You are on page 1of 16

The Global Research-and-

Development Network and Its


Effect on Innovation
Changsu Kim and Jong-Hun Park

ABSTRACT
The authors examine the impact of a firm’s innovation by analyzing its position in a global research-and-development
network and its orientation toward using scientific knowledge (science intensity). Drawing on organizational learning
and network theory research, they focus on the moderating role of a firm’s network position in the relationship between
the firm’s science intensity and the impact of its innovation. Data derived from the global pharmaceutical industry indi-
cate that a firm’s science intensity enhances the impact of its innovation by facilitating effective search. This relationship
is reinforced when the firm has network resources derived from an efficient network position. These results support the
idea that the impact of innovation is strengthened when internal research capability and external network resources are
combined. This study, which is conducted in a global industry network context, also enables the authors to examine the
role of international gatekeepers in bridging ties across U.S., Japanese, and European geographic boundaries.

Keywords: research-and-development alliances, science intensity, network resources, innovation impact

s business becomes increasingly global, corporate Sorescu, Chandy, and Prabhu 2003; Yalcinkaya, Calan-

A innovation strategies are also becoming more


global as companies attempt to lower research-
and-development (R&D) costs and gain international
tone, and Griffith 2007). In the study of innovation,
marketing research has emphasized the need for firms to
exploit existing resources while developing new ones
market access and insight (Jaruzelski and Dehoff 2008; (Atuahene-Gima 2005). This tradition offers a perspec-
Murray and Chao 2005). Therefore, effective inter- tive on innovation that recognizes the value of recom-
national learning and innovation have become critical bining existing and/or new technologies. By definition,
functions of multinational firms (Griffith and Harvey innovations serve as the basis of new products and serv-
2001; Singh 2008), and international marketing ices in the global marketplace and a firm’s global
researchers have drawn attention to innovation issues in competitiveness.
the global marketplace (Lee et al. 2008; Özsomer and
Gençtürk 2003). There is a rich tradition in marketing The extant literature emphasizes two sources of innova-
of studying diverse aspects of innovation and new prod- tion: internal development and external collaborations.
uct development (Atuahene-Gima and Murray 2007; State-of-the-art technologies are often developed from

Changsu Kim is Associate Professor of International Business Journal of International Marketing


(e-mail: cskim@sogang.ac.kr), and Jong-Hun Park is Professor ©2010, American Marketing Association
of Strategic Management (the corresponding author, e-mail: Vol. 18, No. 4, 2010, pp. 43–57
johnpark@sogang.ac.kr), Sogang University, Korea. ISSN 1069-0031X (print) 1547-7215 (electronic)

The Global Research-and-Development Network 43


tacit knowledge that is built internally through invest- complemented by the network approach we take herein,
ments or “learning by doing” (Teece 1992); however, which maps a firm’s position in a global network of
recent research has suggested that a firm’s level of inno- R&D alliances.
vation can and should be evaluated by considering inter-
nal research efforts jointly with external R&D collabo- We organize this article as follows: In the next section,
rations (Fabrizio 2009). More specifically, Fleming and we discuss theory and develop and propose hypotheses.
Sorenson (2004) suggest that a firm must clearly under- In the following sections, we address research design
stand the directions intended for internal research and and testing and provide the results. We conclude with
use external R&D to complement it. However, previous suggestions for scholarship and innovation practice in
studies (Fabrizio 2009) have not thoroughly tested the the international marketing field.
complementary role of internal and external R&D.

The purpose of this study is to examine the issue of THEORY AND HYPOTHESES
learning and innovation in the context of a global R&D Theoretical Model
alliance network to address a gap in innovation research
in the international marketing field. Two important We examine a firm’s innovation impact by drawing
questions to be addressed include the following: How from research on organizational learning and network
does a firm’s internal research enhance its innovation? theory. From a network perspective, we investigate
and How do internal R&D efforts and external R&D innovation within the boundary of the network in which
collaborations interact with each other to improve firm-level technological trajectories influence and are
innovation? influenced by other firms’ trajectories. To capture such
embeddedness in a technological network, it is impor-
In this study, we focus on the impact of a firm’s innova- tant to understand the extent to which innovations gen-
tion, paying particular attention to the firm’s position in erated by a firm are recognized and assimilated by other
a network of R&D alliances. We take a network firms. At a network level, innovations that serve as
approach in our study of innovation impact under the sources for many subsequent innovations by other firms
premise that the impact of innovation is largely a func- can be considered high-impact innovations (Rosenkopf
tion of the quality and diversity of inputs and that a and Nerkar 2001). In this study, the term “innovation
firm’s network position is associated with the quality impact” refers to the degree to which other members of
and diversity of information and resources the firm can a firm’s network have retained and built on that firm’s
acquire from the alliance network (Swaminathan and technology.
Moorman 2009).
In Figure 1, we hypothesize that a firm’s innovation
Despite the importance of the network approach in the impact increases with its science intensity—that is, its
study of learning and innovation, international market- level of efforts and inclination to use scientific knowl-
ing literature has paid little attention to the network of edge. We hypothesize that such an effect is enhanced by
R&D alliances that exists in the global marketplace network position variables such as network efficiency
(Atuahene-Gima and Murray 2007). For example, the (i.e., the extent to which the firm’s partners provide
bulk of past alliance research in the international mar- nonredundant information and resources). Our sample
keting field has taken the firm, or dyad of firms that consists of 32 U.S., 33 Japanese, and 24 European firms,
make up an alliance, as the unit of analysis, examining a multinational sample that enables us to examine the
each alliance partner’s characteristics and the type of role of international gatekeepers (Spencer 2003). We go
alliance to assess the formation, entry modes, and per- further to hypothesize that the effect of a firm’s science
formance of international collaborations such as inter- intensity on its innovation impact will be strengthened
national joint ventures (e.g., Calantone and Zhao when the firm plays the role of an international gate-
2000). However, studies in organizational theory, such keeper by absorbing knowledge from foreign firms and
as that by Powell, Koput, and Smith-Doerr (1996), have disseminating it to local firms.
added a network perspective by suggesting that a firm’s
formation of an alliance and the outcomes of that In Figure 1, the moderation model is specified according
alliance also depend on the position of the firm in the to our view that internally accumulated R&D capabili-
network of R&D alliances. Therefore, firm- and dyad- ties are the basic building blocks of innovation (Cohen
level alliance studies in international marketing can be and Levinthal 1989). We expect R&D alliances to

