Technovation 94–95 (2020) 102094

Contents lists available at ScienceDirect

Technovation
journal homepage: www.elsevier.com/locate/technovation

Reprint of "Performance evaluation of China's high-tech innovation process T

:Analysis based on the innovation value chain"
Xiafei Chen, Zhiying Liu , Qingyuan Zhu
⁎

School of Management, U. Science & Technology of China, Hefei, Anhui 230026, China

ARTICLE INFO ABSTRACT

Keywords: The Chinese high-tech industry has developed greatly since the beginning of China's “National High-tech R&D
Innovation value chain (863) Program” and “China Torch Program”. This paper introduces a conceptual model extended from the
Shared resource innovation value chain model to simultaneously estimate the R&D and commercialization efficiencies for the
High-tech industry innovation high-tech industries of 29 provincial-level regions in China. To match reality, a network DEA incorporating both
Data envelopment analysis
shared inputs and additional intermediate inputs is constructed to open the “black box” view of decision making
units used in single-stage DEA. This study is the first attempt to link the R&D and commercialization with a solid
theoretical foundation and feasible mathematical methods. The empirical findings show that most of the 29
regions have low efficiency in the commercialization sub-process compared to the R&D sub-process, although
there are regional differences in China's high-tech industry. Pearson correlation shows that the R&D sub-process
is not closely correlated to the commercialization sub-process in terms of efficiency. Our analysis can provide
information for the formulation of policies to achieve high innovation efficiency.

1. Introduction
With the advent of the “Third Industrial Revolution”, high-tech in-
novation has reached a highwater mark. On the international stage, the
USA launched its “Star Wars Program” officially called the “Strategic
Defense Initiative” (SDI); France put forward the industrial innovation
plan “create tomorrow's products”; Germany introduced its “2020 -
innovation partnership” and “standard innovation plan”; both the UK
and Singapore launched an “innovation voucher program” (IVS); and
Japan proposed a “digital Japanese innovation plan” (ICT). Having a
transitional economy, China has paid increasing attention to the de-
velopment of high-tech industry. The “Chinese High-tech R&D (863)
Program”, “China Torch Program”, and “Made in China 2025 initiative”
have introduced industrial policies to encourage high-tech industry
development. Against this background, within the analytical framework
of innovation value chain, this study endeavors to analyze the in-
novation efficiency so as to study the improvement path of China's high-
tech industry innovation.
The growth and development of China's high-tech industry gives
evidence that China's commitment is paying off. However, there is still
some way to go before China catches up with high-income countries
such as the USA and Japan. For illustration, Table 1 and Table 2 show
the R&D intensity and the value added respectively. One of the major
scientific and technological development goals of China's 12th Five-
Year Plan (2011–2015) is that “the ratio of value added of high-tech
industry to that of Manufacturing reaches 18%”, which was achieved in
the USA in 2005. The “global competitiveness report 2014–2015”, is-
sued by World Economic Forum, points out that China is in the effi-
ciency-driven stage. The low rate of transformation to productivity and
low level of high-tech industrialization hinder the China's innovation-
driven strategy.
Relevant studies on measurement-oriented high-tech industry ac-
tivities are burgeoning in the literature. Most of the studies focus on R&
D investment and firm performance. Lin et al. (2006) and Lin et al.
(2008) examined the factors influencing firm performance. Hu (2001)
developed an empirical model to study the relationship between gov-
ernment R&D, private R&D, and productivity in Chinese enterprises.
Hong et al. (2015) applied a stochastic frontier analysis model to ex-
plore the relationship of government grants, private R&D funding, and
innovation efficiency of China's high-tech industry. Zhang et al. (2003)
investigated the influence of ownership on the R&D efficiency of Chi-
nese firms. Most of the efficiency evaluation studies use a “black-box”

DOI of original article: https://doi.org/10.1016/j.technovation.2018.02.009

A publisher's error resulted in this article appearing in the wrong issue. The article is reprinted here for the reader's convenience and for the continuity of the special
issue. For citation purposes, please use the original publication details; DOI of original item:http://dx.doi.org/10.1016/j.technovation.2018.02.009
⁎
Corresponding author.
E-mail address: liuzhiyustc@163.com (Z. Liu).

