Global Value Chain Position and Corporate Environmental Performance: A Comparison of Domestic and Foreign Perspectives in China

Abstract

Under the reality of global value chain (GVC) restructuring, the GVC is not only production networks, but also pollution sources that depend on production activities. Based on domestic and foreign markets, this paper proposes the hypothesis that different GVC locations have differential effects on pollution emissions. Using industrial pollution data and two models (a panel fixed-effects model and an instrumental variable two-stage least squares model), we find that increasing exported upstreamness and net upstreamness in GVC by one unit reduces pollution emissions by 3–7%, while increasing imported upstreamness raises pollution emissions by about 5%. We also find that the position in the GVC influences the green production efficiency of enterprises. To mitigate the environmental problems caused by GVC participation, we suggest that external industry regulations and internal technology absorption capacity should be enhanced. This paper provides useful policy implications for adjusting the GVC position of enterprises under the new international division of labor.

1. Introduction

In the context of the current economic globalization, the rapid development of industry has brought many global environmental problems [1]. The Asian Development Bank Working Group’s report “Towards an Environmentally Sustainable Future: A Country Environmental Analysis of the People’s Republic of China” highlights the serious environmental pollution problems caused by China’s rapid industrialization. Examining the pollution status of Chinese industrial enterprises is crucial for addressing global climate and ecological challenges. Moreover, in the 21st century, lower trade costs and improved information technology have facilitated the growth of a global value chain (GVC; see Appendix A for other abbreviations used in this paper). The GVC, as a new form of the international trade division of labor, distributes production across the world, but also creates uneven environmental pollution among different actors involved in various tasks [2].

China is an important participant in the international trade division of labor. Since the implementation of China’s reform and opening-up policy, China’s participation in the GVC has increased to 44.49%, becoming a veritable “world manufacturing factory” [3]. However, the development of the GVC is accompanied by the intensification of environmental pollution, especially in industrial sectors with large export trade. The Statistical Bulletin of China’s National Economic and Social Development in 2020 shows that China’s total energy consumption reached 4.98 billion tons of standard coal. Industrial pollution emissions are still rising year by year. In the process of participating in the GVC, Chinese manufacturing enterprises are locked into low-value-added segments, which not only have limited profit margins but also often become a means of pollution transfer from developed countries [4]. Facing increasingly severe environmental constraints and the pressure of public opinion from domestic and foreign communities, it is particularly imperative to clarify the inherent link between the position of Chinese enterprises in the GVC and their environmental performance. This is beneficial both for understanding the current situation of pollution in the country and for protecting national interests in international trade.

Currently, many scholars refer to the environmental friendliness of enterprises as “corporate environmental performance”. They explore the factors that affect environmental performance from three perspectives: society and culture, government, and business. With the increasing attention towards environmental issues in society, enterprises have gradually incorporated environmental protection factors into company governance and strategic development to maintain their business image [5]. In terms of society and culture, Lai et al. [6] explored the influence of religious culture on the enhancement of corporate environmental performance, considering the Chinese belief system of “the unity of heaven and man”. Chen et al. [7] asserted that social trust encourages members of society to obey social norms and improve corporate environmental performance. In terms of government behavior, Wang et al. [8] pointed out that the privatization process of state-owned enterprises has confirmed that the improving response to the government’s goals will promote the improvement of the company’s environmental performance, which has confirmed the important role of the government to a certain extent [9,10]. Zhang et al. [11] suggested that government supervision is the main source of environmental pressure on corporations, which has contributed to the improvement of companies’ environmental performance. In terms of corporate behavior, most of these studies follow contingency theory, which argues that corporate behavior matches the environment to achieve higher corporate returns [12]. Corporate social responsibility [13], corporate CEO attributes [14], and the political connections of business management [15] can affect corporate environmental performance.

However, the above literature discusses only single-firm-level influences. With trade integration becoming an increasingly dominant feature of global economic development, it is not sufficient to focus solely on the impact of individual firm behavior on corporate environmental performance. In fact, the GVC, as the production division of labor and benefit distribution chain of international trade, is also a pollution-manufacturing chain closely related to the production division of labor. The process of spatial allocation of production factors and trade flows by the GVC through an international division of labor network is also the process of the global placement of environmental pollution [16]. Therefore, it is imperative to explore the influence of the position of firms embedded in the GVC on their environmental performance to reduce environmental pollution and protect the environmental interests of commodity-producing countries.

To address this gap, our study breaks new ground by conducting an empirical study on the impact mechanism and effect of the different positions of the GVC on corporate environmental performance from domestic and foreign perspectives based on the clarification of the theoretical mechanism of the positions of the GVC on corporate environmental performance.

This paper makes the following possible contributions: first, it provides a quantifiable and comparable measure of the upstream degree of each enterprise in the GVC from the perspective of micro enterprises, which are the main actors in pollution prevention and control. It systematically examines how corporate positions in domestic and foreign production affect their pollution emissions. Second, it uses a comprehensive indicator based on various pollutants such as industrial wastewater, exhaust gas, and sulfur dioxide, avoiding the bias of using a single pollutant. Third, it explores the mechanisms of how the import and export positions of enterprises in the GVC influence their green production efficiency and environmental performance, based on endogenous growth theory, low-end capture theory, price-cost-plus theory, and social responsibility theory. It also investigates how the domestic production process of the GVC affects pollution emissions through green production efficiency, based on the reverse mechanism and reverse development capability theory. By explaining these micro-theoretical mechanisms, we deepen our understanding of the intrinsic link between the GVC position and environmental performance. We construct a theoretical framework for analyzing the impact of domestic and foreign value chain division on an enterprise’s environmental performance. Fourth, it explores the ways to mitigate the pollution dilemma from both internal and external aspects of enterprises. Moreover, it empirically analyzes the dynamic impact of GVC position on environmental performance, considering that it changes over time and has different effects on pollution emissions. This provides new insights for enterprises to reduce pollution emissions from a dynamic perspective.

The paper is structured as follows: Section 2 reviews the literature and develops the hypotheses. Section 3 describes the model and data. Section 4 performs an empirical analysis that addresses endogeneity and tests the mechanisms. Section 5 conducts an extended analysis that examines the ways to overcome the corporate environmental performance dilemma. Section 6 discusses the results. Section 7 concludes and provides policy recommendations based on the findings. Section 8 acknowledges the limitations of the paper and suggests directions for future research.

2. Literature Review and Hypothesis Development

2.1. GVC and Pollution Emissions

Scholars have focused on environmental pollution from participation in GVCs mainly at the regional and industry levels. At the national and regional levels, Lin et al. [17] found that the low-end GVC production segment in China is responsible for 17% to 36% of air pollution emissions. Under the policy of retaining core operations in developed countries and outsourcing the rest, the value chain participation process in developing countries can further exacerbate domestic environmental pollution [18]. This is consistent with the hypothesis of a pollution paradise, in which GVC accelerates the relocation of pollution-intensive industries. At the industry level, Lin and Chen [19] found a U-shaped relationship between GVC embeddedness and industrial pollution emissions in Chinese industries. Most of the above literature develops its analysis at the macro level and lacks evidence of firm-level studies. In addition, the effect of GVC is considered from the perspective of the international industrial division of labor, without taking into account domestic production. This makes it difficult to provide a comprehensive picture of the impact of GVC on the environmental performance of enterprises.

