Strategic Management Journal, Vol. 19, 439–459 (1998)
NETWORKS IN TRANSITION: HOW INDUSTRY
EVENTS (RE)SHAPE INTERFIRM RELATIONSHIPS
RAVINDRANATH MADHAVAN
1
, BALAJI R. KOKA
2
and JOHN E.
PRESCOTT
2
*
1
College of Commerce and Business Administration, University of Illinois at
Urbana–Champaign, Champaign, Illinois, U.S.A.,
2
Katz Graduate School of Busi-
ness, University of Pittsburgh, Pittsburgh, Pennsylvania, U.S.A.
Interfirm relationship networks are strategic resources that can potentially be shaped by
managerial action. As a first step towards understanding how managers can shape networks,
we develop a framework which explains how industry networks evolve over time and in response
to specific events. Our main thesis is that industry events may be either structure-reinforcing
or structure-loosening, and that their potential structural impact may be predicted in advance.
We validate our hypotheses with longitudinal data on the strategic alliance network in the
global steel industry.
1998 John Wiley & Sons, Ltd
Strat. Mgmt. J. Vol. 19, 439–459 (1998)
INTRODUCTION
Interfirm
relationships,
especially
strategic
alliances between industry players, represent sig-
nificant flows of knowledge and other resources
that are crucial to industry leadership. This insight
has been the basis for a robust stream of research
in organization theory and strategic management,
beginning with Pfeffer’s (1972) contributions and
continuing to the present day. Interfirm relation-
ships not only help to manage competitive uncer-
tainty and resource interdependency (Pfeffer and
Salancik, 1978), but also serve as conduits for
information and control benefits (Burt, 1992).
Since they are the channels by which goods and
services are accessed, interfirm relationships can
be considered to be resources in their own right
(Freeman and Barley, 1990). While these factors
have been relatively well-understood in the con-
text of dyadic interfirm relationships (i.e., when
relationships are taken one at a time), researchers
Key words:
interfirm relationship networks; industry
events; structural change
*Correspondence to: John E. Prescott, Katz Graduate School
of Business, University of Pittsburgh, 246 Mervis Hall, Pitts-
burgh, PA 15260, U.S.A.
CCC 0143–2095/98/050439–21$17.50
Received 16 January 1996; Revised 3 September 1996
1998 John Wiley & Sons, Ltd.
Final revision received 16 May 1997
in strategic management have recently begun to
acknowledge that multiple dyadic relationships
interconnect
and
bind
firms
into
networks
(Nohria, 1992). The strategic conduct of firms in
an industry is influenced not only by the proper-
ties of their relationships taken one at a time, but
also by the overall structure of interfirm relation-
ship networks (cf. Wellman, 1988). Network ana-
lysts such as Burt (1992) have employed this
logic to argue that well-structured networks are
the basis of superior returns in a variety of
settings, and form valuable ‘social capital.’ Thus,
interfirm relationship networks can be viewed as
strategic resources that play a significant role in
strategic performance. This reasoning is implicit
in recent research that employs the network per-
spective to explore questions of strategy (e.g.,
Nohria and Garcia-Pont, 1991; Shan, Walker, and
Kogut, 1994).
Despite the recent interest in the strategic
implications of interfirm networks, there remains
a need for better theories of ‘how networks
evolve and change over time’ (Nohria, 1992: 15).
The tradition in network analysis has been to
view networks as given contexts for action, rather
than
as
being
subject
to
deliberate
design
(Laumann, Galaskiewicz, and Marsden, 1978;
440
R. Madhavan, B. R. Koka and J. E. Prescott
Nohria, 1992), perhaps a necessary corollary of
the assumption that social structure endures over
time (Barnes and Harary, 1983; Schott, 1991).
However, current network research has begun to
direct attention at what underlies different net-
work structures (e.g., Cook, 1987; White, 1992).
In line with this research thrust, we take the
position that managerial action can potentially
shape networks so as to provide a favorable
context for future action. In order to understand
how managers may do this, research needs to
move beyond asking how networks constrain and
shape action, to examining what factors constrain
and shape networks. This paper seeks to address
the consequent need for an evolutionary perspec-
tive on interfirm networks. Specifically, we draw
upon the structural perspective (Wellman, 1988)
to theorize about how and why interfirm networks
change over time. We argue that interfirm net-
works evolve in response to key industry events
that may be either structure-reinforcing or struc-
ture-loosening. Further, based on the properties
of each event, it may be possible to identify
an event, in advance, as potentially structure-
reinforcing
or
potentially
structure-loosening.
Identifying the potential impact of an event in this
way alerts managers to the evolutionary nature of
their industry networks, and provides guidelines
for strategic maneuvering. We validate our frame-
work using longitudinal data on interfirm relation-
ships (strategic alliances initiated between 1977
and 1993) in the global steel industry.
NETWORKS IN TRANSITION
The structure of an industry network plays an
important role both in firm performance and in
industry evolution. Since external relationships
provide access to key resources, the structure of
relationship networks describes the asymmetric
access that rivals have to raw materials, infor-
mation, technology, markets, or other crucial per-
formance requirements. Consequently, network
structure provides the context for competitive
action, and structural properties such as network
centrality (e.g., Galaskiewicz, 1979) or autonomy
(Burt, 1992) have been proposed as being corre-
lates of strategic performance. Analogously, it
may be argued that network structure also influ-
ences industry evolution. By directing and limit-
ing access to key resources, the network structure
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
enhances or constrains the ability of specific firms
to shape industry developments. A firm that is
an isolate (i.e., disconnected from the industry’s
network) may find itself without significant input
into,
or
timely
information
about,
crucial
decisions affecting the industry’s future. For
example, Apple Computer Inc. is increasingly
finding itself in this position (Wall Street Journal,
12 December 1995).
Because network structure is a key influence
on performance and industry evolution, managers
engage in strategic maneuvering to secure key
positions in their industry’s network, such as
entering into strategic alliances in order to ensure
access to key technologies or other resources.
Thus, periods of strategic change in the history
of an industry are often marked by observable
flurries of interfirm activity. For example, given
the current state of flux and potential strategic
change in the telecommunications industry, a
number of interfirm relationships have been
announced over the last few years (e.g, Elec-
tronics Times, 17 June 1993; Wall Street Journal,
16 November and 29 November 1995). Similarly,
in the biotechnology industry 246 alliances were
announced in one year (Wall Street Journal, 21
November 1995). Though industry networks can
often be observed to be in a state of flux, the
current literature provides little guidance on what
types of network changes to expect consequent
to major industry events. As previously noted,
network analysts usually assume that social struc-
ture endures over time (Schott, 1991; Barnes
and Harary, 1983), although there are significant
exceptions (e.g., Doreian, 1986; Burt, 1988; Bur-
khardt and Brass, 1990). While this is often a
necessary assumption that facilitates investigation
of how structure influences action, turning it into
an
empirical
question
(i.e.,
does
structure
endure?) can lead to valuable insights. This is
especially true if, as we argue, interfirm networks
are strategic resources that managers design and
develop over time in order to meet their objec-
tives. Accordingly, we propose that the structure
of interfirm relationship networks changes in
potentially predictable ways over time and in
response to specific industry events. A dynamic
perspective on interfirm networks will demon-
strate that networks change over time as the
network participants take advantage of opportuni-
ties to improve their individual positions in the
network. A network at a given point in time is
Networks in Transition
441
a ‘snapshot’ that shows interactions as they cur-
rently exist. The crucial question then becomes:
How will the pattern of relationships change if
the firms were to get the interaction they ‘want’?
(Schott, 1991).
This was the motivation behind the current
investigation of how industry networks evolve
over time and in response to industry events.
Understanding how industry events shape net-
works has actionable managerial implications,
including: (1) articulating what network changes
to expect when specific types of industry events
take place; and (2) determining how to use the
anticipated network changes to a specific firm’s
advantage. Thus, we felt that efforts to develop
a theory of structural change in networks would
be a valuable addition to the strategy literature.
However, the broad claim that network struc-
tures change over time does not hold much theo-
retical interest in itself. On the other hand, the
nature of such change, the ‘occasions’ in which
it takes place, and the direction of the change
are topics worthy of serious investigation, and
form the focus of this paper. We characterize the
nature of structural change in terms of centrality,
centralization,
and
interblock
relations.
The
‘occasions’ of structural change are characterized
in terms of the properties of specific industry
events that set off the change. The direction of
structural change is characterized as being either
structure-reinforcing or structure-loosening. These
three
facets
(i.e.,
nature,
‘occasions,’
and
direction) jointly describe the process of struc-
tural change in interfirm networks.
THE PROCESS OF STRUCTURAL
CHANGE IN NETWORKS
Before defining the nature, ‘occasions,’ and direc-
tion of structural change in networks, we must
first indicate what does not constitute structural
change. First, since network structure is viewed
as the relatively enduring pattern of relationships,
it does not change merely because some actors
leave a network position and some others enter
it. This is consistent with the focus of network
analysis on identifying structural patterns that go
beyond the specific actors that may occupy a
network position at any given point in time.
