Trade Liberalization and Pollutant Emissions: Micro Evidence from Chinese Manufacturing Firms

Abstract

Trade liberalization has enabled China to become a global manufacturing hub. However, an increasing focus on pollutant emissions has accompanied China’s rapid economic growth. This paper uses the Annual Survey of Industrial Firms and the Annual Energy Survey of Industrial Firms from 1998 to 2007 to identify the effects of trade liberalization in final goods and intermediate goods on pollutant emissions of Chinese manufacturing enterprises. The difference-in-difference method is used to analyze the data, with China’s accession to the World Trade Organization serving as an exogenous policy shock that brought trade liberalization. The paper’s findings indicate the following: (1) Trade liberalization has reduced the tariffs on final goods, which has led to a notable reduction in the intensity of pollutant emissions from Chinese manufacturing firms. (2) Trade liberalization in intermediate goods, meanwhile, has led to a significant increase in the pollutant emission intensity of manufacturing firms. However, the emission reduction effect observed in final goods is more pronounced. (3) The difference-in-difference-in-difference method was used to examine the heterogeneity of the pollutant emission effect of trade liberalization across manufacturing firms. Our analysis revealed that trade liberalization has significantly enhanced the pollutant emission intensity of state-owned enterprises while exhibiting no significant effect on foreign invested enterprises. Furthermore, trade liberalization has intensified pollutant emissions among exporting firms.

1. Introduction

Since China acceded to the World Trade Organization (WTO) in 2001, its degree of economic openness has progressed, significantly contributing to its rapid economic development [1,2,3]. As stipulated in its WTO accession agreement, China has notably reduced tariff and non-tariff barriers. China’s accession to the WTO has enabled the country to emerge as the ‘world factory’ for global production and the export of goods [4]. However, the question remains regarding the environmental implications of this economic expansion. Some studies have shown that trade liberalization reduces pollution [1,5,6,7], while others have reached the opposite conclusion [8,9,10]. Hence, the impact of trade liberalization on pollution levels remains unclear [11]. Therefore, our research question was as follows: What are the environmental effects of trade liberalization in the Chinese context?

The study of the relationship between trade and the environment has garnered significant attention. There are three broad perspectives on whether and how trade impacts the environment. First, the Pollution Haven Hypothesis (PHH) posits that developed countries shift polluting aspects of production to developing countries seeking economic growth with comparatively lenient environmental regulations. Consequently, due to weak environmental oversight, pollution-intensive industries cluster, aggravating environmental pollution in the producing countries. However, Duan et al. [12] did not find evidence that reducing trade barriers leads to the concentration of pollution-intensive industries in countries with less stringent environmental regulations. Second, trade liberalization can effectively promote resource allocation [13], enabling technological advancements to accelerate a country’s economic and social development, thereby mitigating environmental pollution. Third, given the varying development scenarios among countries, trade’s economic and environmental impacts cannot be generalized. Following the rapid development of foreign trade after the reform and opening up, China’s ecological environment has faced unprecedented challenges. Drawing upon the theory of comparative advantage, numerous scholars have utilized frameworks encompassing scale, structural, and technological effects to evaluate and test the impact of trade on the environment [14]. For a considerable time, these efforts have focused on testing the PHH, which has long dominated the discourse on this crucial topic. These studies have profoundly enriched our understanding of the nexus between trade and the environment. However, the existing research still harbors significant limitations, particularly in two respects. On the one hand, the existing studies are prone to encountering endogeneity issues during the estimation process of macroeconomic data, thereby introducing potential bias when identifying the nexus between trade and the environment. On the other hand, the existing studies predominantly rely on national, regional, or industry-level data, overlooking the influence of firm heterogeneity on firms’ pollutant emissions. This approach constrains the exploration of the link between trade and the environment within the framework of heterogeneous trade theory.

This paper examines the policy impacts of trade liberalization in final and intermediate goods on the pollutant emissions of Chinese manufacturing firms, utilizing the Annual Survey of Industrial Firms (ASIF) and the Annual Energy Survey of Industrial Firms (AESIF) from 1998 to 2007. Employing the difference-in-difference (DID) method with China’s accession to the WTO as an exogenous policy shock that brought trade liberalization, this study also delves into the heterogeneous effects of trade liberalization on firm pollutant emissions from the perspectives of ownership and export. This approach effectively identifies the micro-mechanisms through which trade liberalization affects firm pollutant emissions.

By comprehensively assessing the impact of trade liberalization policies on the pollutant emission intensity of Chinese firms, this paper makes the following contributions in three crucial aspects. Firstly, this paper adopts the combination of Principal Component Analysis (PCA) and DID, which accurately identifies the causality between trade liberalization and the pollutant emission intensity of firms, thus overcoming the endogeneity problem in the existing studies. Secondly, this paper adopts enterprise-level data on pollutant emission monitoring, which not only mitigates the measurement bias issue of existing studies that consider only single-pollutant emissions but also allows us to examine the heterogeneous impact of trade liberalization on the pollutant emissions of enterprises more adequately, thereby enabling a precise identification of the underlying mechanisms. Finally, this paper distinguishes the differences between the impacts of trade liberalization concerning final and intermediate goods. Also, it considers firm heterogeneity, including the nature of firms’ ownership and export characteristics. Furthermore, the environmental effects of trade liberalization are analyzed comprehensively.

This paper is organized as follows: Section 2 provides a comprehensive literature review; Section 3 introduces the data, econometric model, and variables’ measurement; Section 4 presents our empirical analysis and robustness test; Section 5 offers a deeper analysis of trade liberalization in intermediate goods and the heterogeneity among firms; and finally, this paper concludes with policy recommendations.

2. Literature Review

The relationship between international trade and environmental pollution remains a pivotal issue in environmental economics, with numerous scholars engaging in discussions on the topic. However, a consensus has yet to be achieved. In early studies, scholars decomposed the impact of international trade on pollutant emissions into three components: the scale effect, structural effect, and technological effect. They further conducted numerous empirical studies utilizing cross-country data. Antweiler et al. [14]’s study is the most representative of the literature. Drawing upon a partial equilibrium framework, they decompose the impact of trade on pollution into three components: the scale effect, structural effect, and technological effect. They leverage cross-country data to demonstrate that the relationship between trade and the environment is not solely linear; instead, it is contingent upon a country’s comparative advantage and classification. In this way, they elucidated the varying impacts of trade on the environment across different contexts. Numerous studies have scrutinized the environmental implications of China’s foreign trade, employing domestic data, and have arrived at numerous significant conclusions. Song et al. [15] used provincial data to determine that China’s environmental problems have become increasingly severe since it acceded to the WTO, and that there are significant differences in the factors influencing environmental efficiency across provinces. Chen et al. [16] examined the impact of China’s accession to the WTO on air pollution, utilizing NASA’s air pollution data. However, the analytical framework proposed by Antweiler et al. [14] encounters challenges in disentangling the causal relationship between trade and the environment due to endogeneity issues. Consequently, the accuracy of causality inference using this framework is compromised.

With environmental issues becoming one of the most important kinds of issues in global economic governance, the PHH has become a hotspot for many studies. Numerous studies have used cross-country data to test the PHH, and the results of these studies have produced little evidence in support of the PHH. Based on data covering 27 EU countries from 2004 to 2017, Ponce et al. [17] found that liberalization of the internal energy market is negatively associated with carbon dioxide (${\mathrm{C}\mathrm{O}}_2$) emissions. However, the relationship between trade liberalization and pollutant emissions can change depending on the selection of different pollution indicators and, thus, the results can also change. Onwachukwu et al. [18] used data from 196 countries from 1970 to 2016 based on the DID model to investigate the impact of trade liberalization on environmental degradation. WTO accession resulted in ${\mathrm{C}\mathrm{O}}_2$ and methane emissions in the treatment group 18.5% and 3.3% higher than in the control group, respectively. However, the treatment group’s nitrous oxide (${\mathrm{N}\mathrm{O}}_{\mathrm{x}}$) emissions were 6% lower than in the control group. Based on 20 years of panel data for 179 significant countries, Ma and Wang [19] found that international trade participation reduced the emission intensity of ${\mathrm{C}\mathrm{O}}_2$ but not sulfur dioxide (${\mathrm{S}\mathrm{O}}_2$). Trade liberalization may be linked with different pollutant emission indicators due to the differences in each country’s trade structure, industrial system, and trade policy. Therefore, the study of trade liberalization in China needs to take into account the specific situation of China.

