Has the Carbon Trading Pilot Market Improved Enterprises’ Export Green-Sophistication in China?

Abstract

This study empirically examined the effect of a carbon trading pilot market on export green-sophistication of Chinese listed enterprises by adopting a difference-in-difference method. Findings show that a carbon trading pilot market can improve enterprises’ export green-sophistication after using robustness tests to overcome endogeneity. The impact mechanism test shows that a carbon trading pilot market can improve export green-sophistication by increasing green technology innovation. Further research on the system design of carbon trading pilot markets shows that the greater the total carbon quota allocation, the larger the reduction in the trading volume of Chinese certified emissions. Furthermore, the weaker the punishment for an enterprise’s default in the pilot areas, the less favorable it is for enterprises to improve their export green-sophistication. Compared with the grandfather and historical intensity methods, benchmarking used in the allocation of carbon quotas is conducive to the improvement of the export green-sophistication of enterprises.

1. Introduction

After China implemented its gradual opening-up strategy, relying on the advantage of its low-cost labor force, the annual average growth rate of its export volume reached more than 10% [1]. In 2009, China’s total export volume reached approximately USD 1.2 trillion, and China became the world’s first export power. In 2020, China’s total export volume reached nearly USD 2.7 trillion despite the influence of trade protectionism, trade friction, and the Coronavirus disease (2019) (COVID-19) [2]. With the continuous expansion of trade scale and the rapid economic development, China has become a major trading country. However, China’s export quality is not high [3]. In terms of quantity, export growth accounted for approximately 70%, whereas quality improvement is relatively small, accounting for only 19% [4,5,6]. The huge export growth has exacerbated environmental pollution in China [7]. Approximately 1 billion tons of China’s annual carbon emissions are caused by net exports, severely destroying the ecological environment [8,9]. Export sophistication is an important index to measure a country’s export quality and to determine a country’s future international strategic competitive advantage, and its improvement can help the economy achieve sustainable development [10]. Broadening the concept of export sophistication, Li and Lu [11] introduced its green aspect and proposed export green-sophistication (EGS), which refers to the green technology elements in export products. The higher the EGS, the more green-technology elements contained in the exports and the higher the added value of products. EGS can not only reflect whether a country’s export trade has comparative advantages, but also reflect its type, and whether an unsustainable resource input or sustainable green technology is added. Against the background of increasing international trade friction, improving EGS is an important path for China to change its extensive trade growth mode, become a foreign trade power, and achieve high-quality economic development.

As a result of the economic development of all countries, carbon dioxide emissions are gradually increasing worldwide. At present, several types of environmental regulations (ER) are widely used globally to deal with climate change, namely, control command, market incentive, and voluntary planned ER [12]. Among these, the carbon emission trading market (hereinafter referred to as the carbon market) is considered to be the most economical and effective policy means to deal with climate change [13]. By the end of 2020, a total of 24 carbon markets on four continents were in operation, accounting for over 16% of global total carbon dioxide emissions, and another 24 carbon markets were in the process of implementation [14]. In China, the government implemented the carbon trading pilot market (CTPM) in seven provinces and cities in 2011. This policy aims to establish a national carbon emission trading market (NCET) on the basis of local pilot experience. Shenzhen was the first to launch a carbon market in June 2013, and the remaining six carbon markets were launched sequentially within a year. China’s NCET officially launched online trading in 2021.

Whether environmental regulation can improve export sophistication has been widely discussed in literature, and the conclusions differ. CTPM plays a positive role in protecting the environment, but inevitably affects technological innovation and productivity, and hence the development of enterprises [15,16]. Thus, does the CTPM impact the EGS of enterprises? Given their differences, pilot areas adopt varying systems. Is there any difference in these systems’ impact on the EGS of enterprises? Research on these issues can provide a reference for China to optimize the policy design of NCET, and also provide valuable experience for the policy design of other developing countries and regions, to improve export sophistication while protecting the environment and promoting the sustainable development of global trade.

The remainder of this paper is arranged as follows. Section 2 presents literature reviews on CTPM and EGS. Section 3 describes the research design. Section 4 discusses the regression results. Section 5 provides the conclusion and policy recommendations.

2. Literature Review

2.1. Research on China’s CTPM

China’s CTPM has been operating for many years, and, since its launch, its effect on pilot areas has attracted extensive attention from the academic community. In terms of environmental effects, based on provincial-, city-, industry-, and enterprise-level data, existing studies have analyzed the environmental effects caused by CTPM from the perspective of carbon dioxide emission, energy intensity, and carbon intensity, using the Computable General Equilibrium (CGE) model, difference-in-difference (DID) method, or Propensity Score Matching (PSM-DID) method. Most of the conclusions suggest that the CTPM can result in significant positive environmental effects, that is, reduce carbon dioxide emissions, energy intensity, and carbon intensity. Empirical studies found that the CTPM can reduce carbon dioxide emissions in the pilot areas through China’s provincial panel data [17,18,19,20,21]. Chen et al. [22] found that the CTPM can significantly improve energy efficiency in a pilot area by promoting enterprise innovation. Tang et al. [23] found that the CTPM can significantly lower the pilot areas’ carbon intensity, and that this impact persists. Among Chinese enterprises, Cui et al. [24] found that the CTPM can effectively reduce carbon emissions by reducing the use of coal.

Scientific evaluations of the economic effect of the CTPM show a significant positive result. Zhu et al. [25] found that the CTPM can improve China’s industrial enterprises’ low-carbon innovation level. Wang et al. [26] found that the CTPM can aid China’s economy to achieve low-carbon transformation. Peng et al. [27] found that the CTPM increased the employment and return on capital of the industry. Qi et al. [28] used the difference-in-difference-in-difference (DDD) method to empirically study the effect of the CTPM on industrial low-carbon international competitiveness, which showed improvements.

