ABSORPTIVE CAPACITY IN STRATEGIC ALLIANCES: INVESIGATING THE EFFECTS
OF INDIVIDUALS‘ SOCIAL AND HUMAN CAPITAL ON INTER-FIRM LEARNING
Shawn M. Lofstrom
Management and Organization Department
Robert H. Smith School of Business
3340 Van Munching Hall
College Park, MD 20742
Presented at the Organization Science Winter Conference, 2000 in Keystone, Colorado
The last two decades of the 21st Century have witnessed a significant increase in the
frequency and depth of inter-firm collaboration (Bleeke and Ernst, 1991; Contractor and
Lorange, 1988; Fortune, 1992; Hagedoorn, 1995: Hladik, 1985). Increased globalization of
markets and rapid technological change have made it increasingly difficult for firms to develop
internally the skills necessary to compete in many product markets. The received wisdom
suggests that alliances — defined as collaborative arrangements between two firms which
involve the exchange and sharing of multiple resources for the co-development of technologies
or products — are absolutely essential to a company‘s strategy and competitive future (e.g.
Ohmae 1989, Gulati, 1996; Houghton in Fortune, 1992). Consequently, the study of strategic
alliances is a major area of interest that bears significance for contemporary organizations.
A prominent view of strategic alliances suggests that inter-firm collaboration is a
mechanism by which a firm can leverage its skills, acquire new competencies, and learn (e.g.
Kogut, 1989; Hamel, Doz, and Prahalad, 1989; Huber, 1991; Larsson, Bengtsson, Henriksson,
and Sparks, 1998; Lyles, 1988; Powell and Brantley, 1992). For the partnering firms, alliances
represent interfaces with its environment that provide access to valuable external information and
knowledge (Arora and Gambardella, 1990; Powell, Koput and Smith-Doerr, 1996; Teece, 1992).
As such, these arrangements can provide opportunities for firms to assimilate information,
internalize skills, and develop new capabilities. Moreover, research has suggested that social
networks, competencies, and the relative configuration of skills and organizational practices of
the partnering firms can influence the level of learning through alliances (e.g. Hamel, 1991; Lane
and Lubatkin, 1998; Mowery, Oxley and Silverman, 1996; Shan, Walker and Kogut, 1994).
However, although studies have recognized the importance of individuals for alliances
and learning more generally, few studies have incorporated the role of individuals into
explanations for firm learning in alliances. Research has found that the bonds between key
individuals are central mechanisms that initiate alliance formation (e.g. Larson, 1992) and
sustain inter-firm relationships (Seabright, Levinthal and Fichman 1992). Individuals also
embody the knowledge-based resources that evoke problem solving and learning and contribute
the most to a firm‘s ability to utilize information (Allen, 1977; Simon, 1985). Moreover, the
primary basis of the firm‘s ability to capitalize on external information rests on the ability of
individuals to access, assimilate and utilize information (Cohen and Levinthal, 1990; 131).
Despite these insights, researchers of strategic alliances have placed much greater emphasis on
environmental conditions, and organizational level resources, practices and tendencies than
individuals‘ social or human capital as explanations for learning in alliances.
This paper integrates the theory of absorptive capacity (Cohen and Levinthal, 1989;
1990; 1994) and network theory (Burt, 1992; Granovetter, 1973; 1985) to examine the context
embedded within the alliance in order to augment our understanding of what leads to learning in
alliances. The combination of these two perspectives accentuates issues of information access
and assimilation. Network theory highlights the importance individuals‘ social capital that
resides in the relationships that connect key individuals to others as primary resources that effect
access to information. The theory of absorptive capacity specifies the nature of human capital or
knowledge-based resources that allows individuals, and firms through their key employees, to
capitalize on external information and learn through alliances1. Consequently, the framework
suggests that firm‘s differ in their ability to value, assimilate and utilize external information and
their ability to access information. Moreover, these differences influence firm learning in
This paper opens the ‗black box‘ of alliances and systematically investigates the
relationships between variations in individuals‘ networks and knowledge and the firm‘s ability to
learn in alliances. I propose that the networks represented by the ties between key individuals
and their information/advice contacts provide access to information. Moreover, differences in
the strength of these relationships are also critical sources of variation that affect access to
information and the potential for learning. In addition, individuals‘ knowledge influences the
extent to which individuals assimilate and utilize external information (Cohen and Levinthal
1990; 1994). I suggest that differences in the knowledge-based resources held by individuals
influences learning to the extent that it tempers individuals‘ ability to assimilate and utilize new
information. My central proposition is that, in addition to environmental, industry and
organizational factors, the networks and knowledge of individuals have distinct and important
roles for learning through alliances.
Although learning has been conceptualized at many levels, my investigation focuses on firm learning as
represented by the extent to which key individuals involved in the alliances learn. I assume that the extent to which
the key individuals involved in the alliance learn not only influences the ceiling of firm-wide learning as a result of
the alliance, but also represents critical new knowledge which these individuals can apply in other contexts, even if
the organization as a whole is not aware of the knowledge gained. Thus, what individuals learn can benefit the firm
and result in appearance that the firm learned without the lesson being widely shared, understood or internalized.
Theory and Hypotheses
Social Capital: Individuals’ Networks
Bourdieu first defined social capital as ―the aggregate of … resources which are linked to
the possession of a durable network of institutionalized relationships‖ (1985; 248). Consistent
with this definition, researchers have suggested that social capital is derived from both structural
(Burt, 1992; 1997; 1998; Nahapiet and Ghoshal, 1998) and relational dimensions of individuals‘
networks (Coleman 1988; Granovetter, 1973; 1985; Nahapiet and Ghoshal, 1998). Individuals‘
networks, however, vary along these dimensions and these differences influence the value or
level of social capital in their networks. Thus, I suggest that differences in network structure and
relational strength influence the extent to which individuals are able to access and discover new
information, and ultimately learn in ways that can benefit their firm. Specifically, network
theory indicates that networks rich in non-redundant ties provide greater information access
benefits (Burt, 1992). Further, networks rich in strong trusting ties (Granovetter, 1985;
Krackhardt, 1992) also facilitate access to valuable external information and influence the
likelihood that learning occurs within alliances.
Network Structure: Non-redundant Ties
Networks that are rich in non-redundant ties can be defined as networks that link a focal
actor to contacts that are not otherwise connected (Burt, 1992) and provide access to ―valuable
pieces of information‖ (1992: 13). While the value of social capital includes information and
control benefits (Burt, 1992) in this paper I focus on information benefits, and variations in
network configurations believed to influence information access. It follows that the networks of
key individuals‘ that represent different configurations and provide differential access to
information. Moreover, given that new information is needed for learning (Anderson, 1993), the
access benefits derived from these networks are a key source of variation that influences the
extent to learning occurs in alliances.
Burt (1992) suggests that networks characterized by greater non-redundancy in ties
provide access to a greater range of new, unique and different information than networks that
lack non-redundant ties. This access, however, rests on the premise that non-redundant ties tap
into fundamentally different domains or pools of information (Burt, 1992). Non-redundant
networks are more suited to accessing a greater range of information because the disconnection
between contacts implies that individuals are connected to unique, novel and non-overlapping
sources of information. In contrast, networks with redundant ties link the individual to contacts
that ―run in the same social circle‖, belong to a common pool or domain of information and
circulate the same information. As a result, the networks characterized by redundant ties are
likely to provide similar or overlapping information while non-redundant network ties provide
new or novel information.
Since learning requires access to new information (Simon, 1991; Anderson, 1993), the
networks through which each individual gains access to information influence the chances that
the individual, and consequently the organization will learn. Redundant network structures,
because they limit access to new information, limit the level of learning that can occur. In
contrast, non-redundant network structures in the individual‘s networks tend to involve access to
and exchange of a greater diversity of ideas and information that can facilitate learning.
H1 Network non-redundancy is positively related to learning.
Network Relational Dimensions: Trust
Granovetter (1985) suggested that a primary value of social networks stems from the
strength of network ties. Strong ties can be defined in terms of the ―combination of the amount
of time, emotional intensity, mutual confiding, and reciprocal services‖ (Granovetter 1973; 1361)
in a relationship between two actors. Moreover, as interaction develops between actors,
―…economic relationships … become overlaid with social content that carries strong
expectations of trust and abstention from opportunism,‖ (Granovetter, 1985: 113).
Although virtually every exchange has some degree of trust (Arrow, 1962), trust defined
as the expectation that individuals will fulfill obligations in predictable, fair and reliable ways
(Anderson and Weitz, 1989; Cummings and Bromiley, 1996), has important effects on alliances,
and in particular information access. The benefits of trust have been variously characterized as
lowering transaction costs (Dyer, 1996), reducing the need for formal contracts (Larsen, 1992),
making it easier to adapt to changing circumstances (Doz and Hamel, 1998) and facilitating
conflict resolution (Ring and Van de Ven, 1994). Most important for this discussion is the belief
that trust provides access to information. Trust between individuals reduces or eliminates the
risk that individuals withhold information and increases the willingness of individuals to take
risk, thereby allowing open exchanges of information (Nahapiet, 1996; Ring and Van de Ven,
1992; Starbuck, 1992; Zaheer, McEvily and Perrone, 1998). Specifically, research suggests that
trust stimulates collaboration, promotes cooperation and facilitates access to tacit (Axelrod,
1984; Doz and Hamel, 1998; Krackhardt, 1992; Nelson and Winter, 1982; Uzzi, 1996). The
implication of trust is that individuals are more likely to exchange deeper information and
experiment with new combinations (Nahapiet and Ghoshal, 1998). Such information is often
embedded in firm specific routines or specialized knowledge that is difficult to articulate
knowledge but can be critical for learning (Cohen and Levinthal, 1990; Polyani, 1967; VanLehn,
1990). In other words, trust encourages deeper access to information and because these
exchanges facilitate learning, trust increases the level of learning.
