Does the Emissions Trading System Promote Clean Development? A Re-Examination based on Micro-Enterprise Data

Abstract

In 2007, the SO₂ emissions trading pilot policy was established to offer a framework for the management of the industrial environment. The evaluation of the effect of this policy on the industrial enterprise environment is expected to be of great importance for the development of the industrial economy. Our paper aimed to analyze the implementation effects and mechanisms of emissions trading systems using data collected from the China Industrial Enterprise Database and China Industrial Enterprise Pollution Discharge Database from 1998 to 2012. It was found that the policy decreased the emissions intensity of industrial enterprises; moreover, the emission reduction effect was most apparent in the eastern region, in non-state-owned enterprises, in large-scale enterprises, and in low-pollution industries. The findings of the intermediate effect test revealed that the emissions trading system positively affects the environment through the “innovation compensation” effect and “resource allocation” effects. Based on these findings, we make the following recommendations for policy: we should continue to comply with the improvement strategy of joining “market decision” with “government regulation”, actively encourage the construction of an emissions trading system, and guide industrial enterprises to fabricate a plan for working on environmental performance under the motivation of technological innovation.

1. Introduction

As the economy shifts from fast development to high-quality growth, pollution prevention and environmental management are important hurdles to be crossed in China. In recent years, the Chinese government has adopted different strategies for effectively advancing ecological administration [1]. From national legislation to government-led initiatives, including climate planning, energy saving and carbon emissions reduction, the low-carbon society, and promoting the decisive role of markets in resource allocation, the Chinese government is actively exploring environmental governance mechanisms and effective ways to reduce pollution and emissions, so as to effectively bring about improvements in China’s environmental quality [2,3,4]. Regarding the history of environmental management, in the early days, the Chinese government mainly enforced command-oriented environmental laws. In the early 21st century, the central government began its initial efforts to explore the policy of SO₂ emissions trading, and in 2007, 11 regions were approved to carry out pilot emissions trading. In 2014, the General Office of the State Council proposed a timetable for the institution of the emissions trading system, and the establishment of a paid-use and trading system for emissions rights in pilot areas by 2017. The “14th Five-Year Plan” explicitly states that, to improve environmental quality, it is necessary to thoroughly enforce the emissions permit system and continuously push forward the trading mechanisms of carbon emissions rights, so as to promote the effective management of environmental pollution and improve environmental efficiency by market-based means. To date, the emissions trading system has been an essential tool in the fight against pollution in China.

As the main carrier of social production activities, enterprises are both the primary consumers of environmental resources and the main polluters of the environment; their activities significantly impact the ecological environment [5]. The pilot promotion process for the emissions trading system is built on the basis of an ever-improving emissions permit system and a strict total emissions control system, providing enterprises with clear expectations for reducing emissions [6]. It transpires that, for enterprises with more advanced pollution abatement technology than their peers, exercising these capabilities is more profitable; meanwhile, for enterprises with a lower standard of pollution abatement technology than their peers, production costs increase. Therefore, companies must rethink the profitability of their current production and continue to introduce emission reduction technologies and invest in pollution reduction equipment. However, an emissions trading system remains in the development stage in China, and problems such as inadequate market-based trading mechanisms, high transaction costs, and incomplete information persist; these issues may have a direct impact on the effectiveness of the emissions trading mechanism [7]. Questions such as to whether the multiyear emissions trading system has had an effect on pollution reduction, and how ecological conservation and economic development can result in a win–win scenario need to be verified using empirical evidence. It is clear that answering these questions will contribute to an objective understanding of the connection between market-based trading mechanisms and polluting emissions behaviors, thus providing a reference for relevant decision-making processes.

We use the SO₂ emissions trading pilot policy released in 2007 as a “quasi-natural experiment” to identify how the policy affects the pollution emissions behaviors of enterprises. A firm-level pre-assessment found that the emissions trading system may have a “pro-cleaning” impact upon enterprises’ production behaviors. On this basis, the parallel trend test, the placebo test, PSM-DID, and standard error treatment all confirm that the emissions trading system significantly diminishes the discharge strength of regulated industrial firms; this effect is mainly due to the increased innovation of enterprises and the improved efficiency of resource allocation. In addition, heterogeneity analysis shows that enterprises in eastern China, as well as low-pollution enterprises, non-state-owned enterprises, and large-scale firms have preferable responses to the emissions trading system. This paper’s main contributions concern the areas listed below: First, this paper advances the research into the results of enforcing the emissions trading system from the medium level to the micro level, and offers deeper insight into the influence of market-oriented environmental policies on enterprises’ environmental behavior. Second, we explore the micro-transmission mechanisms of emissions trading systems on enterprises’ pollution discharge behaviors through the lens of innovation and resource allocation efficiency, elucidating the theoretical mechanisms of environmental governance. Finally, the DID method is used for empirical analysis in this paper; this method not only effectively evaluates the results of enforcing the emissions trading system, but also effectively alleviates endogenous bias, making the policy evaluation more accurate.

The remaining parts of this study are structured as follows: The associated literature is summarized in Section 2. Section 3 introduces the institutional background and carries out the theoretical analysis. Section 4 presents the research design. The empirical results, including robustness tests, parallel trend tests, and baseline regressions are discussed in Section 5. Section 6 comprises the mechanism’s analysis. Section 7 discusses the heterogeneous influence of emissions trading policies. Section 8 presents the findings and strategy implications.

2. Literature Review

In order to achieve the aim of obligatory pollution reduction by each economic agent, the government usually designs institutional frameworks according to the optimization of the overall welfare of society as a whole when devising relevant regulations [8]. In other words, environmental oversight is a regulatory tool implemented to achieve sustainable development goals by the government. Research on this issue remains a relevant issue within the domain of environment, innovation, and economics [9,10,11]. To date, the majority of studies have shown that environmental oversight is the most fundamental institutional arrangement that motivates firms to fulfill their environmental responsibilities [12,13]. Theories regarding the influence of environmental oversight on pollution fall into two broad categories based on various economic theories. On the one hand, the pollution haven hypothesis proposes that environmental regulation policies internalize external costs and increase the burden on businesses to combat pollution, possibly forcing industries to relocate to areas with fewer and less stringent environmental regulations, resulting in a “pollution haven” [14], an effect known as the compliance cost effect [15].

