Professional Documents
Culture Documents
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=jwiley.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.
John Wiley & Sons is collaborating with JSTOR to digitize, preserve and extend access to Strategic
Management Journal.
http://www.jstor.org
Strategie Management Journal
Strat. Mgmt. J., 24: 293-316 (2003)
Published online 20 December 2002 inWiley InterScience (www.interscience.wiley.com). DOI: 10.1002/smj.297
GAINING FROMVERTICALPARTNERSHIPS:
KNOWLEDGETRANSFER, RELATIONSHIP
DURATION, AND SUPPLIER PERFORMANCE
IMPROVEMENTINTHE U.S. AND JAPANESE
AUTOMOTIVE INDUSTRIES
MASAAKI KOTABE,1XAVIERMARTIN2* and HIROSHIDOMOTO3
1
Institute of Global Management Studies, Temple University, Philadelphia, Pennsyl
vania, U.S.A.
2
Stern School of Business, New York University, New York, New York, U.S.A.
3 and Management
Department of Business Information, Tokyo University of
Information Science, Japan
relationship duration increases. Other results show relevant contrasts consistent with heteroge
neous sourcing behavior between the two countries. The findings the role of relational
highlight
assets and show that it is important to distinguish between and
simple techniques higher-level
technological capabilities when studying interfirm relationships. This research extends the lit
eratures on knowledge transfer, and the performance
buyer-supplier partnerships, dynamics
of interfirm and intrafirm relationships in general. Copyright ? 2002 John Wiley & Sons,
Ltd.
In recent years, researchers have paid increas quality. Influential studies concluded that buy
ing attention to the effects of supplier relation ers should foster high-involvement relationships
ships on buyers' competitive advantage. Studies with suppliers (Womack, Jones, and Roos, 1990;
argued that by involving suppliers extensively Clark and Fujimoto, 1991). As such practices
in product and process development, assemblers require ongoing knowledge exchange, Clark and
(buyers) could gain faster product development Fujimoto (1991) and Martin, Swaminathan, and
cycles, lower input costs and higher end-product Mitchell (1998) argued for an information-based
perspective on buyer-supplier relationships (see
also Takeishi, 2001).
Key words: vertical partnerships; interfirm collabora Another oft-mentioned factor benefiting buyers
tion; knowledge transfer; relation-specific assets; strategic is long-established supplier links. Prior duration
chain management
sourcing; supply of a buyer-supplier relationship (link duration,
Correspondence to: Xavier Martin, Stern School of Business,
New York University, 44 West Fourth Street, New York, NY henceforth) matters because what is effective in
10012, U.S.A. E-mail: xmartin@stern.nyu.edu long-established relationships may not be in newly
established ones (Stigler and Becker, 1977; Fich practices can be replicated in more recent links
man and Levinthal, 1991a). Link duration mea and which cannot.
sures the amount of experience that the supplier
and the buyer have in dealing with each other,
the resulting routines being described as relation PREVIOUS LITERATURE
specific assets
(Levinthal and Fichman, 1988).
While many firms held as of success A shift occurred during the 1980s in the
examples
ful knowledge management have developed com locus for conceptual strategy research on vertical
interfirm Earlier research drew
paratively long-lasting supplier links (Asanuma, relationships.
a sharp distinction between a firm and its
1989), research has shown little about whether
or how firms with shorter links can benefit from buyers or suppliers and described the search
for as a distributive
knowledge-intensive sourcing, and generally how competitive advantage
the benefits of knowledge transfer vary with link game (e.g., Williamson, 1975; Porter, 1980).
duration (see Martin, Mitchell, and Swaminathan, More recent research has emphasized the scope
and for value-adding relationships. Porter (1985)
1994; Dyer, Cho, Chu, 1998).
between suggested that complementarities might arise
study examines
This the connection
and link and raises between successive stages of the industry
knowledge transfer duration,
chain. Wernerfelt (1984, 1985) emphasized the
the issue of what it takes to enhance supplier
role of inputs as a source of competitive
performance. While past studies show that buyers
advantage and argued that a buyer can benefit
benefit when suppliers are intensively and durably
from long-established links allowing effective
involved in knowledge exchange, it is less clear
communication with suppliers. Research has
under what conditions this improves suppliers'
also examined ways of overcoming obstacles
operational performance. to durable buyer-supplier Axelrod
the following arise: What cooperation.
Thus, questions a long
(1984) showed how agents that develop
sourcing-related knowledge transfer practices can
term track record of cooperation and reciprocate
improve a supplier's operational performance?
partners' behaviors could sustain cooperation
And how do the
benefits of these practices
despite short-term incentives to defect. Williamson
vary with link duration? In addressing these
(1985) argued that asset-specific investment could
questions, we define a buyer-supplier relationship, be sustained among nonintegrated firms provided
or partnership, as the set of practices and
it was reciprocated. Granovetter (1985) argued
routines that support economic exchanges between to ongoing
that embeddedness lends strength
the two firms. A buyer-supplier link refers to
relationships in the face of uncertain information.
the fact that two
firms have been doing
the Fichman and Levinthal (1991a) and Asanuma
business continuously for a given period of
(1989) used the concept of relation-specific skills
time (the link duration). A supplier's operational or assets to describe how a buyer and supplier can
refers to the combination of product
performance develop over time distinctive routines that make
development efficiency, process improvements, ongoing collaboration more effective.
quality conformity, and short lead time (Womack In summary, research now
conceptual sug
et ai, 1990). For simplicity, we refer to this below that a firm can benefit from
gests harnessing
as with Such benefits
performance. complementarities suppliers.
We develop and test a multivariate model to can accrue more to firms that foster
strongly
examine what
knowledge transfer practices are durable linkages. However, it remains for empir
associated with supplier performance improve ical research to establish how complementarities
ment. The analysis contrasts two forms of should be pursued in practice, or how their effects
knowledge transfer: relatively simple technical may vary with link duration. Among the most
exchanges, and technology transfer associated with wide-ranging empirical studies are those by Wom
higher-level capabilities. We also examine the ack et al (1990) and Clark and Fujimoto (1991) on
moderating effects of prior link duration. By the automotive industry. Foremost, they described
studying how the benefits of the two types of two related features of the supplier arrangements
knowledge transfer vary with the duration of the made by effective assemblers. First, the division of
buyer-supplier link, we seek to identify which labor can be extensive. Able suppliers are involved
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J, 24: 293-316 (2003)
Gaining from Vertical Partnerships 295
exchange a
between buyer and a supplier. It trust matters (Zaheer, McEvily, and Perrone, 1998), though
past research suggests that it would operate independently of
takes time to develop the familiarity and expertise
relationship duration.
required for each partner to know when and how 2
For operational clarity, this concept can be contrasted with
to draw on the other's resources and when and Williamson's (1985) notion of plain 'specific assets' (omitting
'relation'). Relation-speci?c assets arise from the continued
how to contribute resources 1985;
(Wernerfelt, association between a given buyer and a given supplier in a
Asanuma, 1989). As two firms sustain
their routine-building collaboration process (Fichman and Levinthal,
business over a 1991a; Dyer and Singh, 1998). Thus relation-specific assets
relationship time, they develop are knowledge-based and team-embodied (rather than physical
joint understanding that is highly idiosyncratic but
artifacts). They exhibit (plain) asset specificity in that their value
allows uniquely efficient communication (Stigler is hampered if either partner is substituted (Wernerfelt, 1985).
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 24: 293-316 (2003)
296 M. Kotabe, X. Martin and H. Domoto
prior duration increases, especially when the tasks the occurrence and consequences of various forms
given the supplier are more complex and uncer of knowledge transfer (e.g., Galbraith, 1990).
tain (Levinthal and Fichman, 1988; Fichman and Our second premise is that gains in performance
Levinthal, 1991b). This suggests that as link dura arise from intentional and organized knowledge
tion increases, a buyer and a supplier develop transfer between a supplier and a buyer. The third
relation-specific routines so they become better premise is that the ability to benefit from knowl
able to share hard-to-transfer knowledge. How edge transfer depends on prior link duration. Build
ever, to our knowledge, previous research has ing on the second premise, we discuss the effects
not examined systematically how the impact of of two forms of knowledge transfer. Then, con
knowledge transfer on performance varies with sistent with the third premise, we examine how
link duration. the benefits from knowledge transfer can vary with
Finally, past research points to the relevance of the prior duration of the buyer-supplier relation
studying supplier
performance. Descriptive stud ship. Our empirical analyses control for several
ies have pointed out that even if the sharing other factors affecting knowledge flows and sup
of technical tasks and information benefits buy plier performance. Figure 1 presents a conceptual
ers, suppliers may receive no net benefit (e.g., framework for this study.
