Knowledge Transfer in Intraorganizational Networks: Effects of Network Position and

Absorptive Capacity on Business Unit Innovation and Performance

Author(s): Wenpin Tsai
Reviewed work(s):
Source: The Academy of Management Journal, Vol. 44, No. 5 (Oct., 2001), pp. 996-1004
Published by: Academy of Management
Stable URL: http://www.jstor.org/stable/3069443 .
Accessed: 01/10/2012 22:05

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

Academy of Management is collaborating with JSTOR to digitize, preserve and extend access to The Academy
of Management Journal.

http://www.jstor.org
3 Academy of Management Journal
2001, Vol. 44, No. 5, 996-1004.

KNOWLEDGE TRANSFER IN INTRAORGANIZATIONAL NETWORKS:

EFFECTS OF NETWORK POSITION AND ABSORPTIVE CAPACITY ON
BUSINESS UNIT INNOVATION AND PERFORMANCE
WENPIN TSAI
Pennsylvania State University

Drawing on a network perspective on organizational learning, I argue that organiza-

tional units can produce more innovations and enjoy better performance if they occupy
central network positions that provide access to new knowledge developed by other
units. This effect, however, depends on units' absorptive capacity, or ability to suc-
cessfully replicate new knowledge. Data from 24 business units in a petrochemical
company and 36 business units in a food-manufacturing company show that the
interaction between absorptive capacity and network position has significant, positive
effects on business unit innovation and performance.

Inside a multiunit organization, units can learn absorb knowledge by increasing R&D intensity
from each other and benefit from new knowledge (e.g., Cohen & Levinthal, 1990), much less attention
developed by other units. Knowledge transfer has been focused on the process of gaining knowl-
among organizational units provides opportunities edge access. Getting access to new knowledge re-
for mutual learning and interunit cooperation that quires networking effort that is different from in-
stimulate the creation of new knowledge and, at the vesting in R&D. In a multiunit organization, a unit
same time, contribute to organizational units' abil- can access new knowledge through a network of
ity to innovate (e.g., Kogut & Zander, 1992; Tsai & interunit links (Hansen, 1999). In this research, I
Ghoshal, 1998). However, knowledge is often conceptualize an organization as a network ar-
"sticky" and difficult to spread (Szulanski, 1996; rangement and investigate a unit's access to knowl-
Von Hippel, 1994). How can an organizational unit edge by analyzing its network position in its in-
gain useful knowledge from other units to enhance traorganizational network. In addition, I argue that
its innovation and performance? both external knowledge access and internal learn-
Prior research has suggested that organiza- ing capacity are important for a unit's innovation
tional units not only hold specialized knowledge and performance. Although a central network po-
but also have the opportunity to learn from other sition allows a unit to access new knowledge de-
units (Huber, 1991). However, not every unit can veloped by many other units, high learning capac-
learn from all other units in the same organiza- ity permits a unit to successfully apply or replicate
tion. A unit may want to obtain knowledge from new knowledge.
other units but may not be able to access it. Even
though the knowledge is available, the unit may
ORGANIZATIONAL LEARNING AND
not have the capacity to absorb and apply it for its
INTERUNIT KNOWLEDGE TRANSFER:
own use. Organizational units require external
A NETWORK PERSPECTIVE
access and internal capacity to learn from their
peers. Because of their differential external ac- Inside an organization, learning involves the
cess and internal capacity, organizational units transfer of knowledge among different organiza-
differ in their abilities to leverage and benefit tional units. Such knowledge transfer occurs in a
from knowledge developed by other units. shared social context in which different units are
Although the organizational learning literature linked to one another. Organizational units are em-
has highlighted the importance of the capacity to bedded in a network coordinated through pro-
cesses of knowledge transfer and resource sharing
(Galbraith, 1977; Gresov & Stephens, 1993). Such a
The author would like to thank Don Bergh, Dan Brass,
network of interunit links enables organizational
Sumantra Ghoshal, Ranjay Gulati, Mathew Hayward, units to gain critical competencies that contribute
Martin Kilduff, Harry Korine, Don Sull, Kevin Steensma, to their competitiveness in the marketplace.
Linda Trevifno, Anne Tsui, Greg Northcraft, and three Interunit links and networks are an important
anonymous reviewers for their very helpful comments part of a learning process in which organizational
and suggestions on drafts. units discover new opportunities and obtain new
996
2001 Tsai 997

