Can pollution charges reform promote industrial SO 2 emissions reduction?

Abstract

This paper evaluates the real effects of pollution charges reform on ${SO}_{2}$ emissions at the city level. Using the establishment of Comprehensive Work Plan for Energy Conservation and Emission Reduction (The Plan) in China as a quasi-natural experiment, our difference-in-differences estimation shows that: (1) increasing pollution charges has a significant and positive effect on controlling industrial waste ${SO}_{2}$ , and this relationship is robust to different specifications and alternative measures; (2) ${SO}_{2}$ emissions can be reduced to achieve reduction targets by forcing companies to strengthen the degree of process production controls and end-of-pipe treatment; (3) eastern and western cities have a better pollution control effect than the other regions, and large cities have better emission reduction effects than smaller cities. Overall, this paper sorts out the evolution of China’s pollution charge policy, and reveals the mechanisms behind the real effects of pollution charges reform on ${SO}_{2}$ emissions, thus providing timely implications for policymakers concerned with environmental protection.

Keywords

Pollution charges environmental pollution difference-in-differences propensity score matching

Introduction

Since the reform and opening up of China, China’s economy has accomplished world-renowned achievements, but the contradiction between economic development and environmental protection has become increasingly acute. When exploring the policy path of balanced economic and environmental development, the government realized the traditional “command-control” environmental regulation was far less efficient than the market-based incentive environmental regulation.^1,2 In view of this, China has gradually started to experiment with incentive-based regulatory tools whose essence is economic and voluntary means, and its representative is the pollution charges. Assessing the reduction effects of pollution charges is of great significance for improving China’s environmental policies, promoting ecological environment construction, and achieving green development.

Pollution charges (or pollution tax) is an environmental regulation system that follows the “polluter pays” principle. Its principle is consistent with the “Pigou tax,” which internalizes external costs and promotes the reduction of pollution by enterprises.³ Pollution charges has exhibited significant effects on industrial pollution control in developed countries, but “endogenous enforcement” are common in developing countries.^4,5 China, as the largest developing country, has been widely questioned about the control effect of pollution charges. Some scholars believe that the sewage charge system has achieved good results, which can significantly reduce pollution emissions and improve the firms’ production efficiency.^6–9 However, some scholars believe that the pollution charges have not achieved the expected results, and even have side effects on urban green innovation and development.^10,11 At present, most literatures use provincial data instead of prefecture-level city data, so most researches on domestic emission trading market have the defects of small sample size and weak robustness of results. At the same time, most scholars use the index of actual levy rate of sewage charges to evaluate the effect of pollution control. The main problem of using this index in the regression equation is the endogenous problem that cannot effectively control the actual levy rate. Actually, there is a big difference in pollution levy rates in different regions, which are affected by a variety of factors, such as local economy, culture, etc.^12–14

The existing literature has the following problems. First, most authors use the indicator of the actual rate of pollution charges to evaluate the effect. Due to the serious endogenous enforcement problems in China, the indicator has endogenous problems, that is, the actual rate of pollution charges and environmental pollution affected by economic, policy and other factors. Second, the statistics on pollution charges in China are only provincial-level data, so the sample size is limited, and the research focuses on the differences in policy effects among various provinces. Third, a few researchs use dummy variables to measure the impact of pollution charges reform, but the data are only available at the provincial level, and no consensus conclusions have been reached on the impact of pollution charges reform on environmental governance.

In an attempt to fill this gap, we evaluate the effect of pollution charges by taking advantages of a natural experiment with a country setting i.e. the issue of the The Plan in 2007 in China, which pronounces some provinces doubled the ${SO}_{2}$ emission levy rates (increasing from 0.63 yuan per unit pollutant to 1.26 yuan per unit pollutant). Since only some provinces implemented the policy before 2015, this provided an excellent quasi-natural experiment. We uses the difference-in-difference (DID) method to test the effect of pollution charges reform on ${SO}_{2}$ emissions in prefecture-level cities, which can more effectively identify the causal relationship between the pollution levy system and the outcome variables.

Specifically, we collect information on the adjustment of ${SO}_{2}$ emission levy standards in various prefecture-level cities from 2003 to 2014 and constructed panel dataset. Using this data, we find that raising pollution charges can significantly reduce industrial ${SO}_{2}$ emissions. Since The Plan does not involve industrial smoke (powder) and solid waste, the regression results for industrial smoke (powder) show that raising ${SO}_{2}$ emission charge rate has a limited effect on reducing it, which supports our research conclusions. The increase in pollution charges may not be random, but will be affected by indicators such as regional economic development level, environmental quality, and energy consumption. Therefore, the outcome variables of the treatment and control groups may be unbalanced due to the non-random selection process. Simply using DID method may be biased. To solve this problem, we combine the propensity score matching (PSM) method with the DID approach.¹⁵ The results show that the conclusion is still reliable.

This study makes several contributions to the literature. First, according to the reform of the pollution charges in 2007, this paper compares the ${SO}_{2}$ emissions of the treated group and the controlled group, and uses the DID method to accurately identify the effects of the pollution charges under certain conditions. Second, we collect information on the adjustment of the SO₂ emission levy rates at the prefecture-level cities from 2003 to 2014, and construct prefecture-level panel dataset to make up for inconsistent research results caused by provincial data. Third, we provide timely implications for policymakers to further improve the environmental protection tax system and achieve the energy conservation and emission reduction goals. We also call on local governments to strengthen environmental protection.