Focusing on China’s trade liberalization and pollutant emissions, these studies’ results have produced little evidence supporting the PHH. Trade liberalization has significantly reduced pollutant emissions [17,18]. Specifically, Tan et al. [20] found that bilateral trade has reduced global carbon emissions, and it has been reported that trade liberalization can contribute to reducing ${\mathrm{C}\mathrm{O}}_2$ emissions [21,22,23]. Moreover, Xu et al. [1] found that trade liberalization has significantly reduced haze pollution. In further work, it was reported that trade liberalization can reduce ${\mathrm{C}\mathrm{O}}_2$, ${\mathrm{S}\mathrm{O}}_2$, and ${\mathrm{N}\mathrm{O}}_{\mathrm{x}}$ emissions, with trade in intermediate goods having a more significant reduction effect than trade in final goods [24]. However, some scholars also endorse the PHH. For instance, Hu et al. [25] found that trade liberalization in environmental goods does not necessarily benefit the environment. Moreover, Bombardini and Li [26] found that the rapid expansion of Chinese exports from 1990 to 2010 led to a deterioration in China’s environmental quality. Meanwhile, He et al. [27] found that trade liberalization significantly changes Chinese export behavior and exacerbates pollutant emissions. An increase of 1% in foreign direct investment (FDI) leads to a 0.058% increase in ${\mathrm{C}\mathrm{O}}_2$ emissions [8] and a 0.098% increase in industrial ${\mathrm{S}\mathrm{O}}_2$ emissions [28]. Trade liberalization has made China the worst pollution haven through its trade in final products, resulting in 829.4 Mt ${\mathrm{C}\mathrm{O}}_2$, 4.5 Mt ${\mathrm{S}\mathrm{O}}_2$, and 2.6 Mt ${\mathrm{N}\mathrm{O}}_{\mathrm{x}}$ emissions in 2014, which have since increased and may continue to increase [24]. López [29] found that Chinese exports increased world emissions to 1.28 Gt ${\mathrm{C}\mathrm{O}}_2$ in 2008. Song et al. [15], meanwhile, used provincial data to determine that China’s environmental problems have become increasingly severe since it acceded to the WTO. Moreover, Yi et al. [30] found that trade liberalization in intermediate goods exacerbated firms’ pollutant emissions, and Chen et al. [31] found that export liberalization led to a relative increase in water pollutant emissions from previously low-polluting areas in China. Further scholars have argued that the relationship between trade liberalization and pollutant emissions is not monolithic, and that the results vary under different conditions. Zheng and Shi [5], for example, posited that the validity of the PHH is contingent upon the type of environmental policy and industry characteristics. Overall, based on the available literature, it seems the research on the relationship between trade liberalization and the environment needs to be developed in order to become more conclusive.

Trade theory continues to evolve, and many scholars have approached the study of related environmental issues with a trade theory approach. Numerous studies have used input–output (IO) tables to account for the environmental pollution of trade in different industries and analyze the pollution effects of international trade based on factor endowment theory. A representative study is that of Wang et al. [7], who used multi-regional IO analyses and found that domestic trade accounts for about one-third of China’s total ${\mathrm{S}\mathrm{O}}_2$ emissions, and that inter-provincial transfers of ${\mathrm{S}\mathrm{O}}_2$ in trade depend mainly on the trading pair’s populations, economic development, coal consumption, and distance. Moreover, using data from China’s IO tables, Bombardini and Li [26] created alternative models for pollution and income export shocks. Although these studies have enhanced our understanding of the relationship between trade and the environment, their conclusions are limited by the general assumption that technical coefficients remain constant over time in IO analyses. This includes within it several key assumptions that significantly restrict the applicability of the findings. Furthermore, a significant issue with testing the PHH and measuring the pollution of international trade is a disregard for enterprise heterogeneity. This paper aims to address those gaps in the research. With the continuous development of heterogeneous trade theory, represented by Melitz [32], modern international trade research has increasingly taken firm heterogeneity as a central factor in determining the sources of trade and its welfare allocation, and on this basis, a large number of studies have tested and analyzed the welfare effects of trade liberalization, starting from heterogeneous trade theory. At the same time, the large amount of firm-level microdata that is now available provides better opportunities than in the past to identify the causes of international trade and its economic effects.

3. Data, Econometric Model, and Variables’ Measurement

3.1. Data

To reveal the impact of trade liberalization on the pollutant emissions of Chinese manufacturing enterprises, this paper uses the following sources of data on the characteristics of enterprises, their pollutant emission intensity, and trade liberalization: the ASIF, the AESIF, and the China Tariff Database:

This paper uses the ASIF from 1998 to 2007 to characterize Chinese enterprises’ fundamental operational and performance traits. The National Bureau of Statistics compiles this database based on quarterly and annual reports submitted by all state-owned enterprises (SOEs) and non-SOEs exceeding a specified size threshold, including those with annual sales exceeding RMB 5 million. The database encompasses data on sales, value added, assets, liabilities, equity, and other pertinent production and financial metrics. We exclude non-manufacturing firms, according to the National Economic Industry Classification, to ensure data reliability. Furthermore, we classify the manufacturing industry by harmonizing industry codes from before and after 2003 into a unified sub-category industry code (totaling 482 industries), following the adjusted code proposed by Brandt et al. [33]. Consistent with Brandt et al. [33], sample data exclude firms exhibiting financial inconsistencies, such as having total assets less than the fixed or current assets, negative value added, adverse employment or sales figures, or employing fewer than eight individuals. This approach ensures the reliability of our sample data.

This paper utilizes the AESIF from 1998 to 2007 to delineate the fundamental characteristics of enterprise pollutant discharge intensity. The Ministry of Environmental Protection uses the pollutant emission monitoring system of vital industrial enterprises, encompassing comprehensive reporting of industrial wastewater emissions, chemical oxygen demand (COD), ammonia nitrogen emissions, exhaust gas emissions, ${\mathrm{S}\mathrm{O}}_2$ emissions, ${\mathrm{N}\mathrm{O}}_{\mathrm{x}}$ emissions, soot emissions, industrial dust emissions, and pertinent details about pollutant emissions and enterprise management. Data collection is achieved by installing automated monitoring equipment and proceeding with continuous and manual daily monitoring. The database includes information on enterprise names and corresponding legal person codes, which this paper employs to establish linkages with the ASIF. Firstly, this study employed the enterprise’s legal person code as a primary matching criterion to identify enterprises with identical codes in both databases. Secondly, the enterprise name was utilized as an additional matching criterion to further identify enterprises with identical names. Finally, we computed the similarity ratio of enterprise names in the two databases. Subsequently, we manually verified enterprises with a similarity ratio of 0.7 or higher. A matched database encompassing the ASIF and the AESIF from 1998 to 2007 was established, encompassing 265,296 information samples from 157,195 enterprises. Consistent with Sun et al. [34] and Le et al. [35], samples with negative enterprise discharge values were excluded, resulting in a final dataset of 115,004 enterprise discharge samples.

