This article appeared in a journal published by Elsevier.

The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/authorsrights
 Author's personal copy

China Economic Review 28 (2014) 123–134

Contents lists available at ScienceDirect

China Economic Review

International trade, FDI (foreign direct investment) and

embodied CO2 emissions: A case study of
Chinas industrial sectors
Shenggang REN ⁎, Baolong YUAN 1, Xie MA 1, Xiaohong CHEN 2
Resource-saving and Environment-friendly Society Research Center, School of Business, Central South University, Changsha 410083, Hunan, China

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

Article history: This paper calculates CO2 emissions embodied in China's international trade using an input–output
Received 28 March 2013 analysis, for the period 2000–2010. Based on industrial panel data, the two-step GMM estimation
Received in revised form 13 January 2014 is used to test the impacts of FDI, trade openness, exports, imports and per capita income on CO2
Accepted 13 January 2014
emissions. The results suggest that: (1) China's growing trade surplus is one of the important
Available online 21 January 2014
reasons for the rapidly rising CO2 emissions; (2) large FDI inflows further aggravate China's
CO2 emissions; and (3) the industrial sector's per capita income and CO2 emission relationship
JEL classification: show inverted-U environmental Kuznets curve. Therefore, in order to achieve environmentally
F18
sustainable development of the economy, China should make efforts to transform its trade
Keywords: growth mode, adjust foreign investment structure, strengthen energy efficiency and develop a
International trade low-carbon economy.
FDI
© 2014 Elsevier Inc. All rights reserved.
Embodied CO2 emissions
EKC hypothesis

1. Introduction

Who should take the primary responsibility for greenhouse gas emissions? Countries producing the products and services or
the consuming countries? The developing countries will naturally suffer environmental degradation when they emit CO2. These
issues have led to extensive and heated debates all over the world. At present, research on the relationship between economic
opening up (globalization) and environment has focused on the test of pollution haven hypothesis (PHH) which predicts that
freer trade will result in the creation of PHH in developing countries as pollution-intensive firms are transferred from developed
countries that impose stringent environmental regulations to developing countries. As a result, developing countries become
more polluted with opening up and FDI inflows.
A report from the Ministry of Commerce of China (2012) showed that, in 2011, China's exports and imports accounted for
global shares at 10.4% and 9.5%, which respectively remained the largest and the second largest for three consecutive years in the
world. After entering into the new millennium, China experienced a steady FDI inflow increase. According to OECD (2011), China
was the second FDI recipient in 2010 with an annual inflow of FDI amounting to about 1,476,400 million US dollars behind the
United States on the same year. However China's remarkable openness process seems to be accompanied by obvious environmental
pollution problems. CO2 emissions have accompanied the rapid economic growth. Since China's reform and opening up began. The
emissions increased from 1460 million tons (MT) in 1980 to 6499 MT in 2007 (Guan, Hubacek, Weber, Peters, & Reiner, 2008; Guan,
Peters, Weber, & Hubacek, 2009). According to the International Energy Agency (IEA) statistics, China's carbon emissions accounted

⁎ Corresponding author. Tel.: +86 13875856216.

E-mail addresses: [email protected], [email protected] (S. Ren), [email protected] (B. Yuan), [email protected] (X. Ma), [email protected]
(X. Chen).
1
Tel.: +86 731 88877722.
2
Tel.: +86 731 88879803.

1043-951X/$ – see front matter © 2014 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.chieco.2014.01.003
 Author's personal copy

124 S. Ren et al. / China Economic Review 28 (2014) 123–134

for 26.9% of global emissions. Due to the coal-dominated energy consumption structure, China's industrial sector is still the main
source of carbon emissions. During 1995–2008, industrial consumption of coal continued to rise with industry consuming 95% of all
coals consumed in 2008.
Aiming at obtaining a better understanding on the trade and FDI–environment nexus, this paper analyzes CO2 embodied in
trade and constructs a single multivariate framework, the variables of which include the per capita income, import and export
trade and FDI with respect to the impact of FDI and trade on pollution. This model is then tested by the panel data of 18 industrial
sectors during the period 2001–2010, during which trade and FDI inflow experienced the most important increase, to determine
whether China has become a pollution haven. The remainder of this paper is organized as follows.
Section 2 shows the literature review; the input–output analysis (IOA) and econometric model are described in Section 3;
Section 4 describes data sources and data processing methods used in this research; and Section 5 discusses the empirical results.
Conclusions drawn and the direction for future work are finally summarized in Section 6.

