Energy Economics 108 (2022) 105931

Contents lists available at ScienceDirect

Energy Economics
journal homepage: www.elsevier.com/locate/eneeco

Do carbon ETS pilots improve cities' green total factor productivity?

Evidence from a quasi-natural experiment in China
Changsheng Li a, b, Yaping Qi b, Shaohui Liu a, *, Xu Wang c
a
Institute of Climate Change and Energy Sustainable Development, Qingdao University of Science and Technology, Qingdao 266061, China
b
School of Economics and Management, Qingdao University of Science and Technology, Qingdao 266061, China
c
School of Economics and Management, China University of Mining and Technology, Xuzhou 221116, China

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

Keywords: Improving green total factor productivity (GTFP) is required for China's economic catch-up towards the global
China's carbon ETS pilots frontier. As a vital market-driven environmental regulation instrument, China's carbon emissions trading scheme
Green total factor productivity (ETS) plays a great role in achieving this task. In this paper, we constructed a Difference-in-Difference (DID)
Heterogeneity
model to examine the impact of China's ETS pilot policy on city-level GTFP. The results show that the GTFP in
Spatial spillover effect
ETS pilot cities is increased by 11.4% on average compared with non-pilot cities. However, the positive impact
exists only in a short term and lasts five years since policy implementation. Heterogeneity analysis shows that,
compared with non-industry-dominated cities, the GTFP in industry-dominated cities has been greatly promoted
due to the ETS pilots. In addition, China's ETS pilots produce significant spatial spillover effects on neighboring
non-pilot cities. Finally, several recommendations concerning China's ETS policy and GTFP improvement are put
forward.

1. Introduction Carlén et al., 2019; Gao et al., 2020) and a national climate policy
overlapping with EU ETS was prone to reduce the total emission (Jaraite
China has witnessed a remarkable growth of the economy in the past and Maria, 2016; Borenstein et al., 2019; Herweg, 2020). Besides, EU
few decades. However, the current average productivity is only at about ETS has reduced total consumption of natural gas and petroleum prod­
one-third of the global frontier (Asia and Pacific, 2021). Given the ucts in manufacture sector (Colmer et al., 2020). As a key pathway for
increasing interest in further improving its domestic environmental China pursuing global warming below 1.5 ◦ C (Duan et al., 2021), the
quality and keeping global warming below 1.5 ◦ C, China has to focus synergy effects of China's ETS pilots have also attracted much attention.
more on green development on the catch-up course towards the “fron­ The majority of existing literature found that China's ETS pilot policy
tier”. The carbon emission trading scheme (ETS), regarded as a vital significantly air pollutant emissions and mitigation cost (Aunan et al.,
market-driven carbon mitigation instrument, could trigger technology 2004; Dong et al., 2015) or their concentration level(Yan et al., 2020).
innovation and accelerate green economy transition (Rogge et al., 2011; The total emission and emission intensity can be decreased by 16.7%
Liu et al., 2020). It can help narrow China's productivity gaps from the and 9.7% respectively through switching to green fuels (Cui et al.,
global frontier by improving its green total factor productivity (GTFP). 2021), and the cumulative emission will be reduced by 22.73 billion ton
Early in 2013, China has already launched its seven regional ETS pilots. until 2050 (Mo et al., 2021). Whereas the carbon emissions are trans­
However, they are far from perfect in practice and might impair their ferred from the pilot provinces and cities to other non-pilot provinces
effectiveness (Zhu et al., 2020). This provides us an incentive to explore and cities, resulting in carbon leakage (Aichele and Felbermayr, 2015;
the impacts of China's pilot carbon ETS on GTFP. Böhringer et al., 2017; Gao et al., 2020).
Due to the increasingly significant role in climate policy and ever- Another branch of research is about its economic impacts. Free
expanding global coverage, a growing number of studies have allocation of emission allowances could bring higher cash flows and
explored the mitigation effects of ETS on carbon emissions mitigation form a carbon premium of EU ETS (Oestreich and Tsiakas, 2015). The
and its synergy effects or co-benefits. A bunch of articles found that EU electricity sector could benefit from emission trading due to the increase
ETS could bring emission reduction (Martin et al., 2014; Schaefer, 2018; of the electricity price (Bode, 2006; Veith et al., 2009). Springer et al.

* Corresponding author.
E-mail address: [email protected] (S. Liu).

https://doi.org/10.1016/j.eneco.2022.105931
Received 2 August 2021; Received in revised form 16 January 2022; Accepted 22 February 2022
Available online 3 March 2022
0140-9883/© 2022 Elsevier B.V. All rights reserved.
C. Li et al. Energy Economics 108 (2022) 105931

