Energy 258 (2022) 124789

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Assessing the impact of the carbon market on the improvement of China’s

energy and carbon emission performance
Anhua Zhou a, b, *, Ling Xin a, Jun Li c
a
School of Science, Hunan University of Technology and Business, Changsha, 410205, China
b
Key Laboratory of Hunan Province for Statistical Learning and Intelligent Computation, Changsha, 410205, China
c
Business School, Hunan Normal University, Changsha, 410081, China

A R T I C L E I N F O A B S T R A C T

Keywords: All countries are actively taking measures to address carbon neutrality. As a policy measure for China to respond
Carbon market to climate change, the carbon market is necessary and significant for assessing its impact on energy and the
Energy intensity environment. This study uses data from 30 provinces in China from 2000 to 2017 and employs a difference-in-
Difference-in-difference model
differences model to estimate the impact of the carbon market on energy and environmental performance. The
Carbon emission efficiency
Carbon neutrality
empirical results show that the carbon market has significantly improved energy and environmental perfor­
mance, as shown by a 14.14% reduction in energy intensity and an increase in carbon emission efficiency by
4.21%; a series of robustness test results show that the carbon market only has long-term stability characteristics
in reducing energy intensity, but is unstable in improving emission efficiency; market reform and technological
innovation have played a significant intermediary role in the impact of the carbon market on energy and
environmental performance; and the carbon market has significant heterogeneity effects; only Beijing,
Chongqing, and Hubei have significant energy intensity reduction effects. Finally, the results revealed that the
carbon market is beneficial for improving energy and environmental performance, which provides a universal
reference for countries worldwide to implement carbon emission reduction policies.

1. Introduction carbon market on the reduction of carbon emissions. Most scholars have
demonstrated that carbon trading systems significantly reduced carbon
In recent years, the trend of global warming has intensified, and dioxide emissions [8,9]. Second, we analyzed the economic benefits of
global warming has caused significant damage to the natural ecological carbon market policies. For direct effects, carbon trading policies can
environment, on which mankind depends on survival. Extreme weather promote the development of China’s low-carbon economy, and the ef­
occurs frequently, and greenhouse gas (GHG) emission reduction is ur­ fect of low-carbon transformation in the western region is better than
gent [1]. In particular, the current global epidemic is spreading rapidly, that in the eastern and central regions [10]. For indirect effects, carbon
urging the entire human society to deeply reflect on the relationship trading policies induce low-carbon urban technological innovation [11].
between humans and nature and climate governance issues [2]. In With the increase in innovation input and output, the carbon market
response to current difficulties, China actively assumes the re­ effectively stimulates the innovation effect, promoting green economic
sponsibility of a major country, and the continuous development of growth [12]. Third, this study examined the comprehensive effects of
carbon market (carbon emissions trading) model has been piloted in carbon trading policies on social and economic development. The cur­
seven provinces and cities, including Beijing, Tianjin, and Hubei [3,4]. rent research perspectives on the comprehensive effects of carbon
The carbon market is considered an effective means of reducing global trading policies on emissions reduction and efficiency enhancement are
GHG emissions and mitigating climate change. different, and the conclusions are different. Some scholars have revealed
Since China’s carbon market pilot project, many scholars have con­ from a provincial perspective that carbon market policies can provide
ducted extensive discussions on whether it can promote sustainable employment and are also environmentally friendly [3,13]. It was also
social and economic development [5–7]. The existing literature mainly found that the overall level of the carbon trading mechanism has
focuses on the following aspects: First, research on the impact of the improved economic and environmental benefits based on the industry

* Corresponding author. Hunan University of Technology and Business, Changsha, 410205, China.
E-mail address: [email protected] (A. Zhou).

https://doi.org/10.1016/j.energy.2022.124789
Received 1 November 2021; Received in revised form 17 June 2022; Accepted 8 July 2022
Available online 20 July 2022
0360-5442/© 2022 Elsevier Ltd. All rights reserved.
A. Zhou et al. Energy 258 (2022) 124789

perspective [7]. However, some scholars have found that the risk level in 2. Literature review
China’s carbon market is still relatively high, revealing that there is no
herding behavior in China’s overall carbon market [14,15]. 2.1. The comprehensive effect of carbon market on emission reduction
At present, there is limited academic research on the carbon market. and economic society
First, the research evaluating the impact of the carbon market is insuf­
ficient. Some studies concerning the impact of carbon trading on carbon At present, many scholars and policymakers have conducted rela­
emissions [9,16,17] do not involve efficiency. The gap between carbon tively rich research on China’s carbon trading policy; however, these
emissions and emissions efficiency was relatively large. Emission effi­ studies are mainly reflected in carbon emission reduction, economic
ciency is based on an environmental perspective, emphasizing the benefits, and social development.
impact of carbon emissions on the environment as an output appendage. First, since the implementation of the carbon market policy, some
This indicator is considered from an economic perspective, hoping to scholars have paid attention to the emission reduction effect of the
obtain the maximum economic output with the smallest carbon resource carbon market policy earlier. Zhang et al. [17] used 39 industrial sectors
input, which is an important indicator of a low-carbon economy. Sec­ to assess the environmental benefits from the carbon market, showing
ond, most of the research literature only focuses on the emission effect that emissions were much lower than before the policy. Some scholars
[6,8,18] while ignoring the synergistic impact of the carbon market on found carbon market pilots have significant carbon emission reduction
the reduction of energy intensity. This study investigated the role of the effects at the national and regional levels. There are mediating effects of
carbon market in energy intensity and its impact mechanism for the first innovation, industrial upgrading, and configuration optimization [8],
time to reveal the co-benefits of the carbon market on energy and and [19] It was found that the carbon market system can reduce carbon
environmental performance. Third, research on the influence mecha­ emissions by 11%. Some scholars have found that carbon emissions can
nism is insufficient in many studies [2,19,20], and we will extend this be reduced by 15.5%.
analysis. Finally, existing studies do not accurately grasp the year of Second, regarding the implementation effects of carbon market
implementation of the carbon emission policy. Most of the literature policies, some scholars have examined the economic benefits of carbon
uses 2013 as the year of policy implementation [3,12,19]. Many prov­ market policies and found that carbon trading policies can promote the
inces had already implemented it quietly in 2012 because this is development of China’s low-carbon economy [10]. Some scholars have
foreseeable. found that with an increase in innovation input and innovation output,
The contributions of this study to existing research are as follows: (1) the carbon market effectively stimulates the innovation effect, promot­
Contribution of research perspective: this study comprehensively ana­ ing the growth of the green economy [24]. Some scholars have found
lyzes the role of the carbon market in energy performance and carbon that pilot carbon market policies cannot stimulate companies’ green
emission efficiency. Compared with existing research that focuses only innovation because of reduced output [12]. Yang et al. [6] found that
on the impact of the carbon market on carbon emissions [16,17], this the performance of green production varies significantly between
study expands the limitations of existing research from two aspects: provinces and that carbon trading significantly improves the efficiency
comprehensive energy intensity and efficiency. This study is more of green production.
comprehensive and in depth. Therefore, this study expands the dividend Finally, in addition to the carbon emission reduction effect and
effect of the carbon market in terms of emissions from the perspective of economic benefits of the carbon market, some scholars have paid
energy for the first time and deeply analyzes its internal mechanism. (2) attention to the social development benefits of carbon market policies.
Contribution of the research method: most of the methods for measuring Yan et al. [3] found that the carbon market from a provincial perspective
emission efficiency are based on data envelopment analysis (DEA) and is conducive to reducing emissions and improving employment condi­
its extended methods, such as the slacks-based measure (SBM) model; tions. Some scholars have found that carbon market policy has generally
however, these methods are analyzed from a single perspective, such as reduced the investment of companies covered by the carbon market [4].
a non-radial perspective, and ignore the role of the radial perspective Yu et al. [7] found that the overall level of the carbon trading mecha­
[21–23]. Considering these limitations, we adopted the nism improved economic and environmental benefits from an industry
super-epsilon-based measure (EBM) model, which includes two radial perspective. In contrast [5], carbon trading significantly reduced green
distance functions simultaneously, making the results closer to the real total factor productivity and enterprise total factor productivity from the
situation. Compared to the existing literature, this article is novel in the corporate perspective.
carbon emission efficiency measurement method. (3) Contribution of Based on the above analysis, this article proposes the following
research data processing: this article adjusts the policy implementation hypotheses:
year from 2013 to 2012. Since the relevant policies were promulgated in
Hypothesis 1. The carbon market significantly improves energy and
2011, the formal implementation of the policy was in 2013. However,
environmental performance.
many provinces began to implement it quietly in 2012, because it is
predictable and a sign of precaution. Therefore, compared to 2013 as a
policy implementation point, 2012 was more appropriate and accurate. 2.2. The mechanism of carbon market on energy carbon emission
This study used panel data from 30 provinces in China from 2000 to performance
2017. This study aimed to reveal the impact of the carbon market on
energy and environmental performance. The difference-in-differences As a government-guided and market-led environmental regulation
(DID) model was used to analyze the impact of energy and carbon tool, the carbon market can effectively achieve the role of “reducing
emission performance in the carbon market, and a mechanism analysis carbon and promoting economic growth” and positively affect energy
was implemented. Our research provides an important reference for and environmental performance; its influence channels are
policymakers regarding the effectiveness of carbon market systems, multifaceted.
improving energy and environmental performance, and achieving sus­ First, a carbon market is a market-based environmental rule. In
tainable development. practice, imperfect competition generally exists in the market. Some
The remainder of this study is organized as follows. A review of the companies may purchase and store emission rights that exceed their
previous research literature is presented in Section 2. Section 3 describes emission quotas to obtain monopoly benefits or for future use, such as a
the measurement model and data, and Section 4 presents the results. decline in product market output [25]. Some scholars use US data to find
Finally, Section 5 summarizes the conclusions and policy that reform of the electricity market is conducive to improving energy
recommendations. efficiency and renewable energy use, thereby playing an important role
in reducing carbon emissions [26]. Some scholars have found that the

