Applied Energy 187 (2017) 489–500

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Economic growth model, structural transformation, and green

productivity in China
Ke Li a,b,c, Boqiang Lin b,⇑
a
College of Mathematics & Computer Science, Hunan Normal University, Changsha, Hunan 410081, PR China
b
Collaborative Innovation Center for Energy Economics and Energy Policy, China Institute for Studies in Energy Policy, Xiamen University, Xiamen, Fujian 361005, PR China
c
Center for Electric Pricing Research, Changsha University of Science and Technology, Changsha, Hunan 410076, PR China

h i g h l i g h t s

Two total factor productivities of China are estimated by super-DEA models.

Investment-driven growth model produces the negative effect on green productivity.

Various structural transformation have different impacts on green productivity.

a r t i c l e i n f o a b s t r a c t

Article history: This study investigates the impacts of investment-driven economic growth model, as well as rationaliza-
Received 19 September 2016 tion and upgrading of the industrial structure on green productivity in 30 Chinese provinces over the per-
Received in revised form 14 November 2016 iod 1997–2010. Two total factor productivities (TFP), namely energy adjusted TFP and energy and carbon
Accepted 18 November 2016
dioxide emissions adjusted TFP (denoted as TFEE and TFCE respectively), are estimated using super-
efficiency DEA models, and used as indices to reflect green productivity performance in China. The main
results of the empirical study are as follow: (1) China’s economic growth model does not improve both
Keywords:
TFEE and TFCE; (2) the flow of laborers from the primary, secondary, and tertiary industries helps to
Economic growth model
Rationalize of industrial structure
improve TFEE and TFCE, while capital transformation does not produce the same effect; (3) the structural
Upgrading of industrial structure changes in the manufacturing industry produce negative and positive effects on TFEE and TFCE
Total factor productivity respectively.
Energy efficiency with carbon dioxide Ó 2016 Elsevier Ltd. All rights reserved.
emissions

1. Introduction
It is well known that China’s economic growth is mainly driven
by huge investment, with investment growth is higher than output
growth, and contributing to economic growth more than consump-
tion and net exports. Correspondingly, China’s industrial structure
is dominated by the industrial sector, particularly energy-intensive
industries. According the National Bureau of Statistic of China,
fixed asset investment, total industry output, and heavy industry
output of the country increased by 18.3%, 10.8%, and 22.7% respec-
tively over the period 1997–2010. These exceeded the annual
growth rate of GDP (9.9%), agriculture output (3.9%), services out-
put (10.7%) and light industry output (17.7%). These figures show
that the heavy industry is a key player in the industrial sector. In
⇑ Corresponding author.
E-mail addresses: likekent1208@163.com (K. Li),
bqlin2004@vip.sina.com (B. Lin).
1997, heavy industry output accounted for 51.0% of the total indus-
trial output, which increased to 71.4% in 2010.
The concepts of the ‘Low-Carbon Economy’, or ‘Green Economic
Growth’ has received increasing policy and media attention in
recent years [1]. Enhancing green productivity is an important
way of achieving ‘Green Economic Growth’. However, the
investment-driven economic growth model and the energy-
intensive industrial structure of China have increased energy con-
sumption and deteriorated the environmental resources of the
country, thereby undermining green productivity [2–5]. Because
of rapid increase in energy consumption, China’s CO2 emissions
became the highest in the world in 2007 [6]. The need for green
development has propelled the government to switch from an
extensive economic growth model to an intensive one, and to
speed up the rate of industrial structural transformation [7]. This
study aims to investigate how the investment-driven economic
growth model and structural transformation contribute to the
bqlin@xmu.edu.cn, enhancement of green productivity. It also seeks to determine
how a country can switch from an economic growth model to

