How GVC Embeddedness Affects Firms’ Innovation Level: Evidence from Chinese Listed Companies

1 Introduction

With the development of production and trade globalization, the global value creation system has experienced unprecedented vertical separation and restructuring on a global scale (Gereffi et al., 2005). The development of global value chains (GVCs) has promoted international trade and cross-border investment, dispersing production stages of individual products across different countries (Johnson, 2018), with countries benefiting from the dividends brought by GVCs (Gereffi, 2014). The expansion of GVCs and the intensification of direct and indirect competition among economies have made participation in GVCs an important factor influencing economic development (Auer et al., 2017). Furthermore, the level of enterprise innovation is a key factor for business development, with its importance to enterprises being crucial (Fernandes et al., 2013). Enterprises continuously seek product upgrades and technological advancements to explore new markets (Paladino, 2007). China is an indispensable part of the world trade and global governance system. According to data from the General Administration of Customs of China, China’s import and export volume exceeded $4 trillion in 2022, maintaining its position as the world’s top goods trading nation for six consecutive years. Through deep involvement in the global production network, an innovation consciousness is gradually emerging in China (Li et al., 2020). To improve or resolve the relationship between traditional business and government practices, as well as to meet the urgent need for innovation and entrepreneurship in China (Ahlstrom et al., 2018), China has adopted “mass entrepreneurship and innovation” as a national strategy for economic structural adjustment. In recent years, innovation has been seen as a key driver of China’s economic growth (Salike et al., 2022). According to McKinsey data, China needs to achieve 2–3% annual GDP growth directly through innovation and new enterprises to maintain 5.5–6.5% annual GDP growth in the future (Woetzel et al., 2015). Given China’s position among emerging economies, discussions on China’s issues hold universal significance for developing countries and emerging economies to address problems (Peng et al., 2008).

Influenced by the “low-end lock-in effect” and “innovation spillover effect,” past studies hold two views on the innovation level of enterprises in the global production network, either positive or negative (Ito et al., 2023; Li et al., 2019). This article challenges the unidirectional linear relationships in past literature and validates the non-linear impact of enterprise participation in GVCs on innovation levels.

This article quantifies the innovation levels of Chinese listed companies from 2007 to 2016 by constructing an evaluation system for enterprise innovation levels and comprehensively summarizes the relationship between the two. Drawing on the method of Upward et al. (2013) and based on the OECD’s World Input-Output Tables, the article decomposes the GVC embeddedness of Chinese listed companies and analyzes the forward and backward linkages of enterprises. Finally, market forces and technological progress are indispensable factors in the evolution of GVCs. This article explains the transmission mechanism through which firms enhance their innovation levels by participating in GVCs from the perspectives of marketization and digitalization.

The structure of this article is as follows. The next section reviews the measurement methods of the degree of participation in GVCs and research related to innovation and GVCs, proposing theoretical hypotheses based on the framework of previous literature. The third section briefly introduces the data measurement method and sources, the fourth section presents our empirical framework and regression results, the fifth section tests the mediating effects of marketization and digitalization, and the final section discusses our main conclusions.

2 Literature Review

Firm innovation is a critical topic in the modern business environment (Kline & Rosenberg, 2009). With the constant changes in the market and intensified competition, companies must continuously seek innovative ways to maintain their competitive advantage (Qiu et al., 2020). Various scholars have used different measures to assess the level of innovation in companies, such as patents (Yi et al., 2021), R&D investment (Usai et al., 2021), and Tobin’s Q ratio (Thakur-Wernz & Wernz, 2022). However, due to the high-risk nature of technological innovation activities, effectively translating R&D inputs into innovation outputs can be extremely challenging, leading to potential overestimation of a company’s technological innovation capability when using such indicators (Zhu et al., 2020). Therefore, it is necessary to construct a reasonable evaluation system to quantitatively measure the level of innovation in companies.

Currently, research on GVCs mainly focuses on two areas (Antràs, 2020). The first aspect is the measurement of GVCs, which mainly focuses on the participation and division of labor of countries and regions within GVCs. The basic method is the decomposition of value-added in exported products, which is based on input-output models among industries worldwide. Vertical specialization is used to measure the degree of participation of a country or region in the international division of labor (Hummels et al., 2001). Many scholars attempt to explain trade accounting issues, trace the sources of value-added in total exports, and study the division of labor in value chains or trade gains (Hummels et al., 2001; Koopman et al., 2014). Fally (2012) suggests measuring production segmentation by the number of production stages from commodity production to consumption. Antràs et al. (2012) introduce the concept of “upstreamness” and measure the depth of participation in international division of labor by comparing the distance of various industry products to the final product, yielding results consistent with Fally (2012). In 2017, Wang et al. (2017a, b) further expanded the accounting framework of GVCs to production stages, decomposing value-added trade from the perspectives of intermediate input supply (forward linkages) and usage (backward linkages), and analyzing the comprehensive characteristics of national and sectoral embeddedness in GVCs from the aspects of participation, location, and competitiveness.