This paper uses the tariff rates reported by the WTO (via 6-digit HS codes) in 1997 and 2000–2007. Then, it maps the tariff codes to economic sectors, to identify the average 4-digit tariff rates of the corresponding economic sectors, thereby identifying trade liberalization in different economic sectors. In addition, the IO table of 42 industries in China in 2002 is used to calculate the corresponding tariff rates for intermediate products in each industry.

3.2. Econometric Model

This paper tests and analyses the causal effects of trade liberalization and adopts the cost-plus identification strategy. The DID method analyzes China’s accession to the WTO as a proposed natural experiment. This paper estimates the impact of China’s accession to the WTO, and the associated trade liberalization policy, on pollutant emissions. The estimation strategy involves calculating the DID between Chinese manufacturing firms before and after the experiment. Here, we estimate the DID between the experimental group and the control group. With this approach, the effects of endogenous problems can be avoided to a certain extent, and unbiased and efficient estimators can be obtained.

The DID model is an essential strategy and methodology for effectively estimating the economic effects of trade liberalization. When constructing a DID model, it is essential to consider the pre- and post-experiment periods and the division between the experimental and control groups. These factors are crucial for accurate estimation. On the one hand, some scholars have identified China’s accession to the WTO in December 2001 as creating an exogenous policy shock [36,37,38]. Numerous studies have shown this to be consistent with the exogenous nature of the related policy. On the other hand, to distinguish between the experimental and control groups amid the policy shock, this paper uses the high-tariff industry before the policy shock as the experimental group and the low-tariff industry as the control group [37]. Figure 1 shows a positive correlation between the 2001 tariff level and the change in the tariff level from 2001 to 2007. This suggests that firms with high tariffs experienced more significant reductions in their tariffs after China joined the WTO than firms with low tariffs. The baseline model presented in this paper is as follows:

$${Pollution}_{fit}=\alpha +\beta {Tariff}_{i,2001}\times {Post}_{2002}+{\gamma }_1{X}_{ft}+{\gamma }_2{Z}_{it}+{\epsilon }_{it}$$

(1)

where ${Pollution}_{fit}$ is the pollutant emission intensity of firm f in industry i in year t; ${Tariff}_{i,2001}$ is the average tariff level on final goods in sector i in 2001; ${Post}_{2002}$ is a dummy variable for the time of China’s accession to the WTO, taking 1 for 2002 and later, and 0 otherwise; $X_{ft}$ is a time-varying firm-level control variable, and $Z_{it}$ is a time-varying industry-level control variable used to avoid endogenous problems associated with the omission of important explanatory variables. The estimated coefficients of $X_{ft}$ and $Z_{it}$ are ${\gamma }_1$ and ${\gamma }_2$, respectively. The central explanatory variable of interest in this paper is the DID term ${Tariff}_{i,2001}\times {Post}_{2002}$, whose estimated coefficient $\beta$ tests the effect of trade liberalization on the pollutant emission intensity of firms. If the estimated coefficient of $\beta$ is significantly less than 0, this indicates that trade liberalization significantly reduces firms’ pollutant emissions. $\alpha$ is the constant term; ${\epsilon }_{it}$ is the error term.

Furthermore, this paper draws on Mao and Xu’s [39] test of the resource allocation and employment effects of trade liberalization, to identify the impact of trade liberalization on the pollutant emission intensity of firms that use intermediate goods. The DID approach is used for this purpose. The estimating equations are as follows:

$${Pollution}_{fit}=\alpha +{\beta }_1{Tariff}_{i,2001}\times {Post}_{2002}+{\beta }_2{Input}_{i,2001}\times {Post}_{2002}+{\gamma }_1{X}_{ft}+{\gamma }_2{Z}_{it}+{\epsilon }_{it}$$

(2)

where ${Input}_{i,2001}$ is the average tariff level of intermediate goods in industry i in 2001; this paper uses the average tariff level of intermediate goods before the trade liberalization policy shock to distinguish between the experimental group and the control group in different industries, and the industries with higher tariffs on intermediate goods before the policy shock also had greater tariff reductions under the policy shock, thus identifying the DID effect. In the baseline model (2), this paper focuses on the estimated coefficients of ${\beta }_1$ and ${\beta }_2$. The two estimated coefficients correspond to the pollutant emission effects of the liberalization of trade in final goods and intermediate goods, respectively.

3.3. Variables’ Measurement

The main focus of this paper is measuring the intensity of pollutant emissions from firms. Drawing on the measurements of pollutant emissions from Xue et al. [40], this paper employs seven indicators: industrial wastewater emissions, COD, industrial exhaust emissions, ${\mathrm{S}\mathrm{O}}_2$ emissions, ${\mathrm{N}\mathrm{O}}_{\mathrm{x}}$ emissions, soot emissions, and dust emissions. These indicators are utilized to construct a comprehensive corporate pollutant emission intensity variable by applying PCA. To eliminate the influence of scale differences on industrial pollutant emissions, this paper uses the ratio of pollutant emissions to the value added for industrial enterprises as the principal component. This approach constructs a comprehensive and integrated enterprise pollutant emission index, avoiding the problem of measurement bias. Specifically, this paper standardizes the principal components to construct the sample criteria matrix:

$$Z_{fs}={\displaystyle \frac{x_{fs}-{\overline{x}}_s}{{\sigma }_s}}$$

(3)

where $x_{fs}$ is the sth pollutant emission intensity of firm f; and $Z_{fs}$ is a standardized indicator for each pollutant emission from the enterprise. In this paper, based on the standardized matrix $Z_{fs}$ used to find the correlation coefficient matrix R, we solve the characteristic equation of the sample correlation coefficient matrix R and find its characteristic root, according to the variance, to determine the number of principal components, the principal components of the weighted summation, and the weight is the contribution of the variance of the principal components, and ultimately, ascertain the comprehensive pollution index of each enterprise, which can reflect the intensity of the enterprise’s pollutant emissions comprehensively.

In order to avoid the impact of firm- and industry-level omitted variables on the estimation results, this paper includes the following control variables in the baseline model: the total factor productivity (TFP) of firms, which is estimated in this paper using the methodology of Olley and Pakes [41]; the size of the firm (Size), which is measured in this paper as the logarithm of the number of employees in the firm; the number of years the firm has been in business (Age), which in this paper takes the logarithmic form of the number of years the firm has been in existence as a proxy variable; the firm’s capital–labor ratio (KL), measured in this paper using the logarithm of the ratio of the firm’s fixed capital stock to the number of employees; the firm’s export intensity (Export), where the logarithmic form of the ratio of the firm’s export delivery value to sales is used as a proxy variable in this paper; state-owned enterprises (SOE), which are defined as dummy variables in this paper through the type of enterprise registration, i.e., enterprises with enterprise registration form codes of 110, 141, 143, and 151 are defined as SOEs, and the SOE variable is set to be 1, or 0 otherwise; foreign invested enterprises (FIE), which are defined as a dummy variable in this paper through the type of enterprise registration, i.e., enterprises with enterprise registration form codes labeled 310, 320, 330, 340, 210, 220, 230, and 240 are defined as FIEs, and the dummy variable is set to be 1, or 0 otherwise; the proportion of SOEs in the industry (State), where this paper uses the proportion of SOEs in the industry as its proxy variable; the intensity of government subsidies to firms (Subsidy), expressed in a logarithmic form as the ratio of firms’ subsidies to firms’ value added; competition in the industry (Competition), for which this paper uses the industry Herfindahl index calculated on the basis of enterprise sales—the larger the index, the stronger the degree of competition in the industry—in order to control for the impacts of the market structure and competition in the domestic market on the enterprise’s pollutant emissions.

Table 1. Variable descriptions, metrics, and data sources.

