You are on page 1of 21

Innovation

Management, Policy & Practice

ISSN: 1447-9338 (Print) 2204-0226 (Online) Journal homepage: https://www.tandfonline.com/loi/rimp20

An international comparison of R&D efficiency


of multiple innovative outputs: The role of the
national innovation system

Chiang-Ping Chen, Jin-Li Hu & Chih-Hai Yang

To cite this article: Chiang-Ping Chen, Jin-Li Hu & Chih-Hai Yang (2011) An international
comparison of R&D efficiency of multiple innovative outputs: The role of the national innovation
system, Innovation, 13:3, 341-360, DOI: 10.5172/impp.2011.13.3.341

To link to this article: https://doi.org/10.5172/impp.2011.13.3.341

Published online: 17 Dec 2014.

Submit your article to this journal

Article views: 270

View related articles

Citing articles: 19 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=rimp20
Copyright © eContent Management Pty Ltd. Innovation: Management, policy & practice (2011) 13: 341–360.

An international comparison of
R&D efficiency of multiple
innovative outputs: The role of the
national innovation system
CHIANG-PING CHEN
Department of Economics, National Central University, Jhongli, Taoyuan County, Taiwan

JIN-LI HU
Institute of Business and Management, National Chiao Tung University, Hsinchu, Taiwan

CHIH-HAI YANG
Department of Economics, National Central University, Jhongli, Taoyuan County, Taiwan

ABSTRACT
This paper compares research and development (R&D) efficiency across nations based on various
output-oriented R&D efficiency indices that are developed by the data envelopment analysis
approach. Empirical results show that nations have similar R & D efficiency in terms of patents and
royalties, while their performance for journal publications is quite different. We then investigate
how the innovation environment, especially in the national innovation system, affects these output-
oriented R&D efficiency indices. R&D intensity, intellectual property rights protection, knowledge
stock, and human capital accumulation all have significantly positive effects on efficiency indices.
Private sector R&D, funded by either foreign sources or funded and performed by private businesses,
plays an important role in improving scores on the output-oriented R&D efficiency index for patents
and for royalties and licensing fees. The R&D intensity performed by higher education institutions
has a positive effect on the journal-oriented R&D efficiency index.

Keywords: output-oriented R&D efficiency index; national innovation system; data envelopment analysis

INTRODUCTION environment by enforcing intellectual property

I nnovation is widely recognized as the primary


driving force of sustainable economic growth1.
Most nations have therefore gradually devoted
rights (IPR) to promote innovations. Whether
R&D can serve as a major source of competitive
advantage depends on efficient and productive
more efforts to research and development (R&D) innovation processes (Werner & Souder 1997).
and have tried to create a favorable innovation Using R&D resources inefficiently may result in

1
See Acemoglu et al. (2006) for a comprehensive survey of theoretical as well as empirical literature on the
innovation–economic growth nexus.

Volume 13, Issue 3, December 2011 INNOVATION: MANAGEMENT, POLICY & PRACTICE 341
Chiang-Ping Chen, Jin-Li Hu and Chih-Hai Yang

slow technological progress and reduce the com- higher education graduates. Although these stud-
plementary effect brought on by other innova- ies contain interesting results, some potential
tive activities. Thus, understanding the nature of room for improvement remains.
R&D efficiency and its determinants is essential To incorporate various outputs and produce
for designing R&D policies that effectively foster a single measure for R&D efficiency, the rela-
innovation and promote technological progress. tive importance of outputs must be determined
Innovation includes both technological and and fixed. However, no matter what sets of
non-technological innovation, suggesting that weights are applied to the evaluation of vari-
measures of national innovative outputs are multi- ous innovative outputs, they cannot determine
dimensional, such as patents, royalties and licens- the individual output-specific R&D efficiency.
ing fees (hereafter, RLF), journal articles, and so Moreover, neither of the above mentioned
on. While patents are widely recognized as a proxy studies considered a scenario in which the same
for innovative output (Griliches 1990), various input factors could simultaneously produce
decision-making units (DMUs) such as firms, multiple outputs in a national innovation pro-
industries, and nations may produce distinct out- duction, nor did they control for the influences
put more efficiently by locating in particular tech- of external environments on R&D efficiency,
nological environments. For instance, some R&D particularly the role of the national innovation
programs are essentially basic-research-oriented system (NIS).
and have the primary objective of publishing aca- Innovation environment is an important fac-
demic journal papers. These programs are partic- tor that affects the efficiency and productivity of
ularly relevant to non-technological innovations. the innovation process for all kinds of DMUs. A
Applied-technology-oriented R&D programs can better innovation environment enables DMUs
alternatively aim at licensing or selling technolo- to conduct R&D more efficiently and conse-
gies. Hence, it is difficult to compare the R&D quently stimulates more innovations, whereas
efficiency of various national R&D programs in an inferior one may induce the inefficient use of
the same context, meaning that comparing R&D R&D resources and result in slower technological
efficiency across nations by focusing on one out- progress. The NIS proposed by Lundvall (1992),
put could lead to a narrow view of national R&D Nelson (1993), and Edquist (1997) focuses on
efficiency. the interplay among the government, higher edu-
While this point is well known among econo- cation institutions, and industries that contrib-
mists, few studies comparing the relative effi- ute to innovative behavior in a specific nation,
ciency of R&D activity across nations measure which may significantly affect R&D efficiency of
such relative efficiency by considering multiple various innovative outputs. Due to differences in
outputs. Lee and Park (2005) and Wang and the composition of national R&D and the inter-
Huang (2007) both evaluated R&D efficiency play among various innovation performers, an
across nations by considering three outputs (pat- individual nation has different compositions of
ents, technology balance of receipts, and journal innovative outputs, given the same R&D input.
articles) and two outputs (patents and SCI and For instance, the objective of university faculties’
EI articles), respectively. Lee et al. (2009) used R&D is the publication of more academic papers,
the Data envelopment analysis (DEA) approach whereas government R&D often focuses on basic
to measure and compare the performance of research and pays less attention to patents and
national R&D programs in South Korea. To royalties than the private sector.
consider the heterogeneity of national R&D This paper aims to compare the relative effi-
programs’ objectives, they included 10 outputs, ciency of R&D across nations and attempts
mainly academic publications, patents, and toww provide the following distinct types of

342 INNOVATION: MANAGEMENT, POLICY & PRACTICE Volume 13, Issue 3, December 2011
Role of the national innovation system in R&D efficiency

empirical evidence: First, this study provides NIS on R&D Efficiency’ we further adopt the
new evidence on national R&D efficiency Tobit model to examine the impacts of environ-
in terms of various innovative outputs. This mental factors, especially NIS, on influencing
nation-level study can lend implications for various indices of national R&D efficiency. The
R&D management as well as innovation policy. final section contains concluding remarks and
Specifically, we provide a more complete picture policy implications.
of the national R&D performance by measur-
ing R&D efficiency with three innovative out- METHODOLOGY AND OUTPUT-
puts: patents, RLF, and journal articles. Second ORIENTED R&D EFFICIENCY INDEX
and importantly, by utilizing the concepts of Methodology of DEA and data
DEA-computed actual and target outputs, we Data envelopment analysis is a mathematical pro-
define an individual output-oriented R&D effi- cedure using the linear programming technique
ciency index (OREI), enabling us to assess the to assess the efficiencies of DMUs. This approach
individual R&D efficiency of various outputs. constructs a non-parametric piecewise frontier of
In evaluating an R&D efficiency score, out-of- DMUs that owns optimal efficiency over datasets
date technology and an inefficient innovation for the comparative efficiency measurement. DMUs
process both generate an insufficient portion located at the efficiency frontier represent the most
of production of individual innovative output efficient DMUs that generate maximum outputs by
which needs to be further adjusted. The amount taking a fixed amount of inputs. DEA has two pri-
of total adjustments, including slack and radial mary advantages: it does not require a specification of
adjustments, is computed by DEA. We obtain either the production function form or the weights of
the target level of individual innovative output different inputs and outputs, and it provides detailed
when the amount of total adjustment increases information on the efficiency of the unit relative to
from the actual individual innovative output and specific efficient units used as comparators (Lovell &
then define the ratio of actual individual innova- Schmidt 1993; Coelli et al. 1998).
tive output to target individual innovative out- Charnes et al. (1978) developed the first DEA
put as: the output-oriented R&D efficiency index. model by assuming constant returns to scale (CRS),
This approach allows us to compare the R&D while Banker et al. (1984) proposed another model
efficiency of individual innovative output across by assuming variable returns to scale (VRS). Both
nations. Third, this study further employs the these two models are well known as the CCR and
Tobit model to examine environmental factors BCC models in the efficiency literature, respectively.
affecting R&D efficiency, especially for the role As the scale of a nation is hard to adjust in the short
of the NIS on impacting the individual OREI. run, most existing studies assume an innovation
This remainder of this paper is organized as production with CRS between inputs and outputs,
follows: section ‘Methodology and Output- e.g., Griffith et al. (2006), Hashimoto and Haneda
Oriented R&D Efficiency Index’ addresses the (2008), Lee and Park (2005), and Parisi et al. (2006).
estimating methodology and describes the data- As the objective of R&D lies in increasing outputs
set used in this study. We first introduce the tech- rather than decreasing inputs, this paper therefore
nique of DEA and then describe how different employs the output-oriented CCR model.
indices of R&D efficiency in terms of innova- In the output-orientated CRS DEA model,
tive outputs are calculated from DEA solutions. there are K inputs and M outputs, represented
Section ‘Estimation of Output-Oriented R&D by the column vectors xi and yi, for each of the
Efficiency Indices’ displays and discusses the N DMUs. The K × N input matrix X and the
empirical results on various indices of R&D effi- M × N output matrix Y represent the data for all
ciency across nations. In section ‘The Role of the N DMUs. The output-oriented CRS DEA model