The second aspect of research focuses on the economic impacts of countries and regions deeply embedded in GVCs. Studies on the economic impact of value chains can be summarized into two levels: macro and micro. At the macro level, research suggests that being embedded in GVCs can effectively narrow the technology gap (Ye et al., 2020) and promote labor employment and human capital development (Shepherd, 2013). Furthermore, the degree of embedding in GVCs has varying impacts on food security (Lee et al., 2012) and regional development (Boudreau et al., 2023). Particularly for the least developed countries, their participation in GVCs can effectively improve productivity and enhance innovation capabilities (Flentø & Ponte, 2017; Morrison et al., 2008; Zhan, 2021). One reason is that the export of capital-intensive intermediate products plays an important role in the economic development of the least developed countries (Suvannaphakdy & DiCaprio, 2021).

Research on the micro-level relationship between GVCs and innovation suggests that joining GVCs can improve firm productivity and promote innovation over a certain period (Opazo-Basáez et al., 2022). Qian et al. (2022) used provincial-level data and concluded that participating in GVCs is an important factor influencing firm efficiency and has a positive impact. However, some studies argue against excessive firm integration into GVCs, suggesting that it can exacerbate risks and uncertainties. Gereffi (2011) points out that participating in GVCs means facing deeper international competition, which leads to a long process of transformation and upgrading for emerging economies. Van Assche (2017) argues that firms face significant challenges and costs in establishing GVCs, which can weaken the positive link between GVCs and innovation. Hummels and Lee (2018) suggest that firms deeply embedded in value chains in a region may face higher import competition, leading to lower incentives for R&D, and more firms may choose outsourcing, sacrificing their own innovation capabilities. Ito et al. (2023), using Japanese patent data as a measure of innovation, find that backward participation in GVCs has a negative impact on innovation, while forward participation has a positive impact. Although the relationship between innovation and GVCs has been widely discussed in existing research, the conclusions remain inconsistent. Therefore, this study uses Chinese listed companies as a sample to explore the impact and mechanisms of GVC integration on firm innovation levels. Furthermore, it examines the evolution of GVCs from the perspectives of market forces and technological progress, thereby shedding light on the changes in firm innovation levels.

3 Theoretical Analyses

In modern business theory, technological progress and market forces are key factors driving enterprise innovation. With the rise of Industry 4.0, the external competition in the market and internal digital transformation have become important factors in the analysis of the GVC (Arenkov et al., 2019). The digital transformation of enterprises has a significant impact on deepening their participation in the GVC (Gopalan et al., 2022). This is because digital technology can disrupt the positioning and organizational methods of activities within the GVC, enabling the acquisition of additional value and promoting the enhancement of innovation levels within enterprises (Strange & Zucchella, 2017). The Industry 4.0 revolution further emphasizes digital technology as a core component (Mubarak & Petraite, 2020). Technological advancements represented by digital technology have significantly improved the production efficiency and competitiveness of firms participating in GVCs, thereby promoting their innovation progress (Frank et al., 2019).

Furthermore, there is ample evidence to show that multinational enterprises face increasing competition within the GVC, which has a significant impact on innovation. On one hand, the dependence on overseas investment reduces the competitiveness of enterprises in the global market, thereby affecting innovation and value chain upgrading (Raj-Reichert, 2020). On the other hand, market-oriented competition within the GVC leads to a decline in employment rates in the manufacturing sector with smaller polarizations, ultimately affecting innovation efficiency (Breemersch et al., 2017). Following China’s integration into the GVC, the share of overseas trade by Chinese manufacturers in global export trade has sharply increased, creating a strong import competition effect for manufacturers and subsequently altering their innovation decisions (Bloom et al., 2015). However, in recent years, influenced by changes in the global geopolitical landscape, the risks faced by enterprises in GVCs have been gradually rising. These risks can have a serious impact on trade liberalization and corporate governance, reducing enterprises’ ability to profit in the global production network and further affecting their innovation level (Anderer et al., 2020). This article summarizes the above risks for analysis under the themes of digitization and marketization.

Assume that there are N firms deeply embedded in GVC, there are M products participating in GVC, and the demand for product i in the market is q i . The consumer utility function is:

where α 1 and α 2 are non-negative parameters. The consumer will choose to maximize consumption utility when the consumer faces a product price of P i :

Based on equation (2), the consumer’s inverse demand function can be obtained:

where λ = ∫ 0 M U ′ ( q i ) q i d i > 0 is a Lagrange multiplier that reflects the degree of competition and deepening of the market. Equation (3) shows that when faced with deep market-based competition, firms that do not have a low-cost advantage are at risk (Mondliwa et al., 2021). Assume that the firm’s profit function is as follows when the firm faces marginal cost c and degree of marketization λ :

Based on the production theory, the derivation of equation (4) can be obtained:

The optimal level of output is obtained at this point:

Equation (6) shows the optimal output of the vendor, and the price under optimal output is:

When the degree of marketization is further deepened, manufacturers face higher risks and lower market prices, and some of them gradually withdraw from the market at this time (Figure 1).

