The Impact of GVC Participation on China’s Trade-Implicit Carbon Emission Intensity: A Moderating Effect Based on Industrial Digitalization

Abstract

Based on relevant data from WIOD database from 2010 to 2014, this article calculates the trade-implied carbon emission intensity of various industrial sectors in China, analyzes the impact of GVC embedding on the trade-implied carbon emission intensity of Chinese industrial sectors, and further explores the moderating effect of industrial digitalization on this basis. Research has shown that, on an overall level, as the degree of forward embedding of the GVC deepens, the trade-implied carbon emission intensity of China’s industrial sectors shows an inverted “U”-shaped change of first increasing and then decreasing, while the backward embedding of the GVC promotes trade-implied carbon emissions. From the perspective of industry heterogeneity, there is an inverted “U”-shaped relationship between forward participation in non-pollution-intensive and non-technology-intensive industries and trade-implicit carbon emissions intensity. In technology-intensive industries, there is a positive “U”-shaped relationship between forward participation in the GVC and trade-implicit carbon emissions intensity. The increase in forward participation in pollution-intensive industries effectively suppresses and promotes trade-implicit carbon emissions. At the same time, the improvement of industrial digitalization can promote the early entry of China’s industrial sector’s trade-implicit carbon emission intensity into the decline stage. Therefore, enhancing the forward participation of the GVC and the level of industrial digitalization is an effective measure to promote the low-carbon development of trade in China’s industrial sectors.

1. Introduction

The rapid development of global industrialization has provided an abundant material basis for humanity, but it has also caused a sharp increase in carbon dioxide emissions. Subsequently, extreme weather events have occurred frequently, posing a threat to the global ecological balance and sustainable human development. In order to address global climate change, countries have signed a series of global documents such as the Paris Agreement, announced their nationally determined contributions, and ensured the achievement of global warming control goals (Li et al., 2025) [1]. But, with the further deepening of the international division of labor, economic globalization has enabled international production to break through geographical limitations. Developed countries that occupy the “upstream” links of global value chain(abbreviated as GVC)production have shifted high-energy-consumption and low-added-value production links to developing countries that are in the “downstream” links of the global value chain, resulting in unequal benefits and environmental damage between developed and developing countries (Hermida et al., 2024; Hu et al., 2021) [2,3]. As the largest developing country, China has gained technological spillover from developed countries through its complete industrial system and participation in the GVC division of labor, achieving rapid economic development. However, it has also undertaken a large number of high-energy-consumption and low-value-added industrial transfers from developed countries. The increasing domestic carbon emissions caused by trade have become an issue that cannot be ignored. Therefore, energy conservation and emission reduction are no longer limited to domestic production and consumption but also focus on foreign trade. Therefore, studying whether China’s participation in GVC activities promotes or suppresses trade-implied carbon emissions is a topic worth researching.

The existing research has shown that the continuous integration of the digital economy and the real economy can expand the speed of knowledge and information dissemination, break through existing innovation boundaries, effectively promote technological progress and innovation, optimize resource allocation, break down barriers to cooperation between industries, and promote continuous optimization and upgrading of industrial structures (Chang, 2024) [4]. This technological innovation and industrial structure transformation and upgrading can improve the energy efficiency and production efficiency of enterprises and eliminate outdated high-carbon production capacity and have become an important force in China’s low-carbon transformation. Therefore, it is necessary to incorporate the digital economy into the analytical framework of this article and analyze the role of the digital economy in the embedded GVC in China’s trade-implicit carbon emissions, which has certain practical significance for China to smoothly achieve its “dual carbon” target strategy. This article aims to analyze the changes in the implied carbon emissions intensity of China’s industrial sector trade with the increase in GVC participation, and whether the development of industrial digitalization can reduce the implied carbon emissions of China’s industrial sector trade, in order to provide theoretical reference for relevant departments’ policy formulation. The marginal contribution of this article is that, through empirical research and analysis, it is found that industrial digitalization not only changes the inflection point position of the GVC carbon emission curve but also weakens the carbon emission increasing effect of backward embedding and strengthens the carbon emission reducing effect of forward embedding. This finding provides effective reference for understanding how industrial digitalization reshapes GVC environmental performance.

2. Literature Review

The implicit carbon emissions from trade, as a hot topic in the intersection of environment and trade, have been studied by many domestic and foreign scholars. After Leontief (1970) [5] proposed the input–output analysis method for economy and environment, the accounting method for trade-implicit carbon emissions began to mature. The existing research mainly uses single-region input–output models, MRIO models, and world input–output models to calculate the trade-implicit carbon emissions of a country as a whole or an industrial sector from the production side, consumption side, and other aspects (Li et al., 2020; Xiao et al., 2020; Kim et al., 2021) [6,7,8]. Some scholars have clarified the roles of net exporting and net importing countries in trade-implied carbon emissions from the perspectives of developed and developing countries and distinguished the emission responsibilities on the production and consumption sides (Li et al., 2022; Deng et al., 2017; Wu et al., 2025) [9,10,11]. Furthermore, some scholars have used structural decomposition methods such as SDA, SPA, and LMDI to explore the influencing factors of trade-induced carbon emissions, including trade scale and structure, energy structure and intensity, industrial structure, green technology efficiency, etc. (Li et al., 2023; Pan et al., 2022; Li et al., 2018; Chen et al., 2019) [12,13,14,15]. Scholars have also studied the relationship between trade-implied carbon emissions and economic growth, verifying the existence of the Kuznets curve (Atici et al., 2009) [16].

With the deepening of China’s integration into the global division of labor, how GVC embedding will affect trade-implied carbon emissions has become a focus of current scholars’ attention. The existing research mainly characterizes how GVC status and participation affect trade-implicit carbon emissions. Some scholars believe that the low-end GVC division of labor status and embeddedness are one of the reasons for China’s high trade-implicit carbon emissions, and improving GVC division of labor status and embeddedness will to some extent reduce trade-implicit carbon emissions (Li et al., 2024; Shi et al., 2022) [17,18]. However, some scholars also believe that the impact of GVC status and participation on trade-implicit carbon emissions is not traditional linear but may have a “double threshold effect” (Tao et al., 2019; Lv et al., 2019) [19,20]. Regarding the heterogeneity of GVC embeddings and the implicit carbon emissions from trade, some scholars believe that, as the degree of forward embeddedness of the GVC increases, the implicit carbon emissions from export trade show a reduction effect, while the increase in backward embeddedness will promote the implicit carbon emissions from export trade (Hou et al., 2022) [21]. With the vigorous development of the digital economy in recent years, the speed of knowledge and information dissemination has promoted technological innovation, industrial structure upgrading, and production efficiency improvement. Therefore, it is necessary to study the low-carbon effect of the digital economy on the trade-implicit carbon emissions embedded in the GVC. However, few scholars have included the digital economy, global value chain (GVC) participation, and carbon emissions in a unified analytical framework in the existing research. The existing research results mainly focus on two aspects: One is the relationship between the digital economy and the status or participation of the GVC. Scholars generally believe that the digital economy has improved the level of GVC embedding through technological innovation, industrial structure upgrading, and other paths (such as Gao, 2025; Li et al., 2024; Zhu et al., 2024) [22,23,24]. The second is the impact of the digital economy on carbon emissions, with most empirical evidence supporting its promoting effect on carbon reduction (such as Pu et al., 2025; Zang et al., 2025) [25,26]. In the study of the interaction mechanism among the three, industrial digitalization as a moderating variable of the “GVC carbon emissions” relationship has gradually received attention, but further empirical exploration is needed. For example, Chen et al. (2023) [27] found that industrial digitalization can alleviate the inhibitory effect of GVC bidirectional participation on carbon productivity; Huang et al. (2024) [28] further empirically demonstrated that digitalization enhances the carbon emission reduction effect of forward embedding of the GVC, while weakening the carbon emission enhancement effect of backward embedding, providing key evidence for regulating pathways.

