sustainability

Article
Effect of Population Structure Change on Carbon
Emission in China
Wen Guo 1,2, *, Tao Sun 2 and Hongjun Dai 2,3
1 Accounting Department, College of Accounting, Nanjing University of Finance & Economics,
Nanjing 210046, China
2 College of Economics and Management, Research Institute of Financial Development,
Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China; [email protected] (T.S.);
[email protected] (H.D.)
3 College of Economics and Management, Huainan Normal University, Huainan 232038, China
* Correspondence: [email protected]; Tel.: +86-25-8489-3060 (ext. 925)

Academic Editors: Malin Song and Seunghwan Myeong

Received: 25 October 2015; Accepted: 17 February 2016; Published: 4 March 2016

Abstract: This paper expanded the Logarithmic Mean Divisia Index (LMDI) model through the
introduction of urbanization, residents’ consumption, and other factors, and decomposed carbon
emission changes in China into carbon emission factor effect, energy intensity effect, consumption
inhibitory factor effect, urbanization effect, residents’ consumption effect, and population scale effect,
and then explored contribution rates and action mechanisms of the above six factors on change
in carbon emissions in China. Then, the effect of population structure change on carbon emission
was analyzed by taking 2003–2012 as a sample period, and combining this with the panel data
of 30 provinces in China. Results showed that in 2003–2012, total carbon emission increased by
4.2117 billion tons in China. The consumption inhibitory factor effect, urbanization effect, residents’
consumption effect, and population scale effect promoted the increase in carbon emissions, and their
contribution ratios were 27.44%, 12.700%, 74.96%, and 5.90%, respectively. However, the influence of
carbon emission factor effect (´2.54%) and energy intensity effect (´18.46%) on carbon emissions
were negative. Population urbanization has become the main population factor which affects carbon
emission in China. The “Eastern aggregation” phenomenon caused the population scale effect in the
eastern area to be significantly higher than in the central and western regions, but the contribution
rate of its energy intensity effect (´11.10 million tons) was significantly smaller than in the central
(´21.61 million tons) and western regions (´13.29 million tons), and the carbon emission factor effect
in the central area (´3.33 million tons) was significantly higher than that in the eastern (´2.00 million
tons) and western regions (´1.08 million tons). During the sample period, the change in population
age structure, population education structure, and population occupation structure relieved growth
of carbon emissions in China, but the effects of change of population, urban and rural structure,
regional economic level, and population size generated increases in carbon emissions. Finally, the
change of population sex structure had no significant influence on changes in carbon emissions.

Keywords: population structure; carbon emission; LMDI; urbanization; residents’ consumption

1. Introduction
With the intensified change in the global climate, the problem of carbon emissions from burning
fossil energy has become an important focus. The fourth assessment report of IPCC pointed out
that human factors were the main reasons causing a dramatic increase of global carbon emission;
therefore, the effect of expansion of the population scale and change in demographic structure has
become an essential academic issue. Due to historical reasons, the one-child policy led to the current

Sustainability 2016, 8, 225; doi:10.3390/su8030225 www.mdpi.com/journal/sustainability

Sustainability 2016, 8, 225 2 of 20

Sustainability 2016, 8, 225 2 of 21

special population structure in China, and also provided typical sample data for research into the
effect of change
effect of of demographic
change structure
of demographic ononcarbon
structure emission.
carbon emission. The
The change
change of Chinese
of Chinese demographic
demographic
structure (including population age structure, population sex structure, population education
structure (including population age structure, population sex structure, population education structure,
structure, population urban and rural structure and population occupation structure) is shown in
population urban and rural structure and population occupation structure) is shown in Figure 1.
Figure 1.

60.00 125.00

50.00 120.00

115.00
40.00
110.00
30.00
105.00
20.00
100.00
10.00 95.00

0.00 90.00
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

cdr odr psh pca ur pes tsr bsr

Figure 1. Change of Chinese population structure from 2000 to 2013.

Figure 1. Change of Chinese population structure from 2000 to 2013. Data source: (1) Data of children
dependency Dataratio (cdr),
source: elderly
(1) Data dependency
of children dependency ratio
ratio(odr), proportion
(cdr), elderly of theratio
dependency population in senior high
(odr), proportion
school (psh), proportion
of the ofsenior
population in the population in college
high school (psh), andofabove
proportion (pca), ratio
the population of population
in college urbanization
and above (pca),
ratio of
(ur), sex ratio of population urbanization
total population (tsr),(ur),
andsexsex
ratio of total
ratio population
at birth (tsr), and
(bsr) were sex ratiofrom
collected at birth (bsr)Statistical
China
were collected from China Statistical Yearbook [1] for 2001–2014, while data of children dependency
Yearbook [1] for 2001–2014, while data of children dependency ratio and elderly dependency ratio in
ratio and elderly dependency ratio in 2000–2001 were obtained from China Commonly Used
2000–2001Population
were obtained from China Commonly Used Population Data Set Since 1990 [2]. (2) Data of
Data Set Since 1990 [2]. (2) Data of the proportion of employees in state‐owned units in
the proportion of employees
total China in state-owned
urban employment unitstheinChina
(pes) were from totalLabour
China urbanYearbook
Statistical employment (pes) were from
[3] in 2001–2013.
the China Labour Statistical Yearbook [3] in 2001–2013.
At present, there are significant features regarding aspects of age structure, urban and rural
structure, education structure, sex structure, and occupational structure in the Chinese population.
At present, there
Firstly, China are tosignificant
began carry out thefeatures
one‐child regarding
policy in the aspects
1970s, whichof age structure,
was officially urban asand rural
designated
structure,the
education
basic national structure,
policy insex
1982.structure,
This policyandcausedoccupational
low birth rates structure
over a long inperiod.
the Chinese population.
In addition,
improved
Firstly, China began living standards,
to carry out advances
the one-childin medical
policytechnology, and other
in the 1970s, whichfactors
was also caused a designated
officially rapid as
the basic decline
national in the death rate. The phenomenon of low birth rates and low death rates led to the main features
policy in 1982. This policy caused low birth rates over a long period. In addition,
of age structure in China as being “declining birthrates” and “aging populations”. According to
improvedFigure
living 1, itstandards,
was found thatadvances
the child in medicalratio
dependency technology, and dependency
and the elderly other factors ratioalso caused a rapid
continuously
decline inincreased
the death sincerate. Themet
2010, and phenomenon of low birth
the above characteristic rates andage
of the population low deathSecondly,
structure. rates led to the main
China
features ofhas entered
age a rapidindevelopment
structure China as beingstage of populationbirthrates”
“declining urbanization and since “aging
the 1980s. In 2000–2013, the
populations”. According
population urbanization ratio in China increased from 36.22% to 53.73% (see
to Figure 1, it was found that the child dependency ratio and the elderly dependency ratio continuously Figure 1), and maintained
an average annual growth rate of more than 1%. Wang [4] indicated that population urbanization in
increased China
sinceis2010, and met the above characteristic of the population age structure. Secondly, China
still in the middle stage, and exhibits disparity with the world’s average level. We can expect
has entered
thata Chinese
rapid development
population urban stageandofrural
population
structure urbanization
will change with since
the therapid1980s. In 2000–2013,
development of the
population urbanization
population ratio in China
urbanization Chinainincreased from 36.22%
the future. Thirdly, Figure 1toshowed
53.73%that (seetheFigure
Chinese1), and maintained
population
an averageeducation
annualstructure
growthalso changed
rate of more greatly since
than 1%.theWang
new century, the proportion
[4] indicated that of the population
population of
urbanization
senior high school and proportion of the population in college and above, respectively, rose from
in China is still in the middle stage, and exhibits disparity with the world’s average level. We can
13.20% and 4.28% to 16.52% and 11.32% in 2000–2013, indicating the increase of population quality in
expect that Chinese
China. Fourthly,population urban and
Chinese population rural structure
sex structure will change
was in long‐term imbalancewith duethe rapid
to the development of
child‐bearing
population urbanization in China in the future. Thirdly, Figure 1 showed that the
viewpoint of “preference for sons over daughters”. Figure 1 also showed that the Chinese total population Chinese population
educationsexstructure alsoabout
ratio has been changed greatly
105 since since the new
2000, significantly highercentury,
than the the proportion
normal level. On the of other
the population
hand, of
the sex
senior high ratio at
school andbirth in China wasof
proportion higher
the than 115 in 2000–2013;
population the cumulative
in college and above, effectrespectively,
will undoubtedly rose from
lead to more serious population gender imbalances in the future. Finally, due to historical reasons, the
13.20% and 4.28% to 16.52% and 11.32% in 2000–2013, indicating the increase of population quality in
state‐owned economic component played an important role in the Chinese economy, with the depth
China. Fourthly,
of market Chinese population sex structure was in long-term imbalance due to the child-bearing
economic reform, and the proportion of the state‐owned economy gradually reducing.
viewpoint of “preference for sons over daughters”. Figure 1 also showed that the Chinese total
population sex ratio has been about 105 since 2000, significantly higher than the normal level. On the
other hand, the sex ratio at birth in China was higher than 115 in 2000–2013; the cumulative effect will
undoubtedly lead to more serious population gender imbalances in the future. Finally, due to historical
reasons, the state-owned economic component played an important role in the Chinese economy,
with the depth of market economic reform, and the proportion of the state-owned economy gradually
reducing. The proportion of employees in state-owned units in total Chinese urban employment
rapidly declined in recent years (see Figure 1), which confirmed changes of population occupational
structure in China.
Sustainability 2016, 8, 225 3 of 20

Regarding the effect of population structure change on carbon emission, the conclusions of
scholars are varied. Liddle [5] found that the effect of population age structure on carbon emission
from transport energy consumption presented a change tendency as an inverted “U” type, while
Okada [6], and Menz et al. [7] believed that aging populations would reduce the overall carbon
emission from energy consumption. Zhu et al. [8], Knight et al. [9], and many other scholars found
that the level of urbanization significantly affected overall carbon emission. Katircioğlu [10] took
energy consumption in Turkey as a case; his study showed that the development of high education
level and improvement of population quality will promote growth of the power consumption and
the petroleum consumption regardless of long term or short term timeframes. Obviously, population
structure change significantly affected change of carbon emission; therefore, this paper examined
change of carbon emission of 31 provinces in China during 2003–2012 as the research object, focusing
on the analysis of the effect of China’s population structure change on carbon emission from the
perspectives of development of urbanization and residents’ consumption, and attempted to explain
the pattern of change in carbon emission in China from the perspective of population structure change.
The following content of this paper was organized as follows. In Section 2, we summarize the
literature about this issue. Section 3 discusses the model, methodology, and the data which was
used in this paper. Results were calculated and empirical results are summarized in Sections 4 and 5
respectively. Finally, Section 6 concludes this paper.

