Going Green: Evidence from Product-Level Exports and Firms’ Environmental Performance in China

Abstract

The environmental impact of exporting is a topic of significant concern to both scholars and policymakers. Analyzing how exporting influences firms’ environmental performance concerning the product level is vital for better understanding its implications. Drawing on the data obtained from Chinese industrial firms, this study employs the PSM-DID (Propensity Score Matching-Difference in differences) model to establish a causal relationship between exporting intermediate goods or final goods and firms’ emissions. The findings indicate that, on the one hand, exporting intermediate goods leads to a rise in firms’ production scale and total factor productivity, ultimately significantly increasing total pollution emissions and reducing pollution emission intensity. On the other hand, exporting final goods decreases total pollution emissions and emission intensity through the technical effect. In general, this study provides important micro-level evidence of how exports affect pollution emissions at the product level in China. This evidence is valuable for policymakers in shaping environmental regulations and export policies.

1. Introduction

As the world’s leading exporter and manufacturing hub, China occupies a pivotal role in the global supply chain of resources and environmental elements. It exports a substantial volume of both intermediate and final products, meticulously tailored to satisfy the voracious appetites of international markets. This export prowess, however, comes at a cost as China grapples with heightened pollution emissions, particularly air quality degradation, due to its rapid export expansion [1]. Environmental pollution carries adverse implications for public health and societal well-being [2,3]. Given the present scenario of intensifying global competition and heightened environmental regulations, firms focused on exports are compelled to mitigate pollution to secure environmental certifications from developed nations. However, many Chinese firms have long benefited from export trade, maintaining a high degree of export dependency and thus, are unwilling to forsake the benefits gained from exports for the sake of environmental preservation. It is worth noting that economic growth and environmental protection need not be a “zero-sum game”. By exploring greener export products and pathways, we can enhance firms’ environmental performance. Therefore, from a product-oriented perspective, examining the impact of exporting intermediate and final goods on firms’ pollution emissions holds significant relevance.

The trade-environment correlation has always been a crucial area of interest. Earlier literature on trade and environment mostly uses industry and country-level data and suggests that the impact of trade can be categorized into technical effect, factor composition effect, and scale effect [4,5]. However, recent research has shifted towards analyzing the environmental consequences of trade at the micro-level, yielding inconclusive results. Some studies suggest that exporting firms demonstrate superior environmental performance [6,7], while others argue that trade liberalization exacerbates pollution [8]. We can observe that current research by scholars directly treats the total export value or export quantity of all exported products as a proxy for a firm’s export behavior, without taking into consideration the distinct characteristics of different product exporters. This may be one of the reasons why scholars’ research findings sometimes appear contradictory. Intermediate goods trade, as a key indicator of global value chains, predominantly focuses on labor-intensive and pollution-intensive industries [9]. Most firms in labor-intensive and pollution-intensive sectors are large in scale, leading to relatively higher total emissions of pollutants. On the other hand, final goods which are meant for end users are subject to swift swings in demand. As a result, exporters of final goods are less likely to change their manufacturing scale at random. Additionally, the majority of firms that produce final goods are capital-intensive and boast relatively greater levels of productivity. They may, therefore, have lower overall pollutant emissions and emission intensity [10]. An intriguing issue arises in this context: which exported product is more environmentally friendly? Addressing this puzzle, this paper explores the impact of exporting intermediate goods or final goods on firms’ pollution emissions at the product level.

However, to date, there has been limited empirical evidence regarding the influence of trade on the environment at the product level, particularly concerning the ramifications of exports. Most scholars have primarily investigated the effects of trade liberalization on firm environmental performance through the prism of intermediate or final goods tariffs [10,11]. A closely relevant study to this research is that of He and Huang [12], which explored the environmental consequences of importing intermediate and capital goods, adopting a PSM-DID (Propensity Score Matching-Difference in differences) approach from an import-oriented perspective. Their findings revealed that the importation of intermediate and capital goods results in a reduction of firms’ pollution emission intensity but concurrently amplifies their overall pollution emissions. However, this study exclusively examined the import perspective and did not comprehensively elucidate the influence of trade in final goods on the environment. Therefore, our study aims to investigate the influence of exporting intermediate goods or final products on two key aspects of firm environmental performance: total pollution emissions and emission intensity. Furthermore, we seek to elucidate the mechanisms underlying these impacts.

The significance of this study can be elaborated from both theoretical and practical perspectives. In theoretical terms, to further elucidate the impact of export trade on firms’ pollution emissions, we examine their relationship from a product-oriented standpoint. In practical terms, we aim to identify greener export products that can promote eco-friendly trade practices. Therefore, we have defined our research direction as the analysis of the influence of export trade on firms’ pollution emissions from a product-level perspective, specifically exploring the effects of exporting intermediate goods and final goods on firms’ pollution emissions.

To provide a more comprehensive understanding, we address the following research questions:

(1)

What is the effect of exporting intermediate goods or final products on a firm’s total pollution emissions and emission intensity?

(2)

What are the driving mechanisms behind these effects?

(3)

Do the relationships between exports and environmental performance vary among firms with distinct characteristics and pollution intensity levels?

To answer the questions posed above, we utilize data from Chinese manufacturing firms to investigate the influence of exporting intermediate and final goods on firms’ pollution emissions. Specifically, this study combines data from the China Industrial Firm Database, China Environmental Statistics Database, and China Customs Database. This comprehensive process results in the construction of a micro-level dataset spanning the years 2000 to 2013, which comprises detailed insights into firms’ production, trade activities, and pollution emissions. Sulfur dioxide (${SO}_2$) emissions are selected as the key pollution emission indicator. In practice, it is essential to acknowledge that firm exports do not occur randomly. To effectively mitigate the challenges posed by sample selection bias and endogeneity, this study employs a dual-pronged approach, incorporating Propensity Score Matching (PSM) and Difference-in-Differences (DID) methods, to disentangle the causal relationship between firm exports and firm-level pollution emission intensity. This analysis reveals that exporting intermediate and final goods has various environmental effects. Specifically, exporting intermediate goods amplifies the total amount of pollution emissions but lowers its intensity, whereas exporting final goods reduces both the total pollution emissions and emission intensity. Our research highlights the significance of recognizing the differential effects of various exported products on pollution emissions when assessing the environmental impact of firms’ exports. It is noteworthy that, in terms of total firms’ total pollution emissions, the export of intermediate and final goods has an inverse relationship. Thus, if one solely analyzes China’s environmental impact of exports from an overall perspective and disregards the heterogeneity of exported products, the environmental performance of Chinese firms’ exports may be biased.