Variables	Descriptions	Metrics	Data Sources
${Tariff}_{2001}$	Final product tariff data	Simple arithmetic means of tariff levels in 2001 taken in logarithms	China Tariff Database
${Post}_{2002}$	Years after tariff reductions	1 for 2002 and onward; 0 otherwise
Pollution	Pollutant emissions from enterprises	Use of PCA to construct a composite indicator of pollutant emissions	AESIF
TFP	Total factor productivity of enterprises	Total factor productivity of enterprises based on the OP method of calculation	ASIF
Size	Enterprise size	Logarithmic value of the number of employees in the enterprise	ASIF
Age	Number of years in business	Years of business establishment	ASIF
KL	Enterprise capital–labor ratio	Logarithm of the ratio of the firm’s capital stock to the number of employees	ASIF
Export	Export intensity of enterprises	Logarithm of the ratio of firms’ exports to sales	ASIF
SOE	State-owned enterprises	1 for state-owned enterprises; 0 otherwise	ASIF
FIE	Foreign invested enterprises	1 for foreign invested enterprises; 0 otherwise	ASIF
Subsidy	Intensity of corporate government subsidies	Logarithm of business subsidies as a share of value added	ASIF
State	Degree of nationalization of the industry	Percentage of the number of state-owned enterprises in the industry	ASIF
Competition	Degree of competition in the industry	Herfindahl index based on sales	ASIF

describes the variables relevant to this paper and explains their data sources.

4. Results and Analysis

4.1. Baseline Results

This paper estimates the baseline model (1) using ASIF and AESIF data from 1998 to 2007, to examine the impact of trade liberalization on the emission intensity of the Chinese manufacturing sector. To ensure the robustness of the estimation results and avoid the effects of omitted explanatory variables, we included a set of proxy variables that affect firms’ emissions intensity in the model using stepwise regression. Additionally, we controlled the model’s fixed effects of year, enterprise, region, and industry to avoid the influence of unobservable effects. In this way, we obtained unbiased estimators for the DID term.

In column (1) of

Table 2. Baseline results: The effect of trade liberalization on firms’ pollutant emissions.

Explaining Variables	$\mathbf{Explained} \mathbf{Variables}: {\mathit{P}\mathit{o}\mathit{l}\mathit{l}\mathit{u}\mathit{t}\mathit{i}\mathit{o}\mathit{n}}_{\mathit{f}\mathit{i}\mathit{t}}$
	(1)	(2)	(3)	(4)
${Tariff}_{i,2001}\times {Post}_{2002}$	−0.040 *** (0.005)	−0.047 *** (0.004)	−0.045 *** (0.004)	−0.045 *** (0.004)
TFP		−0.388 *** (0.004)	−0.388 *** (0.004)	−0.388 *** (0.004)
Size		−0.090 *** (0.004)	−0.091 *** (0.004)	−0.091 *** (0.004)
Age		0.007 *** (0.002)	0.006 *** (0.002)	0.006 *** (0.002)
KL		−0.013 *** (0.002)	−0.013 *** (0.002)	−0.013 *** (0.002)
Export		0.009 (0.013)	0.009 (0.013)	0.009 (0.013)
SOE			0.003 (0.003)	0.003 (0.003)
FIE			−0.006 * (0.003)	−0.006 * (0.003)
State			0.172 *** (0.033)	0.170 *** (0.033)
Subsidy			0.021 (0.049)	0.022 (0.049)
Competition				−0.218 *** (0.084)
Year fixed effect	Yes	Yes	Yes	Yes
Firm fixed effect	Yes	Yes	Yes	Yes
Regional fixed effect	Yes	Yes	Yes	Yes
Industry fixed effect	Yes	Yes	Yes	Yes
$R^2$	0.818	0.873	0.873	0.873
Observations	174,634	174,634	174,634	174,634

The symbols *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. The robust standard deviations are shown in parentheses. This paper includes industry fixed effects at the four-digit industry level of the National Economic Industry Classification and area fixed effects at the prefecture level.

, the coefficient of the DID term is −0.040, signifying statistical significance and negative at the 1% level without including control variables. This suggests that trade liberalization significantly reduces the pollutant emission intensity of China’s manufacturing enterprises, thereby contributing to improved energy efficiency and reduced emissions among these enterprises. Subsequently, additional control variables, such as the total factor productivity of enterprises, enterprise size, enterprise operating years, enterprise capital–labor ratio, and enterprise export intensity, are introduced in column (2). The estimated coefficient of the DID term is −0.047, which is statistically significant and negative at the 1% level. Furthermore, additional control variables, such as dummy variables for SOEs, dummy variables for FIEs, the industry share of SOE, and subsidy intensity, are introduced in column (3). The estimated coefficient of the DID term is −0.045, which is significantly negative at the 1% significance level. In column (4), the regression results include the degree of competition in the domestic market, to control the effect of competition on the intensity of firms’ pollutant emissions. The estimated coefficient of the DID term is −0.045, which is significantly negative at the 1% significance level. The baseline results indicate that China’s accession to the WTO significantly reduced import tariffs. Furthermore, trade liberalization significantly reduced firms’ pollutant emission intensities. This estimation remains robust after controlling for various control variables and fixed effects at all levels.

In the baseline regression, this paper methodically adds control variables to avoid the problem of missing explanatory variables. The estimated coefficient of enterprise operating years is significantly positive, indicating that a longer operating time of enterprises correlates positively with the pollutant emission intensity of enterprises. The estimated coefficient of the enterprise capital–labor ratio is significantly negative, indicating that enterprises’ energy saving and emission reduction are positively associated with the increase in the capital of enterprises. Meanwhile, the estimated coefficient of enterprise export intensity is not significant. Notably, the baseline regression results remain unaffected by including control variables related to firm characteristics.

In order to avoid omitting the potential impacts of government behavior on pollutant emissions, this paper adds enterprise ownership, the proportion of SOEs, and subsidies as control variables. Among them, the estimated coefficients of the dummy variable for SOEs are insignificant. In contrast, the estimated coefficients of the dummy variable for FIEs are significantly negative at the 10% significance level. This suggests that enterprise ownership has a differentiated influence on the intensity of enterprise pollutant emissions, indicating that FIEs are more advantageous in energy saving and emission reduction. The estimated coefficients of the proportion of SOEs are significantly positive at the 1% significance level, indicating that an increase in the government’s control over the market will significantly increase the pollutant emission intensity of enterprises. The estimated coefficient of the subsidy variable is not significant. The baseline results remain unaffected by the inclusion of the government marketization factor. Competition is a pivotal factor in shaping the pollutant emission intensity of enterprises. When the degree of industry competition is introduced as a control variable, the resulting estimated coefficient exhibits a significant negative value at the 1% significance level. This finding underscores that as the level of competition within an industry intensifies, enterprises tend to reduce their pollutant emissions, thus revealing a pronounced competition effect.

4.2. Robustness Checks

4.2.1. Other Policy Shocks and Omission of Important Variables

To ensure the robustness of the baseline results is not influenced by other policy shocks or the omission of crucial explanatory variables, this paper incorporates controls for three policy shocks: special economic zones (SEZ), environmental legislation, and industrial policy. Additionally, foreign direct investment (FDI) is included as an essential explanatory variable to remove the risk of its omission. These control variables are subsequently incorporated into the baseline model (1) for the regression analysis, and the resulting estimations are presented in

Table 3. Robustness estimation.

