An empirical analysis of the role of China’s exports on CO2 emissions

Applied Energy 104 (2013) 258–267

Contents lists available at SciVerse ScienceDirect

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

An empirical analysis of the role of China’s exports on CO2 emissions
Nyakundi M. Michieka ⇑, Jerald Fletcher, Wesley Burnett
Resource Management Program – Environmental and Natural Resource Economics, West Virginia University, Morgantown, WV 26505-6108, United States

h i g h l i g h t s
" We attempt to correct China’s coal consumption data.
" We discover Granger Causality running from exports to CO2 emissions.
" We discover Granger Causality running from exports to trade-openness.
" Policies aimed at controlling exports can control CO2 emissions.
" Policies aimed at controlling coal consumption will affect exports and CO2 emissions.

a r t i c l e

i n f o

Article history:
Received 14 March 2012
Received in revised form 5 October 2012
Accepted 20 October 2012

Keywords:
China
Emissions
Vector autoregression

International trade
Coal consumption

a b s t r a c t
China is one of the world’s most rapidly growing countries and the largest consumer of energy in the
world. As a result, China’s pollution emissions almost doubled from 2002 to 2007, and in 2006 it surpassed the United States to become the world’s top carbon dioxide emitter. Understanding the sources
of emissions is essential towards designing policies aimed at curbing carbon emissions in China. The
surge in China’s exports has been partially blamed for this increase in emissions. To understand the
sources of emissions, this study uses a vector autoregression model to examine the relationship among
exports, CO2 emissions, coal consumption and trade openness in China for the years 1970–2010. The
study uses a modified version of Granger Causality developed by Toda and Yamamoto [56]. The main
findings within the study indicate: (1) Granger Causality running from exports to emissions; (2) Granger
Causality running from coal consumption to exports; and (3) GDP determines future variability in exports
and CO2 emissions. Results suggest that governmental policies aimed at controlling coal consumption
could affect CO2 emissions and exports. Results from this study should assist in formulating policies to
mitigate both CO2 emissions and coal consumption.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction
China is one of the most rapidly growing countries in the world
and is the largest energy consumer [21]. In 2006, China surpassed
the US as the world’s top gross carbon dioxide (CO2) emitter with
6.1 billion tons of annual emissions and by 2008 had already outdistanced the US by 1.5 billion tons [40].1 China’s CO2 emissions
grew at 3.3% per year between 1990 and 1999, accounting for 13%
of global emissions. These emissions doubled over the next decade,
growing at 8.9% per year between 2000 and 2007, and accounting
for 17% of global emissions. Presently, China emits 21.3% of global
CO2 emissions [17,55]. The rapid expansion of the Chinese economy,
coupled with a coal-oriented energy structure, has made coal
⇑ Corresponding author.

E-mail addresses: (N.M. Michieka), jerry.
fl (J. Fletcher), (W. Burnett).
1
The term ‘‘CO2 emissions’’ and ‘‘emissions’’ are used interchangeably throughout
this paper.
0306-2619/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
/>
consumption a major source of emissions [63]. The country is moving from a predominantly agricultural economy to one that is
increasingly urbanized and industrialized [1]. Moreover, growth in
coal–fired electricity generation has been cited as a reason for the
surge in emissions.
New research indicates that about a third of all Chinese carbon
dioxide emissions are the result of producing goods for export to
developing and developed countries. Weber et al. [58] found that
in 2005, around one-third of Chinese CO2 emissions were generated by the production of goods for export while Wang and Watson
[57] concluded that net exports from China accounted for 23% of its
total CO2 emissions in 2004. Shui and Harris [50] estimated that in
2003, close to 14% of China’s CO2 emissions came from producing
goods for export. This problem is likely to persist owing to the rising popularity of China’s exports which account for 10% of global
exports. Theory implies that for some countries a comparative
advantage for production exists directly because of differences in
environmental regulations. The pollution heaven hypothesis posits
that ‘‘pollution havens’’ will attract polluting industries that


N.M. Michieka et al. / Applied Energy 104 (2013) 258–267

relocate from more stringent locales, although studies have found
little evidence to support this theory [13,23,9].
China’s exports are inexpensive and attractive because of the

low wages and low cost of raw materials that minimize production costs. In addition, costly pollution controls are often not
implemented in China [57]. This partially explains why China
has overtaken Germany to become the world’s top exporter
[62]. Given the link between China’s exports and carbon emissions, studies have sought to analyze this relationship using economy-wide modeling and econometric methods. These studies
have not adequately addressed this analysis using accurate data;
i.e., some studies have employed data with questionable reliability. This study contributes to the existing literature by analyzing
the relationship between exports and emissions using improved
coal consumption data in China as a number of studies have cast
doubt on the validity of China’s reported coal consumption figures. This issue is circumvented by employing a data-driven interpolation technique that gives more accurate figures over the past
three decades. Given these accurate estimates the relationship between exports, coal consumption, GDP, CO2 emissions and trade
openness is modeled within a multivariate time series model.
The findings, from a global point of view, may provide ideas for
emission reduction strategies that China can adopt. The analysis
will forecast the variability of emissions and coal consumption
to 10 years, while other studies have focused on constructing
forecasts of longer horizons of 50 or 100 years [48]; McCarthy
[38].
This study is important in some ways. Understanding the key
drivers behind China’s growing CO2 emissions is critical for designing climate policies. They provide an insight into how other emerging economies like India and Vietnam may develop low emission
economies in the future. In addition, CO2 emissions in China
have a global impact, making its engagement in global climate
change mitigation essential. Numerous studies have highlighted
the domestic and global environmental impacts of China’s
coal usage [25,53,69]. Finally, given China’s integration in the
global economy, the study of China’s economic development and
energy infrastructure is vital to the health of global economic
development.
This paper also seeks to investigate the relationship between
China’s exports and CO2 emissions. A fraction of emissions are produced by the manufacturing, electric power and transportation
industries needed for producing goods for export. Others are

embodied in the exports themselves. China’s energy infrastructure
is primarily coal based which accounts for 74% of total national energy consumption. Consequently, large quantities of greenhouse
gases are emitted contributing to global climate change. Further,
trade liberalization has been touted as a reason for China’s growth
in GDP. The adoption of the open-door policy in China propelled
the country to become one of the fastest growing economies
in the world. The direction of causality between economic
growth and trade openness has been a subject of extensive debate.
Thus, the relationship between trade openness and GDP is sought
to be examined. Trade openness is defined as the ratio of external
trade (imports plus exports) to GDP as used in the literature [34].
Therefore, a variable for trade openness is explored in this
analysis.2
The remainder of the study is organized as follows. The next
section reviews the literature pertaining to this study while the
third section presents the model and data. Section 4 presents the
results and Section 5 offers the conclusions.

2
Other energy sources of energy contribute to emissions but this study seeks to
examine the contribution of coal and exports to CO2 emissions. Future studies in this
area may look into the effect of these sources on emissions.