https://doi.org/10.1016/j.technovation.2019.102094
Received 25 November 2015; Received in revised form 12 September 2017; Accepted 23 February 2018
Available online 19 November 2019
0166-4972/ © 2019 Published by Elsevier Ltd.
X. Chen, et al.
Table 1
R&D intensity of high-tech industry in selected countries.
Source: National Bureau of Statistics, China Statistics Yearbook on High
Technology Industry (2013).
China USA Japan Germany UK (2006) Korea
(2012) (2009) (2008) (2007) (2006)
1.68 19.74 10.5 6.87 11.1 5.86
Note: R&D intensity is calculated as the ratio of R&D expenditure to Gross
Industrial Output Value of high-tech industry.
Table 2
Ratio of value added of high-tech Industry to that of Manufacturing in selected
countries.
Source: National Bureau of Statistics, China Statistics Yearbook on High
Technology Industry (2013).
China USA Japan Germany UK (2007) Korea
(2007) (2009) (2008) (2007) (2006)
12.7 21.2 15.4 12.8 17.1 23
framework, which is not consistent with the theory that innovation is a
multistage sequential process (Hage and Hollingsworth, 2000; Hansen
and Birkinshaw, 2007; Porter and Millar, 1985). Porter (2008) pro-
posed that the “value chain” divides a company's activities into design,
production, marketing, delivery, and other related strategic activities; it
is not a collection of independent activities, but rather a system of in-
terdependent activities. Hage and Hollingsworth (2000) developed the
concept of the “idea innovation network”, which has six areas reflecting
research: basic research, applied research, product development re-
search, production research, quality control research, and commercia-
lization/marketing research. Hansen and Birkinshaw (2007) proposed
the “Innovation Value Chain” and recommended viewing innovation as
a sequential, three-phase process that involves idea generation, idea
development, and the diffusion of developed concepts. Therefore, it is
necessary to identify different components of the strengths and weak-
nesses in the innovation process.
The Innovation Value Chain (IVC) is an effective instrument applied
to analyze innovation activities (Hansen and Birkinshaw, 2007; Roper
and Arvanitis, 2012; Roper et al., 2008). For the convenience of eva-
luation, based on the fundamental characteristics of China's high-tech
innovation process, this paper simplifies the innovation activity into
two sub-processes: the “R&D process” and the “commercialization
process”. The R&D process generates knowledge and applies the
knowledge to innovation including basic research, applied research,
and product development research, as detailed in China's National Bu-
reau of Statistics. R&D activity is a catalyst for innovative industrial
activities and ultimately it is responsible for the growth in productivity
and total revenue (Shefer and Frenkel, 2005). The “R&D process” can
be treated as the linkage of “idea generation” and “idea conversion”.
The commercialization process is viewed as introducing innovations
into the market, involving economic activities such as manufacturing
and marketing. The commercialization stage can be treated as the
linkage of “idea conversion” and “idea diffusion”.
Considering that tacit and asymmetric knowledge is hindered by
regional boundaries (Li and Tellis, 2016) and that the social culture and
the specific governance rules in China vary from one region to another,
the regional innovation system (RIS) gives insight into using adminis-
trative regions as the unit of analysis. RIS is an innovation network in
which the innovation actors in a geographic area interact with each
other to achieve knowledge generation, diffusion, and exploitation,
involving various innovation activities operating under the innovation
environment shaped by formal and informal institutions. Originating
from the national innovation system (NIS), RIS, proposed by Cooke
(1992), has been an adequate approach for exploring innovation
2
Technovation 94–95 (2020) 102094
activities of geographical entities (Buesa et al., 2010, 2006; Fritsch,
2002; Wang et al., 2015). Previous studies demonstrate the significance
of conducting research in subnational regions (Guan and Chen, 2010;
Li, 2009; Wang et al., 2015). From a practical point view, the statistical
data for innovation activities are available in the provincial-level re-
gions. Following the common practice, this paper studies the innova-
tion disparity of the high-tech industries in provincial-level regions of
China.
To evaluate efficiency, there are two main approaches in the prior
studies: the stochastic frontier analysis (SFA) approach and the data
envelopment analysis (DEA) approach. Both methodologies are fre-
quently used in empirical analysis and have their individual strengths
and limitations. SFA is only applied to the scenarios with a concrete
form for the production function (Aigner et al., 1977; Battese and Corra,
1977; Meeusen and Van den Broeck, 1977). Data envelopment analysis
(DEA), proposed by Charnes et al. (1978), is a mathematical pro-
gramming approach for analyzing the relative efficiency of peer deci-
sion making units (DMUs) which have multiple inputs and multiple
outputs. As a nonparametric technique, DEA has been applied in the
efficiency analysis in various areas including commercial banks, re-
gional innovation, agricultural economics, hospitals, and enterprises
(Chen et al., 2012; Fu et al., 2007; Guan and Chen, 2010; Liang et al.,
2006). The traditional single-stage DEA model treats the DMUs as a
“black box” without consideration of the internal structure. Due to the
complex internal structure of the DMUs, a number of scholars have
endeavored to develop models with a two-stage internal structure. In
recent years especially, the emerging literature has put forward ap-
proaches for two-stage DEA modeling from various perspectives. The
extended or modified models include the linear DEA models (Chen
et al., 2006), network DEA models (Cook et al., 2010; Halkos et al.,
2015; Kao, 2014; Liu et al., 2013), value-chain DEA models (Chen and
Zhu, 2004; Chiu et al., 2012), and relational two-stage DEA models
(Chen and Guan, 2012; Kao, 2009; Liang et al., 2006). Given the flex-
ibility of the DEA model, in this paper, we select DEA for the evaluation
of R&D and commercialization efficiency of the high-tech industries in
29 Chinese provincial-level regions based on the concept of innovation
value chain.
The integrated conceptual framework in this paper expands the
theory and the method of innovation research and provides a new
perspective on the evaluation of innovation performance of the high-
tech industry. By comparing innovation performance between regions
and identifying the determinants of innovation efficiency, this paper
draws upon and contributes to three streams of research: innovation
value chain, high-tech innovation efficiency at the regional level, and
network DEA. The remaining parts of this paper are organized as fol-
lows. Section 2 presents the literature review and constructs the con-
ceptual framework of high-tech industry innovation. Section 3 provides
a two-stage DEA model for our study. In Section 4, the innovation
performance of the high-tech industries of China's 29 provincial-level
regions is analyzed. Section 5 discussed the implications to theory,
practice and policy. Conclusions are summarized in Section 6.
2. Literature review and conceptual framework
2.1. Literature review
Since the pioneering work of Schumpeter (1934) who placed in-
novation at the core of his “Theory of Economic Development”, many
scholars have endeavored to explore innovation activities. An emerging
body of literature suggests that innovation is vital to creating sub-
stantial and sustainable competitive advantages. The innovation lit-
erature covers various levels such as firms, industries, and areas
(Bernstein and Singh, 2006; Hong et al., 2015; Lee et al., 2014; Wang
et al., 2015). Yang and Liu (2006) found that the adoption of innovation
diffusion in high-tech firms was significantly related to the competition
intent of these firms including aggressive technology posture and
X. Chen, et al.
product development frequency. Tseng et al. (2009) identified a new set
of financial and nonfinancial performance indicators and developed a
business performance evaluation model to measure business perfor-
mance in Taiwan's high-tech manufacturing industry. Zhang and Lv
(2012) applied the quantile regression model to investigate various
relevant factors that impact the innovation performance of high-tech
enterprises. The results showed the discriminate impact of factors such
as enterprise scale, R&D expenditure, net assets debts ratio, and tech-
nical efficiency at different quantile performance points. Lee et al.
(2014) found that the negative relationship between R&D intensity and
a firm's tendency to explore was found to be alleviated in the presence
of technological opportunity or financial slack. These studies provide us
a better understanding of the factors influencing innovation activities
from different perspectives. However, these researchers treat the in-
novation production process as a “black box”, which may ignore some
deficiencies affecting the process. Such methods fail to reveal the
multistage property of the high-tech innovation process. They also give
no consideration to the complexity of innovation elements in the
transformation process of knowledge to revenue.
Some researchers have begun to pay attention to two-structure
systems for studying the operation in a disaggregated sense. Hung and
Wang (2012) examined the managerial performance of 367 manu-
facturing firms including 199 high-tech firms and 168 older-established
firms in Taiwan utilizing the two-stage DEA technique. Their model
assumed that the inputs in the second stage all were outputs from the
first stage, meaning complete linearity of innovation elements. Guan
and Chen (2010) introduced external non-R&D technological innova-
tion inputs in the second stage and suggested that the intertwined ef-
forts of the R&D and non-R&D technological innovation inputs are what
eventually results in the final technological innovation outcomes in the
marketplace. Empirical measurement provides an in-depth look at high-
tech innovation inefficiency for China's provinces in 2002 and 2003.
Chiu et al. (2012) constructed a two-stage evaluation framework to
calculate the R&D efficiency and operation efficiency for 21 of China's
high-tech business. Aside from the patents generated from the first
stage, employees and fixed assets were also used as inputs in the pro-
duction efficiency measure. Wang et al. (2013) proposed an R&D value
chain framework to explore the effects of R&D efforts and basic pro-
duction activities on the market valuations of 65 high-tech firms in
Taiwan. These studies involve “additional inputs” in the second stage,
but they did not consider shared resources, whereas, in reality, common
resources are required to produce innovation revenue through colla-
boration between sub-processes in the high-tech industry.
Shared resources are those that are applied to different segments.
Cook et al. (2000) studied the allocation of shared resources to both
service and sales functions within the banking industry. Zha and Liang
(2010) proposed a product-form cooperative efficiency model for
studying shared flow in a serial two-stage production process. Wu et al.
(2015) built a two-stage DEA model with shared inputs to open the
black box of efficiency measurement in traditional energy efficiency
methods. Wu et al. (2016) treated a transportation system as a parallel
system including two parallel transport subsystems (passenger and
freight). Considering prior research, although the DEA method with
shared inputs is widely used, no studies on the high-tech industry are
available. Therefore, it is necessary to construct a new two-stage DEA
model with shared resources for a cross-region performance evaluation
of China's high-tech industry.
2.2. Conceptual model
As Roper et al. (2008) suggested, the innovation value chain high-
lights the structure and complexity of the process of translating
3
Technovation 94–95 (2020) 102094
knowledge into business value and emphasizes the role of skills, capital
investment, and firms’ other resources in the value creation process.
Bernstein and Singh (2006) emphasized that an integrated approach to
managing the innovation process is required to better understand the
interaction between internal organization activities. Also, Rothwell
(1994) pointed out that the innovation process is not a continuous
process that can be divided into a series of functionally distinct entities,
but rather it is made up of interacting and interdependent stages. In-
novation is vital to knowledge and greatly affects improvements in
technology, especially in the high-tech industries (Tseng et al., 2009).
To evaluate the innovation performance of China's high-tech industry,
the priority is to develop a model which answers the following ques-
tions: (1) What is the basic framework of the innovation process? (2)
What are the key elements involved in the framework? (3) How are
these key factors interrelated?
As mentioned above, firms usually obtain knowledge and transform
it into technologies and products, then exploit their innovation to
generate revenue. In this paper, we consider innovation activity in the
high-tech industry as an innovation process including two sub-pro-
cesses: the R&D sub-process and the commercialization sub-process.
The former sub-process is directed toward advancing basic science and
technology, while the latter aims to realize and refine the commercial
applications of certain technologies which are produced in the former.
This paper selects appropriate indicators for measuring innovation
performance by referring to previous works about the characteristics of
China's high-tech industry (Chiu et al., 2012; Guan and Chen, 2010;
Wang et al., 2013).
Prior studies typically use R&D expenditure and R&D personnel to
measure the inputs of innovation (Cruz-Cázares et al., 2013; Griliches,
1979; Hong et al., 2016; Wang et al., 2015). Consistent with the lit-
erature, we select the R&D personnel and R&D expenditure as the core
innovation inputs. As for the output of the R&D sub-process, the
number of patents granted may be the most appropriate proxy (Bronzini
and Piselli, 2016; Buesa et al., 2006; Cappelen et al., 2012; Guan and
Chen, 2010; Hong et al., 2016; Wang et al., 2013). Some researchers
have pointed out that patents cannot entirely explain the actual in-
novation quality because, as Griliches (1990) comments, “not all in-
ventions are patentable, not all inventions are patented, and the in-
ventions that are patented differ greatly in quality.” Despite such
objections, the correlation between patent activity and innovation is
close (r = 0.934) (Feldman and Florida, 1994). Compared to alternative
measurements, patents can guarantee originality and are more likely to
achieve market value (Bronzini and Piselli, 2016; Buesa et al., 2010).
Because of these considerations, it is reasonable to use the patent count
as an indicator of innovation performance. In addition, Lev (2000) in-
dicated that patents are one important indicator of R&D innovation
outcomes. In this paper, patent applications are regarded as a measure
of R&D output. Since new product development is an important factor
which can reflect R&D efficiency, it is also chosen as a new R&D output
in this paper.
Concerning the commercialization process, as suggested by Guan
and Chen (2010), we take into account commercialization-dedicated
inputs as additional intermediate inputs. Based on practical operations
and the availability of data, these additional intermediate inputs are
measured by expenditure on new product development, expenditure for
technical renovation, and employed personnel. According to common
practices (Czarnitzki et al., 2011; Maietta, 2015; Spanos et al., 2015),
the final realization of commercialization is measured by the sales
revenue from new products.
It is evident that in the high-tech industry, fixed assets and tech-
nology resources are shared inputs. We cannot determine precisely each
stage's share of these two kinds of inputs but we can obtain total inputs
X. Chen, et al.
Shared inputs
R&D Process
Dedicated inputs(I)
Dedicated inputs(II)
Idea generation
Innovation value chain
Fig. 1. The conceptual framework of high-tech industry innovation.
for the whole innovation process. Considering the path dependency
effect, we use knowledge capital to represent the technological re-
sources. As was done in Hu and Mathews (2008) and Guan and Chen
(2010), this paper employs the number of invention patents as an in-
dicator of technological resources.
We propose a conceptual model which has the innovation value
chain as the backbone. The R&D sub-process and commercialization
sub-process are incorporated within this model. The input-output con-
ditions of each process and the distribution structure for shared re-
sources between the sub-processes are taken into consideration. Fig. 1
shows the basic conceptual model, which contributes to a better un-
derstanding of the high-tech innovation process for each region in our
study.
3. Data and methodology
3.1. Data
This research studies the high-tech innovation activities of 29 pro-
vincial-level regions in mainland China (excluding Tibet and Qinghai
due to incomplete data). The statistical information is from the China
Statistical Yearbook on the High Technology Industry, which focuses on
China's High Technology Industry (Manufacturing Industry)
Classification, and the Annual Report of Patent Statistics (ARPS). Figures
for R&D personnel (R&DP), intramural expenditure on R&D (R&DE),
patent applications (PA), the number of new product development
(NPD), expenditure on new product development (ENPD), expenditure
for technical renovation (ETR), and employed personnel (EP), sales
revenue from new products (SRND), and fixed assets (FA) are extracted
from the China Statistical Yearbook on the High Technology Industry. The
invention patents stock totals (IPS, meaning the number of invention
patents that can be used by the province) are collected from the ARPS.
The definitions for each variable are detailed as follows.
Intramural expenditure on R&D (R&DE): The expenses of an en-
terprise for internal R&D activities, including researching and devel-
oping technology, patents, and the administrative expenses related to R
&D activities.
R&D personnel (R&DP): This usual international indicator refers to
the number of research workers, engineers, designers, and R&D re-
levant staff who are important for motivating firms to become involved
in R&D innovation activities.
Patent applications (PA): Refers to the number of patent applica-
tions by an enterprise per year. The patent application types include
inventions, utility models, and designs.
Technovation 94–95 (2020) 102094
Final
Intermediate Commercialization outcomes
outputs
Process
Idea conversion Idea diffusion
New product development (NPD): The planning and development of
products, which result in new product launches or new process im-
provements.
Expenditure for technical renovation (ETR): The expenses of an
enterprise to integrate novel technological achievements into its pro-
duction process, which can improve the product quality, update the
product, and improve the synthetic economic benefit.
Expenditure on new product development (ENPD): The expenses
used by an enterprise to develop new products.
Employed personnel (EP): Refers to the number of employees in-
volved in the operation of production processes, including the em-
ployees devoted to production and marketing activities.
Sales revenue of new products (SRND): The annual total value of
new product sales in the high-tech industry, which represents the
market value associated with the new technology.
Invention patents stock (IPS): The total invention patents at the
provincial level, which indicates the technology resources that can be
used in the high-tech industries of one province. IPS is used both for the
R&D stage and the commercialization stage.
Fixed assets (FA): Refers to the cumulative total fixed asset invest-
ment, including typical resources that support high-tech innovation
activities. FA is used both for the R&D stage and the commercialization
stage.
It is important to keep in mind the existence of time lags when
analyzing innovation activities. Following Hong et al. (2015) and Guan
and Chen (2010), this paper selects a 1-year lag for the R&D process and
a 1-year lag for the commercialization process in the high-tech industry.
Because of the dissimilarity of statistical scales, continuous data for
China's high-tech industries innovations is only applied to the moving
average of three years for 2010–2012 and 2011–2013. Therefore, for
the sake of comparison, we choose two consecutive cross-sectional
databases in 2010 and 2011. For the shared inputs, we take the average
value of the three years. The fixed assets, which include the accumu-
lated resources that support high-tech innovation activities, are com-
puted with the Perpetual Inventory Method proposed by Goldsmith
(1951). Regarding the year 2000 as the base period and the deprecia-
tion rate as 8%, we take the base period data divided by 10% as the
base levels (Hall and Jones, 1999).
All the variables are measured in China's high-tech industry at the
provincial level and are summarized in Table 3. The descriptive sta-
tistics for the inputs and outputs of 29 regions in China for 2010–2012
and 2011–2013 are presented in Table 4. From this table, it can be
observed that the mean values and the standard deviations of all ten
variables increased, which indicates that the high-tech industry
4
X. Chen, et al. Technovation 94–95 (2020) 102094