2. Literature review

The environmental Kuznets curve (EKC) hypothesis is a theoretical tool which has been widely used for measuring the
relationship between environmental and economic variables. It is claimed that the relationship between economic growth and
environmental pollution shows an inverted U-curve (Arrow, Bolin, Costanza, et al., 1995; Cole, Rayner, & Bates, 1997; De Bruyn, S. et
al., 2002; Grossman & Krueger, 1991; Roberts & Grimes, 1997). In their pioneering study, Grossman and Krueger (1991) investigated
the EKC hypothesis, and found evidence of an inverted U-shaped relationship between real income and environmental degradation
with the estimated turning points at US $4772 to US $5965. Following Grossman and Krueger (1991), though the EKC relationship has
been intensively analyzed empirically over the past two decades, the empirical results appear to be inconclusive to date: inverted
U-shaped (Ansuategi, 2003; Panayotou, 1993), N-shaped (Friedl & Getzner, 2003; Vincent, 1997) and inverted N-shaped (Shao, Yang,
Yu, & Yu, 2011). Here we use CO2 emission data to test the problem in China.
With the continuous strengthening of the world economy and trade activities, the pollution haven hypothesis is becoming the
focus of scholars. Grossman and Krueger (1991) were the earliest ones to establish the theoretical framework for analyzing
environmental effect of foreign trade. In their study on the North American Free Trade Agreement (NAFTA), they divided the
influence of trade on the environment into three effects: scale effect, structure effect and technology effect and found that
economic growth is the main factor to influence the environment. Since then, a small number of studies, which combine
multiplicative terms of per capita income with other characteristic variables such as trade, FDI in EKC estimations, have also tried
to make adjustments to the basic EKC model. In doing so, researchers have been able to show more directly how the coefficients of
the multiplicative terms can capture the impact of characteristic variables on EKC shapes.
Copeland and Taylor (1994, 1997) developed the North–South trade model to examine linkages between pollution and
international trade and showed that free trade improves the developed countries' environment while the developing countries'
environment exacerbated. In general, world trade has a negative impact on the environment. Antweiler, Copeland, and Taylor
(2001) employed multiplicative terms of openness degree and income in their analysis of SO2 emission determinants. Cole (2004)
augments the basic EKC model with additional explanatory variables meant to capture manufacturing share, trade openness, and
trade in goods produced by dirty industries and found evidence that trade openness and the proportion of dirty imports affect
emissions in developed countries, offering support for the PHH. However, he notes that the effect is small relative to other
determinants of emissions. Kankesu, Reetu, and Liu (2012) followed the methodology of Ang and Liu (2007) and Halicioglu
(2009) who adopted: (1) economic growth CO2 emissions, (2) economic growth-energy consumption and (3) trade-CO2 emissions
(nexus), into a single multivariate framework, to examine the relationship between pollution emissions, economic growth, energy
consumption and trade liberalization in China and India.
Likewise, some studies link environmental pollution and FDI. Frankel and Romer (1999) found that financial liberalization and
development may attract FDI, which in turn can speed up economic growth and hence affect the dynamic of the environmental
performance. Pao and Tsai (2011) showed a long-run relationship between CO2 emissions, energy consumption, FDI and
economic growth in a linear logarithm quadratic form, with a view of testing the validity of the EKC, and concluded that there
are strong dynamic interrelationships between output, energy consumption, environmental pollutants and FDI, which should be
investigated in the same multivariate framework. Jalil and Feridun (2011) use economic growth, energy consumption, trade
openness, FDI and environmental pollution in a single multivariate framework, to analyze the effects of economic growth, financial
development and energy consumption on carbon emissions in the case of China.
But most previous articles employed FDI as the financial development indicator, not the PHH indicator, thus contributing little
to PHH. For China's trade-CO2 emission research, including the EKC, import and export trade and FDI empirical in one multivariate
model, little literature has comprehensively analyzed income–CO2 emission problems basing on the influence of trade and FDI
inflows, and instead, their trade or FDI data is mostly geographical, rarely industries' panel data. Moreover, articles on PHH have failed
to reach a consensus. Some empirical studies have verified the PHH (Chew 2009; Dean, Lovely, & Wang, 2009; Emilson & Zhu, 2009;
Machado, Schaeffer, & Worrell, 2001; Pao & Tsai, 2011). However, some scholars have rejected the PHH after empirical analyses
(Eskeland & Harrison, 2003; Dong et al., 2010; Cole and Fredriksson, 2009; Perkins and Neumayer, 2009).
China is a large trade nation and CO2 emission emitter. Current research focuses on the analysis of estimating international
trade implicit CO2 emissions (Lin & Sun, 2010; Yan & Yang, 2010). And the present stage of study about China's CO2 emission
influence factors is less most studies use mathematical model decomposition analysis (Dong, Masanobu, Liu, & Wang, 2010; Xu,
Li, Crittenden, & Chen, 2011). Considering that the estimate of its international trade embodied CO2 emissions is not enough and
 Author's personal copy

S. Ren et al. / China Economic Review 28 (2014) 123–134 125

the empirical literature analyzing the influence of international trade especially for FDI on China's CO2 emissions is rare, further
research, therefore, on a developing country like China, is needed.
Firstly, in this paper, we propose a comprehensive, multiplicative EKC model through which trade and FDI variables can be
integrated into the analyses of pollution–income with cross-sector panel data. Compared to the basic EKC model, the multiplicative
model in general has greater explanatory power and more flexibility in simulating a CO2 pollution–income relationship that is
economy-specific and dynamic. And we can be able to show more directly how the coefficients of the multiplicative terms can capture
the impact of characteristic variables on EKC shapes. The direct inclusion of economy- and policy-related characteristics into the
estimation of the pollution–trade and FDI relationship can also help turn the simple coefficients into policy suggestions. Secondly, few
studies have focused on FDI industrial data by sector to test FDI impacts on CO2 emissions. Here, we use 18 industries' FDI and trade
data to verify their impacts on carbon emissions through a multiple regression model which may offer further evidence of PHH.
Finally, we improve the EAI assumption3 (Sanchez Choliz & Duarte, 2004) by adopting an average emission factor of “virtual
importing countries”4 to replace imported goods' emission factors, which may be more convincing.

3. Methodology

3.1. Input–output analysis of embodied carbon

The input–output analysis (IOA) was theorized and developed by economist Wassily Leontief between the 1930s and
the 1970s (Leontief, 1970). The IOA is based on the IO model, which represents monetary transactions between supply chains
in mathematical form. The IO model indicates which part of the output of an industry is used as the intermediate input by other
industries. Assuming an economy includes n industries, the equation can be represented as

x ¼ Ax þ y ð1Þ

where x is the total output of the entire economy and is a column vector. Ax represents the intermediate input and y is the final
demand. A represents a direct requirement coefficient matrix. Its element Aij = Xij / xj represents the amount of output from
industry i required directly to produce unit output of industry j.
The relationship between the final demand y and the total output x can be represented as

−1
x ¼ ðI−AÞ y ð2Þ

where I is the unit matrix and (I − A)−1 is called the Leontief inverse matrix. Its element αij represents the amount of output
of the industry required directly and indirectly to produce one unit of final demand from industry j.
In Eq. (3), the purpose of extending the IO model to non-econometric fields is to measure externalities generated by unit
change in output. F is the direct CO2 emission factor, which means industry's direct CO2 emissions per unit of output. Then the
sector-wise total supply-chain emissions f can be calculated as

−1
f ¼ F ðI−AÞ y: ð3Þ

Based on 2006 IPCC (Intergovernmental Panel on Climate Change) Guidelines for National Greenhouse Gas Inventories,
we adopt the reference method to estimate energy-related CO2 emissions by each sector in the following way:

X
18
44 X18
CO2i ¼ Ei  NCV i  CE F i  CO F i  ¼ E  θi ð4Þ
i¼1
12 i¼1 i

where Ei represents the consumption of energy type i. NCVi, CEFi and COFi represent net calorific value, carbon content and carbon
oxidation factor of ith energy, respectively. θi means CO2 emissions intensity of ith energy. Therefore, direct CO2 emission factors
by sector can be expressed as

CO2i
Fi ¼ : ð5Þ
Xi

3
It supposed the production technology of imported goods is the same with domestically produced goods and substitute the domestic complete CO2 emission
coefficient for the importing countries.
4
Assume that a “virtual importer”, which is the only domestic trading partners. And China's imported products all adopt virtual importer production
technology for production. Then with the major trade countries' related data, we calculate complete carbon emission coefficient and carbon emissions embodied
in imported products. We select the main import trade partners of China in the past ten years, the ten countries and regions are the United States, Japan, Korea,
Germany, Australia, Malaysia, India, Britain, Hong Kong (China) and Taiwan (China).
 Author's personal copy

