Impact of Resource-Based Economic Transformation Policy on Sulfur Dioxide Emissions: A Case Study of Shanxi Province

Abstract

Air pollution, particularly SO₂ emission, has become a global problem, seriously threatening the sustainable development and health of mankind. Based on the panel data of 248 prefecture-level cities in China during 2003–2018, this study used the Propensity Score Matching-Difference in Difference (PSM-DID) method within the counterfactual framework to evaluate the treatment effect of the policy made by the National Resource-Based Economic Transformation Comprehensive Supporting Reform Pilot Zone (CRZ) on sulfur dioxide (SO₂) emissions. The results show the following. (1) The benchmark regression results demonstrate that the CRZ policy has significantly decreased per capita SO₂ emissions (PCSO₂) and SO₂ emissions per unit of GDP (PGSO₂) in the pilot zone, and the placebo test indicates that the evaluation of the policy effect is robust. (2) The dynamic effect test indicates that there is a lag in the effect of the CRZ policy on reducing SO₂ emissions. The policy effect of the CRZ policy on PCSO₂ and PGSO₂ was not obvious in the first stage (2011–2015), the CRZ policy significantly reduced the PCSO₂ and PGSO₂ in the second stage of policy implementation (2016 and beyond), and the reduction effect of CRZ policy on SO₂ emissions is increasing over time. (3) The mechanism analysis shows that optimizing industrial structure, increasing human capital, strengthening technological innovation, and expanding opening to the outside world are the main ways for the CRZ policy to reduce SO₂ emissions. The study will help promote SO₂ emissions reduction in Shanxi Province, providing a reference for the transformation and development of other resource-based cities in China and the world and contributing to accelerating the achievement of regional emission reduction targets and sustainable development.

1. Introduction

Following the increase in economic growth and industrialization, the contradiction between resources, environment, and economic development has become more and more prominent, posing severe challenges to global ecological governance and sustainable development [1]. SO₂ has been a key source of air pollution [2] and has resulted in more than one million premature deaths every year [3]. As the largest developing country in the world, China has generated huge energy demand and pollutant emissions in the process of pursuing economic development. According to the statistics of the International Energy Agency, China’s coal consumption was 82.27 EJ in 2020, accounting for 54.3% of the total global coal consumption. Owning to the heavy dependence on fossil energy, particularly coal, SO₂ became the most important air pollutant in China, and the SO₂ emissions amounted to 36.6 million tons in 2007 [4,5]. Hence, the Chinese government has implemented a series of policies to control SO₂ emissions, such as the Two-Control Zone Policy in 1998, the Air Pollution Prevention and Control Action Plan in 2013, and the Three-Year Action Plan for the Defense of the Blue Sky in 2018. The 12th Five-Year Plan (2011–2015) in China proposed to reduce the SO₂ emissions by 10%, and the SO₂ emissions actually reduced by approximately 30% from 2011 to 2015. Furthermore, the 13th Five-Year Plan (2016–2020) in China also proposed to reduce SO₂ emissions by 15% and that emission reduction target was also exceeded by 2020. The SO₂ emissions in China have dropped significantly. However, owing to its high total emissions, China continues to be a major SO₂ emitter in the world, and the volume of SO₂ emissions has become an important index of the environment [6]. Greenpeace pointed out in 2020 that China is the third-largest emitter of SO₂ in the world. Yale University and Columbia University released the 2020 Environmental Performance Index (EPI) in August 2020, which shows that China obtains an EPI score of 37.3 and ranks 120th out of 180 countries [7]. Therefore, reducing SO₂ emissions is critical to China and global air governance; the environmental protection and ecological governance in China need to be improved.

In Shanxi Province, the most typical coal resource-based province in China, the main industries were coal, coking, metallurgy, electric power, and other high energy consumption and high pollution industries, which have resulted in serious air pollution in the province. The SO₂ emission level in Shanxi Province is much higher than the average level of other regions in China. In 2005, the SO₂ emissions of Shanxi Province ranked second among the 31 provinces in China. Moreover, the Shanxi Province was also clearly stated in the 12th Five-Year Plan (2011–2015) to achieve the goal of reducing the SO₂ emissions by 11.3% and then by 20% in its 13th Five-Year Plan (2016–2020). In order to accelerate the optimization and upgrading of the industrial structure, adjust the resource-based economic structure in Shanxi Province, promote the sustainable development of Shanxi Province, and realize the ultimate goal of constructing a resource-saving and environment-friendly society [8,9], in December 2010, the State Council proposed to establish the “National Resource-Based Economic Transformation Comprehensive Supporting Reform Pilot Zone” (hereafter referred to as CRZ) in Shanxi Province. Reducing the emissions of air pollutants is an important part of the CRZ policy. Under the CRZ policy framework, Shanxi Province has successively promulgated the “Contingency Plan for Heavily Polluted Weather in the Shanxi Comprehensive Reform Demonstration Zone” and the “Special Action Plan for the Environmental Protection of the Shanxi Comprehensive Reform Area” to strengthen the control of air pollutants.

Considering the population and economy, the per capita SO₂ emissions (hereinafter referred to as PCSO₂) and SO₂ emissions per unit of GDP (hereinafter referred to as PGSO₂) of prefecture-level cities in Shanxi Province and non-Shanxi prefecture-level cities from 2003 to 2018 are reported in Figure 1. The PCSO₂ and PGSO₂ in Shanxi are much higher than the average values of non-Shanxi regions, indicating that environmental problems are particularly prevalent in Shanxi Province. Observing the trend of lines, the PCSO₂ in Shanxi and non-Shanxi regions maintained a high level before 2011 and has declined since 2011. Besides, the gap in PCSO₂ between the Shanxi Province and non-Shanxi Province has gradually narrowed since 2011. At the same time, the PGSO₂ always shows a steady downward trend during the observation interval.

Figure 1 illustrates that both the PCSO₂ and PGSO₂ in Shanxi Province decreased after 2011. Can we infer from this that the CRZ policy effectively lowered the PCSO₂ and PGSO₂ in Shanxi Province? The answer is no. The reason is that the emissions of SO₂ are affected by numerous factors. As shown in Figure 1, the PCSO₂ and PGSO₂ of the non-Shanxi province where the CRZ policy has not been implemented also show a similar trend with the Shanxi Province where the CRZ policy has been implemented since 2011. Moreover, if Shanxi Province did not implement the CRZ policy in 2011, the changing trend of PCSO₂ and PGSO₂ may not follow the changing trend in Figure 1. Therefore, we need to exclude the influence of other factors to accurately examine whether the CRZ policy has impacted the SO₂ emissions in Shanxi Province and if the policy works, we need to further analyze when and how the policy works [10]. However, few articles have conducted empirical research on the effects and mechanisms of CRZ policy. The implementation effect and mechanism of CRZ policy are not clear, which are crucial for policy optimization and provide reference experience of transformational development for other resource-based cities. Therefore, this study aims to study whether the CRZ policy is effective in reducing the SO₂ emissions in Shanxi Province, analyze the changes of the impact of CRZ policy on regional SO₂ emissions over time, and explore the mechanism by which the CRZ policy affected the SO₂ emissions in Shanxi Province. The study will contribute to a better understanding of the impact and mechanism of resource-based economic transformational policy on the atmospheric environment, provide new ideas and theoretical support for the reduction in SO₂ emissions, and also, as is of vital importance, achieve the sustainable development.