Through reviewing the existing research, it is found that further exploration is needed on the relationship between GVC participation and trade-implied carbon emissions. Firstly, the existing research analyzes the impact of China’s participation in GVC activities on China’s overall, industrial sectors, and trade-implicit carbon emissions in various industrial sectors from the perspective of GVC participation, but heterogeneity analysis is relatively scarce. From the forward and backward perspectives of the GVC, it is possible to explore in depth the impact of GVC participation in various industrial sectors on their trade-implied carbon emissions, which has practical significance for promoting the high-quality development of their trade. Secondly, the digital economy has become an important lever for achieving China’s “dual carbon” goals, but there are few studies that incorporate the three issues of digital economy, GVC participation, and trade-implicit carbon into a unified framework for research. Therefore, studying the impact of GVC embedding on trade-implicit carbon emissions under the trend of digital economy development is of great significance for China to enhance its international competitiveness in the low-carbon field and achieve its “dual carbon” goals.

3. Theoretical Mechanism Analysis and Research Hypotheses

(1)

The impact of GVC forward participation on the implied carbon emission intensity of China’s industrial sector trade

Kuznets (1995) [29] proposed the “Environmental Kuznets Curve” and found that there is not a single linear relationship between economic growth and carbon emissions, and there may be complex relationships such as inverted “U” and “N” shapes. Therefore, the impact of GVC participation on trade-implied carbon emissions may not be linear. According to the “pollution haven” hypothesis, for developing countries, their initial participation in GVC activities is mainly to provide intermediate products at the cost of energy consumption to developed countries (Yu et al., 2021) [30]. As a developing country, China mainly engages in production processes with relatively backward technology in GVC production activities, which easily leads to path dependence on developed countries. At the same time, it faces technological monopolies from developed countries and is prone to the dilemma of “low-end lock-in”, exacerbating carbon emissions in China’s industrial sector (Yin et al., 2022) [31]. With the deepening of China’s participation in international production, it has gained “technology spillover” from developed countries in the process of undertaking industrial transfer, optimized resource allocation, and enhanced clean production capacity (Doryń et al., 2024) [32]. At the same time, with the continuous deepening of environmental regulations, Chinese enterprises participating in international production activities will face increasingly fierce competition and the growing environmental regulatory constraints of developed countries. Therefore, “forced emission reduction” forces Chinese industrial sectors to consider clean production issues in the production process, which further reduces the implicit carbon emission intensity of Chinese industrial sectors’ export trade (Sinn, 2008; Hu et al., 2022) [33,34]. In summary, hypothesis 1 of this article proposes that there may be an inverted “U”-shaped relationship between the forward participation of the GVC and the implied carbon emission intensity of China’s industrial sector trade.

(2)

The impact of GVC backward participation on the implied carbon emission intensity of China’s industrial sector trade

According to the “pollution haven” hypothesis, developed countries tend to transfer their production activities to developing countries with lower levels of environmental regulation through cross-border production or trade transfer, as they have stronger environmental regulations and relatively higher costs of producing high-energy-consuming and highly polluting products, in order to achieve higher economic benefits through lower environmental costs. As a developing country, China has a relatively high proportion of “processing trade”. From a backward perspective, participating in GVC activities may involve more in low-value-added, low-tech, and high-emission international production processes such as processing and assembly, which will inevitably increase China’s carbon emissions burden. At the same time, due to path dependence and technological blockade by developed countries, China will be trapped in the quagmire of “low-end lock-in” for a long time, making it difficult to break through the shackles of low added value and high emissions through technological improvement, innovation, and other means (Li et al., 2024; Shi et al., 2022) [17,18]. Therefore, hypothesis 2 of this article is proposed: the backward participation of the GVC promotes the trade-implicit carbon emissions of China’s industrial sectors.

(3)

The moderating effect of industrial digitalization on GVC participation and the implicit carbon emission intensity of China’s industrial sector export trade

Compared to the environmental pollution that may arise from the advantages of traditional resource endowments, the digital economy has better environmental friendliness. Firstly, as a production factor, data itself has advantages such as easy replication, no loss, and low marginal cost, which can further expand the potential marginal production capacity and improve production efficiency (Chang, 2024) [4]. Secondly, in international production activities, industrial digitalization reduces the information asymmetry between demand and supply sides and enhances the efficiency of resource allocation on a global scale, thereby reducing production costs. At the same time, industrial digitalization weakens the importance of traditional production factors to industrial sectors, enhances the level of intelligent and collaborative production, improves the energy utilization efficiency of industrial sectors, reduces the energy intensity of industrial sectors, and further reduces the implicit carbon emissions of export trade (Wei Yueling et al., 2024) [35]. Therefore, hypothesis 3 of this article is proposed: the improvement of industrial digitalization level will promote the reduction effect of GVC participation on the trade-implicit carbon of China’s industrial sectors.