2.2. Research on Environmental Regulation and EGS

The existing literature shows that the effect of the CTPM has not been explored from the perspective of export sophistication. In relation to this, we discuss the impact of ER on trade exports based on the “Pollution Paradise Hypothesis” and the “Porter Hypothesis”, and mainly obtain two conclusions. One is the disadvantage theory; that is, strict environmental regulation can reduce enterprises’ export competitiveness. Cole et al. [29] found that environmental regulation inhibited the industrial export competitiveness of 41 industries in Japan from 1989 to 2003. Geeenstone et al. [30] found that ER had a certain negative impact on the competitiveness of US manufacturing enterprises from 1972 to 1993. Du and Li [31] found that strict ER inhibited export trade using the Heckman model. The other conclusion is the advantage theory; that is, flexible environmental regulation can reduce enterprises’ production costs through the adoption of new technology, which can be conducive to exports [32,33]. Ge et al. [6] independently constructed the ER index, empirically studied the relationship between ER, financing constraints, and EGS, and found that ER has a positive impact on EGS.

Theoretical and empirical analyses of the CTPM have also been carried out from the perspectives of environmental and economic effects, and deeply examined the definition and influencing factors of export sophistication, obtaining valuable and meaningful conclusions. However, the existing literature lacks direct research on the content of the CTPM and EGS, and still has deficiencies. First, analysis of different carbon market system designs has been neglected. Research on the impact of the CTPM is limited to the overall level, and does not include in-depth analysis of different system designs, which thus may not be conducive to providing more guidance for NECT optimization. Second, research has largely ignored the green aspect of China’s export sophistication, paid insufficient attention to improving its EGS, and lacked corresponding mechanism analysis, which is inconsistent with the nation’s strategy of becoming a trading power.

The marginal contributions of this study are as follows. First, the existing literature mainly focuses on measurement methods and important influencing factors of export sophistication, but ignores its green aspect and pays insufficient attention to the path required to enhance China’s EGS. This study examines the effect of the CTPM on China’s EGS for the first time, and discusses the intermediary channel, which verifies Porter’s theory to a certain extent, and makes up for the shortcomings of existing theoretical and empirical fields. Second, this study examines the implementation effect of the CTPM from the perspective of micro-enterprise export technology, and expands the related research. In addition to providing new empirical evidence, this study also further examines its impact mechanism, which helps to understand the transmission between macro-environmental policy and micro-enterprise export behavior. Third, this study carries out an in-depth analysis of the varying effects of different system design characteristics of each pilot carbon market on the EGS of enterprises, enriches the relevant research on the CTPM, and provides targeted reference for the design of China’s NECT. Fourth, this study analyzes the moderating effect of government behavior in the region where enterprises are located. Furthermore, this study analyzes whether the implementation of the CTPM has significantly improved the competitiveness of enterprises with the help of green innovation channels, and provides theoretical guidance about how to deal with environmental regulation; that is, enterprises must make full use of their own resources and actively carry out high-quality green innovation activities to improve their competitiveness and environmental protection.

3. Research Design

3.1. Model

3.1.1. DID Model

The existing literature mainly uses the DID model when evaluating the economic and environmental effect of the CTPM. This model can avoid the endogenous problem, and the results are highly stable. Therefore, this study constructs an impact model of CTPM on EGS, as follows:

$${\mathrm{lnegs}}_{i\mathrm{t}}={\alpha }_0+{\alpha }_{1}pilo{t}_i\times pos{t}_t+{\alpha }_2{X}_{it}+{\mu }_i+{\gamma }_{\mathrm{t}}+\mathrm{province}\ast \mathrm{year}+\mathrm{sector}\ast \mathrm{year}+{\epsilon }_{it}$$

(1)

where lnegs_it is the natural logarithm of EGS; pilot_i represents the dummy variable of CTPM, which is set as 1 for pilot enterprises and 0 for non-pilot enterprises; post_t represents a dummy variable of CTPM’s policy implementation time in 2013, and post_t = 1 represents after policy implementation (t ≥ 2013), whereas post_t = 0 represents before policy implementation (t < 2013); province*year is the province time trend item, which can eliminate the potential confounding factors at the province level that change with time; and sector*year is the sector time trend, which can eliminate confounding sector-level factors that change with time.

3.1.2. Extensions of DID Model

(1)

Mediating Effect Model

The existing literature has studied whether the CTPM can produce the Porter effect, and reported results that enterprises are encouraged to actively improve the level of technological innovation [16]. Thus, this study uses the standard mediating effect model to test this impact mechanism, and to assess whether the CTPM can promote the increase in EGS of enterprises by improving green technology innovation:

$$\begin{array}{l}{\mathrm{gi}}_{\mathrm{i}\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\beta }}_1\mathrm{p}\mathrm{i}\mathrm{l}\mathrm{o}{\mathrm{t}}_{\mathrm{i}}\times \mathrm{p}\mathrm{o}\mathrm{s}{\mathrm{t}}_{\mathrm{t}}+{\mathsf{\beta }}_2{\mathrm{X}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\gamma }}_{\mathrm{t}}\\ +\mathrm{province}\ast \mathrm{year}+\mathrm{sector}\ast \mathrm{year}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}\end{array},$$

(2)

$$\begin{array}{l}{\mathrm{lnegs}}_{\mathrm{i}\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\lambda }}_1\mathrm{p}\mathrm{i}\mathrm{l}\mathrm{o}{\mathrm{t}}_{\mathrm{i}}\times \mathrm{p}\mathrm{o}\mathrm{s}{\mathrm{t}}_{\mathrm{t}}+{\mathsf{\lambda }}_2{\mathrm{gi}}_{\mathrm{it}}+{\mathsf{\lambda }}_3{\mathrm{X}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\gamma }}_{\mathrm{t}}\\ +\mathrm{province}\ast \mathrm{year}+\mathrm{sector}\ast \mathrm{year}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}\end{array},$$

(3)

where gi_it is the green technology innovation level.

