Content uploaded by Hsin-Yu Shih
Author content
All content in this area was uploaded by Hsin-Yu Shih on Aug 21, 2021
Content may be subject to copyright.
International diffusion of embodied and disembodied technology: A network
analysis approach
Hsin-Yu Shih
a,
⁎, Tung-Lung Steven Chang
b
a
Department of International Business Studies, National Chi Nan University, Puli, Taiwan
b
Department of Marketing and International Business, Long Island University, C. W. Post, Brookville, NY, USA
article info abstract
Article history:
Received 6 July 2008
Received in revised form 1 September 2008
Accepted 4 September 2008
This study proposes a quantitative method for investigating the structure of international
technology diffusion. By using network analysis, this study defines the structural configuration
of each country within the international diffusion network by measuring its degree, closeness,
and betweenness centralities. In addition, this study distinguishes between embodied
technology diffusion, measured by multilateral trade, and disembodied technology diffusion,
measured by patent citations, in individual countries. This study empirically tests a sample data
set of international technology diffusion taken from 48 countries. The empirical results show
that the structural configuration of countries exhibits similar patterns in both embodied and
disembodied diffusion networks. Significant global stratification patterns exist in the capability
of national international technology exportation and brokerage advantages. Moreover, this
study distinguishes four blocks of countries that play different roles in international technology
diffusion: the leading countries provide a source of technological knowledge; an intermediate
group diffuses the knowledge acquired from the source; a third group is in the process of
initiating the export of technological knowledge; and a final group of countries absorbs
technological knowledge without reciprocal exportation. Finally, this study identifies two types
of catch-up strategies that newly industrialized or developing countries can use to move up the
international technology stratification.
© 2008 Elsevier Inc. All rights reserved.
Keywords:
International technology diffusion
Network analysis
Embodied technology
Disembodied technology
International trade
Patent citation
1. Introduction
The contemporary growth theory suggests that technological progress is the primary driver of national economic growth [1–3].
Nations can adopt two approaches for achieving technological progress: endogenously enhancing their technological innovation
capabilities or alternatively by acquiring advanced technology through international technology diffusion. Although the capacity
for original innovation is one of the main sources of economic growth, the real impetus derives from the capacity to exploit the
economical potential and opportunities of inventions via widespread diffusion [4]. Recent research has demonstrated that in most
countries the main sources of technological progress leading to productivity gain are from abroad rather than domestically [5,6].
International technology diffusion thus has become a popular topic in the literature on economics and technology policy (e.g., [6–
11]). International technology diffusion is a comprehensive term dealing with mechanisms for shifting technological knowledge
across borders and the effective diffusion of technology into recipient economies [12]. Thus, it involves numerous complex
processes, leading to policy making in this area being especially complex and requiring careful consideration both at the individual
Technological Forecasting & Social Change 76 (2009) 821–834
⁎Corresponding author. Department of International Business Studies, National Chi Nan University, 1, University Rd., Puli, Nantou 545, Taiwan. Tel.: +886 49
2910960; fax: +886 49 2912595.
E-mail address: hyshih@ncnu.edu.tw (H.-Y. Shih).
0040-1625/$ –see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.techfore.2008.09.001
Contents lists available at ScienceDirect
Technological Forecasting & Social Change
country and multilateral levels. However, most extant studies have focused on the influence of international technology diffusion
on domestic productivity gain from the individual country perspective, but the network structure of technology diffusion within
the international economy from a multilateral perspective has been relatively ignored.
A practical method of studying international technology diffusion from a multilateral perspective involves investigating the
network characteristics of international technology diffusion by using “network analysis”, which is a well-developed set of
methods for systematically studying social networks. Although primarily developed for sociological study, the indicators and
techniques of network analysis are ideally suited for examining the structural features of multilateral technology diffusion. This
study discusses the techniques and indicators of network analysis used for examining the network characteristics of international
technology diffusion, and then we apply these techniques to explore the network structure of international technology diffusion
among countries.
International technology diffusion involves several different channels. As demonstrated by Kraay et al. [13],firms do not merely
conduct a single type of international activity associated with technology transfer; but rather perform several such activities, such
as trade, foreign direct investment, licensing, joint ventures, movement of personnel, and so on. Although a few recent studies have
simultaneously considered several technological channels, the majority of literature examines just one channel in isolation [9].
There is a growing need for research to simultaneously consider various channels of technology diffusion. Based on the argument
of Griliches [14] that two types of knowledge spillovers are generated by R&D activities, this study classifies international
technology diffusion channels into embodied and disembodied forms, enabling a distinction to be made between the benefits
associated with the use of imported capital goods in production, and those resulting from inwards flows of technological
knowledge to facilitate domestic R&D.
The purpose of this study is to investigate the structural differences between embodied and disembodied technology diffusion
networks. By using network analysis, we are able to reveal the blocks of countries that internationally originate, mediate, or absorb
technological knowledge, and also to examine the relationships that exist between embodied and disembodied diffusion
networks. This study aims to make to following contributions to the literature on international technology diffusion: First, this
study is the first to employ a multilateral perspective to examine the cross-country network structure of international technology
diffusion. The findings identify the structural configuration of technologically advanced countries and the competitive positions of
each country within the two international diffusion networks. Second, this study presents a systematic approach, which is rarely
adopted in the literature, to study international technology diffusion, which involves various complex processes. Third, the
findings regarding the relationships between embodied and disembodied international technology diffusion networks can help
explain the different approaches for countries upon the international technology diffusion market. Finally, this study addresses
policy implications for countries interested in acquiring foreign technology for productivity gain.
2. International technology diffusion: embodied vs. disembodied form
Technology has been viewed as a certain type of knowledge [3,6,7,12,15]. Maskus [12] defined technology as the information
necessary to achieve certain outcomes produced from a particular means of combining or processing selected inputs. A given
technology can generate multiple outputs; in addition, many technologies may generate the same outcome but with variable
efficiency. A technology can either be highly specific or can encompass multiple sub-processes, which produce finished and/or
intermediate goods within an overall value chain. Existing technology related research emphasizes two aspects of technology [6]:
(1) Technology is non-rival in the sense that the marginal costs involved in an additional agent using the technology are negligible;
(2) The return to technological investments is a mixture of private and public. The first aspect can be considered as the infinitely
expansible aspect of technology, indicating that technology can be used by other users for minimal additional cost [15–17]. Unlike
human and physical capital, which can only be used by one firm at a time, technology has unlimited potential to be simultaneously
adopted by numerous firms [9]. The second aspect demonstrates that while private returns must be strong enough to keep
innovation ongoing, technological investments frequently create benefits to individuals other than the inventor [6]. These external
effects are known as technology, or knowledge, spillovers.
Technology spillover is defined as technological knowledge being learned and absorbed into competition in such a way that
the benefits do not fully accrue to the original owner of the technology [12]. The benefits from technology spillovers for rivals
refer to lower costs, increased productivity, advantageous follow-on innovation, and other structural elements for which the
technology owner cannot charge the full value of their inventions through private transactions. Griliches [14] identified two
types of potential knowledge externality generated by R&D activities: rent spillovers and pure knowledge spillovers. Rent
spillovers occur when the prices of imported intermediate and capital goods do not completely embody the quality
improvements resulting from innovation activities in other firms or countries. Firms can enjoy the R&D of other firms by
buying one of the intermediate goods of such firms at a price below their practical value; the firm will then adopt a portion of
the rents of the innovating firm [18]. The rent spillovers occur because of imperfect and asymmetric information and the
impossibility of price discrimination by the innovating firm. Pure knowledge spillovers arise because of the imperfect
appropriability of knowledge associated with innovations. Knowledge diffusion occurs when the knowledge generated by a
firm contributes to innovation by other firms [18].