4. Model Setting, Variable Selection, and Data Sources

(1)

Model construction

Based on the research content of this article and combined with the theoretical mechanism analysis in the previous text, the following model is assumed to characterize the relationship between GVC forward and backward participation and trade-implicit carbon emission intensity.

$$\mathrm{L}\mathrm{N}\mathrm{E}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\mathrm{\alpha }}_0+{\mathrm{\alpha }}_1\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}+{\mathrm{\alpha }}_2\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}^2+{\mathrm{\alpha }}_3\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}+{\mathrm{\delta }}_{\mathrm{i}}+{\mathrm{\upsilon }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(1)

$$\mathrm{L}\mathrm{N}\mathrm{E}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{b}}_{\mathrm{i}\mathrm{t}}+{\mathrm{\beta }}_2\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}+{\mathrm{\delta }}_{\mathrm{i}}+{\mathrm{\upsilon }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(2)

Here, i represents a certain industry sector in China, t represents represents a certain year, and $\mathrm{L}\mathrm{N}\mathrm{E}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}$ represents the logarithm of the trade-implied carbon emission intensity of a certain sector in China for a given year; $\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}$ indicates the forward participation of a certain department in China’s GVC for a given year; $\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{b}}_{\mathrm{i}\mathrm{t}}$ indicates the backward participation of the GVC in a certain department in China for a given year; $\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}$ represents the control variables; ${\mathrm{\delta }}_{\mathrm{i}}$ and ${\mathrm{\upsilon }}_{\mathrm{t}}$ represent individual and time fixed effects, respectively; ${\mathrm{\epsilon }}_{\mathrm{i}\mathrm{t}}$ is a random error term.

(2)

Indicator construction

1. GVC Participation Index

This article mainly draws on the value chain decomposition approach proposed by Wang et al. (2017b) [36]. Under this framework, the GVC participation index is subdivided into two dimensions at the national or industry level: forward (production) and backward (consumption), thereby comprehensively measuring the level of GVC embedding in a country or industry. Forward production linkage mainly refers to the products produced by a certain country being used by downstream manufacturers and exported to other countries, which is represented as $\widehat{\mathrm{V}}\mathrm{B}\mathrm{Y}$ in the input–output model. The backward production linkage reflects the demand for upstream products in the production process of a country, which is expressed as $\mathrm{V}\mathrm{B}\widehat{\mathrm{Y}}$ in the input–output model. Therefore, the calculation formulas for forward participation in the global value chain ($\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_\mathrm{f}$) and backward participation in the global value chain ($\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_\mathrm{b}$) can be expressed as follows:

$$\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_\mathrm{f}={\displaystyle \frac{\mathrm{V}\_\mathrm{G}\mathrm{V}\mathrm{C}}{\mathrm{V}{\mathrm{a}}^{\prime }}}$$

(3)

$$\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_\mathrm{b}={\displaystyle \frac{\mathrm{Y}\_\mathrm{G}\mathrm{V}\mathrm{C}}{{\mathrm{Y}}^{\prime }}}$$

(4)

Here, $\mathrm{V}\_\mathrm{G}\mathrm{V}\mathrm{C}$ can be understood as the situation of a country’s value-added exports from a certain sector to other countries from the past; $\mathrm{Y}\_\mathrm{G}\mathrm{V}\mathrm{C}$ can be understood as the situation where a country’s final output in a certain industry needs to consume products from a certain sector of other countries when viewed from a backward perspective. Forward participation in the global value chain refers to the proportion of added value created by a country’s intermediate goods exports and absorbed by foreign countries in a certain industry, while backward participation in the global value chain reflects the proportion of intermediate product imports needed as inputs for final product production.

2. Implied carbon emission intensity in trade

This article draws inspiration from Zhao et al. (2024) [37] and Duan et al. (2022) [38] and assumes that a country’s direct emission coefficient vector is $E^{rr}$. Using the input–output method, the trade-implicit carbon emission intensity of a certain industry in a country can be obtained. The calculation formula for the trade-implicit carbon emission intensity of country r is as follows:

$$\mathrm{E}{\mathrm{C}}_{\mathrm{r}}={\mathrm{E}}^{\mathrm{r}\mathrm{r}}{(\mathrm{I}-{\mathrm{A}}^{\mathrm{r}\mathrm{r}})}^{-1}$$

(5)

Here, $\mathrm{E}{\mathrm{C}}_{\mathrm{r}}$ represents the trade-implied carbon emission intensity of a certain industry in a country, and the data is sourced from the WIOD environmental account; $\mathrm{I}$ represents the identity matrix; ${\mathrm{A}}^{\mathrm{r}\mathrm{r}}$ represents the mutual demand between domestic production sectors of various countries; and the data is sourced from the WIOD World Input Output Table.

3. Industrial digitalization level

Firstly, drawing on the research approach of Calvino et al. (2018) [39], this study divides industries into different levels of digital intensity based on multidimensional indicators such as “IT access and usage intensity” and “robot usage rate”. Next, we match Calvino’s definition of medium- to high-intensity digital industries with the industry classification system of the World Input Output Table (WIOD). In the matching process, both international authoritative institutions (EU, OECD, UN International Telecommunication Union) and Chinese authoritative institutions (China Academy of Information and Communications Technology, China Academy of Electronic Information Industry, Shanghai Academy of Social Sciences) were consulted to define the core categories of the digital economy. Finally, three dimensions representing the core supply side of the digital economy (Digital infrastructure, Digital Public Services, Innovation in Digital Technology) were identified and determined, totaling 10 specific industries, as shown in

Table 1. Industry selection for measuring the digitalization level of industries.

Index	Secondary Indicators	Select by
Digital infrastructure	C1, C39, C40	Based on indicators such as OECD intelligence level, ICT development index released by ITU, and digital infrastructure of the China Academy of Communications and matching the selection with the world input–output table.
Digital Public Services	C34, C35, C41, C45, C50	Based on indicators such as the service-oriented digital economy of the China Electronics Information Industry Research Institute and digital governance of the Shanghai Academy of Social Sciences and matching them with the world input–output table.
Innovation in Digital Technology	C47, C52	Based on indicators such as OECD and EU ICT human capital and Shanghai Academy of Social Sciences digital competitiveness and matching them with the world input–output table.

. Drawing on the research method of Yao (2022) [40], we use the complete consumption coefficient in the input–output table to quantify the digitalization level of various industrial sectors. The specific calculation formula is as follows:

$$\mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{d}\mathrm{j}}={\displaystyle {\sum }_{\mathrm{d}}{\mathrm{B}}_{\mathrm{d}\mathrm{j}}/{\displaystyle {\sum }_{\mathrm{i}}^{\mathrm{I}}{\mathrm{B}}_{\mathrm{i}\mathrm{j}}}}$$

(6)

Among them, $\mathrm{d}$ represents departments related to the digital economy, $\mathrm{i}$ represents various industrial sectors, and $\mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{d}\mathrm{j}}$ represents the digitalization level of each industrial sector; ${\mathrm{B}}_{\mathrm{d}\mathrm{j}}$ represents the complete consumption coefficient of a certain industry sector j towards digital economy-related sectors, represents the total value of all “digital economy-related sector” products (services) that need to be fully consumed (directly and indirectly) by the final product of the production unit j sector, and reflects the overall dependence of the j sector on digital technology, products, and services throughout its entire production chain; $B_{ij}$ represents the complete consumption coefficient of a certain industry sector j to industry sector i and represents the total value of all industry sectors (including its own) products (services) that need to be fully consumed (directly and indirectly) by the final product of the production unit j sector, reflecting the overall intermediate input intensity of the j sector’s production activities. This ratio measures the proportion of inputs from the core departments of the digital economy in the total intermediate inputs of the final products produced by the j department, reflecting the ability and comprehensive level of the j department to use digital technology and data resources. The higher the ratio, the deeper the dependence of the department’s production process on the core elements of the digital economy and the wider and deeper its digital application.

