Analysis of the Impact of Carbon Trading Policies on Carbon Emission and Carbon Emission Efficiency

Abstract

As the carbon trading scheme has a significant impact on China’s sustainable economy and environmental protection, the policy influence of carbon emissions and carbon emission efficiency in pilot provinces has become a key research topic. Based on the data of 30 provinces and cities in China from 2007 to 2018, this paper estimates carbon emission efficiency by using a super-efficiency SBM model, and the difference-in-difference method is adopted to investigate the policy’s influence. The results show that: (1) carbon trading policies have a significant carbon emission reduction effect and a positive effect on carbon emission efficiency in pilot areas. (2) There is a dynamic effect that increases year by year, and the policies have a synergistic emission reduction effect on CO₂ and SO₂. (3) The carbon trading policy has different effects on carbon emission efficiency depending on pilot areas. Before and after the implementation of the policy, carbon emission efficiency in Tianjin remained almost unchanged, while the carbon emission efficiency in Hubei and Chongqing increased significantly. Although the efficiency of Shanghai and Guangdong remains at the forefront, they fluctuate greatly. Beijing is the only city to remain a frontier every year, showing significant policy impact.

1. Introduction

High concentrations of carbon dioxide have led to many environmental problems. The global average temperature in 2019 increased by 1.1 °C compared with that before industrialization [1]. Ocean warming and two-stage glacier melting have led to sea level rise, floods, droughts and hurricanes. Controlling the continuous increase in greenhouse gases such as carbon dioxide has become an important problem for all countries. Compared to developed countries, developing counties have more pressure on carbon dioxide emission reduction, especially China. In recent decades, China’s economy has grown at an incredible speed, which has a close relationship with the use of non-green energy [2]. By 2012, China was consuming nearly as much coal as the rest of world, contributing to more than 10% of China’s share of global CO₂ emission from 2000 to 2012 [3]. Therefore, China has accepted the great responsibility of reducing CO₂ emissions. In addition to signing the Kyoto Protocol and the Paris Agreement, China also pledged to reach its carbon peak by 2030 and become carbon neutral by 2060 in the seventy-fifth session of the United Nations.

Based on this background, China has introduced many policies related to carbon emission reduction, among which the most important is the implementation of the carbon trading policy. The carbon trading market has been proven to be an effective way to reduce carbon emissions, among which the EU’s carbon emission trading system is the most successful one [4]. After three stages of development, the EU’s carbon emission trading system has become the most mature carbon emission trading market in the world, and includes electricity, petroleum, the chemical industry, aerospace [5], and other main industries. Learning from the European Union, China has launched its own pilot carbon trading market [6,7]. In October 2011, the Chinese National Development and Reform Commission issued the “Notice on Carrying out the Pilot Work of Carbon Emission Trading” in preparation for the establishment of a carbon trading market. From June 2013 to April 2014, carbon trading pilot markets in seven locations, including Beijing and Tianjin, were established and are now open for trading. In 2020, the cumulative transaction volume of carbon dioxide in each pilot market reached 445 million tons, and the transaction volume exceeded CNY 10.431 billion. By the end of 2017, China further promulgated the “National Carbon Emission Trading Market Construction Plan (Power Generation Industry)”, and in 2021, the national carbon emission trading market was established.

Obviously, CO₂ emission is the most direct indicator to measure whether the market is working well or not [8]. Reducing carbon emissions as soon as possible has become an important environmental goal in China, but blindly reducing production and emission reduction will also cause obstacles to China’s economic development. As early as 2009, at the United Nations Climate Change Conference, China acknowledged that although reducing carbon emissions to prevent global warming is an urgent matter, social and economic development is still the primary goal for developing countries such as China. In the report of the 19th National Congress of the Communist Party of China, China also stressed that Chinese economic development should first adhere to quality and promote the reform of efficiency. Hence, the government should also consider sustainable development while making policies and a regular index to measure carbon emissions efficiency (CEE) [9,10].

There are two opposing theories about the effect of environmental regulation. The traditional one is that environmental regulation often comes with the cost of reduced economic growth, which has been evidenced in China [11]. On the other hand, the Porter hypothesis believes that strict environmental regulation can stimulate the improvement of enterprise efficiency and technology innovation [12]. Therefore, it is crucial to study the impact of the carbon trading market on CEE. This paper combines both CO₂ emissions and CEE as a policy evaluation index, which is a comprehensive way to analyze policy effect.

This article is divided into five parts. First, this paper summarizes the related literature about the carbon trading market. Second, based on the background of the carbon market and economy, four theoretical research hypotheses are proposed. Third, this paper describes relevant data and models. The empirical analysis in the fourth part is divided into CEE analysis, regression results, a parallel trend hypothesis test and dynamic effect analysis, placebo inspection, and a collaborative emission reduction test. Corresponding policy suggestions are put forward according to the empirical results in the last part.

2. Literature Review

Many countries have established their own carbon trading markets since the Kyoto Protocol came into force, which could be meaningful to China’s market. The EU’s carbon emission trading system is the largest one [4], and was established around 2005. As the oldest carbon market, the EU’s market has a mature structure and influence on energy industry [13]. Compared to the EU, China’s carbon market started relatively late. In the pilot stage, China’s carbon market system was not perfect, resulting in a low carbon price which was not attractive to enterprises [14], with little impact on other markets. For China, the Australian carbon market is relatively referential, since Australia has a similar energy structure to China, using coal as the main energy source [15]. The Australian carbon trading market has a mature legal foundation and has had a great impact on energy use [16]. China’s pilot carbon market is similar to the American California carbon trading market to some extent. During the initial period, both markets were dominated by carbon quotas and supplemented by disposals, which were also regional markets. Although both markets started closely, it is clear that the California carbon trading market is more efficient and has more of an impact on energy cost [17].

Nowadays, carbon market consolidation among regions has become a trend, such as the American California carbon trading market and the Canadian Quebec carbon trading market. China’s carbon market started relatively late and only the electric power industry has been included in the national market. Lagging behind most carbon markets around the board, it will be difficult for China to integrate into any international market. Therefore, it is crucial to study the effect of pilot markets and improve the national market with this experience.