Second, the network structure does not change in
a fundamental sense merely because the rate of
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
network activity increases or decreases. For
example, if actors that were previously in contact
were merely to increase their interaction with
each other, while not initiating any new contacts
with others or different types of contacts with
current partners, the pattern of relationships does
not change. In contrast, true structural change
would be evidenced by significant variation over
time in the underlying pattern of relationships
that bind a given set of actors. This was the logic
employed in defining the nature, ‘occasions,’ and
direction of structural change in industry net-
works.
Nature of structural change
While an array of techniques have been recently
proposed to characterize the nature of structural
change (Burt, 1992; Sanil, Banks, and Carley,
1995), we chose to focus on changes in network
centrality and interblock relations, for the follow-
ing reasons (cf. Doreian, 1986; Snijders, 1990).
As far as centrality was concerned, our interest
in understanding how networks are strategic
resources dictated that we begin by focusing on
the bases of partner attractiveness. This was on
the reasoning that one cannot design and build an
efficient–effective network (Burt, 1992) without
being able to attract the right partners. The causal
force behind centrality resides in the direct and
indirect ‘demand’ for relations with an individual
actor (Burt, 1991). This suggested to us that we
should begin by analyzing structural change in
terms of what made firms in an industry more
or less ‘in demand’ as partners over time.
With respect to interblock relations, we rea-
soned that structural change should be manifest
in changing relations between groups of firms, as
well as between individual firms. In the strategy
literature, the group level of analysis has always
been significant, (e.g., strategic groups) and we
felt that structural change should be investigated
at this level as well. In network terms, firms can
be divided into blocks, with each block being
composed of firms that are structurally similar,
and relations among blocks can be analyzed (e.g.,
Nohria and Garcia-Pont, 1991). Thus, our dis-
cussion characterizes structural change in terms
of the relative centrality of participating firms
and in patterns of relationships between blocks
of firms.
442
R. Madhavan, B. R. Koka and J. E. Prescott
Centrality and centralization
The network notion of centrality reflects the sig-
nificance of a firm in its network. At the simplest
level of conceptualization, a highly central firm
is likely to be connected to many more partners
than the less central firm. Centrality comes from
being the object of relations from other contacts
(Burt, 1991). In the network literature, centrality
has been empirically linked to effects such as
power
(Krackhardt,
1990),
reputation
(Galaskiewicz, 1979), and the early adoption of
innovation
(Rogers,
1971).
Being
a
well-
connected and significant player in a network can
be a crucial strategic advantage. Each contact
is a potential conduit for relevant information,
resources, or influence. As Galaskiewicz (1979)
noted, an organization’s power is determined less
by its internal resources than by the set of
resources it can mobilize through its contacts.
The more such contacts the firm has, the better
it is ‘plugged in’ to the key task and influence
processes of the industry, and the stronger is its
strategic advantage. Thus, the relative centrality
of firms is an important factor for the network
analyst interested in describing structural change
in networks. Firms that realize the value of cen-
trality may constantly be attempting to improve
their centrality by connecting with more and more
central partners. In the process, they may abandon
relationships with partners who are perceived as
being less valuable, reducing the latter’s cen-
trality. Since firm centrality is evidently such a
key resource, we argue that changes in the rela-
tive centrality of firms are important indicators
of structural change. If a given set of firms
retains, over the years, their relative levels of
centrality, there is no indication of a structural
change in the network. On the other hand, a
network undergoing true change would be charac-
terized by either an increase or a decrease in the
firms’ levels of centrality. Attempting a centrality-
based assessment of structural change is consis-
tent with our focus, as strategic management
researchers, on the firm-level implications of such
change. Thus, we argue that before-and-after mea-
sures of centrality can be used to assess structural
change at the firm level. At least one prior inves-
tigation of network evolution (Doreian, 1986) has
used changes in actor centrality as a means to
estimate structural change in networks.
Centrality is a firm-level construct, but the
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
distribution of centrality among firms in a net-
work gives rise to the construct of centralization,
which indexes the tendency of a single firm to
be more central than all others in the network
(Freeman, 1979). In a highly centralized network,
for example, there will be a few very central
firms and many peripheral firms. On the other
hand, a network with many firms at relatively
similar levels of centrality is less centralized.
Thus, centralization provides an aggregate meas-
ure for the distribution of centrality among firms
in a network. Analogous to centrality, before-
and-after measures of centralization can be used
to assess structural change at the network level
(Doreian, 1986).
Interblock relations
In analyzing centrality and centralization, we are
mainly concerned with understanding the patterns
of relationships established by individual firms
with other firms. A complementary concern for
intergroup relations will allow us to explicate
general types of relations maintained by groups
of firms (Scott, 1991). For purposes of analysis,
firms may be divided into blocks in one of three
ways, using the respective criteria of structural
equivalence (e.g., Lorrain and White, 1971), gen-
eral equivalence (e.g., Burt, 1976), and contextual
equivalence (e.g., Freeman and Barley, 1990).
While the criteria of structural equivalence and
general equivalence are based on patterns of
relationships between firms, that of contextual
equivalence
is
based
on
attribute
similarity
between firms (e.g., Barley, 1988). We chose the
criterion of contextual equivalence as it allowed
us to capture change in patterns of relationship
between strategically meaningful categories of
firms. Each block in our analysis consists of firms
that are similar on the attribute of technology.
There are two advantages to using the criterion
of contextual equivalence. First, it is consistent
with the concern of strategic management with
performance
and
other
differences
between
‘groups’ of firms, as evidenced by the wealth of
research on strategic groups (McGee and Thomas,
1986). Second, contextually equivalent blocks
allow us to keep the block membership constant
and thus to observe how relations within and
between those blocks change over time. The
underlying logic was that changes in relationship
patterns within and between blocks are an
Networks in Transition
443
important indicator of structural change. For
example, Nohria and Garcia-Pont (1991) analyzed
strategic alliance activity in the world automobile
industry in terms of relationships between stra-
tegic groups. A longitudinal treatment of similar
data would bring out shifts in the pattern of
relationships between groups with differing capa-
bilities, in conjunction with changes in the com-
petitive dynamics as a key indicator of struc-
tural change.
‘Occasions’ for structural change
In defining the ‘occasions’ that trigger structural
change, we adopted an event focus. Recent work
in organizational research has underlined the
value of a theory of how organizational phenom-
ena are structured (Barley, 1986). A promising
line of reasoning suggested by this work is to
investigate specific events as ‘occasions for struc-
turing’ (Barley, 1986). For example, based on
his study of how new technologies such as medi-
cal imaging help to recreate role structures among
radiologists and radiological technologists, Barley
(1986: 80) proposed that major restructuring only
occurs
when
organizations
face
‘exogenous
shock.’ In a network context, Burkhardt and Brass
(1990) investigated how the organizational power
structure changes in response to a change in
technology. Their study showed that early adop-
ters of a new technology increased their power
and centrality to a greater degree than did later
adopters (Burkhardt and Brass, 1990).
An event focus tracks the evolution of an
industry network over time by examining struc-
ture through various ‘windows’ of time (e.g.,
Doreian, 1986). The length of any ‘window’
depends on specific events. A key advantage of
this approach is that researchers and managers
are equally likely to agree that industry events
provide more relevant ‘check points’ for network
evolution than arbitrary time periods. An anal-
ogous logic may be found in the use of a
‘decision focus’ for case study research in organi-
zations (e.g., Allison, 1971). In the absence of
an event focus, the choice of an appropriate
‘window’ length becomes a thorny problem. As
Doreian (1986) suggests, we need a time-specific
theory rather than one which merely states that
network processes are located in time. Adopting
an event focus provides an elegant solution to
this problem: the timing of industry events allows
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
for a setting in which the length of ‘windows’ is
automatically determined for the researcher.
Building on this reasoning, we propose that
key industry events provide occasions for network
restructuring. The events of interest are specific
occurrences
that
knowledgeable
observers
acknowledge as impacting competition on an
industry-wide basis and in the long run. Illus-
trations include major technological develop-
ments, the entry of a resourceful and determined
competitor, changes in regulatory infrastructure,
or dramatic shifts in consumer preferences.
As we argued earlier, network position deter-
mines access to key resources, such as infor-
mation or the potential for control (Burt, 1992),
and network structure is a crucial influence on
firm performance and industry evolution (cf.
Pfeffer, 1981). Specific industry events provide
opportunities for firms to attempt to improve their
positions in their industry network. Some industry
events, such as fundamental regulatory reform or
radical technological change, potentially change
the basis of competition in an industry. This will
have an observable impact on the network of
relations within the industry. Firms may find that
they need access to a different set of resources
than provided by their current partners. They
may then be prompted to initiate a new set of
relationships with a different set of partners. Thus,
some industry events will be followed by substan-
tial change in the industry’s network structure.
While all technology shocks need not result in
fundamental industry reorientation (Tushman and
Anderson, 1986), there is ample evidence that
some technology events reshape industries. For
example, Piore and Sabel (1984) argued that
changes in production technology underlie shifts
in the way economic activity is organized. Simi-
larly, Glasmeier (1991) documented the radical
impact that technology changes wrought on the
production networks of Swiss watchmakers. On
the other hand, all industry events may not result
in such drastic change. There may be some types
of changes that help to reinforce the current
strategic trajectory of the industry (e.g., by
removing barriers to the achievement of current
industry goals). Such an event could increase the
value of the firm’s current set of relationships,
and actually serve to strengthen them.