Building on our previous analysis, we delve deeper into the potential mechanisms underlying the link between intermediate and final goods exports and firms’ pollution emissions. The findings indicate that two key mechanisms are at work: the scale effect and the technical effect. On the one hand, while firms that export final goods do not expand their production scale, those exporting intermediate goods do so to keep up with the expanding market demand. On the other hand, both intermediate and final goods exporting firms enhance their total factor productivity, which lowers the pollution emission intensity. The increased productivity also benefits firms that export final goods by lowering their environmental emissions. Recognizing that firms possess diverse characteristics, we further develop sub-samples based on various factor intensities and pollution intensities to reassess our results.

Our research stands out from previous studies by providing notable contributions in two primary areas. Firstly, from a research perspective, the prevailing literature on trade and the environment primarily revolves around analyzing the impact of trade liberalization, specifically in terms of tariffs imposed on intermediate and final goods. In contrast, this study takes a product-level approach to distinguish between exports of intermediate and final goods and thoroughly examines their causal relationship with firms’ pollution emissions, which contributes fresh insights into the connection between exports and the environment. Secondly, in terms of research results, our study shows that exporting intermediate goods and final goods has an opposing effect on a firm’s total pollution emissions, but decreases pollution intensity. This implies that the diversity of exported goods may help to clarify the ambiguous environmental impact of exports, offering insightful information into the complex link between exporting activities and environmental issues.

This paper adopts a structured framework. Section 2 presents a thorough analysis of the research-related literature. Section 3 outlines our empirical model and data. The findings derived from our empirical analysis are presented in Section 4. Lastly, Section 5 summarizes the study’s main insights and contributions, concluding the research.

2. Literature Review

Our work aligns closely with the current scholarly literature in three significant aspects. Firstly, there is a growing interest in analyzing the connection between firm performance and its export activities. Firm heterogeneity theory, as proposed by Melitz [13], has made the performance differential between exporting and non-exporting firms a central proposition of firm-level international trade research. Empirical studies have confirmed that firms engaged in export activities exhibit superior productivity and scale in comparison to their non-exporting counterparts [6,14]. Various theories, such as self-selection and export learning effects, have also emerged [15,16]. While numerous studies have extensively explored the effects of exporting on dimensions such as innovation, social welfare, and related factors [17,18], the impact of exporting on firms’ environmental performance has received limited attention, with inconclusive findings [8,19]. Therefore, this study aims to enrich the current scholarly literature by examining the influence of exporting on firm-level environmental performance.

Second, the literature on the environmental implications of international trade offers three viewpoints with no clear consensus. The first viewpoint suggests that trade can benefit the environment. This viewpoint is supported by Antweiler et al. [20], who find that trade liberalization leads to higher GDP and lower pollution concentrations. The second viewpoint, illustrated by Chichilnisk’s [21] North-South trade model and studies by Guan et al. [22] and He et al. [8], argues that trade exacerbates resource distribution inequalities and environmental deterioration. The third viewpoint suggests that the connection between trade and the environment is intricate and unclear. Scholars such as Grossman and Krueger [23] and Cole and Elliott [24] argue that the influence of international trade on environmental pollution cannot be definitively judged as either positive or negative.

The potential impact mechanisms have also been examined in previous research. Grossman and Krueger [23] introduce the concept of three pathways through which trade affects pollution emissions: the scale effect, factor composition effect, and technical effect, which have significantly influenced subsequent research in this field. Several scholars, including Copeland and Taylor [4] and He et al. [8], have expanded on this theory. Nonetheless, current studies mainly focus on developed countries, and microscopic data limitations have hindered research on developing countries. In addition, although recent studies analyze pollution emissions at the firm level involved in international trade, they overlook the heterogeneity of export products, and further analytical studies are necessary.

Third, the scholarly literature has analyzed the economic consequences of intermediate goods trade and final goods trade. The concept of intermediate goods, introduced by Sanyal and Jones [25], refers to products that require further processing and transformation before reaching the final consumers, while final goods are products that do not require further processing and are directly used for consumption or investment purposes. Studies have adopted diverse perspectives on this topic. Some scholars conduct studies from the perspective of tariffs, analyzing the effects of tariffs on intermediate goods and final goods. For instance, Cherniwchan [11] measures trade liberalization in terms of tariffs on intermediate and final goods and analyzes its impact on firms’ pollution emissions, while Caliendo and Parro [26] analyze the impact of exogenous tariff changes on welfare changes after the establishment of the North American Free Trade Area. Additionally, Xi et al. [27] and Zhou et al. [28] investigate the relationship between intermediate and final goods trade and labor force employment. Despite the considerable literature in this area, few studies have explored the influence of trade involving intermediate and final goods on firms’ pollution emissions, highlighting a research gap that requires further exploration and analysis.

In conclusion, synthesizing the existing body of research reveals key insights:

Firstly, in the literature examining the link between export behavior and firm performance, micro-level investigations offer a comprehensive understanding of how firms’ export activities influence production performance. This body of work convincingly demonstrates that exporting significantly impacts various facets of firm performance, including productivity and production scale. While this research has established a robust foundation for comprehending the nexus between export behavior and firm performance, it primarily focuses on the influence of export behavior on production. However, there has been relatively less attention directed toward elucidating the implications of export behavior for firms’ environmental performance. Our study aims to delve into the intricate relationship between export behavior and environmental performance, augmenting the literature that predominantly centers on exports and firm performance.

Secondly, in the literature addressing the trade-environment interplay, research on the impact of international trade on pollution emissions exhibits diversity from a macroeconomic perspective. Notably, a substantial portion of this research primarily focuses on developed countries, with limited attention to developing nations due to data constraints. Given the environmental challenges in developing countries, it is imperative to scrutinize how firms’ export behavior intersects with their environmental performance, leveraging data from these regions. Furthermore, recent scholarship probing pollution emissions characteristics of firms engaged in international trade often overlooks the heterogeneity of export products, necessitating further empirical and analytical exploration. Our study strategically addresses these gaps by focusing on China, the largest developing country, and examining the trade-environment nexus from a micro-level perspective, considering distinctive attributes of various export products.

Thirdly, in the literature examining the environmental impact of intermediate and final goods trade, the predominant focus revolves around tariff-related aspects, with limited direct analysis of different traded product categories. Therefore, this paper seeks to directly distinguish between exporters of intermediate and final goods, enabling an in-depth exploration of their respective impacts on firms’ pollution emissions.