4. Control variables

This article selects labor remuneration, number of employees, per capita output, industrial structure, energy intensity, and energy structure as control variables. Labor remuneration (LAB) is expressed as the income of employees in various sectors of the manufacturing industry. The number of employees (EMP) is expressed in terms of the number of employees in each industry sector. Per capita output (PV) is expressed as the total output produced by employees in various industrial sectors. The industrial structure (IS) is approximately represented by the proportion of output from various sectors of the manufacturing industry. Energy intensity (EI) is expressed as the proportion of the total energy consumed by each industry sector to the added value. The Energy Structure (ENS) is represented by the proportion of fossil fuels such as coal, oil, and natural gas consumed by various industrial sectors to the total energy consumption.

(3)

Data source.

This article selects data from 47 industries in China from 2000 to 2014 (excluding 9 industries lacking carbon emission data), and the core explanatory variable of GVC forward and backward participation data is sourced from the UIBE database of the University of International Business and Economics. In order to maintain data consistency, the data for the dependent variable was selected from the World Input Output Table and environmental account data published in the WIOD database from 2000 to 2014. The control variables were selected from the economic and social accounts and energy accounts data in WIOD, and the missing data were filled in as much as possible using linear interpolation. Considering the magnitude differences in the values of the control variables and other reasons, logarithmic processing was applied to the relevant data, and the statistical description results are shown in

Table 2. Statistical description.

Variable Type	Variable	Symbol	Mean Value	Standard Deviation	Median	Minimum Value	Maximum Value
Explained Variable	GVC forward engagement	GVCpt_f	0.117	0.074	0.109	0.002	0.417
	GVC backward engagement	GVCpt_b	0.146	0.069	0.139	0.018	0.447
Core explanatory variables	Implied carbon emission intensity	EC	0.720	0.124	0.314	0.064	10.353
Adjusting variables	Industrial digitalization level	Dig	0.165	0.124	0.116	0.043	0.579

.

5. Empirical Result Analysis

(1)

Benchmark regression results

Based on the results of the Hausman test, this article chose fixed effects as the final report, according to the econometric model established earlier, as shown in

Table 3. Benchmark regression results.

Variable	LNEC
	(1)	(2)	(3)	(4)
GVCpt_f	3.016 ***	3.153 ***
	(4.294)	(4.814)
GVCpt_f2	−5.381 ***	−5.968 ***
	(−3.996)	(−4.615)
GVCpt_b			2.553 ***	2.504 ***
			(6.751)	(6.591)
LNDVA		−0.001		0.002
		(−0.066)		(0.127)
LNENS		−0.038		−0.034
		(−1.188)		(−1.252)
LNLAB		−0.090 ***		−0.089 ***
		(−2.970)		(−3.233)
LNEMP		−0.038		−0.046
		(−0.986)		(−1.411)
LNPV		−0.066 ***		−0.113 ***
		(−2.676)		(−4.222)
LNEI		−0.039 **		−0.039 ***
		(−2.472)		(−3.019)
Industry fixed	Yes	Yes	Yes	Yes
Fixed year	Yes	Yes	Yes	Yes
N	687	626	687	626
R-squared	0.973	0.975	0.975	0.976

Note: ***, ** represent significance levels of 1% and 5%, respectively. The numbers in parentheses represent t-values, as shown in the table below. All regressions include industry and year fixed effects and use robust standard errors.

.

From Table 3, it can be observed that the square terms of GVC forward participation in columns (1) and (2) are negative at a significance level of 1%, and the primary terms are positive at a significance level of 1%, indicating a preliminary inverted “U” shape. In order to further verify whether the inverted “U” relationship holds, this paper draws on Haans’ (2016) [41] discriminant method. The results are shown in columns (1) and (2) of

Table 4. Inverted “U” relationship test.

U Test	Column (1)	Column (2)
t value	2.700	3.230
p value	0.004	0.001
Inflection point	0.280	0.264
Left slope	3.000	3.135
Right slope	−1.473	−1.827
95% confidence interval	[0.231; 0.354]	[0.217; 0.327]
Result	Inverted “U” shape	Inverted “U” shape

. Regardless of whether control variables are added, there is a significant inverted “U” relationship between GVC forward participation and trade-implied carbon emission intensity, which further verifies hypothesis 1.

This is mainly because when China first participated in GVC activities, its production capacity and technological level were relatively backward, and it undertook the industrial transfer of many high-carbon industries from developed countries, resulting in an increase in the implied carbon emission intensity of export trade. With the continuous development of the Chinese economy and the continuous improvement of the industrial system, the “learning effect” and “forcing effect” have led to a decrease in the implied carbon emission intensity of export trade in various industrial sectors in China. Therefore, the forward participation of China’s GVC shows an inverted U-shaped relationship with the implied carbon emission intensity of industrial sector trade, which is similar to the conclusion drawn by Chen et al. (2023) [27]. According to the Table 4 analysis, the inflection point of the inverted U-shaped curve is located at 0.264. Currently, only 7 industrial sectors (accounting for 14.9%) have crossed this threshold, while 40 sectors (accounting for 85.1%) are still below the inflection point. The sample average forward GVC participation (0.117) is significantly lower than the inflection point value, indicating that the vast majority of industries are still in the upward curve stage, and the increase in their GVC forward participation will continue to exacerbate carbon emissions. This is mainly due to the fact that most of China’s industrial sectors have not yet broken through the technological blockade of developed countries and are still providing high-energy-consuming and high-emission intermediate products to developed countries, resulting in higher carbon emissions.

The coefficients of GVC backward participation in columns (3) and (4) of Table 3 are positive and significant, indicating that GVC backward participation is beneficial for the trade-implicit carbon emissions of China’s industrial sectors, which is consistent with the conclusion drawn by Chen et al. (2023) [27]. This is mainly because, as a developing country, China mainly participates in the international division of labor through extensive production activities such as assembly and processing, relies more on new technologies and products from developed countries, has weak innovation capabilities, and has been locked in low-end production links for a long time. As a result, the trade-implicit carbon emission intensity of China’s industrial sector continues to increase with the deepening of its participation in the international production division of labor.