(2)

Moderating effect model

Considering local economic development, several local governments may over-intervene in enterprises, affecting the enterprises’ enthusiasm to carry out carbon quota trading in the market. At the same time, the CTPM is entrusted to the local government for specific implementation. However, given the existence of local government assessment guidance and a competition mechanism, and other factors, the implementation of the CTPM may be alienated by the interference of government behavior; for example, officials’ promotion pressure and fiscal decentralization are important influencing factors. Following Wang et al. [34], this study empirically examines the above moderating effect through the following model:

$$\begin{array}{l}{\mathrm{lnegs}}_{\mathrm{i}\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1\mathrm{p}\mathrm{i}\mathrm{l}\mathrm{o}{\mathrm{t}}_{\mathrm{i}}\times \mathrm{p}\mathrm{o}\mathrm{s}{\mathrm{t}}_{\mathrm{t}}\times {\mathrm{external}}_{\mathrm{k}}+{\mathsf{\alpha }}_2\mathrm{p}\mathrm{i}\mathrm{l}\mathrm{o}{\mathrm{t}}_{\mathrm{i}}\times \mathrm{p}\mathrm{o}\mathrm{s}{\mathrm{t}}_{\mathrm{t}}+{\mathsf{\alpha }}_3{\mathrm{external}}_{\mathrm{k}}\\ +{\mathsf{\alpha }}_4{\mathrm{X}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\gamma }}_{\mathrm{t}}+\mathrm{province}\ast \mathrm{year}+\mathrm{sector}\ast \mathrm{year}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}\end{array}$$

(4)

Equation (4) adds the interaction term between the external government behavior and the CTPM into the DID model. The variable external represents the officials’ promotion pressure and fiscal decentralization. The core variable is pilot × post × external, and according to its coefficient significance, we analyze the moderating effect of government behavior on the CTPM.

3.1.3. Policy Design Analysis Model

Varying policies are adopted given the different environmental backgrounds of each pilot; therefore, their final effects also differ. According to the research of Li et al. [35], this study analyzes the heterogeneous impact of different policies on enterprises’ EGS through the following model:

$$\begin{array}{l}{\mathrm{lnegs}}_{\mathrm{i}\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1\mathrm{lnpe}+{\mathsf{\alpha }}_2\mathrm{lnnumber}+{\mathsf{\alpha }}_3\mathrm{lnccer}+{\mathsf{\alpha }}_4\mathrm{punish}+{\mathsf{\alpha }}_5{\mathrm{intensity}}_{\mathrm{it}}+{\mathsf{\alpha }}_6{\mathrm{grandfather}}_{\mathrm{it}}\\ +{\mathsf{\alpha }}_7{\mathrm{benchmark}}_{\mathrm{it}}+{\mathsf{\alpha }}_8{\mathrm{X}}_{\mathrm{it}}+{\mathsf{\mu }}_{\mathrm{c}}+{\mathsf{\gamma }}_{\mathrm{t}}+{\mathsf{\delta }}_{\mathrm{s}}+\mathrm{province}\ast \mathrm{year}+\mathrm{sector}\ast \mathrm{year}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}\end{array}$$

(5)

where lnpe and lnnumber are the logarithms of the total carbon quota allocation and of the number of enterprises covered in different pilot areas, respectively; lnccer is the logarithm of Chinese certified emission reduction (CCER) transaction volume in different pilot areas; punish indicates the economic punishment given by each pilot to enterprises that fail to submit the verification report and perform the contract on time, set according to the different punishment intensity as the natural numbers (1, 2, 3, 4, 5); and intensity, grandfather, and benchmark refers to historical intensity, grandfather, and benchmarking methods. If the pilot provinces for i enterprises in year t adopt benchmarking as the carbon quota allocation method, then the value is 1, and 0 otherwise. The interpretation for the other two methods is the same. In Equation (5), the city, year, and industry fixed effects are controlled. The samples used in this regression are from 2011 to 2016 in the pilot areas.

3.2. Data Description

3.2.1. Sample Selection

The selected sample study period to examine the impact of the CTPM on the EGS of listed enterprises is from 2005 to 2016. The main reasons for not choosing more recent latest years for the study are as follows. First, the data used to calculate the EGS of enterprises are mainly from China customs data (CCD) and the China Stock Market & Accounting Research (CSMAR) Database. CCD records the import and export and other relevant trade data of enterprises in detail, whereas CSMAR contains the financial and corporate governance data of Chinese listed enterprises. This study integrates and matches the two databases to obtain the export volume data of listed enterprises. The export volume data of CCD at the enterprise level is only available in 2016, and not the latest year. As such, the study period can only acquire data until 2016. Second, in 2017, the Chinese government issued the Construction Plan of National Carbon Emission Trading Market (Power Generation Industry), marking the official launch of NCET, which may interfere with the implementation effect of the CTPM. To better identify the effect of the CTPM on the EGS of enterprises and eliminate the potential impact of the NECT launch, we believe that it is also more important to limit the sample study period to 2016. When evaluating the impact of CTPM, the existing literature has also set the sampling period to 2016 for reasons of data availability and avoiding interference from other factors [25,35].