In summary, trust brings individuals together and increases the chances that individuals
access, exchange and share information. A lack of trust in advice or information contacts
reduces the likelihood that individuals will access and exchange information that is often critical
for learning to take place. In contrast, higher levels of trust in these contacts increases
individuals‘ access to information and increases the chances that individuals will learn.
H2 Strong ties (trust in one’s contacts) is positively related to firm learning.
Social Capital: Individuals’ Knowledge
Learning, however, requires more than access to information. Building on research of
problem solving and cognition at the individual level (e.g. Bower and Hilgard, 1981), Cohen and
Levinthal (1989, 1990, 1994) have suggested that firms differ in their ability to value, assimilate
and utilize external information. Specifically, this ability or absorptive capacity depends on the
cumulative experience within the firm and the extent to which this knowledge is related to
external information. This theory of absorptive capacity highlights two important criteria of
knowledge – it‘s cumulative nature points to issues of expertise or competence and the relevance
of internal knowledge suggests that a firm‘s knowledge must be complementary to the external
Previous research in other contexts (e.g. Allen, 1977) suggest that a large portion of a
firm‘s knowledge or human capital is accumulated through experience, originates and is applied
in the minds of individuals and becomes embedded in the capabilities and practices individuals
use to accomplish tasks. Each individual‘s knowledge constitutes what the individual knows as a
result of his or her specific experiences that are likely to vary. Further, the knowledge of
individuals influences their ability to utilize external information (Anderson, 1993; Newell and
Simon, 1972; Weick, 1979). The implication is that differences in the human capital allocated
and embodied in individuals are sources of variability that influence learning in the alliance.
Studies have found that the ability to learn from others depends upon the similarity of the
knowledge bases involved (Boisot, 1995; Campbell, 1969). Related knowledge facilitates the
internalization of new information because the basis of relatedness provides common rules for
communication and facilitates the exchange of ideas and information (Boland and Tenkasi, 1995;
Dearborn and Simon, 1958; Cohen and Levinthal, 1990; Lane and Lubatkin, 1998). Related
knowledge ensures that individual are able to recognize and assess the value of new information
and eases the process of assimilating new information or learning. However, at the same time,
diversity of knowledge is critical element that allows knowledge to advance and learning to
occur (Nonaka and Takeuchi, 1995). Diverse knowledge ensures that new information, different
opinions, and new insight is available. Further, when different areas of expertise interact the
chances for making new combinations and learning increase (Cohen and Levinthal, 1990).
The ideal structure of knowledge in an individual‘s network ―should reflect only partially
overlapping knowledge complemented by non-overlapping diverse knowledge (Cohen and
Levinthal, 1990: 134). The presence of only related knowledge ensures that individuals will
communicate, but limits the opportunities for learning. In contrast, knowledge diversity by itself
makes new and different ideas available but limits opportunities for learning because
communication is stifled. I use the term ‗knowledge complementarity‘ to refer to the extent to
the knowledge of individuals is related to and at the same time is different from the knowledge
of contacts in their information/advice networks. The presence of knowledge complementarity
provides for communication, the exchange of diverse information and increases the chances that
learning will occur. Consequently, when individuals‘ have higher levels of knowledge
complementarity with information/advice contacts learning is more likely to occur than when
knowledge complementarity is lacking.
H3 Knowledge complementarity is positively related to firm learning.
A second predictor of individuals‘ ability to utilize new knowledge is their own level of
human capital or expertise within a specific domain. Expertise can be defined, generally, as the
extent to which an individual understands a particular domain of knowledge. High levels of
expertise develop over a long period of intense involvement in or commitment to an area and
yields a comprehensive understanding of why and how something works or is done (de Groot,
1978; Simon, 1990). This level of understanding produces an inventory of knowledge regarding
solutions that ‗work‘ for familiar problem types and skills for solving technical problems
(Anderson, 1993). Put differently, individuals with high levels of expertise are more likely to
understand with great familiarity the laws, logic and rationale underlying the function or process
of a specific domain. This familiarity provides individuals with the ability to identify critical
configurations or ―complexes‖ that contain several pieces of information, including the
relationship between different elements and information about the solution in a complex situation
(Bohn, 1994; Camerer and Johnson, 1991: 23; de Groot, 1978). Thus, they develop deep
knowledge structures that allow them to more fully comprehend and more quickly make
connections between their knowledge and new, external information.
‗Experts‘ or individuals with higher levels of competence are more suitably skilled for
integrating knowledge and information than individuals with less experience and expertise. High
levels of expertise enable individuals to articulate their knowledge and beliefs about the
processes driving performance and to think creatively and critically about problems (Nonaka,
1994). As a result, individuals with higher levels of expertise find it easier to assimilate and
apply new information to solve problems more quickly and effectively (e.g. Singley and
Anderson, 1989). In contrast, individuals with lower levels of expertise lack the understanding
that enables effective problem solving and learning. As a result, when the firm assigns
individuals with higher levels of expertise to work in the alliance it is more likely that learning
will occur then when these individuals have less expertise.
H4 Individuals’ expertise is positively related to firm learning.
The sample for this study was drawn from the population of alliances that focus on the
development or extension of medical device technologies and were formed between U.S. private
and public companies from January 1989 and January 19982. Ward‘s Business Directory of U.S.
Private and Public Companies and the Medical Device Register (a comprehensive business-to-
business reference database on medical-related companies) were used to identify a complete list
The Food and Drug Administration (FDA) defines a medical device as a tool intended for use in the diagnosis,
cure, mitigation, treatment, or prevention of disease to affect the structure or function of the body. Medical devices
include in vitro diagnostic products (e.g. catheters, reagents and bio-markers), radiation-emitting products (e.g.
of companies in the biotechnology, medical device and pharmaceutical industries3. Industry
sources including The Grey Sheet, Windhover‘s Pharmaceutical Strategic Alliance database,
Recombinant Capital‘s database, as well as standard public sources including Lexus-Nexus and
company web pages were then searched to identify and triangulate information on the alliances
these companies formed during the past ten years. Alliances were included only if they involved
the development of medical device technologies focused on human therapeutics and diagnostics.
Following the practice of other researchers, alliances focused solely on marketing or distribution
and ‗as-is‘ license exchanges, and joint ventures which establish a separate legal entity were
excluded to hold constant issues related to technology or industry (e.g. Lane and Lubatkin, 1998;
Shan, Walker & Kogut, 1994). These search efforts identified 571 alliances formed among 533
companies during the period from January 1989 through January 1998. From this population,
212 alliances were selected such that each alliance involved two unique partners, thus ensuring
independence among the firms involved.
Research Design and Data Collection
The primary data source used to collect the data supporting this study were
questionnaires mailed to key individuals involved in these alliances. Key individuals include
scientists, physicians, and engineers that are central to the alliance effort. These individuals were
identified from the various public sources mentioned above and company directories and
contacted by telephone to verify the appropriateness of the alliance, the individual‘s role in the
alliance, and to obtain their agreement to participate in the study. Confidentiality of responses
was emphasized in these conversations, and each company received an explicit nondisclosure
The survey used in this study draws upon existing measures and where these were not
available, draws upon the literature for the development of new measurement scales. A
preliminary version of the survey was pre-tested by thirty individuals who have been involved in
medical technology alliances but were not included in the set of alliances selected for this study.
ultrasound products, x-ray machines and medical lasers) and implantible devices (e.g. pacemakers, transgenic organs
and tissues, and artificial tissues).
These industries include the four digit Standard Industrial Classification (SIC) codes of 3841, 3842, 3843, 3844,
3845 2834, 2835, 2836, and 8731, representing the areas of surgical and medical instruments, surgical appliances
and supplies, dental equipment and supplies, x-ray apparatus and tubes, and electromedical equipment,
pharmaceutical preparations, diagnostic substances, biological products except diagnostics and companies in
commercial physical research.
Feedback from these individuals and comments from several colleagues versed in the art of
survey design were incorporated into the final version of the survey.
The final survey was mailed to 262 key individuals that agreed to participate in the study.
The individuals receiving the survey included people from both sides of 90 alliance dyads and
these individuals were asked to act as key informants on their respective firms and the alliance of
interest 4. The number of key individuals surveyed per company ranged from two to four. This
approach, of surveying individuals prominently involved in the activity of interest, is consistent
with the selection of key informants based on their knowledge by virtue of position (John and
Weitz, 1988). To maximize the response rate in this study, I adopted Dilman‘s (1978) ‗total
design method‘ for this study. In addition to the initial telephone contact and discussion of
confidentiality, extensive follow-up steps were taken. Specifically, 1) reminder/thank you cards
were sent two and four weeks after the initial mailing, 2) additional surveys were sent to non-
respondents, and 3) the remaining non-respondents were contacted by telephone.