Table 3
Input-output variables and Source.
High-tech innovation process Variable Unit Source

R&D-dedicated inputs R&D personnel (R&DP) 10,000 persons China Statistics Yearbook on High Technology Industry 2011, 2012
Intramural Expenditure on R&D (R&DE) 10,000 yuan China Statistics Yearbook on High Technology Industry 2011, 2012
Intermediate outputs Patent applications (PA) Piece China Statistics Yearbook on High Technology Industry 2012, 2013
New product development(NPD) item China Statistics Yearbook on High Technology Industry 2012, 2013
Commercialization-dedicated inputs Expenditure on New product development 10,000 yuan China Statistics Yearbook on High Technology Industry 2012, 2013
(ENPD)
Expenditure for Technical Renovation 10,000 yuan China Statistics Yearbook on High Technology Industry 2012, 2013
(ETR)
Employed personnel (EP) Person China Statistical Yearbook on Science and Technology 2012, 2013
Final outputs Sales revenue of new products (SRND) 10,000 yuan China Statistics Yearbook on High Technology Industry 2013, 2014
Shared inputs Invention patents stock (IPS) Item Annual Report of Patent Statistics (ARPS) by State Intellectual Property
Office of China (2011–2014)
Fixed assets (FA) 100 million yuan China Statistics Yearbook on High Technology Industry: 2001–2014

investments and outputs increased. The high standard deviations imply assume region j being evaluated is denoted by DMU0 (DMU0 DMUj ),
that there may be uneven innovation performance across different the R&D efficiency by E10 , and the commercialization efficiency by E20 .
provinces. This corollary is reasonable in real life because of the im- This paper presents the two-stage DEA model based on the constant
balanced regional development in China, including China's differ- returns to scale (CRS) model introduced by Charnes et al. (1978). The
entiation development strategy, the unbalanced regional economies, CRS efficiency scores for the evaluated DMU in the first and second
and the industrial structure disparity. stage can be written by the models (1) and (2) as follows.
2 1
k1= 1 v k1 Yk1 0
3.2. Methodology Max E10 = 2 2 1
i1= 1 ui1 Xi1 0 + s = 1 us s 0 X s 0
2 1
k1= 1 v k1 Yk1 j
3.2.1. Overall process performance assessment model s. t . E1j = 2 2 1 1
i1= 1 ui1 Xi1 j + s = 1 us sj X sj
Fig. 1 shows the two-stage network process of a high-tech industry
system with shared inputs associated with both stages. For notational Lsj1 sj Lsj2
purposes, each region is viewed as a DMUj (j = 1,2, …, n). Let v k11 , ui1, us1
Yk1 j (k1 = 1,2) , Yk2 j (k2 = 1) denote respectively the outputs of the R&D j = 1, 2, …, n . (1)
and commercialization sub-processes. The intermediate outputs flow
from sub-process 1 to become inputs of sub-process 2. On the input side, v k 2 Yk 2 0
Xi1 j (i1 = 1,2) , Xi2 j (i2 = 1,2,3) , and Xsj (s = 1,2) denote the R&D dedi- Max E20 = 3 2 2 2 2
i2 = 1 ui2 Xi2 0 + s = 1 us (1 s 0) Xs 0 + k1= 1 v k1 Yk1 0
cated-, commercialization dedicated-, and shared inputs, respectively. v k 2 Yk 2 j
Let an allocative factor sj (0 s. t . E2j = 1
sj 1) denote the proportion of shared 3 2 2 2 2
i2 = 1 ui2 Xi2 j + s = 1 us (1 sj ) X sj + k1= 1 v k1 Yk1 j
inputs to be assigned to the R&D sub-process, i.e. sj Xsj and (1 sj ) Xsj
correspond to the portions of shared inputs allocated to the first and Lsj1 sj Lsj2
second stage, respectively. According to Cook and Hababou (2001) and v k2, v k21 , ui2, us2
Wu et al. (2015), certain inequality constraints should be considered, j = 1, 2, …, n . (2)
namely Lsj1 sj Lsj2 .
Following the general practice (Chiu et al., 2012; Wu et al., 2015), Lsj1 and Lsj2 are the lower and upper bounds for the shared resources.

Table 4
Descriptive statistics for the inputs and outputs of 29 regions in China.
R&DP R&DE PA NPD ENPD ETR EP SRND IPS FA

2010
Mean 15,977.69 333,687.21 2296.31 1869.24 617,519.38 10,5077.17 39,5291.55 8,816,755.03 11,289.93 857.45
Std. 32,984.20 683,255.12 3172.32 4745.61 1,154,836.55 177,673.27 740,726.13 18,182,013.64 14,917.15 1060.83
Max 174,644.00 3,630,850.40 15,310.00 25,961.00 5,812,831.00 960,814.00 3,614,903.00 85,195,533.00 59,813.67 5455.95
Min 114.00 3671.70 36.00 16.00 7060.00 3621.00 5612.00 2253.00 342.67 18.04
2011
Mean 20,610.77 480,283.63 2774.20 2228.53 709391.23 127,200.24 422,853.07 10,409,088.20 14,420.42 1061.30
Std. 37,831.26 912,061.24 4013.28 5291.32 1,337,499.35 223,967.17 777,639.27 20,106,742.01 19,336.83 1336.36
Max 198,187.00 4,809,951.00 19,177.00 29,210.00 6,502,335.00 1,225,605.00 3,842,156.00 97,687,741.60 77,675.00 7036.95
Min 45.00 361.00 16.00 2.00 1374.00 1449.00 3843.00 10,065.00 287.67 19.75

5
X. Chen, et al. Technovation 94–95 (2020) 102094

The certain intervals are used to avoid extreme bias toward one of the 2
k1= 1 v k1 Yk1 0 + v k 2 Yk2 0
Max E0 =
sub-processes (Chen et al., 2010; Wu et al., 2015). Similar to Liang et al. 2
u X
i1= 1 i1 i1 0 + 2 3
s = 1 us Xs 0 + i2 = 1 ui2 Xi2 0 +
2
s = 1 v k1 Y k1 0
(2008) and Chen et al. (2009), we assume that v k11 = v k21 = v k1 for all 2
k1= 1 v k1 Yk1 j
k1 = 1,2 in models (1) and (2). As for the shared inputs, we assume that s. t . 2 2 1
i1= 1 ui1 Xi1 j + s = 1 us sj Xsj
us1 = us2 = us for all s = 1,2 because they pertain to the same types of
v k 2 Yk 2 j
inputs. This is also consistent with the principle that the multiplier of a 3 2 2 1
i2 = 1 ui2 Xi2 j + s = 1 us (1 sj ) Xsj + k1= 1 v k1 Yk1 j
factor is always the same, no matter where or how it is used (Kao,
2 2
2009). w1 =
i1= 1 ui1 Xi1 0 + s = 1 us s 0 X s 0
a
2 2 3 2
Following Chen et al. (2010), this paper combines the two sub-stage i1= 1 ui1 Xi1 0 + s = 1 us Xs 0 + i2 = 1 ui2 Xi2 0 + k1= 1 v k1 Yk1 0

efficiencies in a weighted average of efficiency scores as the overall 2 2

i1= 1 ui2 Xi2 0 + s = 1 us (1
2
s 0) Xs 0 + k1= 1 v k1 Yk1 0
w2 = b
efficiency E0 . 2 2
i1= 1 ui1 Xi1 0 + s = 1 us Xs 0 +
3 2
i2 = 1 ui2 Xi2 0 + k1= 1 v k1 Yk1 0

