.0170.0008 0.007 0.003 0.002
position (ENTREP) (0.008) (0.007) (0.008)
(0.008) (0.008) (0.008) (0.01)
PROMIN YR8486 0.053 0.05 0.053
(PROM8486) (0.008)
(0.009) (0.006)
ENTREP YR8486 0.024 0.019 0.017
(ENTR8486) (0.007)
(0.007) (0.007)
PROMIN YR8789 0.008 0.014 0.023
(PROM8789) (0.006) (0.007) (0.006)
ENTREP YR8789 0.021 0.0123+ 0.017
ENTR8789) (0.007) (0.0068) (0.007)
PROMIN Strg 0.98 0.68 0.06
(PROMSTRG) (0.15) (0.16) (0.03)
ENTREP Strg 0.111 0.058 0.006
(ENTRSTRG) (0.054) (0.054) (0.013)
PROMIN ENTREP 0.055 0.027 0.044
(PROMENTR) (0.014) (0.014) (0.013)
YEAR EFFECTS Included Included Included Included Included Included
FIRM EFFECTS Included Included Included Included Included Included
R-SQUARE 0.905 0.915 0.908 0.913 0.908 0.923 0.248
F-value 56.7 62.33 57.15 60.77 57.4 64.88 69.1
Deg. Of freedom 501 499 499 499 500 494 573
"R-sq (from Model 1) 0.01 0.003 0.008 0.003 0.018
F-value for "R sq. 29.29 8.12 22.9 16.27 17.94

- p < 0.0001;
- p < 0.001;
- p < 0.01; - p < 0.05; +
- p < 0.1; - Random effects : Analyzer strategy = 1

This was unexpected because we had predicted we conducted an alternative analysis that examined
that such a position would positively affect firm the interacting effect of each year in the dataset
performance following an environmental change. with the two position measures. The results are
However, the results suggest that following a tech- depicted in Table 3 (Model 8). As can be seen,
nology event, firms that were entrepreneurially the interaction variable for each year and net-
positioned also experienced adverse effects on per- work prominence position is consistently signifi-
formance. In other words, following the technol- cant and negative throughout the analysis period.
ogy change, both network positions had an adverse This was consistent with our expectations and our
effect on firm performance. hypothesis that in environments in flux, firms that
We conducted additional analysis to understand had a prominent position in the network were
why the results differed between the two events. at a disadvantage regardless if the event was
We considered whether the difference was due to legislative- or technology-based. That the steel
the way we had operationalized our dummy vari- industry was experiencing massive restructuring
able for the two events that assumed a three-year during this period has been the subject of many
window. Instead of using the window approach, studies (Madhavan et al., 1998; Mangum et al.,
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
656 B. R. Koka and J. E. Prescott
Table 3. Panel analysis results for productivity as the deal with changing competitive dynamics. In other
dependent variable: full model with year-wise interacting words, such events were changing the bases of
variables
competition, which directly affected the promi-
Model 8 nent firms competitive position. Their existing
network, however, was unlikely to provide them
Estimates S.E with the information necessary to compete in the
new environment resulting in adverse effects on
Intercept 0.748 0.096
Firm size (LNEMP) 0.038 0.0125 their performance.
Strategy (STRG) 0.34 0.19 The results for the interacting effect of entrepre-
Network prominence (PROMIN) 0.033 0.015 neurial position with the individual years are
Entrepreneurial position 0.01 0.011 mixed. As predicted, the interacting variable was
(ENTREP) positive and significant in the first three years of
PROMIN 1983 0.112 0.015
PROMIN 1984 0.112 0.015 the analysis period, which was comparable to the
PROMIN 1985 0.089 0.014 period following the legislative change. Firms that
PROMIN 1986 0.08 0.014 were pursuing an entrepreneurial position in the
PROMIN 1987 0.05 0.013 network seemed to have positioned themselves to
PROMIN 1988 0.04 0.01 acquire the new information necessary to compete
PROMIN 1989 0.05 0.011
PROMIN 1990 0.023 0.011 in the environment following the legislative events
PROMIN 1991 0.028 0.01 of 1984. Such firms exhibited higher performance
PROMIN 1992 0.018 0.0095 in the years 1984 and 1985 immediately follow-
ENTREP 1983 0.049 0.014 ing the change. Following the technology change
ENTREP 1984 0.048 0.013 of 1987, however, an entrepreneurial position was
ENTREP 1985 0.034 0.013
ENTREP 1986 0.013 0.012 not significant in explaining firm performance. In
ENTREP 1987 0.007 0.012 sum, while our hypotheses were strongly supported
ENTREP 1988 0.009 0.012 for the legislative change, there was mixed support
ENTREP 1989 0.009 0.012 for the technology change.
ENTREP 1990 0.003 0.012 We looked to the literature to understand why
ENTREP 1991 0.005 0.013
ENTREP 1992 0.004 0.013 our predicted effect for an entrepreneurial position
PROMIN STRG (PROMSTRG) 0.38 0.15 differed for the two events. The answer seems to
ENTREP STRG (ENTRSTRG) 0.01 0.05 lie in the nature of the eventsin other words
PROMIN ENTREP 0.026 0.013 not all environmental changes are the same. The
(PROMENTR) legislative change of 1984 provided the impe-
YEAR EFFECTS Included
FIRM EFFECTS Included tus for a series of actions that emphasized the
R-SQUARE 0.937 benefits arising from an entrepreneurial position.
F-value 65.25 Restricting imports into the United States forced
Deg. Of freedom 478 Japanese and European firms to seek markets and
alliances in other parts of the world. It thus ben-