259

2. Literature review
Energy consumption in China has attracted considerable research interest due to the environmental ramifications caused by
the extensive use of coal, which has propelled high economic
growth for the past two decades. This interest spans many subbranches of economics. Past studies have looked into sources and
ways of reducing CO2 emissions while others have analyzed the

relationship between emissions and exports. The methods used
vary – a majority employ economy-wide modeling that include input–output techniques while others use econometric techniques
for analysis and forecasting. Given the breadth of interests, this literature review explores different literatures which relate directly
or indirectly to this particular study. Since this analysis is broad,
we present a brief summary in Table 1 of the major themes within
the literature and key findings within particular studies.3
A number of regional and nationwide studies have sought to
analyze the sources of emissions at the global and regional levels.
Using the Kaya identity, Raupach et al. [45] found that population
and GDP are the main drivers of emissions. The Kaya identity is a
decomposition analysis that is used to explore the differing trends
of factors contributing to carbon dioxide emissions [26]. The identity is comprised of three primary factors: economic growth, population growth, and energy consumption. A majority of studies on
China concluded that structural changes across sectors of the economy are one of the main causes of emissions before the 1990s,
whereas technological change within sectors has had a greater impact more recently [37]. Streets et al. [54] used an atmospheric
transport model to study emissions in the Pearl River Delta region.
This region is known to produce goods for export to North America,
Europe and Asia. Streets et al. [54] discovered that pollution in the
area is caused by the manufacturing and transportation industries.
Their study found that the region is responsible for 5–30% of the
ambient concentrations of various emissions. A study by Gregg
et al. [17] recommended that controlling fossil fuel combustion
from electricity power generation and the cement manufacturing
in China would reduce emissions.
Fang et al. [14] found that simple improvements to small industrial boilers could reduce CO2 emissions in China by as much as
63 Mtons and save 34 Mtons of coal at an estimated cost of $10
per ton of CO2. Liang et al. [32] investigated stakeholder and public
perceptions of deploying carbon capture and sequestration (CCS)
technologies in China. They found that a majority of those surveyed perceived climate change to be a major problem and viewed
CCS as a necessary mechanism to reduce CO2 emissions in China.
Choi et al. [7] use a data envelope analysis to estimate the potential

reductions of CO2 emissions through production and technological
efficiencies. They find that by adopting better efficiencies, CO2
emissions can, on average, be reduced by approximately 56.1M
tons in each province and 1683 Mtons nationwide; but, the efficiencies will be easier to achieve in the well-developed eastern
part of the country as opposed to the less well-developed western
part.
Two fairly recent studies highlight how the adoption of biofuels
in China may reduce greenhouse gas (GHG) emissions including
CO2. Hu et al. [20] consider the adoption of cassava-based ethanol
as an alternative automotive fuel in China and conduct a lifecycle
analysis (LCA) to examine the impacts on energy consumption
and CO2 emissions from the adoption of such a policy. They find
that 10% mandate of blending the ethanol with conventional automotive fuel could significantly lower CO2 emissions, but a nationwide program would require an untenable 42% of total farmland in

3
This summary is not meant to be an exhaustive review of the literature, but rather
key findings as they relate to this particular study.


260

N.M. Michieka et al. / Applied Energy 104 (2013) 258–267

Table 1
Existing studies on China’s emissions.
Author(s)
1. Cause of emissions
1.1. Raupach et al. [45]
1.2. Streets et al. [54]
1.3. Gregg et al. [17]

Approach

Region

Findings

Time series analysis and
kaya identity
Atmospheric transport
model
Time series analysis

China, region and the world

Population and GDP were the main drivers of emissions

Pearl river delta region in China

Pollution in the region is caused by the manufacturing and
transportation industries
Fossil fuel combustion from electricity power generation and
the cement manufacturing in China

2. Economy-wide modeling techniques
2.1. Lin and Polenske [33]
Structural decomposition
analysis
2.2. Weber et al. [58]
Input–output model

2.3. Chen and Chen [6]
Ecological input–output

China

China
China
China

2.4. Yunfeng and Laike [66]

Input–output

China and the European Union

2.5. Chung et al. [8]

Index decomposition
analysis and energy input–
output model

South Korea

Integrated econometric
model
Using dynamic models with
spatial dependence and
provincial – level panel data

China

Malaysia

4.2. Sami [47]

Toda and Yamamoto [56],
and Dolado and Lütkepohl
[11], VAR
VECM

4.3. Shiu and Lam [49]

ECM

China

3. Forecast emissions
3.1. ZhiDong [68]
3.2. Auffhammer and
Carson [3]
4. Time series analysis
4.1. Lean and Smyth [30]

China

Japan

China in order to meet the fuel demand. Xunmin et al. [65] extend
the analysis of Hu et al. [20] by considering six different biofuel
pathways. Like the previous authors, Xunmin et al. [65] conduct

a LCA to examine the impacts on energy consumption and GHG
emissions. They argue that different biofuels pathways are required for China because each provincial region is geographically
unique for the country. Ultimately, they find that these biofuel
pathways can reduce both energy consumption and GHG
emissions.
The Hecksher–Ohlin theory of trade suggests that given free
trade, a developing country will specialize in the production of
goods in which it is abundantly endowed [15]. This endowment
then will determine how intensive certain factors will be used in
the developing country. China’s endowment constitutes an abundant labor force. China’s adoption of economic reforms in the latter
1970s, coupled with cheap labor, led to a huge increase in manufacturing, which in turn induced trade with foreign nations. While
economic growth has benefitted from this large increase in manufacturing, a negative side effect is the pollution intensiveness of
manufacturing. Not only is manufacturing pollution intensive in
output, but also in the pollution intensiveness in the coal-fired,
electricity inputs. These combined intensities have led to the increase in China’s CO2 emissions over the past few decades.
Researchers have employed economy-wide modeling techniques such as input–output and structural decomposition analyses to study the relationship between exports and emissions. Lin
and Polenske [33] used a structural decomposition analysis to explain changes in China’s energy use between 1981 and 1987. They
found that increased expenditures on capital products were the
main cause of a rise in emissions. They added that emissions could
be reduced by importing more goods than are currently exported.

Increased expenditures on capital products were the main
cause of a rise in emissions.
The export industry generated 33% of China’s total emissions.
International trade played a significant role in redistributing
carbon emissions.
Machinery and manufacturing sectors substantially
contributed to emissions embodied in exports
The intermediate demand sector was largely responsible for
GHG emissions within the country.