Figure 1

Elevated risk from marketization.

The profit function of the firm at the optimal level of output is:

Equation (8) indicates that when facing market-driven exogenous shocks, firms can increase profits by reducing costs. Additionally, digital technologies can effectively facilitate the transformation of traditional industry dynamics and promote collaborative innovation within the value chain (Agostini et al., 2020). Digitalization also plays a moderating role in cost adjustments (Wang & Bai, 2021). Therefore, it is important to incorporate digitalization into the analysis. We therefore further assume that the average cost of the industry is c ̅ and the firm’s innovation level is I .

where D denotes the degree of digital transformation of the firm, both μ and θ are sensitivity coefficients, and μ , θ ∈ [ 0 , 1 ] . As more and more companies become deeply embedded in GVC, there is an additional cost to the digitization as well as marketization of companies. Also, after a firm joins GVC, intense marketization competition reduces the firm’s drive to innovate, which in turn reduces the benefits from firm innovation. Therefore, the total cost function of the firm can be obtained as follows:

Here, C ( D ) measures the firm’s investment in digitization, C ( λ ) denotes the additional cost that the firm pays in market-based competition, and C ( φ ) denotes the other unobserved cost. Since firms have a propensity to digitally transform and intensify competition, it is assumed that | C ′ ( D ) | < 1 and | C ′ ( λ ) | < 1 . We can obtain the total industry profit function as:

The optimal innovation level can be obtained by substituting equations (8) and (9) into equation (11) and taking a first-order partial derivation of the innovation level I :

In equation (12), the left and right terms represent the marginal returns to innovation as well as the marginal costs of innovation.

The nonlinearity of firms’ optimal innovation level can be derived from equations (13) and (14). As the degree of GVC participation changes, firms will adjust their innovation level based on their MRI and MCI.

From the perspective of open innovation theory, GVC participation provides firms with opportunities for cross-border knowledge flows and technological collaboration, enabling them to acquire and integrate diverse external knowledge resources, thereby enhancing their innovation capabilities. From the perspective of absorptive capacity theory, GVC participation offers firms abundant opportunities for knowledge transfer; however, the extent to which these external knowledge resources can be effectively utilized depends on the firms’ absorptive capacity. Additionally, GVC participation influences innovation through two key mechanisms: first, by facilitating resource acquisition (e.g., technology, capital, and talent), which supports firms’ innovation activities; and second, by exerting competitive pressure from international markets, which drives firms to accelerate their innovation efforts. However, as the level of participation deepens, excessive reliance on GVCs may lead to technological lock-in and innovation path dependency, ultimately hindering innovation efficiency. Based on these mechanisms, we hypothesize a non-linear relationship between GVC participation and firms’ innovation levels, which will be further validated in the empirical analysis.

The hypothesis states that the initial optimal innovation level of the firm is I 1 . Figure 2 illustrates that when the firm becomes deeply embedded in the GVC and undergoes digital transformation, it leads to an increase in the firm’s marginal innovation costs, thereby affecting the firm’s innovation level. In this scenario, the firm’s optimal innovation level decreases from I 1 to I 2 . Due to the positive incentive for digital transformation ( | C ′ ( D ) | < 1 ), even with a lower optimal innovation level, the firm will continue its digital transformation. At this point, the MRI curve rises, and the firm’s optimal innovation level increases from I 2 to I 3 .

Figure 2

The influence of digitization.

Furthermore, let’s analyze the impact of marketization on firms’ innovation levels during the process of GVC embeddedness, as shown in Figure 3. In the short term, as the level of marketization increases, firms face intense competition, leading to a decrease in innovation incentives. Consequently, the firm’s optimal innovation level decreases from I 1 to I 4 . Influenced by intense competition, some firms without low-cost advantages exit the market. However, the firms that possess low-cost advantages and survive will experience a positive innovation-driven incentive in the long run. This will result in an increase in the marginal returns on innovation and cause the MRI curve to shift upward.

Figure 3

The influence of marketization.

Based on the analysis, we can formulate Hypothesis 2.

4 Description of Data Sources and Variables

Due to the occurrence of the China–US trade war in 2018 and to avoid the impact of data fluctuations (Wu et al., 2021), this study analyzes data from the years 2007–2016. The data for this study were obtained from the WIND database, OECD database, China Customs database, and China Industrial Enterprise database. To ensure the integrity and validity of the sample, the following procedures are employed: firstly, exclusion of data from ST and ST* listed companies; second, exclusion of financial and real estate enterprises; third, removal of samples with missing primary variables; fourthly, conducting a 1% two-tailed trimming of continuous variables to eliminate the influence of extreme values on empirical analysis.