2.2. The Impact Mechanism of GVC on the Environment

The existing literature on the mechanism of the environmental impact of GVC has mainly focused on two aspects: structural effect and technological effect [20]. In terms of structural effects, Lu et al. [21] argued that low-end lock-in is the main reason that restricts industrial restructuring, and most of the product segments transferred by GVC have low technology content, which easily forms enterprises’ GVC dependence and hinders their transformation to high value-added segments [22]. In terms of the technology effect, some scholars suggested that GVC has positively contributed to energy saving and emission reduction through the technology effect. Yang and Chen [23] pointed out that imports from developing countries have a certain technology spillover effect, and technology improvement is achieved through the demonstration effect, competition effect, and training effect of foreign direct investment (FDI). In addition, firms participating in GVC can improve productivity and reduce environmental pollution by copying and improving foreign technologies. In addition to the competition effect and push-back mechanism, leapfrog industrial growth can be achieved by shortening pollution-manufacturing chains and reducing pollution emissions [24]. Another group of scholars suggested that the process of enterprises embedding in GVC deteriorates pollution emissions. According to the value chain capture hypothesis, innovative technologies acquired by developing countries have certain limitations. It is usually undertaken with the vested interest of the GVC leader, which tends to create a technology-absorption threshold and makes it difficult for developing countries to catch up technologically [25]. A review of the above-mentioned literature shows that due to the complexity of the GVC and the lack of research on corporate pollution data, it is not only difficult to clarify the mechanism of the GVC on environmental pollution but the literature also fails to provide a specific approach to solving the pollution dilemma.

2.3. The Theoretical Impact Mechanism of the Position of GVC on the Environmental Performance of Enterprises

In our research, the upstreamness indicator is used to describe the GVC position of a firm, that is, to measure the distance from production to final demand [26,27]. For a given firm, it may be both an importer of intermediate goods and an exporter of final goods. Therefore, two indicators, upstreamness of import and upstreamness of export, are used to measure the participation of firms in the production of global value chains.

As shown in Figure 1, from the perspective of imports, on the one hand, according to the theory of endogenous growth dynamics of enterprises, enterprises can improve the quality and diversity of imported intermediate goods through trade liberalization, thus enabling them to achieve endogenous growth, increase productivity, reduce pollution emissions, and realize the cost-saving effect and technology spillovers [28]. However, according to research by Antràs and Chor [29], firms with high upstream imports are mostly engaged in the processing and production of primary raw materials such as energy and chemicals, while downstream products have more information and technology from upstream links. Therefore, the technology spillover effect is more likely to occur in enterprises with lower upstream imports. On the other hand, the import upstream index in our research is not the same as the upstream link of the actual industrial chain. According to the theory of low-end capture, enterprises engaged in the production of primary products are likely to encounter a technology-transfer threshold and technology blockage, which is not conducive to the productivity improvement of enterprises and less likely to achieve the simultaneous improvement of economic and environmental performance.

Based on the preceding arguments, we propose the following hypothesis:

Hypothesis 1.

In the position of the value chain, it is difficult to achieve productivity improvement with a high import upstreamness, which is not conducive to reducing pollution emissions.

From the export perspective, on the one hand, firms are subject to greater critical productivity in the upstream segment compared to the downstream segment [30]. According to price-cost markup theory, in a more demanding survival environment, the average productivity is greater in industries with a high upstreamness of export, as firms enter and exit. Firms with high productivity can overcome the fixed costs of exporting and obtain higher productivity, thus reducing firm pollution. On the other hand, a higher upstreamness of export implies that firms transform from exporting only final goods to exporting a “combination of intermediate and final goods”. This will expand the sphere of influence of companies on the market. According to management strategy theory and social responsibility theory, these companies will seek environmental cost leadership strategies to create a distinctive competitive advantage. In addition, they will build a favorable social and environmental reputation.

Based on these arguments, we propose the following hypothesis:

Hypothesis 2.

In the value chain position, enterprises with high upstreamness of export have higher production efficiency, thus contributing to lower pollution emissions.

Net upstreamness is the difference between import upstreamness and export upstreamness. When export upstream is the same, higher net upstream means greater import upstream, which also indicates that the firm is engaged in more domestic production. The intermediate goods used in domestic production may be either produced internally by the firm or imported by purchasing domestic intermediate goods instead. In the former case, the firm will extend the production chain upstream, and through the “forcing mechanism”, driven by the large value-added share of the upstream link, the firm will try to improve its technology and productivity to match the demand [31]. In the latter case, lower productivity may result from the lower quality of domestic intermediate products. According to the reverse development theory, this would discourage firms from raising productivity through the reverse development of high-quality intermediate goods [32], which in turn would be detrimental to environmental performance.

Based on this theory and the preceding arguments, we propose the following hypothesis:

Hypothesis 3.

The improvement of the net upstreamness of the enterprise means that it is engaged in more domestic production links. Whether this can increase productivity and reduce the level of corporate pollution depends on the source of intermediate goods used by the firm.

3. Data Sources and Methodology

3.1. Data Sources

The data in our research are mainly obtained from the China Customs Database, the World Input–Output Tables, and the China Industrial Enterprises Database. Among them, the Chinese customs database records the product-level imports and exports of Chinese enterprises. The world input–output table adopts the version released by the Organization for Economic Co-operation and Development (OECD) in 2016, which records the input-output of 56 sectors in 43 countries and regions. Considering the changes in the statistical standard of industrial enterprise data and the missing key variables such as the enterprise legal person code, added industrial value, and intermediate inputs in the Chinese industrial enterprise database in 2008 and after, it is not possible to match and measure across years with the Chinese customs database according to the method of [33]. In addition, the data from the Chinese customs database started in 2000, so the sample interval was selected as 2000–2007 [4]. By combining and cleaning the data from the above three sources, we finally obtained a total of 264,839 observation samples from 2000–2007 at the enterprise level. We merged the province-level data on this basis to construct district-level control variables. The above economic indicators were deflated to eliminate the variation caused by different years. To mitigate heteroskedasticity, explained variables and some control variables were taken as logarithms.

3.2. Variable Selection

3.2.1. Explained Variable

The scale of corporate pollution emissions is a key indicator to reflect the environmental performance of enterprises [34]. To comprehensively and truly reflect the scale of corporate pollution emissions, a total of six pollutants, including industrial wastewater emissions, chemical oxygen demand, total industrial waste gas emissions, sulfur dioxide emissions, soot emissions, and industrial dust emissions, were selected for measurement in our research, which can effectively overcome the shortcomings of single pollutant indicators to measure the environmental performance of enterprises. Meanwhile, considering the different units of variables for various pollutants, the comprehensive index of each pollutant was constructed by drawing on the research by Su [4]. The specific calculation formula is as follows:

$$po{l}_{it}=\frac{1}{N}{\displaystyle \sum _{c=1}^N{W}_{cit}{A}_{cit}}$$

(1)

where $W_{cit}$ is the ratio of the emissions of each pollutant of the enterprise to the national average for the same period, and $A_{cit}$ is the result of standardizing the emissions of each pollutant into dimensionless form.