Sakai, 1990; Ramsay, 1996). Dyer (1996) reported
a composite measure of co-specific investments
Exchanges of knowledge
made by automotive buyers and suppliers that was
correlated with the returns on assets If knowledge transfer is expected to affect the
positively
of buyers, but unrelated with those of suppliers. outcome of buyer-supplier relationships, then the
Our study focuses on supplier performance, about extent and complexity of the knowledge trans
which comparatively little is known. fer deserve attention. Existing research suggests
a distinction between relatively simple technical
exchanges and higher-level sharing or transfer of
whole technological capabilities (Teece, 1977; von
PREDICTIONS these two forms of
Hippel, 1988). Conceptually,
exchange differ in the scope and level of the
Our predictions use
an evolutionary perspective involved. A consists of dis
knowledge technique
whereby boundedly rational firms stand to improve crete know-how to solve a
required particular oper
performance by seeking and exploiting techno ational problem, so technical communications per
logical knowledge (Nelson and Winter, 1982). tain to the relatively narrow and simple informa
They build on three linked premises. The first tional resources necessary to handle engineering
is that operational performance improvement is issues case by case. By contrast, a technology is
an ongoing process responding to ongoing tech a broader body of knowledge a set
encompassing
nological opportunities. Thus, studies have docu of related techniques, methods, and designs appli
mented continuing change among automotive firms cable to an entire class of problems (Rosenberg,
(Womack et aU 1990; Clark and Fujimoto, 1991; 1982; Arora and Gambardella, 1994). Thus its
Helper and Sako, 1995; Kotabe, 1998). Relat sharing or transfer involves higher-level capabil
edly, for a buyer-supplier relationship to endure, ities(Granstrand and Sjolander, 1988; Dussauge,
each partner must remain satisfied with the other's Hart, and Ramanantsoa, 1992; Szulanski, 1996).
past performance and outlook (Stigler and Becker, The associations between technology and
1977; Fichman and Levinthal, 1991a). Thus, all capability, and between technique and resource,
else equal, the average absolute performance of add meaning to the distinction between technical
suppliers in longer-established relationships should and technological knowledge. Relative to discrete
be higher than that of suppliers that have yet to resources, capabilities are higher-order, more
prove themselves over time. To avoid this potential complex sets of routines with broader applications
survival bias, we focus instead on the recent trend (Nanda, 1996). Technical knowledge, consisting
in a supplier's performance. Specifically, the out of narrower and more independent pieces of
come of interest is the improvement in a supplier's information, is a form of resources (Nanda, 1996).
operational performance over the last 2-3 years. A By contrast, capabilities are related to higher
2-3 year period is useful and sufficient to assess level, cumulative technology that is harder to
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 2A. 293-316 (2003)
Gaining from Vertical Partnerships 297
Technical Exchanges
Technology Transfer
Other variables:
Crucial Components
Supplier Size
Supplier Tier
develop and mold (Dosi, 1988; Nelson, 1991; as large and functionally diverse groups interact
Helfat, 1997). Transferring technology, sharing both within and across firms for sustained peri
it, or using it in support of a partner is a ods of time (Teece, 1977; Galbraith, 1990). The
challenging collaborative process. For simplicity, greater the scope of a knowledge transfer project,
we use the term technology transfer to refer and the more complex the knowledge, the more
to scenarios where firms transfer, share, and/or tacit and causally ambiguous the knowledge tends
deploy technology on their partner's behalf. to be (Kogut and Zander, 1993). This renders tech
These scenarios all involve complex challenges transfer particularly 1977;
nology costly (Teece,
of codification and communication capacity, and Szulanski, 1996).
are intertwined in practice (Mogavero and Shane,
1982; Godkin, 1988). While exchanging technical
knowledge can also be difficult, the challenge is Technical exchanges
more circumscribed.
The distinction between technical exchanges and Buyer-supplier relationships involve ongoing
transfer also has organizational mutual adjustment between the buyer's and
technology impli
cations. The coordination the supplier's design and production
required for exchanging operations.
small-scale technical knowledge is typically sim Many such adjustments are made readily by
ple. Arranging regular meetings or long-term per personnel sharing explicit engineering knowledge.
sonnel visits, for example, is straightforward if it This entails small-scale exchanges of technical
involves autonomous individuals or small work information (technical exchanges, in short). Past
units (von Hippel, 1988; Nishiguchi, 1994). As studies have argued that these
help improve
technical information tends to be explicit or at the buyer's performance (Clark and Fujimoto,
least codifiable, its exchange is a matter of ver 1991; Lamming, 1993). Suppliers stand to
bal or written communication (Kogut and Zander, benefit likewise when the partners steadily
1992). By contrast, technology transfer involves a share technical knowledge to solve problems
greater scope of activities and higher-level orga and enhance products and processes (Takeishi,
nizing principles (Rosenberg, 1982; Collis, 1991). 2001). Therefore, we expect a positive association
It requires extensive and dedicated coordination, between supplier performance and
improvement
Copyright ? 2002 John Wiley & Sons, Ltd. Strut. Mgmt. /., 24: 293-316 (2003)
298 M. Kotabe, X. Martin and H. Domoto
the involvement of the partners in ongoing projects akin to technical exchanges. This suggests
technical exchanges. that letting relationships mature, and thereby
accumulate relation-specific assets, fosters more
Hypothesis 1: The more technical exchanges successful knowledge transfer from the buyer's
between the buyer and the supplier, the higher standpoint. We examine next how the impact
the supplier performance improvement relative of knowledge transfer stands to vary with link
to 2-3 years earlier. duration, focusing on supplier performance.
In other cases, small-scale technical exchanges do We do not expect the trend in a supplier's perfor
not suffice to solve a particular problem or exploit mance, by itself, to be inherently higher in longer
a particular opportunity. Instead, more lumpy and links. Indeed, if parties to a longer-established rela
complex technology is required. Such knowledge tionship develop a longer-term horizon shielding
transfer requires larger-scale commitments of time the link from interruption in the presence of short
and groups of experts, with extensive coordination term performance dips (Granovetter, 1985; Fich
(Galbraith, 1990). Technology transfer, as we label man and Levinthal, 1991b), the main effect of link
it in short, refers to conceited projects that allow duration may appear weak or slightly negative.3
one partner to access or replicate complete tech This will also be true insofar as newer relationships
nological capabilities the other partner. Prop
of are able to exploit easy opportunities for improve
erly implemented, technology transfer enables a ment while older partnerships must build past those
more efficient division of labor (Clark and Fuji by tackling less obvious improvement projects.
moto, 1991). It also allows distinct improvements Hence, we do not predict a main effect of link
in technological competence throughout the indus duration on supplier performance improvement.
try chain (Lamming, 1993). Technology transfer However, we expect longer-established links,
should therefore be positively associated with sup because they allow more relation-specific assets
plier performance improvement. to develop, to magnify some of the perfor
mance effects hypothesized above. The longer a
Hypothesis 2: The more technology transfer buyer-supplier link endures, the more the relation
between the buyer and the supplier, the higher specific assets between the two parties stand to
the supplier performance improvement relative make pairwise knowledge transfer efficient relative
to 2-3 years earlier. to alternative sets.
partner
In simple technical exchanges, the scope of each
Beyond these main effects, research suggests knowledge transfer is narrow. Under these cir
that buyers seek to further the benefits of cumstances, assets need not be at
relation-specific
knowledge transfer by shaping the balance of their most useful. Nevertheless, Uzzi (1996, 1997)
technical exchanges and technology transfer. found that solidly embedded relationships facil
Asanuma (1985, 1989) argued that Japanese itated the continuous exchange of 'fine-grained'
automotive assemblers may engage in ongoing information, akin to simpler technical exchanges,
narrow exchanges but prefer to delay more
demanding collaboration until the suppliers have 3
An alternative possibility, as an anonymous reviewer pointed
proven themselves and become familiar (see also out, is that firms with links become and
long-term complacent
Nishiguchi, 1994; Dyer and Nobeoka, 2000). We under-perform. This presumes that the buyer and the supplier
face relatively weak incentives in relation to each other. The
argue that time-bound communication efficiencies
buyers we study have increasingly learned to combat compla
can explain such patterns. Knowledge transfer and eliciting continuous
cency by monitoring suppliers closely
is more difficult when buyers and suppliers improvement (Helper and Sako, 1995; Mudambi and Helper,
1998). Indeed, long-tied partners are plausibly better able to
lack familiarity. Thus MacDuffie and Helper
overcome hazards of opportunism (Delios and Henisz, 2000).