126 S. Ren et al. / China Economic Review 28 (2014) 123–134

Combining Leontief inverse matrix deformation and direct CO2 emission factors by sector, CO2 emissions embodied in exports
(EEE) can be calculated as
 
e d −1
f ¼ F I−A ex ð6Þ

where Ad is the direct requirement coefficient matrix of the domestic intermediate input. ex represents goods exported from
China and is a column vector.
Similarly, CO2 emissions embodied in imports (EEI) can be expressed as
 
m d −1
f ¼ F m I−Am im ð7Þ

where Fm is the imported goods' direct CO2 emission factor. Adm represents direct requirement coefficient matrix of the domestic
intermediate input from other countries. im is goods imported into China and is also a column vector.
Therefore, the net CO2 emissions embodied in trade (EEB) can be estimated from

   
b e m d −1 d −1
f ¼ f −f ¼ F I−A ex− F m I−Am im: ð8Þ

3.2. Econometrics regression model

Grossman and Krueger (1995) examined the basic relationship between the size of an economy and the intensity of its emissions,
finding that pollution tends to rise during the first stages of a country's development, and then to decrease after reaching a certain
income level. The basic model is:

2
Eit ¼ β0 þ β1 Y it þ β2 Y it þ εit : ð9Þ

Following the previous empirical literature, it is plausible that we introduce trade and FDI variables to establish the
econometric regression Model (9) to test the impacts of FDI, trade openness, export, import and per capita income on CO2
emissions. The effect of FDI is uncertain, if FDI inflow brings cleaner technologies, we can expect a negative sign for this variable,
but if FDI inflow increases pollution, it may have a pollution-haven effect. TR stands for trade openness (that is, the ratio of
exports plus imports to GDP). EX is the ratio of exports to GDP; IM is the ratio of imports to GDP. The sign β3, β4 could be positive,
and β5 positive; the two positive associations reflect a pollution haven hypothesis, which suggests that an increase in trade
openness will increase pollution due to a comparative advantage in dirty production under weaker environmental regulations;
the negative association indicates that the source of imports be less polluted and Chinese import is clean. Since per capita income
reflects a country's income and level of development, it is expected that the initial level of income induces pollution; however,
as income rises, environmental degradation decreases because of the level of environmental consciousness and technology that
utilities cleaner energy. Under the EKC hypothesis, the sign of β6 is expected to be positive, whereas a negative sign is expected for
β7 (Table 1).

2
CO2it ¼ β0 þ β1 CO2it−1 þ β2 FDIit þ β3 TRit þ β4 EXit þ β5 IMit þ β6 ln Y it þ β7 ln ðY it Þ þ μ i þ ν it : ð10Þ

Our panel data have the characteristics of “small time span but large cross-sectional dimensions” and include the lagged
explanatory variables and this result in stochastic disturbance terms relating to explanatory variables and the other explanatory
variables may also be endogenous. So we adopt the two-step GMM method to test the model. When we use the GMM estimation
method, we adopt Sargan test for reliability of the instrumental variable. If the test value is small, we accept the null hypothesis

Table 1
Meaning of each factor in function 9.

Variable Definition Representation

i Sector 18 industrial sectors

t Year Years 2001–2010
CO2 CO2 emissions CO2 emissions by sector/industrial GDP
FDI Foreign direct investment FDI in sector/industrial GDP
TR Trade openness Total goods exports and imports/industrial GDP
EX Export Exports by sector/industrial exports
IM Import Imports by sector/industrial imports
Y Per capita income Annual value added of constant price by sector/annual average working staff
μ Individual variability –
v Stochastic disturbance term –
 Author's personal copy

S. Ren et al. / China Economic Review 28 (2014) 123–134 127

Table 2
Input–output sector classification.

Code Sector Code Sector

1 Mining and quarrying 10 Basic metals

2 Food, beverages and tobacco 11 Fabricated metal products
3 Textiles and leather 12 Machinery and equipment
4 Wood and wood products 13 Transport equipment
5 Pulp, paper, and printing 14 Electrical machinery
6 Coke, and refined petroleum products 15 Communication equipment
7 Chemicals 16 Office machinery and precision instruments
8 Rubber and plastics products 17 Other manufactures
9 Other non-metallic mineral products 18 Electricity, gas and water supply

that the instrumental variable is suitable. We use first-order and second-order serial correlations of first difference transformation
equation to test AR (1) and AR (2) to decide whether random disturbance is serial correlation.

4. Data

4.1. Data sources

There are six types of data: IO tables, direct CO2 emission factors, import and export trade statistics, FDI flows and per capita
income. The data cover the period from 2000 to 2010. The 2000 and 2005 Chinese IO tables are from OECD (2010)5. Considering
that a country's production and technology structure remain almost stable in the short term and most countries' IO tables in OECD
database are compiled once every five years, we use the 2000 and 2005 IO tables to replace the 2001–2004 periods and
2006–2010 periods, respectively. The OECD IO table is divided into the total table and domestic table and import table. Then we
can get A and Ad through direct consumption coefficient calculation formula. Trade data from OECD bilateral trade database (BTD)
are correlated with the IO tables sector-wise. FDI data are from the China Statistical Yearbook 2000–2010.6 In the absence of FDI
industry data, we adopt ownership interest data from Main Indicators of Industrial Enterprises with Hong Kong, Macao, Taiwan
and Foreign Funds by Industrial Sector. Based on the China Statistical Yearbook, we divide the industrial economy into 18 sectors
(Table 2). In order to eliminate the effects of inflation, time series monetary values of input–output table, GDP and FDI data are
transformed into constant prices (year 2000).7

4.2. Data processing

4.2.1. Direct CO2 emission factors by sector

CO2 emission factor of each fuel type (except electricity) is fixed, so we can calculate the CO2 emission factors using the parameters
of average net calorific value, carbon emission factor, CO2 conversion coefficient, carbon oxygenation efficiency and the coefficient to
convert energy consumption into standard coal equivalent (refer to IPCC 2006 computational method). The results are shown in
Table 3.
The CO2 emission factor of electricity is quite special since electricity consumption does not cause CO2 emissions directly;
during the electricity generation process a lot of fundamental energy is consumed and, therefore, electricity consumption also
causes CO2 emissions indirectly and the quantity of emissions is quite huge. CO2 emissions caused by electricity consumption
depend upon the inputs used (coal, nuclear, solar and wind, etc.) and the standard rate of coal consumption in the case of thermal
electricity. As the data of composition of electricity generation in China are deficient and above 90% of Chinese thermal electricity
comes from raw coal, we estimate the CO2 emissions factor of electricity approximately, through the CO2 emission factor of raw
coal. The results are shown in Table 4.

