The political economy of R&D collaboration: micro- and macro-level implications

PDF

Download

Open document

Transcript (if available)

Content

THE POLITICAL ECONOMY OF R&D COLLABORATION:

MICRO- AND MACRO-LEVEL IMPLICATIONS

by

Matthew A. Shapiro

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(POLITICAL ECONOMY & PUBLIC POLICY)

August 2008

Copyright 2008 Matthew A. Shapiro
Dedication

This work is dedicated to my wife, Yoonshil, and children, Leo and Rachel, who are
among my best teachers.
ii

Acknowledgements

This project could not have been completed without the support of a number of
individuals and agencies. First and foremost, I am entirely indebted to Jeffrey Nugent
for his devotion to my research over the last several years. The comparative approach
extensively applied in this dissertation is a direct reflection of Jefferey Sellers’
instruction and suggestions. Early on in this project, Jim Moore provided first-hand
insight into the phenomena of university-based research in the U.S. along with a
number of other researchers at USC, UCLA, and Caltech. Based on these initial
findings, I met with Chris Moudling at USC’s Office of Technology Licensing, who
described the U.S.’s Association of University Technology Managers. Field research
on this subject was extensive and funded through the Urban Initiative Fellowship, the
Strategic Theme Fellowship, the Foreign Language and Area Studies Fellowship, ETRI,
and ITRI. In Korea, I wish to express my particular gratitude to Sang Sub Cho, Ki-Sik
Park, and Pil Sun Heo, of ETRI’s IT Strategy Research Group, and Chan-goo Yi, of
ETRI’s Intellectual Capital Team. The following individuals were all also extremely
supportive: Jang-Jae Lee and Jung-Jae Lee of KISTEP, Chang Hwa Woo and Bum Jun
Ko of ITEP, Chun-Kyung Park and In-Ho Kim of KOSEF, and Charles CY Park, and
Steve Yang of the IITA. Bong-Cheol Kang’s and Tae-Ho Song’s generosity and
logistical help must also be acknowledged. In Taiwan, I am particularly grateful to
Hubert Chen and, especially, Jian Hung Chen of ITRI’s Industrial Economics and
Knowledge Center, for their research support and help with logistics. Within each of
ITRI’s laboratories and departments, a short list of very helpful individuals includes
iii

Ming-Jinn Tsai, Jung-Pin Yu, Willy W.L. Chiang, Diana YP Chen, Peter Chen, Da-
Tung (DT) Liang, David Hsu, Mao Sheng Huang, Hui-Chung Ma, John Chih-Hung
Hsieh, and Benjamin Wang. Ken Hsieh, of Macronix, was very helpful in explaining
key features from the private perspective. At National Tsing Hua University, I wish to
thank Y.-C. Jack Chang, for his invitation and comments on my presentation. Also at
National Tsing Hua University, I am very grateful to Chintay Shih, for his discussion
on ITRI, university-based incubators, and Taiwan’s NIS as a whole. I have not listed
the names of the interviewees from the first-stage field research for this dissertation,
but would like to thank each and every one for their valuable time and patience. I have
also benefited extensively from the comments and criticism of conference panel
discussants, anonymous referees, and our mini-development seminar attendees in
USC’s Department of Economics. Key individuals include Werner Pascha, Thomas
Roediger-Schluga, Thomas Sattler, Jonathan Krieckhaus, Henry Etzkowitz, Tun-jen
Cheng, Joseph Wong, Christine Cooper, Dollie Davis, and Laura Armey. Uffe
Bergeton and Iris Bergeton provided excellent translation services for the Chinese
version of the questionnaire.

iv

Table of Contents

Dedication

ii
Acknowledgements

iii
List of Tables

vii
List of Figures

xi
Abbreviations

xiii
Abstract

xvi
An Introduction to the Topic
Overview 1
The Nature of R&D Collaboration 3
The Comparative Method 12
Fieldwork & the KORTAI R&D Dataset 34
Conclusion 54

Chapter 1: Information Flows and State-led Technological
Innovation in Korea and Taiwan

Introduction 56
Method & Data 70
Results 79
Conclusion

94
Chapter 2: The Triple Helix Paradigm in Korea and Taiwan: A Test
for New Forms of Capital

Introduction 103
Literature Review & Case Description 106
Research Questions & Data 109
Empirical Specification & Results 120
Conclusion

131
Chapter 3: A Cross-National Study of Governance and the Sources
of Innovation: The Determinants and Effects of International
R&D Collaboration

Introduction 137
Modeling International R&D Collaboration & Institutions 147
Empirical Specification 158
Results 165
Conclusion

187
v

Concluding Thoughts

191
Bibliography

197
Appendix 0.1 213
Appendix 1.1 214
Appendix 1.2 215
Appendix 1.3 216
Appendix 2.1 217
Appendix 2.2 218
Appendix 3.1 224
Appendix 3.2 225
Appendix 3.3 226
Appendix 3.4 227
Appendix 3.5 228
Appendix 3.6 229
Appendix 3.7 230
Appendix 3.8 231

vi

List of Tables

Table 0.1

Research types 9
Table 0.2

Survey response rate information 46
Table 1.1

Definitions of R&D success between sectors 58
Table 1.2

Descriptive statistics for dependent variables by sub-group:
Number of collaborative patents and number of collaborative
publications in 2005

77
Table 1.3a

Pairwise correlation coefficients by sub-group: number of
collaborative patents (2005) and methods of information
transfer

78
Table 1.3b

Pairwise correlation coefficients by sub-group: number of
collaborative publications (2005) and methods of
information transfer

78
Table 1.4

OLS results for impact of information transfers upon
collaborative patenting

84
Table 1.5

OLS results for impact of information transfers – factor
scores – upon collaborative patenting

85
Table 1.6

OLS results for impact of information transfers – factor
scores excluding patents and publications – upon
collaborative patenting

86
Table 1.7

OLS results for impact of information transfers upon
collaborative publications

87
Table 1.8

OLS results for impact of information transfers – factor
scores – upon collaborative publications

88
Table 1.9

OLS results for impact of information transfers – factor
scores excluding patents and publications – upon
collaborative publications

89
Table 1.10

OLS results for impact of information transfers (fa_trans)

92
Table 1.11

OLS results for impact of information transfers (fa_trans),
full model

93
vii

Table 1.12

Determinants of information transfers: inclusion of OTL
formation dummy

96
Table 2.1a

Collaborative tendencies: aggregate level, Korea

111
Table 2.1b

Collaborative tendencies: aggregate level, Taiwan

111
Table 2.2a

Collaborative Tendencies: sub-group level, Korea 112
Table 2.2b

Collaborative Tendencies: sub-group level, Taiwan 112
Table 2.3

Average number of patents through cross-sector R&D
collaboration: by sub-sector: Korea

117
Table 2.4

Average number of patents through cross-sector R&D
collaboration: by sub-sector: Taiwan

118
Table 2.5

Average number of total patents excluding cross-sector R&D
collaboration: by sub-sector, Korea

119
Table 2.6

Average number of total patents excluding cross-sector R&D
collaboration: by sub-sector, Taiwan

120
Table 2.7

OLS results for new and pre-existing capital’s effects upon
collaborative output

123
Table 2.8

OLS results for new and pre-existing capital’s effects upon
non-collaborative output

125
Table 2.9

OLS results for new and pre-existing capital’s effects upon
collaborative and non-collaborative output, by sector sub-
groups

126
Table 2.10

OLS results for new and pre-existing capital’s effects upon
collaborative and non-collaborative output, by sector sub-
groups

127
Table 2.11

Rankings of weighted reasons for repartnering: Korea 129
Table 2.12

Rankings of weighted reasons for repartnering: Taiwan 130
Table 2.13

Rankings of weighted source of personal ties: Korea 131
Table 2.14

Rankings of weighted source of personal ties: Taiwan 131
viii

Table 3.1

Modes of international technology transfer 143
Table 3.2

Descriptive statistics for IPR by patenting group 155
Table 3.3

The changing composition of Tier-1 countries 160
Table 3.4

Testing for patenting effects on TFP, linear and nonlinear,
accounting for country-specific time trends and percentage
of all patenting done in collaboration

170
Table 3.5

Testing for Tier-1 collaboration effects on TFP, linear and
nonlinear, accounting for country-specific time trends, per
capita total patents, and the percentage of all patenting done
in collaboration

171
Table 3.6

Testing for patenting effects on TFP, linear and nonlinear,
accounting for country-specific time trends and percentage
of all patenting done in collaboration

172
Table 3.7

Testing for exclusive patenting effects on TFP, linear and
nonlinear, accounting for country-specific time trends and
percentage of all patenting done in collaboration

173
Table 3.8

Testing for exclusive patenting effects on TFP, linear and
nonlinear, accounting for country-specific time trends,
percentage of all patenting done in collaboration, and patent-
income interactions

174
Table 3.9

Testing for institutional effects on the log of Tier-1
collaboration, accounting for possible interactions between
IPRs and democracy, country-specific time trends

180
Table 3.10

Testing for institutional effects on the log of Tier-1
collaboration, accounting for possible interactions between
IPRs and checks-and-balances, country-specific time trends

181
Table 3.11

Testing for institutional effects on the percentage of total
patents represented by Tier-1 collaboration, accounting for
possible interactions between IPRs and democracy, country-
specific time trends

182
ix

Table 3.12

Testing for institutional effects on the percentage of total
patents represented by Tier-1 collaboration, accounting for
possible interactions between IPRs and checks-and-balances,
country-specific time trends

183
Table 3.13 Testing for institutional effects on the log of non-Tier-1
collaborative patents, accounting for possible interactions
between IPRs and democracy, country-specific time trends
184

Table 3.14

Testing for institutional effects on the log of non-Tier-1
collaborative patents, accounting for possible interactions
between IPRs and checks-and-balances, country-specific
time trends
185

Table 3.15

Testing for effects of overall patents, collaborative patents,
and non-collaborative patents upon TFP, instrumenting for
IPR-POLCON interaction term

186

x

List of Figures

Fig. 0.1

Patenting growth rates 18
Fig. 0.2

Difference in patenting growth rates: 1981-1993 to 1994-
2006

19
Fig. 0.3

Share of worldwide patenting, 1986 (inner ring) and 2006
(outer ring)

20
Fig. 0.4

Relationship between R&D expenditure and total researchers 21
Fig. 0.5

Distribution of researchers by country: 2004 22
Fig. 0.6

Technology balance of payments: 2004 24
Fig. 0.7

Connections between ministries and research entities, Korea 43
Fig. 0.8

Connections between ministries and research entities,
Taiwan

44

Fig. 0.9

Research emphases 48
Fig. 0.10

Industrial affiliation 50
Fig. 0.11

Defining success 52
Fig. 0.12

Respondent information 53
Fig. 1.1

Research emphases by sector and country 61
Fig. 1.2

Returns to R&D and information channels without
collaboration

66
Fig. 1.3

Returns to R&D and information channels with public-
private collaboration

69
Fig. 1.4

Number of patents and publications: Korea and Taiwan
(1981-2006)

90
Fig. 1.5

Duration of OTL: Korea (row 1), Taiwan (row 2), public
(column 1), private (column 2)

95
Fig. 1.6

Percentage of GERD funding by industry and government 98
xi

Fig. 1.7a

Korea: BERD financed by government: percentage and in
2000 constant dollars

100
Fig. 1.7b

Taiwan: BERD financed by government: percentage and in
2000 constant dollars

100
Fig. 1.8

Share of GOVERD plus HERD represented by industry:
constant 2000 dollars

101
Fig. 1.9

GERD per researcher: constant 2000 dollars 101
Fig. 2.1

Tracing the effects of personal ties and repartnering 116
Fig. 2.2

Average number of patents through cross-sector R&D
collaboration: by sub-sector: Korea

117
Fig. 2.3

Average number of patents through cross-sector R&D
collaboration: by sub-sector: Taiwan

118
Fig. 2.4

Average number of total patents excluding cross-sector R&D
collaboration: by sub-sector, Korea

119
Fig. 2.5

Average number of total patents excluding cross-sector R&D
collaboration: by sub-sector, Taiwan

120
Fig. 3.1

Total patents and the percentage represented by Tier-1
countries

140
Fig. 3.2

The Hall and Jones (1999) and international R&D-specific
institutional analyses

146
Fig. 3.3

Distribution of patenting between Tier-1 and non-Tier-1
countries

161
Fig. 3.4

Number of collaborative patents 162
Fig. 3.5

Percentage of total patents represented by collaborative
patents

162
Fig. 3.6 The relationship between Tier-1-all patent ratio and GDP per
capita
175

xii

Abbreviations

ATP

Advanced Technology Program
AUTM

Association of University Technology Managers
CoE

Center of Excellence
CRDC

Center for R&D Commercialization
DoIT

Department of Industrial Technology
ERSO

Electronics Research and Services Organization
ESPRIT

European Strategic Program for Research and Development in
Information Technology

ETRI

Electronics & Telecommunications Research Institute
FDI

Foreign direct investment
GLS

Generalized least squares
GRI

Government research institute
IARCRC

Industry-Academy-Research Institute Cooperative Research Center
IC

Integrated circuit
IDB

Industrial Development Bureau
IITA

Institute of Information Technology Assessment
IPR

Intellectual property rights
ITEP

Institute of Industrial Technology & Evaluation Planning
ITRC

Information Technology Research Center
ITRI

Industrial Technology Research Institute
KIET

Korea Institute for Electronics Technology
KIST

Korea Institute of Science and Technology
xiii

KOSEF

Korea Science & Engineering Foundation
MIC

Ministry of Information & Communication
MOCIE

Ministry of Commerce, Industry and Energy
MOEA

Ministry of Economic Affairs

MOST

Ministry of Science & Technology
MSTI

Main Science and Technology Indicators
NIE

Newly industrialized economy
NIS

National innovation system
NSC

National Science Council
OECD

Organization for Economic Cooperation and Development
OLS

Ordinary least squares
OTL

Office of technology licensing
PCB

Printed circuit board
PNGV

Partnership for a New Generation of Vehicles
R&D

Research and development
RJV

Research joint venture
S&T

Science & technology
SBIR

Small Business Innovation Research Program
SEMATECH

Semiconductor Manufacturing Technology Program
SME

Small and medium enterprise
STEPI

Science & Technology Policy Institute
TDP

Technology Development Program
xiv

TDPA

Technology Development Programs with Academia
TFP

Total factor productivity
TTO

Technology transfer organization
USITC

United States International Trade Commission
USPTO

United States Patent & Trademark Office
VLSI

Very Large Scale Integration Program
WDI

World Development Indicators

xv

Abstract

This dissertation speaks to three separate but related issues on R&D collaboration. The
first chapter details the various ways in which entities from the public and private
research sectors transfer information. In Korea and Taiwan, information transfers are
shown to have a positive impact on public-private R&D output in both countries, but
government funding is a much stronger predictor of such transfers in Taiwan. The
second chapter focuses on the research-based links between the government,
universities, and firms, and tests the Triple Helix-based hypothesis that new capital
arises from the interactions between the public and private research sectors. Focusing
again on Korea and Taiwan, repartnering tendencies (as a proxy for new capital) are
found to be strong predictors of R&D output, with virtually no differences between
these two countries. In the third chapter, attention is drawn to R&D collaboration
between countries, and it is shown that international R&D collaboration has a positive
impact on the growth residual (TFP), and that it has a bonus effect to a country’s
patenting efforts. It is also shown that developing countries derive particular benefit
from international R&D collaboration, but such benefits are largely dependent on the
presence of strong political institutions and intellectual property rights. In terms of data
and methods, Chapters 1 and 2 utilize a unique dataset (KORTAI R&D), which is
specific to the Korean and Taiwanese cases, while the macro-level analysis of Chapter
3 uses data for a maximum of 150 countries drawn from the USPTO, WDI, Penn
World Table, Barro-Lee education data, the Ginarte-Park IPR index, POLITY IV, and

xvi

POLCON. OLS and ordered logit statistical methods are applied in Chapters 1 and 2,
and fixed effects GLS methods are used in Chapter 3.

xvii
An Introduction to the Topic
Overview
Now, more than ever before, research and development (R&D) efforts are
oriented toward R&D collaboration between different research entities. Public funding
plays a role in such collaboration and three separate analyses are offered in this
dissertation to describe the varied nature and effects of R&D collaboration, given the
general view that it is beneficial, despite coordination problems and potential risks.
Government-led instances of public-private R&D collaboration are a relatively
recent phenomena, restricted to the world’s most developed countries. Japan was the
first country to deliberately set up the infrastructure for manufacturing, trade, and
finance in R&D, presenting the first real case of public-private R&D collaboration in
the Very Large Scale Integration Program (VLSI). The U.S. and European countries
later followed Japan’s lead in its R&D processes and specific R&D programs and
structures, ultimately creating public-private R&D collaboration programs such as the
ATP Program (U.S.), Silicon Valley, the Small Business Innovation Research Prgoram
(SBIR) Program (U.S.), the Avery Project (U.K.), ESPRIT (Europe), EUREKA
(Europe), and SEMATECH (U.S.). Given the high levels of research capabilities in
these countries, public-private R&D collaboration may be viewed as an indicator of
advanced development. At the same time, one can attribute public-private R&D
collaboration to state intervention, which is consistent with Japan’s model of an
interventionist state bureaucracy. It should come at no surprise, thus, that public-private
R&D collaboration is now well instituted in the two countries which have both made

1
the leap to “highly industrialized” status and closely followed the Japanese
developmental state model: Korea and Taiwan.
The government’s involvement in fostering public-private R&D collaboration
raises an immediate concern regarding the efficacy of such policies and a question of
whether or not market-failure exists. Here, in this introductory chapter, the
determinants of market-failure are presented, while a test for the effectiveness of
government policies calling for public-private R&D collaboration is offered in Chapter
1, in the context of the Korean and Taiwanese cases. Beyond developmental statism,
there are many other factors to consider, especially the nature of the relationship
between the public and private sectors. Chapter 2 extends this to the three-way
interaction of the Triple Helix paradigm, which focuses particularly on the research
interactions among government research institutes (GRIs), universities, and private
firms. With continued focus on Korea and Taiwan, these interactions are quite complex,
given traditional networking patterns. If family and personal ties are more heavily
weighted than legitimate partnership qualifications, they would also pose a challenge to
the efficacy of policies calling for public-private R&D collaboration, Chapter 3
attempts to complete the progression from the two-country, micro-level comparative
analysis to the macro-level. Korea and Taiwan, however, continue to stand out here,
given their phenomenal increase in patenting over the last thirty years.
A call is also made in this introductory chapter for the revival of discussion on
the East Asian developmental state. Government intervention in Korea and Taiwan is
no longer focused solely on industrialization, for its function is now to maintain an

2
innovation-based developmental state. Because of the large number of shared
characteristics between Korea and Taiwan, an R&D-specific comparison will reveal
that this modified developmental state is now the standard.
Since the data used in this dissertation are largely unique, the fieldwork and
questionnaire that give rise to this data are described in this introductory chapter. Also
presented in this introductory chapter is a look at Korea and Taiwan in a global context,
using the OECD’s Main Science and Technology Indicators (MSTI) database. This
serves as background for the macro-level approach to international R&D collaboration
presented in Chapter 3 and verifies that Korea and Taiwan are now among the world’s
technology leaders. As the “East Asian Miracle” was once the model of
industrialization for developing countries, the innovation-based successes of Korea and
Taiwan is a policymaking blueprint which can be considered by nations with a
sufficient level of R&D infrastructure.

The Nature of R&D Collaboration
There is an immense literature on the subject of R&D collaboration.
Researchers from various disciplines including economics, public policy, industrial
organization, and business management have analyzed this phenomenon. In much of
this analysis, firms are identified as the primary unit of analysis. Patel and Pavitt (2000)
conclude that the empirical evidence shows that there is a positive effect upon each

3
country’s national innovation system from the linkages between firms and academia.
1

This explains in OECD countries the correlation between high quality basic research
and high levels of technology overall. Where knowledge is transmitted through
codified information (scientific reports and publications) or personal interaction, there
are a number of channels for knowledge spillovers.
2
Knowledge creation occurs in the
context of a fluid and evolving community. This social construction process leads
Powell, et al. (1996) to conclude that R&D-type activities in a single formal
organization may be a poor method for learning. Rather, sources of innovation can be
found where firms, universities, and GRIs intersect.
3
Innovation practices are largely assigned in the literature along public-private
lines, with the private firm focusing on applied technology while the public research
entity does basic research. Of course, the firm can invest in R&D to create an
environment entirely similar to a university laboratory where researchers work on their
own projects, publish their findings, interact with scientists in academia and elsewhere,
and host postdoctoral fellows (Powell et al., 1996). Conversely, GRI and university
institutes can develop manufacturing and production capabilities to commercialize
innovations. These methods on both public and private sides are not the norm, however,

1
Patel and Pavitt (2000) point to linguistic and geographic constraints which are enforced from person-
to-person exchanges and transfers of tacit (i.e., basic) knowledge.

2
Such channels include: education of students; carrying out of contract research or innovation-related
services such as testing, consulting, and training of personnel; joint R&D projects of research institutions
and private firms; and informal exchanges of know-how Fritsch and Schwirten (1999).

3
Powell, et al. (1996) also point out that customers are also important here, in addition to firms,
universities, and GRIs. Because we make the assumption that firms are closely familiar with the
demands of the market and prospects for commercializing R&D results, this “customer effect” can be
bundled within firms.

4
which makes for a rich dynamic when collaboration between public and private
researchers occurs.
4
Public and private research entities are not mutually exclusive in
orientation, but their efforts are largely complementary.
Promoting R&D collaboration is not without limits, given that firms have a
disincentive to share R&D results and possibly lose the rights to proprietary knowledge.
When results from R&D efforts are uncertain and difficult to appropriate, R&D
collaboration is not preferred (Pisano, 1990; Oxley, 1997; Cohen, 1994). Nevertheless,
there are clear benefits to public-private R&D collaboration,
5
particularly through
knowledge spillovers, reducing duplication, and exploiting scale economies in R&D.
Such collaboration has also been found to accelerate commercialization (Mowery,
1998). Powell, et al. (1996) have isolated four factors lying behind the upsurge in R&D
collaboration, although these are largely confined to coordination strictly in the private
research sector: risk sharing, obtaining access to new markets and technologies,
speeding products to market, and pooling complementary skills.

Market failure concerns
There appears to be strong empirical support for government subsidization of
R&D efforts. In particular, the private sector will not conduct R&D at levels which

4
This may also be related to available resources. Consider Fritsch and Schwirten (1999), who found that
universities supported production innovations much more often than process innovations, while private
firms were significantly more often involved in process innovations. This could be due to the limited
capacities of those universities that prevent support for complex production processes because of limited
funding and available machinery.

5
Consult Dodgson (1993) for complete details.

5
meet the social optimum (Jones and Williams, 2000). R&D support policies, however,
must not serve as a substitute for private investment in R&D, but should encourage
private investment. Branscomb and Keller (1998), thus, favor the use of market
mechanisms, such as tax incentives, but also acknowledge that direct investments in
research are required by the government, given that private firms are likely to
underinvest in the kinds of research which might satisfy public needs. The government
can also cap public funding to government institutes to incentivize the search for
corporate sponsorship.
6
The notion of market failure is traditionally rooted in economic theory, although
it has close parallels to the interdisciplinary fields of political economy and state-led
growth. For the purposes of this dissertation, an analysis of public R&D and the
dynamics between public-private R&D efforts must look beyond neoclassical
economic theory’s inability to address particular “systemic aspects” (Lall, 2000).
Specifically, the perfect competition paradigm avoids dealing with various kinds of
externalities such as ways of doing business and supporting institutions. The
evolutionary approach, on the other hand, is not limited by restrictive assumptions and
recognizes that many requirements of learning may involve serious market failure. In
other words, investments in innovation and R&D may require policies to overcome
these market failures, such as those involved in tackling learning costs, promoting
linkages, coordinating factor market improvements with technological needs, and

6
Additional effects include improved pay and recognition for GRI-based researchers. Dahlman and
Andersson (2000) emphasize these points with regard to the Korean case.

6
develop institutions. In large part, the findings of Chapter 3 are consistent with this
approach, as it is shown that the growth residual (TFP) is a positive function of
patenting, and that successful patenting is a positive function of strong political
institutions and intellectual property rights (IPRs).
This dissertation is ultimately related to the performance of government,
although its direct impacts are limited to innovation, S&T infrastructure, and
international R&D collaboration. This is the specific added-value of this project, as we
build upon previous research which directly measures government performance (La
Porta, et al., 1999) and address the hazards and benefits of government intervention.
7

Ultimately, when there is balance in allocating costs, there is balance in the flow of
information. These balances, Branscomb and Keller (1998) claim, will enable a result
which can be shared by the private and public sectors in proportion to each party’s
interest, investment, and willingness to share results with others.

Typology of public-private R&D collaboration
There is no single source of R&D; different actors engage in different types of
research, including universities, government research institutes (GRIs), and firms. For
the purposes of this dissertation, public R&D includes that originating from GRIs and
institutions of higher education. GRI-based R&D is specifically that of all entities of
government which furnish but do not sell services, other than higher education, which

7
This literature is represented by Pack (2000).

7
cannot otherwise be conveniently and economically provided. R&D of organizations of
higher education includes that of all universities, colleges of technology and other
institutions of post-secondary education, whatever their source of finance or legal status.
Private R&D is produced by firms, organizations, and institutions whose primary
activity is the market production of goods and services for sale at an economically
significant price. These classifications closely follow those offered by the Frascati
Manual (OECD 2002).
8
Definitions of public-private R&D collaboration are not limited to one single
structure, although the focus here is firm-GRI and firm-university collaboration.
9

Accompanying this focus is a series of somewhat strict, albeit realistic, assumptions
about the alignment between R&D types and sources, the former of which is presented
in Table 0.1. First, public R&D is the source of basic research, particularly university-
based research but also GRI-based research.
10
Second, applied and developmental
research is done primarily in the private sector,
11
and it is also assumed that private
efforts at basic R&D are largely done by firms which are of sufficient size to partition

8
A recent update is offered in OECD (2004), which discusses policies related to public-sector research,
government support for private-sector R&D and innovation, and collaboration and networking among
innovating organizations. Also discussed is public-private R&D collaboration for innovation and the
optimal implementation of public-private partnership programs.

9
It should be noted that these two collaboration structures are not exclusive of one another. I.e.,
collaborative projects between the private sector and both types of public-sector research entities (GRI
and university) are included under the “public-private R&D collaboration” classification.

10
This is true in spite of the fact that the historical role of GRIs to engage in basic research has been
considerably modified and now includes applied research as a goal. Mansfield (1972) pointed this out for
the U.S. case over three decades ago, but it also reflects recent changes in Korea.

11
See Table 0.1 for elaboration on these research types.

8
their R&D departments.
12
Smaller firms, such as those under analysis in the following
discussion, attempt to create niche innovations for ease of commercialization.
Table 0.1 Research types
Basic research Experimental or theoretical work undertaken primarily to acquire new
knowledge of the underlying foundations of phenomena and observable facts,
without any particular application or use in view.
Applied research Original investigation undertaken in order to acquire new knowledge. It is
directed primarily towards a specific practical aim or objective.
Developmental
research
13
Systematic work, drawing on knowledge gained from research and practical
experience, that is directed to: (1) produce new materials, products and devices,
(2) install new processes systems and services, or (3) improve substantially
those already produced or installed.
Source: OECD (2002).

In the context of these different research types and the list of possible benefits
from public-private R&D collaboration, there are several points of concern. The
government, for example, has a preference for diffused knowledge while the private
firm strives to maintain control over R&D results. Some, such as Branscomb and
Keller (1998) claim that this will more easily bring research results to society while
costs are shared between research entities.
14
As people move, interact, and share
information between firms, research becomes less proprietary and more long-range.
15

12
This point counters Holmstrom (1989), who points out that small firms engage in innovative projects,
but that the high cost of managing such projects for large firms prompts a lower incidence.

13
The Frascati Manual uses the term “experimental development” for this third classification of R&D.

14
In Branscomb and Keller’s (1998) study, collaboration between private firms is the sole unit of
analysis, but the conclusions are still relevant for this discussion.

15
Importantly, Branscomb and Keller (1994) state that the ease of diffusion within firm-based research
networks is accompanied by less concern that there will be a non-market-based disruption of the market.
Such disruptions are actually a source of great concern for those who support limited government
intervention to prevent market failure in R&D efforts.

9
Relevant studies and themes in the literature
The increase and facilitation of knowledge generation is an underlying theme of
technology development. Often discussed in terms of the direction in which
technological knowledge flows within or between countries, the “linearity hypothesis”
states that information flows from upstream research entities to downstream research
entities in a uni-directional fashion Tassey (1997). This issue is related to the content of
Chapter 1, which focuses expressly on information flows between the public and
private research sectors. A gross challenge to the linearity hypothesis is offered in
Chapter 2, which discusses multiple linkages among various research entities,
regardless of whether they are in the public or private research sectors. As such, this
dissertation can offer a direct challenge to the linearity hypothesis, as information flows
bi-directionally in the context of public-private R&D collaboration.
Another area of research related to public-private R&D collaboration is that
concerning its potential welfare benefits. As shown in Boorstin’s (1983) description of
how the market for clocks developed, innovation did not occur for its own sake. While
economic motives may be important in innovation, there are other instances in which
innovations and discoveries take place without general concern for economic gain, such
as the case of Newton (Easterly, 2001). The issue of public versus private benefits of
innovation and technological progress is fundamental to this discussion. Increased rates
of return for R&D efforts are, after all, a desired result for all involved in public-private
R&D collaboration. Such desired outcomes, however, are not always realized.

10
Dasgupta and Maskin (1987) emphasize the public nature of scientific and
technological knowledge. That is, basic research-oriented efforts largely generate R&D
results for the general public, which contrasts with efforts by private firms to increase
the flow of rents from applied research efforts.
16
The authors continue to say that this is
just as true for basic research efforts (which the authors describe as “pure sciences”) as
it is for applied research. In the literature, there is no explicit challenge to the public
goods aspect of R&D. Yet, there is an overwhelming emphasis (in studies rooted in
industrial organization and business management theory) on how basic research and
public funded research can lead to increases in private sector research. This is a worthy
area of study,
17
but the results largely hinge on the uni-directional flow of research
consistent with Tassey’s linearity hypothesis.
There is also a large body of research aiming to measure these effects. Four
studies worthy of mention have ties to one or more of the chapters in this dissertation.
Griliches, et al. (2000) quantify the effects of government research funding programs
on firm productivity, using the Cobb-Douglas production function, and conclude that,
for the Israeli case, government support leads to greater firm-based R&D productivity.
Cohen and Walsh (2000) test for the effects of appropriability and knowledge flows on
innovation and find that information flows complement ATP program participants’

16
The nature of the profits which are to be reaped from research efforts do not necessarily have to be
financial in nature. The university or GRI can experience considerable benefit from reputation effects
through publications or symposia presentations. While these points seem to confirm the complementary
nature of the public and private research sectors, it is still an indicator of the need for market-failure
corrections.

17
See Sakakibara and Branstetter (2003), Griliches, et al. (2000), and Mansfield (1996), for example.

11
own R&D efforts. Hall, et al. (2000) use multivariate regression analysis (probit and
tobit estimators) to study the effects of universities in public-private R&D collaboration,
again looking at the U.S. ATP program. They conclude that firms are challenged in
assimilating the basic knowledge of university partners, although prior experience
working with a university partner alleviates this difficulty. Finally, Kogut and
Gittelman (2001) focus on biotechnology firms in the U.S. to determine the effects of
public-private R&D collaboration upon firm research output. Through negative
binomial regression estimations, Kogut and Gittelman make a number of interesting
findings which speak specifically to the biotechnology field: firms publish more when
they are research-intensive, the benefits of collaboration are greater for firms with weak
research capabilities, and commercialization is a function of public-private R&D
collaboration.

The Comparative Method
The practice employed in the three subsequent chapters of this dissertation
supplements statistical analyses with qualitative discussion. Through this extension, we
can offer an interpretation which remains equally valid and is in fact richer.
18

18
In a systematic, cross-sectional analysis of cases of R&D collaboration, one may choose to emphasize
industry and firm characteristics of R&D collaborators. Riccaboni, et al. (2003), for example, engage in a
comparative study on the innovation system in the life sciences between the U.S. and Europe. Particular
emphasis is given to the linkages between research universities, public research institutes, and the private
sectors, and Riccaboni, et al. (2003) conclude that network relations in the United States are concentrated
in regional clusters with considerable interactions across organizations and disciplines; European
networks, on the other hand, are less dense with smaller numbers of actors in highly specialized
communities. Riccaboni, et al. (2003) suggest that European policy should not attempt to copy American
policy, but should focus on the fostering of new networks and interactions between small firms and
universities. Alternatively, one may compare instances of cooperative R&D, such as Aldrich and
Sasaki’s (1995) analysis of R&D consortia in the U.S. and Japan.

12
Conclusions offered in this dissertation are therefore drawn from (1) interviews in the
field, (2) an extensive theoretically- and empirically-driven literature, (3) qualitative
measures of the KORTAI R&D dataset, and (4) the OECD MSTI dataset, in the case of
the macro-level study in Chapter 3.
19
What is ultimately being compared are the
dynamics of R&D collaboration in Korea and Taiwan, necessitating a discussion of
these two countries’ relevant S&T policies.

Case specifics - general
Korea and Taiwan are now implementing cross-sector R&D collaboration at the
policy level, but this set of policies is a reflection of several decades of technology-
based growth. It should be noted at the outset that there is an unmistakable sense that
these two countries will continue to initiate policies which advance their technological
capabilities and output. Indeed, this is the connection between the comparative studies
of Chapters 1 and 2 and the cross-national study which follows. Korea and Taiwan’s
technological output (measured by patents and publications) has enabled both to be
classified as “Tier-1” technology countries, moving from serious trailing positions to a
shared leadership position.
In line with the comparative method, our first task is to consider – in the context
of public-private R&D collaboration – the differences and similarities between Korea
and Taiwan. Kuznets’ (1988) East Asian development model, based on Korea, Taiwan,

19
Strong reference is made here to the work of George and Bennett (2005) and their claim that the case
study and statistical approaches are complementary.

13
and Japan, identifies no real distinctions between the former two cases. Investment
ratios, public spending, and export expansion in these two countries were not very
different from one another. In addition, both practiced substantial government
intervention, enabling the adoption of long-run policies with a much more lasting
impact than short-run, reactive policies (Kuznets, 1988). For Park (1990), differences
in government intervention in Korea and Taiwan are more subtle. The Korean
government has been “interventionist,” while the Taiwanese government has been
“supportive,” which refers to evidence in Korea of domestic market protection and
industrial targeting while, in Taiwan, medium-term economic plans do not provide
policymakers with the authority to allocate credit.
What is clear, between these two countries, is that R&D has been a key focus in
the industrial development process. Pack (2000) specifically identifies the abilities of
countries such as Korea and Taiwan to tap the backlog of technology and efficiently
absorb it. This was done through movement towards international best practice via
openness to foreign knowledge, pressure to increase exports rather than extract rents
from the domestic economy, and the presence of an educated domestic labor force.
Pack (2000), however, claims that domestic innovation efforts are beneficial, while
attempts to create original R&D detract from attempts to improve imported processes
and products, which can discourage local capabilities and their long run benefits. Given
that public-private R&D collaboration is one of the key structures facilitating the
creation of original R&D, Pack’s (2000) concerns must be reevaluated in light of the

14
national S&T plans abundant now in both Korea and Taiwan, not to mention the
shifting pattern throughout developing countries.
In terms of technological progress, both Korea and Taiwan share the
characteristic of creative imitation, which distinguishes it from other NIEs (e.g.,
Thailand, Malaysia, Indonesia, Vietnam, and the Philippines). As Lee, et al. (1988)
state, the latter group undertakes duplicative imitation of foreign products which are at
the mature technology stage with inexpensive labor. In Singapore, capabilities have
developed, but mainly in operating imported technologies, so there is relatively little
design and development activity. In contrast, high-tech exports from Korea and Taiwan
have significant local linkages and far more technological input up to the basic design
stages (Lall, 2000). The continuing emphasis on R&D activities in Korea and Taiwan,
including strong efforts to initiate public-private R&D collaboration, is consistent with
these observed patterns.
With data from the U.S. Patent Patent and Trademark Office (USPTO) database
and the OECD Main Science and Technology Indicators (MSTI) dataset,
20
a
descriptive analysis of patenting (growth rates and the global share), R&D expenditures,
distribution of researchers across sectors, and the technology balance of payments
provides evidence of the transition to first-tier status in both Korea and Taiwan. From
the mid-1970s to the present, these two countries are the only two countries to have

20
An alternative and arguably more balanced source of patent data are Triadic Patents, which account for
those patents which have been simultaneously applied for and granted by the world’s three largest
patenting bodies: the USPTO, the European Patent Office, and the Japan Patent and Trademark Office.
This data is considerably more limited than strictly USPTO data, both in terms of quantity as well as
scale (i.e., over time).

15
risen from relatively no patent generation to being among the top five patenting
countries in the world. Fig. 0.1 describes the scale of the growth rate of patenting in the
two periods, 1981-1993 and 1994-2006, where each bar is the average rate of patenting
over its respective time period. In the first period, the patenting growth rate was no
more than 2 and 6 percent in Korea and Taiwan, respectively, and no more than 16 and
23 percent in the second period. Although such growth might not be considered very
substantial, Korea and Taiwan are in fact among only a handful of countries which
experienced positive growth in patents between these two periods. Indeed, as shown in
Fig. 0.2, Korea and Taiwan share this characteristic with only seven other countries,
based on the OECD (2006) data.
21
Ultimately, Korea and Taiwan seem to reflect the
strong patent orientation of the world’s technological leaders.
In terms of the shares of worldwide patenting, presented with the annual values
for 1981 and 2006 in Fig. 0.3, the United States, Japan, and Germany have clearly
dominated in terms of total output. France, Canada, and the United Kingdom – the
latter particularly in 2006 – also have a presence. From 1981 to 2006, Korea and
Taiwan are the most notable cases, representing less than two-hundredths of a percent
and two-tenths of a percent of total worldwide patenting in 1981, respectively, but 3.57

21
Japan and the United Kingdom are considered Tier-1 patenting countries. (“Tier-1” classification
denotes the top five R&D output generating countries. This category is discussed in much more detail in
Chapter 3 of this dissertation.) France, another Tier-1 patenting country, has virtually no difference in
patenting growth over these two periods, while Germany and the United States, two other Tier-1
countries, show decreased growth. This change for the United States is difficult to interpret here, given
the country’s overall scale of patenting. For the Germany case, we attribute this downward shift to the re-
unification of East and West Germany.

16

17
and 4.43 percent of total worldwide patenting in 2006. None of the other top-end
patenting countries experienced such a sizeable change.

