A Systems Perspective on the Embeddedness of Foreign-Invested Enterprises and Functional Upgrading: Evidence from China’s Manufacturing Sector

Abstract

This study conceptualises the local production network as a complex system composed of domestic enterprises (DOEs) and foreign-invested companies (FIEs) that are interconnected and co-evolving. We define the embeddedness of FIEs according to three types—DOEs–FIEs, FIEs–DOEs and FIEs–FIEs—and examine how the different types of FIE embeddedness influence the functional upgrading of domestic value chains. Using the 2024 OECD database on multinational enterprises’ activities, we empirically assess the embeddedness of FIEs and the functional upgrading of manufacturing industries in 14 sub-sectors from 2003 to 2020. The results show that the embeddedness of FIEs facilitates overall functional upgrading, particularly in R&D and management, though no significant effect is found in marketing. Mechanism analysis reveals that FIE embeddedness in China’s manufacturing value chain primarily drives functional upgrades through productivity and creation effects. Heterogeneity analysis shows that FIE activities of both types, “DOEs–FIEs” and “FIEs–DOEs”, positively influence R&D upgrading, while those of the “FIEs–FIEs” type promote management upgrading. In contrast, “FIEs–DOEs” activities hinder marketing upgrading. This study provides empirical evidence of the role of FIE embeddedness in functional upgrading and offers a theoretical basis to develop policies that guide foreign capital toward higher value functions.

1. Introduction

As China’s economic growth model shifts from a factor-driven stage to an innovation-driven stage, the value chain of China’s manufacturing industry has evolved from low-end processing to diversification and increased high-value inputs [1]. Thus, the functional upgrading of the manufacturing sector has become key to China’s economic transformation. The concept of functional upgrading was first introduced by Humphrey and Schmitz (2002) in summarising the advancement of global value chains, which includes process upgrading, product upgrading, functional upgrading and inter-sectoral upgrading [2]. Coveri et al. (2024) and Capello et al. (2024) more recently defined it as ‘the shift of a country’s enterprises or industries to the higher-end production functions, beyond the fabrication function, within the value chain’ [3,4]. Timmer et al. (2018) divided the functions of the value chain into four areas: R&D, management, fabrication, and marketing [5]. They highlighted that the competitiveness of developed countries in the functions of R&D, marketing and management has been improving, with the corresponding Functional Specialisation (FS) Index rising from 1.09, 1.03 and 1.01 at the end of the 20th century to 1.12, 1.06 and 1.07 in 2011, respectively. Industrialised nations such as Germany and the United States have historically been more competitive in R&D, management and marketing. In contrast, China has long maintained a comparative advantage in the fabrication function, with relatively low value-added gains. Therefore, many researchers and policymakers have called for China to upgrade its functions promptly and shift its production chain toward higher value-added activities [6,7].

In recent years, the attraction of foreign-invested enterprises (FIEs) has been widely regarded as a key way for China to integrate into the global value chain, expand exports and promote industrial transformation and upgrading [8]. According to the data from OECD, China’s foreign direct investment regulatory restrictiveness index (FIRPI) declined from 0.62 in 1997 to approximately 0.25 in 2018 [9]. The Global Investment Report indicates that the scale of China’s actual utilisation of inward foreign direct investment (FDI) increased from less than US$50 billion in 2000 to around US$190 billion in 2022, with its share of global FDI rising from less than 4% in 2000 to approximately 15% in 2022. As FDI inflows into China continue to grow, multinational firms (MNEs) have become deeply embedded in China’s industrial chain through the establishment of subsidiaries, co-production and mergers and acquisitions, cementing their position in China’s manufacturing production, purchasing and sales networks. In 2022, the total value of China’s imported goods was $2.71 trillion, of which FIEs contributed $953 billion, accounting for 35.1% of market share [10]. The total value of exported goods was US$3.57 trillion, with FIEs contributing US$1.12 trillion, accounting for 31.5%. With the growth of the FDI stock size, the participation of FIEs in China’s industrial chain activities has been deepening, representing increased embeddedness. Understanding whether the embeddedness of FIEs has an impact on the functional upgrading of China’s value chain is therefore crucial. Existing studies have not yet explored the effects of increased FIE embeddedness on China’s value chain nor the mechanisms behind it.