Lsj1 sj Lsj2
E0 = 1 E10 + 2 E20
v k1, vk2, ui1, ui2, us
Note that 1 and 2 are the weights for the R&D sub-process and the j = 1, 2, …, n (3)
commercialization sub-process, and that w1 + w2 = 1. Sub-process 1 (2)
with efficiency E1j (E2j) is efficient if and only if E1j = 1 (E2j = 1). DMU0 To transform model (3) into a standard linear program, we can
is overall efficient if and only if E0 = 1, and by the equation above, this apply the usual Charnes and Cooper transformation so that the above
is possible if and only if each of its sub-processes is efficient. fractional program can be converted to the following linear model (4).
1 and 2 represent the relative importance or contribution of the Let t = 2
1
, µ i1 = ui1 t ,
2 3 2
efficiency of each sub-process to the overall performance of the given i1= 1 ui1 Xi1 0 + s = 1 us X s 0 + i2 = 1 ui2 Xi2 0 + s = 1 v k1 Yk1 0

DMU. One reasonable weight choice of each sub-process is the pro- µs = us t , µ i2 = ui2 t , k1 = vk1 t , k2 = v k2 t , sj = µs sj

portion of total resources devoted to each sub-process, reflecting the Then model (4) can be written as:
relative size and importance of a sub-process. According to Chen et al. 2
Max E0 = k1 Yk1 0+ k 2 Yk2 0
(2009) and Cook et al. (2010), to be more specific, we then have: k1= 1

2 2 2
2 2 s. t . k1= 1 k1 Yk1 j µ X
i1= 1 i1 i1 j
+ X
s = 1 sj sj
0
u X
i1= 1 i1 i1 0
+ u X
s =1 s s0 s0
1 = 2 2 3 2
u X +
i1= 1 i1 i1 0
uX
s= 1 s s0
+ u X
i2 = 1 i2 i2 0
+ v Y
k1= 1 k1 k1 0
3 2 2
k 2 Yk 2 j µ X
i2 = 1 i2 i2 j
+ s=1
(µs sj ) Xsj + k1= 1 k1 Yk1 j 0
2 2 2
u X
i1= 1 i2 i2 0
+ u (1
s=1 s s 0 ) Xs 0 + v Y
k1= 1 k1 k1 0 2 2 3 2
2 = 2 2 3 2
µ X +
i1= 1 i1 i1 0
µX
s= 1 s s0
+ µ X
i2 = 1 i2 i2 0
+ s= 1 k1 Yk1 0 =1
u X
i1= 1 i1 i1 0
+ uX
s= 1 s s0
+ u X + k1= 1 v k1 Yk1 0
i2 = 1 i2 i2 0 2 2
µ X +
i1= 1 i1 i1 0
X
s = 1 s0 s0
a
2 2 3 2 2 2 2
where u X
+
i1= 1 i1 i1 0
+ uX
s = 1 s s0
+ u X
i2 = 1 i2 i2 0
re- v Y
k1= 1 k1 k1 0
u X +
i1= 1 i2 i2 0 s=1
( µs s0
) Xs0 + k1= 1 k1 Yk1 0 b
presents the total amounts of input consumed by the whole two-stage Lsj1 sj sj Lsj2 sj
network structure with intermediate outputs used, while k1 , k2 , µi1 , µi2 , µs
u X + s = 1 us s 0 Xs 0 and i1= 1 ui2 Xi2 0 + s = 1 us (1 sj ) Xs0 + k1= 1 v k1 Yk1 0
2 2 2 2 2
i1= 1 i1 i1 0 j = 1, 2, …, n (4)
represent the sizes of the R&D and commercialization sub-processes,
respectively.
By solving linear program (4), the optimal solutions
Then the overall efficiency E0 can be written in the form
(wk*1, w k*2, µ *i1, µ *i2, µ*s , sj*) will be obtained. Since sj = µs sj , we have
*
(s = 1,2j = 1,2, …, n) . Through the above steps, we can obtain
2 sj
k1= 1 v k1 Yk1 0 *
sj =
E0 = w1 + w2 µs*
each DMU's optimal overall efficiency and the optimal proportions for
2 2
i1= 1 ui1 Xi1 0 + s = 1 us s 0 Xs 0
v k2 Yk 2 0 the shared inputs
3 2 2
i2 = 1 ui2 Xi2 0 + s = 1 us (1 s 0 ) Xs 0 + k1= 1 v k1 Yk1 0
2
k1= 1 v k1 Yk1 0 + v k2 Yk 2 0
= 2
i1= 1 ui1 Xi1 0 +
2 3
s = 1 us X s 0 + i2 = 1 ui2 Xi2 0 +
2
s = 1 v k1 Yk1 0 3.2.2. Sub-process assessment model: efficiency decomposition
Once we obtain an optimal solution to model (4), the efficiency
To overcome the possible drawback that some i = 1(i = 1,2) at scores of individual sub-processes can be calculated accordingly.
optimality, we impose bounding restrictions w1 a and w2 b as in However, model (4) may have multiple optimal solutions, so the in-
Amirteimoori (2013), where the a and b represent the minimum weight dividual efficiency may also not be unique. Therefore, we follow Kao
for R&D sub-process and commercialization sub-process respectively in and Hwang's (2008) approach to find a set of multipliers which produce
calculating the overall efficiency. The overall efficiency of the whole the highest R&D or commercialization sub-process efficiency score
two-stage network structure with intermediate outputs used for a while maintaining the overall efficiency score of the whole process.
common DMU0 can be evaluated by the following model. Denote the optimal overall efficiency score for DMU0 obtained from
model (4) as E0* . The maximum efficiency value achievable for sub-
process 1 (R&D stage), denoted E10 * while maintaining the overall effi-
ciency score can be determined via the following model (5).

6
X. Chen, et al. Technovation 94–95 (2020) 102094

2
k1= 1 v k1 Yk1 0 According to Kao and Hwang (2008) and Halkos et al. (2014), the
Max E10 = 2 2
i1= 1 i1 i1 0 + s = 1 us s 0 Xs 0
u X efficiency of sub-process 2 (commercialization) can be calculated as
E* *E *
2
k1= 1 v k1 Yk1 j
E20 = 0 *1 10 , where 1* and 2* are the optimal weights based upon
s. t . E1j = 1 2
2 2
i1= 1 ui1 Xi1 j + s = 1 us sj Xsj model (4), and E10 * represents the optimal efficiency of sub-process1
v k2 Yk2 j while maintaining the overall efficiency in model (6). However, be-
E2j = 1
3
i2 = 1 ui2 Xi2 j +
2
s = 1 us (1 sj ) Xsj +
2
k1= 1 v k1 Yk1 j cause of the multiple optimal solutions of model (4), 1* and 2* may
2
k1= 1 v k1 Yk1 0 + u1 + v k 2 Yk2 0
also not be unique. Therefore, consistent with previous research (Wu
E0* = 2 2 3 2 et al., 2015), we propose to produce the highest efficiency score for sub-
i1= 1 ui1 Xi1 0 + s = 1 us X s 0 + i2 = 1 ui2 Xi2 0 + s = 1 v k1 Y k1 0
2 2 process 2 while maintaining the overall efficiency score E0* and sub-
i1= 1 ui1 Xi1 0 + s = 1 us s 0 X s 0
w1 = 2 2 3 2 a process 1's maximum achievable efficiency E10 * . The maximum achiev-
i1= 1 ui1 Xi1 0 + s = 1 us Xs 0 + i2 = 1 ui2 Xi2 0 + k1= 1 v k1 Yk1 0
able value of E20
1*
can be determined via the following linear model (7).
2 2 2
w2 =
i1= 1 ui2 Xi2 0 + s = 1 us (1 s 0) Xs 0 + k1= 1 v k1 Yk1 0
b This is an approach of efficiency decomposition which has not pre-
viously been applied in a chain system.
2 2 3 2
i1= 1 ui1 Xi1 0 + s = 1 us Xs 0 + i2 = 1 ui2 Xi2 0 + k1= 1 v k1 Yk1 0

Lsj1 sj Lsj2
ui1, ui2, us , vk1, vk 2
j = 1, 2, …, n (5)

Model (5) can be converted into the following linear program (6) as
explained above.

2
* =
Max E10 k1= 1 k1 Yk1 0
2 2 2
s. t. k1= 1 k1 Yk1 j ( µ X
i1= 1 i1 i1 j
+ X )
s = 1 sj sj
0

3 2 2
k2 Yk2 j µ X
i2 = 1 i2 i2 j
+ s= 1
(µs sj
) Xsj + k1= 1 k1 Yk1 j 0

2 2 2 2 3 2
k1= 1 k1 Yk1 0+ k2= 1 k2 Yk2 0 + uB E0* ( µ X
i1= 1 i1 i1 0
+ µX
s =1 s s0
+ µ X
i2 = 1 i2 i2 0
+ s= 1 k1 Yk1 0 ) =0
2 2 3 2 2 2
a ( µ X
i1= 1 i1 i1 0
+ µX
s =1 s s0
+ µ X
i2 = 1 i2 i2 0
+ k1= 1 k1 Yk1 0 ) ( µ X
i1= 1 i1 i1 0
+ X ) 0
s =1 s0 s0
2 2 3 2 2 2 2
b ( µ X
i1= 1 i1 i1 0
+ µX
s =1 s s0
+ µ X
i2 = 1 i2 i2 0
+ k1= 1 k1 Yk1 0 ) ( µ X
i1= 1 i2 i2 0
+ s=1
(µs s0
) Xs0 + k1= 1 k1 Yk1 0 ) 0
2 2
µ X
i1= 1 i1 i1 0
+ X
s =1 s s0
=1
1
Lsj sj sj Lsj2 sj

k1, k2 , µi1 , µi2 , µs

j = 1, 2, …, n (6)

1*
Max E20 = k2 Yk2 0
2 2 2
s. t. k1= 1 k1 Yk1 j ( µ X
i1= 1 i1 i1 j
+ X )
s = 1 sj sj
0