(2)

Endogeneity and robustness testing

There may be a “bidirectional causal” relationship between GVC participation and trade-implied carbon emission intensity, mainly manifested as the higher the GVC participation, the higher the motivation and demand for Chinese enterprises to transform and upgrade towards the global value middle and high end, which will promote the reduction of trade-implied carbon emission intensity. Enterprises with higher trade-implied carbon emission coefficients have relatively lower technological levels and production capabilities, making it difficult to meet the needs of GVC-dominant enterprises. Their deeper participation in GVC activities is limited, so there may be a “two-way causal” relationship between GVC participation and trade-implied carbon emission intensity. At the same time, there may be a possibility of “omitted variables” during the model building process, which can lead to endogeneity issues in the model due to “bidirectional causality” and “omitted variables”, resulting in biased and inconsistent estimation results. In order to reduce the influence of endogeneity on regression results, this article refers to the existing literature and selects the lagged period of the core explanatory variable as the instrumental variable. This is mainly because the lagged period of GVC participation has a strong correlation with the current period, but it is difficult to affect the error of the current period. Therefore, this article selects the lagged one period of GVC forward and backward participation as the instrumental variable and uses the GMM method for endogeneity testing. Through the Hausman test, it was found that, rejecting the hypothesis that all core explanatory variables are exogenous variables, the LM test and Wald test both reject the null hypothesis of unidentifiable and weak instrumental variables, indicating a strong correlation between the lagged one period of GVC forward and backward participation and endogenous explanatory variables. From columns (1) and (2) of

Table 5. Endogeneity- and robustness-related tests.

Variable	Endogeneity		Robustness
	(1)	(2)	(3)		(4)
	LNEC	LNEC	Control Variable Lag		Excluding the Impact of Financial Crisis
GVCpt_f	3.513 ***		3.201 ***		3.519 ***
	(4.454)		(4.729)		(4.985)
GVCpt_f2	−5.732 ***		−5.908 ***		−6.847 ***
	(−3.653)		(−4.370)		(−4.581)
GVCpt_b		3.248 ***		2.506 ***		2.562 ***
		(7.195)		(6.168)		(6.741)
control variable	Yes	Yes	Yes	Yes	Yes	Yes
Industry fixed	Yes	Yes	Yes	Yes	Yes	Yes
Fixed year	Yes	Yes	Yes	Yes	Yes	Yes
N	587	587	588	588	583	583
R-squared	0.905	0.913	0.976	0.977	0.974	0.976

Note: *** represents significance level of 1%. The numbers in parentheses represent t-values, as shown in the table below. All regressions include industry and year fixed effects and use robust standard errors.

, it can be observed that, at a significance level of 1%, as the forward participation of the GVC increases, the implied carbon emissions intensity of China’s industrial sector trade shows an initial increase followed by a decrease. Meanwhile, the backward participation of the GVC promotes the implied carbon emissions of China’s industrial sector trade at a significance level of 1%, which is consistent with the results of the baseline regression.

In order to enhance the credibility of the regression results, this paper first used lagged control variables and excluded the impact of financial crises to further demonstrate the robustness of the regression results (Dressler et al., 2023) [42]. Columns (3) and (4) of Table 5 show the regression results with control variables lagged by one period and after excluding the impact of the 2008 financial crisis. It can be seen that the relationship between GVC participation and trade-implied carbon emission intensity is consistent with the baseline regression results, proving that the regression results are robust.

In order to further demonstrate the robustness of the regression results, this article adopts two methods, replacing the core explanatory variable and replacing the dependent variable, to explore whether the regression results are robust. Due to the fact that the forward production length and backward production length of the GVC can to some extent reflect the participation of a certain industry in the global value chain production activities of a country, this article uses them to replace the core explanatory variables of GVC forward participation and backward participation. The regression results are shown in columns (1) and (2) of

Table 6. Robustness test.

Variable	LNEC
	Replace Core Explanatory Variables		Replace the Explained Variable		Change the Model
	(1)	(2)	(3)	(4)	(5)
GVCpt_f	0.069 ***		−2.703 ***		3.214 ***
	(4.099)		(−3.11)		(4.463)
GVCpt_f2	−0.223 **		6.657 ***		−6.615 ***
	(−1.974)		(3.59)		(−4.389)
GVCpt_b		0.795 ***		−2.718 ***	2.622 ***
		(15.165)		(−2.85)	(7.245)
control variable	Yes	Yes	Yes	Yes	Yes
Industry fixed	Yes	Yes	Yes	Yes	Yes
Fixed year	Yes	Yes	Yes	Yes	Yes
N	586	586	628	628	687
R-squared	0.978	0.987	0.969	0.969	0.976

Note: ***, ** represent significance levels of 1% and 5%, respectively. The numbers in parentheses represent t-values, as shown in the table below. All regressions include industry and year fixed effects and use robust standard errors.

. It can be found that, after replacing the core explanatory variables, the forward production length and carbon emission intensity of the GVC show an inverted “U”-shaped relationship, and the backward production length of the GVC promotes carbon emission intensity, indicating that the benchmark regression is robust.

Secondly, carbon emission efficiency can directly reflect the level of carbon emissions in various industrial sectors of a country and can therefore be used to replace carbon emission intensity to a certain extent. This article draws on the research method of Lin et al. (2025) [43] and constructs a super-efficiency SBM model that includes unexpected outputs to measure the implicit carbon emission efficiency of China’s industrial sectors, thereby replacing the dependent variable. The implicit carbon emission efficiency reflects the overall carbon emission efficiency of industrial sectors from production to final consumption. The input variables are selected as capital stock, employment, and energy consumption. The expected output is the added value of each industrial sector, and the unexpected output is the trade-implicit carbon emissions of each industrial sector. As shown in Table 6 (3) and (4), there is a U-shaped relationship between GVC forward participation and carbon emission efficiency. GVC backward participation suppresses carbon emission efficiency, further verifying the robustness of the baseline regression.

Considering that an industry can participate in the global value chain in both forward and backward directions, if the forward and backward participations of the GVC are interrelated and interact, then separately considering the forward or backward participation of the GVC may affect the regression results. Therefore, this article incorporates both forward and backward participation of the GVC into the model to examine whether the inverted “U”-shaped relationship between the GVC forward participation and carbon emissions, as well as the positive linear effect between GVC backward participation and carbon emissions, hold when controlling each other. The specific model settings are as follows:

$$\begin{array}{cc}\mathrm{L}\mathrm{N}\mathrm{E}{\mathrm{C}}_{\mathrm{i}\mathrm{t}}& ={\mathrm{\alpha }}_0+{\mathrm{\alpha }}_1\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}+{\mathrm{\alpha }}_2\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}^2+{\mathrm{\alpha }}_3\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{b}}_{\mathrm{i}\mathrm{t}}\\ & +{\mathrm{\alpha }}_4\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}+{\mathrm{\delta }}_{\mathrm{i}}+{\mathrm{\upsilon }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{i}\mathrm{t}}\hfill \end{array}$$

(7)

From column (5) of Table 6, it can be observed that when incorporating forward GVC participation and its squared term, as well as backward GVC participation, into the model, the relationship between forward GVC participation and carbon emissions still exhibits an inverted “U” shape, while backward GVC participation promotes carbon emissions. To test multicollinearity, we calculated the variance inflation factor (VIF). The mean VIF of all variables is 4.59, and the maximum VIF is 9.2 (below the strict threshold of 10), indicating that collinearity does not affect the reliability of the conclusion.