3. Empirical Model and Data

3.1. Empirical Model

A potential reverse causal relationship between exports and firms’ pollution emissions may exist, implying the presence of a self-selection effect in firms’ export behavior [29,30]. Firms characterized by high pollution emission efficiency may opt to export their products to foreign markets owing to their inherent competitive advantages. Consequently, when endeavoring to discern the causal impact of firms’ export behavior on their pollution emissions, it becomes imperative to effectively confront the endogeneity challenge inherent in the relationship between exports and firms’ pollution emissions. In this context, this study draws inspiration from the methodologies articulated by De Loecker [31], Heckman et al. [32], and He and Huang [12]. It combines the Propensity Score Matching (PSM) and Difference-in-Differences (DID) methods to scrutinize whether firms’ exportation of intermediate or final goods precipitates alterations in their environmental performance. This integrated analytical framework serves to enhance the management and control of sample selection bias and endogeneity issues that do not arise randomly from firms’ export behavior. Consequently, it effectively disentangles the causal impact of firms’ exports on their pollution emissions, yielding findings of heightened credibility.

Specifically, our research design identifies a treatment group of firms exporting intermediate or final goods and a carefully constructed control group of non-exporting firms, addressing selection biases through propensity score matching. This approach allows us to attribute any differences in pollution emission behavior to exporting. Following the differentiated outcome variables approach within the PSM-DID method as outlined by Heckman et al. [32] and more recently by He and Huang [12], we establish a causal link between exporting and pollution emissions while controlling for common trends. Our treatment variable is defined as whether a firm exports at a given period (t), while the outcome variable captures changes in pollution emissions or pollution emission intensity.

This study employs the Logit Model to initially estimate the propensity score. The probability that firm i begins exporting intermediate or final goods in period t is expressed by Equations (1) and (2), respectively:

$$P^{Inter}\left(X_{it-1}\right)=\mathrm{Pr}\left({Interexport}_{it}=1\right)=\mathsf{\Phi }\left\{h\left(X_{it-1}\right)\right\}$$

(1)

$$P^{Final}\left(X_{it-1}\right)=\mathrm{Pr}\left({Finalexport}_{it}=1\right)=\mathsf{\Phi }\left\{h\left(X_{it-1}\right)\right\}$$

(2)

$$X_{it-1}=\left\{{lnSO}_{2it-1},{lneSO}_{2it-1},{Profit}_{it-1},{Finance}_{it-1}{, lnAge}_{it-1}\right\}$$

(3)

where $\mathsf{\Phi }\left(.\right)$ denotes the normal cumulative distribution function. ${Interexport}_{it}$ and ${Finalexport}_{it}$ are dummy variables for the export of intermediate and final goods, respectively. These variables take a value of 1 when a firm initiates the exportation of intermediate or final goods at period t and 0 otherwise. $X_{it-1}$, on the other hand, represents a set of probabilities used to estimate the probabilities of a firm exporting intermediate or final goods. To address the issue of dissimilar traits between the treated and control groups, we used one-period lagged variables as covariates, which is illustrated by Equation (3). The covariates include firm-level production characteristics such as profitability (${Profit}_{it-1}$), financing constraints (${Finance}_{it-1}$), and firm age (${ lnAge}_{it-1}$), as well as emission characteristics, including total emissions (${lnSO}_{2it-1}$) and emission intensity (${lneSO}_{2it-1}$).

Then we employ Average Treatment Effects to the Treated (ATT) to estimate the effects of exporting intermediate goods or final goods on firms’ pollution emissions $Z_{it}$ as follows:

$${\tau }_{t+1}^{Inter}=E\left[Z_{it+1}-Z_{it-1}|{Interexport}_{it}=1,P^{Inter}\left(X\right)\right]-E\left[Z_{kt+1}-Z_{kt-1}|{Interexport}_{kt}=0,P^{Inter}\left(X\right)\right]$$

(4)

$${\tau }_{t+1}^{Final}=E\left[Z_{it+1}-Z_{it-1}|{Finalexport}_{it}=1,P^{Final}\left(X\right)\right]-E\left[Z_{kt+1}-Z_{kt-1}|{Finalexport}_{kt}=0,P^{Final}\left(X\right)\right]$$

(5)

By comparing the ATT for exporting intermediate goods (denoted as ${\tau }_{t+1}^{Inter}$) and exporting final goods (denoted as ${\tau }_{t+1}^{Final}$), we establish a causal relationship between both types of exports and firms’ pollution emissions. Equations (4) and (5) can be split into two parts. The first part examines the variation in pollution emissions within the treated group by comparing pollution emissions one year after exporting intermediate or final goods to the previous year. The second part compares these differences to emissions in the control group during the same period. The comparison between the year after exporting and the previous year accounts for the time necessary to adjust production after exporting.

To estimate the ATT, we adopt a nearest neighbor matching method with a sample size of n = 4 as Equation (6):

$$ATT=\frac{1}{N}{\sum }_{i\in N}[\left(Z_{it+1}-Z_{it-1}\right)-{\sum }_{j\in C\left(i\right)}w(i,j)(Z_{jt+1}-Z_{jt-1})]$$

(6)

We utilize the PSM method to construct a control group (denoted as $C\left(i\right)$) that closely matches each firm i in the treated group, to determine the number of matching firms (denoted as $N$ initiating export and calculate the corresponding weight (denoted as $w(i,j$)). Equation (6) can be divided into two components. The first component captures the alteration in pollution emissions for firms in the treated group after one year of engaging in exports. The second component represents the corresponding change for the control group. By comparing these changes between the treated and control groups, we can ascertain the Average Treatment Effects (ATE).

3.2. Varibles

This study uses information from three databases: the China Customs Database, the China Industrial Firm Database, and the China Environmental Statistics Database for the sample period of 2000 to 2013.

3.2.1. Outcome Variables

First, we choose sulfur dioxide as the measure of pollution emissions because it remains a global environmental concern for humanity [33]. Furthermore, China, as the world’s largest producer and consumer of coal, grapples predominantly with atmospheric pollution driven by coal combustion, which constitutes a primary facet of environmental pollution in the country [34]. Notably, sulfur dioxide stands as the principal atmospheric pollutant emanating from coal combustion in China. Consequently, this study employs total sulfur dioxide emissions and emission intensity as metrics to evaluate firm pollution behavior.

Next, we utilize both total sulfur dioxide pollution emissions and emission intensity as metrics to evaluate firms’ environmental performance. The selection is primarily driven by the common approach taken by nations to achieve their environmental quality objectives, which involves regulating both the overall volume of pollution and the intensity of emissions. The utilization of these two indicators aligns closely with practical considerations. To begin with, total pollution emissions represent the aggregate quantity of pollutants released by a firm. In this context, our primary focus is on measuring total sulfur dioxide emissions from these firms. Secondly, pollution emission intensity pertains to the pollution emission levels relative to the unit value of output, specifically calculated as the ratio of sulfur dioxide emissions to the total output value of the firm. This metric effectively normalizes the influence of firm size and provides a more precise assessment of a firm’s pollution emission levels. Thus, this study adopts total sulfur dioxide emission and emission intensity indicators to assess firms’ pollution behavior. The different values of these indicators are employed as outcome variables.

