Examining the Impact and Influencing Channels of Carbon Emission Trading Pilot Markets in China

Abstract

As China has an important role in global climate change, the Chinese government has set goals to improve its environmental efficiency and performance and launched carbon emission trading pilot markets in 2013, aiming to reduce CO₂ emissions. Based on panel data of 30 provinces from 2005 to 2017, this paper uses the difference-in-difference method to study the impact of China’s carbon emission trading pilot markets on carbon emissions and regional green development. The paper also explores possible influencing channels. The main conclusions are as follows: (1) China’s carbon emission trading policy has promoted a reduction in CO₂ emissions and carbon emission intensity and has increased green development in the pilot areas. (2) The main path for China’s carbon emission trading policy to achieve carbon emission reduction and regional green development is to promote technology adoption. (3) China’s carbon emission trading policy achieves green development through synergistic SO₂ emission reduction. The pilot carbon markets have reduced both the amount of SO₂ emissions and SO₂ emission intensity.

1. Introduction

China has been the world’s largest emitter of CO₂ since 2006 [1]. In recent years, in addition to facing increasing global pressure to reduce carbon emissions, China’s regional air pollution problems have also been increasing. The emission of multiple pollutants such as SO₂, NO_x, and PM2.5 in urban agglomerations has continued to increase. To achieve cost efficiency and emission reduction, more flexible market-oriented environmental policies, such as SO₂ emission rights trading and carbon emission trading, are gradually being applied in China. SO₂ emission trading started relatively early in China. Many scholars such as Tu and Shen [2] and Xu and Masui [3] successively studied its environmental and economic effects. By contrast, research on carbon emission trading is relatively scarce, since carbon emission trading pilots have only been launched gradually since the end of 2013. Furthermore, there are fewer studies that consider both carbon emission reduction and the synergy between carbon emission reduction and other pollutants.

As China’s carbon emission trading market moves from the regional pilot stage to the national stage in 2021, creating one of the world’s largest carbon markets, it is important to assess its impacts on carbon emissions. Nonetheless, given the differences between the Chinese carbon market and existing ones such as the European Union Emissions Trading Scheme (EU-ETS) and the Regional Greenhouse Gas Initiative (RGGI), it is difficult to extrapolate results from the existing literature on the impacts of carbon markets to the Chinese context. Given the relative scarcity of such studies in China, this paper contributes to the literature by providing an empirical assessment of the Chinese carbon markets using data from regional pilot markets. The paper also contributes to the literature by incorporating both global pollutants (i.e., CO₂) and regional pollutants (i.e., SO₂) to measure the green total factor productivity (GTFP) instead of the more common measures of green development that rely on single pollutant emission efficiency or pollutant emission intensity [4,5,6,7]. GTFP is further subdivided into efficiency change and technological change to evaluate the influencing channels of the effects from the carbon emission trading system. By examining the impacts of the pilot markets on the reduction in both CO₂ and SO₂, the paper also contributes to the literature by studying the synergy between global and local pollutants.

This paper attempts to answer the following questions: (1) Do carbon emission trading pilot markets reduce CO₂ emissions and carbon emission intensity and increase GTFP in China? (2) Is there any synergistic effect from carbon emission trading plot markets in China? (3) What are the influencing channels of carbon emission trading pilot markets in China?

2. Background Information and Literature Review

2.1. Background Information

Carbon markets use CO₂ emission rights as a commodity for trading, which has the advantage of controlling the total amount of carbon emissions. Faced with the dual pressures of international emission reduction commitments and the domestic environmental situation, the Chinese government announced its intent to establish carbon emissions trading markets in 2011. Eight regional emission trading schemes have been operating since 2013 (seven pilot markets during the 12th Five-Year Plan period (2011–2015) and one pilot market during the 13th Five-Year Plan period (2016–2020)). They are in the provinces of Guangdong, Hubei, and Fujian and in the cities of Beijing, Tianjin, Shanghai, Shenzhen, and Chongqing. Some of the key features of the eight pilot trading markets are shown in Figure 1 (full summary is in

Table A1. Key features of China’s carbon emission trading pilot markets.

Region	Launch Year	Allocation	Sectors Covered	Inclusion Thresholds
Shenzhen	June 2013	Free allocation	Power, water, gas, manufacturing sectors, buildings, port and subway sectors, public buses, and other non-transport sectors.	Annual emissions of 3000 tce/year for enterprises and large public buildings and 10,000 m2 for government buildings.
Beijing	November 2013	Free allocation	Electricity providers, heating sector, cement, petrochemicals, other industrial enterprises, manufacturers, service sector, and public transport.	5000 tCO2/year, considering both direct and indirect emissions.
Shanghai	November 2013	Free allocation, auctioning	Airports, aviation, chemical fibers, chemicals, commercial, power and heat, water suppliers, hotels, finance, iron and steel, petrochemicals, ports, shipping, nonferrous metals, building materials, paper, railways, rubber, and textiles.	10,000 tCO2/year or 5000 tce/year.
Tianjin	December 2013	Free allocation	Heat and electricity production, iron and steel, petrochemicals, chemicals, and exploration for oil and gas.	20,000 tCO2/year, considering both direct and indirect emissions.
Guangdong	December 2013	Free allocation, auctioning	Power, iron and steel, cement, papermaking, aviation, and petrochemicals.	20,000 tCO2/year or energy consumption 10,000 tce/year.
Hubei	April 2014	Free allocation	Power and heat supply, iron and steel, nonferrous metals, petrochemicals, chemicals, textile, cement, glass and other building materials, pulp and paper, ceramics, automobile and equipment manufacturing, food, beverage, and medicine producers.	Annual energy consumption more than 60,000 tce in any year during 2010–2011 and 10,000 tce in any year during 2014–2016.
Chongqing	June 2014	Free allocation	Power, electrolytic aluminum, ferroalloys, calcium carbide, cement, caustic soda, and iron and steel.	20,000 tCO2/year or energy consumption 10,000 tce/year.
Fujian	September 2016	Free allocation	Electricity, petrochemical, chemical, building materials, iron and steel, nonferrous metals, paper, aviation, and ceramics.	Energy consumption 10,000 tce/year for any year between 2013 and 2016.