3.2.2. Explanatory Variables

Based on the work by Antràs, Chor, Fally and Hillberry [26], which used the world input–output tables to construct industry upstreamness indicators to describe the physical coordinates of the value chain in which the industry is located, we first use the Chinese customs database to construct industry upstreamness indicators to indicate the distance of the industry from final demand (U_j). The specific calculation is as follows:

$$U_j=1\times \frac{F_j}{Y_j}+2\times \frac{{\displaystyle \sum _{k=1}^N{d}_{jk}{F}_k}}{Y_j}+3\times \frac{{\displaystyle \sum _{k=1}^N{\displaystyle \sum _{t=1}^N{d}_{jt}{d}_{tk}{F}_k}}}{Y_j}+\dots$$

(2)

where $F_j$/$Y_j$ represents the share of final demand in the sector’s total output. The number multiplied by each term in the formula is the length of the production chain. For example, in a single-region input–output table, the length of the production chain is 1 when the value-added input of industry 1 is used directly in the production of final consumer goods in the industry. If a portion of the value-add of industry 1 is input into the production of industry 2 as intermediate goods and industry 2 produces them as final goods, the production chain length of industry 1 is 2. The length of the production chain is 3 when a portion of the value-add of industry 1 is invested in the production of industry 2 as intermediate goods and industry 3 produces them as final goods, and so on; extending this to a multi-regional and multi-sector input–output table is the source of the production chain length in Equation (2).

During data processing, it was found that the Chinese customs database is the eight-digit code of the Chinese tax code. However, the OECD 2016 version of the world input-output data is based on ISIC4 codes, and the two codes are not consistent. We refer to Alfaro et al. [35] for table coding conversion to match the Chinese customs data with the world input–output table data.

Following this, the upstreamness of the enterprise was calculated. After the upstreamness of the industry was calculated, the product data codes of the HS2002 version of the customs data are matched with the BEC product codes. This is to identify the information on intermediate and final goods of the enterprise. The value of the total exports (imports) of the enterprise to the industry export share (import share) was used as a weight to calculate the upstreamness index of the enterprise. The specific measures of import upstreamness (gvc_imp), export upstreamness (gvc_exp) and net upstreamness (gvc_ne) are as follows:

$$gvc\_im{p}_{ft}={\displaystyle \sum _{j=1}^N\frac{M_{fjt}}{M_{ft}}{U}_j}$$

(3)

$$gvc\_ex{p}_{ft}={\displaystyle \sum _{j=1}^N\frac{X_{fjt}}{X_{ft}}{U}_j}$$

(4)

$$gv{c}_{-}n{e}_{ft}=gv{c}_{-}im{p}_{ft}-gv{c}_{-}ex{p}_{ft}$$

(5)

where $M_{fjt}$ denotes the total imports of intermediate and final products of the enterprise. $X_{fjt}$ denotes the total exports of intermediate and final products of the enterprise.

3.2.3. Control Variables

Referring to Su [4] and Ouyang et al. [36], the control variables selected in our research were shown below: (1) firm size was calculated as the logarithm of the enterprise’s fixed assets; (2) firm age was calculated as the firm’s current business year minus the firm’s opening year; (3) profitability ratio; (4) debt ratio was determined using the debt ratio data in the industrial enterprise database; (5) according to the type of enterprises, state-owned enterprises and foreign-funded enterprises, state-owned and collective enterprises were combined as domestic enterprises, while Hong Kong, Macao and Taiwan enterprises are defined as foreign-funded enterprises; (6) regional GDP per capita; (7) regional foreign direct investment; (8) regional per capita wage level.

The results of descriptive statistics of the explained variables, explanatory variables and control variables are shown in

Table 1. Variable index and descriptive statistics results.

Variable Name	Variable Index	Mean	Std	Min	Max
Scale of firm pollution emissions	pol_sac	0.078	0.373	−0.050	8.348
Imported upstreamness of firm	gvc_imp	2.709	0.771	0	4.780
Exported upstreamness of firm	gvc_exp	0.086	0.488	0	4.791
Net upstreamness of firm	gvc_ne	−2.624	1.139	−4.781	4.791
Firm age	age	2.354	0.528	0	4.963
Firm size	size	7.510	1.468	0	12.145
Profitability ratio	lrl	0.031	0.178	−9.821	80.381
Debt ratio	fzl	0.515	0.250	−0.116	3.288463
Regional GDP per capita	pgdp	10.44906	0.5694283	7.94591	11.93229
Regional foreign direct investment	fdi	11.74503	1.05519	0	16.15131
Regional per capita wage level	pwage	9.9526	0.3649005	2.776332	10.80592

Source: author’s own processing.

.

3.3. Model Construction

The baseline model for our research was calculated as follows:

$$po{l}_{it}={\alpha }_0+{\alpha }_{1}gv{c}_{it}+{\alpha }_2\overrightarrow{X}+{\theta }_j+{\lambda }_t+{\delta }_c+{\epsilon }_{it}$$

(6)

where i and t denote firm and year, respectively. $po{l}_{it}$ is the explained variable. Environmental pollution emissions are selected as a proxy variable for a firm’s environmental performance. The scale of pollution emissions (pol_sac) is denoted in the baseline regression model and replaced with pollution-emission intensity (pol_str) in the robustness test. $gv{c}_{it}$ is the position indicator of the firm’s participation in the GVC, including the imported upstreamness (gvc_imp) and exported upstreamness (gvc_exp) of the firm’s participation in the international division of production, and the net upstreamness of the firm’s participation in the domestic division of production (gvc_ne). $\overrightarrow{X}$ denotes the control variables. The model controls for industry fixed effects ${\theta }_j$, regional fixed effects ${\lambda }_t$ and time fixed effects ${\delta }_c$. ${\epsilon }_{it}$ is the random error term.

4. Analysis of Empirical Results

4.1. Estimation Results for the Basic Model

In the process of the study, we found that even if there is some definite relationship between the impact of the GVC position of a firm on its environmental performance, the endogeneity problem caused by reverse causality cannot be excluded. At the same time, although our study included as many factors affecting the environmental performance of firms as possible, it still cannot avoid the endogeneity problem caused by omitted variables, so it is necessary to choose appropriate instrumental variables to mitigate the endogeneity effect.

For instrumental variables, some studies have used explanatory variables lagged terms or mean values, such as Su [4], who uses industry trichotomous level means and time-lagged terms as instrumental variables for the GVC. However, considering that international trade is often influenced by geographical characteristics, the GVC position in which a firm is located is an important expression of participation in international trade. Therefore, geographical factors can meet the relevance requirements as instrumental variables. Moreover, geographical features do not cause the aggravation or mitigation of environmental pollutant emissions; namely, they do not affect the environmental performance of enterprises. Therefore, the exogeneity requirement for instrumental variables has been met. Specifically, the distance between the enterprise and the coastline was calculated by obtaining the latitude and longitude coordinates of the enterprise and the coastline using ArcGIS 10.2 software “domain analysis”, following work by Hong and Guo [37]. The cross product of total exports, imports, and net exports and the distance were regressed as instrumental variables of the upstreamness of exports, upstreamness of imports, and net upstreamness, respectively.