4.2.2. Imported goods' CO2 emission factors

We improve the EAI assumption by assuming that the “virtual importing country” is our only trading partner, meaning that
exports and imports are to and from this virtual country. Then we adopt an average emission factor of 10 major trading countries
to replace imported goods' emission factor. Since getting imported goods' emission factor of the 10 major countries is very
difficult, we make some amendments. The technical adjustment factor formula is as follows, and the results are shown in
Table 5.

X
10
∂¼ λi δi ð11Þ
i¼1

5
Data source: OECD database (http://www.oecd.org/statistics/).
6
Data source: National Bureau of Statistics (NBS), 2004. China Statistical Yearbook 2000–2010. China Statistics Press, Beijing.
7
Data source: the price indexes are from OECD.
 Author's personal copy

128 S. Ren et al. / China Economic Review 28 (2014) 123–134

Table 3
CO2 emission factor of each energy type except electricity.

Energy type CO2 emission factor (tCO2/toe) Energy type CO2 emission factor (tCO2/toe)

Raw coal 2.769 Kerosene 2.104

Cleaned coal 2.769 Diesel oil 2.168
Other washed coal 2.776 Fuel oil 2.265
Briquettes 2.862 Liquefied petroleum gas 1.846
Coke 3.134 Refinery gas 1.687
Coke oven gas 1.299 Natural gas 1.642
Other gases 1.299 Other petroleum products 2.148
Crude oil 2.147 Other coking products 2.365
Gasoline 2.029 Heat 3.249

Notes: Average net calorific value and the coefficient to convert it into standard coal equivalent are collected from “Chinese Energy Statistics Yearbook (2011)”,
carbon emission factors and carbon oxygenation efficiency are collected from IPCC (2006). CO2 emission factor of each energy type except electricity = (average
net calorific power ∗ carbon emission coefficient ∗ carbon conversion coefficient (44/12) ∗ carbon oxygenation efficiency) / the coefficient to transform standard
coal.

where λ5 is the proportion of goods imported from country i in China's total imports and δi represents the ratio of carbon intensity
of country i. Therefore, the CO2 emission embodied in imports (EEI) can be expressed as

 
m d ‐1
f ¼ ∂F I−A im ð12Þ

5. Result

5.1. Scale analysis of CO2 emissions in China's industry and trade

During the Tenth and Eleventh Five-Year Plan periods, China always had a trade surplus; it had aggregate trade surplus
of $1329.94 billion during these two periods. Because of the huge trade surplus, China has been under increasing pressure
of environmental pollution. From Fig. 1, we can see that China's industrial embodied carbon emissions kept growing during
2000–2010. Industrial carbon emissions increased to 5767.37 MT in 2005 from 3494.49 MT in 2000 and then continuously rose to
7755.83 MT in 2010.
Fig. 2 shows that EEE and EEI kept an upward trend on the whole during 2000–2010. But both EEE and EEI declined in 2005
because we used the 2000 and 2005 IO tables to replace the 2001–2004 and 2006–2010 periods, respectively, making carbon
emission data lose continuity. Analyzing the two periods separately indicates the same upward trend. Especially, we focus on
analyzing the embodied carbon during the last five years (2005–2010). Fig. 2 shows that CO2 emissions embodied in exports
(EEE) and imports (EEI) are growing but EEE are always greater than EEI. Our calculations of EEE account for approximately
2499.08 MT in 2005, increasing to 3311.97 MT in 2010 or 37.78%–50.02% of China's annual CO2 emissions. However, China's EEI
were 368.52 MT in 2005 and 490.36 MT in 2010, accounting for only 5.48% (2005) and 6.97% (2010) of total annual CO2 emissions.
Obviously, the net carbon exports (EEB) are positive. That is to say, as the world's largest carbon emitter, some fraction of China's CO2
emissions is produced during the manufacture of export goods destined for foreign consumers. Finally, it is easy to find that EEE and
EEI showed a turning point in 2009. That is due to the subprime crisis in US, which gave rise to a financial tsunami in the latter half of
2008, and the world economic downturn led to significant decline of international trade. (See Fig. 3.)

5.2. Sector analysis of CO2 emissions in China's industry and trade

We take the year 2005 as an example to analyze CO2 emissions embodied in 18 sectors of the Chinese industry (see Fig. 2.).
Clearly, EEBs of all sectors are positive except mining and quarrying (1).8 In 2005, China's total EEB were 2126.6 MT, of which
industrial EEBs were 1964.3 MT, accounting for 92.4% of the total EEBs. This means China's industrial sector is almost the main
body of net carbon exports. Mining and quarrying, where EEI are greater than EEE are thus a “clean type” industry. This is mainly
because mining and quarrying as a basic industry produces low value-added products and materials which are inputs for other
downstream industries. In addition, the trade balance of mining and quarrying was negative in 2005, which further deepened the
EEI greater than the EEE.
The three largest EEE sectors are machinery and equipment (12), textiles and leather (3) and communication equipment (15).
FDI inflows invested in these three industries accounted for 32.6% of the total industrial sector and their EEE accounted for 55.2%
of the total EEE of these sectors. As FDI inflows stimulate trade of the host country, it further increased the EEE. In 2005, China's
largest EEE sector was machinery and equipment (12), which accounted for 36.7% of industrial emissions. Because large parts of
these sectors' exports belong to processing trade, machinery and equipment production needs a great deal of intermediate inputs

8
“()” indicates the number of the sector code.
 Author's personal copy

S. Ren et al. / China Economic Review 28 (2014) 123–134 129

Table 4
Shares of different sources of energy in electricity generation and average CO2 emission factor of electricity for the years 2005–2009.

Year Hydro-power Thermal electricity Nuclear power Wind power Other RES Standard coal equivalent Emission factor of electricity
(kg/kWh) (tCO2/toe)

2005 15.97% 81.54% 2.11% 0.00% 0.38% 0.343 6.307

2006 14.55% 83.31% 1.92% 0.10% 0.11% 0.342 6.425
2007 14.44% 83.34% 1.93% 0.17% 0.12% 0.332 6.240
2008 16.39% 81.22% 2.01% 0.38% 0.00% 0.322 5.898
2009 16.39% 81.22% 2.01% 0.38% 0.00% 0.322 5.898

Notes: China's 2001–2009 power structure data and thermal power standard coal consumption rate data are from the Chinese Electric Power Yearbook
(2002–2010). As Chinese Electric Power Yearbook (2011) is not available as yet, all data for the year 2010 have been assumed to be the same as 2009. CO2
emission factor of electricity = (proportion of thermal electricity ∗ standard rate of coal consumption of thermoelectricity ∗ CO2 emissions factor of raw coal) / the
coefficient of electricity to convert it into standard coal equivalent.