Resource: Adapted from ICAP ref. [14].

of Appendix A).

In addition to China, countries that use emission trading markets to reduce carbon emissions include those in the European Union, the United States, New Zealand, Japan, and South Korea. The European Union Emissions Trading Scheme (EU-ETS) is the largest ETS worldwide. Starting in 2005, the EU-ETS is now in its fourth phase and has been reformed several times. There are different characteristics of the covered sectors and allocation methods in different phases. The Regional Greenhouse Gas Initiative (TGGI), covering the states of Connecticut, Delaware, Maine, New Hampshire, New Jersey, New York, and Vermont, is the first mandatory CO₂ emissions trading scheme in the U.S. The RGGI covers the power sector with a capacity equal to or greater than 25 MW and only trades under auctioning. Given the differences between the Chinese carbon market and the EU-ETS and the RGGI, it is worth exploring the impacts of carbon markets in the Chinese context.

2.2. Review of the Carbon Markets Literature

In testing for the impacts of carbon markets, the main focus of the majority of papers is carbon emissions outcomes, with only a few studies examining the impacts of other measures. These latter studies include that of Li and Zhang [8], which uses industrial carbon emissions data from 30 provinces in China to investigate the impact of carbon trading on industrial carbon emissions and carbon intensity. A few scholars have also conducted research on the synergistic effects of carbon trading on other pollutants and social welfare. Cheng et al. [9] found that carbon trading pilots can reduce carbon emissions and promote the reduction in SO₂ and NO_x. Fujimori et al. [10] found that the emissions reduction from the establishment of carbon markets reduced welfare losses from 0.7–0.9% to 0.1–0.5%. The existing literature on the effect of carbon trading focuses mainly on a single pollutant, CO₂. However, a few scholars have begun to pay attention to the impacts on other regional pollutants and social welfare. However, research using comprehensive indicators to evaluate the impact of carbon trading on the overall environmental improvement of a region, i.e., green development, is relatively rare. This is reviewed in the next section.

2.3. Review of the Green Total Factor Productivity (GTFP) Literature

Studies in the literature mostly use single pollutant emission efficiency or pollutant emission intensity to measure the level of green development. However, a good measure of green development should also take into account undesired outputs, such as pollution. This is possible using green total factor productivity (GTFP), which accounts for global and regional pollutants. Failure to include such pollutants in the measurement of green development may lead to bias in the results. Since undesirable outputs are inevitable byproducts of economic activities, GTFP measures total factor productivity (TFP) in a way that incorporates undesirable outputs.

The Malmquist–Luenberger (ML) index is often used to measure TFP through the DEA framework (GTFP can also be measured using the parametric stochastic frontier analysis (SFA) approach. However, we adopt the DEA approach here since it allows for the evaluation of the influencing channels of the effect of carbon markets.). For example, Abbott [11] used the approach to estimate the total factor productivity of the Australian electricity supply industry during 1969–1999. The paper finds a substantial improvement in the industrial performance since the mid-1980s. Coelli and Rao [12] examined the levels and trends of TFP in the agriculture sector in 93 developed and developing countries over the period 1980–2000, using DEA to derive a Malmquist productivity index. The results indicated an annual growth in TFP of 2.1% and a degree of catch-up in productivity levels between high-performing and low-performing countries.

Yaisawarn and Klein [4] used DEA to compute a Malmquist input-based productivity index, incorporating SO₂ as an undesirable output to estimate the GTFP for coal-burning plants in the U.S. electric power industry in the 1980s. They found that GTFP decreased in 1985 and increased from 1985 to 1989. Du et al. [5] used a biennial Malmquist–Luenberger productivity index, treating CO₂ as an undesirable output, to measure green growth at the provincial level in China from 1998 to 2012. Rusiawan et al. [7] studied the GTFP for low carbon economic development towards sustainable development in Indonesia, taking into account the effect of CO₂ emissions. Li and Lin [6] explored the green productivity growth of 36 Chinese industrial sectors from 1998 to 2011, under the constraint of energy consumption and CO₂ emissions.

2.4. Mechanisms for Carbon Markets to Achieve Environmental Outcomes

The transmission mechanism of a carbon emission trading system is shown in Figure 2. There are three main paths to reduce carbon emission and carbon emission intensity: the first is to upgrade energy technology to improve energy efficiency, the second is to rely on the transformation of energy structure to replace fossil energy with low-carbon energy, and the third is to optimize the industrial structure and reduce the proportion of energy-consuming sectors.