http://dx.doi.org/10.1016/j.apenergy.2016.11.075
0306-2619/Ó 2016 Elsevier Ltd. All rights reserved.
490 K. Li, B. Lin / Applied Energy 187 (2017) 489–500
another, and how the industrial structure can be adjusted to
enhance green development.
Energy conservation and emissions abatement is an inherent
requirement for green development. Enhancing total factor pro-
ductivity (TFP), especially the energy and carbon dioxide emissions
adjusted TFP, is consistent with the connotation of energy conser-
vation and emissions abatement. In this sense, improving TFP
while considering energy input and carbon dioxide emissions
reflects green development.
The effect of investment-driven economic growth model on
green productivity is among the critical economic issues faced by
China. Enhancing energy efficiency is one of the aspects of green
productivity improvement. The investment-driven economic
growth model in China reduces economic value added (EVA) rate
and raises energy consumption growth rate. This makes the
growth model unfavorable for the enhancement of energy effi-
ciency [8]. In 1995, the Chinese government claimed that the eco-
nomic growth model of the country must be gradually switched
from extensive to intensive. Cai et al. [9] believed that low per cap-
ita income and rapid industrialization and urbanization may hin-
der the necessary transformation, and undermine energy
conservation and CO2 emission reduction. Lin and Su [10] argued
that long-term interest and cost of energy should be maintained
at low levels to benefit investment-driven economic growth, which
would subsequently increase energy consumption and reduce
energy efficiency. Chen and Golley [11] believed that much invest-
ment and energy consumption was undertaken by large, inefficient
state-owned enterprises, and thereby undermined green TFP. Pre-
vious studies on investment-driven economic growth model are
merely qualitative and conceptual, and neglect its effect on energy
and carbon dioxide emissions adjusted TFP or green productivity,
which presents a research gap that this study aims to fill.
Energy efficiency is influenced by industrial structure [12–15].
Adjusting the industrial structure directly affects energy efficiency
because the input factors, including the energy consumed by vari-
ous industrial sectors, are reconfigured in the process. Numerous
studies that apply the index decomposition method (IDM) have
examined the effects of industrial structural changes on energy
efficiency [16]. Numerous studies found that during structural
changes, the main energy-consuming sectors were upgraded from
low-efficiency to high-efficiency sectors, and their energy effi-
ciency was improved as a ‘‘structural bonus” [17–22]. However,
Lin and Su [10] argued that structural transformation in China, par-
ticularly in the manufacturing sector, might hinder or even dimin-
ish the enhancement of energy efficiency in the long term. As
energy price and capital cost are maintained at low levels to pro-
mote extensive economic growth, various resources, including
energy, iron, steel, nonferrous metals, and petrochemicals, are allo-
cated to industries that consume large amounts of energy and emit
large amounts of CO2 [14].
Structural changes in the industrial sector can influence the
input–output efficiency of industries. Timmer and Szirmai [23]
found that such structural changes could affect economic growth
rate and energy efficiency in the long run. Fan et al. [24] empiri-
cally discovered that structural changes could improve economic
growth rate and significantly affect energy efficiency. Zheng,
et al. [25] found that structural transformation had direct and indi-
rect effects on economic growth and energy efficiency respectively.
Zhou et al. [26] found industrial structural adjustments can effec-
tively reduce current CO2 emissions.
Although numerous studies have proven that industrial struc-
tural changes can affect the productivity of industries, they failed
to describe how such changes affect green productivity through
technology spillovers. Fagerberg [27] argued that a flexible indus-
trial structure could promote technology spillover among different
industries, which could improve their productivities. Ark and Tim-
mer [28] and Oulton and Srinivasan [29] found that by improving
their level of coordination with other industries, the effects of tech-
nology spillover from the information and communication indus-
tries could be extended to other industries, thereby enhancing
their productivities. Isaksson [30] noted that structural changes
usually improve the effects of reallocation, which could also
increase productivity. By adopting an input–occupancy–output
factor analysis model, Chai et al. [31] revealed that structural
changes in the industrial sector could indirectly affect energy effi-
ciency through technological spillovers. Yang et al. [32] found that
indigenous R&D and interregional R&D spillovers can decrease CO2
intensity, but this effect depends on local R&D expenditure. Boyd
and Curtis [33] found most management techniques had beneficial
spillovers on energy efficiency, but an emphasis on generic targets
would result in decline in energy efficiency.
Schäfer [13] and Hofman and Labar [34] investigated the
changes in the ratio of the primary, secondary, and tertiary indus-
tries to GDP, and Zhan et al. [35] investigated the structural
changes in the manufacturing sectors. They proved that structural
transformation could significantly affect energy efficiency in China.
However, these studies did not completely cover the influence of
structural changes on energy efficiency. Since the 1990s, the struc-
tural changes in China mainly occurred among three of its sectors
and manufacturing, particularly from light to heavy industries.
Lü and Zhou [36] believed that structural changes could affect
productivity through the optimization and upgrading of industries.
Optimization affects productivity through allocation efficiency,
which stems from the flow and transfer of laborers, capital, and
other factors among industrial sectors. Upgrading influences pro-
ductivity through the development of high-degree and high-
value-added industries. An eccentric industrial structure distorts
the configuration of resources among industries, which decreases
the ‘‘structural bonus” of productivity and may even produce a
‘‘structural burden”. Moreover, such a structure may be detrimen-
tal to the technology spillover among different industries, which
diminishes the effects of technological progress on productivity.
This study aims to quantitatively analyze the effects of
investment-driven economic growth model and energy-intensive
industrial structure on green productivity using China as a case
study. A simple but novel index is adopted to measure
investment-driven economic growth model, which reflects the
contribution of investment to economic growth. For structural
changes, two aspects of structural transformation in China are con-
sidered. The first is the flow and transfer of inputs among sectors
which reflects China’s switch from an agricultural country to an
industrial country. The second is the structural changes among
manufacturing sectors which reveals the transformation and
upgrading of industrial structure in China in recent years. This
study distinguishes itself from previous literature by employing
an econometric model or a quantitative analysis rather than a qual-
itative analysis to investigate the effects of investment-driven eco-
nomic growth model and structural changes on green development
in China.
2. Material and methods
This paper first provides the definitions of green productivity,
investment-driven economic growth and structural changes, and
then specifies the model.
2.1. Variables and data source
2.1.1. Green productivity
There are different ways to measure green productivity. In the
current study, we measure it by two indices of TFP—the energy
 K. Li, B. Lin / Applied Energy 187 (2017) 489–500
adjusted TFP and the energy and carbon dioxide emissions
adjusted TFP. For the former, total-factor energy efficiency (TFEE),
which was introduced by Hu and Wang [37] and constructed as the
ratio of the target energy input that is suggested from data envel-
opment analysis (DEA) to the actual energy inputs, is a suitable
index [38]. In recent years, reducing CO2 emissions has attracted
wide attention and became a new ambition for the Chinese govern-
ment. Thus, this study uses input-oriented and constant returns to
scale (CRS)1 to calculate the TFEE while considering CO2 emissions.
This is used to measure the energy and carbon dioxide emissions
adjusted TFP.
DEA can be easily applied to a multiple input-output framework
to compute the index of TFEE. Most studies however fail to account
for one important drawback of DEA: it does not allow for a ranking
of the efficient units themselves. In order to overcome it, we apply
the super-efficiency DEA model. In this paper, the super-efficiency
DEA model assumes labor, capital stock, and energy as inputs, the
GDP of each province as the desirable output, and the CO2 emission
as the undesirable output. This model uses the following variables:
1. Labor (L). The quality of labor significantly affects the output of
the firm or the entire sector. This study defines L based on the
years of education of the laborers and is measured by millions
of persons, and calculated as follows,
lit  Eit
Lit ¼ ð1Þ
Et
where lit is the quantity of employment, Eit is the average years of
education of lit for province i in year t, Et is the average years of edu-
cation of the laborers in China for year t. The data are collected from
China Labor Statistical Yearbook.
2. Capital Stock (K). Since the 1990s, technological innovations in
China have been mainly reflected in pieces of equipments, par-
ticularly advanced information equipments [39]. The capital
stock of China, especially that of the equipment production sec-
tor, is instantaneously renewed. Therefore, the average service
life of these pieces of equipments is short, which shows the cap-
ital allotted for technological progress [40,41]. The perpetual
inventory method which is often used in previous literatures
to calculate capital stock cannot reflect the technological pro-
gress embedded in capital stocks. Therefore, K is computed in
this study as follows (measured by 100 million yuan):
kit  Ait
K it ¼ ð2Þ
At
where kit is the capital stock computed by the perpetual inventory
method”,2 Ait is the average age of the capital stock in year t for pro-
vince i, which is computed based on the method of Limam and Miller
[42]; At is the average service life of the capital stock of China in year
t, which is calculated by weighing the proportion of the capital stock
in year t for province i on the capital stock of China in the same year.
All data are collected from the Data of GDP of China (1952–1995 and
1952–2004) and the China Statistical Yearbook.
1
Although the variable returns to scale (VRS) might be better than CRS in the
Chinese context, many results calculated by VRS in this study were infeasible.
2
The perpetual inventory method is computed as follows:
kit ¼ Iit þ ð1  di Þ  ki;t1 , where Iit is the gross fixed capital formation and the
constant 2000 prices. The price index for investment in fixed assets (constant 2000
prices, 2000 = 1) is calculated by using the method of Shan [43]. Each province is
assumed to have different depreciation rates di of gross fixed capital formation, which
is calculated by Wu [44]. 1978 is set as the initial year and the capital stock is
calculated by dividing the gross fixed capital formation in 1978 (constant 2000 price)
by di .
491
3. Energy consumption (E). The primary energy consumption of
each province is taken as the energy input, which is measured
by 104 tce. Energy consumption data are collected from China
Energy Statistical Yearbook and from China Compendium of Statis-
tics, 1949–2008.
4. Gross regional product (GRP). GRP is measured by 108 yuan (in
constant 2000 prices). Data on GRP are collected from the data-
base of the National Bureau of Statistics of China.
5. CO2 emissions (C). The Chinese government does not release
official CO2 emissions data. Many institutions such as the Car-
bon Dioxide Information Analysis Center (CDIAC) and Interna-
tional Energy Agency (IEA) provide estimates of this data on
China. But they could not estimate CO2 emissions for Chinese
provinces. So researchers usually use the reference method
introduced by the Intergovernmental Panel on Climate Change
(IPCC) to calculate CO2 emissions [45,46]. Based on this
method, the data on CO2 emissions for each province are esti-
mated from the breakdown of energy consumption by each fuel
category, and it is measured in million tons. The formula for the
calculation of CO2 emissions is as follows:
X
8 X
C¼ CO2g ¼ Eg  CEF g  COF g  ð44=12Þ ð3Þ
g¼1 g
where CO2g are CO2 emissions from energy source g (including coal,
coke, crude oil, gasoline, kerosene, diesel oil, fuel oil, natural gas).
CO2 emissions depend on the consumption of the gth energy (Eg)
as well as CO2 emissions factors for each energy source (CEFg). COFg
represents the carbon oxidisation factors. The ratio 44/12 is the
molecular weight ratio of carbon dioxide to carbon. Data on CEFg
and COFg are from IPCC guidelines for national greenhouse gas
inventories[47]. Although these factors may not be accurate for
China [48], formula (3) is an alternative method and widely used
in the literature. Each fuel consumption data is collected from China
Energy Statistical Yearbook.
The super-efficiency DEA model estimates the total factor
energy efficiency (TFEE) using the aforementioned variables, disre-
garding CO2 emissions. Afterwards, this model constructs an envi-
ronmental production function [49] and incorporates the good
output (GRP) and the bad output (C).3 We adopt the environmental
directional distance function to measure the total factor energy effi-
ciency, considering the bad output (TFCE). This is calculated by the
following Eqs. (4) and (5):
D! o ðxtk ; GRP tk ; C tk ; g tGRP;k ; g tC;k Þ ¼ max½b : ðGRPtk þ bg tGRP;k ; C tk  bg tC;k Þ 2 Pðxtk Þ
8 P30
t t
> j¼1;j–k zj  GRP j P ð1 þ bÞ  GRP k
t
>
>
> P30
>
< t t
j¼1;j–k zj  C j ¼ ð1  bÞ  C k
t
s:t: P30
>
> j¼1;j–k ztj  xtj 6 xtk
>
>
>
:
x ¼ ðL; K; EÞ 2 Rþ3 ; ztj P 0; 8j
ð4Þ
!
TFCEtk ¼ 1=½1 þ Do ðxtk ; GRPtk ; C tk ; g tGRP;k ; g tC;k Þ ð5Þ
Fig. 1 shows the mean of TFEE and TFCE of the 30 provinces.
Although TFEE and TFCE tend to rise, both are far from the efficient
level. Moreover, TFCE is always higher than TFEE, which implies that
the potential output of TFCE is smaller than that of TFEE under the
current technologies. One possible reason is that the reduction of
CO2 emissions enhances the performance of development, i.e., the
benefits of emissions abatement to green development.
3
The environmental production function should satisfy the following assumptions:
the output set is compact, the inputs and the desirable output are freely disposable,
and the undesirable output is weakly disposable. Färe et al. (2005) and Färe et al.
(2007) explained each of these assumptions.
492 K. Li, B. Lin / Applied Energy 187 (2017) 489–500

0.78

0.76

0.74

0.72

0.70

0.68

0.66

0.64

0.62

0.60
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

TFEE TFCE

Fig. 1. China’s energy efficiency during 1997–2010.

