The Impact of Carbon Emissions Trading Policy on Regional Economy and Pollution Reductions in Chinese Provinces

Abstract

Carbon emissions trading policy is an important tool to achieve carbon peaking and carbon neutrality goals. In order to explore the effectiveness of carbon emissions trading policy, this paper adopts the difference-in-differences model to analyze the effects of China’s pilot policy on regional economic development and pollution reductions. The results show that carbon trading policy can significantly promote economic development and reduce total carbon emissions, industrial SO₂ emissions and solid wastes production of pilot regions. Further research finds that there is significant regional difference in policy effectiveness, with the policy more effective in western areas. In addition to direct effects, carbon trading policy could exert an indirect effect on carbon emissions, air quality, industrial solid wastes and regional economic development by optimizing energy consumption structures and industrial structures and increasing technological investment. This result verifies the “Porter hypothesis”. China should improve the construction of national carbon trading markets, covering more industries, considering the regional differences and negative spillover effect.

1. Introduction

With rapid economic development and increasing energy consumption, the excessive emission of greenhouse gases (GHGs) has caused climate change issues, such as global warming and other extreme weather, which threaten the survival and development of human beings [1,2,3,4,5], and bring serious economic losses [6,7]. For mitigating the harmful effects of climate change and reducing GHGs emissions, countries across the world developed the United Nations Framework Convention on Climate Change (UNFCC) and participated in the annual Conference of the Parties (COPs) to formulate international agreements about abatement obligations. Market-based approaches are favored by many countries to control environmental pollution, reduce energy consumption and promote firms to improve technology and production efficiency. Carbon emissions trading (CET) systems aimed at reducing carbon emissions with less abatement costs are widely adopted in the world. By 2023, 36 CET systems had been established globally, and another 22 systems are currently under consideration and development. The existing 36 systems cover 18% of GHG emissions, 58% of global GDP in jurisdictions and one third of the population [8].

China, as one of the biggest developing countries, relies on resource-intensive industries to develop the economy, which produce large carbon emissions. In 2006, China became the largest carbon emitter in the world, accounted for 27.6% of the total global carbon emissions [9]. China plays an important role in global climate change, and the Chinese government has set several carbon abatement targets to actively respond to climate change. For example, China’s five-year plan required that national average carbon emission intensity decreased by 18% in 2020 compared with 2015. In 2020, the “double carbon” goal was proposed, achieving carbon peaks around 2030 and carbon neutrality by 2060. All these goals demonstrated the Chinese government’s determination to reduce carbon emissions. For achieving the goals of carbon emissions abatement, the Chinese government had issued and implemented a series of laws, policies and regulations. In 2010, China began to consider establishing a carbon emissions trading system. Then, in 2011, the National Development and Reform Commission (NDRC) officially approved a carbon trading pilot project to be carried out in seven provinces and cities, including Beijing, Tianjin, Shanghai, Chongqing, Guangdong, Hubei and Shenzhen. In 2016, the Fujian province began to establish a carbon market as the eighth pilot. By the end of 2017, the NDRC issued the National Carbon Emission Trading Market Construction Plan (Power Generation Industry) and the carbon market began to expand from its pilot to nationwide. In 2020, the Ministry of Ecology and Environment of the People’s Republic of China (MEP) published the Measures for the Administration of Carbon Emissions Trading (trial), and in July 2021, the national unified CET market was officially opened, which involved more than 2000 power plants and carbon dioxide (CO₂) emissions of 4.5 billion tons, covering about 40% of the annual carbon emissions [10]. Since then, China’s national carbon market has become the biggest market in the world. The mode of the CET system in China is cap and trade. The regulator sets a carbon emissions cap, and distributes initial allowances to emitters based on a certain standard. Then, economic agents can trade their allowances in the carbon market to meet government requirements and maximize their profits or minimize abatement costs. In theory, the CET system could reduce pollution emissions, save costs, encourage emitters to improve their control technologies, and promote economic growth to achieve low-carbon transformation. However, in practice, the effectiveness of the carbon market was directly affected by market liquidity [11,12], market information asymmetry and uncertainty [13]. So, it is important to explore the level and mechanism of the actual carbon trading system in China to reduce carbon emissions and promote low-carbon economy development. The pilots’ evaluations are helpful to provide important references for the optimization and implementation of the national carbon emissions trading market.

In addition to global climate change, China also faces severe environmental pollution. Taking air pollution as an example, in 2023, the annual average PM_2.5 concentration was 30 μg/m³ in 339 cities of China [14], which was much higher than the annual average concentration of 5 μg/m³ regulated by the WHO’s global air quality guidelines, which can bring health damages [15,16,17,18]. Environmental pollution is mainly from the discharge of industrial wastewater, waste gas emission and solid wastes production. Some industrial pollution emissions and CO₂ emissions have the same source, both mainly from the combustion of fossil energy [19,20]. The measures of controlling carbon emissions may also reduce industrial pollution emissions. Some existing studies found that the carbon abatement policy could improve local air quality and bring health benefits [21,22,23,24].

In the framework of the carbon emissions trading policy, enterprises need to meet the constraint of carbon emissions, and meanwhile maximize their benefits. In order to avoid government penalties, they need to reduce their carbon emissions. In the short term, enterprises could choose to reduce product outputs and energy consumption to achieve carbon emissions goals, which would reduce other pollutants emissions at the same time. As to the enterprises with high energy consumption and high pollution emissions, their behaviors to reduce outputs and energy consumption would optimize energy structures and industrial structures. Based on the above analysis, Hypothesis 1 and 2 are proposed.

Hypothesis 1. 

Carbon emissions trading policy can effectively reduce regional industrial pollution and carbon emissions.

Hypothesis 2. 

Carbon emissions trading policy can effectively reduce regional industrial pollution and carbon emissions by the adjustment of energy structures and industrial structures.

In the long term, the behaviors to reduce outputs and energy consumption would damage the economic performance of enterprises and limit their sustainable development [25]. Enterprises can adopt green and low-carbon technologies, promote technological innovation to promote energy utilization efficiency and reduce long-term emission reduction costs. The contradiction between enterprise production and emission reduction can be fundamentally alleviated. Therefore, Hypothesis 3 is proposed.

Hypothesis 3. 

Carbon emissions trading policy can effectively reduce regional industrial pollution and carbon emissions by technological innovation.