Explaining Variables	$\mathbf{Explained} \mathbf{Variables}: {\mathit{P}\mathit{o}\mathit{l}\mathit{l}\mathit{u}\mathit{t}\mathit{i}\mathit{o}\mathit{n}}_{\mathit{f}\mathit{i}\mathit{t}}$
	(1)	(2)	(3)	(4)
	SEZ	Environmental Legislation	Industrial Policy	FDI
${Tariff}_{i,2001}\times {Post}_{2002}$	−0.047 *** (0.005)	−0.045 *** (0.004)	−0.047 *** (0.004)	−0.045 *** (0.004)
Legislation		−0.017 *** (0.004)
Policy			0.022 *** (0.005)
FDI				0.006 (0.013)
TFP	−0.386 *** (0.004)	−0.388 *** (0.004)	−0.388 *** (0.004)	−0.389 *** (0.004)
Size	−0.087 *** (0.005)	−0.091 *** (0.004)	−0.092 *** (0.004)	−0.091 *** (0.004)
Age	0.006 ** (0.003)	0.006 *** (0.002)	0.006 *** (0.002)	0.007 *** (0.002)
KL	−0.012 *** (0.002)	−0.013 *** (0.002)	−0.013 *** (0.002)	−0.015 *** (0.002)
Export	0.018 (0.015)	0.008 (0.013)	0.008 (0.013)	0.009 (0.013)
SOE	0.006 (0.004)	0.003 (0.003)	0.004 (0.003)	0.003 (0.003)
FIE	−0.005 (0.004)	−0.006 * (0.003)	−0.006 * (0.003)	−0.006 * (0.003)
State	0.193 *** (0.037)	0.175 *** (0.033)	0.170 *** (0.033)	0.162 *** (0.033)
Subsidy	0.078 (0.048)	0.021 (0.049)	0.022 (0.049)	0.016 (0.049)
Competition	−0.246 ** (0.101)	−0.221 *** (0.084)	−0.215 ** (0.084)	−0.214 ** (0.084)
Year fixed effect	Yes	Yes	Yes	Yes
Firm fixed effect	Yes	Yes	Yes	Yes
Regional fixed effect	Yes	Yes	Yes	Yes
Industry fixed effect	Yes	Yes	Yes	Yes
$R^2$	0.873	0.873	0.873	0.873
Observations	128,918	174,634	174,634	174,634

The symbols *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. The robust standard deviations are shown in parentheses. The models include DID cross-multiplier terms, single terms, and controls for fixed effects.

.

In column (1) of Table 3, the regression results exclude data about SEZs from the panel, and the estimated coefficient of the DID term is significantly negative at the 1% significance level, indicating that a firm’s location within an SEZ does not significantly influence the firm’s pollutant emission intensity. Regions with a greater implementation of local environmental legislation exhibit more significant environmental benefits post-legislation [38]. Therefore, the regression results include the interaction term (whether there is environmental legislation at the provincial level and whether the legislation’s enactment year precedes the firm’s production activity) as the environmental legislation policy control variable. In column (2), the estimated coefficient is significantly negative at the 1% significance level, indicating that environmental legislation significantly inhibits corporate emissions. In column (3), the regression incorporates a dummy variable indicating whether the industry receives industrial policy support. The estimated coefficient is significantly positive at the 1% significance level, suggesting that industrial policy support exacerbates the intensity of corporate pollutant emissions. FDI is introduced in column (4) of the regression analysis. The estimated coefficient of FDI is rendered insignificant, notwithstanding the significant impact of FDI on the output.

4.2.2. Experimental Group Identification

Figure 1 illustrates a positive correlation between the average final product tariffs in 2001 and the tariff changes observed between 2001 and 2007. This correlation suggests that industries with initially high tariffs experienced greater tariff reductions following China’s accession to the WTO than those with initially low tariffs. In columns (1) to (4) of

Table 4. Identification of the experimental group results.

Explaining Variables	$\mathbf{Explained} \mathbf{Variables}: {\mathit{P}\mathit{o}\mathit{l}\mathit{l}\mathit{u}\mathit{t}\mathit{i}\mathit{o}\mathit{n}}_{\mathit{f}\mathit{i}\mathit{t}}$
	(1)	(2)	(3)	(4)	(5)	(6)
					50%	33%
$d{Tariff}_i\times {Post}_{2002}$	−0.004 *** (0.000)	−0.004 *** (0.000)	−0.004 *** (0.000)	−0.004 *** (0.000)
$I\left(Tariff\right)\times {Post}_{2002}$					−0.036 *** (0.005)	−0.043 *** (0.005)
TFP		−0.387 *** (0.004)	−0.387 *** (0.004)	−0.387 *** (0.004)	−0.387 *** (0.004)	−0.387 *** (0.004)
Size		−0.093 *** (0.004)	−0.093 *** (0.004)	−0.094 *** (0.004)	−0.094 *** (0.004)	−0.094 *** (0.004)
Age		0.007 *** (0.002)	0.006 ** (0.002)	0.006 ** (0.002)	0.006 ** (0.002)	0.006 ** (0.002)
KL		−0.013 *** (0.002)	−0.013 *** (0.002)	−0.013 *** (0.002)	−0.013 *** (0.002)	−0.013 *** (0.002)
Export		0.008 (0.013)	0.007 (0.013)	0.007 (0.013)	0.007 (0.013)	0.008 (0.013)
SOE			0.004 (0.003)	0.004 (0.003)	0.004 (0.003)	0.003 (0.003)
FIE			−0.007 ** (0.003)	−0.007 ** (0.003)	−0.006 * (0.003)	−0.006 ** (0.003)
State			0.136 *** (0.033)	0.134 *** (0.033)	0.172 *** (0.032)	0.173 *** (0.032)
Subsidy			0.015 (0.050)	0.016 (0.049)	0.017 (0.049)	0.013 (0.049)
Competition				−0.215 ** (0.085)	−0.244 *** (0.085)	−0.260 *** (0.085)
Year fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Firm fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Regional fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Industry fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
$R^2$	0.819	0.873	0.873	0.873	0.873	0.873
Observations	173,305	173,305	173,305	173,305	173,305	173,305

The symbols *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. The robust standard deviations are shown in parentheses. The models include DID cross-multiplier terms, single terms, and controls for fixed effects.

, the change in the tariff level of final goods $d{Tariff}_i$ between 2001 and 2007 replaces ${Tariff}_{i,2001}$ of the DID term in the baseline model (1). In columns (5) and (6) of Table 4, a dummy variable, $I\left(Tariff\right)$, replaces ${Tariff}_{i,2001}$ of the DID term in the baseline model (1). When the rate of change in tariff levels between 2001 and 2007, ${\displaystyle \frac{{Tariff}_{i,2001}-{Tariff}_{i,2007}}{{Tariff}_{i,2001}}}$, is greater than the critical value ζ, $I\left(Tariff\right)=1$; otherwise, $I\left(Tariff\right)=0$. The critical value ζ is taken using the median division and Sivadasan division [42], i.e., the median and 33rd percentile of the tariff rate of change, respectively.

In columns (1) to (4) of Table 4, the coefficients of the core explanatory variables are all significantly negative at the 1% significance level after gradually adding control variables and controlling for fixed effects. The results show that the more significant the decline in tariff levels, the lower the intensity of firms’ discharges. The results in columns (5) and (6) reconstruct the core explanatory variables based on the median division and Sivadasan division [42]. The regression results for both are significantly negative at the 1% significance level, verifying the inhibitory effect of trade liberalization on the intensity of firms’ pollutant emissions.

4.2.3. Placebo Testing

This paper randomly generates the tariff level from 0 to 70% to ensure the validity of the core explanatory variable in the baseline model. The year of occurrence of the trade liberalization policy is also randomly selected between 1998 and 2007. The DID term is constructed with ${Tariff}_r$ and ${Post}_r$, double randomly. Then, 1000 regressions are conducted according to the baseline model (1) setting. Figure 2 illustrates the distribution after 1000 regressions, and

Table 5. Descriptive statistics for regression coefficients.