Since the establishment of the pilot carbon trading market in China, many scholars have conducted relevant research. Based on two different perspectives, they have analyzed the operation mechanism and influence of the carbon trading market. Regarding the impact of the carbon market, Hu et al. [18] used the difference-in-difference method to explore the energy-saving and emission-reducing effects of carbon trading policies. Research shows that carbon trading policies have reduced the energy consumption of regulated industries in pilot areas by 22.8% and CO₂ emissions by 15.5%. From a new perspective, Wen et al. [19] found that implementing carbon trading policies can increase the excess returns of enterprises participating in the carbon market by using difference in difference (DID). Actually, because the establishment of the carbon market can be considered a quasi-natural experiment, DID is the most common model that has been used in the studies of carbon trading policy [20,21]. Despite a nationwide emission trading scheme being more efficient, using a computable general equilibrium (CGE) model, Lin Boqiang et al. [22] predicted that this would reduce welfare in less developed areas. Wen Hong-Xing et al. [23] used the synthetic control method, which is often used to evaluate policies, to show that the carbon trading policy would restrain economic development, but this effect will decrease annually. Yang et al. [24] explored the impact of the carbon trading market on carbon emission reduction from the perspective of employment scale. Zhang Wei et al. [25] used the data envelopment method (DEA) to evaluate the operating efficiency of the carbon trading market and found it has increased annually. Yaxue Yan et al. [26] focused on the collaborative effect of carbon policy and concluded that there is a significant ‘reduction effect’ on the haze pollution concentration level.

From another angle, many scholars have focused on the operation mechanism of the carbon market. Regarding the carbon price, Chang-Jing Ji et al. [27] thought that carbon prices are low and have been fluctuating greatly since Chinese carbon emissions trading scheme pilots started operating. Using the DEA model, Zhenling Chen et al. [14] conducted research on the national carbon trading market, and found that the main reason for market inefficiency was the low carbon price. Zhang et al. [28] reached a similar conclusion as Zhenling Chen and suggested that imperfect laws and regulations were also a main reason for market inefficiency. Dong et al. [29] proposed a fair and efficient initial carbon allocation method aiming at the problem of regional differences.

Carbon emission efficiency is a crucial indicator to measure the balance of economic development and environment protections. In the early days, scholars always used the single-factor method [30], which cannot reflect accurate information compared to the total factors method. Now, when multiple inputs and outputs are taken into consideration, SFA (Stochastic Frontier Approach) [31] and DEA are the two main methods used to calculate carbon emission efficiency. Using DEA and a random forest regression model, Feng Dong et al. [32] discussed the impact of renewable energy development on regional carbon emission efficiency. The main difference between SFA and DEA is the adoption of a parameterized method. Compared to SFA, DEA is a nonparametric method that is relatively flexible [33]. It does not need any specific function or parameter and can also be applied with small samples. Therefore, DEA is more suitable for calculating CEE. Based on its basic idea, many models have been created to applied to different occasions, such as CCR, BCC, SBM, super-efficiency SBM, and the Malmquist Index. Among many DEA models, super-efficiency SBM has more advantages than others in calculating the efficiency of undesirable outputs [34]. Du Qiang [35] used it to explore the carbon emission efficiency of the construction industry. Carbon mission efficiency is always used to evaluate an environmental policy. Yu et al. [36] studied the impact of a low-carbon city pilot policy based on carbon emission efficiency. In contrast, Zhang et al. [37] showed that there was a significant regional differentiation of agricultural carbon emissions among the provinces along the belt and road.

In summary, the literature has various perspectives from a research perspective, but most of them are focused on the effectiveness of carbon emission reduction. In terms of research methods, CGE, the synthetic control method, DID, SFA, and DEA are the main methods used. In terms of research content, most scholars have only explored whether carbon emission reductions have been achieved after the implementation of the carbon trading policy in 2013. Only several scholars have added a discussion of synergistic emission reductions or conducted a dynamic analysis after the implementation of a carbon trading policy. Most scholars have focused on the change and causes of carbon efficiency in a certain area or a certain industry, but few scholars have compared the impact of the carbon trading market on carbon emission and CEE.

From the perspective of exploring the impact of the pilot market, this paper evaluates the impact of the carbon market considering two aspects: the effectiveness of carbon emission reduction and the efficiency of carbon emissions. First, this paper calculates carbon emission efficiency by using the super-efficiency SBM, and adopts a difference-in-differences model to analyze the influence of carbon trading policy on carbon emissions and carbon efficiency in the pilot area. Meanwhile, this paper also adds the analysis of the dynamic effect and the sulfur dioxide synergistic emission reduction. Analyzing the effectiveness of the carbon market from two perspectives will reach more comprehensive analysis results and provide reference for improving the national carbon trading market.

3. Theoretical Framing and Research Hypothesis

The carbon trading market is an environmental policy that uses market regulations to reduce carbon emissions. Its specific emission reduction principles can be divided into three aspects. First, the carbon trading market system realizes carbon emission reduction through clear property rights. Since air resources are public resources, their property rights are not clear. According to the Coase theorem, benefits can be maximized after the property rights of carbon dioxide; other gases are clearly defined and their pollution rights are traded in the market as commodities [38]. In China’s carbon trading market, total carbon emissions are first fixed, and then the amount of carbon emissions of each enterprise is specified according to historical emissions and industry baselines. Enterprises can trade their carbon emissions in the carbon trading market. If their emissions exceed their carbon quota by the end of the year, they will be punished according to local regulations. For example, Beijing will impose a fine of 3–5 times the market price. The mode of production in each industry is vastly different, which makes the cost of reducing emissions in some enterprises high and others relatively low. For energy-intensive industries, the marginal cost is often higher than the market price of carbon. Through the carbon trading market, enterprises can buy carbon quotas to maximize profits. However, many enterprises whose marginal cost of emission reduction is lower can sell carbon emissions to increase profits. In this case, the carbon market is consistent with the concept of sustainable development, which will reduce CO₂ emission and improve CEE theoretically.