Rent spillovers occur through embodied knowledge flows, such as trade or direct investment flows; pure knowledge spillovers
are performed via disembodied knowledge flows, such as licensing or outsourcing agreements. The above phenomenon implies
that technology may also manifest in either embodied or disembodied forms. Information may be embodied in the form of
particular products, which might be reversely engineered to discover the underlying processes. On the other hand, technology may
822 H.-Y. Shih, T.-L.S. Chang / Technological Forecasting & Social Change 76 (2009) 821–834
be disembodied: it may be presented in its uncodified form, in the sense of requiring implicit know-how from personnel, or as
codified technology in the form of formulas, blueprints, drawings, patent applications, and so on [12]. Technology diffusion thus
can be addressed through either embodied or disembodied forms.
International technology diffusion refers to all the mechanisms through which countries acquire technological knowledge from
overseas rather than generating such knowledge domestically. Two fundamental mechanisms leading to international technology
diffusion have been identified [9]: (1) Employing specialized and advanced intermediate products that have been invented abroad;
and (2) Direct learning about foreign technological knowledge. The first mechanism involves the implicit usage of the
technological knowledge embodied in foreign intermediate goods for final-output production. A rent spillover occurs in this
process of international technology diffusion to the extent that the intermediate goods cost less than its opportunity costs, which
include the R&D costs of product development [6]. However, this is a relatively weak form of international technology diffusion,
due to the technological knowledge embodied in the importing intermediates which thus are not available to domestic inventors.
The embodied form of international technology diffusion is therefore regarded as a passive technology spillover. Conversely, the
second mechanism is termed an active technology spillover. Direct international learning or purchasing foreign technological
knowledge involves the explicit usage of disembodied knowledge in the forms of formulas, blueprints, drawings, and patent
applications. Pure knowledge spillover occurs internationally if technological knowledge is obtained from abroad for less than the
original cost to domestic inventors. Direct learning regarding foreign technological knowledge increases the domestic
technological stock of knowledge that can be actively adopted for innovation. This type of pure knowledge spillover can be
called the disembodied form of international technology diffusion.
3. Measuring international technology diffusion
As discussed previously, international technology diffusion consists of two forms: embodied diffusion and disembodied
diffusion. Empirically, embodied and disembodied diffusion are not very distinguishable, but the measurement in terms of
empirical data can mainly capture either embodied or disembodied diffusion. That is, work on international flows through
trade data is closer to modeling embodied diffusion, while work on international flows based on patents more closely
resembles disembodied diffusion [19].Thefirst form is the most notable mechanism related to the issue of international
technology diffusion in the literature. Most of these studies confirmed the existence of a significant and positive link between a
country's total factor productivity and its international trade, which they took to be evidence of international R&D spillovers
(e.g., [6,7,20,21]). However, the effect of international R&D spillovers differs among various types of trade. Coe et al. [22] found
that international trade in capital goods measures the trade-related spillover channel more accurately than total trade.
Moreover, Xu and Wang [23,24] used imports of machinery and equipment as a proxy for imports of capital goods. This study
thus considers the international import trade of machinery and equipment as a carrier of international technology diffusion in
embodied form.
Disembodied diffusion is the second form of international technology diffusion. Following Eaton and Kortum [8], Hu and Jaffe
[10], and Jaffe et al. [25], this study considers international patenting as a proxy of the channel of international technology diffusion
in disembodied form. Patent citations provide an appropriate method for measuring disembodied diffusion among various parties,
since for example if patent A is cited by patent B, this implies that the first patent represents a piece of prior existing knowledge
upon which patent B is based. Following Hu and Jaffe [10], this study assumes that the frequency with which inventors in a given
country cite the patents of another country serves as a proxy for the intensity of knowledge flow from the cited to the citing
country. The patent citation link can be viewed as an indicator of technological relevance, leading to pure knowledge spillovers
[18].
Regarding the measurement of international technology diffusion, this study adopts the most common method for
measuring the spillovers, that is, R&D expenditure of other country multiplied by different share weights, as defined by
Griliches [26]:
ITDij =wijRDi;i≠j:ð1Þ
Here, ITD
ij
denotes the extent to which country idiffuses technological knowledge to country j. Moreover, RD
i
represents the R&D
expenditure of country i. Additionally, the weight w
ij
represents the fraction of country i's knowledge that spills over to country j.
To differentiate embodied and disembodied technology diffusion requires constructing two different weighting formulas. In the
case of embodied technology diffusion, the foreign stock of knowledge consists of import weighted foreign R&D capital stocks, and
thus the weights (w
ij
) adopted to measure the stock of foreign knowledge are derived from economic transaction matrices. This
study defines the weights (w
E,ij
) of embodied technology diffusion as:
wE;ij =Mij
X
n
i=1 X
n
j=1
Mij
;i≠j;ð2Þ
where M
ij
denotes the imports of capital goods from country ito country j, and nindicates the number of countries involved. Based
on this measurement, this study assumes that as a given country increases imports of capital goods from another country with
relatively high R&D capital stock, the former countrywill increasingly benefit from embodied technology diffusion stemming from
the latter country.
823H.-Y. Shih, T.-L.S. Chang / Technological Forecasting & Social Change 76 (2009) 821–834
For disembodied technology diffusion, a patent citation represents a link to prior existing knowledge upon which inventors
build, so the weights (w
ij
) used to measure the stock of foreign knowledge can be calculated from international patent citation
matrices. The weights (w
D,ij
) of disembodied technology diffusion thus can be defined as:
wD;ij =Cij
X
n
i=1 X
n
j=1
Cij
;i≠j;ð3Þ
where C
ij
denotes the frequency with which patents of country iare cited by inventors of country j, and nindicates the number of
countries involved. Based on this measurement, this study assumes that the more a given country cites foreign patents from
another country with relatively high R&D knowledgestock, the more the former country will benefit from disembodied technology
diffusion derived from the latter country.
4. International technology diffusion networks: network analysis
This study aims to discuss the structural differences between embodied and disembodied diffusion networks with the help of
quantitative methodologies derived from network analysis. Based on graph theory, network analysis describes the structure of
interactions (displayed by edges) between given entities (displayed by nodes), and applies quantitative techniques to yield
relevant indicators for studying network characteristics and the position of individuals in a network structure. The related ideas of
network analysis have been extensively employed to study R&D networks [27,28], inter-organizational networks [29–31], and
technological systems [32,33]. This study employs network analysis to examine and compare the structural characteristics of
embodied and disembodied diffusion networks among countries, where the countries are treated as nodes and embodied or
disembodied technology diffusion among countries is treated as a series of edges.
Network analysis uses matrices to represent information on relationships among nodes. Using Eq. (1), a matrix ITD(n×n)of
international technology diffusion can be built. Each cell, ITD
ij
, of the matrix represents the extent to which country idiffuses
technological knowledge to country j.ITD(n×n) denotes a valued matrix, meaning that the edges of the network measure linkages
with different magnitude and thus must be dichotomized. Concerning the structure of empirical application, after building the
valued matrix, a single proper cut-off value must be chosen for dichotomizing the cells of the valued matrix to apply the binary
data to the indicators of network analysis. As a result of the dichotomizing process, the (i,j)th cell of the valued matrix becomes 0
when the value of international technology diffusion from country ito country jis below the selected cut-off value, and otherwise
it becomes 1. The dichotomized matrix is thus yielded. In this manner, the dichotomized matrix can offer binary data for measuring
network analysis indicators. The remainder of this section presents the appropriate indicators of network analysis for examining
the network characteristics of international technology diffusion.
One of the primary implications of network analysis is to identify the important or prominent nodes in a social network [34].