4.1.1. First Stage Results for 2SLS Regression

The results of the first stage of the 2SLS regression are shown in

Table 2. The first-stage results for 2SLS regression.

Variable	GVC Position
	gvc_imp	gvc_exp	gvc_ne
Instrument variable	−0.9605 ***	0.9485 ***	−0.9523 ***
	(0.0175)	(0.0204)	(0.0194)
Control variables	YES	YES	YES
Time fixed effect	YES	YES	YES
Provincial fixed effect	YES	YES	YES
Industry fixed effect	YES	YES	YES
F	3296.01	2178.89	2436.11
R2	0.8101	0.5238	0.6509
Observations	264,839	264,839	264,839

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: Author’s own processing.

. This examines the effect of the distance of the firm from the coastline on the position of the firm’s value chain. The F-statistics were all greater than 10, indicating that there are no weak instrumental variables present in this analysis. The longer the distance from the coast, the higher the transportation cost of imported intermediate goods. It was advantageous for firms to import products that cannot be processed locally, that is, to import products that are less upstream. Thus, the two showed a negative correlation. From the export point of view, the lower the upstreamness of exports was, the more enterprises exported more final products. These final products were more affected by the price on the market. The proximity to the coast can help enterprises reduce transport costs to a certain extent. As a result, there was a positive correlation between the two. In summary, it was advantageous for firms to have a lower upstreamness of export and a higher upstreamness of import (namely, a higher net upstreamness) if they are closer to the port. Therefore, the regression coefficient of net upstreamness was negative.

4.1.2. Regression Results for OLS and Second Stage of 2SLS

Table 3. Regression results of OLS and second stage of 2SLS.

Variable	Corporate Pollution Emissions: pol_sac
	OLS			2SLS
	(1)	(2)	(3)	(4)	(5)	(6)
gvc_imp	0.0115 ***			0.0517 ***
	(0.0019)			(0.0145)
gvc_exp		−0.0118 ***			−0.0709 ***
		(0.0020)			(0.0182)
gvc_ne			−0.0059 ***			−0.0319 ***
			(0.0010)			(0.0085)
lnage	0.0189 ***	0.0189 ***	0.0189 ***	0.0182 ***	0.0182 ***	0.0182 ***
	(0.0014)	(0.0014)	(0.0014)	(0.0015)	(0.0015)	(0.0015)
lnsize	0.0506 ***	0.0506 ***	0.0506 ***	0.0518 ***	0.0518 ***	0.0518 ***
	(0.0005)	(0.0005)	(0.0005)	(0.0009)	(0.0009)	(0.0009)
lrl	0.0304 ***	0.0304 ***	0.0304 ***	0.0300	0.0300	0.0300
	(0.0039)	(0.0039)	(0.0039)	(0.0217)	(0.0217)	(0.0217)
fzl	0.0535 ***	0.0535 ***	0.0535 ***	0.0519 ***	0.0519 ***	0.0519 ***
	(0.0029)	(0.0029)	(0.0029)	(0.0028)	(0.0028)	(0.0028)
lnpgdp	0.0041	0.0041	0.0041	0.00432 *	0.00432 *	0.00432 *
	(0.0024)	(0.0024)	(0.0024)	(0.0021)	(0.0021)	(0.0021)
lnfdi	−0.0010	−0.0010	−0.0010	−0.0003	−0.0003	−0.0003
	(0.0013)	(0.0013)	(0.0013)	(0.0010)	(0.0010)	(0.0010)
lnpwage	0.0378 ***	0.0378 ***	0.0378 ***	0.0396 ***	0.0396 ***	0.0396 ***
	(0.0046)	(0.0046)	(0.0046)	(0.0044)	(0.0044)	(0.0044)
Time fixed effect	YES	YES	YES	YES	YES	YES
Provincial fixed effect	YES	YES	YES	YES	YES	YES
Industry fixed effect	YES	YES	YES	YES	YES	YES
R2	0.115	0.115	0.115	0.141	0.139	0.120
Observations	264,839	264,839	264,839	264,839	264,839	264,839

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: author’s own processing.

reports the regression results of the GVC position of the firm and the pollution emissions of the firm. Columns (1)–(3) in Table 3 are the regression results of the ordinary least squares model. Columns (4)–(6) in Table 3 are the regression results of the two-stage least squares second-stage model (2SLS). The comparison between the two shows that although the regression results have the same direction of influence, the OLS regression results greatly underestimate the influence of the position of the firm’s value chain on environmental pollution. This indicates that the OLS regression model has certain endogeneity problems, and the regression results of 2SLS are more consistent with the actual situation.

According to the regression results for 2SLS, combined with the theoretical mechanism analysis in the previous section, it can be seen that the estimated coefficient of the imported upstreamness of firms (gvc_imp) on the scale of corporate pollution emissions was significantly positive, indicating that enterprises increasing the imported upstreamness will increase pollution emissions, which initially verifies Hypothesis 1. Enterprises with higher imported upstreamness produce low-value-added products, which are less likely to achieve endogenous growth and are subject to low-end capture from GVC leaders. It was thus harder to increase green productivity, which in turn is detrimental to emission-reduction activities. The estimated coefficient of exported upstreamness of firms (gvc_exp) was significantly negative. For every unit increase in the exported upstreamness of firms, the firms’ emissions would decrease by 7.09%, tentatively verifying Hypothesis 2. Firms in industries with higher exported upstreamness enter and exit frequently. The surviving enterprises tend to have higher productivity and greater market power. Under the increasingly severe external environmental constraints, such firms will obtain new competitive advantages by implementing environmental management strategies and take social responsibility through emission reduction to win a positive reputation. The regression results in column (6) in Table 3 show that the greater the net upstreamness, the lower the pollution emission of enterprises, with other factors held constant. This indicates that exporting the same type of products will decrease the pollution emission when firms produce more products in the country. According to the analysis of the theoretical mechanism in the previous section, firms may choose to expand upstream in the production chain within the firm when they retreat to domestic production, namely, reducing the production cost by self-production in the firm. This would provide greater profit margins for reducing pollution emissions and help achieve improved corporate environmental performance.