The rest of the paper is structured as follows. Section 2 reviews the literature related to SO₂ emissions and resource-based economic and policy evaluation. Section 3 consists of a methodological framework, variable selection, and data sources. Section 4 presents the results and discussion of empirical research and other analytical tests. Conclusions and suggestions are summarized in Section 5.

2. Literature Review

The related research on SO₂ mainly focuses on the factors that affect SO₂ emissions. Sinha studied the influence of energy intensity and economic development on SO₂ emissions and verified the existence of the environmental Kuznets curve of SO₂ [11]. Yao et al., according to the index decomposition analysis method, pointed out that engineering intensity and regulatory supervision can most significantly reduce SO₂ emissions [12]. Han et al., combined the Grossman decomposition model with the Logarithmic Mean Divisia Index (LMDI) and found the scale effect is an important factor influencing SO₂ emissions [13]. Yang et al., employed LMDI to analyze the critical elements of SO₂ emissions in China and found that energy consumption is the primary reason for SO₂ increase, while technology is crucial to reducing the SO₂ emissions [14]. Shen found a U-shape curve rather than an inverted U-shape curve between sulfur emissions and GDP, which was consistent with the results of Fodha and Zaghdoud [15,16]. Wang et al., studied the relationship between economic development, urbanization, and SO₂ emissions and found an inverted U-shape curve between economic development and SO₂ emissions [17]. Zhou et al., verified an inverse N-shape relationship between SO₂ emissions and economic development; the progress of technology plays an important role in reducing SO₂ emissions [18]. Chen and Yan found that the development of e-commerce can significantly reduce SO₂ emissions in cities in China [19]. Zhong and Shen et al., found that activity growth has been the major factor in driving global SO₂ emission changes overall, but the control measure deployed is playing an increasingly important role [20]. Besides, many scholars found that coal-fired power stations are the main cause of the increase in SO₂ emissions [21,22]. There is also some empirical research on policy to control SO₂ emissions. Cheng et al., set a regional CGE model and found that the sulfur dioxide emissions trading pilot scheme (SETPS) reduced SO₂ emissions in Guangdong province by 33% in 2020 [23]. Ren et al., and Peng et al., consulted the panel data of Chinese industrial enterprises and applied the Difference-in-Differences (DID) method to assess the effects of SETPS on SO₂ emissions. Both found the SETPS policy contributes to reducing SO₂ emissions [24,25]. Chen and Huang et al., used the synthetic control method to study the effectiveness of SETPS and found that the effect of the SETPS has regional heterogeneity; Tianjin has achieved the desired reduction effect and the SETPS implemented in other areas did not reduce the SO₂ emissions [26].

Resource-based cities and resource-based economies have received widespread attention from scholars, and the studies were mainly focused on the causes and problems in resource-based cities, resource-based economic transformation paths, and the evaluation of the effect of transformational policies. In the middle of the 1980s, some scholars found that the economic growth rate of countries or regions with abundant resources (especially mineral and energy resources) is much lower than resource-poor countries or regions [27]. Auty and Warhurst first defined this phenomenon as the “Resource Curse” [28]. Subsequently, many scholars did a lot of work to verify the hypotheses [29,30] and try to explain the mechanism of the “curse” [31,32]. After realizing the harm of the “Resource Curse”, how to solve the problem became particularly important. Many scholars have carried out a lot of studies on the transformation path of resource-based cities from different perspectives. There are already some successful cases about the transformational development of resource-based cities, such as Pittsburgh and the Ruhr [33]. Zhou and Liu et al., analyzed the influencing factors and mechanisms of resource-based economic transformation and development and proposed suggestions for the enterprise, industry, and system to promote the transformation and industrial upgrading in resource-based cities [18]. Some scholars have evaluated the effect of the transformation and development of resource-based cities in terms of economic development, environment, industrial structure, and innovation. Chen and Shen et al., studied the level of transformation and upgrading of six resource-based cities in Shaanxi province. The results indicate that the environmental quality has been effectively improved, but the structural contradictions in the economy and industry have not been fundamentally resolved [34]. Zhang and Zhang et al., used a five-dimensional evaluation method to evaluate the performance of the economic transformation in northeast China since 2003. The results show that the effects of economic transformation and environmental improvement are significant, but technological progress and innovation are insufficient [35]. Zhang and Zhao et al., studied the coupling relationship and interactive response mechanism between the transformation of economic development and environmental quality of the resource-based cities in northeast China and found the interaction effects are different in resource-based cities at different development stages [36]. Du et al., pointed out that resource dependence has a negative impact on high-quality economic development and has a crowding-out effect on innovation investment and human capital [37]. The study of Yang et al., shows that government supervision and policy guidance are important means to promote the transformation of a resource-based economy [38].

Policy evaluation is a scientific tool to study the effect of the policy. At an early stage, the related research is mainly based on qualitative and theoretical research, such as establishing an evaluation index system to evaluate the effect of policy implementation [39,40] or pointing out the deficiencies of policy implementation and putting forward countermeasures and suggestions [41,42]. Those articles often lack data support, and the data analysis is relatively simple and unconvincing. With further research, quantitative and empirical policy evaluation research gradually emerged and produced many theoretical research achievements. At present, some relatively mature policy evaluation methods have been widely used in empirical research, such as Instrumental Variables [43,44], Regression Discontinuity Designs [45,46], Difference-in-Differences (DID) [47,48], Propensity Score Matching (PSM) [49,50], PSM-DID [51,52,53], and the Synthetic Control Method [54,55]. Among them, the PSM-DID combines the advantages of PSM and DID, which can effectively avoid selection bias and endogeneity problems and has been widely used to study macro policy in China. The most studied policies in China are the Western Development [56,57], the Northeast Revitalization [58,59], the Rise of Central China [60,61], and the Low-Carbon Pilot policies [62,63,64].

According to the above literature review, we found that the literature research on SO₂ emissions, the transformation of resource-based cities, and policy evaluation is relatively abundant. However, the studies are scarce in some aspects. First, the research on SO₂ emissions mainly focuses on the relationship between SO₂ emissions and various variables, such as economic development, energy structure, and government regulation, whereas the study of evaluating the impact of macro-policies on SO₂ emissions should be enhanced. Second, the transformation dilemmas, transformation paths, and transformation effects of resource-based cities have been studied further. Among the existing literature on evaluating the effect of transformational policy in resource-based cities, scholars mainly use the entropy method, analytic hierarchy process, and other statistical methods to simply evaluate the implementation effect of transformation policy. However, the literature that adopts econometric methods to directly and quantitatively evaluate the effect of the transformational policy is lacking. The degree and the trend of policy effect and the mechanism by which the policy works are unclear. Finally, most of the previous studies of SO₂ emissions focused on the concentration and total amount of SO₂ emissions and paid less attention to the SO₂ emission intensity. However, economic development and population are the important factors we need to consider, thus, we think that compared with the total pollutant discharge, the emission intensity indicators are more suitable for the current research needs.