2. Literature Review
At present, structural decomposition analysis (SDA) and index decomposition method (IDA)
are the two main decomposition methods to analyze carbon emission. Although the two methods
are widely applied [11,12], Hoekstra et al. [13], and Ang [14] believed that the IDA method adapts
to the empirical analysis at all levels as macro, regional, and industry based on a comparison of the
above two methods. The IDA method also has advantage in data acquisition; therefore, it was more
popular in academic study. The IDA method can be further subdivided into two types as the Laspeyres
Index Decomposition Method and the Divisia Index Decomposition Method. Ang et al. [15] found
that the Logarithmic Mean Divisia Index (LMDI) decomposition method has advantages of handling
null, independent data paths compared with the Laspeyres Index Decomposition Method, and it
was more applicable to empirical research. Therefore, the application of the LMDI method in the
field of carbon emission measurement and decomposition was more extensive in recent years [16,17].
Many scholars also put forward many corresponding improved forms of LMDI on the basis of specific
study objects [18–20].
Regarding the relationship between carbon emission and demographic factors, the early studies
mostly focus on the influence of population scale on carbon emission. Rosa et al. [21] collected the
cross-sectional data and found that the population elasticity was close to 1. O’Neill et al. [22] also
believed that the impact of population on carbon emission was only manifested by the scale effect,
while controlling other influencing factors, with elasticity close to 1. Ping [23] also obtained a similar
conclusion. However, Satterthwaite [24] found that the growth of population scale could promote the
growth of carbon emission, while the contribution of consumption level and consumption scale, which
grows with the growth of population scale, on promoting the growth of carbon emission was greater.
Yao et al. [25], and Wang et al. [26] took time-series data at the national level in China as a sample; their
studies found that growth of consumption scale and rising consumption levels were important factors
influencing the growth of carbon emission.
In recent years, the influence of population structure change on carbon emission has drawn more
and more attention. Firstly, change of population age structure (aging populations) is a kind of typical
population structure change, and has significant impact on change of carbon emission. York [27]
believed that population size and age structure affected carbon emission significantly through affecting
total energy consumption. Liddle et al. [28] analyzed the effect of two age groups, 20–34 and 35–60,
on carbon emission. The results showed that elasticity of carbon emissions in the two samples were
Sustainability 2016, 8, 225 4 of 20

contradictory; elasticity of the former group is positive, but the other is negative. Lugauer et al. [29]
found that carbon emission increased rapidly with the increasing of the proportion of the population
aged 35 to 49. Liddle [5] analyzed the impact of age structure on carbon emission from residents’
electricity consumption and transportation energy consumption. He believed that the impact of age
structure on carbon emission from transportation energy consumption appeared as an inverted “U”
type; namely, the influence coefficients of the minimum and maximum age groups were negative,
and the influence coefficient of the age group of 20–34 is positive, while the influence direction of age
structure on carbon emission from residents’ electricity consumption was the inverse. Zagheni [30]
argued that total carbon emission would decrease with the growth of the proportion of the population
over the age of 60; namely, an increase of elderly population could reduce carbon emission, and the
conclusion of the study of Okada [6], Menz et al. [7] was similar. In addition, Laureti et al. [31] took
attention of the effect of population structure on nitrogen oxides; this study also supports the above
view’s conclusion.
Secondly, many scholars studied the impact of population urban and rural structure change
(urbanization) on carbon emission, and there were two methods of research for this topic. The first
assumed that there was a single causal relationship from urbanization to carbon emission, to explore
the influence of urbanization on carbon emission. However, due to differences between their study
samples and other factors, they did not reach the same conclusion. Jorgenson [32] took OECD
countries as samples and found that the effect of urbanization on carbon emission in the national
level was not significant, Liddle et al. [28] and Jorgenson [33] obtained a similar conclusion. Jorgenson
et al. [34,35] found that urbanization had a significant effect on carbon emission, while its effect
coefficient is small based on a mixed sample with data of developed countries and developing countries.
Studies by Poumanyvong et al. [36], Martinez et al. [37], Poumanyvong et al. [38], Zhu et al. [8],
and Knight et al. [9] supported the above conclusion. Fang et al. [39] argued that urbanization
did not have a significant effect on carbon emission in low-income countries, but it was able to
significantly reduce carbon emission in high-income countries. The second tested the two-way causality
relationship between urbanization and carbon emission. Mishra et al. [40] found that urbanization was
a one-way Grainger reason to change carbon emission in Pacific island countries in the short-term.
Hossain [41] believed that there was no causal relationship between urbanization and carbon emission
in newly-industrialized countries in the short-term. Al-mulali [42] verified the two-way and long-term
positive causal relationship between urbanization and carbon emission in the Middle East and North
African countries.
Although some literature considered the effect of population scale change, population structure
change, and change of population age structure on carbon emission [43,44], most literature still ignored
the effect of population sex structure change and population education and occupation structure
change mentioned previously. In addition, although the positive effect of increasing the size and level
of consumption on carbon emission has become a consensus, there were few studies related to the
effect of change of consumption structure on carbon emission. Finally, the early one-child policy caused
China to face the transformation period of population structure change, as mentioned previously;
population age structure, population sex structure, population education structure, population urban
and rural structure, and population occupation structure significantly changed from 1997 to 2013,
and it provided a typical sample for this study. Therefore, we attempted to expand the study of this
topic via the following two aspects: (1) introducing factors of demographic urbanization and residents’
consumption into Kaya identity, and decomposed change of carbon emission into six effects as carbon
emission factor effect, energy intensity effect, consumption inhibitory factor effect, urbanization
effect, residents’ consumption effect, and population scale effect through using LMDI method; their
contribution rate and action mechanism on change of carbon emission was then discussed; and
(2) empirically analyzed effects of regional population structure change on carbon emission and its
decomposition effects by using panel data of 29 provinces in China from 2003 to 2012.
Sustainability 2016, 8, 225 5 of 20

3. Methodology and Data

3.1. LMDI Decomposed Method

According to Kaya identity, the relationship between population, energy, economic, and carbon
emission can be expressed as Equation (1):

C E G
C“ ˆ ˆ ˆP (1)
E G P
where C, E, G, and P express total carbon emission, total energy consumption, GDP, and total
population, respectively. We attempted to analyze the effect of population structure change on
carbon emission from the perspectives of urbanization and residents’ consumption based on Kaya
identity, and used RC to express residents’ consumption, and Equation (1) can be further extended as:

C E G RC
C“ ˆ ˆ ˆ ˆP (2)
E G RC P

In addition, Pu and Pr were used to express total population, RC bu , RCrb to express basic living
consumption of urban and rural residents, including five classes as consumption of food, clothing,
housing, household appliances and services, transportation, and communication, RCib ib
u , RC r to express
non-basic living consumption, including three categories: health care, education and entertainment
products and services, and other goods and services. Then, residents’ consumption RC can be
converted as: « ff „
RC bu RCib RCrb RCrib Pu Pu Pr Pr T

u
RC “ ˆ ˆP (3)
Pu Pu Pr Pr P P P P

We can deduce the expanded Kaya equation which is based on perspectives of urbanization and
residents’ consumption through substituting Equation (3) above into Equation (2):
« ff „ T
C E G RC bu RCib
u RCrb RCrib Pu Pu Pr Pr
C“ ˆ ˆ ˆ ˆ ˆP (4)
E G RC Pu Pu Pr Pr P P P P

C
Assumed: ci “ expresses carbon emission per unit of energy, namely carbon emission factor,
E
E G
ei “ expresses energy consumption per unit of GDP, namely energy intensity, ci f “
G RC
expresses the inverse of residents’ consumption ratio, namely consumption inhibitory factor,
RC bu ib RCib
u RCrb ib RCrib
rcbu “ , rcu “ , rcrb “ , rcr “ expresses basic and non-basic living consumption
Pu Pu Pr Pr
per urban resident, basic and non-basic living consumption per rural resident, respectively.
Pu Pr
pu “ , pr “ expresses the proportions of the urban and rural population, respectively. Then,
P P
Equation (4) can be translated to Equation (5):
” ı
C “ ci ˆ ei ˆ ci f ˆ rcbu rcib
u rcrb rcrib ˆ rpu pu pr pr s T ˆ P (5)

We can deduce Equation (6) through using the log and differential to Equation (5):
´ ¯
d rcbu pu ` rcib
u up ` rc b p ` rcib p
r r r r
d lnC “ d lnci ` d lnei ` d lnci f ` ` d lnP (6)
rcbu pu ` rcib b ib
u pu ` rcr pr ` rcr pr
Sustainability 2016, 8, 225 6 of 20

rcbu pu
Assumed: sbu “ expresses the proportion of basic living
rcbu pu ` rcib b ib
u pu ` rcr pr ` rcr pr
rcib
u pu
consumption in urban residents’ total living consumption, sib u “
rcu pu ` rcu pu ` rcrb pr ` rcrib pr
b ib
expresses the proportion of non-basic living consumption in urban residents’ total living consumption,
rcrb pr
srb “ b expresses the proportion of basic living consumption in rural
rcu pu ` rcib b ib
u pu ` rcr pr ` rcr pr
rcib
u pu
residents’ total living consumption, sib u “ expresses the proportion
rcu pu ` rcu pu ` rcrb pr ` rcrib pr
b ib
of non-basic living consumption in rural residents’ total living consumption. We can deduce
Equation (7) as:
ˆ
b ib b ib
´ ¯sbu ´ ¯sibu ´ ¯srb ´ ¯srib ˙
d lnC “ d ln ci ˆ ei ˆ ci f ˆ pu su ˆ pu su ˆ pr sr ˆ pr sr ˆ rcbu ˆ rcid
u ˆ rcrb ˆ rcrib ˆP (7)

Equation (7) can be translated to Equation (8); Equation (8) is the multiplicative form of carbon
emission based on factor decomposition:

b ib b ib
´ ¯sbu ´ ¯sibu ´ ¯srb ´ ¯srib
C “ ci ˆ ei ˆ ci f ˆ pu su ˆ pu su ˆ pr sr ˆ pr sr ˆ rcbu ˆ rcib
u ˆ rcrb ˆ rcrib ˆP (8)

In order to measure the contribution value of carbon emission factor and other variables on carbon
emission, we should decompose change of carbon emission into the sum or product of the above
variables. This paper adopted the LMDI plus decomposition method to solve the above problem which
was referred to by Ang [14]. The total effect of change of carbon emission (∆C) from period 0 to period T
can be decomposed into six effects as carbon emission factor effect (∆Cci ), energy intensity effect (∆Cei ),
consumption inhibitory factor effect (∆Cci f ), urbanization effect (∆Cu´r ), residents’ consumption effect
(∆Cb´ib ), and population scale effect (∆CP ). Their calculation formulas are as follows:

´ ¯ ci T
∆Cci “ L C T , C0 ˆ ln 0 (9)
ci
´ ¯ ei T
∆Cei “ L C T , C0 ˆ ln 0 (10)
ei
´ ¯ ci f T
∆Cci f “ L C T , C0 ˆ ln 0 (11)
ci f
˘ !”´ b T T
¯ ´ 0 0
¯ ı ”´ T T
¯ ´ 0 0
¯ ı)
∆Cu´r “ L C T , C0 ˆ su ` sib lnpu T ´ sbu ` sib lnpu 0 ` srb ` srib lnpr T ´ srb ` srib lnpr 0 (12)
`
u u
!” T T T T 0 0 0 0
ı ” T T T T 0 0 0 0
ı)
∆Cb´ib “ L C T , C0 ˆ sbu lnrcbu ` srb lnrcrb ´ sbu lnrcbu ´ srb lnrcrb ` sib ib ib ib ib ib ib ib (13)
` ˘
u lnrcu ` sr lnrcr ´ su lnrcu ´ sr lnrcr

´ ¯ PT
∆CP “ L C T , C0 ˆ ln 0 (14)
P
∆C “ ∆Cci ` ∆Cei ` ∆Cci f ` ∆Cu´r ` ∆Cb´ib ` ∆CP (15)

where: $ ` ˘ ´ ¯
’ C T ´ C0 { lnC T ´ lnC0 , C T ‰ C0
´ ¯ &
L C T , C0 “ C0 , C T “ C0 (16)
’
0, C T “ C0 “ 0
%