A total of 215 key individuals returned complete and usable surveys. These responses
come from individuals at each partnering firm involved in 82 alliances. The responses and final
data set, therefore, are based on the complete and useable surveys from at least one key
individual from each partner on 82 alliances. This number is approximately 39% of the original
212 alliances selected and the corresponding 424 firms identified for this study. Based on the
number of key individuals who had agreed to participate in the study and received a survey, the
response rate equals approximately 82%. The companies involved in these alliances include 82
medical device companies, 51 pharmaceutical firms, and 31 biotechnology firms. On average,
these firms were founded since 1974 and have been involved in five alliances. These companies
reported 1997 sales in medical devices that range from $100,000 to $14 billion and employed
between three and 54,000 employees. As reported by the respondents, all of the alliances
involved less than 50 individuals, 75% involve less than 25 individuals and on average, the
alliances involve five participants from each firm.
Testing for nonresponse bias
Although the response rate of 39% is acceptable relative to similar studies, the possibility
remains that the sample of alliances and firms represented in the study differ systematically from
It was not uncommon for individuals to decline participation in the study. The most frequent reason stated related
to issues of confidentiality or time constraint. This study sought to get participants from both sides of the alliances,
and individuals from one firm that participated in 128 alliances declined the invitation to be included in this study.
those declining to participate. In order to assess this potential bias, I examined potential
differences in alliances represented by participating firms versus alliances not represented in the
study. Specifically, I compared the represented alliances to the non-represented alliances in
terms of the mix of alliance partner companies (Medical Device, Biotechnology and
Pharmaceuticals). I also conducted analysis to test for nonresponse bias at the company level, to
ensure that the firms participating in the study are not systematically different from other firms
that might have participated in this study. These comparisons are based on prior organizational
alliance experience, company size5 and sales. The differences between participants and non-
participants are evaluated using t-tests to compare the mean value on each item.
The results of these comparisons revealed no statistically significant differences between
the typical pairing of companies (t = - 1.49) for the alliances included in this study versus those
which are not represented. The comparisons at the firm level also show no statistical differences
in mean allying experience (t = 1.29), mean sales (t = 0.62) or mean size (t = 0.89) when
comparing the participating companies relative to 150 randomly selected firms involved in 75
alliances not represented in this study. This analysis suggests that the alliances and the
companies involved in them are representative of the population of firms participating in medical
technology alliances. Thus, the threat of non-response bias is limited and there is some
assurance that the sample of alliances participating in this study are representative of the
alliances which declined participation in the study.
In this study, I use a multi-measure approach. Whenever possible, measurement
instruments available from existing research were used to operationalize the theoretical
constructs. In some cases, these instruments were modified to make them more suitable for the
present research setting. Where instruments were not available, new measurement instruments
were developed based on the literature and reviewed for content validity by a panel of colleagues
and included in the pretest of the survey. Table 1 presents for each construct, the details of the
measurement instruments, the scales used, and the Cronbach reliabilities, and the results of a
factor analysis (discussed later in the section on construct validity). Each of the constructs were
at or above the value of 0.70 (Nunnally, 1978).
Size is measured as the number of employees.
insert Table 1 about here
The dependent variable of this study is firm learning or the extent to which individuals
gain knowledge that can benefit the firm. Firm learning in alliances has been measured by the
post hoc production of patents or expert assessment (Mowery, Oxley and Silverman, 1996; Lane
and Lubatkin, 1998). In this study, learning was assessed through a survey measurement
instrument developed to capture the level of learning that occurred which key individuals felt
was beneficial to their firm. Two items, shown in Table 1, were developed, modified according
to pretest comments, and included in the final survey to key individuals. The Cronbach
reliability for these items is .73.
The structure of networks captured by non-redundancy draws on the work of Burt (1992)
which highlights the importance of non-overlapping ties for accessing and discovering new
information and opportunities. To operationalize network non-redundancy, this study uses a
modified version of a measurement instrument developed by Aldrich, Rosen and Woodward
(1986) which has also been used in the study of firm level networks (McEvily and Zaheer, 1999).
The instrument asks respondents to identify up to nine individuals to whom they go to for advice
or information related to the alliance efforts. Respondents were asked to include the a) three
most important individuals from their own firm, b) the three most important individuals from the
partnering firm, c) the three most important individuals from outside either of the partnering
firms that they go to for information or advice related to the alliance6.
The inclusion of this third group draws from previous studies that suggest that this third
group is not only commonly used, but also important given the technologically intensive basis of
the alliances studied. For example, researchers on innovation have noted that it is quite common
for engineers and scientists to form networks and exchange technical information as the result of
With a limit of three key contacts in each group, there may be a bias toward listing three contacts in each area. In
fact, less than 25% of the respondents listed three contacts for each area. Approximately 80% of the time
respondents listed three individuals from their own firm, 67% of the time they listed two individuals from the
partnering firm. Contacts to individuals outside the partnering firms were fairly common. In approximately 58% of
the time they listed one contact, and 45% of the time they listed two contacts from outside the partnering firms.
conferences, publications, and prior training or employment (Almeida and Kogut, 1999;
Freeman, 1991; Liebeskind, 1996: Frost, 1995). Moreover, checking with external specialists is
particularly important and individuals commonly consult experts or contacts in specific scientific
communities to evaluate the limits and validity of developments (Zolo, 1989).
To measure whether these contacts are connected, the instrument then asks respondents
to report whether each of these contacts know one another. In network theory terms, this
instrument produces ego-centered network data. Using the matrix of information reported by
respondents about whether the individuals they go to for information or advice concerning the
alliance know each other, a non-redundancy score is computed for each alter7. This score
reflects the extent to which each alter identified in the ego-network provides ego with new or
unique information that is not available from others in the network. Specifically, for each alter
listed in the ego network,
Non-redundancy = 1- ((Alter‘s Actual Ties) / (Number of Advisors –1))
Alter‘s Actual Ties = the total number of ties each Alter has to other alters listed (0 to
Number of advisors = total number of advisors listed (0 to 9)
This computation defines non-redundancy as the extent to which each alter is not
connected to the other information/advice contacts in ego‘s network. The mean of the scores for
each alter produces a score which reflects, on average, the extent to which ego will have access
to non-redundant or novel information. This aggregated score, because it is a function of the
pairwise ties between actors or contacts in the ego network, it is similar to a network density
measure (Wasserman and Faust, 1994). The values for this variable range from 0 to 1 with lower
numbers indicating greater redundancy and higher numbers indicating higher levels of novelty or
non-redundancy. In the case where the alters all know one another, the score equals 0 (there is
complete redundancy). In contrast, if none of the respondent‘s contacts know one another, the
non-redundancy score is 1, reflecting a network of ties that is entirely non-overlapping. The
Cronbach reliability of the non-redundancy measurement instrument is not reported since this
construct is represented with a single variable.
Trust is defined as the expectation that specific individuals (i) make good faith efforts to
behave in accordance with commitments, both explicit and implicit, (ii) are honest in whatever
negotiations proceeded such commitments, and (iii) do not take excessive advantage of another
even when the opportunity is available (MacAllister, 1995; Cummings and Bromiley, 1996).
These dimensions of trust (reliability, predictability and fairness) are consistent with those used
in other inter-organizational studies (e.g. Zaheer, McEvily and Perrone, 1998). To operationalize
this construct, five measurement items are adapted from Cumming‘s and Bromiley‘s (1996)
scale on trust. The mean score of these items are used to reflect the level of trust. The Cronbach
reliability for these items is .90.
In this study I use the term ‗knowledge complementarity‘ to refer to the extent to which
the knowledge in each individual‘s network is both related and diverse across the network of
contacts. To operationalize this construct, five new measurement items to were developed to
capture the level of complementary knowledge. These items were reviewed by colleagues and
examined for content validity, and included in the pretest of the survey. Comments from these
two groups suggested minor changes to the wording of these items. These resulting items ask
the respondent to assess whether they perceive each individual listed in their ego network to have
backgrounds and training in functional and technical areas that are similar or different to their
own. In order to capture the simultaneous presence and level of similarity and diversity of
knowledge across each individual‘s contact network, the composite variable used for this study is
the product of the mean (to capture the level of similarity) and the deviation (to capture the
diversity as represented by the variability or distribution of similarity-difference) across each
This computation of knowledge complementarity has the effect of assigning lower scores
where knowledge relatedness is on average quite similar and little diverse or range across these
scores. Thus, low scores are given when the knowledge of all contacts is very related or similar
and there is little deviation across the level of similarity in knowledge. The highest scores are
given to cases where there is less knowledge relatedness on average but higher diversity or range
in the knowledge relatedness across the network contacts. The Cronbach reliability for these
items is .85.
Since larger networks may correspond to greater opportunity for non-redundancy, the measure of non-redundancy
is normalized by the size of the network (number of contacts listed).
Several alternative measures were examined including a score based simply on averages,
Expertise can be defined as the perception that an actor possesses a certain degree
of specialized knowledge or technical proficiency with respect to a specific discipline or applied
field (Barber, 1983). For this study, I adapted five items from Barber‘s (1983) scale on
competence to reflect the level of expertise of individuals from the focal firm. The mean score
produces the variable used in this study. The Cronbach reliability for these items is .87.