Compared with previous works, this paper makes the following innovations and contributions: (1) A PSM-DID method within the counterfactual framework was used to evaluate the influence of CRZ policy on regional PCSO₂ and PGSO₂ and provide a reference for other resource-based cities. Sulfur has a significant impact on the atmospheric environment and biology; it is already the main pollutant of ecology and the environment. The SO₂ emissions of resource-based cities are often much higher than those of other types of cities. Therefore, more attention should be paid to controlling SO₂ emissions, especially in resource-based cities. (2) We further analyzed the trend of the policy effects of the CRZ on PCSO₂ and PGSO₂ over time and the mechanism by which the CRZ policy affected PCSO₂ and PGSO₂. This provided theoretical support and development experience for transformational development and policy formulation in other developing countries. (3) The PSM-DID combines the advantage of PSM and DID; it can effectively reduce the selection bias of the sample and alleviate the endogeneity problem. However, the PSM-DID requires more samples and, thus, we collected the panel data of 248 prefecture-level cities in China. Besides, considering the economic development and population of the city, we adopted the intensity index rather than a total index to ensure the rationality of the policy effect evaluation.

3. Materials and Methods

3.1. Methodology

Quantitative study and qualitative study are the two opposing research paradigms of social sciences, and most early research adopted qualitative study. However, with further research, qualitative analysis alone often lacks persuasiveness. Thus, more and more scholars began to propose and adopt quantitative methods to conduct research. In the quantitative policy evaluation method under the counterfactual framework, DID is one of the most widely used methods [65,66], but a common trend assumption needs to be met when we use DID to perform a quantitative evaluation. However, in the actual study, we cannot guarantee the selection of the treated group is completely random, and the implementation of a policy is always not random because it often requires considering many factors, such as regional economy, society, and the environment. To make up for the deficiency, Heckman et al. [67] improved the DID method and proposed PSM-DID. The PSM method can alleviate the sample selection bias and solve the problem of the treated group and control group not meeting the common trend assumption before policy intervention, ensuring the reliability of results.

PSM-DID can be divided into the following steps [68]: (1) Assign the samples into a treated group and a control group. (2) Run a logit model with the covariates to calculate the propensity score of each sample. (3) Check the common support region, and a large region of common support is generally associated with a higher probability of achieving better quality matches. If the common support region is narrow, the covariates included in the logit model and the samples in the control group should be reselected. (4) Match the samples from the treated group with those from the control group based on the propensity score. (5) A balance test should be performed to check whether the treated group and control group are statistically similar after matching. If the covariates are significantly different, then the logit model should be reconstructed. (6) If all the above steps are passed, then evaluate the treatment effects based on the differences between each pair of matched samples.

PSM uses non-experimental data or observational statistical data to remove selection biases in non-experimental research and evaluate the intervention effects [69]. In this study, we divided 248 cities into a treated group (11 prefecture-level cities in Shanxi Province) and a control group (237 non-Shanxi Province prefecture-level cities) and then ran a logit model to calculate the propensity score. The PSM logit model is set as follows:

$$P_i=P(A=T|Z_{it})$$

(1)

where A = {T, C} represents all the 248 cities in this paper, which includes 11 treated samples and 237 control samples. $Z_{it}$ represents a series of observational matching variables that can influence the probability of a city being selected for the treated group. $P_i$ represents the probability of city i being selected for the treated group. Then we used the nearest neighbor matching to select the samples that have the most similar characteristic to the treated samples to evaluate the counterfactual value. The matching variables should have no significant difference between the treated group and the control group after excluding the samples that are not in the common support region.

According to the PSM, we selected appropriate samples to relieve the selection bias and ensure the treated group and control group meet the common trend assumption. Then we took the CRZ policy implemented in Shanxi Province in 2010 as a quasi-natural experiment and used the DID method to capture the differences between the treated group and the control group. Referring to the related research, we constructed the benchmark regression model of DID as follows:

$$Y_{it}={\alpha }_0+{\alpha }_{1}Perio{d}_{it}\times Treate{d}_{it}+{\alpha }_{2}Treate{d}_{it}+{\alpha }_{3}Perio{d}_{it}+{\alpha }_4{X}_{it}+{\epsilon }_{it}$$

(2)

where $Y_{it}$ is the explained variables in this study, including PCSO₂ and PGSO₂. $Treate{d}_{it}$ and $Perio{d}_{it}$ represent individual dummy variables and time dummy variables. If the city i belongs to the treated group, then $Treate{d}_{it}$= 1, otherwise, $Treate{d}_{it }$= 0. If the year t is between 2011 and 2018 (including 2011 and 2018), then $Perio{d}_{it}$= 1, otherwise, $Perio{d}_{it }$= 0. $Treate{d}_{it}\times Perio{d}_{it}$ is the interaction term, and its coefficient represents the estimator in which we are interested, indicating the average treated effects of the CRZ policy. $X_{it}$ represents the observable covariates, ${\epsilon }_{it}$ is a random error term.

The benchmark regression model can evaluate the average treatment effect of the CRZ policy on outcome variables, but it cannot reveal the dynamic effect of the CRZ over time, and we cannot understand the trend of the treatment effect of the CRZ policy year by year. Thus, we introduced the year dummy variable and extended the regression model in Equation (2) to measure the dynamic effect of the CRZ policy implemented in Shanxi Province year by year. The model is constructed as follows:

$$Y_{it}={\alpha }_0+{\alpha }_j\sum Perio{d}_{it}\times Treate{d}_{it}\times yea{r}_j+{\alpha }_{2}Treate{d}_{it}+{\alpha }_{3}Perio{d}_{it}+{\alpha }_4{X}_{it}+{\epsilon }_{it}$$

(3)

where $j=2011, 2012, \dots , 2017, 2018$ and $yea{r}_{\mathrm{j}}$ represent the annual dummy variable and the coefficient ${\alpha }_j$ represents the dynamic effect of the corresponding year. All the other variables are the same as those in Equation (2).

Furthermore, to provide helpful experience for further deepening reforms and other resource-based economic transformations, this study references related literature [70,71], constructs the following model, and combines it with Equation (2) to analyze the mechanism of the impact of CRZ policy on regional PCSO₂ and PGSO₂.

$$X_{it}={\alpha }_0+{\alpha }_{1}Perio{d}_{it}\times Treate{d}_{it}+{\gamma }_i+{\tau }_t+{\epsilon }_{it}$$

(4)

where $X_{it}$ represents the matrix-vector of control variables in Equation (2) and i and t represent the city and the year, respectively. $\gamma$ represents the city fixed effect, $\tau$ represents the time fixed effects, and $\epsilon$ represents the stochastic disturbance term. All of the other variables are the same as those in Equation (2).