3.2. Variables and Data

Our work focused on regional carbon emission and its decomposition effects of 31 provinces in
China from 2003 to 2012. Due to the lack of data for Tibet, we took panel data of the other 30 provinces as
the sample. We selected variables and collected data from two routes. Firstly, in the study of calculation
Sustainability 2016, 8, 225 7 of 20

and decomposition of regional carbon emission, we used 20 kinds of terminal energy consumption in
various regions and a CO2 emission coefficient to determine regional total carbon emissions (C); this
method can overcome the defect of unjust allocations in regions compared with using the three kinds of
primary energy. The above 20 kinds of terminal energy consumption in various regions was shown in
China Energy Statistical Yearbook [45] to express regional total energy consumption, and was converted
into standard coal equivalent. Data of regional total GDP (G), regional total residents’ consumption (RC),
regional total population (P), the proportion of urban population (pu ) and rural population (pR ) were
from the China Statistical Yearbook [1]. The used proportion of consumption on food, clothing, housing,
household appliances and services, transportation, and communication of per person in residents’ per
capita total consumption to express the proportion of basic living consumption of rural residents (srib )
data was also from China Statistical Yearbook [1], and the proportion of non-basic living consumption of
rural residents (srb ) can be calculated by 1 ´ srib . The calculation of these two items of data for urban
residents is the same for rural residents. Secondly, the empirical study used the children dependency
ratio (children) and the elderly people dependency ratio (elderly) to refer to regional population age
structure, used the sex ratio of total population (sex) to refer to regional population sex structure, and
used proportion of the population of senior high school (senior) and proportion of the population in
college and above (college) to refer to the regional population education structure. The data for the above
variables are from the China Statistical Yearbook [1]. The used proportion of employees in state-owned
units in total Chinese urban employment (occupation) refers to regional population occupation structure,
its data is from the China Labour Statistical Yearbook [3], and used the proportion of urban population
to refer to population of urban and rural (urban) structure. In addition, regional total GDP (GDP),
regional total population (population) were selected as control variables. After the collection, sorting,
and screening of the data, its descriptive statistical results were shown in Table 1.

Table 1. Descriptive statistic results of the variables.

Variables Unit Max Min Mean Standard Deviation

C ten thousand tons 61,360.5100 1257.0579 17,153.4307 11,620.5029
E ten thousand tce 27,650.3057 618.6491 7503.2948 5020.2217
G (GDP) hundred million RMB 44,217.1630 390.2000 8827.2409 7853.7926
RC hundred million RMB 14,386.5689 161.7782 2572.7959 2111.0474
pu % 89.3000 21.0472 48.4049 14.7594
pr % 78.9528 10.7000 51.5951 14.7594
sbu % 72.3332 35.2805 54.9927 6.8414
sib
u % 28.0889 9.8375 16.9141 3.3583
srb % 43.3378 4.0469 22.8320 7.9634
srib % 9.9767 0.8790 5.2612 1.7912
P (population) ten thousand persons 10,594.0000 534.0000 4363.0800 2631.6981
children % 44.6500 9.6400 24.6453 7.3184
elderly % 21.8800 7.4400 12.2259 2.3825
sex Male/Female 115.2300 94.9200 103.9320 3.4372
senior % 29.0483 4.8842 14.3981 4.0935
college % 37.3503 1.8284 8.4561 5.5171
occupation % 62.9000 11.5000 36.1880 11.6511
urban % 89.3000 21.0472 48.4049 14.7594
Data sources: the author sorted and obtained data through collecting initial data from the China Statistical
Yearbook [1] in 2004–2013, China Energy Statistical Yearbook [45] in 2004–2013, and the China Labour Statistical
Yearbook [3] in 2004–2013. Data of regional GDP and regional total residents’ consumption was adjusted by
taking 2003 as the base period.

4. Decomposed Results and Discussion

According to the above statistical results, these were combined with the decomposed method and
Equations (9)–(16) to calculate change of carbon emission in China from 2003 to 2012. The calculated
results were shown in Figure 2.
 Sustainability 2016, 8, 225 8 of 21
Sustainability 2016, 8, 225 8 of 20
4. Decomposed Results and Discussion
According
According to the above
to Figure 2, it statistical
was found results,
thatthese
totalwere combined
carbon with the
emission decomposed
(tce) in Chinamethod
increasedand by
Equations (9)–(16) to calculate change of carbon emission in China from
4.2117 billion tons from 2003 to 2012. Consumption inhibitory factor effect, urbanization effect, 2003 to 2012. The calculated
results were shown in Figure 2.
residents’ consumption effect, and population scale effect, which related factors of population and
According to Figure 2, it was found that total carbon emission (tce) in China increased by 4.2117
residents’ consumption, promoted the increasing of carbon emission in China. The influence of
billion tons from 2003 to 2012. Consumption inhibitory factor effect, urbanization effect, residents’
residents’ consumption
consumption effect, effect (rce) is the
and population largest
scale effect,among
which them,
related followed
factors of by consumption
population inhibitory
and residents’
factor effect (cife). These
consumption, led carbon
promoted emissionsoftocarbon
the increasing rise byemission
3.7222 andin 1.1595
China. billion tons during
The influence the sample
of residents’
period,consumption
respectively. effect (rce) is the largest among them, followed by consumption inhibitory factor effectemission
Urbanization effect (ue) and population scale effect (pce) drove carbon (cife).
increases
Theseof led
0.5842
carbonand 0.2985 billion
emissions tons,
to rise by respectively.
3.7222 However,
and 1.1595 billion the impact
tons during of carbon
the sample period, emission
respectively. factor
Urbanization
effect and energy effect (ue) and
intensity population
effect on carbonscaleemission
effect (pce)was
drovenegative
carbon emission increasesThe
in 2003–2012. of 0.5842 and of
influence
energy0.2985
intensitybillion tons,
effect respectively.
(eie) was more However,
obvious, andthe itimpact
causedof carbon
carbon emission
emission factor effectofand
reduction energy
1.4161 billion
intensity effect on carbon emission was negative in 2003–2012. The influence of
tons from 2003 to 2012, while the contribution value of the carbon emission factor effect (cefe) was only energy intensity effect
(eie) was more obvious, and it caused carbon emission reduction of 1.4161 billion tons from 2003 to 2012,
´0.1366 billion tons. Wang [4] believed that the carbon emission factor and energy intensity embodied
while the contribution value of the carbon emission factor effect (cefe) was only −0.1366 billion tons.
a technical level of energy utilization to some degree. Therefore, the above result showed that energy
Wang [4] believed that the carbon emission factor and energy intensity embodied a technical level of
utilization
energy technology
utilization towas some improved during the
degree. Therefore, theabove
sample period
result showed inthat
China,
energyand reachedtechnology
utilization the effect of
“emission
was improved during the sample period in China, and reached the effect of “emission reduction due
reduction by technology”. However, overall carbon emission in China was increasing by to
the influence of a However,
technology”. series of external variables
overall carbon as total
emission in Chinaconsumption
was increasing of due
residents, population
to the influence scaleofand
of a series
urbanization, and so on.
external variables as total consumption of residents, population scale and urbanization, and so on.

45
Carbon emissions (hundred million tons)

35

25

15

-5 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

Years
-15
cefe cie cife ue rce pse tce

Figure
Figure 2. Decomposed
2. Decomposed results
results of of total
total carbonemission
carbon emission in
in China
Chinafrom
from2003
2003toto
2012.
2012.

Figure 3 showed the mean contribution ratio of each decomposed effect of carbon emission. During
Figure
2003 to32012,
showed the mean ratio
the contribution contribution
of residents’ratio of each decomposed
consumption effect was 74.96% effect
andof carbon
it was emission.
the largest
Duringfactor,
2003 followed
to 2012, bytheconsumption
contributioninhibitory
ratio of factor
residents’ consumption effect was 74.96% and
effect and energy intensity effect—their contribution it was the
largestratios
factor,
werefollowed
27.44% andby consumption inhibitory the
−18.46%, respectively—and factor effect and
contribution energy
ratios of theintensity effect—their
urbanization effect,
population
contribution scalewere
ratios effect,27.44%
and carbonandemission
´18.46%, factor effect were 12.70%, the
respectively—and 5.90%, and −2.54%. ratios
contribution Residents’
of the
consumption
urbanization effect,effect and consumption
population inhibitory
scale effect, factor effect
and carbon were the
emission largest
factor affected
effect werefactors
12.70%, which
5.90%,
affect the
and ´2.54%. change of consumption
Residents’ regional carboneffect
emission
and at the regional level.
consumption However,
inhibitory there
factor waswere
effect a significant
the largest
difference between carbon emission factor effect, energy intensity effect, and population size effect in three
affected factors which affect the change of regional carbon emission at the regional level. However,
regions. The contribution ratio of the energy intensity effect in the eastern region was significantly smaller
there was a significant difference between carbon emission factor effect, energy intensity effect, and
than in central and western regions, but its contribution ratio of population scale effect was significantly
population size the
larger than effect
otherintwo
three regions.
regions. The contribution
The carbon emission factorratio
effect of thecentral
of the energy intensity
region effect in the
was significantly
easternlarger
regionthan that of the eastern and western regions, and its population scale effect was almost 0. We wouldratio
was significantly smaller than in central and western regions, but its contribution
of population scale effect
analyze in‐depth was significantly
the decomposed results oflarger
carbon than thefrom
emission other twoaspects
three regions. The carbon
as technology, emission
residents’
factor effect of the central
consumption, region was
and population significantly
factors due to thelarger
above than that of the
characteristics of eastern
decomposedand western
effects inregions,
the
and itsthree regions. scale effect was almost 0. We would analyze in-depth the decomposed results of
population
carbon emission from three aspects as technology, residents’ consumption, and population factors due
to the above characteristics of decomposed effects in the three regions.
Sustainability 2016, 8, 225 9 of 20
Sustainability 2016, 8, 225 9 of 21