2.1.2. The contribution of investment to economic growth 1.20

In China, capital formation contributes to economic growth
1.00
more than the final consumption expenditure or net exports of
goods and services. However, because of unavailability of adequate 0.80
data, it is impossible to estimate real gross capital formation for the 0.60
provinces. Gross capital formation includes the gross fixed capital
0.40
formation and change in inventories, and is dominated by the for-
mer. Therefore, this study uses the contribution of gross fixed cap- 0.20
ital formation to the growth of GDP to measure the contribution of
0.00
investment to economic growth. Similar to the share of the contri- 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
bution of gross capital formation to the growth of GDP, which is
gff
defined by the China Statistical Yearbook, we define the contribu-
tion of investment to economic growth (gff) as follows: Fig. 2. The contribution of investment to economic growth of China during
1997–2010.
gffit ¼ ðDIit =DGRP Eit Þ  100% ð6Þ

where DIit is the fixed capital formation (constant 2000 prices),

DGRP Eit is the increase in GRP by expenditure approach (constant
2000 prices), i and t are the province and time respectively. A higher
gff means that investment or capital formation contributes more to
economic growth, indicating investment-driven economic growth.
Fig. 2 shows the average value of gff from 1997 to 2010, which
has been maintained at a high level, even reaching as high as
113.2% in 2009. Based on the empirical results, this paper shows
that high gff will be disadvantageous to energy efficiency (TFEE)
improvement, especially to green development (TFCE).
2.1.3. Structural transformation
These data pertain to the resource allocation efficiency in the
primary, secondary, and tertiary industries, and the structural
changes in the three manufacturing categories (i.e., first, second,
and third),4 which describe the optimization and upgrading of the
industrial structure respectively.
2.1.3.1. Allocation efficiency. Structural transformation is mainly
manifested by the flow and transfer of labor and capital between
4
According to [50], manufacturing industries could be divided into three
categories. As the industrialization process advances, the manufacturing industry
gradually transfers to the third category, i.e., high-end manufacturing. Basing from
the manufacturing classification standards of China, Wang & Chen [51] claimed that
the first manufacturing category comprised the food, beverage, tobacco, textile,
leather, wood, furniture, paper, printing, and culture manufacturing sectors; the
second category comprised the petroleum, raw chemical material, medicine, chemical
fiber, rubber, plastic, non-metal, and metal manufacturing sectors; and the third
category comprised the machinery, transport equipment, electrical machinery and
equipment, communication equipment, and measuring instrument manufacturing
sectors.
industries, which also serve as the foundations for the production
of a ‘‘structural bonus” on energy efficiency. The allocation effi-
ciency is similar to the mean of optimization. According to the
shift–share analysis of Timmer and Szirmai [23], the growth of
labor or capital productivity can be expressed as:
P3 P3
XP T  XP 0
T
j¼1 ðXP j  XPT0 ÞS0j T
j¼1 ðSj  ST0 ÞXP 0j
0
¼ 0
þ
XP XP XP 0
P3 T T T T
j¼1 ðSj  S0 ÞðXP j  XP 0 Þ
þ ð7Þ
XP 0
In XPT ¼ Y T =X T , Y is the GRP, X is the labor or capital stock, and Sj is
the share of industry j (j = 1, 2, 3 and represent the primary indus-
try, secondary industry, and tertiary industry, respectively) inputs
(either labor or capital) in the total inputs of all three industry cat-
egories. The superscripts 0 and T denote the beginning and end of
the phase period respectively. The second and the third terms on
the right side of formula (4) reflect the effects of the industrial
structure on the productivity of the whole industry (i.e., labor or
capital allocation efficiency, which are denoted as labef and capef,
respectively). In Fig. 3, labef changes from negative to positive and
increases yearly, whereas capef is always negative. Therefore, the
allocation efficiency of labor may be conducive for the enhance-
ment of energy efficiency, whereas the allocation efficiency of cap-
ital may be unfavorable for the enhancement of energy efficiency.
2.1.3.2. Structural transformation in manufacturing. Industrial struc-
tural transformation, particularly the structural changes in the
manufacturing sector, must be given further attention when ana-
lyzing structural transformation in China. According to the
 K. Li, B. Lin / Applied Energy 187 (2017) 489–500
0.5
0.4
0.3
0.2
0.1
0 for the intervention level of government in the industrial sector. It
-0.1
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
capef labef ois
Fig. 3. Structural transformation of China during 1997–2010.
structural change index [52], structural transformation in the man-
ufacturing sector (ois) is computed as follows:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi!
X
3 X3 X3
oisit ¼ arccos W ijt W ij0 = j¼1
W 2ijt  j¼1
W 2ij0
j¼1
where W ijt is the share of output of the jth industry in the year t for
province i in the output value of the entire manufacturing sector.
Year 1997 is represented as t = 0. Fig. 3 shows that the value of
ois increases yearly. China’s manufacturing sector gradually shifts
from the first category to the third category, representing an
upward trend that reflects the level of structural upgrading. This
trend produces two effects on energy efficiency. First, the main
manufacturing sectors turn from light industries into heavy indus-
tries, which is unfavorable for the improvement of energy effi-
ciency. Second, the further expansion of energy-intensive
industries can be hindered by upgrading the level of manufacturing,
which can reduce CO2 emissions and subsequently improve TFCE.
2.2. Empirical model
To analyze the effects of the contribution of investment to eco-
nomic growth and industrial structure transformation on green
productivity, we propose the following empirical model for the
panel data:
EEit ¼ c þ a1 gffit þ a2 labefit þ a3 capefit þ a4 oisit þ X 0it b þ t
þ li þ eit
where EEit is the two indices of TFP (i.e., TFEEit or TFCEit), c is con-
stant term, Xit is the control variables. Given that TFEEit and TFCEit
are calculated by the super-efficiency DEA model, these variables
have unlimited range, which allows the use of standard panel
regression methods rather than the truncated data regression
[53]. t is a time trend variable that controls time-varying factors,
such as the improvement of energy-saving technology and the
adjustment of environment policy; li is the individual effect, which
reflects the differences between provinces in terms of their environ-
mental regulations and energy resource benefits; and eit is a random
term.
Four control variables, namely, urbanization (urban), foreign
direct investment (fdi), energy price (price), and ownership struc-
ture (ps), are added into the model. These variables are measured
by the proportion of non-agricultural population in the total pop-
ulation, the proportion of foreign direct investment in the total
GRP, the purchasing price indices of raw material, fuel, and energy,
and the proportion of state-owned industrial output in the total
industrial output respectively.
Urbanization is a phenomenon of economic and social modern-
ization, and it has great impact on environmental issues, including
493
energy use and emissions. There are many studies that investigate
the relationship but the results show conflicting results, suggesting
that the impact is complex [54,55]. As one of the main paths of
technological progress, China has become the largest recipients
of foreign investment in the world. Foreign direct investment plays
an important role in productivity improvement [56,57]. Although
energy price can promote market mechanisms to ensure energy
conservation and emissions abatement, its distortion weakens this
effect or even has counteract it [58,59]. Ownership structure stands
is widely accepted that reducing the proportion of state-owned
industrial output in the total industrial output is helpful for energy
conservation and emissions abatement, or green productivity [60].
TFEEit or TFCEit may observe lag effects. That is, the previous
TFEEit or TFCEit may affect the present TFEEit or TFCEit. Therefore,
a lag term of the dependent variables is added into the model,
which is now expressed as follows:
EEit ¼ c þ cEEi;t1 þ a1 gffit þ a2 labefit þ a3 capefit þ a4 oisit
þ X 0it b þ t þ li þ eit ð10Þ
ð8Þ
Before using formulas (9) and (10), we conduct a correlation
analysis to test the multicollinearity among all independent vari-
ables. The correlation coefficients between each two variables are
below 0.45 (the highest being 0.448 between urban and fdi), which
indicates the absence of high multicollinearity.
3. Results
Table 1 shows the estimations of formulae (9) and (10) using
TFEE as the dependent variable. Formula (9) can be estimated with
either a fixed- or random-effects model, which is determined
through a Hausman test.5 After running the test, the preferred
method is a fixed-effect model. gff, labef, capef, ois, urban, fdi, price,
and ps are considered as endogenous variables. Hausman specifica-
tion tests are conducted to test for simultaneity bias. Based on the
Hausman test results, the null assumption that ‘‘all variables are
exogenous variables” cannot be rejected, which implies the absence
of endogeneity. To eliminate the influence of heteroskedacity on the
group-wise or cross-sectional dependence, feasible generalized least
squares (FGLS) is applied to re-estimate formula (9) and enhance the
validity of the estimations. The first and second column of Table 1
(model 1, 2) shows the estimations of FGLS.
ð9Þ Model 1 shows that while labor allocation efficiency has a sig-
nificant positive effect on TFEE, other factors, such as the contribu-
tion of investment to economic growth, capital allocation
efficiency, and manufacturing structural changes, produce negative
effects on TFEE. The contribution of investment to economic
growth has a statistically insignificant coefficient. To test the
robustness of these results, four control variables, namely, urban-
ization, foreign direct investment, energy price, and ownership
structural changes, are added into model 1. Table 1 shows the esti-
mations of model 2, which indicate that the contribution of invest-
ment to economic growth produces a negative effect on TFEE at the
15% level. Labor allocation efficiency and manufacturing structural
changes produce a positive and negative effect respectively at the
1% level. However, capital allocation efficiency produces a statisti-
cally insignificant negative effect.
Wooldridge test is conducted to determine if autocorrelation
exists in the panel data models after estimating formula (9)
through the ordinary fixed-effects or the random-effects model
[61]. There is autocorrelation in the panel data models, which
necessitates the estimation of formula (10). This formula can be
5
Hausman test assumes that both li and eit are independent and equally
distributed.
494 K. Li, B. Lin / Applied Energy 187 (2017) 489–500