On the other hand, according to the Porter hypothesis, a reasonable regulatory policy can force firms to modify their production processes, methods, and techniques, which encourages them to innovate and results in innovation compensation effects; this then stimulates resource efficiency and resource allocation optimization, and these gains counteract the adverse consequences of environmental regulation [16,17].

In more recent research, numerous scholars have begun to focus on the policy implications of environmental supervision. The emissions trading system has been extensively addressed by scholars as a market-incentivized environmental policy [18,19,20]. In existing empirical research into the policy effects of emissions trading systems, the primary approach is to use the emissions trading policy as a quasi-natural experiment, comparing control and treatment groups using a DID model to obtain the net effect of the policy intervention [21]. By summarizing the findings of this research, two typical but contrasting views are presented. The effectiveness of ETS is backed up by the research of Ren et al. (2020) and Peng et al. (2021) [21,22]. Ren et al. (2020) indicate that the emissions trading system significantly reduces SO₂ emissions and promotes higher levels of innovation in firms [20]. When assessing the effects of ETS implementation, Peng et al. (2021) determined that ETS has a beneficial influence in terms of productivity gains [22]. However, the findings of some scholars differ from the literature described above. Based on macroenvironmental data, Tu and Shen (2014) evaluated the influence of the emissions trading system on costs associated with pollution reduction [23]. Overall, the emission trading scheme did not significantly reduce pollution abatement costs, perhaps because of the underestimation of policy effects due to the selection of control groups when using the DID approach [24].

In summary, research into the matter of how environmental regulation affects corporate pollution emissions behaviors is scarce, and the findings are not uniform. Most of the studies on the emissions trading system have focused on its innovative effects and macro-level pollution reduction effects, while the policy effects of the emission reductions of enterprises are less extensively studied, and their impact pathway is not analyzed in depth. As a policy that is important for market-oriented environmental supervision, it is worth investigating the specific influence of the emissions trading system in depth. We examine the emission reduction effect of the system at the micro level to fill the gaps between previous studies. In addition, the heterogeneity effect and the mediating effect of innovation and resource allocation on ambient behavior and emissions trading systems are examined; this deepens our understanding of the long-term contributions made by emissions trading systems to pollution control and economic development

3. Institutional Background and Theoretical Mechanism

3.1. Institutional Background

Under the conditions in which total pollutant emission control targets are established, an emissions trading system makes use of market mechanisms to establish statutory rights to discharge pollutants, i.e., emission rights, and allows such rights to be bought and sold like commodities as a means of pollutant emissions control, thereby achieving the goals of lowering emissions and protecting the environment. The significance of the emissions trading system is demonstrated in two main ways. On the one hand, the emissions trading system relies on environmental laws to legalize the right to discharge pollutants, and to control the discharge of pollutants by issuing emissions permits based on the quantity of pollutants emitted by the emission units, the production capacity of the enterprises, and other factors. On the other hand, given the large gap in the cost of environmental pollution treatment, firms with lower treatment costs can use these advantages to reduce environmental pollutants and resell the surplus emissions rights to firms that have insufficient pollution rights, thus forcing polluting firms to reduce their emissions of environmental pollutants in order to reduce treatment costs.

Since 2007, pilot emissions trading has been carried out in 11 regions; emissions trading systems with different characteristics have formed in the different pilots. Compared to mandatory environmental systems, the emissions trading system is an important economic tool for achieving environmental pollution control. The emissions trading system has three advantages. First, the emissions trading system enables enterprises to conduct their production and operations with superior resources without having to make significant investments to achieve the targets set by law or by the government for emissions reduction; this makes it possible to achieve a win–win scenario in terms of both pollution control and enterprise profitability. Second, the cost–benefit mechanism generated by the emissions trading system can promote innovation in production technology by enterprises, which can, in turn, reduce pollution. Finally, the emissions trading system is a market-oriented way of solving environmental problems, reducing government interventions in the market and improving the market’s economic efficiency. The question remains as to whether the emissions trading policy incentivizes Chinese companies to improve their environmental standards.

3.2. Theoretical Mechanisms

In recent years, the Chinese economy has experienced unprecedented levels of growth. At the same time, the ecological and environmental situation has deteriorated, which has attracted the attention of the government; much of this attention is directed towards the environmental regulation of heavily polluting firms. The traditional method of environmental supervision requires compliance with government orders to achieve control over enterprises; this method suffers from an information mismatch between governments and enterprises, and often requires a great deal of human, material, and financial resources, making regulation inefficient. In the context of the current environmental situation, the emissions trading system was created in order to internalize external costs. This kind of system refers to the transfer of emissions between internal sources of pollution in a certain area by means of a monetary exchange, on the premise that the aggregate volume of pollutants emitted will not surpass the permitted level. Through an emissions trading system, low-cost pollution control can be achieved, and environmentally friendly performance can be promoted. Theoretically, the emissions trading system affects firms’ ability to emit pollutants in two different ways: innovation incentives and resource allocation [25]. In this article, we analyze the influence of the emissions trading system on pollution intensity according to the “innovation compensation” effect, as well as the resource allocation effect.