(2019) found that the China ETS could reduce firms' mitigation costs and enterprises. In June 2021, the cumulative volume of all carbon trading
accelerate the process of economic transition. In addition, Nong et al. pilots reached 480 million tons of carbon dioxide equivalent, with a
(2020) explored the impact of Vietnam ETS on the economy and found turnover of about 11.4 billion Yuan.1
that restricting the number of industrial sectors in the emissions trading In mid-2021, a national ETS only covering power sector entered into
market significantly decreased real GDP by 4.57%. operation after three years of preparation since the political launch. It
In recent years, the impacts of ETS on technology innovation and courrently covers more than four billion tCO2 accounting for about 40%
productivity have also been paid much attention. Jaraite and Maria of national carbon emissions. However, according to the latest rules, the
(2012) and Zhu et al. (2020) demonstrated that ETS led to an seven official ETS pilots are allowed to continually coexist alongside the
improvement in environmental efficiency and a shift in technological national ETS as long as they do not overlap. Furthermore, provincial
frontiers in the public power generating sector, and most pilot regions authorities are free to expand pilot systems to below-threshold firms and
have great potentials to improve their green production performance firms in uncovered sectors as a means of supporting the national ETS.
(Yang et al., 2021). Zhu et al. (2019) and Qi et al. (2021) found that This implies that the pilot ETSs will still plays an important role in the
China's pilots increase low-carbon innovation of regulated firms by achievement of provincial CO2 mitigation targets.
5–10% without crowding out other innovations, and the effect is mainly
driven by mass-based allowance allocation. However, Lundgren et al. 2.2. Theoretical framework
(2015) found that EU ETS had a modest influence on the technological
development in the pulp and paper industry, and the effect on firms' GTFP is effected by many factors, such as foreign direct investment,
total factor productivity was negative significantly (Commins et al., environmental regulation, industrial cluster, and among others. As a
2011). cost-effective alternative to command-and-control system, it has become
Based on the above literature, we find that the impact evaluation of one of the most popular environmental regulation instruments in the
China's regional pilot ETS has been extensively investigated from world. The theoretical analysis argues that environmental regulation
various aspects. However, some shortcomings still exist. First, the can stimulate innovative activities to offset compliance costs and pro­
impact of carbon trading pilots on regional GTFP has not received much mote average productivity in the long term (Porter and Claas, 1995).
attention, especially from the perspective of the city. In China, cities Emission constraints in the ETS pilots could trigger regulated parities to
contribute to 85% and 75% of the total CO2 emissions and energy adopt low-carbon, energy-saving and emission-reducing technologies
consumption, respectively (Cheng et al., 2019). Second, the spatial and production equipments. These advances could not only help miti­
spillover effects among different regions have been ignored in most gate carbon emissions, but also to reduce pollutant emissions, and
literature. Third, the heterogeneous effects in pilots with different consequently increase industrial and regional emission performance. On
schemes have not been fully investigated. To fill the above gaps, this the othe hand, compared with non-pilot regions, the pilot regions are
paper constructs a DID model to examine the impacts of China's ETS on more friendly to low-energy-intensive and new, high-growth industries.
cities' GTFP, and the heterogeneous and spatial spillover effects are They are believed to be the driving force of economic growth and pro­
further investigated. ductivity and accelerate decoupling carbon emissions from the tradi­
Our paper differs from the previous studies in three ways. First, in the tional economic development, which also leads to regional GTFP
slacks-based measure (SBM) framework, energy input is calculated from inprovement.
the total energy consumption rather than electricity consumption or The introduction of ETS add extra costs for fossil energy use due to
nighttime light to get a better measurement of GTFP at the city level. restricted emissions. Manufacturers, especially energy-intensive indus­
Second, the average treatment effect of regional ETS pilots on GTFP is trial ones, will be guided to adjust their production input factors, such as
estimated by employing a multi-period DID method. The heterogeneous capital, energy, labor, and information. The optimization process will
effects are evaluated in terms of industry characteristics, which is lead to a higher green productivity. On the other hand, under the con­
conductive to identify the mechanism between ETS and regional GTFP strained emission, heavily polluting plants that cannot adjust their
improvement. Third, the spatial spillover effects of ETS pilots are esti­ production behaviors to bear the extral costs enforced by ETS have to
mated based on geographical and economic relations, which is helpful to relocate to unregulated areas or close. This will results in the relocation
better understand the overall effect of ETS. of production input factors across different sectors in pilot regions,
The remainder of this paper is organized as follows: Section 2 which benefits the GTFP improvement in pilot cities.
introdues the theoretical framework. Section 3 introduces the research Furthermore, the establishment of pilot ETS would generate the
methodology and data. Empirical results and discussions are reported in spatial spillover effects. Firstly, the efficient energy-saving and low-
Section 3. Section 4 conducts some further analyses. Conclusions and carbon technologies induced by pilot ETSs could spread to non-pilot
policy implications are presented in the last section. cities. Second, in the context of Chinese institutions, non-pilot city
governors have the strong momentum to learn the knowledge, ideas,
2. Theoretical framework and experience of carbon emission regulation from the pilot cities to
achieve a higher administrative evaluation. The above technology
2.1. China carbon emissions trading scheme (ETS) diffusion and peer learning behavior could produce positive spillover
effects and are contributed to the improvement of GTFP in non-pilot
As the cornerstone of China's climate change policies, China's carbon cities. However, if the cap in pilot regions is too stringent, energy-
emissions trading system has been seen a the gradual development over intensive and emission-intensive plants are forced to move their pro­
the past decade. Seven pilot ETSs were lunched in two provinces duction to non-pilot regions, leading to an increase in carbon emissions
(Guangdong and Hubei) and five cities (Beijing, Shanghai, Tianjin, in non-pilot regions. Obviously, carbon leakage is harmful to the
Shenzhen and Chongqing) in 2013–2014, which were initially seen as a improvement of GTFP in non-pilot cities. The theoretical framework of
cost-effective alternative to command-and-control approaches for pilot ETS and GTFP improvement is shown in Fig. 1.
addressing the implementation of national climate change mitigation
goals. Provincial or municipal authorities of each pilot region has
tailored policies to determine the coverage of regulated emissions, sec­
tors and firms, introduce different allowance allocation methods, and
enforcement mechanisms. Until the end of 2020, more than 20 in­
dustries are involved in China's seven carbon trading pilot programs,
such as power, cement, steel, and chemicals, with nearly 3000 1
The data are obtained from the website of China Economic Weekly.

2
C. Li et al. Energy Economics 108 (2022) 105931

ETS pilot Non-pilot

green technology and

green and low carbon
low carbon
technologies
technology innovation
Innovation diffusion Induced
innovation effect
effect
production equipment production equipment

polluting firms exit

Factor carbon leakage polluting firms and
reallocation activities entry
non-polluting firms
development
learning relevant skills and
knowledge

spillover effect
GTFP GTFP
improvement improvement

Fig. 1. Theoretical framework of pilot ETS and GTFP improvement.

3. Methodology and data

Table 1
Inputs and outputs for calculating GTFP.
3.1. Difference-in-difference model
Type Variable Detail
Difference-in-Differences (DID) method is often used in policy eval­ The number of employed people at the end of year in
Labor
uation in economics, public policy, and other fields (Abadie, 2005; each city.
Kj, t = (1 − δ)Kj, t− 1 + Ij, t, Kj, t, Ij, t, δ represent capital
Imbens and Wooldridge, 2008; Fredriksson and Oliveira, 2019; Zhu
Input Capital stock, depreciation rate and investment. Following
et al., 2020; Liu et al., 2021). It can control systematic differences be­ Shan (2008), the depreciation rate is 10.96%.
tween the treatment and non-treatment group and eliminate the bias Eit = EPit × GDPit, Eit, EPit present total energy
Energy
caused by omitted variables (Wan et al., 2019), and is conductive to consumption and energy intensity.
solve endogenous problems (Meyer, 1995). However, the start-up time Desirable
output
of seven carbon ETS pilots is not the same. With reference to Beck et al.
Real GDP Using 2007 as the base year.
(2010) and Wang (2013), a multi-period DID model is constructed as Undesirable
follows: output
Output
CO2 emission Industrial CO2 emissions
lnGTFPit = α0 + α1 didit + ωXit + vi + μt + εit (1) SO2 emission Industrial SO2 emissions
smoke dust
Industrial smoke dust emissions
where didit is a dummy variable, and if city i started the ETS pilot in year emssion
t, then didit equals to 1 since year t, and otherwise it equals to 0. vi and μt effluent emission Industrial effluent emissions