When matching CCD and CSMAR, the export volume data of listed enterprises were obtained with reference to Qi and Cheng [3]. Given its high degree of pollution emission, the industrial sector is most vulnerable to the impact of the CTPM. Most of the listed companies that carry out the majority of export trade are industrial enterprises. Therefore, the listed enterprises in non-industrial industries were removed from the sample. Enterprises with missing data and suffer serious losses during the sample period (marked as ST or *ST) were also excluded. This study reduced the tail of all continuous variables at 1% and 99% to avoid the influence of extreme values. When measuring the level of green technology innovation of enterprises, their cited data relating to green patents were examined. The cited data of listed enterprises’ patents were collected and the green patents were then manually sorted according to the classification number published by the world intellectual property organization. The cited data of patents were derived from the incoPat Database.

3.2.2. Variable Definition

(1)

Explanatory variable

Carbon trading pilot policy (CTPM): CTPM is obtained by the multiplication of the policy implementation time and area. pilot_i is the dummy variable of the implementation areas of CTPM, which are Beijing, Shanghai, Guangdong, Shenzhen, Chongqing, Tianjin, and Hubei, according to Zhu et al. [25]. The enterprises in these areas are regarded as the treatment group with a value of 1, and the others are set as 0. post_t is the dummy variable of the implementation time of the CTPM. After June 2013, seven domestic carbon markets sequentially began operations. Therefore, when evaluating the impact of China’s CTPM, this study follows the existing literature that takes 2013 as the policy implementation node when building the DID model [16,28].

(2)

Explained variable

Export green-sophistication (EGS): The micro measurement subject of EGS comprises listed enterprises, which differ from the existing literature using macro-data at the regional or industry level. According to the definition and calculation of Li and Lu [11], EGS is mainly composed of three parts of the enterprise: the revealed export comparative advantage, the technical level, and the green degree. The specific calculation of EGS is shown in Appendix A.

(3)

Mediating variable

Green technology innovation (gi): Most of the existing literature uses the number of green patent applications of enterprises to measure green technology innovation. However, the different patents vary greatly in importance. Therefore, measuring technology innovation based on the number of patents may lead to errors in the conclusions. This study considers the ratio of the cited number of green patents applied by enterprises to the cited green patents in their industry as the proxy index of the green technology innovation level, according to Wang et al. [34]:

$${\mathrm{gi}}_{\mathrm{it}}=\frac{{\mathrm{gpc}}_{\mathrm{it}}}{{\mathrm{gpc}}_{\mathrm{st}}},$$

(6)

where gpc_it is the cited number of green patents applied, and gpc_st is the cited number of green patents applied by all enterprises in industry s, to which enterprise i belongs in year t. gi is green innovation quality, which is used to express green technology innovation.

(4)

Moderating variable

Officials’ promotion pressure (opp): Under China’s current administrative system, GDP is still the local officials’ performance evaluation index. Therefore, based on the measurement method of the opp index by Wu et al. [36], this study uses the ratio of the difference between the GDP growth rates of the province and its area. If the value is greater than 0, then the economic performance of the province is poor and the promotion pressure on officials is high.

Fiscal decentralization (fd): fd refers to the delegation by central government of specific responsibilities and fiscal revenue management authority to local governments. The core of fiscal decentralization is that local governments have a certain degree of fiscal autonomy, and can independently formulate policies to a certain extent. Following Chen and Gao [37], we represent fiscal decentralization by the ratio of fiscal revenue and expenditure in the regional budget. In this study, the logarithm of fiscal decentralization is used and expressed by lnfd.

(5)

Policy design variable

Total carbon quota allocation: The coverage of the carbon market determines the basic carbon emissions of the market, and the total carbon quota allocation is the allowable total carbon emissions determined on the basis of the basic carbon emissions and combined with the emission reduction targets of the trading system in future years. Each carbon pilot area considered its own economic development when setting its total carbon quota allocation, and the total quota is different; this is expressed by lnpe after taking the logarithms.

Number of covered enterprises: Each carbon pilot area determines the number of covered enterprises in the carbon market according to its own emissions each year. In this study, the number of covered enterprises is expressed by lnnumber after taking logarithms.

Chinese certified emission reduction (CCER): CCER refers to the certified voluntary emission reduction registered in the carbon market by quantifying the emission reduction of China’s forestry carbon sequestration, renewable energy, and methane utilization projects. Each carbon pilot area takes CCER trading as an important supplementary form for the operation of its own carbon market. The CCER trading volume of each carbon pilot area is expressed by lnccer after taking logarithms.

Economic punishment intensity: If the covered enterprise fails to submit its verification report and fulfill the contract on time, then it is punished according to each pilot area. The pilot carbon markets having fines include Beijing, Shanghai, Hubei, Shenzhen, and Guangdong. In terms of punishment intensity, Shanghai has the strongest whereas Chongqing and Tianjin have the weakest.

Table 1. Economic punishment intensity.

Beijing	Tianjin	Shanghai	Hubei	Guangdong	Shenzhen	Chongqing
Fine of 3–5 times the market price	Mandatory rectification in a limited period	Fine of 5–10 times the market price	a Fine of 1–3 times the market price	Fine of 1–3 times the market price	Fine of 3 times the market price	Mandatory rectification in a limited period

shows the punishment intensities in the pilot areas. In this study, punishment intensity is expressed by punish, set as a natural number (1, 2, 3, 4, 5) according to the economic punishment intensity in each pilot area; that is, the values are 1 for Tianjin and Chongqing carbon markets, 2 for Hubei and Guangdong carbon markets, for which fines range from 1–3 times the market price, etc.