- p < 0.0001;
- p < 0.001;
- p < 0.01; - p < 0.05; efited those firms that had formed alliances with
+ - p < 0.1
firms in markets such as China, Australia, South
Asia, and South America. In addition, the legisla-
1996; Ghemawat, 1993). For instance, the top six tive change also spurred the formation of several
U.S. steel companies experienced reductions in the alliances between foreign and domestic firms in
workforce from approximately 100,000 employees the United States, such as those between U.S.
in the early eighties to about 20,000 employees Steel and Kobe, and Armco and Acerinox. On the
in the early nineties. This was part of a series of one hand, the legislative change helped U.S. firms
restructuring that forced them to close obsolete acquire continuous casting technology, as well
plants, acquire new technologies, and compete as the process information necessary to increase
in more value-added markets. In the European productivity and efficiency. It also helped them
Union, this period saw similar consolidation activ- enter markets for more value-added products such
ity. The two events that we hypothesized thus were as specialty steels. To the extent that knowledge
symptomatic of the changing environment that cre- and information regarding continuous casting tech-
ated pressures for new information necessary to nology, opportunities regarding new products and
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
Designing Alliance Networks 657

markets, and changes in business practices existed firms pursuing an analyzer strategy, firms that were
in the network, firms that were positioned to bene- more prominent would exhibit higher performance
fit from diversity and unconnected networks exhib- (Hypothesis 3). The hypothesis was supported in
ited increased performance. Our results support both Model 4 and Model 6 (the full model), as
that conclusion. can be seen from the significance of the interacting
However, the technology change was different variable PROMSTRG. Firms that pursue an ana-
in that it was initiated by two firms, Nucor (a lyzer strategy and that are more prominent in the
minimill) and SMS Mannesman (a machinery sup- network exhibit higher performance. This is again
plier), both of which had relatively little presence consistent with prior research. As Miles and Snow
in the erstwhile network. The new Compact Strip (1978) pointed out, analyzers focus their efforts on
Casting technology (CSP) signaled the entry of collecting information from a large and broad base.
minimills into the flat steel segment (Ghemawat, They create administrative structures that enable
1993: Madhavan et al., 1998). With a potential them to access information that reduce costs and
to dramatically increase productivity to about one increase efficiencies. We also tested our Hypoth-
man-hour per ton in comparison to the productiv- esis 4 that proposed that for firms pursuing an
ity of five man hours for the integrated steel plants, analyzer strategy, entrepreneurial positioning pos-
such a technology event placed the integrated mills itively affects performance. We found weak sup-
at a significant competitive disadvantage. There port for this hypothesis (Model 4) which becomes
is little doubt that the technology change engen- insignificant in the full model (Model 6).
dered significant uncertainty for firms in the indus- As indicated earlier, we also conducted the same
try. Adopting the new technology, however, was analysis using a dichotomous variable for firm
not really an option for the firms since commit- strategy as shown in Model 7. We found support
ments to existing technological bases, legacy costs, for Hypotheses 3, indicating that firms pursuing an
and erstwhile labor practices prevented the inte- analyzer strategy exhibited higher performance if
grated mills from embracing the new technology. they were more prominently positioned in the net-
By engendering an uncertainty that enveloped both work. In line with the earlier analysis, we found no
assessment and potential actions, the technology support for Hypothesis 4 that entrepreneurial posi-
change made the strategic options for the dominant tioning leads to higher performance for analyzer
firms in the industry less clear. This would explain strategies. We found identical results when we ran
why an entrepreneurial position was not signifi- separate analyses for the two clusters. Again firms
cant in affecting firm performance. Implicit in the pursuing an analyzer strategy exhibited higher per-
arguments underlying the entrepreneurial position formance if they were more prominently posi-
is the idea that all the information necessary to tioned.9
compete in the industry is available somewhere There may be three reasons for the weaker sup-
in the network, and thus a network position that port for Hypothesis 4. First, it may be possible that
best accesses such information through nonredun- the moderate diversity that analyzers need could be
dant and diverse ties is likely to be effective. Our achieved through their relations with other analyz-
results suggest that there is limit to this argument. ers in the industry. Since most of the analyzers
More importantly, it suggests that there could be were integrated steel mills, forming multiple rela-
limits to the information available in the network. tionships with other integrated steel mills increases
In other words, when events occur that are radi- a firms prominent position while simultaneously
cal in nature and that are initiated by outsiders in providing some diversity benefits because the mar-
the industry, either network position may have lit- ket offerings of these mills were similar. Sec-