(3)

Heterogeneity analysis

Drawing on the research method of Hao et al. (2022) [44], this article divides various industrial sectors in China into pollution-intensive and non-pollution-intensive categories, based on pollution levels, and technology-intensive and non-technology-intensive categories, based on technological differences. Furthermore, from the perspective of industry heterogeneity, this article explores the impact of GVC forward and backward participation on the trade-implicit carbon emission intensity of Chinese industrial sectors.

As shown in

Table 7. Regression results of heterogeneity of pollution levels in industrial sectors.

Variable	Pollution-Intensive Industrial Sectors		Non-Pollution-Intensive Industrial Sectors
	LNEC	LNEC	LNEC	LNEC
GVCpt_f	−1.905 ***		3.497 ***
	(−5.132)		(5.122)
GVCpt_f2			−5.655 ***
			(−4.506)
GVCpt_b		8.545 ***		3.039 ***
		(3.995)		(6.541)
control variable	Yes	Yes	Yes	Yes
Industry fixed	Yes	Yes	Yes	Yes
Fixed year	Yes	Yes	Yes	Yes
N	105	105	523	523
R-squared	0.994	0.992	0.967	0.973

Note: *** represents significance level of 1%. The numbers in parentheses represent t-values, as shown in the table below. All regressions include industry and year fixed effects and use robust standard errors.

, according to the U-shaped relationship test, the relationship between GVC forward participation and the implied carbon emission intensity of trade in pollution-intensive industries is not inverted U-shaped. For pollution-intensive industries, the increase in GVC forward participation significantly promotes the low-carbon effect of their industrial sectors. This is mainly because, with the deepening participation of pollution-intensive industries in the GVC, they are more susceptible to the constraints of environmental regulations in developed countries. In order to maintain their overseas market size, enterprises have to accelerate the pace of research and innovation in clean production technology, which in turn leads to a decrease in their carbon emission coefficient as they continue to participate in the international division of labor. At a significance level of 1%, the backward participation of the GVC promotes trade-implicit carbon emissions, which is related to China’s main participation in low-tech extensive international division of labor. As the participation continues to deepen, the intensity of trade-implicit carbon emissions also increases. For non-pollution-intensive industries, with the increasing participation in the GVC, the implied carbon emissions intensity of trade shows a trend of first increasing and then decreasing, gradually moving towards a low-carbon development path. This may be due to the fact that China’s non-pollution-intensive industries, after experiencing initial extensive growth, have gained technology spillovers in the process of continuously deepening their participation in the international production division of labor. In addition, the increasing attention to the environment at home and abroad has promoted enterprises to pay more attention to clean production while paying more attention to technological innovation, thereby improving energy utilization efficiency and production efficiency and achieving a shift from initial extensive development to green and low-carbon development. At a significance level of 1%, the backward participation of the GVC in non-pollution-intensive industries also increased the intensity of trade-implied carbon emissions, but its promotion effect on trade-implied carbon emissions was lower than that of pollution-intensive industrial sectors.

Table 8. Heterogeneity regression results of technological differences in industrial sectors.

Variable	Technology-Intensive Industrial Sectors		Non-Technology-Intensive Industrial Sectors
	LNEC	LNEC	LNEC	LNEC
GVCpt_f	−6.677 **		2.911 ***
	(−2.550)		(4.295)
GVCpt_f2	8.065 ***		−5.839 ***
	(6.150)		(−4.355)
GVCpt_b		12.379 ***		2.546 **
		(10.906)		(2.025)
control variable	Yes	Yes	Yes	Yes
Industry fixed	Yes	Yes	Yes	Yes
Fixed year	Yes	Yes	Yes	Yes
N	90	90	538	538
R-squared	0.980	0.979	0.975	0.982

Note: ***, ** represent significance levels of 1% and 5%, respectively. The numbers in parentheses represent t-values, as shown in the table below. All regressions include industry and year fixed effects and use robust standard errors.

shows the regression results of GVC participation on trade-implied carbon emission intensity under different technological differences in industrial sectors. It can be found that the relationship between the forward participation of the GVC in technology-intensive industries and the implicit carbon emission intensity of trade in technology-intensive industries is U-shaped. When the forward participation of technology-intensive industries in the global value chain is low (such as primary intermediate product suppliers), international high-end technology transfer (such as introducing low-carbon equipment and management experience) can quickly improve energy efficiency and clean production levels, significantly reducing the implicit carbon emission intensity per unit output. However, with the continuous increase in participation in the forefront of the global value chain, the industry is facing dual pressures: On the one hand, developed countries, in order to maintain their dominant position in the global value chain, have restricted the development of China’s industrial sector through clean production technologies, environmental regulations, and other means, resulting in China’s technology-intensive industrial sector being locked in relatively high-energy-consuming and high-emission processing and manufacturing links (Zhang, 2023) [45], which increases the difficulty and cost of upgrading China’s technology-intensive industrial sector, making it more likely to be locked in relatively high-energy-consuming and low-emission processing and production links at specific stages. On the other hand, the cost reduction effect brought about by the improvement of technological efficiency will stimulate further expansion of economic scale (Gereffi et al., 2015) [46]. However, some key raw materials and components in technology-intensive industries still rely on high-emission industries upstream of the supply chain, such as steel and chemicals. With the expansion of the industry scale, the demand for these high-emission upstream products will also increase accordingly, indirectly leading to an increase in the carbon emission intensity of the entire industry.

The regression structure of the forward participation of non-technology-intensive industrial sectors in the GVC is similar to that of benchmark regression, mainly because China’s non-intensive industrial sectors mainly participate in the global value chain division of labor by providing intermediate products to developed countries, while non-intensive industrial sectors initially participate in low-value-added and high-energy-consumption production links such as assembly and processing in the GVC division of labor. As their participation deepens, the trade-implied carbon emission intensity continues to increase. But, with the deepening of China’s participation in the international division of labor, non-technology-intensive industrial sectors in China have gained more technology spillovers from developing countries. At the same time, with the upgrading of technology and the improvement of industrial chains, the optimization of resource allocation, and the highlighting of competitive effects, the implied carbon emission intensity of trade has crossed the turning point and begun to show a downward trend. At the same time, it can be found that the backward participation of non-technology-intensive industrial sectors in the GVC also significantly promotes the increase in trade-implied carbon emission intensity.