Second, carbon markets can promote technological progress to reduce emissions [39,40]. As early as in the 1990s, Porter proposed the hypothesis that environmental regulation would stimulate enterprise technological innovation [12]. In the carbon market, CO₂ emission quota is a scarce resource which is taken into account in the production plan. Due to the existence of market incentives, enterprises are more inclined to use green and low-carbon technologies to achieve emission reduction by optimizing the energy structure or improving the energy utilization rate. There are multiple ways to carry out technology innovation, such as improving productive processes and adopting green energy. For example, in the electric power industry, many enterprises have begun to use natural gas for power generation to replace original coal technology. Through technology innovation, enterprises will constantly reduce CO₂ emission and improve resource utilization rate (improving CEE) to reach profit maximization.

Third, carbon markets will also promote industrial upgrading to reduce carbon emissions [41]. In the short term, the profits of energy-intensive industries would be squeezed as they have to pay high carbon costs. Investment and production will incline to green industries. In the long run, the reallocation of resources will cause industrial transfer. Highly polluting enterprises will switch to new green enterprises through technological innovation, and the rest will be gradually eliminated by the market. Meanwhile, technology innovation will also reduce the cost of industrial upgrading. Therefore, the proportion of high energy consumption industries will gradually decrease.

Based on theoretical analysis, carbon trading policy plays an important role in emission reduction. At the same time, compared with mandatory environmental regulations, carbon trading policy has less of an impact on production overall and can reduce the impact of carbon emission reduction on economic development to the greatest extent. Carbon emission efficiency is an important indicator to judge whether low-carbon environmental protection and economic development can be achieved simultaneously. In this paper, carbon emission efficiency is taken as an evaluation indicator of whether provinces and cities actively develop their economies while reducing carbon emissions, and the following assumptions are proposed:

Hypothesis 1.

Carbon trading policy will effectively reduce carbon emissions in the pilot area.

Hypothesis 2.

The carbon trading market will improve carbon emission efficiency in the pilot area.

To realize the above theory, a mature carbon trading market is first required. In the first several years, carbon market do not work well because of too much carbon quotas and ambiguous laws [42]. As more enterprises are included, the carbon quota system and the review system (MRV) are gradually improved. Meanwhile, the carbon price gradually tends to become stable, and trading volume and turnover are increased year by year. Generally, a pilot market system for carbon trading gradually takes shape [43]. While the pilot market is becoming mature, the carbon emission reduction effect of the pilot market should be gradually strengthened. At the end of 2017, China started to prepare for the national carbon trading market of the electric power industry. As other regions begin to implement carbon market policies, the emission reduction effect of the pilot provinces and cities would be relatively weakened. Therefore, this paper puts forward the following hypothesis:

Hypothesis 3.

The carbon trading pilot policy has a cumulative dynamic effect on carbon emission reduction and carbon emission efficiency. After the implementation of the national carbon trading policy at the end of 2017, the cumulative effect of the experimental group is relatively weaker.

From a theoretical point of view, two atmospheric pollutants, carbon dioxide and sulfur dioxide, have the same source. The burning of most kinds of fossil energy, especially coal, is accompanied by a large amount of sulfur dioxide emissions [44]. If the carbon trading mechanism can increase the use of clean energy and reduce the burning of coal, it will also reduce sulfur dioxide in a synergistic manner, which means reducing sulfur dioxide emissions while reducing carbon emissions [26]. Based on this theoretical analysis, the following hypothesis is put forward:

Hypothesis 4.

Carbon trading market policy will have a synergistic emission reduction effect on sulfur dioxide.

4. Models and Data

4.1. Models

This paper adopts two models to analyze the policy effect of carbon trading policy. The super-efficiency SBM is adopted to calculate the carbon emission efficiency. As one of the most effective models for evaluating policy effects, the difference-in-difference method is used in this paper to analyze whether the policy has improved the carbon emission efficiency and the effectiveness of carbon emission reduction in pilot areas.

4.1.1. Difference-in-Difference Method

This paper regards the implementation of carbon trading policy as a ‘quasi-natural experiment’, taking seven pilot provinces and cities as the treatment group and the other nonpilot provinces and cities (with the exception of Tibet, Hong Kong, Macao and Taiwan) as the control group. It should be noted that because Shenzhen is prefecture-level, Shenzhen is incorporated into Guangdong to ensure the consistency of data hierarchy and scope [45]. The basic principle of the difference-in-difference method is to eliminate the changes caused by time and individual factors using two subtractions, and obtain the changes caused by policy reasons, which eliminates the interference of systematic errors and facilitates effective analysis. Based on a study by Liu Guixian et al. [46], this paper establishes the model as follows:

$$lnC{O}_{2it}={\alpha }_0+{\alpha }_1{P}_t+{\alpha }_{2}trea{t}_i+{\alpha }_3{P}_t\times trea{t}_i+{\alpha }_i{{\displaystyle \sum }}_i^h{X}_i+{\xi }_{it}$$

(1)

$${\delta }^{\ast }{}_{it}={\beta }_0+{\beta }_1{P}_t+{\beta }_{2}trea{t}_i+{\beta }_3{P}_t\times trea{t}_i+{\beta }_i{{\displaystyle \sum }}_i^h{X}_i+{\xi }_{it}$$

(2)

where i represents provinces and t represents years. Regarding the explained variable, $lnC{O}_{2it}$ and ${\delta }^{\ast }{}_{it}$, are the logarithms of the CO₂ emissions and the carbon emission efficiency, respectively. ${\alpha }_0$ and ${\beta }_0$ are the constant terms and $trea{t}_i$ is the policy dummy variable. $trea{t}_i$ equals 1 if the seven provinces or cities are covered by the policy; otherwise, it equals 0. $P_t$ is the time dummy variable, which equals 1 after 2013; otherwise, it equals 0. The interaction term, $P_t\times trea{t}_i$, represents whether region i in year t implements the policy.${\alpha }_{1 }$ and ${\beta }_1$ represent the difference in the explained variable before and after the implementation of carbon trading in the nonpilot area. ${\alpha }_2$ and ${\beta }_2$ represent the time-fixed effect. ${\alpha }_{3 }$ and ${\beta }_3$ are the interaction term coefficients which represent the net effect of carbon policy. $X_i$ is a group of control variables, and ${\xi }_{it}$ is a random error term.