refer to the city and year fixed effects respectively. εit is the expression of
error term. labor endowment, industrial structure, population density, technology
GTFPit represents the green total factor productivity of city i in year t. progress, financial development, government size, infrastructure, and
With reference to Tone (2001), the city-level GTFP is calculated by no- openness degree. The details are shown in Table 2.
radial and no-angle slacks-based measure (SBM) model and Malmquist-
Luenberger (ML) index. The details are presented in Table 2. In this
paper, labor, capital and energy are used as input variables, and output 3.2. Data and variables
indicators include desirable output and undesirable output. All their
measurement is specified in Table 1. Especially, the energy input is Our study sample contains 284 prefecture-level cities located in the
calculated by the total energy consumption rather than electricity con­ whole country from 2008 to 2019, and the major data of this research
sumption or nighttime light to get more precise results. are derived from China City Statistical Yearbook (2008-2020) and China
In order to consider other influencing factors, with reference to Zhu Provincial Statistical Yearbook (2008-2020). When calculating city-
et al. (2020) and Shi and Li (2020), we also control some factors (Xit) level GTFP, the CO2 emission data from 2008 to 2017 come from
that may influence productivity, including economic development, research published by Chen et al. (2020), while the data from 2018 to

3
C. Li et al. Energy Economics 108 (2022) 105931

Table 2
Variables and explanation.
∑
6
lnGTFPit = β0 + βm Dmit + λXit + vi + μt + εit (2)
Variable Definition Data source − 4

Dependent variable The final results are presented in Table 4. We can find that all the
China Provincial and coefficients are insignificant before the policy introduction (m < 0). It
GTFP The city-level green total factor productivity City Statistical
indicates that the trend between the treatment group and control group
Yearbooks
is similar prior to 2013 and the parallel trend assumption is valid.
In addition, the marginal and dynamic impact of China's ETS is
Explanatory variable
didit Policy dummy variable. /
further captured. The results show that the coefficients are significantly
positive when the policy was introduced. The positive effect increases
over time and then reduces by 12.64% in 2017. However, the positive
Control variable
EL Economic development: per capita GDP. effect is no longer significant after 2018.
Industrial structure: the ratio of secondary
IS
industry value added to GDP. 4.2. The effect of carbon ETS on the GTFP
Labor endowment: the percentage of
LA
employment in the total population.
Openness degree: the percentage of actual Based on the DID model, the net effect of China's ETS pilot on the
FDI
foreign investment in GDP. city-level GTFP is evaluated and the empirical results are reported in
Financial development: the percentage of Table 5. Column (1) shows the estimation results when controlling for
FD loans of national banking system at year- China City Statistical
several factors, such as economic development. In Column (2), some
end in GDP. Yearbook
Population density: the permanent other influencing factors and city and year fixed effects are further
DE considered to capture more reliable results.
population per square kilometer.

SC
Technology: the science and technology In each of the regression, the coefficient of the interaction term is
expenditure. positive and statistically significant at the 10% and 5% critical level,
Government size: the share of local
GV respectively. The GTFP in ETS pilot will increase by 11.4% compared
budgetary expenditure in GDP.
Infrastructure: the ratio of road area to total with untreated cities. It can be inferred that China's pilot is undergoing a
IF
population. greener shift to reduce pollutant emissions while promoting economic
growth. When considering the common origin of fossil fuels, the syn­
ergistic effect of carbon mitigation and pollutant reduction induced also
2019 are estimated by using interpolation algorithms. The energy in­
contributes to GTFP improvement (Higgins and Higgins, 2005; Ohno
tensity is calculated by its decline rate, which is manually collected from et al., 2021). These findings are similar to the conclusions of Yang et al.
the Province Statistical Yearbook, City Statistical Yearbook, and Na­
tional Economic and Social Development Bulletin from 2008 to 2020.
Table 4
Some missing values are added by the total energy consumption in the
Parallel trend test.
industrial sector from the National Economic and Social Development
Bulletin and City Statistics Bureau Energy Statistics. Variable Coefficient Standard error

Table 3 summarizes the average levels of the main variables for the Dit− 4 0.003 0.033
overall samples and the classification of samples by treatment. The study Dit− 3 0.036 0.043
Dit− 2 0.032 0.050
involved 3408 samples, comprising 444 samples in the treatment group
Dit− 1 0.100 0.061
and 2964 samples in the control group. Dit0 0.147* 0.077
Dit1 0.162** 0.080
4. Results and discussions Dit2 0.168* 0.087
Dit3 0.182** 0.081
Dit4 0.159* 0.089
4.1. Parallel trend test Dit5 0.050 0.093
Dit6 0.060 0.103
In this paper, the DID model is constructed to evaluate the effects of Control Variable Y
China's carbon ETS on GTFP, but the DID designs rely on the assumption City FE Y
Year FE Y
that the time series of attributes in each group should exhibit a common
N 3306
set of changes (Fredriksson and Oliveira, 2019). To check this, referred R2 0.452
to Jacobson et al. (1992) and Li et al. (2016), we adopt the event-study
Note: *, **, and *** indicate statistical significance at the levels of 10%, 5%, and
approach to test the lags and leads of the ETS influence based on Eq. (2).
1%, respectively.

Table 3
Descriptive statistics of the variables.
variable The full samples Treatment group Control group

Obs. Mean Std. Obs. Mean Sd. Obs. Mean Std.

GTFP 3408 1.413 2.813 444 1.394 1.100 2964 1.415 2.986
lnEL 3408 13.53 37.41 444 13.15 8.631 2964 13.59 39.98
lnLA 3408 − 2.309 0.644 444 − 2.047 0.915 2964 − 2.348 0.583
lnIF 3408 2.184 0.774 444 2.175 0.811 2964 2.185 0.768
lnIS 3408 3.835 0.282 444 3.822 0.215 2964 3.836 0.290
lnDE 3408 5.742 0.924 444 6.268 0.699 2964 5.663 0.928
lnSC 3408 10.11 1.440 444 10.97 1.796 2964 9.985 1.332
pubper 3408 0.222 0.231 444 0.177 0.143 2964 0.229 0.241
FDI 3408 0.0029 0.0039 444 0.0033 0.003 2964 0.0029 0.004
FD 3408 1.047 0.912 444 0.902 0.610 2964 1.069 0.947

4
C. Li et al. Energy Economics 108 (2022) 105931

Table 5
Benchmark results on the influence of China's carbon ETS on GTFP.
(1) (2)

lngtfp lngtfp

did 0.098* 0.114**

(0.058) (0.054)
Control Variable N Y
City FE Y Y
Year FE Y Y
constant 0.013 0.628**
(0.173) (0.243)
Obs. 3408 3408
Number of city 284 284
R2 0.429 0.445

Note: standard error for coefficients are in parentheses; *, **, and *** indicate
statistical significance at the levels of 10%, 5%, and 1%, respectively.