This paper proceeds as follows. The next section provides a brief history of sewage charges policy and research hypothesis. Section “Results” describes the research design. Section “Results and analysis” presents and analyses the regression results, and Section “Mechanism analysis” provides mechanism transmission. Conclusions and policy implications are presented in Section “Conclusions and policy implications.”

Institutional background and research hypothesis

Institutional background

Pollution charges were first introduced in OECD countries in the early 1970s. Drawing on the experience of environmental management in developed countries, China promulgated the Environmental Protection Law of the People’s Republic of China (Trial) as early as September 1979, establishing China’s pollution charge system. The Interim Measures on Pollution Charges, introduced in 1982, clearly stipulated that excessive emissions of pollutants was charged for the first time. In 1988, the Interim Measures for the Payment Use of the Special Fund for Pollution Source Control was promulgated to standardize the use of funds for pollution control. The Measures for the Administration of Collection of Pollution Charges, which came into effect in 2003, divided the pollution charges into “instance pollution charge” and “double charge for exceeding the standard.” Nevertheless, the excessive growth of energy-intensive and pollution-intensive industries has greatly worsened the environmental situation in China. Six major industries, including electric power, steel, construction materials, chemical industries, etc., account for 70% of the country’s ${SO}_{2}$ emissions and industrial energy consumption.

In order to achieve the goals of the 11th Five-Year Plan, The Plan issued by China in 2007 doubled the standard for ${SO}_{2}$ emissions charges (increase from 0.63 yuan/kg to 1.26 yuan/kg). Several provinces have carried on The Plan at different times, and the specific description is listed in Table 1. There are 16 provinces (municipalities) raised the ${SO}_{2}$ emission charges over the period from 2007 to 2014, including Jiangsu, Shanghai, Hebei, Shanxi, Shandong, Guangxi, Inner Mongolia, Yunnan, Guangdong, Liaoning, Beijing, Tianjin, Heilongjiang, Xinjiang, Ningxia, Zhejiang. Seven provinces (municipalities), including Shanxi, Shanghai, Hebei, Shandong, Guangxi, Inner Mongolia, and Yunnan, enforced The Plan between 1 July 2008 and 30 July 2009. Since only some provinces implemented the policy before 2015, it provides an excellent quasi-natural experiment to identify the emission reduction effect of the pollution charges.

Table 1.

Provincial revised standard of ${SO}_{2}$ pollution charges from 2007 to 2014.

Implementation time	Provinces (municipality)
2007	Jiangsu
2008	Shanxi, Hebei, Shandong, Inner Mongolia
2009	Shanghai, Guangxi, Yunnan
2010	Guangdong, Liaoning, Tianjin
2012	Heilongjiang, Xinjiang
2014	Beijing, Ningxia, Zhejiang

Research hypothesis

According to the “Pigou tax” theory, the pollutants emitted by enterprises have negative externalities. When environmental regulations are looser, enterprises have no incentive to convert pollutants into production costs and reduce pollution emissions, so pollutant emissions will be excessive. When the environmental regulation is strict, the social cost of environmental pollution will be internalized into the production cost of enterprises. The increasing of the unit cost of production and emissions bring the cost of enterprises closer to the social cost, thereby forcing enterprises to effectively allocate resources and maintaining the discharge of pollutants at the best of society.

When industrial ${SO}_{2}$ pollution charges increase, the intensity of environmental regulations increases as well, which raises the marginal cost of ${SO}_{2}$ emissions. The increase production costs of company’s expectation will reduce profits. Therefore, companies tend to reduce industrial ${SO}_{2}$ emissions. Enterprises with high pollution and high energy consumption are even forced to withdraw, which promotes industrial upgrading and increases the proportion of emerging industries.¹¹

In terms of mechanism, enterprises may reduce the emission of ${SO}_{2}$ in the production process through two ways.

First, enterprises abate ${SO}_{2}$ pollution intensity of process production by pollution prevention. In general, they reduce the generation of ${SO}_{2}$ by improving production efficiency and adopting other control measures. For example, enterprises use clean energy to replace traditional coal combustion, which is the main source of industrial ${SO}_{2}$ . Another method is to improve the production technology of process personnel.

Second, enterprises can reduce ${SO}_{2}$ pollution by strengthening the degree of end-of-pipe treatment. In order to increase the ${SO}_{2}$ removal rate, enterprises make efforts to improve the efficiency of emission abatement, including supporting advanced pollution control equipment and innovation in pollution treatment technology.¹⁶

Intuitively, companies can evade the increased payment of pollution charges by relocating to other areas where they did not enforce the reform, but in fact, all provinces have indeed completed the adjustment of ${SO}_{2}$ emission charges standards by 2016. As it is expected that all provinces will complete the adjustment in the near future, enterprises will lack incentives for relocation.⁸ Based on the above analysis, We therefore formulate:

H1: Increasing the pollution charges can significantly reduce industrial ${SO}_{2}$ emissions.

H2: Enterprises can effectively reduce industrial ${SO}_{2}$ emissions by strengthening end-of-line governance and process production control.