44 Journal of International Marketing


Figure 1. Theoretical Model

(+)
Science Intensity Innovation Impact
of a Firm of a Firm

(+)

R&D Alliance Network Position

• Firm network efficiency


• International gatekeeper

complement internal R&D capabilities (Teece 1992) some firms in the industry conduct more science-driven
because each alliance partner is a potential conduit for R&D than others (Cockburn, Henderson, and Stern
valuable information and resources. Thus, being a well- 2000).
connected and significant player in an R&D alliance
network can be a crucial strategic advantage in acquir- Our prediction for the direct effect of science intensity
ing external technologies and related capabilities (Pow- on innovation impact is based on the two lines of logic
ell, Koput, and Smith-Doerr 1996). Therefore, we that underlie science-driven R&D: those of exploration
hypothesize that the diverse technologies and capabili- and focus. First, science intensity reflects a firm’s ten-
ties acquired from alliance partners contribute to a dency to engage in an explorative distant search (Cock-
firm’s internal ability and effort in producing new high- burn, Henderson, and Stern 2000). Science-based R&D
impact technology. is uncertain and costly, but the payoffs can be high when
it is successful, which is a typical pattern for an explo-
Predicting Innovation Impact from Science rative search (Atuahene-Gima 2005; March 1991).
Intensity When a firm is more science driven in its R&D, it is
more likely to appreciate the value of exploration and,
Innovation builds on knowledge gleaned from scientific consequently, be more willing to engage in exploration
studies (Gittleman and Kogut 2003). In particular, the in R&D. As such, the potential for innovative techno-
pharmaceutical industry, which is our research setting, is logical developments in R&D will be greatest when the
dependent on a complex and always-evolving scientific firm’s tendency to search for explorative technological
research base (Pisano 2006; Sorescu, Chandy, and developments is high.
Prabhu 2003), largely because of an increasing reliance
on biotechnology for its R&D activities. As a result, the Second, a firm’s science search tends to lead to more
pharmaceutical industry has become one of the most focused R&D. A firm’s focus may help it employ the
science-intensive sectors in the economy (Pisano 2006). results of scientific developments more quickly and
Although the pharmaceutical industry as a whole is sci- effectively. The capacity of a firm to take advantage of
ence driven, we still expect to find firm-level variations the results of scientific developments better and faster
in the level of reliance on science for innovation and that than competitors is a major source of competitive

The Global Research-and-Development Network 45


advantage (Cockburn, Henderson, and Stern 2000). the novelty and diversity of information and resources
Fleming and Sorenson (2004) show that science-guided the firm can acquire from its network (Swaminathan
search is closely associated with high-impact innova- and Moorman 2009).
tions, especially for the highly coupled, complex innova-
tions in pharmaceuticals, in contrast to innovations that Partner redundancy inevitably increases with growth in
work with modular components (e.g., in computers). the number of a firm’s alliances; therefore, increasing
Fleming and Sorenson articulate that scientific knowl- the number of alliances without considering partner
edge enables researchers to focus their searches in the diversity can create inefficient configurations that yield
most likely areas of opportunity, thus ruling out costly less diverse information and resources at greater cost
and time-consuming trial-and-error local searches, and than a smaller, nonredundant set would yield. A highly
motivates researchers to continue distant searches redundant configuration may even prevent a firm from
despite early failures. In summary, scientific knowledge obtaining new or novel information by limiting the
provides an understanding of the underlying fundamen- number of ties a firm can afford to make. Therefore,
tal properties involved in generating the observed out- more efficient alliance configurations that provide
come (i.e., theories about why it occurs), and this under- access to more diverse information and resources with
standing facilitates more focused and effective searches minimum costs of redundancy will prove most benefi-
(Fabrizio 2009). Therefore, we hypothesize that a firm’s cial for a firm’s performance (Baum, Calabrese, and Sil-
science intensity improves its chances of producing high- verman 2000; Musteen, Francis, and Datta 2010). This
impact innovations: argument is based on the concept of network resources,
in which alliances are knowledge-accessing relationships
H1: The greater a firm’s science intensity, the and the advantages a firm derives from a portfolio of
greater is the firm’s innovation impact. alliances depend on the resource profiles of its alliance
partners (Swaminathan and Moorman 2009). However,
The Moderating Role of Network Position managing a large portfolio of alliances can be challeng-
ing, and alliances also bring with them the risk of losing
In this study, we are particularly interested in examining proprietary technologies to the partner. Thus, it is in a
how a firm’s network position and its structural rela- firm’s best interest to be affiliated with a small number
tionships with others affect its opportunities to access of nonredundant firms rather than a large number of
distinctive external information and resources. The net- redundant firms.
work concept of range, which describes the extent to
which an actor contacts a diversity of other actors, can There is reason to expect that a firm’s network efficiency
be measured as volume or quality of contacts (Burt and its level of science intensity are complementary. A
1983). Volume measures the total number of contacts, firm’s ability to generate high-impact innovations is
and quality measures the extent to which the contacts largely a function of the quality and diversity of inputs.
provide nonredundant information. We focus on quality A science-guided search deepens the firm’s knowledge
as a key network attribute associated with the impact of base by increasing its understanding of the cutting-edge
innovation and examine the concept of network effi- scientific developments that may provide the foundation
ciency. Furthermore, we examine the network concept for the focused research needed for innovation impact.
of range with the three regional boundary sets: the With efficient exposure to a network of R&D alliances
United States, Japan, and Europe. In the case of the (i.e., network efficiency), a firm will gain access to
global R&D network, some firms take on the role of diverse and novel external knowledge (Baum, Cala-
international gatekeeper by bridging ties across regional brese, and Silverman 2000) that facilitates the link
geographic boundaries. between science intensity and innovation. As internal
scientific knowledge accumulates, it tends to generate
Firm Network Efficiency. The concept of network effi- path dependence, and diverse and novel external knowl-
ciency explicitly focuses on the relationship patterns of edge helps break such path dependence and leads to new
a firm’s partners rather than the firm’s direct ties. By paths of discovery. Moreover, internally accumulated
definition, firm network efficiency arises when a firm’s scientific knowledge can help a firm absorb and use new
partners are separated from one another and implies external knowledge (Cohen and Levinthal 1989). In this
access to mutually unconnected partners and, conse- way, science intensity is a necessary input, and efficient
quently, an efficient, information-rich network. We network position generates the positive interaction. In
focus on network efficiency because it is associated with other words, the effect of a firm’s science intensity in

