Do CO2 emissions impact energy use? An assessment of China evidence from 1953 to 2017

Accepted Manuscript Do CO2 emissions impact energy use? An assessment of China evidence from 1953 to 2017

Wen-Wen Zhang, Shi-Chun Xu, Basil Sharp PII: DOI: Article Number: Reference:

S1043-951X(19)30101-4 https://doi.org/10.1016/j.chieco.2019.101340 101340 CHIECO 101340

To appear in:

China Economic Review

Received date: Revised date: Accepted date:

5 December 2018 22 July 2019 19 August 2019

Please cite this article as: W.-W. Zhang, S.-C. Xu and B. Sharp, Do CO2 emissions impact energy use? An assessment of China evidence from 1953 to 2017, China Economic Review, https://doi.org/10.1016/j.chieco.2019.101340

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Do CO2 emissions impact energy use? An assessment of China evidence from 1953 to 2017

Wen-Wen Zhanga [email protected], Shi-Chun Xub,* [email protected],Basil

a

Energy Center, University of Auckland, Auckland 1010, New Zealand

b

PT

Sharpa [email protected]

Management School, China University of Mining and Technology, Xuzhou 221116, China

*

AC

CE

PT E

D

MA

NU

SC

RI

Corresponding author.

ACCEPTED MANUSCRIPT Abstract Bidirectional causality between energy use and carbon emissions has widely been explored, but the results in the existing research are still controversial. Focusing on China, we reveal the bidirectional causality between energy use and carbon emissions, and then take carbon emissions as a factor influencing energy use into decomposition

PT

analysis by the Logarithmic Mean Index (LMDI) approach. The study span 1953-2017 is divided into two main developing stages based on multiple breakpoint

RI

tests, stage one: 1953-2001, before entering World Trade Organization (WTO); and stage two: 2001-2017, after entering WTO. The results show that emission effect and

SC

per capita emission effect indeed promoted energy use and even surpassed output effect at the beginning of stage two (2001-2005). The changes in energy use followed

NU

those in output effect at stage one, emission effect and per capita emission effect at the beginning of stage two, and energy intensity effect at the latter period of stage two.

MA

It indicates that energy efficiency has gradually been becoming a dominate role in energy use changes. The impact of energy mix effect was comparatively small.

AC

CE

PT E

D

Keywords: Carbon emissions; energy use; LMDI model; multiple breakpoint tests

ACCEPTED MANUSCRIPT 1. Introduction The design of policy aimed at addressing energy issues related to the environment and economy is a major challenge. A rapid increase in energy demand and closer interdependence of countries enhance its essential and vital role in economic growth. Climate change associated with the use of fossil fuels threatens sustainable

PT

development. Policy design aimed at reducing greenhouse emissions will benefit from identifying factors that influence energy use. China, as one of the rapid growing and

RI

urbanizing economics, suffers from serious extreme weather events and air pollution. Primary energy consumption in China was about 3.1 billion tons of oil equivalent in

SC

2017, accounting for 23% of global energy use (National Bureau of Energy Statistics of China, 2018), and is increasing (Xiong et al., 2014).

NU

Numerous studies in recent years have investigated factors that affect energy use in China. Sun (1998) decomposed factors that influence changes of energy use into

MA

the scale of economic activities (activity effect), sectorial technological level (intensity effect), and economic structure (structure effect) during the period of 1973

D

to 1990. The activity effect and structure effect were associated with increased energy

PT E

use throughout the study period, while the intensity effect decreased energy use from 1980 to 1990. Increased energy use is associated with economic growth, foreign trade, and changes in industrial structure1 (Li et al., 2014; Liu et al., 2018).

CE

However, the results of statistical tests on the causal relationship between energy consumption and gross domestic product (GDP) differ across countries (Soytas and

AC

Sari, 2003). China and India show a long-run bi-directional causality between coal use and economic growth during 1980-2011 (Bildirici and Bakirtas, 2014). An interesting result from Li et al. (2011) shows that economic growth is one the most important factors affected China’s energy use, exhibiting a negative impact from 1990 to 1994 and a positive impact between 1994 and 2008. Possible explanations for this result include different specifications of decomposition variables such as GDP and energy. Changes in energy structure can have a positive impact on energy use in both developing and developed countries (Neto et al., 2014). Fan and Xia (2012) showed 1

Industrial structure refers to the share of each sector in the total output.

ACCEPTED MANUSCRIPT that energy input mix had a major active influence on energy demand in China from 1987 to 2007. Zhao et al. (2012) also revealed that a structural change to clean energy, driven by a change in relative prices, resulted in a negative impact on residential energy consumption in China during the period of 1998-2007. Improving energy consumption structure is considered to be one of the crucial approaches for realizing green transformation of economic development (Anshasy and Katsaiti, 2014; Yang et

PT

al., 2018). Focusing on China’s 30 provinces from 2007 to 2012, however, Liu et al.

RI

(2018) found that energy mix2 had little effect on changes in energy use. Energy intensity, as measured by energy use per unit GDP, is a factor associated

SC

with energy consumption (e.g., Román-Collado and Colinet, 2018; Xu et al., 2012). A decrease in China’s energy use between 1981 and 1987 was mainly the result of a

NU

decline in energy intensity driven by improvements in energy efficiency(the degree of energy that effectively used) (Lin and Polenske, 1995). Garbaccio et al. (1990)

MA

attributed the fall in China’s energy use in the following period 1987-1992 to a decline in energy intensity. Zhang (2003) also found that 88% of the total energy

D

saving in China’s industrial sectors from 1990 to 1997 was a result of the change in

PT E

energy intensity. From both of the production and final-demand side, Zhang and Lahr (2014) explored driving forces of energy consumption in 7 seven regions of China (including Northeast, North China, East China, South China, Central China,

CE

Northwest, and Southwest) from 1987 to 2007 and found that reducing energy intensity is key to achieve regional energy conservation. Distinguishing domestic and

AC

interregional trade, Zhao et al. (2017) unveiled the drivers of changes in provincial energy consumption from 1997 to 2011. They found per capita final energy demand, energy intensity, and production structure, had a positive effect on provincial energy consumption. The relationship between carbon emissions and energy has been the focus of recent studies. Using a sample drawn from 58 countries, for the period 1990-2012, Saidi and Hammami (2015) found that carbon emissions had a significant, positive 2

In this study, energy mix refers to the share of each primary energy type in the total energy use as defined in section 3.