Numerous studies have shown that by attracting FIEs, it is possible to promote productivity upgrades and output increases in China by leveraging the technology transfer, technology spillover, competition, demonstration and vertical linkage effects of FIEs [11,12]. However, as pointed out by Van et al. (2018), such industrial upgrading, although reflected in improvements in productivity and value-adding, is primarily confined to process upgrading, which involves the introduction of new production technologies, or product upgrading, which focuses on improving product design and quality [13]. Different from process upgrading and product upgrading, functional upgrading emphasises the upgrading of workers’ skills and structural changes as the driving force of industrial advancement [14]. Since the key to determining the direction of functional specialisation and ultimately the impact of functional upgrading on a country’s value chain is workers and their skills, existing studies mostly focus on the impact of FIEs on product quality or export structure. However, they do not provide any theoretical basis to explain the structural changes in the value chain triggered by FIEs, thus failing to explain how the embeddedness of FIEs affects functional upgrading. In addition, existing studies do not provide insight into the relationship between the embeddedness of FIEs and the skills of workers, which determines the direction of functional upgrading [15,16,17,18]. Existing studies have been unable to elucidate whether the embeddedness of FIEs promotes or inhibits the upgrading of workers’ skills [19,20]. This study, therefore, seeks to analyse the relationship between the embeddedness of FIEs and the functional upgrading of China’s manufacturing industry.

The purpose of this study is to adopt a systems perspective to deconstruct the complex system of FIEs and DOEs and, on this basis, examine the relationship between the embeddedness of FIEs and the functional upgrading of China’s manufacturing value chain. In this study, the embeddedness of FIEs is identified based on the Activity of Multinational Enterprises (AMNE) database released by OECD in 2024. Referring to the studies by Stöllinger et al. (2021) and Kordalska et al. (2022) [16,21], measurement indicators of functional upgrading at the overall industry level and sub-functional level are constructed, so as to explore the impact of the embeddedness of FIEs on functional upgrading of the value chain. It is found that, first, the embeddedness of FIEs has a clear positive effect on the overall functional upgrading of China’s manufacturing value chain. At the sub-functional level, it inhibits the specialization of fabrication functions but has a significant positive effect on the upgrading of R&D and management functions, while its impact on marketing functions is not significant. These results remain robust to changes in estimation models and to a one-period lag of the explanatory variables. Second, the mechanism analysis shows that the embeddedness of FIEs has a creative effect on the demand for skills related to R&D and management functions, which extends the related production chain. The embeddedness of FIEs also has a productivity effect on R&D and management skills. Finally, the heterogeneity analysis shows that “DOEs–FIEs” and “FIEs–DOEs” value chain activities have a positive effect on the upgrading of R&D functions, while “FIEs–FIEs” activities positively influence management upgrading, and “FIEs–DOEs” activities negatively affect marketing upgrading.

In summary, the main contributions of this study include the following three aspects. First, in terms of research content, this study starts with examining the embeddedness of FIEs in the Chinese manufacturing value chain, analyses its impact on functional upgrading and bridges and expands two groups of literature on FIE embeddedness and the functional upgrading of value chains. In the literature on value chain upgrading, although the influence of trade, technology and other factors is fully considered, the role of FIE embeddedness is neglected. More focus is given to the characteristics of the embeddedness of FIEs and their impact on the structure of labour skills and inequality, paying little attention to impacts on functional upgrading. Second, in terms of research perspectives, this study introduces ‘task-biased technological progress’ as a theoretical mechanism to analyse the embeddedness of FIEs affecting the functional upgrading of value chains, combining the ‘technology spillover effect’ of FIEs with the ‘functional’ specialisation of the manufacturing value chain. It also examines the impact of the embeddedness of FIEs on the functional upgrading of the value chain at the overall level and in each functional dimension, thus expanding the scope of ‘task-biased technological progress’. Third, from a systems perspective, this study identifies and distinguishes three types of FIE embeddedness within the host-country production network—DOEs–FIEs, FIEs–DOEs and FIEs–FIEs—and, on this basis, explores in-depth their differentiated impacts on the functional upgrading of China’s manufacturing industry. This provides new theoretical perspectives and empirical evidence to explain the inconsistency between observations of the embeddedness of FIEs and findings from industrial upgrading-related studies [22,23].