444
R. Madhavan, B. R. Koka and J. E. Prescott
Direction of structural change
The third facet of structural change revolves
around the question: What types of network
changes are likely to follow key industry events?
Our main thesis is that the structural impact of
industry
events
may
be
either
structure-
reinforcing or structure-loosening. We view the
structure of a network as being reinforced if, in
general, hitherto powerful firms increase their
network power while hitherto less powerful firms
become even less powerful. In other words, the
network structure is reinforced if the existing
distribution of network power is strengthened,
benefitting the network ‘rich’ at the expense of
the network ‘poor.’ On the other hand, the struc-
ture of a network is loosened if hitherto powerful
firms decrease their network power while hitherto
less powerful firms become more powerful than
they were. To continue the ‘rich–poor’ analogy,
the network structure is loosened if network
power is redistributed, benefitting the network
‘poor’ at the expense of the network ‘rich.’ Thus,
the direction of structural change hinges on
whether the existing distribution of network
power is reinforced or loosened.
Accordingly, we now proceed to develop the
framework to arrive at a strategically meaningful
conceptualization of when to expect what kind
of structural change in industry networks. Our
thesis is that the characteristics of the ‘occasion’
for structural change may be used as a basis for
predicting the direction of such change. Armed
with this knowledge, managers may take appro-
priate steps in order to protect and enhance their
network positions and associated ‘social capital.’
The answers to three questions help to determine
whether
an
event
is
potentially
structure-
reinforcing or structure-loosening (See Table 1):
(1) How does the event affect the currently
Table 1.
Characteristics of events affecting network structure
Characteristics of events
Structure–reinforcing event
Structure-loosening event
Effect on the bases of competition
Enhances and strengthens existing
Radically changes the bases of
bases of competition
competition
Who benefits?
Dominant players with high cen-
Peripheral players with low cen-
trality in current network
trality in current network
Who initiates?
Dominant players in current net-
Peripheral players in current net-
work
work
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
accepted bases of competition in the industry?
(2) Who does the event potentially benefit?
(3) Who initiated the event? Structure-reinforcing
events serve to strengthen the current structure
of the network, because they typically sustain,
rather than disrupt, current industry trajectories.
Structure-reinforcing events mainly benefit those
players who are already in key positions in the
network, and are likely to be initiated by those
same dominant players. In contrast, structure-
loosening events trigger change in existing pat-
terns of relationships, as they are more likely to
be disruptive of accepted practices and industry
definitions. Structure-loosening events tend to
benefit peripheral players who use the event to
improve their positions, and are less likely to be
initiated by dominant industry players. Based on
the characteristics of each event, it may be pos-
sible to predict whether it will reinforce or loosen
the current structure. Such prediction will be
crucial in determining the nature of structural
impact to expect after key industry events, and
what moves to make in response.
Our framework thus characterizes the process
of network evolution in terms of the nature,
‘occasions’ and direction of structural change.
The nature of structural change is characterized
in terms of shifts in centrality, centralization,
and relationships between contextually equivalent
technology
blocks.
‘Occasions’
of
structural
change are characterized in terms of key industry
events that provide opportunities for network
restructuring. The direction of structural change
is
characterized
as
being
either
structure-
reinforcing or structure-loosening. These three
aspects of structural change are the building
blocks for a description of the overall process of
structural change and provide the framework for
developing our hypotheses.
Networks in Transition
445
PROPERTIES AND HYPOTHESIZED
EFFECTS OF STRUCTURE-
REINFORCING EVENTS
Structure-reinforcing events will have the follow-
ing three properties:
1. They build upon and extend currently accepted
bases of competition in the industry. For
example, Tushman and Anderson (1986) have
argued that incremental technological improve-
ments enhance and extend the underlying tech-
nology, thus reinforcing the existing order.
Bower and Christensen (1995) have put for-
ward the notion of performance trajectories,
which is useful here. The performance tra-
jectory of a product class is defined by the
rate at which the performance of a product
has improved, and is expected to improve,
over time (Bower and Christensen, 1995: 45).
A structure-reinforcing event will typically
help to sustain the performance trajectory of
a
product
class—for
example,
thin
film
components in computer disk drives replaced
conventional ferrite heads and oxide disks,
enabling data to be recorded more densely. A
regulatory event that removes strategic barriers
that previously existed would be another type
of structure-reinforcing event. For example,
under current laws, the regional telephone util-
ities are seriously constrained in their inter-
national investments. A change in this regula-
tory
framework
would
increase
the
international activity of such firms, without
introducing radical change in the underlying
competitive logic of their industry. In contrast,
the proposed changes that would allow tele-
phone and cable utilities to compete with each
other would probably introduce radical change
in the basis of competition (e.g., telephone
utilities would have to make decisions on what
types of programming to carry). From a cogni-
tive perspective, structure-reinforcing events
would not result in any substantial change in
the managerial recipes (Spender, 1980) in an
industry.
Thus,
structure-reinforcing
events
will typically extend and strengthen a current
competitive regime.
2. Firms that are already in powerful network
positions will benefit more from such events
than more peripheral firms. There are two
reasons why this is the case. First, the domi-
nant firms in the industry are likely to be the
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
ones that are also in powerful network posi-
tions. Since the event reinforces the current
competitive paradigm, the dominant firms are
likely to become even more dominant, becom-
ing more desirable partners in the process.
Second, their powerful position in the network
allows them a better opportunity to capitalize
on the event than the more peripheral firms.
The causal force behind this advantage is pro-
vided by the information and control benefits
(Burt, 1992) of their network positions.
3. They are more likely to be initiated by firms
that are currently dominant. Given that a struc-
ture-reinforcing event is likely to benefit them
more than their weaker counterparts, dominant
players are more likely to initiate such events.
A variety of reasons combine to make this
so. In the field of technology, for example,
incremental progress occurs through the inter-
action of many firms, unlike initial break-
throughs (Tushman and Anderson, 1986).
Since the central players in the network are,
by definition, placed at the center of such
interaction, they are likely to be the source of
such improvements. Similarly, events such as
a regulatory initiative that removes strategic
barriers have to be coordinated and orches-
trated through a lobbying process—again, a
task that dominant firms are likely to be more
effective at (e.g., New York Times, 29 Nov-
ember 1994). In interpersonal networks, Burk-
hardt and Brass (1990) found that if early
adopters are more central than late adopters
before a technological event, the existing pat-
tern of relationships is likely to be strength-
ened. Finally, Bower and Christensen (1995)
have put forward the argument that established
companies tend to concentrate on technologies
that sustain their current performance trajector-
ies because those technologies serve their cur-
rent customer base better. Thus, the established
and dominant firms are less likely to initiate
radical and disruptive change (Henderson and
Clark, 1990).
Hypothesized effects
Centrality
Structure-reinforcing events provide highly central
firms with an opportunity to increase their cen-
trality. Since their desirability as partners is now
446
R. Madhavan, B. R. Koka and J. E. Prescott
higher, they will be able to attract more partners.
At the same time, the peripheral players will now
find it even more difficult to attract partners.
Thus, highly central firms are in the best position
to benefit from a structure-reinforcing event, and
we would expect them to remain central after the
event. Similarly, peripheral players will remain
peripheral. In sum, any structural change would
be an increase in centrality for the already central
firms and increasing marginalization of peripheral
firms. Thus, we would expect a significant and
positive correlation between centrality indicators
before and after the structure-reinforcing event.
Accordingly:
Hypothesis 1a:
The centrality of individual
firms before a structure-reinforcing event will
be significantly and positively correlated with
their centrality after the event.
Centralization
The results at the level of the overall network
will be analogous. Since hitherto central firms
will increase their centrality, and hitherto periph-
eral firms will remain marginalized, a structure-
reinforcing event will increase the overall cen-
tralization of the network. Thus:
Hypothesis 1b:
A structure-reinforcing event
will have the effect of increasing the centrali-
zation of the network.
Interblock relations
The causal forces outlined above would also
imply that a structure-reinforcing event would
benefit groups of firms which are hitherto in
attractive positions. Since the fundamentals of
competition do not shift, the pattern of relation-
ships between the blocks should be reinforced
over time. In other words, the bases of partner
attractiveness and the bonds between blocks of
firms will remain, or indeed, strengthen. We
expect to observe an increasing set of relation-
ships among those blocks that potentially benefit
from the structure-reinforcing event.
Explicit hypothesis testing regarding the sta-
bility or change in the pattern of interblock
relations is hampered by the lack of appropriate
methodological and statistical tools (Snijders,
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
1990). Thus our approach is grounded in a set
of procedures (described in the Results section)
which allow us to draw conclusions regarding the
role of a structure-reinforcing event on the pattern
of interblock relations.