China will sustain a 6% economic growth rate in the coming
years, presenting challenges for CO2 emission reductions.
The magnitude of the projected increase in Chinese emissions
is several times larger than reductions embodied in the Kyoto
Protocol.
Bi-directional Causality running between aggregate output
and electricity consumption. They also found Granger
Causality running from exports to aggregate output.
Causality running from exports and real GDP per capita to
electricity consumption. The study also established a
cointegrating relationship among electricity consumption,
economic growth and exports.
Real GDP and electricity consumption for China are
cointegrated. There is Granger Causality running from
electricity consumption to real GDP

Weber et al. [58] used a standard input–output model of the
Chinese economy which reflected the amount of money flowing
between sectors and found that in 2005, the export industry generated 33% of China’s total emissions. Chen and Chen [6] studied
carbon emissions and resource use in the Chinese economy using
an ecological input–output model. They found that international
trade plays a significant role in redistributing carbon emissions.
More specifically, 3.59 or 5.54 gigatonnes of carbon dioxide
equivalents of GHG emissions are embodied in exported products.
A recent study by Yunfeng and Laike [66] examined the quantity of
CO2 emissions embodied in the trade between China and the
European Union. The paper identifies the sectors contributing
most to these embodied CO2 emissions using the input–output
approach; they find that the machinery and manufacturing

sectors substantially contribute to emissions embodied in exports.
Chung et al. [8] conduct a similar input–output analysis but instead
consider South Korea. Through the use index decomposition
analysis and an energy input–output model they found that the
intermediate demand sector including the industrial sector
accounted for approximately 85% of final energy demand, and as
a result the intermediate demand sector was largely responsible
for GHG emissions within the country.
In the future, China’s CO2 emissions are projected to grow faster
than the economy Hohne et al. [19]. It is therefore important to
forecast emissions to predict future paths of CO2 emissions in
China. Several econometric techniques have been employed for
forecast analyses. ZhiDong [68] used an integrated econometric
model to perform a long-term simulation study in China for a
30 year period and found that China will sustain a 6% economic
growth rate in the coming years, presenting challenges for CO2
emission reductions. Using a panel dataset from 1985 to 2004,


261

N.M. Michieka et al. / Applied Energy 104 (2013) 258–267

Auffhammer and Carson [3] explore alternative econometric specifications for forecasting China’s CO2 emissions. Using dynamic
models with spatial dependence and provincial–level panel data,
they found that the magnitude of the projected increase in Chinese
emissions (out to 2015) is several times larger than reductions
embodied in the Kyoto Protocol.
A number of studies have incorporated time series analysis to
study interactions among variables that fuel increasing emissions.

Lean and Smyth [30] examined the causal relationship among
aggregate output, electricity consumption, exports, labor and capital using a multivariate model for Malaysia. They found bi-directional Granger Causality running between aggregate output and
electricity consumption. They also found Granger Causality running from exports to aggregate output. A study by Sami [47] reviewed the relationship among electricity consumption, exports
and real income per capita in Japan. Using a vector error correction
model (VECM), their results indicated causality running from exports and real GDP per capita to electricity consumption. The study
also established a cointegrating relationship among electricity consumption, economic growth and exports. A similar study by Shiu
and Lam [49] established similar findings in their study, and, more
specifically that electricity consumption and economic growth in
China are cointegrated.
Halicioglu [18] finds that carbon emissions are determined by
energy consumption, income and foreign trade. Other studies have
stated that coal consumption and emissions have been inextricably
linked in China for decades. Increased coal consumption accelerates environmental pollution, so there is very important practical
significance to study the causal relationship between the two Ma
[36]. Coal is also used to produce electricity which plays a huge
role in China’s growth and GDP. Therefore the method employed
by Wolde-Rufael [60] and Li et al. [31] is used to test the relationship between coal consumption and GDP. In addition, empirical
evidence from cross country comparisons suggests that there is a
relationship between economic growth and environmental outcomes [46].
The studies here offer a comprehensive overview of the structure of China’s emissions and their sources; however, there is a
paucity of studies that use time series analysis to determine the
influence of exports on China’s emissions.
3. Model and data description
3.1. Model
The vector autoregressive (VAR) system is constructed using the
following five variables: exports (xt), emissions (et), coal consumption (ct), GDP (mt) and trade-openness (pt). The system of equations
with one lag is expressed as:

X t ẳ a10 ỵ a11 xt1 ỵ a12 et1 ỵ a13 ct1 ỵ a14 mt1 þ a15 pt1 þ ext
et ¼ a20 þ a21 xt1 þ a22 et1 þ a23 ct1 þ a24 mt1 þ a25 pt1 ỵ eet

ct ẳ a30 ỵ a31 xt1 ỵ a32 et1 ỵ a33 ct1 ỵ a34 mt1 ỵ a35 pt1 ỵ ect
mt ẳ a40 ỵ a41 xt1 ỵ a42 et1 ỵ a43 ct1 ỵ a44 mt1 ỵ a45 pt1 þ emt
pt ¼ a50 þ a51 xt1 þ a52 et1 þ a53 ct1 þ a54 mt1 þ a55 pt1 þ ept :
Eq. (1) can be rewritten in matrix notation as:

3

2

3
a11
a10
7 6
6 7 6
a21
6 et 7 6 a20 7 6
7 6
6 7 6
6
6 ct 7 ẳ 6 a30 7 ỵ 6 a31
7 6
6 7 6
7
6 7 6
.
4 mt 5 4 a40 5 6
4 ..
pt
a50
a51

xt

2

Y t ẳ A0 ỵ A1 yt1 ỵ    ỵ Ar ytr ỵ et ;

3ị

where A0 is a (5  1) column vector of intercepts and A1 a (5  5)
matrix of estimated coefficients on the first lag of the explanatory
variables. The system of equations can be extended to multiple lags,
as follows:

xt ẳ a10 ỵ
ỵ

r
X

r
X

r
X

jẳ1

jẳ1

jẳ1

a11;j xtj þ

a12;j etj þ

r
X

r
X

j¼1

j¼1

a14;j mtj þ
r
X

et ¼ a20 þ

j¼1
r

a23;j ctj

j¼1

a25;j ptj þ eet

j¼1

ct ¼ a30 þ

r
X

r
X

r
X

j¼1

j¼1

j¼1

a31;j xtj þ

a32;j etj þ

r
X

r
X

j¼1

j¼1

a34;j mtj ỵ

mt ẳ a40 ỵ

a33;j ctj
4ị

a35;j ptj ỵ ect

r
X

r
X

r
X

jẳ1

jẳ1

jẳ1

a41;j xtj ỵ

a42;j etj ỵ

r
X

r
X

jẳ1

jẳ1

a44;j mtj ỵ

pt ẳ a50 ỵ
ỵ

r
X

r
X

jẳ1

ỵ

a22;j etj ỵ

jẳ1

ỵ m a24;j mtj ỵ

ỵ

a15;j ptj ỵ ext

r
X

a21;j xtj ỵ

a13;j ctj

a43;j ctj

a45;j ptj ỵ emt

r
X

r
X

r
X

jẳ1

jẳ1

jẳ1

a51;j xtj ỵ

a52;j etj ỵ

r
X

r
X

jẳ1

jẳ1

a54;j mtj ỵ

a53;j ctj

a55;j ptj ỵ ept

which implies the following generalization,

yt ẳ A0 ỵ A1 yt1 ỵ    ỵ Ar ytr ỵ et :

ð5Þ

Given multiple lags a generalization of the coefficient matrix, Ar,
would indicate the rth lag of the explanatory variables [12].