knowledge through interacting with one another. lems (e.g., Dougherty & Hardy, 1996; Ibarra, 1993;
The importance of interunit links has been docu- Van de Ven, 1986). However, such knowledge is
mented in the strategy literature. For example, re- usually distributed unevenly within an organiza-
search on diversification has emphasized the ben- tion. As Szulanski (1996) argued, knowledge is dif-
efits, for multiunit companies, of pursuing synergy ficult to spread across different units within an
through knowledge transfer and resource sharing organization in which preexisting relationships
among their strategic business units (SBUs). As among units are absent. Indeed, innovative ideas
Gupta and Govindarajan noted, "The potential for are often at the nexus of interunit links. To foster
synergistic benefits from resource sharing varies innovation, information and knowledge should be
across strategic contexts, and the realization of deliberately distributed. A network of interunit
these potential synergistic benefits depends on links provides channels for distributing informa-
how effectively linkages between SBUs are actually tion and knowledge in such a way as to stimulate
managed" (1986: 696). In addition, research on the and support innovative activities. A central net-
knowledge-based view of the firm has suggested work position is associated with innovation out-
that social networks facilitate the creation of new comes for individual units within an organization.
knowledge within organizations (e.g., Kogut & As several scholars have argued, a unit's network
Zander, 1992; Tsai, 2000). Through the develop- position is an important aspect of "social structure"
ment of interunit network links, horizontal transfer that can enhance the unit's ability to create new
of knowledge broadens organizational learning. As value and to achieve economic goals (e.g.,
Huber (1991) suggested, a learning organization is Coleman, 1990; Tsai & Ghoshal, 1998). An organi-
characterized by motivated units that are inti- zational unit occupying a more central position in
mately connected to one another. By linking differ- its intraorganizational network is likely to produce
ent units together, a network arrangement provides more innovations. Hence,
a flexible learning structure that replaces old hier-
archical structures. Hypothesis la. The centrality of an organiza-
Drawing on a network perspective on organiza- tional unit's network position is positively re-
tional learning, I examined two important con- lated to its innovation.
cepts, network position and absorptive capacity,
that determine the effectiveness of interunit learn- Organizational units differ in their internal
ing and knowledge transfer. Network position, a knowledge, practices, and capabilities. Networks of
unit's location in an interunit network, describes interunit links allow organizational units to access
its access to knowledge; absorptive capacity, a new knowledge from each other and may increase
unit's R&D investment, describes its capacity to their cost efficiency through dissemination of "best
learn. Organizational units are not identically ca- practices" within organizations. As Hill, Hitt, and
pable of acquiring knowledge; they are not equally Hoskisson noted, networks of knowledge transfer
efficient or effective learners. Because of differ- among organizational units "enable the diversified
ences in their knowledge access and learning ca- firm either to reduce overall operating costs in one
pacity, organizational units have differing learning or more of its divisions, or to better differentiate the
capabilities that in turn have a significant impact products of one or more of its divisions" (1992:
on their innovation and performance. 502). The centrality of a unit in the intraorganiza-
tional network may determine the unit's access to
different knowledge, thus affecting its ability to
Network Position
recognize and respond to new market opportuni-
Different network positions represent different ties. A unit occupying a central network position
opportunities for a unit to access new knowledge can gain competitive advantages in the marketplace
that is critical to developing new products or inno- because of its unique access to other units' knowl-
vative ideas. An organizational unit's network po- edge or practices. Such a central unit may enhance
sition reveals its ability to access external informa- its profitability by applying other units' knowledge
tion and knowledge. By occupying a central or practices to adapt its products to market needs,
position in the interunit network, a unit is likely to to respond to emerging market trends, and to deal
access desired strategic resources. Such resources with competitive challenges. Moreover, a central
will fuel the unit's innovative activities by provid- unit is likely to improve its business operations as
ing the external information necessary to generate it can enjoy the benefits of scope economies by
new ideas. Equally, the innovative work of the unit sharing the knowledge developed by other units.
will benefit from access to the new knowledge nec- As a result, performance differences among organi-
essary to resolve design and manufacturing prob- zational units may be attributable to the differences
998 Academy of ManagementJournal October

in their intraorganizational network positions. new products. In addition, a unit with high absorp-
Hence, tive capacity is likely to apply new knowledge to
improve its business operations. Increments to an
Hypothesis lb. The centrality of an organiza- organizational unit's knowledge base enhance the
tional unit's network position is positively re- unit's business performance in that it can profit
lated to its business performance. from the new knowledge it has absorbed. As a
result, higher absorptive capacity is related to bet-
Absorptive Capacity ter business performance. Accordingly,