0.9

0.7

Sustainability 2016, 8, 225 9 of 21

0.5

0.9
0.3

0.7
0.1

0.5
-0.1 Overall Eastern Central Western

0.3
-0.3
cefe cie cife ue rce pse
0.1
Figure 3. Contribution ratios of six decomposed effects on the increasing of regional carbon emission.
Figure 3. Contribution-0.1
ratios ofOverall
six decomposed effects on Central
Eastern the increasingWestern
of regional carbon emission.
Firstly, the technical factors include two effects of carbon emission factor effect and energy
-0.3
intensitythe
Firstly, effect, their change
technical factorstendencycefe is shown
include twoin effects
cie Figure
cife 4.of The influence
uecarbon of carbon
rce emission
pse emission factor and
factor effect and energy
energy
intensity effect,intensity
their on change
change tendency of the regional carbon
is shown in emission
Figure was negative overall during the sample
Figure 3. Contribution ratios of six decomposed effects on the4. The influence
increasing of regionalof carbon
carbon emission factor
emission.
period. Where, in 2003–2006, the contribution value in the three regions was around 0, they showed an
and energy intensity on change of the regional carbon emission was negative overall during the sample
obvious downward tendency in 2007–2012. It indicated that the effect of “technical emission reduction”
period.inWhere,Firstly, 2003–2006,
the technical the factors include two effectsinofthe carbon emission factor effect and 0, energy
the threein regions was gradually contribution
emerging andvalue strengtheningthree regions
since 2007. was around
The possible explanation they
is showed
intensity effect, their change tendency is shown in Figure 4. The influence of carbon emission factor and
an obvious downward
that after tendencyofinthe2007–2012.
the implementation Kyoto Protocol, It indicated
which wasthat drawntheup effect
at the ofglobal
“technical
climateemission
energy intensity on change of the regional carbon emission was negative overall during the sample
reduction” in
conference the three
of Montreal regions
in 2005,was gradually
the Chinese emerging
government and
positively strengthening
implemented a low
period. Where, in 2003–2006, the contribution value in the three regions was around 0, they showed since
carbon2007. Theanpossible
economy
policy is
explanation onthatthe after
two the levels of technology and
implementation of thepolicy,
Kyoto and obtainedwhich
Protocol, the initial
was effect in
drawn up 2007.
at the global
obvious downward tendency in 2007–2012. It indicated that the effect of “technical emission reduction”
In addition, carbon emission factor effect and energy intensity effect of the central region was always
climate conference
in the three regionsof Montreal
was graduallyin 2005, the Chinese
emerging government
and strengthening positively
since 2007. The implemented
possible explanation a low is carbon
lower than in the eastern and western regions from 2006 to 2012. Namely, contribution of carbon
economy that after the
policy on implementation
the two levelsofofthe Kyoto Protocol,
technology which was
and policy, anddrawn
obtainedup atthe the initial
global effect
climatein 2007.
emission factor effect and energy intensity effect on reduction of carbon emission in the central region
conference of Montreal
In addition, in 2005, the Chinese government positivelyeffect
implemented a low carbon economy
was thecarbon
largest.emission
It showed factor effect
that technological and energy intensity
progress of carbon emission ofinthethecentral
central region
region was was always
policy on the two levels of technology and policy, and obtained the initial effect in 2007.
lower higher
than in than thein eastern
the eastern and andwestern regions
western regions. Thefrom 2006
possible to 2012.
reason was thatNamely, contribution
the foundation of energyof carbon
In addition, carbon emission factor effect and energy intensity effect of the central region was always
utilization
emission factor technology
effect and inenergy
the eastintensity
was better,effect
and led on toreduction
technological ofprogress
carbon becoming
emission more
in thedifficult.
lower than in the eastern and western regions from 2006 to 2012. Namely, contribution of central
carbon region
While
was theemissionthe foundation
largest.factor It showed of the
that economy in
technological the western
progress region was
of carbonpoor, it caused
emission a lack of conditions
effect and energy intensity effect on reduction of carbon emissionin inthe centralregion
the central region was
conducive to technological progress.
higher than
was thein the eastern
largest. and western
It showed regions. progress
that technological The possible reason
of carbon was that
emission in thethe foundation
central region was of energy
higher
utilization than in thein
technology eastern
the
0.0100 and
east western
was regions.
better, and The
led possible
to reason
technological was that
progressthe
0.5000 foundation
becoming of energy
more difficult.
Carbon emissions (hundred million tons)
Carbon emissions (hundred million tons)

utilization technology in the east was better, and Years

led to technological progress becoming more difficult.
While the foundation 0.0000 of the economy in the western region was poor, it caused a lack of conditions
While the foundation of 2003 the economy
2004 2005 in the2007
2006 western
2008 region was2011
2009 2010 poor,
2012 0.3000 a lack of conditions
it caused
conducive to technological -0.0100 progress.
conducive to technological progress.
-0.0200 0.1000
0.0100 0.5000
Carbon emissions (hundred million tons)

-0.0300
Carbon emissions (hundred million tons)

Years -0.1000
0.0000
-0.0400
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 0.3000
-0.0100
-0.0500 -0.3000
-0.0200
-0.0600 0.1000
-0.5000
-0.0300
-0.0700
-0.1000
-0.0400
-0.0800 -0.7000
cefe_Eastern cefe_Central cefe_Wastern
-0.0500 -0.3000
eie_Eastern eie_Central eie_Wastern
-0.0600
-0.5000
Figure 4. Change-0.0700
tendencies of energy intensity effect and carbon emission factor effect in three areas
in 2003–2012. -0.0800 -0.7000
cefe_Eastern cefe_Central cefe_Wastern
Carbon emission factor eie_Eastern
effect was much lower than energy intensity
eie_Central effect; the main reason was
eie_Wastern
that carbon emission coefficients of different types of energy were constant in this paper, and change
Figure 4. Change tendencies of energy intensity effect and carbon emission factor effect in three areas
of carbon
Figure emission
4. Change factor wasoffrom
tendencies the intensity
energy adjustment of energy
effect structure,
and carbon while factor
emission we could findinthat
effect theareas
three
in 2003–2012.
energy consumption
in 2003–2012. in China was still mainly coal and petroleum during the sample period, according
to statistical data from the China Statistical Yearbook [1] in 2004–2013. The proportion of consumption of
Carbon emission factor effect was much lower than energy intensity effect; the main reason was
coal and petroleum in Chinese total energy consumption decreased 2.5% and 2.2%, respectively, in
that carbon
Carbon emission emission coefficients
factor effect wasof different types of
much lower energy
than were constant
energy intensity ineffect;
this paper,
the and
mainchange
reason was
of carbon emission factor was from the adjustment of energy structure, while we could find that the
that carbon emission coefficients of different types of energy were constant in this paper, and change
energy consumption in China was still mainly coal and petroleum during the sample period, according
of carbon emissiondata
to statistical factor
fromwas fromStatistical
the China the adjustment
Yearbook [1]ofinenergy structure,
2004–2013. whileofwe
The proportion could findofthat the
consumption
energy consumption
coal and petroleumin China was still
in Chinese totalmainly
energy coal and petroleum
consumption decreasedduring
2.5% andthe2.2%,
sample period, according
respectively, in
to statistical data from the China Statistical Yearbook [1] in 2004–2013. The proportion of consumption
of coal and petroleum in Chinese total energy consumption decreased 2.5% and 2.2%, respectively,
Sustainability 2016, 8, 225 10 of 20

in 2012 compared to 2003, while the proportion of clean energy (non-fossil energy) rose only 1.2%,
indicating that adjustment of energy structure was minimal. At the provincial level, the proportion of
coal consumption declined 14.29% in Shanxi province, which was far higher than the national average.
This led its carbon emission factor effect to become larger than other provinces, and this result was also
in accordance with the above conclusion. On the other hand, the regional energy intensity
Sustainability 2016, 8, 225
continued
10 of 21
to decline significantly in recent years due to the strict implementation of “energy saving and emission
2012 compared
reducing” policy. It led to to 2003, while the reduction
the effective proportion ofofclean regional (non‐fossil
energycarbon energy) rose
emission; theonly
energy1.2%,intensity
indicating that adjustment of energy structure was minimal. At the provincial level, the proportion of
effect was much higher than carbon emission factor effect.
coal consumption declined 14.29% in Shanxi province, which was far higher than the national average.
Residents’
This ledconsumption
its carbon emission factors include
factor effect consumption
to become inhibitory
larger than other provinces,factor
and thiseffect and
result was alsoresidents’
consumption effect. Figure
in accordance with the 5 presented the change
above conclusion. tendency
On the other hand, theofregional
the above
energytwo effects
intensity from 2003–2012.
continued to
decline significantly in recent years due to the strict implementation
Influence of consumption inhibitory factor effect on the change of carbon emission in three regions of “energy saving and emission
reducing” policy. It led to the effective reduction of regional carbon emission; the energy intensity effect
exhibited an upward trend during the sample period, and did not show its inhibitory effect because its
was much higher than carbon emission factor effect.
impact was positive.Residents’ According
consumption to factors
the expenditure approach
include consumption definition
inhibitory factorofeffect
GDP,and regional
residents’ GDP was
equal to theconsumption
sum of residents’
effect. Figure consumption,
5 presented thegovernment
change tendency consumption,
of the above twofixed-asset investment, and
effects from 2003–2012.
net export, Influence
so the change of consumption
direction inhibitory
of thefactor effect on the inhibitory
consumption change of carbon
factoremission
was the in three
same regions
as summing
exhibited an upward trend during the sample period, and did not show its inhibitory effect because its
government consumption, fixed-asset investment, and net export of domestic demand in Chinese
impact was positive. According to the expenditure approach definition of GDP, regional GDP was equal
economic development, and caused
to the sum of residents’ the reverse
consumption, governmentchange relation fixed‐asset
consumption, betweeninvestment,
regions. Calculation
and net export,results of
consumption so inhibitory factor effect
the change direction of the show that the
consumption carbonfactor
inhibitory emission
was theintensity of the industry
same as summing government dimension
consumption, fixed‐asset investment, and net export of domestic
of residents’ consumer goods may be lower than the other three economic dimensions. Namely, the demand in Chinese economic
development, and caused the reverse change relation between regions. Calculation results of
reduction of the consumption ratio would cause the rise of government consumption, fixed-asset
consumption inhibitory factor effect show that the carbon emission intensity of the industry dimension of
investment,residents’
and net consumer
export when goodsthe may economic
be lower than outputthe was
otherfixed, and thendimensions.
three economic promote the increase
Namely, the of total
carbon emissions.
reductionHowever, according
of the consumption ratio the China
to would causeStatistical
the rise of Yearbook
government [1],consumption,
although the total residents’
fixed‐asset
consumption investment,
and perand net export
capita when the economic
consumption increased,outputrespectively,
was fixed, and then
6.8084promote the increase
trillion and 4.8810of totalthousand
carbon emissions. However, according to the China Statistical Yearbook [1], although the total residents’
RMB from 2003 to 2012, the residents’ consumption ratio dropped from 34.25% to 27.29%, which
consumption and per capita consumption increased, respectively, 6.8084 trillion and 4.8810 thousand
indicated that
RMB it was
from a lack
2003 to residents’ consumption
2012, the residents’ ratioratio
consumption anddropped
carbon from
emission.
34.25%Therefore, the residents’
to 27.29%, which
consumption ratio that
indicated dropped
it was ato lack3.15% in the
residents’ central region,
consumption whichemission.
ratio and carbon was much higher
Therefore, than the eastern
the residents’
and westernconsumption
areas. Its ratio dropped to 3.15%
consumption in the central
inhibitory factorregion, which
effect was was much
also higher than thehigher
significantly eastern and
than that of
western areas. Its consumption inhibitory factor effect was also significantly higher than that of the
the eastern and western areas during this period.
eastern and western areas during this period.

1.6000
Carbon emissions (hundred million tons)

1.4000
1.2000
1.0000
0.8000
0.6000
0.4000
0.2000
Years
0.0000
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
cife_Eastern cife_Central cife_Wastern
rce_Eastern rce_Central rce_Wastern

Figure 5. Change tendencies of consumption inhibitory factor effect and residents’ consumption effect
Figure 5. Change
in threetendencies of consumption inhibitory factor effect and residents’ consumption effect
areas in 2003–2012.
in three areas in 2003–2012.
The contribution value of residents’ consumption on change of regional carbon emission was the
maximum in absolute terms, and it was the most important influencing factor. Figure 5 showed that
The contribution value of effect
residents’ consumption residents’
in threeconsumption on change
regions were 0.1419, of regional
0.1413, and carbon
0.0936 billion tons inemission
2012, and was the
maximum in absolute
they terms,
clearly rose duringand it wasperiod;
the sample the most important
namely, influencing
the rising of factor. Figure
per capita consumption caused 5 showed that
carbon
emission to increase 0.1419, 0.1413, and 0.0936 billion tons, respectively,
residents’ consumption effect in three regions were 0.1419, 0.1413, and 0.0936 billion tons in in the eastern, central, and2012, and
western areas. Residentsʹ consumption effect in the western region was always much lower than that
they clearly rose during the sample period; namely, the rising of per capita consumption caused carbon
of the eastern and central areas in 2003–2012. The possible explanation could be that the economic
emission todevelopment
increase 0.1419,
level and0.1413,
its growth andrate0.0936 billionarea
in the western tons,
wasrespectively,
much lower thaninthe the eastern,
other central, and
two regions,
western areas.
whichResidents’ consumption
caused its growth effect
rate of per capita in the western
consumption region
to be much wasthealways
less than much
eastern and lower
central areas.than that
of the eastern and central areas in 2003–2012. The possible explanation could be that the economic
development level and its growth rate in the western area was much lower than the other two regions,
which caused its growth rate of per capita consumption to be much less than the eastern and central areas.
Sustainability 2016, 8, 225 11 of 20

Sustainability 2016, 8, 225 11 of 21

Equation (13) in this paper divided residents’ consumption effect into two parts as residents’ basic
living and non-basic living
Equation (13) consumption
in this paper dividedeffect (brce,
residents’ nbrce). Residents’
consumption effect intobasic living
two parts consumption
as residents’ basic effect
represents theand
living effects of five
non‐basic kinds
living of basiceffect
consumption living consumption
(brce, (food,
nbrce). Residents’ basicclothing, housing,effect
living consumption household
represents
appliances the effects
and services, of five kinds of basic
transportation, living consumption on
and communication) (food, clothing,
changes in housing, household while
carbon emission,
appliances
residents’ non-basicandliving
services, transportation,
consumption and represents
effect communication) on changes
the effects in carbon
of other kindsemission, while
of consumption on
residents’ non‐basic living consumption effect represents the effects of other kinds of consumption on
change in regional carbon emission. The results are shown in Figure 6.
change in regional carbon emission. The results are shown in Figure 6.