3.2.1. The “Innovation Compensation” Effect

A topic widely addressed in academic research concerns government regulation of enterprises’ discharge behavior because of the strong externalities of environmental issues. The Porter hypothesis suggests that well-designed government policies on environmental oversight can lead to an “innovation compensation” effect, in which firms are encouraged to change their technology by strict but flexible environmental regulations [26]. As a direct consequence, environmental regulations are designed and implemented to force or encourage enterprises to invest more in R&D, effectively promoting their innovation activities and thus improving their pollution control capabilities [27]. Specifically, businesses typically take the following two steps to cut down on pollution emissions in order to maximize profits and reduce the costs associated with contamination control in the face of stringent environmental supervision policies: The first is to improve industrial production technology to cut down on the amount of pollutants released during the production process. In this case, although the initial pollution emissions are not reduced, improving the production process can increase the efficiency of the enterprise, which can then cover the cost of contamination control. In addition, technological innovations that improve production processes can eventually mitigate or offset the increased costs of maintaining the environment, thus generating an “innovation compensation” effect. Second, the final discharge of pollutants can be reduced by means of improved technologies for pollution control and transformation technologies. It is evident that environmental supervision drives up the cost of emissions for firms and encourages them to improve their manufacturing processes and pollution treatment capabilities, which eventually leads to an increase in overall corporate performance. Among environmental regulation policies, the emissions trading system constitutes a pollution management tool that uses market mechanisms to mobilize emissions units. The emissions trading system can, on the one hand, achieve the effect of pollutant reduction by stipulating the amount of emissions produced by polluting enterprises [28]. On the other hand, enterprises that exceed their quotas must reduce their pollution emissions or purchase emission allowances, and when the purchase cost is high, enterprises are more inclined to engage in technological innovation [29].

3.2.2. The Resource Allocation Effect

The optimal distribution of resources is the primary manifestation of the allocation effect. The theory behind emissions trading policies, a market-based regulation based on Coase’s theorem, is essentially based on the commodification of environmental resources and cutting down the aggregate cost of contamination control for the entire society through market transactions, based on clear property rights to emissions, in order to achieve optimal resource allocation [30]. Tradability is a typical feature of emissions trading, and it is through repeated market transactions that the deviation between the market price and the relative price of emissions rights is corrected, allowing a reasonable price for emissions rights to be formed in the market [31]. Thus, emissions trading is a market pricing system, and the process of setting prices in the market is the process of optimizing the allocation of resources. The first part of this process concerns the effective allocation of capital and labor. Typically, firms compare the trading price of emissions rights to the marginal treatment cost of treating the pollutant, and thus decide on the firm’s pollution control measures [32]. If the trading price of emissions rights is in excess of the firm’s marginal cost of abatement, the firm takes environmental protection measures to decrease pollutant emissions, so that it can sell the remaining emissions rights to gain economic benefits or reduce the amount of emissions rights purchased. The firm chooses to purchase the remaining emissions rights from other enterprises or obtain emission rights from the government through public bidding when the trading price of emission rights is inferior to the marginal cost of emission reduction. The flow of funds is essentially determined during this procedure. Second, the emissions trading system has an impact on checking contamination outflows by coordinating the designation of energy inputs. Previously, energy prices were not accurately and reasonably reflected, due to the need for economic development and the issue of externalizing the benefits of energy use [33]. The emissions trading system has facilitated the rationalization of emissions prices through market-based adjustments in terms of property rights, attributes, and abatement costs [34]. Therefore, based on the aforementioned theoretical considerations, we put forward the two following hypotheses:

Hypothesis 1.

The emissions trading system induces Chinese enterprises to become more environmentally friendly in their production practices.

Hypothesis 2.

The emissions trading system promotes environmentally friendly behaviors in enterprises, largely due to the “innovative compensation” effect and the resource allocation effect.

4. Research Design

4.1. Samples and Data

The majority of the firm-level variables used in our paper are derived from the Chinese Industrial Enterprise Pollution Discharge Database of the Ministry of Ecology and Environment and the Chinese Industrial Enterprise Database of the National Bureau of Statistics. The Chinese Industrial Enterprise Database provides the enterprise-level financial data, and the Chinese Industrial Enterprise Pollution Discharge Database provides the pollution emissions data. The Chinese Industrial Enterprise Pollution Discharge Database (1998–2012) is the most significant source of data in this paper; it includes basic information, such as the codes and names of corporate representatives, as well as indicators of the levels of production and emissions of major pollutants including industrial wastewater, chemical oxygen demand, and sulfur dioxide. This database covers industrial enterprises whose pollution emissions are responsible for more than 85 percent of the total emissions in each region.

This paper further merges the two previously mentioned types of data, microdata and prefecture-level municipal data, to investigate the connection between the emissions trading system and the polluting practices of enterprises. According to Brandt et al. (2014), the industrial enterprise database was preprocessed prior to merging [35]. The overlapping information, such as the codes and names of corporate representatives and the enterprise names, was merged using the two aforementioned micro databases, and the enterprises that were not merged were identified by manually searching for key information. Accordingly, a series of processes were applied to the raw data: (1) enterprises with zero or negative values for fixed assets and operating profits were excluded from the sample; (2) enterprises with zero or negative values for key variables, such as product sales revenue and paid-in capital, were not included in the sample; (3) the sample does not include enterprises with fewer than eight employees; (4) the enterprises established before 1949 were deleted. Finally, a panel of firms covering 277 prefecture-level cities from 1998 to 2012 was obtained by combining firm-level data with city-level data, according to the firm’s administrative code. The city-level control variables were mainly obtained from the Chinese city statistical yearbook.