3 2 2
k2 Yk2 j µ X
i2 = 1 i2 i2 j
+ s= 1
(µs sj
) Xsj + k1= 1 k1 Yk1 j 0

2 2 2 3 2
k1= 1 k1 Yk1 0+ k2 Yk2 0 E0* ( µ X + s= 1 µs Xs0
i1= 1 i1 i1 0
+ µ X
i2 = 1 i2 i2 0
+ s=1 k1 Yk1 0 ) =0
2 2 2
k1= 1 k1 Yk1 0 * (
E10 µ X + s = 1 s0 Xs0 ) 0
i1= 1 i1 i1 0
2 2 3 2 2 2
a ( µ X
i1= 1 i1 i1 0
+ µX
s =1 s s0
+ µ X
i2 = 1 i2 i2 0
+ k1= 1 k1 Yk1 0 ) ( µ X
i1= 1 i1 i1 0
+ X ) 0
s =1 s0 s0
2 2 3 2 2 2 2
b ( µ X
i1= 1 i1 i1 0
+ µX
s =1 s s0
+ µ X
i2 = 1 i2 i2 0
+ k1= 1 k1 Yk1 0 ) ( µ X
i1= 1 i2 i2 0
+ s=1
(µs s 0 ) Xs 0 + k1= 1 k1 Yk1 0 ) 0
2 2
µ X
i1= 1 i1 i1 0
+ X
s =1 s s0
=1
3 2 2
µ X
i2 = 1 i2 i2 0
+ s=1
(µs s0
) Xs0 + k1= 1 k1 Yk1 0 =1
Lsj1 sj sj Lsj2 sj

k1, k2 , µi1 , µi2 , µs

j = 1, 2, …, n (7)

7
X. Chen, et al. Technovation 94–95 (2020) 102094

Similarly, the following linear program can be established to max- score. The highest achievable efficiency scores E21j* and E12j* of the
imize sub-process 2's efficiency score E20
* while maintaining the overall commercialization and R&D sub-processes based upon models (7) and
efficiency score of E0* . (9) are specified in columns 4 and 6, respectively. Columns 7 and 8

2
Max *
E20 = k1= 1 k1 Yk1 0
2 2 2
s. t. k1= 1 k1 Yk1 j ( µ X
i1= 1 i1 i1 j
+ X )
s = 1 sj sj
0

3 2 2
k2 Yk2 j µ X
i2 = 1 i2 i2 j
+ s= 1
(µs sj
) Xsj + k1= 1 k1 Yk1 j 0

2 2 2 3 2
k1= 1 k1 Yk1 0+ k2 Yk2 0 E0* ( µ X
i1= 1 i1 i1 0
+ µX
s =1 s s0
+ µ X
i2 = 1 i2 i2 0
+ s=1 k1 Yk1 0 ) =0
2 2 3 2 2 2
a ( µ X
i1= 1 i1 i1 0
+ µX
s =1 s s0
+ µ X
i2 = 1 i2 i2 0
+ k1= 1 k1 Yk1 0 ) ( µ X
i1= 1 i1 i1 0
+ X ) 0
s =1 s0 s0
2 2 3 2 2 2 2
b ( µ X
i1= 1 i1 i1 0
+ µX
s =1 s s0
+ µ X
i2 = 1 i2 i2 0
+ k1= 1 k1 Yk1 0 ) ( µ X
i1= 1 i2 i2 0
+ s=1
(µs s0
) Xs0 + k1= 1 k1 Yk1 0 ) 0
2 2
µ X
i1= 1 i1 i1 0
+ X
s =1 s s0
=1
Lsj1 sj sj Lsj2 sj

k1, k2 , µi1 , µi2 , µs

j = 1, 2, …, n (8)

The highest efficiency score E10

2*
for sub-process 1 while maintaining indicate the optimal proportions of shared resources (technology re-
the overall efficiency score E0* and sub-process 2's maximum achievable sources and fixed assets respectively) for the R&D sub-process. In ad-
* can be determined via the following linear model (9).
efficiency E20 dition, the last two rows of Table 5 show the average efficiency scores

2* 2
Max E10 = k1= 1 k1 Yk1 0
2 2 2
s. t. k1= 1 k1 Yk1 j ( µ X
i1= 1 i1 i1 j
+ X )
s = 1 sj sj
0

3 2 2
k2 Yk2 j µ X
i2 = 1 i2 i2 j
+ s= 1
(µs sj
) Xsj + k1= 1 k1 Yk1 j 0

2 2 2 3 2
k1= 1 k1 Yk1 0+ k2 Yk2 0 E0* ( µ X
i1= 1 i1 i1 0
+ µX +
s =1 s s0
µ X
i2 = 1 i2 i2 0
+ s=1 k1 Yk1 0 ) =0
3 2 2
k2 Yk2 0 * (
E20 µ X
i2 = 1 i2 i2 0
+ s=1
(µs s 0 ) Xs 0 + k1= 1 k1 Yk1 0 ) =0
2 2 3 2 2 2
a ( µ X
i1= 1 i1 i1 0
+ µX
s =1 s s0
+ µ X
i2 = 1 i2 i2 0
+ k1= 1 k1 Yk1 0 ) ( µ X
i1= 1 i1 i1 0
+ X ) 0
s =1 s0 s0
2 2 3 2 2 2 2
b ( µ X
i1= 1 i1 i1 0
+ µX
s =1 s s0
+ µ X
i2 = 1 i2 i2 0
+ k1= 1 k1 Yk1 0 ) ( µ X
i1= 1 i2 i2 0
+ s=1
(µs s0
) Xs0 + k1= 1 k1 Yk1 0 ) 0
2 2
µ X
i1= 1 i1 i1 0
+ X
s =1 s s0
=1
1
Lsj sj sj Lsj2 sj

k1, k2 , µi1 , µi2 , µs

j = 1, 2, …, n (9)

Solving the above models, we can obtain each model's optimal value and standard deviations of the 29 provincial-level regions. The order of
* , E20
E10 1*
, E20
* , and E10
2*
. If the results satisfy E10 2*
* = E10 and E20 1*
* = E20 , we data for the year 2011 is the same as that for 2010. To help the readers
arrive at the conclusion that a unique efficiency decomposition is ob- understand, detailed explanations of the efficiencies under different
tained. conditions are given as follows.

4. Results E j*: The maximum overall system efficiency score of the high-tech
innovation in each provincial-level region.
Based on the methodology provided in Section 3, we calculate the E1*j : The highest achievable efficiency score of the R&D sub-process
innovation efficiency of the high-tech industries in 29 of China's pro- of the high-tech innovation in each provincial-level region while
vincial-level regions. For the shared inputs, we set 0.2 sj 0.8 to maintaining the maximum overall system efficiency score (E j*).
conform to reality. In addition, we require that the weights of each E21j*: The highest achievable efficiency score of commercialization
subsystem within the overall system are not smaller than 0.3, that is, sub-process of the high-tech innovation in each provincial-level re-
a = b = 0.3. gion while maintaining the maximum overall system efficiency
The CRS results from models (4), (6), (7), and (8) for 2010 and 2011 score (E j*) and the maximum R&D efficiency (E1*j ).
are reported in Table 5. For the year 2010, the second column reports E2*j : The highest achievable efficiency score of commercialization
the overall systems’ efficiency scores E j* of high-tech innovation process sub-process of the high-tech innovation in each provincial-level re-
of these 29 provincial-level regions in mainland China. Columns 3 and 5 gion while maintaining the maximum overall system efficiency
show the highest achievable efficiency scores E1*j and E2*j of R&D and score (E j*).
commercialization respectively while maintaining the overall efficiency E12j*: The highest achievable efficiency score of the R&D sub-process

8
X. Chen, et al. Technovation 94–95 (2020) 102094

Table 5
Innovation efficiencies of high-tech industry in 29 regions of China from 2010 to 2011.
2010 2011