0.15 1.5
Carbon emissions (hundred million tons)

Carbon emissions (hundred million tons)

0.1
0.05
1
Years
0
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
-0.05
0.5
-0.1
-0.15
-0.2 0
brce_Eastern brce_Central brce_Wastern
nbrce_Eastern nbrce_Central nbrce_Wastern

Figure 6. Change tendency of consumption structure in three areas from 2003 to 2012.
Figure 6. Change tendency of consumption structure in three areas from 2003 to 2012.
Figure 6 showed that there was a significant difference between impacts of residents’ basic living
consumption
Figure 6 showed andthat
residents’
therenon‐basic living consumption
was a significant differenceon change
between in regional
impacts carbon emission. basic
of residents’
The change tendency of residents’ basic living consumption effect was similar to residents’
living consumption and residents’ non-basic living consumption on change in regional carbon
consumption effect. Contribution values of residents’ basic living consumption on change of regional
emission. The change tendency of residents’ basic living consumption effect was similar to residents’
carbon emission in eastern, central, and western areas were 0.1409, 0.1284, and 0.0983 billion tons by
consumption effect.residents’
2012. While Contributionnon‐basicvalues
livingofconsumption
residents’ basic effect living
in three consumption
areas changedon as change
graphic of “N” regional
carbon emission
around 0; itinexhibited
eastern,a central,
fluctuating and western
upward trendareas were 0.1409,
in 2003–2005 0.1284, and
and in 2009–2012, and0.0983
appeared billion
to tons
by 2012.fluctuate down in 2006–2008.
While residents’ non-basic Influence
livingofconsumption
residents’ non‐basic
effectliving
in threeconsumption
areas changedeffect onas regional
graphic “N”
around carbon emission presented
0; it exhibited a fluctuating as a weak
upward negative
trend effect, so the main and
in 2003–2005 source inof2009–2012,
residents’ consumption
and appeared to
effect was the residents’ basic living consumption effect. There may be two reasons for this
fluctuate down in 2006–2008. Influence of residents’ non-basic living consumption effect on regional
phenomenon. First, the proportion of residents’ non‐basic living consumption in residents’ living
carbon emission
consumption presented
was very as asmall.
weakAccording
negative effect,
to the so the main
statistical source
data, it wasof residents’
less than 20% consumption
in many effect
was the provinces,
residents’and basicstillliving consumption
kept decreasing. Second,effect. There
residents’ may be
non‐basic twoconsumption
living reasons forcorresponded
this phenomenon.
First, the proportion
with education, of residents’and
entertainment non-basic living
other service consumption
industries. Zhang [46] in found
residents’ livingefficiency
that energy consumption
was very of small.
education, entertainment
According andstatistical
to the other service industries
data, it waswas lesshigh;
thannamely,
20% unit consumption
in many provinces,of theseand still
industries consumed less energy, and then carried less carbon emission. Therefore, impact of change
kept decreasing. Second, residents’ non-basic living consumption corresponded with education,
of residents’ non‐basic living consumption on change of carbon emission was relatively weak.
entertainment and other service industries. Zhang [46] found that energy efficiency of education,
Figure 7 offered impact of population factors on change of carbon emission in three regions,
entertainment and other
mainly including two service
aspects:industries
urbanization was high;
effect namely, unit
and population scaleconsumption
effect. According ofto these
Figure industries
7,
consumed theless energy, and
contribution then carried
of urbanization on less carbon
carbon emission.
emission Therefore,
in the eastern, impact
central, and of change
western of residents’
areas in
non-basicChina
livingwasconsumption
0.0236, 0.0199, on andchange
0.0143 billion
of carbontons,emission
respectively,was from 2003 to 2012.
relatively weak.Contribution of
change of population size on carbon emission was 0.0196,
Figure 7 offered impact of population factors on change of carbon emission in 0.0030, and 0.0024 billion tons, respectively.
three regions, mainly
Obviously, the urbanization effect was much larger than population scale effect in all the three
including two aspects: urbanization effect and population scale effect. According to Figure 7, the
regions, and demographic urbanization had become the main population factor which affects the
contribution of urbanization on carbon emission in the eastern, central, and western areas in China
change of carbon emission during the sample period. In addition, although the population scale effect
was 0.0236,
was 0.0199, and
positive in 0.0143
central andbillion
westerntons, respectively,
regions from 2003was
overall, its contribution to 2012.
alwaysContribution
close to 0, which ofwas
change of
population size on carbon
significantly lower than emission was 0.0196,
the population 0.0030,
scale effect and
in the 0.0024
eastern billion
area, and ittons, respectively.
manifested Obviously,
as a negative
effect in theeffect
the urbanization centralwasarea from
much 2005 to 2008.
larger thanAccording to the conclusion
population scale effect of Cai [47],the
in all there was always
three regions, and
demographic urbanization had become the main population factor which affects the change the
an obvious population “Eastern aggregation” phenomenon in China since 1990s, especially of carbon
central region which was close to the eastern region, and the phenomenon of population flow to the
emission during the sample period. In addition, although the population scale effect was positive in
eastern area was more obvious. It may be the main reason for the above characteristics in regional
central and western
population scaleregions
effect. overall, its contribution was always close to 0, which was significantly
lower than the population scale effect in the eastern area, and it manifested as a negative effect in the
central area from 2005 to 2008. According to the conclusion of Cai [47], there was always an obvious
population “Eastern aggregation” phenomenon in China since 1990s, especially the central region which
was close to the eastern region, and the phenomenon of population flow to the eastern area was more
obvious. It may be the main reason for the above characteristics in regional population scale effect.
Sustainability 2016, 8, 225 12 of 20
Sustainability 2016, 8, 225 12 of 21

0.25

Carbon emissions (hundred million tons)

0.2

0.15

0.1

0.05
Years
0
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
-0.05
ue_Eastern ue_Central ue_Wastern
pce_Eastern pce_Central pce_Wastern

Figure 7. Change tendencies of urbanization effect and population scale effect in three areas from
Figure 7. Change tendencies of urbanization effect and population scale effect in three areas from 2003
2003 to 2012.
to 2012.
5. Empirical Results and Discussion
5. Empirical Results and Discussion
According to the previous study, population structure change has a great impact on the change of
carbon emission.
According to the Inprevious
this paper,study,
we took seven indicators
population structure to represent
changepopulation
has a great structure.
impact Firstly,
on thewe change
of carbon emission. In this paper, we took seven indicators to represent population structure. to
selected the children dependency ratio (children) and the elderly people dependency ratio (elderly) Firstly,
refer to the regional population age structure. The existing studies showed that, due to higher income
we selected the children dependency ratio (children) and the elderly people dependency ratio (elderly)
level and the strong consumption demand of the middle‐aged population, their per capita carbon
to refer to the regional population age structure. The existing studies showed that, due to higher
emission was obviously higher than that of children and the elderly [5,28,29]. Therefore, we expected
incomethatlevel and the coefficients
the regression strong consumption
of these two demand
indicators of werethenegative.
middle-aged Secondly, population,
we used thetheir perofcapita
sex ratio
carbonthe total population (sex) to refer to the regional population sex structure. Due to the different we
emission was obviously higher than that of children and the elderly [5,28,29]. Therefore,
expected that thepreferences
consumption regressionbetween
coefficients
men andof these
women, twopopulation
indicators sex were negative.
structure Secondly,
also affected we used
regional
the sex ratioemission.
carbon of the total population
Thirdly, we used (sex) to refer
proportion of to
thethe regionalofpopulation
population sex structure.
senior high school (senior) and Due theto the
proportion
different of the population
consumption preferences of college and above
between men (college)
and women, to referpopulation
to the regional sexpopulation
structureeducation
also affected
structure.
regional carbon Generally,
emission. income of residents
Thirdly, we wasusedincreased
proportion with the increasing
of the of a high‐degree
population of senioreducation
high school
population, thus causing the growth of residents’ consumption ability
(senior) and the proportion of the population of college and above (college) to refer to the regional and carbon emissions. However,
high‐degree education would also improve residents’ environmental protection consciousness.
population education structure. Generally, income of residents was increased with the increasing of
Therefore, we believed that the effect of population education structure on carbon emission was
a high-degree education population, thus causing the growth of residents’ consumption ability and
uncertain. The data for the above five indicators are from the China Statistical Yearbook [1]. Fourthly, we
carbonusedemissions.
the proportion However, high-degree
of employees education
in state‐owned units would also improve
in total Chinese residents’ (occupation)
urban employment environmental
protection consciousness. Therefore, we believed that the effect of population
to refer to the regional population occupation structure. Compared with other employees, employees education structure
in on
carbon emissionunits
state‐owned washaveuncertain. The income
more stable data for the
and abovesecurity.
old‐age five indicators
It leads toare the from China Statistical
theconsumption
different
Yearbook [1]. between
attitude Fourthly, themwe andused
otherthe proportion
employees, which ofaffected
employees regional incarbon
state-owned
emission.units inistotal
Its data fromChinese
the
China Labour Statistical Yearbook [3]. Finally, we used the proportion
urban employment (occupation) to refer to the regional population occupation structure. Compared of urban population (urban) to refer
with to the employees,
other population structure
employees of urban and rural areas.
in state-owned unitsObviously,
have more thestable
income level and
income andconsumption
old-age security.
capacity of rural residents were far behind the urban residents.
It leads to the different consumption attitude between them and other employees, This suggests that per capita
which carbon
affected
emission of rural residents was significantly less than urban residents, so the population structure of
regional carbon emission. Its data is from the China Labour Statistical Yearbook [3]. Finally, we used the
urban and rural areas was also an important factor which affected regional carbon emission [33−35].
proportion of urban population (urban) to refer to the population structure of urban and rural areas.
This data is also from the China Statistical Yearbook [1].
Obviously, In the
orderincome level and
to validate the consumption capacity
relationship between of rural
change in residents
populationwere far behind
structure and carbonthe urban
residents. Thiswe
emission, suggests
needed that per capita
to conduct carbon
regression emission
analysis. Weof rural
used theresidents
total effectwas significantly
of change of regionalless than
urbancarbon
residents,
emission so the
(∆C)population structure
as the dependent of urban
variable, and took andthe rural
seven areas was also
indicators an important
(children, elderly, sex,factor
whichsenior,
affected regional
college, carbon
occupation, emission
urban) which [33–35].
referred to This data of
change is regional
also from the Chinastructure
population StatisticalinYearbook
China [1].
In order to validate the relationship between change in population structure and carbon scale
as the independent variables. We then used regional GDP (GDP) and regional population emission,
(population)
we needed as controlled
to conduct variables,
regression conductedWe
analysis. regression
used the analysis
total of the effect
effect of population
of change structure
of regional carbon
change on the change in carbon emission. The regression model was
emission (∆C) as the dependent variable, and took the seven indicators (children, elderly, sex, senior,as Equation (17):
college, occupation, urban) C     1children
which referred  to2elderly
change of  3regional
sex   4senior   5collstructure
population ege  in China as the
independent variables. Wethen 6occupation
used  
regional
7urba n
GDP   GDP
(GDP)
8  
and 9 popul
regionalation  
population scale
(17)
(population)
as controlled variables, conducted regression analysis of the effect of population structure change on
the change in carbon emission. The regression model was as Equation (17):

∆C “ α ` β 1 children ` β 2 elderly ` β 3 sex ` β 4 senior ` β 5 college`

(17)
β 6 occupation ` β 7 urban ` β 8 GDP ` β 9 population ` ε
Sustainability 2016, 8, 225 13 of 20

where: α was a constant, βi (i = 1, . . . , 9) and represented the regression coefficient of the variable, and
ε represented random error.
Table 2 shows the correlation matrix of the variables. We find that the dependent variable
is strongly correlated with each independent variable, and the correlation coefficients between
independent variables are relatively weak.