Emissions trading can promote emission reduction by promoting technological progress. Research in the fields of energy and climate change shows that two types of technological progress may have important impacts on carbon intensity: one type is technological progress with spontaneous energy efficiency improvement, whereby technological change naturally emerges over time; the other type is induced technological progress, which refers to factors such as energy prices, carbon prices, and emission reduction policies that lead to a change in the direction of technological progress. In general, induced technological progress has a stronger effect on energy savings and emission reduction [13]. Emissions trading can induce technological progress by making carbon emission more expensive and making carbon reduction more attractive. Enterprises with low carbon emissions not only can save on the costs of purchasing emission allowances but can also gain profits from reducing carbon emissions. This makes investing in technological improvements for carbon emission reduction an attractive option. Therefore, carbon markets make firms more likely to develop and adopt low carbon emission technologies.

Emissions trading also provides incentives for enterprises to upgrade and optimize their energy structure so as to use cleaner energy to reduce carbon emissions. For example, in China, the power industry accounts for a large proportion of carbon emission reductions in the carbon emission trading pilot markets [14]. This is mainly because carbon emissions reduction can often be achieved with lower abatement costs in the power sector, such as through shifting from coal to natural gas. If participating in the carbon emission trading system, the power sector will switch to less carbon-intensive fuels (such as natural gas) since this is more cost-effective than using coal and purchasing carbon emission quotas. Liu et al. [15] finds that carbon trading propels provinces to optimize their energy structure. The results show that carbon trading policies improve the efficiency of green innovation through their effects relating to technological innovation, energy substitution, and structural upgrading.

Environmental policies such as carbon emission trading to control carbon emissions may also have a synergistic effect on the reduction in regional pollutants due to the fact that CO₂ and SO₂ are derived mainly from the burning of fossil fuels [16,17]. Therefore, the implementation of China’s carbon emission trading policy may also reduce the emission of other regional pollutants and promote the overall green development of the region. Tan et al. [18] employed a bottom-up optimization model based on a technology system to assess the overall co-benefits of a CO₂ mitigation policy on local air pollution, including dust, NO_x, and SO₂ emissions, in China’s cement industry and found that co-benefits for local air quality exist with the reduction in CO₂ emissions. He et al. [19] believe that energy policies are an important point for realizing the two-way synergistic effect between greenhouse gas emission reduction and pollutant control. They evaluated the synergistic effect of China’s energy policy by constructing a comprehensive model that combines an energy prediction model, an emission assessment model, an air quality simulation model, and a health benefit assessment model. The results showed that active energy policies have brought huge benefits, including 1469 million tons of CO₂ emission reduction, a 12–32% reduction in air pollutant concentration, and more than USD 100 billion of health benefits [19]. Xiao et al. [20] also finds interactions between CO₂ and SO₂ policies.

3. Methodology, Variables, and Data

This paper aimed to examine the impact of pilot carbon markets in China on carbon emissions and carbon intensity and to see whether there are corollary benefits on other regional air pollutants. Channels through carbon markets that can bring about overall green development in a region were also explored. An estimation of the impacts of carbon trading systems was conducted using the difference-in-difference (DID) method. This was combined with the decomposition approach to further study the influencing channels of the impact of carbon emission trading pilot markets on green development. Robustness tests using the propensity score matching (PSM)–DID method and dynamic effect test were also applied.

3.1. Difference-in-Difference Method

The difference-in-difference method is a widely used policy effect evaluation method that can quantify the effect of a specific policy on the policy implementation object. The core idea is to select an experimental group (treatment group) that implements the policy and a control group that does not implement the policy and then compare the changes in a specific indicator of the experimental group before and after the policy implementation with the control group. The difference between the experimental group and the control group in terms of the change in the indicator before and after the policy implementation is regarded as the true impact of the policy on the experimental group—that is, the actual effect of the implementation of the policy.

The standard DID model is expressed as Equation (1) as follows:

$$y_{it}={\beta }_0+{\beta }_1\times t_t+{\beta }_2\times pilo{t}_i+{\beta }_3\times t_t\times pilo{t}_i+{\mu }_i+{\gamma }_t+{\epsilon }_{it}$$

(1)

where $i$ represents the provinces;

$t$ represents the year; $t_t$ represents a time dummy variable, where $t_t=0$ represents the pre-pilot period and $t_t=1$ represents the pilot period;

$pilo{t}_i$ represents the treatment variable, where $pilo{t}_i=0$ represents non-pilot regions and $pilo{t}_i=1$ represents pilot regions;

${\mu }_i$ represents a province-fixed effect;

${\gamma }_t$ represents a time-fixed effect; and

${\epsilon }_{it}$ represents the disturbance term. Since the pilots did not begin at the same time, the treatment variable was omitted and only the effect through the interaction term was estimated. The specified DID model is expressed as in Equation (2):

$$y_{it}={\beta }_0+{\beta }_1\times t_t\times pilo{t}_i+{\mu }_i+{\gamma }_t+{\epsilon }_{it}$$

(2)

This paper followed the practices of Li and Zhang [15], Ren et al. [21], and Tu and Shen [2] by adding a series of control variables to the basic DID model, including GDP per capita, pollution control investment, technology introduction, and factor endowment. The variables are discussed in Section 3.3. The DID model was constructed as follows:

$$y_{it}={\beta }_0+{\beta }_1\times tpilo{t}_{it}+{\beta }_2\times X_{it}+{\mu }_i+{\gamma }_t+{\epsilon }_{it}$$

(3)

where $tpilo{t}_{it}$ represents the interaction term $t_t\times pilo{t}_i$, and $X_{it}$ represents the control variables.