4.2. Variables

4.2.1. Dependent Variables

For a developing country such as China, which is undergoing rapid industrialization, the intensity indicator, excluding the scale factor, has strong policy implications. Therefore, this paper uses the pollution emissions intensity of enterprises, namely pollution emissions per unit of output, to measure enterprises’ pollution emission behaviors [36]. In order to fully capture the pollution emissions of enterprises, we selected six major pollutants that cause water and air pollution in China, namely industrial wastewater, industrial waste gas, sulfur dioxide, chemical oxygen demand, soot, and dust, to construct a comprehensive indicator of the intensity of corporate pollution emissions, which is calculated as:

$${\mathrm{cep}}_{\mathrm{it}}=\frac{1}{\mathrm{n}}{\displaystyle \sum _{\mathrm{m}=1}^{\mathrm{n}}({\mathrm{W}}_{\mathrm{mit}}}\times {\mathrm{emission}}_{\mathrm{mit}})$$

(1)

The economic meaning of the above equation is the summation of each pollutant emission of an enterprise, according to its emission intensity. Here, m represents the type of pollutant, and n denotes the number of pollutant types. Since the emission units of each pollutant vary between enterprises, emission represents the dimensionless intensity of the emission of each pollutant after linear standardization. W is the adjustment factor, expressed as the ratio of the emission intensity of each pollutant to the national average.

4.2.2. Core Explanatory Variable

In this study, the policy variable of the emissions trading systems is captured using an interaction term between the time dummy variable and the between-group dummy variable. The between-group dummy variable indicates that, if the firm is located in one of the 11 pilot regions of the pilot policy, it is considered to belong to the treatment group with a value of 1; meanwhile, if the firm is located in one of the other 19 non-pilot regions, it is considered to belong to the control group, which has a value of 0. The reason that the number of non-pilot provinces is 19 is because there are insufficient data. This paper excludes the four provinces and regions of Hong Kong, Macau, Taiwan, and Tibet.

4.2.3. Control Variables

This paper considers control variables in terms of both internal firm characteristics and external influences. Internal characteristics include firm size, firm ownership, firm age, financial leverage, and gross industrial output. Firm size (size) is defined by the logarithm of firm-level total assets. Ownership (owner), which is divided into state-owned firms and non-state-owned firms based on the type of registration, is given a value of 1 for state-owned firms and 0 otherwise. Firm age (age), represented by the present year minus the year the enterprise opened plus 1, is logarithmically included in the equation. Financial leverage (lev) is a measure of the financing ability of the firm, which is calculated by dividing the total assets according to its total liabilities. The gross industrial output value of the enterprise (out) is expressed as the logarithm of the present value of the corporate gross industrial output.

The economic development level, technological level, industrial structure, and foreign direct investment are all control variables at the city level. Economic development (pgdp) is expressed as the log value of GDP per capita in the company’s registered location. The technological level (tech) is measured as the log value of the quantity of patents granted in the company’s registered location. Industrial structure (ind) is estimated by the extent of the GDP of the secondary industry, and by substituting its logarithmic form into the equation. The ratio of actual foreign investment in GDP is used to represent foreign direct investment (FDI) and its logarithmic form is substituted into the equation.

Table 1. Descriptive statistics of variables.

Variable	N	Mean	SD	Min	Max
lncep	242,437	−0.361	1.823	−8.649	7.353
did	242,437	0.218	0.413	0	1
size	242,437	11.03	1.631	0	19.38
owner	242,437	0.123	0.328	0	1
age	242,437	2.793	0.669	0	4.127
lev	242,437	0.624	0.344	0	48.65
out	242,437	11.15	1.569	0	19.14
pgdp	4155	9.500	0.950	4.595	12.12
tech	4155	0.650	1.690	−4.610	6.740
ind	4155	3.820	0.270	0.640	4.510
fdi	4155	0.110	1.420	−6.280	4.020

displays descriptive statistics for all variables.

4.3. Empirical Model

This paper estimates the effect of the ETS on the environmental conduct of firms using the DID approach, a common tool for policy evaluation, and designs a basic econometric model according to the research hypotheses of this paper, as follows:

$$\ln {\mathrm{cep}}_{\mathrm{it}}={\alpha }_0+{\alpha }_1{\mathrm{did}}_{\mathrm{it}}+{\alpha }_2{\mathrm{Control}}_{\mathrm{it}}+{\delta }_{\mathrm{i}}+{\gamma }_{\mathrm{t}}+{\displaystyle \sum \mathrm{ind}}+{\displaystyle \sum \mathrm{city}}+{\epsilon }_{\mathrm{it}}$$

(2)

where $\ln {\mathrm{cep}}_{\mathrm{i}\mathrm{t}}$ is the enterprise i’s pollution emission intensity in year t; the dummy variable ${\mathrm{did}}_{\mathrm{it}}$ denotes the location of the pilot city for emissions trading; α₁ indicates the policy effect of emissions trading pilot program considered in this paper; and ${\mathrm{Control}}_{\mathrm{it}}$ is a set of control variables. ${\delta }_{\mathrm{i}}$ is the firm-fixed effect; the year-fixed effect is represented by ${\gamma }_{\mathrm{t}}$; ${\displaystyle \sum \mathrm{ind}}$ is the industry-fixed effect; and ${\displaystyle \sum \mathrm{city}}$ is the city-fixed effect. ${\epsilon }_{\mathrm{it}}$ refers to random error terms.

5. Empirical Analysis

5.1. Baseline Regressive Results

The benchmark estimation results of the impact of the ETS on the pollution intensity of manufacturing firms are presented in

Table 2. Baseline regression results.

	(1)	(2)	(3)	(4)
	Lncep	Lncep	Lncep	Lncep
did	−0.079 ***	−0.031 **	−0.039 ***	−0.026 **
	(0.009)	(0.012)	(0.012)	(0.012)
size			0.117 ***	0.119 ***
			(0.006)	(0.006)
owner			0.113 ***	0.116 ***
			(0.017)	(0.017)
age			0.06 ***	0.057 ***
			(0.01)	(0.01)
lev			−0.019 *	−0.02 *
			(0.011)	(0.011)
out			0.154 ***	0.158 ***
			(0.005)	(0.005)
pgdp				−0.118 ***
				(0.021)
tech				−0.035 ***
				(0.008)
ind				−0.063 *
				(0.037)
fdi				0.009 **
				(0.004)
_cons	−0.344 ***	−0.31 ***	−3.503 ***	−2.083 ***
	(0.004)	(0.003)	(0.073)	(0.205)
Year fixed	No	Yes	Yes	Yes
Firm fixed	No	Yes	Yes	Yes
Industry fixed	No	Yes	Yes	Yes
City fixed	No	Yes	Yes	Yes
N	242,437	242,437	242,437	242,437
R2	0.03	0.765	0.768	0.768

Note: The values in parentheses are reported as standard errors; *, **, *** represent significant at the 10%, 5%, and 1% levels, respectively.