tle effect on firm performance. More likely, they ond, it may be argued that the hypothesis may
could affect the firms negatively because they are be better framed for prospectors than for ana-
locked into the older technology base. lyzersentrepreneurial positioning may benefit
prospector firms more than analyzers. However,
examining our measures leads us to the conclu-
Firm strategy sion that the context of the steel industry prevented
We proposed that firm strategy would interact with us from testing that hypothesis. In the absence of
the network positions to affect firm performance
(Hypotheses 3 and 4). We hypothesized that for 9
Results are available from the first author.

Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
658 B. R. Koka and J. E. Prescott

many conglomerates and innovators in this dataset, that they are likely to impact firm performance dif-
we leave this hypothesis for future research. Third, ferently under different contexts. We thus extend
it may suggest that even for analyzers that put conventional theory by asking: under what con-
emphasis on both positions, higher performance text are different positions effective? We address
may not result. This indicates the difficulty of this pivotal question by developing contingency
managing two distinct positions that require dif- hypotheses at two levels of analysis: environ-
ferent capabilities. In this regard it reinforces the mental change and firm strategy. In addition, we
results of Hypothesis 5 discussed below. Strong examined the interaction of these positions on
and consistent results for both measures of strat- firm performance. Our results provide support for
egy enhances our confidence in the results. Not our hypotheses that different types of network
only are the specific strategy hypotheses supported positions interact with environmental change and
in different analyses, we find that the effects of strategic contingencies to impact productivity.
other variables in the model are consistent both in Our contingency findings support the call for a
terms of their direction and significance. This indi- strategic approach to the design and management
cates that both the strategy variables have similar of strategic alliances. Since different information
effects raising our confidence in the robustness and benefits are derived from a firms set of alliances,
validity of the results. it is imperative that alliance managers strategi-
cally evaluate their information benefits and limi-
tations. That is, managers have multiple alliance
The interaction of the two network positions
design choices. The strategic design of alliance
Finally, there is strong support for Hypothesis 5 networks focuses managers attention on asking
that proposes that the interaction of the two net- questions such as: (1) what types of benefits do
work positions will adversely affect firm perfor- we want from our alliance network; and (2) are
mance. The interacting variable PROMENTR in we interested in developing networks that allow
both the individual Model 5 and the full Model 6 us to transfer tacit knowledge at the expense of
is negative and highly significant, providing strong diversity, or are we more interested in developing
support for our argument that firms that are high a set of nonredundant relationships that provide
on both of these positions are likely to experience us with diverse information but at a cost to effi-
adverse effects on firm performance. The results ciency and exploitation? These types of trade-offs
suggest that firms have to make trade-offs in terms are important considerations and, as our results
of what position they want to achieve. If being high illustrate, have performance implications. Firms
on both positions results in high performance, the are adversely affected if they try to extensively
prescription for managers is to form many ties with do both, resulting in the need to make a conscious
many diverse partners that will then provide them decision regarding the nature and type of network
with the benefits of each position. Our results sug- position they should pursue.
gest precisely the oppositethat the fit between As we have illustrated, not only do differ-
strategic context and network position is essential ent types of network positions affect performance
for higher productivity. Managers need to be cog- differentially, these network positions adversely
nizant of the strategic context in order to design the affect performance if they do not fit contingency
most appropriate network position to enhance per- requirements. Sometimes the dysfunctional effects
formance. Overall the results provide strong sup- of network positions are more likely than the
port to a contingency theory of network structure positive effects. In other words, possessing the
on firm performance. inappropriate network position may have more
severe consequences than having the right posi-
tion. For example, we found that defenders that
CONTRIBUTIONS AND IMPLICATIONS are highly prominent in the steel industry alliance
network have lower relative performance. One lim-
The network literature identifies and delineates that itation of this study is the lack of prospectors
firms pursue two different types of network posi- in our dataset. We cannot draw any conclusions
tionsnetwork prominence and entrepreneurial regarding the functional or dysfunctional perfor-
positions. We argue that these two network posi- mance consequences of a prospector strategy with
tions provide different benefits thereby implying an entrepreneurial position. It is worthwhile to
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
Designing Alliance Networks 659