4.1.2. Super-Efficiency SBM Model

This paper uses the super-efficiency SBM to calculate the carbon emission efficiency of each province. The super-efficiency SBM model, proposed by Tone [34], is the combination of the super-efficiency DEA and SBM models. In traditional DEA models, such as CCR and BCC, there are usually multiple effective decision units, which means there are multiple cases where the efficiency is 1. To better distinguish these decision units, the super-efficiency DEA model replaces effective decision units with the linear combination of other decision units. Super-efficiency SBM adds slack variables based on the super-efficiency DEA model so that it can better distinguish different decision units; therefore, it is widely used in the calculation of unexpected output efficiency. Referring to the study of Tang Kai et al. [47], this paper added the undesirable outputs based on the original super-efficiency SBM model proposed by Tone. Assuming that a province is a decision unit, the input factor of each decision unit is m, the expected output factor is S₁, and the unexpected output factor is S₂. The specific model is as follows:

$${\delta }^{\ast }=min\delta =\frac{\frac{1}{m}{{\displaystyle \sum }}_{i=1}^m\overline{x}/x_{ik}}{\frac{1}{S_1+S_2}\left\{{{\displaystyle \sum }}_{r=1}^{S_1}{\overline{y}}_r/y_{rk}+{{\displaystyle \sum }}_{p=1}^{S_2}{\overline{y}}_p/y_{pk}\right\}}$$

Subject to

$$\overline{x}⩾{{\displaystyle \sum }}_{j=1,j\ne k}^n{\lambda }_j{x}_{ij},i=1,2,\dots ,m$$

$$\begin{gathered} {\overline{y}}_r⩽{{\displaystyle \sum }}_{j=1,j\ne k}^n{\lambda }_j{y}_{rj},r=1,2,\dots ,S_1 \\ {\overline{y}}_p⩾{{\displaystyle \sum }}_{j=1,j\ne k}^n{\lambda }_j{y}_{pj},p=1,2,\dots ,S_2, \overline{x}⩾x_{ik},i=1,2,\dots ,m \\ {\overline{y}}_r⩽y_{rk},r=1,2,\dots ,S_1 \\ {\overline{y}}_p⩾y_{pk},p=1,2,\dots ,S_2 \\ {\lambda }_j,{\overline{y}}_{rk},{\overline{y}}_{pk}⩾0,j=1,2,\dots ,n,j\ne k \end{gathered}$$

(3)

where ${\delta }^{\ast }$ is carbon emission efficiency, n is the provinces’ number, m is the number of input variables, and S₁ and S₂ represent the number of desirable outputs and undesirable outputs, respectively. x is input factor, $y_{rk}$ is desirable output, and $y_{pk}$ is undesirable output. k indicates the evaluated decision units.

4.2. Data

This paper collects relevant data from 30 provinces and cities in China (excluding Tibet, Hong Kong, Macao, and Taiwan) from 2007 to 2018. The original data come from the China Statistical Yearbook and the China Energy Statistics Yearbook.

4.2.1. Carbon Dioxide Emissions and Sulfur Dioxide Emissions

For Model (1), this paper takes the logarithms of the carbon dioxide emissions of each province as the explained variable. Because there are no official data on carbon dioxide emissions, this paper collects the energy balance tables of provinces and cities from the China Statistical Yearbook. According to the calculation method in the ‘2006 IPCC Guidelines for National Greenhouse Gas Inventories’, this paper estimates the carbon dioxide emissions of each province from 2007 to 2018 based on the energy balance sheet. At the same time, to verify the synergistic emission reduction effect, this paper also uses the provincial energy balance table and ‘2006 IPCC Guidelines for National Greenhouse Gas Inventories’ to estimate the panel data of sulfur dioxide emissions from 2007 to 2018.

4.2.2. Control Variables

(1) Population size (lnpop): Population size increase contributes to carbon dioxide emissions. In this paper, the logarithm of the year-end population in the China Statistical Yearbook is taken as population size data.

(2) Economic growth (lnpgdp): Economic growth is also one of the factors for the carbon emission increase [48]. In this paper, the logarithm of GDP per capita taken from the China Statistical Yearbook represents economic growth [49].

(3) Industrial structure (IS): According to Tang Kai et al. [50], changes in industrial structure will affect carbon emissions in Chinese cities. This paper uses the ratio of the added value of the tertiary industry and the added value of the secondary industry to represent the industrial structure.

(4) Technical level (RD): According to Chen Shuo et al. [51], green technology has a great influence on carbon emissions. The ratio of R&D investment to GDP represents technical level [49]. Data are taken from the China Statistical Yearbook.

(5) Urbanization level (urban): This paper uses the ratio of the total urban population and the total population of provinces and cities at the end of the year to express the level of urbanization [52]. Data are taken from the China Statistical Yearbook.

To test whether the relationship between carbon emissions and economic development conforms to the environmental Kuznets hypothesis, this paper also adds the quadratic term of per capita GDP (lnpgdp²) as a variable representing economic growth.

Table 1. Descriptive statistics.

Variables	(1)	(2)	(3)	(4)	(5)
	n	Mean	Std. Dev.	Min	Max
lnCO2	360	5.465	0.743	3.077	6.816
urban	360	54.69	13.37	28.24	89.60
lnpop	360	8.186	0.742	6.314	9.337
lnpgdp	360	10.55	0.562	8.841	11.85
IS	360	1.125	0.354	0.230	2.002
RD	360	0.00921	0.00519	0.000421	0.0219
lnpgdp2	360	111.6	11.82	78.17	140.4

shows descriptive statistical results for all variables.