Volume 13, Issue 3, December 2011 INNOVATION: MANAGEMENT, POLICY & PRACTICE 343
Chiang-Ping Chen, Jin-Li Hu and Chih-Hai Yang

then solves the following linear programming likely to be biased due to the home advantage
problem for firm i in each year: for patent applications. Thus, we adopt the
max φ, number of patents applied for in the European
θ ,λ Patent Office (EPO) and the US Patent and
s.t. − φyi + Y λ ≥ 0 (1) Trademark Office (USPTO), because both
xi − X λ ≥ 0 provide the clearest indication of innovation
λ ≥ 0, performance and can be treated as ‘new-to-the-
where φ is scalar, and λ is an N × 1 vector of world’ innovations4. Moreover, RLF reflect the
constants. ability to sell newly developed and invented
The value of φ is the inverse of the efficiency score technologies, and they can be treated as one of
for the i-th firms, with φ ≥ 1. The value of unity a nation’s the direct commercial outputs from
for φ indicates a point on the frontier and therefore innovation activities. Journal articles pub-
a technically efficient firm, according to Farrell’s lished in scientific and engineering fields are
(1957) definition. A higher value of φ implies a considered as another major output of research
lower efficiency score. The vector λ consists of the and widely used to evaluate the performance
reference weights of DMU i over all DMUs. of researchers (OECD 2001). In this study
The data utilized in this study make up we tally articles published in journals cover-
a panel dataset of 24 nations, including 16 ing physics, biology, chemistry, mathematics,
European nations, 4 Asian nations, and 4 North clinical medicine, biomedical research, engi-
and South American nations during the neering and technology, and earth and space
1998–2005 period2. Based on the innovation sciences. Information concerning outputs and
production function proposed by Pakes and inputs is collected from the database of World
Griliches (1984) and Griliches (1990), this Development Indicators (WDI, World Bank)
paper consider R&D as production process and Main Scientific Technology Indicators
and treats individual nation as a DMU. As (MSTI, OECD). Table 1 displays the basic sta-
for the output and input variables, we include tistics for inputs and outputs.
three outputs containing patents, RLF, and sci-
entific journal articles and two inputs includ- Construction of the output-oriented
ing full-time researchers and technicians and R&D efficiency index
R&D expenditure. Due to unavailable data on After evaluating the efficiency of 24 nations
R&D capital stock in any statistical yearbook through the DEA approach, we can identify the
or database, this study thus converts R&D most efficient point on the frontier as the target
expenditures into R&D capital stocks using the for inefficient DMUs. For the i-th DMU, the dis-
perpetual inventory method3. tance from an inefficient point on the frontier by
While patent is the most widely adopted radially adjusting the level of outputs, φ−1, is called
measure of innovative output, simple counts of ‘radial adjustment’. Moreover, there exists a sec-
patents filed at an intellectual property office is ond stage shift from the projected point to a point

2
The 24 nations include Belgium, the Czech Republic, Finland, France, Germany, Hungary, Ireland, Italy, the Netherlands,
Poland, Portugal, Spain, Romania, Slovenia, the United Kingdom, Israel, Japan, South Korea, the Russian Federation,
Singapore, Argentina, Mexico, Canada, and the United States.
3
Referring to previous studies, we assume a depreciation rate of 15% for R&D capital (Hall & Mairesse 1995) and the
average rate of R&D growth for the initial growth rate (Guellec & van Pottelsberghe de la Potterie 2004).
4
For the pitfalls and advantages associated with equating patent counts with innovation, please see Furman et al. (2002) for
a survey.

344 INNOVATION: MANAGEMENT, POLICY & PRACTICE Volume 13, Issue 3, December 2011
Role of the national innovation system in R&D efficiency

TABLE 1: SUMMARY STATISTICS OF OUTPUT AND INPUT VARIABLES (1998–2005)


Name Mean SD Maximum Minimum

Output variables
Patentsa 16556.1 43277.0 240902.8 15.71
b
Scientific journal articles 21428.6 39326.1 205320 623
b
Royalty and licensing fees 346.2 925.1 5631.103 0.143
(million US dollars in year 2000)
Input variables
Total R&D manpowera 230297.2 344040.9 1415873 6805
(full-time equivalent units)
R&D expenditure stocksc 137276.3 309116.1 1600433 2080.073
(million US dollars in year 2000)
a
Main Scientific Technology Indicators, Paris: OECD.
b
World Bank: World Development Indicators database. cR&D expenditure flows are collected from Main Scientific
Technology Indicators, Paris: OECD, and then transformed into R&D expenditure stocks by using the perpetual inventory
method.

at the practical maximum level of output on the target output level given its inputs. Therefore,
frontier in the piecewise linear form of the non- three output-orientated R&D efficiency indi-
parametric frontier. The distance of the shift along ces are defined as follows: the patent-oriented
the frontier is called ‘slack’ (Coelli et al. 1998). The R&D efficiency index (PREI), the RLF-oriented
summation amount of slack and radial adjustment R&D efficiency index (RREI), and the Journal-
for outputs is called the amount of total adjust- oriented R&D efficiency index (JREI) for nation
ments, denoting the total amount of outputs that i at time t:
can be adjusted by a DMU to reach its optimal
production efficiency without using additional
Actual Patent Output (i , t )
inputs. Therefore, the total adjustments of output PREI(i , t ) = (2)
should increase while the input level is minimized Target Patent Output (i , t )
so that the DMU operates at the frontier position
Actual RLF Output (i , t )
of production efficiency. The practical maximum RREI(i , t ) = (3)
Target RLF Output (i , t )
output level is thus called the ‘target output level’
for a DMU (Coelli et al. 1998). Actual Journal
When incorporating three outputs into Article Output (i , t )
JREI(i , t ) = . (4)
the innovation production framework of each Target Journal
nation, the target level of the three outputs is cor- Article Output (i , t )
respondingly named the ‘target patent output’,
the ‘target journal articles output’, and the ‘tar- The above output-orientated R&D efficiency
get RLF output’. They represent a practical max- indices range between 0 and 1. When a nation’s
imum level to be used as a target in each nation innovation production reaches the target output
in order to perform at the optimal efficiency of based on the frontier, its R&D efficiency score
output production. As the amount of adjust- achieves 1. Conversely, if the actual innovative
ment increases, it implies a lower efficiency in output of a nation is far below the target output
a national R&D. Thus, if no output adjustment level, then the index approaches to 0, representing
occurs, then this nation is producing at the an extremely low R&D efficiency.