The specific indicators used in this study are as follows:

5 Empirical Analysis

5.2 Robustness Tests

To ensure the robustness of the results and demonstrate the applicability of the U-shaped non-linear relationship between GVC integration and firm innovation level, we will conduct robustness tests in this section.

5.2.1 Replacing Explanatory Variables

Upward et al. (2013) assumed that the foreign value added in a firm’s exports originates from imported or indirectly imported intermediate inputs when calculating GVC participation. However, a portion of the domestically sourced inputs used by firms may contain foreign components. Koopman et al. (2012) estimated this share to be between 5 and 10%. In this study, we will recalculate GVC participation based on this perspective.

(17) GVCPT = M A P + X O M Am O ( D + X O ) + 0.05 M T − M A P − M Am O ( D + X O ) X ,

where M T represents the intermediate input of the enterprise. This equation assumes that domestic intermediate inputs of the enterprise include 5% of foreign value added. The results of Models 1 and 2 in Table 4 demonstrate that even when changing the measurement method of the core variable, GVC participation, the U-shaped non-linear results are still obtained. The baseline regression in this article is robust.

Table 4

Robustness test

	(1)	(2)	(3)	(4)	(5)
	Koopman et al. (2012)		PatentsSum		HDFE of provinces
	Linear	Non-linear	Linear	Non-linear
GVCPT_NEW	−0.005809	−1.036766***
	(0.02285)	(0.11762)
GVCPT_NEW_2		1.670100***
		(0.19355)
GVCPT			456.680	−1794.714*	−0.879***
			(418.76636)	(1025.13267)	(0.18281)
GVCPT_2				2833.389**	1.699***
				(1177.70032)	(0.21140)
Control variables	YES	YES	YES	YES	YES
N	10,400	10,400	6,818	6,818	10,229
R 2	0.074	0.098	0.13	0.13	0.43
Firm FE	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES
Province FE	NO	NO	NO	NO	YES

Note: The values in parentheses are the robust standard errors. The symbols *, **, and *** indicate significance at the 10, 5, and 1% levels, respectively. Due to the excessive redundancy of control variables, these variables are omitted in this table.

5.2.2 Replacing Dependent Variables

This article replaces the dependent variable of innovation level based on DEA with the total number of patents for robustness testing. While the DEA model provides a comprehensive measure of innovation efficiency, the number of patents reflects the scale of innovation output. The results of Models 3 and 4 indicate that the core conclusions of this article are once again supported.

5.2.3 High-Dimensional Fixed Effects of Provinces

Provincial fixed effects effectively control regional differences such as policy environment, market conditions, and technological development levels, which can significantly affect firm innovation levels. By incorporating fixed year and firm effects, followed by provincial fixed effects, we can further eliminate the interference of regional heterogeneity on the results, ensuring that the non-linear relationship we capture is driven by the intrinsic link between GVC participation and innovation levels, rather than regional differences. The results of Model 5 show that the high-dimensional fixed effects still confirm the core conclusions of this article.

5.2.4 U-Shaped Curve Test Based on Discontinuity Regression

To further test the U-shaped relationship, we conducted a discontinuity regression to measure the regression patterns on both sides of the turning point. The preliminary regression results indicate that β 1 and β 2 have opposite signs and are statistically significant. At this point, we define the lowest point of the U curve as X min .

(18) X min = − β 1 2 β 2 .

Interrupted variable from X min

(19) X high = X − X min if X ≥ X min , otherwise 0 ,

(20) X low = X − X min if X ≤ X min , otherwise 0 ,

(21) High = 1 if X ≥ X min , otherwise 0 .

Regression models were constructed based on the above analysis.

(22) FIL it = a + b X high + c X low + d High + ε it .

The results of the discontinuity regression are as follows:

Based on the regression results an image can be plotted as shown in Figure 4:

Figure 4

Nonlinear U-test.

The regression results in Table 5 demonstrate that the regression coefficient on the left side of the breakpoint is negative, while the regression coefficient on the right side of the breakpoint is positive. Moreover, both regression coefficients on the two sides are highly significant. This robustly confirms the U-shaped relationship (Lind & Mehlum, 2010). The overall low range of the fitted curve also confirms the relatively low level of innovation among Chinese firms.

Table 5

Discontinuity regression for U-test

	(1)
	Discontinuity regression model
X_high	4.82***
	(0.57235)
X_low	−4.821***
	(0.58863)
High	−0.341***
	(0.10808)
Constant	−2.523397***
	(−35.82)
N	10,777

Notes: *** denotes significance level with p < 0.01. Numbers in parentheses are standard errors.