2
A. Zhou et al. Energy 258 (2022) 124789

level of marketization significantly improves energy efficiency in all Rhodes (CCR) model [35–38]. In addition, another method was used in
quantiles [27]. [28] They analyzed the welfare effects of Mexico’s the existing literature is the SBM model, which is analyzed from a
market-oriented reforms and found that energy-market-oriented re­ non-radial perspective [39–41]. These DEA models and their extended
forms are not conducive to low-income groups. Based on the importance versions have different degrees of defects. The traditional DEA model
of market reforms, we propose the following hypothesis: cannot calculate efficiency scores at different scales of input and output.
However, this restriction led to a reduction in all input factors in the
Hypothesis 2. The carbon market is conducive to the improvement of
same proportion, which is not conducive to accurately and objectively
energy and environmental performance through market reform
assess the level of efficiency. Although the SBM model considered from a
channels.
non-radial perspective can effectively improve the impact of different
Second, the carbon emission trading system affects the resource input ratios, this method also has a major flaw: it ignores the relation­
allocation, investment, and research and development behavior of en­ ship ratio between the boundary and the projection [13,42,43].
terprises [29]. Several studies investigated whether the pollution To overcome the limitations of the efficiency scores measured by the
trading system can produce the Porter effect [30]. [31] It was found that CCR and SBM models. The EBM model proposed in Ref. [42] overcomes
a carbon emission trading system can induce green innovation activities, the limitations of the two models. Combining the two situations makes
marked by the growth of green invention patents. The carbon emission the EBM model more advantageous because the model we use considers
trading system imposes more severe survival pressure on companies many aspects; it effectively overcomes the problems of slack and
with relatively high pollution control costs or low energy utilization reduced efficiency. The linear relationship between the two results in
efficiency, forcing them to focus on energy conservation and emission lower efficiency than the actual efficiency level [44]. The model is as
reduction in the production process, which in turn forces them to follows:
improve their energy efficiency. Therefore, we propose the following
∑m
b−i φ−i
hypotheses: ζ* = minp − ax (1)
p,η,φ− i=1
ci0
Hypothesis 3. Innovation plays an intermediary role in the carbon
market’s impact on energy and environmental performance. ∑
n
pci0 − ηj cij − φ−i = 0, i = 1, ⋯, n
Third, government departments actively emphasize the restriction of j=1
GHG emissions and steadily promote the transformation of economic ∑
n

and industrial structure [32]. Upgrading the industrial structure is an { ηj yrj ≥ yr0 , r = 1, ⋯, n (2)
indispensable part of the carbon market operation mechanism, and the j=1

pressure of emission reduction brought about by policies has caused the ηj ≥ 0

initial elimination of high-energy-consuming industries [33]. Specif­ φ−i ≥ 0
ically, in the process of optimizing the industrial structure, emerging
industries are gradually replacing traditional industries, such as high where, ζ refers to the best efficiency value calculated by us, p refers to
energy consumption and pollution, and a set of green industrial systems the efficiency level in the radial case, φi refers to the i-th element in a
featuring low carbon have been formed. Therefore, we propose the non-radial perspective, η refers to the relative weight of the input ele­
following hypotheses: ments, (ci0, cr0) refers to the situation where the vector is in the 0-th
position, bi refers to the weight of the i-th element, and satisfies
Hypothesis 4. The carbon market is conducive to improving energy ∑m
and environmental performance through industrial structure i=1 bi = 1; ax refers to the parameter to be estimated in the case of
radial and non-radial. If ax = 1, the equation can be transformed into a
adjustment.
traditional DEA model. If ax and p = 1, the equation can be transformed
The mechanism analysis of the research content of this article is into an extended SBM model. The size of bi and ax was needed in
shown in Fig. 1. We analyze the three channels through which the car­ calculation. We calculated ζ* = 1, and found that the efficiency we used
bon market affects energy and carbon emission performance. was effective. Therefore, this model was useful to measure China’s
carbon emission performance. The model after expansion is as follows:
3. Econometric model and empirical data
∑
m −
bi φ−i
p − ax ci0
3.1. Method *
ζ = min i=1
(3)
φ + +
∑ q s−
∑
br φr bv φs−v
m + ay yr0
+ as sv0
3.1.1. Measurement method r=1 v=1

Measurement of carbon emission efficiency has been widely dis­

∑
n
cussed. Most studies measure efficiency using the DEA method, which ηj cij + φ−i − pci0 = 0, i = 1, ⋯, n
includes information regarding input and output variables [34]. By j=1

summarizing most existing research literature, the existing research can ∑

n

be broadly divided into two categories. The first method was based on ηj yrj − φ+r − myr0 = 0, r = 1, ⋯, s
s.t.{ (4)
radial distance function measurement, such as the Charnes, Cooper, and
j=1

∑
n
ηj svj + φb−v − msv0 = 0, v = 1, ⋯, q
j=1

ηj ≥ 0, φ−i ≥ 0, φ+r ≥ 0, φs−v ≥ 0

where, sv0 refers to the v-th undesired result of the 0-th individual, and
s−
(φ+r , φv ) refers to the relaxation vector of the r-th expected result and
the v-th undesired result. When the level of these two values is greater
than zero, the real specific value does not exceed the production frontier
level, implying that efficiency needs to be improved. bi refers to the
∑ -
weight of the i-th element, and satisfies m i=1 bi = 1, br and bv refer to the
+
Fig. 1. Diagram of the mechanism of the carbon market affecting energy and
r-th expected result and the v-th undesired result, respectively; and the
carbon emission performance.

3
A. Zhou et al. Energy 258 (2022) 124789

other symbols are the same as in Equation (1). Considering that the measurement of this indicator includes the input and output variables.
value calculated using the EBM model is ≤ 1, it is impossible to distin­
guish the most effective unit. To overcome this limitation [45], proposed (1) Capital: This variable is used as an input element when we
a super-efficiency method, that can sort effective decision-making units, calculate carbon emission performance. Based on the existing
and the calculated value can exceed 1. If the value is > 1, it indicates that research [47], we use the total fixed asset investment in different
carbon emission efficiency is the most effective unit; otherwise, it is regions (unit: 100 million RMB) to measure the variable. Using
invalid. Therefore, this study combined the super-efficiency method to the year 2000 as the base period, the conversion method was used
construct a super-EBM model to measure and empirically analyze car­ for variable processing.
bon emission efficiency. (2) Labor: This variable is used as an input element, because labor
input is a necessary element, we use the final employment status
3.1.2. Difference-in-differences model at the end of the year in different regions as an indicator of labor
The carbon emission trading policy was considered a quasi-natural input (unit: 10 thousand).
experiment. In 2011, the National Development and Reform Commis­ (3) Energy: This variable was also used as the input variable. The
sion decided to launch pilot carbon emission trading projects in seven energy is an important input factor. Based on the existing liter­
provinces and cities, including Beijing, Shanghai and Tianjin, and offi­ ature [36], we used the total energy consumption of different
cially launched it in 2013. In 2012, many provinces began to implement regions (unit: ten thousand tons of standard coal) as the energy
carbon trading policies, which is predictable. If 2013 is still a policy input.
point, then the estimated results may contain large deviations. There­ (4) Economic output: This variable is regarded as an output variable
fore, we adjusted the policy implementation year from 2013 to 2012 to because economic output reflects the economic benefits brought
make the results more accurate; therefore, this study used 2012 as the by the input of the factors. We used the total output value of each
starting year of policy implementation. The provinces that were piloted region in a year to express the economic output (unit: 100 million
after 2012 were used as the experimental group, and the provinces RMB).
without the pilot were used as the control group. According to the (5) Environmental output: Since the carbon market policy is mainly
existing research [46], the form of the DID model used in this study is as for carbon emission behavior; thus, our undesired output was
follows: selected as the carbon emissions. The data on of carbon emissions
in various provinces were based on the estimation method of CO2
∑
n
eiit = α0 + β1 didit + βi xit ​ + γ t + δi + εit , i = 1, 2, ⋯, N; t = 1, 2, ⋯, T in the IPCC [48]. Coal, coke, crude oil, gasoline, fuel oil, kero­
i=2 sene, diesel, natural gas, electricity and nine other types of energy
(5) consumption were used to calculate carbon emissions, and the
specific calculation method is shown in equation (7).
∑
n
ceeit = b0 + b1 didit + bi xit ​ + ψ t + λi + μit , i = 1, 2, ⋯, N; t ∑
9
44
i=2 CO2it = En × EFn × OFn × NCVn × (7)
12
= 1, 2, ⋯, T (6) i=1