2. Literature Review and Research Hypotheses

2.1. Literature Review

Research closely related to this study includes two groups of literature: first, the definition of functional upgrading and its influencing factors, and second, the value chain upgrading effect of the embeddedness of FIEs.

2.1.1. Defining Functional Upgrading and Its Influencing Factors

At the outset of this study, it is important to clarify the definition of ‘functional upgrading’. Although existing literature has focused on different aspects of “functional upgrading’, most of the literature has maintained a relatively unified definition, namely the process of a national enterprise or industry shifting to higher-end production functions in the value chain [3,4]. In this definition, ‘function’ usually refers to the completion of specific functions undertaken by the production links in the value chain, such as R&D, design, manufacturing, logistics, management, sales, etc., and the term ‘upgrading’ emphasises the process of shifting from low-end ‘fabrication’ functions in the value chain division of labour to high-end production functions. In short, functional upgrading refers to the specialisation of functions undertaken in the functional division of labour towards higher-end production functions. Unlike traditional skill upgrading represented by the increase in the proportion of high-skilled workers in the traditional labour market, functional upgrading emphasises the change in the skills required to perform different occupations. In most studies, this upgrading has been reflected in changes in the structure of ‘occupations’ in the labour market [24,25].

The concept of functional upgrading reveals dynamic changes in both the skill structure and the value added within an enterprise or industry’s value chain. Therefore, in order to analyse functional upgrading, it is necessary to understand the relationship between the skill structure of the labour market and the distribution of value added within it. Riccio et al. (2023) highlighted that the distribution of gains in the value chain depends on the change in the relative wages of workers with different skill sets within the labour market [26]. A visualisation of the distribution of gains from value added can be given by the ‘smile curve’, showing ‘high end and low middle’ production stages and gains from value added: the closer the link is to the consumer side (e.g., sales) or the producer side (e.g., R&D and design), the higher the gains from value added; while the link that focuses on OEM or simple assembly in the middle tends to have lower gains [27]. In the context of GVCs with integrated production processes, an increase in the share of skills related to the above high value added to a country’s skill structure usually implies skill upgrading. Therefore, it is essentially a matter of exploring the upgrading of workers’ skills to analyse the functional upgrading of value chains, and it requires paying attention to the factors influencing skill upgrading in the labour market.

Skill upgrading has been analysed from two main perspectives: technological progress and trade. The literature focusing on the impact of technological progress on skill upgrading suggests that technological progress is biased towards the relative wages of workers with different skills; namely, productivity gains from new technologies (e.g., automation, the Internet) differ across factors of production (e.g., high-skilled versus low-skilled labour) and are generally more skewed in favour of higher-skilled workers [28]. Some studies empirically analyse the biases introduced by these technologies, such as the use of computers and the number of robots [29,30,31].

The literature examining the impact of trade on skills upgrading argues that cross-border movement of factors of production triggered by economic integration centred on international trade and investment leads to changes in the relative demand for factors of production, and thus explains skills upgrading [32]. This kind of literature focuses on empirical assessments from the perspectives of the share of outsourcing activities and the location of production activities. In addition, some scholars have analysed this phenomenon from the perspective of global value chain embeddedness related to trade activities [33,34]. Overall, the established literature explains skill upgrading from the perspectives of both technological progress and trade activities in the global value chain.