Results

Model

Since the introduction of the pollution levy system, China has raised the charge standard three times, in July 2003, May 2007, and September 2014, respectively. Considering the influence of the other two pollution charges reform, we use the panel dataset at at prefecture level from 2003 to 2014. Also, we choose a DID design to identify The Plan’s net effect of ${SO}_{2}$ emission reduction in China and introduce two dummy variables: (1) Group dummy variables: all samples are divided into two groups according to whether the charge rate is increased or not. The cities coded 1 if they have raised the pollution charge standard are the treated group, while those cities that have not raised the pollution charge standard are the control group coded 0; (2) Time dummy variable: the sample time span is divided into two periods: before the enforcement of The Plan (2003–2008) and after the implementation of The Plan (2009–2014). Specifically, we choose seven provinces, who implemented The Plan between 1 July 2008 and 30 June 2009, as the treated group, and we select 15 provinces, who never executed The Plan during the period from 2007 to 2014, as the control group.^a Due to the availability of data, 68 prefecture-level cities are selected as the treated group and 121 prefecture-level cities are included in the control group. We construct the following DID model: ${SO}_{2 it} = α_{0} + β_{1} Reform + {γX}_{it} + μ_{i} + λ_{t} + ε_{it}$ where ${SO}_{2 i t}$ is the ${SO}_{2}$ emission per unit gross industrial output value, which represents the level of environmental pollution. $Reform$ is the estimator of the DID method, whose regression coefficient $β_{1}$ reflects the net effect of the implementation of The Plan on the ${SO}_{2}$ emission per unit GDP. The equation also includes control variables ( $X_{i t}$ ), including the economic level, foreign direct investment, urbanization rate, and other variables that affect environmental pollution. $μ_{i}$ , is the city fixed effects, and it is controlled for all permanent unobserved determinants of pollution across cities. $λ_{t}$ represents the year fixed effects, and ε_it stands for the error term.

In order to accurately identify the effect of The Plan, we also use the PSM method to solve the sample selection bias problem. We combine the PSM and DID (PSM-DID) method to carry out robust estimation. The basic idea is as follows: when evaluating the effect of a policy, as long as we find and match the samples whose features are closest to the samples in the treated group, then we use the matched samples for DID estimation so that we can greatly solve the sample selection bias problem caused by the estimation error and improve the accuracy of the empirical results. The specific model is as follows:

{SO}_{2 i t}^{PSM} = α_{0} + β_{1} Reform + γ X_{i t} + μ_{i} + λ_{t} + ε_{i t}

(2)

where the variables have the same meaning as before.

Variables and data description

In this paper, the dependent variable is the industrial ${SO}_{2}$ emissions, divided by the gross industrial output value of the city, which measures the ${SO}_{2}$ emission intensity.¹⁷ Other variables affecting pollution are also controlled in this paper, which are listed as follows:

(1) GDP per capita ( $Pgdp$ ): The level of economic growth is an important factor of regional pollution,^18,19 expressed in the logarithm of regional per capita GDP. We also conclude squared logarithm of per capita GDP ( $Rpgdp$ ) to test the existence of environmental Kuznets curve (EKC). (2) Foreign direct investment ( $Fdi$ ): The impact of foreign direct investment on the environment is still controversial,^20,21 so we use the actual utilization of foreign direct investment in regions to test whether the Pollution Paradise Hypothesis is valid at city-level in China. (3) Industrial structure ( $Secind$ ): The influence of industrial structure on the environment will change with time and environmental regulation intensity, showing a nonlinear relationship.^16,19 This paper controls the proportion of secondary industry and its square term ( $Rsecind$ ) to identify the role of industrial structure on pollution. (4) Population density ( $P d$ ): We introduce regional population density to test the impact of population size and population agglomeration on the emission of industrial ${SO}_{2}$ . (5) Urbanization rate ( $Urbanr$ ): Researchers have different opinions on the impact of urbanization on pollution. Some scholars think that urbanization aggravates environmental pollution,^22,23 while others believe that pollution tends to be alleviated with the expansion of urban scale.^24,25 Urbanization is defined as the proportion of non-agricultural population in the urban population.

This paper collects relevant data of 189 cities in China from 2003 to 2014, all of which are taken from the CEIC Database, China Macro Database, the China Urban Statistical Yearbook, the China Environmental Yearbook, and the China Regional Statistical Yearbook. Table 2 shows the descriptive statistics of the panel data.

Table 2.

Descriptive statistics of main variables.