Organizational units also differ in their ability to Hypothesis 2b. An organizational unit's ab-
assimilate and replicate new knowledge gained sorptive capacity is positively related to its
from external sources. Cohen and Levinthal (1990) business performance.
labeled such ability "absorptive capacity." In dis-
cussing how it contributes to innovation, they ar-
Interaction between Network Position and
gue that absorptive capacity tends to develop cu-
Absorptive Capacity
mulatively and builds on prior related knowledge.
Organizational units that possess relevant prior Absorptive capacity is also likely to moderate the
knowledge are likely to have a better understanding effect of network position on business unit innova-
of new technology that can generate new ideas and tion and performance. Although a central network
develop new products. Organizational units with a position provides important access to new knowl-
high level of absorptive capacity are likely to har- edge, its impact on business unit innovation and
ness new knowledge from other units to help their performance may depend on the extent to which a
innovative activities. Organizational units must unit can absorb such new knowledge. A unit may
have the capacity to absorb inputs in order to gen- be able to access certain new knowledge, but not
erate outputs. Without such capacity, they cannot enhance its innovation and performance if it does
learn or transfer knowledge from one unit to an- not have enough capacity to absorb such knowl-
other. For example, in a study of 122 "best prac- edge. The better a unit can access other units'
tice" transfers in eight companies, Szulanski (1996) knowledge, the more it needs absorptive capacity
found that lack of absorptive capacity was a major to benefit from such knowledge. An organizational
barrier to internal knowledge transfer within organ- unit occupying a central network position can ac-
izations. Absorptive capacity results from a pro- cess new knowledge from many other units. Such a
longed process of investment and knowledge central network position will have a more positive
accumulation. An organizational unit's absorptive impact on the unit's innovation output and busi-
capacity for learning depends on its endowment of ness performance if the unit has high absorptive
relevant technology-based capabilities (Mowery, capacity with which to effectively transfer knowl-
Oxley, & Silverman, 1996). R&D investment is a edge from other units. The interaction between net-
necessary condition for the creation of absorptive work position and absorptive capacity is critical to
capacity. As Cohen and Levinthal suggested, the intraorganizational knowledge sharing. Without a
ability to utilize external knowledge is often a by- simultaneous consideration of its network position
product of R&D investment. Organizational units and absorptive capacity, a unit is likely to encoun-
with a high level of absorptive capacity invest more ter a "search-transfer problem" in which it cannot
in their own R&D and have the ability to produce transfer the knowledge it identified through its net-
more innovations. Hence, work search (Hansen, 1999). The more central a
unit is in an intraorganizational network, the
Hypothesis 2a. An organizational unit's ab- broader the knowledge sources the unit has and the
sorptive capacity is positively related to its in- higher the absorptive capacity needed to transfer
novation. such knowledge. Hence,
An organizational unit's absorptive capacity also Hypothesis 3a. The centrality of an organiza-
affects its business performance. According to Co- tional unit's network position is more posi-
hen and Levinthal (1990), absorptive capacity in- tively related to innovation when the unit has
volves not only the ability to assimilate new exter- high absorptive capacity than when the unit
nal knowledge, but also the ability to apply such has low absorptive capacity.
knowledge to commercial ends and, thus, create
the opportunity for profits. Having good research Hypothesis 3b. The centrality of an organiza-
and development, a unit with high absorptive ca- tional unit's network position is more posi-
pacity is likely to successfully commercialize its tively related to business performance when
2001 Tsai 999