(such as electricity and metals) for processing the raw and auxiliary materials imported from abroad. At the same time, this sector
is also the largest net carbon exporting industry as its EEBs were 805.9 MT in 2005.

5.3. Analysis of econometrics regression model

In industry panel data, we use ADF (augment Dickey–Fuller test) method to test the variables of unit root; test results are
shown in Table 6. Two sets of data have passed the test with 1% and 5% significance levels, satisfying the stationarity assumptions,
and thus we can conduct least-squares regression test.
With respect to the dynamic panel data models, Bond (2002) pointed out that the lagged dependent variable is correlated
with the stochastic error term, which results in endogeneity. More specifically, the ordinary least squares (OLS) level estimator is
biased upward, while within groups tend to give a downward-biased estimate of the coefficient on the lagged dependent variable.
There is potential endogeneity for the independent variables as well. As GMM estimators can use information from both difference
and level equations, they work well for panel data with large sample size and small time span. Corresponding to different weight
matrixes, GMM estimators can be divided into one-step GMM estimators and two-step GMM estimators.
In comparison with one-step estimators, two-step GMM estimators are less likely to be affected by heteroskedasticity.
However, the presence of a finite sample may yield significant downward biased standard errors and skew the inferences.
As Windmeijer (2005) showed, the biases can be corrected by paying attention to the extra variation in small samples. With this
in mind, we use robust two-step GMM procedure to calculate the standard errors. As two-step GMM estimators entail testing
the validity of the instruments and whether there are serial correlations among the residuals, we use the Sargan test to assess
the validity of the instruments. If the hypothesis is accepted, it means that the instrument is valid. In addition, we conduct tests in
autoregressive models AR (1) and AR (2) to learn whether the residuals are serially correlated.
Based on inclusion of different independent variables, we estimate nine regression models. Table 7 reports the instruments
and their lagged values in the GMM estimation of the dynamic panel data, as well as results from tests in autoregressive models
and those from the Sargan test. The consistency of the GMM estimator requires zero second-order serial correlation among the
residuals. The AR (2) value of each dynamic panel model indicates that we cannot reject the hypothesis of zero second-order
serial correlation among the residuals. The results of the Sargan test demonstrate that the instruments and the lagged values are
appropriate. The F-value shows that all models are fitting better.
Models 1, 2 and 3 test the impact of the lag and FDI on CO2 emissions, with two-Step GMM estimation, random-effects GLS
regression and pooled OLS regression method, respectively. The results of the three are nearly the same, but the GMM estimation
method yields a relatively significant result. In Model 1, the pre-CO2 emissions and the current period's CO2 emissions are related,
which means CO2 emissions are a process of continuous and cumulative adjustment. The coefficient of FDI is 0.241 at 1% level
of significance, which means that 1 percentage increase in the FDI net inflows will increase CO2 emission by 24.1%. This is opposed
to Zhou, Zhang, and Li (2012) results that indicate negative associations between FDI and CO2 emissions, which support the Porter
Hypothesis: FDI can reduce CO2 emissions by inducing technical innovation.
Models 4, 5 add the variable trade openness (TR) on the basis of Models 1, 2 to test the impacts of international trade on
carbon emissions. The coefficient of TR is 0.004 and statistically significant. As the negative effects of scale, structure and
technique outweigh the positive effects, the total effect of trade is negative. This is consistent with conclusions in Section 5.1. In
Model 4, trade openness and FDI are positively related to CO2 emissions. 1 percentage increase in total trade will increase CO2
emissions by 0.4%. And 1 percentage increase in the FDI net inflows will increase CO2 emissions by 22.3%. So we have evidence
against the results of Zhou et al. (2012). Their results indicate negative associations between FDI and CO2 emissions, which

Table 5
Technical adjustment factor.

Adjustment factor 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

∂ 0.264 0.274 0.266 0.253 0.219 0.222 0.228 0.236 0.235 0.221 0.224

Notes: Trading data during 2000–2010 come from OECD (BTD). The carbon intensity data of 10 countries come from EIA.
 Author's personal copy

130 S. Ren et al. / China Economic Review 28 (2014) 123–134

9000

8000

7000

6000

5000

4000

3000

2000

1000

0
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

CO2 emission embodied in China's industry (MT)

Fig. 1. 2000–2010 China's industrial embodied carbon emissions. Data sources: (1) The 2000 and 2005 Chinese IO tables are from OECD database (http://www.
oecd.org/statistics/). (2) Average net calorific value and the coefficient to convert it into standard coal equivalent are collected from Chinese Energy Statistics
Yearbook (2011), carbon emission factors and carbon oxygenation efficiency are collected from IPCC (2006). (3) China's 2001–2009 power structure data and
thermal power standard coal consumption rate data are from the Chinese Electric Power Yearbook (2002–2010).

support the Porter Hypothesis, that is, FDI can reduce CO2 emissions by inducing technical innovation. On the whole, we can't deny the
possibility of PHH existing in China. Although the rush of foreign investment can promote China's economic development, it contributes
to high levels of pollution. We found that large FDI inflows have gone into such carbon-intensive industries such as chemicals (7 and 8),
communication equipment (15), paper industry (5), transport equipment (13), basic metals (11) during 2000–2010. FDI inflows
further aggravate China's CO2 emissions. In addition, a large number of foreign capital-intensive major industries (such as 15, 13, etc.)
are equipment manufacturing. As downstream industries, they produce a large pull effect on carbon-intensive upstream industries,
thereby indirectly increasing CO2 emissions.
In Model 6, we further consider the impact of per capita income on CO2 emissions to test the EKC hypothesis. Coefficients of Y
and Y2 are 0.108 and − 0.013, respectively, but the coefficient of Y2 becomes non-significant.
Model 7 further considers the impacts of variables exports (EX) and imports (IM) on emissions. Their coefficients are 0.106
and −0.069, respectively, but are not significant. So, it is not clear whether exports increase emissions or imports do reduce them.
This may because the other factors (such as CO2it − 1, FDI, TR, etc.) in the equation have a greater impact on carbon emissions.