		All the samples			The treated group			The control group
Variable	Symbol	Obs	mean	Std. Dev.	Obs	mean	Std. Dev.	Obs	mean	Std. Dev.
${SO}_{2}$ emissions per unit of gross industrial value	${S O}_{2}$	2268	1.167	1.91	816	1.438	2.39	1452	1.016	1.558
GDP per capital	$Pgdp$	2268	9.642	0.702	816	9.778	0.774	1452	9.565	0.646
Foreign direct investment	$Fdi$	2268	1.813	1.984	816	1.525	1.676	1452	1.974	2.122
Industrial structure	$Secind$	2268	49.351	10.412	816	50.399	10.188	1452	48.762	10.494
Population density	$P d$	2268	5.775	0.834	816	5.554	0.925	1452	5.899	0.751
Urbanization rate	$Urbanr$	2268	24.708	17.849	816	21.453	20.515	1452	26.537	15.877

Results and analysis

Impact of increasing sewage charges on environmental pollution

Table 3 examines the effect of increasing charge rate on industrial ${SO}_{2}$ emissions. The estimated coefficient of interest is that for Reform, a dummy variable that equals one if a given city implemented The Plan. We use the two-way fixed effect model to conduct DID estimation, and we also add mixed OLS regression and one-way fixed-effect estimation as robustness comparison. The results of these three estimation methods are consistent. This paper mainly discusses the two-way fixed effect regression results, column (5) shows the average effect of The Plan on ${SO}_{2}$ emissions while controlling for fixed effects of city and year. The coefficient of Reform is –0.409 and significantly negative; indicating that the increase of pollution charges significantly reduces industrial ${SO}_{2}$ emissions. In column (6), we include control variables, and find that coefficient of Reform increases only slightly to –0.830 and remain positively significant. These results demonstrate the validity of Hypothesis 1.

Table 3.

Basic regression results of pollution charge reform and ${SO}_{2}$ emissions.

	OLS		One-way FE		Two-way FE
	(1)	(2)	(3)	(4)	(5)	(6)
Reform	–0.694***	0.090	–1.694***	–0.553***	–0.409***	–0.830***
	(0.087)	(0.137)	(0.116)	(0.112)	(0.106)	(0.106)
Pgdp		0.934***		11.117***		16.787***
		(0.167)		(1.193)		(1.128)
Rpgdp		–0.011		–0.436***		–0.396***
		(0.007)		(0.052)		(0.038)
Fdi		–0.082***		–0.084***		–0.096***
		(0.030)		(0.012)		(0.020)
Secind		–0.017		–0.045		–0.029
		(0.042)		(0.030)		(0.027)
Rsecind		0.000		0.000		0.000
		(0.000)		(0.000)		(0.000)
Pd		–0.194		–1.912***		–2.358***
		(0.133)		(0.610)		(0.764)
Urbanr		0.016***		–0.001		0.001
		(0.004)		(0.002)		(0.002)
Constant	1.292***	12.775***	1.442***	61.341***	2.918***	126.232***
	(0.105)	(1.717)	(0.186)	(5.898)	(0.088)	(8.795)
Year fixed effect	No	No	No	No	Yes	Yes
City fixed effect	No	No	Yes	Yes	Yes	Yes
Observations	2268	2268	2268	2268	2268	2268
R-squared	0.020	0.221	0.494	0.668	0.370	0.464

Note: Robust standard errors are in parenthesis. Controls include per capita GDP ( $pgdp$ ), foreign direct investment ( $fdi$ ), secondary industry proportion ( $secind$ ), population density ( $p d$ ), and urbanization rate ( $urbanr$ ). ***, ** and * respectively represent significant at the levels of 1%, 5%, and 10%.

We simply discuss the control variables. The results exhibit an inverted-U-shaped relationship between economic growth and pollutant emissions, suggesting the existence of EKC. Along with economic growth, the ${SO}_{2}$ emissions will rise at early stage, then fall. The Pollution Paradise Hypothesis is not valid at the city level in China. On the contrary, the increase of FDI will reduce ${SO}_{2}$ emissions, which is consistent with the conclusions of He²⁶ and Xu and Deng.²¹ Because production activities of FDI, like pollution control, also have the nature of increasing returns on scale.²⁷ Pollution Paradise is only a short-term phenomenon, and FDI can affect income and technology levels, thereby improving environmental quality.²⁸ No evidence shows that there is an non-linear relationship between pollution intensity and industrial structure. The increase in population density significantly abates industrial ${SO}_{2}$ emissions, indicating that population agglomeration will promote environment governance. Urbanization will aggravate the problem of air pollution, but the impact is not significant. The reasons for this may be that: the urbanization process leads to urban overcrowding and aggravates environmental pollution. However, due to the urban scale effect, it improves the utilization efficiency of public environmental resources and promotes pollution control. The two effects offset each other.

Balance trend and dynamic effect analysis

Before using the DID method, it is necessary to test whether the balance trend hypothesis is supported; that is, in the absence of policy impact, the ${SO}_{2}$ emission intensity of the treated group and the control group should have a consistent trend. Moreover, the basic regression results cannot reflect the difference in emission reduction effect of the implementation of The Plan in different periods. Some policies have a hysteresis effect and play a positive role at the beginning, but the dynamic effect of policy changes will decrease significantly over time.²⁹ Therefore, this paper uses the event study method proposed by Jacobson et al.³⁰ to test the dynamic effect of the changing pollution charges. We estimate

{SO}_{2 i t} = α_{0} + \sum_{2004}^{2014} β_{t} Reform \times λ_{t} + γ X_{i t} + μ_{i} + λ_{t} + ε_{i t}

(3)

where the base year of The Plan is 2009, and other variables have the same meaning as above.