CO2 represents the carbon emissions of province i in year t; n

where eiit and ceeit stand for energy performance and carbon emission respectively refers to nine types of energy consumption; En represents
performance for province i at time t, respectively, and did refers to the the physical consumption of nine types of energy; EFn represents the
carbon market (carbon emissions trading system) indicators (did = dt × carbon emission coefficient per calorific value of energy n; OFn repre­
du, dt and du are both dummy variables that indicate whether the time is sents the carbon of energy n oxidation rate; NCVn represents the average
greater than 2012, and whether the province has performed carbon low calorific value of energy n. The last term is the conversion factor.
emission trading pilots. If yes, the value is 1; otherwise, the value is 0);
β1 and b1 are the coefficients of interest. We expect β1 to be negative and 3.2.2. Independent variables
b1 to be positive, implying that the carbon market is conducive to The carbon market (did) is the key variable in this study. In 2011, the
improving energy and environmental performance; β and b are the National Development and Reform Commission issued the “Notice on
variable parameter, and x is the control variable; γt and ψ t refer to the Carrying out Pilot Carbon Emission Trading” and simultaneously
time fixed effects; δi and λi capture the province fixed effect; εit and μit approved the pilot work of carbon emissions trading in Shanghai, Bei­
refer to the random error terms. jing, Guangdong, Shenzhen, Tianjin, Hubei, and Chongqing, which was
officially launched at the end of 2013. At the end of 2013, these seven
regions officially launched a pilot work on the carbon market, which
3.2. Variables marked the initial formation of the carbon market in China. These
provinces started in 2012, considering that the actual implementation
3.2.1. Explained variable was in 2012. This study selected the year 2012 as the policy processing
To explore the energy and environmental effects of the carbon year, the value of the pilot provinces after 2012 was set to 1, and the
market, the variables used to measure energy and carbon emission value before 2012 was set to 0. Because Guangdong and Shenzhen
performance in this study are divided into two types: one is energy belong to the same province, we unified these two regions as Guang­
performance. We used the energy intensity (ei) to reflect the energy dong. Some scholars have used a similar approach [16,49]. Therefore,
performance. Energy intensity reflects the dependence of regional eco­ the carbon market was used as a dummy variable in this study.
nomic growth on energy. Reducing energy intensity is the goal of low-
carbon economic development, actively promoted by the country. 3.2.3. Control variable
Therefore, the use of energy intensity to express energy performance is
important. Several studies have also used these indicators for measure­ (1) Economic Growth (gdp): The existing literature on economic
ment [32]. This indicator is determined by the ratio of energy to eco­ growth and emissions shows that economic growth has a signif­
nomic output. The other is carbon emission performance; we used icant effect on the environment [18,50]. Following existing
carbon emission efficiency (cee) to replace carbon emission perfor­ research, this study uses GDP per capita as the gdp indicator (unit:
mance. Carbon emission performance is more significant than carbon RMB). Moreover, to explore the nonlinear effects of economic
emissions, because the former considers the carbon emission efficiency growth, we set the square term of economic growth (gdp2).
of all elements, whereas the latter is only an absolute number. The

4
A. Zhou et al. Energy 258 (2022) 124789

(2) Foreign direct investment (fdi): This variable is used as a control 4. Results
variable because fdi has a close relationship with the environ­
ment, some scholars believe that it has a negative relationship 4.1. Calculation results of energy and environmental performance
[51–55] or a positive relationship [56–60], and the other is un­
stable relationship [61]. Therefore, this study used the fdi of each This study calculated the energy intensity and carbon emission ef­
region to measure this indicator (unit: RMB). ficiency according to the pilot and non-pilot areas, respectively, as
(3) Energy Structure (esc): esc has a significant impact on the effi­ shown in Fig. 2. The implementation time of the policy set in this article
ciency of the ecological environment [62]. Based on these this was 2012, because it started quietly before the formal implementation in
studies, we used the proportion of electricity consumption to 2013. Fig. 2 (a) shows that the energy intensity of the pilot and non-pilot
reflect this indicator (unit: %). areas declined. The trends of changes in policy implementation and non-
(4) Population (pop): Several scholars have used population variables implementation areas were relatively similar before the pilot policy in
to analyze the impact of population growth on the ecological 2012, and the non-pilot areas were higher than those in the non-pilot
environment [50]. To reflect the impact of population variables areas. After the implementation of the policy, the energy intensity of
on ecological efficiency, this study used the total population at the pilot areas declined faster than that of the non-pilot areas. Fig. 2 (b)
the end of the year to replace population indicators (unit: ten shows that the carbon emission efficiency of the pilot and non-pilot
thousand people). areas is basically U-shaped; the change trends of the pilot and non-
(5) Environmental regulations (er): The use of environmental regu­ pilot areas before the pilot policy in 2012 were basically parallel, and
lations as a control variable was mainly due to the interference of the pilot areas were higher than the non-pilot areas. After the imple­
environmental policies on the empirical results of this study. mentation of the policy, the carbon emission efficiency of the pilot areas
Many existing studies have discussed the effects of government suddenly increased, and the growth rate was higher than that of non-
regulations on the environment [30,63]. Therefore, we used implementation areas. Compared to non-pilot areas, the increase in
market-based environmental regulations as a measurement in­ carbon emission efficiency in pilot areas may be due to the 2012 pilot
dicator and the proportion of environmental protection invest­ policy of emissions trading.
ment in the output value of each region (unit: %). Table 2 summarizes the trend values of ei over time. Overall, the
(6) Industrialization (ind): The industrialization variable was used as value of ei slowly decreases over time, indicating that China’s energy
the control variable in this study because the environmental intensity has improved. From the data at the provincial level, there are
damage caused by the industrialization process is evident to all clear differences in ei in different provinces. In some areas, ei is still at a
[64,65]. To eliminate the impact of industrialization, we used the relatively high level. In contrast, in other areas, it is at a low level,
proportion of industrial added value to the gross product of each indicating that different provinces have different intensities of energy
region (unit: %). dependence.
Table 3 summarizes the change trend values of cee over time.
3.3. Data Overall, the value of cee shows wave-shaped changes over time, indi­
cating that China’s carbon emission efficiency fluctuates. From the data
For the empirical analysis provincial-level data from China were at the provincial level, there were clear differences in cee in different
used. Our research data was from 30 provinces in China, and the time provinces. In some areas, cee exceeds 1, whereas in other areas, it is at a
span was 2000–2017. Our data did not include Tibet, Taiwan, Hong lower level, indicating that different provinces have different carbon
Kong and Macau as data from these provinces were not available. The emission efficiencies.
data were obtained from the following databases: China Statistical
Yearbook [66], China Environmental Yearbook [67], the official website
of the Ministry of Ecology (http://www.mee.gov.cn/) and Environment, 4.2. Estimated results of the model
and the Bureau of Statistics of China (http://data.stats.gov.cn/). The
variable definitions and descriptive statistics were presented in Table 1. Table 4 summarizes the estimation results of the DID model with
The variable descriptive statistics of the treatment group and the control dependent variables lnei and lncee. The results in columns (1)–(2) show
group were shown in Appendix A, its results show that the average en­ that after controlling for individual fixed effects and time fixed effects,
ergy performance of the treatment group (0.834) was lower than that of the estimated coefficient of the core explanatory variable did is signifi­
the control group (1.362), while the carbon emission efficiency (0.803) cantly negative, regardless of whether the control variable was added.
was higher than that of the control group (0.580), which implied that The results in column (2) show that ei decreased by 14.14% in the pilot
the carbon emission efficiency in the treatment group was better than area after the carbon trading policy, indicating that the carbon trading
that in the control group, but there was still much room for improve­ policy reduced energy intensity. Thus, Hypothesis 1 was verified.
ment. In addition, both economic growth and foreign direct investment The results in columns (3)–(4) show that the estimated coefficient of
in the treatment group were higher than those in the control group. the core explanatory variable did is positive and significant at 1% and
5%, indicating that compared with non-pilot provinces and cities, car­
bon emission efficiency has been effectively improved after the