Quota allocation: Grandfather and historical intensity methods refer to the allocation of carbon quota according to carbon emissions or emission intensities of enterprises in previous years. By comparison, the benchmarking method determines the amount of carbon quota according to the emission benchmark (average or advanced value) of products or industries. intensity, grandfather, and benchmark indicate the historical intensity, grandfather, and benchmarking methods, respectively.

(6)

Control variable

This study selects the control variables according to Li and Lu [11], Cui et al. [24], and Zhu et al. [25]. Enterprise export volume: We select the export volume of enterprises as the control variable. To reduce heteroscedasticity, we logarithmically process the export volume of enterprises, expressed by lnex. Enterprise size (lnsize): The logarithm of the total assets is used to represent enterprise size. Enterprise age (lnage): The logarithm of the years of listing is used to represent the age of the enterprise. Tobin Q value (tobinq): The ratio of enterprise market value to capital replacement cost. Asset liability ratio (dar): The ratio of total assets to total liabilities is used to represent asset liability ratio. Total assets profit margin (roa): The ratio of the total profits of the enterprise to the total assets is used to represent total assets profit margin after taking the logarithm. Capital intensity (capint): The total income of the enterprise divided by the total assets of the enterprise, indicating the capital intensity. Shareholding ratio of the largest shareholder (top1): The ratio of the number of shares held by the largest shareholder to the total share capital. Lending capacity (lend): The ratio of the net value of fixed assets to the total assets of the enterprise.

Table 2. Descriptive statistics.

Variables	Obs	Mean	Std. Dev.	Min	Max
Lnegs	10,225	0.6235	3.7822	−18.5805	21.5221
Lnex	10,225	16.0208	2.6018	1.0986	23.1742
Lnage	10,225	1.6561	0.8279	0	3.2188
Lnsize	10,225	21.6269	1.1398	17.8786	29.6532
Roa	10,225	0.0367	0.0971	−3.9944	1.2068
Tobinq	10,225	1.973	1.3271	0.7488	48.5053
Dar	10,225	0.4141	0.2223	0.0075	4.2762
capint	10,225	2.2686	12.5457	0.1314	1039.127
Lend	10,225	0.2537	0.1532	0	0.8494
top1	10,225	0.3703	0.1481	0.0431	0.8855

shows the descriptive statistics of the variables.

4. Empirical Results and Discussion

4.1. Regression Result of DID

Table 3. Impact of CTPM on EGS.

	(1)	(2)	(3)	(4)
Variables	lnegs	lnegs	lnegs	lnegs
Pilot × post	0.294 ***	0.271 ***	0.259 ***	0.243 **
	(0.0931)	(0.0935)	(0.094)	(0.0911)
lnex		0.106 **	0.109 **	0.112 **
		(0.0446)	(0.0444)	(0.0445)
lnzge		0.958 ***	0.960 ***	0.969 ***
		(0.0865)	(0.0862)	(0.0861)
lnsize		0.0121	0.00540	0.00602
		(0.0745)	(0.0741)	(0.0741)
roe		0.0209	0.0201	0.0212
		(0.0191)	(0.0195)	(0.0195)
tobinq		−1.053 ***	−1.072 ***	−1.069 ***
		(0.378)	(0.379)	(0.38)
dar		0.0260	0.0221	0.0221
		(0.0245)	(0.0244)	(0.0245)
capint		−0.0443	−0.0487	−0.0543
		(0.255)	(0.255)	(0.255)
lend		0.000314	0.000135	0.000186
		(0.00390)	(0.00386)	(0.00384)
top1		0.314	0.351	0.379
		(0.309)	(0.309)	(0.310)
Year FE	Yes	Yes	Yes	Yes
Enterprise FE	Yes	Yes	Yes	Yes
Province time trend	No	No	No	Yes
Sector time trend	No	No	Yes	Yes
Observations	10,225	10,225	10,225	10,225
R-squared	0.797	0.807	0.807	0.808

Note: *** and ** represent signifificance at the 1% and 5% levels, respectively.

presents the DID results obtained on the basis of Equation (1). Control variables are excluded in Column 1, but included in Columns 2–4. The explained variables in Columns 1–4 are EGS values. The results show that the DID coefficients from Columns 1–4 are significantly positive, which indicates that the CTPM can effectively improve EGS. We compared the main results with previous studies and find them similar. Although the CTPM is not combined with EGS, the literature presents a near consensus that market incentive ER improves international competitiveness. Zhou and Zhou [2] found that the CTPM can increase China’s export sophistication using provincial level data. Qi and Cheng [3] concluded that a pollution emissions trading system can increase China’s export product quality using listed enterprises’ level data. Qi et al. [28] showed that the CTPM increases China’s low carbon international competitiveness using industry level data.

4.2. Regression Result of Robustness Test

4.2.1. Parallel Trend Test and Counterfactual Analysis

Based on the parallel trend test model in Appendix B, we empirically tested whether the control and treatment groups have parallel trends before the implementation of the CTPM. In

Table 4. Parallel trend test and counterfactual analysis.