3.2.2. Treatment Variables

The treatment variables are the intermediate goods export dummy variable (${Interexport}_{ijst}$) and the final export dummy variable (${Finalexport}_{ijst}$), which indicate whether firm i initiates export activities of intermediate goods or final goods in period t. These binary treatment variables take on a value of 1 if a firm initiates exporting the corresponding type of goods and 0 otherwise.

3.2.3. Covariates

To address potential self-selection bias and improve the matching effect, we select the firm’s production and emission characteristics lagging one period as covariates. The production variables encompass: Profitability (${Profit}_{ijst-1}$), quantified as the difference between operating income and operating costs divided by operating income; capital indebtedness (${Finance}_{ijst-1}$), computed as the ratio of total liabilities to total assets; The logarithmic form of firms’ age (${ lnAge}_{ijst-1}$), calculated as the natural logarithm of the difference between the observation year and the year of establishment. The pollution emissions variables consist of: Total sulfur dioxide emission (${lnSO}_{2ijst-1}$), derived by taking the logarithm of the firm’s total sulfur dioxide pollution emissions; Sulfur dioxide emission intensity (${lneSO}_{2ijst-1}$), ascertained by taking the logarithm of the ratio of the firm’s sulfur dioxide emissions to the actual total output of the firm (adjusted for the industrial goods ex-factory price index).

3.2.4. Data

This study relies on three databases for empirical analysis. The first dataset utilized is the China Industrial Firm Database, covering the period from 2000 to 2013. We utilize the approach Brandt et al. [35] suggested for processing this database. The second dataset employed is the China Environmental Statistics Database, widely acknowledged as an authoritative source in this domain [36]. Finally, we utilize the China Customs Database, which records trading transactions and HS code information. We convert HS codes to Broad Economic Categories (BEC) and classify goods into three distinct categories—intermediate goods, capital goods, and consumption goods—enabling us to further categorize capital goods and consumption goods as final goods. Finally, by aggregating the export data of intermediate and final goods per firm within each year, we obtain information on firms’ different export product categories.

This paper adopts the matching methodology proposed by Yu [37] to combine the China Customs Database and the China Industrial Firm Database. The matching process involves two steps. First, according to the firm names and years for initial matching, the same firm is first identified in the two databases. Second, we further verify the first seven digits of the firm’s telephone number and postal code to ensure accuracy. In addition, we implement an outlier removal process to generate a reliable sample consisting of firms’ basic information and trade data.

To obtain our final dataset, we match the new dataset we have constructed with the China Environmental Statistics Database. The matching process involves a two-step approach. First, we match the two databases using the unique firm code for each year. Second, we proceed with the matching process by using the firm names and finally obtain an unbalanced panel database.

Descriptive statistics for major firm characteristics are presented in

Table 1. Summary statistics.

Variable	All Samples	Intermediate Goods Exporter	Final Goods Exporter	Non-Exporter
	(1)	(2)	(3)	(4)
${Interexport}_{jist}$	0.153		0.309
	(0.360)		(0.462)
${Finalexport}_{jist}$	0.114	0.231
	(0.318)	(0.422)
${Profit}_{jist}$	0.178	0.172	0.187	0.179
	(0.143)	(0.131)	(0.143)	(0.145)
${Finance}_{jist}$	0.558	0.55	0.529	0.563
	(0.240)	(0.224)	(0.231)	(0.244)
${lnAge}_{jist}$	2.413	2.473	2.507	2.394
	(0.767)	(0.678)	(0.669)	(0.792)
${lnSO}_{2jist}$	10.055	10.046	9.401	10.130
	(2.108)	(2.281)	(2.169)	(2.068)

Notes: Table 1 summarizes variable means and standard deviations. Columns (1), (2), and (3) present means and standard deviations for full, intermediate, and final goods exporting firms. Column (4) shows non-exporting firms’ means and standard deviations.

. The mean values of the dummy variables for firms exporting intermediate goods and final goods are 15.3% and 11.4%, respectively, indicating a greater prevalence of firms exporting intermediate goods in our sample. Notably, the logarithmic mean value of sulfur dioxide emissions is 10.055, with a standard deviation of 2.108, highlighting significant differences in pollution emission behavior among firms in the sample. Noteworthy is the finding that the logarithmic mean of sulfur dioxide emissions for firms exporting intermediate goods is marginally higher at 10.046 compared to firms exporting final goods, which records a mean of 9.401. This subtle distinction implies that firms exporting final goods exhibit slightly superior environmental performance in comparison to their counterparts engaged in the export of intermediate goods, with non-exporting firms positioned in between. Furthermore, Table 1 provides essential information on firms’ profitability, capital indebtedness, and age.

4. Empirical Results

4.1. Logit Regression and Propensity Score

To ensure reliable estimation, it is imperative to initially estimate the propensity scores.

Table 2. Logit regression of exporting.

	${\mathit{I}\mathit{n}\mathit{t}\mathit{e}\mathit{r}\mathit{e}\mathit{x}\mathit{p}\mathit{o}\mathit{r}\mathit{t}}_{\mathit{j}\mathit{i}\mathit{s}\mathit{t}}$		${\mathit{F}\mathit{i}\mathit{n}\mathit{a}\mathit{l}\mathit{e}\mathit{x}\mathit{p}\mathit{o}\mathit{r}\mathit{t}}_{\mathit{j}\mathit{i}\mathit{s}\mathit{t}}$
	(1)	(2)	(3)	(4)
${lnSO}_{2jist}$	0.1092 *** (0.0041)		−0.0275 *** (0.0047)
${lneSO}_{2jist}$		−0.1466 *** (0.0039)		−0.2603 *** (0.0045)
${Profit}_{jist-1}$	0.3398 ***	0.1983 ***	1.0834 ***	0.9792 ***
	(0.0669)	(0.0671)	(0.0731)	(0.0748)
${Finance}_{jist-1}$	−0.3291 ***	−0.2097 ***	−0.4182 ***	−0.3157 ***
	(0.0329)	(0.0330)	(0.0370)	(0.0380)
${lnAge}_{jist-1}$	0.0585 ***	0.0833 ***	0.1411 ***	0.1399 ***
	(0.0094)	(0.0094)	(0.0108)	(0.0111)
Year FE	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
$LR {\chi }^2$	19,350.42	20,004.59	18,517.81	21,887.03
$Pseudo R^2$	0.1415	0.1463	0.1630	0.1927
Log likelihood	−58,691.329	−58,349.099	−47,550.936	−45,856.162
Observations	159,950	159,900	159,774	159,724

Notes: Table presents Logit regression results on exporting. Dependent variables are the firm’s intermediates exporting dummy (columns 1–2) and final goods exporting dummy (columns 3–4). Standard errors are represented in the parentheses. Significance levels: *** 1%, ** 5%, * 10%.

presents the estimation results of the Logit model. Observing columns (1) and (3), firms with higher pollution emission levels show a tendency to export intermediate goods, while those with lower pollution emissions prefer exporting final goods. Our findings in columns (2) and (4) indicate that firms with lower pollution emission intensity favor exporting both intermediate and final goods. This confirms the presence of a self-selection effect where firms’ pollution emission characteristics influence their choice of export strategy. Additionally, we note that firms with higher profitability, lower financing constraints, and greater age are more inclined to export intermediate and final goods.