5.3 Heterogeneity Analysis

The basic regression results indicate a non-linear relationship between GVC participation and firm innovation level, exhibiting an overall trend of initial inhibition followed by promotion. Considering the significant variations in the depth of GVC participation among firms with different characteristics, this disparity may result in different manifestations of the innovation level brought about by GVC participation across different companies. To conduct a more in-depth analysis, this study divides the sample into two levels: company characteristics and bidirectional participation in the GVC, for further subgroup analysis. The overall research logic is shown in Figure 5.

Figure 5

Heterogeneity analysis.

5.3.1 Forward-Backward Linkage Heterogeneity

Although direct measurement of the pre- and post-participation in the GVC of firms is currently limited due to data availability constraints, it can be indirectly estimated through the sales of intermediate and final goods. A firm participating in the GVC may have four types of participation: importing intermediate goods, exporting intermediate goods, importing final products, and exporting final products. Based on Ju and Yu (2015), two measures of value chain participation length are constructed depending on the firm’s mode of participation in the GVC. Data on intermediate and final goods trade are sourced from the China Customs database, which distinguishes trade types and import-export transactions.

The first step calculates the forward Ricardian traditional trade production participation:

(23) GV C fwd i RT = ∑ j = 1 N GV C fwd i RT × C o n Trade ij C o n Trade i ,

where C on _Trade ij denotes the final goods import or export of enterprise i in industry j , and C on _Trade i denotes the total import and export of final goods of enterprise i . The second calculation is the participation of the forward intermediate goods trade production.

Second, we calculate the degree of production participation in forward intermediate goods trade.

(24) GV C fwd i M = ∑ j = 1 N GV C fwd i M × I n t Trade ij I n t Trade i ,

where I nt _Trade ij denotes the import or export of intermediate goods of enterprise i in industry j , and C on _Trade i denotes the total amount of imports and exports of intermediate goods of enterprise i .

As a result, the forward linkage of a firm’s participation in the GVC can be obtained as the weighted average of two components.

GVC _fwd i = ω RT GVC_ fwd i _RT + ω M GVC_ fwd i _ M .

Similarly, we can calculate the backward value chain production participation for different trade modes.

(25) GV C b i RT = ∑ j = 1 N PL v i RT × C o n Trade ij C o n Trade i ,

(26) GV C b i M = ∑ j = 1 N PL v i M × I n t Trade ij I n t Trade i ,

(27) GVC _ b i = γ RT GV C fwd i RT + γ M GV C fwd i M .

Based on this calculation method, a longer forward production chain for a firm indicates that it is further away from the final consumption stage and positioned more upstream in the value chain. Conversely, a longer backward production chain for a firm implies that it has more upstream production stages and is further away from the production end, indicating a downstream position in the value chain. By replacing the overall GVC participation of the firm with the calculated results, regression analysis can be conducted in Table 6.

Table 6

GVC forward and backward participation heterogeneity test

	(1)	(2)	(3)	(4)
Variables	FIL
GVC_fwd	0.007874	−1.195568***
	(0.03686)	(0.13783)
GVC_fwd2		4.455038***
		(0.54817)
GVC_b			−0.142359***	−1.801483***
			(0.04421)	(0.19853)
GVC_b2				5.088661***
				(0.63855)
Control variables	YES	YES	YES	YES
N	10,777	10,777	10,777	10,777
R 2	0.074	0.0942	0.076	0.112
Firm FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES

Note: The values in parentheses are the robust standard errors. The symbol *** indicate significance at the 1% levels.

After conducting group regression analyses on forward and backward participation in the GVC, Models 1 and 2 indicate that the impact of forward participation on firm innovation levels exhibits a nonlinear effect, presenting a U-shaped curve. The results of Models 3 and 4 show that while the U-shaped trend for backward participation in the GVC is significant, the linear term is negatively correlated with firm innovation levels. Therefore, this article concludes that backward participation is negatively correlated with firm innovation levels.

These results indicate that in the initial stages, as firms engage in more forward activities in the GVC, their innovation level significantly decreases. However, once the firm’s forward participation reaches a certain threshold, further increasing the level of forward participation has a positive impact on innovation, promoting an increase in innovation level. This suggests that in the initial phase, firms may focus more on acquiring and applying technology and knowledge from other participants in the global supply chain. This process of technological catch-up and application may somewhat suppress the development of independent innovative capabilities. However, as firms engage more deeply in the GVC, they accumulate more knowledge and skills through collaboration and exchange with other participants. This knowledge accumulation and skill transfer provide a better foundation for subsequent-stage innovation, thereby enhancing the overall innovation level.