4500 1800

4000 1600
CO2 embodied in trade (Mt)

3500 1400
Trade value (billion USD)

3000 1200

2500 1000

2000 800

1500 600

1000 400

500 200

0 0
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Export Import EEE EEI

Fig. 2. The CO2 emissions embodied in China's foreign trade. Data sources: (1) The 2000 and 2005 Chinese IO tables are from OECD database (http://www.oecd.
org/statistics/). (2) Average net calorific value and the coefficient to convert it into standard coal equivalent are collected from Chinese Energy Statistics Yearbook
(2011), carbon emission factors and carbon oxygenation efficiency are collected from IPCC (2006). (3) China's 2001–2009 power structure data and thermal
power standard coal consumption rate data are from Chinese Electric Power Yearbook (2002–2010). (4) Trade value data from OECD bilateral trade database
(BTD) are correlated with the IO tables sector-wise.
 Author's personal copy

S. Ren et al. / China Economic Review 28 (2014) 123–134 131

Fig. 3. The CO2 emissions embodied in China's industrial sectors in 2005. Data sources: (1) The 2000 and 2005 Chinese IO tables are from OECD database (http://www.
oecd.org/statistics/). (2) Average net calorific value and the coefficient to convert it into standard coal equivalent are collected from Chinese Energy Statistics Yearbook
(2011), carbon emission factors and carbon oxygenation efficiency are collected from IPCC (2006). (3) China's 2001–2009 power structure data and thermal power
standard coal consumption rate data are from Chinese Electric Power Yearbook (2002–2010). (4) Trade value data from OECD bilateral trade database (BTD) are
correlated with the IO tables sector-wise. Notes: Industrial sectors arranged in order of small to large EEB.

When we add variables EX and IM to Model 6, we find that the robustness is improved, with significant coefficients of Y and Y2.
The correlation between per capita income of China and industry and emissions shows an inverted U-shaped curve, and the EKC
hypothesis gets demonstrated. It is consistent with the recent findings of Jalil and Feridun (2011), Kankesu et al. (2012) who
confirm the existence of an environmental Kuznets curve in the case of China.
In the last model, we put Y, Y2, EX and IM into Model 4, but this had no effect. The result suggests that exports are associated
with higher CO2 emissions, and a negative coefficient for the import variables, which is consistent to Kearsley and Riddel's (2010)
result. Comparing all the models, we come to the conclusion that Model 4 is the most robust and coefficients of CO2it − 1, FDI and
TR are 1.002, 0.223 and 0.004, respectively.

6. Conclusions and recommendations

We adopt the single regional input–output analysis model to analyze carbon emissions embodied in China's international
trade during the period 2000–2010 and empirically examine the main factors influencing carbon emissions. Our main research
conclusions are as follows.

Table 6
ADF test results.

Variables (unit) Statistic Concomitant probability Results

CO2 (t/10,000 RMB) 37.9539⁎⁎⁎ 0.0013 Stationary

FDI (%) 37.0499⁎⁎⁎ 0.0022 Stationary
TR (%) 45.8776⁎⁎⁎ 0.0013 Stationary
EX (%) 31.4155⁎⁎ 0.0108 Stationary
IM (%) 31.4147⁎⁎ 0.0108 Stationary
Y (10,000 RMB/person) 29.1687⁎⁎ 0.0158 Stationary

Notes: (1) CO2 is CO2 emissions, denoted by CO2 emissions by sector per industrial GDP. The value is calculated by the author himself and the data are from OECD
database (http://www.oecd.org/statistics/), Chinese Energy Statistics Yearbook (2011, IPCC (2006), Chinese Electric Power Yearbook (2002–2010)) and China
Statistical Yearbook (2000–2010).
(2) FDI is foreign direct investment, expressed as the ratio of FDI in sector to industrial GDP. The data are from China Statistical Yearbook (2000–2010). In the
absence of FDI industry data, we adopt ownership interest data from Main Indicators of Industrial Enterprises with Hong Kong, Macao, Taiwan and Foreign Funds
by Industrial Sector.
(3) TR is trade openness, denoted by total goods exports and imports per industrial GDP. The data are from OECD bilateral trade database (BTD).
(4) EX is export, expressed as the ratio of exports by sector to industrial exports, and IM is import, expressed as the ratio of imports by sector to industrial imports.
The data are from OECD bilateral trade database (BTD).
(5) Y is per capita income, denoted by annual value added of constant price by sector per annual average working staff. The data are from China Statistical
Yearbook 2000–2010.
⁎⁎⁎ Denotes statistical significance at 1% level.
⁎⁎ Denotes statistical significance at 5% level.
 Author's personal copy

132 S. Ren et al. / China Economic Review 28 (2014) 123–134

Table 7
Empirical results of the dependent variable.1

Independent variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9

(Unit)

CO2it − 1 1.005⁎⁎⁎ 0.995⁎⁎⁎ 0.087⁎⁎⁎ 1.002⁎⁎⁎ 0.992⁎⁎⁎ 1.009⁎⁎⁎ 1.005⁎⁎⁎ 1.009⁎⁎⁎ 1.006⁎⁎⁎
(t/10,000 RMB) (120.04) (72.35) (90.31) (114.60) (71.01) (109.04) (123.91) (100.26) (91.10)
FDI 0.241⁎⁎⁎ 0.152⁎⁎ 0.193⁎⁎⁎ 0.223⁎⁎⁎ 0.181⁎⁎⁎ 0.224⁎⁎⁎ 0.190⁎⁎⁎ 0.220⁎⁎⁎ 0.230⁎⁎⁎
(%) (4.71) (2.37) (3.11) (4.60) (2.70) (3.80) (4.64) (3.44) (3.77)
TR 0.004⁎⁎⁎ 0.004⁎ 0.004⁎⁎⁎
(%) (3.41) (1.38) (3.03)
lnY 0.108⁎ 0.109⁎ 0.447
(10,000 RMB/person) (1.62) (1.65) (0.72)
lnY2 −0.013 −0.014⁎ −0.006
(10,000 RMB/person) (−1.45) (−1.52) (−0.72)
EX 0.106 0.052 0.150
(%) (0.86) (0.36) (0.92)
IM −0.069 −0.025 −0.025
(%) (−0.99) (−0.38) (−0.37)
AR (1) −1.017 [0.31] −0.99 [0.32] −0.99 [0.32] −1.03 [0.03] −0.99 [0.3] −0.99 [0.03]
AR (2) −1.18 [0.24] −1.09 [0.28] −1.18 [0.24] −1.09 [0.28] −1.18 [0.24] −1.09 [0.28]
Sargan test 16.71 [1.00] 9.47 [1.00] 16.01 [1.00] 15.81 [1.00] 16.16 [1.00] 15.11 [1.00]
Observation 180 180 180 180 180 180 180 180 180
Individual 18 18 18 18 18 18 18 18 18
Estimation GMM GLS OLS GMM GLS GMM GMM GMM GMM
method
F-value 7520.23 6048.44 560.83 7310.19 6081.73 4933.57 5459.11 3888.44 4082.33
[0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00]