Table 1
Estimation results: dependent variable is TFEE.

Variables FE-GLS(1) FE-GLS(2) Diff-GMM(3) Diff-GMM(4) Sys-GMM(5) Sys-GMM(6)

*** *** ***
gff 0.0013 (0.001) .
0.0031 (0.002) 0.0037 (0.001) 0.0034 (0.001) 0.0054 (0.001) 0.0060*** (0.001)
labef 0.0520*** (0.005) 0.0370*** (0.009) 0.0267*** (0.005) 0.0168*** (0.004) 0.0111** (0.006) 0.0036* (0.002)
capef 0.2709*** (0.049) 0.0132 (0.092) 0.1587 ***
(0.049) *
0.4576 (0.254) .
0.1098 (0.065) 0.2657* (0.156)
ois 0.0641*** (0.011) 0.0664*** (0.026) 0.0460*** (0.014) 0.0219. (0.017) 0.0576*** (0.021) 0.0516** (0.021)
urban 0.1739* (0.093) 0.4633*** (0.062) 0.0993* (0.058)
fdi 0.2442. (0.159) 0.2040 (0.178) 0.1762 (0.160)
price 0.0180 (0.013) 0.0303*** (0.007) 0.0424*** (0.006)
ps 0.0438* (0.025) 0.0223 (0.023) 0.0880*** (0.029)
t 0.0060*** (0.000) 0.0055*** (0.001) 0.0004 (0.000) 0.0029** (0.001) 0.0011** (0.000) 0.0045 (0.001)
Lag.TFEE 0.6941*** (0.033) 0.3545*** (0.038) 0.9477*** (0.028) 0.9526** (0.037)
Constant 1.3107*** (0.054) 1.0228*** (0.080) 0.2043*** (0.018) 0.4825*** (0.033) 0.0408** (0.018) 0.0334 (0.043)
Wald test 5850.10*** 4768.70*** 1229.14*** 2579.53*** 2950.95*** 5577.55***
Robust Hausman test 24.84*** 69.74***
Hausman specification test 0.0300 (1.000) 14.1000 (0.1189)
Autocorrelation test 29.2680 (0.000) 29.0370 (0.000)
AR(1) test 2.4560 (0.014) 1.9323 (0.0533) 2.5460 (0.011) 2.6503 (0.008)
AR(2) test 0.1621 (0.871) 0.1628 (0.870) 0.0770 (0.939) 0.1159 (0.908)
Sargan test 27.0142 (1.000) 24.1258 (0.999) 25.6336 (1.000) 22.8140 (1.0000)
Sample 420 360 390

Notes: Standard errors of variables and p-values of test statistics are shown in parentheses.
.
P < 0.15.
*
P < 0.10.
**
P < 0.05.
***
P < 0.01.

estimated through two widely used methods, namely difference-
GMM [62] and system-GMM [63]. Both can eliminate individual
effects and other immeasurable, time-constant independent vari-
ables as well as partly solve the problems of omitted variables.
The system-GMM estimator is more efficient than the difference-
GMM estimator if the dependent variable tends to be non-
stationary [64]. Both methods are applied to estimate formula
(10), and the results are shown in columns 3–6 of Table 1 (model
3–6).
Table 1 shows that the null hypothesis, ‘‘there is no second-
order serial correlation for the random term of model 3–6,” cannot
be rejected. The feasibility of the instrumental variables is con-
firmed through the Sargan test. Based on these tests, the effective-
ness and consistency of the GMM estimator in estimating model
3–6 are verified.
Based on the coefficients of LagTFEE in model 3–6, the previous
TFEEit positively affects the current TFEEit, which implies that such
effect will accumulate over time. The results of models 3 and 5
show that the contribution of investment to economic growth, cap-
ital allocation efficiency, and manufacturing structural changes
produce significant negative effects on TFEE, whereas labor alloca-
tion efficiency produces a significant positive effect on TFEE. After
adding four control variables, the estimations (model 4 and 6)
show that the signs of the coefficients of the four core variables
(gff, labef, capef, and ois) are consistent with those that are esti-
mated by models 3 and 5.
The estimations in Table 1 show that when CO2 emissions are
ignored, contribution of investment to economic growth hinders
energy efficiency, laborers in the three industries are conducive
for the enhancement of energy efficiency (creating a ‘‘structural
bonus” on energy efficiency), and capital transfers among the three
industries as well as the manufacturing structural changes are
unfavorable for the enhancement of energy efficiency (creating
‘‘structural burdens” on energy efficiency).
To compare these results, formulae (9) and (10) are re-
estimated using TFCEit as the dependent variable, and the results
are shown in Table 2.6 The results of model 7–12 shows that contri-
6
All of the test results in Table 2 are just the same with those in Table 1.
bution of investment to economic growth and capital allocation effi-
ciency are negatively related with TFCEit, whereas labor allocation
efficiency and manufacturing structural changes are positively
related to TFCEit. Based on Table 1 and Table 2, structural changes
in the manufacturing sector negatively affect TFEEit but positively
affect TFCEit.
4. Main findings and discussion
The following findings are derived by comparing Tables 1 and 2.
(1) The investment-driven economic growth of China is not con-
ducive for the enhancement of green productivity (TFEE/
TFCE).
The contribution of investment to economic growth of China is
so high that it affects energy efficiency in different ways. First,
most of the inputs are derived from capital, and all of the sources
are concentrated on the asset sector. Although the embodied tech-
nological innovations are improved through this process, the ben-
efits of technological innovation from real industries are hindered
[8]. Knowledge (which is represented by the education of the
laborer) and embodied technological progress (which is repre-
sented by the service life of the capital stock) are considered as
inputs when estimating TFEE and TFCE. Therefore, TFEE and TFCE
are mainly driven by broad technological progress. Given the lack
of information on broad technological progress, the growth rate
of the proportion of research and development (R&D) expenditure
in the total GDP (R&D intensity) and patent number of the inven-
tion (patent) are used to represent such progress. Table 3 shows
that the growth rates of R&D intensity and patent reduce when
gff exceeds 50%. The correlation coefficient between gff and R&D
intensity and between gff and patents are –0.291 and –0.112
respectively. As economic growth increases, its dependence on
capital inputs, R&D intensity and patent growth rates are further
reduced (i.e., the technological progress is slowed, which reduces
the maximum output of the input factors and subsequently inhi-
bits the total factor energy efficiency). An investment-driven eco-
nomic growth usually leads to overexploitation of resources,
which increases environmental pollution. According to the results
 K. Li, B. Lin / Applied Energy 187 (2017) 489–500 495

Table 2
Estimation results: dependent variable is TFCE.