Included in this study and shown in Table 2 are seven control variables that previous
research suggests influence learning or alliance processes.
insert Table 2 about here
Researchers (Lane and Lubatkin, 1998) have suggested that firm learning is negatively
influenced by asymmetries in partner organizational structures or positively influenced by
similarities in the partners‘ ―dominant logics‖ (Prahalad and Bettis, 1986). As a proxy to capture
differences in organizational structure, I include a measure of firm size (measured as the number
of employees) relative to the size of its partner firm. This measure is computed as the ratio of the
focal firm‘s number of employees to the partner firm‘s number of employees. In the analysis, I
use a standardized measure for relative firm size since this item does not produce a normal
distribution. To control for differences in logics, I include a dichotomous measure of industry
similarity to reflect whether the firm and its partner are from the same (value = 1) or different
(value = 0) industry, based on four digit SIC code groupings. A third control variable related to
learning is a measure to capture how important the alliance is to the focal firm. Similar to the
notion that innovation search and adaptation is often triggered in response to failure of the firm
to reach its aspiration level (e.g Cohen and Levinthal, 1989; Lant, 1990), this control variable is
intended to capture and control for increased learning because the alliance is vital to the
Previous research on alliance success also suggests that environmental factors, prior
allying experience, the age of the alliance, and expectations that the alliance will continue (i.e.
the shadow of the future) are also related to alliance outcomes (e.g. Axelrod, 1984; Harrigan,
1986; Gulati, 1995; Heide and Miner, 1992; Parkhe, 1993). To control for these effects, I
and one based on the standard deviation of scores within each network. The computation used proved to be the
include a single item to capture each of these areas. For the analysis, I use a standardized score
for alliance age, prior allying experience, the shadow of the future since this data is not normally
In order to assess the discriminant properties of the measures across the constructs used
in this research and convergent validity of the multiple items used for each construct, I conducted
a confirmatory factor analysis of the items used to indicate the dependent and independent
variables. This analysis used the principal component method of extraction, varimax rotation
with Kaiser normalization, and an unconstrained factor solution. Table 1 presents the item
commonalties, the Eigenvalue, percent and cumulative variance explained from this factor
analysis. These results provide support for a multiple-factor solution with each of the indicators
loading strongly on those factors predicted by measurement theory.
This research also relies on ego-centered network data as described above. The validity
of the network based constructs relies on the respondent‘s ability to assess accurately the ties
between pairs of alters (Krackhardt, 1996). In order to validate the non-redundancy construct
used in this study, I conducted two tests on two randomly selected sets of ego-networks. In the
first test to validate these measures, 15 surveys from different alliances were randomly selected
of which five respondents gave permission to speak with their listed alters. Calls to the 25 alters
listed by these respondents confirmed with 100% accuracy the ties among alters as reported by
the five respondents.
A second step to verify a wider portion of the data involved randomly selecting 25
alliances and comparing all of the returned surveys related to each alliance for consistency of
reports about connections between alters when respondents had common alters listed. When
there were alters common to more than one respondent (the overlap averaged 76% across the
alliances), the consistency of reports of a tie or no tie averaged between 92% and 97%. Based on
these analyses, it is reasonable to conclude that the measure of non-redundancy is acceptable.
Results of Hypotheses Tests
Multiple regression analysis was used to test the hypotheses developed above. The
analysis included estimating the models based on three different sets of firms. In other studies
there is mixed guidance as to whether the factors contributing to learning have similar effects for
most robust and is consistent with the theoretical argument presented.
both partner firms. Hamel (1991) and Mowery (1988) have suggested that firm level differences
will lead to different rates of learning. In contrast, Lane and Lubatkin (1998) make the
assumption that learning in alliances is symmetrical: ―the factors effecting one-way learning also
effect two-way learning‖(1998; 464). The analysis presented here includes the results for 1) the
full set of 162 firms, 2) the larger firm in each alliance (based on number of employees), and 3)
the smaller firm in each alliance (based on number of employees).9
Table 3, 4 and 5 present the descriptive statistics and zero-order correlations among the
constructs for the results based on the full set of firms, the larger firms, and the smaller firms,
respectively. Table 6 presents the results of the regression analysis, with results based on the full
set of firms in the first column, results for the larger firms in the second column, and results for
the smaller firms in the third column. Based on the adjusted R2 for each sample (full set R2 = .45,
larger partners R2 = .58, smaller partner R2 = .29), the variables in the model account for a
considerable portion of the variance in the level of learning that can be used to benefit the firm,
although the model is less robust for smaller partners. Further, across the three samples, there is
statistically significant support for the model, with the exception of the one hypothesized
relationships when examining the smaller partners. However, contrary to other studies on
learning in alliances and alliance success, only the control variable the is significant is alliance
importance, and then this only matters for the full sample and larger partners.
insert Table 3,4,5 and 6 about here
In terms of individuals‘ social capital, both structural and relational network dimensions
are important predictors of learning. Hypothesis 1 is supported for the combined set of firms (
= .200, t = 3.154, p = .002), larger partners ( = .180, t = 2.290, p = .025), and smaller partners
( = .211, t = 2.014, p = .048). These results suggest that the networks rich in non-redundant
ties positively influence learning in alliances but this effect is less significant for smaller
partners. Trust in the contact ties also positively influences the level of learning. Hypothesis 2 is
supported and the effect of trust on learning is nearly equally significant for all partners
With the exception of relative size and importance of the alliance, this breakout does not distinguish the firms on
any factor examined in the study. Smaller partners are relatively smaller, and they also view the alliance as more
critical. The size of smaller partners ranges from 4 to 3,700 with a mean of 260 employees. 75% of these firms
have less than 210 employees and 50% have less than 70. In contrast, the larger partners range in size from 10 to
57,000 with a mean of 5523 employees. 75% of these firms, however, have less than 4,625 employees and 50%
have less than 200.
(combined set of firms - = .336, t = 4.965, p = .000; larger partners - = .310, t = 3.198, p =
.002; smaller partners - = .322, t = 3.126, p = .003).
This study also finds partial support for the hypotheses that argued that human capital
contributes to learning in alliances. Knowledge complementarity is positively and significantly
related to firm learning for the combined set of firms ( = .141, t = 2.266, p = .025) and larger
partners ( = .238, t = 2.902, p = .005). However, for smaller partners knowledge
complementarity plays an insignificant role in determining learning ( = .090, t = .897, p =
.373). Hypothesis 4 proposed that the competence of key individuals is an important predictor of
learning. This hypothesis is supported, although it is less statistically significant for smaller
firms (combined set of firms - = .321, t = 4.567, p = .000; larger partners - = .467, t = 4.731,
p = .005; smaller partners = .243, t = 2.172, p = .033).
The results stemming from the control variables provide partial support for the aspiration
rationale that when firms place greater importance on the alliance they are more likely to learn.
The importance of the alliance is positively related to learning for the combined set of firms ( =
.181, t = 2.730, p = .007) larger partners ( = .228, t = 2.749, p = .000) and but not for the
smaller partner ( = .176, t = 1.662, p = .101). In this study, each of the remaining control
variables has an insignificant relationship with learning.
Extending the Analysis
Overall, these results provide support for the framework that brings both individuals'
social capital and human capital into the analysis that predicts firm learning in alliances. For
reasons noted earlier, the contacts that individuals go to for information or advice can include
other individuals at their own firm, at the partnering firm, and individuals outside either partner
firm. However, the analysis to this point does not reveal whether there are any specific patterns
across these specific groups within individuals‘ information/advice networks that are particularly
important for learning. The question remains: What is the value of having network non-
redundancy and knowledge complementarity distributed equally among the contacts of key
individuals? Conversely, should the connections and level of knowledge complementarity to
different groups of information/advice contacts vary in order to contribute to learning in
alliances. More generally, is there any value to using outside contacts?
To explore these questions, additional analyses were conducted. Based on the surveys
returned, the key individuals of 96 firms (47 of the larger partners and 49 of the smaller partners)
utilize external consultants as part of their information/advice networks. To address the question
of value attributed to outside contacts, an analysis of variance was conducted to compare firms
that use outside contacts with those that do not. The results, shown in Table 7, suggest that
higher levels of learning are associated with the inclusion of outside contacts in individuals‘
information/advice networks. In addition, when outside consultants are used, individuals‘
networks to contacts at their own firm and the partner firm have higher levels of non-
redundancy. The key individuals at these firms also report higher levels of expertise for their
contacts at their own firm, view uncertainty to be higher and report that their firm has less prior
organizational alliance experience.
Insert Table 7 about here
The results of the regression analysis to further explore the pattern of network non-
redundancy and knowledge complementarity across groups of contacts is shown in Table 8. As
in the earlier analysis, column 1 reports the results for the combined set of firms, column 2
reports the results for larger partners, and column 3 reports the findings for smaller partners.10
Overall, these models are significant and based on the adjusted R2 for each sample (full set R2 =
.43, larger partners R2 = .51, smaller partner R2 = .36), the variables in the model account for a
considerable portion of the variance in the level of learning.
These results suggest that variations in the levels of non-redundancy for specific groups
of contacts influence learning. The ideal structure of the network varies slightly across smaller
and larger partners, and also varies across the different groups of contacts. Specifically, non-
redundancy in ties to contacts at one‘s own firm is related to learning for the combined set of
firms ( = .333, t = 3.243, p = .002) and for smaller partners ( = .448, t = 2.826, p = .007) but is
only marginally significant for larger partners ( = .273, t = 1.966, p = .057). In contrast, non-
redundancy in ties to contacts at the partner firm is not related statistically to learning for any of
the samples. Moreover, non-redundancy in ties to outside contacts is significant for larger
The control variables that were not significant in the previous analysis are excluded in these regressions.
Analysis, not shown here, did not find any significant relationships between these excluded control variables and the
results are essentially the same as those shown in table 8.
partners ( = .339, t = 3.009, p = .005), but is not predictive of learning for smaller partners ( =
- .111, t = -0.674, p = .504) or the combined set of firms ( = .106, t = 1.197, p = .234).