(1997) found that Honda, though a highly the alternative scenario may explain the relative
Nevertheless,
capable assembler, experienced mixed success in inefficiency attributed to in-house parts divisions of some auto
motive assemblers that are shielded from outside competition.
transferring practices to three local suppliers that
Our sample excludes such closely held divisions. It consists of
it had added relatively recently in the United (and buyers) that have been selected against competi
suppliers
States. In response, Honda focused on narrower tion over time so they could not afford durable complacency.
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 24: 293-316 (2003)
Gaining from Vertical Partnerships 299
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 24: 293-316 (2003)
300 M. Kotabe, X. Martin and H. Domoto
presidents/CEOs of 400 randomly selected firms of 21.3 percent. The final sample, upon exclu
and their major affiliates, asking them to provide sion of cases with missing data, consists of 105
names and addresses of managers directly respon responses (18.2%).
sible for the respective firms' supply relations with In defining both samples, we excluded sub
other parts suppliers and/or with automobile man sidiaries where an automotive assembler was a
ufacturers. At this stage 185 firms expressed will major shareholder (20% or more). Thus, the res
can be considered autonomous
ingness to participate in the study and provided a pondents of the
manager-in-charge
to contact. A questionnaire
was buyers. We excluded suppliers owned by foreign
then mailed to the 185 managers, along with a firms, including Japanese-owned suppliers in the
cover letter that the nature United States and the few U.S.-owned facilities in
personalized explained
of the study and informed the managers that their Japan. Because we asked the respondents about
names had been provided by their respective com their major domestic customer, all Japan-based
with to a Japanese
pany heads. All survey solicitation letters indicated respondents replied respect
that companies in the survey would buyer in Japan. Likewise all U.S. respondents
participating
receive a summary of the research and comparative reported locally based buyers, except one response
benchmark results tailored to individual firms. The that was excluded from the sample to ensure
maximum contrast. The cover a wide
managers were asked to choose their major buyer responses
letter, 104 questionnaires were received, responses are demographically comparable. Most
follow-up
respondents held upper-management positions. In
for a response rate of 26.0 percent of the original
the United States and Japan respectively, they had
sample or 56.2 percent of those who were identi
4.66 and 5.58 years of experience in their cur
fied to participate. After eliminating responses with
rent positions = 1.19, =
(t p 0.23) and 14.01 and
missing data, the sample for analysis consists of = 3.68,
19.90 years with the same company (t
97 questionnaires, a rate of 24.3 percent or 52.4
=
of p 0.001). This exceeds the 2-3 year timeframe
percent first-stage respondents. our dependent
In Japan, Teikoku Databank's COSMOS data
for measuring variable, adding to
confidence that respondents were knowledgeable.
base provided the sampling frame. Teikoku Data
Given Japanese life-long employment practices,
bank, founded as a corporate credit research firm in
the difference in corporate tenure is understand
Japan in 1900, produces the largest and most rep
able; however, the respondents had comparable
utable corporate database in Japan. We purchased
experience in their current positions. On aver
the list of the 600 largest automotive components
on age, the U.S. and Japanese suppliers are similar
suppliers, based annual sales. Twenty-three
in annual sales
(about $440 million) and in total
records were unusable due to missing addresses,
employment (3100-4500, not statistically differ
leaving an initial sample of 577. Given the costly
ent). These data show that the respondents have
nature of international survey research, we con
requisite expertise and involvement in supply chain
tracted out survey execution
(questionnaire mail
management, and that the U.S. and Japanese sam
out, collection, and data entry) to a highly rep
ples match well.
utable independent marketing research company
Two issues commonly raised concerning survey
in Tokyo, Japan. The survey solicitation letters are nonresponse
methodology bias and common
were sent to the presidents/CEOs of those Japanese method variance. We examined the nonresponse
suppliers and offered the same incentives as for
U.S. respondents. Following the first mail-out and 5
For the U.S. data, although the source directory does not
a follow-up letter a few weeks later, 123 ques small jobshop-type it represents a
represent parts suppliers,
tionnaires were collected with a response rate significant portion of the automobile parts industry. For the
Japanese data, the 600 largest suppliers account for more than
90 percent of sales to the Japanese automobile manufacturers.
As a result, each sample may under-represent the smallest firms.
vital demographic characteristics were excluded. These criteria However, our focus was on representativeness of the economic
were used to generate the sampling frame, from which 400 impact of buyer-supplier relationships in the industry?which
companies were randomly selected. this achieves?rather than pure randomness of the population.
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J, 24: 293-316 (2003)
Gaining from Vertical Partnerships 301
Marketing/
Manufacturing/
Operations/General
Manager?24
Unreported?19
U.S. President/CEO?9 4.66 years 14.01 years $440.0 mil. 3110.9
Vice President?36
Director?13
Marketing/
Manufacturing/
Operations/General
Manager?39
Unreported?0
Note Japanese and U.S. job titles are not =
1.19,
=
3.68, The average t= 0.70,
In some = 0.23 = 0.001 annual sales = 0.48
necessarily comparable. p p p
cases, director or functional would be
manager positions in Japan are comparable
comparable to vice president at ?162/$
positions in the United States
issue with the procedure suggested by Armstrong based on 5-point Likert scales. In order to assess
and Overton (1977). We performed ?-tests compar construct validity, items across the scales were
ing early and late respondents on key demographic subjected to principal components factor analysis
variables, namely supplier tier, sales volume, and with varimax rotation. Appropriate factor solutions
employment. In each sample we found no signif were obtained after iterated removal of items that
icant differences between early and late respon failed to load on any single factor. The pattern
dents. This suggests that nonresponse would not of observed loadings indicates that the multi-item
likely bias the findings. We used Harman's one scales measure independent constructs, thus fur
factor test to address the issue of common method ther supporting unidimensionality and discriminant
variance. If that were a serious problem, we would validity of the scales. Each multi-item measure
expect a single factor to emerge from a factor anal was obtained by computing the average score pro
ysis or one general factor to account for most of the vided by the respondent across the relevant items.
covariance in the independent and criterion vari Table 2 reports descriptive statistics.
ables (Podsakoff and Organ, 1986). We performed
factor analysis on items related to the predictor
Criterion variable
variables and criterion measure. No general fac
tor was apparent in the unrotated factor structure. The measure of
supplier performance captures
Therefore, no common method variance problem a supplier's performance improvement relative
was detected. to its position 2-3 years earlier. Such a trend
measure avoids the relationship-survival bias
that might result from comparing the absolute
Measures and statistical methods
performance of buyer links with sharply different
The measures used for this study are presented durations. The measure includes dimensions of
in the Appendix. All multi-item measures are operational performance that are central concerns
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 24: 293-316 (2003)
302 M. Kotabe, X. Martin and H. Domoto
O o
o o
o o
o o o ^
O 00
o o 88
^ ? -h ?
I
Ooo? O ^t en
O O Tt O ^h en
O CM^ O O en
-< ? ?
O en cn ^t o r- cn i-H
O m Os cN O vo ^O CN
O O O 1-H O O O O
-h' O ? ? -h' ? o o
O in VOOs oo O en ^o en en
o m m cN os o o ti- in o
O ^HO O ^H O ^h O O CN
-h' ? O O ? -h' o ? ? c5
I
o oo m i?io tj O On in ^h en en
o m oo en o
-H -H O O CN
oo O cn r- oo ^o en
o o ^t -H o o -H
o o o o o O O ? ? ?
O (N ^h m in on ^o O Os o in en o O
o vo ^o in cn cn m O oo m ^O en
O CN -H O -H O O o m in CNO O en
? ? ? ? ? O O O O O
O vo en ^- in oo r- cN O en O ^o r- ^ in tj
O ^h m o r- tj
O ^t <N O O 1-H *?I T-H en cn CNo en
^ o ? ? ? ? ? ? ^H*
? ? ? O ? ? ?
o m on m r- r- ^- ^h oo o r-Hr- in
o oo o o en O ^o CNO O OS O T-H cn r? ^h Tt en
O en CN CN O O O ^h en
-h' O ? ? ? ? ? ? ? nodo ? ? ? ? ?
go ^
o o en r ^sO
O CN in O en
VO OOo o en O CN rt oo os m ^O
oo ^ en in 00M^\? r- ^t oo tj
CCd ? ? o o d ? h (s o ? ? ? ? dd^ood
CNOs
Os ON
00
cn r? oo on oo ^ m ^o CNO m
cd m cn *-hos o ?n cN r m os cn Os
oo en en ^si ^ 00 ONh OrnoN
en en cn en en O vo ^h ? en en CN en en ? vo t-- ?
CN ^h en CN
in
en
T3
Gcd
cd cd
PL,
4)
> > O
O Oh c
ex ? Oh
Oh
S 00 in g
<D
>^
? C/3 O O
?