Figure 1 illustrates the estimation results of the coefficient $β_{t}$ under 95% confidence interval. $β_{t}$ was not significant from 2004 to 2008, indicating that there is no significant difference in ${SO}_{2}$ emissions between the treated group and the control group before the change of pollution charges, which is satisfied with the balance trend hypothesis. After changing the charges in 2009, $β_{t}$ began to be significant and gradually increased, indicating that the cities that raised the ${SO}_{2}$ emission charges in 2009 effectively reduce the emission of ${SO}_{2}$ unit of gross industrial output value. Moreover, the absolute value of the coefficient is also increasing along with time, indicating that the emission reduction effect of this policy increases continuously.

Figure 1.

Dynamic effect diagram of pollution charge reform and ${SO}_{2}$ emissions.Note: Vertical bands represent +(−)1.96 times the standard error of each point estimate.

PSM estimation

Data deviation and confounding variables may lead to inaccurate estimation results. In order to control the systematic differences between the treated group and control group, we further uses a combination of DID and PSM (PSM-DID) design to re-examine the empirical results. Specifically, the cities with the closest propensity scores will be matched to form a matched treated group and a control group, according to the probability of adjusting pollution charges in each city. Then, we carry out the DID estimation again with the matched groups.

The premise of using the PSM-DID method must satisfy the co-supporting hypothesis; that is, there should be no significant difference in the mean of covariates between the treated group and the control group before and after the matching. Table 4 shows that, almost all covariate coefficients are not significant, that is, there is no significant difference between all the covariates before and after matching, but there is a significant difference between the result variables, which confirms the rationality of the use of PSM-DID. In this paper, we use the kernel matching method to test the robustness of the impact of changing pollution charges on industrial ${SO}_{2}$ emission reduction. Figure 2 shows the density function diagram of the propensity score of matching effect test between the treated group and the control group. We can conclude that the propensity score probability density of the treated group and the control group is relatively close after matching. Table 5 presents the results of PSM-DID method, where the DID coefficient is –0.716 and remains significant, indicating that raising pollution charges can significantly reduce the ${SO}_{2}$ emissions. The PSM-DID test results are consistent with DID estimation results, suggesting that the above conclusions have certain robustness.

Table 4.

Balance test of matched samples.

Variable	Control Mean	Treated Mean	Diff.	\|t\|	P > \|t\|
${SO}_{2}$	1.361	2.314	0.953	6.58	0.0000***
Pgdp	9.422	9.408	–0.013	0.38	0.7029
Rpgdp	116.103	115.096	–1.007	0.85	0.3950
Fdi	1.808	1.828	0.020	0.17	0.8641
Secind	49.853	49.312	–0.541	0.82	0.4113
Rsecind	2605.631	2548.952	–56.679	0.89	0.3762
Pd	5.586	5.543	–0.042	0.79	0.4297
Urbanr	29.574	27.846	–1.728	1.87	0.0614*

Note: The original hypothesis was that there was no significant difference between the covariates of the treated group and the control group. ***, **, and *, respectively, represent significant at the levels of 1%, 5%, and 10%.

Figure 2.

Probability distribution density function of the propensity score. (a) before matching, (b) after matching.

Table 5.

Estimation results of pollution charge reform and ${SO}_{2}$ emissions using PSM-DID method.

	Diff.	S. Err.	\|t\|	P > \|t\|
Difference between treated group and control group before policy	0.953	0.105	9.07	0.000***
Difference between treated group and control group after policy	0.237	0.104	2.28	0.023**
Diff-in-Diff	–0.716	0.148	4.84	0.000***

Note: Controls including per capita GDP ( $pgdp$ ) and squared term, foreign direct investment ( $fdi$ ), secondary industry proportion ( $secind$ ) and squared term, population density ( $p d$ ) and urbanization rate ( $urbanr$ ) have been controlled. ***, ** and * respectively represent significant at the levels of 1%, 5%, and 10%.

Placebo test

We use the counterfactual method and construct a treated group to carry out the placebo test.³¹ There are many kinds of industrial pollutants. Since the adjustment in The Plan is mainly for industrial ${SO}_{2}$ , and industrial smoke (powder) is not the emission reduction target, it is expected that The Plan may not affect the emissions of smoke (powder). Because of this, we replace the dependent variable in the benchmark regression model with smoke (powder) emissions, divided by gross industrial output value, and conduct a placebo test (Table 6).^b

Table 6.

Estimation results of placebo test.

	DID	DID	PSM-DID
	(1)	(2)	(3)
Reform	0.025	–0.131	–0.128
	(0.111)	(0.118)	(0.134)
Constant	1.774***	54.199***	0.767***
	(0.092)	(9.777)	(0.067)
Controls	No	Yes	Yes
Year fixed effect	Yes	Yes	Yes
City fixed effect	Yes	Yes	Yes
Observations	2268	226	2191
R-squared	0.145	0.177	0.052

Note: Robust standard errors are in parenthesis. Controls include per capita GDP ( $pgdp$ ) and squared term, foreign direct investment ( $fdi$ ), secondary industry proportion ( $secind$ ) and squared term, population density ( $p d$ ), and urbanization rate ( $urbanr$ ). ***, **, and *, respectively, represent significant at the levels of 1%, 5%, and 10%.