6. Further Analysis: The Regulatory Effect of Industrial Digitalization

Through the previous analysis, it was found that for the Chinese industrial sector, most of the inflection points that have not yet crossed the inverted “U” shape are still in the process of promoting the increase in trade-implicit carbon emissions with the deepening of forward participation in the GVC, while the backward participation in the GVC exacerbates the trade-implicit carbon emissions of the Chinese industrial sector. Therefore, how to enable more points to cross the turning point ahead of time in China’s forward participation in the GVC production process, reduce carbon emissions from backward participation in the GVC production process, and promote the low-carbon effect of China’s industrial sector’s participation in the international division of labor is a focus worthy of further research and discussion. With the development of the digital economy, the development of various industrial sectors in China has also incorporated more and more digital elements. The deep integration between the digital economy and physical industries can guide resource reallocation, promote technological innovation, improve energy quality and efficiency, and optimize clean production, thereby promoting green and low-carbon economic development and helping China achieve its “dual carbon” goals. In order to further explore how the development of the digital economy can enhance the low-carbon effect of GVC participation, this article, based on hypothesis 3, incorporates industrial digitalization, GVC participation, and trade-implicit carbon emission intensity into a unified framework for analysis, with a focus on the moderating effect of industrial digitalization. The specific econometric model is set as follows:

$$\begin{array}{cc}{\mathrm{LNEC}}_{\mathrm{it}}& ={\mathrm{\rho }}_0+{\mathrm{\rho }}_1\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}+{\mathrm{\rho }}_2\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}^2+{\mathrm{\rho }}_3\mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{it}}+{\mathrm{\rho }}_4\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}\times \mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{it}}\\ & +{\mathrm{\rho }}_5\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}^2\times \mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{it}}+{\mathrm{\rho }}_6\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}{\mathrm{l}}_{\mathrm{it}}+{\mathrm{\mu }}_{\mathrm{i}\mathrm{t}}+{\mathrm{\delta }}_{\mathrm{i}\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{i}\mathrm{t}}\hfill \end{array}$$

(8)

$$\begin{array}{cc}{\mathrm{LNEC}}_{\mathrm{it}}& ={\mathrm{\omega }}_0+{\mathrm{\omega }}_1\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{b}}_{\mathrm{i}\mathrm{t}}+{\mathrm{\omega }}_2\mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{it}}+{\mathrm{\omega }}_3\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{b}}_{\mathrm{i}\mathrm{t}}\times \mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{it}}\\ & +{\mathrm{\omega }}_4\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}{\mathrm{l}}_{\mathrm{it}}+{\mathrm{\mu }}_{\mathrm{i}\mathrm{t}}+{\mathrm{\delta }}_{\mathrm{i}\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{i}\mathrm{t}}\hfill \end{array}$$

(9)

Here, Dig represents the level of digitalization in the industry; $\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}\times \mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{it}}$ represents the interaction term between GVC forward participation and industrial digitalization level; $\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}^2\times \mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{it}}$ represents the interaction term between the square of GVC forward participation and the level of industrial digitalization; $\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{b}}_{\mathrm{i}\mathrm{t}}\times \mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{it}}$ represents the interaction term between GVC backward participation and industrial digitalization level. It is worth noting that this article believes that there is no significant endogenous relationship between industrial digitalization and GVC participation. The main reason is that, on the one hand, industrial digitalization reflects the actual dependence of industrial sectors on the core elements of the digital economy in the production process, rather than directly related to the degree of participation in the global value chain. On the other hand, industrial sector participation in the global value chain is mainly integrated into the global production network through trade, investment, technological cooperation, and other means. Its investment in digital technology is more based on objective market and technological development needs and policy guidance factors, rather than the degree of participation in the global value chain directly leading to the improvement of digitalization level. On the contrary, industrial digitalization can significantly enhance the efficiency and competitiveness of industrial participation in the global value chain. At the same time, this article tries to control for other factors that affect carbon emissions as much as possible (control variables are the same as baseline regression) and uses a fixed effects model to control for the effects of individual fixed effects and time invariant variables, thus ensuring the exogeneity of industrial digitalization indicators to a certain extent.

The regression results in columns (1) to (2) of

Table 9. The moderating effect of industrial digitalization on GVC participation in export trade and implicit carbon emissions.

Variable	(1)	(2)	(3)	(4)
	LNEC	LNEC	LNEC	LNEC
GVCpt_f	5.839 ***	5.220 ***
	(6.769)	(6.870)
GVCpt_f2	−14.042 ***	−12.972 ***
	(−6.577)	(−6.561)
Dig	−0.028	9.427	0.102	18.195 *
	(−0.161)	(0.810)	(0.688)	(1.954)
$\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}\times \mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{it}}$	−21.488 ***	−18.917 ***
	(−6.212)	(−5.548)
$\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}^2\times \mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{it}}$	65.949 ***	58.785 ***
	(5.722)	(5.203)
GVCpt_b			12.257 ***	12.021 ***
			(12.634)	(10.983)
${\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_\mathrm{b}}_{\mathrm{i}\mathrm{t}}\times \mathrm{D}\mathrm{i}\mathrm{g}$			−8.271 ***	−8.387 ***
			(−4.119)	(−3.987)
control variable	NO	Yes	NO	Yes
Industry fixed	Yes	Yes	Yes	Yes
Fixed year	Yes	Yes	Yes	Yes
N	687	687	687	687
R-squared	0.975	0.977	0.981	0.982

Note: ***, * represent significance levels of 1% and 10%, respectively. The numbers in parentheses represent t-values, as shown in the table below. All regressions include industry and year fixed effects and use robust standard errors.