. The average impact of the emissions trading system on firms’ pollution intensity, without controlling for any influencing factors, is shown in column 1 of Table 2. The estimates suggest that did is negatively significant, indicating that the pilot policy on emissions trading in China significantly reduces the pollution intensity of firms. Based on column (1), columns (2)–(4) gradually add control variables and control for the year-fixed effect, the firm-fixed effect, the industry-fixed effect, and the city-fixed effect; the results are unchanged, that is, the estimated coefficient of did is still statistically significant with negative values. The results of the four models are generally consistent, suggesting that the estimates are somewhat robust. The above results provide preliminary evidence for the validity of Hypothesis 1 in this paper, i.e., that the emissions trading policy induces Chinese enterprises to become more environmentally friendly in their production behaviors.

5.2. Parallel Trend Test

A parallel trend test is required in order to precisely determine how the emissions trading system affects pollution emission intensity. Specifically, the trend in emissions intensity for the treatment group and the control group is examined through a factual study, using 2007 as the base year for the emissions trading policy. In Figure 1, the time preceding the enforcement of the emissions trading system is depicted on the left side of the dashed line, and the period following the implementation of the emissions trading system is depicted on the right side of the dashed line. Figure 1 shows that the emissions trading system trends of the treatment group and the control group are more-or-less the same during the period before the enforcement of the policy, and the estimated coefficient values are around 0, which fulfills the general assumption of parallel trends. In addition, in the period following the implementation of the emissions trading policy, a disparity in emissions intensity between the treatment group and control group gradually emerges. The estimated coefficients remain significant over time, demonstrating that the influence of the system on the pollution intensity is somewhat sustainable.

5.3. Robustness Tests

5.3.1. Placebo Test

A placebo test is required to make sure that the findings of our paper are driven by the emissions trading system, and to eliminate the impact of other unidentified factors on the choice of pilot cities. By randomly selecting dummy treatment groups from across the sample to carry out regressions that are in line with the baseline regression, the placebo test lends robustness to the initial findings. Specifically, the randomization of the policy shocks is conducted first, and then a regression is run on each random set of samples separately, after the randomization process is repeated 1000 times to obtain 1000 random sets of samples. Figure 2 depicts the t-values and distributions of the estimated regression coefficients following the 1000 random assignments. Where, the blue line represents the distribution of t-values, the vertical solid line is the real t-value, the dashed line is the mean value, and the two short dashed lines indicate t = −1.65 and t = 1.65 respectively. The fact that all of the distributions are centered around 0, and that the majority of the estimated coefficients for the samples have t-values that are below 2, indicates that the emissions trading system has no notable influence on any of these samples. Therefore, the findings drawn in our paper can be affirmed by the placebo test, and the influence of the emissions trading system on the pollution emission intensity of firms in the pilot cities is shown not to be causally related to other unobserved factors, i.e., the baseline estimation results are robust.

5.3.2. PSM-DID

During the research design process when conducting quasinatural experiments, if the policy shocks are strictly exogenous, causal identification can be achieved directly by setting up treatment and control groups, according to policy shocks, using the DID approach. However, since the selection of provinces designated as emission trading regions is not generated by random exogenous shocks in the design process of the emissions trading system but is influenced by factors such as the development of ETS markets in these pilot regions, the possible exogeneity of the policy needs to be removed using the PSM-DID method.

Our research further applies the PSM method to conduct DID model regression analysis to test the influence of the emissions trading system on the pollution emission intensity when the data characteristics and trends of two groups are essentially the same. The steps are as follows: First, all control variables are matched as covariates using the nearest neighbor coordinating strategy with a matching proportion of 1:2, and the matched samples are retained. Second, the matched samples are regressed using the DID model. Figure 3 and Figure 4 illustrate the kernel density distributions before and after the propensity score matching, respectively. The results are then further regressed using the DID method, and the findings are shown in column (1) of

Table 3. Robustness tests.

	(1)	(2)	(3)
	PSM-DID	Exclusion of Corporate Samples	Standard Error Handling
did	−0.028 *	−0.024 *	−0.026 *
	(0.016)	(0.012)	(0.013)
_cons	−2.038 ***	−2.084 ***	−2.083 ***
	(0.207)	(0.206)	(0.25)
Firm-level control variables	Yes	Yes	Yes
City-level control variables	Yes	Yes	Yes
Year fixed	Yes	Yes	Yes
Firm fixed	Yes	Yes	Yes
Industry fixed	Yes	Yes	Yes
City fixed	Yes	Yes	Yes
N	242,400	238,329	242,437
R2	0.698	0.77	0.768

Note: The values in parentheses are reported as standard errors; *, *** represent significant at the 10%, and 1% levels, respectively.

. It can be observed that, after propensity score matching, the difference between the kernel density distribution of the experimental and control groups is obviously reduced, indicating the effectiveness of the propensity score matching method used in this paper.

5.3.3. Other Robustness Tests

The results of alternative forms of robustness regression are depicted in columns (2)–(3) of Table 3. First, firms that entered the exit market three years after 2007 are excluded. As shown in column (2) of Table 3, the core conclusions of this paper have not fundamentally changed. Second, standard errors are treated. Because the estimation results presented in this paper are likely to be affected by the problem of within-group correlation, the standard errors are tested by clustering them at the firm level. From the estimates in column (3), the conclusions of this paper still hold. In the ways elucidated above, the robustness tests indicate that the results of our paper are highly robust, i.e., the emissions trading system can inhibit enterprise pollution.