(2021) and Zhu et al. (2020) at the macro level.

In addition, consistent with the researches of Du et al. (2019) and Li Fig. 2. Placebo test.
and Chen (2019), the GTFP is determined by many factors, such as in­ Note: The x-axis is the estimated coefficient of did for 1000 random samples.
dustrial structure, labor endowment, etc. More investments in advanced The curve is the kernel density distribution of the estimated coefficient, and the
energy-efficient and low-carbon technologies could contribute to pro­ dot is the corresponding P-value.
ductivity improvement, and some energy-intensive manufacturers with
outdated techniques will be phased out. The government regulation the estimated results.
could enhance productivity, and continuous changes in policy design are
needed to adjust in practice. 4.3.2. PSM-DID method
On the other hand, the DID method requires randomized assignment
of group status. Whereas the setting of ETS pilots may be self-selected in
4.3. Robustness test this paper, which will cause the estimates to be biased. The propensity
score matching (PSM) is a powerful procedure for removing the bias
4.3.1. Placebo test (Heckman et al., 1998). The logit model is constructed with the eco­
The placebo test is commonly adopted to eliminate the potential nomic development, labor endowment, industrial structure, population
influence of some unobservable characteristics. In this section, we density, technology progress, financial development, government size,
perform this test from two aspects: 1) We identify the effect by artifi­ infrastructure, and foreign direct investment as the matching variables.
cially changing the time point of ETS implementation. 2) An alternative The kernel matching method and nearest neighbor matching method are
placebo test artificially exposes ETS pilot, which is randomized in the used to match the sample.
untreated cities. As shown in Table A1, the balance test shows that there is no sig­
First, we advance ETS pilot policy and set a time point, and then we nificant difference between the treatment group and the control group
re-estimate the DID model. Column (1) of Table 6 reports the corre­ after matching, and the balanced data are used to re-estimate in the
sponding results. The coefficients of the interaction term (did) cannot regression. Columns (2) and (3) of Table 6 present the PSM-DID results.
pass the significance level test. This placebo test verifies the robustness As we can see, both of the coefficients of interaction term (did) are
of the benchmark results. significantly positive when different matching algorithms, including
Second, we create a placebo treatment where the treatment group is nearest neighbor matching and kernel matching are used. The results
randomly assigned in the control subclass and vice versa, and re- show that China's ETS has a robust and positive effect on regional GTFP.
estimate the main basic regression 1000 times. The results are shown
in Fig. 2. The mean value of the coefficient of DID estimator is close to 5. Further analysis
zero. Most of the corresponding p-values are larger than 0.1, which in­
dicates random correlation across explanatory variables cannot explain 5.1. Heterogeneous analysis

Table 6 The foregoing analysis has shown that China pilot ETSs have
Results of placebo test and PSM-DID method. significantly promoted their GTFP. However, there are obvious differ­
(1) (2) (3) ences in the design aspects among different ETS pilots, such as the target
industries of service sectors in Shenzhen and the power and heating
Set 2009 and 2010 as Nearest neighbor Kernel
policy year matching matching industries in Beijing, which means that the effects of ETS on regional
GTFP may be different.
lngtfp lngtfp lngtfp
To examine whether pilot ETS exerts equal influence on GTFP across
did 0.081 0.088* 0.100* cities, we devide the sample cities into two different subclasses:
(0.057) (0.051) (0.053)
industry-dominated cities and non-industry-dominated cities based on
Control
Variable
Y Y Y city-level industrial structure, and construct interaction terms to capture
City FE Y Y Y the different effects. Then we re-estimate the impact with the DID
Year FE Y Y Y model.
constant 0.609** 0. 842** 0.831**
The results of heterogeneous analysis are reported in Table 7. The
(0.235) (0.336) (0.333)
Obs. 3408 3288 3297
coefficient is positive at 1% level of significance in Column (1), whereas
R2 0.444 0.454 0.456 it does not pass the significance test in Column (2). The results show that
compared with non-industry-dominated cities, the GTFP in industry-
Note: standard error for coefficients are in parentheses; *, **, and *** indicate
dominated cities has been greatly promoted due to the pilot ETSs,
statistical significance at the levels of 10%, 5%, and 1%, respectively.

5
C. Li et al. Energy Economics 108 (2022) 105931

Table 7 is an NT × k matrix of control variables, ε is an error vector, W is an N ×

Results of heterogeneous analysis. N spatial weight matrix.
Variables (1) (2) Considering both geographical and economic relations between
different cities, in this paper, we mainly conduct economic-distance
lngtfp lngtfp
spatial weight matrix defined as follows.
did*citydum 0.132** 0.046
(0.060) (0.088) ⎧ 0, if i = j
⎪
⎨
Control Variable Y Y
City FE Y Y Wij = 1 pgdpi
⎪
⎩ d × ∑N , if i ∕
=j
Year FE Y Y ij pgdp
Constant 0.622*** 0.607** i=1
(0.240) (0.234)
N 3408 3408 where dij represents the geographic distance between city i and j, pgdp
R2 0.445 0.443 represents the city-level average per capita GDP.
Note: standard error for coefficients are in parentheses; *, **, and *** indicate The results of the spatial spillover effect are presented in Table 8. It is
statistical significance at the levels of 10%, 5%, and 1%, respectively. clear that the coefficient of treatment remain positive at 1% level of
significance. The results indicate that China ETS could still contribute to
which is consistent with our expectations. GTFP improvement while taking local spatial interaction into consid­
Our estimation results also suggest that the GTFP growth in pilot ETS eration. Besides, the coefficient of WNT, TDID is significantly positive as
regions is mainly promoted through the improved emissions perfor­ well. It indicates that there are positive spillover effects between pilot
mance in the industrial sector. Several major pollutants including CO2, cities and non-pilot cities, and the effect of the ETS in the treatment
SO2, effluent and dust have been reduced directly through the techno­ group can be propagated to neighboring untreated cities.
logical progress, and firms exist (Cao et al., 2021; Qi et al., 2021). Thus, The conclusions indicate that the pilot cities will make progress in
the emissions productivity is dramatically improved both in the energy- green technology, and positive spatial spillover of technology occurs
intensive and emission-intensive sectors and at the city level. On the from pilot cities to non-pliot cities. Then energy-intensive and emission-
contrary, the service sector is not energy-intensive and then faces less intensive firms will reduce their carbon emissions by employing
stringent regulation (Liu et al., 2021). In addition, under the stringent advanced technology from pilot cities. Besides, economic analysis of
regulation, polluting enterprises could shrink their production size, climate policy often overestimated the extent of carbon leakage due to
leading to production resources reallocation across industrial sector the relative price changes in trade caused by climate policies could affect
(Fullerton and Heutel, 2010; Burton et al., 2011). It can be manifested innovation incentives of corporations, thus it can help to reduce carbon
that China ETS facilitates GTFP mainly through the industrial sector, leakage (Maria and Werf, 2008), which will offset the negative spillover
and the green technical progress and resource reallocation. effect. Therefore, although carbon leakage may diminish the effective­
ness of the carbon ETS policy, the indirect learning and technology
diffusion can produce larger positive effects to offset the possible
5.2. The spatial spillover effect of carbon ETS on the GTFP