The regression results of the DID and PSM-DID model show that the change of ${SO}_{2}$ pollution charges has no significant effect on smoke (powder) emissions. This test has two effects. For one thing, it rules out the influence of other environmental regulation policies, such as “two control zones” policy and “emissions trading” policy. If these policies work, air pollutant emissions excluding ${SO}_{2}$ should also be reduced. The results of the placebo test do not support this conclusion. It also enhances the convincing of the empirical results in this paper. For another, it reflects the limitation of the emission charge system, which only has a significant emission reduction effect on ${SO}_{2}$ , but it does not seem to have significant emission reduction effect on other conventional air pollutants.

Heterogeneity analysis

Regional heterogeneity

In order to test the relationship between the geographical location of different cities and the effect of changing pollution charges, we divided the sample into three parts, eastern, central and western cities.^c Table 7 columns 2–4 report the regression results. Results show that doubling ${SO}_{2}$ emission charges can significantly abate the ${SO}_{2}$ emission in eastern and central cities. In contrast, for western cities, it promotes industrial ${SO}_{2}$ emission, but not significantly. The reason may be that there are location advantages and better development environments in eastern and central cities, which can attract more foreign capital and talents. It is easier to introduce pollution treatment technology in these kinds of cities, and also more likely to create new industry. The above analysis shows that the pollutant charges system can reduce ${SO}_{2}$ emissions through technological innovation and industrial upgrading. The innovation ability of western cities is relatively weak. When sewage charge is raised, the suddenly rising pollution cost of enterprise may inhibit technological progress, so the policy does not play a role in pollution control.

Table 7.

Heterogeneity analysis of the effect of changing pollution charges on ${SO}_{2}$ emissions.

	Geographical position			City size
	Eastern cities	Central cities	Western cities	Medium-sized cities	Large cities	Type II cities	Type I Cities	Megacities and above
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Reform	–0.332***	–0.937***	–0.095	0.165	–0.837***	–1.471***	–0.781***	–0.273*
	(0.105)	(0.178)	(0.291)	(0.236)	(0.109)	(0.275)	(0.174)	(0.143)
Constant	–4.235	30.787**	155.158***	1.611	121.673***	165.612***	144.589***	73.709***
	(10.766)	(12.510)	(23.820)	(17.381)	(9.875)	(24.104)	(19.033)	(15.528)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
City fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	480	1056	732	60	2196	672	70	816
R-squared	0.599	0.684	0.455	0.845	0.457	0.407	0.567	0.477

Heterogeneity of city size

The impact of changing emissions charges on environmental pollution may also be related to city size. This paper conducts a robustness test on the emission reduction effect of The Plan in cities of different sizes.^d Columns 5–9 in Table 7 report the results, where we only report the results of cities above medium size because the sample size of small cities is too small. Results show that if medium-sized cities raise ${SO}_{2}$ emission charges, it will aggravate ${SO}_{2}$ emissions, while the revision of pollution charges in large cities will significantly abate the emission intensity of ${SO}_{2}$ . Among them, the effect of the policy is best in type II cities, followed by type I cities. Doubling the ${SO}_{2}$ emission charges is less effective than II type and type I cities in oversized cities. The reasons may be as follows: on the one hand, larger cities have economic agglomeration effect and are more likely to reduce pollution emissions; on the other hand, larger cities may be overcrowded and are more likely to aggravate urban diseases and environmental problems.^9,32

Mechanism analysis

The above empirical results show that changing ${SO}_{2}$ emission charges can significantly reduce the level of industrial ${SO}_{2}$ pollution, so what is the mechanism of the policy to reduce environmental pollution? From the above theoretical analysis, it can be seen that the implementation of policies will increase the pollution costs and production costs of enterprises. In order to maximize profits and reduce costs, enterprises respond strongly to the charges by either abating ${SO}_{2}$ pollution in the production process or strengthening end-of-pipe treatment for ${SO}_{2}$ emissions.

Emission governance

The increase of ${SO}_{2}$ emission charges force firms to strengthen terminal treatment, which reduce industrial ${SO}_{2}$ emissions by increasing its removal amount when ${SO}_{2}$ generated, that is, increasing the removal rate: ${SO}_{2}$ removal rate = ${SO}_{2}$ removal amount $/ {SO}_{2}$ generation amount. We use industrial ${SO}_{2}$ removal amount generated by coal combustion to test the impact of increasing ${SO}_{2}$ charges on the end-of-pipe governance of firms. In Table 8 column 2, the Reform coefficient is positive and significant, indicating that raising ${SO}_{2}$ emission charges can indeed promote enterprises to strengthen terminal treatment. As a result, the removal rate of industrial ${SO}_{2}$ increase. Column 3 shows the regression results of the removal rate of industrial smoke (powder) as a comparison. The Reform coefficient is negative and not significant, indicating that increasing ${SO}_{2}$ emission charge does not affect the removal rate of industrial smoke (powder).

Table 8.

Mechanism test of the impact of sewage charges on environmental pollution.