Table 1
Descriptive statistics of national samples.
Variables Obs Mean Std.Dev. Min Max

Energy performance (ei) 540.000 1.257 0.774 0.252 4.672

Carbon emission efficiency (cee) 540.000 0.625 0.213 0.241 1.029
Carbon market (did) 540.000 0.067 0.250 0.000 1.000
Economic growth (gdp) 540.000 12,850.980 14,078.900 263.700 91,648.720
Economic growth squared (gdp2) 540.000 3.630e+08 9.020e+08 69537.690 8.400e+09
Industrialization (ind) 540.000 0.455 0.080 0.190 0.590
Population (pop) 540.000 4392.387 2648.591 517.000 11,169.000
Environmental regulations (er) 540.000 0.002 0.002 0.000 0.035
Energy structure (esc) 540.000 0.959 0.363 0.069 2.212
Foreign direct investment (fdi) 540.000 516,087.900 662,207.400 1283.000 3575955.000

5
A. Zhou et al. Energy 258 (2022) 124789

Fig. 2. Comparison of the annual mean values of energy performance and carbon emission efficiency between the policy treatment group and the control group.

Table 2
The average annual energy performance of each province in China from 2000 to 2016.
2000 2002 2004 2006 2008 2010 2012 2014 2016 Mean

Beijing 1.311 1.044 0.852 0.727 0.569 0.493 0.401 0.320 0.271 0.665
Tianjin 1.642 1.405 1.188 1.008 0.798 0.739 0.637 0.518 0.461 0.933
Hebei 2.220 1.925 2.046 1.900 1.519 1.350 1.138 0.997 0.929 1.558
Shanxi 3.645 4.017 3.150 2.890 2.143 1.827 1.596 1.556 1.487 2.479
Neimenggu 2.306 2.349 2.507 2.270 1.660 1.441 1.246 1.030 1.073 1.765
Liaoning 2.282 1.942 1.960 1.611 1.302 1.135 0.947 0.762 0.893 1.426
Jilin 1.930 1.853 1.795 1.382 1.124 0.957 0.791 0.620 0.542 1.222
Heilongjiang 1.957 1.651 1.572 1.406 1.200 1.083 0.932 0.795 0.798 1.266
Shanghai 1.153 1.066 0.917 0.840 0.725 0.653 0.563 0.470 0.416 0.756
Jiangsu 1.007 0.906 0.910 0.876 0.718 0.622 0.534 0.459 0.401 0.715
Zhejiang 1.068 0.923 0.929 0.841 0.704 0.608 0.521 0.469 0.429 0.721
Anhui 1.681 1.510 1.264 1.156 0.941 0.785 0.660 0.576 0.520 1.010
Fujian 0.920 0.781 0.945 0.900 0.763 0.666 0.568 0.503 0.429 0.719
Jiangxi 1.251 1.061 1.103 0.967 0.772 0.672 0.559 0.513 0.473 0.819
Shangdong 1.363 1.075 1.306 1.222 0.988 0.889 0.778 0.614 0.569 0.978
Henan 1.567 1.425 1.528 1.313 1.053 0.928 0.799 0.655 0.571 1.093
Hubei 1.768 1.593 1.619 1.450 1.134 0.948 0.794 0.596 0.427 1.148
Hunan 1.146 1.215 1.347 1.376 1.069 0.928 0.756 0.567 0.501 0.989
Guangdong 0.880 0.841 0.806 0.751 0.638 0.585 0.511 0.436 0.336 0.643
Guangxi 1.283 1.182 1.224 1.136 0.925 0.827 0.702 0.607 0.551 0.937
Hainan 0.911 0.937 0.905 0.863 0.755 0.658 0.591 0.520 0.495 0.737
Chongqing 1.356 1.435 1.209 1.374 1.117 0.991 0.813 0.602 0.466 1.040
Sichuan 1.659 1.589 1.677 1.494 1.202 1.041 0.862 0.697 0.507 1.192
Guizhou 4.155 3.595 3.589 2.639 1.989 1.776 1.442 1.048 0.802 2.337
Yunnan 1.724 1.703 1.691 1.660 1.320 1.201 1.012 0.816 0.721 1.316
Shaanxi 1.514 1.648 1.504 1.292 1.014 0.877 0.735 0.634 0.625 1.094
Gansu 2.861 2.450 2.314 2.083 1.688 1.437 1.240 1.100 1.019 1.799
Qinghai 3.402 2.991 2.926 2.934 2.237 1.902 1.861 1.733 1.598 2.398
Ningxia 3.996 4.672 4.323 3.899 2.682 2.179 1.949 1.797 1.765 3.029
Xinjiang 2.441 2.246 2.223 1.986 1.690 1.525 1.576 1.610 1.689 1.887

implementation of the carbon trading policy. Therefore, the empirical well in various provinces.
results verify the Hypothesis 1. Fig. 3(b)shows that before the policy year, the confidence interval of
the regression coefficient crosses the horizontal axis, indicating that
4.3. Robustness analysis there is no significant difference in carbon emission efficiency between
pilot and non-pilot regions, and the parallel trend hypothesis is estab­
4.3.1. Dynamic effect test lished. After the policy implementation year in 2012, the regression
Fig. 3 (a) shows that before the policy implementation year, the coefficients include negative and positive values, and the marginal effect
confidence interval of the regression coefficient intersects the horizontal line is wavy, indicating that the carbon market has a significant
axis, indicating that the energy intensity of the pilot and non-pilot areas continuous volatility effect on the promotion of cee. The promotion of
has a common trend, and the parallel trend hypothesis is established. cee is also susceptible to interference from other factors; therefore, the
After the policy year 2012, the regression coefficients are all signifi­ policy effect has an unstable wave shape over time.
cantly negative, and the marginal effect line is obviously inclined to the
lower right, which indicates that the reduction effect of the carbon 4.3.2. Placebo test
trading policy on ei is stable and increases year by year. The reason for We conducted a placebo test to ensure the robustness of the results.
the year-by-year increase may be that carbon markets are functioning Some existing studies have used placebo tests for robustness analyses

6
A. Zhou et al. Energy 258 (2022) 124789

Table 3
The average annual carbon emission efficiency of each province in China from 2000 to 2016.
2000 2002 2004 2006 2008 2010 2012 2014 2016 Mean

Beijing 1.013 1.009 1.021 1.017 1.008 1.006 1.005 1.002 1.001 1.009
Tianjin 0.867 0.810 0.747 0.688 0.613 0.587 0.588 0.611 0.572 0.676
Hebei 0.701 0.761 0.727 0.560 0.487 0.372 0.337 0.303 0.304 0.506
Shanxi 0.883 0.795 0.744 0.633 0.613 0.418 0.342 0.327 0.282 0.560
Neimenggu 0.962 0.764 0.551 0.435 0.409 0.420 0.419 0.412 0.421 0.533
Liaoning 1.002 0.952 0.696 0.507 0.427 0.380 0.390 0.394 0.687 0.604
Jilin 0.796 0.684 0.649 0.418 0.358 0.359 0.367 0.347 0.335 0.479
Heilongjiang 0.930 0.865 0.890 0.724 0.577 0.397 0.349 0.349 0.299 0.598
Shanghai 1.011 1.001 1.010 1.007 1.001 1.002 1.002 1.000 1.001 1.004
Jiangsu 0.948 0.873 0.879 0.794 0.743 0.745 0.702 0.687 0.733 0.789
Zhejiang 0.824 0.895 0.861 0.809 0.753 0.780 0.717 0.633 0.600 0.764
Anhui 1.000 0.990 0.818 0.660 0.608 0.631 0.607 0.558 0.532 0.712
Fujian 1.024 1.001 0.943 0.821 0.707 0.709 0.680 0.613 0.616 0.790
Jiangxi 1.002 0.770 0.715 0.700 0.685 0.682 0.665 0.579 0.540 0.704
Shangdong 0.906 0.810 0.599 0.551 0.539 0.495 0.432 0.414 0.399 0.572
Henan 0.940 0.972 0.833 0.618 0.530 0.491 0.461 0.450 0.447 0.638
Hubei 0.721 0.732 0.754 0.683 0.602 0.540 0.482 0.515 0.614 0.627
Hunan 0.955 0.858 0.823 0.709 0.604 0.540 0.477 0.503 0.502 0.663
Guangdong 1.007 1.001 1.007 1.005 1.001 0.939 0.869 0.750 0.778 0.929
Guangxi 0.929 0.936 0.823 0.621 0.545 0.500 0.465 0.425 0.410 0.628
Hainan 0.970 0.833 0.871 0.822 0.693 0.688 0.622 0.564 0.524 0.732
Chongqing 0.858 0.708 0.674 0.542 0.487 0.476 0.471 0.507 0.560 0.587
Sichuan 0.734 0.692 0.689 0.585 0.532 0.462 0.446 0.433 0.510 0.565
Guizhou 0.653 0.545 0.580 0.568 0.553 0.413 0.321 0.299 0.321 0.473
Yunnan 0.782 0.805 0.746 0.562 0.517 0.465 0.432 0.400 0.395 0.567
Shaanxi 0.748 0.687 0.646 0.573 0.516 0.514 0.501 0.462 0.414 0.562
Gansu 0.683 0.651 0.688 0.649 0.528 0.383 0.317 0.258 0.242 0.489
Qinghai 0.481 0.412 0.475 0.435 0.454 0.320 0.257 0.293 0.313 0.382
Ningxia 0.511 0.465 0.429 0.421 0.425 0.310 0.323 0.349 0.339 0.397
Xinjiang 0.673 0.612 0.618 0.607 0.566 0.446 0.383 0.411 0.343 0.518

supported the conclusions of this study.