4.2.3. Input and Output Variables

This paper chooses capital stock, labor force and energy input as input variables. Regarding output variables, carbon emissions and GDP are taken as the undesirable outputs and expected outputs, respectively.

Since there are no statistics on China’s provincial capital stock in the official yearbook, referring to the estimation method and estimated data of capital stock provided by Zhang Jun et al. [53], this paper uses the perpetual inventory method to estimate the capital stock from 2007 to 2018. The calculation formula is as follows:

$$K_{it}=K_{it-1}\left(1-{\delta }_{it}\right)+I_{it}$$

(4)

where i represents provinces and t represents years. K_it and I_it represent the capital stock and investment, respectively. δ_it represents the depreciation rate. According to the research of Zhang Jun et al. [53], the depreciation rate is 9.6%. The specific data are calculated using 2000 as the base year.

Labor force data are based on end-of-year employment numbers in the China Statistical Yearbook, while energy input is based on the total energy consumption in the China Energy Statistics Yearbook. Among the input variables, the capital stock and energy input from Tibet are considered missing data, so this paper does not consider the data from Tibet.

To calculate the efficiency of carbon emissions, this paper takes the carbon emissions of each province and city as undesirable outputs. For expected output, this paper uses the GDP from each region as economic output. Specific data are taken from the China Statistical Yearbook, using 2000 as the base year for reductions to ensure accuracy. As mentioned above, this paper merges Shenzhen into Guangdong.

5. Empirical Analysis

5.1. Provincial Carbon Emission Efficiency Analysis

In this paper, the super-efficiency SBM model was used to obtain the carbon emission efficiency of each province from 2007 to 2018. The specific results are shown in

Table 2. The carbon emission efficiency of each province from 2007 to 2018.

Region	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018
Beijing	1.015	1.001	1.030	1.036	1.102	1.057	1.076	1.184	1.523	1.062	1.215	1.221
Tianjing	0.793	0.736	0.782	0.803	0.781	0.759	0.832	0.763	0.713	0.641	0.680	0.639
Hebei	0.617	0.501	0.506	0.698	0.570	0.585	0.601	0.503	0.646	0.461	0.542	0.508
Shanxi	0.563	0.501	0.454	0.561	0.510	0.478	0.396	0.439	0.418	0.339	0.427	0.449
Inner Mongolia	0.647	0.630	0.656	0.649	0.665	0.558	0.531	0.567	0.507	0.419	0.409	0.453
Liaoning	0.652	0.584	0.588	0.777	0.659	0.639	0.609	0.633	0.657	0.426	0.491	0.524
Jilin	0.571	0.545	0.493	0.594	0.526	0.704	0.427	0.492	0.449	0.566	0.413	0.476
Heilongjiang	0.615	0.581	0.502	0.650	0.541	0.674	0.403	0.471	0.435	0.503	0.397	0.460
Shanghai	1.092	1.249	1.043	1.147	1.030	1.203	1.046	0.832	0.820	1.118	0.833	0.976
Jiangsu	0.844	0.762	0.690	1.131	0.796	1.258	1.100	0.834	1.146	1.090	1.013	1.336
Zhejiang	0.787	0.694	0.633	0.910	0.690	0.901	0.675	0.666	0.770	0.802	0.778	0.793
Anhui	0.474	0.438	0.425	0.604	0.467	0.651	0.418	0.456	0.497	0.532	0.490	0.524
Fujian	0.731	0.612	0.702	0.742	0.739	0.704	0.567	0.589	0.478	0.671	0.786	0.650
Jiangxi	0.558	0.465	0.518	0.622	0.595	0.670	0.417	0.491	0.420	0.566	0.602	0.510
Shandong	1.012	0.664	0.800	1.005	1.018	0.767	1.071	0.576	0.525	0.746	1.022	0.678
Henan	0.572	0.498	0.545	0.643	0.553	0.539	0.508	0.388	0.341	0.489	0.559	0.452
Hubei	0.493	0.469	0.522	0.643	0.582	0.608	0.490	0.505	0.416	0.578	0.639	0.528
Hunan	0.518	0.489	0.522	0.658	0.570	0.605	0.456	0.480	0.397	0.553	0.588	0.492
Guangdong	1.197	1.359	1.269	1.096	1.249	1.050	1.157	2.317	0.642	1.101	1.144	0.873
Guangxi	0.507	0.519	0.488	0.466	0.484	0.394	0.369	0.527	0.331	0.519	0.465	0.408
Hainan	1.939	1.268	1.395	1.197	1.042	1.231	0.315	1.339	1.717	1.498	1.200	1.748
Chongqing	0.489	0.489	0.559	0.502	0.591	0.487	0.438	0.663	0.447	0.665	0.614	0.489
Sichuan	0.480	0.445	0.500	0.480	0.532	0.466	0.481	0.558	0.406	0.575	0.600	0.488
Guizhou	0.371	0.369	0.415	0.400	0.434	0.403	0.342	0.490	0.358	0.464	0.454	0.361
Yunnan	0.356	0.418	0.414	0.390	0.480	0.362	0.332	0.424	0.271	0.403	0.625	0.277
Shanxi	0.416	0.489	0.493	0.464	0.599	0.442	0.455	0.556	0.358	0.497	0.787	0.401
Gansu	0.386	0.446	0.429	0.435	0.515	0.433	0.323	0.485	0.349	0.428	0.599	0.351
Qinghai	1.064	1.004	1.043	1.114	1.231	1.101	0.255	1.145	1.144	1.076	1.575	1.161
Ningxia	0.441	0.554	0.522	0.505	0.511	0.494	0.312	0.498	0.431	0.474	0.480	0.363
Xinjiang	0.422	0.491	0.475	0.449	0.546	0.412	0.378	0.492	0.300	0.358	0.474	0.297