6. Mechanism Analysis

Benchmark analysis and robust research show that the emissions trading system diminishes the emissions intensity of enterprises, and finally achieves the target for reducing emissions. The question that follows concerns the mechanisms through which the emissions trading system influences the environmental performance of enterprises. As discussed in the previous theoretical analysis, the emissions trading system may improve the environmental performance of industrial firms through two channels: technological innovation and resource allocation. First, the emissions trading system reduces pollution emissions through innovation. We use the share of new product output to total output as a proxy variable for technological innovation to determine whether or not this transfer mechanism is true. Second, resource allocation efficiency is improved through the emissions trading system, which reduces pollution emissions intensity. This paper draws on Hsieh and Klenow (2009) to test whether this transmission mechanism holds, using total factor productivity rather than resource allocation efficiency [37].

The next step is to construct a mediating effects model by introducing the above two mediating variables, and to verify the applicability of the relevant impact mechanism by combining realistic data from Chinese manufacturing enterprises. The estimation process is divided into three steps: (1) regressing the dependent variable (firms’ pollution emission intensity) to the basic independent variable (pilot policy on emissions trading); (2) regressing the mediating variables (firms’ technological innovation and resource allocation efficiency) to the basic independent variable; (3) regressing the dependent variable to the independent and mediating variables.

$${\mathrm{innovation}}_{\mathrm{it}}={\varphi }_0+{\varphi }_1{\mathrm{did}}_{\mathrm{it}}+{\varphi }_2{\mathrm{Control}}_{\mathrm{it}}+{\delta }_{\mathrm{i}}+{\gamma }_{\mathrm{t}}+{\displaystyle \sum \mathrm{ind}}+{\displaystyle \sum \mathrm{city}}+{\epsilon }_{\mathrm{it}}$$

(3)

$${\mathrm{tfp}}_{\mathrm{it}}={\mu }_0+{\mu }_1{\mathrm{did}}_{\mathrm{it}}+{\mu }_2{\mathrm{Control}}_{\mathrm{it}}+{\delta }_{\mathrm{i}}+{\gamma }_{\mathrm{t}}+{\displaystyle \sum \mathrm{i}\mathrm{n}\mathrm{d}}+{\displaystyle \sum \mathrm{city}}+{\epsilon }_{\mathrm{i}\mathrm{t}}$$

(4)

$$\ln {\mathrm{cep}}_{\mathrm{i}\mathrm{t}}={\lambda }_0+{\lambda }_1\mathrm{d}\mathrm{i}{\mathrm{d}}_{\mathrm{i}\mathrm{t}}+{\lambda }_2{\mathrm{innovation}}_{\mathrm{i}\mathrm{t}}+{\lambda }_3{\mathrm{Control}}_{\mathrm{i}\mathrm{t}}+{\delta }_{\mathrm{i}}+{\gamma }_{\mathrm{t}}+{\displaystyle \sum \mathrm{i}\mathrm{n}\mathrm{d}}+{\displaystyle \sum \mathrm{city}}+{\epsilon }_{\mathrm{i}\mathrm{t}}$$

(5)

$$\ln {\mathrm{cep}}_{\mathrm{i}\mathrm{t}}={\rho }_0+{\rho }_1\mathrm{d}\mathrm{i}{\mathrm{d}}_{\mathrm{i}\mathrm{t}}+{\rho }_2\mathrm{t}\mathrm{f}{\mathrm{p}}_{\mathrm{i}\mathrm{t}}+{\rho }_3{\mathrm{Control}}_{\mathrm{i}\mathrm{t}}+{\delta }_{\mathrm{i}}+{\gamma }_{\mathrm{t}}+{\displaystyle \sum \mathrm{i}\mathrm{n}\mathrm{d}}+{\displaystyle \sum \mathrm{city}}+{\epsilon }_{\mathrm{i}\mathrm{t}}$$

(6)

where ${\mathrm{innovation}}_{\mathrm{i}\mathrm{t}}$ is the measure of technological innovation. Through R&D innovation and technological progress, emissions reduction targets can be met to a certain degree, and the sustainable operations and growth of enterprises can be ensured. ${\mathrm{tfp}}_{\mathrm{i}\mathrm{t}}$ and $\mathrm{d}\mathrm{i}{\mathrm{d}}_{\mathrm{i}\mathrm{t}}$ represent the efficiency of enterprise resource allocation and the policy variables of the emissions trading pilot, respectively. Model (1) of

Table 4. Mediation effect model.

Variables	(1)	(2)	(3)	(4)	(5)
	Lncep	Innovation	Tfp	Lncep	Lncep
innovation				−0.026 ***
				(0.025)
tfp					−0.007 ***
					(0.008)
did	−0.026 **	0.036 ***	0.021 ***	−0.027 **	−0.03 **
	(0.012)	(0.007)	(0.004)	(0.013)	(0.012)
_cons	−2.083 ***	8.083 ***	−3.059 ***	−2.359 ***	−2.08 ***
	(0.205)	(0.188)	(0.059)	(0.255)	(0.206)
Firm-level control variables	Yes	Yes	Yes	Yes	Yes
City-level control variables	Yes	Yes	Yes	Yes	Yes
Year-fixed	Yes	Yes	Yes	Yes	Yes
Firm-fixed	Yes	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes	Yes
City-fixed	Yes	Yes	Yes	Yes	Yes
N	242,437	242,437	242,437	242,437	242,437
R2	0.768	0.974	0.915	0.768	0.766

Note: The values in parentheses are reported as standard errors; **, *** represent significant at the 5%, and 1% levels, respectively.

corresponds to the baseline regression model. Models (2) and (3) correspond to equations (3) and (4), i.e., regression analysis of the mediating variables on the underlying independent variables. Models (4) and (5) correspond to Equations (5) and (6). Table 4 provides estimation results of the mediating effects.