ACCEPTED MANUSCRIPT effect, on energy use. Using a panel provincial data for China, Li et al. (2018) found that hotter summers exhibit more influence than colder winters, indicating that global warming increases electricity consumption all year round. This result supports an earlier claim by Mansur et al. (2008) that climate change promotes more energy use and increases more electricity consumption. Climate change caused by energy-related

PT

carbon emissions influences use patterns and purchasing decisions of firms and households for heating and cooling appliances, and further impacts energy use.

RI

The dashed-line in Figure 1 illustrates the research path applied in these studies and raises the question as to whether carbon emissions is factor influencing energy

SC

use. Causality between energy use and carbon emissions based on the Grander-causality test has widely been explored, but the results are not conclusive.

NU

Alam et al. (2012) investigated dynamic causality between energy consumption, carbon emissions and economic growth and found that feedback causality exists from

MA

carbon emissions to energy use in Bangladesh in the long term. That is to say, in the long-run CO2 emissions result in an increase in energy use. Antonakakis et al. (2017)

D

found that carbon emissions Granger-cause energy consumption only in high-income

PT E

countries from 1971 to 2011. Apergis and Payne (2010) found evidence of bidirectional causality between energy use and carbon emissions for 11 countries of the Commonwealth of Independent States during 1992-2004. Chang (2010) also

CE

found evidence in support of bidirectional causality between carbon emissions and

AC

energy use in Hong Kong from 1981 to 2006.

Most of the above studies use factor decomposition methods, Index Decomposition Analysis (IDA) and Structure Decomposition Analysis (SDA). Refer to Wang et al. (2017), Hoekstra and Van den Bergh (2003), and Hatzigeorgiou et al. (2008) for more detail on these methods. IDA is more commonly used. Two IDA methods include the Logarithmic Mean Index (LMDI) method and the Arithmetic Mean Divisia Index (AMDI) method (e.g., Wang and Feng, 2017; Liu et al., 2016; Moutinho et al., 2015). Causality tests between energy use and carbon emissions for mainland China

ACCEPTED MANUSCRIPT have yet to be undertaken. Although Saidi and Hammami (2015) incorporated carbon emissions as an explanatory variable into a statistical analysis of changes in energy use and found that carbon emissions had a significant and positive effect on energy use, lack related explanation and further discussion for this result. This study endeavors to fill the gap and make the following contributions. First, we examine

PT

Granger-causality between carbon emissions and energy use in China spanning the study period 1953-2017. Second, we estimate the impact of carbon emissions on

RI

China’s energy use. Based on the LMDI decomposition model, we disaggregate factors that influence China’s energy use during 1978-2015 into an emission effect

SC

and per capita emission effect as well as three conventional factors viz. energy mix, energy intensity, and output. We find that bi-directional causality exists from carbon

NU

emissions to energy use in the long-run. Although carbon emission mainly plays a positive effect on energy use, the negative effect shows up during 2014-2016. In

MA

addition, we measure and compare the contribution of each factor to energy use. Third, the cumulative effect of each factor during 1953-2017 is analyzed first (holistic

D

analysis); and then we divide the study period 1953-2017 into two stages in 2001

PT E

(stage analysis).

The rest of this study is organized as follows. Section 2 tests the statistical relationship between energy use and carbon emissions. Section 3 describes the

CE

methodology of LMDI and Section 4 presents the empirical results and further discussion. Conclusions and policy implications are presented in Section 5.

AC

2. Econometrics analysis of energy use and carbon emissions We first examine causality between energy use and carbon emissions in China for the time series spanning the period 1953-2017 using the following: unit root tests, cointegration test, Granger-causality test, and Variance Decomposition (VD) analysis. Table 1 describes the data.

(1) Stationarity Table 2 displays results of the Augmented Dickey-Fuller (ADF), Dickey-Fuller GLS

(DF-GLS),

Kwiatkowski-Phillips-Schmidt-Shin

(KPSS),

and

ERS

ACCEPTED MANUSCRIPT Point-Optimal (ERS) unit root tests for energy use and carbon emissions. Lagged ranks of variables are obtained by the Akaike Information Criterion (AIC) which is normally used to select the preferred model (Wang, 2000). Results indicate that both series are non-stationary at their level but stationary at their first differences. We accordingly use first differenced variables in following cointegration and

PT

Granger-causality tests. (2) Cointegration test

RI

The linear cointegration test of Johansen (1991) is applied to verify whether there is a long-term equilibrium relationship between energy use and carbon emissions.

SC

Table 3 shows that the Trace Statistic is consistent with the Max-Eigen Statistic, exceeding the corresponding bounds of critical values. The null hypothesis of no

NU

cointegration is rejected at 5% significant level, confirming that there is one linear cointegration relationship between energy use and carbon emissions.

MA

(3) Causality

The cointegration test only confirms a long-term equilibrium relationship, so we

D

proceed with implementing Granger-causality test to verify whether there is a

PT E

long-run relationship between energy use and carbon emissions. Table 4 illustrates that the null hypothesis of Granger non-causality between energy use and carbon emissions is rejected at the 10% significance level. That is, there exists bi-directional

CE

causality between energy use and carbon emissions. This result strengthens the case to incorporate carbon emissions as a factor influencing energy use.