46 Journal of International Marketing


bringing in high-impact innovations can be stronger impact. Participation in global R&D activities is a
when it is combined with a more varied and distinctive unique source of innovation for international firms
set of resources available to the firm because of its net- (Jaruzelski and Dehoff 2008), and success in a global
work efficiency. competition for R&D activities rests on the ability to
gain international market access and insight and to fore-
H2a: The positive relationship between a firm’s see the scientific and technological developments at a
science intensity and its innovation impact is global level (Singh 2008). The international boundary-
strengthened by the firm’s R&D network spanning advantage available to international gatekeep-
efficiency. ers helps them identify global developments, providing
additional opportunities for innovation in the global
International Gatekeepers. One of the ideas underlying marketplace. Therefore, we expect that the benefits
the network concept of range—the extent to which an resulting from science intensity are greater when the
actor contacts a diversity of other actors (Burt 1983)— firm also engages in international boundary-spanning
is that rational actors attempt to maximize benefits out activities.
of network configurations by modifying the way they
are embedded. We continue to examine the issue of net- H2b: The positive relationship between a firm’s
work range but focus on a firm’s brokerage roles in con- science intensity and innovation impact is
necting international groups. Our sample of 89 firms is strengthened if the firm plays an inter-
based in three regions (the United States, Japan, and national gatekeeper role.
Europe), and we employ with these three regions the
concept of gatekeepers, which refers to Tushman’s METHODS
(1977) classic work, in which gatekeepers represent the Sample and Data Collection
key people in mediating the flow of knowledge from one
group to another. We chose the pharmaceutical industry as the empirical
research context in which to test our hypotheses for
Applied to our study of a global R&D alliance network, several reasons. First, it is one of the most dynamic
the term international gatekeepers refers to the few key and globalized industries in the world. The nature of
firms that bridge geographic boundaries by streamlining the industry requires pharmaceutical firms to engage in
and organizing knowledge flows across regions within a wide range of R&D activities, such as DNA and
the global network. Thus, international gatekeepers molecular manipulation, as well as traditional empirical
appropriate information and resources from foreign approaches to drug discovery. Pharmaceutical firms face
firms, convey them to local firms (Spencer 2003), and strong pressures to develop medicines for a global mar-
have access to unique sources of knowledge because of ket and exploit economies of scale and scope at the
their international boundary-spanning roles. Reaching global level (Pisano 2006). However, the growing inter-
across borders through R&D alliances provides a way dependence of previously discrete technologies makes it
to gain access to both technological capabilities embed- increasingly difficult for any single firm to stand alone
ded in partner firms and capabilities embedded in the in this industry (Fleming and Sorenson 2004), and thus
partner’s national environment (Jaruzelski and Dehoff R&D alliances play a critical role in this industry, in
2008). National environments are idiosyncratic with which alliance-based teams race toward the creation
respect to knowledge bases (Bartholomew 1997), which and commercialization of similar end products and a
provides boundary-spanning firms with broad learning winner-takes-all situation often exists for new develop-
opportunities that may not be available to their domes- ments (Powell, Koput, and Smith-Doerr 1996). Second,
tic and international rivals (Singh 2008). Moreover, the pharmaceutical industry is one of the most science-
international gatekeepers are in a strong position to intensive industries in the world economic picture
apply, filter, and reframe knowledge as they pass it from (Pisano 2006). Biotechnology has enabled pharmaceuti-
foreign to local firms (Spencer 2003). cal firms to move from a random approach to conduct-
ing a rational design approach, in which compounds are
Because of the increasing diversity of idiosyncratic developed from scientific theories regarding the origins
national (or regional) knowledge bases, international and evolution of diseases. The latter approach means
gatekeepers that serve as a conduit of knowledge across that pharmaceutical firms must rely on science more
borders benefit more than other firms from their than ever before (Cockburn, Henderson, and Stern
science-based focused searches in terms of innovation 2000).