Table 4
Benchmark regression results.
4.3.3. Propensity score matching-DID test
Variables Lnei Lncee Considering the adverse effects of sample selection errors on esti­
（1） （2） （3） （4） mation results, some scholars have used propensity score matching
did − 0.129*** − 0.141*** 0.124*** 0.042** (PSM) methods to estimate sample selection problems. This method has
(0.024) (0.027) (0.017) (0.017) been used by some scholars to evaluate the effectiveness of environ­
Control No Yes No Yes mental policy [68]. In reference to existing research [69], this study
Year fixed Yes Yes Yes Yes used the PSM-DID technology formed by the combination of the PSM
Province fixed Yes Yes Yes Yes
and DID models to verify the impact of did on lnei and lncee.
cons 0.532*** 0.802 0.850*** − 5.584***
(0.019) (1.033) (0.013) (0.687) Table 6 presents the balance test results obtained after PSM match­
N 540 540 540 540 ing. We find that the standard deviations of the matching feature vari­
Adj-R2 0.921 0.936 0.753 0.815 ables are significantly reduced, and the results of the t-statistic are not
Note: * 10% significance, ** 5% significance, *** 1% significance. Year fixed is a significant after matching, indicating that the PSM method is effective in
time fixed effect and province fixed is an individual fixed effect. Control refers to matching. Thereafter, DID estimation is performed based on the samples
control variables, which are explained in the variable description. after PSM matching is completed.
Table 7 presents the regression results of the PSM-DID model.
[3]. Our approach is to conduct 500 samplings in all 30 Regardless of whether control variables are added, we find that the
prefecture-levels; each sampling randomly selects six provinces as hy­ carbon market has significant negative and positive effects on lnei and
pothetical treatment groups, and the rest as hypothetical reference lncee, respectively, proving that the carbon market can reduce lnei and
groups. increase lncee. Therefore, the conclusion is robust and verifies the val­
Fig. 4 shows the kernel density distribution diagrams of the two idity of Hypothesis 1.
explained variables. We found that the absolute value of the t-value of
most sampling estimation coefficients was within 2 and the p-value was 4.4. Mechanism analysis
above 0.1, indicating that the carbon market did not exert a significant
effect in 500 random samples. Therefore, our research conclusions The purpose of this section is to explore the mechanism by which did
passed the placebo test. affects the energy and carbon emission performance. According to the
Moreover, according to the existing literature [3], a placebo test can literature analysis, we believe that industrial structure adjustment,
also be performed by changing the time point. We advanced the time of technological innovation, and market reform are the channels through
the treatment group by 2 years and then performed regression analysis which did affects energy and carbon emission performance. Therefore,
again to observe the regression coefficient of did. If the coefficient of did this study uses industrial restructuring (indr), technological innovation
is significant, this implies that the empirical conclusions are reliable (ti) and market reform (mr) as intermediary variables. Based on existing
otherwise, the results are biased. We set the time after 2008 as the research [70], we designed a mechanism model as follows:
treatment group for reestimation. Table 5 summarizes the estimation
results and shows that the coefficient is not significant, implying that the
empirical results are robust. Therefore, the results verified and further

7
A. Zhou et al. Energy 258 (2022) 124789

Fig. 3. Schematic diagram of dynamic effect results, (a) and (b) represent the coefficients diagrams of energy performance and carbon emission efficiency,
respectively (red line indicates the policy point, dashed line represents the 95% confidence interval).

Fig. 4. Kernel function distribution diagram of 500 random sampling, (a) and (b) represent the distribution diagrams of energy performance and carbon emission
efficiency, respectively (The red line indicates when the t value is equal to the absolute value 2).

∑
n
Table 5 Mit = h0 + c1 didit + ci controlit ​ + σt + ϑi + εit
(8)
Placebo test results. i=2
​ i = 1, 2, ⋯, N; t = 1, 2, ⋯, T
Variables Time heterogeneity analysis

lnei Lncee ∑
n
ln eiit = d0 + χ 1 didit + χ 2 Mit + χ i controlit ​ + υt + τi + ηit
did2008 − 0.097 0.109 i=3
(9)
(0.061) (0.042) ​ i = 1, 2, ⋯, N; t = 1, 2, ⋯, T
Control Yes Yes
Year fixed Yes Yes
∑
n
Province fixed Yes Yes ln ceeit = g0 + ϖ 1 didit + ϖ 2 Mit + ϖ i controlit ​ + κt + ρi + μit
cons 1.072*** − 9.946*** i=3
(10)
(3.431) (2.856) ​ i = 1, 2, ⋯, N; t = 1, 2, ⋯, T
N 540 540
Adj-R2 0.681 0.143
where Mit is an intermediary variable (indr, ti, or mr) for province i at
Note: * 10% significance, ** 5% significance, *** 1% significance. time t. The coefficient c1 of equation (8) is the effect of did on the in­
termediate variable (M), the coefficient χ 2 or ϖ 2 of equations (9) and
(10) are the effect of M on the dependent variable (ei or cee) after con­
trolling for the influence of the did; the coefficient χ 1 or ϖ 1 of equations
(9) and (10) are the direct effect of did on the dependent variable, and
the coefficient interaction c1 *χ 2 or c1 *ϖ 2 is the mediating effect, σt , νt ,
and κt refer to the time fixed effects, ϑi , τi , and ρi capture the province
fixed effect, εit , ηit and μit refer to the random error terms.

8
A. Zhou et al. Energy 258 (2022) 124789

Table 6
Balance test results after PSM matching.
Variable Unmatched Mean %bias %reduct t-test

Matched Treated Control Bias t-test p-value

lngdp U 9.320 8.819 47.800 0.100 4.220 0.000

M 9.547 10.048 − 47.800 − 0.890 0.389
lngdp2 U 87.760 79.081 46.400 − 2.500 4.180 0.000
M 92.727 101.620 − 47.500 − 0.830 0.420
lnind U − 0.885 − 0.786 − 43.800 22.200 − 4.580 0.000
M − 0.803 − 0.726 − 34.100 − 1.660 0.120
lnpop U 8.009 8.191 − 24.600 61.600 − 2.250 0.025
M 8.777 8.707 9.400 0.350 0.731
lner U − 7.082 − 6.370 − 80.000 74.900 − 7.370 0.000
M − 6.812 − 6.991 20.100 0.390 0.703
lnesc U − 0.470 − 0.026 − 116.800 70.500 − 11.510 0.000
M − 0.271 − 0.140 − 34.500 − 0.950 0.357
lnfdi U 13.213 11.795 94.900 58.900 7.810 0.000
M 13.049 13.632 − 39.000 − 0.780 0.449

(2) Technological innovation was used as an intermediary variable.