To test the null hypothesis that there is ‘‘Granger Causality’’
from exports to emissions, the null: H0:a21j = 0 is tested, where
the a21j’s are the coefficients of xt1, xt1, . . ., xtj respectively in
the second equation in the VAR system. The causality from emissions to exports can be established through rejecting the null
hypothesis which requires finding the significance of the Modified
Wald (MWald) statistic for the group of the lagged independent
variables identified above. To complement the VAR, vector decompositions were developed to check whether the variables affect one
another in the ‘‘future,’’ which assist in confirming the results of
Granger Causality. For completeness, the impulse response
functions are presented to provide a visual depiction of variable’s
responses to shocks.
3.2. Coal consumption data

ð1Þ
2

The multivariate generalization of the process is:

3

a12 a13 a14 a15 2 xt1 3 2 ext 3
7 6
7
a22 a23    a25 7
76
et1 7 6 eet 7
76
7 6
6
7

a32 a33    a35 76 ct1 7 ỵ 6 ect 7
7:
7 6
7
6
7 6
7
..
..
..
.. 7
76
e
c
5
4
4
mt 5
.
.
. . 5 t1
ept
a52 a53    a55 pt1
ð2Þ

The past literature has reported that data on China’s coal consumption suffer from under-reporting. Sinton [52] offers a comprehensive overview relating to the accuracy and reliability of China’s
coal statistics. The point of contention in the data relates to the
period between late 1990s and early 2000s. During this time, official energy statistics showed a significant decrease in coal consumption despite increases in Chinese CO2 emissions. Further,
satellite data suggest that there was significant under-reporting
of coal consumption, which lead Akimoto et al. [2] to conclude that

the official statistics should not be used for emission inventories.


262

N.M. Michieka et al. / Applied Energy 104 (2013) 258–267
Table 2
Summary table.

COAL CONSUMPTION
(Million tonnes oil equivalent)

1800
1600
1400
1200
1000

e (emissions
in million
metric
tonnes)

c (coal
consumption
in million
tonnes oil
equivalent –
Mtoe)

x
(exports in
current
US$)

m (GDP
in
current
US$)

p (trade
openness)

2996522
2460744
8241707
771617.5
3.91497E+12
1978629

610.863
529.879
1713.5
196.494
148103.365
384.842

2.86E+11
5.74E+10
1.75E+12

2.39E+09
2.21E+23
4.70E+11

1.03E+12
3.57E+11
5.93E+12
9.15E+10
1.99E+24
1.41E+12

0.326138
0.309783
0.704653
0.053142
0.036063
0.189903

800
600
400
Interpolated coal consumption
Reported coal consumption

200
0
1970

1975

1980

1985

1990

1995

2000

2005

2010

YEAR

Mean
Median
Maximum
Minimum
Variance
Standard
deviation

Fig. 1. China coal consumption 1970–2010.

To account for the potential under-reporting for that period, coal
consumption was scaled up to reflect a more accurate historic
trend.
Values for coal consumption for the period between 1990s and

early 2000s were to be plotted. However, the problem of fitting the
curve through finite sequence of points while preserving the shape
of the data was experienced. The literature states that a piecewise
polynomial curve offers much more flexibility than a single polynomial in preserving the shape of the data [16]. In addition, piecewise cubic polynomials are used because their plots are smooth
and are the lowest degree polynomials that support inflection
points. Given the nature of the data and shape of the curve, data
for the years 1995–2008 was constructed using a piecewise cubic
hermite interpolating polynomial. This method was also employed
by Auffhammer and Carson [3]. The interpolated data is shown in
Fig. 1.
3.3. The data
Data on coal consumption was obtained from BP statistical data
[4]. Real gross domestic product (GDP), imports and exports were
obtained from the World Bank indicators database [61]. Data on
emissions was obtained from the Carbon Dioxide Information
Analysis Center [5]. The data set ranges from 1970 to 2010. In this
study, the relationship between exports, emissions, coal consumption, GDP and trade openness are investigated within a (VAR)
framework. A statistical summary of these variables is shown in
Table 2.
4. Empirical results
The Granger no-causality test method applied in this analysis is
based upon the work of Toda and Yamamoto [56]. This procedure
is expected to improve the standard F-statistic in the causality testing procedure. In determining whether some parameters of the
model are jointly zero, the traditional F-test is not valid when
the variables are integrated or cointegrated; in this case, the joint
distribution of the variables is not characterized by a normal distribution. In other words, if the data is integrated or cointegrated, the
usual tests for exact linear restrictions on the parameters (e.g. the
Wald test) do not have their usual asymptotic normal distributions. The procedure proposed by Toda and Yamamoto [56] ensures that the usual test statistics for Granger Causality have
standard asymptotic distributions. This procedure can be used to
avoid the pre-testing distortions associated with prior tests for

non-stationarity and cointegration. The basic idea of the approach
is to artificially augment the correct order, k, by the maximal order
of integration, dmax [44]. Once this is done, a (k + dmax)th order of

VAR is estimated and the coefficients of the last lagged dmax vectors
are ignored. To use this approach, the true lag length (k) and the
maximum order of integration (dmax) of the series need to be obtained. The advantage of using the Toda and Yamamoto [56] method is that it does not require a priori knowledge of cointegration
within the system [67].
To ensure that the time series within the VAR model satisfy the
assumption of normality, a number of stationarity tests were conducted. Time series data is often characterized by unit root [39]. In
both raw and log-transformed data, it is found that all the variables
have a non-zero mean. Tests for unit roots were conducted using
the Augmented Dickey and Fuller [10], Phillips–Perron [43] and
Kwaitkowski–Phillips–Schmidt–Shin [27]. The Phillips Perron
(PP) test was used to complement the standard augmented dickey
fuller (ADF) test in testing for unit root. The PP procedure tests for
unit roots in the presence of structural change [43]. The Kwaitkowski–Phillips–Schmidt–Shin test (KPSS) was also used to complement the ADF and PP tests; KPSS tests the null hypothesis of
non-stationarity against the alternative of trend stationarity [41].
The PP and KPSS tests are used together with the ADF tests for
the sake of robustness. The results of the unit root tests are reported in Table 3. Test results indicate that all the time series were
I(1) except for coal consumption which was I(2).
The next step was to find out the appropriate lag length. The approach by Lütkepohl [35] was employed in which the optimal lag
length (mlag) is based upon the number of endogenous variables
in the system (m) and the sample size (T) according to the formula:
m  mlag = T1/3. With a sample size of 40 this rule implies a maximal lag length of one.4 Using the Toda and Yamamoto [56] approach, the Granger Causality tests were conducted using three
lags (Using k = 1 and dmax = 2) and the results presented in Table 4.
The Granger Causality results appearing in Table 4 indicate oneway causality from coal consumption to exports. They also indicate
unidirectional ordering from coal consumption to emissions, confirming hypotheses that past and present values of coal consumption help explain emissions. Also, one-way causality running from
GDP to coal consumption was discovered. Regarding the energygrowth literature, results favored the conservation hypothesis
which asserts that economic growth leads energy consumption,

implying that energy conservation policies may not adversely affect GDP. No causality was discovered between trade openness
and economic growth.
The results appearing on Table 4 also indicate causality running
from exports to emissions; this finding is consistent with other
findings in the literature [42,58,64,66]. The direction of causality
between exports and emissions implies that the government can

4
The unit root test results provide the value of dmax while the method by Lütkepohl
[35] is used to calculate k. The sum of these values was used to select the number of
lags to test for Granger Causality.