4.1.3. Explanation for Underestimation of OLS Regression Results

The above regression results illustrated the problem of endogeneity due to omitted variables or measurement errors in the model. Here, the causes and directions of OLS bias were analyzed and explained, and the following two equations were set to determine.

$$pol\_sa{c}_{it}={\beta }_0+{\beta }_{1}gv{c}_{it}+{\rho }_1{O}_{it}+{\beta }_2\overrightarrow{X}+{\theta }_j+{\lambda }_t+{\delta }_c+{\epsilon }_{it}$$

(7)

$$\widehat{gv{c}_{it}}={\gamma }_0+{\gamma }_{1}gv{c}_{it}+{\gamma }_2{O}_{it}+v_{it}$$

(8)

Equation (7) sets the $O_{it}$ term based on Equation (1), and when used to represent omitted variables, the term is related to both the core explanatory variable $\widehat{gv{c}_{it}}$ (embodied as ${\gamma }_2\ne 0$) and the explained variable $pol\_sa{c}_{it}$ (embodied as ${\rho }_1\ne 0$), as required by the omitted variables. When $O_{it}$ term is used to represent measurement error, the true upstreamness of the firm will be represented by $\widehat{gv{c}_{it}}$ in Equation (8). Integrating the above two equations will reveal that the bias of the estimation results is determined by ${\rho }_1\times {\gamma }_2$. For example, the reason why the OLS regression results are underestimated for exported upstreamness of firm is that, firstly, if the effects of certain variables are ignored, such as trade opening policy shocks, the policy variable will prompt more enterprises to obtain a more advantageous upstreamness of exports. At this point, the omitted variable shows a positive relationship with the upstreamness of exports (${\gamma }_2>0$). In addition, a more open trade environment will increase the order density, and the expanded production scale will inevitably cause an increase in pollution (${\beta }_1>0$), thus leading to an underestimation of the role of upstreamness of exports on emission reduction (${\beta }_1\times {\gamma }_2>0$). Secondly, if there is a measurement error in the OLS regression model, such as the underestimation of exported upstreamness of enterprises, this will also give the estimated coefficients in the OLS model ${\beta }_1$ a downward bias. Therefore, the following empirical sections followed the 2SLS regression model to test the effect of firms’ GVC position on enterprise environmental pollution.

4.2. Mechanism Test

The above analysis revealed that the position of the firm’s value chain has different directions of effect on the firm’s pollution emissions. The theoretical mechanism analysis suggested that firms’ GVC position can influence firm pollution emissions by affecting firms’ green productivity. In this section, this mechanism was tested empirically.

In addition to traditional total factor productivity, green total factor productivity (GTFP) was calculated based on energy input constraints and pollution emissions as an unwanted output constraint. The change in GTFP was calculated using the Malmquist index. This index being greater than 1 (or less than 1) indicates green productivity growth (or regression, respectively). We drew on work by Antràs and Chor [29] and used a two-step mechanism test model to examine the effect of firms’ GVC position on GTFP.

Columns (1)–(3) in

Table 4. Test results of the mechanism of action of GTFP.

Variable	GTFP			pol_sac
	(1)	(2)	(3)	(4)	(5)	(6)
gvc_imp	−0.00171 ***
	(0.000333)
gvc_exp		0.00843 ***
		(0.000644)
gvc_ne			0.00223 ***
			(0.000249)
gvc_imp × GTFP				0.523 ***
				(0.0725)
gvc_exp × GTFP					−2.312 ***
					(0.140)
gvc_ne × GTFP						−0.636 ***
						(0.0541)
GTFP				−2.729 ***	−1.216 *	−2.988 ***
				(0.514)	(0.477)	(0.497)
Control variables	NO	NO	NO	YES	YES	YES
Time fixed effect	YES	YES	YES	YES	YES	YES
Provincial fixed effect	YES	YES	YES	YES	YES	YES
Industry fixed effect	YES	YES	YES	YES	YES	YES
R2	0.581	0.581	0.581	0.169	0.169	0.169
Observations	261,317	261,317	261,317	257,547	257,547	257,547

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: author’s own processing.

show the regression of Equation (10), which tests the effect of firm’s GVC position on GTFP. The results showed that the estimated coefficients of the exported upstreamness and net upstreamness of firms on GTFP are significantly positive. In contrast, the regression results of the imported upstreamness of firms are significantly negative. This was consistent with the previous theoretical analysis that firms with lower imported upstreamness are affected by technology spillover and thus have greater productivity. Firms with higher exported upstreamness have greater market power and higher average industry productivity. Firms with higher net upstreamness seek to expand their production links to form a more competitive green production efficiency.

Further, columns (4)–(6) in Table 4 show how firms’ pollution emissions are influenced by the above effects. First, the regression results for GTFP on firms’ pollution emissions were significantly negative, which suggests that enhancing GTFP will reduce firms’ pollution emissions and enhance their environmental performance. The estimated results of the interaction terms of exported upstreamness and net upstreamness with GTFP, respectively, were negative. This indicates that enhancing the exported upstreamness of firms or firms returning their production operations to the country will help enhance the positive effect of GTFP on emission reduction. On the contrary, the significant positive results of the estimation of the imported upstreamness of firms and the GTFP interaction term indicated that increasing the imported upstreamness of firms is detrimental to emission reduction. The regression results confirmed the effect of firm’s GVC position on firm’s pollution emissions under the role of GTFP. At this point, Hypotheses 1–3 have been more thoroughly tested.

4.3. Robustness Test

Considering that the scale of corporate pollution emissions is often closely related to the business conducted by the enterprise, the scale of the enterprise, and so on, we may not be able to accurately depict the level of pollution emission by enterprises using absolute emissions indicators, nor can we examine the environmental performance of enterprises from the perspective of technology, energy, and efficiency. Therefore, we used the pollution-emission-intensity indicator as a proxy variable to remeasure the environmental performance of enterprises. Pollution-emission intensity is the amount of pollutant emissions caused per unit of economic output. As shown in

Table 5. Robustness test results.

Variable	pol_str
	(1)	(2)	(3)
gvc_imp	0.0103 ***
	(0.00303)
gvc_exp		−0.0616 ***
		(0.00438)
gvc_ne			0.0121 ***
			(0.00226)
Control variables	YES	YES	YES
Time fixed effect	YES	YES	YES
Provincial fixed effect	YES	YES	YES
Industry fixed effect	YES	YES	YES
R2	0.034	0.022	0.033
Observations	264,039	264,039	264,039

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: author’s own processing.

, the robustness test showed that the estimated coefficients of the upstreamness of imports were significantly positive. In contrast, the estimated coefficients of the upstreamness of exports and net upstreamness were negative. The regression results were consistent with the core findings of our study.

4.4. Heterogeneity Analysis

The theoretical mechanism analysis and regression results in the previous section showed that the different positions of firms in the value chain at home and abroad have a significant impact on firms’ environmental performance. On the one hand, China, as a socialist country, has a wide range of ownership enterprises. On the other hand, China is a vast country with large disparities in resource endowments and trade advantages between different regions. Focusing on the impact of these differences on the relationship between GVC position and corporate environmental performance is conducive to a more comprehensive and profound improvement of corporate environmental performance from the perspective of GVC position. Based on this, the next part of this study will analyze the heterogeneity in three aspects, including enterprise ownership, enterprise trade mode, and enterprise region, respectively.

4.4.1. Different Enterprise Ownership

The different ownership of enterprises may lead to differences in pollution-treatment technology and supporting sewage equipment. This leads to heterogeneity in the impact of firms’ GVC position on pollution emissions. As shown in

Table 6. Heterogeneity regression results distinguishing between firm ownership.