Variables	Obs.	Mean	S. D.	Min.	Max.
Regression coefficient	1000	6.09 × 10−6	0.0006703	−0.0020433	0.0017311

shows descriptive statistical information on the regression coefficients. The robustness of the effect of trade liberalization on reducing the pollutant emission intensity of firms is verified by a placebo test.

4.2.4. Parallel Trend Test

The unbiased estimation of the results by the DID method necessitates satisfying a precondition wherein the experimental and control groups exhibit comparable trends of change prior to the event. Otherwise, the DID method may overestimate or underestimate the effect of the policy occurrence. This paper examines whether there were significant changes in the pollutant emission intensity of enterprises before and after China’s accession to the WTO due to tariff adjustment. In Figure 3, it can be seen that before China acceded to the WTO, the experimental and control groups’ tariff levels largely overlapped, with a positive intensity of corporate pollutant emissions. However, following China’s accession to the WTO, the tariff level of the experimental group underwent a more substantial decrease than that of the control group, resulting in a substantial downward trend and negative intensity of corporate pollutant emissions. This result demonstrates that the tariff level of the experimental group decreased more substantially than that of the control group after China acceded to the WTO, thereby enhancing the inhibitory effect on the pollutant emission intensity of enterprises compared to the control group and satisfying the parallel trend hypothesis.

4.2.5. Metrics for Sub-Indicators of Enterprise Pollutant Release

To ascertain the heterogeneity of trade liberalization’s impacts on various types of pollutant emissions, this study conducted regression analyses by segregating pollutant types and sub-indicators, incorporating all control variables, and accounting for year, firm, region, and industry fixed effects. In columns (1) to (6) of

Table 6. Metrics for sub-indicators of enterprise pollutant release.

Explaining Variables	$\mathbf{Explained} \mathbf{Variables}: {\mathit{P}\mathit{o}\mathit{l}\mathit{l}\mathit{u}\mathit{t}\mathit{i}\mathit{o}\mathit{n}}_{\mathit{f}\mathit{i}\mathit{t}}$
	(1)	(2)	(3)	(4)	(5)	(6)
	Wastewater	COD	Exhaust	${\mathbf{S}\mathbf{O}}_2$	Soot	Dust
${Tariff}_{i,2001}\times {Post}_{2002}$	−0.008 (0.007)	−0.116 *** (0.006)	−0.005 (0.004)	−0.022 *** (0.007)	−0.049 *** (0.007)	0.086 *** (0.007)
TFP	−0.551 *** (0.005)	−0.296 *** (0.005)	−0.265 *** (0.004)	−0.450 *** (0.005)	−0.352 *** (0.005)	−0.122 *** (0.004)
Size	−0.117 *** (0.007)	−0.092 *** (0.006)	−0.049 *** (0.004)	−0.095 *** (0.007)	−0.089 *** (0.007)	−0.006 (0.006)
Age	0.017 *** (0.004)	−0.000 (0.003)	0.000 (0.002)	0.007 * (0.004)	0.011 *** (0.004)	0.000 (0.004)
KL	−0.014 *** (0.004)	−0.015 *** (0.003)	−0.005 ** (0.002)	−0.021 *** (0.003)	−0.015 *** (0.004)	0.006 * (0.004)
Export	−0.007 (0.023)	0.020 (0.019)	0.006 (0.010)	0.032 * (0.019)	−0.006 (0.018)	−0.021 (0.013)
SOE	0.002 (0.005)	−0.004 (0.005)	0.010 *** (0.003)	0.005 (0.005)	0.013 ** (0.005)	−0.010 ** (0.005)
FIE	−0.018 *** (0.005)	−0.005 (0.005)	0.003 (0.002)	−0.001 (0.005)	−0.005 (0.005)	0.005 * (0.003)
State	0.267 *** (0.052)	0.317 *** (0.051)	−0.127 *** (0.026)	0.069 (0.049)	0.200 *** (0.047)	−0.170 *** (0.032)
Subsidy	0.096 (0.095)	0.087 (0.084)	−0.014 (0.051)	−0.094 (0.079)	−0.055 (0.067)	−0.102 ** (0.050)
Competition	−0.289 ** (0.146)	−0.138 (0.133)	−0.149 ** (0.065)	−0.275 ** (0.126)	−0.193 (0.118)	−0.259 *** (0.072)
Year fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Firm fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Regional fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Industry fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
$R^2$	0.863	0.816	0.832	0.834	0.760	0.877
Observations	174,634	174,634	174,634	174,634	174,634	174,634

The symbols *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. The robust standard deviations are shown in parentheses. The models include DID cross-multiplier terms, single terms, and controls for fixed effects.

, the explained variables are industrial wastewater emissions, COD, exhaust emissions, ${\mathrm{S}\mathrm{O}}_2$ emissions, soot emissions, and dust emissions, respectively. These variables are measured by calculating the ratio of pollutant emissions to the value added by industrial enterprises. In columns (2), (4), and (5), the coefficients of the DID terms for the three pollutant indicators of enterprise COD, ${\mathrm{S}\mathrm{O}}_2$, and industrial soot are significantly negative at the 1% significance level. This result indicates a significant inhibitory effect of trade liberalization on the emission intensity of these three pollutants. However, the estimated coefficients of the DID term for the indicator of industrial dust are significantly positive at the 1% significance level in column (6). This result indicates that trade liberalization increases the intensity of the industrial dust emissions of enterprises. In columns (1) and (3), the estimated coefficients of industrial wastewater and exhaust fumes are insignificant.

4.2.6. Other Robustness Tests

Referring to the study conducted by Lu and Yu [37], this paper presents additional robustness tests. In Model 1, the regression results include the interaction term between the DID term and the number of HS 8-digit coded products under the CIC 4-digit tariff rate, to construct the difference-in-difference-in-difference (DDD) term. The purpose of this is to mitigate the potential absorption of the impact of tariff variations among products within a given industry. In column (1) of

Table 7. Other robustness tests results.

	$\mathbf{Explained} \mathbf{Variables}: {\mathit{P}\mathit{o}\mathit{l}\mathit{l}\mathit{u}\mathit{t}\mathit{i}\mathit{o}\mathit{n}}_{\mathit{f}\mathit{i}\mathit{t}}$
	(1)	(2)	(3)	(4)	(5)
Explaining Variables	Model 1	Model 2	Model 3	Model 4	Model 5
${Tariff}_{i,2001}\times {Post}_{2002}$	−0.040 *** (0.006)	−0.159 *** (0.055)	−0.159 *** (0.057)	−0.159 *** (0.057)
${Tariff}_{i,2001}\times {Post}_{2002}\times PNUM$	−0.000 (0.000)
${\ln Tariff}_i$					0.040 *** (0.006)
TFP	−0.388 *** (0.004)	−0.388 *** (0.007)	−0.392 *** (0.009)	−0.349 *** (0.009)	−0.390 *** (0.004)
Size	−0.091 *** (0.004)	−0.135 *** (0.010)	−0.138 *** (0.013)	−0.063 *** (0.009)	−0.092 *** (0.004)
Age	0.006 *** (0.002)	0.006 (0.007)	0.003 (0.009)	0.014 ** (0.005)	0.006 ** (0.002)
KL	−0.013 *** (0.002)	−0.013 *** (0.004)	−0.014 *** (0.005)	−0.016 *** (0.005)	−0.013 *** (0.002)
Export	0.009 (0.013)	−0.039 (0.032)		−0.009 (0.033)	0.010 (0.013)
SOE	0.003 (0.003)	0.001 (0.011)	0.000 (0.015)	−0.014 (0.009)	0.004 (0.003)
FIE	−0.006 * (0.003)	−0.007 (0.006)	−0.004 (0.007)	−0.004 (0.010)	−0.003 (0.003)
State	0.170 *** (0.033)	0.129 (0.169)	0.097 (0.210)	0.050 (0.049)	0.210 *** (0.033)
Subsidy	0.021 (0.049)	0.302 ** (0.147)	0.421 *** (0.153)	0.133 (0.205)	0.022 (0.050)
Competition	−0.221 *** (0.084)	0.253 (0.236)	0.022 (0.244)	−0.229 (0.187)	−0.238 *** (0.089)
Year fixed effect	Yes	Yes	Yes	Yes	Yes
Firm fixed effect	Yes	Yes	Yes	Yes	Yes
Regional fixed effect	Yes	Yes	Yes	Yes	Yes
Industry fixed effect	Yes	Yes	Yes	Yes	Yes
$R^2$	0.873	0.922	0.921	0.922	0.302
Observations	174,634	47,180	32,364	43,618	171,726

The symbols *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. The robust standard deviations are shown in parentheses. The models include DID cross-multiplier terms, single terms, and controls for fixed effects.