00 C
C "O t? *V
G 1(D?' (D43
Oh 2 cd Oh 43 o ^-^ Oh
?3 XM o <DOh O , <DOh
pq >, c s-S >> C i
cd ' ai l? W)M
O*
!
i
Ci, <D.a o ^ *-! s SIQ
s si ?-i 3 ^?^h D
C C *-i
? ? ?,? Oh Oh i 'Oh+L <DO Oh ?^
: Oh <*> S? Oh O Oh
r^ r^ S D .0
Co&
^ 00 U J 00 (?h
g^oo H PQ Oh
cn en ^t in ^o r^ oo os in vo o oo os
I
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 24: 293-316 (2003)
Gaining from Vertical Partnerships 303
for supplier firms and that the respondents could index (PNFI, the common measure of compara
expertly assess. Interviews with executives of tive fit parsimony) both exceed 0.65 in the two
supplier firms indicated that improvements in factor model but fall below 0.63 in the alternative.
product design, process design, product quality, The root mean squared error of approximation
and lead time are key indicators of their (RMSEA, a standard indicator based on residu
performance and that a 2-3-year timeframe is als) is on the 'good' side of the 0.10 cut-off with
suitable to assess knowledge transfer and its our two-factor specification (0.09) but poor with a
consequences. These performance dimensions are single factor (0.14). Finally, interfactor correlation
also consistent with those used by assemblers is significantly below 1 (p < 0.001), indicating
to assess suppliers (Cusumano and Takeishi, discriminant validity to our model that the single
1991; Choi and Hartley, 1996). Therefore, factor model would lack (Bagozzi and Yi, 1993).
we use items measuring trends in product In summary, the basic and confirmatory factor
process and lead time analyses alike indicate a sharp and valid distinc
design, design, quality,
achievements to represent tion between Technical Exchanges and Technology
supplier performance
Transfer.
improvement.6 These four items, when introduced
in the factor analysis all the items The moderating variable, Link Duration, was
alongside
that make up our independent loaded measured as the number of years since the buyer
variables,
onto a single measure (Cronbach's alpha = 0.83). and the supplier began their business relation
This variable was labeled Supplier Performance ship. Some studies argue that because experi
ence tends to accumulate at a decreasing rate, a
Improvement.
nonlinear transformation of link duration may be
appropriate (e.g., Heide and Miner, 1992). Thus
we report two sets of estimates. One uses the
Predictor variables
untransformed value of link duration. In the other
The measure of Technical Exchanges contains we take its natural logarithm (Cohen and Cohen,
six items pertaining to common, informal 1983).
communication between engineers (Cronbach's
= The measure of
alpha 0.83). Technology Other variables
Transfer consists of five items describing
transfer of higher-level technological capabilities Control variables are described in the Appendix.
= Each may affect supplier performance and may
(Cronbach's alpha 0.83).
As the distinction between technique (Techni also explain firms' propensities to transfer technol
cal Exchange) and technology (Technology Trans ogy and/or develop a long-term link. Two multi
fer) is important to the study, we used confirma item measures represent the buyer's familiarity
tory factor analysis to verify that these two mea with the supplier's past performance and the sup
sures are distinct. We our two-factor plier's technical influence. Their alphas exceed
compared
model against a single-factor model that would 0.70. A binary measure controls for the class
apply if the distinction were blurred in measure of components the supplier makes, based on the
ment (Steenkamp and van Trijp, 1991). The results importance of the underlying expertise for assem
favor the two-factor blers. Supplier size is measured as the number of
unambiguously specification.
A large x2 difference indicates that a single fac the firm's employees. This measure provides con
tor cannot approximate the two measures (x?[{{
= sistency across countries independent of currency
143.96, p < 0.001). The parsimonious goodness exchange rates. To ascertain that any skewness did
of fit index (PGFI, the common measure of abso not affect our results, we replicated our analyses
lute fit parsimony) and parsimonious normed fit after log-transformation of the supplier size mea
sure. Finally, a binary variable indicates first-tier
supply relationships.
6
The existing literature argues consistently that near
simultaneous improvement along multiple performance dimen
sions is possible and indeed expected, due to the systemic nature Regression methods
of automotive production. See, for example, Womack et al.
(1990); Clark and Fujimoto (1991); Dyer and Ouchi (1993); Estimates were obtained by ordinary least squares
and Sako residuals
Lamming (1993); Helper (1995). regression. Upon examining visually and
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J, 24: 293-316 (2003)
304 M. Kot abe, X. Martin and H. Domo to
conducting White's (1980) test, there is no evi mentioned, the F-tests simply confirm the univari
dence of heteroskedasticity in the data. OLS regres ate results. The discussion below reports interac
sion is appropriate under these circumstances.7 tions between our continuous variables. We also
We conducted a Chow test to examine whether replicated the analyses with up to four splits of
the pattern of coefficients is the same for the U.S. each key variable (e.g., dummies and interactions
and Japanese samples. For basic models exclud for quartiles of Link Duration) and found consis
ing interaction effects, the Chow test F(8,195) tent results.
was 3.82. This indicates that the regressions differ
between the two samples, but not overwhelmingly
Results for the U.S. sample
so. Accordingly, we report separate analyses of
the U.S. and Japanese samples. This etic approach The adjusted R2 for the U.S. results in Table 3
makes it possible to observe similarities and dif is in a satisfactory range (0.173-0.209). The
ferences between countries in response to simi base analysis with untransformed link duration is
lar factors (Craig and Douglas, 1999). It allows reported as Model 1.1. The coefficient for Techni
cal Exchanges is positive < 0.01). The coef
estimated slopes and intercepts to vary, and there (p
fore permits a more of the ficient for Technology Transfer is positive but
thorough comparison
effects of knowledge and link duration smaller in magnitude and not statistically signifi
exchange
in the two samples. Etic research designs have cant. Model 1.2 adds the interaction between Tech
been commonly used in our research context (e.g., nical Exchanges and Link Duration. The main
Cusumano and Takeishi, 1991; Helper and Sako, effect of Technical Exchanges and its cross-term
with Link Duration are both positive. However,
1995; Dyer et a/., 1998).
only the main effect is statistically significant.
These results support Hypothesis 1 but not Hypo
theses 2 and 3. Only Technical Exchanges has
RESULTS a main effect, and that positive effect does not
become substantially stronger as link duration
Table 3 reports the results for the U.S. sam increases.
ple. Table 4 reports the same specifications for Model 1.3 reports the interaction between Tech
the Japanese sample. For each country, two sets nology Transfer and Link Duration. The main
of analyses are reported: one where link dura effect of Technology Transfer remains nonsignifi
tion is untransformed and the other where it is cant, as does that of Link Duration. The cross-term,
log-transformed. Each table reports four analyses meanwhile, is positive and statistically significant.
for each specification of link duration. The first This provides support for Hypothesis 4. Model
analysis contains a base model without interac 1.4 includes both interaction effects. It confirms
tion terms. The second includes a single interac the interactive effect of Technology Transfer and
tion term, between Link Duration and Technical Link Duration after controlling for the interaction
Exchanges. The third contains a single interac of Technical Exchanges and Link Duration. The
tion term, between Link Duration and Technol latter interaction is not significant. An F-test of
ogy Transfer. The fourth analysis includes both the two cross-terms indicates that they are not sig
interaction terms. Below we discuss each variable nificant as a pair (F = 1.34). However, the inter
in turn. In addition, we conducted comprehensive action of Technology Transfer and Link Duration
is. Since the two-interaction models
multivariate F-tests of matched variables (e.g., a throughout
main effect and its interaction). Unless otherwise Tables 3 and 4 simply confirm the simpler mod
els, the remainder of the discussion emphasizes the
interpretation of more parsimonious models with
7
Given that some variables have multiple indicators, an alterna zero or one interaction.
tive method to test the full model might be structural equation
The last four models of Table 3 report results
modeling (SEM). However, sample size considerations preclude
the use of SEM. The ratio of observations to estimated for the U.S. sample when link duration is log
param
eters would fall far short of accepted guidelines (Bentler and transformed. They are generally similar to the first
Chou, 1987). This is even more so when it comes to our interac cross-term
four. One difference is found when the
tion hypotheses, since testing such hypotheses by SEM custom
of Technical Exchanges and Link Duration (log
arily requires splitting each sample into at least two subsamples
(preferably more), and doing so for each interaction of interest. transformed) is introduced (Models 1.6 and 1.8).