represent the moderating effect of industrial digitalization development level on the forward participation of GVC and the implied carbon emission intensity of China’s industrial sector trade. The regression results in columns (3) to (4) represent the moderating effect of the industrial digitalization development level on the backward participation of GVC and the implied carbon emission intensity of China’s industrial sector trade. Haans (2016) [41] provided a detailed interpretation of the discriminant method for quadratic moderation effects. Based on this method, it can be inferred that, if the product of the coefficients before $\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}$ and $\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}^2\times \mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{it}}$ minus the product of the coefficients before $\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}^2$ and $\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}\times \mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{it}}$ in the model is greater than 0, it indicates that the inflection point of the inverted “u” curve has shifted to the left. If the coefficient before $\mathrm{G}\mathrm{V}\mathrm{C}\mathrm{p}\mathrm{t}\_{\mathrm{f}}_{\mathrm{i}\mathrm{t}}^2\times \mathrm{D}\mathrm{i}{\mathrm{g}}_{\mathrm{it}}$ is greater than 0, it indicates that the inverted “U” curve becomes flatter. According to the regression results in columns (1) to (2) of Table 9, it can be found that industrial digitalization shifts the inverted “U” curve of the relationship between GVC forward participation and China’s industrial sector trade-implicit carbon emissions to the left and slows down the slope of the inverted “U” curve. This indicates that, with the improvement of industrial digitalization development level, the inflection point of the inverted “U” curve between GVC forward participation and China’s industrial sector trade-implicit carbon emission intensity has been advanced, promoting the low-carbon effect of GVC forward participation. On the other hand, it also weakens the promoting effect of GVC forward participation on trade-implied carbon emissions on the left side of the inflection point of the inverted “U” curve, which to some extent verifies hypothesis 3 of this paper. According to the previous analysis, during the sample period, the forward participation of the GVC in most industrial sectors has not yet crossed the inflection point. Therefore, the development of industrial digitalization has slowed down the promotion effect of GVC forward participation on trade-implied carbon emission intensity. Meanwhile, according to columns (3) to (4) in Table 9, it can be observed that the improvement of industrial digitalization level can to some extent weaken the promotion effect of GVC backward participation on trade-implied carbon emissions, further verifying hypothesis 3. Overall, industrial digitalization significantly weakens the positive correlation between global value chain participation and carbon emissions by reducing the carbon intensity of unit GVC activities, reshaping the evolutionary path of the relationship between the two.

7. Conclusions and Policy Recommendations

This article uses the WIOD database from 2000 to 2014 to analyze the impact of forward and backward participation in the GVC on the implied carbon emissions intensity of China’s industrial sector trade and further analyzes the moderating effect of industrial digitalization. In this study, it was found that, firstly, from the overall level of industrial sectors, with the increase in forward participation in the GVC, the implied carbon emissions intensity of China’s industrial sector trade shows an inverted “U”-shaped trend, and currently most industrial sectors in China are still on the left side of the inverted “U”-shaped curve. The backward participation of the GVC will promote China’s trade-implicit carbon emissions. Secondly, from the perspective of industrial sector heterogeneity, the increase in GVC forward participation in pollution-intensive industrial sectors will reduce the implied carbon emissions intensity of trade. However, with the increase in GVC forward participation in non-pollution-intensive and non-technology-intensive industrial sectors, the implied carbon emissions intensity of trade shows an inverted “U”-shaped trend of first increasing and then decreasing. For technology-intensive industrial sectors, with the increase in GVC forward participation, the implied carbon emissions of trade show a “U”-shaped relationship of first decreasing and then increasing. However, regardless of how the heterogeneity of industrial sectors is divided, the increase in China’s backward participation in the GVC has always increased the implied carbon emission intensity of China’s trade. Thirdly, after incorporating the regulatory effect of industrial digitalization, the promotion effect of China’s backward participation in the GVC on the trade-implicit carbon emissions of industrial sectors weakens. The forward participation of China’s GVC and the inverted U-shaped inflection point of implicit carbon emissions from industrial sector export trade have advanced, and the slope on the left side of the inverted U-shaped curve has decreased, indicating that the improvement of industrial digitalization can promote the early decline of implicit carbon emissions from China’s industrial sector trade, while slowing down the growth rate of implicit carbon emissions from China’s industrial sector trade.

Based on the previous research findings, in order to address the increasingly severe green trade barriers in developed countries and promote high-quality development of China’s trade, this article further proposes the following policy recommendations: Firstly, to transform the development mode of foreign trade and enhance the level of participation in the international division of labor. For a long time, China has mainly relied on its traditional endowment advantages such as cheap labor, land, and natural resources to participate in the international production division of labor. This way of developing the economy at the expense of the environment is difficult to obtain more benefits from the global value chain. At present, the “talent dividend” has become one of the powerful driving forces for China’s current and future economic transformation and development. Therefore, it is necessary to further tap into China’s growing “talent dividend”, cultivate new foreign trade competitive advantages of the “talent dividend”, update the traditional foreign trade model of “processing trade”, promote China’s technological innovation and production efficiency improvement, drive China to move towards the middle and high-end links of the GVC, improve China’s participation in the GVC division of labor, and thus reduce the implicit carbon emissions of products at the doorstep of various industries (Qian et al., 2021) [47]. Secondly, by deeply integrating into the regional value chain, we can further enhance China’s participation in the global value chain. As the “world factory”, China should make full use of open platforms such as the “the Belt and Road” and RCEP, give full play to its economic size and industrial complementary advantages with the countries along the “the Belt and Road” and RCEP member countries, participate more deeply in the regional value chain, and drive the technological progress of other countries in the region while strengthening its own leading ability. Taking this as an opportunity, we aim to build industrial clusters with international competitiveness, encourage enterprises to accelerate their learning and absorption of international technology while increasing research and development investment, cultivate their own innovation capabilities, and enhance their core competitiveness. Thirdly, to accelerate the construction of digital infrastructure and enhance the level of industrial digitalization. Digital infrastructure is the foundation for the digital development of industries, but the construction of digital infrastructure has the characteristics of high costs, long construction periods, and long investment return periods. It is difficult to achieve significant results in the short term solely through the participation of enterprises. Therefore, the government should take the lead in formulating and refining support policies for digital infrastructure while stimulating social capital participation, in order to improve the level of digital infrastructure construction and promote the digital development of industries. At the same time, relevant industrial policies and tax incentives will be introduced to encourage enterprises to use industrial digitalization to enhance their level of smart production and manufacturing, guide resources to cluster in low-carbon and efficient fields, improve energy management efficiency, reduce energy intensity, enhance production efficiency, and promote the green and high-quality development of China’s industrial sector export trade.

8. Shortcomings

This study has the following limitations: Firstly, the limitation of endogeneity processing. This study used the WIOD World Input Output Table (2010–2014) to construct the core variables, mainly due to its advantage of having multidimensional economic and social account data (such as energy consumption and employment structure), which can effectively control the interference of confounding factors on the regression results. Although endogeneity interference is reduced by incorporating control variables as much as possible and using GMM method for endogeneity testing (limited by the availability of instrumental variables), endogeneity risks caused by potential issues such as omitted variables may still exist. The second is the constraint of data timeliness. The WIOD database used by the research institute (2010–2014) failed to cover China’s digital acceleration period after 2015 (such as the deepening of the “Internet plus” policy), which made it impossible to quantify the dynamic impact of forward participation in the global value chain on carbon emission intensity through graphics or numerical methods. The third issue is the lack of instrumental variables. There is a lack of strong exogenous instrumental variables in global value chain participation, and, although theoretical analysis has found that industrial digitalization indicators have not yet shown significant endogeneity, further verification of their reliability is still needed through instrumental variables. The improvement direction for future research is to expand the data period to after 2015 (pending database updates) and to explore more effective instrumental variables to optimize endogeneity testing.