4.2. Treatment Effects of Different Exported Products on Pollution Emissions

Following the estimation of the PSM method as presented in Table 2, we proceed to derive the Average Treatment Effects. Specifically, we employ the nearest neighbor matching method with n = 4, which is presented in

Table 3. Exporting and pollution emissions.

	Intermediate Goods		Final Goods
	(1)	(2)	(3)	(4)
ATT	0.0684 ***	−0.1225 ***	−0.0391 ***	−0.2709 ***
	(0.0100)	(0.0116)	(0.0117)	(0.0142)
Covariates	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Control group	134,705	134,815	140,212	139,942
Treated group	24,429	24,425	18,274	18,272
Observations	159,134	159,240	158,486	158,214

Notes: Table presents the Average Treatment Effect of Treated (ATT) of exporting on firms’ emissions. The outcome variable is the difference in total emissions and emission intensity, shown in columns (1) (3) and (2) (4), respectively. Standard errors are represented in the parentheses. Significance levels: *** 1%, ** 5%, * 10%.

. The effects of intermediate goods exporting are given in columns (1) and (2), while columns (3) and (4) present on final goods exporting. Additionally, columns (1) and (3) present the impact of exporting intermediate or final goods on total ${SO}_2$ emissions, and columns (2) and (4) show the effect of changes in firms’ ${SO}_2$ emission intensity after exporting intermediate or final goods.

By analyzing the coefficients of ATT in Table 3, it is evident that firms that export intermediate goods experience a rise in their total pollution emissions, with an ATT value of 0.0684, relative to non-exporting firms. Conversely, exporting final goods results in a decline in total pollution emissions, with an ATT of −0.1125. In terms of pollution emissions intensity, exporting both intermediate and final goods results in decreased pollution emissions intensity for firms, with ATT values of −0.0391 and −0.2709, respectively.

The observed differences in pollution emissions may stem from firms’ distinctive characteristics. On the one hand, firms exporting intermediate goods expand their production scale to fulfill foreign demand, thereby increasing total pollution emissions. On the other hand, exporters of final goods face a more diverse and variable consumer market and are less likely to alter their production scale, thus preventing an increase in total pollution emissions. Moreover, both intermediate and final goods exporters show a reduction in pollution emissions intensity, likely attributable to the technical effect of exporting. Through access to advanced foreign production technologies, exporters can enhance their productivity, thereby lowering pollution emissions intensity. Final goods exporters show a higher technical effect of exports relative to the scale effect, resulting in declining total pollution emissions and emission intensity. Our analysis highlights the potential role of scale and technical effects as mechanisms through which exports affect firms’ pollution emissions. However, rigorous empirical testing is necessary to fully elucidate these mechanisms.

To evaluate the distinctions between the treated and control groups, a variable balance test is conducted as part of this study. The findings of this test, reported in

Table 4. Variable balance test.

	${\mathit{I}\mathit{n}\mathit{t}\mathit{e}\mathit{r}\mathit{e}\mathit{x}\mathit{p}\mathit{o}\mathit{r}\mathit{t}}_{\mathit{j}\mathit{i}\mathit{s}\mathit{t}}$		${\mathit{F}\mathit{i}\mathit{n}\mathit{a}\mathit{l}\mathit{e}\mathit{x}\mathit{p}\mathit{o}\mathit{r}\mathit{t}}_{\mathit{j}\mathit{i}\mathit{s}\mathit{t}}$
	(1)	(2)	(3)	(4)
${lnSO}_{2jist}$	−0.2 (−0.17)		−1.3 (−1.32)
${lneSO}_{2jist}$		0.3 (0.34)		0.8 (0.72)
${Profit}_{jist-1}$	−0.2	−0.4	−1.2	−1.2
	(−0.20)	(−0.49)	(−1.08)	(−1.08)
${Finance}_{jist-1}$	−0.2	−0.3	−1.0	−0.3
	(−0.23)	(−0.30)	(−0.94)	(−0.31)
${lnAge}_{jist-1}$	−1.4	−0.8	−1.6	−1.1
	(−1.59)	(−0.88)	(−1.59)	(−1.09)
Year FE	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Control group	134,705	134,815	140,212	139,942
Treated group	24,429	24,425	18,274	18,272
Observations	159,134	159,240	158,486	158,214

Notes: Table presents the results of benchmark model’s balance test. The outcome variable is the difference in total emissions and emission intensity, shown in columns (1) (3) and (2) (4), respectively. T values are represented in the parentheses. Significance levels: *** 1%, ** 5%, * 10%.

, suggest that there are no noteworthy differences between the groups in each control variable after matching. These results indicate that the two groups have similar ex ante characteristics. Based on this finding, it can be inferred that export is the primary cause of the notable divergence in pollutant emissions.

4.3. Robustness Checks

To bolster the precision and dependability of the results displayed in Table 3, supplementary methods are implemented in our study.

4.3.1. Changing the Construction Method of Outcome Variables

In the benchmark regression, we use the $Z_{it+1}-Z_{it-1}$ construction method, which represents the difference between total pollution emissions or pollution emission intensity one year after the firm’s export and one year before the export. In the robustness test, we change the construction method of outcome variables, and adopt the $Z_{it}-Z_{it-1}$ to calculate the difference in total pollution emissions or pollution emission intensity between the year of the firm’s export and the year preceding the export. The results of

Table 5. Robustness check: Changing the outcome variables.

	Intermediate Goods		Final Goods
	(1)	(2)	(3)	(4)
Panel A: Changing the construction method of outcome variables
ATT	0.0528 ***	−0.0862 ***	−0.0221 **	−0.1922 ***
	(0.0083)	(0.0097)	(0.0098)	(0.0120)
Covariates	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Control group	133,487	133,651	138,999	138,732
Treated group	24,342	24,342	18,196	18,196
Observations	157,829	157,993	157,195	156,928
Panel B: Changing other pollutants
ATT	0.1078 ***	−0.1308 ***	−0.0381 ***	−0.2638 ***
	(0.0117)	(0.0133)	(0.0132)	(0.0157)
Covariates	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Control group	120,767	120,809	125,605	124,727
Treated group	22,405	22,391	16,633	16,640
Observations	143,172	143,200	142,238	141,367

Notes: Table presents the robustness check results. The outcome variable is the difference in total emissions and emission intensity, shown in columns (1) (3) and (2) (4), respectively. Standard errors are represented in the parentheses. Significance levels: *** 1%, ** 5%, * 10%.

are consistent with the findings of the benchmark regression, and further provides evidence of the robustness of this study’s conclusions.