PROPERTIES AND HYPOTHESIZED
EFFECTS OF STRUCTURE-
LOOSENING EVENTS
We use a reverse logic to articulate the properties
and effects of structure-loosening events. Struc-
ture-loosening events will have the following
three properties:
1. They radically change the basis of competition
in the industry. One of the best examples is
provided by what Tushman and Anderson
(1986) have labeled ‘competence-destroying
discontinuity.’ Such discontinuities mark a rad-
ical shift in technological paradigms, and tend
to obsolete the skills and knowledge base pre-
viously required. Nontechnology events can
be structure-loosening, too. For example, the
proposed regulatory move to dismantle the
natural monopolies given to telephone and
cable utilities is widely expected to change
the competitive dynamics of the telecommuni-
cation industry. In network terms, structure-
loosening events imply that firms may need
to look for a sharply different set of partners
who will provide them with a new set of
resources.
2. Firms that are already in powerful network
positions have no automatic benefit from such
an event. The more peripheral firms are equa-
lly likely to benefit. Previously powerful firms
may be constrained in their ability to quickly
find new partners, largely because of prior
commitments to existing partners. On the other
hand, the relatively peripheral firms may have
greater leeway. Further, the previously domi-
nant firms may have difficulty adopting the
new paradigm (Henderson and Clark, 1990),
making them less desirable partners under the
new competitive regime. These factors serve
to dampen whatever information and control
benefits prior network centrality may have
offered.
3. They are more likely to be initiated by firms
that
are
currently
peripheral.
The
initial
impetus, such as the adoption of a radical new
Networks in Transition
447
technology or a fundamentally new pricing
strategy, is more likely to come from smaller
players acting alone rather than from the inter-
actions of highly central firms (Henderson and
Clark, 1990). The peripheral players also have
the motivation to introduce structure-loosening
events, as they can only gain—unlike the cur-
rently central players, who stand to lose their
dominance. By the same token, the currently
dominant players are unlikely to initiate struc-
ture-loosening events that can radically rede-
fine the competitive dynamics of the industry.
For example, consider IBM’s bitter and long-
lived opposition to the UNIX operating system
during the 1980s.
Hypothesized effects
Reversing the logic of the previous set of hypoth-
eses, we argue as follows.
Centrality
The structure-loosening event will reduce the
benefit of previous centrality. Since hitherto
highly central firms are no longer in the best
position to benefit from the event, we would
expect them to become less central. Peripheral
players, on the other hand, may use the oppor-
tunity to improve their position, and increase their
centrality. Thus, it may appear that the centrality
of individual firms before a structure-loosening
event will be negatively correlated with their
centrality after the event. However, this is an
extreme characterization of structure-loosening,
representing a case ‘in the limit.’ In reality, we
expect some countervailing forces, because there
are several reasons why prior centrality could be
‘sticky’ even after a structure-loosening event.
First, existing contacts may help to mitigate the
negative impact of the structure-loosening event,
even for those firms whose current competencies
are made less valuable. With the help of those
contacts, these firms may succeed in salvaging
the situation and remain more or less dominant—
at least for the time being. This is one more
powerful reason why networks should be viewed
as a strategic resource: they shield the firm from
being buffeted by industry shocks, and increase
the possibility of immediate survival (Stopford
and
Baden-Fuller,
1990).
Second,
structure-
loosening events need not benefit all peripheral
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
firms—many previously peripheral firms may
remain peripheral even after the event. Some
previously peripheral firms that were well-posi-
tioned to benefit from a particular structure-
loosening event may increase their centrality (e.g.,
those who possess a competency that has now
become
valuable).
Because
of
these
countervailing movements, the net correlation will
not be negative. What we expect, instead, is that
the correlation between before-and-after centrali-
ties will be considerably less in the case of
structure-loosening events than in the case of
structure-reinforcing events. Thus:
Hypothesis
2a:
The
correlation
between
before-and-after centralities
will be signifi-
cantly less for a structure-loosening event than
it will be for a structure-reinforcing event.
Centralization
Structure-loosening events provide less central
firms with an opportunity to increase their cen-
trality. Since their desirability as partners is now
higher than it was earlier, they will be able to
attract more partners. At the same time, the hith-
erto central players will now find it more difficult
to attract partners, as their competencies may be
less valuable under the new competitive regime.
This will have the effect of temporarily decreas-
ing the overall centralization of the network.
1
Eventually, the network may revert to a highly
centralized state, with a new set of firms being
in highly central positions. However, during the
transition phase, the gap between the hitherto
central firms (whose centrality now decreases)
and the hitherto peripheral firms (whose centrality
now increases) is likely to decrease. Thus:
Hypothesis 2b:
A structure-loosening event
will have the effect of temporarily reducing
the centralization of the network.
Interblock relations
The causal forces outlined above would imply
that a structure-loosening event would signifi-
cantly change the established pattern of relations
between blocks of firms. The shift in competitive
1
We are grateful to an anonymous reviewer for pointing out
that the decrease in centralization may be temporary.
448
R. Madhavan, B. R. Koka and J. E. Prescott
regime undermines the current bases of partner
attractiveness, potentially leading to a ‘shuffling’
period which will rearrange the bonds between
blocks of firms. During this period of ‘shuffling,’
blocks which previously had little reason to inter-
act may need to draw on each other’s resource
bases to succeed in the new competitive order.
Conversely,
blocks
which
have
established
relationships may find them to be less important
for the emerging competitive environment. We
expect to observe an increasing set of relation-
ships involving blocks who are likely to be the
beneficiaries of the structure-loosening event. At
the same time, we expect to observe fewer inter-
actions among hitherto connected blocks.
METHODS AND DATA
We tested our hypotheses by analyzing and inter-
preting changes in the structure of the strategic
alliance network in the global steel industry
between 1977 and 1993. The steel industry is an
appropriate context for the study both because it
has witnessed significant strategic changes over
the last few decades (e.g., Abe and Suzuki, 1991)
and because interfirm relationships have been an
acknowledged influence on the industry’s evolu-
tion (e.g., Knoedler, 1993). The strategic alliance
network is an appropriate network to focus on
because (1) strategic alliances represent relation-
ships between rivals, which are acknowledged
sources of influence on industry evolution (Porter,
1980); (2) they represent flows of knowledge
and access to markets, which have been key
success factors in the industry (Grant et al.,
forthcoming); and (3) there has been a significant
amount of strategic alliance activity in the global
steel industry (Madhavan and Prescott, 1995).
Choice of industry events
Two key industry events, the regulatory shock of
1984 and the technology shock of 1987, are used
as the anchor points for our analysis. In 1984,
several regulatory decisions (the Cooperative
Research Act, the approval of the National-NKK
venture) in the United States signaled a lowering
of the traditional institutional barriers to coopera-
tive strategy. In the same year, the Voluntary
Restraints Agreement, by limiting the amount
of steel foreign producers could export into the
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
American market, also provided some impetus to
the cooperative production of steel in the United
States—e.g., production joint ventures between
Japanese and U.S. steel producers. The 1987
technological event was the decision by Nucor to
adopt the Compact Strip Production technology
for its Crawfordsville, Indiana plant, which had
two major implications: (1) a potential threat to
the integrated players who had a monopoly of
the high-margin sheet steel market; and (2) an
acceptance of the Compact Strip Production tech-
nology as the apparent winner over other technol-
ogies such as the Hazelett process. The literature
on the steel industry suggests that these two
events are widely perceived as being turning
points in the history of the industry (e.g., Wall
Street Journal, 16 February 1984; Ghemawat,
1993; Hogan, 1991).
We reasoned that the particular regulatory
event that took place in 1984 has the character-
istics of a structure-reinforcing event. First, it did
not change the basis of competition in the steel
industry. Rather, what it did was to remove bar-
riers to cooperative activity, while leaving in
place the relative strengths of industry players
(e.g., Wall Street Journal, 16 February 1984).
Second, the already central players were in the
best position to benefit from the event, as they
could use their contacts to find new partners
or to strengthen existing relationships. Third, as
regulatory events are often the result of the lobby-
ing efforts of powerful industry actors (e.g., New
York Times, 29 November 1994), it is quite likely
that this event was initiated by dominant players.
One likely scenario is that this regulatory event
was driven by the need for U.S. and Japanese
integrated steel producers to join together in order
to service the steel needs of Japanese transplant
automobile manufacturers in the United States.
On the other hand, the particular technology event
that took place in 1987 has the characteristics of
a structure-loosening event. First, it acted as a
competence-destroying
discontinuity
(Tushman
and Anderson, 1986) for the central players (the
integrateds), and radically changed the basis of
competition in the industry. The primary impact
was that it potentially opened up the attractive
sheet steel market (hitherto the preserve of the
integrateds) to competition from the minimills
(Ghemawat, 1993). The introduction of the com-
pact
strip
production
process
substantially
changed the operating economics of the steel
Networks in Transition
449
industry, representing a 10–15 percent drop in
production
cost
and,
more
significantly,
a
reduction in capital cost from $1000 per ton of
capacity to $300 a ton (Christensen and Rosen-
bloom, 1991). Second, the prior centrality of the
integrated manufacturers did not give them any
advantage in exploiting the new technology.
Further, with the emergence of a viable new
technology, the hitherto central integrateds were
no longer as attractive (as potential partners) as
they used to be. Third, the technology event was
initiated by Nucor, a hitherto (very) peripheral
firm in the network. (For an additional reason to
believe that the 1987 event was a structure-
loosening event, see the Results section, wherein
we discuss the impact of the events on the num-
bers of strategic alliances initiated.)