2.1.2. The Value Chain Upgrading Effect of FIE Embeddedness

The embeddedness of FIEs is defined by Wang et al. as the value chain activities related to production, sourcing and sales of FIEs in host countries [35]. With the continuous advancement of globalisation, the layout of value chain activities of FIEs in host countries has become increasingly complex and diversified. On the one hand, part of the local sales of foreign affiliates of transnational corporations (TNCs) is used as intermediate inputs for local firms or as final products directly supplied to the local market to satisfy the final demand of consumers in the host country; on the other hand, the foreign affiliates of TNCs have to purchase local intermediate inputs for production. According to Wang et al. [35], within the complex production network composed of FIEs and DOEs, value chain activities related to FIEs can be understood according to three types of relational embeddedness: (1) DOE–FIE embeddedness, where the upstream value-added providers are domestic enterprises, and the downstream producers are foreign-invested enterprises; (2) FIE–DOE embeddedness, where the upstream value-added providers are foreign-invested enterprises, and the downstream producers are domestic enterprises; (3) FIE–FIE embeddedness, where both upstream value-added providers and downstream producers are foreign-invested enterprises. The structure of the embeddedness of FIEs, deconstructed from a systems perspective, is illustrated in Figure 1.

The existing literature mainly focuses on two aspects of FIE embeddedness and its effects on the value chain, namely, the technology spillover effect and ‘capital–skill’ complementarity. The former emphasises the mechanism of technology diffusion and knowledge learning in the host country, while the latter focuses on how the capital of FIEs affects the upgrading of workers’ skills and the accumulation of human capital.

First of all, when examining the effects of FIE embeddedness on the host country’s value chain upgrading, technology spillover is mainly reflected in the following three aspects. Firstly, a demonstration effect can be observed. Local enterprises can imitate or introduce the advanced technology and management methods of the FIEs and quickly improve their own production efficiency. Secondly, a competition effect exists. The entry of foreign capital often brings fierce market competition. This competitive pressure will force local enterprises to recruit highly skilled personnel and strengthen training and independent R&D to rapidly upgrade their skills, thus realising product and technology upgrades [36]. Thirdly, a personnel mobility effect takes place. FIEs often hire local employees and provide them with on-the-job training. With the normal mobility of employees, those who have experience with FIEs will bring advanced skills and management experience to local enterprises, thus promoting the diffusion of knowledge and skills.

Second, considering FIEs as a form of capital, the ‘capital–skill’ complementarity mechanism emphasises that the financial inputs, machinery and equipment and R&D support brought by FIEs in the host country complement the skills of the local workforce. This complementarity mechanism has both positive and negative impacts on value chain upgrading. In terms of negative impacts, in order to maintain their monopoly in the host country, FIEs usually force OEMs in developing countries to invest in large-scale fixed assets. They take advantage of the lock-in of specialised production investments created by FIEs to squeeze the profit margins of processors at the lower end of the value chain by consistently undercutting purchasing prices [37].

Specifically for China, many empirical studies examining whether FIEs promote or inhibit the upgrading of China’s manufacturing value chain have been conducted, but the overall conclusions are inconsistent [38]. Some scholars found that FIEs have a positive impact on innovation activities in China [39], especially through upstream and downstream pulling [40]. Another group of scholars found that FIEs can have negative spillover effects [41,42]. Van Assche et al. pointed out that existing studies on the impact of FIEs on the upgrading of the value chain are still confined to product and process upgrading, and few studies explore the relationship between FIEs and functional upgrading [13]. At the same time, relevant studies show that the impact of FIEs on skills upgrading is uncertain. Therefore, the embeddedness of FIEs and its effects on functional upgrading need to be tested empirically.

2.1.3. Research Shortcomings and Opportunities for Theory Development

Summarising the above two kinds of literature on the factors influencing functional upgrading and the relationship between the embeddedness of FIEs and value chain upgrading, the following points can be made. First, research on the factors influencing functional upgrading is still in its infancy. When focusing on skill upgrading, most of the existing studies are confined to the traditional classification of skills, such as different levels of education, productive and unproductive skills, etc., but ignore the much-talked-about functional division of labour in the value chain, as well as the distribution of skills in specialisations that undertake different functions. Second, existing research fails to take into account FIEs in domestic value chains, and there is a lack of theoretical and empirical research on the link between such FIEs in value chains and functional upgrading.