3.2. Dependent Variables

This paper studied the impacts of the carbon emissions trading pilots on carbon emissions and carbon intensity. Furthermore, we also wanted to test whether carbon markets have a synergistic effect on regional air pollutants and whether the market helps to foster overall green development in the region. CO₂ emissions were measured in 10⁴ tons. Carbon emission intensity was measured by the amount of CO₂ emissions per unit of gross domestic product (GDP)—that is, the ratio of CO₂ emissions to GDP.

To study the synergistic effect of carbon markets on regional air pollutants, we used green development as a dependent variable. To measure green development, we chose to use green total factor productivity (GTFP), incorporating both global pollutants (i.e., CO₂) and regional pollutants (i.e., SO₂) instead of the more common measure of green development, which relies on single pollutant emission efficiency or pollutant emission intensity [4,5,6,7]. We also separated GTFP into its various components (see Equation (4)) to study the impact pathways of carbon emission trading markets. Green development captured in GTFP was measured using the Malmquist–Luenberger (ML) index with a directional distance function (DDF) [22]. The larger the ML index is, the higher the level of green production is. In the calculation process, GDP was used as the desirable output; two major gas emissions, CO₂ and SO₂, were used as undesirable outputs; and capital stock, labor input, and energy inputs were used as input variables. Relative efficiency score can also be calculated by the DDF. The ML index was further broken down as follows [23]:

$$\mathrm{ML}=\mathrm{EC}\times \mathrm{TC}=\left(\mathrm{PEC}\times \mathrm{SEC}\right)\times \left(\mathrm{PTC}\times \mathrm{STC}\right)$$

(4)

The ML index can be subdivided into two components, one measuring efficiency change (EC) and the other measuring technological change (TC). TC can also be equivalently considered as the change in the frontier technology. Taking the efficiency change component calculated relative to the constant returns to scale (CRTS) technology, EC was further broken down into pure efficiency change (PEC), calculated relative to the variable returns to scale (VRTS) technology, and scale efficiency change (SEC), which captures the change in the deviation between the VRTS technology and CRTS technology. Following the same process, TC was also further divided into a pure technological change (PTC) component and a scale technological change (STC) component. By using EC, TC, PEC, TEC, PTC, and STC as dependent variables, we attempted to evaluate the influencing channels of the effect from carbon emissions trading systems. This is discussed in detail in Section 5.

3.3. Control Variables

We selected key variables that directly and indirectly affect the environmental outcomes associated with carbon markets, including GDP per capita, pollution control investment, factor endowment, technology introduction, industry structure, and energy structure, as control variables. The corresponding indicators were selected as follows:

1.

GDP per capita was selected to measure the level of economic development to control for the potential differences in CO₂ emissions, CO₂ intensity, and green development in different economic development stages. The impact was expected to be positive on CO₂ emissions and carbon intensity.

2.

Pollution control investment was measured by the ratio of waste gas control investment to GDP. Pollution control investment reflects the level of effort that provinces make to reduce pollutants, including carbon emissions and other air pollutants. Thus, pollution control investment is an important factor that affects CO₂ emissions and was included in the control variables to improve the accuracy of the regression results. The impact was expected to be positive on CO₂ emissions and carbon intensity.

3.

Liu and Wei [24] believe that there is a correlation between resource factor endowments and CO₂ emission reduction efficiency. Factor endowment refers to the different resources a region can utilize for manufacturing. The factor endowment hypothesis asserts that differences in resource endowments determine a region’s comparative advantage and the type of products for manufacturing [25]. The capital/labor ratio was selected to measure the factor endowment, since the emission intensities of capital-intensive products and labor-intensive products are significantly different. The impact was expected to be negative on CO₂ emissions and carbon intensity.

4.

The proportion of foreign direct investment (FDI) in fixed asset investment was selected to measure technology introduction, since it is mainly realized in the form of foreign direct investment. The impact was expected to be negative on CO₂ emissions and carbon intensity.

5.

Industrial structure was measured by the output added value of secondary industry as a percentage of the GDP, because secondary industry contains high energy consumption and emission sectors (in China, primary industry refers to agriculture, forestry, animal husbandry, and fishery industries; secondary industry refers to mining and quarrying, manufacturing, production and supply of electricity, heat, gas and water, and construction; and tertiary industry refers to all other economic activities not included in the primary or secondary industries [26]). At present, China’s industrial structure is in a period of transformation and adjustment. The greater the adjustment of the industrial structure, the greater the reduction in CO₂ emissions will be [27]. Therefore, industrial structure was introduced into the model as a control variable. The impact was expected to be positive on CO₂ emissions and carbon intensity.

6.

Energy structure was expressed as the proportion of coal consumption in the total energy consumption. The impact was expected to be positive on CO₂ emissions and carbon intensity.