extend our findings in an industry setting where the RISC standard were those that had no/little
there are prospectors. Another fruitful area of connections with the previous CISC network base.
future research would be the examination of the This is similar to the position of Nucor in the steel
performance implications of adding, eliminating, industry network, which had just one joint venture
strengthening, and weakening a firms network with another minimillYamato Kogyoprior to
of alliances. Alliance researchers can look to the the introduction of CSP technology. However, that
stream of research in the valuation literature that is not to suggest that all firms that are not well
evaluates the impact of restructuring on market connected in the current network would succeed
value for guidance. following a competence-destroying event. In other
Our findings have implications for theory build- words, while being entrenched in the current net-
ing. Following an environment-changing event, work has adverse implications for firm perfor-
there are limitations to the information benefits that mance, the mere fact of being a peripheral player
are available from the network. Even though the does not guarantee success in the changed envi-
integrated firms, in particular Japanese integrated ronment.
mills, were well positioned on both types of net- More importantly, radical changes also indicate
work position, their alliances reflected their current the potential for radical restructuring of the net-
technological base. These alliances included joint work as new firms enter, existing firms exit, and
ventures in coal and iron ore mines, joint devel- the basis of competitive advantage shifts. In the
opment programs with machinery companies, and steel industry, the high levels of chronic overca-
technical assistance agreements that were geared pacity have led to significant downward pressures
to enhance their blast furnace and continuous cast- on prices all through the nineties. As a result
ing technological base. In the wake of a radical of this and other industry dynamics, there was
change in technology, the ability of these firms to a considerable level of bankruptcy, mergers, and
compete in the new environment was severely con- acquisitions during this period. Significant indus-
strained. The new CSC technology introduced by try restructuring of this type will certainly modify
Nucor was based on the use of iron carbides and the industry alliance structure as well as individual
steel scraps as raw materials and utilized electric firm alliance networks. While this type of restruc-
arc furnace technology. Well-connected firms were turing presents challenges to network analytical
therefore unlikely to exhibit higher relative per- techniques, research examining the evolution of
formance following such a competence-destroying networks would provide valuable theoretical and
event. managerial insights.
Thus, the results suggest that the type of change The theoretical development of the two types
matters. Changes that reinforce the importance of network positions raises the possibility that
of information, knowledge, and technology that there could be contradictory forces operating on
already exist in the network, such as our leg- the firms ability to dynamically design its net-
islative change, strongly support our hypothe- works to meet particular contexts. On the one
ses. Changes of a more radical nature, such as hand, entrepreneurial positions focus on the ability
our technology change, provide only mixed sup- of the firm to strategically form and dissolve ties
port for our hypotheses suggesting that networks as it maneuvers within the network. On the other
benefits may be limited in situations where all hand, network prominence suggests the need to
the information necessary to compete may not develop, manage, and extend the relationships over
reside in the network. Evidence for the inher- considerable periods of time in order to achieve
ent limitations to the novelty of information that the benefits arising out of the accumulated invest-
can potentially be available in the network can ments in the relationship-building process. This
be garnered from Afuahs (2000) study in the implies that the longer the firms relationships sur-
microprocessor industry. He found that successful vive, the more likely it is that the firm could get
microprocessor firms operating within the Com- locked into its current network of relationships.
plex Instruction Set Computer (CISC) technologi- Alternatively, the more the firm focuses on devel-
cal base and well-embedded in the CISC network oping its entrepreneurial position, the less likely it
were less successful in their transition to the new is to invest in its existing relationships resulting in
Reduced Instruction Set Computer (RISC) tech- little coordinated action. Our results indicate that
nology. More significantly, the successful firms in firms need to manage these trade-offs over time if
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
660 B. R. Koka and J. E. Prescott