Variable	Domestic Enterprises			Foreign-Owned Enterprises
	(1)	(2)	(3)	(4)	(5)	(6)
gvc_imp	0.0160 *			0.0321 ***
	(0.00747)			(0.000977)
gvc_exp		−0.0184 **			−0.103 ***
		(0.00710)			(0.00174)
gvc_ne			−0.0135 ***			−0.00302 ***
			(0.00363)			(0.000793)
Control variables	YES	YES	YES	YES	YES	YES
Time fixed effect	YES	YES	YES	YES	YES	YES
Provincial fixed effect	YES	YES	YES	YES	YES	YES
Industry fixed effect	YES	YES	YES	YES	YES	YES
R2	0.276	0.276	0.276	0.274	0.274	0.274
Observations	7703	7703	7703	257,136	257,136	257,136

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: author’s own processing.

, dividing the sample into two categories of domestic and foreign enterprises, the direction of influence was consistent with the results of the baseline regression. This indicates the robustness of the regression results. The upstreamness of import and upstreamness of export reflected the foreign value chain position of enterprises. Obviously, the impact of foreign-owned enterprises was greater and the estimation results were more significant in terms of overseas placement. The net upstreamness reflected the domestic value chain position of enterprises, and the influence of domestic enterprises was more significant on the domestic production layout.

4.4.2. Different Enterprise Trade Modes

Different trade modes reflect significant differences in the characteristics of a company’s production process of traded goods. The different characteristics of energy consumption and pollution-emission intensity of enterprises are mirrored by the cleaning effect and technology effect. As China participates in the global division of production, processing trade accounts for half of the total exports. Therefore, it is necessary to distinguish enterprises between general trade and processing trade types for analysis. As shown in

Table 7. Heterogeneity regression results for distinguishing trade modes.

Variable	Domestic Enterprises			Foreign-Owned Enterprises
	(1)	(2)	(3)	(4)	(5)	(6)
gvc_imp	0.0160 *			0.0321 ***
	(0.00747)			(0.000977)
gvc_exp		−0.0184 **			−0.103 ***
		(0.00710)			(0.00174)
gvc_ne			−0.0135 ***			−0.00302 ***
			(0.00363)			(0.000793)
Control variables	YES	YES	YES	YES	YES	YES
Time fixed effect	YES	YES	YES	YES	YES	YES
Provincial fixed effect	YES	YES	YES	YES	YES	YES
Industry fixed effect	YES	YES	YES	YES	YES	YES
R2	0.276	0.276	0.276	0.274	0.274	0.274
Observations	7703	7703	7703	257,136	257,136	257,136

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: author’s own processing.

, the regression results were all significant at the 1% level and generally consistent with the hypothesis. Overall, the estimated coefficients of general trade enterprises were larger than those of processing trade enterprises. This indicates that the influence of firms’ GVC position on corporate pollution emissions is mainly that of general trade enterprises.

4.4.3. Different Enterprise Affiliation Regions

The economic development status and environmental carrying capacity of different regions in China vary greatly, and the effect of firms’ GVC position on corporate pollution emissions may also vary. According to the regional classification standard of the National Bureau of Statistics of China, the sample is divided into four regions: east, middle, west, and northeastern according to the provinces where the enterprises are located. As shown in

Table 8. Regression results of upstreamness of import of differentiated affiliation regions.

Variable	Regions
	East	Middle	West	Northeast
gvc_imp	0.00372 ***	0.0468 ***	0.0434 ***	0.0122 ***
	(0.00101)	(0.0130)	(0.0120)	(0.00217)
Control variables	YES	YES	YES	YES
Time fixed effect	YES	YES	YES	YES
Provincial fixed effect	YES	YES	YES	YES
Industry fixed effect	YES	YES	YES	YES
R2	0.275	0.117	0.487	0.382
Observations	249,510	2500	2500	10,713

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: author’s own processing.

,

Table 9. Regression results of upstreamness of export of differentiated affiliation regions.

Variable	Regions
	East	Middle	West	Northeast
gvc_exp	−0.0132 ***	−0.0387 **	−0.0984 ***	−0.194 ***
	(0.00205)	(0.0125)	(0.0177)	(0.0190)
Control variables	YES	YES	YES	YES
Time fixed effect	YES	YES	YES	YES
Provincial fixed effect	YES	YES	YES	YES
Industry fixed effect	YES	YES	YES	YES
R2	0.277	0.385	0.486	0.389
Observations	249,510	2500	2116	10,713

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: author’s own processing.

and

Table 10. Regression results of net upstreamness of differentiated affiliation regions.

Variable	Regions
	East	Middle	West	Northeast
gvc_ne	−0.00660 ***	−0.0217 ***	−0.0494 ***	−0.0555 ***
	(0.00102)	(0.00627)	(0.00883)	(0.0119)
Control variables	YES	YES	YES	YES
Time fixed effect	YES	YES	YES	YES
Provincial fixed effect	YES	YES	YES	YES
Industry fixed effect	YES	YES	YES	YES
R2	0.477	0.386	0.486	0.494
Observations	249,510	2500	2116	10,713

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: author’s own processing.

, the regression results indicated that the effects of firms’ GVC position on corporate pollution emissions are all significant at the 1% level. This shows a robust regression result overall. In particular, in central and western regions, higher upstreamness of imports of enterprises has an inhibitory effect on emissions reduction. Central and western regions are less open to the outside world. It is more difficult to improve the quality and diversity of products and not easy to obtain endogenous growth, but the low-end capture effect of pollution reduction is difficult to achieve. Secondly, northeast China has more heavy industrial bases and more serious pollution. The regression results showed that if the northeast region improves the upstreamness of export of enterprises, from the production of industrial final products to portfolio production, this will help alleviate the pollution problem. In addition, the western and northeastern regions show a greater impact on pollution reduction when the production chain is located in the country. In contrast, the eastern region is more open to the outside world. According to the theoretical analysis in the previous section, firms in this region are more likely to adopt import rather than self-production or domestic purchase for production, so the estimated coefficient of net upstreamness on environmental pollution is smaller.

4.5. Dynamic Analysis of Firms’ GVC Position and Corporate Pollution Emissions

Considering that the GVC position of enterprises will change over time, the effect of the GVC position on environmental pollution will also change accordingly. Our study draws on the work by Yu and Duan [20] to conduct a dynamic analysis of firms’ GVC position and corporate pollution emissions over time. As shown in

Table 11. Dynamic regression results of upstreamness of import.

Variable	Time Periods
	2001–2004	2002–2005	2003–2006	2004–2007
gvc_imp	0.079 ***	0.053 ***	0.0132	0.014 **
	(0.018)	(0.012)	(0.0129)	(0.005)
Control variables	YES	YES	YES	YES
Time fixed effect	YES	YES	YES	YES
Provincial fixed effect	YES	YES	YES	YES
Industry fixed effect	YES	YES	YES	YES
R2	0.106	0.101	0.087	0.123
Observations	125,200	167,162	205,414	190,588

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: author’s own processing.

,

Table 12. Dynamic regression results of upstreamness of export.