Insert Table 8 about here
The results also find that the ideal configuration of knowledge complementary is not
consistent across the three groups of network contacts and also varies – to some extent – by
partner size. Knowledge complementarity with contacts at one‘s own firm is positively related
to learning for the combined set ( = .211, t = 2.512, p = .014), and marginally significant for
smaller partners ( = .269, t = 1.995, p = .053), but insignificant for large firms ( = .156, t =
1.384, p = .175). In contrast, knowledge complementarity with contacts at the partner firm is not
significant for any of the samples, and complementarity with outside contacts has a negative and
significant relationship for learning in the combined set ( = -.222, t = -2.554, p = .012) and for
smaller firms ( = -.305, t = -2.069, p = .045), but is insignificant for large firms ( = -.215, t = -
1.886, p = .067).
Before discussing the contributions of this study, there are limitations that must be
recognized. The primary limitation of this study stems from its reliance on survey data. This
study uses a sampling strategy – it includes those individuals who are most critical and involved
in the alliance – and does not include responses from all individuals working on the alliance.
The results of this study might differ with additional respondents for each alliance. An
additional risk includes potential perceptual biases in the measures. One bias which stems from
using a single survey instrument to measure several constructs, a so-called ‗halo effect‘, inflates
the covariance among two variables ―presumably because the respondent is using a common set
of rules or schematic framework to evaluate items or scales that represent conceptually distinct
constructs‖ (Avolio, Yammarino and Bass 1991: 571). To the extent possible, this threat to
internal reliability of the results was addressed by collecting data from multiple respondents.
However, respondents completed survey that included both the dependent and independent
factors, suggesting potential for bias. Given this limitation, the results of this study should be
considered preliminary, and reader should interpret the result in light of the possibility of inflated
The cross-sectional nature of the research design, represents a second limitation in that it
does not capture the dynamics of evolving networks or the dynamics commonly associated with
alliances. The cross-section nature of this study also warrants caution in inferring the direction
of causality among the key constructs. It is possible that the direction of causality is reversed in
some cases. In addition, there may be reciprocal relations among some constructs. For example,
trust may lead to joint action and in the process of jointly working on a task, joint action may
influence interpersonal trust. A longitudinal study design would help to establish the temporal
ordering of such processes.
This study examines a specific group of technology intensive alliances that occur in a
specific industrial sector. Therefore, while the study increases our understanding of alliance
success in this specific context, it may have limited applicability for different types of
relationships including buyer-supplier relationships, joint ventures, and alliances that occur in
less technology intensive settings.
Finally, this study has measured and implicitly equated individuals‘ knowledge based
resources with R&D measures used in previous studies of absorptive capacity. Thus, while the
findings are suggestive, they do not make a direct comparison between individual-based
resources and Cohen and Levinthal‘s firm level R&D indicators of absorptive capacity.
Contributions and Future Research
This study contributes to the management literature by blending network theory (Burt,
1992; Granovetter, 1985) and the theory of absorptive capacity (Cohen and Levinthal, 1990) in
order to suggest critical structural and relational elements of social capital that provide access to
information while also recognizing that human capital contributes to firms‘ ability to utilize this
information. This integration provides an integrative lens that is useful for examining how firms
learn from interfacing with their environment. The empirical analysis lends strong support for
this integrated framework as a parsimonious means of examining how firms learn through their
interfaces with their environment. Moreover, the results suggest that firm learning depends on a
mosaic of connections and resources patterned among key individuals and their
information/advice contacts that span functional, divisional, disciplinary and organizational
This study makes an important contribution to the management literature interested in
issues of capital and the literature on strategic alliances in particular. The model tested supports
both a structural (Burt, 1992) and a relational (Granovetter, 1985) perspective of social capital.
Specifically, the study is supportive of the argument that non-redundant networks access new
information and this access yields learning benefits. However, it also adds a caveat to Burt‘s
(1992) theory of structural holes suggesting that the locus of network non-redundancy has
important effects for learning. This study suggests that the ideal configuration within non-
redundant networks ought to include vary according to different groups of contacts. For both
larger and smaller partners, non-redundancies with the primary contacts at one‘s own firms
influence learning, while this relationship does not hold for contacts at the partner firm, or
For researchers interested in network effects, the results of this study suggest that the
premature sizing of the network to include only the most obvious or primary contacts (contacts at
one‘s own firm and the partner firm) may miss important effects. Moreover, for studies that
investigate networks where there is a potential for cliques, the disaggregation of the network may
uncover important patterns that not only test theoretical propositions, but also allow for
expansion of theory. Future research might examine these variations, and extend this line of
inquiry by including other indications of social capital. This study includes one structural and
one relational measure of social capital, but several other network based measures (e.g. network
closure, frequency of interaction, or multiplexity of ties) along with non-network measures (e.g.
depth of information accessed/exchanged, motivation) would increase our understanding of the
relationship between social capital and learning.
The findings from this study are also consistent with the general notion of Cohen and
Levinthal‘s theory of absorptive capacity that firms differ in their ability to value, assimilate and
utilize new information and these differences depend on the knowledge or human capital of the
firm. However, the framework developed and tested also supports the conclusion that firms also
differ in their ability to access external information. For this study, the results suggest that the
theory of absorptive capacity is better suited for the larger partners than it is for smaller partners
when knowledge complementarity is based on the aggregate source environment. In contrast,
when consideration is given for the multiple layers of knowledge complementarity among
identifiably different sources of information, there is evidence of critical variations. Specifically,
for smaller partners in alliances (and to a smaller and less significant degree the larger partner),
higher levels of knowledge complementarity with outside contacts and (to a lesser extent)
contacts at one‘s own firm, influence alliance learning. Additional variation exists in the
direction of influence. Consistent with the use of outside consultants in uncertainty and rapidly
changing environments (Eisenhardt, 1989) the influence of related and diverse knowledge to
outside contacts is negative. Outside contacts have a positive influence on learning in alliances,
but this effect depends on higher levels of knowledge similarity in the presence of less diversity.
In contrast, knowledge complementarity with contacts at one‘s own firm has a positive influence
on learning. In light of the finding that knowledge complementarity across all contacts has a
positive influence on learning in alliances, these refined results begin to unveil the intricate
pattern of knowledge which is distributed among network contacts that contributes to learning in
This study also finds that learning in alliances is asymmetrical, supporting earlier
explanations of learning in alliances (Grant and Baden-Fuller, 1995; Hamel, 1991; Khanna,
Gulati and Nohria, 1998; Mowery, Oxley and Silverman, 1996). Although the study does not
measure intent, transparency or reciprocity (Hamel, 1991), it is based on only U.S. firms,
suggesting that these variations are not isolated to cross-cultural alliances. Further, it also
suggests that the factors effecting learning for the partner firms are not isomorphic. An
important avenue for future research might build on these findings to understand more fully the
factors that constrain learning in alliances.
The results of this study support a framework that includes social and human capital as
predictors of learning in alliances. This framework integrates network theory and absorptive
capacity to examine the effect of social and human capital on learning in alliances. This study
shows that variations in these resources exist across alliances and these differences matter for
explaining and understanding learning in alliances. Specifically, this research suggests that the
knowledge and networks of key individuals have distinct effect for learning in alliances, when
controlling for other factors. These findings suggest that adding the micro-level factors of
individuals‘ networks and knowledge increases our understanding of what leads to learning in
alliances. Future research on alliances, which inherently couple effects of individuals‘ and their
resources, would benefit from a multi-level approach to the study of alliances.
Aldrich, H., B. Rosen, and W. Woodward. 1986. ―Social Behavior and Entrepreneurial
Networks, Summary,‖ in P. Nystrom and W. H. Starbuck (Eds.) Handbook of
Organizational Design, London: Oxford University Press, 385-408.
Allen, T. 1977. Managing the Flow of Information, MIT Press: Cambridge.
Almeida P. and Kogut B. 1999. ―Localization of Knowledge and the Mobility of Engineers in
Regional Networks,‖ forthcoming in Management Science.
Anderson, J. 1993. ―Problem Solving and Learning,‖ American Psychologist, 48:1, 35-44.
Anderson, E. and B. Weitz 1989. ―Determinants of Continuity in Conventional Industrial
Channel Dyads, ― Marketing Science, 8: 4, 310-323.
Arora, A., and A. Gambardello. 1990, ―Complementarity and External Linkages: The Strategies
of Large Firms in Biotechnology,‖ Journal of Industrial Economics, 38: 361-379.
Arrow, K.J., 1962. ―Economic welfare and the allocation of resources for invention.‖ In R.R.
Nelson (ed.), The Rate and Direction of Inventive Activity: 609-625. Princeton, NJ:
Princeton University Press.
Avolio, B.J., F.J. Yammarino, and B.M. Bass 1991. ―Identifying Common Method Variance
with Data Collected from a Single Source: An Unresolved Sticky Issue,‖ Journal of
Management, 17:3, 571-587.
Axelrod, R. 1984. The Evolution of Cooperation. New York: Basic Books.
Balakrishnan, S. and Koza, M.P. 1993. ―Information Asymmetry, Adverse Selection and Joint
Ventures: Theory and Evidence,‖ Journal of Economic Behavior and Organization, 20,
Barber, B. 1983. The Logic and Limits of Trust, New Brunswick, NJ: Rutgers University Press.