Volume 13, Issue 3, December 2011 INNOVATION: MANAGEMENT, POLICY & PRACTICE 345
Chiang-Ping Chen, Jin-Li Hu and Chih-Hai Yang

In order to evaluate the overall output-oriented nations usually lack high-quality R&D personnel,
R&D efficiency of each nation, this paper uses the advanced techniques of knowledge management,
average of three output-oriented R&D efficiency and other innovation complements, preventing
indices as a proxy index for overall national R&D them from having better R&D efficiency.
efficiency. The average output-oriented R&D Comparing the regional performance in R&D
efficiency index (AOREI) is defined as follows: efficiency, the bottom of Table 2 shows that
Asian nations seem to have a higher overall R&D
AOREI(i , t ) = Average (PREI, RREI, (5)
and JREI) efficiency than their European and American coun-
terparts. As the Asian countries contained in the
Compared with the efficiency index of a sin- dataset include Japan and emerging ‘technology
gle innovative output, the above index is more tigers’, such as Israel, South Korea, and Singapore,
precise, because it considers various innovative it leads to the result that Asian countries are more
outputs produced within a nation. Using the efficient than the other two regions. Furthermore,
calculated AOREI, the government can consider it is obvious that the average R&D efficiency of
how to improve its efficiency through technology OECD nations is significantly higher than that
policies. Furthermore, this index is flexible such of non-OECD nations. This result is reasonable,
that it can incorporate more innovative outputs. because OECD members are advanced countries
with better innovation management skills.
ESTIMATION OF OUTPUT-ORIENTED R&D To further look at individual R&D efficiency
EFFICIENCY INDICES of various innovative outputs, Tables 3–5 respec-
Table 2 displays the results of the average output-ori- tively display the scores of three output-oriented
ented R&D efficiency indices for each nation during R&D efficiency indices. These indices enable us to
the 1998–2005 period. The average overall R&D understand the strengths and weaknesses of each
efficiency is 0.6857, suggesting a 31.43% potential nation with respect to various R&D efficiencies.
for improvement, on average. In terms of the time By comparing the scores of the three specific
trend, it increased steadily from 0.6131 to 0.7354 in R&D efficiency indices, several interesting findings
2004 before decreasing to 0.6693 in 2005. emerge. First, the average score on the individual
Which nations experienced the best and worst output-oriented R&D efficiency index varies sub-
efficiency in R&D investments? Table 2 displays stantially. While the patent-oriented R&D effi-
that the United Kingdom, Israel, the United ciency is similar (usually a little higher) to that of
States, and, surprisingly, Hungary reached an overall R&D efficiency, ranging between 0.6131
AOREI value of 1 every year, suggesting that they and 0.7354 in various years, the efficiency scores on
had the best output-oriented R&D efficiency con- RLF and journals are much lower and higher than
sidering multiple innovative outputs, compared to the overall R&D efficiency, respectively. The RLF-
other sampled nations. Finland, Ireland, and the oriented R&D efficiency scores range from 0.5168
Netherlands experienced the best overall output- to 0.6413, whereas journal-oriented R&D efficiency
oriented R&D efficiency and ranked high scores scores span the interval between 0.7090 and 0.8589.
in specific years. Spain experienced the best overall These results suggest that the sample nations are
output-oriented R&D efficiency during the period most efficient at producing journal articles, relatively
1999–2002, while its efficiency declined sharply efficient on patenting, and less efficient in gaining
since 2003. On the other hand, nations that RLF given the fixed amount of R&D expenditure.
experienced, on average, a low overall R&D effi- This finding is consistent with the arguments in
ciency were Poland, Romania, Russia, and Mexico. Lee and Park (2005) and Wang and Huang (2007)
Specifically, Romania experienced the lowest whereby the computed mean of technical efficiency
R&D efficiency, 0.2387, on average. Developing is higher when SCI/EI publications are adopted as

346 INNOVATION: MANAGEMENT, POLICY & PRACTICE Volume 13, Issue 3, December 2011
Role of the national innovation system in R&D efficiency

TABLE 2: AVERAGE OUTPUT-ORIENTED R&D EFFICIENCY INDEX (AOREI)


Nation OECD Region 1998 1999 2000 2001 2002 2003 2004 2005

Belgium Yes Europe 0.7118 0.7118 0.6748 0.6741 0.7025 0.7095 0.8573 0.7321
Czech Yes Europe 0.4239 0.4434 0.3959 0.4186 0.4652 0.6186 0.4743 0.2953
Finland Yes Europe 0.6622 1.0000 0.9458 1.0000 1.0000 0.9506 0.9737 0.9219
France Yes Europe 0.5351 0.5245 0.5316 0.5370 0.5554 0.5515 0.5239 0.5472
Germany Yes Europe 0.6902 0.6879 0.6753 0.7291 0.7845 0.7224 0.8358 0.8338
Hungary Yes Europe 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Ireland Yes Europe 1.0000 1.0000 1.0000 0.9491 1.0000 0.8333 0.8925 1.0000
Italy Yes Europe 0.6087 0.6548 0.6372 0.6913 0.5796 1.0000 1.0000 0.5893
Netherlands Yes Europe 1.0000 0.9151 0.9515 0.9916 1.0000 1.0000 1.0000 1.0000
Poland Yes Europe 0.2046 0.2321 0.2793 0.3267 0.3896 0.3971 0.3737 0.3927
Portugal Yes Europe 0.3752 0.4067 0.3855 0.3756 0.4393 0.4124 0.4324 0.4366
Romania No Europe 0.1632 0.1925 0.2299 0.3009 0.3226 0.2328 0.2354 0.2324
Russia No Europe 0.3100 0.2598 0.2632 0.2595 0.3894 0.2043 0.1993 0.2187
Slovenia No Europe 0.3774 0.4846 0.6885 0.5552 0.6183 0.9999 1.0000 0.5440
Spain Yes Europe 0.5918 1.0000 1.0000 1.0000 1.0000 0.7177 0.6117 0.5109
UK Yes Europe 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Israel No Asia 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Japan Yes Asia 0.5970 0.6521 0.7296 0.8021 0.7846 0.7836 0.8312 0.7843
S. Korea Yes Asia 0.4282 0.4831 0.5256 0.6055 0.6525 0.7372 1.0000 1.0000
Singapore No Asia 0.5247 0.5843 0.5873 0.7042 0.7028 0.7324 0.9096 0.7123
Argentina No America 0.3208 0.3715 0.4660 0.5193 1.0000 0.4825 0.3737 0.3442
Canada Yes America 0.8342 0.8313 0.8627 0.8371 0.8702 0.8938 0.8911 0.7601
Mexico Yes America 0.3561 0.2552 0.2909 0.2879 0.2766 0.3380 0.2334 0.2065
USA Yes America 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Average Total 0.6131 0.6538 0.6717 0.6902 0.7305 0.7216 0.7354 0.6693
Europe 0.6034 0.6571 0.6662 0.6755 0.7029 0.7094 0.7131 0.6409
Asia 0.6375 0.6799 0.7106 0.7780 0.7850 0.8133 0.9352 0.8742
America 0.6278 0.6145 0.6549 0.6611 0.7867 0.6786 0.6245 0.5777
OECD 0.6677 0.7110 0.7159 0.7348 0.7500 0.7592 0.7739 0.7228
Non-OECD 0.4494 0.4821 0.5392 0.5565 0.6722 0.6086 0.6197 0.5086

the output rather than patents. However, one point researchers are apt to exhaust their energies publish-
worth emphasizing is that journal-oriented R&D ing journal papers even if they do not obtain suf-
efficiency may suffer overestimate bias, because ficient research funding5.
5
Indeed, the publication of journal articles helps researchers to establish their academic reputation and, more importantly,
is the critical criterion for researchers’ career promotion in most nations.

Volume 13, Issue 3, December 2011 INNOVATION: MANAGEMENT, POLICY & PRACTICE 347
Chiang-Ping Chen, Jin-Li Hu and Chih-Hai Yang

TABLE 3: PATENT-ORIENTED R&D EFFICIENCY INDEX (PREI)


Nation OECD Region 1998 1999 2000 2001 2002 2003 2004 2005

Belgium Yes Europe 0.7613 0.7628 0.7110 0.7140 0.7122 0.7862 0.8573 0.7625
Czech Yes Europe 0.1115 0.3133 0.2369 0.3044 0.3182 0.3863 0.3092 0.2337
Finland Yes Europe 0.8737 1.0000 0.9973 1.0000 1.0000 0.9506 0.9737 0.9597
France Yes Europe 0.6307 0.6622 0.6437 0.6417 0.5926 0.6045 0.5892 0.5709
Germany Yes Europe 0.9309 0.9373 0.9356 0.9603 0.9427 0.8863 0.8691 0.8338
Hungary Yes Europe 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Ireland Yes Europe 1.0000 1.0000 1.0000 0.9491 1.0000 0.8333 0.8925 1.0000
Italy Yes Europe 0.8488 0.9112 0.8877 0.9824 0.6286 1.0000 1.0000 0.6738
Netherlands Yes Europe 1.0000 0.9022 0.9515 0.9916 1.0000 1.0000 1.0000 1.0000
Poland Yes Europe 0.0385 0.0707 0.0941 0.1360 0.2424 0.2347 0.2318 0.2662
Portugal Yes Europe 0.0678 0.2653 0.2025 0.1946 0.2357 0.2577 0.2081 0.2814
Romania No Europe 0.0982 0.0679 0.1056 0.1714 0.2651 0.1645 0.1711 0.1915
Russia No Europe 0.2735 0.1788 0.1957 0.2493 0.3812 0.1571 0.1361 0.1915
Slovenia No Europe 0.2800 0.5342 0.8208 0.5449 0.8228 1.0000 1.0000 0.6475
Spain Yes Europe 0.2592 1.0000 1.0000 1.0000 1.0000 0.7393 0.6950 0.5530
UK Yes Europe 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Israel No Asia 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Japan Yes Asia 0.8431 0.8644 0.9446 0.9933 0.9361 0.9304 0.9410 0.9065
S. Korea Yes Asia 0.7270 0.6172 0.6374 0.6730 0.7977 0.9603 1.0000 1.0000
Singapore No Asia 0.7089 0.8030 0.7899 0.8670 0.8433 0.8677 0.9136 0.6767
Argentina No America 0.2018 0.3272 0.4773 0.4911 1.0000 0.4804 0.3011 0.3022
Canada Yes America 0.9295 0.9820 0.9425 0.8868 0.9037 0.8938 0.8911 0.9170
Mexico Yes America 0.1264 0.1770 0.2445 0.2404 0.2227 0.2980 0.1966 0.2085
USA Yes America 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Average Total 0.6130 0.6823 0.7008 0.7080 0.7435 0.7263 0.7157 0.6740
Europe 0.5734 0.6629 0.6739 0.6775 0.6963 0.6875 0.6833 0.6353
Asia 0.8198 0.8212 0.8430 0.8833 0.8943 0.9396 0.9637 0.8958
America 0.5644 0.6216 0.6661 0.6546 0.7816 0.6681 0.5972 0.6069
OECD 0.6749 0.7481 0.7461 0.7593 0.7518 0.7645 0.7586 0.7315
Non-OECD 0.4271 0.4852 0.5649 0.5540 0.7187 0.6116 0.5870 0.5016