, the estimated coefficient of the DDD term is not statistically significant, thereby suggesting that industries with a greater number of HS 8-digit coded products do not differ significantly from those with a smaller number. To test whether the regression results are affected by foreign markets, the regression model excludes the data of processing trade firms and all exporting firms in Model 2 and Model 3, respectively. In columns (2) and (3), the DID term estimated coefficients are significantly negative at the 1% significance level. In Model 4, a two-period DID is constructed for regression analyses by partitioning the panel data into two halves and averaging the resulting subsets centered around 2002. In column (4), the regression results indicate that following 2002, in conjunction with the progressive deepening of trade liberalization, the pollutant emission intensity of Chinese enterprises underwent a significant reduction.

We used the DID method to test the research question in the baseline estimation. Although the tariff level was substantially reduced in 2001, we found that tariffs continued to be cut as trade liberalization deepened. In order to explore the impact of tariff changes on firms’ pollutant emissions, we now directly use the annual tariff level as the core explanatory variable for estimation. The estimation results in column (5) of Table 7 show that as the tariff level rises, firms’ pollutant emissions will deepen, and a higher tariff level implies lower trade liberalization, which is consistent with the baseline estimation results.

5. Further Analysis

As China is a large processing trade country, its imports of intermediate goods play a vital role. Therefore, we explore whether trade liberalization in intermediate goods will have corresponding environmental effects. This paper adds cross-multiplier terms of the 2001 tariff level data on intermediate goods, ${Input}_{i,2001}$ and ${Post}_{2002}$, for a regression analysis to test the impact of trade liberalization on the pollutant emission intensity of firms. The estimation results are shown in

Table 8. Further estimation results.

Explaining Variables	$\mathbf{Explained} \mathbf{Variables}: {\mathit{P}\mathit{o}\mathit{l}\mathit{l}\mathit{u}\mathit{t}\mathit{i}\mathit{o}\mathit{n}}_{\mathit{f}\mathit{i}\mathit{t}}$
	(1)	(2)	(3)	(4)
${Tariff}_{i,2001}\times {Post}_{2002}$	−0.059 *** (0.007)	−0.057 *** (0.006)	−0.059 *** (0.006)	−0.059 *** (0.006)
${Input}_{i,2001}\times {Post}_{2002}$	0.008 *** (0.001)	0.004 *** (0.001)	0.005 *** (0.001)	0.005 *** (0.001)
TFP		−0.387 *** (0.004)	−0.386 *** (0.004)	−0.387 *** (0.004)
Size		−0.089 *** (0.004)	−0.089 *** (0.004)	−0.090 *** (0.004)
Age		0.007 *** (0.002)	0.006 *** (0.002)	0.006 *** (0.002)
KL		−0.013 *** (0.002)	−0.013 *** (0.002)	−0.013 *** (0.002)
Export		0.009 (0.013)	0.009 (0.013)	0.009 (0.013)
SOE			0.003 (0.003)	0.003 (0.003)
FIE			−0.006 ** (0.003)	−0.006 ** (0.003)
State			0.092 *** (0.033)	0.090 *** (0.033)
Subsidy			0.017 (0.049)	0.017 (0.049)
Competition				−0.222 *** (0.086)
Year fixed effect	Yes	Yes	Yes	Yes
Firm fixed effect	Yes	Yes	Yes	Yes
Regional fixed effect	Yes	Yes	Yes	Yes
Industry fixed effect	Yes	Yes	Yes	Yes
$R^2$	0.816	0.871	0.871	0.871
Observations	172,234	172,234	172,234	172,234

The symbols *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. The robust standard deviations are shown in parentheses. The models include DID cross-multiplier terms, single terms, and controls for fixed effects.

, in which year, firm, regional, and industry fixed effects are controlled in each column. Subsequently, additional control variables (firm-level and industry-level) are introduced (shown in columns (2)–(4)) to the regression analysis. Notably, the estimated coefficients on the core explanatory variables are consistently significantly negative at the 1% significance level. Trade liberalization in final goods significantly diminishes the intensity of corporate pollutant emissions. However, trade liberalization in intermediate goods has a substantial positive effect on the intensity of corporate pollutant emissions, indicating that it increases their intensity.

When we consider the liberalization of trade in intermediate goods, on the one hand, the lower cost of intermediate goods brought about by the liberalization of their trade allows firms to scale up production on a broad scale, which leads to an increase in pollutant emissions. On the other hand, the liberalization of trade in intermediates may lead to a lengthening and complication of the production chain, particularly leading to an increase in pollutant emissions as domestic firms take on production stages that foreign firms are trying to transfer, which may involve highly polluting activities. In addition, regulators’ difficulty in effectively controlling polluting emissions is exacerbated by more diversified sources of intermediate goods. For final goods, trade liberalization lowers market entry barriers and increases competition between international and domestic firms. In order to remain competitive in the international market, firms must innovate technologically, which leads them to adopt production technologies that are more efficient and cleaner. At the same time, trade liberalization in final goods may lead to an international convergence of environmental standards, requiring firms to produce goods that meet stricter environmental standards in order to enter the market. Overall, the estimated coefficients for trade liberalization in intermediate goods are much smaller than those for trade liberalization in final goods, suggesting that the increase in firms’ pollution intensity due to tariff reductions in intermediate goods is offset by the suppression of firms’ pollution intensity due to tariff reductions in final goods.

Ownership in China has often been the focus of existing research, especially as SOEs play a vital role in China’s markets. Accordingly, China’s diverse ownership types are likely to differ significantly in terms of productivity, strategy, and performance and to respond differently to trade liberalization. Moreover, exporting and non-exporting firms face different market environments. The former have direct access to international markets, meaning they are more susceptible to the impacts of relevant trade policies, while non-exporting firms mainly serve the domestic market, and the impact of trade policies is relatively indirect. As a result, there may also be differences in the performances of these two types of enterprises under trade liberalization policies. This paper introduces a DDD term to the regression model, to test the impact of trade liberalization on firms’ pollutant emission intensity and the heterogeneity of firms’ ownership and export. The model is constructed as follows:

$$\begin{array}{cc}\hfill {Pollution}_{fit}=\alpha & +{\beta }_1{Tariff}_{i,2001}\times {Post}_{2002}+{\beta }_2{Tariff}_{i,2001}\times {Post}_{2002}\times SOEs+{\beta }_3{Tariff}_{i,2001}\times {Post}_{2002}\times FIE\hfill \\ & +{\gamma }_1{X}_{ft}+{\gamma }_2{Z}_{it}+{\epsilon }_{it}\hfill \end{array}$$

(4)

$${Pollution}_{fit}=\alpha +{\beta }_1{Tariff}_{i,2001}\times {Post}_{2002}+{\beta }_2{Tariff}_{i,2001}\times {Post}_{2002}\times Export+{\gamma }_1{X}_{ft}+{\gamma }_2{Z}_{it}+{\epsilon }_{it}$$

(5)

Table 9. Results of DDD estimation.