4.3.2. Changing Other Pollutants

In the benchmark regressions of this study, we employ sulfur dioxide as the indicator of firms’ pollution. To assess the reliability and robustness of the findings, we substitute other pollutants by using smoke and dust emissions and their emission intensity as the measures. The estimation outcomes are displayed in Panel B of Table 5, where the coefficient of ATT remains robust.

4.3.3. Changing the Treatment Variables

Table 3 presents the benchmark regression results that distinguish between intermediate goods exporters and final goods exporters to explore their impact on firms’ pollution emission behavior. In the robustness test, we extend our analysis by categorizing the sample into three distinct subsets: firms exclusively exporting intermediate goods, firms exclusively exporting final goods, and firms engaged in the exportation of both intermediate and final goods.

Table 6. Robustness check: changing the treatment variables.

	Only Intermediate Goods		Only Final Goods		Mixed Goods
	(1)	(2)	(3)	(4)	(5)	(6)
ATT	0.0862 ***	−0.0487 ***	−0.0487 ***	−0.2175 ***	−0.0249	0.2908 ***
	(0.0109)	(0.0124)	(0.0135)	(0.0163)	(0.0200)	(0.0231)
Covariates	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES	YES	YES
Province FE	YES	YES	YES	YES	YES	YES
Control group	140,143	140,215	145,640	142,969	149,301	149,619
Treated group	18,778	18,771	12,618	12,620	5652	5652
Observations	158,921	158,986	158,258	155,589	154,953	155,271

Notes: Table presents the robustness check results. The outcome variable is the difference in total emissions and emission intensity, shown in columns (1), (3) and (5) and (2), (4) and (6), respectively. Standard errors are represented in the parentheses. Significance levels: *** 1%, ** 5%, * 10%.

lists the estimation findings. The estimation results reveal that firms exporting only intermediate goods significantly increase total pollution emissions but decrease pollution emission intensity. In contrast, firms exporting only final goods reduce both total pollution emissions and emission intensity, which is consistent with the benchmark regression findings. Interestingly, for firms exporting both intermediate and final goods, there is no statistically significant impact on total pollution emissions but a significant reduction in pollution emission intensity. This result also shows that there are varying effects of exporting various items on firms’ pollution emissions; hence, it is vital to differentiate between various products for discussion. Moreover, the pollution emission intensity after exporting intermediate or final goods decreases significantly, so for firms exporting both types of products, their pollution emission intensity also decreases significantly.

4.3.4. Changing the Construction Method of Outcome Variables

To evaluate the robustness of the findings, we vary the matching method utilizing the nearest neighbor matching method with n = 3 to match the samples. The coefficients and significance of

Table 7. Robustness check: changing the matching method.

	Intermediate Goods		Final Goods
	(1)	(2)	(3)	(4)
ATT	0.0673 ***	−0.1245 ***	−0.0372 ***	−0.2688 ***
	(0.0102)	(0.0120)	(0.0120)	(0.0147)
Covariates	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Control group	134,705	134,815	140,212	139,942
Treated group	24,429	24,425	18,274	18,272
Observations	159,134	159,240	158,486	158,214

Notes: Table presents the robustness check results. The outcome variable is the difference in total emissions and emission intensity, shown in columns (1) (3) and (2) (4), respectively. Standard errors are represented in the parentheses. Significance levels: *** 1%, ** 5%, * 10%.

are also similar to our benchmark regression results.

4.3.5. Dropping Samples of Small Exporting Volume

In this paper’s sample, some firms have relatively smaller export sizes, which may have less impact on pollution emission behavior. Analyzing these firms together with larger ones could result in biased conclusions. To address this, we exclude the sample of exporting firms with export sizes less than USD 500 in the robustness test. The results of

Table 8. Robustness check: dropping samples of small exporting volume.

	Intermediate Goods		Final Goods
	(1)	(2)	(3)	(4)
ATT	0.0761 ***	−0.1200 ***	−0.0433 ***	−0.2743 ***
	(0.0101)	(0.0116)	(0.0119)	(0.0143)
Covariates	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Control group	134,357	134,475	139,860	139,698
Treated group	24,163	24,159	17,801	17,799
Observations	158,520	158,634	157,661	157,497

Notes: Table presents the robustness check results. The outcome variable is the difference in total emissions and emission intensity, shown in columns (1) (3) and (2) (4), respectively. Standard errors are represented in the parentheses. Significance levels: *** 1%, ** 5%, * 10%.

indicate that the coefficients and statistical significance of ATT remain robust, and consistent with the benchmark regression.

4.3.6. Dropping Samples from the Years 2000 and 2001

This study’s sample period spans from 2000 to 2013. Given China’s accession to the World Trade Organization in December 2001, which might prompt increased export activities by firms and potentially influence our findings, we introduce year-fixed effects into the baseline regression model as a partial mitigation measure. Nonetheless, to further address the potential interference resulting from China’s WTO accession, we conduct robustness tests by excluding the data for the years 2000 and 2001 from the sample, effectively commencing the sample period after China’s WTO accession. In

Table 9. Robustness check: dropping samples from the years 2000 and 2001.

	Intermediate Goods		Final Goods
	(1)	(2)	(3)	(4)
ATT	0.0739 ***	−0.1089 ***	−0.0488 **	−0.2555 ***
	(0.0102)	(0.0120)	(0.0120)	(0.0145)
Covariates	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Control group	125,810	125,896	130,593	130,178
Treated group	22,942	22,938	17,173	17,171
Observations	148,752	148,834	147,766	147,349

Notes: Table presents the robustness check results. The outcome variable is the difference in total emissions and emission intensity, shown in columns (1) (3) and (2) (4), respectively. Standard errors are represented in the parentheses. Significance levels: *** 1%, ** 5%, * 10%.

, the robustness test results are consistent with those of the baseline regression, affirming the reliability of our conclusions.

4.4. Mechanism Tests

After conducting benchmark regressions and a series of robustness analyses, we address the inquiry of whether the export of intermediate or final goods affects firms’ pollution emissions and emission intensity. This section extends the analysis by exploring the specific transmission mechanism.