Second, as depicted in Figure 1, the dynam- in 2004 and then decreased sharply to 0.5168 in
ics of various output-oriented R&D efficiency 2005. In contrast, the mean score on the JREI
indices exhibit similar trends except for the remained relatively stable during the post-2001
score on RLF. The mean efficiency score on RLF period. Why did these sample nations experience
increased stably from 0.5174 in 1998 to 0.6413 a sharp decrease in RLF-oriented R&D efficiency

348 INNOVATION: MANAGEMENT, POLICY & PRACTICE Volume 13, Issue 3, December 2011
Role of the national innovation system in R&D efficiency

TABLE 4: RLF-ORIENTED R&D EFFICIENCY INDEX (RREI)


Nation OECD Region 1998 1999 2000 2001 2002 2003 2004 2005

Belgium Yes Europe 0.6127 0.6099 0.6023 0.5943 0.5829 0.5231 0.8573 0.6095
Czech Yes Europe 0.4892 0.3329 0.1818 0.1756 0.2233 0.7348 0.2286 0.0280
Finland Yes Europe 0.4315 1.0000 0.9973 1.0000 1.0000 0.9506 0.9737 0.9597
France Yes Europe 0.3439 0.2491 0.3075 0.3276 0.4515 0.4454 0.3934 0.4998
Germany Yes Europe 0.5582 0.4694 0.3995 0.5505 0.4679 0.6382 0.7693 0.8338
Hungary Yes Europe 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Ireland Yes Europe 1.0000 1.0000 1.0000 0.9491 1.0000 0.8333 0.8925 1.0000
Italy Yes Europe 0.1287 0.1419 0.1362 0.1090 0.1500 1.0000 1.0000 0.1533
Netherlands Yes Europe 1.0000 0.9216 0.9515 0.9916 1.0000 1.0000 1.0000 1.0000
Poland Yes Europe 0.0892 0.0972 0.1005 0.1627 0.0629 0.0358 0.0204 0.0291
Portugal Yes Europe 0.5167 0.3585 0.2677 0.2041 0.2498 0.1643 0.1118 0.0697
Romania No Europe 0.0887 0.0842 0.0357 0.1861 0.1061 0.0083 0.0107 0.0384
Russia No Europe 0.0831 0.0583 0.0599 0.0437 0.3152 0.0308 0.0210 0.0152
Slovenia No Europe 0.2775 0.2902 0.3840 0.3232 0.2092 0.9998 1.0000 0.0470
Spain Yes Europe 0.5268 1.0000 1.0000 1.0000 1.0000 0.5601 0.2308 0.0919
UK Yes Europe 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Israel No Asia 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Japan Yes Asia 0.5846 0.6504 0.7545 0.9350 0.6599 0.9304 0.9410 0.9065
S. Korea Yes Asia 0.2161 0.3714 0.3951 0.6231 0.3622 0.7309 1.0000 1.0000
Singapore No Asia 0.1562 0.1467 0.1821 0.3785 0.4084 0.4619 0.9015 0.6061
Argentina No America 0.2005 0.2119 0.2355 0.3318 1.0000 0.1346 0.0804 0.0421
Canada Yes America 0.6437 0.5298 0.7031 0.7377 0.8032 0.8938 0.8911 0.4461
Mexico Yes America 0.4710 0.1026 0.0967 0.0571 0.0875 0.2333 0.0687 0.0271
USA Yes America 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Average Total 0.5174 0.5261 0.5330 0.5700 0.5892 0.6379 0.6413 0.5168
Europe 0.5091 0.5383 0.5265 0.5386 0.5512 0.6203 0.5943 0.4610
Asia 0.4892 0.5421 0.5829 0.7342 0.6076 0.7808 0.9606 0.8782
America 0.5788 0.4611 0.5088 0.5317 0.7227 0.5654 0.5101 0.3788
OECD 0.5896 0.6019 0.6052 0.6343 0.6167 0.7041 0.6877 0.5919
Non-OECD 0.3010 0.2986 0.3162 0.3772 0.5065 0.4392 0.5023 0.2915

scores? A closer look at the performance of indi- output. However, the efficiency scores of some
vidual nations can help determine potential nations vary significantly across these indices. As
causes. shown in Table 6, Japan’s mean patent-oriented
Hungary, the UK, Israel, and the US remained and RLF-oriented R&D efficiency scores are
100% efficient in all measures of innovative higher, but journal-oriented R&D efficiency

Volume 13, Issue 3, December 2011 INNOVATION: MANAGEMENT, POLICY & PRACTICE 349
Chiang-Ping Chen, Jin-Li Hu and Chih-Hai Yang

TABLE 5: JOURNAL-ORIENTED R&D EFFICIENCY INDEX (JREI)


Nation OECD Region 1998 1999 2000 2001 2002 2003 2004 2005

Belgium Yes Europe 0.7613 0.7628 0.7110 0.7140 0.8124 0.8193 0.8573 0.8243
Czech Yes Europe 0.6709 0.6838 0.7689 0.7757 0.8542 0.7349 0.8851 0.6242
Finland Yes Europe 0.6814 1.0000 0.8428 1.0000 1.0000 0.9506 0.9737 0.8462
France Yes Europe 0.6307 0.6622 0.6437 0.6417 0.6222 0.6045 0.5892 0.5709
Germany Yes Europe 0.5814 0.6570 0.6908 0.6764 0.9427 0.6426 0.8691 0.8338
Hungary Yes Europe 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Ireland Yes Europe 1.0000 1.0000 1.0000 0.9491 1.0000 0.8333 0.8925 1.0000
Italy Yes Europe 0.8488 0.9112 0.8877 0.9824 0.9601 1.0000 1.0000 0.9407
Netherlands Yes Europe 1.0000 0.9216 0.9515 0.9916 1.0000 1.0000 1.0000 1.0000
Poland Yes Europe 0.4861 0.5283 0.6432 0.6815 0.8636 0.9208 0.8690 0.8829
Portugal Yes Europe 0.5411 0.5963 0.6863 0.7280 0.8323 0.8152 0.9772 0.9587
Romania No Europe 0.3028 0.4255 0.5484 0.5452 0.5964 0.5256 0.5244 0.4674
Russia No Europe 0.5735 0.5424 0.5340 0.4856 0.4719 0.4249 0.4407 0.4495
Slovenia No Europe 0.5748 0.6294 0.8606 0.7976 0.8228 1.0000 1.0000 0.9374
Spain Yes Europe 0.9896 1.0000 1.0000 1.0000 1.0000 0.8536 0.9093 0.8878
UK Yes Europe 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Israel No Asia 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Japan Yes Asia 0.3635 0.4415 0.4897 0.4781 0.7578 0.4899 0.6117 0.5400
S. Korea Yes Asia 0.3416 0.4605 0.5442 0.5205 0.7977 0.5203 1.0000 1.0000
Singapore No Asia 0.7089 0.8030 0.7899 0.8670 0.8566 0.8677 0.9136 0.8541
Argentina No America 0.5602 0.5754 0.6852 0.7351 1.0000 0.8324 0.7396 0.6883
Canada Yes America 0.9295 0.9820 0.9425 0.8868 0.9037 0.8938 0.8911 0.9170
Mexico Yes America 0.4710 0.4861 0.5315 0.5662 0.5195 0.4826 0.4348 0.3840
USA Yes America 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Average Total 0.7090 0.7529 0.7813 0.7926 0.8589 0.8005 0.8491 0.8170
Europe 0.7277 0.7700 0.7981 0.8106 0.8612 0.8203 0.8617 0.8265
Asia 0.6035 0.6763 0.7060 0.7164 0.8530 0.7195 0.8813 0.8485
America 0.7402 0.7609 0.7898 0.7970 0.8558 0.8022 0.7664 0.7473
OECD 0.7387 0.7830 0.7963 0.8107 0.8815 0.8090 0.8756 0.8450
Non-OECD 0.6200 0.6626 0.7364 0.7384 0.7913 0.7751 0.7697 0.7328

score is much lower than its overall R&D effi- licensing rather than publishing journal articles6.
ciency, suggesting that R&D activities in Japan In contrast, developing nations, such as Argentina,
focus more on patent applications and technology Romania, and Russia, and developed countries,
6
Traditionally, Japanese scholars are apt to write articles in Japanese and publish papers domestically rather than submitting
papers to SCI journals. This may explain why the Japanese mean score on the JREI is lower than on the PREI and RREI.