Task-biased technological progress is an ideal perspective to compensate for the shortcomings of the above research because: (1) it emphasises the bias and degree of technological progress made in different value chain ‘tasks’, which helps to reveal the specialisations and directions of the value chain functions; (2) combined with the technological spillover mechanism of FDI, it can provide in-depth analysis of the effects of the embeddedness of FIEs on functional upgrading.

First, the theory of task-biased technological progress reveals the differential impact of technological progress on wage growth in different occupations; namely, wage growth in middle-skill occupations, which are dominated by routine tasks, slows down significantly, while wage growth in high-skill and low-skill occupations occurs at a greater rate [43]. According to the theory of task-biased technological progress, enterprises’ business activities can be split into a series of tasks that can be performed either by labour or by capital (e.g., machinery and equipment). Regarding the division of tasks, based on the widespread use of automation technology since the 1980s, Autor et al. (2003) distinguished between two types of tasks based on the technical parameters of different tasks and their labour preference parameters, namely routine tasks and non-routine tasks [44]. The former are characterised by repetition, regularity, proceduralism and ease of substitution, so that the advent of automation will lead to a rapid increase in the technological parameters of these tasks, and thus to a decline in the wages of the corresponding skills; the latter involves judgment, innovation and complex interactions, and the technological parameters are slower to grow and less susceptible to technological substitution. If we shift our perspective from an emphasis on the classification of tasks under automation technology to a focus on the functional division of labour in the value chain under the functional specialisation model, then the skills corresponding to the realisation of the functional division of labour in the value chain become a new criterion for classifying tasks. Applying this criterion to the specific functions in the production process of the value chain, we can obtain the classifications proposed by Timmer et al., namely, R&D function, management function, marketing function and fabrication function [5]. It can be seen that the theory of task-biased technological progress, which emphasises the differential impact of technological progress on different skills/occupations and shapes the structure of skill upgrading based on the division of tasks, aligns with an upgrading path based on the functional division of labour. Therefore, task-biased technological progress is an effective tool for analysing functional upgrading.

Second, in conjunction with the theoretical mechanisms by which the technology spillover of FIEs and capital-skill complementarities affect skill upgrading, the theory of task-biased technological progress provides theoretical support for the idea of assessing how the embeddedness of FIEs affects functional upgrading. As mentioned above, the embeddedness of FIEs brings about the technology spillover and capital-skill complementarities that differentiate the demand for different skills, which in turn affects the direction of functional upgrading through the skill premium.

In short, the theory of task-biased technological progress provides a powerful theoretical tool for exploring the embeddedness of FIEs and value chain functional upgrading. In addition, the share of FIE-related value chain activities in the GDP of China’s manufacturing sector continues to rise, which provides an ideal empirical context to fill the above research gap.

2.2. Research Hypotheses

Applying task-biased technological progress to the functional division of labour, the main focus is on the change in the centre of gravity of FIEs in the specialisation of the functional division of labour in the demand-side value chain and the direction of its impact. Combined with the previous hypothesis about the lower added value of fabrication functions in the distribution of value-added gains, if the embeddedness of FIEs has a greater impact on the demand for fabrication than on the demand for manufacturing in China’s manufacturing value chain, the embeddedness of FIEs is conducive to the promotion of value chain function upgrades. Based on the analytical framework of task-biased technological progress, theoretically, the embeddedness of FIEs does not have a consistent effect on labour demand, i.e., there is a negative substitution effect and positive productivity and creation effects.