the unit has high absorptive capacity than Ghoshal (1998), to measure interrater agreement for
when the unit has low absorptive capacity. relational data. The mean percentage agreement is
defined as the number of responses selected by
both respondents in a unit divided by the number
METHODS of responses selected by at least one of the two
respondents in a unit. The value of the mean per-
Data Collection and Research Site
centage agreement can range from 0.0 (perfect in-
This research was conducted in two large multi- consistency) to 1.0 (perfect consistency). In this
national corporations, here given the fictitious study, the mean percentage agreement was 0.93 in
names Taiplex Corporation and Resident Enter- Taiplex and 0;77 in Resident for my network mea-
prise. Each had a typical multiunit organizational sure. The mean percentage agreement was calcu-
structure in which each unit was responsible for lated before I cross-validated the responses. For my
developing, manufacturing, and selling products. statistical analyses, only validated data were used.
Although similar in organizational structure, the The method for data cross-validation is detailed in
two companies specialized in different businesses the next section.
and thus differed in many aspects of their opera-
tions. Taiplex specialized in the petrochemical in-
Dependent Variables
dustry and had annual revenues of $10.7 billion
and total assets of $15 billion at the time of the There were two dependent variables in this
study. Resident specialized in food manufacturing study: innovation and performance. Both were
and had annual revenues of $4.1 billion and total measured at the business unit level. Because units
assets of $3.8 billion. The two companies also tar- may specialize in different industries and have dif-
geted very different markets. Taiplex's products, ferent strategic priorities, innovation and perfor-
which were mainly for industrial markets, in- mance data needed to be adjusted to evaluate each
cluded plastic raw materials, plastic secondary unit (Gupta & Govindarajan, 1984). To do so, I used
products, and industrial equipment. Resident's an innovation achieved rate, or the number of new
products, which were mainly for consumer mar- products introduced in a unit in a particular year
kets, included edible oil, beverages, fast foods, and divided by the unit's target number in that year,
dairy products. and a profitability achieved rate-a unit's return on
A questionnaire was distributed to all business investment in a particular year divided by its target
units in the two companies in 1996. I used socio- return in that year-to measure innovation and per-
metric techniques in the questionnaire to collect formance, respectively. The innovation and profit-
relational data that described how units interacted ability targets were assessed and negotiated be-
with one another within each company. At the time tween unit managers and corporate managers each
of the survey, Taiplex had 24 business units and year. These managers considered business unit
Resident had 36 business units. For each of these strategic priorities and industry-related factors
units, I contacted two individuals, the director and when they set the targets. The achieved rates for all
the most senior deputy director, to respond to my the units in this study were collected for the time
questionnaire. Therefore, I had a total of 120 poten- period 1997-98 through corporate internal records.
tial respondents. Because top management in both
companies approved and supported the study, all
Independent Variables
the contacts completed and returned my question-
naire. To ensure confidentiality, I promised that I Absorptive capacity. Following Cohen and
would not reveal the true names of the companies, Levinthal (1990), I used R&D intensity (defined as
the units, and the respondents involved in this R&D expenditure divided by sales) to measure ab-
research. Respondents were asked to return their sorptive capacity at the business unit level. Data on
completed questionnaires directly to me instead of R&D expenditure and unit sales were obtained
routing them through corporate headquarters. In through corporate internal records. Consistent with
addition to the questionnaire survey, corporate in- the data collection period of other independent
ternal records were also used to collect data on variables in this study, 1996 R&D intensity was
business unit R&D intensity, innovation, and per- used here.
formance. Network position. To identify a business unit's
Because I had multiple respondents in each unit, intraorganizational network position, I developed a
I calculated interrater agreement to examine how questionnaire item asking the respondents, "Which
responses varied within each unit. I used the mean units provide your unit with new knowledge or
percentage agreement, as suggested by Tsai and expertise when your unit is seeking technical ad-
1000 Academy of ManagementJournal October