Traditional economics supposed that economic growth and environmental protection were not mutually beneficial and even had a negative impact on a country’s economic development, as environmental regulation would increase the cost expenditures of enterprises and crowd out the investment of technological innovation [26]. Unlike traditional schools, Porter believed that appropriate environmental regulation could encourage companies to conduct technological innovation, and the benefits of such technological innovation could partially offset or even exceed the cost of complying with environmental regulations, and achieve a win-win economic and environmental situation, which is the “Porter hypothesis” [27]. Based on the above analysis, Hypothesis 4 and 5 are proposed.

Hypothesis 4. 

Carbon emissions trading policy can effectively promote regional economy.

Hypothesis 5. 

Carbon emissions trading policy can effectively promote regional economy by technological innovation.

As a market-based environmental regulation, can a carbon emissions trading policy lead to the win-win situation of pollution reductions and regional economic development? Also, how does this policy affect economy and pollution reduction? Answering the above two questions would be important for promoting the implementation of a national carbon trading policy and the design of related supporting policies. Therefore, based on the panel data from 2008 to 2020 of 30 provincial-level administrative regions in China, this paper adopted the difference-in-differences method (DID) to analyze the impact and mechanisms of a carbon emissions trading system on emissions reduction and regional economic development. The rest of this paper is arranged as follows. Section 2 presents the literature review. Section 3 shows the methods and data. Section 4 analyzes the direct and indirect effects of carbon emissions trading on pollution reduction and reginal economy development, and provides a series of robustness tests. The conclusions and policy implications are in Section 5.

2. Literature Review

The direct aim of the CET system is to reduce carbon emissions. Scholars have extensively researched carbon emissions reduction effects from ex-ante and ex-post evaluation in worldwide areas. During the initial phase of the CET system, some researchers use macroeconomic models such as CGE models to simulate the reductions and economic effects [28,29,30]. Research results have mostly indicated that CET would decrease carbon emissions and carbon intensity, but have a certain negative impact on the economy. An agent-based model was also used in this period to simulate the trading process and explore the carbon emissions and economic effects [31,32,33]. As the CET system has been implemented for several years, econometric methods have been widely employed to assess the policy effects and reduction mechanisms, mainly including DID, PSM-DID and the Synthetic Control Method (SCM). Empirical results showed that the CET pilot system had significant effects on the reductions of carbon emissions and carbon intensity [34,35,36,37,38,39]. Furthermore, these reductions were achieved through reductions in total energy consumption [40], technological innovation [39,41] and adjustments in the structure of the energy sector [9].

As an environmental regulatory measure, the CET system could generate synergistic effects, co-control carbon emissions and other pollutants [42,43], especially air pollution [23,44,45], improving environmental quality [46]. Chen et al. (2024) [47] used the DID method to investigate the influence of CET on the synergistic emission reduction of carbon dioxide and various pollutants, including sulfur dioxide (SO₂), nitrogen oxides (NO_X), ambient fine particulate matter (PM_2.5), ammonia nitrogen (AN), chemical oxygen demand (COD) and solid waste (SW), and found that the CET system had a reduction effect on these pollutions through technological progress and structure adjustment. Dong et al. (2022) [48] demonstrated that CET significantly affected the co-benefits of total carbon emissions reduction and air quality improvement, and this system also can indirectly affect carbon emissions and air quality by changing the innovation ability of cities and location choice of local industries. Lu et al. (2024) [49] found that the CET pilot policy can reduce urban environmental pollution including industrial wastewater emission, SO₂ emission and soot emission, and the government’s environmental awareness and technological innovation are the channel of action. Some studies analyzed the co-benefits of the CET system [45,50]; Chang et al. (2020) [45] explored the CET system and found that it could decrease PM_2.5 concentration, reduce morbidity and mortality and the health benefits can partially offset GDP loss.

As a market-based economic measure, the CET system significantly impacted the economy, technology, society and other aspects. Some studies found that the CET system could bring economic loss [51,52], while some others argued that it had contributed to economic development [53,54], and the mechanisms are mainly promotion of technological innovation and adjustments of industrial structure. Pang et al. (2021) [55] found that the emission trading system would cause 0.3–0.5% loss of GDP in 2030 from the baseline at the national level. At the provincial level, the GDP impacts vary from 4.6% (Shanxi) to 1.8% (Hainan). The research of Dong et al. (2024) [53] showed that CET can effectively promote urban economic development through innovation. There is no consensus within the academic community on this matter. Some studies have demonstrated that CET significantly improved the quality of low-carbon economy in county-level regions through administrative intervention and technological progress [56], and promoted low-carbon economic transformation [13]. In addition, the impact of CET on energy consumption is a hot research field, such as energy structures [57] and energy efficiency [41]. Some results found that the CET system accelerated the transformation of energy consumption structures [57], and promoted energy transition [58]. The CET had a positive effect on improving the efficiency of green innovation by affecting the industrial structure, energy structure, human capital and FDI [59], and significantly promoted technical change towards clean energy [60]. Some research has focused on the effect on technological innovation [59,61,62] and green development performance of enterprises [63,64]. Liu et al. (2023) [62] found that CET had significantly encouraged green technological innovation in industrial enterprises and this positive effect is greater for firms with large capital scales, with better corporate governance. However, some literature have suggested that a CET policy has not promoted technological innovation [65] and failed to generate Porter effects in the short term [66]. In addition to economy, energy and technology, some scholars have also paid attention to the effects of a CET system on society, especially employment [67]. The CET policy in China has generated some employment dividend, and the impact of a carbon emission trading policy on employment is gradually increasing [68]. The study by Cong et al. (2023) [69] found that although energy consumption and technology levels effectively dampen CET’s negative employment effects and promote employment in the pilot areas, the high carbon price levels within the range of carbon price fluctuation have inhibited the employment levels of the pilot industries. As future carbon prices increase, the policy’s regional employment dividend may decrease or disappear.

Carbon trading policy may result in carbon leakage and pollution transfer [48,70]. The “pollution haven hypothesis” indicates that polluting firms may choose to move to areas that have lax environmental regulations. Therefore, a carbon emissions trading policy could also have a third-party effect on neighboring cities, resulting in the degradation of their environmental quality. The design and implementation of carbon trading policy is necessary to consider its potential effect on other regions, especially the negative spillover effect, which is vital to achieving environmental justice and reduce the inefficiency of the policy.