Fig. 0.1 Patenting growth rates
Patenting growth rates
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
AU AT BE CA DK FI FR DE GR HU IS IE IT JP KR LU MX NL NZ NO PT ES SE CH TR GB US AR CN IL RU SGSI ZA TW
1981-1993 1994-2006

Source: OECD (2006).
Note: See Appendix 0.1 for country abbreviation details. Iceland and Russia’s patent data is available only for the latter period (194-2006).

18

Fig. 0.2 Difference in patenting growth rates: 1981-1993 to 1994-2006
Difference in patenting growth rates: 1981-1993 to 1994-2006
AU
AT
BE
CA
DK
FI
FR
DE
GR
HU
IS
IE
IT
JP
KR
LU
MX
NL
NZ
NO
PT
ES
SE
CH
TR
GB
US
AR
CN
IL
RU
SG
SI
ZA
TW
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4

Source: OECD (2006).
Note: See Appendix 0.1 for country abbreviation details. Iceland and Russia’s patent data is available only for the latter period (194-2006), so the
difference could not be determined.

19
Fig. 0.3 Share of worldwide patenting, 1981 (inner ring) and 2006 (outer ring)
CA
FR
DE
JP
KR
GB
US
TW
CA
FR
DE
JP
KR
GB
US
TW

Source: OECD (2006).
Note: See Appendix 0.1 for country abbreviation details.

While Korea and Taiwan remain relatively average in terms of the relationship
between R&D expenditures and total researchers, over the 1981 to 2006 period (Fig.
0.4), there is a much stronger presence of private researchers in these two countries vis-
à-vis the public research sector (the sum of government and university researchers) (Fig.
0.5). Figs. 0.1 to 0.5 are all significant pieces of evidence, showing close alignment
among Japan, Korea, and Taiwan. This offers a challenge to the literature which aligns
Japan with the coordinated market economies of Western Europe and only draws loose
parallels between Japan and both Korea and Taiwan. Hall and Soskice’s (2001)
varieties of capitalism approach,
22
in particular, does not treat the East Asian system
fully. These boundaries can be redrawn, in the context of this new evidence.

22
This approach focuses on differences in economic and political institutions as a comparative basis,
accounting for whether the economy is market-led or “coordinated” through non-market relationships.

20
Fig. 0.4 Relationship between R&D expenditure and total researchers
AU
AT
BE
CA
CZ
DK
FI
FR
DE
GR
HU
IE IT
JP
KR
LU
MX
NL
NZ
NO
PL
PT
ES
SE
CH
TR
GB
US
AR
CN RO
RU
SG
SI
ZA
TW
-250 0 250 500 750 1000
Average per capita R&D expenditure (current PPP dollars)
0 2 4 6 8 10 12
Average total researchers per thousand employed
95% CI Fitted values
Average GE RD per capita
Source: OE CD M STI 2006
Relationship between R&D expenditure and total researchers: 1981-2006

Source: OECD (2006).

21
Fig. 0.5 Distribution of researchers by country: 2004
Distribution of researchers: 2004
0% 20% 40% 60% 80% 100%
TR
AR
PL
GR
PT
HU
NZ
ES
MX
IT
CH
SI
RO
CZ
IE
NO
BE
DE
FR
AT
FI
NL
DK
TW
SG
JP
RU
KR
Business/total researchers
Government/total researchers
University/total researchers

Source: OECD (2006).

22
The final component of this macro-level overview is the technology balance of
payments in 2004 (Fig. 0.6). Among all countries, Korea and Taiwan represent the two
countries with the greatest technology balance of payments deficit. This is not
necessarily reflective of these two countries’ inability to compete with other leading
countries such as the U.S., the U.K., Japan, or France. Korea and Taiwan are, after all,
generating a considerable share of overall patents, which could be licensed to other
countries and create positive contributions to the technology balance of payments. This
simply indicates that, given these net flows, technology receipts are much smaller than
technology sales, and that Korea and Taiwan are purchasing massive amounts of
technology relative to the amount which they sell. It is expected that this purchased
technology is used in the generation of additional product or process R&D, or as a
component in a patent portfolio which provides additional leverage when Korea and/or
Taiwan are faced with intellectual property rights claims against other countries.

23
Fig. 0.6 Technology balance of payments: 2004
Technology balance of payments - millions of US dollars
-5000 0 5000 10000 15000 20000 25000 30000
US
GB
JP
FR
BE
LU
CA
NO
FI
RO
CH
DE
AR
PT
CZ
RU
MX
IT
TW
KR

Source: OECD (2006).

A final point of similarity among several of the East Asian countries, Korea and
Taiwan included, is the principle of shared growth which was established by the
country leaders. In essence, all groups were promised benefits as the economy
expanded, while the government established institutions and bureaucracies in support
of this principle (World Bank, 1993). “Shared growth” is currently represented in
Korea and Taiwan with efforts to coordinate the public and private research sectors.
The principle of shared growth, however, may be less strongly adhered to now, after
certain sectors were left unprotected following the 1997-98 financial crisis.

24
Nevertheless, the nature of the national R&D programs provides very much a sense of
“national progress” and unity within Korea and Taiwan.

Case specifics – developmental state
In the 1950s and 1960s, the government played a vital role in overcoming
market failures (Stiglitz 1986). Government failures in the 1970s were viewed as more
problematic, but revisionists of the old political economy school emerged, arguing that
economic success in East Asia was based on active and market-friendly intervention
policies, particularly in Korea and Taiwan. Despite the logic of the World Bank (1993)
policy recommendation of getting the prices right, for countries at the early stages of
economic development, the real world experience demonstrates severe institutional and
economic constraints to price controls and economic liberalization. These experiences
suggest that a period of neo-classical interventionism has to precede laissez-faire
capitalism. The key to success is that interventionist measures are market-conforming
rather than market-distorting (World Bank, 1993).
This same pattern is present in Korea and Taiwan with regard to R&D policies.
Their governments have incentivized public-private R&D collaboration with the
subsidization of research in an attempt to implement market-conforming interventionist
measures.
23
In the context of the developmental state, discussion on the function of
public finance by Skocpol (1982), Johnson (1982), Woo (1991), and Haggard and

23
R&D promotion activities by the government in Korea and Taiwan are not identical: in Korea, there
are tax credits, accelerated depreciation is allowed, and import tariffs have been lowered; in Taiwan,

25
Cheng (1998) typically emphasizes the importance of loans in the East Asian
developmental state.
24
, but there is also a need to determine whether these subsidies are
properly market conforming. A proper test for this is offered in Chapter 1.
25
Korea and Taiwan have differed in several ways. The most apparent distinction
between Korea and Taiwan has been differences in firm size, which is a function of the
developmental state in each country. Essentially, Korea exercised development via
large scale firms, while the small and medium enterprise sector (SME) dominated in
Taiwan (Pack, 2000). In Korea, the strong leadership the government exhibited in its
industrial drive later became its liability, as it was hindered by the political collusion
between the state and conglomerates. This led to an irrational allocation of resources
and impeded the growth of small and medium enterprises (Ernst, 2000). Along these
lines, Levy (1991) documents the stronger role of SMEs in Taiwan rather than Korea
for the footwear manufacturing industry. The effects of these two parallel yet
alternative strategies is acknowledged and incorporated into the analysis of Chapter 2.
After the currency crisis of 1997-98, Korean conglomerates reduced investment
in R&D, which enabled SMEs to increase their overall share in R&D in Korea. The
number of venture capital firms in Korea also increased from 100 (pre-crisis) to more

R&D expenses are fully deductible, accelerated depreciation is also allow, and large firms are
encouraged to invest (Yusuf, 2003).

24
R&D subsidies are clearly distinct from financing through loans, reflecting the need to update the East
Asian developmental state discourse

25
The KORTAI R&D dataset used in Chapters 1 and 2 of this dissertation is focused expressly on
government programs in both countries which subsidize public and private research entities on the
condition that cross-sector R&D collaboration ensues.

26
than 7,000 in June 2000 (Kim, 2001), as investors sought out innovative research
projects outside the conglomerate mainstream. This may represent a structural change
in Korea, as some predicted that over-diversification of the Korean conglomerates
prevented maximum accumulation of knowledge (Ernst, 2000). SMEs in Korea and
Taiwan are valid mechanisms for innovation, which some attribute to the inter-
organizational linkages facilitating knowledge transfer in both directions (Ernst, 2000).
In Chapter 2, with the KORTAI R&D dataset, there is a specific focus on the
collaborative tendencies of SMEs and the collaborative tendencies of GRIs and
universities to partner with large firms and SMEs.
To conclude this sub-section, mention of the close connection between the
“national innovation systems” approach and the underpinnings of the developmental
state is necessary. An innovation system is a set of institutions, the interaction of which
determines innovative performance. These institutions can be deliberately created by
the government or spontaneously arise from society or the market (Yusuf, 2003), and
an analysis of the developmental state must certainly attend to those institutions
resulting from government policies – in our case, policies affecting the propensity for
public-private R&D collaboration. The first two chapters of this dissertation show that,
in the case of Korea and Taiwan, the national innovation system is a function of and
dependent on the nature of the developmental state. Theoretical deficiencies of the NIS
have been identified by Edquist (2001), who claims that the NIS approach lacks a
theoretical view of the state, despite the connection between it and alternative models
of innovation, such as the Triple Helix paradigm (see Chapter 2). To this end, Edquist

27
emphasizes the importance of the interaction between different organizations in the
pursuit of innovation.

Case specifics – developmental state with an R&D focus
Industry policy can be thought of as having two main elements: functional
interventions and selective interventions. Functional interventions, according to Lall
(1994), are interventions which remedy market failure without favoring any one
activity over another. Selective interventions are designed to favor individual activities
or groups of activities in order to correct suboptimal resource allocation.
26
It is in this
context of selective interventions that we now turn to the specific details of R&D-
related, S&T policies of Korea and Taiwan over the last thirty years.
From the 1960s and 1970s, imitation was the source of rapid industrialization.
This imitation, or “reverse engineering,” of existing foreign technologies required
minimal investment in R&D for the production of simple products. However, reverse
engineering rarely occurs in a vacuum, requiring multi-level interactions among firms,
universities, and public R&D institutes (Kim and Nelson, 2000). The Korea Institute of
Science and Technology (KIST) was established in 1966 as Korea’s first GRI. KIST’s
specific purpose was to provide solutions for less complicated technological issues as
well as to help internalize foreign technologies. As the industrial focus expanded and
the demand for technical support increased in the 1970s, KIST was spun-off into

26
Selective interventions include trade policy (protection from imports and/or promotion of exports),
financial sector policies (affecting the demand and supply of industrial credit), tax benefits and

28
various specialized institutes, such as the Electronics and Telecommunications
Research Institute (ETRI, which is the GRI of focus in the subsequent analysis).
In the initial period, GRIs played a key role in the acquisition of technology
from abroad, and Korean firms often collaborated with GRIs to receive knowledge
about technology as well as about foreign suppliers. Kim (2000) describes GRIs as the
anchor of national R&D projects since 1982, although they are much less dynamic than
corporate R&D centers in recent years, and their role has been weakened in light of
university- and firm-based R&D efforts. Kim (2000) speculates that they may have
been better off adjusting their roles and targeting specific areas of agriculture, public
health, the environment, and/or nuclear energy. What is clear, at least in terms of the
sample which comprises the KORTAI R&D dataset, is that the industrial focus of GRIs
and universities in Korea does not overlap that of the private sector.
Electronics became the emphasis in Korea, following the monumental Heavy
and Chemical industry Plan of 1973. The government helped foster public and private
research institutes for electronics and informatics and actively encouraged information
transfer through licensing and consultations, and the Korea Institute for Electronics
Technology (KIET) helped coordinate and disseminate semiconductor R&D with firms.
KIET also facilitated with inflows of international technology transfer and market
research (UNIDO, 2005). By the mid-1980s, the Korean conglomerates had surpassed
KIET’s R&D capabilities, so the institute sold off much of its fabrication infrastructure

investment incentives, direct government investment and ownership, highly selective foreign investment
regimes, measures to encourage industrial agglomeration, and labor market regulation.

29
to the private sector, was renamed ETRI, and was retasked to focus on basic research
rather than commercialization (Wade, 1990).
The development of private firms’ in-house research capabilities diminished the
need for GRIs, and they were restructured in 1982 as part of the Ministry of Science
and Technology’s (MOST) National R&D Program. GRIs now complemented industry
research by engaging in upstream tasks, thereby preventing the weakening of the
science knowledge base and duplication of efforts. The National R&D Program was the
first indication that policy directives acknowledged the significance of moving beyond
simple imitation or reverse engineering of foreign-based technologies. It also reflected
the government’s involvement in promoting upstream research at GRIs rather than at
universities.
From the mid-1980s, the Korean government began technology planning and
evaluation of public R&D programs, to minimize duplication of research efforts and
delineate R&D specializations between different ministries. This attempt to develop a
long-term, strategic approach to R&D called for the participation of both the public and
private research sectors. All research interests were given consideration, ultimately
leading to the first concrete cases of publicly instigated public-private R&D
collaboration (Chung, 2001). Public-private R&D collaboration increased in frequency
from 1989 as the government implemented programs modeled on the U.S. Small
Business Innovation Research (SBIR) program and the Advanced Technology Program
(ATP). (Korean counterparts for the SBIR and ATP programs are the bases for the
private and university sub-groups of the Korean data used in Chapters 1 and 2.)

30
The New Economic Five Year Plan of 1993 was the first set of policies
clarifying the government’s role in the transfer of technology from the public to the
private research sectors (Choi and Lee, 2000). The Plan aimed to facilitate
collaboration among research entities from both sectors, and it was followed by the
1995 Support Act for Starting SMEs, which facilitated technology transfer specifically
from the public to the private sector. From 1992-94, the Industry-Academy-Research
Institute Cooperative Research Center (IARCRC) was established under STEPI (KIST
at that time) to transfer technology from GRIs to industries, provide free technological
consultation, and to provide specialized information in various technological fields.
The financial crisis of 1997-98 prompted a marked shift in Korea’s policy orientation
to address some of these alignments, particularly the 21
st
Century Frontier Technology
R&D Program. With goals of raising the quality of researchers, the number of
researchers, and reforming Korean universities, the effects of this program on public-
private R&D collaboration will have continued significance.
27

27
Lee (2000) describes the evolution of S&T policies in Korea in the context of three stages of
technological development: the imitation stage, the internalization stage, and the generation stage.
During the imitation stage, foreign technology imitation is the predominant means of acquiring
technological capability. The internalization stage starts when local engineers are capable of developing
products or constructing new plants through indigenous efforts, or when domestically manufactured
products become technically superior to products manufactured in foreign countries. The generation
stage begins when a nation is capable of introducing the market-leading products and state-of-the-art
core technology. Lee (2000) identifies the three stages in Korea’s technology development in the
following way: imitation stage, 1962-79; internalization stage, 1980-89; generation stage, 1990-present.
Innovation policy in the internalization stage was based on the implementation of functional incentives,
such as tax-based incentives for R&D, rather than sectoral incentives. These incentives enhanced the
private firm’s capacity for innovation and accumulating in-house R&D capabilities, and were coupled by
public R&D investments under the National R&D Program (NRDP) of MOST. Universities and private
firms could both participate in such a program, competing against GRIs to acquire R&D project funding.
The programs targeted in the KORTAI R&D database, however, are reflective of the policy changes
which occurred during the generation stage. In an attempt to construct an NIS on par with other highly
advanced countries, during the generation stage, S&T policies target a more balanced development of
research capabilities within universities, GRIs, and private firms.

31
Turning now to Taiwan, government support for industrial development began
in the early 1960s, with the founding of numerous technology development centers for
metals, chemicals, mining, energy, glass, textiles, and food processing. This was
followed with a number of separate but coordinated efforts, particularly the
consolidation of three centers in 1973 into the Industrial Technology Research Institute
(ITRI), which continued to expand with the formation of new laboratories. One of these
laboratories, the Electronics Research and Services Organization (ERSO),
28

exemplified early forms of information transfer, identifying appropriate foreign
partners, establishing pilot plants, and developing capabilities which were subsequently
spun off to create new enterprises. The practice of licensing technology was also
initiated at this time, drawing in at times the expertise of university researchers who
trained firm-based researchers.
29
The entire ITRI model eventually led to other
government institutes focusing on personal computers or international business needs
(UNIDO, 2005)
The true beginning for Taiwan’s fostering of R&D efforts was in 1928 with the
creation of Academia Sinica, under the office of the president.
30
This was the primary
effort to foster basic research. The National Science Council (NSC) was established
subsequently to coordinate national S&T policy as well as oversee funding for some

28
ERSO was merged with the Opto-Eletronics and Systems Laboratory in early 2006 to create the
Electronics and Optoelectronics Research Laboratories.

29
This practice was later mirrored in Korea (UNIDO, 2005).

30
Academia Sinica’s beginnings were on the mainland and the institute transferred to Taiwan at the end
of the civil war (c. 1950).

32
GRIs and science parks. The MOEA, on the other hand, became the national economic
agency in charge of the Industrial Development Bureau (IDB) and the Department of
Industrial Technology (DoIT). The former administers policies for industrial
development, while the latter plans and stimulates industrial R&D.
The laws leading up to the current standard of public-private R&D
collaboration in Taiwan were initiated in 1977, when S&T programs were designed to
develop pre-competitive basic research and “infra-technology” to disseminate results to
the private sector. Under the MOEA, the Industrial TDP, the SBIR Promoting Program,
and the IT Applications Promotion Project were initiated in 1997. In January 1999, the
Fundamental Science and Technology Laws were passed and implemented, which
stated that IPRs and achievements of S&T R&D that are funded or subsidized by the
government will be given or will be authorized to institutes or private firms and not
constrained by the National Property Law (Mathews and Hu, 2007).
31
In 2001, the
Industrial Technology Development Alliance Program and the Academic TDP were
initiated to foster greater public-private links, the latter program specifically addressing
university-based R&D collaboration.
In many ways, particularly with regard to the content of Chapter 3, these
changes in the intellectual property rights are highly significant. Countries must
accumulate sufficient indigenous capabilities and have an extensive science and

31
This is consistent with the effects of the Bayh-Dole Act in the U.S., which subsequently allowed
public funded research to become the property of the researching entity, especially when such entities are
university-based.

33
technology infrastructure to undertake creative imitation (Kim, 2003). Among other
countries, Kim (2003) identifies both Korea and Taiwan as having such capabilities,
but observes that they would not have achieved their present levels of technological
complexity had strong IPR regimes been applied in the early stage of their respective
industrializations. This point is supportive of Lall and Albaladejo (2001), who have
determined that developing countries experience long-term benefits from strong IPRs
only after reaching a set level in their industrialization. These points and others will be
addressed in detail in Chapter 3.

Fieldwork & the KORTAI R&D Dataset
Fieldwork
The unique dataset used in the first two chapters of this dissertation – the
KORTAI R&D dataset – is based on my field research while I was a visiting researcher
at ETRI and ITRI, which are now two of the world’s largest GRIs. Thirty-eight and
twenty-five interviews were conducted with research directors and heads of projects in
Korea and Taiwan, respectively, all of whom received public funds leading to cross-
sector R&D collaboration.
The notable findings of these first-stage field research interviews are presented
here in general terms, revealing many consistencies between the Korea and Taiwanese
cases. First, the impetus for collaboration is primarily structured around improving the
use of a particular technology and ultimately commercializing the results. To this end,
industrial consortia and industrial researcher education programs have been established

34
to promote the research agenda of the individual public research institute. Technical
advice on general or detailed issues may be provided to companies, which may or may
not be incentivized with an honorarium or fee. Collaboration also provides private
firms with the opportunity to use facilities which are not otherwise available.
In other instances, public-private R&D collaboration develops new technology,
both machines and processes. Specific components are supplied by the private partner,
which facilitates the research efforts of the public institute. The private partner, usually
an SME, is specialized in the manufacturing of such components and may also generate
specific components which more directly contribute to the finished R&D result. Many
interviewees, in fact, stressed that the private firm was an invaluable source of
components which would have been extremely difficult and costly for the public R&D
institute to produce.
Public R&D directors also referenced the effects of the IPR-related institutions
upon public-private R&D collaboration. Often, the private firm engages in
collaboration solely to generate patentable R&D results. Royalties from these patents
are returned to the public research institute, but ambiguities in royalty policy can
disincentivize public-private R&D collaboration. When arrangements are not clarified
beforehand, there can be, for example, stand-offs between collaborators following the
completion of an R&D project, with great costs to both sides.
In terms of project design, approximately one quarter of interviewees stated that
projects are designed in the public sector while recognizing the demands of industry.
Another third of interviewees stated that the projects are designed primarily at the

35
public institute, but that the private firm may be involved in project design in varying
degrees from the early stages. Specifically, GRIs typically design projects in response
to a firm’s specific problem or request, while the GRI determines the research method.
The GRI evaluates the request and presents this evaluation to the firm. The firm can
then either provide the funding to the GRI for the research project, or it can use the
evaluation as the basis for a proposal for government research funds. Regardless, the
project is always designed to create a new innovation and satisfy some deficiency in the
available R&D.
The basis of this project design structure is rooted in the various forms of
discussion which may result between partners. Basic research concepts may be
provided by the public research institute, for example, which are then confirmed by the
private firm. The line of communication may also be initiated when public institutes
inform the private firm of their capabilities and present proposals for the private firm’s
review (unless the private firm first approaches the public institute with a specific
problem), as described above. Regardless of which entity takes the first step towards
collaboration, partner selection is ultimately based on the identification of capabilities
held by potential partners.
Complementarity between the public and private research sectors is another
important issue and is a component of the capabilities described in the previous
paragraph. Interviewees particularly emphasized differences in knowledge between the
public and private sectors. Precisely, the firm provides information about current
application areas and estimates of technology demand, while the public research entity

36
takes a more long-term perspective of a particular technological development.
32
The
practical information provided by the private sector was viewed by interviewees as
essential for the generation of practical and commercializable results, which is a stated
goal of the public funding programs detailed earlier. Even when private collaborators
have minimal knowledge about particular technologies, their proximity to market
provides them with insight that is still unfamiliar to public researchers. Likewise, the
solutions provided by public researchers are usually things which had never been
considered by the private firm.
Complementarity in assets also plays a role here. Public researchers are
identified as responsible for the generation of core research results or prototypes,
although prototypes may also be more efficiently produced at the firm level, while
private researchers facilitate the process toward commercialization. This is because the
private firm’s R&D orientation is primarily oriented toward improving manufacturing
techniques, such as reducing the size of a particular product. The public research entity
(specifically the GRI, in this case) is attempting to present a new and better technology
to young companies. Because the knowledge is so different for each sector, the
collaborators have an initial meeting and determine the knowledge that is known by
each group. Usually, the private sector is responsible for mass production activities.
Because of funding concerns, this identification of capabilities is important for the
private firm, which will be responsible for paying the GRI for services rendered.

32
This long-term perspective is present both in terms of the duration of research as well as the effects of
research results.

37
Alternatively, the private firm may establish linkages with other industrial partners and
divide up tasks in an effort to streamline the production process and exploit economies
of scale.
The private firm is always oriented towards increasing market share for itself.
For GRI researchers, however, they are interested in improvements at the national level.
That is, they are interested in generating improvements across an industry, not just for a
particular firm. If one company is doing very well, collaboration between a GRI and
that company will not likely occur. To ensure that this pattern continues in Taiwan, the
government limits public-private R&D collaboration between the GRI and very
successful firms, while collaboration is promoted between the GRI and small
companies. In other cases, the government has focused on the establishment of an
entirely new industry, such as nanotechnology.
33
There are a number of remaining issues which can also be considered in the
context of public-private R&D collaboration, such as the function of previous
employment in the opposite sector. For the public researcher, this can foster a more
applied orientation. In one case, for example, the researcher formerly worked in the
private sector in the printed circuit board (PCB) design and integrated circuit (IC)
design industry. At the time of the interview, however, the researcher worked in the
public sector, in the package design industry. This movement between parallel

33
The science adviser within the government is responsible for making these decisions, not GRI
researchers, it should be noted. In Taiwan, there is a council that responds to the president and generates
proposals to foster innovation and growth. Members include the president of NSC, Academic Sinica,
perhaps foreign researchers in a particular area, and the GRI president (since the project will ultimately
occur at the GRI).

38
industries should expedite the transfer of knowledge between research sectors and is
facilitated by changing market conditions and the availability of funding, especially for
the public sector.
The affiliated technology transfer organization (TTO) also plays an important
role, especially for GRIs, but also for universities and private firms. In the U.S., the
Bayh-Dole Act helped formalize the on-campus TTO’s, and there is now a body of
university technology managers – AUTM – which maintains standards and helps
identify how much technology and revenue is being generated from university-based
researchers. In other countries, TTOs are now becoming increasingly needed and
recognized, given the increased emphasis on IPR strength and R&D output in the form
of patents. Especially for GRIs in Korea and Taiwan, where the packaging and
marketing of patent portfolios to companies are targeted at an increasing rate, TTOs are
all the more important.
34,35

The KORTAI R&D dataset
Empirical estimates are based on data collected from a closed-ended survey
questionnaire distributed to public and private research directors who have received

34
There is an institutional aspect of patent portfolios which needs to be addressed. Because of the
growing trend towards identifying copyright or intellectual property rights infringement, patent packages
may not expand an entity’s knowledge base but function as a tool to wage legal appeals and generate
income through the courts. The generation of funds through patents in this way is counterintuitive to the
original research project and has extensive negative externalities.

35
Indeed, TTOs now function to distribute results domestically as well as abroad. At the time of the first-
stage field research interviews for this dissertation, ITRI was initiating its first ever international division
within the ITRI TTO.

39
public funding for cross-sector R&D collaboration – the Korea-Taiwan (KORTAI)
R&D database – which was compiled by the author. This study attempts to adhere to
the sampling methods of Sakakibara (1994), Branstetter and Sakakibara (1997), and the
bulk of quantitative studies surveyed in Ruegg and Feller (2003), all of which have also
focused on specific government research funding programs.
36
The design of this
questionnaire was facilitated by the first-stage field research interview results,
described in abridged form in the previous section.
37
The existing literature also served
as an important framework in the creation of a tailored questionnaire design exclusively
for Korea and Taiwan’s public and private research directors and managers.
White papers, formal directives, and personal interviews with analysts at
government agencies in both countries led to the selection of several funding programs
which call for public-private R&D collaboration as a stressed or necessary condition of
receiving research funds. In Korea, KOSEF’s (of MOST) Centers of Excellence
Program (CoE) and IITA’s (of MIC) Information Technology Research Center (ITRC)
Program provide funding to university-based researchers. The Taiwanese counterpart to
university-based programs is the TDP’s Technology Development Program for
Academia (TDPA). GRI-based project leaders were selected from Korea’s Electronics
and Telecommunications Research Institute (ETRI) and Taiwan’s Industrial
Technology Research Institute (ITRI). Representing private sector participants in

36
This dataset was collected in the winter and spring of 2005-06, following field research and interviews
by the author with public and private research directors in Korea (summer 2005) and Taiwan (winter
2005). The averaged response rate for both countries and all sectors is just under 50 percent.

37
In spite of repeated and failed attempts to meet with private research directors in person for first stage
interviews, the second stage results have generated a healthy response from the private sector.

40
public-private R&D collaboration are participants in Korea’s Mid-term Technology
Development Program and Communal Core Technology Development Program (ITEP,
for short; both of MOCIE) and Taiwan’s Small Business and Innovation Research
(SBIR) Program (of MOEA).
In Korea, sources of government funds for public-private R&D collaboration
include, but are not limited to, MOST, the Ministry of Commerce, Industry, and Energy
(MOCIE), and the Ministry of Information and Communication (MIC). These three
ministries are representative of much of the federal obligation for R&D, and the
conditions of its provision affect the structure of public-private R&D collaboration. In
the case of MOCIE and MIC, funds are provided for R&D on the condition that results
are commercialized. MOST, with its stronger emphasis on basic research, orients its
programs around research results with both long-term and commercializable prospects.
In any case, the provision of funds to public institutes is conditioned by the stipulation
that relations with industry are fostered.
These relations between the funding ministry and the research entities of
interest are illustrated for Korea in Fig. 0.1, and for Taiwan in Fig. 0.2. The most
apparent distinction is that private firms are given a much greater opportunity in
Taiwan to receive funding in terms of its source, although its quantity is much less than
in the Korean case (see conclusions for Chapter 1). The SBIR Program, which is the
source of the private firm respondents in the KORTAI R&D dataset, is but one
component of such public funding, but it is also made available through a number of
other programs and facilities, particular the strong emphasis on university-based and

41
GRI-based start-up firm incubation centers. While these are themselves a phenomena,
these SMEs and start-ups were not selected as potential recipients of the KORTAI
R&D questionnaire for fear that they were not necessarily engaging in cross-sector
R&D collaboration or did not have sufficient expertise.

42
Fig. 0.7 Connections between ministries and research entities, Korea

43
Fig. 0.8 Connections between ministries and research entities, Taiwan

KOSEF outlined the policy orientation of the CoE over fifteen years ago, but
there is recent evidence of programs calling for cross-sector R&D collaboration, such
as MIC’s “839” policy. This policy changes the focus of public research institutes from
basic research to research that will enlarge the market for particular products. Applied
and developmental research is the focus, and public-private R&D collaboration has
been identified as a necessary means.
In terms of bureaucratic organization, despite fluctuations over time between
these two countries, the centralized government in Korea has resulted in a shift of
policymaking from the bureaucracy to the president’s office, while technological
oversight diminished. At the same time, leaders from the private sector are now
involved on the National Science and Technology Council, which is responsible for
S&T policies and government R&D funding. Similarly, the private sector is also

44
involved in the management of GRIs by participating on the relevant research council
boards.
38
In Taiwan, the organizational structure has been considerably looser but
coordinating mechanisms still play a role (Cheng, et al.,1998).
39
Yet, as was stated
earlier, the Taiwanese government more intensely promotes collaboration between
GRIs and SMEs rather than with large, successful firms. The science advisor within the
government is responsible for making the relevant R&D-related decisions, and there is
also a council which is accountable to the president and produces proposals to foster
innovation and growth. Members of this council include the president of the National
Science Council (NSC), Academia Sinica, foreign research entities with particular
expertise, and the GRI president.
The KORTAI R&D dataset is based on a maximum of 325 questionnaire
responses from research project leaders and research institute directors, all of whom are
recipients of public research funds in one of the aforementioned programs. The overall
response rate is 43 percent, described in detail in Table 0.2. As the sample is not
randomly generated, the response rate may appear problematic, but the selection issue
is not as serious as it might look. Indeed, the omission of these potential respondents is
expected to create an even more impressive selection of participants in public-private
R&D collaboration. Respondents determined their qualifications based on the
questionnaire’s cover page instructions and requirements, specifically that s/he must be

38
Cited from www.oecd.org/dataoecd/30/60/34242958.pdf.

39
Cheng, et al. (1998) list the following coordinating mechanisms: U.S. aid, the strong central bank, and
a number of organizing structures and bodies peripheral but connected to the government.

45
a project leader or a research center director. Since the TDPA, SBIR, CoE, ITRC, and
ITEP programs have all experienced large increases in participation in the years
immediately preceding the questionnaire distribution, it is quite likely that many non-
respondents self-determined that they lacked sufficient relevant experience and hence
chose not to complete the questionnaire. A similar rationale can be applied to ITRI and
ETRI, where internal restructuring occurred immediately following the first-stage field
research. To some degree, this represents shortcomings in questionnaire distribution
methods and the inaccessibility to classified program-level data to target the best
available list of research entities engaging in public-private R&D collaboration. But, it
was ultimately decided that it would be preferable that the research entities be given the
opportunity to self-select their involvement. This was also in the interest of preserving
an environment of anonymity.
40
Table 0.2 Survey response rate information
Distributed Completed Response rate
ITRI 74 46 0.62
TDPA 79 18 0.23
SBIR 363 119 0.33
Taiwan overall 0.39
ETRI 99 49 0.49
CoE 66 24 0.36
ITRC 43 27 0.63
ITEP 118 42 0.36
Korea overall 0.46

40
Additional controls are employed in the subsequent statistical analysis to account for sample
selectivity bias, including respondent’s age, number of years at present job, and number of years in
present industry. In total, 2,903 instances of public-private R&D collaboration have been supervised by
this sample of research directors and project leaders, from 1997 to 2005. The questionnaire was
distributed and collected by the author, ministry-level officials, GRI-based directors, and government
agency officials. Based on the existing research in all related languages (English, Korea, and Chinese),
no other survey of this type has been conducted.

46
Some additional reasons for believing the non-response problem to be less
serious than it might appear are the following: First, the number of distributed
questionnaires is an over-estimate, representing the maximum number of respondents
available as of the final drafting of the KORTAI R&D questionnaire, based on the
available information on each program and research institute. Second, some project
managers and directors would not have been able to respond due to the fact that they
were no longer available. They may have left the GRI or lost their government
funding.
41
In some respects, confidentiality issues limited access to specific, historical
information about individual institutes and firms. This was particularly true with regard
to the private sector-based programs, SBIR and ITEP.
Finally, with regard to the university-based research centers – TDPA, CoE, and
ITRC – each government program allocates a portion of its overall funding to more
basic research efforts. This is consistent with the U.S., Japanese, and European
counterparts to the Korean and Taiwanese R&D funding programs which emphasize
university-firm R&D collaboration. Basic research is distinct from applied research and
experimental development, as Table 0.1 describes, and typically occurs at the
university rather than the GRI or private firm.
42
Among the TDPA, CoE, and ITRC
funding recipients, those who engage in basic research will experience little or no

41
This is, again, tied to limits of the available information.

42
When firms are large enough, they will often possess the infrastructure and capabilities to engage in
basic R&D, but these large enterprises are not considered samples in the KORTAI R&D dataset, and
there will be no discussion of their impact and influence.

47
cross-sector R&D collaboration, and therefore may well have not self-determined that
it would not be appropriate for them to response to the survey.
Questionnaire responses are structured in several different ways, ranging from
raw numbers, selections from zero to ten (for percentages), selections from one to
seven (Likert scale), and dummy variables.
43
For informative purposes and to provide a
general idea about the nature of the KORTAI R&D dataset, responses for research
emphases, industrial affiliation, definitions of R&D success, and respondent
information is presented here. This information is included in robustness checks for the
hypothesis tests of Chapters 1 and 2.
Fig. 0.9 Research emphases
0 1 2 3 4 5 6
1-7 Likert score
public private
Korea Taiw an Korea Taiw an
Average of research emphases by sector & country
basic research em phasis applied research emphasis
patenting em phasis publications em phasis

Source: KORTAI R&D dataset

43
A variable list is presented for each chapter in the appendix. Please refer to each subsequent chapter
for details.

48
Along a 7-point Likert scale, research emphases (Fig. 0.9) are not different
between research sectors in both countries with regard to an applied research emphasis
and an emphasis on patenting. In both countries, at this general level, basic research
and publications are emphasized more by the public sector than the private sector,
which is expected, given the complementarity assumption.
In Korea, based on our questionnaire responses, research sectors are industry-
specific. The public sector is primarily focused on computer and electronics
manufacturing as well as electrical equipment, appliance, and component
manufacturing. Respondents from the Korean private sector share an emphasis on
electrical equipment manufacturing, but also include respondents with substantial
emphases on machinery manufacturing, fabricated metal product manufacturing, and
chemical manufacturing. In Taiwan, on the other hand, the public and private research
sectors possess virtually identical research emphases, focusing primarily on computer
and electronics manufacturing, followed by electrical equipment manufacturing,
machinery manufacturing, and chemical manufacturing.

49
Fig. 0.10 Industrial affiliation

Korea – public sector Korea – private sector
chem
plasrub
mach
compel
elequip

chem
plasrub
fabmet
mach
compel
elequip
transeq

Taiwan – public sector Taiwan – private sector
chem
plasrub
mach
compel
elequip
transeq

chem
plasrub
fabmet
mach
compel
elequip
transeq

Note:
“chem” - chemicals
“plasrub” - plastics & rubber
“fabmet” - fabricated metals
“mach” - machinery
“compel” - comp. & electronics
“elequip” - electrical equip.
“transeq” - transportation equip.
Source: KORTAI R&D dataset

Based on Fig. 0.10, cross-sector R&D collaboration is more likely in Korea
than in Taiwan to draw upon complementarities in industrial affiliation. It may be

50
overly strict to assume that these results are sufficiently representative of public and
private research entities in Korea and Taiwan. For one, these seven industrial
classifications are general approximations,
44
especially when compared to the
specificity of using three or four digit subject identification codes of the USITC. It is
quite likely, therefore, that the similarities between the Taiwanese public and private
sectors illustrated in Fig. 0.10 will lessen as the industrial fields are narrowed down,
revealing affiliations that align the public sector with process technology and the
private sector with product technology.
To control for variance and provide greater explanatory power to the empirical
conclusions offered in this dissertation, one method of analysis is to focus on a single
sector engaged in R&D in both Korea and Taiwan. A number of studies conducted in
this fashion have been mentioned already. Unfortunately, the specific focus upon the
impact of state involvement in the generation of public-private R&D collaboration
greatly limits the sample size. If we were to focus specifically on, for example, the
biotechnology sector or the information technology sector, this dissertation would be
reduced to a case study method of analysis.

44
The actual questionnaire includes a total of ten classifications, including a category for non-listed or
miscellaneous industries (“other”). The remaining three classifications – nonmetallic mineral product
manufacturing, primary metal manufacturing, and the miscellaneous category – were excluded from Fig.
0., because they were not significantly representative of the respondents’ industrial affiliation.

51
Fig. 0.11 Defining success
0 .2 .4 .6 .8 1
Korea Taiwan
public private public private
By Country & Research Sector
Definitions of R&D Success
publications of results
patenting of results
comm ercialization of results

Source: KORTAI R&D dataset

There are statistically significant differences between the definitions of R&D
success for each research sector, based on Fig. 0.11. This pattern is essentially the same
for each country, demonstrating a clear emphasis on commercialization in the private
sector. In the public sector of Korea and Taiwan, commercialization is similarly
emphasized, although patenting and publications are also defined as successful R&D
outcomes by over 60 percent and 25 percent of respondents, respectively.

52
Fig. 0.12 Respondent information
0 10 20 30 40 50
years
Korea Taiwan
public private public private
By Country & Research Sector - public sector subdivided
Personal Information of Respondent
age
years at job
years in industry

Source: KORTAI R&D dataset

In Fig. 0.12, the remarkable observation is that, in Korea, respondents have
worked for a greater number of years on average in a particular industry vis-à-vis the
number of years worked at his/her current job. In Taiwan, however, respondents on
average worked a greater number of years at his/her job. In other words, Korean
researchers experienced a change in work while still in the same industry. In Taiwan,
researchers’ industries changed while working at the same job. This is because the
company or the public institute dictates the respondent’s orientation, creating a shift in
emphasis. The responsible party for such shifts is ultimately the government, given its
direct impact on KORTAI R&D respondents’ research goals. Again, these measures
will be included as controls in the subsequent empirical studies.