263

N.M. Michieka et al. / Applied Energy 104 (2013) 258–267
Table 3
Unit root test results.
Variable

Series

C

CT

C

CT

ETA (mu)

ETA (tau)

Ln_c

Ln_c
Ln_Dc
Ln_D2c

1.9068
0.9245
6.5066***

0.159317
1.7214
6.7633***

2.8454*
0.9864
6.5066***

0.6201
1.7214
6.7633***

0.7987***
0.5333**
0.2031

0.1949**
0.09516
0.0826

Ln_e

Ln_e
Ln_De

0.4469
4.0062***

2.2769
3.9952**

0.0253
4.0062***

1.9746
3.9952**

0.7938***
0.0978

0.0862
0.0915

Ln_x

Ln_x

Ln_Dx

0.4107
4.8267***

2.3579
4.7635***

0.4106
4.6908***

2.6266
4.5789***

0.7964***
0.1156

0.1192*
0.1158

Ln_m

Ln_m
Ln_Dm

2.5210
5.1810***

0.0875
3.9261**

2.9174*
5.2379***

0.0034
5.9409***

0.7841***
0.4877**

0.1993⁄⁄
0.0789

Ln_p

Ln_p
Ln_Dp
Without trend
1%
3.58
0.216

2.4521
4.5250***

2.7305
4.7969***

3.4755**
4.2828***

0.7491***
0.3421

0.7491***
0.34205

5%
3.22
0.146

10%
2.60
0.119

1.4542
5.2529***
With constant and trend
1%
4.15
0.739

5%
3.80
0.463

10%
3.18
0.347

ADF and PP
KPSS

ADF

PP

KPSS

ADF = Augmented dickey fuller Test, PP = Phillips Perron Test, KPSS = Kwiatkowski–Phillips–Schmidt–Shin test; C = Constant, CT = Constant and Trend.
Denote significance at 10%.
**
Denote significance at 5%.
***
Denote significance at 1%.
*

Table 4
VAR Granger Causality Tests.

*
**

Hypothesis:

F-statistic

Prob.

Ln_x does not Granger Cause Ln_c

Ln_c does not Granger Cause Ln_x
Ln_c does not Granger Cause Ln_e
Ln_e does not Granger Cause Ln_c
Ln_c does not Granger Cause Ln_m
Ln_m does not Granger Cause Ln_c
Ln_c does not Granger Cause Ln_p
Ln_p does not Granger Cause Ln_c
Ln_x does not Granger Cause Ln_e
Ln_e does not Granger Cause Ln_x
Ln_m does not Granger Cause Ln_x
Ln_x does not Granger Cause Ln_m
Ln_p does not Granger Cause Ln_x
Ln_x does not Granger Cause Ln_p
Ln_m does not Granger Cause Ln_e
Ln_e does not Granger Cause Ln_m
Ln_p does not Granger Cause Ln_e
Ln_e does not Granger Cause Ln_p
Ln_p does not Granger Cause Ln_m
Ln_m does not Granger Cause Ln_p

1.0290
10.2759
2.7051
0.5776
0.74482
2.6610
1.1696
0.9932
2.56188
1.6371

1.41354
1.3394
1.6119
3.1236
1.7111
1.2567
0.4046
1.1353
1.8222
1.5556

0.3933
0.0000⁄⁄⁄
0.0623⁄
0.6341
0.5336
0.0654⁄
0.3371
0.4089
0.0728⁄
0.2009
0.2575
0.2795
0.2066
0.0399**
0.1851
0.3063
0.7507
0.3501
0.1637

0.2199

Denote significance at 10%.
Denote significance at 5%.
Denote significance at 1%.

***

regulate emissions by designing policies that regulate exports. In
addition, policies aimed at curbing coal consumption can also control exports and emissions. Finally, the results demonstrate causality running from exports to trade openness. Therefore, it is possible
to forecast the future levels of trade openness from the past levels
of exports. No relationship between trade openness and economic
growth was found.
The causality tests indicate only Granger Causality within the
sample period, and do not allow us to gauge the relative strength
of the Granger Causality among the series beyond the sample period. Thus, to complement the above analysis, the forecast error variance of exports, consumption, emissions, GDP and trade openness
were decomposed into proportions attributed to shocks in all variables in the system. This allows us to provide an indication of the
Granger Causality beyond the sample period [51]. The variance
decomposition results are presented in Table 5.

The cells in the variance decomposition represent percentages
of the forecast variance (error) in one variable at different time
periods induced by innovations of the other variables. These percentages help determine the relative contribution the innovations
make towards explaining movements in the other variables.5
Table 5 also shows that GDP and coal consumption have the greatest
effect on export variability over the forecast period. GDP explains
20.34% variability at five years and 24.76% variability at a ten year
horizon while coal consumption explains 14.81% and 19.19% of export variability at five and ten year horizons respectively. The shocks
explained by coal consumption confirm the Granger Causality tests.
Emissions have an increasing effect on export variability, explaining

6.83% and 15.26% at the end of five and ten year horizons. Trade
openness has a decreasing effect on export variability of exports
with time.
While coal consumption and exports appear to be an important
factor to emissions in China, the variance decomposition analysis
shows that they explain 4.6% and 2.8% of variability at the end of
the forecast period. Furthermore, GDP is the most important factor
in explaining emissions variability. It is beyond the scope of the paper to thoroughly examine the underlying reasons behind these
weak consumption–emissions and export-emission relationships
but it can be surmised that in the future, China will find more efficient ways of using of coal, which in turn affect electricity production and exports, and thereby reduce CO2 emissions. The table also
indicates that GDP has an increasing effect on coal consumption
explaining 11.59% and 44.44% of consumption variability in the
short and long-run respectively. In addition, exports explain
7.33% of consumption variability in the long run. The relatively
low contribution may be an indication that the causal relationship
between exports and coal consumption is relatively weak over the
long run compared to that of GDP and coal consumption. For the
forecast variance decomposition of trade openness, GDP has an
increasing effect on variability, explaining 52.69% at the end of
the tenth year – up from 43.74% in the fifth year. Exports also have
a decreasing effect on trade openness, explaining 16.29% at the end
of the forecast period. This result confirms the earlier Granger
Causality tests. Finally, Table 5 shows that trade openness has a
5
Each entry is the percentage of forecast error variance. Due to rounding, the
numbers in the rows do not always sum to 100.


264

N.M. Michieka et al. / Applied Energy 104 (2013) 258–267

Table 5
Variance decomposition results.
Variance decomposition of Ln_m:
Period

S.E.