Variable	Time Periods
	2001–2004	2002–2005	2003–2006	2004–2007
gvc_exp	−0.006 **	−0.013 **	−0.052 ***	−0.087 ***
	(0.002)	(0.005)	(0.013)	(0.019)
Control variables	YES	YES	YES	YES
Time fixed effect	YES	YES	YES	YES
Provincial fixed effect	YES	YES	YES	YES
Industry fixed effect	YES	YES	YES	YES
R2	0.106	0.101	0.087	0.123
Observations	125,200	167,162	205,414	190,588

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: author’s own processing.

and

Table 13. Dynamic regression results of net upstreamness.

Variable	Time Periods
	2001–2004	2002–2005	2003–2006	2004–2007
gvc_ne	−0.007 **	−0.013 **	−0.053 ***	−0.068 **
	(0.003)	(0.005)	(0.012)	(0.025)
Control variables	YES	YES	YES	YES
Time fixed effect	YES	YES	YES	YES
Provincial fixed effect	YES	YES	YES	YES
Industry fixed effect	YES	YES	YES	YES
R2	0.106	0.101	0.087	0.123
Observations	125,200	167,162	205,414	190,588

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: author’s own processing.

, the regression results indicated that the effect of firms’ GVC position on corporate pollution emissions gradually increases in strength. The reason is that as China continues to deepen its domestic and foreign production division of labor cooperation, the inhibitory effect of emission reduction brought about by the higher upstreamness of enterprise imports tends to decrease. In addition, Chinese enterprises continue to expand their market power and enhance their technological level in production cooperation. They also gradually incorporate the concept of green development in their strategic management, so the emission-reduction effect of the upstreamness and net upstreamness of enterprise exports gradually increases.

5. Extended Analysis: Cracking the Dilemma of Corporate Pollution Emissions from Different Perspectives

The above findings show that GVC position in which a firm is located has a significant effect on the environmental performance of the firm. Among them, the upstreamness of import has a significant contribution to the pollution emission. In contrast, the upstreamness of export and the upstreamness of net export have a significant inhibitory effect on the pollution emission. From both internal and external levels, are there any measures that can weaken the promoting effect of GVC position on pollution emission or strengthen the inhibiting effect of GVC position on pollution emission to improve the environmental performance of enterprises? The extended analysis in this section tries to answer these questions by focusing on two aspects: industry environmental regulation and firms’ technological absorption capacity. In this way, we explore the effective measures for the government and enterprises to break the dilemma of corporate pollution emissions and thus improve environmental performance.

5.1. External Factor: Industry Environmental Regulations

We constructed the variable cs in the industry-level analysis using the industry cleaner production standards issued by the Chinese Ministry of Ecology and Environment, taking a value of 1 for industries with established standards and 0 otherwise. The regression model was shown in Equation (9), and the regression results were shown in

Table 14. Regression results for the effect of industry environmental regulation.

Variable	pol_sac
	(1)	(2)	(3)
gvc_imp	0.0123 ***
	(0.0020)
gvc_imp × cs	−0.0524 ***
	(0.0020)
gvc_exp		−0.0359 ***
		(0.0021)
gvc_exp × cs		−0.1290 ***
		(0.0030)
gvc_ne			−0.012 ***
			(0.0010)
gvc_ne × cs			−0.0465 ***
			(0.0013)
cs	−0.3400 ***	−0.1720 ***	−0.3107 ***
	(0.0061)	(0.0020)	(0.0041)
Control variables	YES	YES	YES
Time fixed effect	YES	YES	YES
Provincial fixed effect	YES	YES	YES
Industry fixed effect	YES	YES	YES
R2	0.147	0.151	0.149
Observations	264,839	264,839	264,839

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: author’s own processing.

.

$$pol\_sa{c}_{it}={\alpha }_0+{\alpha }_{1}gv{c}_{it}+{\alpha }_{2}gv{c}_{it}\times cs+{\alpha }_{3}cs+{\alpha }_4\overrightarrow{X}+{\theta }_j+{\lambda }_t+{\delta }_c+{\epsilon }_{it}$$

(9)

After introducing the industry environmental regulation factor, the regression results of firms’ GVC position on corporate pollution emissions were still robust. The estimated coefficient of the cross product term between the upstreamness of import and industry environmental regulation was negative, suggesting that setting industry environmental regulation weakly suppresses the inhibitory effect of the upstreamness of import on emission reduction. This indicates that the constraints imposed by external environmental regulations on enterprises can force them to engage in green production to a certain extent. The estimated coefficients of the cross products of upstreamness of export and net upstreamness were positive. This indicates that for enterprises with higher upstreamness of export, the increase in regional environmental regulations provides them with more incentives to improve their strategic management mechanism. They are also more likely to take the initiative to assume social responsibility, thus leading to a further decrease in their emissions. For firms with higher net upstreamness, industry environmental regulations may have improved firms’ ability to reverse development, further enhancing their environmental performance under the push-back mechanism.

5.2. Internal Factor: Enterprise Technology Absorption Capacity

Drawing on the work by Aghion et al. [38] that used the number of patents to measure the firm’s technological absorption capacity, the mechanism model constructed was as follows.

$$pol\_sa{c}_{it}={\alpha }_0+{\alpha }_{1}gv{c}_{it}+{\alpha }_{2}gv{c}_{it}\times patent+{\alpha }_{3}patent+{\alpha }_4\overrightarrow{X}+{\theta }_j{+}_t+{\delta }_c+{\epsilon }_{it}$$

(10)

The regression results in

Table 15. Regression results for the effect of technology-absorption capability.

Variable	pol_sac
	(1)	(2)	(3)
gvc_imp	0.0162 ***
	(0.0019)
gvc_imp × lnpatent	−0.0108 ***
	(0.0007)
gvc_exp		−0.0168 ***
		(0.0020)
gvc_exp × lnpatent		−0.0114 ***
		(0.0011)
gvc_ne			−0.0090 ***
			(0.0010)
gvc_ne × lnpatent			−0.0068 ***
			(0.0005)
lnpatent	−0.0088 ***	−0.0218 ***	−0.0026
	(0.0021)	(0.0006)	(0.0014)
Control variables	YES	YES	YES
Time fixed effect	YES	YES	YES
Provincial fixed effect	YES	YES	YES
Industry fixed effect	YES	YES	YES
R2	0.120	0.120	0.120
Observations	264,839	264,839	264,839

The tables in brackets are standard errors; *, **, *** are significant under the conditions of 10%, 5%, and 1%, respectively. Source: author’s own processing.

show that the effect of the firm’s technology absorption capacity on the firm’s pollution emissions is significantly negative at the 1% level. This confirms that the firm’s technology-absorption capacity is conducive to reducing pollution emissions. The regression results of firms’ GVC position on pollution emissions remained robust, indicating that the higher upstreamness of enterprises’ imports is not conducive to reducing pollution emissions. The estimated coefficient of the cross product term between the upstreamness of import and technology-absorption capacity was negative, which implies that a greater technology-absorption capacity of enterprises with higher upstreamness of import can inhibit pollution emissions. The regression results of the respective cross products of upstreamness of export and net upstreamness with enterprises’ technology-absorption capacity indicate that the improvement of enterprises’ technology absorption capacity strengthens their emission-reduction ability.

6. Discussion

This study has the following four main findings.