Table 7 The regression coefficient of did was 0.0150, and the regression
PSM-DID results.
coefficients of lnti were 0.1789 and 12.0869, respectively, indi­
(1) (2) cating that did significantly improve lnti, but the mediating effect
Lnei lncee of lnti on lnei and lncee was positive, indicating that tc does not
did − 0.162*** 0.119**
have a positive effect on reducing lnei, but lncee promotion has a
(0.044) (0.065) mediating effect. Therefore, technological innovation only plays
Control Yes Yes an intermediary role in the carbon market’s impact on carbon
Year fixed Yes Yes emission efficiency. Therefore, the empirical results partially
Province fixed Yes Yes
verify Hypothesis 3.
cons − 9.709** − 7.982*
(4.512) (4.645) (3) Market reforms are used as intermediary variables. The regres­
N 440 440 sion coefficient of did was 0704, and the regression coefficients of
Adj-R2 0.946 0.825 lnmr were − 0.2403 and 0.0461, respectively, indicating that did
Note: * 10% significance, ** 5% significance, *** 1% significance. Year fixed is a significantly improve lnmr, and the mediating effect of lnmr on
time fixed effect and province fixed is an individual fixed effect. Control refers to lnei and lncee was negative and positive, indicating that lnmr had
the control variables. a positive effect on reducing lnei, but the mediating effect on lncee
was not significant. Therefore, market reforms only played an
Table 8 summarizes the results of the mechanistic analysis. (1) In­ intermediary role in the impact of the carbon market on energy
dustrial restructuring is used as an intermediary variable. The regression intensity. Therefore, the empirical results partially verify Hy­
coefficient of did is 0.0129, and the regression coefficients of indr are pothesis 4.
0.4756 and 0.0278, respectively, implying that did significantly im­
proves indr, but the mediating effect of indr on lnei and lncee is positive. 4.5. Heterogeneity analysis
This indicates that indr does not have a positive effect on reducing lnei,
but lncee promotion has a mediating effect, but it is not significant. The implementation of carbon market policies has a heterogeneous
Therefore, industrial restructuring does not play an intermediary role in effect, that is, there may be significant differences in the improvement
the carbon market’s effect on energy and environmental performance. effects of energy and carbon emission performance in Beijing, Tianjin,
Therefore, the empirical results do not pass Hypothesis 2. Shanghai, and Hubei. To explore the heterogeneity of policy effects in

Table 8
Mechanism analysis results.
Industrial restructuring Technological innovation Market reform

indr lnei lncee lnti Lnei lncee lnmr lnei lncee

did 0.012** − 0.136*** 0.266*** 0.015*** − 0.147*** 0.207*** 0.070*** − 0.124*** 0.268***
(0.006) (0.024) (0.030) (0.001) (0.027) (0.032) (0.020) (0.026) (0.030)
indr 0.475*** 0.027
(0.165) (0.086)
lnti 0.178* 12.086***
(2.529) (2.600)
lnmr − 0.240*** 0.046
(0.058) (0.075)
Control Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year fixed Yes Yes Yes Yes Yes Yes Yes Yes Yes
Province fixed Yes Yes Yes Yes Yes Yes Yes Yes Yes
_cons 0.117 0.757 − 0.1545* − 0.048*** 0.984 − 0.283*** 1.470* 1.155 − 0.112
(0.408) (1.022) (0.082) (0.018) (1.039) (0.032) (0.787) (1.020) (0.112)
N 540 540 540 540 540 540 540 540 540
Adj-R2 0.856 0.938 0.743 0.787 0.937 0.753 0.813 0.939 0.743

Note: * 10% significance, ** 5% significance, *** 1% significance. Year fixed is a time fixed effect and province fixed is an individual fixed effect. Control refers to the
control variables.

9
A. Zhou et al. Energy 258 (2022) 124789

the pilot areas, this study incorporates the interaction terms of did and Table 9
provinces into the model based on the benchmark model. The results are Results of province heterogeneity effects.
presented in Table 9. (1) (2)
Table 9 summarizes the results for regional heterogeneity. The re­
lnei lncee
sults in the first column show that the regression coefficients of Beijing,
Hubei, and Chongqing are negatively significant, whereas the regression did 0.008 0.173**
(0.053) (0.037)
coefficients of Tianjin and Shanghai are not significant, implying that did×Beijing − 0.269*** 0.194***
the carbon market to reduce energy intensity was only established in (0.074) (0.052)
Beijing, Chongqing and Hubei. In addition, the results in the second did×tianjin − 0.095 0.063
column show that the regression coefficients of Beijing and Shanghai are (0.074) (0.052)
did×shanghai 0.007 0.200***
positively significant, whereas other provinces are not significant,
(0.074) (0.052)
implying that Beijing and Chongqing have a stronger lncee-raising effect did×chongqing − 0.165** 0.087
than Guangdong province. (0.074) (0.052)
did×hubei − 0.307*** 0.043
(0.074) (0.052)
5. Conclusions and policy implications
Control Yes Yes
Year fixed Yes Yes
Achieving carbon neutrality is our main purpose, and it is a major Province fixed Yes Yes
task that China must complete ahead of schedule. To achieve carbon cons 0.532*** 0.850***
neutrality by 2060, China implemented a series of policies, one of which (0.019) (0.013)
N 540 540
is the carbon market. Therefore, the purpose of this study is to assess the
Adj-R2 0.926 0.741
effect of the carbon market on energy and environmental performance.
By evaluating the role of the carbon market in energy and environmental Note: * 10% significance, ** 5% significance, *** 1% significance. Year fixed is a
conditions, we found that the carbon market has significantly reduced time fixed effect and province fixed is an individual fixed effect. Control refers to
the control variables.
energy intensity and improved carbon emission efficiency, indicating
that the carbon market is conducive to the improvement of energy and
carbon emission performance. Furthermore, the series of robustness the uneven distribution of regional carbon markets to play a role in
tests that we implemented also supports this result. We also find that China’s overall carbon market. To achieve carbon peak and carbon
technological progress and market-oriented reform channels have a neutral goals, it is necessary to deal with the imbalance of regional en­
beneficial intermediary effect. Therefore, the conclusions of this study ergy conservation and emission reduction. Similar to some regions, the
provide strong policy recommendations. carbon markets in Beijing and Shanghai have a significant effect on
First, this study found that China’s overall carbon emission efficiency carbon emission efficiency, whereas the effect in other regions is not
is generally low, and improving carbon emission efficiency is the main obvious, and the implementation of carbon market policies also needs to
task of the current government. Therefore, it is necessary to support the consider the economic development of the region, such as appropriately
guiding role of the government, establish a long-term mechanism for adjusting carbon prices to meet the needs of the region. With the
regional scientific and technological cooperation, and strengthen the improvement of carbon emission efficiency, the government needs to
relationship between regions for high-tech development. Establishing an solve the difference in emission efficiency caused by the imbalance of
effective regional interactive development mechanism to avoid the regional development and finally realize the energy-saving and
promotion of homogeneous high-tech development in all regions emission-reduction effect of the national carbon market.
regardless of function and coordination, leads to differences in innova­ This study has obtained significant results on the carbon market
tion capabilities. Reasonably guiding the flow of innovative technolo­ based on a novel perspective; however, there are still some limitations.
gies to underdeveloped regions, jointly building and sharing several First, three input and two output indicators were used to select the input
regional cooperative innovation platforms at the industrial and univer­ and output variables. However, there may be other indicators that
sity research levels and high-tech industry innovation alliances, and require further exploration. Second, the time of data used in this study
relying on the collaborative industrial innovation chain of technological was short. In future research, other variables should be included to
research and development and technological transformation. measure carbon emission efficiency, and further studies with longer
Second, the conclusions of this study found that the technological periods should be selected for empirical analysis.
innovation mechanism is an intermediate channel for promoting the
improvement of China’s energy intensity and carbon emission perfor­ Author statement
mance, indicating that the government should increase scientific
research expenditure and introduce the latest technology to stimulate Anhua Zhou: Manuscript writing, data processing, visualization,
the development of technological innovation in China. Technological financial support, proofreading articles, revision opinions. Ling Xin:
progress is an important part of the carbon market construction mech­ Financial support, theory construction, statistical graph drawing. Jun Li:
anism. Therefore, it is necessary to increase support for emission Project administration, supervision, and investigation.
reduction R&D technologies and encourage green bias in R&D invest­
ment. Moreover, this study found that market reforms are a mechanism Declaration of competing interest
for reducing energy intensity. The government should further improve
regional and national carbon market systems, strengthen the role of The authors declare that they have no known competing financial
market-oriented reforms in reducing energy dependence on the carbon interests or personal relationships that could have appeared to influence
market, and gradually establish market reforms. It is necessary to the work reported in this paper.
correctly handle the role positioning of the government and the market,
scientifically and rationally formulate the role of market-oriented re­ Data availability
form elements in the carbon market, and maximize the energy-saving
effect of market-oriented reforms on the carbon market. Data will be made available on request.
Finally, this study found that there are significant regional differ­
ences in the promotion of energy intensity and carbon emission effi­
ciency in China’s carbon market. Therefore, it is necessary to improve

10
A. Zhou et al. Energy 258 (2022) 124789

Acknowledgements of Hunan Province (21B0583, 19YBA115, 2022JJ40137). We are

grateful for the anonymous review. The opinions of the articles were
This research was funded by the Scientific Research Project of the only our own.
Education Department，Social Science and Natural Science Foundation

Appendix A
Table A
Descriptive statistics of variables in treatment group and control group

Variables Treatment group Control group

Obs Mean Std.Dev. Obs Mean Std.Dev.