Explaining Variables	$\mathbf{Explained} \mathbf{Variables}: {\mathit{P}\mathit{o}\mathit{l}\mathit{l}\mathit{u}\mathit{t}\mathit{i}\mathit{o}\mathit{n}}_{\mathit{f}\mathit{i}\mathit{t}}$
	(1)			(2)
	Model 1			Model 2
${Tariff}_{i,2001}\times {Post}_{2002}$	−0.040 *** (0.005)	−0.047 *** (0.004)	−0.047 *** (0.004)	−0.042 *** (0.005)	−0.052 *** (0.004)	−0.050 *** (0.004)
${Tariff}_{i,2001}\times {Post}_{2002}\times SOEs$	0.009 *** (0.002)	0.005 *** (0.002)	0.007 *** (0.002)
${Tariff}_{i,2001}\times {Post}_{2002}\times FIEs$	−0.003 * (0.002)	−0.000 (0.001)	0.002 (0.002)
${Tariff}_{i,2001}\times {Post}_{2002}\times Export$				0.005 *** (0.002)	0.011 *** (0.002)	0.010 *** (0.002)
TFP		−0.388 *** (0.004)	−0.388 *** (0.004)		−0.388 *** (0.004)	−0.388 *** (0.004)
Size		−0.090 *** (0.004)	−0.091 *** (0.004)		−0.092 *** (0.004)	−0.093 *** (0.004)
Age		0.007 *** (0.002)	0.006 *** (0.002)		0.007 *** (0.002)	0.006 *** (0.002)
KL		−0.013 *** (0.002)	−0.013 *** (0.002)		−0.014 *** (0.002)	−0.013 *** (0.002)
Export		0.009 (0.013)	0.009 (0.013)		−0.016 (0.013)	−0.013 (0.013)
SOE			−0.004 (0.004)			0.003 (0.003)
FIE			−0.009 * (0.005)			−0.006 * (0.003)
State			0.173 *** (0.033)			0.146 *** (0.033)
Subsidy			0.022 (0.049)			0.024 (0.049)
Competition			−0.217 ** (0.084)			−0.201 ** (0.084)
Year fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Firm fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Regional fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Industry fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
$R^2$	0.818	0.873	0.873	0.818	0.873	0.873
Observations	174,634	174,634	174,634	174,634	174,634	174,634

The symbols *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. The robust standard deviations are shown in parentheses. The models include DID cross-multiplier terms, single terms, and controls for fixed effects.

displays the regression results, indicating that the estimated coefficients of the core explanatory variables are significantly negative at the 1% significance level. This paper introduces a dummy variable in Model 1 to account for the nature of enterprise ownership, specifically whether it is SOE or not (SOE) and whether it is FIE or not (FIE). The variable is cross-multiplied and regressed with the DID term. This study finds that the impact of trade liberalization on the intensity of enterprise pollutant emissions varies depending on the type of enterprise ownership. The DDD term’s estimated coefficients for SOEs are significantly positive, while those for FIEs are insignificant. This result suggests that trade liberalization exacerbates the pollutant emission intensity of SOEs and has no significant effect on the pollutant emission intensity of FIEs. Possible reasons for this are, on the one hand, at the beginning of China’s accession to the WTO, the higher technological barriers for SOEs, the production efficiency of FIEs, and the standards FIEs already adhere to in the international market. On the other hand, SOEs are relatively less productive, and need to mature their production technology; nonetheless, these factors and their political affiliations may mean they face relatively lower regulation, which introduces the possibility that their polluting emissions may go unnoticed.

In Model 2, the paper uses a dummy variable, denoting whether a firm engages in export activities or not (Export), cross-multiplied with the DID term to construct the DDD term. The estimated coefficients of the DDD term are significantly positive at the 1% significance level. This result indicates that as tariffs fall, exporting firms increase the intensity of firms’ pollutant emissions, contributing to environmental pollution. The possible reasons for this are that exporting enterprises tend to have larger production scales than non-exporting enterprises, allowing them to respond to the demands of the international market, and such large-scale production activities may lead to more polluting emissions. At the same time, exporters face more substantial competitive pressures in the international market and pay more attention to their cost control and production efficiency. Since the adoption of environmentally friendly technologies will increase the production costs of exporters and thus generate competitive pressures, this may lead exporters to ignore environmental protection and thus increase their pollutant emissions in the production process.

6. Conclusions and Implications

Utilizing data from the ASIF and the AESIF from 1998 to 2007, this paper has empirically analyzed the impacts of trade liberalization on the emission intensities of major pollutants among Chinese enterprises. By identifying the changes in tariff levels of various industries after China acceded to the WTO, and by adopting the DID method, this study has found that (1) after China acceded to the WTO, the technological spillover effect and competition effect contributed to a decline in the intensity of corporate pollutant emissions. The estimation results remained robust through a series of robustness tests. Additionally, there was a certain degree of heterogeneity in the mitigating effects of trade liberalization on enterprises’ emissions of different pollutants. Specifically, there were significant inhibitory effects on COD, ${\mathrm{S}\mathrm{O}}_2$, and the soot intensity, whereas trade liberalization aggravated the intensity of industrial dust emissions. (2) The estimated coefficients of trade liberalization in intermediate goods were significantly positive after introducing a DID term for the tariff level. On the one hand, tariff reductions and the ensuing competition effectively motivated domestic enterprises to enhance their production technology and adopt ‘clean’ production practices. On the other hand, due to tariff reductions and price declines, domestic firms involved in midstream and upstream production imported ‘clean’ intermediate goods from abroad while retaining ‘polluting’ production domestically. This represents a negative externality stemming from the division of labor effect. (3) This study employed the DDD method to determine the influence of enterprise ownership and exporting on the enterprise pollutant emission intensity. This approach revealed that the liberalization of trade in final goods intensified the pollutant emission intensity of SOEs. Meanwhile, it did not significantly affect the pollutant emission intensity of non-SOEs. Additionally, exporting enterprises tended to increase their pollutant emission intensity. These findings suggest that private and non-exporting enterprises play a prominent role in energy conservation and emission reduction.

The findings of this study are in contrast to the PHH since they indicate that trade liberalization led to a significant reduction in the pollutant emission intensity of Chinese firms. Moreover, while processing trade in China contributed a certain level of environmental pollution, the technological spillover and competitive effects resulting from trade liberalization substantially enhanced the production technology of Chinese firms, thereby significantly reducing the environmental pressure caused by production activities. In general, the findings reveal that developing countries may undergo a process of trade liberalization in tandem with addressing excessive corporate pollutant emissions. We wish to highlight three key takeaways: (1) In China, trade liberalization has significantly reduced the intensity of corporate pollutant emissions. This leads us to propose that the Chinese authorities should continue safeguarding China’s trade relations with developed countries, as well as promoting new trade relations between China and developing countries, and encouraging trade liberalization. These efforts may achieve market expansion, technological innovation, and industrial transfer, ultimately leading to improvements in environmental quality. (2) Reducing tariffs on intermediate products is likely to increase the pollutant emission intensity of firms. As the world’s factory, China undertakes a large amount of processing trade, and the Chinese government should pay attention to the potential pollution problems that the processing link of intermediate products’ trade may bring about. Moreover, the Chinese government should set up environmental regulations for the production and processing of intermediate products. (3) Through our analyses, we found that SOEs increase pollutant emissions more so compared to FIEs. Therefore, the Chinese government should increase innovation and investment in SOEs, to improve their productivity. At the same time, it is necessary to change the production concepts of SOE managers, which will encourage green and high-quality development and reduce environmental pollution.