Table 2. Correlation matrix of variables.

Variables ∆C Children Elderly Sex Senior College Occupation Urban GDP Population
∆C 1.000
children ´0.832 1.000
elderly ´0.327 ´0.140 1.000
sex ´0.117 ´0.154 0.159 1.000
senior ´0.452 0.182 ´0.212 ´0.236 1.000
college ´0.367 0.169 ´0.176 ´0.233 ´0.041 1.000
occupation ´0.463 0.220 ´0.118 ´0.263 0.160 0.157 1.000
urban 0.480 ´0.245 0.245 ´0.064 ´0.171 ´0.116 ´0.291 1.000
GDP 0.555 0.286 ´0.300 0.006 0.185 0.251 0.305 ´0.356 1.000
population 0.568 0.342 ´0.294 ´0.113 0.305 0.174 0.218 ´0.163 0.317 1.000

In order to avoid spurious regression, this paper used Levin, Lin, and Chut, Fisher-ADF and Im,
Pesaran, and Shin methods to complete the unit root test. Its results are shown in Table 3. Results
show that the total effect of change of carbon emission (∆C) was 0-stage integration, lower than the
integrated stage of all the other variables. Thus, the sample was suitable for regression analysis.

Table 3. Unit root test of variables.

Level 1st Difference

Test (c, t)
Variables Parameters LLC Fisher-ADF IPS LLC Fisher-ADF IPS
´16.0600 *** 108.6670 *** ´1.5614 * ´14.4271 *** 97.6317 *** ´1.5265 *
∆C
(0.0000) (0.0000) (0.0592) (0.0000) (0.0015) (0.0613)
´9.2836 *** 106.7150 *** ´0.7277 ´32.4082 *** 216.4500 *** ´5.8521 ***
children
(0.0000) (0.0002) (0.2334) (0.0000) (0.0000) (0.0000)
´11.0706 *** 80.8367 ** ´0.7182 ´9.3649 *** 80.6729 ** ´1.4363 *
elderly
(0.0000) (0.0378) (0.2363) (0.0000) (0.0388) (0.0713)
´12.2982 *** 114.0410 *** ´1.7438 ** ´21.3471 *** 109.3150 *** ´2.2791 **
sex
(0.0000) (0.0000) (0.0406) (0.0000) (0.0001) (0.0113)
´10.8482 *** 88.6095 *** ´0.6717 ´8.6645 *** 99.7102 *** ´1.4869 *
senior
(0.0000) (0.0096) (0.2509) (0.0000) (0.0010) (0.0643)
´9.6667 *** 79.6100 ** ´0.3295 ´14.6147 *** 125.5700 *** ´2.0845 **
college
(0.0000) (0.0460) (0.3709) (0.0000) (0.0000) (0.0186)
Statistics ´10.1579 *** 85.4663 ** ´0.8982 ´12.8208 *** 105.0820 *** ´1.3890 *
occupation (Pro.) (0.0000) (0.0171) (0.1845) (0.0000) (0.0003) (0.0824)
´17.9312 *** 112.8150 *** ´3.0377 *** ´26.5032 *** 194.4340 *** ´6.1255 ***
urban
(0.0000) (0.0000) (0.0012) (0.0000) (0.0000) (0.0000)
´4.5980 *** 24.6834 2.2162 ´48.9063 *** 112.7480 *** ´3.1260 ***
GDP
(0.0000) (1.0000) (0.9867) (0.0000) (0.0000) (0.0009)
´7.3813 *** 67.1597 0.4411 ´3.5134 *** 88.2176 ** ´2.0025 **
population
(0.0000) (0.2452) (0.6704) (0.0002) (0.0103) (0.0224)
´26.8766 *** 139.5560 *** ´3.7129 *** ´48.5369 *** 177.8920 *** ´7.6320 ***
∆Cei
(0.0000) (0.0000) (0.0001) (0.0000) (0.0000) (0.0000)
´3.3387 *** 152.1686 *** ´1.4127 * ´14.4752 *** 109.1800 *** ´1.6076 *
∆Cb´ib
(0.0004) (0.0000) (0.0734) (0.0000) (0.0001) (0.0540)
´11.5972 *** 92.0104 ** ´1.5105 * ´16.5986 *** 117.8310 *** ´2.0217 **
∆Cci f
(0.0000) (0.0049) (0.0625) (0.0000) (0.0000) (0.0216)
*, **, ***, respectively represented 1%, 5%, 10% of significance (same in below), (c, t) represented the test included
intercept and trend.

The regression results were shown in Table 4.

Sustainability 2016, 8, 225 14 of 20

Table 4. Regression results by using the panel data.

Dependent Variable ∆C (Model I) ∆Cei (Model II) ∆Cb´ib (Model III) ∆Cci f (Model IV)
Model Random Effect Fixed Effect Random Effect Fixed Effect Random Effect Fixed Effect Random Effect Fixed Effect
α 0.0435 (0.5635) 0.0404 (1.1053) 0.0860 * (1.7391) 0.0893 *** (3.4590) ´0.0507 (´0.9356) ´0.0542 * (´1.9466) 0.0175 (0.6590) 0.0124 (0.7796)
children ´0.5183 (´0.6576) ´0.2921 (´0.3532) 0.0497 (0.0904) ´0.0267 (´0.0457) 0.3410 (0.5736) 0.5643 (0.8955) ´0.9641 *** (´2.9198) ´0.8454 ** (´2.3562)
elderly ´3.7614 *** (´2.9769) ´4.1666 *** (´3.1737) ´0.5438 (´0.6158) ´0.9330 (´1.0068) ´2.7265 *** (´2.8558) ´2.7113 *** (´2.7103) ´0.5854 (´1.0998) ´0.5877 (´1.0318)
sex ´0.6393 (´1.0824) ´0.5301 (´0.8843) ´0.7065 * (´1.7013) ´0.5293 (´1.2508) 0.1372 (0.3060) 0.0893 (0.1954) 0.0035 (0.0139) ´0.0139 (´0.0536)
senior ´0.0132 (´0.0118) ´0.0167 (´0.0146) ´1.3958 * (´1.7748) ´1.4344 * (´1.7861) 0.4019 (0.4731) 0.4584 (0.5288) 1.0580 ** (2.2080) 0.9902 ** (2.0060)
college ´4.3014 *** (´4.5915) ´3.9709 *** (´4.0395) ´2.4658 *** (´3.7756) ´2.2253 *** (´3.2070) ´1.5444 ** (´2.1862) ´1.4575 * (´1.9458) 0.0191 (0.0487) 0.0295 (0.0691)
occupation ´1.7805 *** (´4.0319) ´1.8162 *** (´3.8919) 1.1680 *** (3.7999) 1.3068 *** (3.9672) ´1.4898 *** (´4.4799) ´1.6132 *** (´4.5368) ´1.2401 *** (´6.7425) ´1.3053 *** (´6.4469)
urban 2.1633 *** (2.8974) 2.4480 *** (3.0411) 0.7583 (1.4664) 1.0007 * (1.7612) 1.4307 ** (2.5551) 1.4542 ** (2.3708) ´0.7318 ** (´2.3870) ´0.6403 * (´1.8333)
GDP 1.1435 *** (15.1690) 1.0851 *** (13.6907) ´0.2825 *** (´5.3773) ´0.3258 *** (´5.8226) 1.0508 *** (18.4913) 1.0407 *** (17.2322) 0.3081 *** (9.7790) 0.2924 *** (8.5038)
population 3.9944 *** (3.1241) 3.6307 *** (2.7260) 0.6994 (0.7831) 0.9549 (1.0157) 4.1915 *** (4.3407) 4.1820 *** (4.1208) 4.3960 *** (8.1742) 4.1527 *** (7.1863)
F 165.4951 *** 117.0388 *** 49.2603 *** 30.2704 *** 244.4422 *** 109.7192 *** 77.1463 *** 42.6501 ***
Adj-R2 0.8462 0.9425 0.6175 0.8053 0.8906 0.9389 0.7181 0.8547
Hausman Test 8.7323 (0.4623) 20.3562 ** (0.0158) 5.0647 (0.8286) 4.8763 (0.8450)
LM test 10.9722 (0.2121) 11.3757 (0.2016) 12.9090 (0.1617) 12.5401 (0.1713) 12.1020 (0.1827) 11.9982 (0.1854) 11.7408 (0.1921) 11.8906 (0.1882)
Figures in brackets in the same row as “Hausman Test” indicated accompanying probability; other figures in brackets indicated statistical quantity T of the corresponding variables
(same as below).
Sustainability 2016, 8, 225 15 of 20

Table 4 showed that the Hausman test statistic of model I was 8.7323, and its probability was
0.4623. So the random effect model was more suitable to the sample. Results of the LM test illustrated
that the regression residuals are stable, and the regression result is effective. Firstly, the regression
coefficient of the regional elderly dependency ratio (elderly) was negative, and it had 1% significance,
indicating that the ageing population would reduce carbon emission from energy consumption. Studies
of Liddle et al. [28] and Okada [6] also obtained a similar conclusion. The possible explanation was that
income of the elderly population gradually reduces with the growth of age, so their consumption scale
and consumption levels were much lower than middle-aged or youth population, and then caused
the decreasing of carbon emission. Secondly, the regression coefficient of the regional proportion
of the population of college and above (college) was negative, and it had 1% significance, indicated
that rising of population quality was helpful to decrease regional carbon emission, in contrast with
the conclusion of Katircioğlu [10]. There may be two reasons to explain this result: the first, that
the proportion of the highly educated population not only reflected the level of human capital, but
also was the source of regional technology innovation. Thus, an increase in education was helpful
to renew the energy utilization technology. The second, that this population could effectively fulfill
the development concept of “energy-saving and emission-reducing”, especially in the economic
department of consumption, the highly educated population preferred to choose the environmentally
friendly products. This was helpful to reduce carbon emission. Thirdly, the regression coefficient of
the proportion of employees in state-owned units in total Chinese urban employment (occupation)
was negative, and it had 1% significance, indicating that reducing the proportion of employees
in state-owned units promoted the increasing of carbon emission, which was in contrast with the
expectation of this paper. The proportion of employees in state-owned units reflected the proportion of
the state-owned economy to some extent. Due to the inefficiency of state-owned economic departments,
the rapid development of market economics caused the increasing of energy consumption, and then
caused a rise in carbon emission. This explains why the regression coefficient of occupation was
negative. Finally, the regression coefficients of regional population urban and rural structure (urban),
regional total GDP (GDP) and regional total population (Population) were positive, and were all of 1%
significance. This met our expectation completely, indicating that the development of urbanization,
economic growth, and expansion of the regional population scale would promote the increasing of
carbon emission from energy consumption.
In addition, the carbon emission factor effect (∆Cci ) was decided by the energy emission coefficient
and energy structure; the effect of population structure on it was small. Urbanization effect (∆Cu´r )
expressed the impact of change of urban and rural population on carbon emission. It was only related
to the change of urban and rural population structure (urban), and population scale effect (∆CP ) only
expressed the impact of change of population scale on carbon emission. Obviously, the above three
decomposed effects were hardly affected by the seven indices of population structure change. Therefore,
energy intensity effect (∆Cei ), residents’ consumption effect (∆Cb´ib ) and consumption inhibitory factor
effect (∆Cci f ) may be the main path which conducted the effect of change of population structure
on change in carbon emission. Then we would take three decomposed effects (∆Cei , ∆Cb´ib , ∆Cci f )
as dependent variables respectively to analyze the path which conducted the effect of change of
population structure on change of carbon emission in the content of this paper below. The regression
model was as Equation (18):