Bleeke. J. and D. Ernst 1991. ―The Way to Win in Cross-Border Alliances,‖ Harvard Business
Review, Nov.-Dec., 27-135.
Bohn, R. 1994. ―Measuring and Managing Technological Knowledge,‖ Sloan Management
Boisot, M. 1995. Information space: A framework for learning in organizations, institutions and
culture. London: Routledge.
Boland, R.J., & Tenkasi, R.V. 1995. ―Perspective making and perspective taking in communities
of knowledge‖, Organization Science, 6: 350-372.
Bourdieu, P. 1985. ―The forms of capital‖. In J. Richardson (Ed.), Handbook of Theory and
Research for the Sociology of Education: 241-58. New York: Greenwood.
Bower, G. and E. Hilgard 1981. Theories of Learning, Englewood Cliffs, NJ: Prentice-Hall.
Burt, R. 1992. ―The Social Structure of Competition,‖ in Norhia and Eccles (Eds.) Networks and
Organizations: Structure, Form and Action, Cambridge, MA; Harvard Business Press, 57-
Burt, R. 1997. ―The Contingent Value of Social Capital,‖ Administrative Science Quarterly,
Burt, R. 1998. The Network Structure of Social Capital. Background paper prepared for the
―Social Networks and Social Capital‖ conference November, 1998.
Camerer, C. F. and E. J. Johnson, 1991. ―The Process-Performance Paradox in Expert Judgment.
How Can Experts Know So Much and Predict So Badly?‖ in K. A. Ericsson and J. Smith
(Eds.), Toward a General Theory of Expertise-Prospects and Limits, Cambridge, MA:
Cambridge University Press. 195-217.
Campbell, D. T. 1969, Ethnocentricisms of Disciplines and the Fish-Scale Model of
Omniscience,‖ in M. Sherif and C. Sherif (Eds.), Interdisciplinary Relationships in the
Social Sciences, Chicago: Aldine, 328-348.
Cohen, W.M., and D.A. Levinthal, 1989. ―Innovation and learning: The two faces of R&D‖.
Economics Journal, 99: 569-596.
Cohen, W.M., and D.A. Levinthal, 1990. ―Absorptive Capacity: A New Perspective on Learning
and Innovation,‖ Administrative Science Quarterly, 25, 128-152.
Cohen, W.M., and D.A. Levinthal, 1994. ―Fortune Favors the Prepared Firm,‖ Management
Science, 40, 227-251.
Coleman, J. 1988. ―Social Capital in the Creation of Human Capital,‖ American Journal of
Sociology, 94, 95-120.
Contractor, F.J. and P. Lorange. 1988. Cooperative Strategies in International Business,
Lexington, MA: Lexington Books.
Cummings, L. and P. Bromiley, 1996. ―The Organizational Trust Inventory,‖ Trust in
Dearborn, R., & Herbert A. Simon. 1958. ―Selective perception in executives.‖ Sociometry, 21
de Groot, A. 1978. Chance and Choice in Chess, The Hauge: Mouton, 2nd Edition.
Dilman, D. 1978. Mail and Telephone Surveys: The Total Design Method, NY: Wiley.
Dore R. 1983. ―Goodwill and the Spirit of Market Capitalism,‖ British Journal of Sociology, 34,
Doz, Y.L., & Hamel, G. (1998). Alliance advantage: The art of creating value through
partnering. Boston, MA: Harvard Business School Press.
Dyer, J.H. 1996b. Does governance matter? Keiretsu alliances and asset specificity as sources of
Japanese competitive advantage. Organization Science, 7:649-666.
Eisenhardt, K. M. 1989. ―Making Fast Strategic Decisions in High-Velocity Environments‖,
Academy of Management Journal, 32, pp. 543-559.
Freeman, C., 1991. ―Networks of Innovators: A Synthesis of Research Issues,‖ Research Policy,
Frost, P. 1995. ―Crossroads,‖ Organizational Studies, 132-143
Fortune. 1992. ―Are Strategic Alliances Working?‖, September 21, 77-78.
Galaskiewicz, J. and A. Zaheer, 1998. ―Networks of Competitive Advantage,‖ in S. Andrews and
D. Knoke (Eds.) Research in the Sociology of Organizations, JAI Press.
Granovetter, M. 1973. ―The Strength of Weak Ties,‖ American Journal of Sociology, 78, 1360-
Granovetter, M. 1985. ―Economic Action and Social Structure: The Theory of Embeddedness,‖
American Journal of Sociology, 91:3, 481-510.
Grant, R.M., and C. Baden-Fuller 1995. ―A Knowledge-Based Theory of Inter-Firm
Collaboration‖, Best Paper Proceedings 1995, Academy of Management, pp. 17-21.
Gulati, R. 1995. ―Does Familiarity Breed Trust? The Implications of Repeated Ties for
Contractual Choice in Alliances,‖ Academy of Management Journal, 38, 85-112.
Gulati, R. 1996. ―Social Structure and Alliance Formation Patterns: A Longitudinal Analysis,‖
Administrative Science Quarterly, 40, 619-652.
Gulati, R., and Lawrence, P. 1997. ―Organizing Vertical Networks: A Design Perspective,‖
working paper, J. L. Kellogg Graduate School of Management, Northwestern University.
Hagedoorn, J. 1995. ―Strategic Technology Partnering During the 1980s: Trends, Networks, and
Corporate Patterns in Non-core Technologies,‖ Research Policy, 24, 207-231.
Hamel, G. 1991. ―Competition for Competence and Inter-partner Learning Within International
Strategic Alliances,‖ Strategic Management Journal, 12, 83-103.
Hamel, G, Y. Doz, and C.K. Prahalad. 1989. ―Collaborate with Your Partners and Win,‖
Harvard Business Review, January-February: 133-139.
Harrigan, K.R. 1986. Managing Joint Ventures for Success, Lexington, MA, Lexington Books.
Heide, J.B. and A. Miner. 1992. ‗The Shadow of the Future: Effects of Anticipated Interaction
and Frequency of Contact on Buyer-Seller Cooperation,‖ Academy of Management
Journal, 35, 265-291.
Hladik, K.J. 1985. International Joint Ventures: An Economic Analysis of U.S.-Foreign Business
Partnerships, Lexington, MA: Lexington Books.
Huber, G.P. (1991). ‗Organizational learning: The contributing processes and literatures‘,
Organization Science, 2(1), pp.88-115.
John, G.A. & B.A. Weitz 1988 ―Forward integration into distribution: An empirical test of
transaction cost analysis‖, Journal of Law, Economics, & Organization, 4,pp.337-355.
Khanna, T., Gulati, R. and N. Nohria. 1998. ―The Dynamics of Learning Alliances: Competition,
Cooperation and Relative Scope,‖ Strategic Management Journal, 19:3, 193-210.
Kogut, B. 1988. ―Joint Ventures: Theoretical and Empirical Perspectives,‖ Strategic
Management Journal, 9:4, 319-332.
Kogut, B. 1989. ―The Stability of Joint Ventures: Reciprocity and Competitive Rivalry,‖ Journal
of Industrial Economics, 38, 183-198.
Krackhardt, D. 1992. ―The Strength of Strong Ties: The Importance of Philos in Organizations,‖
In N. Nohria and R. Eccles (Eds.), Networks and Organizations: Structure, Form and
Action. Harvard Business School Press. Boston, 216-239.
Krackhardt, D. 1996. ‗Comment on Burt and Knez‘s third party effects on trust‘, Rationality and
Lane, P.J. and M. Lubatkin 1998. ―Relative Absorptive Capacity and Interorganizational
Learning. Strategic Management Journal, 19, 461-477.
Lant, T. 1992. ―Aspiration Level Adaptation: An Empirical Exploration‖, Management Science,
vol. 38: 5, 623-644.
Larson, A. 1992. ―Network Dyads in Entrepreneurial Settings: A Study of the Governance of
Exchange Relationships,‖ Administrative Science Quarterly, 37, 76-104.
Larsson, R. L, Bengtsson, K. Henriksson and J. Sparks (1998). ―Barriers to Interorganizational
Learning : Developing Collective Knowledge Across Corporate Boundaries,‖ in R. Garud
and J. Porac (Eds.), Cognition, Knowledge, and Organization, Greenwich, CT: JAI Press,
Liebeskind, J.P. 1996. ―Knowledge, Strategy, and the Theory of the Firm,‖ Strategic
Management Journal, 17 (Winter Special Issue), 93-107.
Lyles, M., 1988, Learning Among Joint Venture Sophisticated Firms, in F. J. Contractor and O.
Lorange (Eds.), Cooperative Strategies in International Business, D.C. Heath, Lexington,
MacAllister, D. 1995. ―Affect- and Cognition-Based Trust as Foundations for Interpersonal
Cooperation in Organizations,‖ Academy of Management Journal, 38:1, 24-59.
McEvily, W. & A. Zaheer. 1999. Bridging Ties: A Source of Firm Heterogeneity in Competitive
Capabilities: Strategic Management Journal 20(12), pp.1133-1156.
Mowery, D.C. (ed.) (1988). International Collaborative Ventures in U.S. Manufacturing.
Ballinger, Cambridge, MA.
Mowery, D. C., J. E. Oxley, and B. S. Silverman. 1996. ―Strategic Alliances and Interfirm
Knowledge Transfer,‖ Strategic Management Journal, 17, (Winter Special Issue), 77-92.