they are to design alliance networks that fit their linkages. Strategic Management Journal , March Spe-
particular contexts. How firms manage these con- cial Issue 21: 317343.
tradictory influences has serious implications for Barley S. 1986. Technology as an occasion for
structuring: evidence from observations of CT
the quality and quantum of benefits available and scanners and the social order of radiology departments.
warrants research attention. Administrative Science Quarterly 31(1): 78108.
Finally, the results indicate further avenues for Bonacich P. 1972. Factoring and weighting approaches
research. While the context of this study has been to social status and clique identification. Journal of
the steel industry, the basic findings of the con- Mathematical Sociology 2(2): 113130.
tingent effect of the alliance network would seem Borgatti SP, Everett MG, Freeman LC. 1999. Ucinet 5
for Windows: Software for Social Network Analysis.
to be generalizable to other industries and con- Analytic Technologies: Natick, MA.
texts. While the contingencies examined may dif- Burt RS. 1992. Structural Holes: The Social Structure of
fer, understanding their nature and impact may Competition. Harvard University Press: Cambridge,
explain firm performance vis-`a-vis their alliance MA.
relationships in those industries. Other implica- Burt RS. 2000. The network structure of social capital.
Research in Organizational Behavior 22: 345423.
tions include the construction of scores for each
Coleman J. 1988. Social capital in the creation of
of the two types of network positions for every human capital. American Journal of Sociology 94(1):
firm in the network in order to examine the differ- S95S120.
ent alliance strategies pursued by firms, as well as Cyert RM, March JG. 1963. A Behavioral Theory of the
the development of a taxonomy of alliance strate- Firm. Prentice Hall: Englewood Cliffs, NJ.
gies. The potential for examining a wide range of Doty DH, Glick W, Huber G. 1993. Fit, equifinality,
and organizational effectiveness: a test of two
questions, such as the evolution of such strate- configurational theories. Academy of Management
gies over time through simulation techniques, is Journal 36(6): 11961250.
intriguing. While these and other ideas are beyond Ghemawat P. 1993. Commitment to a process innovation:
the scope of this study, it does clearly illustrate Nucor, USX and thin slab casting. Journal of
the far-reaching potential of our approach. Explor- Economics and Management Strategy 2(2): 135161.
ing these and other questions clearly holds exciting Greene W. 1993. Econometric Analysis. MacMillan: New
York.
possibilities for managers and researchers. Gulati R. 1995. Social structure and alliance formation
patterns: a longitudinal analysis. Administrative
Science Quarterly 40(4): 619652.
AKNOWLEDGEMENTS Hagg I, Johanson J, Forsgren M, Hakansson H,
Mattsson L-G. 1982. Firms in Networks: A New View
We gratefully acknowledge the comments and of Competitive Power. SNS: Stockholm, Sweden.
Hambrick DC. 1983. Some tests of the effectiveness and
feedback from Bob Atkin, John Camillus, Bill
functional attributes of Miles and Snows strategy
Glick, David Krackhardt, Ravi Madhavan, and Sue types. Academy of Management Journal 26(1): 526.
McEvily. This study was partially supported by Hogan WT. 1991. Global Steel in the 1990s: Growth or
the Deans Award for Excellence Summer Grant Decline. Lexington Books: Lexington, MA.
Program funded by the Deans Council of 100, Koka B, Prescott J. 2002. Strategic alliances as social
the Economic Club of Phoenix, the Alumni of the capital: a multidimensional view. Strategic Manage-
ment Journal 23(9): 795816.
W.P. Carey School of Business, and the Depart- Larson A. 1992. Network dyads in entrepreneurial
ment of Management Summer Grants Program at settings: a study of the governance of exchange
the Arizona State University. processes. Administrative Science Quarterly 37(1):
76104.
Lawrence P, Lorsch J. 1967. Organizations and Environ-
REFERENCES ments. Harvard University Press: Cambridge, MA.
Levinthal DA, March JG. 1993. The myopia of learning.
Afuah A. 2000. How much do your co-opetitors Strategic Management Journal , Winter Special Issue
capabilities matter in the face of technological change? 14: 95112.
Strategic Management Journal, March Special Issue Locke D. 1999. Some reservations about social capital.
21: 387404. Academy of Management Review 24(1): 89.
Ahuja G. 2000a. Collaboration networks, structural holes, Madhavan R. 1996. Strategic flexibility in the steel
and innovation: a longitudinal study. Administrative industry: the role of interfirm linkages. PhD diss.,
Science Quarterly 45(3): 425455. University of Pittsburgh.
Ahuja G. 2000b. The duality of collaboration: induce- Madhavan R, Koka B, Prescott J. 1998. Networks in
ments and opportunities in the formation of interfirm transition: how industry events (re)shape interfirm
Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj
Designing Alliance Networks 661