Data
Since the focus of our study was the evolution of
networks, we were primarily interested in alliance
relationships within the industry. We accordingly
adopted
a
relationship
criteria
(Laumann,
Marsden, and Prensky, 1983) to define the bound-
aries for our study. Only those firms in the steel
industry that formed at least one strategic alliance
and their partners were included in the study.
The data on strategic alliances in the global steel
industry were obtained from the Dow Jones News
Retrieval Service. These data covered the period
1977–93, and related to all types of strategic
alliances, including joint ventures, joint programs,
licensing arrangements, and long-term supply
relationships. In all, 130 firms participated in the
network, comprising 41 integrated steel firms, 10
minimills, and 9 specialty steel producers, 20
upstream partners (e.g., a coal mining company),
26 downstream partners (e.g., an automobile
manufacturer) and 24 firms from other industries.
In terms of nationality, there were 66 American
firms, 11 Canadian firms, 19 Japanese/Korean
firms, 25 European firms and 9 firms from other
parts of the world. In terms of production output,
30 of the top 50 steel firms in the world partici-
pated in alliance activity and they accounted for
75 percent of the top 50 output in 1990.
Adapting the approach suggested by Contractor
and Lorange (1988) and used by Nohria and
Garcia-Pont (1991), each alliance was assigned a
numerical score indicating its ‘intensity’ on a
continuous 9-point scale. A score of 1 on the
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
scale indicated an alliance of low intensity
(technical training), while a score of 9 indicated
an alliance of high intensity (greenfield joint
venture). Thus, the score for each alliance proxies
the strength, or ‘intensity,’ of the relationship,
thereby indicating the potential for resource flow
(cf. Krackhardt, 1992). If a pair of partners had
more than one alliance, the scores for all alliances
were added to form a composite index. As an
additional validity check, we also conducted other
supplementary analyses, the results of which are
detailed in the Results section.
In order to facilitate analysis of changes
‘around’ each event, the data were captured in
the form of a series of 130
×
130 matrices
depicting the relationship between each potential
pair of partners. Since strategic alliances are bidi-
rectional by definition, these matrices were sym-
metric. The first matrix, Structure-Reinforcing at
time period one (SR1), consisted of all alliances
announced in 1977–83, i.e., before the regulatory
event of 1984—a structure-reinforcing event. The
second
matrix,
Structure-Reinforcing
at
time
period two (SR2), consisted of all alliances
announced in 1984–93, i.e., after the regulatory
event. Comparing these two networks would
enable us to draw some conclusions about the
effect of the structure-reinforcing event. A similar
‘cut’ was made to bring out the effect of the
technology event of 1987—a structure-loosening
event. The third matrix, Structure-Loosening at
time period one (SL1), consisted of all alliances
announced in 1977–86, i.e., before the technology
event of 1987. The fourth matrix, Structure Loos-
ening at time period 2 (SL2), consisted of all
alliances announced in 1987–93, i.e. after the
technology event. The effect of the structure-
loosening event would be assessed by comparing
these two networks.
It must be noted here that our method of
creating separate matrices using pre- and post-
event alliances parallels the ‘sliding window’
approach advocated by Doreian (1986), rather
than the use of an ‘expanding window.’ The
major advantage of such an approach is that it
helps to purge the resultant matrices of historical
components. By ensuring that the post-event
matrix contains only new alliances initiated after
the
event,
the
‘sliding
window’
approach
‘decumulates’ the network, and is thus able to
better
capture
structural
change
(Doreian,
1986:61).
450
R. Madhavan, B. R. Koka and J. E. Prescott
The approach adopted resulted in matrices
which have an overlapping time coverage across
SR1 and SL1, as well as between SR2 and SL2,
which was unavoidable since we were investigat-
ing the effect of two events in the same industry.
The comparisons and statistical tests, however,
were
between
the
nonoverlapping
pairs
of
SR1/SR2 and SL1/SL2. We did investigate a
three-period analysis, comparing and contrasting
the structure in 1977–83, 1984–86, and 1987–
93. However, the short ‘window’ and sparseness
of the second network (1984–86) were causes
for serious concern. In this context, Doreian
(1986: 62) has pointed out that short ‘windows’
are undesirable because of higher ratios of noise
to structural information. Accordingly, we chose
the two-period analysis.
Operationalization of network measures
From among the various measures of centrality
available (e.g., Freeman, 1979), we chose the
flow betweenness measure, as it best reflects the
idea of centrality emerging from the ability to
benefit
from
and
control
the
flow
of
resources/information in the network (Freeman,
Borgatti, and White, 1991). The measure assesses
the amount of flow (m
jk
) between firm j and
firm k that must pass through firm i. The flow
betweenness of firm i is the sum of all m
jk
where
i, j and k are distinctive and is given by:
Centrality (Firm i)
=
冘冘
m
jk
(i)
The flow betweenness measure is therefore a
measure of the contribution of firm i to all pos-
sible maximum flows and depends on the number
and intensity of alliances of firm i. This was
consistent with the way we coded each alliance
for its capacity for resource flow.
Centralization is operationalized as the average
difference between the centrality of the most
central firm and that of all other firms. For a
given network, network centralization is given by:
Centralization of Network
=
冘
C
max
−
C(firm i)
n
−
1
where C
max
is the centrality measure of the most
central firm and n is the number of firms in
the network.
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
We operationalized blocks in terms of the tech-
nology of the firms. Technology was selected as
the attribute for creating contextual equivalent
groups for three reasons. First, the technological
basis (integrated, minimill, specialty) of a firm
significantly affects its cost structure, product
offerings, and customer groups. Second, tradition-
ally there has been a limited degree of compe-
tition in the product–markets between the tech-
nology
groups.
In
recent
years,
however,
competition between integrated and minimills has
intensified. Third, international competition has
been significantly influenced by the implemen-
tation of new-generation technologies since the
end of World War II. Given this set of factors,
changes in the pattern of relationships between
technology blocks should provide insights into
the process of network transition. The 130 firms
were thus divided into six technology blocks—
integrated steel firms, minimills, specialty steel
firms, upstream, downstream, and others—and
we investigated how relations between them had
changed after each event.
Analysis
The above four 130
×
130 matrices were used as
the input to UCINET IV, a widely used network
analysis program. UCINET IV was used to ana-
lyze firm centrality (Hypotheses 1a and 2a) and
overall centralization (Hypotheses 1b and 2b) in
each network. The approach we took for explor-
ing interblock differences was block-model analy-
sis. UCINET IV allows the ‘collapsing’ of firms
into blocks and the mapping of relationships
within and between these blocks. In order to
increase the validity of our conclusions, relations
between or within blocks were considered to exist
if the density was at least equal to or greater
than the overall network density (Gerlach, 1992).
Network density is measured as the total value
of all ties divided by the number of possible ties,
and signifies the ‘average’ level of relationship
in the network. By taking only those relationships
that are equal to or greater than network density,
we consider only those interblock relationships
that are stronger than the ‘normal’ relationship
level in the network. Relations between the blocks
were then examined before and after each event
in an effort to describe and interpret how the
pattern of relations had changed.
Networks in Transition
451
RESULTS
The accompanying charts in Figure 1 present the
number of alliances in the global steel industry
between 1980 and 1993 (the period 1977–80 has
been collapsed into 1980 for ease of depiction).
The data show a steady increase in the cumulative
number of announced strategic alliances. On ana-
lyzing the number of new alliances announced in
any year, however, we find an interesting pattern.
There was a sharp rise in the number of new
alliances
in
1984,
i.e.,
after
the
structure-
reinforcing event. Immediately after the structure-
loosening event of 1987, there was sharp drop in
the number of new alliances, though alliance
activity picked up in subsequent years. We inter-
pret this drop as confirmation that the 1987 event
was a structure-loosening event: steel industry
Figure 1.
Strategic alliances in the global steel industry, 1980–93
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
firms would have taken time to analyze the devel-
opment and assess its implications, as it was
potentially a competence-destroying discontinuity
(Tushman and Anderson, 1986).
Overall, 130 strategic alliances were reported
during 1977–93. Of these, 8 percent were techni-
cal training agreements; 10 percent were patent
licensing
agreements;
3
percent
were
production/buy-back agreements; 3 percent were
‘know-how’ licensing agreements; 1 percent were
service agreements; 6 percent were nonequity
alliances; 5 percent were joint ventures in existing
operations; and 64 percent were greenfield joint
ventures.
Table 2 presents the results of the centralization
and centrality analyses. First, we found that there
was a significant correlation of 0.58 (p
⬍ 0.001)
between firm centralities before and after the
452
R. Madhavan, B. R. Koka and J. E. Prescott
Table 2.
Tests of hypotheses
Hypothesis
Network measures
Tests of significance
Interpretation
1. Hypothesis 1a
Correlation between actor centralities
before and after SR event: 0.58
p
⬍ 0.001
Strongly supported
2. Hypothesis 1b
Centralization before SR event
a
: 30.33%
Centralization after SR event: 64.23%
Difference between the two measures:*
z
=
5.47
Strongly supported
(p
⬍ 0.001)
3. Hypothesis 2a
Correlation between actor centralities
before and after SR event: 0.58
Correlation between actor centralities
before and after SL event: 0.38
Difference
between
the
two
corre-
lations:**
z
=
1.73
Supported
(p
⬍ 0.05)
4. Hypothesis 2b
Centralization before SL event
b
: 52.55%
Centralization after SL event: 44.42%
Difference between the two measures:*
z
=
1.47
Moderately supported
(p
⬍ 0.08)
N
=
130.
a
SR event: structure-reinforcing event.
b
SL event: structure-loosening event.