3.2. Variables and Data

The study collected the panel data of 248 prefecture-level cities in China from 2003 to 2018 and used the PSM-DID to evaluate the effect of the CRZ policy implemented in Shanxi Province in 2010. The select rules of samples are as follows. First, based on the availability and completeness of the data, we excluded Tibet, Xinjiang, Hong Kong, Macao, and Taiwan, whose data are missing. Then, considering the difference in the administrative hierarchy, we excluded the municipalities (Beijing, Tianjin, Chongqing, Shenzhen). In order to eliminate the impact of other national reform policies, cities included in the other 11 national reform zones are excluded. Finally, we sorted out the panel data of 248 prefecture-level cities. The data come from the China City Statistical Yearbook and the Statistical Communiques on National Economic and Social Development of the prefecture-level cities, the statistical yearbooks of provinces and prefecture-level cities. A small number of missing data are supplemented by interpolation. In order to reduce heteroscedasticity, most variables in this study are logarithmically processed.

Key explanatory variables: $Treate{d}_{it}$ and $Perio{d}_{it}$ are two dummy variables. For the cities belonging to the treated group, set the $Treate{d}_{it}=1$, otherwise set as 0. For the years between 2011 and 2018 (including 2011 and 2018), set the $Perio{d}_{it}$= 1, otherwise set $Perio{d}_{it}=$0. As shown in Equation(2), the interaction term $Perio{d}_{it}\times Treate{d}_{it}$ is the key explanatory variable in this study and its coefficient indicates the treatment effect of CRZ policy on PCSO₂ and PGSO₂.

Dependent variable: Taking economics and population into account, this paper adopts PCSO₂ and PGSO₂ as outcome variables to study. PCSO₂ reflects the regional overall air pollution level. The lower the value, the better the air quality in the region. PGSO₂ is an intensity indicator, which can reflect the regional efficiency of SO₂ emissions.

Control variables: We refer to existing relevant literature research and consider data integrity and availability. The selection and definition of control variables are as follows:

(1) Economic development level. The EKC indicates that there exists an inverted U-shaped relationship between the level of economic development and environmental pollution [72,73]. This study uses the logarithm of the GDP per capita (lnpgdp) to measure the level of regional economic development;

(2) Population agglomeration. Population density is an important factor, and it is generally agreed that population agglomeration can reduce air pollution in developed countries, but, in China, the impact is not clear yet. Referring to related literature, this study uses the logarithm of urban population density (pop) to measure population agglomeration [74];

(3) Human capital. Environmental issues and economic development are inseparable from humans, and the human capital can represent individual environmental protection awareness and promote economic high-quality transformation. Referring to the related literature, this study uses the logarithm of the number of students in regional higher education (lnhc) to represent the human capital level [75];

(4) The fiscal expenditure scale of government. Some studies have found that fiscal expenditure can influence the environmental quality and economic development quality, although the influence mechanism is ambiguous [76,77]. This study uses the logarithm of the expenditure in the local fiscal budget (lnfe) to indicate the scale of fiscal expenditure;

(5) Industrial structure. Compared with the service sector and agriculture, the secondary industry can produce more pollutants [78]. The transition from secondary industry to tertiary industry is also an important indicator of economic high-quality development. This study uses the output value of the secondary industry as a percentage of GDP (sec) to characterize the industrial structure;

(6) Fixed asset investment level. Some scholars pointed out that fixed-asset investment may inhibit high-quality economic development and environmental quality [79,80]. This study uses the logarithm of the regional investment in fixed assets (lnfix) to characterize the fixed asset investment level;

(7) The foreign direct investment (FDI) level. The “pollution paradise” and the “pollution halo” are the two main hypotheses about the relationship between environmental quality and FDI [81,82]. Some scholars also pointed out that FDI can promote the efficiency of economic development. This study uses the logarithm of the amount of direct use of foreign investment (lnfdi) denominated in RMB to indicate the FDI level;

(8) Scientific education investment. Many studies have shown that technological innovation and advancement can greatly improve environmental quality and reduce polluting emissions [83]. This study uses the logarithm of the total expenditures of science, technology, and education in government fiscal expenditures (lnse) to represent the degree of emphasis on science education;

(9) The informatization level. Information technology can promote the upgrade of industrial structure and improve environmental quality [84]. This study uses the logarithm of the number of internet broadband access households (lnil) to represent the level of informatization.

The descriptive statistics of variables are presented in

Table 1. Descriptive statistics of main variables.

Var	Treat Group				Control Group
	Mean	Max	Min	Std.dev.	Mean	Max	Min	Std.dev.
PCSO2	328.83	965.84	19.89	221.33	148.46	2749.71	0.20	202.82
PGSO2	1.75	892.27	1.93	1.63	0.70	1599.31	0.19	1.08
lnpgdp	10.05	11.76	8.21	0.66	10.17	15.68	7.77	0.83
pop	5.68	6.28	4.83	0.40	5.91	7.14	3.73	0.61
lnhc	3.06	6.10	−1.12	1.36	3.31	6.99	−2.99	1.36
lnfe	9.13	10.90	5.55	1.20	9.32	12.43	5.22	1.29
sec	53.73	73.71	36.12	8.75	48.08	85.92	14.40	10.27
lnfix	5.98	7.62	3.63	0.92	6.32	9.04	2.81	1.19
lnfdi	6.09	8.80	0.000	1.64	6.90	11.36	1.60	1.84
lnse	7.71	9.27	5.57	0.85	7.83	11.01	4.37	1.06
lnil	5.56	7.59	1.64	0.98	5.68	8.94	−1.44	1.18

.

4. Results and Discussion

4.1. Propensity Score Matching

First of all, in order to overcome the systematic differences between the control group and the treated group and ensure the experimental group and the control group meet the common trend assumption before implementing the CRZ policy, we used the PSM to match the control group and the treated group. As stated above, this study took the control variables as covariates and PCSO₂ as the outcome variable to match. The results are shown as follows. The nearest neighbor matching within a caliper with a ratio of 1:4 and a radius of 0.01 was used in this study for matching. The matching results are reported in

Table 2. Common support test (nearest neighbor matching within a caliper).

Treatment Assignment	Off Support	On Support	Total
Untreated	20	3772	3792
Treated	7	169	176
Total	27	3941	3968

. Only 27 observations did not satisfy the common support test, which accounts for a small part of the total sample and the matching results are satisfactory.

Figure 2 and

Table 3. The balance test results.