350 INNOVATION: MANAGEMENT, POLICY & PRACTICE Volume 13, Issue 3, December 2011
Role of the national innovation system in R&D efficiency

In sum, national R&D


activities are composed of
various target-oriented R&D
programs in various sectors
(industry, higher education
institutions, and government),
implying that nations may
target different innovative
outputs. Similar to previous
studies, this paper compares
R&D efficiency across nations
by constructing an overall
index of several innovative
FIGURE 1: TIME TREND OF VARIOUS OUTPUT-ORIENTED R&D EFFICIENCY INDICES outputs. Importantly, our
unique approach also enables
such as the Czech Republic, Poland, Portugal, and us to simultaneously compare
Mexico, have a relatively strong performance on various output-oriented R&D efficiency indi-
academic publications and experience extremely ces across nations and finds that one nation may
poor R&D efficiency on the RREI. Their mean experience significant differences in the relative
efficiency scores on RLF are lower than 0.3, scores and rankings across various output-ori-
causing the RREI to have the lowest mean score ented R&D efficiency indices.
among the various output-oriented R&D effi-
ciency indices. Moreover, Finland, Ireland, and THE ROLE OF THE NIS ON R&D
the Netherlands experience high and similar effi- EFFICIENCY
ciency scores for all of the output-oriented R&D Empirical specification of
efficiency indices. the output-oriented R&D
In terms of rankings on the various R&D effi- efficiency index
ciency indices, most nations seem to have similar Estimates shown in the previous section clearly
rankings in each index, suggesting that although indicate that nations may experience significant
some nations exhibit considerable differences in differences in R&D efficiency for various out-
the magnitude of their mean scores across vari- put-oriented indices. This section examines how
ous indices, their relative efficiency performances environmental factors, especially the NIS, affect
tend to remain consistent throughout the sample the various output-oriented R&D efficiency indi-
period. Different from the majority of sample ces. First, we specify the empirical model of the
nations, few have a distinct R&D focus, inducing AOREI as follows:
divergent performances on the various output-
AOREIit = β0 + β1 IPR protectionit (6)
oriented R&D efficiency indices. For instance,
+ β2 Per-capita GDPit
Italy ranks 13th out of the 24 nations on the over-
+ β3 Human capitalit
all output-oriented R&D efficiency, but its rela-
+ β4 ICT infrastructureit
tive performance on RLF and journals results in
+ β5 Total RD intensityit + εit
rankings of 17th and 8th, respectively. The case
of Spain echoes this tendency as well. In contrast, The dependent variable AOREIit denotes the score
while Japan ranks high in R&D efficiency scores on the average output-oriented R&D efficiency
on patents and RLF (8th for both), its ranking on index of nation i in year t obtained in Table 2. As
journals is a low 21st among the 24 sample nations. for the explanatory variables, we include mainly the

Volume 13, Issue 3, December 2011 INNOVATION: MANAGEMENT, POLICY & PRACTICE 351
Chiang-Ping Chen, Jin-Li Hu and Chih-Hai Yang

TABLE 6: MEAN SCORES AND RANKING OF VARIOUS R&D EFFICIENCY INDICES, 1998–2005
Nation AOREI Ranking PREI Ranking RREI Ranking JREI Ranking

Belgium 0.7217 12 0.7584 15 0.6240 11 0.7828 13


Czech 0.4419 19 0.2767 19 0.2993 18 0.7497 15
Finland 0.9318 7 0.9694 6 0.9141 7 0.9118 10
France 0.5383 17 0.6169 17 0.3773 16 0.6206 20
Germany 0.7449 11 0.9120 10 0.5859 13 0.7367 16
Hungary 1.0000 1 1.0000 1 1.0000 1 1.0000 1
Ireland 0.9594 6 0.9594 7 0.9594 6 0.9594 6
Italy 0.7201 13 0.8666 11 0.3524 17 0.9414 8
Netherlands 0.9822 5 0.9807 5 0.9831 5 0.9831 5
Poland 0.3245 21 0.1643 23 0.0747 23 0.7344 17
Portugal 0.4080 20 0.2141 22 0.2428 20 0.7669 14
Romania 0.2387 24 0.1544 24 0.0697 24 0.4920 22
Russia 0.2630 23 0.2204 20 0.0784 22 0.4903 23
Slovenia 0.6585 16 0.7063 16 0.4414 14 0.8278 12
Spain 0.8040 9 0.7808 14 0.6762 10 0.9550 7
UK 1.0000 1 1.0000 1 1.0000 1 1.0000 1
Israel 1.0000 1 1.0000 1 1.0000 1 1.0000 1
Japan 0.7456 10 0.9199 8 0.7953 8 0.5215 21
S. Korea 0.6790 15 0.8016 13 0.5874 12 0.6481 19
Singapore 0.6822 14 0.8088 12 0.4052 15 0.8326 11
Argentina 0.4848 18 0.4476 18 0.2796 19 0.7270 18
Canada 0.8476 8 0.9183 9 0.7061 9 0.9183 9
Mexico 0.2806 22 0.2143 21 0.1430 21 0.4845 24
USA 1.0000 1 1.0000 1 1.0000 1 1.0000 1

innovation environment and the composition of adequately enforced or are adequately enforced’
R&D activity among the private sector, academic and the respondents give a score from 0 to 10.
institutes, and the public sector within a nation. A higher value on the index indicates a stronger
Intellectual property rights protection is the level of protection. The IMD IPR index has been
strength of IPR protection, representing a nation’s widely adopted recently and has shown to be
legal environment. How do we measure national highly correlated with the index of Ginarte and
differences in IPR protection? We adopt the IPR Park (1997) (e.g., Nunnenkamp & Spatz 2003).
index surveyed by the International Institute for An environment with strong IPR protection
Management Development (IMD). The IMD is favorable for firms to devote efforts to R&D
survey is related to IPR protection in general. activity and reduce the risk of imitation.
The question included in the IMD survey asks Per-capita GDP is real per-capita GDP that
whether ‘Intellectual property rights are not takes a natural logarithm and is adopted to

352 INNOVATION: MANAGEMENT, POLICY & PRACTICE Volume 13, Issue 3, December 2011
Role of the national innovation system in R&D efficiency

capture the concept of knowledge stock (Furman + β4 ICT infrastructureit


et al. 2002; Hu & Mathews 2008). Accumulating + β5 Business RD intensityit
more knowledge stock represents superior innova- + β6 Public RD intensityit + εit
tion ability and is critical for future innovations,
The empirical specifications regarding the
contributing positively to R&D efficiency.
determinants of PREI, RREI, and JREI are cor-
Human capital denotes the investment of human
respondingly specified as follows:
capital, which is measured by the percentage of
total education expenditure on GDP. A higher INDEXit = β0 + β1 IPR protectionit (8)
ratio of education investment intensity can + β2 Per-capita GDPit
strengthen the accumulation and the quality of + β3 Human capitalit
human capital and, in turn, enhance R&D effi- + β4 ICT infrastructureit
ciency (Furman et al. 2002). Within the NIS, the + β5 Business RD intensityit
infrastructure of information and communica- + β6 Public RD intensityit + εit
tion technologies (ICT) plays an important role
where INDEX represents one of the above three
in spurring R&D efficiency. This study takes the
output-oriented R&D efficiency indices.
number of personal computers per 100 people in
One critical concept of the NIS is the interac-
a nation as a proxy to determine the relationship
tion between various sectors. To further investigate
between ICT infrastructure and R&D efficiency.
how different funding balances between the pri-
Finally, Total RD intensity represents a national
vate sector and public sector affect the R&D effi-
R&D intensity, measured by the ratio of total
ciency index, we divide the variable Business RD
R&D expenditure to GDP. The NIS approach
intensity into the business R&D intensity fund-
focuses on a textured description of the organi-
ing by the business itself, the public sector, and
zation and patterns of activity that contribute
foreign sources9. Furthermore, we construct two
to innovative behavior in a nation7. The NIS
variables that respectively denote the proportions
approach emphasizes the divergent role played
of the Public RD intensity performed by higher
by different sectors and institutions on innova-
education institutions and government. Finally, to
tion within specific nations. For instance, the
examine the R&D interaction between the private
key component and linkages in the NISs of the
sector and higher education institutions, we con-
United States and Germany are quite different8.
struct the variable that is measured by the product
To differentiate the relative importance of
of the business funding share of higher education
R&D contributed by various sectors within one
and higher education R&D intensity.
NIS, this structure further divides Total RD inten-
The publication of journal articles is not the
sity into R&D intensity contributed by the pri-
main output target of the private sector’s R&D
vate sector (Business RD intensity) and that by the
activity, whereas it is the main focus of high educa-
public sector (Public RD intensity). Equation (6)
tion institutes’ R&D. Therefore, in the specification
can be rewritten as follows:
of the JREI in equation (8), we divide the public
AOREIit = β0 + β1 IPR protectionit (7) sector’s R&D activity into the government R&D
+ β2 Per-capita GDPit expenditure to GDP ratio and the higher education
+ β3 Human capitalit institution R&D expenditure to GDP ratio.