1
0.008391
5
0.025876
10
0.05027
Variance decomposition of Ln_c:
1
0.048794
5
0.091488
10
0.113396
Variance decomposition of Ln_x:
1
0.117682
5
0.156152
10
0.193008
Variance decomposition of Ln_p:

1
0.087028
5
0.189721
10
0.264067
Variance decomposition of Ln_e:
1
0.111395
5
0.197998
10
0.242661
Cholesky ordering: Ln_m Ln_c Ln_x Ln_p Ln_e

Ln_c
0
0.002304
0.06292

Ln_e

Ln_x

0
1.087862
2.959298

0
0.048208

0.030711

97.22053
71.90651
37.31687

0
7.088608
10.33621

0
9.24577
7.331531

19.24216
14.81005
19.18808

0
6.834465
15.26063
0
1.716841
2.172582

15.71462
6.997893
9.509195
0.300299
2.849479

4.641847

92.93357
81.65356
69.37177

marginal effect on GDP, explaining 1.53% at the end of the forecast
period.
Next the generalized impulse response functions (GIRFs) were
generated as shown in Fig. 2. The GIRF’s represent the reactions
of the variables to shocks in the system. While the Toda and
Yamamoto [56] method tests the long-run Granger Causality relationships, it does not consider how variables respond to shocks in
other variables. The generalized impulse response function examines how a shock to one variable affects another, and how long the
effect lasts. Ordering of variables in the VAR system is important in
order to calculate the impulse response functions (IRFs) analyses.
Different ordering may result in different IRF results. The generalized GIRFs which are invariant to the ordering of the variables in a
VAR were employed [28]. The charts in Fig. 2 reflect the dynamic
properties of the system where without any shock, the response
plots would be flat. The horizontal line in GIRFs shows the time
period after the initial shock. The vertical line in GIRFs shows the
magnitude of response to shocks.
Fig. 2 shows that the response of the emissions path to a one
standard deviation shock in GDP and coal consumption is positive
over the forecast period, whereas the response of emissions to a
one standard deviation shock in exports drifts around zero over
the forecast period. This implies that GDP and coal will continue
to have a positive effect on emissions over the forecast period
while exports will have a marginal effect on the emissions path.
The emissions path is negative in response to a shock in tradeopenness, suggesting that opening trade in China may reduce
emissions.

The path of exports in response to a one standard deviation
shock in coal consumption is initially negative but then positive;
the effect levels off after the fifth period. The response path is reversed when observing the effect of a shock of trade openness on
exports. The response path is positive over the first four years before crossing zero and decreasing over time. The response of exports to a shock in emissions and GDP is positive, implying that
coal consumption and GDP positively influence the path of exports
in the long run. The path of coal consumption in response to a
shock in exports is negative until the ninth period when it approaches zero while its response from a GDP shock is positive over
the forecast period.
The variance of GDP was shocked with coal consumption and
trade-openness but the path was unresponsive. Of importance is

Ln_m
100
97.24468
95.41664

Ln_p
0
1.616946
1.530436

2.779469
11.58904
44.44783

0
0.170077
0.567567

60.59998

50.71922
35.1931

20.15787
20.34278
24.76041

0
7.293481
5.597784

50.1422
22.31577
16.28869

9.54676
43.73825
52.69731

0.017222
1.795379
2.789918

2.64086
6.113843
17.65177

24.59642
25.23124
19.33222

4.108052
7.587738
5.544688

the finding that the path of GDP from a shock in trade openness
is negative. This implies that some regions, especially those in
the east, may experience a decrease to GDP when exposed to further liberalization of trade. This may occur due to the lack of competitiveness in international markets as pointed by Jin [24].
Nevertheless the GDP path remains positive and unresponsive
when shocked by emissions and exports. The unresponsive nature
of GDP to shocks from all the variables paths confirm the Granger
Causality findings, implying that GDP is impacted by shocks outside the system. Finally, the response path of trade openness in response to GDP shock is negative over the forecast period, while
remaining unresponsive when shocked by emissions. This analysis
was conducted using the interpolated coal consumption data.6

5. Conclusion
China’s economic reforms have liberalized the economy, resulting in remarkable economic growth and energy consumption since
the late 1970s. Evidence that China has overtaken the United States
to take the number one spot has led to renewed calls for China to
act to reduce the environmental impact of its phenomenal growth
[29]. As China observes a rapid increase in emissions, its policy
makers may question why the country is criticized by the very consumers who import relatively inexpensive Chinese goods. It has
been argued that the steep rise in China’s emissions has been
fuelled by exports of cheap goods to the rest of the world.
This study employed a vector autoregressive analysis to investigate the link between China’s exports and carbon dioxide emissions. Based on the empirical analysis, Granger Causality running
from exports to emissions was discovered. These results imply that
the government should consider policies aimed at controlling exports to reduce emissions. For example, the country can implement
policies that place an environmental levy on exports to fund
domestic GHG mitigation programs. Such projects entail the
installation of emission reducing technologies in industries that
6

For a sensitivity analysis, coefficient estimates for the interpolated and the
original coal consumption data were compared [59]. All estimates except those for
exports were similar in signage, magnitude, and significance, implying that the
interpolation method was quite robust. Also, results obtained using the original and
interpolated data sets were compared but not presented to conserve space. These
findings are available from the author.


265

N.M. Michieka et al. / Applied Energy 104 (2013) 258–267

Response to Generalized One S.D.Innovations ±2 S.E.
Response of LN_GDP to LN_CONS

Response of LN_GDP to LN_EMISS

Response of LN_GDP to LN_EXPORTS

Response of LN_GDP to LN_TRADE

.2

.2

.2

.2

.1

.1

.1

.1

.0

.0

.0

.0

-.1

-.1
1

2

3

4

5

6

7

8

9

10

-.1
1

Response of LN_CONS to LN_EMISS

2

3

4

5

6

7

8

9

10

-.1
1

Response of LN_CONS to LN_EXPORTS

2

3

4

5

6

7

8

9

10

1

Response of LN_CONS to LN_GDP

.04

.04

.04

.02

.02

.02

.02

.00

.00

.00

.00

-.02

-.02

-.02

-.02

-.04
1

2

3

4

5

6

7

8

9

10

-.04
1

Response of LN_EXPORTS to LN_CONS

2

3

4

5

6

7

8

9 10

2

3

4

5

6

7

8

9

10

1

Response of LN_EXPORTS to LN_GDP

.15

.15

.15

.10

.10

.10

.10

.05

.05

.05

.05

.00

.00

.00

.00

-.05

-.05

-.05

-.05

-.10
1

2

3

4

5

6

7

8

9

10

-.10
1

Response of LN_TRADE to LN_CONS

2

3

4

5

6

7

8

9

10

2

3

4

5

6

7

8

9

10

1

Response of LN_TRADE to LN_EXPORTS

.2

.2

.2

.1

.1

.1

.1

.0

.0

.0

.0

-.1

-.1

-.1

-.1

-.2
1

2

3

4

5

6

7

8

9

10

-.2
1

Response of LN_EMISS to LN_CONS

2

3

4

5

6

7

8

9 10

2

3

4

5

6

7

8

9

10

1

Response of LN_EMISS to LN_GDP

.08

.08

.08

.06

.06

.06

.06

.04

.04

.04

.04

.02

.02

.02

.02

.00

.00

.00

.00

-.02

-.02

-.02

-.02

-.04
1

2

3

4

5

6

7

8

9

10

-.04
1

2

3

4

5

6

7

8

9

10

9

10

2

3

4

5

6

7

8

9

10

2

3

4

5

6

7

8

9

10

2

3

4

5

6

7

8

9

10

Response of LN_EMISS to LN_TRADE

.08

-.04

8

-.2
1

Response of LN_EMISS to LN_EXPORTS

7

Response of LN_TRADE to LN_GDP

.2

-.2

6

-.10
1

Response of LN_TRADE to LN_EMISS

5

Response of LN_EXPORTS to LN_TRADE

.15

-.10

4

-.04
1

Response of LN_EXPORTS to LN_EMISS

3

Response of LN_CONS to LN_TRADE

.04

-.04

2

-.04
1

2

3

4

5

6

7

8

9

10

1

2

3

4

5

6

7

8

9

10

Fig. 2. Generalized impulse response functions. LN_CONS is Ln_c; LN_GDP is Ln_m; LN_EXPORTS is Ln_x; LN_TRADE is Ln_p and LN_EMISS is Ln_e.