Variables FE-GLS(7) FE-GLS(8) Diff-GMM(9) Diff-GMM(10) Sys-GMM(11) Sys-GMM(12)

⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄
gff 0.0078 (0.001) 0.0150 (0.001) 0.0235 (0.001) 0.0212 (0.002) 0.0018 (0.001) 0.0132⁄⁄⁄ (0.002)
labef 0.0049 (0.004) 0.0012 (0.004) 0.0268⁄⁄⁄ (0.002) 0.0367⁄⁄⁄ (0.004) 0.0775⁄⁄⁄ (0.001) 0.1255⁄⁄⁄ (0.007)
⁄⁄ ⁄⁄⁄ ⁄⁄⁄ ⁄⁄⁄
capef 0.0571 (0.026) 0.1413 (0.041) 0.6403 (0.045) 0.5845 (0.072) 0.5045⁄⁄⁄ (0.039) 0.3955⁄⁄⁄ (0.093)
ois 0.0036 (0.006) 0.0162⁄⁄⁄ (0.005) 0.0258⁄⁄⁄ (0.007) 0.0279⁄⁄⁄ (0.011) 0.0066。 (0.005) 0.0204。 (0.014)
⁄⁄⁄
urban 0.0300 (0.038) 0.1609 (0.032) 0.3472⁄⁄⁄ (0.052)
fdi 0.0514 (0.053) 0.2286⁄⁄⁄ (0.061) 0.7827⁄⁄ (0.389)
price 0.0531⁄⁄⁄ (0.008) 0.0342⁄⁄⁄ (0.010) 0.0189 (0.015)
ps 0.0441⁄⁄⁄ (0.014) 0.2308⁄⁄⁄ (0.022) 0.7293⁄⁄⁄ (0.075)
t 0.0056⁄⁄⁄ (0.000) 0.0038⁄⁄⁄ (0.001) 0.0031⁄⁄⁄ (0.000) 0.0046⁄⁄⁄ (0.001) 0.0032⁄⁄⁄ (0.000) 0.0137⁄⁄⁄ (0.002)
Lag.TFCE 0.1422⁄⁄⁄ (0.003) 0.1382⁄⁄⁄ (0.014) 0.7036⁄⁄⁄ (0.003) 0.4322⁄⁄⁄ (0.032)
Constant 0.9457⁄⁄⁄ (0.084) 0.8997⁄⁄⁄ (0.054) 0.5706⁄⁄⁄ (0.005) 0.3322⁄⁄⁄ (0.024) 0.2172⁄⁄⁄ (0.004) 0.1999⁄⁄ (0.079)
Wald test 4061.31⁄⁄⁄ 12450.65⁄⁄⁄ 4616.57⁄⁄⁄ 2178.61⁄⁄⁄ 508156.1⁄⁄⁄ 26625.66⁄⁄⁄
Robust Hausman test 12.17⁄⁄⁄ 331.53⁄⁄⁄
Hausman specification test 0.1000 (1.000) 2.3400 (0.985)
Autocorrelation test 52.7920 (0.000) 47.3080 (0.000)
AR(1) test 1.1691 (0.242) 1.1582 (0.247) 1.2362 (0.216) 1.1286 (0.259)
AR(2) test 1.3447 (0.179) 1.3464 (0.178) 1.4049 (0.160) 1.3372 (0.182)
Sargan test 28.6081 (1.000) 27.2230 (1.000) 28.9818 (1.000) 23.8246 (1.000)
Sample 420 360 390

Note: Same as Table 1.

Table 3
Investment-driven economic growth, innovation and income gap: main years.

Index 1997 1998 2000 2004 2005 2007

Investment-driven economic growth (gff) 33.11% 54.56% 32.75% 48.15% 66.34% 51.95%
Growth rate of R&D intensity 11.76% 0.09% 18.19% 4.57% 5.57% 2.83%
Growth rate of patent number of invention 10.43% 6.79% 99.24% 57.40% 12.23% 26.76%
Proportion of employee income in GRP 50.84% 50.69% 48.65% 41.57% 41.37% 39.60%
Income gap 0.441 0.452 0.501 0.531 0.542 0.544

Note: For statistical consistency, R&D expenditures, the patents number of invention, labor’s return and China’s GDP are calculated by adding all province’s data (except
XinJiang). The data of income gap are measured by the method of Cheng [67], which is based on Gini coefficients.

of the empirical models, gff produces more negative effects on TFCE
than on TFEE, which confirms that an investment-driven economic
growth seriously causes environmental deterioration in China.
Second, the excessive contributions of investment to economic
growth decrease the proportion of employee income in GRP,
expand the income gap and reduce consumer demand, which
inhibits the enhancement of energy efficiency. Capital increases
at a much faster rate than labor, which decreases the proportion
of employee income in GRP. Table 3 shows that the proportion
has decreased from 50.84% in 1997 to 39.60% in 2007, thereby pro-
moting uneven income distribution. Li, Ye [65] confirmed that the
expansion of the income gap was unfavorable for the enhancement
of energy efficiency. According to the efficiency wage theory, labor
return is extremely low compared to capital return, thereby reduc-
ing labor efficiency and eventually decreasing the total factor
energy efficiency. Relatively low income levels limit the consump-
tion capacity of the residents and inhibit technological progress
given the creative demands of the consumer. Thus, the total factor
energy efficiency is negatively affected.
(2) The allocation of the labor force in the three industry cate-
gories is conducive for the enhancement of green productiv-
ity, whereas the allocation of the capital stock is unfavorable
for the enhancement of green productivity.
As shown in Table 4, the share of employment in the primary
industry of China decreased by 15.654% from 1997 to 2010,
whereas the share of employment in the secondary and tertiary
industries increased by 5.873% and 9.782% respectively. This may
imply that most employees in the primary industry transferred
to the secondary and (particularly) tertiary industries. During the
same period, the share of capital stock decreased by 2.998% in
the primary industry, increased by 4.914% in the secondary indus-
try, and decreased by 1.916% in the tertiary industry. The interac-
tion of the changes in the labor force and capital stock results in an
output structural change. The proportion of the primary industry
in the total GDP decreased by 9.973% while that of the secondary
and tertiary increased by 5.705% and 4.268% respectively.
The three primary industries are ranked as (1) secondary indus-
try, (2) tertiary industry, and (3) primary industry, based on the
level and growth rate of their labor productivity. The huge gap
among these industries instigates a labor force transfer from the
primary industry to the secondary and tertiary industries, which
enables labor structural changes to contribute to the enhancement
of energy efficiency.
Capital productivity is a different case. In the investment-driven
economy, the capital stock increases at a much faster rate than out-
put, thereby decreasing capital productivity. From a static point of
view, the capital productivity of the primary industry is higher
than that of the two other industries, and the capital productivity
of the secondary industry is higher than that of the tertiary indus-
try. Therefore, transferring the capital stock of China does not con-
form to the prerequisite for the production of a ‘‘structural bonus”
on green productivity. From a dynamic point of view, capital pro-
ductivity has negative growth rates, with the primary industry
having the largest slump. This tightens the capital productivity
gap among the three industries. For example, the capital productiv-
ity of the primary industry in 1997 was 2.876 times and 4.376
times larger than that of the secondary and tertiary industries
respectively, but decreased to 2.510 times and 3.340 times in 2010.
The capital transformation in China is in accordance with the
economic laws of the country. Although the capital productivity
496 K. Li, B. Lin / Applied Energy 187 (2017) 489–500

Table 4
Inputs structural, the level and growth rate of productivity.

Index Year Labor force/employment Capital stock

All I II III All I II III
Structural (%) 1997 100 51.935 22.356 25.709 100 5.581 41.388 53.031
2003 100 48.980 21.622 29.399 100 3.897 41.742 54.361
2010 100 36.673 28.735 34.591 100 3.075 46.056 50.869
The level of productivity 1997 1.241 0.416 2.490 1.823 0.532 1.659 0.577 0.379
2003 1.992 0.511 4.366 2.713 0.456 1.470 0.517 0.336
2010 3.514 0.804 6.122 4.221 0.332 0.907 0.361 0.271
The growth rate of productivity (%) 1997 7.866 2.924 11.534 5.437 1.433 1.585 0.934 0.625
2003 8.197 4.736 5.005 4.792 2.681 2.459 1.875 5.325
2010 7.817 5.588 6.249 5.436 5.347 8.556 5.209 4.948
Mean 8.333 5.081 7.599 6.616 3.395 4.060 3.319 2.379

Note: All data are calculated by adding all province’s data (except Xin Jiang). I, II, III represent the primary, secondary and tertiary industries, respectively. Labor productivity is
calculated by the ratio of output to employment and measured by million Yuan per capita; Capital productivity is calculated by the ratio of output to capital stock, and its
indicates how much Yuan created per capital stock inputs. Output and capital stock are measured by constant 2000 prices. The mean growth rates of productivity are the
average growth rates between 1997 to 2010.