vice inside your organization?" A list of all the innovations. Likewise, business units that per-
units was provided in the questionnaire, allowing formed very well in the past are likely to keep up a
respondents to simply select their answers from the good performance. Hence, I included the innova-
list. To validate the data, I also asked the opposite tion and performance measures for previous years
question, that is, who came to them for new knowl- (1993-96) in my statistical analyses.
edge or expertise. I ascertained that there was a
knowledge transfer relationship between units i
and j if unit i indicated it had provided its knowl-
edge to unit j and unit j also confirmed receiving RESULTS
knowledge from unit i (cf. Hansen, 1999; Krack- Table 1 shows the mean values, standard devia-
hardt, 1990). Because I had multiple respondents tions, and correlations for all the measured vari-
per unit, I considered data valid if a knowledge ables for both companies. Since I had two research
transfer relationship (indicated by any respondent sites, I performed a Chow test to examine the con-
of the knowledge source unit) was confirmed by sistency of results; it indicated that the levels of
any respondent of the knowledge recipient unit. significance found for my independent variables
Using validated data, I constructed a relational ma- were not statistically different across the two com-
trix of interunit links for each company-a 24 by 24 panies (business unit innovation, F4, 52= 0.95 p =
matrix for Taiplex and a 36 by 36 matrix for Resi- 0.44; business unit performance,F4 52 = 0.88, p =
dent. In each matrix, the i, jth cell is coded 1 if unit 0.48). Given the result of the Chow test, I pooled the
i provided its knowledge to unit j. data for all subsequent analyses. To see how much
Drawing on these relational matrixes of interunit additional variance was explained by the indepen-
links, I calculated in-degree centrality for each unit. dent variables after controls, I tested my hypothe-
In-degree centrality represents the total number of ses with hierarchical regression analysis, entering
units from which a focal unit has received knowl- control variables in step 1, independent variables
edge. The higher a unit's in-degree centrality, the in step 2, and interactions in step 3 and tracing
more knowledge sources the unit has. As Freeman change in the multiple squared correlation coeffi-
(1979) argued, in-degree centrality is the most suit- cient (R2) from step to step.
able centrality measure for capturing an individual Table 2 shows the results of hierarchical regres-
actor's information or knowledge access. sion analyses estimating the effects of absorptive
capacity and network position on business-unit
innovation. Hypothesis la states that a unit occu-
Control Variables
pying a more central network position is likely to
Size can affect a unit's innovation and perfor- be more innovative. As shown in Table 2, the co-
mance. Large units tend to have more resources efficient for network position is positive and signif-
with which to enhance their innovation and per- icant (p < .01), indicating that a unit's centrality in
formance. They are also usually more powerful its intraorganizational network contributes to its
than small units and have some advantages in gain- innovation. Hence, Hypothesis la is supported. Hy-
ing the headquarters' support for their business pothesis 2a predicts a direct effect of absorptive
operations and innovation activities. In this study, capacity on business unit innovation. The coeffi-
I used the logarithms of unit sales and the number cient for absorptive capacity is positive and signif-
of employees in each unit as indicators of unit size. icant (p < .01), indicating that a unit with high
Since the two size indicators were correlated, I absorptive capacity is likely to be more innovative.
averaged them to create a composite measure. The Hypothesis 2a is confirmed. Hypothesis 3a states
Cronbach's alpha for this composite measure was that absorptive capacity will moderate the relation-
.94 in Taiplex and .95 in Resident. ship between network position and innovation. To
Local competition is another variable that can test this hypothesis, I multiplied network position
affect innovation and performance. To assess the and absorptive capacity and entered the multipli-
extent of competition in different local markets, I cative interaction item into the regression. Follow-
used these two items: "Competition is intense in ing Aiken and West (1991), I mean-centered the
our local environment" and "Our unit has strong variables (transforming the data into deviation
competitors in the marketplace" (1 = "strongly dis- score form with means equal to zero) and reran the
agree," 7 = "strongly agree"; a = .82, Taiplex, and regression to minimize the distortion due to high
.89, Resident; rwg = .92, Taiplex, and .88, Resident). correlations between the interaction term and its
I also controlled for past innovation and past component variables. As predicted, the coefficient
performance. Business units with a strong history of the interaction was positive and significant (p <
of innovation tend to continue producing many .01), indicating that the effect of network position
2001 Tsai 1001

TABLE1
Means, Standard Deviations, and Correlations
Variable Mean s.d. 1 2 3 4 5 6 7

Taiplex
1. Network position 24.64 17.47
2. Absorptive capacity 8.52 4.64 .27
3. Unit size 5.24 0.44 -.28 .20
4. Competition 4.05 1.63 -.30 -.20 .05
5. Prior innovation 94.24 49.78 -.16 .28 .38 .07
6. Innovation 92.04 51.27 .36 .60** .29 -.15 .49*
7. Prior performance 113.83 15.55 .20 -.28 -.04 -.49* .08 -.09
8. Performance 115.29 23.95 .34 .33 -.34 -.39 -.07 .13 .16
Resident
1. Network position 23.49 20.11
2. Absorptive capacity 9.28 3.94 .09
3. Unit size 3.76 0.43 .64** -.06
4. Competition 4.22 1.76 -.26 .19 -.07
5. Prior innovation 106.50 71.44 .31 .37* .23 -.22
6. Innovation 101.39 40.11 .56** .45** .56** .06 .46**
7. Prior performance 103.97 14.67 .16 -.08 .13 -.43** .21 .03
8. Performance 105.11 18.16 .36* .24 .22 -.47** .41* .31 -.08

*p < .05
**p < .01
Two-tailed tests.