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J, 24: 293-316 (2003)
Gaining from Vertical Partnerships 305
enoosONcn^HVO-H r- cN ^OenoocNcNOONOooen ? ^ m
r-enomr-r-r-^H CN ^T i>ONenr-ooooo^-H^ti> *?ir^ oo
M^rHpH(S(NrHrH O ^H oo^^minoom^H rto\ ^
? o ? o ? o ? o ?o?o?o?o?o
^t?
?
en in oo cN en On oo\OHMnooa\m ? ^ oo
in en 00 vo 00 Os ^h CN \ooo>Dh(N\OH-HO^ ^o m oo
en ^h (NMrHO O -H Oo^^Hcnenooin^H cno ^
? O ? O o o ?????????? en^-H* ?
I
On r** CN ?- m Os O en inenoo^H^rcNONOi>^r * ^ os
oo i?i rt m \D ^h en ^t i>o>a\r-oooo^enr- o> ~ . vo
~_. oo
CN ^?T-H -H O -H O -H ooo^H^-inoo"
o o o o o o ? o oooooooooo
r- in en On o ^o ^00\hO\xhO(N^ ? vo en
oo en 00 ^h in CN ^HOhM?HH\o\o in r- o
en ^h O -H O -H OO^h^hOOOOtJ-^h oo^O ^h
O O O o O ? oooooooooo ^' o ?
^rfCNr-^OCNOOrt ONCNinr-r^^ooocNcN * en o
^ooom^ooo ^h en (Na\O\^rH(S|HrH00h OS OS O
cn^OOO^OO O -H OOO^OOOO^r^ rt On CN
?O?O?O?O ? w? OOOOOOOOOO CNO ?
I
00 o ^OOs mooh^^^oao^o # o os
\D CN CN CN CnOOO\^rHrHO^O\^! m ^1" O
en ^h O -H OOO^hOOOO^T?: ^or- CN
? O o o o o ? w? oooooooooo CNO O
I
^h ^h Tf r- tj- o O 00 Tj-eninooor-asOONin t os ^o
^h r- o o ^- cN CN en
en ^h o O O ^h O -H OOO^OOOO^T^h cNO> ^h
? ? ? ? ? ? ? ? OOOOOOOOOO CNO ?
l> CN m On o>ONooo>cN^j-ooino CN
00 CN m ^h m CN enoooo^ooO'-H^Henvo Tto ._
00
en ^h O -H O -H ooo-hoooo^i--h t^vo -h
O o o o oooooooooo ??o
en CN
+ + +
-S
OJO
? V
c ^
il
O
? So G G ? c o
^ gcd &
^ G G ? .?
X XH O ^-^_^ Oh S3V
g -s c
o s
O
>^
?O ^ o ?cdfir-1
Oh G
O U C * e a,
a G S o ?
g So 5-hO
'S-S .S g ? cu V> % ?*-h
5-HoG ^
DO
?4^
43 G G 'i* G Oh
73
a)
3 43 ? 43Jp
S G Q-S .2
G <DOh?t? O
&^ U is
^3 43 ^ t? ^
?) o si KMOh Oh43 G?H
44
G ??"^ 5-h "Sv
G & ex,
0. H O PQ fc 43
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J, 24: 293-316 (2003)
Gaining from Vertical Partnerships 307
\o o in CN in O incN cNooGN^tt^-cNooenvoen *^ ^)
Os CN -H O ^t ^t T-*r* OMhMmmoOHin en ^t *-h
O ^H en in cN ^h cn^h rj-oo-H^-^tooo^H inin in
? ? ? ? ?O ?O?O?O?O?O CN?0
se On ^O os cn ^-Hoom^-r-^ot^-eno^- on cn en
_
s: o en ?h ir) T-* HOOH^o^oon^t m r- o
O -H en i?i CN^h rtO^^H^OOOO^-H o^t in
O O oo ?o?o?o?o?o do o
CN CN ^h o Os se Os tJ- aNOOen^cNt^ooenr-HTj- * o oo
O CN en Tt O O *?i ^h asooascNcN^HOOOm ^hcn o
O CNO O cn^h enoO^OOOOO^H r-n r- in
? ? ? ? ?O ?O?O?O?O?O *-h*
o ?
Os l> ~i> en ^h se m CN O
o o en ?h O" "en ^f oo oo o
O -H CN en in
? ? ? ? ?
en CN
PC X
+
(U?
J? V
c ^
Sicd *
O
G c o
^ G ? cd <D
^ cd G e(D^43 .Sh
O cl V
? ? sO ^00 -Ci O >5 ? Oh ? G^
* I2 *-?
VhO
G G
OS V..<u
o
43 75 Oh
^3 43
s e * Je Q-g
'
?.2
&
3
cd
9? G G<->
IIP
Oh - +1?
G <*>
G is
V
W o a? 'S G Ph Oh43
G ?i
Oh ""
g .a G
g
3 ex,
O PQ ?0 [Ih < c/3 *
derivative of the equation with respect to Tech in Japan appear to be particularly beneficial if tech
nology Transfer: nology flows in a well-settled interface (see also
Branstetter, 2000).
3(Supplier Performance Improvement)/
3 (Technology Transfer)
DISCUSSION
= -0.315 + 0.245* ln(Link Duration)
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 24: 293-316 (2003)
Gaining from Vertical Partnerships 309
generalizing and implementing this recommenda more prone to use technical exchanges, but the
tion.11 Our findings can help explain why U.S. difference between samples is not statistically sig
firms appear more willing to deal with new part nificant. This pattern may explain why supplier
ners, while Japanese firms have been more reluc turnover has generally been higher in the United
tant to
change partners. States, but also why U.S. buyers have shown great
In the U.S. there is evidence that
sample, loyalty to selected suppliers (Helper, 1991b). Over
smaller-scale technical exchanges promote sup all, this clarifies how U.S. firms have come to
improvement. Furthermore, the
plier performance expect and perhaps should still use more flexibil
U.S. model with the link duration variable untrans
ity in partner selection, but also shows that they
formed fits the data slightly better than when link should not ignore the benefits of long-term rela
duration is log-transformed, while the reverse is
tionships.
true for the Japanese data. This suggests that the A further difference pertains to the time that it
distinction between medium- and long-duration takes for technology transfer to start paying off.
links is more relevant in the United States than This takes longer in the United States than in
in Japan, where the primary distinction is between
Japan (5.1 years vs. 3.6 years, different at p <
short links and longer links. However, no signif
0.01). While modest in absolute, this gap indi
icant effect of
technical exchanges is found in
cates a difference in the initial incentives to foster
Japan. The pattern of results for Japan suggests
long-lastingrelationships. Our findings may also
that immediate pay-off from technical exchanges
help explain differences in the propensity to rely
may be elusive (relative to Japanese competition), on extensive transfer.
while transfer is most beneficial once technology Having experi
technology enced higher turnover in the decades
the sourcing relationship has been in place for a leading to
the 1990s, though not previously (Helper, 1991b),
moderate period. Thus, failing to reconduct a rela
U.S. links are more recent on average but also con
tionship past the first purchasing cycle?or after
sist of some very old ties (up to 92 years). The
subsequent cycles?entails a substantial opportu fact that many U.S. suppliers with shorter-lasting
nity cost. This may explain why Japanese buyers
have long been described as comparatively reluc relationships expect little benefits from technology
transfer plausibly helps account for the difference
tant to change suppliers (and vice versa).
in technology transfer emphases between the two
What explains that technical exchanges pay off
countries. Indeed, when both samples are split
in the U.S. sample but not in the Japanese sam
according to the median link duration,
U.S. the
ple? While it is theoretically possible that U.S.
firms are inherently more efficient at sharing such difference between is less among cases
countries
with longer-established links (0.342, p < 0.10 by
explicit knowledge (see Hall and Hall, 1990), prior
research on the automotive the i-test) than among more recent links (0.412,
industry does not sup
port this view Liker et al., 1996). A more p < 0.01).