manufacture goods for export. Another potential policy may
encourage foreign direct investment in domestic, energy efficient
industries which emit less CO2 emissions. Also, unidirectional
causality running from coal consumption to exports, and coal
consumption to emissions was found. China can consider marketbased mechanisms, such as cap-and-trade, which reduce coal
consumption and consequently reduce emissions. Other policies
include renewable energy strategies or the use of clean coal
technologies in the formulation of a long–term emission reduction

portfolio. Vector error decomposition analysis revealed that GDP
had the greatest effect on exports, emissions and coal consumption
variability.
Predictions indicate that the increase in greenhouse gas emissions from 2000 to 2030 in China alone will nearly equal the increase from the entire industrialized world. It is important for
China to take a lead in reducing CO2 emissions [22]. Their efforts
in taking responsibility for reducing the carbon emissions will resonate across countries following similar developmental patterns.
In addition, China can invest returns from exports on projects that

promote use of renewables to mitigate emissions and further
achieve better human health from reduced air pollution.
One limitation of this study is that it uses interpolated data for
coal consumption which may not reflect actual trends. Future work
should consider a different technique improving this data set. Another limitation is that the VAR is a reduced form model; therefore,
IRFs may not capture shocks to the true underlying innovations.

Disclaimer
This report was prepared as an account of work sponsored by an
agency of the United States Government. Neither the United States
Government nor any agency thereof, nor any of their employees,
makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed,
or represents that its use would not infringe privately owned
rights. Reference herein to any specific commercial product,


266

N.M. Michieka et al. / Applied Energy 104 (2013) 258–267

process, or service by trade name, trademark, manufacturer, or
otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors

expressed herein do not necessarily state or reflect those of the
United States Government or any agency thereof.
Acknowledgements
We are very grateful to two anonymous referees whose constructive comments have helped to improve upon the quality of
the paper. We are also grateful to the Editor of the Journal, Prof. Jerry Yan, for his encouragement. Errors and omissions, if any, are our
own.
This material is based upon work supported by the Department
of Energy under Award Number DE-FC26-06NT42804.
References
[1] Adams FG, Shachmurove DY. Modeling and forecasting energy consumption in
China: implications for Chinese energy demand and imports in 2020. Energy
Econ 2008;30(3):1263–78.
[2] Akimoto H, Ohara T, Kurokawa J-I, Horii N. Verification of energy consumption
in China during 1996–2003 by using satellite observational data. Atmos.
Environ. 2006;40(40):7663–7.
[3] Auffhammer M, Carson RT. Forecasting the path of China’s CO2 emissions:
offsetting kyoto – and then some. Berkeley: University of California; 2006.
Paper No. 971.
[4] British Petroleum (BP), 2010. BP statistical review of world Energy; 2010.
< />contentId=7068555>.
[5] Carbon Dioxide Information Analysis Centre (CDIAC); 2010. cdiac.ornl.gov/trends/emis/tre_prc.html>.
[6] Chen GQ, Chen ZM. Carbon emissions and resources use by chinese economy
2007: a 135-sector inventory and input–output embodiment. Commun
Nonlinear Sci Numer Simul 2010;15(11):3647–732.
[7] Choi Y, Zhang N, Zhou P. Efficiency and abatement costs of energy – related
CO2 emissions in China: a slacks-based efficiency measure. Appl Energy
2012;98:198–208.
[8] Chung W-S, Tohno S, Choi K-H. Socio-technological impact analysis using an
energy IO approach to GHG emissions issues in South Korea. Appl Energy

2011;88(11):3747–58.
[9] Cole M. Trade, the pollution haven hypothesis and environmental Kuznets
curve: examining the linkages. Ecol Econ 2004;48(1):71–81.
[10] Dickey DA, Fuller WA. Distribution of the estimators for autoregressive time
series with a unit root. J Am Stat Assoc 1979;74(366):427–31.
[11] Dolado JJ, Lütkepohl H. Making Wald test work for cointegrated VAR systems.
Econ Rev 1996;15:369–86.
[12] Enders W. Multiequation time-series models. Applied econometric time
series. John Wiley & Sons; 2004.
[13] Eskeland G, Harrison A. Moving to greener pastures? multinationals and the
pollution haven hypothesis. J Dev Econ 2003;70:1–23.
[14] Fang J, Zeng T, Yang LIS, Oye KA, Sarofim AF, Beer JM. Coal utilization in
industrial boilers in China – a prospect for mitigating CO2 emissions. Appl
Energy 1999;63(1):35–52.
[15] Flam H, Flanders M. Heckscher–Ohlin trade theory/Eli F. Heckscher and Bertil
Ohlin; Translated, In: Flam Harry, June Flanders M. editors. Cambridge,
Massachusetts.:MIT Press; 1991.
[16] Goodman TNT. Shape preserving interpolation by curves. Algorithms for
approximation IV. In: Proceedings of the 2001 international symposium.
England: University of Huddersfield; 2001.
[17] Gregg JS, Andres RJ, Marland G. China: emissions pattern of the world leader in
CO2 emissions from fossil fuel consumption and cement production. Geophys
Res Lett 2008;35:1–5.
[18] Halicioglu F. An econometric study of CO2 emissions, energy consumption,
income and foreign trade in Turkey. Energy Policy 2009;37(3):1156–64.
[19] Hohne N, Hare B, Schaeffer M, Chen C, Rocha M, Vieweg M, Moltmann S. China
Emission Paradox: Cancun Emissions Intensity Pledge to be Surpassed but
Emissions Higher. Germany: Climate Action Tracker, Koln; 2011.
[20] Hu Z, Fang F, Ben D-F, Pu G, Wang C. Net energy, CO2 emission, and life-cycle
cost assessment of cassava-based ethanol as an alternative automotive fuel in

China. Appl Energy 2004;78(3):247–56.
[21] International Energy Agency (IEA). China overtakes the United States to
become world’s largest energy consumer; 2010a. [accessed August 20, 2011].
< />[22] International Energy Agency (IEA). World Energy Outlook 2010. Paris; 2010b.
[23] Javorcik B, Wei S. Pollution havens and foreign direct investment: dirtysecret
or popular myth? The B.E. journal of. Econ Anal Policy 2005;3(2).
[24] Jin JC. On the relationship between openness and growth in china: evidence
from provincial time series data. World Econ 2004;27(10):1571–82.