Based on the literature review above, we can observe that existing studies provide plenty of methods and directions for evaluating the effectiveness of the CET system, such as carbon emission reduction, pollution reduction, economy and energy. The time period of these studies mainly centers before 2017. However, the CET pilot policy was implemented in 2013 and continued until 2020 (when the national carbon market was established), so the time period is relatively short. In addition, the effect and mechanism of the CET policy among existing studies is still in dispute. There is relatively limited research on the effects of a CET system on pollution reductions and economy simultaneously. Therefore, this paper tries to adopt the DID method with provincial panel data from 2008 to 2020 to empirically examine the pollution reduction and economy effects of the CET policy, discuss the heterogeneity in different regions and furthermore explore the transmission mechanism using a mediation effect model.

3. Methodology and Data

3.1. DID Model

This paper treated the CET policy as a quasi-natural experiment, and used the DID model to estimate the causal effect of the pilot policy on economy and pollution reductions. The six pilot provinces and cities including Beijing, Shanghai, Tianjin, Chongqing, Guangdong (here containing Shenzhen) and Hubei are considered as the experiment group, while the other 24 regions are the control group (except Tibet). By comparing the difference between the non-pilot areas and the pilot areas both before and after the policy’s implementation, we can eliminate the difference between the different areas before the policy’s implementation and obtain the causal effect of the policy. The basic DID model is as follows:

$$Y_{it}={\alpha }_0+{\alpha }_1\left({treat}_i\times {time}_t\right)+{\alpha }_2{Control}_{it}+{\mu }_i+{\gamma }_t+{\epsilon }_{it}$$

(1)

where t indexes the year and $i$ indexes the region. $Y_{it}$ represents the dependent variable value of region $i$ in year t. ${\alpha }_0$ is the constant term. ${treat}_i$ is a dummy variable that equals one if a region belongs to a pilot area and zero otherwise. ${time}_t$ equals one for every year after the policy implementation. ${\alpha }_1$ and ${\alpha }_2$ are the coefficients to be estimated, in which ${\alpha }_1$ on the cross term ${treat}_i\times {time}_t$ captures the causal effect of the CET policy. ${Control}_{it}$ represents the control variables of region $i$ in year t. ${\mu }_i$ is the region fixed effect to control for unobserved factors that changed across regions but were unchanged as time progresses. ${\gamma }_t$ is the year fixed effect to control for unexpected events over time. ${\epsilon }_{it}$ is an error term, representing the influence of unobserved factors in the model.

3.2. Mediation Model

To further explore the mechanisms of CET policy effects, the mediation model was employed to analyze the indirect transmission pathways. The model is presented as follows:

$$M_{it}={\beta }_0+{\beta }_1\left({treat}_i\times {time}_t\right)+{\beta }_2{Control}_{it}+{\mu }_i+{\gamma }_t+{\epsilon }_{it}$$

(2)

$$Y_{it}={\lambda }_0+{\lambda }_1\left({treat}_i\times {time}_t\right)+{\lambda }_2{M}_{it}+{\lambda }_3{Control}_{it}+{\mu }_i+{\gamma }_t+{\epsilon }_{it}$$

(3)

In the above model, $M_{it}$ represents the mediation variable. The meanings of other variables and coefficients are the same with model (1). The test processes of mediation effects are as follows. First, perform a regression on Equation (1) and test the regression coefficient ${\alpha }_1$ of Equation (1). If ${\alpha }_1$ is not significant, the test is stopped. If ${\alpha }_1$ is significant, then proceed with the following test. Second, perform a regression on Equation (2) and test the regression coefficient ${\beta }_1$ of Equation (2). If ${\beta }_1$ is significant, then proceed with the following test. Third, perform a regression on Equation (3) and test the regression coefficient ${\lambda }_2$ of Equation (3). If ${\lambda }_2$ is significant, a mediation effect exists. Then, test the regression coefficient ${\lambda }_1$ of Equation (3). If the coefficient ${\lambda }_1$ is significant, a partial mediation effect exists; if the coefficient is not significant, a full mediation effect exists.

3.3. Variables and Data

3.3.1. Data

A panel dataset consisting of 30 provinces and cities (except Tibet) in China from 2008 to 2020 was used in this study. The carbon emissions data were collected from Carbon Emissions Accounts and Datasets (CEADs). The pollution emission and economic data were collected from the 2008–2020 China Statistical Yearbook. We set a cut-off year for the policy adoption in 2013 in which year the pilot policy of carbon emissions trading started. There were six provinces and cities in the treated group including Beijing, Shanghai, Tianjin, Chongqing, Guangdong and Hubei which were chosen as the pilots for the emissions trading policy. The control group consisted of the other 24 provinces and cities over the country.

3.3.2. Dependent Variable

This paper tries to analyze the effects of CET policy on economy and pollution reductions. The dependent variables reflecting regional economy development include per capita gross regional product (pGRP). The dependent variables representing pollution emissions include total carbon emissions (CEs), industrial sulfur dioxides emissions (SO₂), discharge of chemical oxygen demand in industrial wastewater (COD) and industrial solid waste production volume (SW). The variables of pollution emissions were taken in the logarithm form.

3.3.3. Independent Variables

The core independent variable is the implementation of a CET pilot policy, namely ${treat}_i\times {time}_t$. When a province or city belongs to the pilot areas, and the year is during the policy implementation years (here, we set 2013–2020), the policy dummy variable ${treat}_i\times {time}_t$ is set to 1, otherwise, it is set to 0.

3.3.4. Control Variables

To mitigate potential endogeneity issues arising from omitted variables which impact region economy and pollution emissions, this paper set several control variables based on the existing literature [9,42,48], including population size, energy consumption, industrial pollution control investment and foreign direct investment.

3.3.5. Mediation Variables

In addition to the direct effect on economy and pollution emissions, there are existing indirect mechanisms which may affect the efficiency of the CET policy. This paper explores three such potential mechanisms. First, the CET policy may stimulate enterprises to explore green technologies and innovate to achieve long-term competition. Then, we used the research and experimental development (RD) investment as the proxy for innovation and technological progress. Second, under the pressure of abatement and cost, firms tend to replace fossil fuels with low-carbon energy in order to achieve carbon reduction and save costs by optimizing their energy structure. So, the proportion of coal consumption in total energy consumption was used to show energy structure adjustment. Third, the carbon emissions quota between different industries are different and the CET policy mainly involves key emission industries, such as the chemical industry, steel, non-ferrous metals, papermaking and electricity. By strengthening the constraints on these industries, it is beneficial to guide the transfer of resources such as capital and talent to low-energy and low-emission industries, forcing the transformation, adjustment, and upgrading of industries to achieve carbon reduction. Here, the share of secondary industry value added in GRP is used to express the adjustment of industrial structure.