53
Conclusion
The initial aims of this introductory chapter have been to (1) present R&D
collaboration in a general and historical context, (2) make the connection between
R&D collaboration and market failure concerns, (3) present the relevant literature, and
(4) connect this literature to the Korean and Taiwanese cases. A careful analysis of the
data, coupled with details of government-initiated R&D in these two countries, makes
it clear that the existing East Asian developmental state discourse is not sufficient.
Korea and Taiwan are unmistakably at the forefront of innovation on a global scale,
surpassing many Western European nations.
45
The pattern of state intervention is still
evident, but we are witnessing a modified developmental state in these two countries.
There is a distinct shift toward innovation and R&D targeting, along with industrial
targeting. In this way, Korea and Taiwan are growing as a function of an innovation-
based developmental state.
This chapter also integrates the following three chapters of this dissertation in
the context of the existing literature. Specific themes which have been related to one or
more of the following chapters include market failure in R&D efforts, the linearity
(uni-directionality) hypothesis, information transfers and linkages within and between
countries, and advanced technology development. Despite the increased attention given
to matters of R&D collaboration over the last ten to fifteen years, as a sub-field of
political economy, we still seem to be trying to figure out which topics are most
desirable. I have skirted this issue by incorporating several theoretical, empirical, and

45
See Chapter 3 for further details.

54
analytical approaches with hopes that it makes for a richer body of knowledge on
Korea and Taiwan, both comparatively and in the global context.

55
Chapter 1: Information Flows and State-led Technological Innovation in Korea
and Taiwan
Introduction
There are a number of viable reasons for public-private R&D collaboration to
occur on its own (David, et al., 2000; Scott, et al., 2001; OECD, 2004). The science
and technology (S&T)-oriented policymaker, however, is faced with the task of
creating R&D collaboration targets because of market-failure in R&D. To correct for
market-failure, the government’s objective is to identify and target “winning” projects
that are privately unprofitable but socially beneficial. One measure of social benefit is
the degree of spillovers accompanying these projects (Stiglitz and Wallsten, 1999).
This paper is structured to examine the degree of such spillovers and to examine how
they are determined by government funding programs. The analysis is limited to public
and private recipients of government research funds in Korea and Taiwan. Given that
government grants and subsidies in both countries are provided to generate unique and
commercializable research results, public-private R&D collaboration often occurs.
46

There is a greater propensity for information diffusion to occur when public entities
such as universities collaborate with firms rather than when firms collaborate with each
other (Dasgupta and Maskin, 1987). In this way, science and technology policies
indirectly affect information transfers through public-private R&D collaboration.

46
For now, we define “public” as government research institutes (GRIs) and universities, “private” as
private firms.

56
A sizable literature explicitly addresses the issue of information transfers.
D’Aspremont and Jacquemin (1988) use information transfers as an output measure,
and Griliches (1998) emphasizes how information transfers lead to increases in R&D
productivity.
47
In both of these studies, however, the source and nature of such
transfers is neglected. This lacuna is due to a strict focus on the private research sector,
where knowledge transfers levy significant costs to the research-engaging firm, making
appropriability issues paramount.
48
This is the basic economic and transaction cost
framework for firm-based R&D collaboration. When one approaches the subject from
both public and private sector perspectives, government intervention is of more
consequence in terms of how it can increase cross-sector (public-private) information
flows. This intervention arises for a number of reasons, primary of which is the attempt
to correct for market failure in R&D. It has been found that many requirements of
learning and R&D involve serious market failure, necessitating corrective policies as
well as policies to tackle learning costs and promote externalities and linkages (Lall,
2000).
49
These linkage-promoting policies are identifiable in part through instances of
public-private R&D collaboration.

47
In his survey of the literature, Griliches (1998) considers the content of information transfers, but not
the method of transferring information, upon which we focus.

48
There is also a literature which show a correlation between decreasing costs and increasing
information transfers. For a survey, see Bernstein and Nadiri (1988: 429).

49
Lall (2000) provides an evolutionary approach which is not limited by the restricting assumptions of
accumulation with learning and perfect markets. These assumptions are adopted by the production
function framework, which lacks the necessary depth to show how individuals in the East Asian
countries were able to capitalize on advanced technologies. Nelson and Pack (1998) classify theories
based on the production function as “accumulationist” in nature, as they primarily focus on the stock and
flow of operative inputs.

Market-failure corrective efforts play a minimal role at this level of analysis and,
when they do play a role, are functional rather than selective.

57
To clarify, this study approaches the issue of information transfers in terms of
its advantages rather than its costs. The costs of increased information transfers are
directly connected to the incentives of not conducting R&D, as the sharing of new
knowledge diminishes the marginal benefits for the R&D producing entity. Here, these
costs are assumed to be negligible because the focus is placed on information transfers
between public and private research entities rather than that between private firms.
50

This assumption is based on the fact that public and private research entities are not
direct competitors but provide complementary efforts and skills. There may be some
overlap in terms of patent generating efforts, but this assumption is largely accurate,
shown in a comparison of definitions of R&D success between sectors (see Table 1.1).
Table 1.1 Definitions of R&D success between sectors
-------------------------------------------------------------------------------------
Sub-group Variable | Obs Mean Std. Dev. Min Max
-------------------------------------------------------------------------------------
Korea public defpat | 87 .6896552 .4653167 0 1
defcom | 87 .8045977 .3988087 0 1
defpub | 87 .3103448 .4653167 0 1

Korea private defpat | 41 .4390244 .5024331 0 1
defcom | 41 .9512195 .2180848 0 1
defpub | 41 .097561 .3004062 0 1

Taiwan public defpat | 56 .625 .4885042 0 1
defcom | 56 .8928571 .3120939 0 1
defpub | 56 .3392857 .4777518 0 1

Taiwan private defpat | 115 .3913043 .4901781 0 1
defcom | 115 .973913 .1600915 0 1
defpub | 115 .1304348 .338255 0 1
-------------------------------------------------------------------------------------
Notes: defpat, defcom, defpub indicate dummy variables where 1 is assigned by KORTAI R&D
respondent if successful R&D results are defined as patenting of results, commercialization of results, or
publication of results, respectively.

50
A comprehensive empirical study measuring the methods through which firms capitalize on their
research output is presented by Cohen, et al. (2000).

58
Government Intervention & S&T Policies
Government intervention via S&T policies is a frequently discussed topic,
particularly in East Asia. Lall (2000) and Pack (2000) share a number of concerns
about the role of technology in economic development for the East Asian NIEs. These
two scholars differ, however, with regard to the role of government intervention. Pack
(2000) finds no viable examples of successful government intervention in East Asia,
while Lall (2000) claims that the process of shifting from “know-how” to “know-why”
is the result of national agents promulgating local capabilities to innovate.
51
Captured
within these innovative capabilities are accelerated information transfers through the
mechanism of public-private R&D collaboration. Others conclude that minimized state-
led R&D efforts will make room for an industry-centered NIS (Kim, 2000), but it is
assumed that cross-sector R&D collaboration will not be as strongly emphasized in
such an environment. Rather, only marginal changes to existing technology will occur
through experimental development and basic and applied research efforts will
decrease.
52
There are separate concerns about government intervention for both the public
and private research sectors. Crowding-out is the primary issue for the private sector, as
government interventions via technology policy are meant to create an environment
which encourages private investment rather than substituting for it (Branscomb and

51
The result of public funding may create yet another form of failure: research failure. In contrast to
market failure, research failure is due to conditions of public funding limiting the impulse of independent
inquiry (Butos and McQuade, 2006). This is an extreme and a key political economic concern, but we do
not concede that such conditions are present in either the Korean or Taiwanese cases.

52
See the introductory chapter for details on the types of research available.

59
Keller, 1998). David, et al.’s (2000) thorough survey of the literature, however, is
inconclusive as to the effects of public R&D investments upon private R&D
investments.
Government funding for research at universities and GRIs has been identified as
having the potential to alter the course of innovation (Nelson and Rosenberg, 1993).
The effects of government involvement upon the university are primarily framed by
whether the government has a duty or right to refocus goals of higher education
institutions. Mowery, et al. (2004) identify the ramifications and differences of
“business-oriented” and research/teaching universities. On this count, we agree that the
university’s focus must remain on research and teaching, although our reasons are
specific to the phenomenon of public-private R&D collaboration. If the focus on
commercialization becomes paramount at the university, the retasking of university
goals can potentially limit the advantages of collaborating across sectors. In other
words, the university will excessively embody characteristics of the firm and minimize
the propensity for information diffusion between R&D collaborators. As the box-and-
whisker chart in Fig. 1.1 shows, the public research sector places significantly more
emphasis on basic research and publication of results than the private research sector,
which is more evidence of complementarities between the public and private research
sectors.
53

53
The rectangular boxes in each figure represent those responses between the twenty-fifth percentile
(lower hinge) and the seventy-fifth percentile (upper hinge). The median is found directly in the middle
of the box. Lines (or “whiskers”) extending from the box are capped with adjacent values, beyond which
are outside values, represented by small circles. Adjacent values are calculated by multiplying the

60
Fig. 1.1 Research emphases by sector and country
1 2 3 4 5 6 7
public sector private sector
Korea Taiw an Korea Taiw an
resbas: 1-7 Likert scale (7 being greatest) for basic research em phasis
respub: 1-7 Likert scale (7 being greatest) for publications em phasis
resbas respub

Source: KORTAI R&D dataset

It is also worth mentioning the number of forms in which government
intervention facilitates public-private R&D collaboration. For the American automobile
industry, government involvement helps with the generation of environmental
improvements. Roos, et al. (1998), point out that this is a fundamental difference
between R&D programs such as Semiconductor Manufacturing Technology
(SEMATECH) and Partnership for a New Generation of Vehicles (PNGV). In the
former case, increased competitiveness is the ultimate goal; for the latter, the
government facilitates the partnering of private firms with national laboratories

interquartile range (the difference between the first and third quartile values) by 1.5, and adding or
subtracting it from the upper or lower hinges, respectively.

61
(GRIs).
54
The government is also able to coordinate public-private R&D collaboration
with the intention of overhauling an entire industry, which was intended through the
Lean Aircraft Initiative (LAI).
55
As well, small, university-based research centers, such
as the U.S. Industry-University Cooperative Research Centers (IUCRC), are designed
to facilitate academic-to-firm information transfers. Indeed, IUCRCs have been shown
to play a role in the uni-directional form of information transfer, namely via faculty
consultation, co-authoring of publications, and the hiring of graduate students (Adams,
et al., 2001).

Korea & Taiwan
Turning now to the cases of this study, government intervention through public
funding is nicely aligned with studies of the developmental state in East Asia. In his
challenge to neo-liberal economic arguments against public funding, Chang (1999)
claims that price signals – the cornerstone of neoliberal arguments – fail to
acknowledge long-run development and growth patterns. Indeed, this is especially true
with regard to R&D efforts, where returns are often calculated along a longer than
normal time horizon. While public funding and finance are the essence of the state’s
function (Woo-Cumings, 1999), developmental state studies have failed to examine the
public financing of R&D. This is a significant omission, for market-failure corrections

54
SEMATECH targeted the increased competitiveness of US chip manufacturers, and the hands-off
approach of the government enables SEMATECH participants to effectively manage the R&D consortia.
PNGV was designed to improve national competitiveness in manufacturing through innovations which
would achieve three times the fuel efficiency of 1994 family sedans (Roos, et al., 1998).

55
Now the Lean Aerospace Initiative.

62
in R&D are now prevalent in the East Asian region, Korea and Taiwan in particular. In
the pursuit of explanations for government intervention in areas with economically
beneficial consequences for the economy, attention to the national innovation system is
highly warranted.
56
Existing models of the East Asian developmental state can be generally
structured along two foci (Evans, 1998). The market-friendly model is aligned with the
World Bank’s (1993) seminal text which describes how East Asian governments
managed to maintain macroeconomic essentials throughout the rapid growth period. On
the other hand, the industrial policy model of Johnson (1982) and subsequent analyses
(Amsden, 1989; Wade, 1990) describes the entrepreneurial function of government
policies, in addition to the macroeconomic stability of the market-friendly model.
Policymakers, thus, are responsible for identifying potentially strong and/or weak
sectors and distributing funding to maximize growth prospects.
57
The distinction
between these two models and the present reality is that the functions of policymakers
also include the targeting of skills and effort, so the entrepreneurial policymaker of the
industrial policy model now possesses a requisite degree of scientific and technological
knowledge. The effectiveness of this knowledge is reflected in a number of ways, but

56
Haggard (2004) makes a similar point along these lines.

57
Evans (1998) actually outlines three models. The entrepreneurial function of the second model
embodies the investment concerns of the third model, we believe.

63
the amount of information transfer resulting from public funding offers an additional
layer of analysis to studies of East Asia emphasizing the political environment.
58
This is hardly the first attempt to contrast the Korean and Taiwanese cases with
regard to the role of the state. From an historical perspective and with regard to the
command style of these two countries, Hamilton and Biggart (1988) point out that,
unlike the Korean case, Taiwan’s economic plans in the past have no implementation
procedures, are not supported by controls, and lack credibility. Others subscribe to the
view that Korea is “interventionist” while Taiwan is “supportive” (Park, 1990).
59
It is
claimed here, however, that both Korea and Taiwan’s policymaking structures are
currently designed to bolster their national innovation system, particularly via public-
private R&D collaboration. Details of these policy structures are not offered here,
60
but
it has been found that, from the early 1980s, the Taiwanese government withdrew from
direct intervention while supporting efforts to build a knowledge foundation as a
particular development policy (Hsueh, et al., 2001). It will ultimately be shown here,
however, that this represents only moderate convergence of the S&T structures of
Korea and Taiwan. A relative lack of overall funding in R&D in Taiwan has created an
environment in which R&D fund recipients use disbursements with more determination
and to productive ends, such as an increase in the receipt of information transfers.

58
See Cheng (1990) for an example of such a study.

59
Specifically, there is evidence in Korea of domestic market protection and industrial targeting; in
Taiwan, medium-term economic plans limit policymakers authority to allocate credit (Park, 1990).

60
See Shapiro (2007) and Shapiro (forthcoming) for historical accounts.

64
Modeling Information Transfers
Increased social welfare has been identified in the literature as a positive
function of information transfers between research collaborators (Hagedoorn, et al.,
2000). We contextualize the benefit of information transfers through R&D returns and
apply a similar structure to that used in Jaffe’s (1998) study of the U.S. Advanced
Technology Program (ATP).
61
In this theoretical work, Jaffe’s underlying premise is
that market, knowledge, and network spillovers increase social returns, which are
ultimately measured here as the sum of private returns and the benefits from
spillovers.
62
In Jaffe’s model, knowledge is assumed to originate in either the private or
public research sector and is transferred to other research entities through publications
or patents, indicated by χ
1
in Fig. 1.2. One research entity’s efforts to create returns, in
the form of reputation effects (A
1
) for the public sector or profits (A
2
) for the private
sector, can generate additional returns through other research entities’ reputation effects
(D
1
) or profits (D
2
), or ultimately benefiting the customer (E) through some measure of
increased satisfaction. Both public and private research entities can generate income
through patent licensing (B), determined in part by a functioning technology transfer
office or office of technology licensing. In summation, Fig. 1.2 describes how a

61
The ATP, incidentally, is the basis for several of the research funding programs in Korea and Taiwan
from which the KORTAI R&D dataset (the primary dataset under analysis in this paper) is drawn.

62
Knowledge spillovers occur through reverse engineering or through the reading of other’s findings in
published form, and full compensation is not awarded to the original source of such information. Market
spillovers benefit the customer when the same price is paid for products of higher quality, which are the
result of product innovations. As well, process innovations can lead to decreased production costs which
result in lower prices, again benefiting the consumer. Network spillovers are exemplified by the
successful coordination between research entities to create a new technology.

65
research entity’s attempts to create private returns (the sum of Α
1
, A
2
, or B) leads to
greater social returns (the sum of Α
1
, A
2
, B, C, D, and E.), when information transfers
along χ
1
. Under the assumption that greater information transfers lead to increased
returns, both social and private, it is claimed here that public funding should target
projects with large amounts of information transfers.
The framework presented in Fig. 1.2 makes a strong case for the government to
prioritize and support projects which promise large information transfers, but the
structure is ill-equipped to deal with the specifics of information transfers within a
public-private R&D collaborative framework. In addition, only three forms of
information transfer are considered (patents, publications, and licenses). If we
acknowledge that information can be transferred between public and private research
entities on a number of other levels, Jaffe’s (1998) structure must be modified. The
existing literature and details of the Korean and Taiwanese cases confirm this.
Fig. 1.2 Returns to R&D and information channels without collaboration

66
Hall, et al.’s (2000) research on the ATP program examine the relations
between universities and firms in R&D, qualitatively measuring the interaction
between public and private research entities. They conclude that universities enhance
research efficiency by deepening and expanding the research of ATP-fund receiving
firms. Speaking specifically to the issue of knowledge transfers, Adams, et al.’s (2001)
analysis of the U.S. Industry-University Cooperative Research Centers (IUCRCs)
shows that knowledge transfers are increasing, particularly through faculty
consultations, co-authorship of publications across sectors, and the hiring of former
graduate students. These transfers are unidirectional, however, focusing only on public
to private information flows.
63
Information transfers are bi-directional and are not
linear (Mowery, 2004). We are in alignment with those who identify a more complex
arrangement of inter-organizational R&D based on the character of relationships
between universities and/or GRIs and private firms (Fontana, et al., 2003). There are,
for example, numerous meetings before beginning the actual R&D process, not to
mention the updates and modifications done during the actual research. Therefore, the
linearity assumption of one-way information flows becomes moot, especially when
commercializable research results are pursued and the public research sector
incorporates the market-based research of the private sector.
Given the assumptions of bi-directionality, Jaffe’s (1998) model is limited. A
more appropriate reflection of private and social returns to R&D and information

63
Joly and Mangamatin (1996) show that information flows in public-private R&D contracts were all
one way, from the public to the private sectors.

67

68

channels in the context of public-private R&D collaboration is presented in Fig. 1.3.
This figure captures the information transfers which occur between public and private
research entities at several different levels, based on the assumption that ideas do not
necessarily originate in one sector or the other (Fontana, et al., 2003). The initial stage
of public-private R&D collaboration, represented by δ
1
in Fig. 1.3, involves the bi-
directional sharing of ideas about commercialization and project feasibility.
Information transfers along δ
1
occur through conferences/meetings as well as dialogue
about each collaborator’s previous research results. Also included in δ
1
is previously
documented research in the form of patents and publications. After the goals are
determined, each sector works to complete their specified tasks: public research entities
develop prototypes or conduct experiments and tests while the private sector prepares
its manufacturing facilities for new production. Throughout this period, research
entities from the public and private sector may be in consultation, in conference, and
exchanging researchers. The information transfers occurring at this stage of the project
are represented as δ
2
in Fig. 1.3. Finally, information transfers via χ
∗
1
in Fig. 1.3 are
distinct from those of χ
1
in Fig. 1.2 in that the former arises from public-private R&D
collaboration while the latter does not.
64

64
In other words, the information transferred in χ∗
1
embodies the transfers which occurred in δ
1
, wile
those of χ
1
in Fig. 2 do not.

69
Prototype development/
experiments
Private R&D entity
Other product markets
Better products, lower costs.
More knowledge through interaction
with the public research sector
Public R&D entity
New knowledge
Other R&D entity ’s
knowledge
A
*
: Reputation effects
___________________
B
*
: Returns from
licensing
C
*
: Firm profits
___________________
D
*
: Customer benefit
E
1
*
: Reputation effects
E
2
*
: Firm ’s profits
___________________
F
*
: Customer benefit

Fig. 1.3 Returns to R&D and information channels with public-private collaboration

From this theorizing, it should be clear that the information transfers afforded
under instances of public-private R&D collaboration could be expected to be more
numerous. The distinction on the far right-hand side – outputs measured by returns to
research – is that both public and private entities generate returns from the same project.
Whether these are effectively generated as a result of public fund-instigated R&D
collaboration, which is expected to be a function of cross-sector information transfers,
will be examined in the following section.

Method & Data
The goals of this paper are three-fold. First, we offer an examination of the
channels through which information is received between the public and private
research sectors, and how certain channels affect R&D output. Second, the strength of
the government is considered as an explanatory variable of information transfers. Third,
the similarities and/or differences between Korea and Taiwan are identified.
The significance of cross-sector (public-private) R&D collaboration in its
impact upon information transfers between sectors has already been established, but
there is still much uncertainty as to how information is received, and through which
channels. In the first stage of analysis, and to represent a semblance of the returns
displayed on the right-hand side of Fig. 1.3, the dependent variable is R&D output.
Two relevant measures of R&D output are available: (1) patents and (2) publications
generated through public-private R&D collaboration. To demonstrate the importance of
information transfers, in this first stage of analysis we model both measures of R&D

70

output through the use of a number of different information transfers. This provides a
sense of legitimacy to the aforementioned claims regarding information transfers, and
the results shall show the degree to which patents and publications are positively
predicted by information transfers, in its varied forms.
Government funding is provided under different stipulations, and measurements
of success are not fixed. As such, both collaborative patent output and collaborative
publication output are considered as dependent variables. The general terminology used
in the outlines of these funding programs is that the research outcomes will have lasting
effects, although the continued provision of funding is typically based on a quantifiable
measure, such as the number of patents and publications. Given that there is less likely
to be a proclivity of publications for the private sector, using both collaborative patents
and publications as dependent variables, we can control and test for variation between
the public and private research sectors with the use of a private sector dummy variable
(0 or 1, 1 for private sector). The same can be said about patents with regard to the
public sector, although there now exists a clear incentive for universities and GRIs to
patent, given the implementation of Korean and Taiwanese counterparts to the U.S.
Bayh-Dole Act of 1982: the 2000 Technology Transfer Facilitation Law and the 1999
Fundamental Science and Technology Laws, respectively.
The second stage of analysis approaches the question of public funding of
public-private R&D collaboration and its impact upon information transfers. Thus, the
degree of information transfers is now the dependent variable, which is explained by
the extent to which public funding generates public-private R&D collaboration. After

71

examining the effects of public funding on information transfers, we drop the
assumption that research entities receiving public R&D funding are operating in a
vacuum. Such entities are simultaneously exposed to a number of factors uncorrelated
with the effects of public funding, which may determine public-private R&D
collaboration. In this more constrained but realistic view of the determinants of
information transfers, two additional explanatory variables are considered in addition to
the public funding variable: the strength of the respective office of technology licensing
(OTL) and geographic proximity.
These two additional explanatory variables are included as controls on the
influence of public funding in the generation of public-private R&D collaboration.
Beyond their methodological import, they can provide some very interesting
interpretations of the Korean and Taiwanese cases, which will be attempted in the
concluding section to this paper. The decision to include the OTL variable is based on
the relatively recent arrival of OTLs in these two countries, and the use of a geographic
proximity measure is consistent with an extensive literature, such as that of Saxenian
(1994).
65

In this second stage of analysis, the data is broken into country- and sector-
based sub-groups, to provide a more detailed examination of the Korean and Taiwanese
cases. The literature is virtually replete with comparative analyses of these two
countries, given their similar historical and cultural roots, and especially because of

65
Multicollinearity between the explanatory variables is not an issue, as pairwise correlation coefficients
between them are not significantly high. See Appendix 1.2 for pairwise correlation coefficients of these
second stage-based variables.

72

their mirroring growth patterns. The number of shared details between Korea and
Taiwan effectively controls for a number of explanatory variables, allowing one to
make narrow conclusions about these two cases, which is the third stage of analysis.
Much of this related discussion, again, is reserved for the final section of this paper.

Measuring Information Transfers
Before introducing the data formally, a few words must be said on the nature of
information transfers. This key variable has already been discussed at length in the
context of Fig. 1.3, and it should be apparent that there is no single mechanism to
account for such transfers at the policymaking level.
66
The selection of information
transfers considered here is based on first-hand interviews between the author and
research directors and managers in Korea and Taiwan.
67
The selection has also been
influenced by a number of studies, such as Rahm, et al. (1999), who identify the
following university-firm information transfers: licenses and patents, joint R&D in
shared research centers, start-up firms, campus-based industrial services, R&D
contracts, and faculty consultations. Conferences and consultations are also identified
by Cohen, et al. (2002) as important for public-to-private transfers. These conclusions,
however, are founded on a linear, university-to-firm transfer of information. Lécuyer’s

66
David and Foray (1996) make a solid effort to isolate the “knowledge distribution power” of an
innovation system, which includes institutions supporting information transfer (e.g., the norm of
disclosure), incentives toward codification, and an IPR regime which facilitates disclosure.

67
These interviews preceded the distribution of the KORTAI R&D questionnaire, which is outlined in
the introductory chapter.

73

(1998) study of MIT in the U.S. describes information transfers which do not strictly
flow to the firm. University engineers, for example, can ally themselves with local
corporations to find jobs for graduates and align their curriculum to maximize the
usefulness of students (after graduation) for private firms. The university can also hire
practicing engineers to come to the university to lecture and increase the teaching
competence of the faculty. University-based engineers also train engineers from
industry to address shortcomings in research approaches.
68
Publications are also given
attention in the literature as a potential source of information transfer. Private firms use
publications as a method of detecting expertise within public research organizations,
which is just as important a piece of information as new science or technology results
(Scott, et al., 2001).
We must also acknowledge the number of potential problems arising when
measuring information transfer. Simply put, the mechanisms through which knowledge
flows are absorbed by firms are not always directly and objectively observable (Han,
2001). As well, first-hand interviews with researchers in Korea and Taiwan – and
subsequent quantitative analysis of questionnaire responses confirm that information
transfers are not uniform but complementary. This was an expected finding, as Cohen,
et al. (1998) have already found that, when a given channel of how useful information
moves from universities to industrial R&D facilities is at least moderately important,
based on factor analysis, channels tend to be used together. They found specifically that

68
I.e., the public researchers provided the necessary theoretical and/or practical insight into the research
process, enabling the firm-based engineers to arrive at a solution. In the case of Lécuyer’s (1998) study

74

person-to-person interactions and other forms such as publications and conferences
function as complements, while hires, joint ventures between industry and universities,
patents and contract research are less important. This is reconfirmed in Cohen, et al.
(2002), which indicates that public and personal channels of information transfer –
publication, conferences, and informal interactions – are more important than licenses
and cooperative ventures.
69
Fritsch and Schwirten (1999) also conclude that personal
contacts are very important among universities and firms, with 80 percent of their
survey respondents mentioning that informal contacts are important.
It is clear that multiple forms of information transfer are required for
comprehensive analysis of this phenomenon, but the nature of the research questions
considered here limit deep investigation along the lines of Cohen, et al. (1998, 2002).
Another constraint is the multicollinearity which arises if all forms of information
transfer are considered simultaneously.
70
To remedy this, factor analysis is employed to
create a composite measure of all information transfers (fa_trans). This method
generates communalities set to the squared multiple-correlation coefficients. The first
loading proved to be sufficient, so there was no need to examine any sort of rotation of
the factor-loading matrix.
71

of MIT, this was exemplified through the efforts of Prof. Noyes, who promoted collaborations with
faculty and the laboratories of large science-based firms such as GE and DuPont.

69
These studies, again, are minimized to the uni-directional transfer of information from the university to
the firm.

70
See Appendix 1.1 for pairwise correlation coefficients for the six measures of information transfers.

71
In a pairwise correlation coefficient matrix, the estimates from the factor patterns and a Cronbach’s
alpha measure presented a correlation coefficient of 0.9961, which confirms the validity of the former
estimates.

75

Variables of the KORTAI R&D Dataset
The variables selected from the KORTAI R&D dataset for analysis here come
in three forms: raw numbers of patents and publications in 2005, Likert scores, and
dummy variables.
72
The 2005 data for patents and publications is included because it
offers the largest response rate, given the increasing application of public funding for
the fostering of public-private R&D collaboration in Korea and Taiwan. The output of
these R&D collaborations, based on the revised model presented above in Fig. 1.3,
should be measured in ways which reflect the various overall returns. These differing
returns, denoted by the superscript asterisk, are quite diverse and may at times be
categorized by the public or private nature of the research entity. The model is
structured to show the expected correlation between A* and B* and public research
entities, and between C* and D* for private research entities.
73
These delineations are
not necessarily fixed, as a large firm may have the resources to generate basic research
results, which could have a reputation effect. Likewise, the returns from licensing, B*,
is applicable to both the public and private research sectors, given the increasing role of
patent licensing. The increasing role of offices of technology licensing (OTLs) at the
research entity can potentially have an impact for both public and private research
entities. Such offices are already implemented in the private sector, as part of the R&D

72
See Appendix 1.3 for a complete description of these variables, with their abbreviated notation.

73
E2*, and E1* and F* for that matter, are presented in Fig. 1.2 to illustrate and reemphasize the
importance of spillovers in the public-private collaborative framework (χ∗1). This offers an extension to
the information transfers discussion presented here; however, the expected impact is peripheral to the
S&T policy-related issue of public-private R&D collaboration and will not be tested here.

76

department of firms. At GRIs and universities, OTLs are having an increasing influence,
given the patenting emphasis. This overlap between sectors is evidenced in the
summary statistics for collaborative patents (patcol) and collaborative publications
(pubscol) in 2005.
Table 1.2 Descriptive statistics for dependent variables by sub-group: Number of
collaborative patents and number of collaborative publications in 2005
-----------------------------------------------------------------------------
Sub-group Variable | Obs Mean Std. Dev. Min Max
-----------------------------------------------------------------------------
Korea public patcol | 87 .4252874 .9229904 0 5
Korea private patcol | 41 .5121951 .9518916 0 4
Taiwan public patcol | 56 .6607143 1.947276 0 10
Taiwan privat e patcol | 115 .5304348 2.418212 0 20

Korea public pubscol | 87 .6436782 2.31264 0 20
Korea private pubscol | 41 .4390244 .9232762 0 4
Taiwan public pubscol | 56 .2678571 .8200016 0 4
Taiwan privat e pubscol | 115 .0956522 .4387765 0 4
-----------------------------------------------------------------------------

Given the concerns raised above regarding the variety of information transfers,
the KORTAI R&D dataset includes six separate measures of how useful information is
received by the respondent from the opposing research sector. On a 7-point Likert scale,
“7” being greatest, each public (or private) respondent assigns a number to measure
how much useful information is received from the private (or public) sector through a
particular method. These eight methods of information transfers considered here are
patenting, publications, hires, conferences, contract research, and consultations. This is
consistent with the literature presented in the “Measuring Information Transfers”
section above. As the secondary analysis within this discussion is to determine how
information transfers are predicted in terms of each group of researchers, a number of
similarities and differences within and between each country can be identified in terms

77

of pairwise correlation coefficients between research output and each of the six
methods of information transfer, presented in Tables 1.3a and 1.3b. How such
differences play out while controlling for country and sector variance (via dummy
variable assignments) is shown in the subsequent section.
Table 1.3a Pairwise correlation coefficients by sub-group: number of collaborative
patents (2005) and methods of information transfer
-
Korea public Korea private Taiwan public Taiwan private
----------------------------------------------------------------------------
-----------------------------------------------------------------------------
patcol | 1.0000 1.000 1.000 1.000
consult_trans | 0.3424* 0.1577 0.0139 0.0839
hire_trans | -0.0307 0.0393 0.0479 0.0935*
patents_trans | 0.3396* 0.214 -0.0526 0.4110*
papers_trans | 0.1345 0.2574 0.0430 0.2060*
confer_trans | 0.2022 0.2312 0.1047 0.1945*
contract_trans | 0.3268* 0.2652 0.1218 0.2242*
-----------------------------------------------------------------------------
* significance level <0.05

Table 1.3b Pairwise correlation coefficients by sub-group: number of collaborative
publications (2005) and methods of information transfer
-----------------------------------------------------------------------------
Korea public Korea private Taiwan public Taiwan private
-----------------------------------------------------------------------------
pubscol | 1.0000 1.0000 1.0000 1.0000
consult_trans | 0.1548* -0.0159 0.0604 0.0131
hire_trans | 0.1252 0.1390* 0.0593 0.1934*
patents_trans | 0.1427* -0.0106 0.0386 0.1576*
papers_trans | 0.2146* 0.0900 0.0307 0.1554*
confer_trans | 0.0521 0.0774 0.0487 0.1478*
contract_trans | 0.1264* 0.0771 0.1605* 0.0945*
-----------------------------------------------------------------------------
* significance level <0.05

The variable capturing the impact of public funding upon public-private R&D
collaboration (pubfundinstig) is also on a 7-point Likert scale, “7” being greatest,
representing the degree to which public-private R&D collaborations occur as a result of
receiving public funding. Likewise, the OTL efficiency measure (otl_good) is on a 7-
point Likert scale, capturing the degree to which the affiliated technology licensing

78

office (OTL) satisfies the needs of the respondent, while the geographic proximity
measure (geo_coll) shows on a 7-point Likert scale the degree to which geographic
proximity positively affects the decision to collaborate with research entities from the
opposite sector. Where indicated, and consistent with the underlying assumptions about
Korea, Taiwan, and the public and private research sectors mentioned above, country
(0 or 1, “1” for Taiwan) and sector (0 or 1, “1” for private) dummies are included. In all
regressions, industry dummy variables for the following sectors are also included, but
not reported in the output tables: chemical manufacturing, machinery manufacturing,
computer and electronic product manufacturing, electrical equipment, miscellaneous
manufacturing.

Results
Results of the test for the effects of information transfers on the two measures
of public-private R&D collaborative output – patents and publications – are presented
in Tables 1.4 to 1.9. Based on the literature, first-stage interviews in both Korea and
Taiwan, and the structural model of Figs. 1.1 and 1.2, it was expected that the receipt of
information from the opposing sector would have a positive impact on R&D output.
These expectations were largely met.
Looking first at the effects of individual forms of information transfer upon the
total number of collaborative patents (Table 1.4), the coefficients for each form of
information transfer were all positive and statistically significant. There tended to be
some differences between countries with regard to the strength of the effects of

79

information transfers through patenting, publications, and conferences, but the
coefficient values are not statistically significant. Likewise, the private dummy was in
no case statistically significant, indicating the absence of significant differences in such
assessments between sectors. At this aggregate level, and looking at the coefficient
values for each form of information transfer, the predicted effects of information
transfers are ranked from greatest to smallest as follows: patenting, conferences,
publications, contract research, hires, and consultations. To some extent, these results
are also expected, particularly with regard to the effects of information transfers
through patents. The fact that transfers through publications also predicts collaborative
patent output, however, indicates that basic research output is important. This is
because basic research is primarily published. We do not go so far as to say that there is
a uni-directional transfer of information and innovation, but the connection between
publications and patent output is indeed interesting.
When factor scores of all six methods of information transfer are the
explanatory variable (Table 1.5), the coefficient estimates are largely consistent with
information transfer’s individual effects. At this stage of analysis, examination is
conducted by country and sector sub-sample, and there is considerable variance
between countries.
74
There is also a difference between the effects of the factor scores
of information transfer on collaborative patents within Taiwan, as the effect of
information transfers for the public sector is neither statistically significant nor relative

74
Each loading from the factor analysis of the data at the sub-group level are based on the Likert scores
for the six measures of information transfers for each sub-group. For example, in Table 1.5, the

80

to the effects of information transfer for the private sector. This can be attributed in part
to the overall higher output of collaborative patents in Taiwan’s private research sectors,
shown already in Table 1.4. Such bias could be expected to lead to marginally higher
coefficient estimates, given the overall size of the Taiwan private sector.
It is restrictive to assume that collaborative patenting and collaborative
publications have no impact upon the composite measure of information transfers in
Table 1.5 (fa_trans), particularly as it is based partially on the degree to which
information transfers via patents and publications. To address the endogeneity concerns
which arise from the use of a composite measure of information transfers which
includes transfers via patents and publications as predictors of collaborative patents and
collaborative publications, a revised composite measure of information transfer is
presented in Table 1.6. This revised measure (fa_trans_2) is based only on the Likert
scores for the degree of information transfers through conferences, hires, contract
research, and consultation, excluding transfers through patents and publications. In
Table 1.6, the coefficient estimates for this revised composite measure present a
general pattern which is consistent overall with the results from Table 1.5. A close
examination, however, reveals that the aggregate measure (column 1 in Tables 1.5 and
1.6) is biased upward with the original measure, as are the coefficients for both public
and private sectors in Taiwan (columns 4 and 5 in Tables 1.5 and 1.6). Regardless,
information transfers continue to present a positive impact upon collaborative patents.

coefficient for the composite measure of information transfers for the Korean public sector sub-sample
reflects the 83 factor patterns of the Likert scores for the Korean public sector only.

81

Our second collaborative R&D output measure – publications – is positively
affected by most of the individual measures of information transfers (Table 1.7), but
the coefficient estimates are nowhere near as high as those predicting collaborative
patent output (Table 1.4). As well, information transfers through patents and
conferences are not statistically significant predictors of collaborative publications.
Indeed, these two methods of information transfers present the two lowest coefficient
values, which is in direct contrast to the pattern of individual effects upon collaborative
patents. To a certain extent, this raises a number of related issues which are not to be
examined further in this discussion. The overall finding is that information transfers at
the individual level have an overall greater impact upon collaborative patents than
collaborative publications.
Scores from the factor analysis of all six information transfer methods present
evidence that such transfers are more likely to have a stronger effect upon the
generation of collaborative publications for the Korean public sector (with the Korean
private sector, of course) than for either sector in Taiwan or for the Korean private
sector. A comparison of the impact of factor scores of information transfers shows
significantly smaller effects for all sub-groups except for the Korean public sector.
Indeed, the difference between the coefficients of factor scores for information
transfers in Tables 1.8 and 1.5 for the Korean public research sector (column [2] in
both tables) is 0.277. For the remaining sub-samples – the Korean private sector, the
Taiwanese public sector, and the Taiwanese private sector – a unit increase in the factor
analysis score leads to 0.218, 0.272, and 0.857 more collaborative patents than

82

83
publications, respectively. At the aggregate level, as well, publications are much less
affected by information transfers than patents. Finally, in Korea, collaborative
publication output is greatly impacted by the factor scores for information transfer,
which is reflected in part by differences in overall publication levels, shown in Fig. 1.4.
Specifications controlling for possible endogeneity of the six component-
measure of information transfer are presented in Table 1.9. Rather than use the original
measure of information transfer (fa_trans), the revised composite measure of
information transfer (fa_trans_2) is used, which excludes the Likert scores for the
degree of information transfers through patents and publications, accounting only for
transfers via conferences, hires, contract research, and consultation. In line with our
earlier results, a comparison of Tables 1.8 and 1.9 shows that the original measure of
information transfer (fa_trans), which includes all six methods of information transfer,
could create an upward bias in the effects of transfers upon collaborative patents as a
result of including transfers through patents and publications. In terms of the effects of
information transfers, however, the pattern remains consistent between both the
original and the revised measures: information transfers in both countries have a greater
impact upon collaborative publications for the public sector than for the private sector.