relationships. Strategic Management Journal 19(5): Powell WW, Koput KW, Smith-Doerr L. 1996. Interor-
439459. ganizational collaboration and the locus of innovation:
Mangum GL, Kim S, Tallman S. 1996. Transnational networks of learning in biotechnology. Administrative
Marriages in the Steel Industry: Experience and Science Quarterly 41(1): 116145.
Lessons for Global Business. Quorum: Westport, CT. Rowley T, Behrens D, Krackhardt D. 2000. Redundant
March JG. 1999. The Pursuit of Organizational Intelli- governance structures: an analysis of structural and
gence. Blackwell: Oxford, UK. relational embeddedness in the steel and semiconduc-
Miles R, Snow C. 1978. Organizational Strategy, Struc- tor industries. Strategic Management Journal, March
ture, and Process. McGraw-Hill: New York. Special Issue 21: 369386.
Miller A. 1988. A taxonomy of technological settings, Snow CC, Hrebiniak LG. 1980. Strategy, distinctive
with related strategies and performance levels. competence, and organizational performance. Admin-
Strategic Management Journal , 9(3): 239254. istrative Science Quarterly 25(2): 317336.
Nair A, Kotha S. 2001. Does group membership matter? Uzzi B. 1996. The sources and consequences of embed-
Evidence from the Japanese steel industry. Strategic dedness for the economic performance of organi-
Management Journal 22(3): 221235. zations: the network effect. American Sociological
Nohria N, Garcia-Pont C. 1991. Global strategic linkages Review 61(4): 674698.
and industry structure. Strategic Management Journal , Uzzi B, Gillespie JJ. 2002. Knowledge spillover in
Summer Special Issue 12: 105124. corporate financing networks: embeddedness and
Podolny J. 2001. Networks as pipes and prisms of the firms debt performance. Strategic Management
the market. American Journal of Sociology 107(1): Journal 23(7): 595618.
3360. Von Hippel E. 1988. The Source of Innovation. Oxford
Podolny J, Stuart T, Hannan M. 1996. Networks, knowl- University Press: New York.
edge, and niches: competition in the worldwide semi- Zaheer A, Bell G. 2005. Benefiting from network
conductor industry, 19841991. American Journal of position: firm capabilities, structural holes, and
Sociology 102(3): 659689. performance. Strategic Management Journal 26(9):
Porter ME. 1980. Competitive Strategy. Free Press: 809825.
New York.

Copyright 2008 John Wiley & Sons, Ltd. Strat. Mgmt. J., 29: 639661 (2008)
DOI: 10.1002/smj

Vous aimerez peut-être aussi