Table 1. Explanation of variables.

Variable Types	Variables	Description	Data Sources
Independent variables	pGRP	Per capita GRP (CNY 10,000/person)	National Bureau of Statistics
	CEs	Total carbon emissions (ten thousand ton)	Carbon Emissions Accounts and Datasets (CEADs)
	SO2	Industrial SO2 emissions (ten thousand ton)	National Bureau of Statistics
	COD	Industrial COD emissions (ten thousand ton)	National Bureau of Statistics
	SW	Industrial solid waste production volumes (ten thousand ton)	National Bureau of Statistics
Control variables	POP	Proportion of urban population (%)	National Bureau of Statistics
	FDI	Share of foreign direct investment in GRP (%)	National Bureau of Statistics
	INV	Share of industrial pollution control investment in GRP (%)	National Bureau of Statistics
	EI	Energy intensity (ten thousand tons of standard coal/CNY 100 million)	National Bureau of Statistics
Mediation variables	ES	Proportion of coal consumption in total energy consumption (%)	National Bureau of Statistics
	SI	Share of secondary industry value added in GRP (%)	National Bureau of Statistics
	RD	Research and experimental development investment (CNY 10,000)	National Bureau of Statistics

shows the variable definitions, and

Table 2. Summary statistic.

Variables	Obs	Mean	Min	Max	Std. Dev.
pGRP	390	4.787	0.970	16.416	2.715
lnCEs	390	5.585	3.469	7.563	0.784
lnSO2	390	3.416	−1.715	5.208	1.243
lnSW	387	8.814	5.303	10.727	1.029
lnCOD	390	3.549	0.678	5.290	0.995
INV	390	0.132	0.001	1.103	0.129
POP	390	57.026	29.116	89.583	13.142
lnFDI	390	11.090	7.762	14.825	1.436
EI	390	0.903	0.188	2.835	0.536
ES	390	96.503	1.322	256.061	42.966
SI	390	42.075	15.967	61.960	8.273
lnRD	390	14.097	9.079	17.034	1.416

shows the summary statistics.

4. Results and Discussions

4.1. Baseline Results

Table 3. Baseline regression.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Variables	pGRP	pGRP	lnCEs	lnCEs	lnSO2	lnSO2	lnSW	lnSW	lnCOD	lnCOD
${treat}_i\times {time}_t$	1.486 ***	0.933 ***	−0.243 ***	−0.254 ***	−0.716 ***	−0.415 ***	−0.424 ***	−0.278 ***	−0.148 **	−0.021
	(0.151)	(0.161)	(0.034)	(0.037)	(0.077)	(0.082)	(0.060)	(0.068)	(0.067)	(0.075)
lnFDI		−0.013		0.018		−0.035		−0.022		0.114 **
		(0.100)		(0.023)		(0.051)		(0.042)		(0.047)
INV		1.010 ***		−0.380 ***		−0.146		−0.104		0.373 **
		(0.365)		(0.085)		(0.186)		(0.157)		(0.170)
EI		1.040 ***		0.354 ***		0.144		−0.018		0.312 **
		(0.280)		(0.065)		(0.143)		(0.119)		(0.131)
POP		−0.097 ***		0.013 ***		0.078 ***		0.035 ***		0.042 ***
		(0.018)		(0.004)		(0.009)		(0.008)		(0.009)
Constant	4.604 ***	9.297 ***	5.615 ***	4.398 ***	3.504 ***	−0.720	8.866 ***	7.142 ***	3.567 ***	−0.458
	(0.035)	(1.467)	(0.008)	(0.340)	(0.018)	(0.749)	(0.014)	(0.625)	(0.015)	(0.684)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	390	390	390	390	390	390	387	387	390	390
R-squared	0.959	0.967	0.975	0.978	0.949	0.958	0.955	0.958	0.940	0.946

Standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

shows the baseline results of the effects of a carbon trading policy on the economy and pollution emissions, according to Equation (1). The interaction term ${treat}_i\times {time}_t$ is significantly positive at the 1% level in Column (1), which indicates that the per capita GRP in the pilot areas have experienced larger progress during the policy period compared to the non-pilot areas. This result is relatively stable after adding other factors impacting economic development in Column (2), which confirms Hypothesis 4. Regarding the total carbon emissions, the interaction term ${treat}_i\times {time}_t$ is significantly negative at the 1% level in Column (3) and (4), indicating that the carbon emissions decreased in pilot provinces during the policy period compared to the non-pilot provinces. As to the abatement effect of other pollutants from Column (5) to (10), the interaction terms are significantly negative at the 1% level for SO₂ emissions and solid wastes production except COD emissions, suggesting that the carbon trading policy could have negative effects on SO₂ emissions and industrial solid wastes production. These results verify Hypothesis 1. Regarding the COD emissions, the interaction term is significantly negative at the 5% level in Column (9) but loses its significance after adding control variables in Column (10), which suggests that the carbon emissions trading policy may not exert a significant impact on the COD emissions reduction. This may be because the carbon emissions trading pilot policy pays more attention to the industries emitting waste gases, such as power industries, rather than industries producing high emissions of waste water, such as papermaking, agricultural and sideline food processing and textiles. Some study had the same result [42], which found the carbon trading policy had no significant reduction impact on COD discharge. In addition, this result also indicates that the degree of pollution reduction from the carbon trading policy is closely related to whether it has relationship with CO₂ emissions [71]; for example, SO₂ and CO₂ are both from the combustion of fossil energy. The above results provide evidence that the carbon emission trading policy is effective in achieving the co-benefits of carbon emissions reduction, other pollution control and economic development.

4.2. Robustness Check

We conducted a series of additional analyses to ensure that the results were robust.