Centrally positioned individuals are usually located in strategic locations within a network and definitely enjoy a position of
privilege over those relegated to the circumference [35]. Social network analysts employ the concept of centrality to acquire the
positional features of individual nodes within their networks. In particular, Freeman [36,37] deduced three forms of centrality.
First, degree centrality is the simplest and most straightforward way to measure the centrality in terms of the number of nodes to
which a particular node connects. In directed networks, degree centrality of nodes can be distinguished in terms of their inward
and outward connections to measure their in-degree and out-degree centrality, respectively. The in-degree centrality (C
D,in
) and
out-degree centrality (C
D,out
) of a given node are defined as:
CD;in ni
ðÞ
=X
n
j=1
xij;in ;CD;out ni
ðÞ
=X
n
j=1
xij;out ð4Þ
where x
in
and x
out
respectively indicate one of the inward and outward connections of node i, and ndenotes the number of nodes
within the network. The implications of these two indicators corresponding to the examination of the network features of
international technology diffusion as inward and outward technological linkages of a country represents international technology
acquisitions and exportations, respectively. Comparing the two measures of in-degree and out-degree of a particular country can
display that the focal country act as a role of “source”,“intermediate”,or“absorber”in international technology diffusion market.
Closeness centrality is the second idea of centrality defined by Freeman. The measure of closeness centrality is based on
distance and centers on how close a node is to all other nodes in the set of nodes [34]. This is a global measurement that brings into
play the closeness to all network members, not just connections to immediate neighbors [35]. The index of closeness centrality (C
C
)
of a node is given by:
CCni
ðÞ
=n−1
P
n
j=1
dn
i;nj
:ð5Þ
Here, the d(n
i
,n
j
) denotes the geodesic distance that is defined as the length of the shortest path between nodes iand j, and n
indicates the number of nodes within the network. In directed networks, closeness centrality can be considered in terms of in-
closeness and out-closeness centralities on the basis of inward and outward connections, respectively. This indicator represents the
824 H.-Y. Shih, T.-L.S. Chang / Technological Forecasting & Social Change 76 (2009) 821–834
idea that a node is crucially central if it can be quickly reachable from and to all other nodes. Applying this indicator to the study of
international technology diffusion, as a country has numerous reachable other countries and it is closer apart in distance from
these reachable countries, its closeness centrality must be high, since it is more central and closer to all of the other countries in
international technology diffusion market.
The third measure of centrality is betweenness, which measures the extent to which a particular node lies on the paths between
various two nonadjacent nodes [34]. Betweenness centrality indicates the extent to which individuals are able to potentially have
control over the interactions between pairs of other nodes in the network. The betweenness centrality (C
B
) of a node is defined
as:
CBni
ðÞ=X
n
j=1 X
n
k=1
bjk ni
ðÞ
bjk
;j≠k≠ið6Þ
where b
jk
denotes the number of geodesics between nodes jand k, and b
jk
(n
i
) denotes the number of geodesics linking the two
nodes that contain node i. The betweenness of a node measures the extent to which it can play the role of a gatekeeper or broker
with a potential for control over the interactions between pairs of others [38]. Betweenness centrality is an appropriate indicator
measuring the extent to which nodes broker indirect connections between all other nodes in a network. Additionally, increasing
redundant connections in a network decreases the efficacy of the brokerage advantage of nodes; increasing non-redundant
connections would improve. In the context of international technology diffusion, a particular country with higher betweenness
centrality represents more opportunities to broker the flows of diffusion among other countries since most technology diffusion
will pass through this country, and thus it should possess competitive advantages in terms of brokerage opportunities.
In addition, it is important for sociologists to distinguish subsets
1
within a network in terms of individuals or social positions
according to the relations they keep withothers, which is the way of defining thesocial role of an individual in a social network. We
can expect that individuals belonging to the same subset would exhibit similar behavior even without direct linkages existing
among them. The network concept used to group subsets in terms of individuals sharing the same relations is called equivalence,
and the network technique used to identify such subsets is called blockmodeling [39]. Additionally, a blockmodel of relations
between blocks or positions can be characterized in terms of aggregate relationships between countries in the respective blocks or
positions [40]. Two alternative methods exist for operationalizing blockmodeling. The first method for determining blocks based
on structural equivalence is related to the CONCOR algorithm [41], which has been used by Snyder and Kick [42], and Nemeth and
Smith [43]. Meanwhile, the second method derives from regular equivalence and is associated with the REGE algorithm [44], which
has been adopted by Smith and White [40]. However, Schweizer [45] demonstrated how CONCOR can conflate spatial proximity
with global role structure, while REGE can precisely identify generic structural positions in a network. This study thus uses the
REGE algorithm as the blockmodeling approach for identifying blocks in technology diffusion networks. With the application of the
REGE algorithm corresponding to the study of international technology diffusion, it is possible to distinguish blocks of countries in
the embodied and disembodied diffusion networks on the basis of patterns in their network ties, and furthermore it is possible to
describe the patterns of aggregate relations among these blocks.
5. Data
The study sample consists of 48 countries derived from among the top 60 nations based on the innovation factors of Global
Competitiveness [46]. This study requires the data sets of bilateral trade, patent citations, and R&D expenditure. To avoid bias
resulting from the unstable variances of the annual data sets, this study uses the average of annual data taken from the 1997–1999
period instead of annual data for a certain year. Regarding the bilateral trader data sets, although the data sources provide
information on both exports and imports, this study only uses import data because there is various evidence indicating that import
figures are more accurate than export figures [40,47]. In addition, this study uses imports of machinery and equipment as a proxy
for imports of capital goods (the data is sourced from the World Trade Atlas of the Global Trade Information Services). Next, the
patent data used in this study, including country of assignee and the identification of previous patents cited in a given patent, are
obtained from the utility patents granted by the U.S. Patent and Trademark Office, sourced from the NBER Patent Citations Database
[48]. Finally, the R&D expenditure data for eachcountry is sourced from the World Development Indicators of the World Bank. This
study examines the networks of technology diffusion among countries using all three of the above data sets. For various reasons,
some countries failed to report all of the three data required for all three years. Consequently, the final sample included just 48
countries.
6. Analyses and results
Following the measures of international technology diffusion mentioned in Section 3, we can obtain the valued matrices ITD
(n×n) of the embodied and disembodied diffusion networks, respectively. Next, appropriate cut-off values must be selected for
1
There are two main approaches to distinguish subsets of a network in network analysis. The first approach is to look at existing relations and distinguish
subsets according to their cohesion. The second one is based on individuals sharing the same relations and distinguishes subsets according to their equivalence.
Equivalence is a cognitive operation that sociologist performs to describe relations and social role [35]. For the purpose of this study, we adopt the equivalence
approach as the method of distinguishing subsets.
825H.-Y. Shih, T.-L.S. Chang / Technological Forecasting & Social Change 76 (2009) 821–834
dichotomizing the cells of the valued matrices to apply the binary data to the network analysis indicators. Although the selection of
cut-off value is arbitrary, two steps can be implemented for preliminary sensitivity testing to identify the most appropriate value
[33]. First, since the difference in the structural patterns of the networks are reasonably stable while the cut-off valueschange from
very low to very high, a single cut-off value can be selected for each valued matrix. Second, the appropriate cut-off value must be
chosen on the basis of the heuristic criteria that the structural features of the networks can be detectable, rather than the very high
or very low values that characterize nearly totally unconnected or almost completely connected networks. Additionally, it is not
meaningful to set the same cut-off value for the two valued matrices due to their heterogeneity. However, it is possible to select
two different cut-off values for each matrix to generate the same densityof both networks and thus maintain a comparable base for
both networks. Following these considerations, appropriate cut-off values are chosen for building the dichotomized metrics, with
Table 1
Network indicators of embodied technology diffusion.