	Emission governance		Production process control
	${SO}_{2}$ removal rate	smoke (powder) removal rate	Coal	${SO}_{2}$ production process
	(1)	(2)	(3)	(4)
Reform	0.160***	–0.009	–14.923**	–0.117***
	(0.054)	(0.012)	(6.848)	(0.016)
Constant	–0.979*	–0.703	804.628***	5.328***
	(0.501)	(0.982)	(306.462)	(1.458)
Controls	Yes	Yes	Yes	Yes
Year fixed effect	Yes	Yes	Yes	Yes
City fixed effect	Yes	Yes	Yes	Yes
Observations	2268	2268	1506	1506
R-squared	0.039	0.120	0.032	0.904

Note: Robust standard errors are in parenthesis. Controls, including per capita GDP ( $pgdp$ ) and squared term, foreign direct investment ( $fdi$ ), secondary industry proportion ( $secind$ ) and squared term, population density ( $p d$ ), and urbanization rate ( $urbanr$ ), have been controlled in all regressions. ***, **, and *, respectively, represent significant at the levels of 1%, 5%, and 10%.

Pollution prevention

In order to test whether firms respond to the levy by abating ${SO}_{2}$ pollution in the production process, this paper selects the coal usage and the ${SO}_{2}$ generated amount in the production process, divided by gross industrial output value, to measure the intensity of production process control.^e From columns 4 to 5 in Table 8, DID coefficients are significantly negative, indicating that higher ${SO}_{2}$ emission charges encourage enterprises to strengthen production process control and take adequate measures to improve production efficiency. It is evident that enterprises are motivated to reduce coal consumption and ${SO}_{2}$ production in the production process, thus lowering the emissions. The above results support Hypothesis 2.

Conclusions and policy implications

After constant exploration, on 1 January 2018, China incorporated pollution charges into its tax system, solving the problem of insufficient rigidity in the enforcement of pollution charges. Whether the environmental tax can guide enterprises to reduce pollution emissions is an important issue worth pondering and studying. Taking the practice of China’s pollution charge reform as a quasi-natural experiment, this paper analyzes the environmental effect and mechanism of the pollution levy system by using the data of 189 prefecture-level cities from 2003 to 2014. We find that: (1) the increase of levy rate can significantly reduce ${SO}_{2}$ emissions, and this relationship is robust to different measures, PSM-DID, counterfactual test, etc. It proves that the pollution tax in China can play a positive role in environmental governance at the level of prefecture-level cities and realize green dividend. (2) In terms of the mechanism, enterprises respond to higher levy by either abating ${SO}_{2}$ pollution in the production process or strengthening end-of-line treatment for ${SO}_{2}$ emissions. (3) Further analysis shows that the reduction effect of pollution tax has heterogeneity.

The conclusions of this paper have important implications for improving the environmental protection tax system. Based on China’s current environmental situation, we propose:

First, government should strengthen pollution tax management by raising the levy rates and widening the levy scope. At present, the levy scope of China’s environmental protection tax is relatively narrow, only including four major categories of pollutants. More pollutant emissions, such as ${CO}_{2}$ , should be included in the pollution tax system. At the same time, China should make efforts to reasonably divide the central and local environmental management powers, promote the implementation of the green performance evaluation mechanism for local officials, and strengthened residents’ supervision of channels. Thus, problems such as “collusion between government and enterprises” and “tax shelter” can be solved.

Second, China should set heterogenous pollution tax rates. Taking the differences in economic growth levels, environmental pollution tolerance and emission reduction costs in different regions into account, different tax rates should be set based on local conditions. Government should vigorously promote cross-regional cooperation in environmental governance, breaking through administrative barriers and the “segmentation” system, and then promote the progress of urban pollution control technologies.

Third, supporting mechanism for pollution tax should be established, including tax preferential policies and punishment measures. To encourage production capacity updating and technology innovation, firms who are devoted to cyclic utilization and R&D, can get tax allowance and exemption. Enterprises with excessive emissions should be levied at a higher tax rate. Furthermore, fraudulent companies will get severe penalties.

Footnotes

Authors’ contribution

Conceptualization, D.G, and Y.L.; methodology, D.G.; software, D.G.; validation, D.G., Y.L. and Q.Y.; formal analysis, Y.L.; investigation, Y.L.; resources, D.G.; data curation, D.G.; writing—original draft preparation, L.Y.; writing—review and editing, D.G.; visualization, D.G.; supervision, Q.Y.; project administration, Y.L.; funding acquisition, Q.Y.

Declaration of conflicting of interests

The author(s) declared no potential conﬂicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work is supported by the Graduates’ Innovation Fund (grant number 2019ygscxcy051) of Huazhong University of Science and Technology.

ORCID iD

Da Gao

Notes

Da Gao, PhD, is affiliated with School of Economics, Huazhong University of Science and Technology, Wuhan, China. He focuses on the research of environmental economics and econometrics and has published several papers in related research fields in well-known Chinese and international journals.

Yi Li, Master, is affiliated with School of Economics, Huazhong University of Science and Technology, Wuhan, China. She is interested in the research of industrial policies and environmental economics.

Qiuyue Yang, PhD, is affiliated with School of Economics, Huazhong University of Science and Technology, Wuhan, China. She focuses on the research of development economics, resource and environmental economics and has published several papers in related research fields in CSSCI and SSCI journals.

References

Stavins

RN.

Public policies for environmental protection. Abingdon: Routledge, 2010, pp.41–86.

Freeman

Kolstad

CD.

Moving to markets in environmental regulation: lessons from twenty years of experience. Oxford: Oxford University Press, 2007, pp.3–15.

Pigou

The economics of welfare. Abingdon: Routledge, 2017.