The Global Research-and-Development Network 47


We limited our potential sample to firms that belonged ceutical industry because many cutting-edge and impor-
to Standard Industrial Classification 28 during the tant findings in the industry are reported in scientific
period of our research (1988–1994). Furthermore, we journals (Cockburn, Henderson, and Stern 2000).
confined our sample to companies whose patent data
were available from CHI Research, Inc., because we Moderating Variables. Following the classification of
measured a firm’s science intensity and innovation the Securities Data Company database, we defined
impact using the CHI patent database. This criterion R&D alliances as any form of alliance that involves col-
resulted in a total of 315 potential firms for our sample. laboration in R&D activities. Such alliances include
Then, we combined the CHI data with the R&D technology licensing agreements and bilateral tech-
alliance data obtained from the Securities Data Com- nology transfers, such as joint R&D contracts and R&D
pany database and, because we needed complete panel joint ventures. We constructed a binary adjacency
data on both R&D alliances and patents, restricted the matrix in which we recorded the presence of R&D
sample to R&D alliance pairs that belonged to the CHI alliances between firms i and j longitudinally from the
database. This process left us with a final sample of 89 1988 matrix to the 1994 matrix. The matrix was input
firms (32 U.S., 33 Japanese, and 24 European firms). into UCINET V software, which computes network
Accordingly, our sample consists of 623 firm-year obser- measures. With respect to previous research (Burt
vations (89 firms over the seven years of our research 1983), we constructed a measure of network efficiency
period). that identifies firms as efficient when they maintain
nonredundant ties that provide access to diverse and dis-
Measures tinctive information and resources. More specifically,
we measured the network efficiency of the focal firm i as
Dependent Variable. Following previous studies (Flem- follows:
ing and Sorenson 2004; Gittelman and Kogut 2003), we Network efficiencyi = [ΣjZij – Σj ≠ iΣk ≠ iZjk/
defined the innovation impact of a firm’s patent as the
number of citations the patent received from the other ΣjZij]/ΣjZij,
firms in the five years following the year the patent was
granted. A patent application is required to list the prior where j represents the focal firm i’s partners, k is the
work it builds on, and previous research has demon- common partners of firms i and j, and Zij and Zjk are the
strated that the number of such citations a patent number of ties between firms i and j and between firms j
receives correlates highly with the economic value of the and k, respectively. In other words, we measured the
patent (Jaffe, Trajtenberg, and Henderson 1993). Thus, focal firm’s network efficiency by subtracting the number
we measured the innovation impact of firm i in year t by of R&D alliances the focal firm has from the average
the average number of citations (i.e., citations per number of R&D alliances each partner of the focal firm
patent) cited by the patents of the other firms in our has and then normalizing the difference by the number
sample over the subsequent five years (i.e., year t + 1 to of the focal firm’s R&D alliances. Thus, this measure
year t + 5). For example, Abbott Laboratories issued reveals the proportion of the focal firm’s nonoverlapping
130 patents in 1994, and the 88 other firms in the sam- R&D alliances that provide access to diverse and distinc-
ple cited those 130 patents 160 times during the next tive information and resources.
five years (i.e., 1995–1999); thus, we measured the
innovation impact of Abbott in 1994 as 1.23 (i.e., International gatekeepers bridge the gaps across the
160/130). three regions—the United States, Japan, and Europe—
absorbing information and resources from extraregional
Independent Variable. We measured the science inten- firms and conveying them to intraregional firms. A
sity of firm i in year t using the average number of sci- firm’s score in the international gatekeeper variable
ence references on the front pages of the firm’s patents reflects the number of times it lies on a cross-regional
in that year. This variable measures the closeness of the geodesic, simultaneously linking with an extraregional
firm’s technologies to scientific knowledge. Firms that and an intra-regional firm. Our geographic regional
want to integrate scientific findings into their inventions demarcation is based on the firm’s home base.
are more likely to cite science references in their prior
work (Gittelman and Kogut 2003). Therefore, the level Control Variables. Because a variety of factors can influ-
of science intensity indicates the degree to which the ence a firm’s innovation impact, we included firm-,
firm builds its technology using advances in science. region-, and year-level control variables. For the firm-
This measure is particularly important in the pharma- level control variables, we controlled for the number of

48 Journal of International Marketing


patents, R&D expenditure, the number of R&D effects across the entire sample (Greene 2003). How-
alliances, and the type of firm. To control for the effects ever, care must be taken when conducting statistical
of a firm’s innovation rate on its innovation impact, we tests on panel data because of the risk of cross-sectional
measured the number of patents by counting the total heteroskedasticity and within-unit serial correlation;
number of patents issued to a firm in a given year. We when these are present, the ordinary least squares (OLS)
also controlled for the effect of a firm’s R&D expendi- assumptions of constant variance and uncorrelated
ture on its innovation impact. We collected R&D expen- error terms are violated, rendering OLS inappropriate.
diture from the COMPUSTAT database (for U.S. firms) To correct for heteroskedasticity and autocorrelated
and Global Vantage database (for non-U.S. firms). We error terms, we used the generalized least squares
also controlled for the number of R&D alliances. When procedure.
firms have R&D alliances with many different firms,
they are likely to occupy a central position in the R&D Usually, fixed-effects models are preferred in panel data
alliance network. We also included a firm-type dummy analyses (Greene 2003). However, we were precluded
to determine whether the level of innovation impact sig- from using a fixed-effects approach because our sample
nificantly differs across different firm types. Our sample consists of U.S., Japanese, and European firms, and
consisted of two subsamples based on a firm’s primary there are significant differences in innovation impact
patenting activity: dedicated pharmaceutical firms, such among the three groups of firms. If we adopted a fixed-
as Abbott, GlaxoSmithKline, and Yamanouchi, and effect model, we could not include the regional dummy
diversified pharmaceutical firms, such as DuPont, BASF, variables in the regression equation. This is a common
and Sumitomo. We coded the firm-type dummy as 1 for problem when group-level heterogeneity exists within a
the dedicated pharmaceutical firms and 0 for the diver- sample (Greene 2003). When the fixed-effects approach
sified pharmaceutical firms. is ruled out, a random-effects approach, in which the
fixed effects are uncorrelated with the other independ-
Because our data were derived from three regions, we ent variables, can be used. A Hausman test revealed no
controlled for regional effects. For each year, we meas- significant correlations between our independent
ured regional network density as the number of R&D variables and the firm-level fixed effects, so we used
ties in each region divided by the total number of possi- random-effects models to test our hypotheses.
ble R&D ties in that region; this measure is a reflection
of the abundance of R&D alliances in each region. Fur- Another important issue is that of causality. To address
thermore, to control for differences in innovation this issue, we incorporated temporal precedence by
impact across years, we included fixed-year effects. We using a lag structure to ensure that all measures of inde-
included six such dummy variables with the year 1988 pendent variables preceded the dependent variable in
as a default. Finally, we introduced Japanese and Euro- time. As stated previously, we measured science intensity
pean dummies, with the United States as a default. and network attributes in year t and innovation impact
Because our sample consists of U.S. and non-U.S. firms, over the period from year t + 1 to year t + 5.
using U.S. patent data could raise possible biases favor-
able to U.S. firms. However, U.S. patent data are widely
used in studies of the global pharmaceutical industry RESULTS
(Penner-Hahn and Shaver 2005). Major pharmaceutical
firms tend to patent extensively under the U.S. patent Table 1 shows the descriptive statistics and the correla-
system because the United States is a major market that tion matrix for all the variables included in this study.
accounts for half of global drug sales and U.S. intellec- Although the means and standard deviations reported
tual property rights are considered critical for any firm in Table 1 are for unstandardized measures, we used
in the industry to do business globally (Ahuja and Katila the standardized values of all variables to test the
2004). hypotheses. Standardization reduces multicollinearity
and facilitates the interpretation of interaction terms
Analysis (Aiken and West 1991). To check for multicollinearity,
we ran OLS regressions to generate variance inflation
Our data set is a panel (pooled cross-sectional time factors for all the variables and found that all the
series). Panel data and methods can control for unob- variance inflation factor values were less than 2,
served heterogeneity and improve statistical estimates substantially lower than the recommended cutoff of 10.
by enlarging the sample size and capturing both the Therefore, we concluded that multicollinearity is not
average effects across individual units and the dynamic likely to be an issue.