3.4. Data

This paper includes data from 30 provinces in China from 2005 to 2017 (data do not include Tibet, Hong Kong, Macao, and Taiwan due to data availability). Data on GDP and FDI in the fixed asset investment were obtained from the China Statistical Yearbook; labor input (expressed in the number of employees) was from the statistical yearbooks of each province; energy input (energy consumption) was from the China Energy Statistical Yearbook and the Carbon Emission Accounts & Datasets (CEADs) and was converted to tons of coal equivalent; capital stock was calculated using the perpetual inventory method [28]; CO₂ emission data were obtained from CEADs (alternate measures also exist, such as the Intergovernmental Panel on Climate Change (IPCC) approach [29]; Shan et al. [30] constructed the time-series data of CO₂ emission inventories for 30 provinces in China following the IPCC approach); and the amount of SO₂ emissions, the added value of secondary industry, and waste gas pollution control investment data were from the National Bureau of Statistics of China database. To deduct the impact of price factors, all data related to pricing were adjusted to a constant year 2010 price level, based on the corresponding price index from the National Bureau of Statistics of China. The variable meanings and calculation methods of the input and output variables and control variables are summarized in

Table 1. Variable meaning and calculation method of input and output variables and control variables.

Variable	Variable Meaning	Calculation Method
Energy	Energy consumption	Consumption of 20 types of energy
Capital	Capital stock	Perpetual inventory method [28]
Labor	Number of employees by the end of year	-
CO2	Amount of CO2 emissions	-
CI	Carbon emission intensity	The ratio of CO2 emissions to GDP
SO2	Amount of SO2 emissions	-
SI	SO2 emission intensity	The ratio of SO2 emissions to GDP
GDP	Gross domestic product	Converted to a constant year 2010 price level based on the corresponding price index
GDPper	GDP per capita	Converted to a constant year 2010 price level based on the corresponding price index
WGIratio	Pollution control investment	The ratio of waste gas control investment to GDP
CapI	Factor endowment	The ratio of capital stock to labor
FDIratio	Technology introduction	The proportion of foreign direct investment (FDI) in fixed asset investment
IS	Industrial structure	The output value added of the secondary industry as a percentage of GDP
ES	Energy structure	The proportion of coal consumption in total energy consumption

. Summarized statistics of these variables for provinces in China from 2005 to 2017 are listed in

Table 2. Summary statistics of input and output variables and control variables for provinces in China from 2005 to 2017.

Variable	Measurement Unit	Obs	Mean	Std. Dev.	Min	Max
Energy	104 tons of coal equivalent	390	7712.513	4959.009	580.446	25,316.537
Capital	108 yuan	390	59,408.987	48,308.146	4137	273,417.56
Labor	104 persons	390	2609.871	1721.799	291.04	6766.86
CO2	104 tons	390	27,902.385	18,601.978	1650	84,220
CI	%	390	2.187	1.206	0.367	7.249
SO2	104 tons	390	679,790.25	437,692.82	14,271.49	2,002,000
SI	%	390	70.935	73.217	0.659	534.415
GDP	108 yuan	390	16,543.544	14,392.573	728.843	79,884.952
GDPper	104 yuan	390	3.78	2.209	0.681	12.007
WGIratio	%	390	931.475	924.428	6.781	7985.792
CapI	%	390	26.177	16.552	4.527	97.074
FDIratio	%	390	2.977	2.795	0.006	13.409
IS	%	390	46.384	8.031	19.014	59.045
ES	%	390	27.588	11.064	2.575	56.696

. China set up carbon emissions trading pilot markets in eight pilot provinces and cities—Beijing, Tianjin, Shanghai, Chongqing, Hubei, Guangdong, and Shenzhen—from the end of 2013 to 2014 and in Fujian on 30 September 2016. Therefore, this paper regards the implementation of China’s carbon emission trading pilot markets as a “natural experiment”. Since all pilots except Shenzhen are in provinces and municipalities directly under the central government, in order to unify the scope of the study, Shenzhen was merged into Guangdong province (Shenzhen is a special zone adjacent to Guangdong and thus is merged into Guangdong province). In addition, six carbon emission trading pilot markets, excluding Shenzhen and Fujian, were launched in November or December 2013 and early 2014; this paper defines the years before 2014 as the pre-pilot period ($t_t=0$) and the years from 2014 onward as the pilot period ($t_t=1$) for Beijing, Tianjin, Shanghai, Chongqing, Hubei, and Guangdong. For Fujian, the years before 2017 are the pre-pilot period ($t_t=0$) and the years from 2017 onward are the pilot period ($t_t=1$). Pilot provinces are considered the treatment group ($pilo{t}_i=1$) in this paper, and the remaining non-pilot provinces are considered the control group ($pilo{t}_i=0$).

4. Impacts of Pilot Carbon Markets on Carbon Emissions, Carbon Intensity, and Green Development

4.1. Simple Comparative Analysis of Pre-Pilot and Pilot Periods

The sample in this paper was divided into two phases (2005–2013 as the pre-pilot period and 2014–2016 as the pilot period) to examine the changes in the mean values of the main variables in the pilot regions and non-pilot regions. In order to more intuitively reflect the changes in the mean values of each variable in the pilot and non-pilot regions during the pre-pilot and pilot period, this paper uses the ratio method to measure the changes (full results are in

Table A2. Comparison of the mean values of the variables in the pilot and non-pilot regions during the pre-pilot and pilot periods.