AC

3. Empirical method

Having confirmed bi-directional causality between energy use and carbon emissions, we incorporate carbon emissions into decomposition analysis. Ang (2015) summarized and compared eight LMDI models based on three different dimensions, including method (LMDI-I or LMDI-II), decomposition procedure (additive or multiplication), and aggregate indicator (quantity or intensity indicator). Model 1 (LMDI-I, additive, and quantity indicator) is recognized as a perfect model with a quantity indicator, therefore, we apply it into our study. The additive decomposition analysis in the model presents aggregate change in energy use and decomposition

ACCEPTED MANUSCRIPT results in a physical unit, which is different from the multiplicative decomposition analysis used in Ang and Liu (2001). Energy consumption is expressed as Eq.(1), where j stands for 4 types of energy including coal, crude oil, natural gas, and primary electricity and other energy; E is total energy consumption; Ej denotes the energy consumption of energy type j;

denote total output, population and total energy-related carbon

PT

and

,

emissions, respectively. ∑

RI

(1)

SC

According to Model 1, decomposed factors for changes in energy consumption

(

) [

( )

( )

]

(3)

(

)

]

(

)

]

(5)

)

(

)

]

(6)

(

)

(

)

]

) [ (

)

(

)

]

(

) [

(

)

∑

(

) [ (

)

∑

(

) [

∑

(

∑

(

MA

∑

PT E

(

(2)

(4)

(6*) (7)

CE

) [

(

) is defined as Eq.(8).

AC

Where

(

D

∑

) are computed as follows.

NU

between the base period and target period (

)

(8)

{ in Eq.(6) is interpreted as the negative effect of

as

captured in Eq.(6*). Eq.(2) is transformed to Eq.(2*). Factors influencing energy consumption are decomposed into energy mix, energy intensity, output, per capita emissions, and emission effect.

ACCEPTED MANUSCRIPT ( (1) Energy mix effect (

)

(2*)

): reflects changes in energy use influenced by

energy substitution. Limited by resource endowment (rich in coal, lack of petroleum and natural gas), coal has played a dominate role in energy use and accounts on average for more than 76% of the total energy consumption from 1953 to 2017.

PT

Although there has been an increase in the ratio of renewable energy in total energy

RI

consumption in recent years, it has not substantially changed China’s energy mix. (2) Energy intensity effect (

): reflects changes in energy use associated

SC

with changes in energy consumption per unit output. It also reflects energy efficiency which measures the degree of energy effectively used (Masjuki et al., 2001; Cantore

NU

et al., 2016). Increased energy intensity implies one unit of output needs more energy (known as lower energy efficiency). A decrease in energy intensity (higher energy

(3) Output effect (

MA

efficiency) reduces energy consumption.

): reflects the changes in energy use associated with

D

per capita output. Some studies (e.g., Li et al., 2011) have found that the contribution

PT E

level of output effect depends on the development level. Given the great changes in China’s economy and society since 1953, we explore the impact of output on energy

CE

use at different stages.

(4) Per capita emission effect (

): reflects the changes in energy use

AC

associated with changes in per capita carbon emissions. Compared to the emission effect, the per capita emission effect takes population into account. Although China’s per capital emissions are relatively lower than other countries, they have increased because the growth rate in carbon emissions exceeds that in population. (5) Emission effect (

): reflects the changes in energy use influenced by

total carbon emissions. Energy consumption has been identified as the most important source of carbon emissions in the current studies. However, the reverse impact of carbon emissions on energy use remains unexplored. Data for the consumption of coal, crude oil, natural gas, and primary electricity

ACCEPTED MANUSCRIPT and other energy, population and GDP between 1953 and 2017 come from the China Statistical Yearbook, and GDP is converted into 1953 constant price (1953=100), and data description can be found in Table 1. 4. Results and discussion We break down the changes in energy consumption

(orange curve), energy intensity effect

PT

effects: energy mix effect

(black curve) into five

(purple curve), per capita emission effect

(blue curve), and emission effect

RI

(gray curve), output effect

(red curve). Figure 2 shows the

SC

contribution of each factor over the period 1953-2017. 4.1 Holistic analysis

NU

Total energy consumption increased by approximately 4435 million tce (tons of standard coal equivalent) between 1953 and 2017. Positive impacts from output

MA

effect (142%), emission effect (90%), per capita emission effect (75%) and energy mix effect (<1%) totally offset the negative impact associated with the energy

D

intensity effect (-57%).

PT E

The output effect was positive throughout 1953-2017 (except 1960-196, 1961-1962, 1966-1967, 1967-1968, and 1975-1976) and its cumulative effect was 6318 million tce, confirming that the output effect was the dominate factor

(2018).

CE

stimulating energy use. This finding is consistent with Li et al. (2014) and Liu et al.

AC

It is interesting to find that changes in energy use followed the path of emission effect and per capita emission effect up to the period 2007-2008. Emission and per capita effects were mostly positive and were the second largest contributors pushing energy use to reach 3978 and 3322 million tce from 1953 to 2017, respectively. Per capita emission effect showed a similar trend with emission effect, but there is a growing gap between them. This can be explained by a decreasing trend in population increase and an increasing trend for carbon emission changes. Changes in energy mix also resulted in a slight increase in overall energy consumption from 1953 to 2017. Its cumulative effect on increasing energy use was

ACCEPTED MANUSCRIPT minimal, less than 1 million tce. Energy intensity showing a decreasing trend indicates that it benefits to reduce energy use by improving energy efficiency, which is consistent with Román-Collado and Colinet (2018) and Xu et al. (2012). Between 1953 and 2017, it contributed to a decrease in energy use 57% (2539 million tce). Although energy intensity as a

it seldom suppressed the positive impact of output effect.

RI

4.2 Stage analysis

PT

negative effect played a dominate role in reducing energy use during the study period,

We use multiple-breakpoint tests to examine whether structural breaks exist in

SC

energy use during the period 1953-2017 (Bai and Perron, 2003). The null hypothesis of zero breaks is rejected as both the UDMax statistic (43.23628) and WDMax

NU

statistic (43.23628) are larger than their corresponding critical values (8.88 and 9.91), respectively. One structural break 2001-2002 exists, significant at the 95% confidence

MA

level. It is definitely consistent with the fact that joining WTO was a watershed event for China’s economic development. The whole study period 1953-2017, therefore, is

D

divided into two stages: 1953-2001 (stage one: before entering WTO) and 2001-2017

PT E

(stage two: after entering WTO) which is illustrated by different background colors in Figure 2.