Table 1.4 OLS results for impact of information transfers upon collaborative patenting
------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
patcol patcol patcol patcol patcol patcol
------------------------------------------------------------------------------------------------------------
patents_trans 0.372***
(0.0668)

papers_trans 0.206**
(0.0713)

confer_trans 0.237**
(0.0754)

hire_trans 0.173*
(0.0723)

contract_trans 0.232***
(0.0615)

consult_trans 0.158*
(0.0684)

taiwan_dummy 0.426+ 0.309 0.533+ 0.0260 -0.0212 0.130
(0.243) (0.256) (0.276) (0.251) (0.251) (0.248)

private_dummy -0.130 -0.0125 0.160 0.240 0.137 0.0849
(0.244) (0.260) (0.256) (0.258) (0.253) (0.255)

constant -0.586+ -0.206 -0.611 0.0578 -0.240 -0.130
(0.332) (0.363) (0.438) (0.331) (0.338) (0.381)
------------------------------------------------------------------------------------------------------------
N 284 281 282 272 281 285
R-sq 0.120 0.049 0.055 0.043 0.070 0.038
F 2.838 1.057 1.200 0.885 1.550 0.814
------------------------------------------------------------------------------------------------------------
Standard errors in parentheses, industry sector results omitted.
+ p<0.10, * p<0.05, ** p<0.01, *** p<0.001
84

Table 1.5 OLS results for impact of information transfers – factor scores – upon collaborative patenting
--------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5)
patcol patcol patcol patcol patcol
aggregate Korea,public Korea,private Taiwan,public Taiwan,private
--------------------------------------------------------------------------------------------
fa_trans 0.630*** 0.370** 0.344+ 0.530 0.997***
(0.128) (0.114) (0.193) (0.338) (0.275)

taiwan 0.209
(0.247)

private 0.147
(0.254)

constant 0.295 0.511* 0.115 0.969+ 0.180
(0.268) (0.233) (0.317) (0.559) (0.447)
--------------------------------------------------------------------------------------------
N 263 83 37 44 99
R-sq 0.101 0.157 0.176 0.083 0.182
F 3.585 2.363 1.329 0.560 3.400
--------------------------------------------------------------------------------------------
Standard errors in parentheses, industry sector results omitted.
+ p<.10, * p<.05, ** p<.01, *** p<.001

85

Table 1.6 OLS results for impact of information transfers – factor scores excluding patents and publications – upon
collaborative patenting
--------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5)
patcol patcol patcol patcol patcol
aggregate Korea,public Korea,private Taiwan,public Taiwan,private
--------------------------------------------------------------------------------------------
fa_trans_2 0.536*** 0.391** 0.304 0.681+ 0.704***
(0.134) (0.119) (0.192) (0.337) (0.288)

taiwan_dummy 0.0302
(0.247)

private_dummy 0.236
(0.255)

constant 0.355 0.550* 0.130 1.029+ 0.217
(0.271) (0.234) (0.326) (0.549) (0.459)
--------------------------------------------------------------------------------------------
N 266 83 38 44 101
R-sq 0.073 0.160 0.158 0.120 0.117
F 2.524 2.411 1.201 0.838 2.086
--------------------------------------------------------------------------------------------
Standard errors in parentheses, industry sector results omitted.
+ p<.10, * p<.05, ** p<.01, *** p<.001

86

Table 1.7 OLS results for impact of information transfers upon collaborative publications
------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
pubscol pubscol pubscol pubscol pubscol pubscol
------------------------------------------------------------------------------------------------------------
patents_trans 0.0684
(0.0530)

papers_trans 0.154**
(0.0537)

confer_trans 0.0884
(0.0576)

hire_trans 0.126*
(0.0567)

contract_trans 0.108*
(0.0464)

consult_trans 0.0918+
(0.0511)

taiwan_dummy -0.338+ -0.267 -0.219 -0.374+ -0.496** -0.417*
(0.193) (0.193) (0.211) (0.197) (0.190) (0.185)

private_dummy -0.217 -0.253 -0.159 -0.194 -0.105 -0.120
(0.194) (0.196) (0.196) (0.202) (0.191) (0.190)

constant 0.642* 0.359 0.436 0.562* 0.495+ 0.466
(0.263) (0.273) (0.335) (0.260) (0.256) (0.284)
------------------------------------------------------------------------------------------------------------
N 284 281 282 272 281 285
R-sq 0.042 0.064 0.042 0.053 0.058 0.049
F 0.919 1.406 0.915 1.116 1.266 1.073
------------------------------------------------------------------------------------------------------------
Standard errors in parentheses, industry sector results omitted.
+ p<0.10, * p<0.05, ** p<0.01, *** p<0.001
87

Table 1.8 OLS results for impact of information transfers – factor scores – upon collaborative publications
--------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5)
pubscol pubscol pubscol pubscol pubstcol
aggregate Korea,public Korea,private Taiwan,public Taiwan,private
--------------------------------------------------------------------------------------------
fa_trans 0.296** 0.647* 0.126 0.258+ 0.140**
(0.102) (0.299) (0.190) (0.130) (0.0490)

taiwan_dummy -0.344+
(0.197)

private_dummy -0.181
(0.202)

constant 0.732*** 1.569* -0.0392 0.258 0.0201
(0.213) (0.611) (0.314) (0.214) (0.0796)
--------------------------------------------------------------------------------------------
N 263 83 37 44 99
R-sq 0.064 0.085 0.195 0.183 0.148
F 2.173 1.177 1.507 1.379 2.659
--------------------------------------------------------------------------------------------
Standard errors in parentheses, industry sector results omitted.
+ p<.10, * p<.05, ** p<.01, *** p<.001
88

Table 1.9 OLS results for impact of information transfers – factor scores excluding patents and publications – upon
collaborative publications
--------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5)
pubscol pubscol pubscol pubscol pubscol

89
aggregate Korea,public Korea,private Taiwan,public Taiwan,private
--------------------------------------------------------------------------------------------
fa_trans_2 0.255* 0.549+ 0.0827 0.230+ 0.0925+
(0.105) (0.315) (0.185) (0.134) (0.0508)

taiwan_dummy -0.414*
(0.194)

private_dummy -0.152
(0.200)

constant 0.757*** 1.613* -0.0576 0.249 0.0253
(0.213) (0.620) (0.314) (0.218) (0.0809)
--------------------------------------------------------------------------------------------
N 266 83 38 44 101
R-sq 0.055 0.066 0.176 0.162 0.100
F 1.860 0.893 1.368 1.196 1.737
--------------------------------------------------------------------------------------------
Standard errors in parentheses, industry sector results omitted.
+ p<.10, * p<.05, ** p<.01, *** p<.001

Fig. 1.4 Number of patents and publications: Korea and Taiwan (1981-2006)
0
5000
10000
15000
20000
25000
30000
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
Korea: patents
Korea: publications
Taiwan: patents
Taiwan: publications

Source: USPTO (2008), ISI Web-of-Science (2008).

Korea and Taiwan diverge most, however, when we turn our attention to the
hypothesis that public funding will have a significant, positive effect on the degree of
information transfers. Such funding is measured here by the degree (1 to 7 Likert scale)
to which it leads to public-private R&D collaboration, Table 1.10 shows that, when
considered singly, public funding is significant and positive at the aggregate level and
at the sub-group level, except for the Korean public research sector (Table 1.10,
column [2]). Even, when accounting for other exogenous factors, such as OTL
efficiency (otl_good) and geographic proximity (geo_coll), at the aggregate level
(Table 1.11, column [1]), the results show that public funding still has a strong, positive
effect upon the factor scores of information transfer. The coefficients of OTL
efficiency and geographic proximity are also positive and significant, but there is a
significant effect of the country dummy variable. Further analysis at the sub-sample

90

91
(country and sector) level (Table 1.11, columns [2], [3], [4], and [5]) verify such
differences between Korea and Taiwan and, to a lesser, degree, within countries.
For Korea, the results show that public funding generally has very little effect
upon information transfers when accounting for other non-policy-related factors, such
as the strength of OTLs and geographic proximity. In fact, when these other two
variables are included for the Korean public sub-sample, the coefficients are
statistically significant and positive. The same is true for the Korean private sub-sample,
except that geographic proximity has no significant effect whatsoever. This stands in
stark contrast to the results for both Taiwan sub-samples, which show a positive and
significant effect of public funding instigated public-private R&D collaboration upon
information transfers. Explaining these country-level differences is the primary purpose
of the concluding section of this paper, to which we now turn.

Table 1.10 OLS estimates for public funding effects on information transfers (fa_trans)
--------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5)
fa_trans fa_trans fa_trans fa_trans fa_trans
aggregate Korea,public Korea,private Taiwan,public Taiwan,private
--------------------------------------------------------------------------------------------
pubfundins~g 0.108*** 0.0234 0.204* 0.223** 0.143**
(0.030) (0.065) (0.086) (0.079) (0.045)

taiwan_dummy -0.237*
(0.117)

private_dummy 0.0921
(0.121)

constant -0.533** -0.200 -1.357* -1.434** -0.501*
(0.197) (0.392) (0.536) (0.448) (0.246)
--------------------------------------------------------------------------------------------
N 263 83 37 44 99
R-sq 0.091 0.032 0.357 0.342 0.148
F 3.178 0.413 3.444 3.207 2.661
--------------------------------------------------------------------------------------------
Standard errors in parentheses, industry sector results omitted.
+ p<0.10, * p<0.05, ** p<0.01, *** p<0.001

92

Table 1.11 OLS estimates for determinants of information transfers (fa_trans)
------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
fa_trans fa_trans fa_trans fa_trans fa_trans fa_trans
aggregate aggregate Korea,public Korea,private Taiwan,public Taiwan,private

93
------------------------------------------------------------------------------------------------------------
pubfundins~g 0.0858** 0.0861** 0.0181 0.129 0.198* 0.134**
(0.117) (0.030) (0.060) (0.095) (0.088) (0.049)

otl_good 0.0822* 0.0828* 0.197* 0.305+ 0.120 -0.0230
(0.038) (0.038) (0.080) (0.156) (0.099) (0.061)

geo_coll 0.0975** 0.0968** 0.144* -0.00677 0.115 0.0366
(0.033) (0.033) (0.056) (0.0967) (0.120) (0.055)

taiwan_dummy -0.328** -0.299+
(0.912) (0.167)

private_dummy 0.0333 0.0702
(0.123) (0.190)

taiwan*private -0.0624
(0.250)

constant -1.114*** -1.121*** -1.526** -2.697* -2.076** -0.615
(0.245) (0.247) (0.461) (1.033) (0.709) (0.382)
------------------------------------------------------------------------------------------------------------
N 256 256 81 35 42 98
R-sq 0.171 0.171 0.251 0.458 0.410 0.200
F 3.293 3.080 2.650 2.025 1.897 1.620
------------------------------------------------------------------------------------------------------------
Standard errors in parentheses, industry sector results omitted.
+ p<0.10, * p<0.05, ** p<0.01, *** p<0.001

Conclusions
Information transfers have generally large positive and significant effects on
public-private R&D collaborative output, which affirms the existing literature and the
tenets put forth in Fig. 1.3. This is consistent between Korea and Taiwan, despite
nuanced interpretations. The impact of public fund instigated public-private R&D
collaboration upon information transfers, however, differs between these two countries.
Given the breadth of similarities between Korea and Taiwan, and to justify the results
of Table 1.11, geographic, OTL duration, and R&D funding differences are identified.
Following, the results of Table 1.11 are further reinforced with a robustness check
which includes an OTL-duration dummy variable as an explanatory variable of
information transfers.
Geographic proximity plays a greater role in Korea, evidenced empirically in
Table 1.11, particularly for the Korean public sector (column [3]).
75
Although Korea
and Taiwan both share a number of geographic similarities, such as their size, island
status (virtual for Korea), use of technology parks, and high urban concentration, there
is much greater research concentration in Taiwan, with a focus on the region including
and between Taipei and Hsinchu. In Korea, the heavy chemical industrial drive of the
late 1970s/early 1980s in Pohang and the more recent targeting of the Gwangju
Institute of Technology (founded in 1993) have done much to disperse R&D
throughout the country. Information transfers in Korea, particularly in the public

75
It is also notable that the coefficients for the geographic proximity variable (geo_coll) are considerably
higher for both countries’ public research sector sub-groups, relative to the private sector sub-group. The

94

95
research sector, are more positively affected by geographic proximity given the greater
relative distances vis-à-vis the distances in Taiwan.
There is a notable difference in the duration of OTLs – the length of time in
which OTLs have been in place – between Korea and Taiwan. For approximately half
of the respondents in both the public and private research sectors in Korea, OTLs are in
their infancy. In Taiwan, however, OTLs have been present for quite some time for a
majority of the respondents. Fig. 1.5 represents the dummy variable capturing OTL
duration, with “1” assigned to OTLs older than two years.
Fig. 1.5 Duration of OTL: Korea (row 1), Taiwan (row 2), public (column 1), private
(column 2)
0 20 40 60 80 0 20 40 60 80
0 1 0 1
0, 0 0, 1
1, 0 1, 1
Percent
otlform dum
Graphs by country and sector

Source: KORTAI R&D dataset
effects of geographic proximity upon information transfers tend to be greater for the public research
sectors in Korea and Taiwan.

96
Table 1.12 Determinants of information transfers: inclusion of OTL formation dummy
--------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5)
fa_trans fa_trans fa_trans fa_trans fa_trans
--------------------------------------------------------------------------------------------
pubfundinstig 0.0848** 0.0116 0.128 0.196* 0.136**
(0.0302) (0.0567) (0.0784) (0.0743) (0.0508)

otl_good 0.0851* 0.193* 0.254+ 0.0921 -0.00298
(0.0380) (0.0752) (0.128) (0.0823) (0.0623)

geo_coll 0.103** 0.138* 0.0150 0.0427 0.0441
(0.0323) (0.0526) (0.0843) (0.105) (0.0567)

otlformdum 0.0220 0.0676 0.117 -0.157 -0.0349
(0.109) (0.191) (0.306) (0.265) (0.194)

country -0.333**
(0.120)

sector 0.0538
(0.121)

constant -1.124*** -1.379** -1.822* -1.677* -0.720+
(0.244) (0.429) (0.760) (0.661) (0.395)
--------------------------------------------------------------------------------------------
N 256 81 35 42 98
R-sq 0.160 0.261 0.440 0.389 0.157
F 4.214 2.791 2.549 2.261 1.826
--------------------------------------------------------------------------------------------
Standard errors in parentheses, industry sector results omitted.
+ p<0.10, * p<0.05, ** p<0.01, *** p<0.001

Comparing the results from Tables 1.11 and 1.12, there is no real change to the
predicted effects of our previously assigned explanatory variables upon the receipt of
information transfers from the opposite sector, when a dummy variable for the duration
of OTL is included.
76
In Table 1.12, at the aggregate (column [1]) and sub-sample
levels (columns [2], [3], [4], and [5]), the coefficient for this dummy variable is
insignificant. However, the difference in signs between the Korean and Taiwanese
cases (Table 1.12, columns [2, 3] compared with columns [4, 5]) provides further
support for our explanation as to why the OTL efficiency coefficient is positive and
significant for Korea and not for Taiwan. OTLs are more efficient in Korea because
they are a relatively newer phenomenon, generating a start-up effect. In Taiwan,
however, OTLs are a longer-standing institution, which is shown by the negative (but
insignificant) coefficient values for the OTL duration dummy variable. Relative to the
impact of public funding, OTLs have little impact upon the degree of information
transfers received.
The OTL start-up issue is dwarfed, however, by the differences in funding
patterns between Korea and Taiwan. Essentially, in Taiwan there is greater funding,
which has a stronger predicted impact upon the degree of information transfers, vis-à-
vis the strength (and duration) of OTLs and geographic proximity. Fig. 1.6 displays the
percentage of total national expenditures in R&D (GERD) by industry and government.

76
Again, this OTL dummy variable is assigned a “1” when OTLs are older than two years.

97

In Korea, industry invests more and government less, relative to Taiwan.
77
This only
captures shares of R&D funding rather than overall levels, so a more accurate
representation of the degree to which the government provides R&D funding in these
two countries is presented in Figs. 1.7a and 1.7b.
Fig. 1.6 Percentage of GERD funding by industry and government
0
10
20
30
40
50
60
70
80
90
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Korea: financed
by industry
Korea: financed
by govt
Taiwan: financed
by industry
Taiwan: financed
by govt

Source: OECD MSTI 2006

It is very evident, based on Figs. 1.7a and 1.7b, that the amount of public money
going to industry in Korea greatly exceeds the amount of such expenditures in Taiwan.
The share of business expenditure in R&D (BERD) financed by the government is
nearly three times greater than in Taiwan, while the total amount of money (expressed

77
This proportional difference is attributed to the differing size of firms in these two countries. Korean
large enterprises have the wherewithal to engage in R&D on a different, more involved level than in
Taiwan. If the share of industry financed GERD were removed from Fig. 6, there is no doubt that GERD
financed by industry in Taiwan would have a larger overall share than that of Korea.

98

in 2000 constant dollars) is four to six times greater. At times it is as high as ten times
greater in Korea than in Taiwan. Cross-country funding differences are also present
when we examine the source of funding for government R&D expenditures
(GOVERD) and higher education (university) R&D expenditures (HERD). Expressed
in Fig. 1.8, industry’s funding of GOVERD and HERD is considerably more in Korea
than in Taiwan. Although this does not directly reflect the influence of public funding
for public-private R&D collaboration, it continues to show that R&D funding overall is
in much greater supply in Korea. This pattern has continued over time, creating an
environment in which Taiwanese research entities utilize their own resources more. In
other words, Taiwan research entities receive more information and focus on
information transfers to a greater extent than their Korean counterparts because of the
need to perform in the presence of relative funding shortages. If these funding
differences were the result of significantly fewer researchers (full time equivalent
[FTE]) in Taiwan, the robustness of these claims would be greatly diminished. As Fig.
1.9 shows, however, the difference is not great: approximately twenty thousand dollars
per researcher.

99

Fig. 1.7a Korea: BERD financed by government: percentage and in 2000 constant
dollars
0
1
2
3
4
5
6
7
8
9
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
BERD % financed by govt.
0
200
400
600
800
1000
1200
1400
million dollars
% BERD financed by
govt.
millions of dollars

Source: OECD MSTI 2006

Fig. 1.7b Taiwan: BERD financed by government: percentage and in 2000 constant
dollars
0
0.5
1
1.5
2
2.5
3
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
BERD % financed by govt.
0
50
100
150
200
250
million dollars
% BERD financed by
govt.
millions of dollars

Source: OECD MSTI 2006

100

Fig. 1.8 Share of GOVERD plus HERD represented by industry: constant 2000
dollars
0
200
400
600
800
1000
1200
1400
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
million dollars
Korea: GOVERD &
HERD by industry
Taiwan: GOVERD &
HERD by industry

Source: OECD MSTI 2006

Fig. 1.9 GERD per researcher: constant 2000 dollars
0
20000
40000
60000
80000
100000
120000
140000
160000
1997
1998
1999
2000
2001
2002
2003
2004
2005
constant 2000 dollars .
Korea: GERD per researcher Taiwan: GERD per researcher

Source: OECD MSTI 2006

Having established the function of government funding in the receipt of
information transfers in a public-private R&D collaborative context, a number of
inherent questions of the aforementioned structural model are still left unresolved.
Future projects utilizing the KORTAI R&D dataset can change the analytical

101

framework, mix up the components of information transfers, and explore institutional
differences of information transfers’ effects upon collaborative patenting and/or
publications. The overall conclusion offered here is that public funding has greater
outcomes in Taiwan in terms of information transfers, relative to Korea.

102

Chapter 2: The Triple Helix Paradigm in Korea and Taiwan: A Test for New
Forms of Capital
Introduction
Successful collaborative R&D output is a function of how partnerships are
initiated, so it is particularly important that we examine the nature of the relationships
between GRIs, universities, and firms. Specifically, I consider here whether researchers
tend to partner repeatedly with cross-sector research entities. As was detailed in the
case study presentation of the introductory chapter, cross-sector or public-private R&D
collaboration is increasingly applied and necessary in Korea and Taiwan, given the
complexity of new technologies. Such advancements typically require interdisciplinary
skills, large financial investments, testing, and experimental and production facilities
(Yusuf, 2003), which is reflected in these two countries’ R&D- and S&T-related
policies. In this way, this chapter attempts to examine the Triple Helix-based claim
which strongly emphasizes the benefits of relations between private research entities,
universities, and GRIs. For Korea and Taiwan, this presents a unique opportunity to
determine whether traditional methods of networking dominate, or whether R&D
trumps culture.
For other countries approaching a similar level of technological capacity, this
chapter provides evidence of East Asia’s industrial success beyond the underlying
macro-level differences.
78
The conclusions offered here confirm Triple Helix-based
claims that new forms of relationship-based capital arise through public-private R&D

78
See Westphal (1990) for a discussion on this point.

103

collaboration. Korea and Taiwan are largely homogenous in these findings, despite the
strength of traditional network structures in these two countries. These traditional
structures were expected to limit the formation of new relationships.
79
There is, thus,
support for the application of policies calling for public-private R&D collaboration in
other technology-oriented countries, such as India and China.
80
This line of research
also bridges the divide between the Triple Helix paradigm, the sociologically-based
research on the East Asian organizational structure (Hamilton, 1990; Park, 2000;
Hamilton and Biggart, 1988; Biggart, 1990), and East Asian political economy (Fields,
1995).
Theoretically, this chapter offers a deeper look at the linkages between the three
parts of the Triple Helix structure than is generally presented in the literature. These
three helices between the government, universities, and private firms offer interesting
conclusions but little discussion of differences between sub-groups. Sub-group refers
here to the different types of research entities within the public and private research
sectors. The private sector, for example, is comprised of a range of different sized firms,
each providing unique qualities to the Triple Helix dynamic. Further, when standard
Triple Helix linkages are drawn between universities, firms, and the government, GRIs
are not specifically identified and positioned in the schematic. Knowledge creation
occurs in the context of a fluid and evolving community and formal organizations are

79
This chapter addresses head-on many of the issues which are presented in Evans (1997), which bridges
the topics of social capital and state-led development.

80
This is true especially for China, given its shared Confucian influence with Korea and Taiwan.

104

poor methods for learning (Powell, et al., 1996). It is, thus, all the more relevant to
identify and examine the sources of innovation where firms, universities, and GRIs
intersect. To test for this, GRIs must be included in the Triple Helix paradigm, as
should delineations between different sized firms.
Methodologically, Triple Helix studies are typically conducted at the case study
level, and comparative research is not common. This paper offers alternatives to these
conventions. Indeed, there are a number of stylized facts which are ripe for quantitative
analysis. Using the unique KORTAI R&D dataset, in combination with interview
results between the author and research directors in Korea and Taiwan, this paper offers
a response to a number of program-, sector-, and country-level questions which are
often raised on a case-by-case basis, as the subsequent literature review shows. With
appropriate techniques and a sufficient sample size, generalizations could be tested
with confidence. Here, the context of such tests are instances of public-private R&D
collaboration resulting from a specific policy structure which targets innovations with
long-term potential and commercializability. As it is concluded in the introductory
chapter to this dissertation, the East Asian developmental state is functioning to
facilitate the interaction and collaboration between the public and private research
sectors. The mechanisms utilized for this effect include the continued use of
government funding and R&D subsidization, as well as reinforced targeting of joint
R&D, the sharing of research facilities, more effective tax incentivization, and a
bolstered IPR regime (Chung, 2004). These are the minimal requirements for other
countries attempting to implement a similar structure.

105

The structure of this paper is designed to present a full analysis of the Korean
and Taiwanese cases, test the aforesaid aspects of the Triple Helix paradigm, and make
propositions about the viability of transference of best-practice techniques. To that
effect, the second section briefly reviews the Triple Helix-related literature and presents
background details on the Korean and Taiwanese cases which have not yet been
covered in this dissertation. The third section introduces the relevant variables from the
KORTAI R&D dataset and presents summary statistics to validate a number of key
assumptions. Section four outlines the empirical specification and also presents the
results and robustness checks from this model. Additional data on the sources of pre-
existing and new relationships is also presented here. The fifth section concludes this
chapter with particular reference to the influence of traditional network structures in
these two countries.

Literature Review & Case Description
Despite the breadth of literature on the Triple Helix phenomena, case study
rather than within-region comparative analysis is generally preferred. Studies of the
Triple Helix paradigm in developing countries, moreover, are relative rare. In this
chapter, attention is drawn to the micro-level institutions at work with respect to the
Korean and Taiwanese cases.
A survey of the related literature also reveals an overall absence of
simultaneous assessments of both the public and private research sectors. There is also
an overwhelming focus on the European phenomenon, recently presented in an analysis

106

of increased entrepreneurship at a single Belgian university (Van Looy, et al., 2006)
and in an earlier study of diffusion through an analysis of technology transfer offices at
German universities (Kruecken, 2003). Colombo and Delmastro (2002) generate
micro-level data in a comparison of firms located in Italian science parks, and
Schartinger, et al. (2002) also gathered micro-level data for their study of the
complexity of knowledge interactions between Austrian universities and firms. The
data of these last two studies, it should be noted, is similar in nature to the content of
the KORTAI R&D dataset, which is the source of the data used here. Marques, et al.
(2006) also provide valuable insights with their case study of a Portuguese university’s
effectiveness in promoting innovation and entrepreneurship throughout the local area.
Beyond Europe, the focus turns to Latin America and East Asia with few
exceptions. These studies also fail to simultaneously consider multiple components of
the Triple Helix paradigm, but they do provide deep insight into the particulars of
individual countries. For example, Casas, et al. (2000) show how the presence of key
scientific and engineering fields in Mexico’s Bajio region exhibit relationships along
the lines of the Triple Helix model. Mirroring the aforementioned Portuguese
university case study, Bernasconi (2005) emphasizes the absence of the Triple Helix
paradigm in a case study of Chile’s Pontificia Universidad Catolica, detailing the
transition from a teaching to a research orientation, in spite of shortages in government
funding. The value of these developing country-based studies has led to explicit calls in
support of the Triple Helix paradigm, such as Saad and Zawdie’s (2005) detailed
progression of Algeria’s post-independence industrialisation. This is echoed by

107

Etzkowitz and Leydesdorff (2000), who survey related events in Europe, Latin
America, and Asia, particularly the government’s involvement in altering the
relationship between knowledge producers and users.
Turning now to the most relevant case-specific literature, Korea and Taiwan are
typically presented comparatively with other countries having Triple Helix qualities.
Park, et al. (2005) compare Korea and the Netherlands based on “knowledge
infrastructure.” They conclude that Korea’s scientific and technological output is
greater than that of the Netherlands, measured by webometric, scientometric, and
technometric indicators, but offer little in terms of policy prescription. Etzkowitz and
Brisolla (1999) also study Korea comparatively, using Korea and Brazil as proxies for
their respective regions. The authors ultimately make a connection between
technology-bolstering policies and the international political economy, concluding that
intervention in technology policies is no guarantee of success.
This chapter, thus, extends the aforementioned micro-institution-based literature
with a largely unexplored aspect of the Triple Helix paradigm, all the while focusing on
the Korean and Taiwanese cases. The research question under consideration is a
response to an untested hypothesis by Etzkowitz (2003), which states that new social
arrangements and channels of interaction are needed if industry and government are
joined by universities in knowledge-based economies. Earlier studies allude to this
(Faulkner and Senker, 1994), concluding that cooperation between universities and
private firms is based on personal contact.
81
Others have found that found that

81
For Faulkner and Senker (1994), this contact is a function of scientific publications.

108

interactions between university and firm researchers occur through a dense network of
interpersonal relationships (Dierdonck, et al. 1990),
82
while still others have
determined that the source of innovation-based relationships is captured through the
personal contacts of research institution employees (Fritsch and Schwirten, 1999).
The Korean and Taiwanese cases provide an ideal opportunity to test for
relationship-based capital. As was described in the introductory chapter, the Triple
Helix in Korea and Taiwan is rooted in the 1960s and 1970s, during which imitation
was the source of rapid industrialisation. This imitation, or “reverse engineering,” of
existing foreign technologies required minimal investment in R&D for the production
of simple products. Reverse engineering rarely occurs in a vacuum, and multi-level
interactions among firms, universities, and public R&D institutes were required (Kim
and Nelson, 2000).
83

Research Questions & Data
A test of Etzkowitz’s (2003) new relationship-based capital claim is bound to a
number of related research questions which are intended to advance the overall
understanding of the Triple Helix structure. The additional issues included here address
sub-groups of the public and private research sectors. Private sector sub-grouping is
divided between small and medium enterprises (SMEs) and large firms. Ernst (2000)

82
These findings are particular to the Belgian case.

83
Details of subsequent events in these two countries may be found in the introductory chapter of this
dissertation.

109

notes that the chaebol-dominant industry structure in Korea was accompanied through
“octopus-like diversification” into many unrelated industries. Such over-diversification
minimized specialization, which some claim actually hindered the accumulation of
knowledge (Ernst, 2000). As such, SMEs can now be viewed as the vehicle through
which ideas and technologies germinate. In Taiwan, on the other hand, SMEs have
always been the dominant firm model. To test for this difference, cross-sector sub-
group collaborative tendencies will be included as control variables.
Growth in the research efforts of SMEs has some attributes of a structural
change. Kim (2001) identifies the Asian crisis of 1997 as a factor in the growth of SME
innovation in Korea. The chaebols reduced R&D investment following the crisis,
prompting an SME-based upsurge. The number of new firms in Korea also increased
from 100 (pre-crisis) to more than 7,000 in June 2000, as a result of post-crisis layoffs
by the chaebols. This growth in SMEs and new start-ups is also affected by the
targeting of public funds, either by using the new technology as bank loan collateral,
subsidizing R&D personnel, or providing technical information and services (Chung,
2004).
Collaborative tendencies between public and private groups are captured in the
KORTAI R&D dataset. Exactly, collaborative tendencies measure the percentage of
R&D collaboration done by the respondent with the opposite sector, or with a sub-
group of the opposite sector.
84
Looking first at aggregate patterns in Tables 2.1a and
2.1b, the pattern of collaboration in both Korea and Taiwan is similar, as public sector

84
Responses are based on a 0 to 10 scale, representing increments of 10 from 0 to 100 percent.

110

respondents collaborate more with the private sector (45.2 and 30.9 percent,
respectively) in both countries than private sector respondents collaboration with the
public sector (34.4 and 17.3 percent, respectively). The only major distinction between
Korea and Taiwan in this respect is in terms of scale, as both public and private
respondents in Korea engage in cross-sector R&D collaboration approximately 50
percent more than Taiwanese respondents.
Table 2.1a Collaborative tendencies: aggregate level, Korea

Source
Collaborator Percentage
public sector private sector 45.2%
private sector public sector 34.4
Source: KORTAI R&D database
Note: Percentages do not add up to 100 because of possible overlap in collaborative efforts.

Table 2.1b Collaborative tendencies: aggregate level, Taiwan

Source
Collaborator Percentage
public sector private sector 30.9%
private sector public sector 17.3
Source: KORTAI R&D database
Note: Percentages do not add up to 100 because of possible overlap in collaborative efforts.

Collaborative tendencies between each sub-group and the sub-group of the
opposite sector, presented in Tables 2.2a and 2.2b, shows relatively similar patterns
across countries. GRIs in both countries collaborate more with SMEs than with large
firms, while private firms in both countries collaborate evenly with GRI and university
partners. There are only three major distinctions between Korea and Taiwan. First, the
magnitude of private sector collaboration is higher in Korea, which is consistent with
Tables 2.1a and 2.1b. Second, GRIs collaborate slightly more with SMEs in Korea than
in Taiwan, although the overall pattern is similar in both countries. Third, with regard
to the university sub-group of the public research sector, there is virtually no difference

111

in collaborative patterns with SMEs and large firms in Korea (at around 44 percent),
while universities collaborate more with SMEs than with large firms in Taiwan. This
last point is not an entirely robust finding, however, given that the sample size for the
Taiwan university sub-group is not very large. On the whole, however, we can
conclude that SMEs are the preferred collaborative R&D partner for GRIs in both
countries, and that private respondents are relatively indifferent as to whether they
collaborative with GRIs or with universities.
Table 2.2a Collaborative Tendencies: sub-group level, Korea
Source Collaborator Percentage
GRI SME 54.5%
large firm 36.8
university SME 44.3
large firm 44.0
private sector GRI 32.7
university 36.1
Source: KORTAI R&D database
Note: Percentages do not add up to 100 because of possible overlap in collaborative efforts.

Table 2.2b Collaborative Tendencies: sub-group level, Taiwan
Source Collaborator Percentage
GRI SME 47.8%
large firm 33.6
university SME 47.5
large firm 31.7
private sector GRI 17.6
university 17.0
Source: KORTAI R&D database
Note: Percentages do not add up to 100 because of possible overlap in collaborative efforts.

The measures capturing new and pre-existing relationship-based capital, also
available in the KORTAI R&D dataset and detailed below, are designed to show
whether new forms of capital created through Triple Helix-structured collaborations
have a greater effect upon research output than pre-existing forms of capital. Again, our
test is based on Etzkowitz’s (2003) claim that the Triple Helix generates new forms of

112

capital, which is the logical result of having a new, dynamic research-based
relationship with an entity from the opposing sector.
From the KORTAI R&D dataset, new forms of capital generated through
public-private R&D collaboration are measured by the percentage of collaboration
done with partners from previous projects (new_relations).
85
This proxy for new,
relationship-based capital is delineated by a discrete, time invariant value from zero to
ten, measuring percentage values from zero to 100 in increments of 10. The KORTAI
R&D dataset also accounts for eight different reasons for such repartnering, all based
on a 7-point Likert scale response (“7” being greatest): a lack of other qualified
partners (noother); a stipulated funding condition (fundstip); a shared commitment
(sharecom); a lack of tension (lacktension); ease of communication (easecom);
complementarity in knowledge (compknow); the presence of trust (trust); and expected
commercialization (expcom). Interviews between the author and project managers in
Korea confirm that these reasons all have potential relevancy.
The effects of new forms of capital are held up here in comparison to the effects
of pre-existing forms of relationship-based capital.
86
The variable for pre-existing
capital is measured as the percentage of public-private R&D collaboration originating
from pre-existing connections – personal ties (old_relations) – between the respondent
and collaborators from the opposing sector. The variable is formulated by a number

85
See Appendix 2.1 for complete variable notation.

86
The two measures of relationship-based capital – new_relations and old_relations – do not always
equal 100 percent because of overlap in these two categories. Controls for such overlap are provided in
the subsequent analysis.

113

from zero to 10, representing percentage values from zero to 100 in increments of 10.
87

As before, the KORTAI R&D dataset also includes five dummy variables capturing the
sources of the various forms of personal ties: university-based ties (sameuni), former
university laboratory-based ties (sameunilab), former private firm-based ties
(samefirm), ties through working on multiple previous projects (sameproj), and ties
from meeting at a conference (sameconf).
88
A conceptual framework outlining the possible interactions between these
potential new forms of capital and research output is presented on the right-hand side of
in Fig. 2.1, while the interactions between pre-existing forms of capital and research
output are shown on the left-hand side of Fig. 2.1.
89
The dependent variable is R&D
project output, measured by (1) the number of patents strictly from cross-sector R&D
collaboration and (2) the number of total patents excluding cross-sector R&D
collaboration. Consistency among the predicted effects upon these two dependent
variables will confirm that there is no difference in the generation of both types of
patents. On the other hand, use of both measures enables one to perform a
counterfactual test of the new-versus-pre-existing hypothesis: if new capital is truly
important for collaboration in the Triple-Helix structure, the effects of repartnerships

87
Although the measures of new and pre-existing capital may not be wholly exclusive of each other, an
examination of the reasons for repartnering and the sources of personal ties as a robustness check
confirms that this is not a problematic issue.

88
Again, these five categories were selected as a result of the content of interviews held between the
author and research directors in Korea.

89
It should be noted that the two sides of Fig. 2.1 express new and pre-existing forms of capital as
operating separately upon research output, but they can also have simultaneous as well as interactive
effects. These concerns are treated in a series of robustness checks in the subsequent analysis.

114

(new_relations) should be greater for collaborative patents than for non-collaborative
patents.
As evidence of the growing importance of public-private R&D collaboration as
a source of R&D output, Tables 2.3 to 2.6 and Figs. 2.2 to 2.5 show the time trends for
the average number of patents by sub-sector via collaboration and excluding
collaboration, respectively. The units of these averages are computed from the
KORTAI R&D dataset, and effectively illustrate that output is growing across all sub-
groups, with a couple of anomalies. While these tables and figures show important
changes over time, the subsequent empirical analysis is entirely static in nature (i.e.,
patent measurements will reflect data only for 2005), as all of our explanatory variables
are time invariant.
90

90
The rationale for not having time varying explanatory variables is that questionnaire respondents were
not expected to accurately recall the amount of previous participation or personal ties over time.