Variable	Average of Pre-Pilot Period (2005–2013)			Average of Pilot Period (2014–2016)			(7) = (6) − (3) Change of Ratio
	(1) Non-Pilot	(2) Pilot	(3) Ratio	(4) Non-Pilot	(5) Pilot	(6) Ratio
CO2	27,189.907	20,970.926	0.771	34,331.528	23,222.778	0.676	−0.095
CI	2.641	1.424	0.539	1.977	0.859	0.434	−0.105
Efficiency	0.711	0.882	1.241	0.720	0.917	1.274	0.033
ML	1.020	1.114	1.092	1.015	1.255	1.236	0.144
EC	0.994	1.007	1.013	1.007	1.007	0.999	−0.014
TC	1.027	1.106	1.077	1.008	1.248	1.238	0.161
PEC	0.990	1.005	1.015	1.001	1.005	1.004	−0.011
SEC	1.004	1.002	0.998	1.006	1.002	0.996	−0.002
PTC	0.984	1.136	1.154	0.967	1.395	1.443	0.288
STC	1.115	0.977	0.876	1.161	0.940	0.810	−0.067
SO2	825,978.780	529,764.900	0.641	607,954.680	309,790.300	0.510	−0.132
SI	99.797	45.001	0.451	40.662	12.739	0.313	−0.138
GDPper	2.648	4.978	1.880	4.465	7.669	1.718	−0.162
WGIratio	994.074	592.288	0.596	1247.723	470.518	0.377	−0.219
CapI	17.853	31.134	1.744	34.126	49.093	1.439	−0.305
FDIratio	2.740	6.386	2.331	1.380	4.108	2.977	0.646
IS	48.803	44.510	0.912	44.746	39.001	0.872	−0.040
ES	30.520	23.816	0.780	26.125	16.898	0.647	−0.134

of Appendix B).

For the key variables, namely the amount of CO₂ emissions, carbon emission intensity, scale efficiency change, and scale technological change, the amount of SO₂ emissions and the SO₂ emission intensity of pilot regions were lower than those of non-pilot regions during both the pre-pilot and pilot periods; the efficiency, GTFP, technological change, pure efficiency change, and pure technological change of the pilot regions were higher than those of the non-pilot regions during both the pre-pilot and pilot periods. This preliminarily indicates that the implementation of China’s carbon emission trading pilots has limited the increase in the amount of CO₂ emission and carbon emission intensity and promoted green development in the pilot regions. It should be noted that this is only a simple comparative analysis and does not control for other important influencing factors; whether the implementation of China’s carbon emission trading pilots has truly improved the level of green development is more rigorously tested using econometrics.

From the perspective of explanatory variables, the GDP per capita, factor endowment, and technology introduction of pilot regions were higher than those of non-pilot regions during both the pre-pilot and pilot periods; the pollution control investment, industry structure, and energy structure of the pilot regions were less developed than those of the non-pilot regions during both the pre-pilot and pilot periods. From the point of view of the numerical gap, the gap between the pilot regions and non-pilot regions narrowed after the pilot market implementation in terms of GDP per capita, pollution control investment, factor endowment, industry structure, and energy structure; the gap between the pilot regions and non-pilot regions widened after the pilot market implementation in terms of technology introduction.

4.2. Analysis of Regression Results

We used the DID method to conduct regression analysis on the impact of China’s carbon emission trading pilot markets on CO₂ emissions, carbon emission intensity, and regional green development. In order to accurately estimate the regression model, two-way fixed effects models with and without control variables were successively estimated. Additionally, in order to facilitate the comparison of the size of the coefficients of each variable, non-dummy variables were logged so that the elasticity of each variable to the dependent variable could be intuitively analyzed.

Table 3. DID regression results (CO₂ emissions, carbon emission intensity, and GTFP).

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	lnCO2	lnCO2	lnCO2	lnCI	lnCI	lnCI	lnML	lnML	lnML
tpilot	−0.178 ***	−0.141 **	−0.135 **	−0.192 ***	−0.163 ***	−0.183 ***	0.121 *	0.092 **	0.076 *
	(−3.123)	(−2.48)	(−2.22)	(−4.625)	(−3.364)	(−3.44)	(1.828)	(2.036)	(1.895)
lnGDPper		0.794 ***	0.78 ***		−0.334 *	0.107		0.077	−0.148
		(3.454)	(2.729)		(−1.79)	(0.365)		(1.556)	(−1.149)
lnWGIratio		0.029 ***	0.03 ***		0.049 ***	0.037 ***		0.007	0.006
		(3.098)	(2.964)		(3.805)	(3.3)		(0.801)	(0.676)
lnCapI		−0.393	−0.331		−0.032	−0.291		−0.04	−0.045
		(−1.611)	(−1.212)		(−0.166)	(−1.109)		(−0.881)	(−0.454)
lnFDIratio		−0.011	−0.011		−0.021	−0.005		−0.012	−0.007
		(−0.574)	(−0.519)		(−1.297)	(−0.233)		(−0.957)	(−0.652)
lnIS		0.297 **	0.226		0.283 **	0.179		−0.202 *	0.032
		(2.336)	(1.531)		(2.435)	(1.329)		(−1.93)	(0.386)
lnES		0.099 ***	0.09 ***		0.098 ***	0.097 ***		−0.077 ***	−0.084 **
		(4.851)	(4.633)		(3.826)	(4.827)		(−2.63)	(−2.106)
_cons	9.572 ***	8.427 ***	7.66 ***	0.91 ***	−0.551	−0.353	−0.002	1.051 **	0.867 **
	(64.456)	(10.851)	(9.945)	(11.616)	(−0.718)	(−0.461)	(−0.065)	(2.115)	(2.342)
Observations	390	390	390	390	390	390	390	390	390
Control Variables	NO	YES	YES	NO	YES	YES	NO	YES	YES
Year Fixed	YES	YES	YES	YES	YES	YES	YES	YES	YES
Province Fixed	NO	NO	YES	NO	NO	YES	NO	NO	YES