It is possible that the ETS effect of not only pertain to the pilot cities, Table 8
Results of spatial spillover effect.
but also to other non-pilot cities. Taking regional interrelations into
account, the influence of ETS pilot may be propagated to untreated cities Variable (1)
neighboring the treated cities. This is what the spatial spillover effect lngtfp
means. Intuitively, stricter low-carbon policy could induce carbon
DID 0.174***
leakage through the relocation of emission-intensive industries to re­ (0.044)
gions without emission regulation. The existence of carbon leakage will WT,TDID − 0.929
not only offset the reduced carbon emission in pilot policy cities, but also (0.663)
WNT,TDID 4.395**
diminish the effectiveness of carbon ETS policy. The spillover effect
(1.745)
demonstrates a negative tendency. WlnEL − 0.006
But on the other side, most local governments pay more attention to (0.008)
the pilots of China's ETS and take the initiative to learn from each other. WlnLA 0.332
Thus the policy experimentation on market-based emission trading (0.683)
WlnIF 0.772***
schemes will equip other cities with relevant skills and knowledge about
(0.183)
the management rules, trading procedures, allowance allocation WlnIS − 4.142***
method, etc. Especially, the technical advancement in the treatment (1.399)
cities will produce an indirect effect on both regions, treated and un­ WlnDE − 1.762
treated, which displays positive spatial autocorrelation. In this situation, (1.36)
WlnSC 1.009***
estimating α1 as an average effect might underestimate the real effect of (0.221)
ETS. Wpubper − 0.613
In this section, we further employ a spatial DID model referred to (2.595)
Chagas et al. (2016) to investigate the spatial spillover effect. The model WFDI 101
(89.036)
is specified as follows:
WFD − 0.253
(0.642)
G = ∅ + ϑ + ρWG + αDID + τ1 WT,T DID + τ2 WNT,T DID + Xγ ' + WXδ + ε
Control Variable Y
(6) City FE Y
Year FE Y
where G is an NT × 1 vector of cities' GTFP, ∅ is an NT × 1 vector of city Obs. 3408
fixed-effects, ϑ is an NT × 1 vector of time fixed effects, ρ is the spatial R2 0.075
autocorrelation coefficient. The parameter τ1 captures the direct effect Note: 1. standard error for coefficients are in paren­
of the treatment on the treated city, and τ2 captures the indirect effect of theses. 2. *, **, and *** indicate statistical significance
the treatment on untreated city, conditioned on the neighbor treated. X at the levels of 10%, 5%, and 1%, respectively.

6
C. Li et al. Energy Economics 108 (2022) 105931

negative effect on GTFP. pilot ETS to gain vauable experience before expansion the national ETS.
However, the coefficient of WT, TDID is not significant, and there is no In order to speed up the pilot process, it is suggested that the aimed
direct effect betweent pilot cities. The conclusion is the same as that of sectors should be allocated to seven pilot ETS simultaneously and
Feng et al. (2021), who found that the spillover effects among different reasonably.
cities are quite weak. The possible explanation may be the great dis­ Second, the overall contribution of the ETS to city-level GTFP,
parities between ETS pilot. For example, Guangdong uses the bench­ however, was still limited, suggesting the need for broader program
marking method whereas Beijing uses the historical intensity method. coverage to increase policy impacts, such as expansion to below-
On the other hand, governors of the non-pilot regions have strong in­ threshold firms. The benefits are thirdfold. Firstly, it helps incease the
centives to learn from pilot cities to take climate mitigation measures pilot ETS market liquidity. Secondly, it will gain precious experiences
through interaction and collaboration and gain better assessment from and lessons for national carbon ETS before its expansion of coverage of
higher-level supervisory authorities (Mi et al., 2019). entities in the covered sectors. Lastly, a larger coverage of ETS could
achieve better environmental effectiveness and economic efficiency and
6. Conclusions and policy implications increase the overall contribution of the pilot ETS to city-level GTFP.
Third, the spatial spillover effects among different cities play a great
The China carbon ETS pilot policy is a useful experiment to make use role in the GTFP improvement. On the one hand, local governments of
of market mechanisms to addressing its national climate change miti­ non-pilot regions are encouraged to actively learn good experiences
gation goals in a cost-effective manner. In this paper, the impact of from pilot regions to enhance spillover effects on productivity
China's ETS pilot policy on GTFP at the city level is assessed by improvement. However, the spatial spillover effects are undermined by
employing a multi-period DID model and the heterogeneous impact and carbon leakage. New mechanism should be considered to prevent firms
spatial spillover effect are further investigated, respectively. Based on from avoiding emissions rules. A hybrid mechanism integrating carbon
the above results, we reach the following main conclusions. ETS with other environmental regulation instruments is recommended.
First, China's ETS pilot policy exerts a significantly positive impact For example, a carbon levy on emissions emitted in non-pilot regions to
on city-level GTFP. Compared with non-pilot cities, the GTFP in pilot stop companies moving outside pilot regions.
cities has been promoted greatly. The positive impacts last five years
since ETS policy implementation. This conclusion implies that the pos­ CRediT authorship contribution statement
itive impacts only exist in the short term and the long-term effects are
still not clear. Second, the impacts of China's ETS pilot policy on city- Changsheng Li: Writing – original draft, Funding acquisition, Su­
level GTFP are heterogeneous. GTFP in industry-dominated cities pervision, Validation, Writing – review & editing. Yaping Qi: Concep­
covered by China pilot ETS has been greatly promoted compared with tualization, Data curation, Formal analysis, Methodology, Visualization,
non-industry dominated cities. Third, the impact of China's ETS pilot Writing – original draft. Shaohui Liu: Data curation, Formal analysis,
policy on GTFP exists obvious spatial spillover effects between the pilot Methodology, Supervision, Validation, Writing – review & editing. Xu
city and non-pilot city, which results from indirect policy leanring and Wang: Funding acquisition, Writing – review & editing.
innovative technology diffusion. However, the direct spatial spillover
effects don't exist between pilot cities. Acknowledgments
Based on the above conclusions, several policy implications are
further proposed. First, China's carbon trading pilot policy produces The authors would like to thank the National Natural Science
positive effects on city-level GTFP improvement. This conclusion further Foundation of China (Grant No. 71303126, 71603256), Shandong
confirms the necessity of the coexistence of the pilots alongside the Province College Youth Innovation Technology Support Program (Grant
national ETS. In order to support China's goals of reaching peak CO2 No.2019RWE003), Qingdao Philosophy and Social Science Planning and
emissions by 2030 and of achieving carbon neutrality by 2060, China Management Office, China (Grant No.QDSKL200272), Jiangsu Planned
plans to expand its national ETS to seven other sectors. However, the Projects for Postdoctoral Research Funds (No. 1601086C), the Funda­
expansion will increase complexity and therefore be more challenging. mental Research Funds for the Central Universities (No.
An important question concerns how fast, and in what order, to expand 2021YCPY0112), and Jiangsu Key Laboratory of Coal-based Greenhouse
the coverage of the ETS to other sectors. Considering the complexity of Gas Control and Utilization (No. 2020KF02).
China's carbon ETS, a safe way is to introduce the aimed sectors to the