4.2.1. Parallel Trend Test

The hypothesis of the DID model is to find a proper control group as the counterfactual treatment group, which requires parallel trends between the treated and control groups. Therefore, a parallel trend test was used to explore the trend evolving over time between the two groups. The interaction term is expected to be insignificant before the implementation time, but significant after implementation if the two groups satisfy the hypothesis of the parallel trend.

Table 4. Results of parallel trend test.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Variables	pGRP	lnCEs	lnSO2	lnSW	lnCOD	ES	SI	lnRD
pre_5	−0.354	0.159	0.216	0.375	0.143	3.873	−0.027	0.037
	(0.331)	(0.080)	(0.163)	(0.145)	(0.162)	(7.441)	(1.557)	(0.114)
pre_4	−0.323	0.143	0.225	0.356	0.151	2.12	0.302	0.003
	(0.330)	(0.080)	(0.162)	(0.145)	(0.162)	(7.432)	(1.556)	(0.114)
pre_3	−0.227	0.141	0.172	0.317	0.079	6.32	−0.088	−0.064
	(0.329)	(0.079)	(0.162)	(0.145)	(0.161)	(7.417)	(1.552)	(0.114)
pre_2	−0.067	0.060	−0.013	0.040	−0.010	2.973	−0.443	−0.016
	(0.328)	(0.079)	(0.161)	(0.144)	(0.161)	(7.390)	(1.547)	(0.114)
current	0.156	−0.132 *	0.008	−0.030	0.022	−15.547 **	0.45	0.038
	(0.330)	(0.080)	(0.162)	(0.145)	(0.161)	(7.430)	(1.555)	(0.114)
post_1	0.329	−0.159 **	0.019	−0.014	0.043	−18.035 **	1.194	0.117
	(0.331)	(0.080)	(0.163)	(0.145)	(0.162)	(7.461)	(1.562)	(0.115)
post_2	0.406	−0.110	0.045	−0.077	0.023	−20.473 ***	1.979	0.226 **
	(0.332)	(0.080)	(0.163)	(0.146)	(0.162)	(7.473)	(1.564)	(0.115)
post_3	0.748 **	−0.151 *	−0.420 **	−0.063	0.134	−24.517 ***	1.841	0.282 **
	(0.335)	(0.081)	(0.165)	(0.147)	(0.164)	(7.537)	(1.578)	(0.116)
post_4	1.020 ***	−0.162 **	−0.561 ***	−0.131	0.119	−27.215 ***	1.85	0.259 **
	(0.338)	(0.081)	(0.166)	(0.148)	(0.165)	(7.600)	(1.591)	(0.117)
post_5	1.490 ***	−0.207 ***	−0.743 ***	−0.125	0.005	−33.239 ***	2.471 *	0.276 ***
	(0.285)	(0.069)	(0.140)	(0.125)	(0.139)	(6.410)	(1.342)	(0.099)
lnFDI	−0.026	0.020	−0.023	−0.011	0.113 **	1.900	0.511	−0.046
	(0.097)	(0.023)	(0.048)	(0.042)	(0.047)	(2.174)	(0.455)	(0.033)
INV	0.792 **	−0.382 ***	−0.014	−0.104	0.363 **	−16.778 **	2.127	0.359 ***
	(0.356)	(0.086)	(0.175)	(0.158)	(0.174)	(8.014)	(1.677)	(0.123)
EI	0.997 ***	0.371 ***	0.176	0.029	0.334 **	−14.658 **	−2.873 **	0.398 ***
	(0.272)	(0.065)	(0.134)	(0.120)	(0.133)	(6.112)	(1.279)	(0.094)
POP	−0.061 ***	0.011 **	0.056 ***	0.031 ***	0.042 ***	−1.716 ***	0.340 ***	0.077 ***
	(0.019)	(0.005)	(0.009)	(0.008)	(0.009)	(0.423)	(0.089)	(0.007)
Constant	7.480 ***	4.483 ***	0.381	7.181 ***	−0.439	191.676 ***	19.141 ***	9.790 ***
	(1.450)	(0.350)	(0.713)	(0.637)	(0.709)	(32.629)	(6.830)	(0.502)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	390	390	390	387	390	390	390	390
R-squared	0.970	0.979	0.965	0.960	0.946	0.929	0.916	0.985

Standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

shows the results of parallel trend test. The term pre_i represents the policy dummy variable of i years before the policy implementation, and post_j represents the dummy variable of j year after policy implementation. In Column (1), the estimators during pre_2 to pre_5 are statistically insignificant, which suggests that the trends of economy development between the pilot and non-pilot provinces were parallel before the implementation of the carbon emission trading policy. However, with policy implementation, the estimators of pGRP turned to positive and the ones from post_3 to post_5 were significantly positive, which suggested that the CET policy was effective in promoting economic development in the pilot provinces, and displayed some delay.

As to the carbon emissions in Column (2), before the policy implementation year, the regression coefficients of the carbon emissions are insignificant, which suggests there is no significant difference between the pilot and non-pilot regions. However, after policy implementation, differences began to emerge. The level of CO₂ emissions in the pilot regions exhibited a significant decrease compared to non-pilot regions, and the reduction effect continuously increased, which showed that the impact of the CET policy was persistent through the implementation period and did not fade away over time. In terms of SO₂ emissions, the pollution reduction effects showed some delay. Initially, there was no significant difference between the pilot and non-pilot regions after the policy implementation. However, differences began to emerge from the third year onward. The industrial solid waste production showed a decline following policy implementation, but the results are not notable. As to the industrial COD emissions, the results found that there is no reduction effect. Columns (6)–(8) showed the parallel trend tests for the mediation variables and the three variables all passed the parallel trends test. In Column (6), the estimators during pre_2 to pre_5 were statistically insignificant, which suggested that the trends of energy structures (ES) between the pilot and non-pilot provinces were parallel before the implementation of the carbon emission trading policy. However, the estimators during current to post_5 were significantly negative, which indicated that the CET decreased the proportion of coal consumption in total energy consumption.

In summary, before the implementation of the pilot policy, there was no significant difference in the trend of pollutant levels and the mediation variables between the pilot regions and non-pilot regions. However, significant differences emerged after policy implementation, aligning with the criteria for parallel trends except industrial COD emissions. Because the industrial COD emissions did not pass the parallel test, the following placebo test did not include this indicator.