In contrast, backward participation in the GVC shows a significant negative linear relationship. This implies that as firms participate in more backward activities in the GVC, their innovation level may decline. Over-reliance on backward participation can lead to a lack of independent innovative capabilities and excessive dependence on the technological and knowledge inputs from other stages, thereby restricting the firm’s ability to innovate. Additionally, backward participation may involve more production and operational aspects, where firms primarily play a role in assembly and integration rather than in research and development. This also explains why backward participation is negatively correlated with innovation level.

5.3.2 Heterogeneity in the Nature of Firms

Different firm characteristics have a certain influence on firm innovation (Cohen & Klepper, 1996). Based on these differences, firms are classified into three categories: private enterprises, state-owned enterprises, and foreign-invested enterprises. The study investigates the impact of firm integration into the GVC on the innovation level. The specific results are shown in Table 7.

Table 7

Tests for heterogeneity in the nature of firms

	(1)	(2)	(3)	(4)
Variables	FIL
	State-owned	Private	Foreign-invested (Linear)	Foreign-invested (Non-linear)
GVCPT	−1.113220**	−0.739811**	0.940795	−1.959638
	(0.54423)	(0.33287)	(0.60034)	(1.29658)
GVCPT_2	2.161045***	1.444815***		3.398207**
	(0.78520)	(0.44161)		(1.48518)
Control variables	YES	YES	YES	YES
N	3,188	6,823	313	313
R 2	0.126	0.099	0.183	0.205
Firm FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES

Note: The values in parentheses are the robust standard errors. The symbols **, and *** indicate significance at the 5, and 1% levels, respectively. Due to the excessive redundancy of control variables, these variables are omitted in this table.

The results of Models 1−4 indicate that the U-shaped trend is once again validated in private and state-owned enterprises. However, in foreign-invested enterprises, the U-shaped trend and linear relationships are not significant. This may be because foreign-invested enterprises often undertake high-end tasks in the GVC, such as research and development, design, or brand management, while enterprises located in China may be more involved in low-end manufacturing. Therefore, the marginal benefits that foreign-invested enterprises gain from GVC participation may be small and not significantly change their innovation levels. This high starting point leads to an inconspicuous impact of GVC participation on their innovation.

Compared with state-owned enterprises, private enterprises usually face shortages of funds, technology, and policy support. Therefore, they are easily squeezed in the low-end stage of the GVC and fall into the “lock-in effect,” which is reflected on the left side of the U-shaped curve as a low point in innovation capability. On the right side of the U-shaped curve, private enterprises have strong market adaptability and flexibility and are more willing to actively carry out digital transformation and market-oriented reforms. These measures can effectively make up for the short-term negative effects of GVC participation and promote enterprises to quickly climb to high-end links.

5.3.3 Heterogeneity in Industry Intensity

There are significant differences in the innovation models, technological requirements, and market competition environments across industries. Industry intensity typically determines the role that firms play in the GVC and the nature of their innovation activities. Classifying industry intensity helps identify and analyze the different impact mechanisms of GVC participation on innovation in industries with varying levels of intensity.

The results of the Table 8 show a U-shaped trend in technology-intensive industries, insignificant linear and nonlinear relationships in capital-intensive industries, and a significant positive linear relationship in labor-intensive industries. The reasons for these patterns may be as follows:

Table 8

Tests for heterogeneity in industry intensity

	(1)	(2)	(3)	(4)	(5)
Variables	FIL
	Technology-intensive	Capital-intensive		Labor-intensive
		Linear	Non-linear	Linear	Non-linear
GVCPT	−1.325313***	0.448160	−0.217398	0.467818**	−0.433021
	(0.34197)	(0.27572)	(1.04006)	(0.21564)	(0.64172)
GVCPT_2	2.302760***		0.787167		1.256628
	(0.47168)		(1.38053)		(1.01685)
Control variables	YES	YES	YES	YES	YES
N	5,823	1,830	1,830	3,060	3,060
R 2	0.111	0.118	0.119	0.089	0.091
Firm FE	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES

Note: The values in parentheses are the robust standard errors. The symbols **, and *** indicate significance at the 5, and 1% levels, respectively. Due to the excessive redundancy of control variables, these variables are omitted in this table.

First, technology-intensive industries rely on high-tech and knowledge-intensive innovation processes. Initially, low GVC participation may limit firms’ access to external advanced technologies and knowledge. However, as GVC integration deepens, firms can gain more opportunities for cross-border technology collaboration and knowledge transfer, leading to improved innovation levels. Hence, the relationship in these industries is U-shaped.

Second, in capital-intensive industries, innovation is mainly driven by capital expansion and production capacity rather than external technological or knowledge inputs. GVC participation may have a smaller direct impact on innovation, or the innovation path in these industries is constrained by changes in capital equipment and production methods, which fails to show a distinct nonlinear trend.

Lastly, in labor-intensive industries, innovation mainly focuses on improving production processes, cost control, and management efficiency. GVC participation provides greater market demand and resource acquisition opportunities, promoting continuous investment and improvement in innovation, which leads to a sustained positive relationship.