Appendix A. Appendix A

A.1. Slacks-based measure (SBM) model

SBM model with undesirable output was introduced in this section. Suppose that there are n DMUs (decision-making units) each including three
factors: inputs, desirable output, undesirable outputs, as represented by three vectors x ∈ Rm, yg ∈ Rs1, yb ∈ Rs2, respectively. We define the matrices X,
Yg, Ybas follows:
X = [x1, x2, ⋯, xn] ∈ Rm×n, Yg = [y1g, y2g, ⋯, yng] ∈ Rs1×n, Yb = [y1b, y2b, ⋯, ynb] ∈ Rs2×n, and
, Yg, Yb > 0. The production possibility set P is defined as follows:
{( )⃒ }
P = x, yg , yb ⃒x ≥ Xλ, yg ≤ Y g λ, yb ≥ Y b λ, eλ = 1, λ ≥ 0

The SBM model considering undesirable output is shown as follows:

1
∑
m −
si
1− m xik
ρ* = min ( i=1
)
1
∑
s1 g+
sr ∑
s2 b−
st
1 + s1 +s 2
g
yrk
+ ybtk
r=1 t=1

7
C. Li et al. Energy Economics 108 (2022) 105931

⎧
⎪
⎪
⎪
⎪
⎪ Xλ + s− = xk
⎪
⎪ g
⎪ g g+
⎨ Y λ − s = yk
⎪
s.t. Y λ + s = ybk
b b−
⎪
⎪
⎪
⎪ eλ = 1
⎪
⎪
⎪
⎪
⎪ s−i , sg+
⎩
b−
r , st , λ ≥ 0

where ρ*is the efficiency value of the evaluated DMU, xk, ykg, ykbrepresent the input, desirable output, undesirable output, respectively. si− , srg+, stb−
refer to the slack vector of the input, desirable output, undesirable output, respectively. λis the weight vector.

A.2. Malmquist Luenberger (ML) index

The efficiency determined by the SBM-DEA model is analyzed from a static aspect. To reflect the dynamic change in GTFP, this paper adopts
Malmquist–Luenberger (MI) index including undesirable output. The Malmquist index was firstly proposed by Malmquist (1953) and then developed
by Caves et al. (1982) and Färe et al. (1994). However, the undesirable output is usually produced along with desirable output. To resolve the difficulty
caused by the joint production of desirable and undesirable output, Chung et al. (1997) proposed the Malmquist-Luenberger index considering un­
desirable output. The ML index from period t to t + 1 is shown as follows.
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√ ̅→ ̅̅→
√
t+1 √ 1 + Dt0 (xt , yt , bt ; yt , − bt ) 1 + Dt+1 t t t t
0 (x , y , b ; y , − b )
t
MLt = √ ̅→ ( ) × ̅̅→( )
t t+1 t t+1 t+1 t+1 t+1
1 + D0 x , y , b ; y , − b 1 + D0 xt+1 , yt , bt+1 ; yt+1 , − bt+1

̅→( ) ̅̅→( )
where x, y, and b represent the input, desirable output, and undesirable output, respectively. Dt0 xt , yt , bt ; yt , − bt and Dt+1
0 xt , yt , bt ; yt , − bt represent
̅→ ( )
the direct distance function at t period for the production of technology at t and t + 1, respectively. Dt0 xt+1 , yt , bt+1 ; yt+1 , − bt+1 and
̅̅→( )
Dt+1
0 xt+1 , yt , bt+1 ; yt+1 , − bt+1 represent the direct distance function at t + 1 period for the production of technology at t and t + 1, respectively.
MLtt+1is the GTFP of change index from period t to t + 1. MLtt+1 > 1means the GTFP is improved. MLtt+1 < 1means the GTFP is decreased. MLtt+1 =
1means the GTFP is unchanged.
As shown above, the ML index is the ratio of the productivity of period t to the productivity of period t + 1, reflecting the growth rate of GTFP.
Refers to Qiu et al. (2008), this paper assumed the GTFP in 2007 is 1, then the GTFP in 2008 is the GTFP in 2007 multiplied by the ML index in 2008. By
this method, we obtained the cities' GTFP from 2008 to 2019.

Appendix B. Appendix B
Table A1
Balance test of variables before and after PSM.