*Test for difference between proportions (single-tailed test); **Test for difference between independent correlations (single-
tailed test).
structure-reinforcing event, confirming Hypothesis
1a. Next, we found that network centralization
increased from 30.33 percent to 64.23 percent
(p
⬍ 0.001) after the structure-reinforcing event,
thereby providing support for Hypothesis 1b.
Thus, the data strongly support the expected
effects of a structure-reinforcing event.
The effect of the structure-loosening event was
also as expected. We found that there was a
correlation of 0.38 (p
⬍ 0.001) between firm cen-
tralities before and after the structure-loosening
event, which was considerably less than the corre-
lation for the structure-reinforcing event. The dif-
ference between the two correlations was signifi-
cant at p
⬍ 0.05, thus supporting Hypothesis 2a.
Network centralization decreased from 52.55 per-
cent to 44.42 percent (p
⬍ 0.08) after the struc-
ture-loosening event, thereby providing moderate
support for Hypothesis 2b.
2
2
Our method of coding alliances differentially (on a scale
from 1 to 9) appeared to be the most efficient way to proxy
the overall strength of the relationship between two players,
as it captures, although roughly, the cumulative strength of a
set of alliances. In order to address concerns about the pitfalls
of imputing cardinality, however, we did two supplementary
analyses. First, we ignored the strength of the alliances, and
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
In order to explore the relationships among
the technology blocks, we used an interpretation-
intensive approach. We must note here that both
the composition of the groups, as well as judging
if the patterns of intergroup relations are different,
are both tasks that are heavily dependent on the
industry context of a particular study. The results,
in
our
interpretation,
support
our
position
developed in the hypotheses section (see Figures
2 and 3).
First,
we
examined
how
the
structure-
reinforcing event affected relations between tech-
nology blocks (Figure 2). As noted earlier, only
those relationship arcs that are greater than the
average network density are considered. The
change in the number and the strength of these
expressed the relationship between each pair of players simply
in terms of the number of alliances between them. The
corresponding network was then analyzed in the same way
we analyzed the earlier network. As predicted, the correlation
between SR1 and SR2 centralities (0.66) was higher than that
between SL1 and SL2 (0.45). Second, we constructed a
subnetwork consisting only of greenfield joint ventures and
reran the analysis. Again, as predicted, the correlation between
SR1 and SR2 centralities (0.25) was higher than that between
SL1 and SL2 (0.21), though by a narrower margin.
Networks in Transition
453
Figure 2.
Global steel industry networks: Effect of a structure-reinforcing event on technology blocks
arcs is used to judge if the pattern of relations
between and within blocks is different (Snijders,
1990). The analysis of technology blocks shows
that there were eight arcs before the event, and
seven arcs afterwards. Four of the earlier eight
arcs were dropped and three new ones were
added. The structure-reinforcing event appears to
have resulted in the strengthening of the positions
of the integrated and speciality steel players
(Figure 2). Prior to the event, integrated players
were only connected to each other and to mini-
mills. After the event, the integrated players have
relationships with minimills (much stronger now),
speciality producers, and downstream players, as
well as with each other. Specialty producers have
also established new relationships with minimills
and integrateds after the event. This strengthening
of the integrated and specialty producers was to
be expected, as they were the dominant players
before the event.
Next, we examined how the structure-loosening
event affected relations between the technology
blocks (Figure 3). The impact of the structure-
loosening event appears to have been somewhat
greater and more radical than the structure-
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
reinforcing event. Six out of eight earlier arcs
were dropped, and four new ones were added.
Visually, it can be seen that the structure-
loosening event has resulted in a substantial reori-
entation of the network from a ‘crow’s nest pat-
tern’ of interlocked blocks to a much more
streamlined network with relatively high cen-
trality for the minimill players (Figure 3). This
redesign was expected, especially for the mini-
mills, since the structure-loosening event provided
the ‘occasion’ for less dominant players to
improve their position.
As an additional check on the validity of our
conclusions, we conducted centrality analysis for
the blocks (similar to the analysis conducted ear-
lier at the firm level). For the technology blocks
we found that the centrality correlations before-
and-after
the
structure-reinforcing
event
was
0.757, which was significant at p
⬍ 0.08. As
expected, the centrality correlations before-and-
after the structure-loosening event was 0.647,
which was lower, although not significantly, than
the correlation for the structure-reinforcing event.
In summary, the analysis of technology blocks
shows support for our position. In particular,
454
R. Madhavan, B. R. Koka and J. E. Prescott
Figure 3.
Global steel industry networks: effect of a structure-loosening event on technology blocks
structure-reinforcing events provide conditions
conducive to the established players while struc-
ture-loosening events establish a forum for per-
ipheral players to restructure the network to
their advantage.
DISCUSSION
Our discussion will cover two main topics. First,
we offer our interpretation of the results, and
identify a few limitations. Second, we sketch out
some major implications of our study.
Interpretation of results
On the whole, the results show strong support
for our framework. Three of our hypotheses
(Hypotheses 1a, 1b and 2a) were supported at a
significance level better than 0.05. Hypothesis 2b
was also supported, with results in the hypothe-
sized
direction
and
significant
at
p
⬍ 0.08.
Though appropriate statistical significance tests
are not available, we judged our positions related
to interblock relationships to be supported. Since
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
we hypothesized opposing results for structure-
reinforcing and structure-loosening events, and
found support in both cases, these results suggest
a strong test of our framework. Thus, we con-
clude that the ideas of structure-reinforcing and
structure-loosening events have validity.
An issue worth noting here is the possibility
that the sequencing of change may have an
important effect. In our sample, the structure-
loosening event followed the structure-reinforcing
event. It is quite possible that the firms may have
expended much of their ‘degrees of freedom’ in
responding to the structure-reinforcing event, and
may not have been able to respond very much
to the subsequent event. Barley’s (1986) obser-
vations about the cumulative effect of structuring
are relevant in this context. Because of this path
dependency effect, the net effect of the structure-
loosening event may have been muted. If this is
indeed the case, the sample may have provided
us with a conservative test of our hypotheses,
increasing our confidence in the results.
A related aspect also worth discussion is the
length of time ‘windows’ over which structural
change takes place. In our approach, ‘window’
Networks in Transition
455
length is dictated by the choice of appropriate
event
‘around
which’
structural
change
is
expected to take place. This is consistent with
Doreian’s (1986: 63) advice that we find ‘good
pragmatic reasons’ for selecting a ‘window’
length, given the absence of a ‘theory on the
shelf.’
However, we acknowledge that such a
method does not resolve all issues. For example,
there is the question of how long a structure-
loosening event will have the effect of reducing
centralization—i.e., how long will it be before a
new set of players become highly central, and
the network reverts to a highly centralized state,
if ever? Given the nascency of research into
network evolution, as well as the highly industry-
specific contingencies that might influence the
pace of ‘recentralization,’ there appear to be no
clear answers to such a question at this point.
Another important issue that merits discussion
is whether industry events must necessarily be
followed by network changes of one type or
other. We believe, and our data have demon-
strated, that there are aggregate network changes
that follow key industry events. However, this
does not mean that it is possible to predict how
an industry event will affect the fortunes of every
individual firm in the network. Whether a given
firm is able to improve its position in the network
is dependent on three factors: its ability to attract
desirable partners, its motivation to improve its
position, and whether it has an opportunity to do
so. These factors are consistent with Burt’s
(1992) discussion on the issue of entrepreneurial
motivation and network opportunity. What our
study has shown is that industry events provide
certain firms with the opportunity to improve their
network positions. Firms that are differentially
endowed with the ability or the motivation to
improve their network position will benefit differ-
entially. Clearly, questions of ability and moti-
vation are important (e.g., what constitutes the
ability to improve one’s network position?)—
however, they are beyond the scope of this paper.
Thus, the data on strategic alliances in the
global steel industry show encouraging support
for our main thesis, i.e., that industry events may
be classified as structure-reinforcing and structure-
loosening, and that they impact the subsequent
structure of the network. Our research also shows
that industry events may be identified in advance
as potentially structure-reinforcing or potentially
structure-loosening. This possibility of prediction
has important implications, to be discussed below.
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
Limitations
When interpreting our results, a few cautionary
statements are in order. First, as in most network
studies, our conclusions are based on the experi-
ence of one industry network, i.e., the global
steel industry. In that sense, our study should
perhaps be
viewed
as
ideographic research.
Second, the two events in our study took place
in somewhat close proximity to each other (within
3 years), and some of their effects may have
been intertwined. Third, the effect of the two
events were investigated over relatively short pe-
riods
of
time—9
years
for
the
structure-
reinforcing change and 6 years for the structure-
loosening change—and all the changes may not
have appeared in the network. Finally, it is diffi-
cult to tease out the effect of the path dependency
factor that is ever-present in longitudinal research.