Country Block Degree centrality Closeness centrality Betweenness
centrality
In-degree Out-degree In-closeness Out-closeness
United States 1 27 47 5.85 100 264.19
Germany 1 24 47 5.83 100 232.94
Japan 1 16 47 5.77 100 66.08
United Kingdom 1 22 46 5.82 97.92 88.87
France 1 19 45 5.80 95.92 52.70
Italy 1 17 43 5.78 92.16 33.68
Sweden 1 17 36 5.78 81.03 65.78
Netherlands 1 19 33 5.80 77.05 26.09
South Korea 1 15 32 5.77 75.81 18.75
China 1 17 32 5.78 75.81 17.16
Switzerland 1 12 30 5.75 73.44 5.00
Taiwan 1 14 29 5.76 72.31 8.26
Canada 2 13 19 5.75 62.67 10.10
Belgium 2 13 17 5.75 61.04 1.77
Spain 2 14 16 5.76 60.26 2.23
Singapore 2 17 15 5.78 59.49 8.81
Finland 2 13 13 5.75 58.03 2.77
Austria 2 11 12 5.74 57.32 0.56
Denmark 2 12 9 5.75 55.29 0.19
Australia 2 14 8 5.76 54.65 1.70
Hong Kong 3 16 6 5.77 53.41 0.67
Ireland 3 13 5 5.75 52.81 0
Brazil 3 13 4 5.75 52.22 0.27
Israel 3 8 4 5.71 52.22 0
Malaysia 3 14 3 5.76 51.65 0
Mexico 3 14 2 5.76 51.09 0.44
Czech 3 11 1 5.74 50.54 0
Hungary 3 11 1 5.73 50.54 0
India 3 11 1 5.73 50.54 0
Norway 3 11 1 5.74 45.19 0
Poland 3 16 1 5.77 50.54 0
Thailand 3 13 1 5.75 50.54 0
Chile 4 11 0 6.08 2.08 0
Colombia 4 7 0 6.04 2.08 0
Cyprus 4 5 0 6.03 2.08 0
Estonia 4 5 0 6.03 2.08 0
Greece 4 8 0 6.06 2.08 0
Iceland 4 4 0 6.02 2.08 0
Indonesia 4 11 0 6.08 2.08 0
Lithuania 4 5 0 6.03 2.08 0
Malta 4 6 0 6.03 2.08 0
New Zealand 4 7 0 6.04 2.08 0
Portugal 4 11 0 6.08 2.08 0
Slovakia 4 6 0 6.03 2.08 0
Slovenia 4 6 0 6.03 2.08 0
South Africa 4 12 0 6.09 2.08 0
Turkey 4 14 0 6.10 2.08 0
United Arab Emirates 4 11 0 6.08 2.08 0
Descriptive statistics
Sum (S) 606 606 281.36 2154.79 909.00
Mean (M) 12.63 12.63 5.86 4 4.89 18.94
Variance (Var) 24.37 273.22 0.02 1147.70 2743.84
Standard deviation (S.D.) 4.94 16.53 0.14 33.88 52.38
Min. 4 0 5.71 2.08 0
Max. 27 47 6.10 100 264.19
826 H.-Y. Shih, T.-L.S. Chang / Technological Forecasting & Social Change 76 (2009) 821–834
values being set to 703 (US dollar) for the embodied diffusion network and 0.034 (US dollar) for the disembodied diffusion
network. Individual country network indicators derived from embodied and disembodied diffusion can be calculated to reveal the
structural configuration of each country within the two different forms of diffusion networks by measuring the indicators of degree
centrality, closeness centrality, and betweenness centrality, and by assessing the patterns of aggregate relations among blocks,
which are all calculated by UCINET 6 [49]. The network indicators of embodied and disembodied technology diffusion are shown in
Tables 1 and 2, respectively. In addition, a preliminary visual assessment of global networks can be captured based on the network
graph approach. Figs. 1 and 2 show the network graphs of embodied and disembodied technology diffusion, respectively, where a
set of nodes denote countries, and a set of arcs directed between pairs of nodes represent the directional technology diffusion
Table 2
Network indicators of disembodied technology diffusion.
Country Block Degree centrality Closeness centrality Betweenness
centrality
In-degree Out-degree In-closeness Out-closeness
United States 1 32 47 6.25 100 446.99
Japan 1 27 44 6.21 94.00 146.99
Germany 1 25 43 6.19 92.16 103.02
United Kingdom 1 24 40 6.18 87.04 71.74
France 1 21 40 6.16 87.04 53.67
Canada 1 24 33 6.18 77.05 43.03
Italy 1 19 33 6.14 77.05 17.20
Switzerland 1 20 29 6.15 72.31 11.23
Netherlands 1 19 29 6.14 72.31 10.98
Sweden 2 20 27 6.15 70.15 8.08
South Korea 2 22 25 6.17 68.12 12.72
Taiwan 2 22 25 6.17 68.12 12.37
Australia 2 19 23 6.14 66.20 3.77
Belgium 2 17 20 6.13 63.51 1.16
Israel 2 18 18 6.14 61.84 0.99
Finland 2 18 18 6.14 61.84 0.70
Denmark 2 18 17 6.14 61.04 0.18
Spain 2 15 17 6.11 61.04 0.13
Austria 2 17 17 6.13 61.04 0.06
Norway 2 14 14 6.10 58.75 0
Brazil 2 12 10 6.09 55.95 0
China 2 12 8 6.09 54.65 0
Singapore 3 12 5 6.09 52.81 0
Ireland 3 13 5 6.10 52.81 0
South Africa 3 12 5 6.09 52.81 0
Mexico 3 11 4 6.08 52.22 0
India 3 10 3 6.07 51.65 0
Hong Kong 3 15 2 6.11 51.09 0
New Zealand 3 11 1 6.08 50.54 0
Greece 3 7 1 6.05 50.54 0
Hungary 3 8 1 6.06 50.54 0
Poland 3 6 1 6.04 50.54 0
Portugal 3 6 1 6.04 50.54 0
Chile 4 5 0 6.41 2.08 0
Colombia 4 6 0 6.42 2.08 0
Cyprus 4 2 0 6.39 2.08 0
Czech 4 7 0 6.43 2.08 0
Estonia 4 1 0 6.38 2.08 0
Iceland 4 4 0 6.40 2.08 0
Indonesia 4 5 0 6.41 2.08 0
Lithuania 4 3 0 6.40 2.08 0
Malaysia 4 5 0 6.41 2.08 0
Malta 4 2 0 6.39 2.08 0
Slovakia 4 4 0 6.40 2.08 0
Slovenia 4 6 0 6.42 2.08 0
Thailand 4 5 0 6.41 2.08 0
Turkey 4 4 0 6.40 2.08 0
United Arab Emirates 4 1 0 6.38 2.08 0
Descriptive statistics
Sum (S) 606 606 298.16 2168.51 945.00
Mean (M) 12.63 12.63 6.21 45.18 19.69
Variance (Var) 62.28 219.05 0.02 1001.28 4770.87
Standard deviation (S.D.) 7.89 14.80 0.14 31.64 69.07
Min. 1 0 6.04 2.08 0
Max. 32 47 6.43 100 446.99
827H.-Y. Shih, T.-L.S. Chang / Technological Forecasting & Social Change 76 (2009) 821–834
between countries. The remainder of this section describes and compares the structural features of embodied and disembodied
technology diffusion networks from global and block perspectives.