Existing theoretical studies suggest that trade has a twofold impact on firm production. On the one hand, trade affects the scale of firms’ production [23]. To cater to the foreign market demand, firms produce more intermediate or final goods and consume more energy in the production process, leading to a notable escalation in total pollution emissions. This implies a positive linkage between the growth of production scale and the subsequent increase in pollution emissions. On the other hand, exports can motivate firms to enhance their technology and productivity, resulting in a reduction in pollution emissions, which is known as the technical effect [38]. Therefore, this section aims to examine the potential mechanism by introducing the two channels of “scale effect” and “technical effect.”

It is noteworthy that the productivity data available in the China Industrial Firm Database is only preserved from 2000 to 2008. As a result, we select the sample from 2000 to 2008 for analysis in this section.

4.4.1. Scale Effect

Within this empirical study, we utilize the PSM-DID method to analyze the potential influence of exporting intermediate or final goods on a firm’s production scale. Specifically, we utilize Equation (6) as the primary model, where the difference in firms’ sales revenue serves as the designated outcome variable. The empirical results are presented in

Table 10. Scale effect and technical effect.

	Scale Effect		Technical Effect
	(1)	(2)	(3)	(4)
ATT	0.0160 ***	0.0019	0.0312 ***	0.0340 ***
	(0.0061)	(0.0070)	(0.0068)	(0.0093)
Covariates	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Control group	78,458	79,891	66,030	67,179
Treated group	11,971	9964	9683	8458
Observations	90,429	89,855	75,713	75,637

Notes: Table exhibits mechanism tests for scale effect and technical effect on firm’s emission intensity. Columns (1) and (3) and (2) and (4) display results for exporting intermediate goods and final goods, respectively. Standard errors are represented in the parentheses. Significance levels: *** 1%, ** 5%, * 10%.

, with columns (1) and (2) displaying the findings for exporting intermediate and final goods, respectively. The results reveal that for firms exporting intermediate goods, the coefficients of ATT exhibit a remarkable and statistically significant positive association at the 1% level, which implies that exporting is associated with a higher likelihood of firms expanding their production scale. However, the coefficients are insignificant for exporting final goods, suggesting that firms that export final goods do not expand their production scale. These results demonstrate that exporting intermediate goods stimulates the expansion of firms’ production scale, resulting in an increase in total pollution emissions. Conversely, firms exporting final goods do not increase their production scale, thereby avoiding a corresponding increase in total pollution emissions.

4.4.2. Technical Effect

Next, we apply the PSM-DID method to explore potential effects on firms’ total factor productivity (TFP), verifying the technical effect associated with exporting intermediate or final goods. This paper adopts the labor productivity of firms as the proxy for their productivity, consistent with previous studies [39,40]. Specifically, we employ the difference in firm productivity as the dependent variable. To estimate the technical effect of exporting, we refer to columns (3) and (4) of Table 9 for intermediate and final goods, respectively. Notably, the coefficients of ATT for both export types demonstrate a statistically significant positive relationship at the 1% level. Given that productivity improvements are found to enhance firms’ environmental performance [41], our results suggest that exporting intermediate goods or final goods leads to a boost in firms’ productivity, which contributes to a reduction in pollution emissions intensity.

The paper explores the impact of exporting intermediate and final goods on firms by examining both the scale effect and the technical effect. Our findings suggest that firms exporting intermediate goods tend to increase their production scale and simultaneously enhance their total factor productivity, resulting in higher total pollution emissions but lower pollution emission intensity. However, firms exporting final goods do not exhibit a significant expansion in their production scale but rather improve their total factor productivity, resulting in both a decline in their total pollution emissions and their pollution emission intensity.

4.5. Heterogeneous Tests

To comprehensively examine how the export behavior of firms affects their pollution emissions, we further classify the sample based on the factor and pollution intensities within the industries in which the firms operate. Moreover, to isolate the influence of firms’ total output value, we focus solely on the impact of exporting intermediate or final goods on ${SO}_2$ emission intensity in various subsamples.

4.5.1. Different Factor Intensity

To examine the heterogeneous effects of export firms on pollution emissions, we categorize the sample into labor-intensive, capital-intensive, and technology-intensive subsamples based on industry-level factor intensities [42].

Table 11. Emission intensity changes of different factor intensity.

	Capital-Intensive		Technology-Intensive		Labor-Intensive
	(1)	(2)	(3)	(4)	(5)	(6)
ATT	0.2206 ***	0.4302 ***	0.1061 ***	0.2868 ***	0.0809 ***	0.2137 ***
	(0.0241)	(0.0331)	(0.0183)	(0.0263)	(0.0192)	(0.0194)
Covariates	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES	YES	YES
Province FE	YES	YES	YES	YES	YES	YES
Control group	53,351	55,472	35,798	42,633	44,419	41,901
Treated group	5429	2972	11,154	5455	7840	9840
Observations	58,780	58,444	46,952	48,088	52,259	51,741

Notes: Table presents the Average Treatment Effect of Treated (ATT) of exporting on firms’ emissions intensity across various factor intensities. Outcome variable is the difference in firm’s emission intensity. Columns (1), (3) and (5) examine exporting intermediate goods, while (2), (4) and (6) analyze exporting final goods. Standard errors are represented in the parentheses. Significance levels: *** 1%, ** 5%, * 10%.

presents the empirical findings. Among the three industry types, exporting intermediate or final goods has the greatest impact on pollution reduction in capital-intensive industries, the least impact in labor-intensive industries, and a moderate impact in technology-intensive industries. The varying impact can be attributed to the fact that firms operating in capital-intensive industries rely heavily on energy, enabling greater emission reduction. Additionally, firms operating in technology-intensive industries benefit from the learning effects associated with exporting, enhancing competitiveness [43]. Comparing the coefficients of ATT for intermediate and final goods, firms exporting final goods exhibit a higher potential for emission reduction. This may stem from their need to continually adjust production methods to meet diverse market demands, leading to stronger learning effects and substantial pollution emission intensity reduction. This finding may explain the overall decline in total pollution emissions after exporting final goods.

4.5.2. Different Industrial Pollution Intensity

Previous studies have found that the relationship between exports and firms’ emissions exhibits variability contingent upon the pollution intensity prevailing within the industry [6,11]. Thus, we examine the heterogeneous effects of firms in different pollution-intensive industries, spanning both polluting and clean sectors. Industry classification is based on the First National Census of Pollution Source, distinguishing polluting industries (key polluting and monitoring industries) from clean industries (all others).

Table 12. Emission intensity changes of different industrial pollution intensity.