(e.g.,
is that U.S. buyers and sup Technical exchanges may offer early benefits for
plausible explanation
have increased their commitment to U.S. firms, and should thus be encouraged. How
pliers recently
joint problem solving (Helper and Sako, 1995). It ever, given the contingent benefits of technology
is therefore comparatively easy to find technical transfer, a challenge is to make high-level technol
exchange opportunities that enhance performance ogy transfer succeed in recent relationships. Our
in the United States, whereas that is more diffi results suggest that U.S. firms adopting aggres
cult in Japan where the practice has long been sive technology transfer practices prematurely are
exploited extensively. In both samples, meanwhile, likely to find these efforts ineffective.
the benefits of technology transfer remain contin In Japan, meanwhile, buyer-supplier links have
gent on prior link duration. Consistent with our started to fray in recent years. Relationships may
interpretation, ?-tests show that Japanese firms, now turn over faster, even for strong assem
whose links have been in place longer on average blers like Toyota (e.g., Ahmadjian and Lincoln,
(p < 0.01), also tend to engage in more technol 2001). Thus, a challenge would again be to fruc
ogy transfer (p < 0.01). U.S. firms are slightly tify recent links. While technology transfer starts
paying off sooner, engaging in technology trans
1' fer too early is also problematic. Furthermore, our
Our discussion focuses on the hypothesized effects, but differ
ences regarding control variables are also noteworthy. results show that small-scale technical exchanges
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 24: 293-316 (2003)
310 M. Kotabe, X. Martin and H. Domoto
do not improve performance relative to estab make the information flow bilateral (Arrow, 1974;
lished Japanese competition. This suggests a tran Sengupta and Abdel-Hamid, 1993).
sition challenge when partners are replaced. The In this respect, studies of best practices often
resulting trade-off may explain the relative rigidity focus on the buyer's role as source of knowl
of Japanese buyer-supplier links. Unless relaxed, edge, while paying less attention to the supplier's
such rigidity could hinder adaptation in the face of function other than as a recipient. Yet our survey
global competition, radical technological change, shows some tentative evidence that firms rely on
or simply slower growth (Tezuka, 1997; Robinson bilateral knowledge flows. One of the items load
and Martin, 1997). ing onto Technical Exchanges pertains to bilateral
Given our findings, it is worth asking whether communications, and items loading onto Technol
there is a joint effect of technical exchanges and ogy Transfer include questions about technology
technology transfer. Even if technical exchanges passing from the supplier to the buyer as well as in
have no main effect, can they complement higher the reverse direction (see the Appendix). While we
level technology transfer, and if so how? To cannot statistically separate items describing uni
explore this issue, we examined their interaction lateral and bilateral knowledge flows in our data,
as well as their three-way interaction one plausible way to build up relational assets is
pairwise,
with link duration. We found the new interactions to foster two-way communication, provided each
statistically insignificant, and our other results party has sufficient competence. Future research
robust. Thus, the effect of technology transfer does should thus examine information flows in both
not appear contingent on simultaneous technical directions, as well as through third partners.
(and reciprocally). While Another solution is to leverage firms' capac
exchanges incomplete,
this additional analysis suggests that intense tech ities to transfer technology, holding link dura
nical exchanges would not suffice to make tech tion constant. An extensive literature discussed
transfer beneficial in short-duration rela the challenges of transferring knowledge, espe
nology
if it is tacit and team-embodied (Godkin,
tionships. We believe, rather, that sufficient time cially
would need to elapse for relation-specific assets to 1988). In response, some research has empha
take hold. Technical nevertheless sized the ability of a firm to take in knowledge
exchanges may
from outside (e.g., Szulanski, 1996). Martin and
play a useful role for their own benefits (evidenced
in the United States), and as a means of sustaining Salomon (2002) argue that two distinct capabili
a relationship while assets build ties contribute to successful interfirm knowledge
relation-specific
Steensma and transfer: source transfer capacity, which pertains
up (see Lyles, 2000). Undertaking
technical exchanges may thus be valuable in help to a transferor's ability to transmit knowledge out
assets accumulate over time, ward, and recipient transfer capacity, which per
ing relation-specific
that the relation-specific assets are subse tains to a transferee's ability to assimilate knowl
knowing
quently for
critical beneficial technology transfer. edge from a willing external source. All else equal,
The procedures deployed and adjusted by Honda to the most successful technology transfer will accrue
transfer best practices to U.S. suppliers (MacDuffie in pairs of firms that possess the requisite combi
nation of source and knowledge transfer capacity.
and Helper, 1997) are consistent with this view.
Our results suggest that link duration also matters.
Extending both arguments, it presumably takes
On assets time for partners to fine-tune their transfer capac
building up relation-specific
ity and specialize it to each other. Furthermore,
For U.S. and Japanese firms alike, then, a key chal if bilateral transfer is the purpose, then each firm
lenge is how to build up relational assets to render would need time to develop both source and recip
technology transfer effective. We suggest two plau ient transfer capacity.
sible approaches to complement the passage of Our results also suggest that strong perfor
time. One solution may be to focus on accelerating mance potential exists in very long-lived relation
the idiosyncratic learning process whereby buyer ships. However, the extreme durability of some
and supplier develop joint understanding and rou buyer-supplier links raises the possibility of an
tines. A known way to accelerate learning through alternative explanation for the results. If interfirm
communication, the type of process that gener partnerships endure in the extreme, then link dura
ates relation-specific assets, is to add feedback and tion may be confounded with firm age, which
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J, 24: 293-316 (2003)
Gaining from Vertical Partnerships 311
itself can be associated with superior performance common method artifact to explain why one inter
(Delacroix and Swaminathan, 1991). To examine action set is strong while another is weak. Specif
this alternative, we collected measures of the age, ically, since the terms for both interactions were
in years, of respondent suppliers and their buyers. measured through self-report, common methods
We could only obtain meaningful samples for the alone cannot explain why one effect (Hypothe
= 61 for the measure
Japanese data, where N of sis 4) is statistically significant while the other
= measure is not. the existence of
supplier age and N 95 for the of buyer (Hypothesis 3) Third,
age. Any concern about confusing firm age and effects in one sample but not the other (e.g., regard
link duration would be strongest in the Japanese ing Hypothesis 1) cannot easily be attributed to
sample, where the automotive industry is younger
common variance. Still, we recognize the need
and supplier links tend to endure longer. In this to replicate the results with procedures not con
instance, neither variable was significant, and our sisting of common methods. Such analyses would
further enhance the external of those
key results were essentially unchanged. This shows validity
that link duration moderates reported here.
performance improve
ment independently of supplier or buyer age. More generally, it would be interesting to repli
Furthermore, this supplemental pro cate this research with other measures of supplier
analysis
vides some insight into the nature of the relational performance, preferably objective ones like mar
assets associated with link duration. It is relevant ket share or financial returns. We would expect
to ask just how partner-specific relational assets the mechanisms we described to help explain
are. Would skills developed in dealing with a par such outcomes too, as our mea
performance
ticular partner be useful only while dealing with sure of operational performance plausibly predicts
that partner, i.e., be strictly relation-specific, or a supplier's overall competitiveness. Indeed, our
would the skills readily apply to alternative part items figure among the main supplier selection
ners, i.e., be nonrelation-specific? Our data do not criteria used by buyers (Cusumano and Takeishi,
allow us to conduct the most direct possible test, 1991; Choi and Hartley, 1996). Furthermore, man
the duration of past links with alterna agerial assessments and objective indicators are
whereby
tive partners would be used as predictors of the generally consistent (Venkatraman and Ramanu
of a focal relationship. Nevertheless, jam, 1987), including when it comes to mea
performance
we find that while supplier age is significantly suring impacts of interfirm relations (Simonin,
correlated with focal link duration (+0.30), buyer 1997). We could not use financial or market data
age is not (+0.09). Using age as an (admittedly
because many automotive suppliers are privately
indicator of the cumulative effect of owned, and the two-country design would make
imperfect)
alternative we find that it has no some financial comparisons problematic. Never
relationships,
effect on technology theless, in a suitable context, such replication
moderating transfer, unlike
focal link duration. This indicates that the rela would be worthwhile.
tional assets we measure via link duration are truly It would also be interesting to examine the
relevance of other dimensions of buyer-supplier
relation-specific.
relationships. In particular, there may be deter
minants of relation-specific assets other than link
Other for future research duration, such as the range of products being
opportunities
traded (Levinthal and Myatt, 1994). Our find
One possible shortcoming of this research is com ings about the importance of link duration justify
mon method variance. Although the data were more research on the sources of relation-specific
operationalized through self-reports measured at assets.
one time, there are several reasons to believe that
our findings are not an artifact of common meth
ods. First, one-factor tests showed no evidence of CONCLUSION
a problem. Second, key findings in the study are
interaction effects. Even if respondents tended to This study examines practices that contribute to
associate successive items, common method vari supplier performance improvement among U.S.
ance is unlikely to distort interaction-based anal and Japanese automotive suppliers. It addresses
ysis (Brockner et al.y 1997). It is difficult for a value-adding mechanisms that long-established
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. /., 24: 293-316 (2003)
312 M. Kotabe, X. Martin and H. Domoto
interfirm partnerships can enable. Our predictions vertically or diversify (especially via acquisition),
point to the relevance of knowledge transfer in horizontal technology transfer deals between
in general, and in particular of distinguishing rivals, and in alliances broadly defined. In each
between simple technical exchanges and higher case, separate organizations must share knowl
level technology transfer. They highlight the role edge for joint advantage to develop. This requires
of prior relationship duration and describe how effective knowledge transfer mechanisms. Though
this moderates the impact of the two types of governance and initial knowledge positions may
knowledge flows. Thus, this work contributes to vary, relation-specific assets stand to be critical in
the study of buyer-supplier relationships, while enabling the pooling of corporate capabilities. This
extending prior research on relational assets. makes link duration highly relevant in setting strat
The empirical analysis confirms that knowl egy in these contexts too. For example, merging
edge transfer can be associated with supplier per firms should be cognizant of the state of relational
formance improvement. The interaction effects assets before they implement large-scale knowl
that we uncover work in similar ways in the edge transfers as part of their integration plans.