. In Figure 1, the trend chart of the carbon emission efficiency of each pilot province is shown. Among the pilot provinces and cities, the efficiency value of Beijing is greater than 1 before and after the policy. After the implementation of carbon trading (after 2013), the carbon emission efficiency of Beijing shows a significant upwards trend. The operation of Beijing’s carbon market has a strong comprehensive capacity and good operation management, which ensure the policy’s effect [54]. The carbon emission efficiency in Guangdong and Shanghai fluctuates dramatically and behaves in an unstable manner, but basically remains a frontier. The carbon emission efficiency of Hubei Province is low and fluctuates at the level of 0.53, but it has a significant upwards trend after 2015. The main reason may be the problem of policy adaptability; efficiency has increased after adapting the policy [55]. At the same time, the economic impact of the carbon trading market in Hubei Province is relatively small, only 0.06% [56]. Hubei’s active market and stable carbon price have promoted the upgrading of its industrial structure [57] and improved its carbon emission efficiency. The carbon emission efficiency level of Tianjin was relatively stable from 2007 to 2018. Tianjin’s industry occupies a relatively high proportion in its industrial structure, and its labor productivity of the tertiary industry is relatively low. It is still in the initial stage of industrial structure upgrading, so the policy effect is not obvious [58,59]. The efficiency of Chongqing increased significantly after the implementation of carbon trading, but the overall efficiency was low, ranging from 0.44 to 0.67. Chongqing’s urbanization quality and coupling coordination degree are constantly improving [60], which promotes the improvement of efficiency. Meanwhile, the carbon market can reduce carbon emissions with a minor negative impact on the industrial economy [61]. However, Chongqing is still an area of low carbon efficiency. A large number of industrial enterprises from other provinces are migrating to Chongqing, which makes it difficult to transform to a green and low-carbon city [62]. The unreasonable industrial structure restricts the economic development and carbon efficiency improvement of Chongqing. To promote carbon efficiency as soon as possible, Chongqing should strengthen the construction of the carbon market and keep promoting industrial upgrading.

5.2. Basic Regression Results of the Difference-in-Difference Model

The regression results of carbon trading policies on carbon dioxide emissions and carbon emission efficiency in pilot provinces are shown in

Table 3. Regression results of carbon emission and carbon emission efficiency.

Variables	Carbon Emission			Carbon Emission Efficiency
	(1)	(2)	(3)	(4)	(5)
	No Control Variables	Part of the Control Variables	All Control Variables	No Control Variables	All Control Variables
Pt × treati	−0.185 ***	−0.143 ***	−0.191 ***	0.277 ***	0.115 **
	(−4.99)	(−4.99)	(−6.50)	(5.16)	(1.80)
lnpgdp		1.065 ***	1.675 ***
		(2.81)	(3.96)
lnpgdp2		−0.031 *	−0.060 ***
		(−1.66)	(−2.95)
urban			−0.004		0.860 ***
			(−0.97)		(8.71)
lnpop			0.865 ***		0.020
			(9.70)		(0.87)
IS			0.010		−0.129 **
			(0.28)		(−2.40)
RD			−1.175		−0.124 ***
			(−0.33)		(−3.36)
Constant	5.389 ***	−2.225	−12.287 ***	0.687 ***	−3.226 ***
	(35.66)	(−1.14)	(−5.32)	(13.06)	(−4.83)
Observations	360	360	360	360	360
Regional fixed effect	YES	YES	YES	YES	YES
Time fixed effect	YES	YES	YES	YES	YES
R-squared	0.413	0.705	0.733	0.091	0.375

Note: ***, **, and * represent significant levels at 1%, 5%, and 10%, respectively.

. To reflect the accuracy of the regression results, this paper carried out regressions with and without control variables. In the regression of carbon emission efficiency, economic development was not added as a control variable. Columns (1) and (4) show the regression results without adding any control variables, and show that the regression coefficients of $P_t\times trea{t}_i$ are all significant at the 1% level. There is a negative effect on carbon dioxide emissions and a positive effect on carbon emission efficiency, which preliminarily verifies Hypotheses 1 and 2. Column (2) shows the regression results after adding economic growth as a control variable. The coefficient of $P_t\times trea{t}_i$ was −0.143, which is significant at 1%. The linear term of per capita GDP (lnpgdp) and its quadratic term (lnpgdp²) are significant. The coefficient of the linear term is negative, and the coefficient of the quadratic term is positive, which are 1.065 and −0.031, respectively. It can be inferred that the relationship between carbon emissions and economic development conforms to the environmental Kuznets hypothesis, and the curve is ‘an inverted U’ shape, which agrees with the mainstream conclusions of existing research [63]. In Columns (3) and (5), all control variables are added. For carbon emissions, the significance does not change, and the absolute value of the coefficient increases. This shows that pilot provinces have a more significant carbon dioxide reduction effect than control provinces under the carbon trading market mechanism. At the same time, with regard to the control variables, except for economic growth, population size also shows a positive effect and a significant coefficient at the 1% level. This result indicates that the expansion of population size will increase carbon emissions, which is similar to the study of Wang Feng et al. [48]. For carbon emission efficiency, with the addition of control variables, the interaction term is significant at the level of 5%, and the coefficient is 0.115, which is slightly lower than that without the addition of control variables but does not affect the accuracy of the results. From the regression results, industrial upgrading and technological progress have a significant negative effect on carbon emission efficiency. However, for carbon emission reduction, they are not significant. This paper shows that the effect is not very stable and its credibility is low. Meanwhile, if a negative effect truly exists, it may be caused by the regional coupling mismatch between industrial upgrading and carbon emission efficiency [64], and the technological progress of production yield enhancement [65]. The above analysis confirms Hypothesis 1 and Hypothesis 2.