The Global Research-and-Development Network 49


Table 1. Descriptive Statistics and Correlations

Variable M SD 1 2 3 4 5 6 7 8 9 10 11

1. Innovation impact .85 .61

2. Science intensity 2.42 3.78 .34**

3. Firm network
efficiency .39 .48 .14** .14**

4. International
gatekeeper .14 .68 .04 .08* .22**

5. Number of patents 63.94 88.75 –.05 –.14** .14** .17**

6. R&D expenditure 461.15 429.13 .00 –.06 .14** .07 .77**

7. Number of R&D
alliances .52 .95 .08* .13** .64** .64** .25** .19**

8. U.S. firm dummy .36 .48 .23** .31** .08* .05 .09* –.02 .06

9. Japanese firm dummy .37 .48 –.26** –.28** –.13** –.10** –.29** –.37** –.14** –.57**

10. European firm


dummy .27 .44 .04 –.03 .05 .06 .21** .43** .08* –.45** –.46**

11. Regional network


density .02 .01 .13** .22** .34** .25** .15** .14** .40** .21** –.39** .19**

12. Firm-type dummy .59 .49 .23** .34** .12** –.04 –.31** –.05 .02 .28** –.31** .04 .12**

*p < .05.
**p < .001.

The data in Table 1 demonstrate that science intensity our sample had a higher average level of R&D expendi-
and network efficiency are positively correlated with ture than their counterparts, which may have led Euro-
innovation impact. These data indicate that both inter- pean firms to outperform U.S. and Japanese firms in
nal research efforts and external R&D collaborations terms of the number of patents acquired during our
are associated with high-impact innovation. The U.S. research period. Table 2 also shows that, in general, U.S.
and European firm dummies are positively correlated firms had the highest level of innovation impact, though
with innovation impact, whereas the Japanese firm they produced a smaller number of patents than Euro-
dummy is negatively correlated with innovation impact. pean firms. This result may be due to the difference in
However, the patterns and trends across the three science intensity between U.S. and European firms. We
regions (see Figure 2) corroborate that there is a positive broke the firm types into dedicated pharmaceutical
association between science intensity and innovation firms and diversified firms and found that both U.S. and
impact. European groups were dominated by dedicated pharma-
ceutical firms, whereas the Japanese group was domi-
Table 2 presents the descriptive statistics of the U.S., nated by diversified firms. Firm-type data indicate that
European, and Japanese firms. On average, U.S. firms the difference in innovation impact across the three
employed the highest level (4.03) of science intensity, groups is partially affected by the different business
followed by European firms (2.19) and Japanese firms portfolio profiles of these groups.
(1.02). Network position variables for U.S. and Euro-
pean firms revealed that these firms used external R&D Table 3 shows the results for the hypotheses tests.
collaboration more efficiently than Japanese firms. The Model 1, a baseline model consisting only of control
data in Table 2 also reveal that the European firms in variables, indicates that the Japanese firm dummy

50 Journal of International Marketing


Figure 2. Innovation Impact and Science Intensity of U.S., European, and Japanese Firms

Innovation Impact Science Intensity

1.4 6

1.2
8

1
4

.8
3

.6

2
.4

1
.2

0 0
88 89 90 91 92 93 94
Year
Innovation impact of U.S. firms Innovation impact of European firms
Innovation impact of Japanese firms Science intensity of U.S. firms
Science intensity of European firms Science intensity of Japanese firms

(b = –.517, p < .001) and the firm-type dummy (b = ciency is positive and significant (b = .156, p < .001),
.212, p < .05) have significant effects on innovation and Model 6 shows that the interaction term between
impact. Model 2 shows a significant, positive associa- science intensity and international gatekeeper also is
tion between science intensity and innovation impact positive and significant (b = .064, p < .05). We included
(b = .252, p < .001). The coefficient of science intensity the two interaction terms together in Model 8 and
remains positive and significant regardless of the model found that only the interaction term between science
specification. These results provide strong support for intensity and firm network efficiency remained positive
H1. and significant (b = .149, p < .001). Taken together,
these results support H2a but only weakly support H2b.
Before testing H2, we included network position
variables in the model to control for the main effects of Figure 3 graphically shows the interaction, for which we
these variables. Though not our main focus of interest, relied on Jaccard, Turrisi, and Wan’s (1990) procedure.
the results are presented in Models 3 and 5. Firm net- First, we divided the sample into two groups based on
work efficiency has a significant and positive effect on the median value of firm network efficiency. We further
innovation impact, whereas the effect of international divided each of the two groups (i.e., lower and higher
gatekeeper is insignificant. network efficiency groups) into lower and higher science
intensity groups in terms of the median value of science
To examine H2, we also included the interaction term intensity. Second, we calculated the mean values for
between science intensity and each of the two network each of the four resulting groups and used the results
position variables. Model 4 indicates that the inter- from Model 8 (Table 3) to develop plots for the moder-
action term between science intensity and network effi- ating effect of firm network efficiency. As Figure 3

The Global Research-and-Development Network 51


Table 2. Sample Firm Characteristics

Network Positions
Firm Firm Number Science Firm Network International R&D Number of Innovation
Nationality Type of Firms Intensity Efficiency Gatekeeper Expenditurec Patents Impact

U.S. firms Dedicated 25 (78%) 4.786a .483 .158 400.4 53.5 1.083
pharmaceutical (5.866)b (.496) (.503) (323.1) (47.1) (.747)

Diversified 7 (22%) 1.355 .319 .312 606.6 154.9 .922


pharmaceutical (1.082) (.457) (1.215) (440.8) (131.1) (.618)

U.S. Total 32 (100%) 4.036 .447 .191 445.5 75.7 1.048


(5.396) (.492) (.721) (361.3) (84.8) (.722)