Var		Mean		% Bias	% Bias Reduction	t-Test
		Treated	Control			t	p
lnpgdp	U	10.046	10.172	−16.8	-	−2.00	0.046
	M	10.064	10.113	−6.6	61.0	−0.61	0.544
pop	U	245.24	429.66	−79.9	-	−7.90	0.000
	M	247.64	237.79	4.3	94.7	−0.64	0.524
lnhc	U	3.062	3.311	−18.3	-	−2.38	0.018
	M	3.072	3.066	0.5	97.4	0.05	0.962
lnfe	U	9.129	9.322	−15.6	-	−1.96	0.05
	M	9.161	9.198	−3.0	80.8	−0.29	0.774
sec	U	53.732	48.082	59.2	-	7.18	0.000
	M	53.214	53.121	1.0	98.3	0.08	0.936
lnfix	U	5.980	6.323	−32.3	-	−3.77	0.000
	M	5.998	6.025	−2.5	92.2	−0.24	0.807
lnfdi	U	6.090	6.900	−46.2	-	−5.70	0.000
	M	6.129	6.003	7.2	84.4	0.65	0.515
lnse	U	7.711	7.831	−12.5	-	−1.48	0.138
	M	7.714	7.754	−4.2	66.3	−0.42	0.677
lnil	U	5.562	5.676	−10.5	-	−1.26	0.208
	M	5.575	5.577	−0.1	98.7	−0.01	0.989

show the differences in control variables before and after the PSM between the treated group and the control group. The standardized percentage biases across the control variables have been significantly reduced after matching, all the biases are less than 10%, and the p-values of the t-test for all control variables after matching are greater than 0.1. The results mean there are no significant differences in the control variables between the treatment group and control groups after matching. After the matching, the two groups are very similar and satisfied the balance condition for PSM.

This study also reported the kernel density curve of the propensity scores of the treated group and control group before and after matching, and the result is shown in Figure 3. Figure 3a shows the distribution of propensity scores of the treated group and control group before matching, which is a significant difference. Figure 3b shows the distribution of propensity scores of the treated group and control group after matching, which is highly consistent. The results of Figure 3 indicate that the PSM can significantly improve the comparability between the treated group and the control group. Thus, the matching results based on PSM are more reliable.

4.2. Benchmark Regression Results

The previous section described the results of PSM, and the results have passed a series of related tests; we secured the control group and the treated group that met the parallel trend hypothesis. This section further utilizes the panel data of samples observed in the common support region and adopts the DID to construct a two-way fixed effect model (Equation (2)) to calculate the policy effect of the CRZ policy implemented in Shanxi Province. The results are reported in

Table 4. The average treatment effect of the CRZ policy on PCSO₂ and PGSO₂.

	(1)	(2)	(3)	(4)
	PCSO2	PGSO2	PCSO2	PGSO2
$Perio{d}_{it}\times Treate{d}_{it}$	−64.905 *** (−4.65)	−1.131 *** (−10.34)	−47.803 *** (−3.48)	−113.935 *** (−10.41)
$Treate{d}_{it}$	80.226 ** (2.27)	1.398 *** (5.05)	−32.062 (−0.72)	0.643 ** (1.80)
$Perio{d}_{it}$	−108.982 *** (−13.63)	−1.555 *** (−24.78)	−250.598 *** (−7.93)	0.496 * (1.97)
lnpgdp			63.081 *** (6.84)	−0.523 *** (7.11)
pop			0.122 *** (−3.42)	0.00058 ** (2.05)
lnhc			−34.211 *** (−6.65)	0.034 (0.82)
lnfe			33.673 *** (5.69)	−0.031 (−0.65)
sec			76.307 ** (1.99)	1.101 *** (3.61)
lnfix			28.416 *** (4.67)	0.044 (0.91)
lnfdi			−1.401 (−0.81)	−0.049 *** (−3.55)
lnse			−34.136 *** (−3.36)	−0.251 *** (−3.09)
lnil			−10.206 ** (−2.43)	0.186 *** (−5.57)
cons	152.773 *** (6.69)	1.482 *** (8.28)	−342.966 *** (−3.38)	8.636 *** (10.67)
N	3941	3941	3941	3941
R2	0.8170	0.6219	0.8299	0.6370

Note: * p < 0.1, ** p < 0.05, *** p < 0.01. Below the coefficient () is the standard error.

. Columns (1) and (2) represent the models without any control variables, and columns (3) and (4) represent the models with the control variables added. In Table 4, we focus on the coefficients of the interaction terms $Perio{d}_{it}\times Treate{d}_{it}$. We find that with or without the control variables, the coefficients of the interaction term are negative and all have passed the significance test at the 1% significance level, indicating that the CRZ policy implemented in Shanxi Province has significantly reduced the regional PCSO₂ and PGSO₂. Specifically, the interaction coefficients in column (1) and column (3) are −64.905 and −47.803, respectively, indicating the CRZ policy has reduced the PCSO₂ by 64.905 tons/10,000 people when the control variables are not considered and by 47.803 tons/10,000 people when the control variables are considered. Similarly, the coefficients of the interaction terms in columns (2) and (4) are −1.131 and −113.935, respectively, representing the CRZ policy has reduced the PGSO₂ by 1.131 tons/100 million when the control variables are not considered and by 113.935 tons/100 million when the control variables are considered.

Observing and analyzing the coefficients of other control variables in Table 4, we can also infer the following information: (1) Population density, the proportion of secondary industry, and fixed asset investment level have a significantly positive correlation with PGSO₂ and PCSO₂. This is primarily because the population density may increase the city’s demand for housing and energy consumption and the secondary industry will increase the pollutant emissions, which is consistent with the results of much literature [85,86]. The fixed asset investment means the production, construction, and maintenance of public infrastructure, which may lead to more pollution. (2) Foreign direct investment and scientific education investment can reduce PCSO₂ and PGSO₂. As the Shanxi Province is undergoing industrial restructuring, the focus of FDI investment has shifted from traditional manufacturing to high-end manufacturing or other higher-end tertiary industries, which are more environmentally friendly. Moreover, increasing the investment in science and education can improve the quality of the population and enhance the awareness of environmental protection, and it also can promote technological innovation. (3) Economic development and the scale of fiscal expenditure can effectively reduce PGSO₂ while having a positive influence on PCSO₂. This may mainly be due to the CRZ policy emphasizing economic transformation and promoting the investment shifting from labor-intensive to capital-intensive, thus improving the efficiency of economic development and reducing PGSO₂. Economic development increased PCSO₂, which may be because the population has not changed much, but the total SO₂ emissions varied greatly.

4.3. Placebo Test

In order to ensure the reliability of the results, this study refers to relevant literature to perform a placebo test on the previous calculation results to verify the robustness of the results [87,88]. We randomly selected 11 cities from the 248 prefecture-level cities as the treated group, assuming that they were impacted by the CRZ policy, the other 237 prefecture-level cities were the control group, and we repeated the aforementioned PSM-DID process 1000 times. Correspondingly, 1000 different interaction term coefficients were obtained. Then, the reliability of the previous regression results could be judged according to the significance and distribution of the coefficients of the interaction term.