5.3. The Test of Parallel Trend Hypothesis and Dynamic Effect

An important premise of the difference-in-difference method is to meet the parallel trend hypothesis in that, before the policy shock, the control group and the treatment group should maintain the same trend. In this paper, the time trend graph is constructed to test the parallel trend hypothesis [52]. This paper plots the time diagram of two explained variables with the average of annual carbon dioxide emissions and carbon efficiency of the control group and the treatment group. Each explained variable is divided into a control group and a treatment group and the annual average value of the two groups is calculated. Then, the time trend charts of CEE and CO₂ are drawn and the specific results are shown in Figure 2 and Figure 3, respectively. As shown in Figure 2, from 2007 to 2012, the carbon emissions of the pilot area and the nonpilot area maintain the same growth trend. After 2013, due to the emission reduction effect of the carbon trading market policy, the carbon emissions of the pilot areas began to decline, while the nonpilot areas still maintained an upwards trend. In Figure 3, we can see that the carbon emission efficiency of the pilot and nonpilot areas fluctuated steadily before 2013. After the implementation of carbon trading, the experimental group had a more significant upwards trend than the control group. The above results indicate that carbon emissions and carbon efficiency passed the parallel trend hypothesis test.

To further verify the parallel trend hypothesis and analyze the dynamic effects of carbon trading policies, this paper successively changes the years of policy impact from 2008 to 2018. This paper modifies Models (1) and (2) to establish the following models:

$$lnC{O}_{2it}={\alpha }_0+{\alpha }_1{P}_t+{\alpha }_{2}trea{t}_i+{\alpha }_t{{\displaystyle \sum }}_{t=2008}^{2018}{P}_t\times trea{t}_i+{\alpha }_i{{\displaystyle \sum }}_i^h{X}_i+{\xi }_{it}$$

(5)

$${\delta }^{\ast }{}_{it}={\beta }_0+{\beta }_1{P}_t+{\beta }_{2}trea{t}_i+{\beta }_t{{\displaystyle \sum }}_{t=2008}^{2018}{P}_t\times trea{t}_i+{\beta }_i{{\displaystyle \sum }}_i^h{X}_i+{\xi }_{it}$$

(6)

The regression results of each year are shown in

Table 4. The test of parallel trend hypothesis and dynamic effect.

Variables	Parallel Trend Hypothesis		Dynamic Effect
	Carbon Emission	Carbon Emission Efficiency	Carbon Emission	Carbon Emission Efficiency
	(1)	(2)	(3)	(4)
treat × 2008	−0.155	0.012
	(−0.44)	(0.27)
treat × 2009	−0.163	0.010
	(−0.63)	(0.21)
treat × 2010	2010	0.010
	−0.169	(0.21)
treat × 2011	−0.180	0.021
	(−0.89)	(0.45)
treat × 2012	−0.185	0.024
	(−0.95)	(0.49)
treat × 2013			−0.191 ***	0.115 **
			(−6.50)	(1.80)
treat × 2014			−0.237 **	0.175 ***
			(−2.27)	(3.09)
treat × 2015			−0.240 **	0.131 **
			(−2.19)	(2.07)
treat × 2016			−0.243 **	0.138 *
			(−2.19)	(1.89)
treat × 2017			−0.236 *	0.190 **
			(−1.84)	(2.02)
treat × 2018			−0.187	0.191
			(−1.05)	(1.42)
Control	YES	YES	YES	YES
Fixed effect	YES	YES	YES	YES
Observation	360	360	360	360
R-squared	0.044	0.384	0.770	0.143

Note: standard errors are clustered at the province level. The parentheses are the t-values. ***, **, and * represent significant levels at 1%, 5%, and 10%, respectively. Control means control variables.

. As seen from Columns (1) and (2) of Table 4, before 2013, carbon trading policies had no significant effect on carbon emissions or carbon efficiency, which further verifies the parallel trend hypothesis. Columns (3) and (4) show that, after 2013, the regression results of carbon emissions and carbon efficiency begin to become significant, and the coefficient gradually increases. This indicates that carbon trading policy has significant dynamic effects and gradually accumulates. Since China announced the launch of the national carbon trading market at the end of 2017, the dynamic effect of pilot provinces began to weaken, which led to nonsignificant results in 2018. In the early stage of establishment, the carbon trading market is immature and has many problems, such as excessive carbon quotas, serious fluctuations in carbon prices, imperfect laws and regulations, low market activity, and information asymmetry. However, over time, the carbon market system gradually improves, with the trading volume increasing year by year. After 2016, diversified green financial products were also introduced [66]. The gradual improvement in the carbon market promoted the strengthening of the policy effect, which caused a cumulative dynamic effect on carbon emission reduction and carbon emission efficiency. The national carbon trading market of the electric power industry opened on July 16, 2021. By December 22, 2021, the national carbon market was running smoothly in its first compliance period, with a total transaction volume of 140 million tons and CNY 5.802 billion of carbon emission quotas. Although the pilot markets are retained, the policy effect on pilot provinces will keep declining as the national market gradually improves. Therefore, this paper predicts that the regression result after 2019 will still be nonsignificant, and Hypothesis 3 is verified.

However, China is still facing many challenges. From Table 2 and Figure 1, the performance of CEE is not well enough. From Table 4, both the absolute value of coefficients of carbon emission and CEE increase at a low speed, which indicates that the pilot market is still lacking many specific measures. Many studies show that China’s carbon market has the problems of overly free quotas [67] and an imperfect market mechanism in the pilot stage [68]. China’s carbon market still has a long way to go to catch other markets at the present stage [69,70]. Therefore, the national carbon market should improve its mechanism and absorb more industries as soon as possible.

5.4. Placebo Test

To further verify the robustness of the experimental results, this paper uses a placebo test on the results of carbon emissions and carbon efficiency to test the robustness of the results and exclude the external influence of other policy factors on the results. Based on the method of Xiqian Cai et al. [71], this paper selects 6 of the 30 provinces as a ‘pseudo treatment group’, assuming that the carbon trading policy is implemented, and the remaining 24 provinces are used as the control group. Then, according to Models (1) and (2), 200 benchmark regression tests are conducted again. If the estimators of most pseudo samples are still significant, this indicates that the original estimation results are likely to be biased. The p values and coefficients of the regression results are presented in Figure 4 and Figure 5. It can be seen from the figures that the p value of most coefficients is concentrated near 0, and only a few regression coefficients have p values less than 0.1, which means that most of them are not significant. This indicates that the results of this paper pass the placebo test. The emission reduction effect and the increase in carbon efficiency in pilot provinces are not accidental but robust.