6 Impact Mechanism Analysis

6.1 Theoretical Mechanisms Identification

The analysis of the above sections indicates that deep integration of enterprises into the GVC has a nonlinear impact on innovation levels, and this effect exhibits significant heterogeneity due to the different degrees of pre- and post-participation and the nature of the enterprises. In this section, in order to examine the channels through which the GVC integration affects the innovation levels of enterprises, the analysis framework incorporates two channels: digitalization and marketization, to analyze the mechanisms that lead to the nonlinear relationship. According to Hypothesis 2, the transmission mechanism is illustrated in Figure 6.

Figure 6

Intermediate channeling mechanism.

Digitalization entails significant transformation costs in the early stages. Additionally, when enterprises participate in the GVC, they face intense market competition, which can lead to a decline in innovation levels. However, the formation of economies of scale and the spillover effects brought about by later-stage digital dividends provide positive innovation impetus for enterprises, thereby enhancing innovation levels. Therefore, this study further analyzes the mechanisms through which marketization and digitalization affect the innovation levels of enterprises participating in the GVC.

6.1.1 Digitization Mechanisms

The deep integration of the digital economy and the traditional economy makes it difficult to measure the scale and degree of digitalization in the digital economy, and there are also controversies regarding various viewpoints (Ragnedda et al., 2020). In this study, text analysis and machine learning methods are employed to construct a digitalization index based on corporate annual reports, aiming to overcome the limitations of existing measures of corporate digitalization.

First, based on highly digitized corporate annual reports and digital literature, keywords are extracted using a comprehensive consideration of the common features of words. High-frequency words are then selected as the digitalization lexicon. This approach ensures that the index measurement is not influenced by sample data. After extracting the literature and annual reports, the total word frequency of all company annual reports is further extracted. After standardization, the digitalization index, referred to as Digital, can be obtained.

6.1.2 Marketization Mechanisms

The indicators used to measure the degree of marketization of enterprises in the international market exhibit diversity, but revenue from operations is considered the most effective measure of market competitiveness for enterprises (Hou et al., 2021). Overseas market profits reflect a company’s performance and competitiveness in the international market. High overseas market profits indicate that the company’s products are attractive in the global market, meeting the needs of different countries and regions, thereby enhancing its competitive position in GVCs. Therefore, this study selects the overseas business revenue of enterprises in the CSMAR database for the current year to reflect the level of marketization within the GVC.

7 Endogeneity Test

Due to the high likelihood of endogeneity issues in the nonlinear U-shaped curve, it is challenging to find new instrumental variables that satisfy both exogeneity and relevance. However, in order to mitigate estimation biases caused by omitted variables or reverse causality, this study employs different estimation methods to address the endogeneity issue. Given that the innovation of firms in the GVC may not have an immediate effect and exhibits time lags, the lagged value of GVC participation is chosen as an instrumental variable. This variable represents that a firm’s GVC participation in the previous year may impact the innovation level in the following year. Furthermore, to further correct for unobserved individual heterogeneity, omitted variable bias, measurement errors, and potential endogeneity issues, the system GMM method is used for estimation. In the case of limited samples, the standard errors of the two-step system GMM show a significant decrease, hence this study adopts the two-step system GMM to estimate the lagged one-period and two-period dependent variables.

The relevance test indicates that the lagged one-period GVC participation meets the instrumental variable conditions. Moreover, conducting overidentification and weak instrument tests on this instrumental variable shows a P-value of 0.00 for the overidentification test, and the weak instrument test confirms the rationality of the instrumental variable selection. Continuing with the estimation using the two-step system GMM method, including the lagged dependent variables, the core variables remain significant and exhibit a U-shaped relationship. The tests for Models 2 and 3 in Table 10 reject the hypothesis of second-order autocorrelation, and the Hansen test results indicate the effectiveness of the instrumental variables. The endogeneity test results demonstrate that the relationship between GVC embeddedness and firm innovation level still follows a U-shaped pattern, confirming the robustness of the results.

8 Conclusion and Recommendations

This study uses data from Chinese listed companies from 2007 to 2016 as a sample to examine the impact of GVC integration on firm innovation levels from a division of labor perspective. The main conclusions of this study are as follows:

First, as firms’ participation in the GVC increases, firm innovation levels exhibit a “U”-shaped trend, initially decreasing and then increasing. The novelty of this article lies in integrating the “low-end lock-in effect” and “knowledge spillover effect” perspectives from the existing literature. These results remain robust even after changing estimation methods, incorporating discontinuity regressions, and using instrumental variables. From an empirical perspective, this study confirms the multifaceted nature of innovation, demonstrating that innovation within GVCs is not solely characterized by lock-in effects or knowledge spillovers. This aligns with certain existing viewpoints in the literature (Buciuni & Pisano, 2021; Choi et al., 2019).