of the primary industry is higher than that of the other industries,
such industry accounts for a small share of GDP and cannot absorb
or accommodate large investments. Since 1997, particularly after
entering the World Trade Organization in 2001, China has invested
massively in the secondary industry given the changes in the con-
sumption behaviors of the residents and the then-recent deregula-
tion of investment. Therefore, capital is freely transferred from a
highly productive primary industry to a lowly productive sec-
ondary industry, which is not conducive for the enhancement of
green productivity. The production structure of capital has greatly
improved, as reflected in the reduced difference in the capital pro-
ductivity of the three industries and the annual increase in their
capital allocation efficiency. For example, the capital allocation
efficiency increased from 4.892% to 1.667% in 2010, which
implied that the industrial structure had loosened the constraints
on capital productivity. Such increase effectively illustrates the
optimized industrial structure of China.
(3) Manufacturing structural changes produce a ‘‘structural bur-
den” on the energy adjusted TFP (TFEE) but produce a ‘‘struc-
tural bonus” on the energy and carbon dioxide emissions
adjusted TFP (TFCE).
China is at the metaphase of industrialization, in which energy-
intensive industries rapidly develop. Industrial structures can be
further readjusted and optimized by developing the third manu-
facturing category, particularly high-end manufacturing.
We compare the third and second manufacturing categories in
terms of their productivity based on three criteria. Table 5 shows
the results of the comparison. The value-added rates and labor pro-
ductivity of the third manufacturing category are lower than those
of the second manufacturing category during 1997–2007 and
2003–2010 respectively. The gap between these categories contin-
ues today. By using the method of Lü and Zhou [36]7, the second
manufacturing category has the largest marginal productivity of cap-
ital and is followed by the third and first manufacturing categories.8
7
Different from Lü & Zhou, we use labor, capital, and energy as inputs. The capital
stocks of the three manufacturing categories are estimated by using the method of
Zhang et al. [66]. Additional information on the estimation procedure and results are
available upon request.
8
The gap among the three manufacturing categories in terms of their marginal
productivity of capital is reduced, which implied a reasonable allocation of capital
among the three categories and signaled an improvement in their capital efficiencies.
The marginal productivity of capital of the third manufacturing category exceeded
that of the second category in 2010, which confirmed that the current trend and
direction of the manufacturing structural changes accelerated the development of the
third manufacturing category.
These comparisons imply that resources should be invested in the
second manufacturing category to ensure improvements in the man-
ufacturing sector.9
The low value-added rate of the third manufacturing category
reflects high input costs and low contribution coefficients of inter-
mediate inputs.10 From 1997 to 2007, the average contribution coef-
ficients of the intermediate inputs of the third manufacturing
category were 0.328, which was lower than that of the intermediate
inputs of the second manufacturing category (0.349). This low con-
tribution coefficient resulted from the participation of the third
manufacturing category in international specialization by importing
large-scale core components and capital goods and by selling the
final products internationally (improvement in trade). As China does
not produce core technology, local manufacturers have to import
high value intermediate inputs, which decreases the value-added
rate and labor productivity of the third manufacturing category.
In order to illustrate this conclusion, we estimate the ratio of
intermediate inputs from import to the total intermediate inputs.
However, it is difficult to get the correct information about the
use of imports. In other words, determining whether an import is
used for intermediate production or for end-use production seems
impossible. So, we present three assumptions:
Assumption 1 (A1): All imports are used for intermediate input.
Under this assumption, in 1997–2010, the average contribution
coefficients of intermediate inputs of the second and the third
manufacturing category are 20.580% and 45.140%, respectively.
Assumption 2 (A2): The proportion of processing trade to total
inputs of the second manufacturing category is 42.127%, while the
third manufacturing category is 54.153%.11
Assumption 3 (A3): The proportion of processing trade to total
inputs of the second manufacturing category is the same as the
third manufacturing category, i.e., the proportion of processing
trade to the total imports of China during 1997–2007.12
9
Various resources are invested in the second manufacturing category as the
energy and capital costs are maintained at low levels for a long time.
10
Contribution coefficients of intermediate inputs = industrial value-added/costs of
intermediate inputs.
11
For the second manufacturing category, 42.127% is the average value of the
proportion of the processing trade to import of China in 1997–2007. The third
manufacturing category is dominated by the high-tech sectors. But we lack the
import and export data about the processing trade for high-tech products before
2009. So, this data (54.153%) is the proportion of imports as processing trade to the
total imports of high-tech products in 2009. Historically, the technical level of China’s
high-tech sectors won’t be higher than it in 2009. So, it is reasonable to assume that
the proportion should be higher than 54.153%.
12
The annually data are calculated according to ‘‘China trade and external economic
statistical yearbook”.
 K. Li, B. Lin / Applied Energy 187 (2017) 489–500 497

Table 5
A Comparison table of growth quality between the second manufacturing category and the third manufacturing category.

Index Manufacturing Category 1997 1998 2000 2003 2005 2008 2010
Account for total manufacturing output Second 38.881% 38.024% 39.312% 37.523% 40.669% 42.236% 40.855%
Third 28.720% 30.586% 32.272% 36.628% 35.306% 34.834% 36.048%
Value-added Rate of Industry Second 25.976% 24.837% 25.344% 26.761% 25.631% / /
Third 24.959% 24.720% 25.215% 24.803% 24.325% / /
Overall labor productivity Second 9.069 12.089 17.388 28.951 41.904 61.806 76.030
Third 9.097 12.679 18.643 33.119 37.977 46.898 59.052
Marginal producti-vity of capital Second 3.289% 3.171% 2.652% 1.645% 1.624% 1.567% 1.839%
Third 2.559% 2.462% 2.015% 1.232% 1.263% 1.452% 1.844%
Account for total CO2 emission Second 85.272% 87.425% 89.098% 90.371% 91.521% 92.192% 92.677%
Third 5.712% 3.394% 2.840% 2.524% 2.182% 2.074% 2.049%
Contribution coefficients of intermediate inputs Second 0.351 0.330 0.339 0.365 0.345 / /
Third 0.333 0.328 0.337 0.330 0.321 / /

Note: Based on data available and comparability, all data are calculated by adding varies industries of China; CO2 emission of manufacturing are calculated by reference
method (IPCC, 2006); overall labor productivity is measured by million Yuan (constant 2000 prices) per capita; the 2008–2010 data on industrial added value are not
available.

Table 6
A Comparison table of intermediate inputs and foreign investment profits between the second manufacturing category and the third manufacturing category.

Index Manufacturing category 1997 2000 2002 2005 2007 Mean

The proportion of processing trade to total inputs (%) A1 Second 19.379 22.814 23.772 19.362 15.104 20.580
Third 38.003 46.432 49.159 48.536 39.437 45.140
A2 Second 8.164 9.611 10.014 8.157 6.363 8.670
Third 20.580 25.144 26.621 26.284 21.356 24.444
A3 Second 9.556 9.381 9.842 8.039 5.822 8.659
Third 18.740 19.093 20.352 20.152 15.201 18.898
The proportion of profit from foreign-funded enterprises to Second 18.675 28.114 30.354 26.803 24.669 25.549
the total profits (%) Third 52.514 59.891 60.136 55.323 51.568 54.341