E*j E1*j E21j* E2*j E12j* 1j 2j E*j E1*j E21j* E2*j E12j* 1j 2j

Beijing 1.000 1.000 1.000 1.000 1.000 0.556 0.447 1.000 1.000 1.000 1.000 1.000 0.554 0.423
Tianjin 1.000 1.000 1.000 1.000 1.000 0.390 0.460 1.000 1.000 1.000 1.000 1.000 0.420 0.449
Hebei 0.460 0.858 0.290 0.412 0.507 0.671 0.578 0.426 1.000 0.148 0.219 0.597 0.679 0.573
Shanxi 0.597 0.478 0.686 0.686 0.478 0.649 0.548 0.435 0.705 0.052 0.052 0.705 0.582 0.578
Inner Mongolia 0.506 0.545 0.457 0.457 0.545 0.778 0.663 0.364 0.452 0.169 0.169 0.452 0.563 0.564
Liaoning 0.403 0.413 0.399 0.496 0.283 0.674 0.578 0.366 0.561 0.138 0.185 0.460 0.524 0.578
Jilin 0.623 0.786 0.416 0.416 0.786 0.562 0.554 0.338 0.419 0.149 0.149 0.419 0.200 0.384
Heilongjiang 0.257 0.716 0.060 0.119 0.333 0.645 0.581 0.235 0.675 0.047 0.095 0.310 0.605 0.573
Shanghai 0.514 0.815 0.385 0.592 0.449 0.584 0.626 0.489 1.000 0.209 0.301 0.641 0.654 0.679
Jiangsu 0.980 1.000 0.970 1.000 0.944 0.358 0.567 0.742 1.000 0.593 0.765 0.720 0.711 0.578
Zhejiang 0.733 0.588 0.795 1.000 0.398 0.621 0.651 0.770 0.913 0.708 1.000 0.544 0.588 0.513
Anhui 0.693 1.000 0.548 0.696 0.689 0.694 0.502 0.528 1.000 0.229 0.262 0.837 0.755 0.574
Fujian 1.000 1.000 1.000 1.000 1.000 0.392 0.493 0.917 1.000 0.868 1.000 0.818 0.708 0.567
Jiangxi 0.721 1.000 0.576 0.621 0.877 0.765 0.585 0.697 1.000 0.466 0.484 0.952 0.783 0.568
Shandong 0.803 0.966 0.733 1.000 0.595 0.637 0.552 0.612 1.000 0.443 0.694 0.546 0.742 0.719
Henan 0.379 0.759 0.216 0.310 0.444 0.705 0.526 0.953 0.842 1.000 1.000 0.000 0.200 0.250
Hubei 0.558 0.763 0.463 0.650 0.470 0.608 0.577 0.447 0.887 0.258 0.438 0.453 0.704 0.655
Hunan 0.589 0.867 0.469 0.603 0.572 0.694 0.470 0.533 0.650 0.482 1.000 0.294 0.568 0.525
Guangdong 1.000 1.000 1.000 1.000 1.000 0.441 0.491 1.000 1.000 1.000 1.000 1.000 0.510 0.398
Guangxi 0.582 1.000 0.164 0.164 1.000 0.793 0.796 0.402 0.993 0.149 0.245 0.519 0.648 0.580
Hainan 0.556 1.000 0.192 0.192 1.000 0.800 0.800 0.562 1.000 0.193 0.193 1.000 0.800 0.800
Chongqing 0.652 0.426 0.749 0.749 0.426 0.800 0.787 0.421 0.617 0.102 0.102 0.617 0.586 0.577
Sichuan 0.626 1.000 0.329 0.340 0.962 0.787 0.578 0.586 1.000 0.249 0.249 1.000 0.800 0.800
Guizhou 0.498 1.000 0.161 0.183 0.844 0.743 0.469 0.518 1.000 0.194 0.223 0.839 0.744 0.681
Yunnan 0.489 0.611 0.337 0.337 0.611 0.572 0.571 0.414 0.661 0.231 0.374 0.437 0.616 0.540
Shaanxi 0.331 0.716 0.167 0.345 0.324 0.568 0.568 0.333 0.902 0.089 0.150 0.463 0.654 0.574
Gansu 0.376 0.232 0.437 0.437 0.232 0.800 0.800 0.332 0.442 0.082 0.082 0.442 0.623 0.601
Ningxia 0.542 1.000 0.307 0.326 0.899 0.763 0.571 0.547 1.000 0.153 0.153 1.000 0.800 0.800
Xinjiang 0.414 0.701 0.005 0.005 0.701 0.748 0.574 0.098 0.114 0.060 0.060 0.114 0.202 0.200
Mean 0.617 0.801 0.494 0.556 0.668 0.648 0.585 0.554 0.822 0.361 0.436 0.627 0.604 0.562
S.D. 0.211 0.220 0.299 0.306 0.256 0.128 0.099 0.237 0.238 0.322 0.357 0.276 0.166 0.139

of the high-tech innovation in each provincial-level region while
maintaining the maximum overall system efficiency score (E j*) and
the maximum R&D efficiency (E2*j ).
From Table 5, the following findings can be concluded. First, the
mean values of the overall efficiency scores are 0.617 and 0.554 for
2010 and 2011, respectively, which indicates that there is great po-
tential for improving the high-tech innovation efficiency. Effective
measures should be taken to encourage the high-tech industry in-
novation performance in China. Second, the average R&D efficiency is
larger than the commercialization efficiency in 2010 and 2011. Taking
the year 2010 for example, while maintaining whole process overall
efficiency, if we maximize the R&D sub-process achievable efficiency,
then the average efficiency scores of R&D and commercialization are
0.801 and 0.494, respectively, whereas if we maximize the commer-
cialization sub-process efficiency, the corresponding figures are 0.668
and 0.556, respectively. In other words, the R&D sub-process in China's
high-tech industry performs better than the commercialization sub-
process. Such results indicate that more emphasis should be put on
commercialization activities to stimulate the development of the high-
Table 6
Pearson correlation tests cross each efficiency.
* *
E10 1* E20
E20
E0* 0.635** 0.945** 0.868**
* 0.428*
E10
*
E20
* Indicates that correlation is significant at 0.05 level (2-tailed).
** Indicates that correlation is significant at 0.01 level (2-tailed).
tech industry. Third, the standard deviations of all the efficiencies are
bigger than 0.2, indicating that there are remarkable variations in the
efficiency scores across provinces both in 2010 and 2011. The inter-
regional differences in all the efficiency performances corroborate our
Section 3.1 conjecture about disparity.
For ease of illustration, we take the year 2011 as the research object
of this section, investigating the high-tech innovation efficiency. Since
the whole innovation process is composed of the two connected sub-
processes, it is technically efficient only when both sub-processes are
efficient. Table 5 shows that there is only a small percentage of regions
that were overall efficient in terms of the high-tech innovation process
in 2011: Beijing, Tianjin, and Guangdong. Maybe the good local
economy encouraged these regions to introduce and incorporate more
advanced technology in production, with the result that the innovation
inputs are used more effectively. In addition, there are some regions
with low overall efficiency, such as Jilin (0.338), Shaanxi (0.333),
Gansu (0.332), Heilongjiang (0.236), and Xinjiang (0.098) in 2011. In
the future, these low-innovation-efficiency regions could find out what
is holding them back by looking at the well-developed regions and
could learn from the experience of those regions to improve their own
performance in high-tech innovation. From the optimal proportions of
shared resources for the R&D stage, each region can choose its optimal
division of shared resources.
To illustrate the relationships among the overall efficiency and sub-
process efficiencies, the results of Pearson correlation tests are shown in
Table 6. The significant correlations indicate that the overall efficiency
2*
E10
performance is related to the two sub-processes. The correlation coef-
0.498**
ficient (E10 1*
* < E20 2*
; E10 * ) indicates that the overall efficiency is
< E20
0.469 more related to the commercialization efficiency than to the R&D ef-
ficiency. In other words, the main cause of inefficiency in the innova-
tion process is the commercialization stage. As to the correlation be-
tween the two sub-processes, the smaller value and weaker significance