115

Fig. 2.1 Tracing the effects of personal ties and repartnering
Percentage of
collaboration done
with previous partners
(new_relations)
1) Number of patents
strictly from cross-
sector R&D
collaboration
2) Number of total
patents excluding
cross-sector R&D
collaboration
Ease of
communication
(easecom)
Expectated
commercialization
(expcom)
Complementary
knowledge
(compknow)
No other partner
available (noother)
Trust (trust)
Funding stipulation
(fundstip)
Shared
commitment
(sharecom)
Absence of tension
(lacktension)
Percentage of
collaboration originating
from personal ties
(old_relations)
Met at a
conference
(sameconf)
Same university
ties (sameuni)
Same university
laboratory ties
(sameunilab)
Same firm ties
(samefirm)
Same previous
project(s) ties
(sameproj)

On the whole, the three subsectors in these two countries have followed similar
trends, although there are a couple of anomalies. For the Taiwanese case, the average
number of collaborative patents for universities is much higher after 2002, although
this can be attributed to a couple of outlying cases, not to mention the small sample size
for the Taiwan university sub-group. Another pattern of divergence among the three
groups over time is the increase of average non-collaborative patents by Korean GRIs
from 2001. While this does not necessarily portray an image of a focus on public-
private R&D collaboration in Korea, it does confirm that patenting overall has grown in

116

emphasis. GRIs in Korea are simply not limiting their efforts to cross-sector
collaboration.
Table 2.3 Average number of patents through cross-sector R&D collaboration: by
sub-sector: Korea
1997 1998 1999 2000 2001 2002 2003 2004 2005
GRI .022 .022 .022 .022 .130 .043 .244 .370 .348
university .000 .075 .150 .225 .075 .150 .150 .250 .500
firm .075 .150 .100 .125 .175 .100 .225 .325 .500

Fig. 2.2 Average number of patents through cross-sector R&D collaboration: by sub-
sector: Korea
0
0.1
0.2
0.3
0.4
0.5
0.6
1997
1998
1999
2000
2001
2002
2003
2004
2005
GRI
university
firm

117

Table 2.4 Average number of patents through cross-sector R&D collaboration: by
sub-sector: Taiwan
1997 1998 1999 2000 2001 2002 2003 2004 2005
GRI .000 .000 .045 .227 .023 .023 .114 .205 .477
university .000 .000 .000 .000 .000 .250 .500 1.17 1.33
firm .000 .000 .009 .009 .017 .148 .052 .113 .530

Fig. 2.3 Average number of patents through cross-sector R&D collaboration: by sub-
sector: Taiwan
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1997
1998
1999
2000
2001
2002
2003
2004
2005
GRI
university
firm

118

Table 2.5 Average number of total patents excluding cross-sector R&D
collaboration: by sub-sector, Korea
1997 1998 1999 2000 2001 2002 2003 2004 2005
GRI 0.978 1.809 1.915 2.191 2.383 3.021 3.891 6.532 8.340
university 0.175 0.450 0.725 0.675 1.000 1.125 1.350 1.350 1.750
firm 0.025 0.075 0.050 0.200 0.231 0.615 0.745 0.692 1.462

Fig. 2.4 Average number of total patents excluding cross-sector R&D collaboration:
by sub-sector, Korea
0
1
2
3
4
5
6
7
8
9
1997
1998
1999
2000
2001
2002
2003
2004
2005
GRI
university
firm

119

Table 2.6 Average number of total patents excluding cross-sector R&D
collaboration: by sub-sector, Taiwan
1997 1998 1999 2000 2001 2002 2003 2004 2005
GRI 1.070 1.116 1.116 1.860 0.302 1.023 0.977 1.581 1.405
university 0.000 0.000 0.000 0.000 0.000 0.500 0.333 1.083 1.917
firm 0.026 0.061 0.052 0.783 0.435 0.491 0.596 0.947 1.588

Fig. 2.5 Average number of total patents excluding cross-sector R&D collaboration:
by sub-sector, Taiwan
0
0.5
1
1.5
2
2.5
1997
1998
1999
2000
2001
2002
2003
2004
2005
GRI
university
firm

Empirical Specification & Results
Our first task, the comparing of new forms of capital arising from the Triple
Helix with pre-existing capital, is done through a comparison of the new_relations and
old_relations coefficients’ separate effects upon research output. The first general
model, thus, is
i ik ij
X Y ε + = , (2.1)
where Y is the number of patents generated by respondent i in 2005, and j denotes the
nature of the patent (either accounting only for collaborative projects or excluding such
projects altogether). X represents the relationship-based capital discussed at length in

120

the preceding sections. For respondent i, k is either new capital (new_relations) or pre-
existing capital (old_relations), and ε is the error term for respondent i.
While the simple model in Eq (2.1) considers either new or pre-existing
relationship-based capital (given subscript k), Eq (2.2) tests for their simultaneous
effects
91
:
i i i ij
X X Y ε + + =
2 1
(2.2)
The results of the specification offered in Eq (2.2) are intended to provide a deeper
understanding of the possible effects of relationship-based capital, in contrast to the
restrictions of Eq (2.1).
Among the possible techniques to study the effects of new versus pre-existing
forms of capital in a Triple Helix construct, the ordinary least squares (OLS) statistical
analysis is preferred, given the nature of our dependent and explanatory variables. For
those cases in which j is comprised strictly of collaborative patents, OLS results are
presented in Table 2.7. Columns 1 and 2 reflect Eq (2.1), while column 3 represents Eq
(2.2). For those cases in which j is comprised strictly of noncollaborative patents,
which was determined by taking total patents less collaborative patents, OLS results are
presented in Table 2.8. Robustness checks beyond standard methods include the
inclusion of a number of control variables, as well as a new_relations*old_relations

91
A robustness check for the combined interactive effect of new capital and pre-existing capital was also
conducted, although the results do not have a significant impact. See Appendix 2.2 for details of these
robustness checks.

121

interaction term. All of the results from these checks are available in Appendix 2.1 and
do not present any significant challenges to the results offered in Tables 2.7 and 2.8.
92
The results for the basic hypothesis are presented in Table 2.7. First and
foremost, there are no statistically significant differences between Korea and Taiwan or
between the public and private sectors. Pre-existing relationship-based capital
(old_relations) is positive and significant when accounted for individually (column 2),
as is new relationship-based capital (new_relations; column 1). Both variables continue
to remain positive and statistically significant when considered simultaneously, which
is expected to most accurately reflect the reality for public and private researchers in
these two countries.

92
From the KORTAI R&D dataset, controls are included for industrial sector (see introductory chapter
for details), age of respondent, respondent’s years worked at present job, and respondent’s years worked
in present industry.

122

Table 2.7 OLS results for new and pre-existing capital’s effects upon collaborative
output
--------------------------------------------------------------------
(1) (2) (3)
patcol patcol patcol
--------------------------------------------------------------------
new_relations 0.121** 0.105**
(0.0365) (0.0371)

old_relations 0.1000** 0.0782*
(0.0365) (0.0369)

taiwan_dummy 0.126 0.0117 0.0316
(0.224) (0.229) (0.227)

private_dummy -0.00635 0.0299 0.0424
(0.222) (0.224) (0.222)

Constant -0.00936 0.112 -0.221
(0.225) (0.217) (0.245)
--------------------------------------------------------------------
Observations 299 299 299
R-squared 0.037 0.026 0.052
F 3.789 2.619 3.997
--------------------------------------------------------------------
Standard errors in parentheses
+ p<.10, * p<.05, ** p<.01, *** p<.001

The results of Table 2.8 present regression output for the case in which the
dependent variable is non-collaborative patenting. As it was mentioned above, this
applies a counterfactual test for the primary hypothesis, in that an expected negative
effect of new relationship-based capital on non-collaborative patents reaffirms its
importance in terms of collaborative patent output. As Table 2.8 shows, these
expectations were not met. In fact, the effect of pre-existing relationships
(old_relations) upon these two forms of patenting seems to support the conclusion that
personal ties are most important for public-private R&D collaboration. Pre-existing
personal ties have a positive (and significant) effect on collaborative patenting, but they
have a negative (and insignificant) effect on non-collaborative patenting. New
relationship-based capital in the form of previous participation, on the other hand, has

123

positive effects for both collaborative and non-collaborative output. This dual
importance is attributed to the increased output potential arising from repartnerships
between the public and private research sectors. Indeed, in tandem with the results of
Table 2.7, the results of Table 2.8 confirm that new relationships are a valuable source
of increased patenting, regardless of whether the patenting is done with research
entities of the opposite sector. Still, the fact that pre-existing relationships function
primarily to generate public-private R&D collaborative output is a conceptual
challenge, requiring further investigation.

124

Table 2.8 OLS results for new and pre-existing capital’s effects upon non-
collaborative output
--------------------------------------------------------------------
(1) (2) (3)
nocollpat nocollpat nocollpat
--------------------------------------------------------------------
new_relations 0.724+ 0.785+
(0.415) (0.424)

old_relations -0.127 -0.291
(0.414) (0.422)

taiwan_dummy -0.346 -0.146 0.00387
(2.542) (2.604) (2.595)

private_dummy -3.145 -3.420 -3.326
(2.520) (2.546) (2.536)

Constant 3.840 7.127** 4.628+
(2.557) (2.466) (2.803)
--------------------------------------------------------------------
Observations 299 299 299
R-squared 0.017 0.007 0.019
F 1.728 0.736 1.413
--------------------------------------------------------------------
Standard errors in parentheses
+ p<.10, * p<.05, ** p<.01, *** p<.001

Tables 2.9 and 2.10 are variations on these earlier results, confirming that new
relationship-based capital (new_relations) has a positive effect for each sector.
93
The
coefficient is not always statistically significant, as it is in the aggregate case presented
in Tables 2.7 and 2.8, but the pattern is consistent. At this sub-group level, Table 2.10
attempts to account for the effects of collaborative tendencies upon collaborative and
non-collaborative patenting output. This is, thus, a direct attempt to empirically apply
some of the Triple Helix’s theoretical shortcomings. For the public sub-groups
(columns 1 and 3 of Table 2.8), collaboration with SMEs (coll_sme) and with large
firms (coll_large) positively predicts patents, but is not highly statistically significant.

93
The only difference between the specifications in Tables 2.9 and 2.10 is the dropping of the private
dummy variable.

125

(These variables are all assigned values from zero to ten, measuring the percentage
values from zero to 100 in increments of 10.) For the private sub-groups (columns 2
and 4 of Table 2.10), the only statistically significant result is the percentage of
collaboration with universities (coll_uni), which has a positive and significant effect on
collaborative patents.
Table 2.9 OLS results for new and pre-existing capital’s effects upon collaborative
and non-collaborative output, by sector sub-groups
------------------------------------------------------------------------------------
Public sample Private sample Public sample Private sample
(1) (2) (3) (4)
patcol patcol nocollpat nocollpat
------------------------------------------------------------------------------------
new_relations 0.0340 0.156** 1.448+ 0.326
(0.0453) (0.0563) (0.847) (0.336)

old_relations 0.0235 0.116* -0.555 -0.156
(0.0448) (0.0563) (0.839) (0.336)

taiwan_dummy 0.199 -0.0991 4.110 -4.164+
(0.253) (0.375) (4.739) (2.243)

Constant 0.211 -0.414 1.468 5.571*
(0.260) (0.399) (4.871) (2.382)
------------------------------------------------------------------------------------
Observations 143 156 143 156
R-squared 0.014 0.091 0.025 0.029
F 0.656 5.073 1.177 1.539
------------------------------------------------------------------------------------
Standard errors in parentheses
+ p<.10, * p<.05, ** p<.01, *** p<.001

126

Table 2.10 OLS results for new and pre-existing capital’s effects upon collaborative
and non-collaborative output, by sector sub-groups
------------------------------------------------------------------------------------
Public sample Private sample Public sample Private sample
(1) (2) (3) (4)
patcol patcol nocollpat nocollpat
------------------------------------------------------------------------------------
new_relations 0.0122 0.154** 0.992 0.415
(0.0480) (0.0568) (0.895) (0.342)

old_relations 0.0324 0.103+ -0.389 -0.111
(0.0453) (0.0562) (0.844) (0.338)

coll_sme 0.0403 0.0823
(0.0401) (0.748)

coll_large 0.0550 1.339+
(0.0433) (0.808)

coll_uni 0.175* -0.514
(0.0841) (0.506)

coll_gri -0.0687 -0.389
(0.0737) (0.443)

taiwan_dummy 0.230 0.142 4.773 -5.787*
(0.254) (0.413) (4.743) (2.486)

Constant -0.154 -0.774 -3.072 8.237**
(0.373) (0.491) (6.959) (2.954)
------------------------------------------------------------------------------------
Observations 143 156 143 156
R-squared 0.028 0.118 0.045 0.044
F 0.800 4.002 1.293 1.394
------------------------------------------------------------------------------------
Standard errors in parentheses
+ p<.10, * p<.05, ** p<.01, *** p<.001

A number of robustness checks help establish whether there are persistent
cross-national and cross-sectoral differences. To accommodate the hypothesis that new
relationship-based capital arises from Triple Helix constructs, such as public-private
R&D collaboration, the percentage of collaboration done with partners from previous
projects (new_relations) is the dependent variable and, given that larger values of
previous participation correspond with higher percentages of participation, an ordered
logit model is adopted.
94
By country and sector, age of respondent is shown to have a

94
The statistical output of these results is not included here, and the results described are in log-odds
form.

127

negative impact upon the degree of repartnering in all cases except for the Korean
public sector; i.e., the older the older the researcher, the less likelihood of repartnership.
It is found that repartnering is positively affected by job experience for Korea,
particularly the public sub-group. For Taiwan, however, years of job experience
negatively predicts the tendency for repartnership with former public-private R&D
collaborators. Another distinction between Korea and Taiwan is the predicted effect of
different research emphases upon the propensity to repartner.
95
In Korea, basic research
emphasis is more likely to generate repartnering in Korea, but not in Taiwan. In both
countries, there is evidence that applied research emphasis is more likely to generate
repartnering in the private sector than in the public sector. This is consistent with
assumptions about differing research emphasis between research sectors.
Complementarity between sectors, however, has a uniform positive, significant
predicted effect for all sub-groups.
The richness of the KORTAI R&D dataset allows us to examine details of
relationship-based capital even further. With reference to Fig. 2.1, the sources of new
and pre-existing capital provide enable us to understand how and why each sub-group
in Korea and Taiwan is repartnering or utilizing personal ties with collaborators. Rather
than base this examination on the summary statistics for the various sources of
repartnering and personal ties, weights are assigned. That is, new_relations is weighted
by each reason for repartnering to indicate the precise amount of
impact:( , where is the nth reason for repartnering for )
ni i
Z relations new × _
n
Z

95
See the introductory chapter of this dissertation for details on the nature of different research emphases.

128

respondent i. In this way, if a reason is strong but repartnering in weak, the reason is
given less weight. The rankings of these weighted reasons are presented in Tables 2.11
and 2.12 for Korea and Taiwan, respectively.
Table 2.11 Rankings of weighted reasons for repartnering: Korea
Rank GRI university firm
1 trust
27.91

trust
27.42

trust
28.41

2 expcom
26.35

easecom
24.36

expcom
26.04

3 noother
25.32

compknow
24.00

compknow
25.17

4 compknow
25.26

expcom
22.19

easecom
24.45

5 easecom
24.47

noother
21.14

noother
21.69

6 fundstip
21.88

fundstip
20.69

sharecom
21.66

7 sharecom
22.15

sharecom
19.83

fundstip
21.17

8 lacktension
19.56
lacktension
18.42
lacktension
19.37

129

Table 2.12 Rankings of weighted reasons for repartnering: Taiwan
Rank GRI university firm
1 trust
29.43

trust
21.50

compknow
26.49

2 expcom
26.83

easecom
21.50

trust
26.32

3 compknow
26.06

expcom
21.25

expcom
25.36

4 sharecom
23.89

compknow
19.75

easecom
23.24

5 easecom
22.78

sharecom
14.75

sharecom
19.15

6 fundstip
18.77

fundstip
14.00

fundstip
18.68

7 noother
15.64

lacktension
11.71

noother
16.39

8 lacktension
12.00
noother
9.71
lacktension
14.45

Consistent results across all three sub-samples include the dominance of trust,
and the relative unimportance of fundstip and lacktension. The low ranking of fundstip
is an indicator that other repartnering forces are at work, which has implications for the
conclusions offered in Chapter 1 of this dissertation. A subtle distinction between these
two countries is that sharecom lacks relative emphasis in Korea while, in Taiwan,
noother is ranked among the bottom three reasons for repartnering. What is
unmistakable is that trust is of great importance for both of these countries and across
all research groups, particularly given that trust outranks expected commercialization
of results (expcom) for the private sectors in Korea and Taiwan.
A similar weighted ranking system is also constructed for the nature of personal
ties, although the dichotomous nature of the variables prevents a similarly scaled

130

measurement. The Likert scale response for old_relations is weighted with each
dummy variable (i.e.,[] sameuni relations old × _ , [ ] sameunilab relations old × _ , etc.),
generating the mean values for each relevant group, which are then ranked. What is
most notable among these results for the Korean case, presented in Table 2.13, is the
consistent, high ranking of sameproj for all Korean groups. For the Taiwanese case,
presented in Table 2.14, personal ties arising from attending the same conference is
ranked highest, for GRIs and universities. Taiwanese firms (Table 2.14) are the only
exception, as personal ties arising from university (sameuni) and university-lab
(sameunilab) are ranked highest. This is the likely result of the strong university-based
firm incubation programs in Taiwan.
Based on the weighted rankings of the source of personal ties in Tables 2.13
and 2.14, one can conclude that researchers in Korea and Taiwan are largely
establishing collaborations not through university or firm connections, but through
conferences and work on previous projects.
96
This ranking pattern of personal ties
actually corresponds with repartnering, lending additional support for the “new capital”
hypothesis of Etzkowitz (2003). The overall impression from the high ranking of
personal ties based in former projects (sameproj) particularly shows that repartnerships
are widespread.
97

96
This is correlated with the results from Chapter 1 of this dissertation, which showed that information
transfers through conferences have a positive and significant effect upon the generation of collaborative
patents.

97
Despite this overlap between new_relations and old_relations, there are still clear differences between
these two measures, given the opposite signs of old_relations for collaborative and non-collaborative
patents.

131

Table 2.13 Rankings of weighted source of personal ties: Korea
Rank GRI university firm
1 sameproj
1.96

sameproj
2.25

sameproj
2.08

2 sameconf
1.26

sameconf
1.70

sameconf
1.20

3 sameunilab
1.09

samuni
1.18

sameuni
0.88

4 sameuni
0.38

samefirm
0.85

sameunilab
0.48

5 samefirm
0.33
sameunilab
0.85
samefirm
0.18

Table 2.14 Rankings of weighted source of personal ties: Taiwan
Rank GRI university firm
1 sameconf
3.36

sameconf
4.12
sameuni
1.91
2 sameproj
2.07

sameproj
2.25
sameunilab
1.41
3 sameuni
1.36

sameunilab
1.83
sameconf
1.36
4 sameunilab
1.27

sameuni
1.75
sameproj
1.33
5 samefirm
0.91
samefirm
0.00
samefirm
1.07

Conclusion
Theoretical and methodological contributions have been offered here in an
attempt to bolster an understanding of the Triple Helix paradigm in Korea and Taiwan.
This discussion has confirmed earlier claims which detail the benefits of the Triple
Helix structure, particularly the increase of new forms of capital (Etzkowitz, 2003).
There are a number of qualifiers included here, however, which make it clear that

132

caution must be exercised when breaking down the components of the paradigm on a
number of levels.
The evidence provided in this discussion supports the “new capital” hypothesis
of Etzkowitz (2003) with regard to the Triple Helix structure. Two different dependent
variables (collaborative patents and collaborative publications) are applied to compare
the effects of this new, relationship-based capital upon the Triple Helix and related
methods of R&D. The results clearly show that there is a positive impact of new,
relationship-based capital on patenting, and that there are homogenous effects between
countries and research sectors.
98
What is most notable among the OLS results from the
model presented in Fig. 2.1 and Eqs (2.1) and (2.2) is that personal ties – pre-existing
relationship-based capital – are important for public-private R&D collaborative output
but not for non-collaborative output. This does not necessarily invalidate the
importance of new capital, but it does show that personal ties are most important in
terms of their effects on collaborative output. As it has been shown that personal ties
most strongly reflect work on previous projects, this result does not challenge the “new
capital” hypothesis. Indeed, it provides further validation.
From a policy perspective, it would appear prudent to facilitate those
opportunities for public and private research entities to work together, such as the
funding programs included in the KORTAI R&D dataset. Based on the ranked and
weighted reasons for repartnering, R&D funding policies were found to be less

98
Robustness checks show that there are a number of subtle differences between the Korean and
Taiwanese cases, but these have a negligible effect on patenting output.

133

important than other reasons for repartnering. This does not rule out the possibility that
policy-directed cross-sector R&D collaboration is still a necessary catalyst. In other
words, other reasons for repartnering become salient for the researcher over time. In
this way, policies have a direct effect for the initial partnership and an indirect but still
important effect for repartnering. There are no grounds for cutting public funding, by
any means.
Returning now to the distinct patterns of networking which are evident in East
Asian countries such as Korea and Taiwan, there is a distinction, albeit hazy, between
the effects of new, relationship-based capital and the effects of “traditional” forms of
networking. Traditional organization structures are typically comprised of familial and
alumni ties (Biggart, 1990), where work-related rewards depend on how long and well
one serves the organization rather than individual achievements, and employees are
hired through personal recommendation, family background, education, and
examination scores. (Rozman, 1992). It was precisely the measurement of these
characteristics which was attempted through the data on the sources of personal ties.
Given that the weighted sources of personal ties reveal that university-, university
laboratory-, or firm-ties are relatively less important than relationships which are
established through conference attendance and/or former projects,
99
traditional
networking structures in Korea and Taiwan are superseded by the new relations
highlighted in the Triple Helix paradigm. In addition, the importance of trust as a

99
The only exception here are the high ranking of university ties for Taiwanese private sector, albeit the
weighted score is not high, in relative terms.

134

reason for repartnering represents a major transition from traditional social norms,
given that R&D partnerships are done beyond traditional conventions. It has been
determined, for example, that Korea has a relatively low level of trust in comparison to
other developed nations, such as Japan and Germany (Kim, 2000). Some claim that this
causes low R&D productivity because of poorly coupled links between research
entities (Kim, 2000). In this context, there is considerable added-value from applying
the Triple Helix paradigm to R&D efforts.
Returning to the robustness checks, the effect of age on the propensity to
repartner provides some insight into a change R&D model in Korea and Taiwan. In line
with von Hippel (1988), who emphasizes that the duration of a relationship is positively
linked to the degree of knowledge shared between partners, the age of the researcher
diminishes the likelihood of repartnerships. If age were a factor, one could assume that
older researchers would have a more established network of colleagues with whom s/he
could engage in public-private R&D collaboration. Ultimately, this is further
confirmation of the claim that S&T trumps culture in Korea and Taiwan.
There is still room, however, to interpret the above results as conforming to
traditional norms, in the context of the developmental state discussion offered in the
introductory chapter. Most discussion of the East Asian-Western distinction is couched
in the idea of patrimonialism, where rule by a patriarch and administered by a personal
staff has been the key mechanism to maintain social order, subordinate the self, and
exercise individual self control (Biggart, 1990). In this context, the heavy involvement
of the state is the logical result of an indigenous, partrimonial framework. Despite

135

research which correlates the diminishing function of these traditions with increases in
growth (Park, 2000), the patriarchal function of the government is still evident as it
provides R&D funding in order to advance the country’s science and technology levels.

136

Chapter 3: A Cross-National Study of Governance and the Sources of Innovation:
The Determinants and Effects of International R&D Collaboration
Introduction
R&D collaboration between countries is just one of several available forms of
international coordination. More standardized forms include bilateral or multilateral
trade talks, programs with supra-national (IMF or World Bank) organizations, or
coordination with other countries’ sub-national bodies (e.g., firms). International R&D
collaboration is distinct from these other forms because the transferred product is not
always tangible, the direction of the transfers is not clearly delineated, and there is the
possibility that both (or all) collaborating countries receive benefits. On the whole, the
increase in international R&D collaboration over the last twenty to thirty years has not
seen a concurrent shift in research on this phenomenon. There is definitely not a
complete vacuum in the literature, as it is shown below, but a macro-level study is
lacking.
The initial task here is to show that international R&D collaboration is
significant for a country, at least on a number of levels. On a descriptive level, it is
clear that this phenomenon is on the rise, particularly among developing countries. In
terms of standard growth accounting methods, however, it remains to be seen whether
such collaboration is a predictor of a nation’s level of income. As opposed to income
levels, a more appropriate dependent variable is the residual from national accounting,
which is commonly termed “technology” or total factor productivity (TFP). If
international R&D collaboration is indeed important, especially for countries which

137

have lower levels of GDP, lower levels of physical capital inputs, and lower levels of
human capital inputs, the second task of the paper is to establish the impact of relevant
institutions. The dependent variable now becomes international R&D collaboration and
the explanatory variables are the strength of intellectual property rights (IPRs) and level
of democracy. The decision to engage in cross-border R&D collaboration is ultimately
a matter of political and economic institutions which have a direct impact upon the ebb
and flow of intellectual property, so IPRs and democracy play a significant role.
There is much potential valued added to be gained from these two lines of
analysis. With regard to the economic growth literature, over the long-term,
neoclassical growth models have failed to adequately explain the persistence of non-
steady state interpretations of economic growth. The common explanation is often the
exogenous TFP factor, although there have been some who look specifically at IPRs
and growth.
100
Here, technology and its foundations are the focus and full attention is
devoted to providing an explanation for fluctuations in the growth accounting residual.
In this way, we are not restricted to the strong accumulationist assumptions inherent in
the neoclassical model, which maintain that a country’s output is determined by
increases in physical and human capital,
101
but are aligned with assimilationist
approaches to technology-based growth. The assimilationist approach is best
exemplified by the East Asian case, where the receipt of technology from abroad was

100
See Markusen (1998) for details.

101
See Nelson and Pack (1998) for further details on these two sets of theories.

138

complemented by extensive domestic efforts to understand and improve the technology
(Pack 2001) .
Assimilationist effects may not necessarily be the norm. If we make the
assumption that countries fall into high and low patenting research producing countries
– a technological North and South delineation – the common belief is that high
producing countries innovate and are imitated by the low research output countries.
Chin and Grossman (1988) identified this as a tension between the two types of
countries and offered an explanation in terms of strength of IPRs: weak IPRs enable
greater levels of imitation. Chin and Grossman’s (1988) model does not address other
avenues through which technology can be transferred from the technological North to
the technological South. Others provide alternative explanations for imitation, such as
Lai’s (1998) analysis of IPR strength and foreign direct investment (FDI) by
multinational firms.
102
In this context, our study of international R&D collaboration
provides a potentially useful addendum.
A preliminary analysis shows that the growth of overall patenting is occurring
at a near-exponential rate (Fig. 3.1, histogram), with nearly 27,000 in 1975 and over
91,000 in 2000.
103
In an effort to establish a benchmark for the most research
productive nations, the five most patenting countries are henceforth categorized as the
top tier of patenting. As the rankings in patenting change over time, our definition of

102
Lai’s conclusions on the strength of IPRs in the global South are very important here. Previously,
stronger IPRs have been found to be positively related to innovation when FDI, not imitation, is the
technology transfer method.

103
Patent data for the duration of this analysis is limited to the United States Patent Trademark Office
(USPTO).

139

the “Tier-1” countries also changes over time. The U.S., Japan, and Germany are
constants in the Tier-1 category, but France and the U.K. have been replaced by
Taiwan and Korea beginning in the mid-1990s.
104
Indeed, this is even more evidence of
the East Asia technological growth phenomenon, given the abbreviated time span of
this analysis.
Fig. 3.1 Total patents and the percentage represented by Tier-1 countries
0
100000
200000
300000
400000
500000
600000
700000
800000
900000
1000000
1975-79
1980-84
1985-89
1990-94
1995-99
2000-04
0
10
20
30
40
50
60
70
80
90
percent
Total patents
Tier-1 share

Source: USPTO (2008).

The Tier-1 countries also provide a condition for studying international R&D
collaboration. Rather than count each and every instance of R&D collaboration
between or among countries, we focus strictly on those patents which are generated
between non-Tier-1 and Tier-1 countries. This is justified on three grounds. First, it

104
This point should be of great interest to students of East Asian political economy. Taiwan and Korea’s
strong patenting efforts have enabled these two countries to replace the U.K. (by Taiwan, from 1995) and
France (by Korea, from 2000) in the Tier-1 category.

140

facilitates data collection .
105
Second, the Tier-1 countries represent a large component
of overall patenting; roughly three-quarters of all patents issued (Fig. 3.1, line plot).
Third, with an underlying focus on development, collaboration with the Tier-1 research
leaders is a legitimate measure of international R&D collaboration. This is particularly
true in the context of the Nelson-Phelps catch-up framework (Nelson and Phelps 1966),
where it is shown that increases in education and skills leads to convergence in
growth.
106
Granted, the conceptual leap from education levels to R&D output is not
small; yet, there is reason to believe that R&D collaboration with Tier-1 countries
might well serve as a means of achieving such convergence.
An early attempt to examine internationalization and technological
collaboration was made by Dodgson (1993), where he describes triadic collaboration
among the U.S., Europe, and Japan. Applying a case study approach, Dodgson
concludes that collaboration is done primarily by firms to increase skills or learning
opportunities, and that publicly promoted R&D collaboration generates little cross-
border R&D collaboration, witnessed for example by IBM’s many failed attempts to
join the Europe’s public-funded ESPRIT program.
107
Given that public funding is
meant to generate returns which maximize domestic benefit, technological spillovers to
other countries resulting from public funding may be deemed counterproductive. At the

105
The greatest limitation of the USPTO database is its search parameter cap, especially when U.S. state
codes must be included instead of an aggregated “U.S.” country code.

106
There are, of course, parallel areas of research to Tier-1 collaborative research, especially the effects
of geographic proximity, but this is a topic which is not addressed directly here.

107
The ESPRIT program (European Strategic Program for Research and Development in Information
Technology) was founded in 1983 for a ten year duration.

141

same time, the flow of information could very well work in favor for technology-
trailing countries. This point is left unexplored, and for the duration of this macro-level
study, public policies promoting international R&D collaboration are held constant. Of
course, this does not preclude the possibility that a government may indirectly affect
that country’s propensity for international R&D collaboration by affecting factors
deemed favorable to such R&D collaboration such as its governance and the strength of
IPRs.
Another literature contextualizing the importance of international R&D
collaboration relates to international technology transfer. Such transfers are peripheral
to our understanding of international R&D collaboration, as they are traditionally
applied in the development literature in terms of their uni-directional transfer methods.
International R&D collaboration, on the other hand, represents transfers which are bi-
directional and are assumed to be mutually beneficial for all parties involved. The uni-
directional modes of technology transfer offered by Kim (1999), presented in cells 1-4
in Table 3.1, are divided by market/nonmarket mediation and the active/passive role of
foreign suppliers.
108
Importantly, Kim concludes that informal mechanisms are equally
if not more important than formal mechanisms, most noticeable when the technology

108
Also important are the three stages in which Kim (2000) divides up the technology transfer process:
The first stage is the local development of production processes via the acquisition of “packaged” foreign
technology. Such technology is packaged in the sense that all needed factors are provided for the
production of standardized products, including assembly processes, product specification, production
know-how, technical personnel, and parts. In the second stage, production and product design are
diffused within the country, thereby increasing competition within the market and generating local
technical efforts to assimilate the products. The third stage of technology transfer – the gradual
improvement of mature technologies – requires an emphasis upon export promotion and enhanced
capabilities of local scientific and engineering human resources.

142

receiving country is endowed with high levels of absorptive capacity. Cell 5 in Table
3.1, which is our contribution to Kim’s (1999) framework, is unique in that both
foreign and domestic entities actively engage in the generation of new technology. In
this way, international R&D collaboration functions as a bridge between models of
incoming and outgoing technology transfer.
Table 3.1 Modes of international technology transfer

Active Role of
Foreign Suppliers
Passive Role of
Foreign Suppliers
Active Role of Foreign &
Domestic Suppliers
Market
Mediated
FDI, foreign licensing, turn-
key plants, technical
consultancy, made-to-order
machinery Cell 1
Standard (serial) machinery
purchase

Cell 2
Nonmarket
Mediated
Technical assistance by
foreign buyers, technical
assistance by foreign
vendors
Cell 4
Imitation (reverse
engineering), trade
journals, technical
information service
Cell 3

International R&D
collaboration
Cell 5

There is an even more salient literature on the subject of international R&D
collaboration and TFP. Frantzen (2002), focusing on a number of OECD countries over
time, finds that both international and domestic R&D spillovers increase TFP for large
economies. Frantzen, however, does not control for the expected positive correlation
between domestic R&D intensity and the propensity for international R&D
spillovers.
109
Park (2004) also explores the effects of domestic and foreign
manufacturing R&D upon TFP for fourteen OECD countries but includes also Korea,
Taiwan, and Singapore. In this case, Park identifies international R&D spillovers from

109
Direct measures for domestic R&D intensity are available through the OECD’s MSTI dataset,
although this data is limited to the OECD and a small number of additional countries. To account for
strictly domestic R&D output, we use patenting in the absence of international R&D collaboration with
Tier-1 countries.

143

foreign manufacturing research efforts, although the primary method of analysis is
tracing trade flows and outsourcing between countries’ manufacturing and
nonmanufacturing sectors. Overall, these two TFP-based analyses offer conclusions
only for the world’s most R&D-productive countries. As well, international technology
diffusion measures are based largely on input-output tables, which requires the
assumption that all countries are equally qualified to engage in imitation and/or reverse
engineering.
In terms of the growth accounting literature, research attempts beyond neo-
classical methods have been strong in the last ten to fifteen years, particularly since
Jones’ (1995) focused tests for AK-style and endogenous growth models. Much of the
theoretical framework is based on Jones’ addendums to early R&D-based growth
models, particularly Romer’s (1990) work. There are also applications of more recent
work, such as Barro and Sala-i-Martin’s (2003) endogenous models of technological
progress and diffusion via a dynamic panel dataset, somewhat akin to the approach
taken here. Again, we acknowledge that the various sources of international technology
transfer listed in Table 3.1 are important.
110
However, international R&D collaboration
is most consistent with the endogenous growth literature, given the domestic
capabilities and policies which facilitate such collaboration.

110
Keller (2002) has conducted a couple of important studies on the international transfer of technology,
showing that trade is the operative mechanism through which R&D-driven growth occurs. This trade-
related issue is of tertiary importance here, however.

144

The literature is also replete with studies of “institutional effects” which are
fascinating and valuable on a number of levels,
111
but they are commonly applied as
explanatory variables of income and/or growth. An important study of the effects of
institutions and government policies upon growth in the literature was that conducted
by Hall and Jones (1999), who conclude that high rates of investment in physical and
human capital are a positive function of a country’s “social infrastructure.” Building on
the work of North (1990) and others, the authors measure this key variable by those
institutions which encourage productive activity, limit the diversion of resources and
rent-seeking, and protect property rights. Two indices are combined to create a measure
of social infrastructure, which is then held up at the macro level to GDP per capita.
112

In this paper, as it has been detailed already, institutions are not applied directly to
inputs and productivity variables, but are given a much narrower role. Shown on the
right-hand side of Fig. 3.2, key institutions are studied with regard to their most
immediate impact, rather than their generalized effects. While the model shares a
number of similarities with Hall and Jones (1999) (Fig. 3.2, left-hand side), the
specification is much more restricted.

111
This fascination is largely referring to the varied forms and uses of institution-based data, the national
and regional breakdowns employed, and the long-term trends. Indeed, those patterns persisting over time
offer strong evidence that there are forces at work which are effecting change.

112
The two combined indices are the Index of Government Antidiversion Policies (accounting for
expropriation risk, contract enforcement, government corruption, law and order, and bureaucratic
quality) and the Sachs-Warner index of trade openness (a composite measure based on the degree of
tariffs, non-tariff trade barriers, black market premiums, socialist orientations, and government
monopolization of major exports).

145

Fig. 3.2 The Hall and Jones (1999) and international R&D-specific institutional
analyses

To address the issue of imitation, Helpman (1993) produced a general
equilibrium model to test the effects of IPRs in a North-South framework but the global
R&D landscape has changed since his study was published. It is agreed that the North
and South may be distinct in terms of the degree of schooling and capital investment.
On the other hand, the practice of innovation and R&D is occurring in many more
countries than at the time of Helpman’s (1993) study. Lai (1998) also makes a relevant
contribution to the discussion of IPRs in a North-South framework, concluding that the
strength of IPRs in the South is a function of how production is transferred from the
North to the South. When FDI is the primary channel of production transfer, IPRs in

146

the South are positively related to innovation. Lai finds the opposite to be the case
when production is transferred through imitation. These are crucial concerns for both
the first and second stage analyses of this paper.
The remainder of this paper is divided into four sections to address the
aforementioned issues of TFP, international R&D collaboration, and institutions. In the
next section, a simple R&D-based growth model is presented. The residual from this
model, again, constitutes the primary dependent variable of this analysis. Also in this
section, more specific details are presented about how institutions can be modeled as
explanatory variables of international R&D collaboration. The following section details
the dynamic panel dataset and the empirical specifications for the two stages of
analysis. This followed by a section presenting the empirical results and a concluding
section interpreting the results and suggesting future avenues of related research.

Modeling International R&D Collaboration & Institutions
Growth Accounting
We assume a constant returns to scale production function,
α α −
=
1
) (AH K Y , (3.1)
with GDP per capita, Y, physical capital flows, K, A or TFP, and human capital flows,
H, where the capital share is set at one-third and the labor share (AH) is set at two-
thirds for all developed countries.
113,114
For all remaining developing countries, shares

113
These estimates of 1/3 and (1-1/3) are in conformity with Benhabib and Spiegel (2002).

114
Developed countries are defined by membership in the OECD in 1970.

147

of physical capital and human capital flows are set at one-half. This is consistent with
the stylized fact that human capital investments are given considerably less emphasis in
developing countries. The use of the Cobb-Douglas formulation is consistent with the
majority of all R&D-based growth models, such as Romer (1990) and Jones (1995,
1999). Rather than look at the rates of change between each time period, we examine
the dynamic differences of physical and human capital in each particular time period.
Steady state values are applied, given the basic assumption that countries converge to
their steady state levels of growth if they are not already there.
115

As commonly expressed in the literature, A or TFP represents labor-augmenting
technology, but it is not assumed to grow exogenously. The importance of this residual
effect has been widely documented. Hall and Jones’ (1999) look beyond standard
neoclassical growth accounting techniques and emphasize cross-country differences in
the growth residual. Second, they treat a number of key factors relevant to the present
discussion, such as regional influences and institutional make-up. Given Hall and
Jones’ findings, one can include technology-related measures as proxies for the residual.
We effectively substitute for TFP by the log value of the number of per capita
international collaborative patents to capture some of the residual effects on GDP per
capita, assuming that patents function in the overall economy to generate income,
perhaps through licensing. As well, patents can provide a foundation for subsequent

115
A strong challenge to the convergence hypothesis is made by Pritchett (1995). Easterly and Levine
(2001) also claim that the concept of steady-state growth applies to industrial countries only. Our
revision of the growth accounting parameters for developing countries is an attempt to treat such
differences between developed and developing countries.

148

research (i.e., high levels of patenting engender high levels of patenting), which takes
on the qualities of a virtuous cycle of growth. Attempts to model this directly are not
made here, however, as the dependent variable of the first-stage analysis is TFP rather
than income level.
116
Educational attainment, H, is generally expressed in one of two forms in the
literature: as a measure of school enrollment or as a measure of average years of
schooling attained. Barro and Lee (2000) have the relevant data for both of these
measures, and our basis of educational attainment is the average number of years of
schooling for a country’s population aged fifteen or older, attained in each time period.
Benhabib and Spiegel (1994, 2002) have demonstrated that these average levels of
human capital are significant predictors of per-capita income growth, but that growth
rates thereof are not.
It is generally assumed that individuals accumulate human capital by learning
new skills while not working.
117
We apply the Mincerian formulation and consider

116
Further justification for substituting patenting for TFP is rooted in the concept of absorptive capacity.
Although absorptive capacity is typically considered at the level of the firm, it is assumed here that the
exploitation of existing technology has had less effect upon industrial growth than the ability to tap and
absorb technology. As Cohen and Levinthal (1990) indicate, however, such abilities are conferred by
prior knowledge to identify, value, and assimilate new information originating outside of a particular
country. There is, thus, a degree of overlap between previous and subsequent successful international
R&D collaborations (measure via collaborative patents) and absorptive capacity. As well, if we accept
Saggi’s (2002) point that absorptive capacity is a function of adequate levels of human capital and
investments in R&D, there may be a degree of collinearity between the standard growth inputs and
patent output. As the following model shows, human capital is treated in a labor-augmenting fashion,
and R&D investments are but a portion of overall physical capital investment, which diminish the
concerns about multi-collinearity.