Figure 4 illustrates the placebo test results. The distribution of the interaction coefficients of PCSO₂ obtained through the above random repeated experiments is shown in Figure 4a. The interaction coefficients obtained through the random repeated experiments are distributed symmetrically with a mean value of 0.71. The distribution density of the interaction term coefficient is approximately a normal distribution, indicating that the treated effect of PCSO₂ calculated by randomly selecting the treated group has no significant difference before and after the implementation of the CRZ policy. The interaction term coefficient calculated in the benchmark model is −47.803, which is significantly different from the interaction term distribution map obtained by random repeated experiments in Figure 4a. Thus, the results further prove that the CRZ effectively lowered the PCSO₂. The same placebo test was performed on the PGSO₂, the results are reported in Figure 4b, and similar conclusions were drawn from the analysis. The actual reduction in PGSO₂ in Shanxi Province is significantly different from the results obtained by the random repeated experiment, indicating that the improvement of PGSO₂ in Shanxi Province before and after the policy is caused by the implementation of the CRZ policy rather than other factors. In general, the regression results of the benchmark model are very robust and credible. It can be considered that the implementation of the CRZ policy effectively reduced PCSO₂ and PGSO₂ in Shanxi Province.

4.4. Dynamic Effects Test

The above research only considered the average treatment effect of the CRZ policy on PCSO₂ and PGSO₂ in the pilot zone and did not consider the changing trend of the treatment effect of the CRZ policy over time. Thus, we further added a dummy variable that represents the specific year based on Equation (2) and obtained Equation (3) to evaluate the policy effect of CRZ every year after 2011. The results are reported in

Table 5. Estimation results of the dynamic effects of the CRZ policy.

	(1) PCSO2	(2) PGSO2
$Perio{d}_{it}\times Treate{d}_{it}\times yea{r}_{2011}$	42.423 (1.08)	−12.970 (−0.42)
$Perio{d}_{it}\times Treate{d}_{it}\times yea{r}_{2012}$	18.970 (0.48)	−25.574 (−0.83)
$Perio{d}_{it}\times Treate{d}_{it}\times yea{r}_{2013}$	16.914 (0.44)	−24.360 (−0.81)
$Perio{d}_{it}\times Treate{d}_{it}\times yea{r}_{2014}$	−3.133 (−0.08)	−26.650 (−0.89)
$Perio{d}_{it}\times Treate{d}_{it}\times yea{r}_{2015}$	−36.214 (−0.94)	−31.655 (−1.06)
$Perio{d}_{it}\times Treate{d}_{it}\times yea{r}_{2016}$	−95.012 ** (−2.47)	−51.371 * (−1.72)
$Perio{d}_{it}\times Treate{d}_{it}\times yea{r}_{2017}$	−150.020 *** (−3.80)	−74.821 ** (−2.44)
$Perio{d}_{it}\times Treate{d}_{it}\times yea{r}_{2018}$	−144.701 *** (−3.76)	−74.906 ** (−2.50)
cons	152.534 *** (6.72)	142.617 *** (8.06)
Control	YES	YES
N	3941	3941
R2	0.8190	0.6313

Note: * p < 0.1, ** p < 0.05, *** p < 0.01. Below the coefficient () is the standard error.

.

Analyzing the interaction coefficients and their significance test results in column (1) of Table 5, we find that the interaction coefficients are positive from 2011 to 2013 and not significant, which means the CRZ policy increased PGSO₂ but the boost effect on PCSO₂ was not significant. Besides, the interaction coefficients are negative but not significant in 2014 and 2015, which means the CRZ policy had reduced the PCSO₂ insignificantly. Moreover, the coefficients are negative and the significance test results are significant since 2015, indicating that the CRZ policy significantly reduced the PCSO₂, and we also find the reduction effect gradually strengthening over time. In general, the CRZ policy slightly increased the PCSO₂ before 2013, then showed a reduction effect on PCSO₂ after 2014 and gradually strengthened over time. Analyzing the interaction coefficients and their significance test results in column (2) of Table 5, we find that the interaction coefficients are positive since 2011, which means the CRZ policy has reduced the PGSO₂ in Shanxi Province since 2011. The significance test results show that the reduction effect of CRZ was not significant from 2011 to 2015 and became significant in 2016. In general, we found that the reduction effect of CRZ policy on PGSO₂ shows a trend of increasing year by year. Figure 5 described the dynamic process of the impact of CRZ policy on PCSO₂ and PGSO₂ in Shanxi Province more intuitively.

4.5. Mechanism Analysis

Based on the tests and analysis above, we obtained the treatment effect of CRZ policy on PCSO₂ and PGSO₂ and the changing trend of the treatment effect over time. Furthermore, in order to further promote the reform of resource-based cities in Shanxi Province and provide experience and reference for the subsequent transformation of other resource-based cities, we also need to analyze the mechanism of the influence of the CRZ policy on PCSO₂ and PGSO₂. Thus, this study refers to the related literature and constructs the model of Equation (4) to analyze the influence of the CRZ on control variables. The results are reported in

Table 6. The influence of the CRZ on control variables.

Var	lnpgdp	pop	lnhc	lnlf	sec	lnfix	lnfdi	lnse	lnil
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Treatedit	−0.592 *** (−7.96)	−0.947 ** (−43.93)	−3.092 *** (−27.69)	−0.870 *** (−8.03)	0.192 *** (10.51)	−1.756 *** (−14.7)	−2.47 *** (−7.23)	−0.775 *** (−12.66)	−1.909 *** (−13.59)
$Perio{d}_{it}$	1.799 *** (106.7)	0.093 *** (18.95)	1.142 *** (45.14)	3.320 *** (135.3)	−0.016 *** (−3.95)	2.634 *** (97.12)	1.237 *** (15.98)	2.479 *** (178.8)	2.624 *** (82.47)
$Perio{d}_{it}\times Treate{d}_{it}$	−0.110 *** (−3.81)	−0.006 (−0.69)	0.112 ** (2.55)	−0.118 *** (−2.76)	−0.064 *** (−8.84)	−0.122 *** (−2.59)	0.460 *** (3.41)	0.162 *** (6.70)	−0.026 (−0.46)
cons	9.343 (194.2)	6.829 *** (490.19)	5.098 *** (70.62)	8.112 *** (115.9)	0.441 *** (37.36)	6.187 *** (79.94)	7.34 *** (33.21)	7.383 *** (186.6)	6.007 *** (66.15)
N	3941	3941	3941	3941	3941	3941	3941	3941	3941
R2	0.9485	0.9925	0.9577	0.9549	0.7946	0.9354	0.7809	0.9786	0.9093

Note: ** p < 0.05, *** p < 0.01. Below the coefficient () is the standard error.

. Then, the results in Table 4 and Table 6 are combined to explore the influence mechanism of CRZ.