Nahapiet, J., 1996. ―Managing Relationships With Global Clients: Value Creation Through
Cross-border Networks‖, Paper presented at the 16th Annual Conference of Strategic
Management Society, Phoenix, AZ.
Nahapiet, J. and S. Ghoshal 1998. ―Social Capital, Intellectual Capital and the Organizational
Advantage,‖ Academy of Management Review, 23:2, 242-266.
Nelson, R. R. 1982. ―The Role of Knowledge in R&D Efficiency,‖ Quarterly Journal of
Nelson, R.R. & Winter, S.C. 1982. An Evolutionary Theory of Economic Change. Boston:
Belknap Press of Harvard University Press.
Newell, A. and H. Simon, 1972. Human Problem Solving, Englewood Cliffs, NJ: Prentice Hall.
Nonaka, I. 1994. ―A Dynamic Theory of Organizational Knowledge Creation,‖ Organization
Science, 5:1, 14-37.
Nonaka, I., & Takeuchi, H. 1995. The Knowledge Creating Company. New York: Oxford
Nunnally, J. 1978. Psychometric Theory, New York: McGraw Hill.
Ohmae, K. 1989. ―The Global Logic of Strategic Alliances,‖ Harvard Business Review, 67:2,
Parkhe, A. 1993. ―Partner Nationality and Structure-Performance Relationship in Strategic
Alliances,‖ Organization Science, 4:2, 301-324.
Polanyi, M. 1967. The Tacit Dimension, garden City NY: Anchor Books.
Powell, W. W. 1990. Neither Market Nor Hierarchy: Network Forms of Organization, in L.L.
Cummings and B.M. Staw (Eds.) Research in Organizational Behavior, 12, 295-336.
Powell, W. W. and P. Brantley. 1992. ―Competitive Cooperation in Biotechnology: Learning
Through Networks?‖ In N. Nohria and R. Eccles (Eds.). Networks and Organizations:
Structure, Form and Action, 366-394. Boston: Harvard Business School Press.
Powell, W.W., K.W. Koput, and L. Smith-Doerr. 1996. ―Interorganizational Collaboration and
the Locus of Innovation: Networks of Learning in Biotechnology,‖ Administrative
Science Quarterly, 41, 116-145.
Prahalad, C.K. & R. Bettis, 1986. ‗The dominant logic: A new linkage between diversity and
performance‘, Strategic Management Journal , 7(6), pp. 485-501.
Ring, P., and Van de Ven, A. 1992. ―Structuring Cooperative Relationships Between
Organizations,‖ Strategic Management Journal, 13:7, 483-498.
Ring, P., and Van de Ven, A. 1994. ―Developmental Processes of Cooperative
Interorganizational Relationships,‖ Academy of Management Review, 19;1, 90-118.
Seabright, M.A., D.A. Levinthal & M. Fichman. 1992. ‗Role of individual attachment in the
dissolution of interorganizational relationships‘. Academy of Management Journal, 35(1),
Shan, W., G. Walker and B. Kogut, 1994. ‗Interfirm cooperation and startup innovation in the
biotechnology industry‘, Strategic Management Journal, 15(5), pp.387-394.
Simon, H.A. 1985. ―What we know about the creative process.‖ In R.L. Kuhn (ed.) Frontiers in
Creative and Innovative Management, 3-20. Cambridge, MA: Ballinger.
Simon, H. 1990. ―Invariants of Human Behavior,‖ Annual Review of Psychology, 1-19.
Simon, H. 1991. ―Bounded Rationality and Organizational Learning,‖ Organization Science,
Singley, M.K. and J. R. Anderson, 1989. The Transfer of Cognitive Skill, Cambridge, MA:
Starbuck, W.H. 1992. Learning by knowledge intensive firms. Journal of Management Studies,
Teece, D. 1992. ―Competition, Cooperation, and Innovation: Organizational Arrangements for
Regimes of Rapid Technological Progress,‖ Journal of Economic Behavior and
Organization, 18: 1-25.
Uzzi, B. 1996. ‗The sources and consequences of embeddedness for the economic performance
of organizations: The network effect‘, American Sociological Review, 61,pp. 674-698.
Van Lehn, K. 1990. Mind Bugs, Cambridge Mass; MIT Press.
Wasserman S. and K. Faust 1994. Social Network Analysis: Methods and Applications,
Cambridge UK: Cambridge University Press.
Weick, K. 1979. Enactment and Organizing: The Social Psychology of Organizing, New York:
Zaheer, A., B. McEvily and V. Perrone, 1998. ―Does Trust Matter? Exploring the Role of
Interorganizational and Interpersonal Trust on Performance,‖ Organization Science, 9,
Zolo, D. 1989. Reflexive Epistemology: The Philosophical Legacy of Otto Neurath. Kluwer,
Table 1: Measurement Instruments
1 2 3 4 H2 Alpha
To what extent do the individuals listed in your network
meet their obligations to you 0.862 0.007 0.197 0.150 0.804
deal fairly with you 0.888 -0.127 0.127 0.159 0.847
do not mislead you 0.820 -0.025 0.010 0.108 0.685
withhold relevant information 0.810 0.076 0.089 0.103 0.680
keep their word 0.750 -0.207 0.059 0.190 0.644
7-point Likert scale: 1 = not at all, 7 = full extent
Knowledge Complementarity 0.85
For each individual listed in your network
do you agree or disagree with the following statement.
We have worked in very similar functional areas -0.064 0.745 0.276 0.158 0.660
Overall, our backgrounds are very different 0.002 0.850 -0.020 -0.117 0.736
We have very different areas of expertise 0.017 0.901 0.011 -0.052 0.814
We are in the same profession -0.232 0.618 0.085 0.321 0.545
Our training in different technical areas -0.029 0.824 -0.036 -0.045 0.682
7-point Likert scale: 1 = strongly disagree, 7 = strongly agree
To what extent are the individuals working on
the alliance from your own firm ...
are highly competent 0.164 0.081 0.730 0.250 0.629
are highly skilled 0.152 0.122 0.890 0.115 0.843
are capable specialists 0.051 0.033 0.882 -0.007 0.782
are experts in their fields 0.055 -0.015 0.844 0.043 0.717
7-point Likert scale: 1 = not at all, 7 = full extent
I have learned a great deal in this alliance 0.346 -0.031 0.145 0.746 0.697
What I learn benefits my firm. 0.297 0.071 0.153 0.799 0.755
7-point Likert scale: 1 = strongly disagree, 7 = strongly agree
Eigenvalues 4.761 3.526 2.2162 1.0179
% of Variance 29.757 22.035 13.8522 6.362
Cumulative % 29.757 51.792 65.644 72.006
Table 2: Control Variables
The alliance is critical to your organization.
7-point Likert scale: 1 = strongly disagree, 7 = strongly agree
The technology underlying the alliance is highly uncertain.
7-point Likert scale: 1 = strongly disagree, 7 = strongly agree
Number of Employees in Focal Firm/ Number of Employees at Partner
Based on SIC for Medical Device, Pharmaceutical, and Biotechnology firms
coding 1 = same industry, 0 = different industry
Duration of alliance in years
Organizational Allying Experience
Number of alliances the organization has participated in over the past ten years
Shadow of the Future
Expected length of time that the alliance will continue. In years, months.