First, this paper systematically describes the location of firms’ division of labor from the perspective of domestic and foreign markets. In accordance with the findings of He and Huang (2021) [39], firms can achieve energy savings and emission reductions by increasing upstream exports or retreating to their home countries. Most of the studies concluded that a higher degree of GVC participation is beneficial for energy saving and emission reduction [40]. However, on the one hand, this literature mainly focuses on studying the degree of international division of labor and fails to capture the impact of the location of participation. In a global context, with a high degree of fragmentation of the production chain, the differences in firms’ international operations are increasingly reflected in their position in the value chain division of labor, and not only in the extent to which they carry out export activities [26]. On the other hand, firms are both importers of intermediate goods and exporters of final goods. The existing literature does not provide a more accurate picture of the specific position of firms in the chain in terms of both exports and imports. At the same time, the international division of labor is vulnerable to uncertainty. Most firms choose to expand domestic production, and the number of domestic production links also affects the way firms participate in the international division of labor and further affects environmental performance [41], which is partially confirmed by the empirical evidence in this paper.

Second, this paper enriches the mechanism of GVC on environmental performance. Previous studies have shown that there is a nonlinear effect of GVC participation on pollution emissions [40], but failed to identify the mechanism of action. In this paper, the mechanism is tested by introducing GTFP. Studies have shown that enhancing GTFP contributes to energy savings and emission reduction, which is also confirmed by Wang et al. [42]. Unlike the existing literature, this paper subdivides the effect under different divisions of labor positions. The study finds that firms with a higher import upstream degree will weaken the emission-reduction effect of GTFP. This is because the presence of importing firms with a higher upstream degree implies that most of the imported factors of production are primary products. These products contain limited technological information, rendering it difficult to realize the “technological learning effect” of imports, as well as being dependent on the strategic intentions of the GVC leaders, and being less effective in reducing emissions [43].

Furthermore, this paper is the first to examine the dynamics of the value chain location’s effect on environmental performance through a timeseries regression. The results show that since China’s accession to the WTO in 2001, the gap between the advantages and disadvantages of the trade division of labor has been widening. Firms with higher import upstream are captured by the chain “leaders” at the lower end of the value chain for a long time and their environmental performance deteriorates, which is consistent with previous studies [44]. In contrast to the existing literature, this paper further finds that firms with higher export upstreamness and net upstreamness continue to improve their GTFP by improving their strategic management capabilities and reverse development capabilities, thus contributing to the continuous improvement of environmental performance, which also explains the opposite performance results of the “pollution paradise effect” in different scenarios [45], which implies that firms need to consider the possibility of extending themselves into low-energy production while undertaking polluting industries.

Finally, in order to address the negative impact of import upstreamness on environmental performance, this paper finds that external environmental regulation and internal technology-absorption capacity are conducive to improving firms’ environmental performance, which also confirms the significant correlation between GVC, technological progress, and environmental pollution [42,46], while demonstrating that environmental regulation not only affects the GVC division-of-labor position [47], but also creates a push-back mechanism for firms to save energy and reduce emissions.

7. Conclusions with Policy Implications

With the deepening of economic globalization, the international division of production has changed from the intra-regional division of labor to the global division of labor. This has not only changed the world trade pattern but also renders the problem of pollution emissions arising from the production process more and more complicated. The research sample was selected as Chinese industrial enterprises, and the position of the enterprise value chain was measured from two perspectives: at home and abroad. Based on the clarification of the theoretical mechanisms, our study, for the first time, conducted an empirical study on the impact mechanisms and effects of different positions of GVC on the environmental performance of enterprises from both domestic and international perspectives. The following main conclusions are drawn.

First, based on the analysis of the theoretical mechanisms by which GVC position affects firms’ environmental performance, an empirical test was conducted to confirm the role of firms’ green total factor productivity. Specifically, for enterprises with higher upstreamness of import, it is more difficult to achieve green productivity improvement and corporate environmental performance under the low-end capture effect and unfavorable endogenous growth environment; enterprises with higher upstreamness of export enhance corporate green productivity by taking social responsibility to pursue a good reputation for protecting the environment. Enterprises with higher net upstreamness emit less pollution. The net upstreamness reflects the production linkage of enterprises’ layout in the country. The results showed that enterprises mostly enhance their production efficiency by expanding to the upstream linkage, thus improving their environmental performance. Hence, policy makers should consider the differences in the domestic and foreign production status of enterprises, and optimize the allocation of internal and external production links by adjusting import incentives. They should also encourage enterprises to export more intermediate products to enhance production efficiency and lower pollution emissions.

Second, the heterogeneity analysis found that the environmental performance of domestic firms is influenced to a significant extent by the domestic production layout. In contrast, the environmental performance of firms abroad is more affected by the foreign production layout. The impact of GVC position on the environmental performance of firms is more influenced by general trading firms. The central region is more influenced by GVC position. It has more incentive to achieve corporate environmental performance improvement by adjusting firms’ participation in domestic and foreign production links. The development of environmental policies needs to take into account differences in the level of regional development and the nature of enterprises. For foreign-funded enterprises, the main focus is to improve the control of exports and enhance the ability to capture and absorb the technology of imported intermediate products. For domestic enterprises, the main focus is on learning improvement in domestic industry chain expansion. Actively improving the value chain position of general trade enterprises will bring significantly better environmental performance improvement than processing trade enterprises. One must be wary of the value-chain pollution-transfer effect brought about by regions with a lower degree of openness to foreign trade.

Furthermore, the results of the dynamic analysis showed that as China continues to deepen the domestic and foreign production division of labor cooperation and improve enterprise technology, the inhibitory effect of emission reduction brought about by higher upstreamness of enterprise imports tends to decrease, while the emission-reduction effect of upstreamness of exports and net upstreamness shows a gradually increasing trend.

Finally, to solve the pollution-management dilemma, we further found that external industry regulation and internal improvement of enterprises’ technology absorption capacity will weaken the negative impact of the upstreamness of import on corporate environmental performance and enhance the emission-reduction effect of the upstreamness of export and net upstreamness. From the government’s perspective, industry environmental regulations can force enterprises to improve their environmental performance and reduce the risk of pollution transfer; from the enterprise’s perspective, by encouraging enterprises to invest in R&D and technological innovation to improve their endogenous growth capacity, and by expanding production links to high value-added links such as R&D and design, enterprises can develop in the direction of good environmental performance.

8. Future Research Directions and Limitations

This paper has several limitations. First, it uses data from Chinese industrial firms, which captures some inter-firm variability, but neglects the important role of non-industrial firms in the globalization process and the internal organization of firms. Future studies could expand the scope of the analysis using small- and medium-sized enterprises as respondents, and by conducting in-depth field studies on specific firms in relation to practice. Second, this paper does not account for regional factors. Although using China as the data source helps to generalize the findings to firms in other developing countries in Asia, it does not consider the cross-country effects. Future research could incorporate domestic and international regional aspects in the analysis and use a spatial econometric model to reveal the interaction effects between geographical units. Finally, in the post-pandemic era, the digital economy has had a significant impact on the international division of labor. This paper is constrained by the data availability and ignores this part of the relationship. Future research should explore it in the context of digitalization using the latest data.