4.2.1 Stage one 1953-2001

CE

Changes in energy consumption and all decomposition factors are more stable in stage one compared to stage two. An increase in energy use from 1953 to 2001 by

AC

1501 million tce is a result of the positive output effect (136%), emission effect (94%), per capita emission effect (69%). The combined positive effects dominate the energy mix effect (<1%) and the energy intensity effect (-61%) over the period 1953-2001. However, within this period there is a decrease in energy use during 1960-1963, 1966-1967, and 1980-1981, which is explained as follows. Natural disasters happened during 1960-1963, which resulted in a negative effect for all decomposition factors. This phenomenon also happened over the period 1966-1967 as a result of Cultural Revolution. A decrease in energy use (1980-1981) was caused by a decline in the positive output effect and an increase in the negative energy intensity effect. In the

ACCEPTED MANUSCRIPT initial process of economic system transformation and comprehensively socialist market economic system establishment, the output effect fluctuated between 1978 and 2001as shown in Figure 2. Two sharp decreases in energy use during 1985-1986 and 1989-1990 are associated with a corresponding decrease in output caused by excessive regulation and inflation.

PT

The magnitude of changes in energy consumption are more obvious between 1991 and 2001 shown as an N-shaped trend. During the period 1991-1995, energy use

RI

increased gradually caused by an increase in the output effect, emission effect and per capita emission effect. However, the negative impact of the energy intensity effect on

SC

energy use remained relatively stable over this period. A series of important reforms initiated in 1992, such as adoption of the market mechanism, commodity circulation

NU

system, financial system, and state-owned enterprises transformed the planned economic system. China’s economy was influenced by the South-East Asian

MA

economic crisis in 1997 and gradually recovered since then. Energy use accordingly increased from 1997-1998 and even surpassed the output effect in 2000-2001. This is

D

because big events happened in 2001, for instance, Beijing won the right to host the

PT E

2008 Olympic Games, Shanghai held the APEC summit, and China joined WTO. These events promoted China’s internationalization which definitely stimulates

effect.

CE

energy consumption. Impacts of these events would last to stage two due to a lag

Output effect had a greater positive impact on energy use than other factors,

AC

which leads to changes in energy consumption closely following those in output effect. Since 1978, output effect tended to increase, which was generally higher than other effects. In 1984, the output was the highest because the founding of new China promoted to complete the 6th Five-Year Plan in advance, which was reflected by large energy use in 1983-1984 (88 million tce). Output peaked again in 1992-1993 contributing to 134 million tce of energy use. Deng Xiaoping’s South Tour Speeches in 1992 stimulated the development of coastal zones. However, a serious inflation around 1993 resulted in a fluctuation in output effect. Economy achieved a soft landing adjusted by stabilization policies since then.

ACCEPTED MANUSCRIPT Emission effect and per capita emission effect were the other two important factors stimulating energy use. In addition to the impacts, we observe that changes in energy consumption perfectly coincided with those in emission effect during the period 1953-1992. The impact of per capita emission effect on energy use was similar to that of emission effect, but the former was weaker especially in subsequent years.

PT

This may be explained by the increased population at this stage. Although family planning was proposed as one of the national policies in China since 1982, it really

RI

cannot control population growth as expected. Emission effect and per capita emission effect reduced energy consumption over the period 1996-1998. Based on the

SC

changes in annual average temperature in China3, two sharp increases and two drastic decreases happened between 1991 and 2001. It means that households and all

NU

industries at this stage tended to buy more heating or cooling applications to conquer colder winter or hotter summer. Government provided more central heating for the

MA

northern areas in cold winter. All these measures were driven by energy consumption. In fact, climate in China has been changing at this stage but utterly unconscious. At

D

this important developing point, coal was the main driving force. Energy mix effect

PT E

were minor positive increasing energy use compared to other positive factors during this stage.

Different from the above factors, energy intensity was the dominate factor in

CE

reducing energy use (-921 million tce) at stage one. It is worth mentioning that the positive effect of energy intensity existed primarily at the beginning of stage one, such

AC

as 1953-1960, 1967-1972, and 1973-1977. Before reform and opening up, energy efficiency was relative lower due to backwardness of technology, which resulted in consuming and wasting energy. However, energy efficiency improved since 1978 (reform and opening up), which is illustrated under the abscissa in Figure 2 reflecting the negative impact of energy intensity effect on energy use over the period 1977-2001. At the beginning of reform and opening up, China was urgent to develop its economy, but improvements in energy efficiency were slow due to the limitation of resources and talent; at the period 1992-2001, the comprehensively establishing 3

CEIC data: https://www.ceicdata.com/en

ACCEPTED MANUSCRIPT socialism market economic system, investment in science and technology promoted energy efficiency over the period 1992-2001. During the stage one, coal definitely became China’s first choice because of its resource endowment, accounting for an average of approximately 80% of total energy use between 1953 and 2001. During this initial stage of economic system transformation, changes in energy

PT

consumption were moderate compared to stage two as well as all decomposition factors. It was limited by the background of this era. The crucial contradiction for

RI

China was to realize economy ‘taken-off’ at any costs by the socialist market economic reforms. In 2001, a market economy system was introduced as the

SC

fundamental role in allocating resources. This economy system has brought huge motivation to China’s economy. Furthermore, the world economy was to recover from

NU

a severe depression of 2001, which provided a promising international environment for China. Therefore, this definitely expanded demand for energy in all industries of

MA

China. 4.2.3 Stage two 2001-2017

D

At the stage of the socialist market economic system improvement, changes in

PT E

energy consumption and all decomposition factors at this stage fluctuated greatly compared to stage one. Energy consumption totally increased by 2935 million tce over the period 2001-2017, attributable to increased output effect (146%), emission

CE

effect (87%), per capita effect (77%), and energy mix effect (<1%), and a decreased energy intensity effect (-55%).

AC

A sharp increase in energy use coincided with joining the WTO. In particular, energy intensity effect was positive over the period 2002-2005, which contributed energy use increase reaching a peak. This may be a result of slower technological progress, increased demand and a lack of standard enterprise management.