Nearest neighbor matching kernel matching

Reduction t-test Reduction t-test

Variable Bias(%) |bias|(%) t p > |t| Bias(%) |bias|(%) t p > |t|

lnEL U − 1.5 − 0.23 0.818 − 1.5 − 0.23 0.818

M − 0.1 93.3 − 0.05 0.961 1.1 28.0 0.18 0.599
lnLA U 39.2 9.28 0.000 39.2 9.28 0.000
M 1.4 96.5 − 0.21 0.835 5.8 85.1 0.20 0.378
lnIF U − 1.3 − 0.25 0.800 − 1.3 − 0.25 0.800
M − 6.0 − 377.1 − 0.83 0.405 4.0 − 216.5 − 0.03 0.587
lnIS U − 5.7 − 1.01 0.313 − 5.7 − 1.01 0.313
M − 0.1 98.7 − 0.01 0.991 5.9 − 4.7 0.92 0.330
lnDE U 73.6 13.19 0.000 73.6 13.19 0.000
M 3.1 95.8 0.52 0.601 6.6 91.0 0.41 0.270
lnSC U 62.6 13.8 0.000 62.6 13.88 0.000
M 5.1 91.9 − 0.84 0.403 8.2 86.9 0.31 0.207
pubper U − 26.5 − 4.48 0.000 − 26.5 − 4.48 0.000
M − 1.3 95.0 − 0.25 0.802 − 7.6 71.3 − 0.70 0.174
FDI U 12.8 2.30 0.022 12.8 2.30 0.022
M 6.4 50.2 1.08 0.281 2.0 84.0 0.23 0.738
FD U − 20.9 − 3.59 0.000 − 20.9 − 3.59 0.000
M 4.3 79.4 0.97 0.333 3.7 82.4 0.09 0.433

Appendix C. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.eneco.2022.105931.