117
We will similarly adopt this assumption, although there are also considerable effects of informal
education and/or education while working (on-the-job training, in-servicing, etc.).

149

human capital as a function of s, average years of schooling of the total population aged
fifteen and over,
118
and ψ, the returns to schooling each year:
L e H
s ψ
= . (3.2)
L is the labor force, which should be defined as the population of a country over age
fifteen, but since the correlation between the size of the entire population and the total
population aged fifteen and over is sufficiently high, we assume that L can be proxied
by the population.
119
We will assume that ψ is 0.10, or ten percent, which is consistent
with the estimates provided by Bils and Klenow (2000).
Capital stock, K, is shown in the standard capital accumulation equation,
K I K
K
δ − =
&
, (3.3)
where the change in physical capital stock is the difference between
K
I , the amount of
investment in physical capital, and K δ , the depreciation of capital. If we divide both
sides of Eq (3.1) by AL, defining y, k and h as Y/L, K/L and H/L, respectively, we have
α
α
−
⎟
⎠
⎞
⎜
⎝
⎛
=
1
h
A
k
A
y
. (3.4)
Using (~) notation to show the ratio of a variable to A or TFP,
α α −
=
1
~
~
h k y , (3.5)

118
The use of an infinite-life construct of human capital contradicts the findings of Mincer (1974), as
noted by Klenow and Rodriguez-Clare (1997). At this stage, we assume that people complete their full
time schooling before beginning full time work.

119
Use of the total population rather than the labor force is the result of uniform data availability for a
number of countries.

150

we may rewrite the standard capital accumulation equation in terms of the capital-
technology ratio, shown as
k n I k
K
~
) (
~
δ γ + + − =
&
, (3.6)
where, for a particular period, n is geometric mean of the average growth of the labor
force as defined above, γ average TFP growth and δ is a fixed depreciation rate.
Several studies have been done with varying fixed depreciation rates in a similar
formulation, generating comparable results. For example, Benhabib and Spiegel (2002)
assign δ a value of 0.03, and Jones (1997) assigns γ a rate of 0.03 for developed
countries and 0.01 for less developed countries. We apply these rates to our empirical
analysis, with developed countries defined by OECD membership status in 1970.
Our model is based on the assumption that there is an initial level of capital
accumulation from which capital stock shifts. Since our focus is on the entire period
from 1975 to 2005, the initial 1975-1979 five-year period and each subsequent five-
year period are calculated as flows by the steady-state level of physical capital,
presented in Eq (3.8):
α α
δ γ δ γ
−
+ +
=
+ +
=
1
~
~
~
h k
n
I
y
n
I
k
K K
, as , (3.7) 0
~
= k
&
h
n
I
k
K
α
δ γ
−
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
+ +
=
1
1
*
~
. (3.8)
After substituting this into Eq (3.5) and dividing out
A
y
y =
~
, we have

151

) ( ) (
1
t hA
n
I
t y
K
α
α
δ γ
−
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
+ +
= . (3.9)
Eq (3.9) denotes the level of income as a function of the accumulation of factors of
production, and is no different from the formulation of Jones (1997).
The theoretical specification presented in Eqs (3.1) to (3.9) treats technological
change, A, as exogenously determined, implying nonexcludability and nonrivalry, as
specified by Solow (1957). A, thus, receives no compensation and may be exploited
without limits. Arrow (1962), on the other hand, claims that increases in capital goods,
K, increase knowledge through “learning by doing.”
120
As such, knowledge is treated
similarly to that of the Solow model, but, as Romer (1990) indicates, fails to
acknowledge the tendency for firms to intentionally invest in R&D. We acknowledge
Romer’s claims that technological change is endogenous in that it arises from
intentional actions made in response to market conditions, but we also accept that there
exist other alternative effects upon technological change.

Key Institutions
An examination of international R&D collaboration provides an opportunity to
test for the endogenous effects of institutions. The two sets of institutions expected to
have a direct, positive impact upon the propensity for international R&D collaboration

120
“Learning by doing” is the education process which occurs during production. This education may
occur in a training facility (college- or firm-based) separate from the production floor.

152

are strength of IPRs and the level of democracy. Explanations for such effects, based
on existing literature, are as follows.
IPRs are treated in the literature in terms of technology received by a country.
Mansfield (1995) claims that there is a direct relationship between the strength of a
country’s IPR regime and the kinds of technology transferred to that country,
particularly with regard to high-technology industries. This is also confirmed with
Caselli and Coleman’s (2001) study of imported computing equipment, where IPRs are
a strong predictor of such imports. At the same time, Yang and Maskus (2003) claim
that stronger IPRs may discourage innovation and reduce international technology
transfer in the preliminary stages of a nation’s development. Kim (2003) claims that
technology transfer increases as the returns to innovation resulting from such IPRs
become apparent. These increases are also dependent on international licensing, local
wages, and other aspects of absorptive capacity.
The above studies’ concerns about the stage of development, industrial focus,
duration, and the effectiveness of IPR implementation are framed only by technology
receipts. Yet, unilateral transfers are only one component of the sorts of technology
transfer occurring through international R&D collaboration. This line of reasoning can
be carried to its logical conclusion, leading to the hypothesis that the strength of a
country’s IPR regime is a positive function of its propensity to collaborate in R&D with
other countries. This is supported by Mansfield’s (1995) and Kim’s (2003) conclusions
that effective IPRs lead to greater patenting, if initial patenting output is considered a

153

return to innovation.
121
The claims of Yang and Maskus (2003) are also acknowledged
and will be considered in the analysis to follow, which includes both developed and
developing countries. In any case, it must be remembered that these studies deal only
with unilateral technology transfer flows rather than that occurring through
international R&D collaboration.
For developing countries, there is not a strong set of incentives to implement
policies which foster R&D. This is true even when the direction of international R&D
transfers is solely inward-oriented. Easterly’s (2001) leaks-matches-traps idea helps
elucidate this issue. That is, incentives play a role in the decision to engage in
international R&D collaboration, and these incentives affect and are affected by the
potential for leaks-matches-traps. In the case of leaks, knowledge is a nonrivalrous
good, so international R&D collaboration is more likely to occur between countries
when a system of IPRs is implemented. “Matches” play a role as researchers from
different countries search out viable counterparts with whom they may engage in R&D
collaboration. Finally, the issue of “traps” is also a factor, as there is a major portion of
the world which still has minimal R&D generation, either domestic or via international
collaboration. These statements should be kept in mind when focusing on IPR efficacy
as a function of international R&D collaboration.
A simple descriptive analysis shows that there is a relatively clear correlation
between patenting patterns and IPR scores from the Ginarte-Park (Ginarte and Park,

121
This obviously presents an endogeneity problem in the analysis, but we will reserve further treatment
for the methods section of this paper.

154

1997) index, given a general breakdown by number of overall patents and overall
patents done in collaboration with Tier-1 (the world’s most research-productive)
countries. As shown in Table 3.2, there is a clear correlation between IPR strength and
patenting, especially with regard to collaborative patenting.
Table 3.2 Descriptive statistics for Ginarte-Park IPR scores, by patenting group
----------------------------------------------------------------------------
Number of Overall Patents Obs Mean Std. Dev. Min Max
----------------------------------------------------------------------------
>250 151 3.497883 .8577014 1.233333 4.875
50<249 69 2.282121 .957326 .9215686 4.275
25<49 32 2.232996 .8473742 .9215686 4.416667
10<24 57 1.960277 .7642848 .5882353 3.725
0<9 266 1.861444 .5739642 .5882353 3.475

----------------------------------------------------------------------------
Number of Collab. Patents Obs Mean Std. Dev. Min Max
----------------------------------------------------------------------------
>301 60 4.073472 .5747092 2.266667 4.875
31<300 94 3.120388 .9078376 1.033333 4.666667
6<30 100 2.253499 .8305728 .9215686 4.416667
2<5 80 2.011672 .6262651 .9215686 3.475
0<1 241 1.825619 .5810323 .5882353 3.725
----------------------------------------------------------------------------
Source: USPTO (2008), Ginarte and Park (1997)

We turn now to the second of the two institutions under analysis: level of
democracy. Despite the unique conditions surrounding the phenomena of international
R&D collaboration and the likelihood that experienced researchers in developing
countries are contributing significantly to R&D output, it can be assumed that
spillovers are generally flowing from more advanced to less advanced countries. As
such, international R&D collaboration is expected to be occurring under one of two
conditions. It can arise out of the natural workings of the market; hence, market failure
correcting policies of the government are not necessary. Alternatively, it can arise when
countries with weaker political institutions implement industrial promotion plans with
more concern for long-term market outcomes. It can be assumed that countries

155

encouraging international R&D collaboration in this light are less developed,
122
and
that such plans are implemented with expectations that international R&D collaboration
will generate receipts of technology transfer from abroad.
Regarding the overarching question of political institutions and national income
levels, there has been much said on this subject.
123
The seminal study of La Porta, et al.
(1999) is a key contribution to the literature, claiming that good government fosters
economic growth, where good government is defined by the absence of government
intervention and rent-seeking, public sector efficiency, small government size, and
political freedom.
124
In her review of the literature, Aron (2000) also confirms that key
institutions – particularly civil liberties and property rights – are determinants of
economic growth, although her conclusions are not robust, given simultaneity issues
and lack of quality measures.
Given that the positive effects of political institutions upon economic growth
are neither definite nor persistently robust, we must acknowledge that a number of
caveats applying to the positive democracy-growth relationship. In his examination of

122
This is easily confirmed through an examination of the correlation between GDP per capita and a
measure of political institutions. Based on the data used here, the results were positive and statistically
significant.

123
For a general discussion, see Rodrik, et al (2004).

124
La Porta, et al. also acknowledge that the quality of government is endogenous, given that economic
development can create a demand for good government. North (1981) makes a similar observation,
claiming that increases in economy activity will decrease the overall costs of institutions and facilitate
improvements in government performance. Helliwell (1994) makes this his exclusive focus, as he looks
at the bilateral linkages between economic growth and political institutions. Helliwell concludes that the
strength of political institutions positively predict economic growth, while the effect of democracy on
economic growth is negative and statistically insignificant. For the purposes of this paper, we do not treat
political institutions as a function of economic growth, primarily because our parameters extend beyond
the standard foci of growth and democracy.

156

the connection between growth and democracy, Barro (1998) concludes that there is a
nonlinear relationship: low levels of political rights increase economic growth, but
increases in political rights actually hinder economic growth after a threshold level of
democracy has been reached. Others break down political institutions to explain these
cross-country differences. Feng (2005) identifies and focuses on the intermediating
variables between democracy upon economic growth: political stability, inflation,
investment, education, income distribution, property rights, and population growth.
Feng (2005) concludes that all of these intervening variables play a part in the complex
relationship between democracy and economic growth.
From a methodological standpoint, the use of the POLITY measures may not
accurately capture the effects of the most relevant political institutions for this
discussion. There are several other indices of political institutions which may have
more relevance for this discussion than democracy. Brunetti (1997), for example, looks
at five categories of political variables as determinants of economic growth, concluding
that measures of democracy are much less statistically significant than measures of
policy volatility and subjective perceptions of politics. For the purposes of this
discussion, voting patterns and numbers of political parties in a particular country may
be less related to the reasons for international R&D collaboration than other political
institutions such as checks and balances. For this reason, checks and balances are
included in two forms in the subsequent analysis. The checks and balances measure
from the POLCON dataset, compiled by Henisz (2002), will be compared with
POLITY IV. A robustness check for the POLCON checks and balances measure is the

157

checks and balances variables from the World Bank’s Database of Political Institutions,
constructed by Keefer and Stasavage (2003).

Empirical Specification
The two-staged analysis employed here would seem to suggest the use of an
instrumental variable approach to deal with simultaneity or endogeneity concerns,
based on the literature of the previous section. Fig. 3.2 clearly supports such an
approach, but there are expected correlations between the instruments for institutions
(patenting and international patenting) and growth. One can simply ignore this
possibility and obtain biased and inconsistent estimates, given the correlation between
the R&D output and both income and institutions. The alternative approach employed
here is to examine each causal relationship in turn: first the effects of
patenting/international R&D collaboration upon income levels, then the effects of
institutions upon international R&D collaboration. We cannot simply assume, however,
that these are causal and separate, so a two-stage instrumental variable approach will
also be used.
It is worth mentioning the large literature which attempts to control for the
potential endogeneity of institutional analysis in a growth framework. These efforts are
made for good purpose, as there may indeed be a reverse causal relationship between
increased income and better institutions. Frankel and Romer (1999) instrument trade
openness (expressed as total exports and imports divided by GDP) with a gravity
equation for trade flows. Acemoglu, et al. (2001) instrument for the quality of

158

institutions with the mortality rates of colonial settlers, since property rights and the
rule of law developed with intensity when European settlers had less health problems.
These methods have subsequently been employed by Rodrik, et al. (2004) in their study
which confirms the importance of institutions over all else. Hall and Jones (1999)
account for the potential endogeneity of social infrastructure by instrumenting it with
distance from the equator, colonial language usage, and use of the English language.
Such opportunities to incorporate viable instruments are not yet available for
international R&D collaboration.
Data for the Tier-1 collaborative patents was collected through the USPTO
website’s search function.
125
Tier-1 countries – the top five patenting countries in the
world, according to USPTO patent grants – were first identified based on the number of
total patents in each of the six five-year periods, from 1975 to 2005. For each five-year
period, each country’s number of patents with Tier-1 countries was determined. What
is remarkable about the Tier-1 category is its dynamic composition, shown in Table 3.3.
For the first four periods (1975, 1980, 1985, and 1990), Tier-1 is represented by the
U.S. Japan, Germany, the U.K. and France. From 1995, however, Taiwan gained Tier-1
status as it surpassed the U.K. In similar fashion, Korea produced more patents than
France from 2000, and gained Tier-1 status. The distinction of patenting between Tier-
1 and non-Tier-1 countries is presented graphically over time in Fig. 3.3.

125
Data access is publicly available at http://www.uspto.gov/patft/ or
http://patft.uspto.gov/netahtml/PTO/search-adv.htm.

159

When generating estimates for panel data, we are ultimately faced with the
issue of how to treat country-level effects over time. Fixed effects control for constant
unobserved heterogeneity, such as developed country status, in our case. When
unobserved heterogeneity is not constant over time, such as Tier-1 status, country-
specific differences may be considered random disturbances. Despite the dynamic
nature of the Tier-1 category, shown in Table 3.3, Tier-1 status is largely constant over
time for the entire sample. Indeed, the Hausman specification test confirms that fixed
effects estimation is more efficient than random effects estimation. Simple fixed effects
estimation, however, suffices when determining the growth residual (TFP), but there
are country-specific time trends in patenting. By controlling for them in the TFP
regression, we can control for omitted variables that differ between cases but change
over time.
Table 3.3 The changing composition of Tier-1 countries
Exact dates Member countries
Period 1
Period 2
Period 3
Period 4
1975-1979
1980-1984
1985-1989
1990-1994
U.S., Japan, Germany, U.K., France
Period 5 1995-1999 U.S., Japan, Germany, France, Taiwan
Period 6 2000-2004 U.S., Japan, Germany, Taiwan, Korea

160

Fig. 3.3 Distribution of patenting between Tier-1 and non-Tier-1 countries
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1975-79
1980-84
1985-89
1990-94
1995-99
2000-04
Non-Tier-1 patents
Tier-1 patents

Source: USPTO (2008).
Note: “T1” represents Tier-1 countries, and “Non-T1” are all other countries.

Figs. 3.4 and 3.5 present further details of the changing trend in collaboration
from the non-Tier-1 countries. Not only is there an upward jump in the raw number of
collaborative patents from 1990, but the percentage of total patents represented by
international collaborative patents also increased from this time.
126
The key research –
namely the work of Helpman (1993) – focuses on the levels of research output in the
period immediately preceding this upward trend in the non-Tier-1 countries, so an
update on the innovating North-imitating South structure is needed.

126
Even though the composition of the Tier-1 sector fluctuates over time, the upward shift from 1995
would have still occurred if the U.K. (in 1995) and France (in 2000) were not replaced with Taiwan and
Korean, respectively.

161

Fig. 3.4 Number of collaborative patents
0
5000
10000
15000
20000
25000
1975 1980 1985 1990 1995 2000
T1 collab.
patents
Non-T1
collab.
patents

Source: USPTO (2008).
Note: “T1” represents Tier-1 countries, and “Non-T1” are all other countries.

Fig. 3.5 Percentage of total patents represented by collaborative patents
0
2
4
6
8
10
12
1975 1980 1985 1990 1995 2000
percent
T1 ratio
Non-T1 ratio

Source: USPTO (2008).
Note: “T1” represents Tier-1 countries, and “Non-T1” are all other countries.

We begin with the original model which holds the log of GDP per capita as the
dependent variable. This is the source of the key dependent variable in subsequent
regressions: the residual, or TFP. The measures of physical and human capital are

162

proxies for the portion of GDP spent on each form of capital for each time period. To
smooth out business cycle fluctuations, the time periods adopted here are five year
averages. Labor is measured by the number of workers per time period t, and labor
productivity is found by dividing output in t by the total number of workers.
127
Capital
is a stock measure, but it is based on the flow of investment. Both the physical capital
and human capital measures are taken in per capita terms and in logs, based on
equations (8) and (2), respectively, and the residual from the GLS regression is a
weighted average of the random-error component and the overall error component. Not
all countries have positive TFP, given this specification,
128
so the antilog of TFP was
taken before engaging in regression analysis.
The variable list can be found in Appendix 3.7 and descriptive statistics in
Appendix 3.8. In the first stage of analysis, the dependent variable is TFP, which is
based on the residual from the growth accounting regressions. The number of per capita
patents (all_patents) is the total number of patents generated by a country. This can be
juxtaposed with Tier-1 patents (tier1_collab), which is the number of patents generated
by a country in collaboration with one of the Tier-1 countries. Squared terms are
included to account for potential nonlinear effects of patenting and collaborative
patenting in terms of TFP. The general empirical framework is presented as follows, Eq
(3.10):

127
In the case of this study, the entire population is held as a proxy for the number of workers in country
i.

128
The mean for the residual is very close to zero (5.92e-10) with a standard deviation of 1.243.

163

. (3.10)
it
it
f TFP
controls) trend time
effects, patenting nonlinear
ion, collaborat 1 - Tier with patenting
ion, collaborat 1 - Tier without patenting
patenting, total ( =
To compensate for the larger stock of human capital where country i’s
population level is large, the ratio of Tier-1 patents to the total number of patents
reflects the amount of R&D collaboration done with the leading countries.
129
This is
effectively the percentage of all patents represented by collaboration with Tier-1
countries, and is a measure of the intensity of R&D collaboration with the world’s
technology leaders. The figure, however, is likely to be biased upward with small
numbers of patents, as there is much less variation between total number of patents
issued to country i and the number of patents issued in collaboration with a Tier-1
country. The data show that the greater the amount of patenting, the smaller the ratio
between these two countries. This may prove important when analyzing the predicted
effects of key institutions on international R&D collaboration.
A second set of empirical tests is done along the guidelines of Eq (3.11):
it
it
f ion collaborat nal Internatio
controls) trend
time ns, institutio
political IPRs, ( =
. (3.11)

129
For robustness checks (not presented here), this percentage term was interacted with total patenting
output and included on the right-hand side of the equation, to determine whether the interaction between
the intensity of international R&D collaboration and overall patenting had a significant effect upon TFP.
The results were not significant.

164

The dependent variable here takes two forms. First, we examine the effects of
institutions on the log of per capita international R&D collaboration between country i
and the Tier-1 countries. This is intended to show that the relevant institutions of
international R&D collaboration have a direct, positive impact on the amount of Tier-1
collaborative patenting. Another set of tests will show the predicted effects of these
same institutions upon the intensity of international R&D collaboration measure: the
percentage of total patenting represented by collaborative patents. Given that, as it was
described above, this percentage generally increases with lower number of overall
patents, the expected signs of IPR strength and the level of democracy are not clear.

Results
The panel dataset is only slightly unbalanced, and 125 countries are accounted
for in the TFP-related regression output. When IPRs (with other variables) are included
in the second-stage of the analysis, data for 111 countries is available.
130
These results
include a number of different specifications, the bulk of which serve as robustness
checks for the predicted effects of patents and collaborative patents on TFP.
131

Appendix 3.1 presents generalized least squares (GLS) regression output for the total

130
When democracy (POLITY) is included by itself in the second stage analysis, data for 150 countries
is available. When checks-and-balances (POLCON) is included by itself, data for 153 countries is
available.

131
Although panel data observes changes over time, there were a large number of countries with
negative TFP values. An analysis of the log_y, log_k, and log_h measures (not included here) shows that
the residuals (TFP) are large for a number of low capital-intensive, low education-intensive countries
with high per capita income. This can be the result of population size. Although this problem is partially
alleviated with the application of five-year averages across all specifications (growth accounting as well
as patenting measures), certain aspects of TFP may be biased upwards given the absence of factors
which are more prevalent in developing countries.

165

number of patents (all_patents) and its squared form (all_patents_sq), and Appendix
3.2 has a similar presentation for the number of Tier1-collaborative patents
(tier1_collab) and its squared form (tier1_collab_sq).
132
Appendices 3.3 and 3.4
perform the same operation while controlling for the percentage of all patents done in
collaboration with Tier-1 countries (perc_t1_all). In all of these results (Appendices 3.1
to 3.4), significant coefficients for overall time trends, whether based in a coded year
variable (yearcode) or year dummy variables (y1975, …, y2000),
133
reveal that there is
a time-related consideration and a nonlinear effect which must be treated in this panel
dataset.
134
Overall time trends capture exogenous movements in changes in TFP, but they
fail to isolate some of the country-specific time trends. That is, beyond patents and its
derivatives, which are the only country-specific time varying explanatory variables,
there can also be factors which are varying over time within each country. The overall
time trend measures included in Appendices 3.1 to 3.4 do not address the biasedness
which arises when individual country’s trends counter the dynamic effects of patenting,

132
The GLS fixed effects model is preferred because modified Hausman tests (for the inclusion of
country-specific time effects) provided evidence in support of fixed effects modeling. GLS modeling
techniques allow for higher weights for countries with higher output and smaller disturbance variances
than countries with less output and larger disturbance variance (Greene, 2002). This effectively controls
for volatility which remains after using five year averages of the data.

133
In the regression output, y1975 is omitted for multicollinearity-related reasons.

134
A number of variants on these patent measures have also been included, such as an acceleration of
flows measure, which shows the differences in patents between period t and 1975, and a stock measure
which measures the total stock of patents in period t, building on a base period of 1975. These efforts
largely confirmed the results shown here, and function as a robustness test.

166

etc. on TFP.
135
As such, country-specific time trends are included in the results of
Tables 3.4 to 3.8, offering an alternative method of accounting for time trends without
diminishing the impact of our explanatory variables.
With country-specific time trends, the effect of the total number of patents is
positive in the linear term (all_patents) and negative in the squared term
(all_patents_sq) (Table 3.4). In Table 3.4, column 3, the coefficients of these measures
decrease slightly in scale when the percentage of total patents represented by Tier-1
collaborative patents is included in the specification, but they are still statistically
significant. In this simple analysis of individual effects, the pattern of Tier-1
collaborative patents’ (tier1_collab and tier1_collab_sq) effects on TFP (columns [4] –
[6]) counter those of overall patents, but the coefficients are generally not statistically
significant. This implies that there may be alternative specifications which account for
the joint effects of overall patents and Tier-1 collaborative patents, both in linear and
nonlinear forms.
Table 3.5 presents tests for various relationships of overall patents’ and Tier-1
collaborative patents’ effects on TFP. Both linear and nonlinear forms are used for both
categories of patents, and country-specific time trends are included. When the effects of
overall patenting (all_patents) and international R&D collaboration (tier1_collab) are
analyzed together, the former is positive in its linear term and negative in its squared
term, while the opposite is true for Tier-1 collaborative patenting. This is consistent
across all specifications in Tables 3.5.

135
Inclusion of the country-specific time trends also minimizes autocorrelation in the error term.

167

In Tables 3.6, 3.7, and 3.8, we examine the log-transformed values for overall,
collaborative, and non-collaborative patents.
136
Ultimately, the linear-log model
attempts to show the marginal TFP output of an extra per capita patent, overall,
collaborative, or non-collaborative. Thus, positive coefficients for the linear term show
that the marginal increase in TFP with respect to an increase in patenting is a
decreasing function of patenting. In this way, the use of the linear-log model allows us
not only to generate coefficients which are of a manageable size (given the huge size of
the treated per capita patenting measures), but also to perform a test consistent with
nonlinearity in the effects of patents on TFP.
Based on these results, collaboration creates a bonus effect upon overall
patenting (Table 3.6, column [4]). This also seems to be largely the case when we look
at the joint effects of collaboration (log_tier1_collab) and non-collaboration
(log_notier1), in Table 3.7 (column [4]). In column (5) of these two tables, the
percentage of all patents represented by Tier-1 collaborative patents (perc_t1_all) is
negative and statistically significant. This can be attributed to the fact that countries
with a higher percentage term are developing countries with minimal domestic
technological capabilities. Consider the relationship between the percentage term and
GDP per capita expressed in Fig. 3.6.

136
It should be noted that, in order to achieve increasing values in the squared terms, patents (total and
collaborative) have been multiplied by 1,000,000. This was the minimum factor which could be applied
to per capita patents, given that the smallest untreated per capita overall patent value was around .000002
(China). This scale is the same for per capita collaborative patents.

168

169
Expressed in Table 3.8, the interactive effects of non-Tier-1 patenting with a
lagged income measure (log_notier1_yt1) are generally positive, indicating that non-
Tier-1 collaborative patents have greater effects on TFP when income is higher.
Coefficients for the interaction between Tier-1 patents and lagged income
(log_tier1collab_yt1) are negative, which indicates that countries which collaborate and
have high income will have lower TFP. There may be particular benefits of
collaboration with Tier-1 countries for developing countries, thus, which tend to have
lower levels of income.

Table 3.4 Testing for overall and Tier-1 patenting effects on TFP, linear and nonlinear, accounting for country-specific time
trends and percentage of all patenting done in collaboration
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
tfp tfp tfp tfp tfp tfp
--------------------------------------------------------------------------------------------------------------------
all_patents 0.00000626*** 0.0000142*** 0.0000141***
(0.00000116) (0.00000185) (0.00000186)

all_patents_sq -9.99e-12*** -9.93e-12***
(1.85e-12) (1.86e-12)

tier1_collab -0.0000263 -0.0000126 -0.0000121
(0.0000152) (0.0000176) (0.0000176)

tier1_collab_sq -1.01e-10 -1.01e-10
(6.54e-11) (6.53e-11)

perc_t1_all -0.234 -0.477
(0.411) (0.429)
--------------------------------------------------------------------------------------------------------------------
Observations 662 662 662 662 662 662
chi2 2095.7 2089.1 2086.8 1973.7 1926.4 1928.4
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001
170

Table 3.5 Testing for overall patents and Tier-1 collaboration effects on TFP, linear and nonlinear, accounting for country-
specific time trends, per capita total patents, and the percentage of all patenting done in collaboration
------------------------------------------------------------------------------------
(1) (2) (3) (4)
tfp tfp tfp tfp
------------------------------------------------------------------------------------
all_patents 0.00000966*** 0.0000113*** 0.0000148*** 0.0000161***
(0.00000130) (0.00000148) (0.00000185) (0.00000193)

all_patents_sq -7.64e-12*** -7.35e-12***
(1.93e-12) (1.93e-12)

tier1_collab -0.0000875*** -0.000120*** -0.0000653*** -0.0000943***
(0.0000167) (0.0000219) (0.0000174) (0.0000226)

tier1_collab_sq 1.61e-10* 1.38e-10*
(7.10e-11) (7.04e-11)

perc_t1_all -0.162
(0.406)
------------------------------------------------------------------------------------
Observations 662 662 662 662
chi2 2227.4 2189.2 2152.4 2166.2
------------------------------------------------------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001

171

Table 3.6 Testing for patenting effects on TFP, linear and nonlinear, accounting for country-specific time trends and
percentage of all patenting done in collaboration
----------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5)
tfp tfp tfp tfp tfp
----------------------------------------------------------------------------------------------------
log_allpatents 0.264*** 0.279*** 0.200*** 0.140***
(0.0290) (0.0293) (0.0413) (0.0422)

log_tier1collab 0.269*** 0.104* 0.249***
(0.0341) (0.0478) (0.0553)

perc_t1_all -1.187** -2.309***
(0.405) (0.469)
----------------------------------------------------------------------------------------------------
Observations 662 662 662 662 662
chi2 2345.4 2387.2 2249.2 2366.5 2493.3
----------------------------------------------------------------------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001

172

Table 3.7 Testing for exclusive patenting effects on TFP, linear and nonlinear, accounting for country-specific time trends
and percentage of all patenting done in collaboration
----------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5)
tfp tfp tfp tfp tfp
----------------------------------------------------------------------------------------------------
log_notier1 0.269*** 0.268*** 0.205*** 0.124**
(0.0284) (0.0286) (0.0352) (0.0424)

log_tier1collab 0.269*** 0.125** 0.258***
(0.0341) (0.0414) (0.0571)

perc_t1_all -0.120 -1.838***
(0.401) (0.548)
----------------------------------------------------------------------------------------------------
Observations 662 662 662 662 662
chi2 2377.1 2373.2 2249.8 2421.8 2479.4
----------------------------------------------------------------------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001
173

174
Table 3.8 Testing for exclusive patenting effects on TFP, linear and nonlinear, accounting for country-specific time trends,
percentage of all patenting done in collaboration, and patent-income interactions
----------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5)
tfp tfp tfp tfp tfp
----------------------------------------------------------------------------------------------------
log_notier1 0.220*** 0.212*** 0.0690 -0.0261
(0.0364) (0.0375) (0.0536) (0.0601)

log_notier1_yt1 0.00000801 0.00000907 0.0000232* 0.0000229*
(0.00000624) (0.00000637) (0.00000929) (0.0000106)

log_tier1collab 0.297*** 0.240*** 0.419***
(0.0421) (0.0593) (0.0793)

log_tier1collab_yt1 -0.0000130 -0.0000251* -0.0000292*
(0.00000679) (0.0000104) (0.0000145)

perc_t1_all 0.216 -2.149**
(0.514) (0.674)

perc_yt1 -0.000104 0.00000679
(0.000117) (0.000163)
----------------------------------------------------------------------------------------------------
Observations 556 556 556 556 556
chi2 3282.0 3273.8 3233.4 3411.5 3530.1
----------------------------------------------------------------------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001

Fig. 3.6 The relationship between Tier-1-all patent ratio and GDP per capita
0 .2 .4 .6 .8 1
4 6 8 10 12
log_y
95% CI Fitted values
perc_t1_all

The modeling techniques of IPRs and democracy/checks-and-balances are
similar to the previous TFP-based regression analyses. Tables 3.9 to 3.12 and Appendix
3.5 to 3.12 offer three comparable sets of results: Tables 3.9 and 3.10 (and Appendices
3.5 and 3.6) use the log of Tier-1 collaborative patents (log_tier1collab) as the
dependent variable, Tables 3.11 and 3.12 use the percentage of patents represented by
international R&D collaborative patents (perc_t1_all), and Tables 3.13 and 3.14 have
the log of non-collaborative patents as the dependent variable. Appendices 3.5 to 3.6
are distinct from the results in the main text in that they include an overall time trend

175

variable (yearcode) as an identifier of the strong time trend, while Tables 3.9 to 3.13
control for country-specific time trends.
137
.
Looking first at the determinants of the log of Tier-1 collaborative patents,
changes in IPRs (log_ipr) and democracy (log_polity) were found to have significant
and positive effects on the change in the number of per capita patents generated with
Tier-1 countries. Further, changes in IPR-democracy interactions (log_ipr_polity) were
found to have a considerable, positive impact on changes in the number of per capita
collaborative patents generated by a country. In the case of Table 3.9, column (6), the
impact of IPRs in the full model is much stronger when there are democratic
institutions in place.
138
Where the interaction term is not included in the model
(columns [1] to [3]), there are positive and significant results for both the log of IPRs
and the log of democracy.
When the POLCON checks-and-balances measure is substituted for the
POLITY scores as in Table 3.10 (and Appendix 3.6), the results largely replicate the
results which use the POLITY score. Indeed, the IPR-checks-and-balances interaction
(log_ipr_polconchecks) term is significant in all specifications (Table 3.10, columns [4]
to [6]), again implying that these two sets of institutions are much more effective when
they are both present. There is a sense of interdependence with regard to the impact of
IPRs and political institutions upon collaborative patenting with Tier-1 countries,
whether such political institutions are measured in terms of democracy (POLITY) or

137
GLS fixed effects modeling is used in Appendices 3.9 and 3.12, based on the Hausman test results.

176

checks-and-balances (POLCON). The absence of major differences in the coefficients
of the POLITY and POLCON scores also indicates that they are measuring the same
phenomena, in terms of international R&D collaboration.
When the dependent variable is the percentage of total patents represented by
Tier-1 collaboration (Tables 3.11 and 3.12), the results are very different from those in
which the dependent variable is the log of collaborative patents with Tier-1 countries.
In nearly all cases, IPR strength and political institutions (whether measured as level of
democracy or checks-and-balances) have negative coefficients, when the dependent
variable is the percentage term. This finding extends to the IPR-POLITY/POLCON
interaction term(s). Ultimately, this represents a qualitative difference between per
capita international collaborative patenting and the percentage of patenting represented
by collaborative patenting. These are the poorer countries with virtually no other
patenting at home.
As a final check on the function of institutions in collaboration with Tier-1
countries, we can compare their effects in both collaborative and non-collaborative
contexts. When we compare the impact of institutions upon collaborative patenting
with Tier-1 countries (Tables 3.9 and 3.10) and patenting not done with Tier-1
countries (Tables 3.13 and 3.14), institutions have a greater effect upon non-Tier-1
patents. This is also true when POLITY is replaced with the POLCON checks-and-
balances measure. While institutions are important and have significant and positive

138
This is also confirmed in Appendix 3.9, columns (5) and (6), where fixed effects are only applied. In
this case, both IPRs and democratic institutions had much stronger effects when they are jointly present.

177

effects upon both types of patents, we can conclude that collaborative patents are
affected to a lesser degree than non-Tier-1 patents. This lends further support for our
earlier interpretation that international R&D collaboration has a supplemental or bonus
effect to non-collaborative patents.
The analysis thus far has been focused on two indirectly connected phenomena.
We first accounted for the predicted effects of patenting and international collaborative
patenting upon TFP, which was followed by an analysis of the effects of political
institutions and IPRs upon patenting and collaborative patenting. While these
specifications have produced a number of interesting results, they do not operate in
isolation, so we now present the results from a two-stage analysis using political
institutions and IPRs as instruments for patents and collaborative patents.
139
Among the five different institution-based variables already employed, the only
valid instrument for overall patents and international collaborative patents is the IPR-
POLCON interaction term (log_ipr_polconchecks). The individual and joint
significance of the coefficients for IPRs (log_ipr), POLITY (log_polity), POLCON
(log_polconchecks), and the IPR-POLITY interaction term (log_ipr_polity) precluded
using them as instruments, given that they were found to be correlated with the error
term of the second-stage equation. These tests are necessary, as the causal connections
between these institutions and TFP are loose but not wholly non-existent. The
significant effects of political institutions and IPRs upon income levels, detailed in the

139
Hausman tests show that there is apparent endogeneity of overall patents and international
collaborative patents in the TFP regression, which also supports a two-stage regression analysis.

178

179

earlier literature review, could very well work through the determinants of income (i.e.,
TFP), but this connection has not been well established.
The instrumental variable results in Table 3.15 simultaneously present both
causal chains, from institutions to patenting and from patenting to TFP. The only viable
instrument, for reasons described above, is the interaction term between IPR strength
and checks and balances, which was found to be a largely significant and positive
predictor of patents in all of our earlier results.
140
Given that the results in Table 3.15
capture and instrument for both first-stage operative institutions, albeit in interactive
form, we feel that these results are robust in showing the positive and statistically
significant effects of overall patenting, collaborative patenting, and non-collaborative
patenting upon TFP. Further, based on these two-stage results, collaboration with Tier-
1 countries was found to have a stronger impact upon TFP than either overall patenting
or non-collaborative patenting.

140
Not presented here are individual significant and positive effects of this interaction term upon overall
patents and collaborative patents.