Some of these are limitations that may be
addressed by future work that goes across indus-
tries to investigate the effects of structure-
reinforcing and structure-loosening events.
Implications
As hinted at before, we believe that our results
have significant implications, both for research
and for practice.
Research implications
The increasing popularity of the network perspec-
tive among researchers (Nohria and Eccles, 1992)
and practitioners (e.g., Baker, 1993) has under-
scored the importance of intrafirm and interfirm
networks as contexts for strategic action. While
the network perspective is a recent addition to
strategic management, we believe that it has great
potential to add value. However, before network
thinking can more fundamentally influence our
discipline, we need to address the issue of how
networks may be ‘reengineered’ by managers so
as to facilitate the achievement of strategic goals.
For example, recent work such as Burt’s (1992)
has begun to explore the structure of networks
that are both efficient and effective in ensuring
superior performance. This goal suggests the need
to develop pragmatic guidelines for the strategic
design of networks, much as we now have for
the design of organizations (e.g., Galbraith and
Kazanjian, 1986). In this context, it is instructive
to recall that prescriptions for organization design
were based on systematic study of the contin-
gencies that accompanied specific design attri-
butes (e.g., Lawrence and Lorsch, 1967). Anal-
456
R. Madhavan, B. R. Koka and J. E. Prescott
ogously, we feel that researchers need to direct
attention to network structure as a dependent
variable, in order to understand what brings about
a particular network structure. We perceive our
study as a first step in that direction.
In our empirical analysis, we characterized the
nature of structural change in terms of centrality,
centralization, and interblock relations. What we
have proffered is one way to sketch the process
and drivers of structural change at the network
level. As the nascent literature on network evolu-
tion (Sanil et al. 1995; Burt, 1992) suggests,
there are alternative ways to conceptualize the
nature of structural change. For instance, struc-
tural change can be conceptualized in terms of
range, brokerage, and hole signatures (Burt,
1992). Complementary future studies could inves-
tigate alternatives, or study the process and driv-
ers of structural change at the firm level.
Another aspect of networks that points the way
for fruitful future research is that industry net-
works are composed of many different types of
relationships. Our study focused on strategic
alliances in the global steel industry. However,
the same firms may be tied together by other
types of relations, such as customer–supplier
links, interlocking directorates, bank borrowing,
personnel migration, or membership in industry
associations. Future studies that encompass other
types of relationship networks, or networks com-
prising
multiple
relationships,
could
throw
additional light on the network impact of struc-
ture-reinforcing and structure-loosening events.
For example, it is conceivable that severely struc-
ture-loosening events could reduce the value not
only of relationships with specific partners, but
also of specific types of relationships. The post-
World
War
II
reconfiguration
of
interfirm
relationships in Japan from the zaibatsu system to
the keiretsu system illustrates such a possibility.
Managerial implications
Our perspective of the firm’s network as a stra-
tegic resource suggests that we can think of the
firm’s structural properties as managerial ‘levers’
for enhancing strategic performance. For example,
since centrality is generally held to be a positive
factor, a firm may wish to increase its centrality.
Broadly speaking, some ways to increase cen-
trality in a network are to increase the number
of interfirm relationships that the firm is part of,
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
or to connect with partner firms who are already
well networked. What this study has done in this
regard is to point out that there may be more or
less opportune times to attempt to reengineer
one’s network. We showed that industry events
may be identified in advance as potentially struc-
ture-reinforcing or potentially structure-loosening.
A practicing manager may use our reasoning and
approach to identify the attributes of specific
industry events that may be facing his/her indus-
try and anticipate its potential impact. A structure-
reinforcing event may be a good time for a firm
that is already central to consolidate its position.
On the other hand, a structure-loosening event
may provide the best context for a peripheral firm
to attempt to improve its position. By pointing out
this aspect of timing, our study offers one element
of a guide to the strategic management of inter-
firm networks. There is also a defensive aspect
that needs to be considered. For example, highly
central firms may need to exercise vigilance after
a structure-loosening event if they want to protect
the investments they have made in current
relationships. USX’s recent attempts to overcome
its perceived technological weaknesses by initiat-
ing an alliance with Nucor illustrates this option
(Business Week, 7 November 1994).
Public policy implications
Since industry networks are acknowledged influ-
ences on industry evolution, our study has
important public policy implications as well. The
analysis of potential network changes provides
one more useful way to assess the potential
impact of governmental intervention in order to
regulate competition. For example, as part of the
recent discussion on relaxing the Glass–Steagall
Act, there may have been value to thinking
through what potential impact such a regulatory
event would have on resource networks in the
banking and financial services industries (e.g.,
New York Times, 19 January 1995). From another
viewpoint, the structure-reinforcing impact of the
regulatory events of 1984 may be interpreted as
providing indirect support for the argument that
dominant firms ‘capture’ the regulatory process
in order to further their own interests (e.g., Laf-
font and Tirole, 1991). Since interfirm networks
are a valuable strategic resource, careful research
on the structural impact of policy decisions could
serve to guide the formulation and implemen-
Networks in Transition
457
tation of public policy that stimulates competition
and protects industrial competitiveness.
CONCLUSION
Our investigation of the strategic alliance network
in the global steel industry has suggested that
industry events may be either structure-reinforcing
or structure-loosening. We have argued further
that their characteristics may allow events to be
identified, in advance, as potentially structure-
reinforcing
or
potentially
structure-loosening.
Identifying the potential impact of an event in
this way clearly signals to managers the direction
of industry evolution, and guides them in initiat-
ing strategic moves.
The
unique
features
of
this
study
are:
(1) examination of network patterns over time;
and (2) contribution and empirical validation of
the ideas of structure-reinforcing and structure-
loosening events. One of the intended goals of
the paper is to initiate a discussion of the dynamic
aspects of networks. While the choice of specific
events (and their impact) will vary from industry
to industry, our study illustrates the utility of an
analysis of how networks are reshaped by indus-
try events. The framework is of very high rel-
evance today, in the context of industries such
as telecommunications, banking, and information
technology, which are both heavily ‘networked’
and regularly buffeted by various ‘shocks.’ An
understanding of how networks evolve over time
is the first step in the progression towards under-
standing if and how networks can become a
strategic resource, potentially subject to mana-
gerial design.
ACKNOWLEDGEMENTS
Research support from the Alfred P. Sloan Foun-
dation under the research program ‘Competi-
tiveness in the Global Steel Industry’ is gratefully
acknowledged. Our thanks are due to Kalpana
Biswas, Joseph Mahoney, Mona Makhija, Anju
Seth, Russell Wright and anonymous reviewers
of this journal for valuable contributions that
helped us in this research.
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
REFERENCES
Abe, E. and Y. Suzuki (1991). Changing Patterns of
International Rivalry: Some Lessons from the Steel
Industry. University of Tokyo Press, Tokyo.
Allison, G. (1971). Essence of Decision: Explaining
the Cuban Missile Crisis. Little, Brown & Co.,
Boston, MA.
Baker, W.E. (1993). Networking Smart: How Managers
Build Relationships for Personal and Organizational
Success. McGraw Hill, New York.
Barley, S. R. (1986). ‘Technology as an occasion for
structuring: Evidence from observations of CT scan-
ners and the social order of radiology departments’,
Administrative Science Quarterly, 21, pp. 78–108.
Barley, S. R. (1988). ‘Actions, institutions, and techni-
cal change in radiology: Towards a role-based theory
of technology and the social organization of work’,
unpublished manuscript, School of Industrial and
Labor Relations, Cornell University.
Barnes, J. A. and F. Harary (1983). ‘Graph theory in
network analysis’, Social Networks, 5(2), pp. 235–
244.
Bower, J. L. and C. M. Christensen (1995). ‘Disruptive
technologies: Catching the wave’, Harvard Business
Review, 73(1), pp. 43–53.
Burkhardt, M. E. and D. J. Brass (1990). ‘Changing
patterns or patterns of change: The effect of a
change in technology on social network structure
and power’, Administrative Science Quarterly, 35,
pp. 104–127.
Burt, R. S. (1976). ‘Position in networks’, Social
Forces, 55, pp. 93–122.
Burt, R. S. (1988). ‘The stability of American markets’,
American Journal of Sociology, 93, pp. 356–395.
Burt, R. S. (1991). Structure: A General Purpose Net-
work Analysis Program, Version 4.2. Columbia Uni-
versity, New York.
Burt, R. S. (1992). Structural Holes: The Social Struc-
ture of Competition. Harvard University Press, Cam-
bridge, MA.
Business Week (7 November 1994). ‘The odd couple
of steel’, pp. 106–108.
Christensen, C. and R. S. Rosenbloom (1991). ‘Con-
tinuous casting investments at USX Corporation’,
Inter-collegiate Case Clearinghouse 9-391-121. Har-
vard Business School, Boston, MA.
Contractor, F. and P. Lorange (1988). ‘Competition vs.
cooperation: A benefit/cost framework for choosing
between fully-owned investments and cooperative
relationships’, Management International Review,
Special Issue, pp. 5–18.
Cook, K. (1987). Social Exchange Theory. Sage,
Beverly Hills, CA.