Energy performance (ei) 108.000 0.834 0.382 432.000 1.362 0.810

Carbon emission efficiency (cee) 108.000 0.803 0.201 432.000 0.580 0.191
Carbon market (did) 108.000 0.333 0.474 432.000 0.000 0.000
Economic growth (gdp) 108.000 17,101.230 17,347.780 432.000 11,788.420 12,942.690
Economic growth squared (gdp2) 108.000 5.910e+08 1.330e+09 432.000 3.060e+08 7.520e+08
Industrialization (ind) 108.000 0.425 0.094 432.000 0.462 0.074
Population (pop) 108.000 3949.605 3077.040 432.000 4503.082 2522.080
Environmental regulations (er) 108.000 0.001 0.001 432.000 0.003 0.003
Energy structure (esc) 108.000 0.667 0.204 432.000 1.032 0.357
Foreign direct investment (fdi) 108.000 858,928.000 711,031.400 432.000 430,377.900 621,349.000

References [16] Gao Y, Li M, Xue J, Liu Y. Evaluation of effectiveness of China’s carbon emissions
trading scheme in carbon mitigation. Energy Econ 2020;90:104872. https://doi.
org/10.1016/j.eneco.2020.104872.
[1] IPCC. Climate change. synthesis report. In: Pachauri RK, Meyer LA, editors.
[17] Zhang H, Maosheng D, Peng Z. Analysis of the impact of China’s emissions trading
Proceedings of the contribution of working groups I, II and III to the fifth
scheme on reducing carbon emissions. Energy Proc 2019;158:3596–601. https://
assessment report of the intergovernmental panel on climate change. Geneva: Core
doi.org/10.1016/j.egypro.2019.01.905.
Writing Team; 2014. p. 151. IPCC, https://www.globalchange.gov/browse/rep
[18] Akalpler E, Hove S. Carbon emissions, energy use, real GDP per capita and trade
orts/ipcc-climate-change-2014-synthesis-report.
matrix in the Indian economy-an ARDL approach. Energy 2019;168:1081–93.
[2] Narassimhan E, Gallagher KS, Koester S, Alejo JR. Carbon pricing in practice: a
https://doi.org/10.1016/j.energy.2018.12.012.
review of existing emissions trading systems. Clim Pol 2018;18(8):967–91. https://
[19] Zhang Y-J, Peng Y-L, Ma C-Q, Shen B. Can environmental innovation facilitate
doi.org/10.1080/14693062.2018.1467827.
carbon emissions reduction? Evidence from China. Energy Pol 2017;100:18–28.
[3] Yang X, Jiang P, Pan Y. Does China’s carbon emission trading policy have an
https://doi.org/10.1016/j.enpol.2016.10.005.
employment double dividend and a Porter effect? Energy Pol 2020;142:111492.
[20] Lin B, Jia Z. What will China’s carbon emission trading market affect with only
https://doi.org/10.1016/j.enpol.2020.111492.
electricity sector involvement? A CGE based study. Energy Econ 2019;78:301–11.
[4] Zhang Y-J, Wang W. How does China’s carbon emissions trading (CET) policy
https://doi.org/10.1016/j.eneco.2018.11.030.
affect the investment of CET-covered enterprises? Energy Econ 2021;98:105224.
[21] Wang S, Zhao D, Chen H. Government corruption, resource misallocation, and
https://doi.org/10.1016/j.eneco.2021.105224.
ecological efficiency. Energy Econ 2020;85:104573. https://doi.org/10.1016/j.
[5] Hu Y, Ding Y. Can carbon emission permit trade mechanism bring both business
eneco.2019.104573.
benefits and green efficiency? China Popul Resour Environ 2020;30(3):56–64. in
[22] Xing Z, Wang J, Zhang J. Total-factor ecological efficiency and productivity in
Chinese.
Yangtze River Economic Belt, China: a non-parametric distance function approach.
[6] Yang L, Li Y, Liu H. Did carbon trade improve green production performance?
J Clean Prod 2018;200:844–57. https://doi.org/10.1016/j.jclepro.2018.08.015.
Evidence from China. Energy Econ 2021;96:105185. https://doi.org/10.1016/j.
[23] Carneiro MLNM, Gomes MSP. Energy-ecologic efficiency of waste-to-energy plants.
eneco.2021.105185.
Energy Convers Manag 2019;195:1359–70. https://doi.org/10.1016/j.
[7] Yu Y, Zhang W, Zhang N. The potential gains from carbon emissions trading in
enconman.2019.05.098.
China’s industrial sectors. Comput Econ 2018;52. https://doi.org/10.1007/
[24] Liao Dong XK, Weng M. Economic effect of market-oriented environmental
s10614-017-9724-2.
regulation: carbon emission trading, green innovation and green economic growth.
[8] Chen S, Shi A, Wang X. Carbon emission curbing effects and influencing
China Soft Sci 2020;(6):159–73. in Chinese.
mechanisms of China’s Emission Trading Scheme: the mediating roles of technique
[25] Goeree J, Holt C, Palmer K, Shobe W, Burtraw D. An experimental study of auctions
effect, composition effect and allocation effect. J Clean Prod 2020;264:121700.
versus grandfathering to assign pollution permits. J Eur Econ Assoc 2010;8:
https://doi.org/10.1016/j.jclepro.2020.121700.
514–25. https://doi.org/10.2139/ssrn.1480074.
[9] Xuan D, Ma X, Shang Y. Can China’s policy of carbon emission trading promote
[26] Wakiyama T, Zusman E. The impact of electricity market reform and subnational
carbon emission reduction? J Clean Prod 2020;270:122383. https://doi.org/
climate policy on carbon dioxide emissions across the United States: a path
10.1016/j.jclepro.2020.122383.
analysis. Renew Sustain Energy Rev 2021;149:111337. https://doi.org/10.1016/j.
[10] Zhou C, Qin Y. The impact of a carbon trading pilot policy on the low-carbon
rser.2021.111337.
economic transformation in China: an empirical analysis based on a DID model.
[27] Zhou A, Li J. Investigate the impact of market reforms on the improvement of
Soft Sci 2020;34(10):36–42+55. in Chinese.
manufacturing energy efficiency under China’s provincial-level data. Energy 2021;
[11] Liu Z, Sun H. Assessing the impact of emissions trading scheme on low-carbon
228:120562. https://doi.org/10.1016/j.energy.2021.120562.
technological innovation: evidence from China. Environ Impact Assess Rev 2021;
[28] Moshiri S, Martinez Santillan MA. The welfare effects of energy price changes due
89:106589. https://doi.org/10.1016/j.eiar.2021.106589.
to energy market reform in Mexico. Energy Pol 2018;113:663–72. https://doi.org/
[12] Chen Z, Zhang X, Chen F. Do carbon emission trading schemes stimulate green
10.1016/j.enpol.2017.11.035.
innovation in enterprises? Evidence from China. Technol Forecast Soc Change
[29] Albrizio S, Kozluk T, Zipperer V. Environmental policies and productivity growth:
2021;168:120744. https://doi.org/10.1016/j.techfore.2021.120744.
evidence across industries and firms. J Environ Econ Manag 2017;81:209–26.
[13] Zhu B, Zhang M, Huang L, Wang P, Su B, Wei Y-M. Exploring the effect of carbon
https://doi.org/10.1016/j.jeem.2016.06.002.
trading mechanism on China’s green development efficiency: a novel integrated
[30] Ren S, Li X, Yuan B, Li D, Chen X. The effects of three types of environmental
approach. Energy Econ 2020;85:104601. https://doi.org/10.1016/j.
regulation on eco-efficiency: a cross-region analysis in China. J Clean Prod 2018;
eneco.2019.104601.
173:245–55. https://doi.org/10.1016/j.jclepro.2016.08.113.
[14] Zhou X, Gao Y, Wang P, Zhu B, Wu Z. Does herding behavior exist in China’s
[31] Calel R, Dechezleprêtre A. Environmental policy and directed technological
carbon markets? Appl Energy 2022;308:118313. https://doi.org/10.1016/j.
change: evidence from the European carbon market. Rev Econ Stat 2016;98(1):
apenergy.2021.118313.
173–91. https://doi.org/10.2139/ssrn.2041147.
[15] Zhu B, Tang J, Wang P. Examining the risk of China’s pilot carbon markets: a novel
[32] Li K, Lin B. The nonlinear impacts of industrial structure on China’s energy
integrated approach. J Clean Prod 2021;328:129408. https://doi.org/10.1016/j.
intensity. Energy 2014;69:258–65. https://doi.org/10.1016/j.energy.2014.02.106.
jclepro.2021.129408.
[33] Feng G, , B; ZM, Wen J ea. What determines the Chinese firms’ technological
onnovation: a re-empirical investigation based on the previous empirical literature