6.1. Global level
As shown in Table 1, the embodied diffusion network demonstrates significantly higher variance of out-degree centrality than
in-degree centrality (Var
out,E
=273.22, Var
in,E
=24.37). The out-degree centrality of the network ranges between 0 and 47, leading
Fig. 2. Network graph of disembodied technology diffusion.
Fig. 1. Network graph of embodied technology diffusion.
828 H.-Y. Shih, T.-L.S. Chang / Technological Forecasting & Social Change 76 (2009) 821–834
to average variability from one country to the next is 16.53 (S.D.
out,E
), exceeding the mean (12.63). Considerable variation thus
exists in out-degree centrality in the embodied network. On the contrary, the variance in in-degree centrality appears much more
stable, and has a standard deviation of 4.94 (S.D.
in,E
), less than the mean (12.63). The patterns of in-degree and out-degree
centrality in the disembodied network (Var
out,D
=219.05, Var
in,D
=62.28; S.D.
out,D
=14.80, S.D.
in,D
=7.89) are quite similar to
those in the embodied network (see Table 2). A structural basis for stratification may thus exist to the extent that the links of a
network are not evenly distributed. The high variation in the out-degree centrality demonstrates that the out-degree partitions of
both diffusion networks can be regarded as hierarchic networks.
From the outward linkage perspective, as shown in Tables 1 and 2, both networks are characterized by considerable
concentration or centralization. Specifically, the power of individual countries varies substantially in outward linkages, meaning
that the advantages are rather unequally distributed in the outward linkage parts of both networks. However, from the inward
linkage perspective both networks are quite evenly distributed, but the embodied network is less concentrated than the
disembodied one (Var
in,E
=24.37 vs. Var
in,D
=62.28). That is, individual countries within the embodied network share power and
advantages in the inward linkage more equally than those within the disembodied network.
The out-degree centrality of a country represents the extent to which it exports technological knowledge to other countries. The
empirical results show that a significant global stratification pattern exists in the distribution of technology export capability. As
shown in Table 1, United States, Germany, and Japan export embodied technological knowledge to all of the other countries except
itself, while Chile, Colombia, Cyprus, Estonia, Greece, Iceland, Indonesia, Lithuania, Malta, New Zealand, Portugal, Slovakia,
Slovenia, South Africa, Turkey, and United Arab Emirates have no capability of exporting embodied technological knowledge to
other countries. In the disembodied diffusion network (see Table 2), United States is the only country that contributes to all other
47 countries, but Chile, Colombia, Cyprus, Czech, Estonia, Iceland, Indonesia, Lithuania, Malaysia, Malta, Slovakia, Slovenia,
Thailand, Turkey, and United Arab Emirates can not export disembodied technological knowledge.
The closeness centrality indicates the extent to which a particular country is reachable from and to other countries. As shown in
Tables 1 and 2, the patterns of in-closeness and out-closeness centrality resemble those of degree centrality in both diffusion networks.
Both diffusion networks thus have a hierarchical structure in terms of out-closeness centrality but have an evenly distributed structure
in terms of in-closeness centrality (Var
out,E
=1147.70, Var
out,D
=1001.28 vs. Var
in,E
=0.02, Var
in,D
=0.02). For example, the United
States, Germany, and Japan in the embodied network, and the United States in the disembodied network, have 100% out-closeness
centrality resulting from their comprehensive export links to other countries. Both diffusion networks are quite evenly distributed in
terms of in-degree centrality, and thus no significant difference exists among the distributionof in-closeness centrality in each network.
Betweenness centrality measures the extent to which a country can act as a broker or gatekeeper capable of controlling
technology diffusion between pairs of other countries. The significant variation (Var
B,E
=2743.84, Var
B,D
=4770.87) in the
betweenness centrality of both diffusion networks reveals that the broker advantages of individual countries are quite unevenly
distributed. Countries with higher betweenness centrality, for instance the United States (C
B
= 264.19) and Germany (C
B
=232.94)
in the embodied network, and the United States (C
B
=446.99) in the disembodied network, which act as critical brokers between
pairs of other countries, possess extensive competitive advantages in terms of brokerage opportunities. In contrast, there are 25
countries in the embodied network and 29 countries in the disembodied network with zero value of betweenness centrality,
indicating they are unable to have any advantage in terms of brokerage opportunities.
As discussed previously, a country's brokerage opportunities depend on the non-redundant connections in its existing linkages.
For instance, in the embodied network, Canada (C
B
=10.10) should possess more brokerage opportunities between pairs of other
countries than Switzerland (C
B
= 5.00) or Taiwan (C
B
= 8.26), even though Switzerland (C
D,out
= 30) and Taiwan (C
D,out
= 29) have
much higher out-degree centralities than Canada (C
D,out
= 19). This phenomenon implies that Switzerland and Taiwan possess lots
of redundant connections among their existing linkages, while Canada has fewer redundant connections.
6.2. Block level
Blockmodeling is conducted herein to simplify the patterns of relations between blocks in terms of aggregate relationships
between countries in the respective blocks. This study uses the REGE algorithm as the blockmodeling approach for determining
blocks in both technology diffusion networks. The empirical results of blocks in embodied and disembodied networks are listed in
the second column of Tables 1 and 2.
Four blocks are identified in the embodied diffusion network, and the reduced model in terms of aggregate relationships
between countries is listed in Table 3. International technology diffusion proceeds from row blocks to column blocks. The network
Table 3
The reduced matrix of embodied technology diffusion.
Block 1 Block 2 Block 3 Block 4
Block 1 Full connected block
(nearly) (Density=0.992)
Strong tie (Density= 0.948) Strong tie (Density= 0.882) Medium tie (Density = 0.615)
Block 2 Medium tie (Density =0.656) Sparsely connected block
(Density=0.232)
Weak tie (Density= 0.240) Extremely weak tie (Density= 0.078)
Block 3 Weak tie (Density= 0.174) Extremely weak tie
(Density=0.031)
Non-connected block
(nearly) (Density= 0.007)
Extremely weak tie(Density = 0.0 05)
Block 4 No tie (Density= 0) No tie (Density = 0) No tie (Density= 0) Non-connected block (Density= 0)
829H.-Y. Shih, T.-L.S. Chang / Technological Forecasting & Social Change 76 (2009) 821–834
indicators of the constituent countries in a certain block are quite similar to one another. However, because of the existence of a
global stratification pattern in international technology diffusion capabilities, network indicators differ remarkably between
blocks. Block 1 consists of 12 countries, including United States, Germany, Japan, United Kingdom, France, Italy, Sweden,
Netherlands, South Korea, China, Switzerland, and Taiwan, each of which possesses high values of degree, closeness and
betweenness centralities. Additionally, the countries in Block 1 are closely linked to one another (density of Block 1= 0.992). Block
1 is characterized by the high out-degree and out-closeness centralities of the countries within it, and thus the constituent
countries of this block are strongly capable of exporting embodied technological knowledge. Therefore, this study defines the role
or position of this block in the embodied diffusion network as a source of embodied technological knowledge. The countries within
the source block also have numerous advantages in terms of opportunities for brokerage with other countries due to their high
betweenness centrality.
Next, Block 2 contains 8 countries, including Canada, Belgium, Spain, Singapore, Finland, Austria, Denmark, and Australia, with
the second highest value of indicators in general, and its block density (density of Block 2= 0.232) is the second highest value
among the four blocks but significantly less than that of Block 1. The constituent countries of this block occupy an intermediate role
or position in the embodied diffusion network on the basis of their similar values in terms of in-degree and out-degree centralities.
Block 2 thus can be defined as an intermediate block. On one hand, countries in the intermediate block absorb embodied
technological knowledge from leading countries; on the other hand, such countries are also capable of exporting embodied
technological knowledge. However, based on their low betweenness centrality, the linkages of such countries contain numerous
redundant connections, and thus they have less ability to play the role of a broker to control over other countries.