7
See Nelson (1993) for a comprehensive review of the national innovation system.
8
Please see Figure 2 in Furman et al. (2002). Lichtenberg (1993) indicated also that the R&D activity sourced from the
private and public sectors has different impacts on productivity in the innovation process.
9
Foreign funding sources include foreign governments, foreign business enterprises, the European Commission, and other
international organizations.

Volume 13, Issue 3, December 2011 INNOVATION: MANAGEMENT, POLICY & PRACTICE 353
Chiang-Ping Chen, Jin-Li Hu and Chih-Hai Yang

TABLE 7: DEFINITIONS AND DESCRIPTIONS OF ALL VARIABLES IN THE TOBIT REGRESSION


Variables Definition Mean SD

IPR protection The degree of intellectual property rights protection in a 6.3740 1.7089
nation is collected from World Competitiveness Yearbook
published by IMD. (0–10)
Per-capita GDP The knowledge stock is measured by per-capita gross 4.2814 0.1996
domestic product, collected from World Development
Indicators and then a natural logarithm is taken.
Human capital Human capital investment is measured by the percentage 5.2785 1.4123
of total education expenditure on GDP, collected from
World Development Indicators. (%)
ICT infrastructure Infrastructure of ICT is measured by the number of 30.1958 20.2760
personnel computers per 100 people, collected from World
Development Indicators.
Total RD intensity The overall R&D intensity of a nation is measured by the 0.0263 0.0478
ratio of total R&D expenditure to GDP.
Business RD intensity Business R&D intensity is measured by the Total RD 0.0159 0.0258
intensity performed by the business sector.
Share of private sector The share of private sector funding is measured by the 0.0135 0.0219
funding × Business RD ratio of the business R&D expenditure funded by the
intensity business itself to total business R&D expenditure.
Share of public sector The share of public sector funding is measured by the ratio 0.0014 0.0029
funding × business R&D of the business R&D expenditure funded by the public
intensity sector to total business R&D expenditure.
Share of foreign source The share of foreign source funding is measured by the 0.0009 0.0015
funding × business RD ratio of the business R&D expenditure funded by foreign
intensity sources to total business R&D expenditure.
Public RD intensity Public R&D intensity is measured by the Total RD intensity 0.0104 0.0222
performed by the public sector.
Higher education RD This variable is measured by the Public RD intensity 0.0066 0.0147
intensity performed by the higher education institution.
Business funding share The share of business funding is measured by the ratio 0.0007 0.0017
of higher education RD of the higher education R&D expenditure funded by the
intensity business to total higher education R&D expenditure.
Government RD This variable is measured by the public R&D intensity 0.0038 0.0076
intensity performed by the government.
All detailed information about R&D intensity is collected from Main Scientific Technology Indicators, Paris: OECD.

Table 7 shows the definition and description of Empirical results and discussions
all variables. All nominal variables are transformed Table 8 displays a series of estimates on the deter-
into constant 2000 US dollars by GDP defla- minants of the AOREI, showing that estimates
tors. As mentioned above, the output-oriented are quite similar in various specifications. The
R&D efficiency indices range from 0 to 1, imply- coefficient estimates of IPR protection are signifi-
ing that the data distribution is double censored. cantly positive, suggesting that a nation with stron-
Therefore, we use the Tobit regression to estimate ger IPR protection provides a favorable innovation
the effects of environmental factors on efficiency environment for R&D activity and induces a higher
scores. R&D efficiency compared with other nations,

354 INNOVATION: MANAGEMENT, POLICY & PRACTICE Volume 13, Issue 3, December 2011
Role of the national innovation system in R&D efficiency

TABLE 8: ESTIMATES OF DETERMINANTS OF AVERAGE OUTPUT-ORIENTED R&D EFFICIENCY INDEX


(1) (2) (3) (4)

Constant −1.6837a (0.6353) −1.7044a (0.6322) −1.4155b (0.6348) −1.5836a (0.5290)


IPR protection 0.0381b (0.017) 0.0330c (0.0179) 0.0214 (0.0185) 0.0342c (0.0182)
b b b
Per-capita GDP 0.4068 (0.1714) 0.4145 (0.1706) 0.3605 (0.1694) 0.4000a (0.1462)
Human capital 0.0553a (0.0142) 0.0589a (0.0145) 0.0592a (0.0141) 0.0511a (0.0150)
ICT infrastructure 0.0035b (0.0013) 0.0027c (0.0014) 0.0023 (0.0014) 0.0026c (0.0014)
Total RD intensity 1.1509a (0.3988)
Business RD intensity 7.2922c (3.0989) 5.1079 (3.8640)
a
Share of private sector 12.7401 (4.6444)
funding × business RD
intensity
Share of public sector −25.4307 (16.6612)
funding × business RD
intensity
Share of foreign source 45.5584 (27.7896)
funding × business RD
intensity
Public RD intensity −6.0380 (4.4734) −9.3519c (5.2265) −49.5137a (18.4882)
Higher education RD 55.3183b (25.8728)
intensity
Business funding share 126.0489b (49.0608)
of higher education RD
intensity
Adj. R-square 0.6184 0.6306 0.6510 0.6688
Figures in parentheses are standard error. Notationsa, b, and c represent significance at the 1, 5, and 10% levels,
respectively.

ceteris paribus. The coefficients of Per-capita GDP in column 2 seem to indicate that the R&D
and Human capital are significantly positive at a efficiency-enhancing effect is brought on mainly
conventional statistical level, indicating that the by R&D performed by the private sector rather
more knowledge stock and the higher the invest- the than by the public sector, because the esti-
ment of human resources a nation has, the higher mated coefficients of Business RD intensity are
its R&D efficiency is, on average. Furthermore, the significantly positive whereas the estimated coef-
technological infrastructure also has a positive and ficient of Public RD intensity is not significantly
significant parameter in most estimates. positive. Public sector R&D usually is more
Focusing on the impacts of R&D performed basic-research-oriented and aims to develop new
by various sectors, column 1 shows that the technologies to advance industrial technologi-
estimated coefficient of Total RD intensity is cal capability rather than applying for patents,
positive and significant at the 1% statistical acquiring licensing fees, and publishing journal
level. This result suggests that a nation devot- articles. Thus, given the fixed expenditure on
ing more efforts to R&D activity in terms of R&D activity, a nation with a higher proportion
R&D intensity reaches a better performance of R&D expenditure by the private sector will
on overall R&D efficiency. However, estimates reach a higher R&D efficiency.