8
C. Li et al. Energy Economics 108 (2022) 105931

References Jaraite, J., Maria, C.D., 2012. Efficiency, productivity and environmental policy: a case
study of power generation in the EU. Energy Econ. 34, 1557–1568.
Jaraite, J., Maria, C.D., 2016. Did the EU ETS make a difference? An empirical
Abadie, A., 2005. Semiparametric difference-in-differences estimators. Rev. Econ. Stud.
assessment using Lithuanian firm-level data. Energy J. 37, 1–23.
72, 1–19.
Li, P., Chen, Y., 2019. Environmental regulation, bargaining power of Enterprisesand
Aichele, R., Felbermayr, G., 2015. Kyoto and carbon leakage: an empirical analysis of the
green Total factor productivity. Fin. Trade Econ. 40, 144–160 (in Chinese).
carbon content of bilateral trade. Rev. Econ. Stat. 97 (1), 104–115.
Li, P., Lu, Y., Wang, J., 2016. Does flattening government improve economic
Asia, I., Pacific, D., 2021. People’s Republic of China: 2020 Article IV Consultation–Press
performance? Evidence from China. J. Dev. Econ. 123, 18–37.
Release, Staff Report, and Statement by the Executive Director for the People’s
Liu, C., Ma, C., Xie, R., 2020. Structural, innovation and efficiency effects of
Republic of China. In: IMF Staff Country Reports, 006, p. A001.
environmental regulation: evidence from China’s carbon emissions trading pilot.
Aunan, K., Fang, J., Vennemo, H., Oye, K., Seip, H.M., 2004. Co-benefits of climate
Environ. Resour. Econ. 75, 741–768.
policy—lessons learned from a study in Shanxi, China. Energ Policy 32, 567–581.
Liu, S.H., Liu, C.J., Yang, M., 2021. The effects of national environmental information
Beck, T., Levine, R., Levkov, A., 2010. Big bad banks? The winners and losers from bank
disclosure program on the upgradation of regional industrial structure: evidence
deregulation in the United States. J. Financ. 65, 1637–1667.
from 286 prefecture-level cities in China. Struct. Change Econ. D 58, 552–561.
Bode, S., 2006. Multi-period emissions trading in the electricity sector-winners and
Lundgren, T., Marklund, P., Samakovlis, E., Zhou, W., 2015. Carbon prices and incentives
losers. Energ Policy 34, 680–691.
for technological development. J. Environ. Manag. 150, 393–403.
Böhringer, C., Rosendahl, K.R., Storrøsten, H.B., 2017. Robust policies to mitigate carbon
Malmquist, S., 1953. Index numbers and indifference surfaces. Trabajos De Estadistica 4,
leakage. J. Public Econ. 149, 35–46.
209–242.
Borenstein, S., Bushnell, F.A.J., Wolak, M., 2019. Zaragoza-Watkins, expecting the
Maria, C.D., Werf, E.V.D., 2008. Carbon leakage revisited: unilateral climate policy with
unexpected: emissions uncertainty and environmental market design. Am. Econ.
directed technical change. Environ. Resour. Econ. 39, 55–74.
Rev. 109, 3953–3977.
Martin, R., Muûls, M., Preux, D.B.L., Wagner, U.J., 2014. Industry compensation under
Burton, D.M., Gomez, I.A., Love, H.A., 2011. Environmental regulation cost and industry
relocation risk: a firm-level analysis of the EU emissions trading scheme. Am. Econ.
structure changes. Land Econ. 87 (3), 545–557.
Rev. 104, 2482–2508.
Cao, J., Ho, M.S., Ma, R., Teng, F., 2021. When carbon emission trading meets a
Meyer, B.D., 1995. Natural and quasi- experiments in economics. J. Bus. Econ. Stat. 13,
regulated industry: evidence from the electricity sector of China. J. Public Econ. 200,
151–161.
104470.
Mi, Z., Guan, D., Liu, Z., Liu, J., Viguié, V., et al., 2019. Cities: the core of climate change
Carlén, B., Dahlqvist, A., Mandell, S., Marklund, P., 2019. EU ETS emissions under the
mitigation. J. Clean. Prod. 207, 582–589.
cancellation mechanism - effects of national measures. Energ Policy 129, 816–825.
Mo, J.L., Zhang, W.R., Tu, Q., Yuan, J.H., Duan, H.B., Fan, Y., Pan, J.F., Zhang, J.,
Caves, D.W., Christensen, L.R., Diewert, W.E., 1982. The economic theory of index
Meng, Z.X., 2021. The role of national carbon pricing in phasing out China’s coal
numbers and the measurement of input, output, and productivity. Econometrica 50,
power. iScience 24 (6), 102655.
1414–1939.
Nong, D., Nguyen, T.H., Wang, C., Van, Khuc Q., 2020. The environmental and economic
Chagas, A.L.S., Azzoni, C.R., Almeida, A.N., 2016. A spatial difference-in-differences
impact of the emissions trading scheme (ETS) in Vietnam. Energ Policy 140, 111362.
analysis of the impact of sugarcane production on respiratory diseases. Reg. Sci.
Oestreich, A.M., Tsiakas, I., 2015. Carbon emissions and stock returns: evidence from the
Urban Econ. 59, 24–36.
EU emissions trading scheme. J. Bank. Financ. 58, 294–308.
Chen, J., Gao, M., Cheng, S., et al., 2020. County-level CO2 emissions and sequestration
Ohno, H., Shigetomi, Y., Chapman, A., Fukushima, Y., 2021. Detailing the economy-wide
in China during 1997-2017. Sci. Data 7 (391), 1–20.
carbon emission reduction potential of post-consumer recycling. Resour. Conserv.
Cheng, J., Yi, J., Dai, S., Xiong, Y., 2019. Can low-carbon city construction facilitate
Recycl. 166.
green growth? Evidence from China’s pilot low-carbon city initiative. J. Clean. Prod.
Porter, M.E., Claas, V.D.L., 1995. Toward a new conception of the environment-
231, 1158–1170.
competitiveness relationship. J. Econ. Perspect. 9, 97–118.
Chung, Y.H., Färe, R., Grosskopf, S., 1997. Productivity and undesirable outputs: a
Qi, S.Z., Zhou, C.B., Li, K., Tang, S.Y., 2021. Influence of a pilot carbon trading policy on
directional distance function approach. J. Environ. Manag. 51, 229–240.
enterprises’ low-carbon innovation in China. Clim. Pol. 21 (3), 318–336.
Colmer, J., Martin, R., Muls, M., Wagner, U.J., 2020. Does pricing carbon mitigate
Qiu, B., Yang, S., Xin, P., 2008. Research on FDI technology spillover channels and
climate change? Firm-level evidence from the european union emissions trading
China's manufacturing productivity growth: an analysis based on panel data.
scheme. CEP Discussion Papers dp1728, Centre for Economic Performance, LSE.
J. World Econ. 20–31 (in Chinese).
Commins, N., Lyons, S., Schiffbauer, M., Tol, R.S.J., 2011. Climate policy and corporate
Rogge, K.S., Schneider, M., Hoffmann, V.H., 2011. The innovation impact of the EU
behaviour. Energy J. 32, 51–68.
emission trading system-findings of company case studies in the German power
Cui, J., Wang, C., Zhang, J., Zheng, Y., 2021. The effectiveness of China’s regional carbon
sector. Ecol. Econ. 70 (3), 513–523.
market pilots in reducing firm emissions. Proc. Natl. Acad. Sci. U. S. A. 118, 52.
Schaefer, S., 2018. Decoupling the EU ETS from Subsidized Renewables and Other
Dong, H., Dai, H., Dong, L., Fujita, T., Geng, Y., et al., 2015. Pursuing air pollutant co-
Demand Side Effects Lessons from the Impact of the EU ETS on CO2 Emissions in the
benefits of CO2 mitigation in China: a provincial leveled analysis. Appl. Energy 144,
German Electricity Sector.
165–174.
Shan, H., 2008. Reestimating the capital stock of China :1952-2006. J. Quant. Tech.
Du, L., Zhao, Y., Tao, K., Lin, W., 2019. Compound effects of environmental regulation
Econ. 25, 17–31 (in Chinese).
and governance transformation in enhancing green competitiveness. Econ. Res. 54,
Shi, D., Li, S., 2020. Emission trading system and energy use
106–120 (in Chinese).
efficiency——measurements and empirical evidence for cities at and above the
Duan, H., Zhou, S., Jiang, K., et al., 2021. Assessing China's efforts to pursue the 1.5◦ C
prefecture level. China Ind. Econ. 5–23 (in Chinese).
warming limit. Science. 372 (6540), 378–385.
Springer, C., Evans, S., Lin, J., Roland-Holst, D., 2019. Low carbon growth in China: the
Färe, R., Grosskopf, S., Norris, M., Zhang, Z., 1994. Productivity growth, technical
role of emissions trading in a transitioning economy. Appl. Energy 235, 1118–1125.
progress and efficiency change in industrial countries. Am. Econ. Rev. 84, 66–83.
Tone, K., 2001. A slacks-based measure of efficiency in data envelopment analysis. Eur.
Feng, Y., Wang, X., Liang, Z., Hu, S., Xie, Y., Wu, G., 2021. Effects of emission trading
J. Oper. Res. 130, 498–509.
system on green total factor productivity in China: empirical evidence from a quasi-
Veith, S., Werner, J.R., Zimmermann, J., 2009. Capital market response to emission
natural experiment. J. Clean. Prod. 294.
rights returns: evidence from the European power sector. Energy Econ. 31, 605–613.
Fredriksson, A., Oliveira, G.M.D., 2019. Impact evaluation using difference-in-
Wan, Z., Zhou, X., Zhang, Q., Chen, J., 2019. Do ship emission control areas in China
differences. Rausp Manag. J. 54, 519–532.
reduce sulfur dioxide concentrations in local air? A study on causal effect using the
Fullerton, D., Heutel, G., 2010. The general equilibrium incidence of environmental
difference-in-difference model. Mar. Pollut. Bull. 149.
mandates. Am. Econ. J. 2 (3), 64–89.
Wang, J., 2013. The economic impact of special economic zones: evidence from Chinese
Gao, Y.N., Li, M., Xue, J.J., Liu, Y., 2020. Evaluation of effectiveness of China’s carbon
municipalities. J. Dev. Econ. 101, 133–147.
emissions trading scheme in carbon mitigation. Energy Econ. 90, 104872.
Yan, Y., Zhang, X., Zhang, J., Li, K., 2020. Emissions trading system (ETS)
Heckman, J.J., Ichimura, H., Todd, P., 1998. Matching as an econometric evaluation
implementation and its collaborative governance effects on air pollution: the China
estimator. Rev. Econ. Stud. 65, 261–294.
story. Energ Policy 138.
Herweg, F., 2020. Overlapping efforts in the EU emissions trading system. Econ. Lett.
Yang, L., Li, Y., Liu, H., 2021. Did carbon trade improve green production performance?
193, 109323.
Evidence from China. Energy Econ. 96, 105185.
Higgins, P.A.T., Higgins, M., 2005. A healthy reduction in oil consumption and carbon
Zhu, J., Fan, Y., Deng, X., et al., 2019. Low-carbon innovation induced by emissions
emissions. Energ Policy 33, 1–4.
trading in China. Nat. Commun. 10, 4088.
Imbens, G.W., Wooldridge, J.M., 2008. Recent developments in the econometrics of
Zhu, B., Zhang, M., Huang, L., Wang, P., Su, B., Wei, Y., 2020. Exploring the effect of
program evaluation. J. Econ. Lit. 47, 5–86.
carbon trading mechanism on China’s green development efficiency: a novel
Jacobson, L.S., LaLonde, R.J., Sullivan, D.G., 1992. Earnings Losses of Displaced
integrated approach. Energy Econ. 85, 104601.
Workers. Research Papers in Economics.