United States and Japan. The regression analy Generally, relation-specific assets may precondi
ses also identify some potentially important dif tion the search for intrafirm and interfirm synergy.
ferences between the two samples such that one Thus, research building on the present study will
should not take either country as an unambiguous be well worth pursuing.
basis for generalizing. These findings shed light
on the conditions for any convergence between
U.S. and Japanese practice. Given the accumu ACKNOWLEDGEMENTS
lated difference in mean link duration (10.6 years
in our sample) and the residual difference in The first two authors contributed equally to this
a research. We are thankful for comments from
supplier turnover, adopting straight 'Japanese'
Laurence Capron, Richard Caves, Frederic Dal
model with extensive technological cooperation
sace, Witold Henisz, Paul Ingram, Patrick More
may be difficult and potentially wasteful for vari
ous U.S. firms. However, we find that some U.S. ton, and J. Myles Shaver. The paper benefited
greatly from the insights of the reviewers and
firms are already in a strong position to lever
the Associate Editor. Any remaining errors are
age their technologies and relation-specific assets.
ours. This research was supported by a grant from
If supplier relationships start turning over faster
in Japan, meanwhile, U.S. practice might
the U.S. Air Force's Japan Industry and Man
yield
useful lessons (e.g., regarding technical exchanges agement of Technology Program and by a grant
from the Tenneco Fund Program at the Stern
in recent relationships). Thus, convergence toward
School of Business of New York University. The
an intermediate or third model could occur more
authors acknowledge the support of the University
quickly.
of Texas at Austin, Columbia University, and the
Regardless of the context, two major lessons
from our findings. stand
Mitsubishi Research Institute, where they respec
follow First, suppliers
to benefit from systematic tively conducted a portion of this research.
knowledge exchange
with buyers. Second, prior link duration conditions
the effectiveness of more complex, higher-level REFERENCES
technology transfer. Most importantly, we show
that higher-level technology transfer works best in Ahmadjian CL, Lincoln JR. 2001. Keiretsu, governance,
and learning: case studies in change from the Japanese
long-established buyer-supplier relationships.
automotive industry. Organization Science 12(6):
This confirms the importance of relation-specific
683-701.
assets as a source of competitive advantage (Mar Aiken LS, West SG. 1991. Multiple Regression: Testing
tin et a/., 1998; Dyer and Singh, 1998). Altogether, and Interpreting Interactions. Sage: Newbury Park,
this research is a step toward understanding how CA.
transfer and relational assets Armstrong JS, O verton T. 1977. Estimating nonresponse
knowledge jointly
bias in mail surveys. Journal of Marketing Research
add value between firms. Indeed, the collabora
we describe are not unique to 14(August): 396-402.
tive mechanisms Arora A. 1994.
A, Gambardella Evaluating technological
vertical interfirm relationships. They also stand information and utilizing it. Journal of Economic
to affect performance inside firms that integrate Behavior and Organization 24: 91-114.
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J, 24: 293-316 (2003)
Gaining from Vertical Partnerships 313
Arrow KJ. 1974. The Limits of Organization. W. W. Dyer JH, Cho DS, Chu W. 1998. Strategic supplier
Norton: New York. segmentation: the next 'best practice' in supply chain
Asanuma B. 1985. The organization of parts purchases in management. California Management Review 40(2):
the Japanese automotive industry. Japanese Economic 57-77.
Studies 13(4): 32-53. Dyer JH, Chu W. 2000. The determinants of trust in
Asanuma B. 1989. Manufacturer-supplier relationships supplier-automaker relationships in the U.S., Japan,
in Japan and the concept of relation-specific skill. and Korea. Journal of International Business Studies
Journal of the Japanese and International Economies 31(2): 259-285.
3: 1-30. Dyer JH, Nobeoka K. 2000. Creating and managing a
Axelrod R. 1984. The Evolution of Cooperation. Basic high performance knowledge-sharing network: the
Books: New York. Toyota case. Strategic Management Journal 21(3):
Bagozzi RP, Yi Y. 1993. Multitrait-multimethod matri 345-367.
ces in consumer research: and new devel JH, Ouchi WG. 1993. business
critique Dyer Japanese style
opments. Journal of Consumer Psychology 2(2): partnerships: giving companies a competitive edge.
143-170. Sloan Management Review 35(1): 51-63.
Bentler PM, Chou CP. 1987. Practical issues in struc Dyer JH, Singh H. 1998. The relational view: cooperative
tural equation modeling. Sociological Methods and strategy and sources of interorganizational competitive
Research 16: 78-117. advantage. Academy of Management Review 23(4):
Branstetter L. 2000. Vertical keiretsu and knowledge 660-679.
spillovers in Japanese manufacturing: an empirical Fichman M, Levinthal DA. 1991a. History dependence
assessment. Journal of the Japanese and International and professional relationships: ties that bind. Research
Economies 14(2): 73-104. in the Sociology of Organizations 8: 119-153.
Brockner JP, Siegel JP, Tyler DT, Martin C. 1997.When Fichman M, Levinthal DA. 1991b. Honeymoons and the
trust matters: the moderating effect of outcome liability of adolescence: a new on duration
perspective
favorability. Administrative Science Quarterly 42: dependence in social and organizational relationships.
558-583. Academy ofManagement Review 16(2): 442-468.
Choi TY, Hartley JL. 1996. An exploration of supplier Galbraith C. 1990. Transferring core
manufacturing
selection practices across the supply chain. Journal of in firms.
technologies high-technology California
Operations Management 14: 333-343. Management Review 32(Summer): 56-70.
Clark KB, Fujimoto T. 1991. Product Development Godkin L. 1988. Problems and practicalities of technol
Performance. Harvard Business School Press: Boston, ogy transfer: a survey of the literature. International
MA. Journal of Technology Management 3: 587-603.
Cohen J, Cohen P. 1983. Applied Multiple Regres Granovetter MS. 1985. Economic action and social
sion/Correlation Analysis for the Behavioral Sciences structure: the problem of embeddedness. American
(2nd edn). Erlbaum: New York. Journal of Sociology 91(3): 481-510.
Collis D. 1991. A resource-based analysis of global Granstrand O, Sjolander S. 1988. innovation
Managing
competition: the case of the in multi-technology Research 19:
bearings industry. corporations. Policy
Strategic Management Journal, Summer Special Issue 35-60.
12: 49-68. Hall ET, Hall MR. 1990. Understanding Cultural Differ
Craig CS, Douglas SP. 1999. International Marketing ences. Press: Yarmouth,
Intercultural ME.
Research (2nd edn). Wiley: New York. Heide JB, Miner A. 1992. The shadow of the future:
Cusumano MA, Takeishi A. 1991. Supplier relations effects of anticipated interaction and of
frequency
and management: a survey of Japanese, contact on
Japanese buyer-seller cooperation. Academy of
transplant, and U.S. auto plants. Strategic Management Management Journal 35(2): 265-291.
Journal 12(8): 563-588. Helfat CE. 1997. Know-how and asset complementarity
Delacroix J, Swaminathan A. 1991. Cosmetic, specula and dynamic accumulation: the case of
capability
tive, and adaptive organizational change in the wine R&D. Strategic Management Journal 18(5): 339-360.
a longitudinal
industry: study. Administrative Science Helper S. 1991a. How much has really changed
Quarterly 36: 631-661. between U.S. automakers and their suppliers? Sloan
Delios A, Henisz WJ. 2000. Japanese firms' investment Management Review 15(Summer): 15-28.
strategies in emerging economies. Academy of Helper S. 1991b. Strategy and irreversibility in supplier
Management Journal 43(3): 305-323. relations: the case of the U.S. automobile industry.
Dosi G. 1988. Sources, procedures, and microeconomic Business History Review 65(Winter): 781-824.
effects of innovation. Journal of Economic Literature Helper S, Sako M. 1995. Supplier relations in Japan
26: 1120-1171. and the United States: are they converging? Sloan
Dussauge P, Hart S, Ramanantsoa B. 1992. Strategic Management Review 36(Spring): 77-84.