Volume 13, Issue 3, December 2011 INNOVATION: MANAGEMENT, POLICY & PRACTICE 355
Chiang-Ping Chen, Jin-Li Hu and Chih-Hai Yang

How does the interaction between the private particularly relevant to the development of new
and public sectors affect R&D efficiency? Results in technologies and applications for patents. On the
column 3 show that among various funding sources other hand, firms generally license mature rather
of private R&D, only a higher proportion of R&D than newly developed technologies to other firms,
contributed by private enterprises themselves posi- causing knowledge capital to have relatively slight
tively contributes to R&D efficiency. In contrast, impacts on RLF. Second, a national R&D intensity
as shown in columns 3 and 4, R&D intensity that moving toward a higher proportion of private R&D
includes a higher proportion of public R&D nega- can improve patent-oriented and RLF-oriented
tively impacts overall R&D efficiency. However, R&D efficiency indices, ceteris paribus. This result
when public R&D is composed of a higher propor- is economically intuitive since patents and RLF are
tion of academic R&D, it exhibits a positive impact the primary targeted outputs of the private sector’s
and improves R&D efficiency. This result may arise R&D. Third, the main efficiency-improving effect
from the fact that academic R&D usually pro- on the patent-oriented and RLF-oriented R&D
duces many journal articles, which in turn induce a efficiency indices seems to be attributed to R&D
higher average score on the output-oriented R&D activity produced by private businesses themselves
efficiency index (AOREI). Importantly, the coeffi- and the interaction between business and foreign
cient of the business funding share of higher educa- funding sources. Fourth, a higher proportion of
tion R&D intensity is significantly positive at the public R&D lowers efficiency in patents and RLF,
1% statistical level, suggesting that the interaction because public R&D expenditure is usually not
between private and academic R&D can positively aimed at patent creation and royalties and licens-
influence overall R&D efficiency. ing fee collection, but more toward the provision
Estimates in the previous section indicate that of public goods. Conversely, a higher proportion of
some nations perform quite differently on the higher education R&D expenditure helps improve
assorted output-oriented R&D efficiency indices, the efficiency of patents and RLF as part of their
because their national R&D compositions (pri- performance. Similarly, an increase in the R&D
vate, government, and higher education institu- expenditure donated by private enterprise to higher
tions) vary considerably and have specific-targeted education institutions also helps promote the effi-
innovative outputs. This result implies that the ciency in patents and RLF. As a result, when private
relative share of R&D expenditure contributed enterprises and higher education institutions con-
by the various R&D performers and their interac- stitute a major part of national R&D efforts, R&D
tions may play distinct roles in influencing scores efficiency will improve, because both of these sec-
on the different output-oriented R&D efficiency tors include patents as well as RLF as major por-
indices. To examine how the NIS differenti- tions of their R&D outputs.
ates among individual R&D efficiency indices, As for the determinants of the JREI, results in
Table 9 presents estimates on the determinants of Table 10 show that the estimates exhibit different
the PREI and the RREI. Moreover, Table 10 dis- and interesting findings compared with estimates
plays the estimates of determinants of the JREI. in Table 9.
While the estimates of determinants for the The coefficient of IPR protection is not sig-
patent-oriented and RLF-oriented R&D efficiency nificant from zero in all estimates, suggesting that
indices are quite similar, some points of compari- the legal environment of innovation in terms of
son are worth noting. First, the proxy of knowledge IPR protection is not particularly relevant to
stock (Per-capita GDP) exhibits a positive sign on journal-oriented R&D efficiency performance.
influencing the patent-oriented and RLF-oriented Based solely on this finding, we cannot claim
R&D efficiency indices, but its impact is significant that IPR protection is not important, because we
only on patents. Accumulated knowledge capital is attribute this result to the distinct technological

356 INNOVATION: MANAGEMENT, POLICY & PRACTICE Volume 13, Issue 3, December 2011
Role of the national innovation system in R&D efficiency

TABLE 9: DETERMINANTS OF PATENT-ORIENTED AND RLF-ORIENTED R&D EFFICIENCY INDICES


Dependent variable Patent-oriented R&D efficiency index RLF-oriented R&D efficiency index

Independent (1) (2) (3) (4) (5) (6) (7) (8)


variables

Constant −2.7977a −2.8371a −2.5783a −2.8910a −1.4609c −1.5272c −0.9069 −1.3727c


(0.7293) (0.7116) (0.7174) (0.5887) (0.8258) (0.8137) (0.8003) (0.7589)
IPR protection 0.0449b 0.0348c 0.0228 0.0364c 0.0641a 0.0555b 0.0304 0.0582b
(0.0205) (0.0202) (0.0209) (0.0194) (0.0231) (0.0230) (0.0232) (0.0253)
Per-capita GDP 0.6897a 0.7045a 0.6586a 0.7283a 0.2794 0.3042 0.1852 0.2866
(0.1968) (0.1921) (0.1914) (0.1648) (0.2223) (0.2192) (0.2133) (0.168)
Human capital 0.0297c 0.0372b 0.0370b 0.0311c 0.0535a 0.0562a 0.0595a 0.0454b
(0.0163) (0.0165) (0.0162) (0.0167) (0.0178) (0.0176) (0.0169) (0.0185)
ICT infrastructure 0.0041b 0.0025 0.0019 0.0024c 0.0044b 0.0030c 0.0021 0.0029c
(0.0016) (0.0016) (0.0016) (0.0014) (0.0017) (0.0018) (0.0017) (0.0017)
Total RD intensity 1.0635b 1.0910b
(0.4575) (0.5106)
Business RD intensity 13.8520a 10.0608b 11.6007b 8.2133c
(4.4688) (4.1964) (4.6943) (4.3017)
Share of private 19.3532a 23.8083a
sector funding × (5.2995) (5.6011)
business RD intensity
Share of public −14.2289 −59.9024a
sector funding × (18.8578) (20.9333)
business RD intensity
Share of foreign 63.7980b 94.8598a
source funding × (31.8758) (34.2663)
business RD intensity
Public RD intensity −13.7646a −18.3961a −55.4598a −11.1468b −18.7328a −74.0491a
(5.1066) (5.9561) (19.3105) (5.4002) (6.3482) (21.2480)
Higher education RD 48.1371c 79.4647b
intensity (27.3537) (31.0469)
Business funding 190.3844a 189.1861a
share of higher (52. 9471) (60.4375)
education RD
intensity
Adj. R-square 0.5639 0.5966 0.6100 0.6388 0.4768 0.4986 0.5561 0.5458
a, b c
Figures in parentheses are standard error. Notations , and represent significance at the 1, 5, and 10% levels,
respectively.

environment for publishing journal articles. While ethics, because this type of behavior could destroy
the infringement of patents and copyrights is prev- a scholar’s academic career. Therefore, the legal
alent in the world, especially in some developing environment of IPR protection has a weaker influ-
nations, cases of plagiarism are rare in academic ence on journal-oriented R&D efficiency.
publications. In practice, the standard of aca- While overall R&D intensity is associated with
demic ethics is rigorous and extremely critical for a positive and significant coefficient as shown
scholars. Few scholars attempt to violate academic in Table 9, the impacts of various components

Volume 13, Issue 3, December 2011 INNOVATION: MANAGEMENT, POLICY & PRACTICE 357
Chiang-Ping Chen, Jin-Li Hu and Chih-Hai Yang

TABLE 10: THE DETERMINANTS OF JOURNAL-ORIENTED R&D EFFICIENCY INDEX (JREI)


(1) (2) (3)

Constant −0.9916c (0.5311) −0.9546c (0.5095) −0.8223 (0.5296)


IPR protection −0.0020 (0.0149) 0.0048 (0.0146) 0.0049 (0.0145)
Per-capita GDP 0.3184b (0.1434) 0.3071b (0.1375) 0.2820b (0.1401)
a a
Human capital 0.0749 (0.0121) 0.0685 (0.0685) 0.0649a (0.0121)
ICT infrastructure 0.0016 (0.0011) 0.0027b (0.0011) 0.0027b (0.0011)
Total RD intensity 1.2131a (0.3387)
Business RD intensity −7.6468b (3.0866) −7.4176b (3.1017)
Public RD intensity 11.2545a (3.5509)
Government RD intensity −2.1936 (15.4566)
Higher education RD intensity 17.8610b (8.1873)
Adj. R-square 0.8042 0.8676 0.8739
Figures in parentheses are standard error. Notations a, b, and c represent significance at the 1, 5, and 10% levels,
respectively.

appear to have entirely different influences on JREI Republic, Poland, Romania, and Russia, have a rel-
compared with their influences on the other two atively strong performance on the JREI and poor
indices. As shown in column 2 of Table 10, the R&D efficiency on the PREI and RREI. These
coefficients of Business and Public RD intensity are nations have a long tradition of well-respected
significantly negative and positive, respectively. This higher education systems relative to their industrial
result indicates that, given a fixed R&D intensity, a sectors. Their governments ladle a lot of funding
nation with a lower proportion of private R&D and into universities every year, inducing a relatively
a higher proportion of public R&D will achieve a higher R&D efficiency score on journal articles
higher score on the JREI, on average. Private R&D compared with the other two innovative outputs.
activity purely aims at appropriating the monopoly
advantages brought on by innovations and there- CONCLUDING REMARKS AND POLICY
fore focuses primarily on patents and RLF that IMPLICATIONS
have pecuniary benefits. On the other hand, pub- The importance of innovation on sustainable
lic R&D generally targets on basic research and/ economic growth is widely analyzed and recog-
or aims to develop technology to meet the needs nized in academic research. Therefore, to reach
of industrial upgrading. Thus, public R&D may stable and sustainable growth, most nations have
aggressively publish journal articles rather than aggressively devoted more and more efforts toward
apply for patents. When the allocation of national R&D activity. Specifically, utilizing R&D expen-
R&D activity concentrates more on public R&D, diture efficiently is an essential step to achieve
it will have a positive influence on JREI. this goal. This study compares R&D efficiency
The estimated coefficient of Higher education across nations. Each national R&D activity is
RD intensity in column 3 of Table 10 is positive composed of various sectors (private, government,
and significant at the 5% statistical level, whereas and higher education institutions), implying that
the coefficient attached to Government RD inten- there are different targeted innovative outputs in
sity is negative and not significant. As discussed in various R&D programs. Unlike previous studies
the previous section, some nations, e.g., the Czech through the DEA approach that focus on only one