Column (1) shows the baseline regression results of the previous section. Column (2) uses corporate technological innovation as the explanatory variable, and the findings clearly show that the coefficient of the emissions trading system on innovation is 0.036 and passes a significance test, which implies that the emissions trading system significantly improves technological innovation. Column (4) uses emissions intensity as the explained variable, and introduces both the emissions trading system and innovation variables. The fact that the emissions trading system’s regression coefficient remains significantly negative lends credence to the hypothesis that the “innovation compensation” effect plays a significant role in the ability of the emissions trading system to influence pollution emission intensity. The emissions trading system encourages firms to increase technological innovation inputs to satisfy environmental standards.

Column (3) of Table 4 uses resource allocation efficiency as the explained variable, and the estimate of the emissions trading system is obviously positive, which indicates that the emissions trading system can promote corporate resource allocation efficiency. Column (5) takes pollution emission intensity as the explained variable and adds the emissions trading system and the mediating variable of resource allocation efficiency. The regression estimates demonstrate that the emissions trading system exerts a considerable impact on pollution intensity through the “resource allocation” effect. Firms are forced to reduce capacity investment and maximize resource allocation as a result of the emissions trading system increasing the cost of pollution control, thereby reducing the intensity of emissions, to some extent.

The findings of the mediating effect model reveal that the emissions trading system may improve environmental performance in two ways: stimulating enterprises’ innovation and refining the efficiency of enterprises’ resource allocation, and ultimately achieving the goal of pollution control. That is, the emissions trading system demonstrates both an “innovation compensation” effect and a “resource allocation” effect on pollution emission intensity, which verifies the conclusions of the analysis of the theoretical mechanism.

7. Heterogeneous Analysis

Although we have demonstrated that the emissions trading system works, there may be some variation in the responses of different regions, industries, and enterprises to policy shocks in the pilot area. An exploration of heterogeneity can deepen our understanding of the mechanisms of action and the boundary conditions of the emissions trading system. Therefore, we investigated the heterogeneity of the emissions trading system in terms of the external characteristics and the internal characteristics that affect the pollution emission intensity.

7.1. Heterogeneous Analysis at the Regional Level

The emissions trading system may have heterogeneous effects on corporate environmental performance due to its vast scope and regional differences in development. In order to offer theoretical support for the formulation of more specific environmental policies, we introduce regional characteristics in order to probe the effect of the emissions trading system on environmental performance in various regions. The sample is divided into eastern, central, and western regions, based on China’s regional development characteristics.

Table 5. Regional heterogeneity.

Variables	(1) East	(2) Center	(3) West
did	−0.062 ***	−0.019	0.154 ***
	(0.017)	(0.028)	(0.039)
_cons	−2.951 ***	−4.804 ***	−3.637 ***
	(0.355)	(0.614)	(0.512)
Firm-level control variables	Yes	Yes	Yes
City-level control variables	Yes	Yes	Yes
Year-fixed	Yes	Yes	Yes
Firm-fixed	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes
City-fixed	Yes	Yes	Yes
N	149,637	54,591	38,205
R2	0.77	0.767	0.753

Note: The values in parentheses are reported as standard errors; *** represent significant at the 1% level.

displays the regression results for the subsample. In the eastern region, the estimated coefficient of the emissions trading system effectively reduces pollution intensity. In the central region, the estimated coefficient of the emissions trading system is insignificant. However, the pollution intensity of western regions is significantly increased by the emissions trading system. This may be because the western region faces the task of economic catch-up in the process of economic development. Local governments are more interested in promoting economic growth in a short period, and environmental protection is therefore neglected. Secondary industry is still the driving force of economic development in the western region. Therefore, the emissions trading system fails to bring about an effective curb on corporate emissions in the western region

7.2. Heterogeneous Analysis at the Industry Level

Generally speaking, the government tends to impose stricter environmental regulations on heavily polluting industries as a key concern of the public and the government, which can force the green transformation of heavily polluting industries and thus achieve greater pollution reduction. Therefore, we further examined the industry heterogeneity of pollution levels to verify the differential influence of the emissions trading system on pollution emission intensity. As indicated by the First National Census Program on Pollution Sources, 11 industries are delegated the intensely dirtying enterprises of key concern; examples include industry, the horticulture and food-handling industry, and the synthetic unrefined components and compound item fabrication industry. The study sample is split into two groups based on the industries to which they belong: heavily polluting firms and low-level polluting companies. The estimates are shown in columns (1)–(2) of

Table 6. Industry heterogeneity and firm heterogeneity.

	Industry Heterogeneity		Enterprise Heterogeneity
Variables	Pollution Emission Intensity		Ownership Property		Firm Size
	(1) High	(2) Low	(3) State-Owned	(4) Non-State	(5) Big	(6) Small
did	−0.011	−0.044 ***	−0.002	−0.027 **	−0.037 **	−0.031 *
	(0.022)	(0.016)	(0.052)	(0.013)	(0.017)	(0.019)
_cons	−2.306 ***	−1.556 ***	−4.043 ***	−1.914 ***	−1.247 ***	−2.466 ***
	(0.374)	(0.264)	(0.671)	(0.224)	(0.291)	(0.317)
Firm-level control variables	Yes	Yes	Yes	Yes	Yes	Yes
City-level control variables	Yes	Yes	Yes	Yes	Yes	Yes
Year-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Firm-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Industry-fixed	Yes	Yes	Yes	Yes	Yes	Yes
City-fixed	Yes	Yes	Yes	Yes	Yes	Yes
N	78,155	150,543	31,430	206,695	120,547	115,008
R2	0.775	0.771	0.85	0.758	0.762	0.782

Note: The values in parentheses are reported as standard errors; *, **, *** represent significant at the 10%, 5%, and 1% levels, respectively.