Technology Management. Wiley: New York. KogutB, Zander U. 1992. Knowledge of the firm,
Dyer JH. 1996. Specialized supplier networks as a combinative and the of
capabilities, replication
source of competitive evidence from the Science 383-397.
advantage: technology. Organization 3(3):
automotive industry. Strategic Management Journal Kogut B, Zander U. 1993. Knowledge of the firm
17(4): 271-291. and the evolutionary theory of the multinational
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J, 24: 293-316 (2003)
314 M. Kotabe, X. Martin and H. Domoto
corporation. Journal of International Business Studies Nishiguchi T. 1994. Strategic Industrial Sourcing: The
24: 625-645. Japanese Advantage. Oxford University Press: New
Kotabe M. 1998. Efficiency vs. effectiveness orientation York.
of global sourcing strategy: a of U.S. Podsakoff PM, DW. 1986. in organi
comparison Organ Self-reports
and Japanese multinational companies. Academy of zational research: and Journal
problems prospects. of
Management Executive 12(November): 107-119. Management 12(4): 531-544.
Lamming R. 1993. Beyond Partnership: Strategies Porter ME. 1980. Competitive Strategy: Techniques for
for Innovation and Lean Supply. Prentice-Hall: Industries and Free Press:
Analyzing Competitors.
Englewood Cliffs, NJ. New York.
Levinthal DA, Fichman M. 1988. Dynamics of interor Porter ME. 1985. and
Competitive Advantage: Creating
ganizational attachments: auditor-client relationships. Free Press: New
Sustaining Superior Performance.
Administrative Science Quarterly 33: 345-369. York.
Levinthal DA, Myatt J. 1994. Co-evolution of capabili J. 1996. The case
Ramsay against purchasing partner
ties and industry: the evolution of mutual fund pro International Journal
ships. of Purchasing Manage
cessing. Strategic Management Journal, Winter Spe ment 4(32): 13-19.
cial Issue 15: 45-62. Robinson P, Martin X. 1997. The Japanese economy:
Liker JK, Kamath RR, Wasti SN, Nagamachi M. 1996. the economic Sloan
unpacking Japanese system.
Supplier involvement in automotive component
Management Review 38(3): 6-7.
design: are there really large U.S.-Japan differences?
Rosenberg N. 1982. Inside the Black Box: Technology and
Research Policy 25: 59-89.
Innovation. Cambridge University Press: New York.
MacDuffie JP, Helper S. 1997. Creating lean suppliers: Sakai K. 1990. The feudal world of manufac
Japanese
diffusing lean production through the supply chain. turing. Harvard Business Review 68(6): 38-49.
California Management Review 39(4): 118-134. Sako M, Helper S. 1998. Determinants of trust in supplier
Martin X, Mitchell W, Swaminathan A. 1994. Recreating
relations: evidence from the automotive industry in
and extending buyer-supplier links following inter
Japan and the United States. Journal of Economic
national expansion. In Advances in Strategic Man
Behavior and Organization 34: 387-417.
agement, Vol. 10B, Shrivastava P, Huff A, Dutton J
Sengupta K, Abdel-Hamid T. 1993. Alternative concep
(eds). JAI Press: Greenwich, CT; 47-72.
tions of feedback in dynamic decision environments:
Martin X, Mitchell W, Swaminathan A. 1995. Recreating
an experimental investigation. Management Science
and extending Japanese automobile buyer-supplier
links in North America.
39(4): 411-428.
Strategic Management Simonin B. 1997. The importance of collaborative know
Journal 16(8): 589-620.
how: an test of the learning
Martin Salomon R. 2002. transfer empirical organization.
X, Knowledge capacity
and when should firms Academy ofManagement Journal 40(5): 1150-1174.
cooperative strategies: leverage
In Steenkamp JE, van Trijp HCM. 1991. The use of LISREL
tacit knowledge through licensing and alliances? in validating constructs. International
and Contractor marketing
Cooperative Strategies Alliances, FJ,
Journal of Marketing Research 8: 283-299.
Lorange P (eds). Elsevier Science: New York;
Steensma HK, Lyles MA. 2000. Explaining IJV survival
729-748.
in a transitional economy through social exchange and
Martin X, Swaminathan A, Mitchell W. 1998. Organi
zational evolution in the envi knowledge-based perspectives. Strategic Management
interorganizational
ronment: incentives and constraints on international Journal 21(8): 831-851.
GJ, Becker GS. 1977. De gustibus non est
Administrative Science Stigler
expansion strategy. Quarterly
American Economic Review 67: 76-90.
43(3): 566-601. disputandum.
Szulanski G 1996. Exploring internal stickiness: imped
McMillan J. 1990. Managing suppliers: incentive systems
iments to the transfer of best practice within the firm.
in Japanese and United States industry. California
Journal, Winter Issue
Management Review 32(4): 38-55. Strategic Management Special
Mudambi S. 1998. The 'lose but adversarial' aries: management of supplier involvement in auto
R, Helper
model of relations in the U.S. auto mobile product development. Strategic Management
supplier industry.
Journal 19(8): 775-792. Journal 22(5): 403-433.
Strategic Management
Nanda A. 1996. and Teece DJ. 1977. Technology transfer by multinational
Resources, capabilities, competen
cies. In and firms: the resource cost of technological
Organizational Learning Competitive transferring
B, Edmondson A know-how. Economic Journal 87: 242-261.
Advantage, Moingeon (eds). Sage:
93-120. Tezuka H. 1997. Success as the source of failure?
London;
Nelson RR. 1991. Why do firms differ, and does Competition and cooperation in the Japanese
it matter? Journal, Winter economy. Sloan Management Review 38(2): 83-93.
Strategic Management
Special Issue 12: 61-74. Uzzi B. 1996. The sources and consequences of embed
Nelson RR, Winter SG. 1982. An Evolutionary Theory dedness for the economic performance of organi
Economic Change. Harvard University Press: zations: the network effect. American Sociological
of
Cambridge, MA. Review 61: 674-698.
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J, 24: 293-316 (2003)
Gaining from Vertical Partnerships 315
Uzzi B. 1997. Social structure and competition in Wernerfeit B. 1985. Brand loyalty and user skills.
interfirm networks: the paradox of embeddedness. Journal of Economic Behavior and Organization 6:
Administrative Science
Quarterly 42(1): 35-67. 381-385.
Uzzi B. 1999. Embeddedness in the making of financial White H. 1980. A heteroskedasticity-consistent covari
how social relations and networks benefit firms ance matrix estimator and a direct test for het
capital:
seeking financing. American Sociological Review 64: eroskedasticity. Econometrica 48: 817-838.
481-505. Williamson OE. \915. Markets and Hierarchies: Analysis
Venkatraman N, Ramanujam V. 1987. Measurements of and Antitrust Implications. Free Press: New York.
business economic performance:
an examination of Williamson OE. 1985. The Economic Institutions of
method convergence. Journal of Management 13: Capitalism. Free Press: New York.
109-122. Womack JP, Jones DT, Roos D. 1990. The Machine
von Hippel E. 1988. The Sources of Innovation. Oxford that Changed the World. Rawson Associates: New
All items in multi-item variables were measured on a 5-point Likert scale. The scale was anchored by
?1 and 5?'Strongly unless otherwise indicated.
'Strongly Disagree' Agree'
alpha
Criterion variable
development process3
{continued overleaf)
Copyright ? 2002 John Wiley & Sons, Ltd. Strat. Mgmt. J, 24: 293-316 (2003)
316 M. Kotabe, X. Martin and H. Domoto
alpha
Supplier's Technical A 4-item measure: How much influence does your firm have on 0.74
Influence your partner with respect to each of the following activities:
Product design specifications
?
(anchors: 1 'Our Firm Manufacturing process specifications
Has No Influence' and Procurement of materials and components
? Product control
5 'Our Firm Has quality
Complete Influence')
Crucial For 15 component classes, the respondents were asked to rate how n.a.*
Components
crucial (1?Least Crucial to 5?Most Crucial) it is for
automobile manufacturers' long-term competitiveness to
a
A potential issue with this item is that itmentions both a practice and an outcome. We thank an anonymous reviewer for bringing
this point to our attention. To ensure that this does not change the interpretation, we replicated the results after excluding this item.
The results were materially unchanged.
b Transfer variable. We believe that this is because the words
One might wonder why this item loads onto the Technology 'problems'
and 'often' tend to be associated with larger-scale, higher-stakes knowledge transfer. The results are essentially unchanged when
this ambiguous item is removed.
*
Indicates single-item measures to which Cronbach's alpha is not applicable.
2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 24: 293-316 (2003)
Copyright ?