on innovation is dependent on a unit's absorptive effect of absorptive capacity on business-unit per-

capacity. Hence, Hypothesis 3a is supported. formance. The coefficient for absorptive capacity is
Hypothesis lb states that a unit occupying a more positive and significant (p < .05), indicating that a
central network position is likely to perform better unit with high absorptive capacity is likely to have
than a unit in a less central position. As shown in good performance. Hypothesis 2b is confirmed. Hy-
Table 2, the coefficient for network position is not pothesis 3b states that absorptive capacity will
statistically significant, indicating that a unit's cen- moderate the relationship between network posi-
trality in its intraorganizational network does not tion and performance. The interaction coefficient is
contribute to its performance. Hence, Hypothesis significant (p < .05), indicating that the effect of
lb is not supported. Hypothesis 2b predicts a direct network position on performance is dependent on a

TABLE2
Results of Hierarchical Regression Analysis: Effects of Network Position and Absorptive Capacitya
Business-Unit Innovation Business-Unit Performance

Variable 1 2 3 1 2 3

Unit size 34.50** 27.20** 32.30** -2.41 -6.25 -4.51

Local competition 0.65 2.06 2.56 -5.72** -4.93** -4.69**
Company 57.33** 47.56** 57.78** -10.49 -17.03 -13.63
Prior innovation 0.25** 0.14* 0.13*
Prior performance 0.19 0.13 0.12
Network position 0.73** 0.60** 0.24 0.19
Absorptive capacity 4.12** 4.09** 1.27 1.24*
Network position x absorptive capacity 0.18** 0.06*
R12 0.31 0.56 0.63 0.25 0.37 0.42
AR2 0.24 0.08 0.12 0.05
AF 14.62** 11.01** 4.97** 4.41*

an = 60. Data for the two research sites were pooled.

*p < .05
**p < .01
1002 Academy of ManagementJournal October

unit's absorptive capacity. Hypothesis 3b is sup- units' performance are mediated by their effects on
ported. business units' innovation by entering business-
To better explain the form of interactions re- unit innovation as an additional control variable in
ported in the above hierarchical regression analy- the business unit performance analysis. The results
sis, I plotted the interaction effects in the graphs show that absorptive capacity and its interaction
shown in Figure 1, using one standard deviation with network position remain significant (p < .05)
above and below the mean to capture high and low when business unit innovation has been entered as
absorptive capacity (Cohen & Cohen, 1983). a control, indicating that the effects of network
position and absorptive capacity on business-unit
performance were not mediated by business-unit
Additional Analyses
innovation in this study.
In the above statistical analyses, a business unit's
network position was measured as its in-degree
DISCUSSION AND CONCLUSION
centrality in its firm's intraorganizational network.
I also performed additional analyses using an alter- How can an organizational unit gain useful
native measure of network position based on the knowledge from other units to enhance its innova-
similarity of ties among business units (e.g., Burt, tion and performance? This research suggests that a
1976, 1987). To identify the similarity of ties, I ran unit's external knowledge access and internal
structural equivalence analysis using UCINET IV learning capacity are critical to answering this
(Borgatti, Everett, & Freeman, 1992). This alterna- question.
tive measure yielded the same pattern of results. A unit's external knowledge access is character-
Finally, I also tested whether the effects of net- ized by its network position. By occupying a cen-
work position and absorptive capacity on business tral network position, a unit is likely to access
useful knowledge from other units. The result of
this research indicates that a unit's innovative ca-
FIGURE1 pability is significantly increased by its centrality
Interaction Results in the intraorganizational network, which provides
Innovation Highabsorptivecapacity opportunities for shared learning, knowledge trans-
fer, and information exchange. The result demon-
strates the importance of gaining access to knowl-
edge through networks and, at the same time,
contributes to the literature on networks and inno-
vations (e.g., Ibarra, 1993). Given that vigorous in-
Lowabsorptivecapacity novative activities usually take place in organiza-
tional units, it is indispensable to examine how
internal social processes within organizations af-
fect innovation at the organizational unit level. By
showing how network position affects innovation
in business units, this research provides motivation
to study innovation processes within multiunit or-
NetworkCentrality ganizations. This research, however, does not show
a significant association between a unit's network
Performance position and its business performance. It seems that
Highabsorptivecapacity
the benefits of centrality may not always outweigh
its costs. Although a central unit can gain a lot of
information benefits, maintaining a central posi-
tion may require intensive coordination efforts that
lead to high administrative costs. More research is
needed to investigate the net effect of a unit's net-
Lowabsorptivecapacity work position on its performance.
An organizational unit's internal learning capac-
ity determines the extent to which it can absorb
new knowledge from other units (Cohen &
Levinthal, 1990). Investing in absorptive capacity
allows a unit to effectively assimilate and apply
NetworkCentrality external knowledge for its own use. This research
2001 Tsai 1003