Notes: (1) CO2 is CO2 emissions, denoted by CO2 emissions by sector per industrial GDP. The value is calculated by the author himself and the data are from OECD
database (http://www.oecd.org/statistics/), Chinese Energy Statistics Yearbook (2011), IPCC (2006), Chinese Electric Power Yearbook (2002–2010) and China
Statistical Yearbook (2000–2010).
(2) FDI is foreign direct investment, expressed as the ratio of FDI in sector to industrial GDP. The data are from China Statistical Yearbook (2000–2010). In the
absence of FDI industry data, we adopt ownership interest data from Main Indicators of Industrial Enterprises with Hong Kong, Macao, Taiwan and Foreign Funds
by Industrial Sector.
(3) TR is trade openness, denoted by total goods exports and imports per industrial GDP. The data are from OECD bilateral trade database (BTD).
(4) Y is per capita income, denoted by annual value added of constant price by sector per annual average working staff. The data are from China Statistical
Yearbook (2000–2010).
(5) EX is export, expressed as the ratio of exports by sector to industrial exports, and IM is import, expressed as the ratio of imports by sector to industrial imports.
The data are from OECD bilateral trade database (BTD).
(6) The t-statistics for the coefficients are in parentheses. ***denotes statistical significance at 1% level; **denotes statistical significance at 5% level; *denotes
statistical significance at 10% level. P-values are in square brackets. The data covers 2000–2010, the time span is 10 years and sample size is 180. In the two-Step
GMM estimation, all the endogenous variables in regression is CO2it − 1 endogenous variables' two order and more order lag term is instrumental variable.
1
The software we used in our econometric estimations is “Stata 10.0”.

Firstly, China's growing trade surplus is one of the important reasons for the rapidly growing CO2 emissions. From Fig. 2,
we can see that the net CO2 emissions embodied in China's foreign trade follow the growth trend of China's trade surplus.
During 2005–2010, China's trade surplus increased from $102 billion to $181.76 billion while net CO2 exports increased from
2130.5 MT to 2821.6 MT. This may indicate that the rapid growth of China's CO2 emissions is inseparable from the growing
trade surplus. Consequently, a large part of CO2 emissions occur in China but the related commodities are consumed by
foreigners.
Secondly, from results of the econometrics regression, we can see that the total effect of foreign trade is negative for China
because it increases carbon emissions, and FDI has a significant positive effect on CO2 emissions. The coefficient means that
although the rush of trade openness and foreign investment can promote China's economic development, it contributes to high
levels of pollution. So, we can't deny the possibility of PHH existing in China. Our results seem to not support the halo effect of FDI
which argues that FDI will tend to spread greener technology to host countries for environmental improvement of developing
countries (Pao & Tsai, 2011). Besides we confirm the existence of an environmental Kuznets curve in China under free trade, and
the correlation between per capita income of China and industry emissions shows an inverted U-shaped curve.
Based on the above research conclusions, we put forward the corresponding policy recommendations:

First, strengthen the adjustment of the export product structure and improve the trade growth mode. Through different export
drawback policies, export tariffs and other means, the government should encourage exports of products of low carbon
emission industries and limit or ban exports of products with high carbon emissions. As China's main industries, exports of
mechanical and electronic products and chemicals and textiles are necessary but we ought to upgrade to high value-added
products for exports. Importantly, we should enhance environmental technology's imports and link energy and pollution
intensive products' exports to imports to improve CO2 emissions. In addition, guide the upgrade of the processing trade and
ease the adverse influences on the environment caused by exports' rapid growth.
 Author's personal copy

S. Ren et al. / China Economic Review 28 (2014) 123–134 133

Second, adjust the composition of foreign investment (in terms of industry segments) to make sure industries set up with FDI are
environmentally sustainable. Foreign capital should be used for promoting the industrial technical level and the efficiency of
resource utilization. China should encourage foreign capital to enter the industry with low consumption of resources and little
environmental pollution and guide foreign capital to upgrade and alter traditional industry with heavy pollution and huge resource
consumption (Gao and Zhang 2013). With formulation of unified national environment protection policies, various domestic
regions ought to avoid competition for foreign capital for the sake of local economic development and only choose high-quality
investments. Furthermore, it is advisable to reinforce industrial guidance for foreign investment and encourage foreign investment
to flow into environment-friendly industries, such as ecological agriculture, service industry and so on. Besides, restrict and prohibit
projects which bring about serious pollution (for example, non-metallic mineral industry, textiles, papermaking industry, etc.), and
replace pollution control with pollution prevention to strive to nip pollution in the bud before the production process.
Third, as its energy efficiency is still very low, China should reinforce international cooperation and make good use of market
mechanism to develop a carbon trading and carbon sink economy. On one hand, it should actively participate in the international
CDM project, to reduce the use of coal, improve terminal energy efficiency and reduce the carbon emission coefficient. On the
other hand, use market mechanisms such as carbon emission trading and carbon sink economy to allocate reasonably CO2
emission rights among enterprises; meanwhile environment-friendly products should attract lower taxes and import tariffs,
in order to encourage their consumption.
This study provides a theoretical basis for formulating policies which can help China shape its international trade and introduce
environment-friendly foreign investment. However, there are limitations in this paper, which needs further discussions in the future.
OECD input–output table data were available for only two years (2000, 2005), so any other year's IO tables had to be obtained by
approximation. Definitely, this must have caused some deviations. Overall, research in this field is still in the initial stages at present.
In view of China's regional differences, future discussions can focus on environmental effects from international trade and FDI.

Acknowledgments

This research work is supported by the National Natural Sciences Foundation of China (71221061, 71202055), by the Program
for New Century Excellent Talents in University (NCET-12-0542), by the Education Ministry Social Science of China (13JZD016),
by the Fundamental Research Funds for the Central Universities (2011JQ007), and by the Fundamental Research Funds for the
Central Universities of Central South University (2013zzts003).