Doreian, P. (1986). ‘On the evolution of group and
network structure II: Structures within structure’,
Social Networks, 8, pp. 33–64.
Electronics Times (17 June 1993). ‘T.V. talks go inter-
active: Time Warner will test interactive TV ser-
vice’, p. 6.
Freeman, J. and S. R. Barley (1990). ‘The strategic
analysis of inter-organizational relations in biotech-
458
R. Madhavan, B. R. Koka and J. E. Prescott
nology’. In R. Loveridge and M. Pitt (eds.), The
Strategic Management of Technological Innovation.
Wiley, Chichester, pp. 127–156.
Freeman, L. (1979). ‘Centrality in social networks:
Conceptual
clarification’,
Social
Networks,
1,
pp. 215–239.
Freeman, L. C., S. P. Borgatti and D. R. White (1991).
‘Centrality
in
valued
graphs:
A
measure
of
betweenness based on network flows’, Social Net-
works, 13, pp. 141–154.
Galaskiewicz, J. (1979). Exchange Networks and Com-
munity Relations. Sage, Beverly Hills, CA.
Galbraith, J. R. and R. K. Kazanjian (1986). Strategy
Implementation: Structure, Systems and Process.
West Publishing, St. Paul, MN.
Gerlach, M. (1992). ‘The Japanese corporate network:
A blockmodel analysis’, Administrative Science
Quarterly, 37, pp. 105–139.
Ghemawat, P. (1993). ‘Commitment to a process inno-
vation: Nucor, USX and thin slab casting’, Journal
of Economics and Management Strategy, 2(1),
pp. 135–161.
Glasmeier, A. (1991). ‘Technological discontinuities
and flexible production networks’, Research Policy,
20(5), pp. 469–485.
Grant, J., R. Ahlbrandt, F. Giarratani, and J. E. Prescott
(forthcoming). ‘Restructuring for competition in the
global steel industry’, Long Range Planning.
Henderson, R. M. and K. B. Clark (1990). ‘Architec-
tural innovation’, Administrative Science Quarterly,
35, pp. 9–30.
Hogan, T. (1991). Global Steel in the 1990s: Growth
or Decline? Lexington Books, Lexington, MA.
Knoedler, J. T. (1993). ‘Market structure, industrial
research, and consumers of innovation: Forging
backward linkages to research in the turn-of-the-
century
U.S.
steel
industry’,
Business
History
Review, 67, pp. 98–139.
Krackhardt, D. (1990). ‘Assessing the political land-
scape: Structure, power and cognition in organi-
zations’,
Administrative
Science
Quarterly,
35,
pp. 342–369.
Krackhardt, D. (1992). ‘The strength of strong ties:
The importance of Philos in organizations’. In N.
Nohria and R. Eccles (eds.), Networks and Organi-
zations: Structure, Form and Action. Harvard Busi-
ness School Press, Boston, MA, pp. 216–239.
Laffont, J. and J. Tirole (1991). ‘The politics of govern-
ment decision making: A theory of regulatory cap-
ture’,
Quarterly
Journal
of
Economics,
106,
pp. 1089–1127.
Laumann, E. O., J. Galaskiewicz and P. Marsden
(1978). ‘Community structure as interorganizational
linkages’. In A. Inkeles (ed.), Annual Review of
Sociology, Vol. 4. Annual Reviews, Palo Alto, CA,
pp. 454–484.
Laumann, E. O., P. V. Marsden and D. Prensky (1983).
‘The boundary specification problem in network
analysis’. In R. S. Burt and M. J. Minor (eds.),
Applied Network Analysis. Sage, Beverly Hills, CA,
pp. 18–34.
Lawrence, P. R. and J. R. Lorsch (1967). Organizations
and Environments. Harvard Business School Press,
Boston, MA.
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
Lorrain, F. and H. C. White (1971). ‘Structural equiva-
lence of individuals in social networks’, Journal of
Mathematical Sociology, 1, pp. 49–80.
Madhavan, R. and J. E. Prescott (1995). ‘Market value
impact of joint ventures: The effect of industry
information processing load’, Academy of Manage-
ment Journal, 38(3), pp. 900–915.
McGee J. and H. Thomas (1986). ‘Strategic groups:
Theory, research and methodology’, Strategic Man-
agement Journal, 7(2), pp. 141–160.
New York Times (29 November 1994). ‘U.S. seeks to
loosen bank rules: Comptroller of the currency pro-
poses new rules that would allow banks to enter
new businesses, including trading securities’, p. C1.
New York Times (19 January 1995). ‘Annuity sales
by banks upheld: High Court decision is blow to
insurers’, p. C1.
Nohria, N. (1992). ‘Introduction: Is the network per-
spective a useful way of studying organizations?’.
In N. Nohria and R. Eccles (eds.), Networks and
Organizations: Structure, Form and Action. Harvard
Business School Press, Boston, MA, pp. 1–22.
Nohria, N. and R. Eccles (1992). Networks and Organi-
zations: Structure, Form and Action. Harvard Busi-
ness School Press, Boston, MA.
Nohria, N. and C. Garcia-Pont (1991). ‘Global strategic
linkages and industry structure’, Strategic Manage-
ment Journal, Summer Special Issue, 12, pp. 105–
124.
Pfeffer, J. (1972). ‘Merger as a response to organi-
zational interdependence’, Administrative Science
Quarterly, 17(3), pp. 382–394.
Pfeffer, J. (1981). Power in Organizations. Harper &
Row, New York.
Pfeffer, J. and G. R. Salancik (1978). The External
Control of Organizations: A Resource Dependence
Perspective. Harper & Row, New York.
Piore, M. and C. F. Sabel (1984). The Second Industrial
Divide. Basic Books, New York.
Porter, M. E. (1980). Competitive Strategy: Techniques
for Analyzing Industries and Competitors. Free
Press, New York.
Rogers, E. M. (1971). Communication of Innovations.
Free Press, New York.
Sanil, A., D. Banks and K. Carley (1995). ‘Models for
evolving fixed node networks: Model fitting and
model testing’, Social Networks, 17, pp. 65–81.
Schott, T. (1991). ‘Network models’. In R. S. Burt
(ed.), Structure: A General Purpose Network Analy-
sis Program, Version 4.2. Columbia University, New
York, pp. 107–206.
Scott, J. (1991). Social Network Analysis: A Handbook.
Sage, London.
Shan, W., G. Walker and B. Kogut (1994). ‘Inter-firm
cooperation and startup innovation in the biotechnol-
ogy industry’, Strategic Management Journal, 15(5),
pp. 387–394.
Snijders, T. A. B. (1990). ‘Testing for change in a
digraph at two time points,’ Social Networks, 12,
pp. 339–373.
Spender, J.-C. (1980). ‘Strategy-making in business’,
unpublished Ph.D. dissertation, Manchester Univer-
sity.
Stopford, J. M. and C. Baden-Fuller (1990). ‘Flexible
Networks in Transition
459
strategies: The key to success in knitwear’, Long
Range Planning, 23(6), pp. 56–62.
Tushman, M. L. and P. Anderson (1986). ‘Technologi-
cal discontinuities and organizational environments’,
Administrative Science Quarterly, 31, pp. 439–465.
Wall Street Journal (16 February 1984). ‘Ruling on
steel merger hurts industry consolidation plans’, p.3.
Wall Street Journal (16 November 1995). ‘German and
French monopolies take on Italy: Sprint, Bell Atlan-
tic join’, p.A14.
Wall Street Journal (21 November 1995). ‘Biotech
companies abandon go-it alone approach’, p.B4.
1998 John Wiley & Sons, Ltd.
Strat. Mgmt J., Vol 19, 439–459 (1998)
Wall Street Journal (29 November 1995). ‘FCC adopts
rules that will toughen foreign entry into U.S. phone
market’, p.A3.
Wall Street Journal (12 December 1995). ‘Apple’s
likely loss may spur cutbacks, merger’, p.A3.
Wellman, B. (1988). ‘Structural analysis: From method
and metaphor to theory and substance’. In B. Well-
man and S. D. Berkowitz (eds.), Social Structures:
A Network Approach. Cambridge University Press,
New York, pp. 19–61.
White, H. C. (1992). Identity and Control. Princeton
University Press, Princeton, NJ.
Wyszukiwarka
Podobne podstrony:
network postion and firm performance organziational returns to collaboration in the biotechnological
Neural networks in non Euclidean metric spaces
Applications and opportunities for ultrasound assisted extraction in the food industry — A review
Drugs in the Music Industry
business group affiliiation and firm performance in a transition economy a focus on ownership voids
exploring the social ledger negative relationship and negative assymetry in social networks in organ
Neural networks in non Euclidean metric spaces
Applications and opportunities for ultrasound assisted extraction in the food industry — A review
Isaac Deutscher Russia in Transition (rtf)
Tijan Davy Harwood in Transition
Family in Transition Ch 01
Kwiek, Marek The University and the Welfare State in Transition Changing Public Services in a Wider
In Transit by Lillybellis
(autyzm) Hadjakhani Et Al , 2005 Anatomical Differences In The Mirror Neuron System And Social Cogn
social networking in the web 2 0 world contents
the role of networks in fundamental organizatioonal change a grounded analysis
więcej podobnych podstron