The embeddedness of FIEs produces a substitution effect mainly through the following two aspects. On the one hand, for a fabrication or non-fabrication function, due to the technological spillover of FIEs, work previously accomplished by the labour force can be accomplished by capital, thus reducing the demand for labour related to different functions; that is, technology substitutes labour. As the degree of embeddedness of FIEs deepens, technology spillover along the value chain will also gradually increase, eventually leading to the substitution of labourers with different skills, such as through automation, artificial intelligence, and other technologies to complete the work. Therefore, the technology spillover due to the embeddedness of FIEs can lead to the substitution of the existing labour force. On the other hand, the embeddedness of FIEs can change the relative prices of different skill factors, which in turn reshape the skill structure of the labour market and impact the distribution of specialisations in the value chain. Yuan et al. pointed out that the competition effect and the demonstration effect of FIEs are positively influenced by the application of robots by domestic manufacturers in China, which has significantly reduced labour demand in manufacturing [45]. As a result, FIEs have a greater substitution effect on the fabrication function, which may encourage China’s manufacturing value chain to focus more on the specialisation of non-fabrication functions.

In regard to the productivity effect, FIEs and local enterprises jointly constitute an important resource in the national industrial chain. The combination of personnel mobility, business cooperation and the training of local employees enables local enterprises and the labour force to learn and imitate advanced technology and management skills and then apply them in their own operations to improve productivity [46]. At the same time, FIEs utilise their monopoly to bring competitive pressure on the value chain in the same industry. This pressure may not only prompt domestically owned enterprises (DOEs) to accelerate equipment upgrades and process improvements but also force DOEs to pay attention to management, R&D and organisational and operational capacity-building.

In regard to the creation effect, in replacing some functions, foreign technology spillover will also create new functional needs and thus expand the demand for R&D, management, and other skills. For example, since the 21st century, multinational corporations have set up R&D in China and enhanced the skills of the local workforce through training, which inevitably increases the demand for R&D skills and improves the labour productivity of related skills [47]. This means that the embeddedness of FIEs usually enhances non-fabrication function productivity and increases total output. We can expand the demand for non-fabrication functions in areas where R&D, management and marketing functions need to be utilised to have a comparative advantage.

To summarise, the net effect of the embeddedness of FIEs on functional upgrading depends on the relative proportions of the substitution effect, the creation effect and the productivity effect. If the creation and productivity effects of non-fabrication functions brought about by the embeddedness of FIEs are larger than the substitution effect, functional upgrading will occur. Based on the above analysis, the following two hypotheses are proposed:

H1. 

At the industry level, when the substitution effect of non-fabrication functions of FIEs is dominant, greater FIE embeddedness leads to functional specialisation of the value chain in the fabrication segment, thus inhibiting functional upgrading.

H2. 

When the productivity and creation effects of non-fabrication functions of FIEs are dominant, greater FIE embeddedness leads to functional specialisation of China’s manufacturing value chain in non-fabrication segments, which promotes functional upgrading. This theoretical framework is summarised in Figure 2.

3. Materials and Methods

3.1. Model Construction

In this section, the relationship between the embeddedness of FIEs and the functional upgrading is empirically examined through econometric analysis. This study constructs a fixed effect benchmark model (1) based on panel data to examine the impact of the embeddedness of FIEs on the functional upgrading at the overall industry level.

$${\mathrm{F}\mathrm{S}\_\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}}_{i,\mathrm{t}}={\beta }_0+{{\beta }_1\mathrm{F}\mathrm{I}\mathrm{E}\mathrm{s}}_{i,\mathrm{t}}+X_{i,\mathrm{t}}^{\prime }\mathsf{\theta }+{\mu }_{\mathrm{t}}+{\eta }_{\mathrm{j}}+{\epsilon }_{\mathrm{j},\mathrm{t}}$$

(1)

In the formula, the subscripts i and t represent the manufacturing sub-industry and the year, respectively. The explanatory variable is FS_Total, which denotes the focus of functional specialization in the value chain. The explanatory variable $\mathrm{F}\mathrm{I}\mathrm{E}\mathrm{s}$ is the percentage of value chain activities of FIEs in each industry. $X$ is the control variable of industry level. ${\mu }_{\mathrm{t}}$ and ${\eta }_{\mathrm{i}\mathrm{j}}$ denote the fixed effect of year and industry, respectively, which are used to control the unobservable macro-factor impacts in the time dimension and the industry shocks that do not change over time. ${\epsilon }_{\mathrm{i}\mathrm{j}\mathrm{t}}$ is a random disturbance term.