References

Ang, B. W., & Liu, N. (2007). Handing zero values in the logarithmic mean Divisia index decomposition approach. Energy policy, 35, 238–246.
Ansuategi, A. (2003). Economic growth and trans-boundary pollution in Europe: An empirical analysis. Environmental and Resource Economics, 26, 305–328.
Antweiler, W., Copeland, B. R., & Taylor, M. S. (2001). Is free trade good for the environment? American Economic Review, 91, 877–908.
Arrow, K., Bolin, B., Costanza, R., et al. (1995). Economic growth, carrying capacity, and the environment. Ecological Economics, 15, 91–95.
Bond, S. R. (2002). Dynamic panel data models: A guide to micro data methods and practice. Portuguese Economic Journal, 1, 141–162.
Chinese Electric Power Yearbook (2002-2010). National Bureau of Statistics: Beijing, China.
Chinese Energy Statistical Yearbook (2011). National Bureau of Statistics: Beijing, China.
Chinese Statistical Yearbook (2000-2010). National Bureau of Statistics: Beijing, China.
Chew, G. L. (2009). Foreign direct investment, pollution and economic growth: evidence form Malaysia. Applied Economics, 41, 1709–1716.
Cole, M. A. (2004). Trade, the pollution haven hypothesis and the environmental Kuznets curve: Examining the linkages. Ecological Economics, 48, 71–81.
Cole, M. A., & Fredriksson, P. G. (2009). Institutionalized pollution havens. Ecological Economics, 68, 1239–1256.
Cole, M. A., Rayner, A. J., & Bates, J. M. (1997). The environmental Kuznets curve: An empirical analysis. Environment and Development Economics, 2, 401–416.
Copeland, B., & Taylor, M. S. (1994). North–south trade and the environment. Quarterly Journal of Economics, 109(3), 755–787.
Copeland, B., & Taylor, M. S. (1997). A simple model of trade, capital mobility and the environment. Working paper, No. 5898, National Bureau of Economics
Research.
De Bruyn, S. (2002). Dematerialization and rematerialization as two recurring phenomena of industrial ecology. In R. U. Ayres, & L. W. Ayres (Eds.), A handbook of
industrial ecology (pp. 209–222). Cheltenham Edward Elgar.
Dean, J. M., Lovely, M. E., & Wang, H. (2009). Are foreign investors attracted to weak environmental regulations? Evaluating the evidence from China. Journal of
Development Economics, 90, 1–13.
Dong, Y. -L., Masanobu, I., Liu, X. -B., & Wang, C. (2010). An analysis of the driving forces of CO2 emissions embodied in Japan–China trade. Energy Policy, 38,
6784–6792.
Emilson, C. D., & Zhu, X. (2009). Emissions trading of global and local pollutions, pollution havens and free riding. Journal of Environmental Economics and
Management, 58, 169–182.
Eskeland, G. S., & Harrison, A. E. (2003). Moving to greener pastures? Multinationals and the pollution haven hypothesis. Development Economy, 70, 1–23.
Frankel, J., & Romer, D. (1999). Does trade cause growth? American Economic Review, 89, 379–399.
Friedl, B., & Getzner, M. (2003). Determinants of CO2 emissions in a small open economy. Ecological Economics, 45, 133–148.
Gao, X., & Zhang, W. (2013). Foreign investment, innovation capacity and environmental efficiency in China. Mathematical and Computer Modelling, 58,
1040–1046.
Grossman, G., & Krueger, A. (1991). Environmental impacts of a North American free trade agreement. NBER working paper, No. 3194.
Grossman, G. M., & Krueger, A. B. (1995). Economic growth and the environment. Quarterly Journal of Economies, 353–377.
Guan, D., Hubacek, K., Weber, C. L., Peters, G. P., & Reiner, D. M. (2008). The drivers of Chinese CO2 emissions from 1980 to 2030. Global Environmental Change, 18,
626–634.
Guan, D., Peters, G. P., Weber, C. L., & Hubacek, K. (2009). Journey to world top emitter: An analysis of the driving forces of China's recent CO2 emissions surge.
Geophysical Research Letters, 36, L04709, http://dx.doi.org/10.1029/2008GL036540.
Halicioglu, F. (2009). An econometric study of CO2 emissions, energy consumption, income and foreign trade in Turkey. Energy Policy, 37(3), 1156–1164.
IPCC (2006). IPCC national greenhouse gas inventory guide. Japanese Global Environment Strategy Research Institute.
 Author's personal copy

134 S. Ren et al. / China Economic Review 28 (2014) 123–134

Jalil, A., & Feridun, M. (2011). The impact of growth, energy and financial development on the environment in China: A co-integration analysis. Energy Economics,
33, 284–291.
Kankesu, J., Reetu, V., & Liu, Y. (2012). CO2 emissions, energy consumption, trade and income: A comparative analysis of China and India. Energy Policy, 42,
450–460.
Kearsley, A., & Riddel, M. (2010). A further inquiry into the pollution haven hypothesis and the environmental Kuznets curve. Ecological Economics, 69, 905–919.
Leontief, W. (1970). Environmental repercussions and the economic structure: An input–output approach. Review of Economic Statistics, 52, 262–277.
Lin, B. -Q., & Sun, C. -W. (2010). Evaluating carbon dioxide emissions in international trade of China. Energy Policy, 38, 613–621.
Machado, G., Schaeffer, R., & Worrell, E. (2001). Energy and carbon embodied in the international trade of Brazil: An input–output approach. Ecological Economics,
39, 409–424.
Ministry of Commerce of China (2012). http://english.mofcom.gov.cn/
OECD (2011). OECD database. http://www.oecd.org/statistics/
OECD (2010). OECD database. http://www.oecd.org/statistics/
Panayotou, T. (1993). Empirical tests and policy analysis of environmental degradation at different stages of economic development. Working paper, No. 238,
Geneva Technology and Employment Programme, International Labour Office.
Pao, H. -T., & Tsai, C. -M. (2011). Multivariate Granger causality between CO2 emissions, energy consumption, FDI (foreign direct investment) and GDP (gross
domestic product): Evidence from a panel of BRIC (Brazil, Russian Federation, India, and China) countries. Energy, 36, 685–693.
Perkins, R., & Neumayer, E. (2009). Transnational linkages and the spillover of environment-efficiency into developing countries. Global Environmental Change, 19,
375–383.
Roberts, J. T., & Grimes, P. E. (1997). Carbon intensity and economic development 1962–1991: A brief exploration of the environmental Kuznets curve.
World Development, 25(2), 191–198.
Sanchez Choliz, J., & Duarte, R. (2004). CO2 emissions embodied in international trade: Evidence for Spain. Energy Policy, 32, 1999–2005.
Shao, S., Yang, L. -L., Yu, M. -B., & Yu, M. -L. (2011). Estimation, characteristics, and determinants of energy-related industrial CO2 emissions in Shanghai (China),
1994–2009. Energy Policy, 39, 6476–6494.
Vincent, J. R. (1997). Testing for environmental Kuznets curves within a developing country. Environment and Development Economics, 2, 417–431.
Windmeijer, F. (2005). A finite sample correction for the variance of linear efficient two-step GMM estimators. Journal of Econometrics, 126, 25–51.
Xu, M., Li, R., Crittenden, J. C., & Chen, Y. S. (2011). CO2 emissions embodied in China's exports from 2002 to 2008: A structural decomposition analysis.
Energy Policy, 39, 7381–7388.
Yan, Y. -F., & Yang, L. -K. (2010). China's foreign trade and climate change: A case study of CO2 emissions. Energy Policy, 38, 350–356.
Zhou, X. -Y., Zhang, J., & Li, J. -P. (2012). Industrial structural transformation and carbon dioxide emissions in China. Energy Policy, 57, 43–51.