Table 3.9 Testing for institutional effects on the log of Tier-1 collaboration, accounting for possible interactions between
IPRs and democracy, country-specific time trends
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
log_tier1collab log_tier1collab log_tier1collab log_tier1collab log_tier1collab log_tier1collab
--------------------------------------------------------------------------------------------------------------------
log_ipr 2.371*** 1.972*** -2.057*** -0.0768
(0.337) (0.318) (0.594) (0.891)

log_polity 1.346*** 1.418*** 0.857*** 0.839**
(0.141) (0.158) (0.187) (0.283)

log_ipr_polity 2.200*** 1.063*** 1.099*
(0.250) (0.158) (0.447)
--------------------------------------------------------------------------------------------------------------------
Observations 619 829 587 587 587 587
chi2 1185.4 1577.3 1375.3 1364.9 1398.7 1395.8
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
+ p<.10, * p<.05, ** p<.01, *** p<.001
180

Table 3.10 Testing for institutional effects on the log of Tier-1 collaboration, accounting for possible interactions between
IPRs and checks-and-balances, country-specific time trends
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
log_tier1collab log_tier1collab log_tier1collab log_tier1collab log_tier1collab log_tier1collab
--------------------------------------------------------------------------------------------------------------------
log_ipr 2.371*** 1.668*** 0.230 0.230
(0.337) (0.324) (0.397) (0.397)

log_polcon_checks 1.524*** 1.438*** -0.230
(0.154) (0.171) (0.397)

log_ipr_polconchecks 1.438*** 1.668*** 1.438***
(0.171) (0.324) (0.171)
--------------------------------------------------------------------------------------------------------------------
Observations 619 807 583 583 583 583
chi2 1185.4 1679.3 1368.9 1368.9 1368.9 1368.9
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses.
Note: log_polcon_checks was dropped from column 6 because of multicollinearity.
+ p<.10, * p<.05, ** p<.01, *** p<.001
181

Table 3.11 Testing for institutional effects on the percentage of total patents represented by Tier-1 collaboration, accounting
for possible interactions between IPRs and democracy, country-specific time trends
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
perc_t1_all perc_t1_all perc_t1_all perc_t1_all perc_t1_all perc_t1_all
--------------------------------------------------------------------------------------------------------------------
log_ipr -0.0544+ -0.0571+ -0.0209 0.0346
(0.0312) (0.0326) (0.0607) (0.0919)

log_polity -0.00452 -0.00242 0.0153 0.0235
(0.0132) (0.0162) (0.0193) (0.0292)

log_ipr_polity -0.0183 -0.0330* -0.0492
(0.0256) (0.0163) (0.0461)
--------------------------------------------------------------------------------------------------------------------
Observations 619 829 587 587 587 587
chi2 324.1 391.3 292.8 293.6 294.4 294.0
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses.
+ p<.10, * p<.05, ** p<.01, *** p<.001
182

Table 3.12 Testing for institutional effects on the percentage of total patents represented by Tier-1 collaboration, accounting
for possible interactions between IPRs and checks-and-balances, country-specific time trends
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
perc_t1_all perc_t1_all perc_t1_all perc_t1_all perc_t1_all perc_t1_all
--------------------------------------------------------------------------------------------------------------------
log_ipr -0.0544+ -0.0606+ -0.0458 -0.0458
(0.0312) (0.0319) (0.0390) (0.0390)

log_polcon_checks -0.0108 -0.0148 0.0458
(0.0141) (0.0168) (0.0390)

log_ipr_polconchecks -0.0148 -0.0606+ -0.0148
(0.0168) (0.0319) (0.0168)
--------------------------------------------------------------------------------------------------------------------
Observations 619 807 583 583 583 583
chi2 324.1 433.3 332.7 332.7 332.7 332.7
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses.
Note: log_polcon_checks was dropped from column 6 because of multicollinearity.
+ p<.10, * p<.05, ** p<.01, *** p<.001
183

Table 3.13 Testing for institutional effects on the log of non-Tier-1 collaborative patents, accounting for possible
interactions between IPRs and democracy, country-specific time trends
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
log_notier1 log_notier1 log_notier1 log_notier1 log_notier1 log_notier1
--------------------------------------------------------------------------------------------------------------------
log_ipr 3.128*** 2.611*** -2.959*** 0.0000274
(0.397) (0.365) (0.686) (1.023)

log_polity 2.004*** 1.992*** 1.253*** 1.253***
(0.165) (0.182) (0.215) (0.325)

log_ipr_polity 3.046*** 1.401*** 1.401**
(0.289) (0.182) (0.513)
--------------------------------------------------------------------------------------------------------------------
Observations 619 829 587 587 587 587
chi2 1163.5 1625.2 1463.5 1434.3 1493.8 1490.7
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
* p<.05, ** p<.01, *** p<.001
184

Table 3.14 Testing for institutional effects on the log of non-Tier-1 collaborative patents, accounting for possible
interactions between IPRs and checks-and-balances, country-specific time trends
------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5)
log_notier1 log_notier1 log_notier1 log_notier1 log_notier1
------------------------------------------------------------------------------------------------------
log_ipr 3.128*** 2.265*** 0.125
(0.397) (0.365) (0.447)

log_polcon_checks 2.207*** 2.141*** -0.125
(0.181) (0.192) (0.447)

log_ipr_polconchecks 2.141*** 2.265***
(0.192) (0.365)
------------------------------------------------------------------------------------------------------
Observations 619 807 583 583 583
chi2 1163.5 1669.0 1529.3 1529.3 1529.3
------------------------------------------------------------------------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001
185

186
Table 3.15 Testing for effects of overall patents, collaborative patents, and non-collaborative patents upon TFP,
instrumenting for IPR-POLCON interaction term
--------------------------------------------------------------------
(1) (2) (3)
tfp tfp tfp
--------------------------------------------------------------------
log_allpatents 0.602***
(0.0715)

log_tier1collab 0.812***
(0.104)

log_notier1 0.556***
(0.0645)
--------------------------------------------------------------------
Observations 518 518 518
chi2 1894.8 1631.3 1981.4
--------------------------------------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001

Conclusion
This chapter examines the determinants and effects of the growing phenomenon
of international R&D collaboration, measured through cross-national patenting efforts.
It has been shown that the nonlinearity hypothesis holds both when the overall time
trend (Appendices 3.1 to 3.4) and country-specific time trends are accounted for
(Tables 3.4 to 3.8). These trends should be included, given the strong, increasing trend
in overall collaborative patenting over time and the biasedness which arises when
individual country’s trends counter the dynamic effects of patenting on TFP. In
addition, there are other sources of TFP growth that can vary considerably from
country to country, such as shifts in product mix, and various country-specific shocks.
Our overall interpretation of the negative effect of collaborative patenting on
TFP is that there is a considerable amount of imitation occurring at low levels of
collaborative patenting, primarily by developing countries, leading to considerable TFP
increase independent of factor accumulation as well as patenting. In line with the
literature, patenting is not widely encouraged by developing countries, and it is inferred
that copying technology through Tier-1 collaboration rather than producing original
R&D output in the form of patents is the norm. Once a particular collaborative
patenting threshold is crossed, indicated by the positive and significant nonlinear term
in the regression analysis, technology-oriented countries become patenting centers and
achieve higher TFP levels. This interpretation is largely founded on the issue of
enforcement, where weak IPRs are likely to prevent extensive patenting. Descriptive

187

analyses (Table 3.2) evidenced a clear correlation between patenting trends – both
overall and collaborative – and the strength of IPRs.
In terms of patenting analysis, in the linear-log specifications all measures have
positive and statistically significant effects upon TFP. When we extract collaborative
patents from the total patents measure, the effects of non-Tier-1 collaborative patents
(log_notier1) and Tier-1 collaborative patents (log_tier1collab) upon TFP are
comparable, presented in Tables 3.7 and 3.8.
141
Yet, a two-stage analysis instrumenting
patenting with the IPR-POLCON interaction term shows a greater impact upon TFP
from Tier-1 collaborative patents than from non-collaborative patents. In the context of
the structural model presented in Fig. 3.2, we are led to conclude that international
R&D collaboration is a key determinant of TFP and is a function of political
institutions and IPR strength.
Collaborative patenting clearly plays an important role, indicated also by the
percentage measure. However, this result is challenged by the above results from the
second stage analysis, in which IPRs and democracy/checks-and-balances have
negative predicted effects on the share of patents represented by collaborative patents
(perc_t1_all), presented in Tables 3.11 and 3.12. This may reflect the simple fact that
international R&D collaboration is still not occurring in large enough quantities for IPR
strength to generate more patents. A more convincing interpretation refers to the simple
fact that lower overall patenting efforts are correlated with higher percentages of

141
On the other hand, political institutions and IPRs have greater predicted effects upon non-Tier-1
collaborative patents. Compare the results of Table 3.9 with Table 3.13 and Table 3.10 with Table 3.14.

188

collaboration, which describes the situation in many technology trailing countries.
142

Nevertheless, for the number of per capita international collaborative patents, IPRs and
democracy/checks-and-balances have important, positive effects.
Readers more familiar with the costs and benefits of international technology
transfer are undoubtedly aware that international R&D collaboration, as described here,
is not without limits. There is a strong disincentive to share the research results with
real or potential competitors,
143
and attempts to limit the sharing of knowledge and
expertise with developing countries have parallels to the effects of tariff escalations
between the global north and south. Stronger IPRs help suppress reverse engineering
and imitation efforts, so they may in fact help limit the disbursement of knowledge and
the growth of key capabilities (Maskus, et al., 2005). As Maskus, et al. (2005) note, this
is the balancing act between protectionism and development, and the ramifications for
developing countries are evidenced in support of the vicious cycle of development.
Mention can also be made of the connection between national and global
welfare, which is a peripheral impetus to this study of international R&D collaboration.
Much like Barrett’s (2007) discussion of global public goods, international R&D
collaboration has the potential to not only increase income for individual countries, but
also increase global welfare through the generation of advances in science and
technology which would not have been available under non-collaborating conditions.

142
See Fig. 3.6, for a graph of the relationship between the percentage term and GDP per capita.

143
Disincentives to share information are equally true for both domestic and international competition,
given extensive linkages between countries.

189

This practice reflects the internationalization of externalities which had previously been
isolated to individual countries, particularly shared environmental and economic costs
within regions. Along these lines, increased regional integration will continue to
advance science and technology. Analyses of how international R&D collaboration is
treated within regional pacts should be given full attention as this pattern continues, as
research of economic geography attempts to address how different institutional factors
facilitate and/or hinder flows of information and knowledge across firms, regions, and
nations.
144
The two-stage analysis herein attempted to resolve simultaneity issues, but
future work should continue to focus on potential endogeneity concerns with regard to
collaboration. It is much more likely that pre-existing levels of international R&D
collaboration lead to high levels of current international R&D collaboration. As well, it
is also necessary to qualify and compare the benefits of international R&D
collaboration, vis-à-vis alternative methods of bolstering the national innovation system.
There is an abundant literature which focuses on such alternative phenomena,
145
and
these results must be considered jointly with the international R&D collaborative
effects.

144
See Polenske (2007) , particularly the chapters by Alice Lam and Saxenian.

145
See, for example, two related papers on public-private R&D collaboration in Korea and Taiwan
(contact mattheas@usc.edu). The most recent contribution focusing on developing countries is that of
Hall and Maffioli (2008) which concludes that government R&D funding does not crowd out private
investment, although it does not necessarily generate significant amounts of patenting and other output.

190

Concluding Thoughts
Much of the available work on East Asia’s technological development has been
contextualized in macroeconomic analysis but strewn with policy-, industry-, or firm-
level case studies. In this dissertation, this literature is refocused through the lens of
developmental statism, and it is maintained that S&T policies attempting to correct for
market failure in R&D efforts are now one of the most important areas of political
economy, particularly in East Asia. This is confirmed through a brief review of the
overall methods and findings of each of the previous chapters.

Recapitulation
Chapter 1 studies the effects of science and technology policies upon
information transfers. According to the literature, such transfers are indicators of
greater social returns, particularly when they flow between public and private research
entities. The model presented in Chapter 1 tests for whether government research and
development (R&D) subsidization increases information flows in Korea and Taiwan.
Using the unique KORTAI R&D dataset, it is shown that information transfers have a
positive impact upon public-private R&D output (patents and publications), but that the
effect of government funding is not uniform in both countries. Information flows are
much more strongly predicted by government funding programs in Taiwan, while
exogenous factors such as effectiveness of technology licensing offices (OTLs) and
geographic proximity are key determinants for Korea. This is attributed to the

191

burgeoning importance of OTLs in Korea as well as the relative lack of public funding
in Taiwan.
Chapter 2 begins with the admission that the Triple Helix paradigm has great
value as a model public-private R&D collaboration, given its focus on the research-
based links between the government, universities, and firms. Also shown in this chapter
is the paucity of formal tests for the Triple Helix model. To alleviate this deficiency, an
empirical analysis is conducted of the effects of relationship-based forms of capital,
which has been identified as an attribute of Triple Helix-based R&D. OLS results
utilizing the KORTAI R&D dataset reveal three primary conclusions: (1) there are no
differences between the two countries with regard to the effects of this new form of
relationship-based capital; (2) the effects of pre-existing capital are positive only for
collaborative patenting output; (3) new forms of relationship-based capital have a
positive and significant effect for both collaborative and non-collaborative output.
Robustness checks from ordered logit models provide strong evidence that the Triple
Helix paradigm does indeed have predictive power with regard to relationship-based
capital. For Korea and Taiwan, the importance of new relationships is a particularly
remarkable discovery from Chapter 2, given that traditional methods of networking
typically constrain the expansion of relationships.
R&D output has been increasing exponentially over the last thirty years, but it is
disproportionately represented by developed countries. In terms of total factor
productivity (TFP), Chapter 3 considers the linear and nonlinear effects of overall
patenting, the linear and nonlinear patterns of international R&D collaboration, and the

192

percentage of overall patenting represented by international R&D collaboration. The
nonlinear estimation results support the view that higher patenting and collaborating
countries are more effective patenting hubs, which can be attributed to stronger
institutional bases. For confirmation, an analysis is conducted of the effects of
intellectual property rights (IPRs) and democratic institutions upon international R&D
collaboration. The dataset is an unbalanced panel of five year averages from 1975 to
2005 for a maximum of 150 countries.
146
What is conclusive here is that Korea and Taiwan are extremely consistent,
which makes the most-similar comparative approach of Przeworski and Teune (1970)
tractable. Where differences do arise, such as in the effects of public funding (Chapter
1) or the scale of collaborative tendencies (Chapter 2), they can be attributed to specific
characteristics, such as public funding tendencies or the precedent of SMEs.
Consistencies between Korea and Taiwan are also evident in terms of overall patenting
output, international R&D collaboration, and their ascent to Tier-1 status, which was
covered in the macro-level analysis of this dissertation (Chapter 3).

Policy prescriptions
Policy prescription based on the results and conclusions of Chapters 1, 2, and 3
(and the literature review of the introductory chapter) may be divided into two sets.
Based on evidence offered in Chapters 1 and 2, the various R&D funding programs are

146
The data is drawn from the USPTO, WDI, Penn World Table, Barro-Lee education data, the Ginarte-
Park IPR index, and POLITY IV.

193

largely successful and should continue to experience similar successes if information
transfers are facilitated. Given the importance of conferences as an information transfer
mechanism, public funding can correct for potential market failure in R&D efforts if it
is partially allotted to create conferencing opportunities.
147
In relative terms, however,
based on the ranked and weighted reasons for repartnering offered in Chapter 2, public
funding was found to be less important than other reasons for repartnering. This does
not rule out the possibility that policy-directed cross-sector R&D collaboration is still a
necessary catalyst; i.e., other reasons for repartnering become salient for the researcher
over time. In this way, market failure-correcting policies have a direct effect for the
initial partnership and an indirect but still important effect for repartnering.
It is apparent from Chapter 3 that IPRs and democratic institutions are
extremely important, both in terms of overall patenting efforts and international R&D
collaboration with Tier-1 countries. Precise policy prescriptions are more difficult
when we consider the lessons which can be transferred from Korea and Taiwan to other
developing countries. Because Korea and Taiwan have applied a wholly interventionist
approach to the strengthening of their respective NISs, perfect replication will be a
challenge for other countries. This is yet further confirmation of the unique qualities of
the East Asian growth model, which significantly diminished in the literature following
the financial crisis of 1997-98.

147
The weighted sources of personal ties in Chapter 2 also support the conclusion that conferences are
important for public-private R&D collaboration.

194

Future pursuits
As an increasing number of countries devote larger shares of R&D expenditures
for the generation of innovative results via public-private R&D collaboration and
international R&D collaboration, it is absolutely crucial to conduct both broader and
deeper analyses. To address the point of transferable lessons, future research projects
should include the cataloging of similar programs for other developing countries,
beginning with the East Asian region. This will be the essence of a renewed call for
East Asian developmental statism, with a strong R&D focus. If the evidence does not
permit a broad, regional model, then the pattern of the Korean and Taiwanese cases can
be said to be truly unique. To get at this point further, we can conduct detailed analyses
of geographic proximity and local politics and policies as a function of R&D success,
given the numerous science and engineering parks throughout East Asia and the rest of
the world. It is also believed that there is more which can be said about the impact of
intellectual property rights upon international technology flows beyond the structures
addressed in Chapter 3. Precisely, the deadweight losses from patent portfolio
strategies are expected to be considerable, and such portfolios are becoming
increasingly popular.
Finally, there is a clear relationship between Korea and Taiwan’s R&D output
and that of China, so one must consider the regional dynamics of East Asia. Given
China’s continued growth in terms of traditional measures as well as technological
capabilities, it is of paramount importance for Korea and Taiwan to intensify their
efforts. Korea may be generating more patents, but China is outpacing Korea in terms

195

of basic research (Seong, 2005). As well, shorter-term military service in Taiwan can
be waived in lieu of military service as an engineer at a research institute or a private
firm. This intersection between defense and S&T policies represents the new era of
technology development in response to international competition. Rather than
concentrate their forces in the traditional, military service manner, Taiwan is offering
engineers within the military the option of fulfilling their required military service
duties as a researcher in one of several laboratories (both public and private)
throughout the country. The implication is that the Taiwanese government now
perceives mainland China to be not only a military threat but also a technological threat
in terms of skills, knowledge, and capabilities.
148
In order to sustain the current
knowledge base, thus, qualified Taiwanese soldiers are being removed from the
conventional military to complete a slightly longer service in conditions which are
virtually no different from civilian-sector R&D employment.
149
With this replacement
of military warfare with technological warfare, where success is measure by R&D
returns, the Chinese phenomena cannot be underestimated.

148
It should be noted that the engineers are not engaging in defense-related research.

149
Korea, like Taiwan, also has a required military service. In some respects, the Taiwan-Mainland
relations are quite similar to those between North and South Korea, given the post-civil war environment.
(“Post-civil war” and “civil war” are not entirely accurate verbiage for either the Korean peninsula or
Taiwan-China case. The civil war in Korea is not formally over, as there is currently only a cease fire
between the North and South. For the Taiwanese case, the situation corresponds more with a seceded
province rather than an internal battle, although the early political leaders in Taiwan claimed that Taipei
was the formal seat of power for all of China, despite the irrefutable communist leadership under Mao in
Beijing.) The military threat in Korea, however, does not justify a research-based military service for
engineers. Even without formal data, the R&D generated within North Korea cannot be compared to that
of mainland China.

196

Bibliography
Acemoglu, Daron, Simon Johnson, and James A. Robinson. "The Colonial Origins of
Comaprative Development: An Empirical Investigation." American Economic
Review 91, no. 5 (2001): 1369-401.
Adams, James D., Eric P. Chiang, and Katara Starkey. "Industry-University
Cooperative Research Centers." Journal of Technology Transfer 26, no. 1-2
(2001): 73-86.
Aldrich, Howard E., and Toshihiro Sasaki. "R&D Consortia in the United States and
Japan." 24, no. 2 (1995): 301-16.
Amsden, Alice. Asia's Next Giant: South Korea and Late Industrialization. New York:
Oxord University Press, 1989.
Aron, Janine. "Growth and Institutions: A Review of the Evidence." World Bank
Research Observer 15, no. 1 (2000): 99-135.
Barrett, Scott. Why Cooperate? The Incentive to Supply Global Public Goods. Oxford:
Oxford University Press, 2007.
Barro, Robert J. Determinants of Economic Growth: A Cross-Country Empirical Study.
Cambridge: Cambridge University Press, 1998.
Barro, Robert J., and Jong-Wha Lee. "International Data on Educational Attainment:
Updates and Implications." CID Working Paper, No. 42 (2000).
Barro, Robert J., and Xavier Sala-i-Martin. Economic Growth. Cambridge: Cambridge
University Press, 2003.
Benhabib, Jess, and Mark M. Spiegel. "Human Capital and Technology Diffusion."
FRBSF Working Paper 2003-02 (2002).
———. "The Role of Human Capital in Economic Development: Evidence from
Aggregate Cross-Country Data." Journal of Monetary Economics 34, no. 2
(1994): 143-73.

197

Bernasconi, Andreas. "University Entrepreneurship in a Developing Country: The Case
of the P. Universidad Catolica De Chile, 1985-2000." Higher Education 50, no.
2 (2005): 247-74.
Bernstein, Jeffrey I., and M. Ishaq Nadiri. "Interindustry R&D Spillovers, Rates of
Return, and Production in High-Tech Industries." American Economic Review
78, no. 2 (1988): 429-34.
Biggart, Nicole Woolsey. "Institutionalized Patrimonialism in Korean Business."
Comparative Social Research 12 (1990): 113-33.
Boorstin, Daniel J. The Discoverers. New York: Random House, 1983.
Branscomb, Lewis M., and James H. Keller. "Towards a Research and Innovation
Policy." In Investing in Innovation: Creating a Research and Innovation Policy
That Works, edited by Lewis M. Branscomb and James H. Keller. Cambridge:
MIT Press, 1998.
Branstetter, Lee, and Mariko Sakakibara. "Japanese Research Consortia: A
Microeconometric Analysis of Industrial Policy." NBER Working Paper No.
6066 (1997).
Brunetti, Aymo. "Political Variables in Cross-Country Growth Analysis." Journal of
Economic Surveys 11, no. 2 (1997): 163-90.
Butos, Wilmam N., and Thomas J. McQuade. "Government and Science: A Dangerous
Liaison?" Independent Review 11, no. 2 (2006): 177-208.
Casas, Rosalba, Rebeca de Gortari, and Ma. Josef Santos. "The Building of Knowledge
Spaces in Mexico: A Regional Approach to Networking." Research Policy 29,
no. 2 (2000): 225-41.
Caselli, Francesco, and Wilbur John Coleman. "Cross-Country Technology Diffusion:
The Case of Computers." American Economic Review 91, no. 2 (2001): 328-35.
Chang, Ha-Joon. "The Economic Theory of the Developmental State." In The
Developmental State, edited by Meredith Woo-Cumings. Ithaca: Cornell
University Press, 1999.

198

Cheng, Tun-Jen. "Political Regimes and Development Strategies." In Manufacturing
Miracles: Paths of Industrialization in Latin America and East Asia, edited by
Gary Gereffi and Donald Wyman. Princeton: Princeton Universtiy Press, 1990.
Cheng, Tun-Jen, Stephan Haggard, and David Kang. "Institutions and Growth in Korea
and Taiwan: The Bureaucracy." Journal of Development Studies 34, no. 6
(1998): 87-111.
Chin, Judith C., and Gene M. Grossman. "Intellectual Property Rights and North-South
Trade." NBER Working Paper No. 2769 (1988).
Choi, Younghoon, and Jang-Jae Lee. "Success Factors for Transferring Technology to
Spin-Off Applications: The Case of the Technology Property Rights
Concession Program in Korea." Journal of Technology Transfer 25, no. 2
(2000): 237-46.
Chung, Sung Chul. "Oecd Science, Technology and Industry Outlook 2004: Country
Response to Policy Questionnaire."
http://www.oecd.org/dataoecd/30/60/34242958.pdf: OECD,
http://www.oecd.org/dataoecd/30/60/34242958.pdf, 2004.
———. "The Research, Development and Innovation System in Korea." In Research
and Innovation Policies in the New Global Economy: An International
Comparative Analysis, edited by Philippe Larédo and Philippe Mustar.
Cheltenham: Edward Elgar, 2001.
Cohen, Linda. "When Can Government Subsidize Research Joint Ventures? Politics,
Economics, and Limits to Technology Policy." American Economic Review 84,
no. 2 (1994): 159-63.
Cohen, Wesley M., and Daniel A. Levinthal. "Absorptive Capacity: A New Perspective
on Learning and Innovation." Administrative Science Quarterly 35, no. 1
(1990): 128-52.
Cohen, Wesley M., Richard R. Nelson, and John P. Walsh. "Links and Impacts: The
Influence of Public Research on Industrial R&D." Management Science 48, no.
1 (2002): 1-23.
———. "Protecting Their Intellectual Assets: Appropriabiliy Conditions and Why U.S.
Manufacturing Firms Patent (or Not)." NBER Working Paper, No. 7552, 2000.

199

Cohen, Wesley M., and John P. Walsh. "R&D Spillovers, Appropriability and R&D
Intensity: A Survey Based Approach." draft: ATP, 2000.
Colombo, Massimo G., and Marco Delmastro. "How Effective Are Technology
Incubators? Evidence from Italy." Research Policy 31, no. 7 (2002): 1103-22.
D'Aspremont, Claude, and Alexis Jacquemin. "Cooperative and Noncooperative R&D
in Duopoly with Spillovers." American Economic Review 78, no. 5 (1988):
1133-37.
Dahlman, Carl J., and Thomas Andersson, eds. Korea and the Knowledge-Based
Economy: Making the Transition. Washington, D.C.: World Bank Publications,
2001.
Dasgupta, Partha, and Eric Maskin. "The Simple Economics of Research Portfolios."
The Economic Journal 97 (1987): 581-95.
David, Paul A., and Dominique Foray. "Information Distribution and the Growth of
Economically Valuable Knowledge: A Rationale for Technological
Infrastructure Policies." In Technological Infrastructure Policy: An
International Perspective, edited by Morris Teubal, Dominique Foray, Moshe
Justman and Ehud Zuscovitch, 87-116. New York: Kluwer Academic
Publishers, 1996.
David, Paul A., Bronwyn H. Hall, and Andrew A. Toole. "Is Public R&D a
Complement or Substitute for Private R&D? A Review of the Econometric
Evidence." Research Policy 29, no. 4-5 (2000): 497-529.
Dierdonck, Roland van, Koenraad Debackere, and B. Engelen. "University-Industry
Relationships: How Do the Belgian Academic Community Feel About It?"
Research Policy 19, no. 6 (1990): 551-66.
Dodgson, Mark. Technological Collaboration in Industry: Strategy, Policy and
Internationalization in Innovation. London: Routledge, 1993.
Easterly, William. The Elusive Quest for Growth: Economists' Adventures and
Misadventures in the Tropics. Cambridge: MIT Press, 2001.
Easterly, William, and Ross Levine. "It's Not Factor Accumulation: Stylized Facts and
Growth Models." World Bank Economic Review 15, no. 2 (2001): 177-219.

200

Edquist, Charles. "The Systems of Innovation Approach and Innovation Policy: An
Account of the State of the Art." In DRUID Conference. Aalborg, 2001.
Ernst, Dieter. "Catching-up and Post-Crisis Industrial Upgrading: Searching for New
Sources of Growth in Korea's Electronics Industry." In East-West Center
Working Papers. Washington, D.C.: East-West Center, 2000.
———. "Inter-Organizational Knowledge Outsourcing: What Permits Small
Taiwanese Firms to Compete in the Computer Industry?" In East-West Center
Working Papers: East-West Center, 2000.
Etzkowitz, Henry. "Innovation in Innovation: The Triple Helix of University-Industry-
Government Relations." Social Science Information 42, no. 3 (2003): 293-337.
Etzkowitz, Henry, and Sandra N. Brisolla. "Failure and Success: The Fate of Industrial
Policy in Latin America and South East Asia." Research Policy 28, no. 4
(1999): 337-50.
Etzkowitz, Henry, and Loet Leydesdorff. "The Dynamics of Innovation: From National
Systems and "Mode 2" to a Triple Helix of University-Industry-Government
Relations." Research Policy 29, no. 2 (2000): 109-23.
Evans, Peter B. "Transferable Lessons? Re-Examining the Institutional Prerequisites of
East Asian Economic Policies." Journal of Development Studies 34, no. 6
(1998): 66-86.
———, ed. State-Society Synergy: Government and Social Capital in Development.
University of California International and Area Studies Digital Collection,
Research Series #94, http://respositories.cdlib.org/uciaspubs/research/94, 1997.
Faulkner, Wendy, and Jacqueline Senker. Knowledge Frontiers: Public Sector
Research and Industrial Innovation in Biotechnology, Engineering Ceramics,
and Parallel Computing. Oxford: Clarenden Press, 1994.
Feng, Yi. Democracy, Governance, and Economic Performance. Cambridge: MIT
Press, 2005.
Fields, Karl J. Enterprise and the State in Korea and Taiwan. Ithaca: Cornell
University Press, 1995.

201

Fontana, Roberto, Aldo Guena, and Matt Mireille. 2003. Firm Size and Openness: The
Driving Forces of University-Industry Collaboration. In, A report
commissioned by the OST-DTI, SPRU, University of Sussex, Brighton,
http://www.sussex.ac.uk/Users/prff0/mainpages/home.html.
Frankel, Jeffrey A., and David Romer. "Does Trade Cause Growth?" American
Economic Review 89, no. 3 (1999): 379-99.
Frantzen, Dirk. "Intersectoral and International R&D Knowledge Spillovers and Total
Factor Productivity." Scottish Journal of Political Economy 49, no. 3 (2002):
280-303.
Fritsch, Michael, and Christian Schwirten. "Enterprise-University Co-Operation and
the Role of Public Research Institutions in Regional Innovation Systems."
Industry and Innovation 6, no. 1 (1999): 69-83.
George, Alexander L., and Andrew Bennett. Case Studies and Theory Development in
the Social Science. Cambridge: MIT Press, 2005.
Ginarte, Juan Carlos, and Walter G. Park. "Intellectual Property Rights and Economic
Growth." Contemporary Economic Policy 15, no. 3 (1997): 51-61.
Greene, William H. Econometric Analysis. Upper Saddle River: Prentice Hall, 2002.
Griliches, Zvi. "The Search for R&D Spillovers." In R&D and Productivity: The
Econometric Evidence. Chicago: University of Chicago Press, 1998.
Griliches, Zvi, Haim Regev, and Manuel Trajtenberg. "R&D Policy in Israel: Overview
and Lessons for the Atp." ATP, 2000.
Hagedoorn, John, Albert Link, and Nicholas S. Vonortas. "Research Partnerships."
Research Policy 29 (2000): 567-86.
Haggard, Stephan. "Institutions and Growth in East Asia." Studies in Comparative
International Development 38, no. 4 (2004): 53-81.
Hall, Bronwyn H., Albert N. Link, and John T. Scott. "Universities as Research
Partners." NBER Working Paper No. 7643 (2000).

202

Hall, Bronwyn H., and Alessandro Maffioli. "Evaluating the Impact of Technology
Development Funds in Emerging Economies: Evidence from Latin America."
NBER Working Paper No. 13835 (2008).
Hall, Peter A., and David Soskice, eds. Varieties of Capitalism. Oxford: Oxford
University Press, 2001.
Hall, Robert E., and Charles I. Jones. "Why Do Some Countries Produce So Much
More Output Per Worker Than Others?" Quarterly Journal of Economics 114,
no. 1 (1999): 83-116.
Hamilton, Gary G. "Patriarchy, Patrimonialism and Filial Piety: A Comparison of
China and Western Europe." British Journal of Sociology 41, no. 1 (1990): 77-
104.
Hamilton, Gary G., and Nicole Woolsey Biggart. "Market, Culture, and Authority: A
Comparative Analysis of Management and Organization in the Far East."
American Journal of Socioogy 94, Supplement (1988): S52-S94.
Han, Yongliang. "What Drives R&D Alliances? Evidence from the Biotechnology
Industry." Ph.D. Dissertation, University of Southern California, 2001.
Helliwell, John F. "Empirical Linkages between Democracy and Economic Growth."
British Journal of Political Science 24, no. 2 (1994): 225-48.
Helpman, Elhanan. "Innovation, Imitation, and Intellectual Property Rights."
Econometrica 61, no. 6 (1993): 1247-80.
Henisz, Witold J. "The Institutional Environment for Infrastructure Investment."
Industrial and Corporate Change 11, no. 2 (2002): 355-89.
Holmstrom, Bengt. "Agency Costs and Innovation." Journal of Economic Behavior
and Organization 12, no. 3 (1989): 305-27.
Hsueh, Li-Min, Chen-kuo Hsu, and Dwight H. Perkins, eds. Industrialization and the
State: The Changing Role of Government in Taiwan's Economy, 1945-1998.
Cambridge: Harvard Institute for International Development, 2001.
Jaffe, Adam B. "The Importance of 'Spillovers' in the Policy Mission of the Advanced
Technology Program." Journal of Technology Transfer 23, no. 2 (1998): 11-19.

203

Johnson, Chalmers. Miti and the Japanese Miracle: The Growth of Industrial Policy,
1925-1975. Stanford: Stanford University Press, 1982.
Joly, P.B., and V. Mangamatin. "Profile of Public Laboratories, Industrial Partnerships
and Organisation of R&D: The Dynamics of Industrial Relationships in a Large
Research Organisation." Research Policy 25, no. 6 (1996): 901-22.
Jones, Bryan D. Reconceiving Decision-Making in Democratic Politics: Attention,
Choice, and Public Poliy. Chicago: University of Chicago Press, 1995.
Jones, Charles I. "Convergence Revisited." Journal of Economic Growth 2, no. 2
(1997): 131-53.
———. "Growth: Without or without Scale Effects?" American Economic Review
Papers and Proceedings 89, no. 2 (1999): 139-44.
———. "Time Series Tests of Endogenous Growth Models." Quarterly Journal of
Economics 110, no. 2 (1995): 495-525.
Jones, Charles I., and John C. Williams. "Too Much of a Good Thing? The Economics
of Investment in R&D." Journal of Economic Growth 5, no. 1 (2000): 65-85.
Keefer, Philip, and David Stasavage. "The Limits of Delegation: Veto Players, Central
Bank Independence and the Credibility of Monetary Policy." American
Political Science Review 97, no. 3 (2003): 407-23.
Keller, Wolfgang. "Trade and the Transmission of Technology." Journal of Economic
Growth 7, no. 1 (2002): 5-24.
Kim, Linsu. "The Dynamics of Technological Learning in Industrialisation."
International Social Science Journal 53, no. 168 (2001): 297-308.
———. "Korea's National Innovation System in Transition." In Technology, Learning
and Innovation: Experiences of Newly Industrializing Economies, edited by
Linsu Kim and Richard R. Nelson. Cambridge: Cambridge University Press,
2000.
———. Learning and Innovation in Economic Development. Cheltenham: Edward
Elgar, 1999.

204

———. "Technology Transfer & Intellectual Property Rights: The Korean
Experience." In Intellectual Property Rights and Sustainable Development,
edited by UNCTAD-ICTSD. Geneva: UNCTAD, 2003.
Kim, Linsu, and Richard R. Nelson, eds. Technology, Learning and Innovation:
Experiences of Newly Industrializing Economies. Cambridge: Cambridge
University Press, 2000.
Klenow, Peter, J., and Rodriguez-Clare Andres. "The Neoclassical Revivial in Growth
Economics: Has It Gone Too Far?" In Nber Macroeconomics Annual 1997,
edited by Bill Bernanke and Julio Rotemberg. Cambridge: MIT Press, 1997.
Kogut, Bruce, and Michelle Gittelman. "Public-Private Partnering and Innovation
Performance among U.S. Biotechnology Firms." draft: ATP, 2001.
Kruecken, Georg. "Mission Impossible? Institutional Barriers to the Diffusion of the
"Third Academic Mission" at German Universities." International Journal Of
Technology Management 25, no. 1-2 (2003): 18-33.
Kuznets, Paul W. "An East Asian Model of Economic Development: Japan, Taiwan,
and South Korea." Economic Development and Cultural Change 36, no. 3
(1988): S11-S43.
La Porta, Rafael, Florencio Lopez-De-Silanes, Andrei Shelifer, and Robert Vishny.
"The Quality of Government." Journal of Law, Economics and Organization 15,
no. 1 (1999): 222-79.
Lai, Edwin L.-C. "International Intellectual Property Rights Protection and the Rate of
Product Innovation." Journal of Development Economics 55, no. 1 (1998): 133-
53.
Lall, Sanjaya. "Industry Policy: The Role of Government in Promoting Industrial and
Technological Development." UNCTAD Review (1994).
———. "Technological Change and Industrialization in the Asian Newly
Industrializing Economies: Achievements and Challenges." In Technology,
Learning and Innovation: Experiences of Newly Industrializing Economies,
edited by Linsu Kim and Richard R. Nelson. Cambridge: Cambridge University
Press, 2000.

205

Lall, Sanjaya, and Manuel Albaladejo. "Indicators of the Relative Importance of Iprs in
Developing Countries." ICTSD/UNCTAD; http://www.ictsd.org/unctad-
ictsd/docs/Lall2001.pdf/ (2001).
Lécuyer, Christophe. "Academic Science and Technology in the Service of Industry:
Mit Creates a "Permeable" Engineering School." AEA Papers and Proceedings:
Clio and the Economic Organization of Science 88, no. 2 (1998): 28-33.
Lee, Jinjoo, Zong-Tae Bae, and Dong-Kyu Choi. "Technology Development Processes:
A Model for a Developing Country with a Global Perspective." R&D
Management 18, no. 3 (1988): 235-50.
Lee, Won-Young. "The Role of Science and Technology Policy in Korea's Industrial
Development." In Technology Learning and Innovation: Experiences of Newly
Industrializing Economies, edited by Linsu Kim and Richard R. Nelson.
Cambridge: Cambridge University Press, 2000.
Levy, Brian. "Transaction Costs, the Size of Firms and Industrial Policy: Lessons from
a Comparative Study of the Footwear Industry in Korea and Taiwan." Journal
of Development Economics 34, no. 1 (1991): 151-78.
Mankiw, N. Gregory, David Romer, and David N. Weil. "A Contribution to the
Empirics of Economic Growth." Quarterly Journal of Economics 107, no. 2
(1992): 407-37.
Mansfield, Edwin. "Contribution of R&D to Economic Growth in the United States."
Science 175, no. 4021 (1972): 477-86.
———. "Estimating Social and Private Returns from Innovations Based on the
Advanced Technology Program: Problems and Opportunities." GCR 99-780.
Gaithersburg: National Institute of Standards and Technology, 1996.
———. "Intellectual Property Protection, Direct Investment, and Technology Transfer:
Germany, Japan, and the United States." International Finance Corporate
Discussion Paper, No. 27 (1995).
Markusen, James R. "Contracts, Intellectual Property Rights, and Multinational
Investment in Developing Countries." NBER Working Paper No. W6448
(1998).

206

Marques, Joao, Joao Caraca, and Henrique Diz. "How Can University-Industry-
Government Interactions Change the Innovation Scenario in Portugal? The
Case of the University of Coimbra." Technovation 26, no. 4 (2006): 534-42.
Maskus, Keith E., Kamal Saggi, and Thitima Puttitanun. "Patent Rights and
International Technology Transfer through Direct Investment and Licensing."
In International Public Goods and Transfer of Technology under a Globalized
Intellectual Property Regime, edited by Keith E. Maskus and Jerome H.
Reichman. Cambridge: Cambridge University Press, 2005.
Mathews, John A., and Mei-Chih Hu. "Enhancing the Role of Universities in Building
National Innovative Capacity in Asia: The Case of Taiwan." World
Development 35, no. 6 (2007): 1005-20.
Mincer, Jacob. Schooling, Experience and Earnings. New York: National Bureau of
Economic Research, 1974.
Mowery, David C. "The Roles and Contributions of R&D Collaboration: Matching
Policy Goals and Design." edited by Committee on Science Prepared for the
Hearings of the National Science Policy Study, U.S. House of Representatives,
March 11., 1998.
Mowery, David C., Richard R. Nelson, Bhaven N. Sampat, and Arvids A. Ziedonis.
Ivory Tower and Industrial Innovation: Univeristy-Industry Technology
Transfer before and after the Bayh-Dole Act in the United States. Stanford,
California: Stanford Business Books, 2004.
Nelson, Richard R., and Howard Pack. "The Asian Miracle and Modern Growth
Theory." In Policy Research Working Paper No. 1881: The World Bank
Development Research Group, 1998.
Nelson, Richard R., and Edmund S. Phelps. "Investment in Humans, Technological
Diffusion, and Economic Growth." American Economic Review 56, no. 1/2
(1966): 69-75.
Nelson, Richard R., and Nathan Rosenberg. "Technical Innovation and National
Systems." In National Innovation Systems: A Comparative Analysis, edited by
Richard R. Nelson. New York: Oxford University Press, 1993.
North, Douglass. Growth and Structural Change. New York: Norton, 1981.