4.2.2. Placebo Test

A placebo test was carried out to ensure the robustness of the regression results, and prevent bias stemming from omitted variables, unobservable factors and random influences. Therefore, this paper adopted methods of fabricating an experiment group and control group to test the credibility of the regression findings. Several samples with the same province number were randomly selected as the pilot provinces to form a “fake treat group”. The remaining provinces were classified as the “fake control group”. Then, we repeated the regressions 500 times, and the regression coefficients and their corresponding p-values were obtained. The results are as shown in Figure 1, Figure 2, Figure 3 and Figure 4.

The test results indicated that across 500 random regressions, the average regression coefficients of the policy interaction term among the four dependent variables were approximately 0, exhibiting a normal distribution, which were different from the coefficients of real DID results (the vertical red lines in figures). In addition, the majority of corresponding p-values exceeded 0.1, indicating insignificance in their regression results. The results of regression coefficients and p-values affirmed the reliability of the previous regression outcomes and supported the conclusions that the CET policy has a pollutants reduction effect, indicating they were not merely random outcomes.

4.3. Mechanism Analysis

A stepwise regression approach according to Equations (2) and (3) was employed to examine the mediating effects using three mediating variables, i.e., energy consumption structure (ES), industrial structure (SI) and technology development (RD).

According to the regression results shown in Columns (1), (2), (3) and (4) of

Table 5. Regression results for step 1 and step 2 of mediating effects test.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Variables	pGRP	lnCEs	lnSO2	lnSW	ES	SI	lnRD
${treat}_i\times {time}_t$	0.933 ***	−0.254 ***	−0.415 ***	−0.278 ***	−26.812 ***	1.732 **	0.207 ***
	(5.797)	(−6.823)	(−5.054)	(−4.072)	(−7.664)	(2.398)	(3.861)
lnFDI	−0.013	0.018	−0.035	−0.022	1.760	0.528	−0.041
	(−0.126)	(0.792)	(−0.691)	(−0.516)	(0.811)	(1.180)	(−1.242)
INV	1.010 ***	−0.380 ***	−0.146	−0.104	−19.529 **	2.486	0.399 ***
	(2.771)	(−4.493)	(−0.784)	(−0.665)	(−2.463)	(1.519)	(3.292)
EI	1.040 ***	0.354 ***	0.144	−0.018	−14.796 **	−2.933 **	0.391 ***
	(3.711)	(5.454)	(1.004)	(−0.147)	(−2.428)	(−2.331)	(4.196)
POP	−0.097 ***	0.013 ***	0.078 ***	0.035 ***	−1.245 ***	0.297 ***	0.072 ***
	(−5.269)	(3.051)	(8.299)	(4.408)	(−3.095)	(3.578)	(11.646)
Constant	9.297 ***	4.398 ***	−0.720	7.142 ***	167.206 ***	21.384 ***	10.037 ***
	(6.339)	(12.940)	(−0.962)	(11.434)	(5.243)	(3.248)	(20.574)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	390	390	390	387	390	390	390
R-squared	0.967	0.978	0.958	0.958	0.937	0.927	0.986

Standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

, the regression coefficient of the policy variable ${treat}_i\times {time}_t$ for pGRP was significantly positive, and the ones for CO₂, SO₂ and SW were significantly negative when excluding the mediating variables and only including control variables. Additionally, the regression coefficients of the policy variable for ES, SI and RD in Column (5), (6) and (7) of Table 5 were all significant, demonstrating that the CET pilot policy can promote the adjustment of industrial structure, energy consumption structure and technological progress. As to the ES, the regression coefficient of the policy variable was significantly negative, indicating that CET would decrease the proportion of coal consumption in total energy consumption. As to the SI, the regression coefficient of the policy variable was significantly positive, indicating that CET would increase the share of secondary industry value added in GRP. As to the RD, the regression coefficient of the policy variable was significantly positive, indicating that CET would increase research and experimental development investment.

Table 6. Regression results for step 3 of mediating effects test.

Variables	pGRP			lnCEs			lnSO2			lnSW
	(1)	(2)	(3)	(1)	(2)	(3)	(1)	(2)	(3)	(1)	(2)	(3)
${treat}_i\times {time}_t$	0.653 ***	0.793 ***	0.719 ***	−0.074 **	−0.248 ***	−0.242 ***	−0.300 ***	−0.408 ***	−0.415 ***	−0.152 **	−0.282 ***	−0.231 ***
	(3.845)	(5.236)	(4.653)	(−2.357)	(−6.594)	(−6.366)	(−3.430)	(−4.928)	(−4.934)	(−2.107)	(−4.089)	(−3.359)
lnFDI	0.006	−0.055	0.030	0.006	0.020	0.016	−0.043	−0.033	−0.035	−0.031	−0.023	−0.031
	(0.059)	(−0.593)	(0.319)	(0.360)	(0.882)	(0.685)	(−0.852)	(−0.650)	(−0.691)	(−0.743)	(−0.541)	(−0.750)
INV	0.807 **	0.809 **	0.597 *	−0.248 ***	−0.370 ***	−0.356 ***	−0.062	−0.136	−0.145	−0.012	−0.110	−0.011
	(2.248)	(2.370)	(1.715)	(−3.751)	(−4.369)	(−4.152)	(−0.337)	(−0.730)	(−0.766)	(−0.080)	(−0.698)	(−0.073)
EI	0.886 ***	1.277 ***	0.636 **	0.454 ***	0.343 ***	0.378 ***	0.207	0.132	0.144	0.046	−0.011	0.071
	(3.213)	(4.845)	(2.352)	(8.927)	(5.242)	(5.688)	(1.457)	(0.918)	(0.984)	(0.396)	(−0.089)	(0.589)
POP	−0.110 ***	−0.121 ***	−0.172 ***	0.021 ***	0.014 ***	0.017 ***	0.084 ***	0.079 ***	0.078 ***	0.041 ***	0.034 ***	0.051 ***
	(−6.039)	(−6.910)	(−8.349)	(6.362)	(3.270)	(3.447)	(8.878)	(8.258)	(7.029)	(5.229)	(4.238)	(5.579)
ES	−0.010 ***			0.007 ***			0.004 ***			0.005 ***
	(−4.308)			(15.076)			(3.432)			(4.502)
SI		0.081 ***			−0.004			−0.004			0.002
		(7.202)			(−1.419)			(−0.620)			(0.435)
RD			1.034 ***			−0.060			−0.002			−0.229 ***
			(6.774)			(−1.612)			(−0.028)			(−3.364)
Constant	11.042 ***	7.568 ***	−1.082	3.272 ***	4.482 ***	5.005 ***	−1.436 *	−0.638	−0.697	6.361 ***	7.093 ***	9.437 ***
	(7.427)	(5.446)	(−0.525)	(11.934)	(13.009)	(9.875)	(−1.874)	(−0.839)	(−0.622)	(10.067)	(11.168)	(10.272)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	390	390	390	390	390	390	390	390	390	387	387	387
R-squared	0.968	0.971	0.971	0.987	0.979	0.979	0.960	0.958	0.958	0.960	0.958	0.959

Standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

showed that the regression coefficients of policy dummy variables for each dependent variable were all significant, significantly positive for the economy indicator (pGRP), and significantly negative for pollution emissions indicators. In addition, the coefficients of the mediating variable “ES” for each dependent variable were significant, demonstrating “ES” as a mediating variable. The results indicated that the CET pilot policy could reduce the total carbon emissions, industrial SO₂ emissions and solid wastes production by optimizing energy structures, which partly confirms Hypothesis 2. That is to say, the CET pilot policy could decrease the proportion of coal consumption in total energy consumption to reduce carbon emissions, industrial SO₂ emissions and solid wastes production. Some studies have the same conclusions [42]. As to another mediating variable, such as “RD”, the regressions coefficients is significantly positive for pGRP, and significantly negative for SW, showing “RD” as a mediating variable. The results demonstrated that the CET pilot policy could improve economic development and reduce industrial solid wastes production by increasing technological investment, which verifies Hypothesis 5 and partly confirms Hypothesis 3. However, the introduction of the mediating variable “SI” did not yield significant regression coefficients for the dependent variables of carbon emissions, SO₂ emissions and solid wastes production, except economy development. The results indicated that the share of secondary industry value added in GRP had a significantly positive effect on regional economy, and the CET pilot policy could improve economic development by increasing the proportion of the secondary industry. A study by Fu (2024) [42] also found that “SI” did not yield significant regression coefficients for the dependent variables of pollution indicators. Some other studies found that the carbon trading policy adjusted the proportion of heavily polluting enterprises among all enterprises, which further affects carbon emissions and air pollution in the power, industry, transport and resident sectors [23].

4.4. Heterogeneity Analysis

Due to huge regional differences in China, the effect of a carbon trading policy is expected to be different in different parts of the country. We re-performed the DID analysis based on the different areas (eastern, central and western areas) in China.

Table 7. Heterogeneity analysis in east–central–west regions.

Variables	pGRP			lnCEs			lnSO2			lnSW
	East	Central	West	East	Central	West	East	Central	West	East	Central	West
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
${treat}_i\times {time}_t$	−0.124	0.932 ***	1.101 ***	−0.062	−0.247 ***	−0.515 ***	0.183	−0.204	−0.502 ***	−0.089	0.058	−0.693 ***
	(0.332)	(0.162)	(0.168)	(0.050)	(0.062)	(0.099)	(0.156)	(0.135)	(0.120)	(0.099)	(0.083)	(0.185)
lnFDI	−0.321 **	0.086	0.034	−0.034	−0.064	0.103	0.008	0.227 **	0.063	0.038	−0.255 ***	−0.078
	(0.156)	(0.121)	(0.113)	(0.023)	(0.047)	(0.066)	(0.074)	(0.101)	(0.080)	(0.046)	(0.062)	(0.124)
INV	1.591	−0.544	0.540 **	−0.546 ***	0.100	−0.355 ***	0.644	0.282	0.014	0.469	−0.122	0.084
	(1.165)	(0.507)	(0.223)	(0.174)	(0.194)	(0.131)	(0.550)	(0.421)	(0.158)	(0.376)	(0.259)	(0.245)
EI	6.394 ***	−0.818 **	0.424 **	0.487 ***	0.482 ***	0.394 ***	0.605	1.008 ***	0.009	−0.304	0.213	0.193
	(1.211)	(0.342)	(0.200)	(0.181)	(0.131)	(0.118)	(0.571)	(0.283)	(0.142)	(0.359)	(0.174)	(0.222)
POP	−0.179 ***	0.154 ***	0.100 **	0.027 ***	0.011	−0.016	0.166 ***	−0.067 ***	−0.063 **	0.045 ***	0.005	−0.032
	(0.040)	(0.021)	(0.043)	(0.006)	(0.008)	(0.025)	(0.019)	(0.018)	(0.030)	(0.012)	(0.011)	(0.048)
Constant	18.231 ***	−4.720 ***	−2.424	3.946 ***	5.519 ***	4.832 ***	−8.568 ***	4.045 ***	6.105 ***	5.064 ***	11.612 ***	11.060 ***
	(3.503)	(1.511)	(1.891)	(0.523)	(0.579)	(1.112)	(1.652)	(1.253)	(1.346)	(1.035)	(0.771)	(2.112)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	143	117	130	143	117	130	143	117	130	142	117	128
R-squared	0.974	0.977	0.974	0.992	0.970	0.972	0.976	0.947	0.965	0.985	0.963	0.889

Standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

shows the regression results.

For the economic development results (Column (1)–(3)), the ${treat}_i\times {time}_t$ coefficients are significantly positive in Column (2)–(3), which indicates that the carbon emissions trading policy could promote economic development in central areas and western areas, but there is no significant effect on eastern areas. As to total carbon emissions results, the coefficients of policy dummy variables are significantly negative in central areas (−0.247) and western areas (−0.515), which means that the carbon trading policy could be effective in central and western areas. Moreover, the magnitude of ${treat}_i\times {time}_t$ in the western area is about 2.1 times higher than that in the central areas. This result suggests that the carbon emissions trading policy reduces the carbon emissions in the western area the most. Plus, there is no significant effect in the eastern areas. Some studies have the same results [48]. Columns (7)–(12) demonstrate the effects of the carbon emissions trading policy on industrial SO₂ emissions and solids wastes, respectively. The coefficients of the interaction term are only significant in western areas, indicating that the effect of the carbon emissions trading policy on the SO₂ emissions and solids wastes discharge are mainly due to its effectiveness in western China. The above results show that the effects of carbon emissions trading are most effective in the west areas. This may due to resource endowments. The distribution of coal resources in China presents the characteristics of “rich in the north and poor in the south, with more in the west and less in the east”, and the coal-producing areas are mainly in the central and western areas. The industrial structure in some central and western regions heavily relies on fossil fuels, emitting higher pollution. Therefore, these areas have more reduction potentiality. Then, the policy effects are more pronounced.