It was to

large extent related to the national strategy of realizing the ‘fast and good development’ for the economy. In addition to energy intensity effect, emission effect and per capita effect reached their peak over the period 2002-2005 as well as energy mix effect. It is interesting to find that output effect was positive showing an upward trend but it was exceeded by emission effect, per capita emission effect and energy

ACCEPTED MANUSCRIPT use between 2002 and 2005. The contradiction among economy, energy, and environment obviously surfaced during this period. Output effect reached a peak at 381 million tce in 2006-2007 and then showed a downward trend but not as obvious as energy use and other decomposition factors especially a sharp decrease in 2007-2008. In 2007, national economic strategy

PT

changed to ‘good and fast development’, which emphasized sustainability rather than speed. In fact, insufficient domestic demand and international economic instability

RI

(the global economic crisis) in 2008 contributed to worldwide economic depression. In order to overcome the negative influence of this financial crisis and stimulate

SC

domestic demand, government implemented a large-scale stimulus plan of 4 trillion yuan investment in public infrastructure construction (mainly in the transportation and

NU

power sector) in the same year. This investment resulted in energy use increase reaching another peak in 2010-2011. According to the statistic data, China’s total

MA

production volume of primary energy reached 3 billion tce in 2011, ranking first in the world. Emission effect and per capita emission effect contributed to this increase

D

by reaching their own peak again in 2010-2011. Temperature accordingly increased

PT E

during this time, which changes consumption decisions of households and industries on appliances and energy selection (e.g., Mansur et al., 2008; Li et al., 2018). It will stimulate more energy use in electricity production. Energy intensity had a similar

CE

trend with energy use but in the opposite coordinate over the period 2006-2017. It indicates that energy intensity plays a fundamental role in energy use in recent years.

AC

In addition, the negative effect of energy intensity can be explained by the fact that provinces of China have gradually adopted energy-saving policies (Fujii et al., 2016). In 2006, this program has covered Shandong, Jiangsu, Sichuan, Jilin, Hebei, and Shanxi, etc., which definitely stimulates energy efficiency improvement especially in energy-intensive industries. China’s energy policy issued in 2012 was an important strategy to realize energy sustainability, which requires saving, clean, and safe energy development with high scientific and technological content, low resources consumption, less environmental pollution, high economic benefit and security. In fact, China’s energy and

ACCEPTED MANUSCRIPT environmental problems have been becoming more serious at this time. In 2014, the target decreasing the proportion of non-fossil energy in primary energy consumption to around 15% in 2020 was put forward in the National Plan on Climate Change (2014–2020). Therefore, applying energy-saving technology, improving energy mix, and enhancing energy efficiency have been taken into practice. It is reflected by a

PT

slow increase in energy consumption only at 41 million tce in 2014-2015. Emission effect had an obvious decline over the period 2013-2015 and per capita emissions

RI

even showed a negative impact on energy increase. Approaching 2020, the target year aiming at reducing energy intensity by 15% from 2015 levels, the ‘anti-driving

SC

mechanism’ of the policy would be more obvious. It can also be as a result of seven regional carbon pilot markets subsequently established since 2013. In addition, a slow

NU

growth of population in recent years contributed to the negative impact of per capita emissions. Even more important, ‘The market plays the decisive role in allocating

MA

resource’ was proposed by Third Plenary Session of the Eighteen in 2013, which has been a milestone in China’s economy development.

D

The share of coal consumption was still more than 60%, but that of renewable

PT E

energy was increasing slightly from approximately 8.4% in 2011 to 12.1% in 2015. Therefore, the positive impacts of energy mix effect increasing energy use were weakening in 2014-2015. It definitely contributed to realization of the 2020 target that

CE

non-fossil energy accounting for about 15% of the primary energy consumption. Energy price is a fundamental cause of difficulties in improving energy mix.

AC

Advanced and mature mining technology largely reduces coal production cost, which is opposite to the high cost of renewable energy in terms of storage and transportation. However, renewable energy in China has been developing significantly in recent year. According to Renewable Energy Handbook, renewable energy accounted for the total power generation at 23.2% in 2014. However, changes in energy use and its decomposition factors over the period 2015-2017 showed a recovering trend at different levels. In particular, energy use increased 132 million tce in 2016-2017. This attributes to the enhanced positive impacts of output effects (206%) and emission effect (62%), and the weakened

ACCEPTED MANUSCRIPT negative impact of energy intensity effect (-124%). Coal oversupply directed at meeting extreme demand since 2016. Higher energy efficiency to a large extent results in an energy rebound effect. The constant rise in temperature stimulated more energy consumption. According to the National Climate Center, the average temperature in 2017 was 10.39℃ higher than the history at 0.84℃; there were 12.1

PT

days with average high temperature (≥ 35℃) in 2017,which were four days longer than the history. On the contrary, output effect and energy mix effect are not so

RI

obvious. It reflects the fact that energy intensity effect and emission effect dominate the change in energy use.

SC

Over the period 2001-2017, changes in energy use and all decomposition factors fluctuated and they reached a peak. In fact, industries in China have been

NU

experiencing excessive capacity, which is a result of an unreasonable development pattern. It indicates that a considerable number of energy resources are wasted.

MA

5. Conclusions and policy implications

Econometric suggests that there is evidence of a bidirectional relationship

D

between energy use and carbon emissions over the period 1953-2017. The result that

PT E

energy use Granger causes carbon emissions is the same as existent studies. Causation in the other direction, that carbon emissions Granger causes energy use, suggests that government policy and, or, business management practices focused on emission

CE

reductions, contribute to changes in energy use. The additive decomposition model provided insights into factors that influenced

AC

changes in energy use. Based on multiple breakpoint tests, we divided the study period to reflect two stages of development 1953-2001, before entering WTO; and, 2001-2017, after entering WTO. In general, the output effect played a significant role in increasing China’s energy use during 1953-2017, followed by emission effect, per capita emission effect and energy mix effect. Energy intensity effect reduced energy use. Stage one conclusions: Changes in energy use closely follow the output effect at stage one, emission effect and per capita emission effect at the beginning of stage two, and energy intensity effect at the latter period of stage two. Changes in energy use and

ACCEPTED MANUSCRIPT all decomposition factors fluctuate greatly at stage two relative to stage one. Energy intensity effect even promoted energy use at the beginning of stage two (2002-2005). Stage two conclusions: There is a larger gap between emission effect and per capita emission effect on energy use at stage two. Their positive impacts even surpassed output effect at the beginning of stage two (2001-2005). Negative impact of

PT

emission effect and per capita emission effect on energy use showed up at the end of stage two (2014-2016). The impact of energy mix effect was comparatively minimum

RI

at any stage which can be about to ignore, but it existed an upward trend for increasing energy use.