According to Table 6, we find that the CRZ policy has a significantly positive effect on human capital, FDI, and the science and education investment. Meanwhile, the CRZ policy has produced a significantly negative impact on economic development, fiscal expenditure scale of government, the proportion of the secondary industry, and fixed asset investment. Combining with the results in Table 4, which show the economic development level, the scale of fiscal expenditure, the proportion of the secondary industry, and the fixed asset investment level all have a significant positive correlation with PCSO₂. However, the human capital, FDI, scientific and education investment, and the informatization level can reduce the emission of PCSO₂. Therefore, we can conclude that the CRZ policy mainly reduced the emission of the PCSO₂ by decreasing the level of economic development, the proportion of the secondary industry, and the expenditure of the government and improving the level of human capital and investment in science and education. Similarly, the results in Table 4 demonstrate that the level of economic development, FDI, and investment in science and education are significantly negatively correlated with PGSO₂ emissions, while population, the proportion of the secondary production, and informatization level are significantly positively correlated with PGSO₂ emissions. Therefore, we can conclude that the CRZ policy mainly reduced the PGSO₂ emissions by reducing the proportion of the secondary industry and increasing FDI and investment in science and education. Overall, reducing the proportion of the secondary industry and increasing the human capital, FDI, and the investment in science and education are the main ways for the CRZ policy to improve SO₂ emissions. In addition, the CRZ policy also reduced fiscal expenditure, fixed asset investment, and the proportion of secondary production to make a difference. In other words, in the progress of reform, Shanxi Province sacrificed the speed of economic development in exchange for environmental improvement and the quality of economic development, which is more conducive to the sustainable and healthy development of the economy and society in long run. The exploration of the influence mechanism of CRZ is of great significance for optimizing regional transformation policy and the transformation and development of other similar cities.

4.6. Discussion

4.6.1. Discussion of the Benchmark Regression Results

Table 4 reveals that the CRZ policy has a significant reduction effect on the PCSO₂ and PGSO₂ in Shanxi Province. The result further proves that the policy tool is an effective measure to control pollutant emissions, which is consistent with many existing studies. Anish and Ritesh et al., find that the SO₂ abatement options implemented in India have reduced the air pollutants by 75% (approximate to 4600 kt-SO₂ reductions per year) [89]. The results of this study can be explained through the following aspects.

Firstly, the implementation of the CRZ policy strengthened the attention of the public to the environment, and the administrators set stricter emissions standards and regulatory mechanisms. In the CRZ policy, the government has successively issued a series of environmental regulations to control air pollutant emissions, such as the “Special Action Plan for Environmental Protection in the Comprehensive Reform Zone of Shanxi Province” and the “Emergency Plan for Seriously Polluted Weather in the Comprehensive Reform Demonstration Zone of Shanxi Province”.

Secondly, the government integrated and regulated the coal market and the electricity market, established an ecological environment property rights and polluter-pays system, and developed the paid use of pollutant emission rights and the emission rights trading market. The establishment of emissions trading markets and the increase in emission cost force enterprises to improve technology and production methods to control pollutant emissions.

Thirdly, the CRZ policy emphasized resource conservation, consumption reduction, emission reduction, and pollution control. Re-calculating the cost of resource-based products and increasing the price of resource products, establishing an energy-saving evaluation system and energy-saving benchmark management for high-energy-consuming industries such as metallurgy, coke, electric power, and chemical building materials and expanding the development and utilization of renewable and clean energies. The photovoltaic power generation industry in Shanxi province leads the country. By saving energy and developing renewable clean energy, the CRZ policy has reduced SO₂ emissions effectively.

4.6.2. Discussion of the Dynamic Effects Test

Table 5 and Figure 5 demonstrate the dynamic effect of the CRZ policy. Based on the results of the dynamic effects test, we found that the CRZ policy did not produce a significant effect on SO₂ emission reduction in the first stage of policy implementation (2011–2015) and the policy effect of the CRZ on PCSO₂ and PGSO₂ started to be noticeable and strengthened year by year in the second stage of policy implementation (after 2016). The results can be explained as follows.

Firstly, the resource-dependent economic development path formed in Shanxi Province in the past few decades is very stubborn, and the barriers to the exit of the resource-based enterprises in Shanxi Province are high. This led to the reform process of resource-based enterprises, particularly the state-owned coal enterprises, being slow, and the CRZ policy encountered great resistance in the early stage. The difficulty in the transformation of resource-based industries in Shanxi Province also led to the fact that production factors are still concentrated in resource-based industries, and other non-resource-based industries lack the input of production factors. Thus, the development of emerging industries and service industries is slow, and the CRZ policy did not produce a significant impact on SO₂ emissions in the first stage of policy implementation.

Secondly, implementing a new policy always takes time to work, and it may conflict with the original policy, which also needs time to resolve [88]. The CRZ policy is a comprehensive exploration and reform, including economic, social, environmental, and people’s livelihood. The implementation of the policy requires a transition and preparation stage, such as the plan and construction of industrial parks and the improvement of infrastructure. In the first stage of the implementation of the CRZ policy, owning to a lack of reference experience and the infrastructure, specific reform measures and policy systems are not perfect and need to be gradually improved in practice. Thus, the CRZ policy at this stage did not produce substantial effects on SO₂ emissions.

Finally, we found that the emission reduction effect of the CRZ policy on SO₂ became more and more obvious in the second stage of the implementation of the CRZ policy. This may be due to after the exploration and preparation in the first stage, the policies and systems of various transformation reforms have gradually been improved. Besides, the infrastructure and supporting services also have been gradually improved, and the CRZ policy gradually entered the stage of formal implementation and promotion. In November 2016, the Preparatory Committee of the Management Committee of Shanxi Transformation and Comprehensive Reform Demonstration Zone was formally established, marking that the implementation of the CRZ policy has entered a substantive stage of advancement. Since then, Shanxi Province has made significant progress in industrial transformation, energy conservation, emission reduction, efficiency improvement, and government supervision. For example, promoting the transformation and upgrading of traditional manufacturing industries such as equipment manufacturing, introducing the big data, internet of things, and other information industries, and conducting pilot projects for comprehensive reform of the energy revolution. Therefore, the effect of the CRZ policy on reducing SO₂ emissions began to manifest.

4.6.3. Discussion of the Mechanism Analyses

Combining the results of Table 4 and Table 6, we find that the industrial structure, human capital, opening to the outside world, and investment in science and education are the main ways that CRZ policies affect SO₂ emissions. In fact, Bakhsh and Akmal et al., also found that FDI and population density are significantly related to SO₂ emissions [90]. Contacting the specific implementation of the CRZ policy, we explain the impact mechanism as follows.

Firstly, optimizing and upgrading the industrial structure are important ways to transform a resource-based economy [33]. After the implementation of the CRZ policy, the local government has issued a series of policies to encourage, support, and guide the production factors transfer from the resource-based industries to non-resource-based industries and promote the development of emerging industries, such as big data, medical and health care, new materials, and new energy vehicles. Moreover, the integration and regulation of traditional industries have been accelerated, a large number of small-scale resource-based enterprises with high investment, high consumption, high pollution, and low efficiency have been closed, and the large-scale resource-based enterprises have been pushed to upgrade and transformed in technology and production methods. The CRZ policy promoted the upgrading of the industrial structure by strengthening the transformation and upgrading of traditional industries and developing emerging industries, further reducing pollutant emissions.