Block 3 comprises 12 countries, including Hong Kong, Ireland, Brazil, Israel, Malaysia, Mexico, Czech, Hungary, India, Norway,
Poland, and Thailand, with lower value of individual indicators and block density (density of Block 3 =0.007) than those of block 2.
The fact that the out-degree centralities of the constituent countries in this block are considerably lower than their in-degree
centralities demonstrates that they have just begun to develop the capability of exporting embodied knowledge to other countries,
and still heavily rely upon the absorption of foreign embodied technological knowledge. Consequently, Block 3 is defined as a
beginner block. Generally, as shown in Table 1, the countries in the beginner block are unable to build non-redundant connections
in their linkages, and thus have very limited advantages in terms of brokerage opportunities.
Finally, Block 4 is the largest block in the embodied diffusion network, consisting of 16 countries, including Chile, Colombia,
Cyprus, Estonia, Greece, Iceland, Indonesia, Lithuania, Malta, New Zealand, Portugal, Slovakia, Slovenia, South Africa, Turkey, and
United Arab Emirates, but the countries in this block have the lowest values of individual indicators and do not share any mutual
linkages (density of Block 4= 0). The countries belonging to this block can only absorb embodied technological knowledge from
advanced countries but are unable to export embodied knowledge to other countries owing to the zero value of their out-degree
centrality. Block 4 is thus defined as an absorber block. The countries in the absorber block do not have any brokerage opportunities
because they have no outward linkage, as confirmed by their zero values of betweenness centrality.
Similar patterns exist in the disembodied diffusion network as in the embodied diffusion network, and the reduced model of
the disembodied network is shown in Table 4. Four blocks are identified in the disembodied diffusion network, and the constituent
network indicators and block density decrease progressively (densities of blocks 1, 2, 3, and 4 are 1, 0.673, 0, and 0, respectively).
Block 1 contains 9 countries, including United States, Japan, Germany, United Kingdom, France, Canada, Italy, Switzerland, and
Netherlands, has the highest value of all of the network indicators in general, and can be characterized as a source in the
disembodied technology network. Block 2 consists of 13 countries, including Sweden, South Korea, Taiwan, Australia, Belgium,
Israel, Finland, Denmark, Spain, Austria, Norway, Brazil, and China, with the second highest value of all indicators in general, and
plays an intermediary role in the disembodied technology network. Additionally, block 3 comprises 11 countries, including
Singapore, Ireland, South Africa, Mexico, India, Hong Kong, New Zealand, Greece, Hungary, Poland, and Portugal, all of which are in
the process of developing disembodied knowledge exportation, and thus is defined as the beginner block. Finally, block 4 consists
of 15 countries, including Chile,Colombia, Cyprus, Czech, Estonia, Iceland, Indonesia, Lithuania, Malaysia, Malta, Slovakia, Slovenia,
Thailand, Turkey, and United Arab Emirates, is the largest block in the disembodied diffusion network, and is characterized as an
absorber on the basis of the zero value of the out-degree and betweenness centralities of these countries.
Based on the features of each block, the reduced graphs of embodied and disembodied technology diffusion networks can be
constructed to display the international technology diffusion patterns of inter-block relations. The reduced pattern features and
graphs of both networks are shown in Tables 3 and 4, and Figs. 3 and 4, respectively. The constituent countries of the source blocks
in both networks serve as the main technology suppliers to other countries. Block 1 has strong ties to Blocks 2 and 3, as well as
having medium ties to Block 4. The constituent countries of Block 2 act as an intermediate role in both networks, not only
absorbing technological knowledge from advanced countries but also being capable of exporting knowledge to other countries.
Block 2 has weak ties to Block 3 in both networks and a medium tie to Block 1 in the embodied network but a strong tie to Block 1 in
Table 4
The reduced matrix of disembodied technology diffusion.
Block 1 Block 2 Block 3 Block 4
Block 1 Full connected block (Density = 1) Strong tie (Density= 1) Strong tie (Density =0.899) Medium tie (Density = 0.444)
Block 2 Strong tie (Density = 0.957) Densely connected block (Density = 0.673) Weak tie (Density = 0.154) No tie (Density = 0)
Block 3 Weak tie (Density=0.273) Extremely weak tie (Density=0.014) Non-connected block (Density= 0) No tie (Density=0)
Block 4 No tie (Density = 0) No tie (Density = 0) No tie (Density = 0) Non-connected block(Density= 0)
830 H.-Y. Shih, T.-L.S. Chang / Technological Forecasting & Social Change 76 (2009) 821–834
the disembodied network. Furthermore, block 3 depends on the absorption of technological knowledge from advanced blocks and
thus receives a strong link from Block 1 and a weak link from Block 2. On the other hand, the constituent countries of Block 3 fulfill a
beginner role in both networks and thus start to develop the capability of exporting technological knowledge. However, the main
exporting partners for Block 3 are the constituent countries of Block 1, and Block 3 has extremely weak or even no ties to Blocks 2
and 4. These findings indicate that the late developing countries, namely the countries in Block 3, select the most advanced
countries, namely those in Block 1, as their exporting partners instead of selecting countries with similar technological capacities to
themselves, such as the countries in Blocks 2 or 4. Finally, the constituent countries of Block 4 play an absorber role and have no
outward linkages to other countries in either network. They primarily absorb technological knowledge from the constituent
countries of the source block.
Table 5 examines the consistency between strata of constituent countries in both networks. Most countries consistently stay
between strata in both networks, but there are eight countries whose belonging blocks in the embodied network are upper than
their belonging blocks in the disembodied network, that is, Sweden, South Korea, China, Taiwan, Singapore, Malaysia, Czech, and
Thailand; and another eight countries are positioned oppositely, that is, Canada, Brazil, Israel, Norway, Greece, New Zealand,
Portugal, and South Africa. This phenomenon implies the existence of two distinguishing strategies for cultivating technological
knowledge. The countries whose blocks in the embodied network exceed their corresponding blocks in the disembodied network
take advantage of technologies embodied in capital goods to strengthen their technology capabilities (Embodied Plus). Numerous
previous studies have documented that many newly industrialized developing countries in East Asia adopt export-oriented
development strategies to move up the international technology ladder [50,51]. For developing countries in East Asia,
competitiveness during the catch-up process is not based on basic research or radical new products or processes. Instead, these
countries catch up with the industrial countries through incremental improvements to existing products and processes, using
engineering and technical skills rather than R&D [52]. This argument is confirmed by evidence from the six East Asian countries in
this group, namely South Korea, China, Taiwan, Singapore, Malaysia, and Thailand. New knowledge learned in terms of technology
embodied in capital goods deepens and broadens the technological bases of these countries, leading to opportunities for further
patent-related innovation. On the contrary, the countries whose belonging blocks in the disembodied network are higher than
their corresponding blocks in the embodied network exploit their advantages in patent performance to foster their technology
capabilities through research (Disembodied Plus). Based on the superior innovation knowledge accumulated in the form of
Fig. 4. Reduced graph of disembodied technology diffusion.
Fig. 3. Reduced graph of embodied technology diffusion.
831H.-Y. Shih, T.-L.S. Chang / Technological Forecasting & Social Change 76 (2009) 821–834
patents, countries in this group pursue opportunities for further innovation related to product engineering and development. The
contracted term “R&D”can be divided into demands for scientific research and the development of commercial results [53], where
research leads to the generation of advantages in terms of pure knowledge, while development aims at production. Therefore, two
catch-up strategies can be identified (Table 5): the countries in the Embodied Plus group implement the “development-oriented”
strategy, while the countries in the Disembodied Plus group dothe “research-oriented”strategy. The countries that are consistently
ranked in the lower strata can follow one of these two catch-up strategies to improve their ranking within the international
technological hierarchy.