207

———. Institutions, Institutional Change, and Economic Performance. Cambridge:
Cambridge University Press, 1990.
OECD. Frascati Manual: Proposed Standard Practice for Surveys on Research and
Experimental Development. Paris: OECD, 2002.
———. "Oecd Science, Technology and Industry Outlook 2004." Paris and
Washington, D.C.: OECD, 2004.
Oxley, Joanne E. "Appropriability Hazards and Governance in Strategic Alliances: A
Transaction Cost Approach." Journal of Law, Economics and Organization 13,
no. 2 (1997): 387-409.
Pack, Howard. "Research and Development in the Industrial Development Process." In
Technology, Learning and Innovation: Experiences of Newly Industrializing
Economies, edited by Linsu Kim and Richard R. Nelson: Cambridge, 2000.
———. "Technological Change and Growth in East Asia: Macro Versus Micro
Perspectives." In Rethinking the East Asian Miracle, edited by Joseph E.
Stiglitz and Shahid Yusuf. New York: Oxford University Press, 2001.
Park, Han Woo, Heung Deug Hong, and Loet Leydesdorff. "A Comparison of the
Knowledge-Based Innovation Systems in the Economies of South Korea and
the Netherlands Using Triple Helix Indicators." Scientometrics 65, no. 1 (2005):
3-27.
Park, Jungsoo. "International and Intersectoral R&D Spillovers in the Oecd and East
Asian Economies." Economic Inquiry 42, no. 4 (2004): 739-57.
Park, Tae-Kyu. "Confucian Values and Contemporary Economic Development in
Korea." In Culture and Economy: The Shaping of Capitalism in Eastern Asia,
edited by Timothy Brook and Hy V. Luong. Ann Arbor: University of Michigan
Press, 2000.
Park, Yung Chul. "Development Lessons from Asia: The Role of Government in South
Korea and Taiwan." American Economic Review 80, no. 2 (1990): 118-21.

208

Patel, Pari, and Keith Pavitt. "National Systems of Innovation under Strain: The
Internationalization of Corporate R&D." In Productivity, Innovation and
Economic Performance, edited by Ray Barrell, Geoff Mason and Mary
O'Mahony. Cambridge: Cambridge University Press, 2000.
Pisano, Gary P. "The R&D Boundaries of the Firm: An Empirical Analysis."
Administrative Science Quarterly 35, no. 1, Special Issue: Technology,
Organizations, and Innovation (1990): 153-76.
Polenske, Karen R., ed. The Economic Geography of Innovation. Cambridge:
Cambridge University Press, 2007.
Powell, Walter W., Kenneth W. Koput, and Laurel Smith-Doerr. "Interorganizational
Collaboration and the Locus of Innovation: Networks of Learning in
Biotechnology." Administrative Science Quarterly 41, no. 1 (1996): 116-45.
Pritchett, Lant. "Divergence, Big Time." World Bank Policy Research Working Paper,
No. 1522 (1995).
Przeworski, Adam, and Henry Teune. The Logic of Comparative Social Inquiry. New
York: Wiley-Interscience, 1970.
Rahm, Dianne, John Kirkland, and Barry Bozeman. University-Industry R&D
Collaboration in the United States, the United Kingdom, and Japan.
Netherlands: Kluwer Academic Publishers, 1999.
Riccaboni, Massimo, Walter W. Powell, Fabio Pammoli, and Jason Owen-Smith.
"Public Research and Industrial Innovation: A Comparison of U.S. And
European Innovation Systems in the Life Science." In Science and Innovation:
Rethinking the Rationales for Funding and Governance, edited by Aldo Guena,
Ammon J. Salter and W. Edward Steinmueller. Cheltenham: Edward Elgar,
2003.
Rodrik, Dani, Arvind Subramanian, and Francesco Trebbi. "Institutions Rule: The
Primacy of Institutions over Geography and Integration in Economic
Development." Journal of Economic Growth 9, no. 2 (2004): 131-65.
Romer, Paul. "Endogenous Technological Change." Journal of Political Economy 98,
no. 5 (1990): S71-S102.

209

Roos, Daniel, Frank Field, and James Neely. "Industry Consortia." In Investing in
Innovation: Creating a Research and Innovation Policy That Works, edited by
Lewis M. Branscomb and James H. Keller. Cambridge: MIT Press, 1998.
Rozman, Gilbert. "The Confucian Faces of Capitalism." In The Emergence of Modern
Pacific Asia, edited by Mark Borthwick. Boudler: Westview Press, 1992.
Ruegg, Rosalie, and Irwin Feller. A Toolkit for Evaluating Public R&D Investment:
Models, Methods, and Findings from Atp's First Decade. Washington, D.C.:
National Institute of Standards and Technology & U.S. Department of
Commerce, 2003.
Saad, Mohammed, and Girma Zawdie. "From Technology Transfer to the Emergence
of a Triple Helix Culture: The Experience of Algeria in Innovation and
Technological Capability Development." Technology Analysis and Strategic
Management 17, no. 1 (2005): 89-103.
Saggi, Kamal. "Trade, Foreign Direct Investment, and International Technology
Transfer: A Survey." World Bank Research Observer 17, no. 2 (2002): 191-235.
Sakakibara, Mariko. "Cooperative Research and Development: Theory and Evidence
on Japanese Practice." Doctoral Dissertation, Harvard University, 1994.
———. "Evaluating Government-Sponsored R&D Consortia in Japan: Who Benefits
and How?" Research Policy 26, no. 4-5 (1997): 447-73.
Sakakibara, Mariko, and Lee Branstetter. "Measuring the Impact of Us Research
Consortia." Managerial and Decision Economics 24, no. 2-3 (2003): 51-69.
Saxenian, Annalee. Regional Advantage: Culture and Competition in Silicon Valley
and Route 128. Cambridge: Harvard University Press, 1994.
Schartinger, Doris, Christian Rammer, Manfred M. Fischer, and Josef Frohlich.
"Knowledge Interactions between Universities and Industries in Austria:
Sectoral Patterns and Determinants." Research Policy 31, no. 3 (2002): 303-28.

210

Scott, Alister, Grove Steyn, Aldo Guena, Stefano Brusoni, and Ed Steinmueller. "The
Economic Returns to Basic Research and the Benefits of University-Industry
Relationships: A Literature Review and Update of Findings." In A report
commissioned by the OST-DTI, SPRU, University of Sussex, Brighton.
Available from http://www.sussex.ac.uk/Users/prff0/mainpages/home.html,
2001.
Seong, Somi, Steven W. Popper, and Kungang Zheng. Strategic Choices in Science
and Technology: Korea in the Era of a Rising China. Santa Monica: Rand
Corporation, 2005.
Shapiro, Matthew. "Public-Private R&D Collaboration in Korea and Taiwan: The
Ramifications of Rent-Seeking Behavior on Tournament Efficiency."
unpublished manuscript, 2007.
———. "The Triple Helix Paradigm in Korea: A Test for New Forms of Capital."
International Journal of Technology Management and Sustainable
Development (forthcoming).
Skocpol, Theda. "Bringing the State Back In." Items 36, no. 1-2 (1982).
Stiglitz, Joseph E., and Scott J. Wallsten. "Public-Private Technology Partnerships -
Promises and Pitfalls." American Behavioral Scientist 43, no. 1 (1999): 52-73.
Tassey, George. The Economics of R&D Policy. Westport, Connecticut: Quorum
Books, 1997.
Tushman, Michael L., and Richard R. Nelson. "Introduction: Technology,
Organizations, and Innovation." Administrative Science Quarterly 35, no. 1
(1990): 1-8.
UNIDO. "Capability Building for Catching-Up: Historical, Empirical and Policy
Dimensions." United Nations Industrial Development Organization, 2005.
USPTO. "United States Patent and Trademark Office Database."
http://www.uspto.gov/patft/, 2008.

211

Van Looy, B., K. Debackere, J. Callaert, R. Tijssen, and T. van Leeuwen. "Scientific
Capabilities and Technological Performance of National Innovation Systems:
An Exploration of Emerging Industrial Relevant Research Domains."
Scientometrics 66, no. 2 (2006): 295-310.
von Hippel, Eric. The Sources of Innovation. New York: Oxford University Press, 1988.
Wade, Robert. Governing the Market: Economic Theory and the Role of Government in
East Asian Industrialization. Princeton: Princeton University Press, 1990.
Web-of-Science, ISI. "Isi Web-of-Science Database." isiknowledge.com, 2008.
Westphal, Larry E. "Industrial Policy in an Export Propelled Economy: Lessons from
South Korea's Experience." Journal of Economic Perspectives 4, no. 3 (1990):
41-59.
Woo-Cumings, Meredith. "Introduction: Chalmers Johnson and the Politics of
Nationalism and Development." In The Developmental State, edited by
Meredith Woo-Cumings. Ithaca: Cornell University Press, 1999.
Woo, Jung-en. Race to the Swift: State and Finance in Korean Industrialization. New
York: Columbia University Press, 1991.
World Bank. The East Asian Miracle: Economic Growth and Public Policy. New York:
World Bank, Oxford University Press, 1993.
Yang, Guifang, and Keith E. Maskus. "Intellectual Property Rights, Licensing, and
Innovation." World Bank Policy Research Working Paper, No. 2973, 2003.
Yusuf, Shahid. Innovative East Asia: The Future of Growth. New York: World Bank
and Oxford University Press, 2003.

212

Appendices

Appendix 0.1 Country abbreviation details
AU Australia
AT Austria
BE Belgium
CA Canada
DK Denmark
FI Finland
FR France
DE Germany
GR Greece
HU Hungary
IS Iceland
IE Ireland
IT Italy
JP Japan
KR Korea
LU Luxembourg
MX Mexico
NL Netherlands
NZ New Zealand
NO Norway
PT Portugal
ES Spain
SE Sweden
CH Switzerland
TR Turkey
GB United Kingdom
US United States
AR Argentina
CN China
IL Israel
RU Russia
SG Singapore
SI Slovenia
ZA South Africa
TW Taiwan

213

Appendix 1.1 Pairwise correlation coefficients for six measures of information
transfers

a) Aggregated sample
| patents papers confer hire contract consult
-------------+------------------------------------------------------
patents_trans | 1.0000
papers_trans | 0.5626* 1.0000
confer_trans | 0.3723* 0.4379* 1.0000
hire_trans | 0.3607* 0.3833* 0.2515* 1.0000
contract_trans| 0.2306* 0.3119* 0.3106* 0.3486* 1.0000
consult_trans | 0.3128* 0.3770* 0.3825* 0.3313* 0.6320* 1.0000

b) Korean public sub-sample
| patents papers confer hire contract consult
-------------+------------------------------------------------------
patents_trans | 1.0000
papers_trans | 0.5428* 1.0000
confer_trans | 0.1780 0.2262* 1.0000
hire_trans | 0.2989* 0.4778* 0.1168 1.0000
contract_trans| 0.1340 0.2277* 0.5127* 0.2679* 1.0000
consult_trans | 0.3672* 0.3955* 0.4725* 0.3046* 0.5212* 1.0000

c) Korean private sub-sample
| patents papers confer hire contract consult
-------------+------------------------------------------------------
patents_trans | 1.0000
papers_trans | 0.4884* 1.0000
confer_trans | 0.5181* 0.4454* 1.0000
hire_trans | 0.4074* 0.3077 0.4313* 1.0000
contract_trans| 0.2357 0.3776* 0.4454* 0.5567* 1.0000
consult_trans | 0.4609* 0.4166* 0.7629* 0.4641* 0.4631* 1.0000

d) Taiwan public sub-sample
| patents papers confer hire contract consult
-------------+------------------------------------------------------
patents_trans | 1.0000
papers_trans | 0.5342* 1.0000
confer_trans | 0.3615* 0.5625* 1.0000
hire_trans | 0.4000* 0.5786* 0.3041* 1.0000
contract_trans| 0.2892* 0.2006 0.3477* 0.3429* 1.0000
consult_trans | 0.3109* 0.2186 0.4825* 0.1659 0.7171* 1.0000

e) Taiwan private sub-sample
| patents papers confer hire contract consult
-------------+------------------------------------------------------
patents_trans | 1.0000
papers_trans | 0.5519* 1.0000
confer_trans | 0.4650* 0.5239* 1.0000
hire_trans | 0.3950* 0.2995* 0.3241* 1.0000
contract_trans| 0.3584* 0.4991* 0.4061* 0.3631* 1.0000
consult_trans | 0.2554* 0.4447* 0.3599* 0.3742* 0.7288* 1.0000

Note: See Appendix 1.3 for details on variable notation.

214

Appendix 1.2 Pairwise correlation coefficients for key (dependent and independent)
variables

a) Aggregated sample
| fa_trans pubfun~g otl_good geo_coll
-------------+------------------------------------
fa_trans | 1.0000
pubfundins~g | 0.2198* 1.0000
otl_good | 0.1733* 0.0709* 1.0000
geo_coll | 0.2531* 0.1192* 0.1868* 1.0000

b) Korean public sub-sample
| fa_trans pubfun~g otl_good geo_coll
-------------+------------------------------------
fa_trans | 1.0000
pubfundins~g | 0.0200 1.0000
otl_good | 0.4009* 0.0582 1.0000
geo_coll | 0.4049* -0.0957* 0.3368* 1.0000

c) Korean private sub-sample
| fa_trans pubfun~g otl_good geo_coll
-------------+------------------------------------
fa_trans | 1.0000
pubfundins~g | 0.2335* 1.0000
otl_good | 0.3029* 0.0301 1.0000
geo_coll | 0.1982* -0.0234 0.3077* 1.0000

d) Taiwan public sub-sample
| fa_trans pubfun~g otl_good geo_coll
-------------+------------------------------------
fa_trans | 1.0000
pubfundins~g | 0.3825* 1.0000
otl_good | 0.1353* 0.0425 1.0000
geo_coll | 0.3417* 0.1803* -0.0903* 1.0000

e) Taiwan private sub-sample
| fa_trans pubfun~g otl_good geo_coll
-------------+------------------------------------
fa_trans | 1.0000
pubfundins~g | 0.2859* 1.0000
otl_good | 0.0634 0.1244* 1.0000
geo_coll | 0.1938* 0.3064* 0.1757* 1.0000

Note: See Appendix 1.3 for details on variable notation.

215

Appendix 1.3 Variable notation (Chapter 1)
Variable name Variable description
patents_trans 1-7 Likert scale score (7 being greatest) for the degree to which
information transfers from the opposite sector through patents
papers_trans 1-7 Likert scale score (7 being greatest) for the degree to which
information transfers from the opposite sector through publications
confer_trans 1-7 Likert scale score (7 being greatest) for the degree to which
information transfers from the opposite sector through conferences
hire_trans 1-7 Likert scale score (7 being greatest) for the degree to which
information transfers from the opposite sector through hires
contract_trans 1-7 Likert scale score (7 being greatest) for the degree to which
information transfers from the opposite sector through contract
research
consult_trans 1-7 Likert scale score (7 being greatest) for the degree to which
information transfers from the opposite sector through consultations
patcol Number of patents in year 2005 from cross-sector R&D collaboration
pubscol Number of publications in year 2005 from cross-sector R&D
collaboration
fa_trans Composite measure of six forms of information transfers, based on
factor loadings (factor patterns) for transfers through patents,
publications, conferences, hires, contract research, and
consultations.
fa_trans_2 Same as fa_trans, excluding transfers through patents and
publications.
pubfundinstig 1-7 Likert scale score (7 being greatest) for the degree to which
public-private R&D collaborations occur as a result of receiving
public funding
otl_good 1-7 Likert scale score (7 being gratest) for the degree to which the
affiliated technology licensing office (OTL) satisfies the needs of the
respondent
otlformdum Dummy variable. 0 if OTL duration is two years or less, 1 if it is
greater than 2 years.
geo_coll 1-7 Likert scale score (7 being greatest) for the degree to which
geographic proximity positively affects the decision to collaborate
with research entities from the opposite sector
taiwan_dummy Dummy variable. 0 for Korea, 1 for Taiwan
private_dummy Dummy variable. 0 for public, 1 for private
taiwan*private Interaction term for taiwan_dummy*private_dummy
Industry dummy
variables included for
the following sectors
(except where
indicated):
Chemical manufacturing, machinery manufacturing, computer and
electronic product manufacturing, electrical equipment,
miscellaneous manufacturing.

216

217
Appendix 2.1 Variable notation (Chapter 2)
Variable name Variable description
old_relations Percentage of PPRD collaboration originating from personal ties. (Data
points are from zero to ten, representing percentage values from zero to
one hundred)
sameuni Personal ties based on university ties dummy variable
sameunilab Personal ties based on former university laboratory ties dummy variable
samefirm Personal ties based on former private firm ties dummy variable
sameproj Personal ties based on multiple previous projects dummy variable
sameconf Personal ties based on meeting at a conference dummy variable
new_relations Percentage of PPRD collaboration done with partners from previous
projects (Data points are from zero to ten, representing percentage values
from zero to one hundred)
noother Repartnering affected by a lack of other qualified partners
(seven-point Likert scale response, seven being greatest)
fundstip Repartnering affected by funding stipulation
(seven-point Likert scale response, seven being greatest)
sharecom Repartnering affected by a shared commitment
(seven-point Likert scale response, seven being greatest)
lacktension Repartnering affected by a lack of tension
(seven-point Likert scale response, seven being greatest)
easecom Repartnering affected by ease of communication
(seven-point Likert scale response, seven being greatest)
compknow Repartnering affected by complementarity in knowledge
(seven-point Likert scale response, seven being greatest)
trust Repartnering affected by presence of trust
(seven-point Likert scale response, seven being greatest)
expcom Repartnering affected by expected commercialisation
(seven-point Likert scale response, seven being greatest)
coll_large Percentage of PPRD collaboration done with large firm partners (Data
points are from zero to ten, representing percentage values from zero to
one hundred)
coll_sme Percentage of PPRD collaboration done with SME partners (Data points are
from zero to ten, representing percentage values from zero to one hundred)
coll_uni Percentage of PPRD collaboration done with university partners (Data
points are from zero to ten, representing percentage values from zero to
one hundred)
coll_gri Percentage of PPRD collaboration done with GRI partners (Data points are
from zero to ten, representing percentage values from zero to one hundred)

Appendix 2.2a Robustness checks
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
patcol patcol patcol patcol patcol patcol
--------------------------------------------------------------------------------------------------------------------
prevpart 0.121** 0.105** -0.0123 -0.0487
(0.0365) (0.0371) (0.0505) (0.0595)

persties 0.1000** 0.0782* -0.0416 -0.0663
(0.0365) (0.0369) (0.0486) (0.0572)

prevpart_persties 0.0279*** 0.0308*** 0.0384**
(0.00746) (0.00721) (0.0118)

country 0.126 0.0117 0.0316 0.00262 0.0386 0.0359
(0.224) (0.229) (0.227) (0.222) (0.223) (0.223)

sector -0.00635 0.0299 0.0424 0.0393 0.0228 0.0152
(0.222) (0.224) (0.222) (0.218) (0.218) (0.218)

Constant -0.00936 0.112 -0.221 0.0854 0.139 0.301
(0.225) (0.217) (0.245) (0.222) (0.211) (0.289)
--------------------------------------------------------------------------------------------------------------------
Observations 299 299 299 299 299 299
R-squared 0.037 0.026 0.052 0.081 0.083 0.085
F 3.789 2.619 3.997 6.452 6.635 5.436
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
+ p<.10, * p<.05, ** p<.01, *** p<.001

218

Appendix 2.2b Robustness checks (industry controls included)
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
patcol patcol patcol patcol patcol patcol
--------------------------------------------------------------------------------------------------------------------
prevpart 0.133*** 0.114** -0.0111 -0.0542
(0.0373) (0.0379) (0.0514) (0.0612)

persties 0.109** 0.0848* -0.0486 -0.0772
(0.0372) (0.0376) (0.0499) (0.0595)

prevpart_persties 0.0300*** 0.0339*** 0.0425***
(0.00759) (0.00743) (0.0122)

country 0.140 -0.0118 0.0278 0.0120 0.0612 0.0610
(0.232) (0.239) (0.236) (0.229) (0.232) (0.232)

sector 0.115 0.154 0.182 0.177 0.155 0.141
(0.236) (0.239) (0.236) (0.231) (0.232) (0.232)

Constant -0.0877 -0.0236 -0.364 -0.0513 0.0382 0.215
(0.309) (0.315) (0.330) (0.301) (0.304) (0.364)
--------------------------------------------------------------------------------------------------------------------
Observations 299 299 299 299 299 299
R-squared 0.059 0.046 0.076 0.108 0.111 0.114
F 1.382 1.061 1.666 2.466 2.538 2.419
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
+ p<.10, * p<.05, ** p<.01, *** p<.001

219

Appendix 2.2c Robustness checks (industry, respondent controls included)
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
patcol patcol patcol patcol patcol patcol
--------------------------------------------------------------------------------------------------------------------
prevpart 0.142*** 0.125** -0.00919 -0.0533
(0.0403) (0.0408) (0.0553) (0.0662)

persties 0.116** 0.0922* -0.0493 -0.0778
(0.0408) (0.0409) (0.0537) (0.0644)

prevpart_persties 0.0321*** 0.0361*** 0.0446***
(0.00824) (0.00798) (0.0132)

country 0.176 0.0221 0.0667 0.0578 0.104 0.104
(0.267) (0.274) (0.270) (0.262) (0.264) (0.264)

sector 0.306 0.341 0.378 0.404 0.387 0.382
(0.311) (0.315) (0.310) (0.304) (0.304) (0.304)

Constant -0.787 -0.729 -1.203 -0.688 -0.544 -0.298
(0.879) (0.892) (0.892) (0.856) (0.861) (0.914)
--------------------------------------------------------------------------------------------------------------------
Observations 273 273 273 273 273 273
R-squared 0.070 0.054 0.088 0.122 0.125 0.127
F 1.123 0.862 1.360 1.962 2.013 1.939
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
+ p<.10, * p<.05, ** p<.01, *** p<.001

220

Appendix 2.2d Robustness checks
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
nocollpat nocollpat nocollpat nocollpat nocollpat nocollpat
--------------------------------------------------------------------------------------------------------------------
prevpart 0.724+ 0.785+ 0.558 -0.0175
(0.415) (0.424) (0.587) (0.690)

persties -0.127 -0.291 -1.039+ -1.048
(0.414) (0.422) (0.563) (0.664)

prevpart_persties 0.0347 0.198* 0.201
(0.0867) (0.0836) (0.136)

country -0.346 -0.146 0.00387 -0.500 0.0272 0.0263
(2.542) (2.604) (2.595) (2.575) (2.585) (2.590)

sector -3.145 -3.420 -3.326 -3.088 -3.466 -3.468
(2.520) (2.546) (2.536) (2.528) (2.527) (2.533)

Constant 3.840 7.127** 4.628+ 3.958 7.305** 7.363*
(2.557) (2.466) (2.803) (2.577) (2.448) (3.356)
--------------------------------------------------------------------------------------------------------------------
Observations 299 299 299 299 299 299
R-squared 0.017 0.007 0.019 0.018 0.026 0.026
F 1.728 0.736 1.413 1.333 1.969 1.570
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
+ p<.10, * p<.05, ** p<.01, *** p<.001

221

Appendix 2.2e Robustness checks (industry controls included)
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
nocollpat nocollpat nocollpat nocollpat nocollpat nocollpat
--------------------------------------------------------------------------------------------------------------------
prevpart 0.761+ 0.836+ 0.655 0.0237
(0.426) (0.437) (0.604) (0.717)

persties -0.175 -0.351 -1.143+ -1.131
(0.425) (0.433) (0.584) (0.697)

prevpart_persties 0.0220 0.208* 0.204
(0.0891) (0.0870) (0.143)

country -0.349 -0.173 0.116 -0.443 0.275 0.276
(2.656) (2.727) (2.718) (2.687) (2.711) (2.716)

sector -3.220 -3.705 -3.500 -3.175 -3.703 -3.697
(2.699) (2.733) (2.723) (2.710) (2.711) (2.721)

Constant 7.158* 10.78** 8.299* 7.185* 11.16** 11.08**
(3.529) (3.590) (3.801) (3.536) (3.564) (4.267)
--------------------------------------------------------------------------------------------------------------------
Observations 299 299 299 299 299 299
R-squared 0.027 0.017 0.030 0.028 0.036 0.036
F 0.616 0.380 0.618 0.574 0.768 0.714
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
+ p<.10, * p<.05, ** p<.01, *** p<.001

222

Appendix 2.2f Robustness checks (industry, respondent controls included)
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
nocollpat nocollpat nocollpat nocollpat nocollpat nocollpat
--------------------------------------------------------------------------------------------------------------------
prevpart 0.845+ 0.892+ 0.613 -0.0177
(0.460) (0.469) (0.648) (0.775)

persties -0.0711 -0.244 -1.104+ -1.113
(0.464) (0.471) (0.628) (0.754)

prevpart_persties 0.0491 0.225* 0.228
(0.0966) (0.0933) (0.155)

country 0.459 0.428 0.748 0.279 0.938 0.938
(3.045) (3.111) (3.100) (3.070) (3.090) (3.096)

sector -3.562 -4.019 -3.753 -3.411 -3.733 -3.735
(3.544) (3.583) (3.568) (3.561) (3.552) (3.560)

Constant 4.632 9.129 5.734 4.783 10.28 10.36
(10.02) (10.15) (10.25) (10.03) (10.06) (10.70)
--------------------------------------------------------------------------------------------------------------------
Observations 273 273 273 273 273 273
R-squared 0.051 0.039 0.052 0.052 0.060 0.060
F 0.808 0.603 0.776 0.775 0.903 0.852
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
+ p<.10, * p<.05, ** p<.01, *** p<.001
223

Appendix 3.1 Testing for overall patenting effects on TFP, linear and nonlinear, accounting for overall time trends
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
tfp tfp tfp tfp tfp tfp
--------------------------------------------------------------------------------------------------------------------
all_patents 0.00000309*** 0.00000654*** 0.00000230*** 0.00000520*** 0.00000238*** 0.00000512***
(0.000000345) (0.000000811) (0.000000361) (0.000000833) (0.000000360) (0.000000829)

all_patents_sq -3.12e-12*** -2.54e-12*** -2.41e-12***
(6.67e-13) (6.61e-13) (6.60e-13)

yearcode 0.112*** 0.0993***
(0.0192) (0.0192)

y1980 0.212* 0.185
(0.104) (0.103)

y1985 0.507*** 0.462***
(0.105) (0.105)

y1990 0.557*** 0.505***
(0.105) (0.104)

y1995 0.535*** 0.467***
(0.106) (0.107)

y2000 0.575*** 0.519***
(0.111) (0.110)

Constant 2.265*** 2.153*** 2.037*** 1.972*** 1.912*** 1.862***
(0.0392) (0.0452) (0.0545) (0.0564) (0.0747) (0.0751)
--------------------------------------------------------------------------------------------------------------------
Observations 662 662 662 662 662 662
R-squared 0.130 0.164 0.182 0.204 0.200 0.220
F 80.21 52.64 59.61 45.71 22.14 21.32
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001

224

Appendix 3.2 Testing for Tier-1 collaborative patenting effects on TFP, linear and nonlinear, accounting for overall time
trends
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
tfp tfp tfp tfp tfp tfp
--------------------------------------------------------------------------------------------------------------------
tier1_collab 0.0000233*** 0.0000408*** 0.0000187*** 0.0000342*** 0.0000198*** 0.0000357***
(0.00000218) (0.00000368) (0.00000229) (0.00000393) (0.00000228) (0.00000390)

tier1_collab_sq -9.08e-11*** -7.58e-11*** -7.75e-11***
(1.56e-11) (1.58e-11) (1.56e-11)

yearcode 0.100*** 0.0801***
(0.0187) (0.0188)

y1980 0.227* 0.212*
(0.101) (0.0991)

y1985 0.528*** 0.495***
(0.102) (0.0999)

y1990 0.557*** 0.505***
(0.101) (0.0997)

y1995 0.501*** 0.419***
(0.103) (0.102)

y2000 0.518*** 0.417***
(0.107) (0.107)

Constant 2.332*** 2.269*** 2.108*** 2.100*** 1.961*** 1.948***
(0.0329) (0.0338) (0.0528) (0.0518) (0.0722) (0.0706)
--------------------------------------------------------------------------------------------------------------------
Observations 662 662 662 662 662 662
R-squared 0.176 0.225 0.218 0.250 0.242 0.276
F 114.6 77.66 74.61 59.47 28.22 28.80
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001

225

Appendix 3.3 Testing for overall patenting effects on TFP, linear and nonlinear, accounting for overall time trends and
percentage of all patenting done in collaboration
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
tfp tfp tfp tfp tfp tfp
--------------------------------------------------------------------------------------------------------------------
all_patents 0.00000307*** 0.00000649*** 0.00000230*** 0.00000520*** 0.00000238*** 0.00000512***
(0.000000345) (0.000000811) (0.000000361) (0.000000834) (0.000000361) (0.000000830)

all_patents_sq -3.10e-12*** -2.54e-12*** -2.41e-12***
(6.67e-13) (6.62e-13) (6.60e-13)

perc_t1_all 0.213 0.189 -0.00280 0.00168 0.0125 0.0167
(0.162) (0.159) (0.162) (0.160) (0.161) (0.159)

yearcode 0.112*** 0.0992***
(0.0197) (0.0197)

y1980 0.211* 0.184
(0.104) (0.103)

y1985 0.507*** 0.462***
(0.105) (0.105)

y1990 0.556*** 0.504***
(0.106) (0.105)

y1995 0.533*** 0.465***
(0.108) (0.109)

y2000 0.573*** 0.517***
(0.113) (0.113)

Constant 2.231*** 2.124*** 2.037*** 1.972*** 1.910*** 1.861***
(0.0470) (0.0516) (0.0570) (0.0588) (0.0767) (0.0771)
--------------------------------------------------------------------------------------------------------------------
Observations 662 662 662 662 662 662
R-squared 0.133 0.167 0.182 0.204 0.200 0.220
F 41.03 35.59 39.67 34.22 18.95 18.62
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001

226

Appendix 3.4 Testing for patenting effects on TFP, linear and nonlinear, accounting for overall time trends
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
tfp tfp tfp tfp tfp tfp
--------------------------------------------------------------------------------------------------------------------
tier1_collab 0.0000231*** 0.0000406*** 0.0000187*** 0.0000342*** 0.0000198*** 0.0000357***
(0.00000219) (0.00000369) (0.00000229) (0.00000393) (0.00000228) (0.00000390)

tier1_collab_sq -9.04e-11*** -7.57e-11*** -7.74e-11***
(1.57e-11) (1.58e-11) (1.56e-11)

perc_t1_all 0.119 0.0814 -0.0619 -0.0573 -0.0472 -0.0421
(0.158) (0.154) (0.158) (0.155) (0.156) (0.153)

yearcode 0.102*** 0.0816***
(0.0192) (0.0193)

y1980 0.228* 0.213*
(0.101) (0.0993)

y1985 0.529*** 0.497***
(0.102) (0.100)

y1990 0.561*** 0.509***
(0.102) (0.101)

y1995 0.506*** 0.424***
(0.105) (0.104)

y2000 0.524*** 0.422***
(0.109) (0.108)

Constant 2.313*** 2.256*** 2.114*** 2.106*** 1.966*** 1.953***
(0.0415) (0.0415) (0.0552) (0.0541) (0.0740) (0.0725)
--------------------------------------------------------------------------------------------------------------------
Observations 662 662 662 662 662 662
R-squared 0.177 0.225 0.218 0.251 0.242 0.276
F 57.54 51.80 49.71 44.56 24.16 25.16
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001

227

Appendix 3.5 Testing for institutional effects on the log of Tier-1 collaboration, accounting for possible interactions between IPRs and
democracy
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
log_tier1_~b log_tier1_~b log_tier1_~b log_tier1_~b log_tier1_~b log_tier1_~b
--------------------------------------------------------------------------------------------------------------------
log_ipr 1.283** 1.383*** 0.538 0.735
(0.404) (0.417) (0.643) (0.930)

log_polity 0.0797 0.253 -0.0788 0.0834
(0.164) (0.182) (0.197) (0.284)

log_ipr_polity 0.420 0.621*** 0.326
(0.268) (0.187) (0.418)

yearcode 0.375*** 0.439*** 0.322*** 0.330*** 0.343*** 0.324***
(0.0615) (0.0462) (0.0703) (0.0669) (0.0661) (0.0704)

Constant 2.519*** 2.337*** 2.173*** 2.635*** 2.865*** 2.493***
(0.233) (0.222) (0.353) (0.245) (0.270) (0.541)
--------------------------------------------------------------------------------------------------------------------
Observations 619 829 587 587 587 587
R-squared 0.235 0.159 0.240 0.240 0.240 0.241
F 77.83 63.78 50.22 50.45 50.21 37.79
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
+ p<.10, * p<.05, ** p<.01, *** p<.001

228

229
Appendix 3.6 Testing for institutional effects on the log of Tier-1 collaboration, accounting for possible interactions
between IPRs and check-and-balances
--------------------------------------------------------------------------------------------------------------------
(1) (2) (3) (4) (5) (6)
log_tier1_~b log_tier1_~b log_tier1_~b log_tier1_~b log_tier1_~b log_tier1_~b
--------------------------------------------------------------------------------------------------------------------
log_ipr 1.283** 1.221** 1.334** 1.334**
(0.404) (0.405) (0.444) (0.444)

log_polcon_checks 0.121 -0.114 -1.334** 0
(0.175) (0.197) (0.444) (0)

log_ipr_polconchecks -0.114 1.221** -0.114
(0.197) (0.405) (0.197)

yearcode 0.375*** 0.444*** 0.396*** 0.396*** 0.396*** 0.396***
(0.0615) (0.0430) (0.0674) (0.0674) (0.0674) (0.0674)

Constant 2.519*** 2.518*** 2.768*** 2.768*** 2.768*** 2.768***
(0.233) (0.201) (0.308) (0.308) (0.308) (0.308)
--------------------------------------------------------------------------------------------------------------------
Observations 619 807 583 583 583 583
R-squared 0.235 0.171 0.243 0.243 0.243 0.243
F 77.83 67.25 50.53 50.53 50.53 50.53
--------------------------------------------------------------------------------------------------------------------
Standard errors in parentheses
+ p<.10, * p<.05, ** p<.01, *** p<.001

Appendix 3.7 Variable notation (Chapter 3)
Variable name Variable description
log_y Natural log of GDP per capita
log_k Natural log of physical capital measure, denoted in Eq (3.9)
log_h Natural log of labor-augmenting human capital, denoted in Eq (3.2)
new_tfp Revised TFP measure, based on Eq (3.9), post-antilog
all_patents Number of per capita patents generated in a particular time period,
multiplied by 1 million
all_patents_sq all_patents squared
tier1_collab Number of per capita collaborative patents with Tier-1 countries
generated in a particular time period, multiplied by 1 million
tier1_collab_sq tier1_collab squared
log_allpatents Natural log of all_patents
log_allpatents_sq Square of log_all_patents
log_tier1collab Natural log of tier1_collab
log_tier1collab_sq Square of log_tier1_collab
log_notier1 Natural log of all_patents minus tier1_collab
perc_t1_all Percentage of all patents done in collaboration with Tier-1 countries
(tier1_collab/all_patents)
all_perc_t1_all Interaction term of all per capita patents and the percentage of all
patents done in collaboration (all_patents*perc_t1_all)
yearcode 0 to 5 coded variable for each five-year time period, from 1975 to 2000
y1975, …, y2000 Dummy variables (equaling “1”) for each time period
log_ipr Natural log of the Ginarte-Park IPR index
log_polity Natural log of the POLITY IV score
log_polcon_checks Natural log of the POLCON checks-and-balances score
log_ipr_polity Natural log of the IPR-POLITY interaction term
log_ipr_polconchecks Natural log of the IPR-POLCON interaction term

230

Appendix 3.8 Summary statistics
Variable name Obs Mean Std. Dev. Min Max
log_y 969 8.143594 1.203312 4.816339 10.82645

log_k 969 8.478562 1.873212 3.660103 12.46541

log_h 675 .5536859 .285509 .009 1.205

tfp 662 2.485968 4.106354 .0993158 23.35948

pc_all_patents 1044 .0505335 .1628203 0 1.356272

pc_tier1_collab 1044 .0054518 .022069 0 .3036036

all_patents 1044 50533.53 162820.3 0 1356272

all_patents_sq 1044 2.90e+10 1.51e+11 0 1.84e+12

tier1_collab 1044 5451.783 22069.03 0 303603.6

tier1_collab_sq 1044 5.16e+08 4.36e+09 0 9.22e+10

log_allpatents 1044 5.110431 4.580593 0 14.12025

log_allpatents_sq 1044 47.07825 53.60428 0 199.3815

log_tier1collab 1044 3.322462 3.870907 0 12.62348

log_tier1collab_sq 1044 26.00833 36.25859 0 159.3522

log_notier1 1044 4.787643 4.591596 0 14.1009
perc_t1_all 1044 .1402982 .2353265 0 1

log_ipr 619 .7597804 .4308303 -.5306282 1.58412

log_polity 829 1.499978 .8440558 0 2.302585

log_polconchecks 807 1.168551 .7440447 0 1.94591

231

Abstract (if available)

Abstract This dissertation speaks to three separate but related issues on R&D collaboration. The first chapter details the various ways in which entities from the public and private research sectors transfer information. In Korea and Taiwan, information transfers are shown to have a positive impact on public-private R&D output in both countries, but government funding is a much stronger predictor of such transfers in Taiwan. The second chapter focuses on the research-based links between the government, universities, and firms, and tests the Triple Helix-based hypothesis that new capital arises from the interactions between the public and private research sectors. Focusing again on Korea and Taiwan, repartnering tendencies (as a proxy for new capital) are found to be strong predictors of R&D output, with virtually no differences between these two countries. In the third chapter, attention is drawn to R&D collaboration between countries, and it is shown that international R&D collaboration has a positive impact on the growth residual (TFP), and that it has a bonus effect to a country's patenting efforts. It is also shown that developing countries derive particular benefit from international R&D collaboration, but such benefits are largely dependent on the presence of strong political institutions and intellectual property rights. In terms of data and methods, Chapters 1 and 2 utilize a unique dataset (KORTAI R&D), which is specific to the Korean and Taiwanese cases, while the macro-level analysis of Chapter 3 uses data for a maximum of 150 countries drawn from the USPTO, WDI, Penn World Table, Barro-Lee education data, the Ginarte-Park IPR index, POLITY IV, and POLCON. OLS and ordered logit statistical methods are applied in Chapters 1 and 2, and fixed effects GLS methods are used in Chapter 3.

Linked assets