Moreover, due to the different economic structures and energy consumption structures, the effect of a carbon emissions trading policy on northern and southern areas in China may be different. This paper also categorized the pilot regions into northern areas and southern areas based on the Qinling–Huaihe Line to further explore the heterogeneity. The regression results are shown in

Table 8. Heterogeneity analysis in north–south regions.

Variables	pGRP		lnCEs		lnSO2		lnSW
	North	South	North	South	North	South	North	South
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
${treat}_i\times {time}_t$	1.406 ***	0.439 **	−0.142 *	−0.254 ***	−0.695 ***	−0.293 ***	−0.247 *	−0.148 **
	(0.267)	(0.188)	(0.080)	(0.034)	(0.161)	(0.096)	(0.144)	(0.058)
lnFDI	0.503 ***	−0.337 ***	−0.032	0.024	−0.085	0.056	−0.156 *	0.062
	(0.165)	(0.123)	(0.049)	(0.022)	(0.100)	(0.063)	(0.089)	(0.038)
INV	0.648 *	0.750	−0.310 ***	−0.709 ***	−0.368 *	0.666	−0.004	0.131
	(0.359)	(1.132)	(0.108)	(0.206)	(0.217)	(0.579)	(0.197)	(0.348)
EI	0.361	0.361	0.485 ***	0.270 ***	−0.046	0.538 **	0.064	0.191
	(0.354)	(0.445)	(0.106)	(0.081)	(0.214)	(0.228)	(0.191)	(0.137)
POP	−0.124 ***	−0.115 ***	0.042 ***	−0.009 **	0.052 ***	0.106 ***	0.079 ***	0.015 *
	(0.026)	(0.026)	(0.008)	(0.005)	(0.016)	(0.013)	(0.014)	(0.008)
Constant	5.687 ***	15.061 ***	3.136 ***	5.615 ***	1.528	−3.568 ***	6.022 ***	6.970 ***
	(1.962)	(2.162)	(0.587)	(0.394)	(1.182)	(1.106)	(1.058)	(0.665)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	195	195	195	195	195	195	192	195
R-squared	0.972	0.975	0.981	0.980	0.961	0.962	0.958	0.977

Standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

. The regression coefficients of the policy dummy variables for economy are significantly positive in both the northern and southern areas. As to every pollutant, the coefficients are significantly negative, indicating that the CET pilot policy has induced pollution reduction effects in both northern and southern areas. Meanwhile, the regression coefficients for pGRP, SO₂ and solid wastes production in the northern areas were 1.406, −0.695 and −0.247, which were both higher than those in the southern areas (0.439, −0.293 and −0.148). However, as to the regression coefficients for CO₂, the south (−0.254) is higher than the north (−0.142). The higher effects of emissions reduction in the northern areas may be due to the following reasons. First, economy development in the northern areas mainly relies on the heavy and resource-based industries. Second, the energy consumptions rely more on traditional energy sources, with a greater emphasis on coal, petroleum and natural gas. As a result, the pollutants emissions are relatively higher, and the policy effects are more pronounced, leading to more significant pollution reduction effects in the northern areas.

5. Conclusions and Policy Implications

The Chinese government has implemented several policies to reduce carbon emissions and develop a low-carbon economy. However, the effectiveness of these policies is still uncertain due to the potential adverse effect on the economy and inconclusive effect on carbon emissions. Therefore, this study tries to explore the effect of the carbon emissions trading policy issued in 2013 on the reduction of carbon emissions, and co-benefits of other pollutants, and more importantly, to discuss its effects on regional economic development. Our findings are as follows.

First, the carbon emissions trading policy has significant reduction effects on carbon emissions, industrial SO₂ emissions and solid wastes production, but has no significant reduction effects on industrial COD discharge. This may because the industries targeted by the carbon trading policy are those discharging heavy waste gases and solid wastes, while paying little attention to the industries with large COD emissions such as papermaking, pharmaceutical and food processing. In addition to a direct reduction effect, the carbon emissions trading policy could exert an indirect effect on carbon emissions and air quality by adjusting the energy consumption structure, and reduce industrial solid wastes by improving energy consumption structures and increasing technology investment.

Second, the carbon emissions trading policy has a significantly direct promotion effect on regional economic development in the pilot areas, and this result verifies the “Porter hypothesis”. The policy could also promote regional economic development indirectly by optimizing energy consumption structures, industrial structures and increasing technological investment.

Third, the pollution reductions and economic development effects show significant regional differences. The carbon emissions trading policy is more effective in the western areas than the central and eastern areas. As to the north and south regions, the pollution reduction effects of the pilot policy in northern regions are greater than those in southern regions. This may be because economic development in the northern areas relies more on heavy industries, and takes coal as the main energy in the energy consumption structure.

Based on the above findings, this study has some policy implications. First, the carbon emissions trading policy could indeed reduce total carbon emissions, and bring co-benefits of other pollutants abatement. However, there exists a regional difference in the reduction effectiveness. Therefore, establishing a national carbon market should consider regional characteristics and a negative spillover effect. Due to the higher reduction potentiality in western areas, and higher emissions levels in northern regions, policymakers should consider implementing stricter policies to achieve better effects. Second, the Chinese government should pay attention to optimizing the energy consumption structure and increase investments for researching and promoting the development of green and low-carbon technologies. Third, the coverage of national carbon markets could be expanded to include more industries, such as papermaking, pharmaceutical and food processing industries, to reduce water pollution synergistically.

This paper still has some limitations. Firstly, due to data availability, provincial-level panel data were utilized for analysis. However, the application of data at the municipal level and firm level would yield more comprehensive results. For example, carbon leakage among cities or firms which may result in more carbon and pollution emissions in the neighboring cities. The results are very important for the design and implementation of a national carbon emission trading policy. Secondly, in terms of research methods, although this paper has considered possible control variables, and made robust tests, there remains the possibility of unobserved confounding factors that could affect the regression results. For example, the effect of other policies indirectly on carbon trading policy. Thirdly, this paper focuses on single pilot trading without considering the cross-regional transactions, carbon leakage or market liquidity.