7. Conclusions
This study examines the network structure of international technology diffusion in terms of embodied and disembodied
diffusion networks. The structural configuration of each country within these two networks is defined by measuring the indicators
of network analysis, including degree centrality, closeness centrality, and betweenness centrality. Block modeling is used to define
the position of each country within four distinct blocks and also to distinguish the aggregate relationships between the blocks. This
study thus demonstrates the applicability of network analysis to illustrate blocks of countries that originate, mediate, or absorb
international technology diffusion, and reveals the relationships between embodied and disembodied diffusion networks. The
empirical findings of this study are summarized below.
First, both embodied and disembodied technology diffusion networks are characterized by high centralization and are regarded
as strongly hierarchical networks from the outward linkage perspective. That is, both networks display significant global
stratification patterns in terms of national ability to export technological knowledge. On the contrary, from the inward linkage
perspective, the comparatively even distribution of both networks indicates more equal sharing of power and advantages between
individual countries.
Second, both networks exhibit a highly variable distribution of betweenness centrality. This finding thus indicates an uneven
distribution in individual national brokerage advantages in terms of their capacity to mediate technology diffusion between pairs
of countries.
Third, based on blockmodeling, four blocks are identified in both networks. The constituent countries in each block share
similar network indicators. The constituent countries of the source block have strong mutual linkages with one another and
possess high values of degree, closeness and betweenness centralities. The intermediate block is characterized by similar values
Table 5
The inter-block classification table between embodied and disembodied technology diffusion.
Disembodied Source (Block 1) Intermediate (Block 2) Beginner (Block 3) Absorber (Block 4)
Embodied
Source (Block 1) United States Sweden
Germany South Korea
Japan China
United Kingdom Taiwan
France
Italy
Netherlands
Switzerland
Intermediate (Block 2) Canada Belgium Singapore
Spain
Finland
Austria
Denmark
Australia
Beginner (Block 3) Brazil Hong Kong Malaysia
Israel Ireland Czech
Norway Mexico Thailand
Hungary
India
Poland
Absorber (Block 4) Greece Chile
New Zealand Colombia
Portugal Cyprus
South Africa Estonia
Iceland
Indonesia
Lithuania
Malta
Slovakia
Slovenia
Turkey
United Arab Emirates
832 H.-Y. Shih, T.-L.S. Chang / Technological Forecasting & Social Change 76 (2009) 821–834
between in-degree and out-degree centralities, and thus plays an intermediary role in both networks. The constituent countries of
the beginner block are starting to develop the ability to export technological knowledge to other countries based on their out-
degree centralities being much lower than their in-degree centralities. The absorber block is featured by the zero value of out-
degree centrality, and therefore can only absorb technological knowledge from advanced countries but lacks knowledge related to
reciprocal exportation.
Fourth, on the basis of reduced matrices and graphs of both networks, the source blocks serve as the main technology suppliers
to the other three blocks. The intermediate blocks primarily absorb technological knowledge from the source blocks and export
knowledge back to the source blocks and forward to the beginner blocks. Meanwhile, the beginner blocks mainly depend on the
absorption of technological knowledge from advanced blocks and also export their limited knowledge to the source blocks. The
absorber blocks merely absorb technological knowledge from the source blocks.
Finally, although most of the countries stay consistently between strata of constituent countries in both networks, eight
countries belong to blocks in the embodied network that are higher than their corresponding blocks in the disembodied network,
and another eight countries are inversely positioned. The former eight countries take advantage of technology embodied in capital
goods to improve their technological capability, while the latter eight countries cultivate their advantages in patent performance to
foster their technological capability.
Based on the empirical findings of this study, a significant global stratification pattern of national technological capability exists
in both international technology diffusion networks. Countries with higher positions in the hierarchyare expected to possess more
advanced technological knowledge and enjoy higher productivity gains. Therefore, it is important for countries with a lower
position in the hierarchy to improve their technological capability and thus improve their ranking in the international
technological hierarchy. The findings of this study imply that there are two different types of catch-up strategies that lower-strata
countries can adopt. One strategy, defined as the development-oriented strategy, involves using new knowledge learned through
the spillover of technology embodied in capital goods to enrich national technological base, and subsequently perform innovations
in the form of disembodied technology. The other approach is the research-oriented strategy, inwhich countries first upgrade their
technological capabilities in disembodied form, and then perform further innovation via the embodied forms of product
engineering and development to achieve improved status in both networks. These two distinguishing approaches provide
different strategies and paths for catch-up countries to improve their position in the international technology hierarchy.
It is important to acknowledge the limitations of this study since such acknowledgement can help in identifying future research
directions. First, this study examines data from a limited time period, and thus the changes in the relative positions of countries
over periods of years within the two diffusion networks cannot be traced. Future works can trace these relative positions through
repetitive longitudinal studies using the same methodology, which may demonstrate the potential policy implications of
determining an effective path for technology cultivation. Second, Xu and Chiang [11] examined the impacts of international
technology diffusion via trade and patenting on productivity growth and found that countries at different development stages
benefit from different forms of international technology spillover. Due to the lack of a performance-related index in this study, it is
impossible to compare whether the research-oriented or the development-oriented strategy can perform better than its
alternative at different developmental stages. Future studies can use technological performance-related indices to properly
address the above issues. The third limitation of this study is that it only considers information on patents granted by the U.S.
Patent and Trademark Office. Since data is not obtained for patents granted by non-U.S. organizations, this study is overweighted
on U.S.-granted patents. Future studies could include information on patents granted by other international patent organizations,
such as EPO (European Patent Office) and JPO (Japan Patent Office), to eliminate this bias.
References
[1] R.J. Barro, X. Sala-i-Martin, Capital mobility in neoclassical models of growth, Am. Econ. Rev. 85 (1) (1995) 103–115.
[2] R.J. Barro, X. Sala-i-Martin, Technological diffusion, convergence, and growth, J. Econ. Growth 2 (1) (1997) 1–26.
[3] G. Grossman, E. Helpman, Innovation and growth in the world economy, MIT Press, Cambrid ge, MA, 1991.
[4] E. Helpman, General Purpose Technologies and Economic Growth, MIT Press, Cambridge, 1998.
[5] J. Eaton, S. Kortum, Engines of growth: domestic and foreign sources of innovation, Japan World Econ. 9 (1997) 235–259.
[6] W. Keller, International technology diffusion, J. Econ. Lit. XLII (2004) 752–782.
[7] D.T. Coe, E. Helpman, International R&D spillovers, Eur. Econ. Rev. 39 (1995) 859–887.
[8] J. Eaton, S. Kortum, International patenting and technology diffusion: theory and measurement, Int. Econ. Rev. 40 (1999) 537–570.
[9] G. Gong, W. Keller, Convergence and polarization in global income levels: a review of recent results on the role of international technology diffusion, Res.
Policy 32 (2003) 1055–1079.
[10] A.G.Z. Hu, A .B. Jaffe, Patent citation and international knowledge flow: the cases of Korea and Taiwan, Int. J. Ind. Organ. 21 (2003) 849–880.
[11] B. Xu, E.P. Chiang, Trade, patents, and international technology diffusion, J. Int. Trade Econ. Dev. 14 (1) (2005) 115–135.
[12] K.E. Maskus, Encouraging international technology transfer, International Centre for Trade and Sustainable Development, Geneva, Switzerland, 2004.
[13] A. Kraay, I. Soloaga, J. Tybout, Product quality, productive efficiency, and international technology diffusion: evidence from plant-level panel data, Paper
Presented at the NBER Summer Institute, July 2001.