Bongaerts

Kraemer

Permits and effluent charges in the water pollution control policies of France, West Germany, and the Netherlands. Environ Monit Assess 1989; 12: 127–147.

Wang

and Wheeler

Endogenous enforcement and effectiveness of China’s pollution levy system. Washington: World Bank Publications 2000; Vol. 2336, pp.17–19.

Wang

Wheeler

Financial incentives and endogenous enforcement in China’s pollution levy system. J Environ Econ Manag 2005; 49: 174–196.

Yuan

Geng

The transmission mechanism of environmental policies and sustainable development on environment protect industry of China–based on research of government and pollutant corporation and environment protect corporation. China Indus Econ 2010; 10: 65–74.

Guo

Fang

Yang

Does China’s pollution levy standards reform promote emissions reduction?

World Econ 2019; 42: 121–144.

Xie

JG.

How sewage levy affects enterprise productivity: evidence from Chinese manufacturing enterprises. World Econ 2016; 39: 143–168.

10.

Liu

, et al. Can environmental protection tax achieve’reducing pollution and economic growth’? Based on the change perspective of China’s sewage charges. China Pop Resour Environ 2019; 29: 130–137.

11.

Liu

Zheng

SL.

The impact of environmental regulation on manufacturing upgrading. China Soft Sci 2019; 05: 40–52.

12.

Chen

Maung

Shi

, et al. Foreign direct investment concessions and environmental levies in China. Int Rev Financ Anal 2014; 36: 241–250.

13.

Maung

Wilson

Tang

Political connections and industrial pollution: evidence-based on state ownership and environmental levies in China. J Bus Ethics 2016; 138: 649–659.

14.

Fujii

Managi

Kaneko

Decomposition analysis of air pollution abatement in China: empirical study for ten industrial sectors from 1998 to 2009. J Clean Prod 2013; 59: 22–31.

15.

Caliendo

Kopeinig

Some practical guidance for the implementation of propensity score matching. J Econ Surv 2008; 22: 31–72.

16.

Tong

Liu

Xue

Environmental regulation, factor input structure and industrial transformation. Econ Res 2016; 51: 43–57.

17.

Zhang

Kong

The real effect of legal institutions: environmental courts and firm environmental protection expenditure. J. Environ Econ Manag 2019; 98: 102254.

18.

Gürlük

Economic growth, industrial pollution and human development in the Mediterranean Region.

Ecol Econ 2009; 68: 2327–2335.

19.

Huang

Liu

Wang

Environmental policies of local governments: an panel data analysis of

{S O}_{2}

abatement. World Econ 2016; 39: 166–188.

20.

List

CY.

The effects of environmental regulations on foreign direct investment. J Environ Econ Manag 2000; 40: 1–20.

21.

Deng

YP.

Does FDI contribute to China’s environmental pollution?– Spatial measurement research based on China’s inter-provincial panel data. World Manag 2011; 02: 30–43.

22.

Kharel

Effects of urbanization on environmental resources: a land-use planning perspective. Master thesis, University of Texas, Arlington, USA, 2011.

23.

Yang

Wan

Yang

CG.

Industrial structure, Urbanization, and Environmental pollution–An Empirical Study Based on Shandong. Econ Mana Rev 2014; 30: 67–74.

24.

Satterthwaite

Environmental transformations in cities as they get larger, wealthier and better managed. J Geogr J 1997; 8: 216–224.

25.

Feng

Agglomeration and emission reduction: empirical study on the influence of urban size gap on industrial pollution intensity. World Econ 2014; 37: 86–114.

26.

Pollution haven hypothesis and environmental differences of foreign direct investment: the case of industrial emission of sulfur dioxide (

{S O}_{2}

) in Chinese provinces. J Ecol Econ 2006; 60: 228–245.

27.

Zarsky

Havens, halos, and spaghetti: untangling the evidence about foreign direct investment and the environment. Foreign Direct Invest Environ 1999; 13: 47–74.

28.

Mani

Wheeler

In search of pollution havens? Dirty industry in the world economy, 1960 to 1995. J Environ Dev 1998; 7: 215–247.

29.

Fan

Zhao

Does rule of law promote pollution control? Evidence from the establishment of the environmental court. Econ Res J 2019; 54: 21–37.

30.

Jacobson

LaLonde

Sullivan

DG.

Losses of natural disasters. Am Econ Rev 1993; 20: 685–709.

31.

Moser

Voena

Compulsory licensing: evidence from the trading with the enemy act. Am Econ Rev 2012; 102: 396–427.

32.

Shi

Ding

Wei

, et al. Can smart cities construction reduce environmental pollution. China Indus Econ 2018; 6: 117–135.

Can pollution charges reform promote industrial SO 2 emissions reduction?—Evidence from 189 China’s cities

Abstract

Keywords

Introduction

Institutional background and research hypothesis

Institutional background

Research hypothesis

Results

Model

Variables and data description

Results and analysis

Impact of increasing sewage charges on environmental pollution

Balance trend and dynamic effect analysis

PSM estimation

Placebo test

Heterogeneity analysis

Regional heterogeneity

Heterogeneity of city size

Mechanism analysis

Emission governance

Pollution prevention

Conclusions and policy implications

Footnotes

Authors’ contribution

Declaration of conflicting of interests

Funding

ORCID iD

Notes

References