Table 3 – Means, Standard Deviations and Zero Order Correlations
1 Learning 5.806 1.025
2 Network Non-Redundancy 0.178 0.157
3 Trust 4.860 1.115
4 Knowledge Complementarity 7.024 2.065
5 Expertise 5.585 0.787
6 Importance 4.152 1.715
7 Relative Size 10.000 1.000
8 Industry Mix 1.585 0.494
9 Technological Uncertainty 3.639 1.644
10 Alliance Age 10.003 1.004
11 Org. Allying Experience 10.058 0.973
12 Shadow of the Future 10.016 0.973
1 2 3 4 5 6
2 Network Non-Redundancy .380***
3 Trust .507*** .144
4 Knowledge Complementarity .139 .049 -.076
5 Expertise .542*** .312*** .436*** .079
6 Importance .228** .122 .043 -.079 .013
7 Relative Size -.065 -.006 -.078 .042 -.013 -.212**
8 Industry Mix -.069 -.004 -.131 -.022 -.046 .075
9 Technological Uncertainty .019 .020 .006 .171 .116 -.120
10 Alliance Age .037 .105 .030 -.090 .132 -.011
11 Org. Allying Experience -.102 -.033 -.010 -.150 -.010 -.180
12 Shadow of the Future .186* .159* .130 -.020 .122 .293**
7 8 9 10 11
2 Network Non-Redundancy
4 Knowledge Complementarity
7 Relative Size
8 Industry Mix -.070
9 Technological Uncertainty .076 -.080
10 Alliance Age -.110 -.280** -.100
11 Org. Allying Experience .133 .017 -.040 -.040
12 Shadow of the Future -.120 -.010 -.210** .070 .090
Table 4 – Means, Standard Deviations and Zero Order Correlations
1 Learning 5.759 1.064
2 Network Non-Redundancy 0.162 0.151
3 Trust 4.876 1.084
4 Knowledge Complementarity 6.825 2.052
5 Expertise 5.566 0.717
6 Importance 3.819 1.658
7 Relative Size 10.300 1.353
8 Industry Mix 1.585 0.496
9 Technological Uncertainty 3.632 1.707
10 Alliance Age 10.022 1.014
11 Org. Allying Experience 10.378 1.115
12 Shadow of the Future 10.040 0.963
1 2 3 4 5 6
2 Network Non-Redundancy .399***
3 Trust .580*** .191
4 Knowledge Complementarity .130 -.042 -.199
5 Expertise .682*** .324** .592*** -.017
6 Importance .199 .054 .066 -.108 -.072
7 Relative Size -.073 .038 -.126 .106 -.010 -.236*
8 Industry Mix -.004 .071 -.184 -.009 -.078 .173
9 Technological Uncertainty .044 -.028 -.018 .188 .093 -.156
10 Alliance Age .011 .058 .077 -.213 -.003 -.058
11 Org. Allying Experience -.148 .022 -.080 -.262* -.050 -.126
12 Shadow of the Future .142 .071 .112 -.171 .171 .289**
7 8 9 10 11
2 Network Non-Redundancy
4 Knowledge Complementarity
7 Relative Size
8 Industry Mix -.105
9 Technological Uncertainty .110 -.037
10 Alliance Age -.175 -.282 -.146
11 Org. Allying Experience .044 .047 -.030 -.010
12 Shadow of the Future -.189 .032 -.229* .109 .186
Table 5 – Means, Standard Deviations and Zero Order Correlations
1 Learning 5.854 0.989
2 Network Non-Redundancy 0.194 0.162
3 Trust 4.844 1.152
4 Knowledge Complementarity 7.222 2.072
5 Expertise 5.604 0.856
6 Importance 4.486 1.717
7 Relative Size 9.700 0.006
8 Industry Mix 1.585 0.496
9 Technological Uncertainty 3.646 1.589
10 Alliance Age 9.984 1.000
11 Org. Allying Experience 9.739 0.674
12 Shadow of the Future 9.993 0.989
1 2 3 4 5 6
2 Network Non-Redundancy .358**
3 Trust .437*** .107
4 Knowledge Complementarity .142 .115 .040
5 Expertise .423 .301** .315 .156
6 Importance .250* .150 .030 -.093 .074
7 Relative Size -.019 .130 .024 -.015 .001 .024
8 Industry Mix -.138 -.075 -.080 -.035 -.020 -.017
9 Technological Uncertainty -.010 .067 .030 .155 .138 -.095
10 Alliance Age .068 .155 -.017 .030 .248 -.165
11 Org. Allying Experience .003 -.031 .086 .090 .069 -.125
12 Shadow of the Future .235 .243 .146 .130 .086 .317
7 8 9 10 11
2 Network Non-Redundancy
4 Knowledge Complementarity
7 Relative Size
8 Industry Mix .056
9 Technological Uncertainty .229 -.123
10 Alliance Age .083 -.288** -.046
11 Org. Allying Experience -.004 -.030 -.072 .041
12 Shadow of the Future -.090 -.043 -.195 .032 -.058
Table 6: Results of Regression Analysis
Dependent Variable: Learning All Firms Larger Partner Smaller Partner
Standardized Coefficient Standardized Coefficient Standardized Coefficient
H Beta T-value Sig. T Beta T-value Sig. T Beta T-value Sig. T
Intercept 0.834 0.406 -0.930 0.356 0.342 0.733
Network Non-Redundancy H1 0.200 3.152 0.002 0.180 2.290 0.025 0.211 2.014 0.048
Trust H2 0.336 4.965 0.000 0.310 3.198 0.002 0.322 3.126 0.003
Knowledge Complementarity H3 0.141 2.266 0.025 0.238 2.902 0.005 0.090 0.897 0.373
Expertise H4 0.321 4.567 0.000 0.467 4.731 0.000 0.243 2.172 0.033
Importance 0.181 2.730 0.007 0.228 2.749 0.008 0.176 1.662 0.101
Relative Size 0.006 0.098 0.922 0.006 0.076 0.939 -0.033 -0.324 0.747
Industry Mix -0.025 -0.388 0.699 0.058 0.722 0.473 -0.098 -0.947 0.347
Technological Uncertainty -0.027 -0.436 0.663 0.008 0.106 0.916 -0.067 -0.632 0.530
Alliance Age -0.015 -0.223 0.824 0.061 0.740 0.462 -0.022 -0.198 0.844
Prior Alliance Experience -0.041 -0.657 0.512 -0.007 -0.087 0.931 -0.027 -0.274 0.785
Shadow of the Future 0.021 0.322 0.748 -0.015 -0.175 0.862 0.027 0.251 0.803
F-Score 12.693 11.095 3.976
Significance of F .000 .000 .000
Adjusted R2 .450 .587 .296
Table 7: ANOVA Descriptive Statistics
Item Sum Sq. df MN Sq. F Sig. Contact MN Dev
Learning Between 15.809 1 15.809 16.472 0.000 yes 6.068 0.853
Within Groups 155.482 162 0.960 no 5.438 1.135
Total 171.291 163 all 5.806 1.025
Network Non- Between 0.529 1 0.529 24.638 0.000 yes 0.226 0.141
redundancy (NN) Groups
Within Groups 3.476 162 0.021 no 0.111 0.154
Total 4.005 163 all 0.178 0.157
NN own firm Between 0.134 1 0.134 6.627 0.011 yes 1.169 0.135
Within Groups 3.283 162 0.020 no 1.111 0.152
Total 3.417 163 all 1.145 0.145
NN partner firm Between 0.275 1 0.275 11.380 0.001 yes 1.208 0.151
Within Groups 3.909 162 0.024 no 1.125 0.162
Total 4.183 163 all 1.174 0.160
Trust Between 0.268 1 0.268 0.214 0.644 yes 4.894 1.044
Within Groups 202.487 162 1.250 no 4.812 1.215
Total 202.754 163 all 4.860 1.115
Knowledge Between 10.531 1 10.531 2.492 0.116 yes 7.237 1.947
Complementarity KC Groups
Within Groups 684.573 162 4.226 no 6.723 2.201
Total 695.104 163 all 7.024 2.065
KC own firm Between 3.527 1 3.527 0.997 0.319 yes 4.751 2.014
Within Groups 572.969 162 3.537 no 4.453 1.674
Total 576.496 163 all 4.628 1.881
KC partner firm Between 14.538 1 14.538 2.071 0.152 yes 5.234 2.854
Within Groups 1137.455 162 7.021 no 4.629 2.329
Total 1151.993 163 all 4.983 2.658
Expertise Between 5.393 1 5.393 9.138 0.003 yes 5.738 0.753
Within Groups 95.598 162 0.590 no 5.369 0.790
Total 100.990 163 all 5.585 0.787
Importance Between 0.302 1 0.302 0.102 0.750 yes 4.116 1.816
Within Groups 479.220 162 2.958 no 4.203 1.574
Total 479.522 163 all 4.152 1.715
Relative Size Between 0.336 1 0.336 0.335 0.564 yes 10.038 1.096
Within Groups 162.664 162 1.004 no 9.946 0.852
Total 163.000 163 all 10 1
Industry Mix Between 0.442 1 0.442 1.820 0.179 yes 1.542 0.501
Within Groups 39.363 162 0.243 no 1.647 0.481
Total 39.805 163 all 1.585 0.494
Table 7: ANOVA continued Descriptive Statistics
Item Sum Su df MN Sq. F Sig. Contact MN Dev
Technological Between 12.682 1 12.682 4.803 0.030 yes 3.873 1.568
Within Groups 427.723 162 2.640 no 3.309 1.702
Total 440.404 163 all 3.639 1.644
Alliance Age Between 0.340 1 0.340 0.336 0.563 yes 10.041 0.984
Within Groups 163.900 162 1.012 no 9.949 1.036
Total 164.240 163 all 10.003 1.004
Prior Alliance Between 4.557 1 4.557 4.934 0.028 yes 9.918 0.800
Within Groups 149.622 162 0.924 no 10.256 1.151
Total 154.179 163 all 10.058 0.973
Shadow of the Future Between 0.041 1 0.041 0.044 0.835 yes 10.003 1.002
Within Groups 148.644 156 0.953 no 10.035 0.938
Total 148.685 157 all 10.016 0.973
Table 8: Results of Regression Analysis
Dependent Variable: Learning All Firms Larger Partner Smaller Partner
Standardized Coefficient Standardized Coefficient Standardized Coefficient
Beta T-value Sig. T Beta T-value Sig. T Beta T-value Sig. T
Intercept 0.450 0.654 -0.304 0.763 0.656 0.515
Of own firm contacts 0.333 3.243 0.002 0.273 1.966 0.057 0.448 2.826 0.007
Of partner firm contacts -0.020 -0.166 0.868 0.021 0.135 0.893 0.023 0.111 0.912
Of outside contacts 0.106 1.197 0.234 0.339 3.009 0.005 -0.111 -0.674 0.504
Trust 0.366 4.208 0.000 0.378 2.939 0.006 0.314 2.282 0.028
with own firm contacts 0.211 2.512 0.014 0.156 1.384 0.175 0.269 1.995 0.053
with partner firm contacts -0.080 -0.931 0.354 -0.211 -1.775 0.084 -0.017 -0.116 0.909
with outside contacts -0.222 -2.554 0.012 -0.215 -1.886 0.067 -0.305 -2.069 0.045
Expertise 0.170 1.876 0.064 0.116 0.816 0.420 0.163 1.160 0.253
Importance 0.214 2.609 0.011 0.244 2.035 0.049 0.190 1.490 0.144
F-Score 9.146 6.521 3.997
Significance of F .000 .000 .001
Adjusted R2 .436 .519 0.360