	(1)	(2)	(3)	(4)	(5)
Variables	lnegs	lnegs	lnegs	lnegs	lnegs
pilot × post			0.477	0.321	0.342
			(0.511)	(0.345)	(0.387)
pilot × post2008	0.414	0.443
	(0.529)	(0.531)
pilot × post2009	0.217	0.290
	(0.235)	(0.233)
pilot × post2010	0.334	0.351
	(0.212)	(0.305)
pilot × post2011	0.576	0.557
	(0.593)	(0.589)
pilot × post2012	0.708	0.668
	(0.786)	(0.783)
pilot × post2013	0.700 ***	0.656 ***
	(0.190)	(0.188)
pilot × post2014	0.525 ***	0.528 ***
	(0.199)	(0.200)
pilot × post2015	0.654 ***	0.527 **
	(0.207)	(0.210)
pilot × post2016	0.535 **	0.461 *
	(0.234)	(0.236)
Control variables	No	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes
Enterprise FE	Yes	Yes	Yes	Yes	Yes
Province time trend	Yes	Yes	Yes	Yes	Yes
Sector time trend	Yes	Yes	Yes	Yes	Yes
Observations	10,225	10,225	5188	5188	5188
R-squared	0.798	0.809	0.834	0.833	0.834

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

, Columns 1–2 show the results of the parallel trend test where the DID coefficients were not significant in the five years before the implementation of the CTPM but significantly positive after the implementation of the CTPM; this meets the parallel trend test. This also shows that the CTPM is worthy of further promotion and implementation, and ensures the effectiveness of the DID model established in this study. Using counterfactual tests, that is, assuming that the CTPM was implemented in 2008, 2009, and 2010, and excluding the samples in 2011 and after, the regression was carried out according to Equation (1). The DID coefficients in Columns 3–5 indicate that the policy was implemented in 2008, 2009, and 2010, respectively. The DID coefficients in Columns 3–5 are not significant, which verifies the robustness of results.

4.2.2. Eliminate the Impact of Environmental Regulation

During the study period, China has implemented several environmental regulation policies. At the prefecture city level, China implemented the Plan for Acid Rain and Sulfur Dioxide Pollution Control Zones in 1998, the Pilot Policy of Low-carbon Cities in 2010, and the Ambient Air Quality Standard (2012) in 2012. To eliminate the impact of these environmental regulations implemented by prefecture-level cities on the results of this study, we added the intersection term of city and year fixed effects over time to Model (1). The DID coefficients are significantly positive, regardless of whether control variables are added, in Columns 1 and 2 of

Table 5. Eliminate the impact of environmental regulation.

	(1)	(2)	(3)	(4)
Variables	lnegs	lnegs	lnegs	lnegs
pilot × post	0.285 ***	0.211 ***	0.274 ***	0.207 ***
	(0.067)	(0.069)	(0.066)	(0.071)
Control variables	No	Yes	No	Yes
Year FE	Yes	Yes	Yes	Yes
Enterprise FE	Yes	Yes	Yes	Yes
Province time trend	Yes	Yes	Yes	Yes
Sector time trend	Yes	Yes	Yes	Yes
Observations	10255	10255	10255	10255
R-squared	0.797	0.807	0.798	0.807

Note: *** represents signifificance at the 1% level.

. At the industry level, China issued the Green Credit Guideline in 2012, proposing that banking institutions must restrict loans to enterprises in high pollution and energy consumption industries, and raise their loan thresholds to force them to carry out green transformation and upgrading. This policy may have an impact on enterprises’ EGS, and thus must be eliminated. The cross-multiplication term of sector and year fixed effects were added to Model (1) for control. The DID coefficients are significantly positive in Columns 3 and 4 of Table 5.

4.2.3. Instrumental Variables Estimation

The use of the DID method to evaluate the CTPM includes a premise, that is, the government selection of the carbon trading pilot areas must randomly sample from all provinces or cities in China. However, the actual situation is that the selection of the seven carbon trading pilot areas may be affected by other potential factors, such as economic development, carbon emissions, and energy consumption structure. Thus, the sampling of carbon trading pilot areas may have a self-selection problem, which can have a certain impact on the estimation results of DID method. As such, this study further uses the instrumental variables (IV) method to overcome the endogenous problem as much as possible. The selection of IVs needs to meet two conditions related to endogenous variables and IVs must not be related to random disturbance terms. In accordance with the practices of Hering and Poncet [38], this study selected the ventilation coefficient as IV. The reasons for this selection are as follows. First, when the total pollutant emission is certain, a smaller ventilation coefficient leads to greater pollutant monitoring concentration. Such cities tend to adopt more strict environmental regulations, and thus their probability of being selected as carbon trading pilot cities is higher. This finding is consistent with the correlation assumption of IV. Second, the ventilation coefficient is determined by meteorological and geographical conditions, and can meet the exogenous assumption of IV. In the current study, the ventilation coefficient of each city in China is obtained from the database of the European Center for Medium-Range Weather Forecasts. The two-stage results of IV estimation are reported in Columns 1 and 2 of

Table 6. IV estimation.

	(1)	(2)
Variables	lnegs	lnegs
iv × post	−0.163 ***
	(0.021)
pilot × post		0.411 ***
		(0.108)
Control variables	Yes	Yes
Year FE	Yes	Yes
Enterprise FE	Yes	Yes
Province time trend	Yes	Yes
Sector time trend	Yes	Yes
Observations	10,225	10,225
R-squared	0.755	0.325

Note: *** represents signifificance at the 1% level.

. Column 1 presents the regression result of the first stage. The interaction term coefficient between IV and the time variable is significant, indicating that IV meets the correlation conditions. Column 2 shows the regression result of the second stage. The DID coefficient is also significantly positive, indicating that CTPM can still significantly improve EGS after eliminating the endogenous problem in the selection of cities in the experimental group.

4.2.4. DDD

We further verified the robustness of results using the DDD model in the Appendix B. The DDD coefficients are significantly positive in Columns 1 and 4 of

Table 7. DDD estimation.

	(1)	(2)	(3)	(4)
Variables	lnegs	lnegs	lnegs	lnegs
pilot × post × sector	0.021 *	0.053 *	0.062 *	0.085 *
	(0.012)	(0.029)	(0.034)	(0.045)
Control variables	No	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Enterprise FE	Yes	Yes	Yes	Yes
Province time trend	No	No	No	Yes
Sector time trend	No	No	Yes	Yes
Observations	10,225	10,225	10,225	10,225
R-squared	0.785	0.797	0.797	0.797

Note: * represents signifificance at the 10% level.