t-values are in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1.

shows the results. Models (1), (4), and (7) are benchmark models with time-fixed effects that do not contain any control variables. Models (2), (5), and (8) are models that add control variables to the basic time-fixed effect models. Models (3), (6), and (9) are models with control variables that control for both time effect and province effect. It can be seen from Table 3 that in the process of adding the control variables and fixed effects, the significance of the core explanatory variables and the signs of the coefficients do not fundamentally change.

Columns (1) to (3) of Table 3 show that the regression coefficients of the core explanatory variable—the interaction term—are significantly negative at the 1% level, 5% level, and 5% level, respectively, which indicates that the implementation of China’s carbon emission trading pilots has suppressed the CO₂ emissions of the pilot regions. Columns (4) to (6) of Table 3 show that the regression coefficients of the interaction term are significantly negative at the 1% level, suggesting that the implementation of China’s carbon emission trading pilots has reduced the carbon intensity of the pilot regions. Columns (7) to (9) of Table 3 show that the regression coefficients of the interaction term are significantly negative at the 10% level, 5% level, and 10% level, respectively, indicating that the implementation of China’s carbon emission trading pilots has significantly improved the GTFP. The robustness tests using the propensity score matching (PSM)–DID method that control for GDP per capita, pollution control investment, factor endowment, technology introduction, industry structure, and energy structure confirm the regression findings (full results are in Appendix C). The specific influencing channels of carbon markets, synergistic effects, and mechanism are discussed in Section 5.

It should be noted that inherent differences between the pilot and non-pilot regions could threaten the validity of our results. For example, GDP per capita and technology introduction in the pilot regions is higher than the non-pilot areas (see Table A2). It could also be that the pilots and non-pilots differ in terms of their technological readiness, with the pilots implemented in locations that were ready to adopt new things. While income and technology introduction are controlled for in our PSM–DID regression, technological readiness cannot be measured and is, therefore, not included in our models. This latter factor could make the pilots differ from the non-pilots in a way that cannot be controlled for by our control variables. Thus, if pilot regions are inherently more ready to switch technologies, it would mean that our regressions are overestimating the impacts of carbon markets on emission outcomes.

With China’s national ETS being rolled out in 2021, we venture to draw some im-plications from our findings to be broader context of decarbonizing the Chinese economy. The ETS covers the power section, including combined heat and power and captive power plants from other sectors [31]. These are sectors included in the pilots. Thus, our results suggest that carbon markets can positively contribute to the efforts of decarbonization, albeit with some caveats as discussed in the previous paragraph. As it stands, China’s national ETS only covers 26% of greenhouse gas emissions in the country [31]. This means that even if the included industries are successful in reducing CO₂ emissions, it might not be significant enough to reduce the overall emissions for the Chinese economy. There are plans to include other sectors in the 14th Five-Year Plan (2021–2025). Polluting sectors such as aviation, building materials, chemicals, iron and steel, non-ferrous metals, pulp and paper, and petrochemicals have been planned to be included in the carbon markets [31]. If these sectors were included, then the impact of the carbon markets would be higher.

5. The Synergistic Emission Reduction Effect of Carbon Trading Policy and the Pathway of Impacts

5.1. Synergy Test

To study the synergy between carbon emissions reduction as a result of carbon markets and the reduction in other pollutants, this study used GTFP. In this section, SO₂ emission intensity (ratio of SO₂ emission to GDP) is used as the dependent variable, replacing carbon emission intensity in the regression to test the synergistic emission reduction effects of carbon emission trading pilots. The results are shown in

Table 4. Test results of synergistic effect.

	(1)	(2)	(3)	(4)	(5)	(6)
	lnSO2	lnSO2	lnSO2	lnSI	lnSI	lnSI
tpilot	−0.416 ***	−0.331 ***	−0.318 ***	−0.43 ***	−0.337 ***	−0.365 ***
	(−3.208)	(−3.438)	(−3.274)	(−4.124)	(−3.836)	(−3.895)
_cons	13.379 ***	11.995 ***	11.012 ***	4.718 ***	2.327	2.999 *
	(77.928)	(7.048)	(6.868)	(34.043)	(1.374)	(1.902)
Observations	390	390	390	390	390	390
Control Variables	NO	YES	YES	NO	YES	YES
Year Fixed	YES	YES	YES	YES	YES	YES
Province Fixed	NO	NO	YES	NO	NO	YES

t-values are in parentheses, *** p < 0.01, * p < 0.1.

.