demonstrates that absorptive capacity significantly tively is truly a source of competitive advantage.
affects business units' innovation as well as their By examining the pattern of intraorganizational
performance. The result suggests that high absorp- knowledge transfer and its performance implica-
tive capacity is associated with a better chance to tions, this research contributes to the organiza-
successfully apply new knowledge toward com- tional learning literature and highlights the impor-
mercial ends, producing more innovations and bet- tance of sharing firm-specific knowledge within
ter business performance. The result contributes to organizations.
the research on business unit strategy, given that Providing further evidence that networks play an
improving business performance is one of the most important role in shaping business outcomes, this
important objectives for business units in large, research has significant implications for the grow-
complex organizations (e.g., Gupta & Govindarajan, ing body of research on networks. Specifically, it
1986). indicates that network position can promote social
This research also shows that the interaction be- learning that makes linked units more astute col-
tween network position and absorptive capacity lectively than they are individually (Kraatz, 1998).
significantly affects business units' innovation and Using network analysis, this research indicates a
performance. This finding is interesting, given that way of exploring the relational profiles of organi-
previous research has focused on the direct effect of zational units and the patterns of interunit knowl-
network structure in explaining business outcomes edge transfer. Although a few other studies have
only, without addressing whether the effect might examined attributes of interunit networks, their
be dependent on the extent to which a unit can findings are limited to a specific organization be-
absorb knowledge (e.g., Tsai & Ghoshal, 1998). A cause of a one-site sampling scheme (e.g., Hansen,
central unit may be able to access knowledge 1999). In contrast, this research examined network
through its network links but may not have suffi- structures in two multinational companies special-
cient capacity to absorb such knowledge. Hence, izing in different industries. The present results are
the better a unit can access other units' knowledge, stronger given that a similar pattern was found in
the higher the absorptive capacity the unit should two different interunit networks.
have. The result suggests that a unit has to invest An organizational unit's network position and
significantly in its absorptive capacity when ex- absorptive capacity represent its ability to leverage
panding its network links. useful knowledge residing in other parts of its or-
In this study, I focused on how the interaction ganization. A unit's network position reveals its
between network position and absorptive capacity relative strength in gaining access to new knowl-
affected innovation and performance, respectively. edge, a unit's absorptive capacity reveals its ability
Innovation may mediate the effects of absorptive to replicate or apply such new knowledge. The
capacity and network position on performance. present results show a positive association between
However, this study does not demonstrate this po- network position and business unit innovation,
tential mediation effect. It is possible that there is a and confirm the moderating role of absorptive ca-
significant time lag between innovation and a pos- pacity in this association. The influence of both
itive impact on performance. Future studies, in- network position and absorptive capacity should
cluding full longitudinal histories of business unit be studied simultaneously. Investing in absorptive
performance, could further explore this issue. capacity while expanding network links is critical
Although previous research has elaborated the to the success of organizational units in learning
concept of organizational learning, there is little new knowledge that eventually leads to competi-
systematic understanding of the social processes tive advantage. Investigating network position and
that underlie how organizational units learn from absorptive capacity also provides useful informa-
each other. Critical insights and ideas reside in tion a multiunit firm's corporate headquarters can
organizational units. However, knowledge gener- use to understand the relational profiles and learn-
ated by individual units does not come to bear on ing potential of its units. Although this research
an organization independently (Crossan, Lane, & focuses on learning outcomes at the unit level, col-
White, 1999). Knowledge and ideas are shared and lectively these outcomes may influence the evolu-
common meanings are developed through interac- tionary path of an entire organization. Interesting
tions. Knowledge is socially constructed, and or- results may accrue from examining how interunit
ganizational learning involves a complex social learning affects the development of organizational
process in which different units interact with each capabilities and organization-level outcomes. Fu-
other (Berger & Luckmann, 1966; Huber, 1991). An ture research pursuing this line of inquiry has great
organization is a repository of knowledge. The abil- potential to make significant contributions to man-
ity to access knowledge and to integrate it effec- agement research.
1004 Academy of ManagementJournal October