∆Cei “ α1 ` β 1 children ` β 2 elderly ` β 3 sex ` β 4 senior ` β 5 college`

β 6 occupation ` β 7 urban ` β 8 GDP ` β 9 population ` ε 1

∆Cb´ib “ α2 ` β 1 children ` β 2 elderly ` β 3 sex ` β 4 senior ` β 5 college`

(18)
β 6 occupation ` β 7 urban ` β 8 GDP ` β 9 population ` ε 2

∆Cci f “ α3 ` β 1 children ` β 2 elderly ` β 3 sex ` β 4 senior ` β 5 college`

β 6 occupation ` β 7 urban ` β 8 GDP ` β 9 population ` ε 3
Sustainability 2016, 8, 225 16 of 20

where αi (i = 1, 2, 3) was a constant, βi (i = 1, . . . ,9) which represented the regression coefficient of

the variable, and εi (i = 1, 2, 3) represented random error. The regression results are shown in Table 3.
The results showed that the Hausman test statistical quantity in model II was 20.3562, and it was of
5% significance. Thus, the fixed effect model was more suitable to model II, while we should select
a random effect model in model III and model IV.
Firstly, in model II, regression coefficients of the regional proportion of the population of senior
high school (senior) and the proportion of the population of college and above (college) indices were
negative, and they were of 10% and 1% significance, respectively, indicating that improvement of the
population quality was helpful to reduce the regional energy intensity effect (∆Cei ). Energy intensity
reflected the development of technology of energy utilization. The regression coefficient of the regional
proportion of employees in state-owned units in the total Chinese urban employment (occupation)
was positive, and it showed 1% significance, which indicated that the proportion of employees in
state-owned enterprises hindered the advance of energy utilized technology. This result corresponded
to the inefficiency of state-owned economic components. Regression coefficients of regional population
urban and rural structure (urban) and total GDP (GDP) indices were of 10% and 1% significance,
respectively, and the regression coefficient of regional total GDP (GDP) was negative, which indicated
that improvement of the economy in China was helpful to promote the progress of energy utilized
technology during the sample period.
Secondly, identical to model I, when ∆Cb´ib is taken as the dependent variable, the effect of the
regional elderly people dependency ratio (elderly), proportion of the population of college and above
(college), proportion of employees in state-owned units in total China urban employment (occupation),
population urban and rural structure (urban), total GDP (GDP) and total population (population) indices
were significant, and the sign of their regression coefficients were the same as in model I. This indicated
that the population education structure, population occupation structure, and population urban
and rural structure could further affect regional carbon emission from energy consumption through
affecting residents’ consumption scale and consumption levels.
Thirdly, in model IV, the regression coefficient of children dependency ratio (children) index
was negative, and it showed 1% significance. Due to the low consumption capacity of the juvenile
population, the increasing of the children dependency ratio (children) would inevitably cause the
residents’ consumption ratio to decrease, thereby reducing carbon emission from energy consumption.
Contrary to the former three models, the regression coefficient of regional population urban and
rural structure (urban) was negative, and it showed 10% significance, indicated that development of
urbanization would reduce the consumption inhibitory factor effect (∆Cci f ). Although the development
of urbanization has promoted the rising of the urban population, and increased the amount of residents’
consumption, the residents’ consumption ratio in provinces of China decreased continuously during
the sample period according to the data of the China Statistical Yearbook [43]. Therefore, the contribution
of urbanization development on the overall economy was stronger than on residents’ consumption.
This may explain why the regression coefficient of regional population urban and rural structure
(urban) was negative. In addition, identical to model I and model III, the regression coefficients of
occupation, GDP, and population all showed 1% significance.
Finally, through comparing the results of the above models, we found that population structure
change affected change of regional carbon emission from energy consumption mainly through affecting
residents’ consumption ratio and consumption level. Population education structure affected change of
regional carbon emission mainly by two paths as progress of energy utilized technology and change of
residents’ consumptive conception. Paths of impact of population occupation structure and population
urban and rural structure all included progress of energy utilized technology, scale, and level of
residents’ consumption, residents’ consumption ratio. Although the effect of population sex structure
(sex ratio) index on regional carbon emission was always not significant, the sex ratio at birth in China
was always higher than 115 during the sample period; this may cause serious gender imbalance in
the future. It should catch our attention whether this phenomenon would have significant impact on
carbon emission from energy consumption in China in the future.
Sustainability 2016, 8, 225 17 of 20

Due to the large gap of change of carbon emission and its decomposed effects between three
regions, this paper analyzed effect of population structure change on carbon emission and its
decomposed effects in different regions by introducing dummy variables into Equation (17) and
Equation (18). Dummy variables were set up as follows:
# #
1 pEasternq 1 pCentralq
Eastern “ Central “ (19)
0 pOther areasq 0 pOther areasq

∆C, ∆Cei , ∆Cb´ib , ∆Ci f were taken as dependent variables, respectively, introduced dummy
variables into the regression model, and obtained the results as shown in Table 5.

Table 5. Regression results by using the panel data (including dummy variables).

Dependent Variables ∆C (Model V) ∆Cei (Model VI) ∆Cb´ib (Model VII) ∆Ci f (Model VIII)
Model Random effect Random effect Random effect Random effect
α ´0.0533 (´0.4274) 0.0103 (0.1386) ´0.0645 (´0.7419) 0.0240 (0.5932)
children ´0.4754 (´0.5997) 0.0728 (0.1323) 0.3652 (0.6098) ´0.9501 *** (´2.8577)
elderly ´3.9236 *** (´3.0862) ´0.7000 (´0.7908) ´2.7423 *** (´2.8497) ´0.5351 (´0.9970)
sex ´0.5990 (´1.0116) ´0.6774 (´1.6297) 0.1377 (0.3060) ´0.0025 (´0.0100)
senior 0.0743 (0.0662) ´1.2794 (´1.6241) 0.4196 (0.4919) 1.0231 ** (2.1251)
college ´4.2848 *** (´4.5513) ´2.5008 *** (´3.8324) ´1.5358 ** (´2.1598) 0.0424 (0.1075)
occupation ´1.7906 *** (´4.0323) 1.1457 *** (3.7308) ´1.4996 *** (´4.4761) ´1.2306 *** (´6.6494)
urban 2.2128 *** (2.9322) 0.7640 (1.4727) 1.4406 ** (2.5377) ´0.7102 ** (´2.2847)
GDP 1.1345 *** (14.8856) ´0.2842 *** (´5.3693) 1.0489 *** (18.2049) 0.3058 *** (9.5512)
population 4.0629 *** (3.1490) 0.5210 (0.5798) 4.1958 *** (4.2958) 4.2943 *** (7.8786)
Eastern 0.2061 * (1.7130) 0.1829 * (1.8795) 0.0250 (0.2182) ´0.0352 (´0.6699)
Central 0.0406 (0.1999) ´0.0132 (´0.1109) 0.0138 (0.0987) 0.0413 (0.6451)
F 136.1431 *** 40.7581 *** 199.8641 *** 63.3061 ***
Adj-R2 0.8468 0.6192 0.8905 0.7181
Hausman Test 5.8078 (0.7590) 14.7004 * (0.0995) 3.2680 (0.9527) 3.4778 (0.9423)
According to the results of the Hausman test, the random effect model was more suitable to this study. Due to
space limitations, we only provide the regression results of the random effect model in this paper; the author
could provide regression results of the fixed effect model if necessary.

Table 5 showed that the regression coefficient of Eastern in model V was 0.2061, and it was
significant. This indicated that change of carbon emission from energy consumption in the eastern
region was significantly higher than that in the central and western regions. The regression coefficient
of Eastern in model VI was 0.1829, and it was also significant, while the effect of dummy variables was
not significant in the other models. This result further showed that the foundation of energy utilization
technology in the east was better, and met the above conclusion.

6. Conclusions
On the basis of Kaya identity, this paper expanded the LMDI model through introducing
population urbanization, residents’ consumption and other factors, and decomposed the change
of carbon emission in China into six effects: carbon emission factor effect, energy intensity effect,
consumption inhibitory factor effect, urbanization effect, residents’ consumption effect, and population
scale effect. It then explored contribution ratios and action mechanism of the above six effects on
change of carbon emission. Then, we analyzed the effect of population structure change on carbon
emission by taking 2003-2012 as a sample period, and combined this with the panel data of 30 provinces
in China. The main conclusions of this paper are as follows:

(1) In 2003–2012, total carbon emission in China increased by 4.2117 billion tons. Consumption
inhibitory factor effect, urbanization effect, residents’ consumption effect, and population scale
effect could promote the increasing of carbon emission, while the influence of carbon emission
factor effect and energy intensity effect on carbon emission were negative, indicating that Chinese
energy technology utilized during the samples was improved and reached the effect of “technical
carbon emission reduction” to some extent. Specifically, the contribution ratio of residents’
Sustainability 2016, 8, 225 18 of 20

consumption effect affected change of carbon emission was greatest, which reached 74.96%,
followed by the consumption inhibitory effect and energy intensity effect. Their contribution
ratios were 27.44% and ´18.46%, respectively. The contribution ratio of impact of urbanization
effect, population scale effect and carbon emission factor effect on carbon emission was 12.700%,
5.90%, and ´2.54% respectively.
(2) Firstly, there was a significant difference between the carbon emission factor effect, energy
intensity effect, and population scale effect in the three areas. Contribution ratio of energy
intensity effect in the eastern region was significantly smaller than the central and western areas,
while contribution ratio of its population scale effect was larger than the other two regions.
Carbon emission factor effect of the central area was significantly higher than that in the eastern
and western regions, and its population scale effect was close to 0. Secondly, the influences of
carbon emission factor and energy intensity on the changing of regional carbon emission were
all negative. The carbon emission factor effect and energy intensity effect in the three regions
were around 0 from 2003 to 2006, while in 2007–2012 they all showed obvious decline. Thirdly,
influences of the consumption inhibitory factor effect on changing of regional carbon emission
in three areas showed a fluctuating upward trend, their effect were positive and did not show
their inhibition. The residents’ consumption effect in the three regions showed a clear upward
trend, and the residents’ consumption effect in the eastern, central, and western regions were
0.1419, 0.1413, and 0.0936 billion tons in 2012, respectively. The residents’ consumption effect
became the most important influencing factor which affected change of carbon emission and its
contribution ratio was the greatest. The main source of the residents’ consumption effect in the
three regions was the residents’ basic living consumption effect, while residents’ non-basic living
consumption effect in the three areas changed as a graphic “N” around 0; it showed a fluctuating
upward trend in 2003–2005 and in 2009–2012, and appeared to fluctuate downwards in 2006–2008.
Finally, the “eastern aggregation” phenomenon of population caused the population scale effect
of the eastern area to be obviously larger than the central and western areas. The urbanization
effect in all three regions were obviously larger than the population scale effect, which indicated
that demographic urbanization has become the main population factor which affected the change
of carbon emission.
(3) Secondly, the impact of regional people dependency ratio on regional carbon emission was
negative and significant, as was the proportion of the population of college and above, and the
proportion of employees in state-owned units in total Chinese urban employment, while the effect
of regional population urban and rural structure, total GDP and total population were positive
and significant. Based on the analysis of the decomposition effects, we found that population
structure change affected change in regional carbon emission from energy consumption mainly
through affecting residents’ consumption ratio and consumption levels. Population education
structure affected change of regional carbon emission mainly by two paths, as progress of energy
utilized technology and change of residents’ consumptive conception. While the path of the
impact of population occupation structure and population urban and rural structure all included
progress of energy utilized technology, scale, and the level of residents’ consumption and residents’
consumption ratio, the effect of population sex structure (sex ratio) index on regional carbon
emission was always non-significant. Finally, we found that the change of carbon emission in
the eastern region was significantly higher than that in the central and western regions when
dummy variables were introduced into the regression model. The foundation of energy utilized
technology in the eastern region was also significantly better than in the other two regions.