The coefficients of the core interaction variable $tpilo{t}_{it}$ are all significantly negative at the 1% level. This reflects that the pilot carbon markets reduce both the amount of SO₂ emissions and the SO₂ emission intensity (we also used the amount of NOx emissions and NOx emission intensity (ratio of NOx emission to GDP) as dependent variables to test for synergistic effect; however, the results are not significant (the results are available upon request)). This provides evidence for the existence of a synergistic emission reduction effect of carbon trading on SO₂. This could be driven by the fact that China’s coal production ranks first in the world, and most of it is high-sulfur coal (sulfur content exceeding 2.5%). A large proportion of the total coal is burned directly in China, emitting a great volume of SO₂ during the combustion process. SO₂ emissions from burning coal account for a very high percentage of the total SO₂ emissions. Du et al. [32] analyze the synergistic effect of SO₂ emissions and CO₂ emissions at coal-fired power plants in China. Their study shows that the coal-fired power industry is key to achieving synergistic reduction of SO₂ emissions and CO₂ emissions. Thus, the switch from coal to cleaner energy sources brought about by the carbon trading policy leads to synergistic effects on SO₂ emissions reduction. However, such synergistic effect may only happen initially. If all the power industries switch away from coal and use non-coal sources, the synergistic effect may not happen in the future (the synergistic effect is not guaranteed; it depends on the type of policy employed [32]). It should also be noted that the synergistic effect between SO₂ and CO₂ can lead to overlaps between the policies targeting sulfur and carbon [20].

5.2. Impact Pathways of Carbon Markets

Carbon trading can impact environmental outcomes through a variety of channels. We tested these channels by breaking down GTFP into efficiency change and technological change. Then, efficiency change and technological change were further divided into pure efficiency change, scale efficiency change, pure technological change, and scale technological change. Time-province fixed-effect DID models with control variables were used to estimate these pathways for transmitting positive environmental outcomes. Figure 3 summarizes the results of the DID regression. The full table of results is included in Appendix D.

It shows that the positive impact of the carbon emissions trading pilot markets on the GTFP is from technological change. The positive impact of the technological change is due to the impact of pure technological change, as the impact of the scale technological change is not significant. A possible reason could be that the increase in foreign investment leads firms to take advantage of advanced technology to achieve energy savings and emission reductions with low marginal emission reduction costs (the study of Asghari [33] supports the pollution halo hypothesis that FDI inflow contributes to emission reduction by introducing cleaner environmental technology and improved environmental management practices to the region). Therefore, firms can obtain additional benefits and achieve developments and improvement in both the economy and the environment through quota market transactions in the pilot regions.

However, the results show negative impacts from the carbon emission trading pilot markets on efficiency change. The impact of pure efficiency change is not significant. The negative impact of scale efficiency change is significant but small. The non-significant results could be due to the fact that quotas are not strict in many regions and that many of the permits are freely allocated. Taking Guangdong province as an example, the free quota ratio is 95% for the power sector and 97% for the remaining sectors [14]. Such a high percentage of free quotas weakens the incentives for firms to move towards low emissions for the purpose of trading carbon emission quotas, and thus the role of the carbon emission trading pilot market to promote the optimization of the power structure and energy structure is weakened.

6. Conclusions and Policy Implications

Based on the panel data of 30 Chinese provinces from 2005 to 2017, the difference-in-difference method was used to empirically test the impact of China’s carbon emission trading pilot markets on carbon emissions, carbon intensity, and green development. The paper also studied synergistic effects with other pollutants and examined possible channels for transmitting positive environmental outcomes. The results show the following:

(1)

China’s carbon emission trading policy promoted a reduction in CO₂ emission and carbon emission intensity and resulted in increased green development in the pilot areas.

(2)

Taking into account the possible link between CO₂ and other regional pollutants, it was found that the carbon trading policy has had a positive effect on the GTFP of the pilot area through a synergistic reduction effect on regional pollutants such as SO₂. This indicates that the carbon emissions trading policy is conducive to promoting the overall green development of the pilot regions.

(3)

The main path for China’s carbon emission trading policy to achieve carbon emission reduction and regional green development is to promote technology introduction. Through further differentiation, it could be seen that the positive impact of the carbon emission trading pilot markets on the GTFP was from technological change and mostly from pure technological change. The negative impact on efficiency change was significant but small.

Given these findings, and keeping in mind the potential for overestimation of impacts due to unobserved factors, we conclude that carbon markets have strong potential to contribute to the efforts of decarbonizing the Chinese economy, with added benefits in promoting green development and suppressing SO₂ emissions. Nonetheless, given that China’s national ETS only initially includes about a quarter of CO₂ emissions, the benefits from emissions reduction generated from the ETS may not be enough to reduce overall emissions for the country. Planned expansion of coverage of the ETS should alleviate some of this problem. Additionally, our findings on synergy suggest that there may be overlaps between carbon policies and the regulation of other pollutants such as SO₂ emission rights trading. Therefore, in the process of promoting the national carbon emission trading market, this factor should be considered. Finally, due to the synergistic control of greenhouse gases and regional pollutants, carbon emission reduction projects can potentially be harnessed to achieve the dual goals of carbon emission reduction and regional emission reduction. This could lead to pollution reduction at lower costs. Thus, the interactions between carbon markets and SO₂ reduction should be further studied.