REFERENCES izational subunits. Administrative Science Quar-

terly, 44: 82-111.
Aiken, L. S., & West, S. G. 1991. Multiple regression:
Testing and interpreting interactions. Newbury Hill, C. W. L., Hitt, M. A., & Hoskisson, R. E. 1992.
Park, CA: Sage. Cooperative versus competitive structures in related
and unrelated diversified firms. Organization Sci-
Berger,P., &Luckmann,T. 1966. The social construction ence 3: 501-521.
of reality: A treatise in the sociology of knowledge.
Huber, G. P. 1991. Organizational learning: The contrib-
GardenCity, NY: Anchor Books.
uting processes and the literatures. Organization
Borgatti, S. P., Everett, M. G., & Freeman, L. C. 1992. Science, 2: 88-125.
UCINETIV Version 1.00. Columbia, SC: Analytical
Ibarra, H. 1993. Network, centrality, power, and innova-
Technologies.
tion involvement: Determinants of technical and ad-
Burt, R. S. 1976. Positions in networks. Social Forces, ministrative roles. Academy of Management Jour-
55: 93-122. nal, 36: 471-501.
Burt, R. S. 1987. Social contagion and innovation: Cohe- James, L. R., Demaree, R. G., & Wolf, G. 1993. rwg: An
sion versus structural equivalence. American Jour- assessment of within-group interrater agreement.
nal of Sociology, 92: 1287-1335. Journal of Applied Psychology, 78: 306-309.
Chow, G. 1960. Tests of equality between sets of coeffi- Kogut, B., & Zander, U. 1992. Knowledge of the firm,
cients in two linear regressions. Econometrica, 28: combinative capacities and the replication of tech-
591-605. nology. Organization Science, 3: 383-397.
Cohen, J., & Cohen, P. 1983. Applied multiple regres- Kraatz, M. S. 1998. Learning by association? Interorgani-
sion/correlation analysis for the behavioral sci- zational networks and adaptation to environmental
ences (2nd ed.). Hillsdale, NJ:Erlbaum. change. Academy of ManagementJournal, 41: 621-
643.
Cohen, W., & Levinthal, D. 1990. Absorptive capacity: A
new perspective on learning and innovation. Ad- Krackhardt, D. 1990. Assessing the political landscape:
ministrative Science Quarterly, 35: 128-152. Structure, cognition and power in organizations. Ad-
ministrative Science Quarterly, 35: 342-369.
Coleman, J. S. 1990. Foundations of social theory. Cam-
bridge, MA: HarvardUniversity Press. Mowery, D. C., Oxley, J. E., Silverman, B. S. 1996. Stra-
tegic alliances and interfirm knowledge transfer.
Crossan, M. M., Lane, H. W., & White, R. E. 1999. An Strategic Management Journal, 17: 77-91.
organizational learning framework:From intuition
to institution. Academy of Management Review, Szulanski, G. 1996. Exploring stickiness: impediments to
24: 522-537. the transfer of best practice within the firm. Strate-
gic Management Journal, 17(winter special issue):
Dougherty, D., & Hardy, C. 1996. Sustained product in- 27-43.
novation in large, matureorganizations:Overcoming
innovation-to-organization problems. Academy of Tsai, W. 2000. Social capital, strategic relatedness and
Management Journal, 39: 1120-1153. the formation of intraorganizational linkages. Stra-
tegic Management Journal, 21: 925-939.
Freeman, L. C. 1979. Centralityin social networks: Con-
Tsai, W., & Ghoshal, S. 1998. Social capital and value
ceptual clarification. Social Networks, 1: 215-239.
creation: The role of intrafirm networks. Academy
Galbraith,J. R. 1977. Organizational design. Reading, of Management Journal, 41: 464-476.
MA: Addison-Wesley.
Van de Ven, A. 1986. Central problems in the manage-
Gresov, C., & Stephens, C. 1993. The context of interunit ment of innovation. Management Science, 32: 590-
influence attempts. Administrative Science Quar- 607.
terly, 38: 252-276.
Von Hippel, E. 1994. "Sticky information" and the locus
Gupta, A. K., & Govindarajan,V. G. 1984. Business unit of problem solving: Implications for innovation.
strategy, managerial characteristics, and business Management Science, 40: 429-430.
unit level effectiveness at strategy implementation.
Academy of Management Journal, 27: 25-41. Wenpin Tsai ([email protected]) is an assistant professor of
management in the Smeal College of Business Adminis-
Gupta,A. K., &Govindarajan,V. G. 1986. Resource shar- tration at the Pennsylvania State University. He received
ing among SBUs: Strategic antecedents and admin- his Ph.D. in strategic and international management at
istrative implications. Academy of Management the London Business School. His current research inter-
Journal, 29: 895-714. ests include social capital, organizational learning, busi-
Hansen, M. T. 1999. The search-transferproblem: The ness unit strategy, and the evolution of intra- and inter-
role of weak ties in sharing knowledge across organ- organizational networks.