European Dedicated 15 (63%) 3.030 .430 .133 636.7 38.4 1.087


firms pharmaceutical (2.472) (.492) (.605) (385.8) (29.8) (.643)

Diversified 9 (37%) .800 .438 .349 985.6 191.1 .594


pharmaceutical (.511) (.494) (1.404) (731.9) (161.3) (.355)

European 24 (100%) 2.194 .433 .214 767.6 95.6 .902


Total (2.253) (1.227) (.985) (565.9) (125.3) (.601)

Japanese Dedicated 13 (39%) 1.549 .362 .022 308.7 18.8 .652


firms pharmaceutical (2.226) (.483) (.147) (141.2) (16.5) (.444)

Diversified 20 (61%) .679 .266 .064 217.4 36.4 .640


pharmaceutical (.647) (.440) (.273) (167.1) (33.2) (.382)

Japanese 33 (100%) 1.022 .304 .047 253.4 29.4 .645


Total (1.541) (.459) (.232) (163.2) (29.1) (.406)

aNumbers are the mean value of the variable.


bNumbers in parentheses are the standard deviation of the variable.
cNumbers are expressed in millions of U.S. dollars.

shows, the positive relationship between science inten- academics and practitioners who work to understand
sity and innovation impact is more pronounced in the durable and sustained superior performance (Sorescu,
higher network efficiency group than in the lower net- Chandy, and Prabhu 2003).
work efficiency group. This pattern is consistent with
our argument in H2a. Therefore, the interaction graph In this study, we observed that a firm’s science intensity
presented in Figure 3 provides additional support for has a positive direct effect on the impact of its innova-
H2a. tion. This finding is consistent with Fleming and Soren-
son’s (2004) observation that utilization of scientific
knowledge in patented inventions generates more inno-
DISCUSSION AND CONCLUSION vation impact by facilitating effective search. Further-
more, our data suggest the critical role of science inten-
High-impact innovations represent rare, valuable, and sity in integrating scientific research (R) with
potentially inimitable sources of competitive advantage technological development (D), though previous studies
(Rosenkopf and Nerkar 2001). Understanding how have suggested that the social structures of the science
firms create breakthrough innovations and sustain their world and the technology world are distinct (Murray
preeminence in an industry is of fundamental concern to 2002). Science is characterized by publication and sup-

52 Journal of International Marketing


Table 3. Results of Generalized Least Squares Regression Analysis for Innovation Impact

Independent Variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8

Science intensity .252*** .250*** .201*** .252*** .221*** .249*** .197***


(.041) (.041) (.042) (.041) (.044) (.041) (.043)

Firm network efficiency .116* .110* .123* .115*


(.048) (.047) (.051) (.050)

International gatekeeper –.014 –.024 .024 .014


(.048) (.048) (.051) (.050)

Science intensity × firm network efficiency .156*** .149***


(.039) (.043)

Science intensity × international gatekeeper .064* .010


(.030) (.033)

Number of patents –.136† –.075 –.075 –.070 –.073 –.076 –.077 –.072
(.072) (.071) (.071) (.070) (.071) (.071) (.071) (.070)

R&D expenditure .030 .018 .013 .025 .015 .018 .017 .027
(.070) (.068) (.068) (.067) (.068) (.068) (.068) (.067)

Number of R&D alliances .058 .031 –.037 –.043 .041 .039 –.057 –.056
(.042) (.041) (.050) (.049) (.052) (.052) (.066) (.064)

Japanese firm dummy –.517*** –.337** –.335** –.353** –.337** –.345** –.335** –.358**
(.130) (.130) (.130) (.128) (.130) (.130) (.129) (.128)

European firm dummy –.206† –.099 –.092 –.115 –.098 –.097 –.094 –.114
(.106) (.105) (.104) (.103) (.105) (.104) (.104) (.103)

Firm–type dummy .212* .126 .102 .126 .124 .135 .103 .127
(.096) (.095) (.095) (.094) (.095) (.094) (.095) (.094)

Regional network density .124 .118 .116 .107 .119 .127 .113 .106
(.106) (.103) (.103) (.101) (.103) (.103) (.103) (.102)

Year 1989 dummy –.254 –.304† –.322* –.304† –.306† –.319* –.319* –.305†
(.162) (.158) (.157) (.155) (.158) (.157) (.157) (.156)

Year 1990 dummy –.177 –.218 –.239 –.199 –.222 –.220 –.234 –.197
(.166) (.161) (.161) (.159) (.162) (.161) (.161) (.159)

Year 1991 dummy –.318 –.366† –.410* –.349† –.371† –.369† –.405* –.347†
(.196) (.191) (.191) (.189) (.193) (.191) (.191) (.189)

Year 1992 dummy –.251 –.339** –.349* –.337* –.340* –.335* –.348* –.336*
(.154) (.150) (.149) (.147) (.150) (.149) (.149) (.147)

Year 1993 dummy –.109 –.208 –.218 –.168 –.215 –.210 –.207 –.162
(.226) (.220) (.219) (.217) (.221) (.221) (.220) (.218)

Year 1994 dummy –.451 –.534† –.575† –.504 –.539† –.531† –.567† –.501
(.330) (.321) (.320) (.317) (.321) (.320) (.320) (.317)

The Global Research-and-Development Network 53


Table 3. Continued

Independent Variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8

Constant .343** .358** .390* .332* .362* .351* .385* .329*


(.165) (.160) (.160) (.159) (.161) (.160) (.160) (.159)

σμ .563 .513 .518 .499 .515 .506 .520 .501

σε .755 .755 .750 .744 .756 .755 .751 .745

ρ .349 .315 .323 .310 .317 .310 .324 .311

Wald chi-square 88.54*** 130.71*** 137.65*** 157.17*** 130.81*** 135.98*** 137.92*** 157.42***

†p < .10.

*p < .05.
**p < .01.
***p < .001.
Notes: This table reports standardized regression coefficients, with standard errors in parentheses.

ported by a priority-based reward system, which is in point to the conflicting logic behind innovation rate and
contrast to the world of technology, in which ideas are innovation impact. There is an inherent tension between
produced for economic ends and encoded in patents exploitation and exploration in innovation processes
to facilitate appropriability. Given the two distinctive (March 1991): Exploitation improves innovation effi-
communities of science and technology, Gittelman ciency, whereas exploration is required to produce new
and Kogut (2003) emphasize the bridging role of technologies of high quality and impact (Atuahene-
inventors in a firm as a critical function in the firm’s Gima 2005). Thus, firms must undergo the paradoxical
ability to materialize scientific research into valuable strategic task of balancing between exploitative and
technologies. explorative innovation strategies (He and Wong 2004).