Secondly, the level of human capital reflects the quality of the population, which contributes to improving the environmental protection awareness of enterprises and individuals and improving productivity and production efficiency [91]. In the process of resource-based economic transformation in Shanxi Province, CRZ policy pointed out to attract high-level talents and increase financial support in key fields and improve the salary and promotion system of scientific researchers. The policy also encourages in-depth cooperation between scientific researchers and enterprises, strengthens personnel training, and promotes exchanges and cooperation between universities and enterprises. Therefore, through these initiatives, CRZ policy has effectively improved human capital and further reduced pollutant emissions.

Thirdly, the opening to the outside world can strengthen the exchanges and cooperation of resource-based cities with the developed regions, attract more investment for non-resource-based industries, and improve the enterprise management capabilities and production efficiency [92]. The CRZ policy draws useful development experiences from the “free trade zone”, such as simplifying the administrative approval procedures and improving the service capabilities of the government. The local government issued a series of regulations to promote the facilitation of foreign trade, overseas investment, and the production factors such as capital and technology flow into Shanxi Province. Besides, the CRZ policy promoted the development of low-carbon, green, and circular industries in the provincial-level economic and technological development zones and encouraged the development of high-tech industrial clusters. Therefore, the CRZ policy promoted the FDI, the development of emerging industries, and the introduction of technologies by expanding the opening to the outside world and, thereby, improved the SO₂ emissions in Shanxi Province.

Finally, the investment in science and education is conducive to accelerating regional technological innovation and improving production efficiency and pollution control capabilities [93]. The CRZ policy proposed to build the national independent innovation demonstration zone, especially in terms of emission reduction and new energy, to strengthen technological innovation and advancement and to enhance the capacity of new energy exploration and utilization. In addition, CRZ policy also promoted building a professional technical alliance to speed up scientific and technological research and the transformation of scientific research results and improve development efficiency. Therefore, investment in science education is also an important path for the CRZ policy to reduce regional SO₂ emissions.

5. Conclusions and Suggestions

As a typical air pollutant and greenhouse gas, SO₂ not only caused ecological problems such as acid rain and smog but also caused great harm to human health. To achieve the goals of ecological environment governance and sustainable development, we must pay attention to controlling SO₂ emissions. Meanwhile, in the context of sustainable development, many resource-based cities and developing countries have formed resource-dependent development paths and inefficient economic development models in the process of economic development and are facing dual pressures from resource environment and economic development. How to reasonably solve the contradiction between economic development and resource environment is the dilemma faced by many resource-based cities and developing countries. How to promote the orderly transformation of the economic development model should be an important topic in the field of sustainable development research. As the largest developing country in the world, the environmental problems and development contradictions faced by China are particularly prominent and representative. Therefore, Shanxi Province is the first national comprehensive reform demonstration zone with the theme of a resource-based economy in China. The accurate evaluation of the effectiveness of CRZ policy has substantial policy significance and reference value for the other resource-based cities and developing countries to achieve economic transformation and sustainable development.

This study treated the CRZ policy as a quasi-natural experiment, collected the panel data of 248 prefecture-level cities in China from 2003 to 2018, and applied the PSM-DID method to evaluate the impact that CRZ has on SO₂ emissions in the pilot zone. According to the benchmark regression, placebo test, dynamic effect test, and mechanism analysis, the main conclusions are summarized as follows:

Firstly, based on the results of empirical research, we found that the CRZ policy has significantly reduced the PGSO₂ and PCSO₂. Reliability of results was demonstrated through the placebo test by randomized repeated trials.

Secondly, the results of the dynamic effect test show that there is a lag in policy effect. The impact of CRZ policy on PCSO₂ and PGSO₂ was not significant in the first stage of policy implementation (2011–2015). However, the CRZ policy significantly reduced the PCSO₂ and PGSO₂ in the second stage of policy implementation (2016 and after) and the reduction effect of the CRZ policy on PCSO₂ and PGSO₂ is strengthened over time.

Thirdly, we found that optimizing the industrial structure, improving human capital level, expanding opening to the outside world, and increasing investment in science and education are the main ways for the CRZ policy to reduce the regional SO₂ emissions.

Through the analysis and research of this paper, we found that the CRZ policy is indeed effective in reducing SO₂ emissions, which is of great significance for improving the atmospheric quality and realizing sustainable development in resource-based cities and developing countries. Further, we put forward the following suggestions for the SO₂ emissions reduction in other resource-based cities.

Firstly, the government must pay attention to and accelerate the transformation and upgrading of the industrial structure. On the one hand, the government should increase the policy support and subsidies for resource-based enterprises and the traditional resource-based enterprises should increase the investment in R&D to promote technological upgrading and efficiency improvement in traditional industries. On the other hand, the government should strengthen the effort of investment promotion, actively introduce high-tech industries, and promote regional industrial diversification and structural upgrading.

Secondly, innovation and talent are of vital importance for the reduction in SO₂ emissions of resource-based cities; managers must pay attention to the investment in science and education. Many previous studies have shown that resource-based industries have a crowding-out effect on both human capital and innovation. Therefore, in the process of pushing the economic transformation for resource-based cities, it is necessary to make up for the shortcomings of innovation and talents. The promotion of innovation and talent needs to start from both the government and the enterprise. The government should introduce special policies to attract talents, simplify administrative procedures, and encourage innovation. The enterprises should improve the talent introduction mechanism and improve the welfare and pay of technical talents.

Thirdly, it is necessary for resource-based cities and developing countries to further expand the opening to the outside world and strengthen exchanges with developed countries and regions. The government should strengthen exchanges and cooperation with developed regions, absorb advanced experience and technology, and improve production efficiency. In addition, creating a fair business environment, attracting foreign investment, and expanding import and export trade are of vital importance for the further expanding opening to the outside world.

This paper empirically studied the impact and dynamic trend of the CRZ policy on SO₂ emissions through PSM-DID and further analyzed the impact mechanism. However, there are still some deficiencies in this study. Owing to the changes in the statistical caliber of smoke and dust emissions and the availability of data, this article only studied the impact of the CRZ policy on SO₂ emissions. However, the transformation of the resource-based city is a comprehensive reform, which may have an impact on economic structure, development quality and efficiency, and other pollutant emissions. Therefore, further study should analyze the impact of the transformation of the resource-based city from more perspectives, such as economy, society, and environment. Besides, this study mainly used data at the prefecture-level city level, which is still a macro-level study. Therefore, further research can deep into micro-fields of industries and enterprises to obtain a more precise evaluation and guidance for the transformation of a resource-based economy. Finally, this study used the PSM-DID to alleviate the selection bias and endogeneity to guarantee the accuracy and rationality of the evaluation results as much as possible. However, evaluating the effects of a policy is a complex problem; we cannot completely eliminate the intervention of other factors and the systematic bias. Therefore, to further verify the reliability of the results, further study can consider using other policy evaluation methods to evaluate the policy effect.