, indicating that the DDD estimation passed the requirements.

4.2.5. PSM–DID Estimation

The DID method is prone to selective deviation, that is, the DID method cannot ensure that the enterprises in the control and treatment groups have the same individual characteristics before the implementation of the CTPM. This problem is more common in the case of a large sample. In this study, the sample includes listed enterprises, with 10,225 observed values and differences in terms of the scale, geographical location, and financial status. Therefore, the selected samples clearly have large individual differences. To alleviate the problem of selectivity bias as much as possible, we combine propensity score matching (PSM) with the DID model based on the research of Caliendo and Kopeinig [39] and Zhang and Duan [40]. Kernel, nearest neighbor, and caliper matching methods are used to match the enterprises in the control and treatment groups. Then, the matching results are further regressed by the DID model.

Table 8. PSM-DID estimation.

	(1)	(2)	(3)
Variables	lnegs	lnegs	lnegs
pilot × post	0.167 **	0.185 **	0.106 ***
	(0.079)	(0.081)	(0.046)
Control variables	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
Enterprise FE	Yes	Yes	Yes
Province time trend	Yes	Yes	Yes
Sector time trend	Yes	Yes	Yes
Observations	8986	7526	8084
R-squared	0.633	0.695	0.688

Note: *** and ** represent signifificance at the 1% and 5% levels, respectively.

shows the specific regression results of PSM-DID. Columns 1–3 show the estimation results of kernel, nearest neighbor, and caliper matching, respectively. The DID coefficients are significantly positive.

4.3. Mediating Effect of Green Technology Innovation

This study examined the mediating effect of green technology innovation level based on Equations (2) and (3).

Table 9. Mediating effect of green technology innovation.

	(1)	(2)
Variables	gi	lnegs
pilot × post	0.034 *	0.137 ***
	(0.018)	(0.0001)
gi		3.212 *
		(1.794)
Control variables	Yes	Yes
Year FE	Yes	Yes
Enterprise FE	Yes	Yes
Province time trend	Yes	Yes
Sector time trend	Yes	Yes
Observations	10,225	10,225
R-squared	0.491	0.788

Note: *** and * represent signifificance at the 1% and 10% levels, respectively.

shows the results. The DID coefficient is significantly positive in Column 1, indicating that the CTPM has significantly improved the green technology innovation of enterprises. Column 2 shows that the coefficients of DID and of green technology innovation are significant, indicating the mediating effect. Green technology innovation plays a partial mediating effect between the CTPM and EGS, indicating that the CTPM can improve the EGS of enterprises through green technology innovation. That is, the CTPM improves the expectation of enterprises regarding innovation benefits and reduces their concerns about related risks. Enterprises can also increase R&D in cleaner production, improving green technology innovation, to reduce their carbon emissions [16]. Enterprises can exchange excess carbon quotas in the carbon market for additional benefits, resulting in an innovation compensation effect, offsetting the cost pressure arising from the CTPM, and promoting the improvement in production efficiency. Enterprises can produce more products at a lower pollution cost and gain competitive advantage in the export product market. Finally, the EGS of enterprises can be improved.

4.4. Moderating Effect of Government Behavior

Based on Equation (4), this study analyzed the mediating effect of government behavior on CTPM. Column 1 in

Table 10. Moderating effect of government behavior.

	(1)	(2)
Variables	lnegs	lnegs
polit × post × opp	0.098 ***
	(0.022)
polit × post × fd		0.146 ***
		(0.029)
Control variables	Yes	Yes
Year FE	Yes	Yes
Enterprise FE	Yes	Yes
Province time trend	Yes	Yes
Sector time trend	Yes	Yes
Observations	10,225	10,225
R-squared	0.797	0.797

Note: *** represents signifificance at the 1% level.

shows the mediating effect of officials’ promotion pressure on CTPM. The coefficient of pilot × post × opp is significantly positive, which indicates that the greater the promotion pressure of local officials, the stronger the promotion effect of the CTPM on EGS. The possible reason for this is that environmental protection performance is included in the local officials’ evaluation in the current context of sustainable development, and GDP is no longer the only evaluation index. When the promotion pressure is high, local officials resolutely implement the central environmental protection policy while developing the regional economy, and thus strictly supervise the emission reduction of enterprises [34]. Resource integration is also integrated into the field of green technology development to help enterprises carry out green innovation, such that enterprises can eliminate polluting and backward products, produce clean products, and improve their EGS.

Column 2 shows the mediating effect of fiscal decentralization on the CTPM. The coefficient of pilot × post × fd is significantly positive, indicating that the greater the fiscal decentralization, the greater the promotion effect of CTPM on the EGS of enterprises.

4.5. Impact of Policy Design on EGS

Based on Equation (5), this study empirically analyzed the differential effects of various carbon market systems on EGS. The coefficients of total carbon quota allocation (lnpe) in

Table 11. Policy design.