Acknowledgments: This work was financially supported by Funding of Jiangsu Social Sciences (No. 12EYA001),
Funding of Jiangsu University Philosophy and Social Science (No. 11ZDIXM051), Funding of Anhui Humanities
and Social Sciences (No. SK2014A099), Funding of Jiangsu Innovation Program for Graduate Education (No.
KYZZ_0107 & No. CXLX12_0178), China Scholarship Council: “Graduate Programs of National Construction
High Level University in 2014”.
Author Contributions: Wen Guo and Tao Sun wrote the paper, Wen Guo and Hongjun Dai analyzed the data.
Sustainability 2016, 8, 225 19 of 20

Conflicts of Interest: The authors declare no conflict of interest.

References
1. National Bureau of Statistics of China. China Statistical Yearbook; China Statistics Press: Beijing, China,
2001–2014.
2. Zhuang, Y.E.; Zhang, L.P. China Commonly Used Population Data Set Since 1990; China Population Press:
Beijing, China, 2003.
3. National Bureau of Statistics of China. China Labour Statistical Yearbook; China Statistics Press: Beijing, China,
2001–2014.
4. Wang, Q. Effects of urbanization on energy consumption in China. Energy Policy 2014, 65, 332–339. [CrossRef]
5. Liddle, B. Consumption-driven environmental impact and age structure change in OECD countries:
A cointegration-STIRPAT analysis. Demogr. Res. 2011, 30, 749–770. [CrossRef]
6. Okada, A. Is an increased elderly population related to decreased CO2 emissions from road transportation?
Energy Policy 2012, 45, 286–292. [CrossRef]
7. Menz, T.; Welsch, H. Population aging and carbon emissions in OECD countries: Accounting for life-cycle
and cohort effects. Energy Econ. 2012, 34, 842–849. [CrossRef]
8. Zhu, H.M.; You, W.H.; Zeng, Z. Urbanization and CO2 emissions: A semi-parametric panel data analysis.
Econ. Lett. 2012, 3, 848–850. [CrossRef]
9. Knight, K.W.; Rosa, E.A.; Schor, J.B. Could working less reduce pressures on the environment?
A cross-national panel analysis of OECD countries, 1970–2007. Glob. Environ. Chang. 2013, 23, 691–700.
[CrossRef]
10. Katircioğlu, S.T. Estimating higher education induced energy consumption: The case of Northern Cyprus.
Energy 2014, 66, 831–838. [CrossRef]
11. Cellura, M.; Longo, S.; Mistretta, M. Application of the structural decomposition analysis to assess the indirect
energy consumption and air emission changes related to Italian households consumption. Renew. Sustain.
Energy Rev. 2012, 16, 1135–1145. [CrossRef]
12. Zhao, X.; Li, N.; Ma, C. Residential energy consumption in urban China: A decomposition analysis.
Energy Policy 2012, 41, 644–653. [CrossRef]
13. Hoekstra, R.; Bergh, J.C.J.M. Comparing Structural and Index Decomposition Analysis. Energy Econ. 2003, 25,
39–64. [CrossRef]
14. Ang, B.W. Decomposition Analysis for Policymaking in Energy: Which is the Preferred Method. Energy Policy
2004, 32, 1131–1139. [CrossRef]
15. Ang, B.W.; Zhang, F.Q. A Survey of Index Decomposition Analysis in Energy and Environmental Studies.
Energy 2009, 25, 1149–1176. [CrossRef]
16. Ang, B.W.; Huang, H.C.; Mu, A.R. Properties and linkages of some index decomposition analysis methods.
Energy Policy 2009, 11, 4624–4632. [CrossRef]
17. 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. [CrossRef]
18. Zhang, N.; Choi, Y. Total-factor carbon emission performance of fossil fuel power plants in China:
A metafrontier non-radial Malmquist index analysis. Energy Econ. 2013, 40, 549–559. [CrossRef]
19. Su, B.; Ang, B.W. Structural decomposition analysis applied to energy and emissions: Some methodological
developments. Energy Econ. 2012, 34, 177–188. [CrossRef]
20. Zhang, N.; Zhou, P.; Choi, Y. Energy efficiency, CO2 emission performance and technology gaps in fossil
fuel electricity generation in Korea: A meta-frontier non-radial directional distance function analysis.
Energy Policy 2013, 56, 653–662. [CrossRef]
21. Rosa, E.A.; York, R.; Dietz, T. Tracking the anthropogenic drivers of ecological impacts. AMBIO A J.
Hum. Environ. 2004, 8, 509–512. [CrossRef]
22. O’Neill, B.C.; Liddle, B.; Jiang, L.; Smith, K.R.; Pachauri, S.; Dalton, M.; Fuchs, L. Demographic change and
carbon dioxide emissions. Lancet 2012, 380, 157–164. [CrossRef]
23. Wang, P.; Wu, W.; Zhu, B.; Wei, Y. Examining the impact factors of energy-related CO2 emissions using the
STIRPAT model in Guangdong Province, China. Appl. Energy 2013, 106, 65–71.
24. Satterthwaite, D. The Implications of Population Growth and Urbanization for Climate Change.
Environ. Urban. 2009, 2, 545–567. [CrossRef]
Sustainability 2016, 8, 225 20 of 20

25. Yao, C.; Chen, C.; Li, M. Analysis of rural residential energy consumption and corresponding carbon
emissions in China. Energy Policy 2012, 41, 445–450. [CrossRef]
26. Wang, L.; Chen, Z.; Ma, D.; Zhao, P. Measuring carbon emissions performance in 123 countries: Application
of minimum distance to the strong efficiency frontier analysis. Sustainability 2013, 5, 5319–5332. [CrossRef]
27. York, R. Demographic trends and energy consumption in European Union Nations, 1960–2025. Soc. Sci. Res.
2007, 3, 855–872. [CrossRef]
28. Liddle, B.; Lung, S. Age-structure, urbanization, and climate change in developed countries: Revisiting
STIRPAT for disaggregated population and consumption-related environmental impacts. Popul. Environ.
2010, 5, 317–343. [CrossRef]
29. Lugauer, S.; Jensen, R.; Sadler, C. An estimate of the age distribution’s effect on carbon dioxide emissions.
Econ. Inq. 2014, 2, 914–929. [CrossRef]
30. Zagheni, E. The leverage of demographic dynamics on carbon dioxide emissions: Does age structure matter?
Demography 2011, 1, 371–399. [CrossRef] [PubMed]
31. Laureti, T.; Montero, J.M.; Fernández-Avilés, G. A local scale analysis on influencing factors of NOx emissions:
Evidence from the Community of Madrid, Spain. Energy Policy 2014, 74, 557–568. [CrossRef]
32. Jorgenson, A.K. Does foreign investment harm the air we breathe and the water we drink? A cross-national
study of carbon dioxide emissions and organic water pollution in less-developed countries, 1975 to 2000.
Organ. Environ. 2007, 2, 137–156. [CrossRef]
33. Jorgenson, A.K. The sociology of ecologically unequal exchange and carbon dioxide emissions, 1960–2005.
Soc. Sci. Res. 2012, 2, 242–252. [CrossRef] [PubMed]
34. Jorgenson, A.K.; Clark, B. Assessing the temporal stability of the population/environment relationship
in comparative perspective: A cross-national panel study of carbon dioxide emissions, 1960–2005.
Popul. Environ. 2010, 1, 27–41. [CrossRef]
35. Jorgenson, A.K.; Clark, B. Are the Economy and the Environment Decoupling? A Comparative International
Study, 1960–2005. Am. J. Soc. 2012, 1, 1–44. [CrossRef]
36. Poumanyvong, P.; Kaneko, S. Does urbanization lead to less energy use and lower CO2 emissions?
A cross-country analysis. Ecol. Econ. 2010, 2, 434–444. [CrossRef]
37. Martinez-Zarzoso, I.; Maruotti, A. The impact of urbanization on CO2 emissions: Evidence from developing
countries. Ecol. Econ. 2011, 70, 1344–1353. [CrossRef]
38. Poumanyvong, P.; Kaneko, S.; Dhakal, S. Impacts of urbanization on national transport and road energy use:
Evidence from low, middle and high income countries. Energy Policy 2012, 46, 268–277. [CrossRef]
39. Fang, W.; Miller, S.; Yeh, C.C. The effect of ESCOs on energy use. Energy Policy 2012, 51, 558–568. [CrossRef]
40. Mishra, V.; Smyth, R.; Sharma, S. The energy-GDP nexus: Evidence from a panel of Pacific Island countries.
Resour. Energy Econ. 2009, 3, 210–220. [CrossRef]
41. Hossain, S. Panel estimation for CO2 emissions, energy consumption, economic growth, trade openness and
urbanization of newly industrialized countries. Energy Policy 2011, 11, 6991–6999. [CrossRef]
42. Al-mulali, U.; Fereidouni, H.G.; Lee, J.Y.M.; Sab, C.N.B.C. Exploring the relationship between urbanization,
energy consumption, and CO2 emission in MENA countries. Renew. Sustain. Energy Rev. 2013, 23, 107–112.
[CrossRef]
43. Liddle, B. Impact of population, age structure, and urbanization on carbon emissions/energy consumption:
evidence from macro-level, cross-country analyses. Popul. Environ. 2014, 3, 286–304. [CrossRef]
44. Song, M.L.; Zhou, Y.X. Analysis of Carbon Emissions and Their Influence Factors Based on Data from Anhui
of China. Comput. Econ. 2014, 1, 1–16. [CrossRef]
45. National Bureau of Statistics of China. China Energy Statistical Yearbook; China Statistics Press: Beijing, China,
2004–2013.
46. Zhang, J.F.; Deng, W. Industrial structure change and its eco-environmental influence since the establishment
of municipality in Chongqing, China. Procedia Environ. Sci. 2010, 2, 517–526. [CrossRef]
47. Cai, F.; Lu, Y. Population change and resulting slowdown in potential GDP growth in China.
China World Econ. 2013, 2, 1–14. [CrossRef]

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons by Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).