9
X. Chen, et al.
Table 7
Four administrative areas.
Area Provinces
East Beijing Tianjin Hebei Shanghai Jiangsu Zhejiang Fujian Shandong
Guangdong Hainan
Center Shanxi Anhui Jiangxi Henan Hubei Hunan
West Inner Mongolia Guangxi Chongqing Sichuan Guizhou Yunnan
Shaanxi Gansu Ningxia Xinjiang
Northeast Liaoning Jilin Heilongjiang
indicates that the R&D and commercialization sub-processes are not
significantly related, that is, the efficiency of either one does not greatly
affect the other, which is consistent with the finding of (Chiu et al.,
2012). When we maximize the R&D sub-process achievable efficiency
while maintaining the whole process overall efficiency, 14 provinces
are efficient in the R&D stage and 4 provinces are efficient in the
commercialization stage. The inconsistent result indicates that just fo-
cusing on one sub-process is not enough to improve the overall effi-
ciency. It is practical to open the internal transformation processes and
investigate the sources of inefficiency in the innovation process. An-
other factor to consider is that there are 12 provinces with E10 2*
* = E10
and 20
E * = E 20 , namely Beijing, Tianjin, Jilin, Shanxi, Inner Mongolia,
1*
Guangdong, Hainan, Chongqing, Sichuan, Gansu, Ningxia, and Xin-
jiang. This result indicates that only these provinces obtain a unique
efficiency decomposition, which indicates the usefulness and justifica-
tion of the approach to efficiency decomposition used in this paper.
To analyze the efficiency of the high-tech innovation process from a
larger scale viewpoint, we divide the 29 regions into four adminis-
trative areas: eastern (East), central (Center), western (West), and
northeastern (Northeast). The constituent provinces of each area are
listed in Table 7. To clearly reflect the difference of the four areas, each
area's average efficiency is shown in the following Fig. 2. The average
overall efficiency in the East is the best at 0.752; behind are the Center
at 0.599, West at 0.401, and Northeast at 0.313. As to the overall ef-
ficiency decomposition, the East also has the highest efficiency in each
sub-process. That is, the eastern part of China plays a leading role in the
high-tech innovation process considering the R&D and commercializa-
tion factors. This conclusion certainly agrees with reality. The reason is
mainly a series of preferential policies attracting more skilled laborers
and qualified enterprises. Flexible innovation mechanisms and a strong
market economic role make enterprises in these eastern provinces be-
come innovators.
5. Discussion
The conceptual model and empirical results provide practitioners
and researchers with a better understanding of the innovation activities
in China's provincial high-tech industries. The implications for theore-
tical work and practice are now considered in this discussion.
Fig. 2. The efficiency of the eastern, central, western, and northeastern areas of
China in 2011.
10
Technovation 94–95 (2020) 102094
5.1. Implications to theory
First, this paper constructs a two-stage conceptual framework to
measure the high-tech industry innovation activities, using the in-
novation value chain theory. This is the first attempt to link the R&D
and commercialization activities with a solid theoretical foundation.
The model considers both “shared inputs” and “additional intermediate
inputs”, which is important because it goes a step further than previous
studies that have ignored shared resources and the way they are allo-
cated.
Second, an analysis of the difference between R&D and commer-
cialization efficiency shows that the provincial commercialization effi-
ciencies are relatively low. This result disagrees with some prior lit-
erature: Chiu et al. (2012) show that the average R&D efficiency was
higher than operation efficiency in 2004 and 2007, but the converse
was true in 2005 and 2006. That work does not calculate the overall
efficiency. Guan and Chen (2010) indicate that the commercial effi-
ciency performance was slightly better than the R&D efficiency on
average in 2002 and 2003. Wang et al. (2013) document that the
average R&D efficiency is larger than the operational efficiency for
high-tech firms in Taiwan. These mixed findings may due to different
databases, model specification, or science policies. In addition, our
finding implies that the generally lower innovation efficiency of the
high-tech industry may be attributed to the lower commercialization
efficiency, which indicates that a substantial number of R&D achieve-
ments are not transformed into economic value, as suggested by Chiu
et al. (2012). The findings remind us that stimulating R&D investment
may not be an optimal measure to promote innovation efficiency and
that it would be preferable to have more incentive policy instruments to
improve the conversion rate of R&D achievements.
Third, there is a regional character to China's high-tech industry,
which extends the regional innovation system perspective. This is
consistent with the argument that the disparity of regional environ-
ments within the same country may lead to prominent differences in
innovation activities between regions (Sternberg and Rocha, 2007). Not
all of the 29 studied regions have better R&D efficiency than com-
mercialization efficiency. Some provinces achieve better efficiency in
the commercialization process but cannot perform efficiently in R&D
process, examples being Jiangsu, Shandong, and Henan in 2011.
Therefore, administrators should adjust measures to local conditions as
emphasized in Guan and Chen (2010). From an area perspective, the
eastern part of China performs best among the four areas, so the other
three areas can learn from it to improve themselves.
5.2. Implications to practice
The current study has several practical implications. The two-stage
network DEA model allows the local governments to identify the
sources of inefficiency better and thus devote more effective effort to
improving the overall efficiency. The application of a weighted additive
strategy and the proposal of a deterministic efficiency decomposition
approach is pioneering in the value-chain DEA model, and this tech-
nique offers a beneficial guide to find optimal solutions. Moreover, this
analytical framework is generalizable and can be extended to other
contexts.
This paper evaluated and analyzed the innovation performance for
the high-tech industries of 29 regions in China, and the results show
that innovation is not simply a matter of input-output activities but also
a collaborative system between R&D and commercialization. Therefore,
an appropriate balance of efficiency between R&D and commerciali-
zation is critical for overall efficiency. Beyond this, during the in-
novation process, it is necessary to identify the effect of various factors
and find more efficient options. These findings provide guidelines for
improving high-tech innovation performance considering both sub-
process disparity and regional disparity.
X. Chen, et al.
5.3. Implications to policy
This study gives insight into the cross-region innovation process of
China's high-tech industry for policy makers. Based on this empirical
study, we offer guidelines for enhancing China's high-tech innovation
performance.
First, considering the systematic nature of the relationship between
R&D and commercialization, the policy makers should emphasize the
coordination of the R&D activities and industrial activities.
Additionally, based on the shared resource measurements, the optimal
proportions can provide guidance on resource allocation to the R&D
sub-process and commercialization sub-process. Attention should be
paid not only to the scale of R&D resources but also to the allocation
mechanism, which is beneficial to optimize the use of limited resources.
This suggestion echoes Lee et al. (2014)’s argument for encouraging the
use of the Technology Road mapping technique. From the optimal
proportions of shared resources for the R&D sub-process, we know how
each region can choose its optimal division of shared resources for the R
&D and commercialization sub-processes. Taking Beijing in 2011 as an
example, the optimal proportions of technology resources and fixed
assets for the R&D sub-process are 0.554 and 0.423, respectively. In
other words, the optimal proportions for commercialization sub-process
are 0.446 and 0.577.
Several other factors may contribute to the low conversion of R&D
achievements. “Sleeping patents” may be the major cause, that is, a
substantial share of patents are not used internally nor used in market
transactions. The focus on patent applications rather than patent usage
leads to poor-quality patents (i.e. a “patent bubble”), which deviates
from the market demand and cannot be successfully commercialized.
On the other hand, the lack of skilled professionals and corresponding
technical knowledge reserves may block the transformation of R&D
achievements into effective productivity. In this case, the top priority
for policy makers should be oriented toward the ultimate goal of in-
novation activities, rather than blindly providing tax policies and other
incentives to stimulate R&D investment. Therefore, patent valuation
institutions should be established to improve patent screening me-
chanisms and to promote reviews of granted patents. In addition, there
should be incentives for the transformation to an open innovation
model (Hu et al., 2015; Mortara and Minshall, 2011), including im-
plementing inside-out activities (Cassiman and Valentini, 2016), es-
tablishing technical trading platforms (e.g., the “IP Marketplace” in
Denmark and the “Innovation Market” in Germany), the setup of
crowdsourcing platforms (e.g., Odeskm, Elance, Freelancer), promoting
the development of technology intermediaries, enhancing research co-
operation between enterprises and universities, and so on.
Finally, the findings of this study clearly show that high-tech in-
novation activities differ widely across subnational regions, implying
that it is necessary for decision makers to formulate province-specific
policies. For example, provinces like Hebei, Shanxi, Guangxi, and
Hainan should facilitate the conversion of R&D achievements, while
provinces like Shandong and Jiangsu should focus on improving R&D
efficiency. Other provinces such as Xinjiang, Gansu, Chongqing, and
Yunnan should pay attention equally to the R&D process and the
commercialization process. It is important to analyze the impact of
factors influencing the different sub-processes. In addition, the central
government should assess local government efforts to make appropriate
policy decisions, considering the scientific development information.
6. Conclusions
As a major contributor to innovation-driven strategy, the develop-
ment of China's high-tech industry is becoming an important topic.
Unfortunately, most previous studies in this field have considered the
innovation process as a black box, disregarding the internal structure.
In order to respond to this theoretical and practical problem, the cur-
rent paper develops a conceptual framework for high-tech industry
11
Technovation 94–95 (2020) 102094
innovation process based on innovation value chain theory. A two-stage
DEA model with shared resources freely distributed among the stages is
applied to evaluate the innovation performance of 29 provincial-level
regions in mainland China for the years 2010 and 2011. The empirical
findings show that most of the 29 regions have low efficiency in the
commercialization sub-process compared to the R&D sub-process. To
complement previous publications, this model gives insight into the
appropriate proportion of shared resources in each sub-process, thereby
avoiding the waste and misuse of resources. In addition, this study
contributes to the regional studies literature by analyzing the high-tech
industry innovation activities from the perspective of the regional in-
novation system.
This research is not free of limitations and these offer several useful
directions for future study. First, it would be interesting to add some
characteristics of high-tech innovation into the framework. Future re-
search should consider other factors such as the fast life cycle and a
feedback loop from commercialization to the R&D stage. Second, owing
to the lack of statistics about subindustries, this study focuses on the
regional disparity for the high-tech industry, as does the research of
Guan and Chen (2010). Since there is always some degree of regional