Second, based on different modes of firm participation, this study decomposes the forward and backward linkages of firms, following the approach proposed by Ju and Yu (2015). The study also examines the heterogeneity of the impact of forward and backward participation on firm innovation levels based on firm characteristics. The results show that the impact of forward GVC participation on firm innovation levels exhibits a nonlinear effect, following a U-shaped pattern. In contrast, backward GVC participation shows a significant negative linear relationship with firm innovation levels. The heterogeneity analysis in this study confirms the conclusions of Ito et al. (2023). Moreover, due to the maturity of multinational networks and the low-end tasks undertaken by foreign-invested enterprises in China, the U-shaped relationship is not observed among foreign-invested enterprises. In contrast, a U-shaped relationship in innovation levels is evident among state-owned and private enterprises, with the U-shaped pattern being more pronounced in state-owned enterprises.

Third, marketization and digitization significantly influence firm innovation levels when firms are embedded in the GVC, and both exhibit a nonlinear U-shaped relationship. Marketization, through increased competition, and the costs associated with digitization inhibit firm innovation initially. However, the scale effects resulting from later marketization and the benefits derived from digitization enhance firm innovation levels. In the overall transmission mechanism, the marketization effect is more pronounced than the digitization effect.

Based on the conclusions of this study, the following recommendations are proposed:

First, firms should actively participate in the GVC. In the early stages of GVC integration, firms should seek to establish partnerships to achieve resource sharing and collaborative innovation. By leveraging the technological expertise and experience of partners, firms can pursue high-level technological cooperation to accelerate the innovation process and enhance their innovation levels. As firms become more deeply embedded in the GVC, they should actively establish partnerships with other firms to further promote resource sharing and collaborative innovation.

Second, firms should actively strengthen their cooperation with upstream suppliers, technology partners, and research and development institutions to enhance forward participation. This helps firms gain access to the latest technologies and market information, providing more resources and opportunities for their innovation activities. Although backward GVC participation shows a negative linear relationship with firm innovation levels, it does not imply complete disengagement from the backward linkages. On the contrary, firms should actively participate in the backward linkages, but with a greater focus on supply chain management and control. Optimizing supply chain management can improve efficiency and quality, reduce unnecessary costs, and provide more resources for innovation.

Finally, in the early stages when innovation levels are suppressed, firms should focus on enhancing their competitiveness and cost-effectiveness. By reducing costs, improving efficiency, and offering more competitive products and services, firms can strengthen their position within the GVC. As firms become more integrated into the GVC, scale and income effects will emerge in the later stages. Firms should leverage these effects to further enhance their innovation levels. By expanding scale, improving resource allocation, and optimizing supply chains, firms can achieve higher levels of innovation and growth. Governments, through policy guidance and resource allocation, should encourage firms to increase R&D investment, drive technological innovation, and support industrial upgrading. Firmly implementing an innovation-driven development strategy, building open and cooperative global networks, and formulating differentiated policies to support enterprises of different natures are essential. Additionally, optimizing industrial and supply chains, strengthening risk management, enhancing industrial chain synergy, and promoting digital transformation are critical. At the same time, strengthening international exchanges and cooperation, participating in global governance, and contributing to the creation of a more equitable and reasonable international economic order should be prioritized.

However, this article still has certain limitations. Due to the availability of data, the research period does not cover the most recent years. The study requires data on firms’ intermediate goods purchases and processing conditions, which need to be merged and cleaned with firm-level data. However, the Industrial Enterprise Database has not been updated to the latest year, so the data selected for this study are from Chinese listed companies from 2007 to 2016. As a result, there may be issues with data updates. Additionally, due to the limitations of the database in accurately tracking each firm’s intermediate and final outputs, the decomposition method used to analyze firms’ backward and forward linkages may lead to potential issues of double counting.

Future research could extend to other countries and emerging economies. For example, in India, GVC participation is primarily focused on information technology and service outsourcing. The innovation level of firms in India may exhibit a nonlinear U-shaped relationship, as initially low participation could limit the absorption of external knowledge, but with deeper GVC integration, innovation capabilities are enhanced. In Brazil, as a resource-driven economy, the innovation effects of GVC participation may present a more stable nonlinear relationship, with excessive reliance on resource-based and low-tech external collaborations possibly having a limited effect on innovation. In South Africa, where there are fewer technology-intensive industries, innovation may follow a positive nonlinear relationship with GVC participation, with innovation capabilities gradually improving as foreign investment and technology transfer accelerate. Also, we can test the heterogeneity of the nonlinear relationship between GVC participation and innovation across different economies. This would further deepen our understanding of the mechanisms behind nonlinear relationships and provide targeted policy recommendations for countries at different stages of development and with varying economic structures, thereby fostering innovation and economic development globally.