According to Table 6, the import costs comprise 18.898% to
45.140% of the total intermediate input costs of the third manufac-
turing category, which is higher than that of the second manufac-
turing category.
Profits from foreign-funded companies comprise more than 50%
of the total profit of the third manufacturing category, resulting
from the high investment of foreign funds in the same category.
These foreign-funded companies also develop intermediate goods
on a large scale. The production processes in China are mainly han-
dled by labor-intensive sectors, such as processing and assembly
firms, that do not adopt high-technology equipment or procedures.
This pattern increases the production cost, reduces the contribution
coefficients of the intermediate inputs, decreases the competitive
advantage of the third manufacturing category over the second
manufacturing category, and negatively affects TFEE by increasing
the weight of the third manufacturing category. These results indi-
cate that the manufacturing structure can be effectively upgraded
through a forced adoption of foreign technology and by enhancing
the capacity of China to innovate independently. These measures
can also improve the technological production and the interna-
tional competitiveness of the Chinese manufacturing sector.
Despite outranking the other manufacturing categories in terms
of labor productivity and value-added rate, the second manufac-
turing category is largely energy-intensive and emits large
amounts of CO2. Table 5 shows that the second manufacturing cat-
egory emits CO2 that accounts for more than 85% of total CO2 emis-
sions in the country, but produces an output value that accounts
for about 40% of the total manufacturing output. By contrast, the
third manufacturing category emits a negligible amount of CO2
and produces an output value that accounts for more than 30% of
the total manufacturing output.
These findings reveal the difficulty for China to adjust its man-
ufacturing structure. A forced mechanism must be immediately
developed to accelerate the structural upgrades by conserving
more energy and reducing CO2 emissions. In effect, increasing
the proportion of the tertiary category in the entire manufacturing
sector is conducive for improving the environmental performance
of China.
5. Conclusions and policy implications
This study has empirically investigated the effects of
investment-driven economic growth and structural transformation
(including those of the three industries and three manufacturing
categories, which also indicate the optimization and upgrading of
industrial structure respectively) on green productivity, including
energy adjusted TFP (TFEE) and energy and carbon dioxide emis-
sions adjusted TFP (TFCE). The study employed panel data that
cover 30 Chinese provinces from 1997 to 2010. Investment-
driven economic growth and the capital transformation in the
three industry categories were not beneficial to green productivity,
whereas the flow of laborers in the three categories helped to
enhance green productivity. Structural changes in the manufactur-
ing sector negatively affected the energy adjusted TFP (TFEE) when
carbon dioxide emissions is neglected.
There is no doubt that the developing countries have many
common features about economic growth model and industrial
structure. The economic growth of most developing countries
depends much on inputs rather than technology progress, implying
that the growth is an extensive model rather than an intensive one.
This is similar for most developing countries during the industrial-
ization process, as their industrial structure shows similar dynamic
characteristics as China’s. In this sense, our study presents suffi-
cient contributions to the new body of literature on green develop-
ment from the international perspective rather than focusing on
China.
498 K. Li, B. Lin / Applied Energy 187 (2017) 489–500

Table A1
Definition of the variables used in the analysis.

Variable Definition Unit of measurement Data source

Labor (L) Eq. (1) millions of persons China Labor Statistical Yearbook
Capital Stock (K) Eq. (2) 100 million yuan (in constant China Statistical Yearbook
2000 prices)
Energy consumption (E) The primary energy consumption of each province 104 tce China Energy Statistical
Yearbook
China Compendium of Statistics,
1949–2008
Gross regional product (GRP) Gross regional product 108 yuan (in constant 2000 The National Bureau of Statistics
prices) of China
CO2 emissions (C) Eq. (3) Million tons China Energy Statistical
Yearbook
IPCC
Economic growth model (gff) Eq. (6) % China Statistical Yearbook
Labor allocation efficiency (labef) Eq. (7) Unit China Statistical Yearbook
Capital allocation efficiency (capef) Eq. (7) Unit China Statistical Yearbook
Structural transformation in Eq. (8) Unit China Statistical Yearbook
manufacturing (ois)
Urbanization (urban) The proportion of non-agricultural population in the total % China Statistical Yearbook
population
Foreign direct investment (fdi) The proportion of foreign direct investment in the total % China Statistical Yearbook
GRP
Energy price (price) The purchasing price indices of raw material, fuel, and Unit China Statistical Yearbook
energy
Ownership structure (ps) The proportion of state-owned industrial output in the % China Statistical Yearbook
total industrial output

Table A2
Summary statistics of variables.

Variable Mean Std. Dev. Minimum Maximum

Labor (L) 2347.066 1547.932 155.907 6450.062
Capital Stock (K) 14520.860 12154.610 1648.568 79660.460
Energy consumption (E) 7799.150 5953.178 390.260 34807.770
Gross regional product (GRP) 5498.146 5442.199 205.682 36030.550
CO2 emissions (C) 22075.200 17874.520 445.251 110658.900
Economic growth model (gff) 0.627 0.464 0.988 3.692
Labor allocation efficiency (labef) 0.102 0.339 1.011 2.425
Capital allocation efficiency (capef) 0.021 0.045 0.165 0.084
Structural transformation in manufacturing (ois) 0.181 0.130 0.000 0.633
Urbanization (urban) 0.336 0.159 0.140 0.889
Foreign direct investment (fdi) 0.032 0.029 0.001 0.165
Energy price (price) 1.213 0.259 0.868 2.150
Ownership structure (ps) 0.520 0.206 0.107 0.899