, and suggest that the emissions trading system does not have a significant effect on the pollution intensity of the heavily polluting enterprises, while having a significant inhibitory influence on the pollution intensity of the firms with lower levels of pollution.

7.3. Heterogeneous Analysis at the Enterprise Level

(1) Grouping test based on enterprise ownership: Enterprises with different ownership characteristics differ greatly in terms of resource endowment, business objectives, and institutional logic, and may have differential emissions reduction responses to macro policy shocks. Therefore, this paper conducts group tests based on the nature of enterprise ownership, and the test results are presented in columns (3)–(4) of Table 6. The findings suggest that the mitigation effect of the policy is mainly felt by non-SOEs, as the estimated coefficient of did is significantly negative in the non-SOE group, but it is unable to pass the test of significance in the SOE group. The possible reasons for this are twofold: on the one hand, although SOEs may respond more strongly to the policy with higher credit levels or more favorable interest rate support, they are not sensitive to information on efficiency improvements and competitive opportunities brought about by market-based environmental supervision. As a result, the incentive and constraint mechanisms of the policy do not apply to SOEs, making it difficult to motivate them to adopt measures to combat pollution and reduce emissions. On the other hand, non-SOEs are quicker to capture the competitive opportunities brought about by external environmental changes and use the emissions trading policy to form new competitive advantages, which can stimulate the formation of strong incentives to innovate and thus drive enterprises to reduce emissions.

(2) Grouping test based on enterprise size: We classify businesses according to their median total assets, with large-scale enterprises being those with assets above the median. Columns (5) and (6) of Table 6 display the outcomes. The estimated coefficient of did in the sample of large-scale enterprises is negative, indicating that the emissions trading system effectively encourages large-scale businesses to achieve pollution reduction. However, in the sample of SMEs, although the emissions trading system reduces the intensity of pollution emissions, this reduction is less effective than it is in large enterprises. The reason for this may be that large-scale enterprises have pronounced financial strength, capital market financing, and risk resistance, and the emissions trading system can motivate large-scale enterprises to actively engage in green innovation. Meanwhile, small-scale enterprises are often faced with higher innovation costs due to resource constraints, and thus are not as motivated to engage in green innovation activities. As a result, large-scale enterprises are more willing to radically reduce their pollution intensity through increased R&D investment and green innovation than small-scale enterprises. At the same time, small-scale enterprises are more inclined to respond to environmental regulations through strategic emission reduction measures, such as purchasing pollution control equipment or even reducing production.

8. Conclusions and Implications

8.1. Conclusions

It is essential to clarify the connection between the emissions trading systems and environmental performance in order to ensure the development of a more ecological civilization and strike a balance between economic growth and environmental protection. This paper uses the SO₂ emissions trading policy of 2007 as an exogenous event in a quasi-natural experiment to investigate how the emissions trading system can impact corporate environmental behaviors, and how the central environmental regulation policy is enforced in a pilot area. On this basis, a variety of approaches were used to evaluate the ability of emissions trading systems to influence corporate environmental conduct. In addition, the “innovation compensation” and “resource allocation” effects were tested using mediating effects models. The conclusions are as follows:

(1)

The baseline results of the DID model indicate that the estimated coefficients of the policy variables are negative, regardless of whether the time effect, firm effect, industry effect, and city effect are controlled. This suggests that the emissions trading system reduces pollution emissions to a remarkable degree. After several robustness tests, including a parallel trend test, a placebo test, PSM-DID, the exclusion of firm samples, and standard error treatment, it was confirmed that the emissions trading system has a pro-cleaning impact on the production behaviors of firms.

(2)

Using technological innovation and resource allocation efficiency as mediating variables, an empirical test of the theoretical mechanism by which the emissions trading system influences environmental performance was conducted. The findings show that innovation plays an interceding role in the mechanism by which the outflow exchanging framework influences corporate ecological conduct, confirming the “innovation compensation” effect. The “resource allocation” effect is supported by the fact that efficiency in resource allocation acts as a mediator between the emissions trading policy and pollution intensity.

(3)

According to the findings of the heterogeneity analysis, the emissions trading system has a diverse effect on corporate environmental performance in various regions, with the effect being more pronounced in the eastern region. The emissions trading system has a greater impact on energy savings and emissions reductions in industries that produce lower levels of pollution, as compared to industries produce more pollution. In contrast with state-owned firms, the policy meaningfully affects discharge decreases in non-state-owned firms in the pilot regions. Additionally, large enterprises reduce pollution at a higher rate than smaller enterprises.

8.2. Policy Implications

(1)

A development strategy of combining “market decisions” and “government regulation” should always be followed when controlling environmental pollution and improving environmental performance. First, we should maintain the dominance of the market over resource allocation for pollution control. Through the market, the benefit mechanisms for firms with and without pollution permits are constantly adjusted, and the cost of treatment is internalized into the cost–benefit analysis of firms, enhancing their environmental performance. Second, to compensate for market failures such as monopolies and externality caused by the limitations of market, we should fully utilize government regulation and the auxiliary role, and continuously advance the market environment.

(2)

Promoting the innovation and resource allocation efficacy of firms is crucial to allowing the emissions trading system to encourage improved environmental performance. The government, industry, and society should recognize that the inherent mechanisms of action whereby the emissions trading system promotes environmental performance are technological innovation and resource optimization. To encourage firms to realize green development, R&D investment, technological innovation, and production technology, updates should continue for all firms. The government should support the improvement of corporate environmental behaviors and encourage businesses to change their conventional inefficient production methods.

(3)

Different pilot regions experience notable differences in the enforcement of emissions trading systems due to their own economic development status, technological level, and other factors. As a result, the effect of the emissions trading system on environmental performance is heterogeneous. Therefore, the pilot construction of emissions rights ought to be carried out according to the circumstances at that particular location. In addition, the intermediary effects of technological progress and resource allocation efficiency should be fully realized, to encourage companies to undertake technological research and development.