This study measured science intensity by the extent to Our finding that science intensity interacts with R&D
which a firm cites scientific findings in its patents. By alliance network variables supports the complementary
explicitly tying scientific publishing and patenting roles of internal and external R&D activities in innova-
together, this measure incorporates the degree to which tion. In particular, the data support our network moder-
a firm encourages inventors who are directly involved in ation argument that the effect of science intensity on
the firm’s technology development to participate in the innovation impact is stronger when a firm’s science
scientific community. We conclude that in circumstances intensity is combined with its network resources (i.e.,
in which scientific research is tightly linked to tech- the quality and diversity of information and resources
nology developments, how firms access and practice sci- derived from the firm’s position in the network of R&D
ence plays a critical role in the production of valuable alliances). As Table 3 shows, the direct effect of network
technologies. variables is, by and large, weak or nonexistent; however,
we estimated all the interaction terms as positive and
The data in Table 2 indicate that U.S. firms score rela- significant. Figure 3 also shows that the effect of science
tively high with regard to science intensity and innova- intensity on innovation impact is more pronounced in
tion impact, whereas European firms score high in both the high than the low network efficiency group, provid-
R&D expenditure and number of patents. Therefore, it ing support for our argument on firm heterogeneity in
seems that R&D expenditure is associated with the network resources. Not all firms can enhance their inno-
number of patents produced, though science intensity vation impact by configuring their portfolios of alliances
itself is associated with innovation impact. These results into efficient networks.

54 Journal of International Marketing


Figure 3. Moderating Effect of Firm Network Efficiency

Innovation Impact
2.339
Firm network efficiency: high

1.276
Firm network efficiency: low

.438

.187

Science Industry
Low High

When the two network variables (firm network effi- licensing, joint R&D contracts, and joint R&D
ciency and international gatekeeper) enter an equation ventures. However, there is a substantial difference
simultaneously, only the significance of the efficiency between R&D alliances designed mainly to effect an
variable survives, and that representing the international exchange of existing technologies and those designed to
gatekeeper variable disappears. These findings suggest generate new technologies. In addition, R&D alliances
that leveraging knowledge globally, rather than region- provide access to external complementary technologies,
ally, is more important in producing high-impact inno- facilitating the exploitation of existing technologies
vation because the efficiency variable measures the con- (Teece 1992), and offer a way to explore new technolo-
cept of network range in the global industry context, gies jointly with partners (Powell et al. 1996).
whereas the international gatekeeper variable measures Because the exploitative licensing mode is different
the same concept in the domestic-foreign context, with from explorative joint R&D modes, for an
reference to the United States, Japan, and Europe as ex post check, we eliminated licensing from our sample
regional groups. This finding agrees with extant inter- and reran the analysis but found no significant devia-
national marketing research on the globalization of tions from our results. However, distinguishing a net-
R&D activities (Murray and Chao 2005; Özsomer and work of explorative R&D alliances from that of
Gençtürk 2003). In an era of globalization, configuring exploitative ones remains an area for further research,
and managing a portfolio of alliances worldwide has perhaps with an enriched database on R&D collabora-
become more salient and important. tions. Second, given that cross-country comparisons in
patent indicators are useful areas for study, further
Although this study directly examines the impact of research should be conducted with multicountry patent
innovation in a global industry network context, it has data. Third, although we believe our findings can be
a few limitations. First, notwithstanding the aforemen- generally applicable to other science-intensive indus-
tioned merits of network analysis, with its primary tries in which integrating scientific research with tech-
focus on the network as a whole, network analysis nology development is important, we admit that such a
does not differentiate between the types and modes claim is actually an empirical question. Therefore, we
of alliances within the domain of a R&D network. also suggest that our findings be validated with further
Broadly defined, R&D alliances include technology research in other industrial settings.