SC

Several implications for policy follow from the above results. First, the sustained decline in energy mix from 2002 implies that energy mix effect could be an effective

NU

way to decrease energy use in the future. Given China’s rich endowment of coal resources government could set up medium and long-term targets for the development

MA

of alternative sources of energy. For instance, the target ratio of non-fossil energy in primary energy consumption in the National Plan on Climate Change (2014–2020) is

D

15% in 2020. Given regional disparities, regional government could design local

PT E

policies based on its specific industrial characteristics (Fujii et al., 2016 and Chen et al., 2019). In addition to controlling policies, central government could provide local governments with financial support. Taking flexible and effective measures to

CE

promote the development of renewable energy is vital to local governments. Regulating coal mining and use is necessary. For example, from the perspective of

AC

production, local government should resolutely close illegal mines, fine nonstandard production behaviors, and promote the merger small mines. China is also reliant on imported coal. Considering energy security and independence, a shift to non-fossil energy should be encouraged and the quantity of imported coal could be limited. From the perspective of consumption, coal washing should be strictly required in advance. Financial subsidies could be used to accelerate the development of renewable energy. Although China’s renewable energy development is advanced, there exists considerable room for decreasing generation costs for renewables relative to coal.

ACCEPTED MANUSCRIPT More importantly, escaping coal lock-in by overcoming the inertia of traditional energy use also depends on the political will of the government. Therefore, central and local governments should continue to cooperatively push power system reformation which is crucial for the further development of renewable energy. Endeavoring to improve energy efficiency is still one of the most important

PT

approaches for reducing energy consumption. Maintaining improvements in technology is vital to increasing energy efficiency, especially for energy-intensive

Production-Study-Research

cooperation

for

RI

industries. Technology improvements could flow from mobilizing the module of energy-intensive

industries.

SC

Improvements in the input of capital to realize production factor substitution would contribute to reducing energy intensity, which would also promote industrial

NU

upgrading and restructuring. Energy rebound could result from gains in energy efficiency (Lin and Du, 2015). Energy conservation policies would minimize the

MA

rebound effect. In addition to energy efficiency, optimizing energy mix has great potential contribution to strengthen energy intensity effect.

D

Secondary industry has been the main source of economic growth. Transitioning

PT E

to an economy in which the service industry contributes an increasing share of GDP is a long-term target. However, the most important thing is to maintain stably GDP growth under this fierce international situation.

CE

Climate policies aimed at reducing carbon emissions, such as the peak emissions pledge around 2030, are in effect. The national emissions trading system in the power

AC

industry has been in operation since 2017 and is regarded as an effective way promoting emission mitigation. According to the special circumstances of China, the national emissions trading system should gradually expand to cover all sectors and, possibly, connect to other emission markets. It must be noted that, the positive relationship between emission effect and energy use in this study and results from the current literature confirm the interaction between economy, energy, and environment. That is, solving energy (environmental or economic) problems is no longer a single issue, calling for a comprehensive approach to policy.

ACCEPTED MANUSCRIPT Acknowledgments This study was financially supported by the National Natural Science Foundation of China (grant no. 71573253), China Scholarship Council (grant no. 201706420071), Key Projects of Philosophy and Social Sciences for Universities by Jiangsu Provincial Department of Education (grant no. 2018SJZDI109), and Jiangsu Social Science Fund

AC

CE

PT E

D

MA

NU

SC

RI

PT

Project (grant no. 18EYB014).

ACCEPTED MANUSCRIPT References Alam, M. J., Begum, I. A., Buysse, J., & Van Huylenbroeck, G. (2012). Energy consumption, carbon emissions and economic growth nexus in Bangladesh: Cointegration and dynamic causality analysis. Energy Policy, 45, 217-225. Ang, B. W., & Liu, F. L. (2001). A new energy decomposition method: perfect in

Ang,

B.W.

(2015).

LMDI

decomposition

PT

decomposition and consistent in aggregation. Energy, 26(6), 537-548. approach:

guide

for

RI

implementation. Energy Policy, 86, 233-238.

a

Antonakakis, N., Chatziantoniou, I., & Filis, G. (2017). Energy consumption, CO2

Sustainable Energy Reviews, 68, 808-824.

SC

emissions, and economic growth: An ethical dilemma. Renewable and

NU

Apergis, N., & Payne, J. E. (2010). The emissions, energy consumption, and growth nexus: evidence from the commonwealth of independent states. Energy Policy,

MA

38(1), 650-655.

Bai, J., & Perron, P. (2003). Computation and analysis of multiple structural change

D

models. Journal of applied econometrics, 18(1), 1-22.

PT E

Bildirici, M. E., & Bakirtas, T. (2014). The relationship among oil, natural gas and coal consumption and economic growth in BRICTS (Brazil, Russian, India, China, Turkey and South Africa) countries. Energy, 65, 134-144.

CE

Cantore, N., Calì, M., & te Velde, D. W. (2016). Does energy efficiency improve technological change and economic growth in developing countries?. Energy

AC

Policy, 92, 279-285.

Chang, C. C. (2010). A multivariate causality test of carbon dioxide emissions, energy consumption and economic growth in China. Applied Energy, 87(11), 3533-3537. Chen, J., Xu, C., Managi, S., & Song, M. (2019). Energy-carbon performance and its changing trend: An example from China’s construction industry. Resources, Conservation and Recycling, 145, 379-388. El Anshasy, A. A., & Katsaiti, M. S. (2014). Energy intensity and the energy mix: What works for the environment?. Journal of environmental management, 136,

ACCEPTED MANUSCRIPT 85-93. Fan, Y., & Xia, Y. (2012). Exploring energy consumption and demand in China. Energy, 40(1), 23-30. Feng, C., & Wang, M. (2017). The economy-wide energy efficiency in China’s regional building industry. Energy, 141, 1869-1879.