	(1)	(2)
Variables	lnegs	lnegs
lnpe	0.122 ***	0.165 ***
	(0.0214)	(0.0368)
lnnumber	0.234	0.362
	(0.665)	(0.648)
lnccer	−0.098 ***	−0.101 ***
	(0.0234)	(0.0265)
punish	0.935 ***	1.877 ***
	(0.312)	(0.799)
grandfather	−0.631	−0.635
	(0.699)	(0.743)
benchmark	0.675 ***	1.421 ***
	(0.213)	(0.367)
intensity	1.531	1.602
	(2.101)	(2.936)
Control variables	No	Yes
Year FE	Yes	Yes
Enterprise FE	Yes	Yes
Province time trend	Yes	Yes
Sector time trend	Yes	Yes
Observations	1555	1555
R-squared	0.342	0.426

Note: *** represents signifificance at the 1% level.

are negative, indicating that the high total carbon quota allocation is less conducive to the improvement in EGS. When the total carbon quota is set more loosely, the actual total of the carbon emissions of the region decreases relative to the total amount preset in the carbon market, which leads to a lower carbon price [2]. When the carbon price is too low, the enthusiasm of enterprises participating in carbon trading is not high, especially when the cost of fines for a breach of contract is lower than the carbon price. Thus, enterprises do not reduce emissions, which is not conducive to stimulating their motivation to innovate. Enterprises may ignore the constraints of the carbon market and continue to produce high-polluting products, which is not conducive to improvement of their EGS.

Table 11 shows that the coefficients of lnccer are negative, indicating that more transactions of CCER are less conducive to improvement in EGS. Excessive CCER trading volume may have an impact on the demand for carbon quota, reduce the demand for carbon quota, and lower the carbon price [35], which is not conducive to active performance of enterprises and the improvement of EGS.

Table 11 also shows that the coefficients of economic penalties (punish) are positive, indicating that an economic penalty is conducive to the improvement of EGS. When the pilot area has weak economic punishment, enterprises naturally choose to breach the contract and continue to produce highly polluting products. The result is a loss in the deterrence of the punishment mechanism and the reduced price discovery function of the carbon market [35].

The coefficients of grandfather and intensity are negative, whereas the coefficients of benchmarking are positive, indicating that the benchmarking method is significantly better than grandfather and historical intensity methods in improving the EGS of enterprises. Different quota allocation methods have different incentive effects on the green technology innovation of enterprises. The benchmarking method determines the quota allocation according to the advanced emission benchmark of the industry, and thus can encourage development toward such level [28]. To reach the advanced level of the industry, enterprises obtain more free quotas, and then sell the excess in the market to obtain income. The willingness of enterprises to carry out green technology innovation increases, which helps to improve EGS.

The above results of the policy design of the CTPM are similar to the conclusions of the relevant comparative studies on the effectiveness of China’s carbon trading market. For example, Zhou and Zhou [2] found that an excessive CCER trading volume may have a negative impact on export sophistication. Qi et al. [16] and Peng et al. [27] showed that benchmarking performs grandfather and intensity methods in promoting the economic output of the industry. Therefore, the current conclusion is scientific.

5. Conclusions, Policy Implications and Limitations

5.1. Conclusions

Based on the data of Chinese listed enterprises from 2005 to 2016, we empirically studied the impact of the CTPM on EGS of enterprises using the DID model, and drew the following conclusions.

First, the CTPM can effectively promote the EGS of enterprises through green technology innovation.

Second, the promotion pressure of officials and fiscal decentralization have a positive moderating effect on the CTPM.

Third, different carbon trading policy designs have different effects on the EGS of enterprises. The higher the total carbon quota allocation, the larger the trading volume of CCER, and the weaker the punishment for enterprise default in the pilot areas, the less favorable it is for enterprises to improve their EGS. The use of the benchmarking method in the allocation of carbon quotas is conducive to the improvement in the EGS.

5.2. Policy Implications

From the preceding conclusions, the policy implications of the current work are as follows.

First, the CTPM can stimulate the innovation potential of enterprises and improve the EGS. Therefore, the government must focus on implementation of the CTPM. At present, China does not have a law to deal with climate change, has not fundamentally established a mandatory carbon market, and cannot fully guarantee the interests of all participants in the carbon trading market. As a result, the standardized operation of the NCET is restricted. Therefore, according to international and pilot experiences and China’s specific national conditions, a system of laws and regulations must be formulated specifically for the carbon trading market. The main greenhouse gas covered by China’s seven pilot carbon trading markets is carbon dioxide, and these markets do not include nitrous oxide, methane, and other greenhouse gases. As the carbon market matures, the government must gradually expand the greenhouse gas coverage and industry coverage of NCET.

Second, in the policy design of the carbon market, the allocation mode of the carbon quota must transition from free to public auction, and the quota allocation should be mainly conducted using the benchmarking method. Considering the impact of CCER on the supply–demand relationship of the carbon quota, the proportion of CCER must not be too high and must be controlled within an appropriate range. Enterprises that fail to fulfill their contract on time must be given strict economic punishment by government; in addition, their credit capital support should be cancelled and speculation should be eliminated. For enterprises that fulfill the contract on time, the government can consider giving financial and policy support to their emission reduction projects, to improve the enthusiasm of enterprises to fulfill the contract.

5.3. Limitations

Despite the above findings, this study has the following limitations.

First, given the limitation of the availability of export volume data of listed enterprises, the sampling period of this study is from 2003 to 2016. Although this period may exclude the influence of a few potential factors, the latest years are not examined, and the national carbon trading market has officially started trading. Updating the sampling period may lead to richer conclusions.

Second, intermediary channels for the CTPM may affect the EGS of enterprises. This study only empirically tested green technology innovation. The CTPM may also affect EGS through other ways, which were examined later in this study.

Third, the policy design of China’s pilot carbon market contains many characteristics, including the total amount of carbon market quota allocation, coverage of enterprises, quota allocation methods, offset proportion of CCER, punishment for enterprise default, and monitoring, reporting, and evaluation (MRV). Empirically, it is difficult to quantify the characteristics of several policy elements, such as enterprise coverage, transaction mode, and MRV, hindering the comprehensive analysis of all policies and systems. Thus, the impact of different policy designs should be further and comprehensively examined in subsequent research.