[25] Johnson TM, Li J, Jiang Z, Taylor RP. China: issues and options in greenhouse
gas emissions control, report of a joint study team from the national
environmental protection agency of China, the state planning commission of
China, united nations development programme, and the world bank.
Washington (DC): World Bank; 1996.
[26] Kaya Y, Yokobori K. Environment, energy, and economy: strategies for
sustainability. Tokyo: Bookwell; 1999.
[27] Kwiatkowski D, Phillips PCB, Schmidt P, Shin Y. Testing the null hypothesis of
stationarity against the alternative of a unit root: how sure are we that
economic time series have a unit root? J Econ 1992;54(1–3):159–78.
[28] Koop G, Pesaran MH, Potter SM. Impulse response analysis in nonlinear
multivariate models. J Econ 1996;74(1):119–47.
[29] Kuby M, He C, trapido-Lurie B, Moore N. The changing structure of energy
supply, demand, and CO2 emissions in China. Ann Assoc Am Geogr
2011;101(4):795–805.
[30] Lean HH, Smyth R. Multivariate granger causality between electricity
generation, exports, prices and GDP in Malaysia. Energy 2010;35:3640–8.
[31] Li J, Song H, Geng D. Causality relationship between coal consumption and
GDP: difference of major OECD and non-OECD countries. Appl Energy
2008;85(6):421–9.
[32] Liang X, Reiner D, Li J. Perceptions of opinion leaders towards CCS

demonstration projects in China. Appl Energy 2011;88(5):1873–85.
[33] Lin X, Polenske KR. Input-output anatomy of China’s Energy use changes in the
1980s. Econ Syst Res 1995;7(1):67–84.
[34] Liu X, Song H, Romily P. An empirical investigation of the causal relationship
between openness and economic growth in China. Appl Econ 1997;29(12):
1679–86.
[35] Lütkepohl H. New introduction to multiple time series analysis. 1st ed. New
York: Springer; 2005.
[36] Ma H, Casual relationship among GDP, coal consumption and coal production
in China. Grey systems and intelligent services (GSISs). In: 2011 IEEE
international conference; 2011. p. 58–61.
[37] Ma H, Oxley L, Gibson J. China’s energy economy: a survey of the literature.
Econ Syst 2010;34(2):105–32.
[38] McCarthy JJ. Intergovernmental panel on climate change. Working
Group II. Climate change 2001: impacts, adaptation, and vulnerability:
contribution of working group ii to the third assessment report of the
intergovernmental panel on climate change, Cambridge University Press; 2001.
[39] Nelson CR, Plosser CR. Trends and random walks in macroeconmic time
series: some evidence and implications. J Monetary Econ 1982;10(2):
139–62.
[40] Netherlands Environmental Assessment Agency (NEAA). China Now No. 1 in
CO2 Emissions; USA in second position. Netherlands: Bilthoven; 2007.
[41] Otero J, Smith J. The KPSS test with outliers. Comput Econ 2005;26:
241–9.
[42] Pan J, Phillips J, Chen Y. China’s balance of emissions embodied in trade:
approaches to measurement and allocating international responsibility.
Oxford Rev Econ Policy 2008;24(2):354–76.
[43] Perron P. The great crash, the oil price shock, and the unit root hypothesis.
Econometrica 1989;57(6):1361–401.
[44] Pittis N. Efficient estimation of cointegrating vectors and testing for causality

in vector autoregressions. J Econ Surv 1999;13(1):1–35.
[45] Raupach MR, Marland G, Ciais P, Le Quéré C, Canadell JG, Klepper G, et al.
Global and regional drivers of accelerating CO2 emissions. Proc Natl Acad Sci
2007;104(24):10288–93.
[46] Ravallion M, Heil M, Jalam J. Carbon emissions and income inequality. Oxford
Econ Papers 2000;52(4):651–69.
[47] Sami J. Multivariate cointegration and causality between exports, electricity
consumption and real income per capita: recent evidence from Japan. Int J
Energy Econ Policy 2011;1(3):59–68.
[48] Schmalensee R, Stoker TM, Judson RA. World carbon dioxide emissions: 1950–
2050. Rev Econ Stat 1998;80(1):15–27.
[49] Shiu A, Lam PL. Electricity consumption and economic growth in China. Energy
Policy 2004;32:47–54.
[50] Shui B, Harris RC. The role of CO2 embodiment in US–China trade. Energy
Policy 2006;34(18):4063–8.
[51] Sims CA. Macroeconomics and reality. Econometrica 1980;48(1):1–48.
[52] Sinton JE. Accuracy and reliability of China’s energy statistics. China Econ Rev
2001;12(4):373–83.
[53] Smil V. China shoulders the cost of environmental change. Environ: Sci Policy
Sustain Dev 1997;39(6):6–37.
[54] Streets DG, Yu C, Bergin MH, Wang X, Carmichael GR. Modeling study of air
pollution due to the manufacture of export goods in China’s pearl river delta.
Environ Sci Technol 2006;4(7):2099–107.
[55] The World Bank. World development indicators 2010. Washington (DC); 2011.
[56] Toda HY, Yamamoto T. Statistical inference in vector autoregressions with
possibly integrated processes. J Econ 1995;66(1–2):225–50.
[57] Wang T, Watson J. 2007. Who Owns China’s Carbon Emissions? University of
Sussex and Tyndall Centre for Climate Change Research. Tyndall Briefing Note
No. 23.
[58] Weber CL, Peters GP, Guan D, Hubacek K. The contribution of chinese exports

to climate change. Energy Policy 2008;36(9):3572–7.
[59] White H, Lu X. Robustness checks and robustness tests in applied
economics. San Diego (CA): University of California; 2010.
[60] Wolde-Rufael Y. Coal consumption and economic growth revisited. Appl
Energy 2010;87(1):160–7.


N.M. Michieka et al. / Applied Energy 104 (2013) 258–267
[61] World Bank (WB). World development indicators; 2011. data.worldbank.org/data-catalog>.
[62] World trade organization (WTO). Trade to expand by 9.5% in 2010 after a
dismal 2009; 2010. [accessed November 9, 2010]. < />english/news_e/pres10_e/pr598_e.htm>.
[63] Xu G-Q, Liu Z-Y, Jiang Z-H. Decomposition model and empirical study of carbon
emissions for China, 1995–2004. China Popul, Resour Environ 2006; 6: 158.
[64] Xu M, Allenby B, Chen W. Energy and air emissions embodied in ChinaUS
trade: eastbound assessment using adjusted bilateral trade data. Environ Sci
Technol 2009;43(9):3378–84.
[65] Xunmin O, Zhang X, Shiyan C, Qingfang G. Energy consumption and ghg
emissions of six biofuel pathways by LCA In (the) People’s Republic of China.
Applied Energy 2009; 86(Suppl. 1): S197–S208.

267

[66] Yunfeng Y, Laike Y. China’s foreign trade and climate change: a case study of
CO2 emissions. Energy Policy 2010;38(1):350–6.
[67] Zapata HO, Rambaldi AN. Monte Carlo evidence on cointegration and
causation. Oxford Bull Econ Stat 1997;59(2):285–98.
[68] ZhiDong L. An econometric study on China’s economy, energy and
environment to the year 2030. Energy Policy 2003;31(11):1137–50.
[69] Zhou Z, Wu W, Wang X, Chen Q, Wang O. Analysis of changes in the structure

of rural household energy consumption in northern China: a case study.
Renew Sustain Energy Rev 2009;13(1):187–93.