PT

Fujii, H., Cao, J., & Managi, S. (2016). Firm-level environmentally sensitive productivity and innovation in China. Applied energy, 184, 915-925.

ratio fallen in China?. The Energy Journal, 63-91.

RI

Garbaccio, R. F., Ho, M. S., & Jorgenson, D. W. (1999). Why has the energy-output

SC

Gillingham, K., Kotchen, M. J., Rapson, D. S., & Wagner, G. (2013). Energy policy: The rebound effect is overplayed. Nature, 493(7433), 475.

NU

Greening, L. A., Greene, D. L., & Difiglio, C. (2000). Energy efficiency and consumption—the rebound effect—a survey. Energy policy, 28(6-7), 389-401.

MA

Hatzigeorgiou, E., Polatidis, H., & Haralambopoulos, D. (2008). CO2 emissions in Greece for 1990–2002: a decomposition analysis and comparison of results using

D

the Arithmetic Mean Divisia Index and Logarithmic Mean Divisia Index

PT E

techniques. Energy, 33(3), 492-499. Hoekstra, R., & Van den Bergh, J. C. (2003). Comparing structural decomposition analysis and index. Energy economics, 25(1), 39-64.

Gaussian

CE

Johansen, S. (1991). Estimation and hypothesis testing of cointegration vectors in vector

autoregressive

models. Econometrica:

Journal

of

the

AC

Econometric Society, 1551-1580. Li, F., Song, Z., & Liu, W. (2014). China’s energy consumption under the global economic crisis: Decomposition and sectoral analysis. Energy Policy, 64, 193-202. Li, H., Mu, H., & Zhang, M. (2011). Analysis of China's energy consumption impact factors. Procedia Environmental Sciences, 11, 824-830. Li, J., Yang, L., & Long, H. (2018). Climatic impacts on energy consumption: Intensive and extensive margins. Energy Economics, 71, 332-343. Li, L., Wang, J., Tan, Z., Ge, X., Zhang, J., & Yun, X. (2014). Policies for eliminating

ACCEPTED MANUSCRIPT low-efficiency production capacities and improving energy efficiency of energy-intensive industries in China. Renewable and Sustainable Energy Reviews, 39, 312-326. Lin, B., & Du, K. (2015). Measuring energy rebound effect in the Chinese economy: an economic accounting approach. Energy economics, 50, 96-104.

PT

Liu, L., Wang, S., Wang, K., Zhang, R., & Tang, X. (2016). LMDI decomposition analysis of industry carbon emissions in Henan Province, China: comparison

RI

between different 5-year plans. Natural Hazards, 80(2), 997-1014.

Liu, X., Zhou, D., Zhou, P., & Wang, Q. (2018). Factors driving energy consumption

SC

in China: a joint decomposition approach. Journal of Cleaner Production, 172, 724-734.

NU

Mansur, E. T., Mendelsohn, R., & Morrison, W. (2008). Climate change adaptation: A study of fuel choice and consumption in the US energy sector. Journal of

MA

Environmental Economics and Management, 55(2), 175-193. Masjuki, H. H., Mahlia, T. M. I., & Choudhury, I. A. (2001). Potential electricity

D

savings by implementing minimum energy efficiency standards for room air

PT E

conditioners in Malaysia. Energy conversion and management, 42(4), 439-450. Moutinho, V., Moreira, A. C., & Silva, P. M. (2015). The driving forces of change in energy-related CO2 emissions in Eastern, Western, Northern and Southern

CE

Europe: The LMDI approach to decomposition analysis. Renewable and Sustainable Energy Reviews, 50, 1485-1499.

AC

National Bureau of Energy Statistics of China. (2018). China statistical yearbook 2013.

Neto, A. B. F., Perobelli, F. S., & Bastos, S. Q. (2014). Comparing energy use structures: An input–output decomposition analysis of large economies. Energy Economics, 43, 102-113. Peng, L., Zeng, X., Wang, Y., & Hong, G. B. (2015). Analysis of energy efficiency and carbon dioxide reduction in the Chinese pulp and paper industry. Energy Policy, 80, 65-75. Peng, L., Zhang, Y., Wang, Y., Zeng, X., Peng, N., & Yu, A. (2015). Energy efficiency

ACCEPTED MANUSCRIPT and influencing factor analysis in the overall Chinese textile industry. Energy, 93, 1222-1229. Román-Collado, R., & Colinet, M. J. (2018). Is energy efficiency a driver or an inhibitor of energy consumption changes in Spain? Two decomposition approaches. Energy Policy, 115, 409-417.

PT

Saidi, K., & Hammami, S. (2015). The impact of CO2 emissions and economic growth on energy consumption in 58 countries. Energy Reports, 1, 62-70.

RI

Soytas, U., & Sari, R. (2003). Energy consumption and GDP: causality relationship in G-7 countries and emerging markets. Energy economics, 25(1), 33-37.

SC

Sun, J. (1998). Changes in energy consumption and energy intensity: a complete decomposition model. Energy economics, 20(1), 85-100.

NU

Wang, M., & Feng, C. (2017). Decomposing the change in energy consumption in China's nonferrous metal industry: An empirical analysis based on the LMDI

MA

method. Renewable and Sustainable Energy Reviews, 82, 2652-2663. Wang, H., Ang, B. W., & Su, B. (2017). Assessing drivers of economy-wide energy

D

use and emissions: IDA versus SDA. Energy Policy, 107, 585-599.

PT E

Wang, Z. (2000). Model selection using the Akaike information criterion. STATA technical bulletin, 9(54).

Xiong, P. P., Dang, Y. G., Yao, T. X., & Wang, Z. X. (2014). Optimal modeling and

623-634.

CE

forecasting of the energy consumption and production in China. Energy, 77,

AC

Xu, S. C., Xi, R., & He, Z. X. (2012). Influential factors and policy implications of carbon emissions for energy consumption in China. Resources Science, 34(1), 2-12.

Yang, Z., Shao, S., Yang, L., & Miao, Z. (2018). Improvement pathway of energy consumption structure in China's industrial sector: From the perspective of directed technical change. Energy Economics, 72, 166-176. Zhang, H., & Lahr, M. L. (2014). China's energy consumption change from 1987 to 2007: A multi-regional structural decomposition analysis. Energy Policy, 67, 682-693.