358 INNOVATION: MANAGEMENT, POLICY & PRACTICE Volume 13, Issue 3, December 2011
Role of the national innovation system in R&D efficiency

innovative output and a single index, this study cannot be properly evaluated using a single index.
develops a new methodology to simultaneously When a nation attempts to assess its R&D effi-
assess the individual R&D efficiency of multiple ciency performance, it should acknowledge the
innovative outputs, including patents, RLF, and difference in innovative outputs across sectors.
journal articles. Using a panel dataset of 24 nations Second, since private R&D funded by foreign
over the 1998–2005 period, we first compute vari- sources plays an important role in improving scores
ous output-oriented R&D efficiency indices by on the OREI for patents and RLF, the government
using DEA and then examine the determinants of should cultivate a nurturing environment for inno-
various output-oriented R&D efficiency indices, vation, such as stronger IPR, higher quality human
especially for the role of the NIS. resources, and better technological infrastructure,
Empirical findings show that most nations expe- in order to attract large multinational enterprises to
rienced similar R&D efficiency in terms of their establish R&D labs or to collaborate with domestic
patents and their RLF, while their performance on firms. Third, because R&D intensity performed by
journal publication varied greatly. Hungry, Israel, higher education institutions may have a positive
the United Kingdom, and the United States pre- impact on the JREI, the government should con-
sented the optimal output-oriented R&D efficiency tinuously support universities’ R&D. Moreover,
across the various indices during the sample period. because universities are widely recognized as effec-
Conversely, Romania, Russia, and Mexico performed tive incubators for high-tech start-ups, measures to
the worst in each of the R&D efficiency indices. encourage the collaboration between private enter-
Transition economies, such as Argentina, Romania, prise and universities are critical.
and Russia seemed to have higher efficiency in
publishing journal articles than in applying for pat- ACKNOWLEDGEMENTS
ents and licensing their technologies. Japan is more The authors are indebted to an editor and anony-
accomplished at applying for patents and licensing its mous referees for their helpful comments. Partial
technologies than other nations relative to publishing financial support from Taiwan’s National Science
journal articles. Council (NSC99-2410-H-009-063) is very much
This study further finds that the higher the appreciated.
proportion of private R&D a nation has, the References
higher the nation scores are for the overall OREI, Acemoglu, D., Aghion, P. and Zilibotti, F. (2006)
the PREI, and the RREI. However, this situation Distance to frontier, selection, and economic
reverses when we assess the nations’ scores on the growth, Journal of European Economic Association
JREI. This finding reflects the intrinsic difference 4(1): 37–74.
in targeted innovative outputs between private Banker, R. D., Charnes, A. and Cooper, W. W.
and public (especially higher education institu- (1984) Some models for estimating technical
tions) R&D activities. The interactions between and scale inefficiencies in data envelopment
private R&D and foreign-sourced R&D and analysis, Management Science 30(9): 1078–1092.
between private and academic R&D are found Charnes, A., Cooper, W. W. and Rhodes, E. (1978)
Measuring efficiency of decision making units,
to be positively relevant to most R&D efficiency
European Journal of Operational Research 2(6):
indices. Finally, we find that the higher the pro-
429–444.
portion of higher education R&D a nation has, Coelli, T., Rao, D. S. P. and Battese, G. E. (1998)
the higher the nation scores on the JREI. An introduction to efficiency and productivity
This study derives several key policy implications analysis. Kluwer Academic, Boston.
for innovation. First, because the targeted innova- Edquist, C. (1997) Systems of innovation:
tive outputs of R&D provided by various sectors Technologies, institutions, and organizations.
differ considerably, national-level R&D efficiency Pinter, London, Washington DC.

Volume 13, Issue 3, December 2011 INNOVATION: MANAGEMENT, POLICY & PRACTICE 359
Chiang-Ping Chen, Jin-Li Hu and Chih-Hai Yang

Farrell, M. J. (1957) The measurement of produc- objectives: A DEA approach, European Journal
tive efficiency, Journal of the Royal Statistical of Operational Research 196(3): 847–855.
Society, Series A 120(3): 253–281. Lichtenberg, F. R. (1993) R&D investment and
Furman, J. L., Porter, M. E. and Stern, S. (2002) international productivity differences. In H.
The determinants of national innovative capac- Siebert (Ed.), Economics growth in the world
ity, Research Policy 31(6): 899–933. economy, pp. 47–68, Mohr, Tübingen.
Ginarte, J. C. and Park, W. G. (1997) Lovell, C. A. K. and Schmidt, S. S. (1993) Production
Determinants of patent rights: A cross-national frontiers and productive efficiency, the measurement
study, Research Policy 26(3): 283–301. of productive efficiency: Techniques and applications.
Griffith, R., Huergo, E., Mariesse, J. and Peters, B. Oxford University Press, Oxford.
(2006) Innovation and productivity across four Lundvall, B. (1992) National systems of innovation.
European countries, Oxford Review of Economic Pinter, London, New York.
Policy 22(4): 483–498. Nelson, R. R. (1993) National innovation systems:
Griliches, Z. (1990) Patent statistics as economic A comparative analysis. Oxford University Press,
indicators: A survey, Journal of Economic Oxford.
Literature 28(4): 1661–1707. Nunnenkamp, P. and Spatz, J. (2003) Foreign direct
Guellec, D. and van Pottelsberghe de la Potterie, B. investment and economic growth in developing
(2004) From R&D to productivity growth: Do countries: How relevant are host-country and industry
the institutional settings and the source of funds characteristics? Keil Working Papers No.1176, Kiel,
of R&D matter, Oxford Bulletin of Economics Germany: Kiel Institute for World Economics.
and Statistics 66(3): 353–378. OECD. (2001) OECD science, technology, and
Hall, B. H. and Mairesse, J. (1995) Exploring industry scoreboard. OECD, Paris.
the relationship between R&D and productiv- Pakes, A. and Griliches, Z. (1984) Patents and
ity in French manufacturing firms, Journal of R&D at the firm level: A first look. In Z.
Econometrics 65(1): 263–293. Griliches (Ed.), R&D Patents and Productivity,
Hashimoto, A. and Haneda, S. (2008) Measuring pp. 55–72. University of Chicago Press, Chicago.
the change in R&D efficiency of the Japanese Parisi, M. L., Schiantarelli, F. and Sembenelli, A.
pharmaceutical industry, Research Policy 37(10): (2006) Productivity, innovation and R&D:
1829–1836. Micro evidence for Italy, European Economics
Hu, M. C. and Mathews, J. A. (2008) China’s Review 50(8): 2037–2061.
national innovative capacity, Research Policy Wang, E. C. and Huang, W. C. (2007) Relative effi-
37(9): 1465–1479. ciency of R&D activities: A cross-country study
Lee, H. Y. and Park, Y. T. (2005) An international accounting for environmental factors in the
comparison of R&D efficiency: DEA approach, DEA approach, Research Policy 36(2): 260-273.
Asian Journal of Technology Innovation 13(2): Werner, B. M. and Souder, W. E. (1997)
207–222. Measuring R&D performance: State of the art,
Lee, H. Y., Park, Y. T. and Choi, H. (2009) Research Technology Management 40(2): 34–41.
Comparative evaluation of performance of
national R&D programs with heterogeneous Received 06 June 2010 Accepted 12 July 2011

N O W AVA I L A B L E
Sustainability Innovators: Agents of Change on the Sunshine Coast
by Dana C Thomsen – Pages: vi+74 – ISBN 978-1-921214-49-3
The Sunshine Coast is a contested Australian landscape that has retained much of its natural beauty
and distinctiveness in the face of successive waves of development. This is in no small part due to
the active and vibrant communities that form its social fabric, and the vision and energy of a small
number of passionate individuals prepared to stand out from the crowd and to be advocates for
sustainability.
This collection of case studies is also available in print and online as a course reader.
www.e-contentmanagement.com

360 INNOVATION: MANAGEMENT, POLICY & PRACTICE Volume 13, Issue 3, December 2011

You might also like