The above conclusions produce two significant policy implica-
tions. First, the government must introduce structural transforma-
tion policies to reduce CO2 emission. It must not only focus on the
rational allocation of resources, but also on the forced adoption of
foreign technologies in the manufacturing sector and on enhancing
the capacity of China to innovate independently. Second, the gov-
ernment must actively promote energy market reforms, increase
energy costs, and hinder environmental deterioration. All these
could further stimulate structural changes.
China faces increasing pressures from escalating energy con-
sumption costs and further environmental deterioration. The Chi-
nese government must consider changing the economic growth
model of the country and adjusting the industrial structure to
address these concerns. Industrial structural changes are crucial
for a country to switch from one economic growth model to
another. This study provides the following recommendations for
the industrial structural transformation of China.
First, industrial structures are primarily adjusted for further
optimization. An increase in the labor or capital allocation effi-
ciency increases the likelihood of an industrial structure to opti-
mize. To promote such increase, the government must actively
develop factor markets and eliminate or reduce barriers for certain
factors (particularly capital) in order to allow the entry of firms
into any industry without compromising national security. It must
also permit certain factors to flow freely among competitive
industries. Moreover, the government must improve the quality
of education and launch training programs, particularly profes-
sional training programs, to improve the quality and capacity of
laborers.
Second, manufacturing firms, especially high-tech ones, must
adopt the latest technology and accelerate their independent inno-
vation to optimize and upgrade their manufacturing structure.
Given that China is at the metaphase of industrialization, energy-
intensive industries develop comparative advantages over other
industries, in line with the experience of industrialized countries.
However, the third manufacturing category may also develop com-
parative advantages over other industries by adopting technologi-
cal innovations, which pushes China out of the metaphase and into
post-industrialization.
Third, energy price reforms in China have remained stagnant for
a number of years. Therefore, energy costs are maintained at low
levels, and energy-intensive industries, such as steel and petro-
chemical manufacturing, have rapidly expanded, hindering the
upgrading of the industrial structure. Therefore, the government
 K. Li, B. Lin / Applied Energy 187 (2017) 489–500
must promote energy market reforms and increase the costs of
energy consumption in order to stem environmental deterioration.
This would reduce the advantages of the second manufacturing
category, particularly energy-intensive industries, and upgrade
the manufacturing structure.
Acknowledgements
The paper is supported by Xiamen University - Newcastle
University Joint Strategic Partnership Fund, the Grant for Collabo-
rative Innovation Center for Energy Economics and Energy Policy
(No. 1260-Z0210011), and Xiamen University Flourish Plan Special
Funding (No. 1260-Y07200).
Appendix
See Tables A1 and A2.
References
[1] Mundaca L, Markandya A. Assessing regional progress towards a ‘Green Energy
Economy’. Appl Energy 2015.
[2] Sheehan P, Sun F. Energy use and CO2 emissions in china: retrospect and
prospect; 2006.
[3] Jiang L, Folmer H, Ji M. The drivers of energy intensity in China: a spatial panel
data approach. China Econ. Rev 2014;31:351–60.
[4] Chen S. Environmental pollution emissions, regional productivity growth and
ecological economic development in China. China Econ Rev 2014;35:171–82.
[5] Herrerias MJ, Joyeux R, Girardin E. Short- and long-run causality between
energy consumption and economic growth: evidence across regions in China.
Appl Energy 2013;112:1483–92.
[6] IEA. World Energy Outlook; 2012.
[7] Chen S, Jefferson GH, Zhang J. Structural change, productivity growth and
industrial transformation in China. China Econ Rev 2011;22:133–50.
[8] Shen LS, Wang H. What does it mean that China’s value-added rate goes down.
Econ Res J 2006:59–66 [in Chinese].
[9] Cai F, Du Y, Wang MY. The political economy of emission in china: will a low
carbon growth be incentive compatible in next decade and beyond? Econ Res J
2008:4–11 [in Chinese].
[10] Lin YF, Su J. Analysis on the transformation of economic growth mode. Manage
World 2007:5–13 [in Chinese].
[11] Chen S, Golley J. ‘Green’ productivity growth in China’s industrial economy.
Energy Econ 2014;44:89–98.
[12] Medlock KB, Soligo R. Economic development and end-use energy demand.
Energy J 2001;22(2):77–105.
[13] Schäfer A. Structural change in energy use. Energy Policy 2005;33:429–37.
[14] Li K, Lin B. The nonlinear impacts of industrial structure on China’s energy
intensity. Energy 2014;69:258–65.
[15] Chang T-P, Hu J-L. Total-factor energy productivity growth, technical progress,
and efficiency change: an empirical study of China. Appl Energy
2010;87:3262–70.
[16] Ma H, Oxley L, Gibson J. China’s energy economy: a survey of the literature.
Econ Syst 2010;34:105–32.
[17] Greening LA, William BD, Schipper L, Khrushch M. Comparison of six
decomposition methods: application to aggregate energy intensity for
manufacturing in 10 OECD countries. Energy Econ 1997;19:375–90.
[18] Liu N, Ang BW. Factors shaping aggregate energy intensity trend for industry:
energy intensity versus product mix. Energy Econ 2007;29:609–35.
[19] Wing IS. Explaining the declining energy intensity of the U.S. economy.
Resource Energy Econ 2008;30:21–49.
[20] Weber CL. Measuring structural change and energy use: decomposition of the
US economy from 1997 to 2002. Energy Policy 2009;37:1561–70.
[21] Voigt S, De Cian E, Schymura M, Verdolini E. Energy intensity developments in
40 major economies: structural change or technology improvement? Energy
Econ 2014;41:47–62.
[22] Zeng L, Xu M, Liang S, Zeng S, Zhang T. Revisiting drivers of energy intensity in
China during 1997–2007: a structural decomposition analysis. Energy Policy
2014;67:640–7.
[23] Timmer MP, Szirmai A. Productivity growth in Asian manufacturing: the
structural bonus hypothesis examined. Struct Change Econ Dynam
2000;11:371–92.
[24] Fan S, Zhang X, Robinson S. Structural change and economic growth in China.
Rev Develop Econ 2003;7:360–77.
[25] Zheng RG, Gan CH, Yu DF. The industrial structure and institutional effects of
economic growth in China’s transition period: an analysis based on a
stochastic frontier model. China Indus Econ 2010:58–67 [in Chinese].
[26] Zhou X, Zhang J, Li J. Industrial structural transformation and carbon dioxide
emissions in China. Energy Policy 2013;57:43–51.
499
[27] Fagerberg J. Technological progress, structural change and productivity
growth: a comparative study. Struct Change Econ Dynam 2000;11:393–411.
[28] Ark Bv, Timmer M. Asia’s productivity performance and potential: the
contribution of sectors and structural change; 2003.
[29] Oulton N, Srinivasan S. Productivity growth in UK industries, 1970–2000:
structural change and the role of ICT; 2004.
[30] Isaksson A. Structural change and productivity growth: a review with
implications for developing countries; 2009.
[31] Chai J, Guo J-E, Wang S-Y, Lai KK. Why does energy intensity fluctuate in
China? Energy Policy 2009;37:5717–31.
[32] Yang Y, Cai W, Wang C. Industrial CO2 intensity, indigenous innovation and
R&D spillovers in China’s provinces. Appl Energy 2014;131:117–27.
[33] Boyd GA, Curtis EM. Evidence of an ‘‘energy-management gap” in U.S.
manufacturing: spillovers from firm management practices to energy
efficiency. J Environ Econ Manage 2014;68:463–79.
[34] Hofman B, Labar K. Structural change and energy use: evidence from China’s
provinces; 2007.
[35] Zhan J, Chen SY, Jefferson GH. Structural reform and industrial growth in
China. Econ Res J 2009:4–20 [in Chinese].
[36] Lü T, Zhou SL. Industrial structure upgrading and economic growth mode
transformation in China. Manage World 1999:113–25 [in Chinese].
[37] Hu J-L, Wang S-C. Total-factor energy efficiency of regions in China. Energy
Policy 2006;34:3206–17.
[38] Chang M-C. A comment on the calculation of the total-factor energy efficiency
(TFEE) index. Energy Policy 2013;53:500–4.
[39] Gordon RJ. Does the ‘New Economy’ measure up to the great inventions of the
past? J Econ Perspect 2000;14:49–74.
[40] Zhao ZY, Lü BY, Guo QW, Jia JX. On the dynamic integration of capital
accumulation and technological progress: a stylized fact in China’s economic
growth. Econ Res J 2007:18–31 [in Chinese].
[41] Song DL, Wang LH, Dong ZQ. Capital embodied technological progress and its
contribution to economic growth: 1981–2007. Soc Sci China 2011:91–106 [in
Chinese].
[42] Limam YR, Miller SM. Explaining economic growth: factor accumulation, total
factor productivity growth, and production efficiency improvement; 2004.
[43] Shan HJ. Reestimating the capital stock of China: 1952–2006. J Quant Tech
Econ 2008:17–31 [in Chinese].
[44] Wu Y. Productivity, efficiency and economic growth in China. New
York: Palgrave Macmillan; 2008.
[45] Mi Z-F, Pan S-Y, Yu H, Wei Y-M. Potential impacts of industrial structure on
energy consumption and CO2 emission: a case study of Beijing. J Cleaner Prod
2015;103:455–62.
[46] Mi Z, Zhang Y, Guan D, Shan Y, Liu Z, Cong R, et al. Consumption-based
emission accounting for Chinese cities. Appl Energy 2016.
[47] Intergovernmental Panel on Climate Change. The national greenhouse gas
inventories programme. In: Eggleston HS, Buendia L, Miwa K, Ngara T, Tanabe
K, editors. IPCC guidelines for national greenhouse gas
inventories. Japan: IGES; 2006.
[48] Liu et al. Reduced carbon emission estimates from fossil fuel combustion and
cement production in China. Nature 2015;524:335–8.
[49] Färe R, Grosskopf Jr S. CAP. Environmental production functions and
environmental directional distance functions. Energy 2007;32:1055–66.
[50] Chenery H, Robinson S, Syrquin M. Industrialization and growth: a
comparative study. Oxford University Press; 1986.
[51] Wang YZ, Chen JL. Development report of yangtze river delta in
2005. Shanghai: Shanghai People’s Publishing House; 2005 [in Chinese].
[52] Moore JH. A measure of structural change in output. Rev Income Wealth
1978:105–17.
[53] Coelli TJ, Rao DSP, O’Donnell CJ, Battese GE. An introduction to efficiency and
productivity analysis. 2nd ed. Springer; 2005.
[54] Poumanyvong P, Kaneko S. Does urbanization lead to less energy use and
lower CO2 emissions? A cross-country analysis. Ecol Econ 2010;70:
434–44.
[55] Li K, Lin B. Impacts of urbanization and industrialization on energy
consumption/CO2 emissions: does the level of development matter? Renew
Sustain Energy Rev 2015;52:1107–22.
[56] Elliott RJR, Sun P, Chen S. Energy intensity and foreign direct investment: a
Chinese city-level study. Energy Econ 2013;40:484–94.
[57] Ren S, Yuan B, Ma X, Chen X. International trade, FDI (foreign direct
investment) and embodied CO2 emissions: a case study of Chinas industrial
sectors. China Econ Rev 2014;28:123–34.
[58] Steinbuks J, Neuhoff K. Assessing energy price induced improvements in
efficiency of capital in OECD manufacturing industries. J Environ Econ Manage
2014;68:340–56.
[59] Lin B, Ouyang X. A revisit of fossil-fuel subsidies in China: challenges and
opportunities for energy price reform. Energy Convers Manage
2014;82:124–34.
[60] Li K, Lin B. Impact of energy conservation policies on the green productivity in
China’s manufacturing sector: evidence from a three-stage DEA model. Appl
Energy 2016;168:351–63.
[61] Wooldridge JM. Econometric analysis of cross section and panel data. MA: The
MIT Press; 2010.
[62] Arellano M, Bond S. Some tests of specification for panel data: Monte Carlo
evidence and an application to employment equations. Rev Econ Stud
1991;58:277–97.
500 K. Li, B. Lin / Applied Energy 187 (2017) 489–500

[63] Blundell R, Bond S. Initial conditions and moment restrictions in dynamic
panel data models. J Economet 1998;87:115–43.
[64] Baltagi BH. Econometric analysis of panel data. 3rd ed. England: John Wiley &
Sons Ltd.; 2005.
[65] Li HP, Ye H, Zhang JB. An empirical test on the relationship between income
inequality and environmental quality in China: extensions of environmental.
China Popul Resour Environ 2006;16:46–50 [in Chinese].
[66] Zhan J, Chen SY, Jefferson GH. Structural reform and industrial growth in
China. Econ Res J 2009:4–20 [in Chinese].
[67] Cheng Y. China’s overall gini coefficient since reform and its decomposition by
rural and urban areas since reform and opening-up. Soc Sci China 2007:45–60
[in Chinese].