The Global Research-and-Development Network 55


REFERENCES Griffith, David A. and Michael G. Harvey (2001), “A Resource
Perspective of Global Dynamic Capabilities,” Journal of Inter-
Ahuja, Gautam and Riitta Katila (2004), “Where Do Resources national Business Studies, 32 (3), 597–606.
Come From? The Role of Idiosyncratic Situations,” Strategic
Management Journal, 25 (8–9), 887–907. He, Zi-Lin and Poh-Kam Wong (2004), “Exploration vs.
Exploitation: An Empirical Test of the Ambidexterity Hypo-
Aiken, Leona S. and Stephen G. West (1991), Multiple Regres- thesis,” Organization Science, 15 (4), 481–94.
sion: Testing and Interpreting Interactions. Newbury Park,
CA: Sage Publications. Jaccard, James, Robert Turrisi, and Choi Wan (1990), Inter-
action Effects in Multiple Regression. London: Sage
Atuahene-Gima, Kwaku (2005), “Resolving the Publications.
Capability–Rigidity Paradox in New Product Development,”
Journal of Marketing, 69 (October), 61–83. Jaffe, Adam B., Manuel Trajtenberg, and Rebecca Henderson
(1993), “Geographic Localization of Knowledge Spillovers as
——— and Janet Y. Murray (2007), “Exploratory and Evidenced by Patent Citations,” Quarterly Journal of Eco-
Exploitative Learning in New Product Development: A Social nomics, 108 (August), 577–98.
Capital Perspective on New Technology Ventures in China,”
Jaruzelski, Barry and Kevin Dehoff (2008), “Beyond Borders:
Journal of International Marketing, 15 (2), 1–29.
The Global Innovation 1000,” strategy+business, 53 (Winter),
Bartholomew, Susan (1997), “National Systems of Biotechnol- 1–18.
ogy Innovation: Complex Interdependence in the Global Sys-
Lee, Ruby P., Qimei Chen, Daekwan Kim, and Jean L. Johnson
tem,” Journal of International Business Studies, 28 (2),
(2008), “Knowledge Transfer Between Multinational Corpo-
241–66.
rations’ Headquarters and Their Subsidiaries: Influences on
Baum, Joel A.C., Tony Calabrese, and Brian S. Silverman and Implications for New Product Outcomes,” Journal of
(2000), “Don’t Go It Alone: Alliance Network Composition International Marketing, 16 (2), 1–31.
and Startups’ Performance in Canadian Biotechnology,”
March, James G. (1991), “Exploration and Exploitation in
Strategic Management Journal, 21 (3), 267–94.
Organizational Learning,” Organization Science, 2 (1),
Burt, Ronald S. (1983), “Range,” in Applied Network Analysis: 71–87.
A Methodological Introduction, Ronald S. Burt and Michael
Murray, Fiona (2002), “Innovation as Co-Evolution of Scien-
J. Minor, eds. Beverly Hills, CA: Sage Publications, 176–94.
tific and Technological Networks: Exploring Tissue Engineer-
Calantone, Roger J. and Yushan S. Zhao (2000), “Joint Ven- ing,” Research Policy, 31 (8–9), 1389–403.
tures in China: A Comparative Study of Japanese, Korean, Murray, Janet Y. and Mike C.H. Chao (2005), “A Cross-Team
and U.S. Partners,” Journal of International Marketing, 9 (1), Framework of International Knowledge Acquisition on New
1–23. Product Development Capabilities and New Product Market
Performance,” Journal of International Marketing, 13 (3),
Cockburn, Iain M., Rebecca M. Henderson, and Scott Stern
54–78.
(2000), “Untangling the Origins of Competitive Advantage,”
Strategic Management Journal, 21 (10–11), 1123–45. Musteen, Martina, John Francis, and Deepak K. Datta (2010),
“The Influence of International Networks on Internationaliza-
Cohen, Wesley M. and Daniel A. Levinthal (1989), “Innovation
tion Speed and Performance: A Study of Czech SMEs,” Jour-
and Learning: The Two Faces of R&D,” Economic Journal,
nal of World Business, 45 (3), 197–205.
99 (September), 569–96.
Özsomer, Ayflegül and Esra Gençtürk (2003), “A Resource-
Fabrizio, Kira R. (2009), “Absorptive Capacity and the Search
Based Model of Market Learning in the Subsidiary: The Capa-
for Innovation,” Research Policy, 38 (2), 255–67.
bilities of Exploration and Exploitation,” Journal of Inter-
Fleming, Lee and Olav Sorenson (2004), “Science as a Map in national Marketing, 11 (3), 1–29.
Technology Search,” Strategic Management Journal, 25 (8–9),
Penner-Hahn, Joan and J. Myles Shaver (2005), “Does Inter-
909–928.
national Research and Development Increase Patent Output?
An Analysis of Japanese Pharmaceutical Firms,” Strategic
Gittelman, Michelle and Bruce Kogut (2003), “Does Good Sci-
Management Journal, 26 (2), 121–40.
ence Lead to Valuable Knowledge? Biotechnology Firms and
the Evolutionary Logic of Citation Patterns,” Management Pisano, Gary P. (2006), Science Business. Boston: Harvard Busi-
Science, 49 (4), 366–82. ness School Press.

Greene, William H. (2003), Econometric Analysis. Upper Sad- Powell, Walter W., Kenneth W. Koput, and Laurel Smith-Doerr
dle River, NJ: Prentice Hall. (1996), “Interorganizational Collaboration and the Locus of

56 Journal of International Marketing


Innovation: Networks of Learning in Biotechnology,” Admin- Capabilities: Implications for Product Innovation and Market
istrative Science Quarterly, 41 (1), 116–45. Performance,” Journal of International Marketing, 15 (4),
63–93.
Rosenkopf, Lori and Atul Nerkar (2001), “Beyond Local
Search: Boundary-Spanning, Exploration, and Impact in the
Optical Disk Industry,” Strategic Management Journal, 22 THE AUTHORS
(4), 287–306.

Singh, Jasjit (2008), “Distributed R&D, Cross-Regional Knowl- Changsu Kim is Associate Professor of International
edge Integration and Quality of Innovative Output,” Research Business at Sogang University, Korea. He earned his
Policy, 37 (1), 77–96. doctoral degree at Rutgers University. His research
Sorescu, Alina B., Rajesh K. Chandy, and Jaideep C. Prabhu
interests include cross-border alliances, learning and
(2003), “Sources and Financial Consequences of Radical innovation, and foreign direct investment from emerg-
Innovation: Insights from Pharmaceuticals,” Journal of Mar- ing markets.
keting, 67 (October), 82–102.

Spencer, Jennifer (2003), “Global Gatekeeping, Representation, Jong-Hun Park is Professor of Strategic Management at
and Network Structure: A Longitudinal Analysis of Regional Sogang University, Korea. He obtained his doctoral
and Global Knowledge-Diffusion Networks,” Journal of
degree from the Sauder School of Business, University of
International Business Studies, 34 (5), 428–42.
British Columbia. His research interests center on strate-
Swaminathan, Vanitha and Christine Moorman (2009), “Mar- gic leadership, strategic alliance, and innovation
keting Alliances, Firm Networks, and Firm Value Creation,” management.
Journal of Marketing, 73 (September), 52–69.

Teece, David J. (1992), “Competition, Cooperation, and Inno-


vation: Organizational Arrangements for Regimes of Rapid
ACKNOWLEDGMENTS
Technological Progress,” Journal of Economic Behavior &
Organization, 18 (1), 1–25. The authors thank the anonymous JIM reviewers for
helpful comments on previous drafts of this article. This
Tushman, Michael L. (1977), “Special Boundary Roles in the work was supported by grants, including Korea
Innovation Process,” Administrative Science Quarterly, 22
Research Foundation’s World Class University project
(4), 587–605.
(R31-20002), Korea Research Foundation Grant (KRF-
Yalcinkaya, Goksel, Roger J. Calantone, and David A. Griffith 2009-32A-B00089), and Sogang University Research
(2007), “An Examination of Exploration and Exploitation Grant of 2010.

The Global Research-and-Development Network 57


Copyright of Journal of International Marketing is the property of American Marketing Association and its
content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for individual use.

You might also like