11
A. Zhou et al. Energy 258 (2022) 124789

of nine Chinese economics top journals and a-share listed company data. China [52] Zhu L, Gan QM, Liu Y, Yan ZJ. The impact of foreign direct investment on SO2
Industri Econ 2021;(1):17–35. in Chinese. emissions in the Beijing-Tianjin-Hebei region: a spatial econometric analysis.
[34] Miao C, Fang D, Sun L, Luo Q. Natural resources utilization efficiency under the J Clean Prod 2017;166:189–96. https://doi.org/10.1016/j.jclepro.2017.08.032.
influence of green technological innovation. Resour Conserv Recycl 2017;126: [53] Shahbaz M, Nasreen S, Abbas F, Anis O. Does foreign direct investment impede
153–61. https://doi.org/10.1016/j.resconrec.2017.07.019. environmental quality in high-, middle-, and low-income countries? Energy Econ
[35] Honma S, Hu J-L. Industry-level total-factor energy efficiency in developed 2015;51:275–87. https://doi.org/10.1016/j.eneco.2015.06.014.
countries: a Japan-centered analysis. Appl Energy 2014;119:67–78. https://doi. [54] Wang DT, Chen WY. Foreign direct investment, institutional development, and
org/10.1016/j.apenergy.2013.12.049. environmental externalities: evidence from China. J Environ Manag 2014;135:
[36] Borozan D. Technical and total factor energy efficiency of European regions: a two- 81–90. https://doi.org/10.1016/j.jenvman.2014.01.013.
stage approach. Energy 2018;152:521–32. https://doi.org/10.1016/j. [55] Cole MA, Elliott RJR, Zhang J. Growth, foreign direct investment and the
energy.2018.03.159. environment: evidence from Chinese cities. J Reg Sci 2011;51(1):121–38. https://
[37] Song M, Wang S, Sun J. Environmental regulations, staff quality, green technology, doi.org/10.1111/j.1467-9787.2010.00674.x.
R&D efficiency, and profit in manufacturing. Technol Forecast Soc Change 2018; [56] He J. Pollution haven hypothesis and environmental impacts of foreign direct
133:1–14. https://doi.org/10.1016/j.techfore.2018.04.020. investment: the case of industrial emission of sulfur dioxide (SO2) in Chinese
[38] Charnes A, Cooper WW, Rhodes E. Measuring the efficiency of decision making provinces. Ecol Econ 2006;60(1):228–45. https://doi.org/10.1016/j.
units. Eur J Oper Res 1978;2(6):429–44. https://doi.org/10.1016/0377-2217(78) ecolecon.2005.12.008.
90138-8. [57] Hao Y, Guo Y, Guo Y, Wu H, Ren S. Does outward foreign direct investment (OFDI)
[39] Li J, Lin B. Ecological total-factor energy efficiency of China’s heavy and light affect the home country’s environmental quality? The case of China. Struct Change
industries: which performs better? Renew Sustain Energy Rev 2017;72:83–94. Econ Dynam 2020;52:109–19. https://doi.org/10.1016/j.strueco.2019.08.012.
https://doi.org/10.1016/j.rser.2017.01.044. [58] Xie Q, Sun Q. Assessing the impact of FDI on PM2.5 concentrations: a nonlinear
[40] Wang Y, Sun X, Guo X. Environmental regulation and green productivity growth: panel data analysis for emerging economies. Environ Impact Assess Rev 2020;80:
empirical evidence on the Porter Hypothesis from OECD industrial sectors. Energy 106314. https://doi.org/10.1016/j.eiar.2019.106314.
Pol 2019;132:611–9. https://doi.org/10.1016/j.enpol.2019.06.016. [59] Demena BA, Afesorgbor SK. The effect of FDI on environmental emissions:
[41] Tone K. A slacks-based measure of efficiency in data envelopment analysis. Eur J evidence from a meta-analysis. Energy Pol 2020;138:111192. https://doi.org/
Oper Res 2001;130(3):498–509. https://doi.org/10.1016/S0377-2217(99)00407- 10.1016/j.enpol.2019.111192.
5. [60] Jiang L, Zhou H-f, Bai L, Zhou P. Does foreign direct investment drive
[42] Tone K, Tsutsui M. An epsilon-based measure of efficiency in DEA – a third pole of environmental degradation in China? An empirical study based on air quality index
technical efficiency. Eur J Oper Res 2010;207(3):1554–63. https://doi.org/ from a spatial perspective. J Clean Prod 2018;176:864–72. https://doi.org/
10.1016/j.ejor.2010.07.014. 10.1016/j.jclepro.2017.12.048.
[43] Wu P, Wang Y, Chiu Y-h, Li Y, Lin T-Y. Production efficiency and geographical [61] Huang Y. Environmental risks and opportunities for countries along the Belt and
location of Chinese coal enterprises - undesirable EBM DEA. Resour Pol 2019;64: Road: location choice of China’s investment. J Clean Prod 2019;211:14–26.
101527. https://doi.org/10.1016/j.resourpol.2019.101527. https://doi.org/10.1016/j.jclepro.2018.11.093.
[44] Charnes A, Cooper WW. Programming with linear fractional functionals. Nav Res [62] Wang S, Li C, Zhou H. Impact of China’s economic growth and energy consumption
Logist Q 1962;9(3-4):181–6. https://doi.org/10.1002/nav.3800090303. structure on atmospheric pollutants: based on a panel threshold model. J Clean
[45] Andersen P, Petersen NC. A procedure for ranking efficient units in data Prod 2019;236:117694. https://doi.org/10.1016/j.jclepro.2019.117694.
envelopment analysis. Manag Sci 1993;39(10):1261–4. https://doi.org/10.1287/ [63] Caputo MR. Comparative statics of a monopolistic firm facing price-cap and
mnsc.39.10.1261. command-and-control environmental regulations. Energy Econ 2014;46:464–71.
[46] Jefferson G, Tanaka S, Yin W. Environmental regulation and industrial https://doi.org/10.1016/j.eneco.2014.05.013.
performance: evidence from unexpected externalities in China. SSRN Electron J [64] Dong F, Wang Y, Su B, Hua YF, Zhang YQ. The process of peak CO2 emissions in
2013:20–34. https://doi.org/10.2139/ssrn.2216220. developed economies: a perspective of industrialization and urbanization. Resour
[47] Wang Z-H, Zeng H-L, Wei Y-M, Zhang Y-X. Regional total factor energy efficiency: Conserv Recycl 2019;141:61–75.
an empirical analysis of industrial sector in China. Appl Energy 2012;97:115–23. [65] Liu X, Bae J. Urbanization and industrialization impact of CO2 emissions in China.
https://doi.org/10.1016/j.apenergy.2011.12.071. J Clean Prod 2018;172:178–86. https://doi.org/10.1016/j.jclepro.2017.10.156.
[48] IPCC., IPCC. Guidelines for national greenhouse gas inventories. Institute for [66] National NBS. Bureau of Statistics (NBS) of China (1998-2017) China statistical
Global Environmental Strategies (IGES). Institute for Global Environmental yearbook. Beijing,China: National Bureau of Statistics of China; 2018. https://m.
Strategies (IGES); 2006. https://www.ipcc.ch/report/2006-ipcc-guidelines-for-na ca.investing.com/economic-calendar/nbs-press-conference-1752.
tional-greenhouse-gas-inventories/. [67] MEEC. China. Environmental quality statistical bulletin. In: Ministry of Ecology
[49] Zhang W, Li J, Li G, Guo S. Emission reduction effect and carbon market efficiency and environment of the people’s Republic of China. ed. 2018 Beijing http://www.
of carbon emissions trading policy in China. Energy 2020;196:117117. https://doi. mee.gov.cn/.
org/10.1016/j.energy.2020.117117. [68] Guo X, Xiao B, Song L. Emission reduction and energy-intensity enhancement: the
[50] Zhou A, Li J. Impact of income inequality and environmental regulation on expected and unexpected consequences of China’s coal consumption constraint
environmental quality: evidence from China. J Clean Prod 2020;274:123008. policy. J Clean Prod 2020;271:122691. https://doi.org/10.1016/j.
https://doi.org/10.1016/j.jclepro.2020.123008. jclepro.2020.122691.
[51] Shahbaz M, Balsalobre-Lorente D, Sinha A. Foreign direct Investment-CO2 [69] Du M, Liao L, Wang B, Chen Z. Evaluating the effectiveness of the water-saving
emissions nexus in Middle East and North African countries: importance of biomass society construction in China: a quasi-natural experiment. J Environ Manag 2021;
energy consumption. J Clean Prod 2019;217:603–14. https://doi.org/10.1016/j. 277:111394. https://doi.org/10.1016/j.jenvman.2020.111394.
jclepro.2019.01.282. [70] Wang W, Wang D, Ni W, Zhang C. The impact of carbon emissions trading on the
directed technical change in China. J Clean Prod 2020;272:122891. https://doi.
org/10.1016/j.jclepro.2020.122891.

12