ACCEPTED MANUSCRIPT Zhang, Z. (2003). Why did the energy intensity fall in China's industrial sector in the 1990s?

The

relative

importance

of

structural

change

and

intensity

change. Energy economics, 25(6), 625-638. Zhao, N., Xu, L., Malik, A., Song, X., & Wang, Y. (2018). Inter-provincial trade driving energy consumption in China. Resources, Conservation and Recycling,

PT

134, 329-335. Zhao, X., Li, N., & Ma, C. (2012). Residential energy consumption in urban China: a

AC

CE

PT E

D

MA

NU

SC

RI

decomposition analysis. Energy Policy, 41, 644-653.

ACCEPTED MANUSCRIPT Table 1 Statistical description of data and variables included in research Variables

Definition (units) 4

Mean

Min

Max

SD

N

Total energy use (10 tce)

129587

5411

449000

132914.2

65

CO2

CO2 emission (million ton)

837.6

39.36

2658.55

824.49

65

91211

5104

280999

87792.2

65

23158

206.2

84412

23889.7

65

Annual natural gas use (10 tce)

4686.3

1.08

31430

7447.78

65

Annual primary electricity use (104 tce)

10532

99.56

61962

15183.4

65

Year-end total population (104 persons)

103466

58796

139008

25565.5

65

22023

824.4

116592

30695.2

65

4

Coal

Annual coal use (10 tce) 4

Crude oil

Annual crude oil use (10 tce) 4

Natural gas Primary electricity and other energy Population

8

Gross domestic product (10 yuan)

RI

GDP

PT

Energy Use

Note: SD=standard deviation; N=number of values; data resources: data for Energy Use, Coal, Crude Oil, Natural Gas, and Primary Electricity and Other Energy, population and GDP between 1953 and 2017 come from the China Statistical Yearbook for

SC

each relative year, and GDP is converted into 1953 constant price (1953=100). We updated the carbon emissions inventories for China from 1953 to 2017 using the approach from ‘Comprehensive report on sustainable development of energy and carbon

AC

CE

PT E

D

MA

NU

emissions in China’ (2003).

ACCEPTED MANUSCRIPT Table 2 Traditional unit root tests for energy use and carbon emissions DF-GLS

KPSS

ERS

ADF

DF-GLS

KPSS

ERS

t-statistic

-0.4011

-2.0840

0.2366

76.530

-0.3750

-2.0961

0.2336

103.497

1%

-4.1130

-3.7396

0.2160

4.2320

-4.1305

-3.7396

0.2160

4.2320

5%

-3.4840

-3.1644

0.1460

5.6960

-3.4921

-3.1644

0.1460

5.6960

10%

-3.1701

-2.866

0.1190

6.7760

-3.1748

-2.866

0.1190

6.7760

Prob.

0.9855

t-statistic

-3.4997

-3.4862

0.0927

3.8853

-3.3440

-3.3910

0.0710

2.3112

1%

-4.113017

-3.7244

0.2160

4.2312

-4.1104

-3.7206

0.2160

4.2312

5%

-3.483970

-3.1516

0.1460

5.6976

-3.4828

-3.1484

0.1460

5.6976

10%

-3.170071

-2.854

0.1190

6.7756

-3.1694

-2.851

0.1190

6.7756

Prob.

0.0482

0.9862

0.0686

AC

CE

PT E

D

MA

NU

SC

1st difference

ADF

RI

Level

CO2

PT

EC

ACCEPTED MANUSCRIPT Table 3 Johansen linear cointegration test for energy use and carbon emissions Trace Statistic

0.05 Critical Value

Prob.

Max-Eigen Statistic

0.05 Critical Value

Prob.

None *

0.408200

36.93168

25.87211

0.0014

32.52435

19.38704

0.0004

At most 1

0.068618

4.407327

12.51798

0.6827

4.407327

12.51798

0.6827

AC

CE

PT E

D

MA

NU

SC

RI

PT

Eigenvalue

ACCEPTED MANUSCRIPT Table 4 Granger causality test for energy use and carbon emissions F-Statistic

Prob.

Causality decision

CO2 does not Granger Cause EC

2.43210

0.0969

Reject

EC does not Granger Cause CO2

6.78074

0.0023

Reject

AC

CE

PT E

D

MA

NU

SC

RI

PT

Null Hypothesis

ACCEPTED MANUSCRIPT

Climate Change

Individuals and Firms

Temperature

Chemical Reaction

Demand

?

Carbon Emissions

Energy Use

Consumption

AC

CE

PT E

D

MA

NU

SC

RI

PT

Figure 1 The relationship between carbon emissions and energy use in the social and climate system

ACCEPTED MANUSCRIPT

∆E(Ej/E)

∆E(E/GDP)

∆E(GDP/POP)

∆E(CO2/POP)

∆E(CO2)

∆E

500 400

200 100

2016-2017

2013-2014

2010-2011

2007-2008

2004-2005

RI

1995-1996

1992-1993

1989-1990

SC

1986-1987

1983-1984

1980-1981

1977-1978

1974-1975

1971-1972

1968-1969

1965-1966

1962-1963

1959-1960

1956-1957

-300

1953-1954

-200

2001-2002

-100

PT

0

1998-1999

Million tce

300

AC

CE

PT E

D

MA

NU

Figure 2 Changes in China’s energy consumption (black curve) from 1953 to 2017 influenced by energy mix effect (orange curve), energy intensity effect (gray curve), output effect (purple curve), per capita emission effect (blue curve), and emission effect (red curve)

ACCEPTED MANUSCRIPT

AC

CE

PT E

D

MA

NU

SC

RI

PT

Highlights  We reveal the bidirectional causality between energy use and carbon emissions  Emission effect promoted energy use and surpassed output effect during 2001-2005  Energy use followed the tract of changes in output effect at stage one  Energy use followed changes in emission effect or energy intensity effect at stage two  Energy efficiency has gradually been becoming a dominate role in energy use changes  The impact of energy structure effect was comparatively small