5.5. Synergistic Emission Reduction Testing

To verify the synergistic emission reduction effect on sulfur dioxide, this paper changes Model (1) by taking the logarithm of sulfur dioxide emissions as a new explained variable. The model is as follows:

$$lnS{O}_{2it}={\alpha }_0+{\alpha }_1{P}_t+{\alpha }_{2}trea{t}_i+{\alpha }_3{P}_t\times trea{t}_i+{\alpha }_i{{\displaystyle \sum }}_i^h{X}_i+{\xi }_{it}$$

(7)

The results of the difference-in-difference regression are shown in

Table 5. Test of synergistic emission reduction effect.

Variables	(1)	(2)
	No Control Variables	All Control Variables
Pt × treati	−0.611 **	−0.832 ***
	(−2.30)	(−4.78)
constant	4.029 ***	1.250
	(47.94)	(0.79)
Control	NO	YES
Observations	360	360
R-squared	0.219	0.676
Regional fixed effect	YES	YES
Time fixed effect	YES	YES

Note: standard errors are clustered at the province level. The parentheses are the t-values. ***, ** and represent significant levels at 1%, 5%, respectively. Control means control variables.

. Columns (1) and (2) show the regression results without control variables and with all control variables, respectively. The regression coefficients are significant and the significance and coefficient size are improved after adding control variables. Therefore, carbon trading policy has a significant synergistic emission reduction effect on sulfur dioxide.

To illustrate the feasibility of Model (7), this paper conducted a parallel trend hypothesis test for sulfur dioxide, and the test results are shown in Figure 6. Before the implementation of the policy, the pilot and nonpilot areas had the same decreasing trend. From 2014 to 2018, the decreasing trend of sulfur dioxide emissions in the pilot area was more obvious than that in the nonpilot area. It can be concluded that sulfur dioxide passes the parallel trend hypothesis test. The above analysis shows that Hypothesis 4 is valid.

6. Conclusions and Policy Recommendations

The carbon trading market commercializes carbon dioxide emissions rights and uses market mechanisms to achieve emission reduction. To explore the policy effect of the carbon trading market, based on panel data from 2007 to 2018, this paper first uses super-efficiency SBM to measure the carbon emission efficiency of each province and then analyses the carbon emission reduction and carbon emission efficiency of pilot provinces using the DID model. Finally, the dynamic effect and synergistic emission reduction are analyzed. According to the research in this paper, carbon trading policy has a significant policy effect on the carbon emissions and carbon efficiency of pilot provinces. It is also verified that there was an inverted U-shaped curve between carbon emissions and economic growth, which conforms to the environmental Kuznets Curve hypothesis. Over time, the regression coefficient of carbon emissions and carbon efficiency gradually increases, which indicates that there is a cumulative dynamic effect. Until the end of 2017, after the release of relevant documents to establish a national carbon trading market, the dynamic effect began to weaken. Regarding the carbon efficiency in pilot provinces, Beijing has always been at a frontier and shows significant policy effects. The efficiency of Shanghai and Guangdong fluctuated greatly, while that of Tianjin did not change significantly. Although the carbon efficiency of Chongqing and Hubei is relatively low, they show a significant growth trend after adapting to the policy. The carbon trading policy has a significant synergistic emission reduction effect for sulfur dioxide. In this paper, the mechanism of emission reduction is not considered.

Based on the above empirical conclusions, this paper puts forward the following policy recommendations:

(1) While intensifying the construction of the national carbon trading market, the pilot market should be further improved. At present, China is in the stage of the coexistence of the national carbon trading market and pilot market, and the pilot market still plays a huge role. This paper confirms the significant policy effects of carbon trading pilot markets. The pilot market should still be used as ‘a beacon’, where many further policies can be implemented. For example, increasing the proportion of paid quotas and the projects of carbon finance and CCER [72] could mobilize market enthusiasm and provide valuable experience for the further development of the national carbon trading market;

(2) The national carbon trading market should be gradually built. As can be seen from the empirical results, carbon trading policies have accumulated dynamic effects on pilot provinces and cities. However, for each pilot market, there are still many problems, such as large fluctuations in carbon emission efficiency and poor policy adaptation. The national carbon trading market is more complex than the pilot market, so the national carbon trading market should be established step by step based on the experience of the pilot market. The impact of various problems can be reduced by accumulating the dynamic effects of policies. Regarding the current situation, this article suggests that the national carbon trading market should be gradually improved in stages by the industry. The next step should be to start with industries that are easy to regulate and in which it is easy to measure carbon quotas. At present, the national carbon trading market only includes the electric power industry. In the future, it should further include industrial industries in which it is convenient to calculate carbon emissions, such as fossil and steel. After including key industrial industries, it should turn to the tertiary industry and eventually cover most industries;

(3) The effect of emission reduction should be examined from multiple dimensions and levels and both carbon emission reduction and economic development should be taken into account. At present, the emission reduction target of pilot provinces only remains in the single dimension of carbon emissions. Although such emission reduction targets directly target the current environmental issues, they do not consider the impact of emission reduction on economic development. Therefore, this paper suggests that emission reduction and economic development should be taken as a comprehensive objective and that increasing carbon emission efficiency or green total factor productivity should be taken as an evaluation indicator. At the same time, pilot provinces should further stabilize the carbon price, rationally allocate the carbon quota, strengthen the function of the carbon market, and strive to achieve the common development of the economy and green ecology;

(4) Actively promoting the synergistic emission reduction effect of sulfur dioxide. According to the research in this paper, carbon trading policy has a significant synergistic emission reduction effect on the pilot areas, which reduces the emission of sulfur dioxide while reducing carbon emissions. At present, the carbon market has not achieved complete improvement, and there are also relevant emission reduction policies for sulfur dioxide. Therefore, China should pay attention to the synergistic effect and avoid policy overlap with regard to future policy adjustments and improvements to the national carbon trading market. Additionally, China should actively develop complementary advantages of policies, promote synergistic emission reductions of sulfur dioxide, and reduce policy costs on environmental regulations.