and industrial diversity within a nation, it could be fruitful to consider
the regional industrial differences in future work. Third, following Chiu
et al. (2012), Guan and Chen (2010), and Wang et al. (2013), our focus
is on measuring the internal innovation process of the high-tech in-
dustry itself rather than finding concrete solutions to improve innova-
tion. In real life, there is imbalanced regional development in China due
to aspects such as China's differentiation development strategy, the
unbalanced regional economies, and the disparities in industrial
structure, all of which affect possible strategies. The quest for concrete
solutions for performance improvement deserves more research. The
current paper's province-specific recommendations need to be in-
vestigated using provincial-level analysis, taking into account the
characteristics of R&D and commercialization activities, the regional
policy, and so on. Fourth, the collaborative system between R&D and
commercialization cries out for more research. Finally, a longer time
horizon in the empirical study would help capture a more dynamic
picture of the high-tech industry development in China, which might
provide interesting results.
Acknowledgments
The authors gratefully acknowledge the valuable comments and
suggestions of the three anonymous reviewers and Editor-in-Chief Prof.
Linton. This research is funded by the National Natural Science
Foundation of China (Project no. 71472172) and the Social Science and
Humanity Foundation of Ministry of Education of China (Project no.
14YJA630035).
References
Aigner, D., Lovell, C.K., Schmidt, P., 1977. Formulation and estimation of stochastic
frontier production function models. J. Econ. 6, 21–37.
Amirteimoori, A., 2013. A DEA two-stage decision processes with shared resources. Cent.
Eur. J. Oper. Res. 21, 141–151.
Battese, G., Corra, G., 1977. Estimation of a production frontier model: with a generalized
frontier production function and panel data. Aust. J. Agric. Econ. 21, 169–179.
Bernstein, B., Singh, P.J., 2006. An integrated innovation process model based on prac-
tices of Australian biotechnology firms. Technovation 26, 561–572.
Bronzini, R., Piselli, P., 2016. The impact of R&D subsidies on firm innovation. Res. Policy
45, 442–457.
Buesa, M., Heijs, J., Baumert, T., 2010. The determinants of regional innovation in
Europe: a combined factorial and regression knowledge production function ap-
proach. Res. Policy 39, 722–735.
Buesa, M., Heijs, J., Pellitero, M.M., Baumert, T., 2006. Regional systems of innovation
and the knowledge production function: the Spanish case. Technovation 26,
463–472.
Cappelen, Å., Raknerud, A., Rybalka, M., 2012. The effects of R&D tax credits on pa-
tenting and innovations. Res. Policy 41, 334–345.
Cassiman, B., Valentini, G., 2016. Open innovation: are inbound and outbound knowl-
edge flows really complementary? Strateg. Manag. J. 37, 1034–1046.
X. Chen, et al.
Charnes, A., Cooper, W.W., Rhodes, E., 1978. Measuring the efficiency of decision making
units. Eur. J. Oper. Res. 2, 429–444.
Chen, C., Zhu, J., Yu, J.-Y., Noori, H., 2012. A new methodology for evaluating sus-
tainable product design performance with two-stage network data envelopment
analysis. Eur. J. Oper. Res. 221, 348–359.
Chen, K.H., Guan, J.C., 2012. Measuring the efficiency of China's regional innovation
systems: application of network Data Envelopment Analysis (DEA). Reg. Stud. 46,
355–377.
Chen, Y., Cook, W.D., Li, N., Zhu, J., 2009. Additive efficiency decomposition in two-stage
DEA. Eur. J. Oper. Res. 196, 1170–1176.
Chen, Y., Du, J., Sherman, H.D., Zhu, J., 2010. DEA model with shared resources and
efficiency decomposition. Eur. J. Oper. Res. 207, 339–349.
Chen, Y., Liang, L., Yang, F., Zhu, J., 2006. Evaluation of information technology in-
vestment: a data envelopment analysis approach. Comput. Oper. Res. 33, 1368–1379.
Chen, Y., Zhu, J., 2004. Measuring information technology's indirect impact on firm
performance. Inf. Technol. Manag. 5, 9–22.
Chiu, Y.-H., Huang, C.-W., Chen, Y.-C., 2012. The R&D value-chain efficiency measure-
ment for high-tech industries in China. Asia Pac. J. Manag. 29, 989–1006.
Cook, W.D., Hababou, M., 2001. Sales performance measurement in bank branches.
Omega 29, 299–307.
Cook, W.D., Hababou, M., Tuenter, H.J., 2000. Multicomponent efficiency measurement
and shared inputs in data envelopment analysis: an application to sales and service
performance in bank branches. J. Product. Anal. 14, 209–224.
Cook, W.D., Zhu, J., Bi, G., Yang, F., 2010. Network DEA: additive efficiency decom-
position. Eur. J. Oper. Res. 207, 1122–1129.
Cooke, P., 1992. Regional innovation systems: competitive regulation in the new Europe.
Geoforum 23, 365–382.
Cruz-Cázares, C., Bayona-Sáez, C., García-Marco, T., 2013. You can’t manage right what
you can’t measure well: technological innovation efficiency. Res. Policy 42,
1239–1250.
Czarnitzki, D., Hottenrott, H., Thorwarth, S., 2011. Industrial research versus develop-
ment investment: the implications of financial constraints. Camb. J. Econ. 35,
527–544.
Feldman, M.P., Florida, R., 1994. The geographic sources of innovation: technological
infrastructure and product innovation in the United States. Ann. Assoc. Am. Geogr.
84, 210–229.
Fritsch, M., 2002. Measuring the quality of regional innovation systems: a knowledge
production function approach. Int. Reg. Sci. Rev. 25, 86–101.
Fu, F.-C., Vijverberg, C.-P.C., Chen, Y.-S., 2007. Productivity and efficiency of state-
owned enterprises in China. J. Product. Anal. 29, 249–259.
Goldsmith, R.W., 1951. A Perpetual Inventory of National Wealth, Studies in Income and
Wealth 14. NBER, pp. 5–74.
Griliches, Z., 1979. Issues in assessing the contribution of research and development to
productivity growth. Bell J. Econ. 92–116.
Griliches, Z., 1990. Patent Statistics as Economic Indicators; A Survey. J. Econ. 28,
1661–1797.
Guan, J., Chen, K., 2010. Measuring the innovation production process: a cross-region
empirical study of China's high-tech innovations. Technovation 30, 348–358.
Hage, J., Hollingsworth, J.R., 2000. A strategy for the analysis of idea innovation net-
works and institutions. Organ. Stud. 21, 971–1004.
Halkos, G.E., Tzeremes, N.G., Kourtzidis, S.A., 2014. A unified classification of two-stage
DEA models. Surv. Oper. Res. Manag. Sci. 19, 1–16.
Halkos, G.E., Tzeremes, N.G., Kourtzidis, S.A., 2015. Regional sustainability efficiency
index in Europe: an additive two-stage DEA approach. Oper. Res. 15, 1–23.
Hall, R.E., Jones, C.I., 1999. Why Do Some Countries Produce So Much More Output Per
Worker than Others? Q. J. Econ. 114 (1), 83–116.
Hansen, M.T., Birkinshaw, J., 2007. The innovation value chain. Harv. Bus. Rev. 85, 121.
Hong, J., Feng, B., Wu, Y., Wang, L., 2016. Do government grants promote innovation
efficiency in China's high-tech industries? Technovation 57, 4–13.
Hong, J., Hong, S., Wang, L., Xu, Y., Zhao, D., 2015. Government grants, private R&D
funding and innovation efficiency in transition economy. Technol. Anal. Strateg.
Manag. 27, 1068–1096.
Hu, A.G., 2001. Ownership, government R&D, private R&D, and productivity in Chinese
industry. J. Comp. Econ. 29, 136–157.
Hu, M.-C., Mathews, J.A., 2008. China's national innovative capacity. Res. Policy 37,
1465–1479.
Hu, Y., McNamara, P., McLoughlin, D., 2015. Outbound open innovation in bio-phar-
maceutical out-licensing. Technovation 35, 46–58.
Hung, S.-W., Wang, A.-P., 2012. Entrepreneurs with glamour? DEA performance char-
acterization of high-tech and older-established industries. Econ. Model. 29,
1146–1153.
Kao, C., 2009. Efficiency decomposition in network data envelopment analysis: a rela-
tional model. Eur. J. Oper. Res. 192, 949–962.
12
Technovation 94–95 (2020) 102094
Kao, C., 2014. Network data envelopment analysis: a review. Eur. J. Oper. Res. 239, 1–16.
Kao, C., Hwang, S.-N., 2008. Efficiency decomposition in two-stage data envelopment
analysis: an application to non-life insurance companies in Taiwan. Eur. J. Oper. Res.
185, 418–429.
Lee, C.-Y., Wu, H.-L., Pao, H.-W., 2014. How does R&D intensity influence firm ex-
plorativeness? Evidence of R&D active firms in four advanced countries.
Technovation 34, 582–593.
Lev, B., 2000. Intangibles: Management, Measurement, and Reporting. Brookings
Institution Press.
Li, X., 2009. China's regional innovation capacity in transition: an empirical approach.
Res. Policy 38, 338–357.
Li, Y., Tellis, G.J., 2016. Is China uniform? Intra-country differences in the takeoff of new
products. Technovation 47, 1–13.
Liang, L., Cook, W.D., Zhu, J., 2008. DEA models for two-stage processes: game approach
and efficiency decomposition. Nav. Res. Logist. 55, 643–653.
Liang, L., Yang, F., Cook, W.D., Zhu, J., 2006. DEA models for supply chain efficiency
evaluation. Ann. Oper. Res. 145, 35–49.
Lin, B.-W., Lee, Y., Hung, S.-C., 2006. R&D intensity and commercialization orientation
effects on financial performance. J. Bus. Res. 59, 679–685.
Lin, D., Liang, Q., Xu, Z., Li, R., Xie, W., 2008. Does knowledge management matter for
information technology applications in China? Asia Pacific. J. Manag. 25, 489–507.
Liu, J.S., Lu, L.Y.Y., Lu, W.-M., Lin, B.J.Y., 2013. Data envelopment analysis 1978–2010: a
citation-based literature survey. Omega 41, 3–15.
Maietta, O.W., 2015. Determinants of university–firm R&D collaboration and its impact
on innovation: a perspective from a low-tech industry. Res. Policy 44, 1341–1359.
Meeusen, W., Van den Broeck, J., 1977. Efficiency estimation from Cobb-Douglas pro-
duction functions with composed error. Int. Econ. Rev. 435–444.
Mortara, L., Minshall, T., 2011. How do large multinational companies implement open
innovation? Technovation 31, 586–597.
Porter, M.E., 2008. Competitive Strategy: Techniques for Analyzing Industries and
Competitors. Simon and Schuster.
Porter, M.E., Millar, V.E., 1985. How information gives you competitive advantage.
Harvard Business Review, Reprint Service.
Roper, S., Arvanitis, S., 2012. From knowledge to added value: a comparative, panel-data
analysis of the innovation value chain in Irish and Swiss manufacturing firms. Res.
Policy 41, 1093–1106.
Roper, S., Du, J., Love, J.H., 2008. Modelling the innovation value chain. Res. Policy 37,
961–977.
Rothwell, R., 1994. Towards the fifth-generation innovation process. Int. Mark. Rev. 11,
7–31.
Schumpeter, J.A., 1934. The Theory of Economic Development: an Inquiry Into Profits,
Capital, Credit, Interest, and the Business Cycle. Transaction Publishers.
Shefer, D., Frenkel, A., 2005. R&D, firm size and innovation: an empirical analysis.
Technovation 25, 25–32.
Spanos, Y.E., Vonortas, N.S., Voudouris, I., 2015. Antecedents of innovation impacts in
publicly funded collaborative R&D projects. Technovation 36, 53–64.
Sternberg, R., Rocha, H.O., 2007. Why entrepreneurship is a regional event: Theoretical
arguments, empirical evidence, and policy consequences. Entrep.: Engine Growth 3,
215–238.
Tseng, F.-M., Chiu, Y.-J., Chen, J.-S., 2009. Measuring business performance in the high-
tech manufacturing industry: a case study of Taiwan's large-sized TFT-LCD panel
companies. Omega 37, 686–697.
Wang, C.-H., Lu, Y.-H., Huang, C.-W., Lee, J.-Y., 2013. R&D, productivity, and market
value: an empirical study from high-technology firms. Omega 41, 143–155.
Wang, S., Fan, J., Zhao, D., Wang, S., 2015. Regional innovation environment and in-
novation efficiency: the Chinese case. Technol. Anal. Strateg. Manag. 1–15.
Wu, J., Xiong, B., An, Q., Sun, J., Wu, H., 2015. Total-factor energy efficiency evaluation
of Chinese industry by using two-stage DEA model with shared inputs. Ann. Oper.
Res. 1–20.
Wu, J., Zhu, Q., Chu, J., Liu, H., Liang, L., 2016. Measuring energy and environmental
efficiency of transportation systems in China based on a parallel DEA approach.
Transp. Res. Part D: Transp. Environ. 48, 460–472.
Yang, J., Liu, C.-Y., 2006. New product development: an innovation diffusion perspective.
J. High. Technol. Manag. Res. 17, 17–26.
Zha, Y., Liang, L., 2010. Two-stage cooperation model with input freely distributed
among the stages. Eur. J. Oper. Res. 205, 332–338.
Zhang, A., Zhang, Y., Zhao, R., 2003. A study of the R&D efficiency and productivity of
Chinese firms. J. Comp. Econ. 31, 444–464.
Zhang, Y., Lv, X., 2012. Innovation performance of Chinese high-tech firms and its de-
terminants: an empirical study based on quantile regression. In: Technology
Management for Emerging Technologies (PICMET), 2012 Proceedings of PICMET'12:.
IEEE, pp. 2125–2131.