	Polluted Industry		Clean Industry
	(1)	(2)	(3)	(4)
ATT	−0.1195 ***	−0.2664 ***	−0.1399 ***	−0.3186 ***
	(0.0123)	(0.0152)	(0.0341)	(0.0376)
Covariates	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Industry FE	YES	YES	YES	YES
Province FE	YES	YES	YES	YES
Control group	122,092	127,125	11,282	12,468
Treated group	20,847	14,874	3579	3397
Observations	142,939	141,999	14,861	15,865

Notes: Table presents the Average Treatment Effect of Treated (ATT) of exporting on firms’ emissions intensity across various industrial pollution intensity. Outcome variable is the difference in firm’s emission intensity. Columns (1) and (3) examine exporting intermediate goods, while (2) and (4) analyze exporting final goods. Standard errors are represented in the parentheses. Significance levels: *** 1%, ** 5%, * 10%.

presents the results, indicating that firms in clean industries experience a greater reduction in pollution emission intensity compared to those in polluting industries when exporting intermediate or final goods. This can be attributed to higher levels of clean technology innovation, enabling firms to achieve lower pollution emissions under similar conditions. Overall, this study emphasizes how exporters’ effects on ${SO}_2$ emission intensity vary across several industries with high pollution levels.

5. Conclusions

The present literature indicates an unclear understanding of the environmental implications associated with exporting firms. Moreover, the heterogeneous nature of exported products further highlights the potential variability in the environmental impact across different product types. Therefore, we raise an interesting question: what is the impact of exporting intermediate or final goods on firms’ pollution emissions?

To answer this question, we leverage information on Chinese manufacturing firms spanning the period between 2000 and 2013. In doing so, we establish a causal link between exports and firms’ pollution emission behavior using the PSM-DID method. The empirical findings indicate that the impact of exporting intermediate versus final goods varies concerning firms’ environmental performance. More specifically, firstly, exporting intermediate goods results in an increase in firms’ total pollution emissions but a decrease in pollution emission intensity, while exporting final goods reduces both total pollution emissions and emission intensity. Secondly, the export of intermediate or final goods can impact firms’ environmental performance through two vital mechanisms: scale effect and technical effect. On the one hand, firms that export intermediate goods tend to ramp up their production scale, which increases in pollution emissions, while firms exporting final goods do not expand their production scale. On the other hand, firms exporting either intermediate goods or final goods improve their total factor productivity, thereby reducing their pollution emission intensity. Thirdly, exporting intermediate or final products shows a varying effect on firms’ pollution emissions across industry sectors, with capital-intensive and clean industries benefiting the most in terms of reducing pollution emission intensity and improving environmental performance. Overall, our study highlights the significance of considering the specific type of product exported when evaluating the environmental impact of exporting firms.

This study engenders substantial contributions to both theoretical and practical dimensions. On the theoretical front, prevailing research has predominantly scrutinized the environmental implications of trade in developed countries, with limited attention directed toward developing nations, particularly China. Previous studies have often adopted a national, sectoral, or regional perspective. Although certain endeavors have sought to utilize micro-level firm data, they have regrettably overlooked the nuances of export product heterogeneity and its discrete implications for firms’ environmental performance. This study leverages multi-year, micro-level firm panel data and meticulously distinguishes exports into intermediate and final goods. This approach facilitates a more nuanced investigation into the influence of firms’ export behavior on the environment, thereby enriching our understanding of the trade-environment relationship at the product level. Additionally, this study proactively addresses potential endogeneity concerns inherent in the trade-environment nexus by amalgamating propensity score matching and difference-in-differences methodologies. It systematically subjects the research to robustness tests and heterogeneity analyses from multifaceted perspectives, meticulously dissecting the mechanisms underlying the impact of intermediate and final goods exports on the environment. The resultant findings, thus, furnish more robust and dependable conclusions.

On a pragmatic level, this study furnishes a compelling insight: the export of final goods can concomitantly curtail both total pollution emissions and emission intensity. In simpler terms, encouraging exporters to shift their focus toward final goods can potentially dismantle the “zero-sum game” that has traditionally characterized the interplay between trade and the environment. Furthermore, this study delves into the mechanisms revolving around scale and technological effects, underscoring the pivotal role of heightened productivity in mitigating emission intensity. Consequently, governments and firms can channel heightened efforts into the enhancement of production technology, thereby championing environmentally responsible production. These findings, therefore, cast valuable illumination upon the development of theories designed to harmonize the developmental trajectories of developing countries with sustainable, green initiatives in the future.

Over a prolonged period, numerous Chinese firms derive significant benefits from export activities. The findings of this study have significant ramifications for both researchers and policymaking. For research, this study indicates that the export of intermediate goods or final goods has different implications for firms’ pollution emissions. Therefore, researchers should analyze the environmental performance of Chinese exporters separately for intermediate goods and final goods to avoid biased conclusions. This study presents crucial evidence at the product level, highlighting the environmental implications of different types of exports. Based on the results of our study, we also put forward the following enlightenment:

On one hand, the export of intermediate goods has the dual impact of expanding economic scale and simultaneously giving rise to detrimental environmental consequences. However, it is noteworthy that these adverse environmental effects can be effectively alleviated through enhanced productivity, leading to a reduction in the emission intensity of firms. In light of this, when promoting the exportation of intermediate goods, prioritizing advancements in production technology becomes paramount. Governments are poised to play a pivotal role by bolstering their commitment to supporting research and development (R&D) investments in export-oriented firms, augmenting their R&D capabilities, and actively fostering a climate conducive to green technology innovation. Particular emphasis should be placed on incentivizing firms to augment their labor productivity, thereby resulting in a reduction in pollution emissions per unit of output.

On the other hand, concerning final goods, their exportation offers a unique opportunity to concurrently diminish both overall pollutant emissions and emission intensity through the catalytic effects of technological innovations. Consequently, policymakers are urged to advocate for exporters to focus their efforts on the production of final goods. This entails the introduction of incentive mechanisms and welfare subsidies tailored to the needs of final goods exporters, while simultaneously nurturing an environment conducive to the enhancement of governance and production technologies. The establishment of collaborative networks for shared firm resources holds great promise in facilitating the seamless exchange of green technologies among firms and the communal utilization of emission-reduction equipment. This strategic approach not only propels the evolution of green technology but also accelerates the upgrading of emission-reduction equipment, thereby achieving further reductions in pollutant emissions and contributing to the realization of environmentally responsible trade practices.

While this study provides valuable insights into the impact of export trade on firm environmental performance at the product-level micro perspective, it is important to acknowledge its limitations:

Firstly, this study is constrained by a limited theoretical framework, lacking in-depth theoretical models. It primarily investigates the causal relationship between the export of intermediate or final goods and firm pollution emissions empirically, without extensive theoretical development. Future research can enhance our understanding by integrating theoretical models with empirical analyses to provide a more comprehensive view of this complex issue.

Secondly, due to database constraints, this study’s sample period only covers the years from 2000 to 2013, creating a gap with the current year. This limitation may affect this study’s ability to fully capture the recent environmental performance of export-oriented firms. Subsequent research can overcome this limitation by utilizing more up-to-date data to conduct a more in-depth examination of this phenomenon.