[14] Z. Griliches, Issues in assessing the contribution of research and development to productivity growth, Bell J. Econ. 10 (1) (1979) 92–11 6.
[15] P.M. Romer, Endogenous technological change, J. Polit. Econ. 98 (1990) 71–102.
[16] P. David, Knowledge, property, and the systems dynamics on technological change, Proceedings of the World Bank Annual Conference of Development
Economics, World Bank, Washington, D.C., 1992, pp. 215–255.
[17] D. Quah, Technology dissemination and economic growth: some lessons for the new economy, in: C.E. Bai, C.W. Yuen (Eds.), Technology and the New
Economy, MIT Press, 2002, pp. 95–156.
[18] E. Bascavusoglu, Patterns of technology transfer to the developing countries: differentiating between embodied and disembodied knowledge, TEAM and CNRS
Working Papers, 2004.
[19] G. Papaconstantinou, N. Sakurai, A. Wyckoff, Domestic and international product-embodied R&D diffusion, Res. Policy 27 (1998) 301–314.
[20] J. Eaton, S. Kortum, Trade in capital goods, Eur. Econ. Rev. 45 (7) (2001) 1195–1235.
833H.-Y. Shih, T.-L.S. Chang / Technological Forecasting & Social Change 76 (2009) 821–834
[21] W. Keller, The geography and channels of diffusion at the world's technology frontier, National Bureau of Economic Research Working Paper, vol. 8150, 2001.
[22] D.T. Coe, E. Helpman, A.W. Hoffmaister, North-South R&D spillovers, Econ. J. 107 (1997) 134–14 9.
[23] B. Xu, J. Wang, Capital goods trade and R&D spillovers in the OECD, Can. J. Econ. 32 (1999) 1258–1274.
[24] B. Xu, J. Wang, Trade, FDI, and international technology diffusion, J. Econ. Integr. 15 (4) (2000) 585–601.
[25] A.B. Jaffe, M. Trajtenberg, R. Henderson, Geographic localization of knowledge spillovers as evidenced by patent citations, Q. J. Econ.108 (3) (1993) 577–598.
[26] Z. Griliches, Productivity and technological change: some measurement issues, OECD, Technology and Productivity: The Challenge for Economic Policy,1991,
pp. 229–231.
[27] N. Arranz, J.C. Fdez. de Arroyabe, Governance structures in R&D networks: an analysis in the European context, Technol. Forecast. Soc. Change 74 (5) (2007)
645–662.
[28] N. Arranz, J.C. Fdez. de Arroyabe, Joint R&D projects: experiences in the context of European technology policy, Technol. Forecast. Soc. Change 73 (7) (2006)
860–885.
[29] L. Knoben, L.A.G. Oerlemans, R.P.J.H. Rutten, Radical changes in inter-organizational network structures: the longitudinal gap, Technol. Forecast. Soc. Change
73 (4) (2006) 390–404.
[30] S. Samaddar, S. Nargundkar, M. Daley, Inter-organizational information sharing: the role of supply network configuration and partner goal congruence, Eur. J.
Oper. Res. 174 (2) (20 06) 744–765.
[31] Z. Simsek, M.H. Lubatkin, S.W. Floyd, Inter-firm networks and entrepreneurial behavior: a structural embeddedness perspective, J. Manage. 29 (3) (20 03)
427–442.
[32] R. Leoncini, S. Montresor, Accounting for core and extra-core relationships in technological systems: a methodological proposal, Res. Policy 34 (1) (2005)
83–100 .
[33] P.-L. Chang, H.-Y. Shih, Comparing patterns of intersectoral innovation diffusion in Taiwan and China: a network analysis, Technovation 25 (2) (2005)
155 –169.
[34] S. Wasserman, K. Faust, Social Network Analysis: Methods and Application, Cambridge University Press, Cambridge, 1994.
[35] A. Degenne, M. Forse, Introducing Social Networks, Sage Publications, London, 1999.
[36] L.C. Freeman, Centrality in social network: I. Conceptual clarificati on, Soc. Netw. 1 (1979) 215–239.
[37] L.C. Freeman, The gatekeeper, pair-dependency, and structural centrality, Qual. Quant. 14 (1980) 585–592.
[38] P.V. Marsden, Brokerage behavior in restricted exchange networks, in: P.V. Marsden, N. Lin (Eds.), Social Structure and Network Analysis, Sage Publications,
Beverly Hills, CA, 1982.
[39] F. Lorrain, H.C. White, Structural equivalence of individuals in a social network, J. Math. Sociol. 1 (1971) 49–80.
[40] D.A. Smith, D.R. White, Structure and dynamics of the global economy: network analysis of international trade, 1965-1980, Soc. Forces 70 (1992) 857–893.
[41] R. Breiger, S. Boorman, P. Arabie, An algorithm for clustering relation data with applications to social network analysis and comparison with multidimensional
scaling, J. Math. Psychol. 12 (1975) 328–383.
[42] D. Snyder, E. Kick, Structural position in the world system and economic growth: a multiple network analysis of transnational interactions, Am. J. Sociol. 84
(1979) 1096–1126.
[43] R. Nemeth, D.A. Smith, International trade and world-system structure: a multiple network analysis, Review 8 (4) (1985) 517–560.
[44] K. Reitz, D. White, Rethinking the role concept: homomorphisms on social networks, in: L.C. Freeman, D.R. White, A.K. Romney (Eds.), Research Methods in
Social Network Analysis, George Mason University Press, Fairfax, VA, 1989.
[45] T. Schweizer, Detecting positions in networks: a formal analysis of loose structure in rural Java, Am. Anthropol. 90 (4) (1988) 944–951.
[46] M.E. Porter, A. Lopez-Claros, K. Schwab, X. Sala-i-Martin, Global Competitiveness Report 2006-2007, Palgrave Macmillan, New York, 2006.
[47] S. Kim, E.-H. Shin, A longitudinal analysis of globalization and regionalization in international trade: a social network approach, Soc. Forces 81 (20 02)
445–468.
[48] B.H. Hall, A.B. Jaffe, T. Manuel, The NBER patent citations data file: lessons, insights and methodological tools, National Bureau of Economic Research Working
Paper, vol. 8498, 2001.
[49] S.P. Borgatti, M.G. Everett, L.C. Freeman, Ucinet for Window: Software for Social Network Analysis, Analytic Technologies, Harvard, 2002.
[50] World Bank, The East Asian Miracle: Economic Growth and Public Policy, Oxford University Press, New York, 1993.
[51] P. Krugman, The myth of Asia miracle, Foreign Aff. 73 (6) (1994) 62–78.
[52] Asian Development Bank, Asian Development Outlook 2003, Oxford University Press, New York, 2003.
[53] M. von Zedtwitz, O. Gassmann, Market versus technology drive in R&D internationalization: four different patterns of managing research and development,
Res. Policy 31 (2002) 569–588.
Hsin-Yu Shih is an associate professor at the Department of International Business Studies, National Chi Nan University of Taiwan. His previous articles have
appeared in the Technological Forecasting and Social Change,Psychology and Marketing,International Journal of Service Industry Management, Technovation, Journal of
e-Business. His current research interests are in the areas of Technology Management and Network Analysis.
Tung-Lung Steven Chang is currently Professor of Marketing and International Business at the College of Management, Long Island University - C.W. Post campus.
His research has been published by Journal of World Business, International Marketing Review, Journal of Global Marketing, International Finance Review and
Competitiveness Review. His has centered his research on the strategic marketing and technology aspects of multinationals' global expansion.
834 H.-Y. Shih, T.-L.S. Chang / Technological Forecasting & Social Change 76 (2009) 821–834