In order to further examine the impact of the embeddedness of FIEs on value chain upgrading at the “functional” level, this study constructs model (2) as follows:

$${\mathrm{F}\mathrm{S}}_{\mathrm{i}\mathrm{j},\mathrm{t}}={\beta }_0+{{\beta }_1\mathrm{F}\mathrm{I}\mathrm{E}\mathrm{s}}_{\mathrm{j},\mathrm{t}}+X_{\mathrm{j},\mathrm{t}}^{\prime }\mathsf{\theta }+{\mu }_{\mathrm{t}}+{\eta }_{\mathrm{j}}+{\epsilon }_{\mathrm{j},\mathrm{t}}$$

(2)

In the Formula (2), the subscripts $\mathrm{i},\mathrm{j},\mathrm{t}$ denote function, industry and year, respectively. The explanatory variable FS (i = 1,2,3,4) denotes the shift in the country’s industries in four different functions relative to the focus of production in Fabrication (FS_FAB), R&D (FS_RD), Management (FS_MGT) and Marketing (FS_MAR).

3.2. Selection of Variables

3.2.1. Explained Variable

The explanatory variable of this study is functional upgrading. Two primary approaches are currently used to measure functional upgrading [21]. One approach is the induction approach of enterprise business activity, and the other is the decomposition approach of labor compensation structure. The induction approach, which is based on a micro-level analysis, identifies the functional division of labor within the business activities of enterprises. These activities are then generalized to the industry level, and classified into upstream, midstream, and downstream positions within the value chain. This approach constructs an index of the upstream and downstream functional division of labor in the value chain, thereby reflecting dynamic changes in functional upgrading. A larger index indicates a greater degree of functional upgrading within the industry [48]. The decomposition approach, which is based on the value-added structure of the industry, integrates input-output and labor occupation data. This approach estimates the share of labor remuneration for workers with different occupational skills, allowing the calculation of an index of specialization for each function. This index can then be used to measure the upgrading process across sub-functional dimensions [49]. Both approaches incorporate functional divisions of labor into the industrial structure and are capable of reflecting dynamic changes in the production chain. Therefore, their measurement results are consistent [48]. A key advantage of the induction approach is its ability to directly reflect the functional division of labor through enterprise business activities. However, it does not account for value-added compensation across different functions or structural changes, meaning it can only reflect the overall process of functional upgrading. In contrast, the decomposition approach, while accounting for value-added compensation and structural changes, and analyzing the specialization levels of different functions, provides a more detailed view of the specific structural path of functional upgrading. However, it is less effective at capturing the process of upgrading in its entirety. Thus, the induction approach is particularly useful for measuring the overall process of functional upgrading, while the decomposition approach is better suited to elucidating the specific direction of upgrading. To more accurately capture the process of functional upgrading and understand the relationship between the embeddedness of FIEs and functional upgrading, both approaches are adopted in this study.

(1)

The Induction approach

We use Stöllinger et al. (2021)’ s methodology and the fDi Markets database to link each greenfield investment project’s primary activities in the host country to five functional modules that cover upstream, midstream, and downstream value chain segments, including headquarter economy, R&D and design, production, and logistics [16]. An index of the value chain’s functional division of labor is calculated by comparing China’s manufacturing sector’s greenfield investment projects to the global proportion of upstream and downstream functional activities. The index tracks China’s manufacturing sectors’ functional upgrading toward the top of the “smile curve” in a more comprehensive manner. The calculating formula is:

$${FS\_Total}_{it}={\displaystyle \frac{P_{cit}^{UD}/P_{cit}}{P_{wit}^{UD}/P_{wit}}}$$

In year t, $P_{wit}^{UD}$ represents the total number of greenfield investment projects in global manufacturing sector i covering upstream headquarters economy, R&D, design, downstream logistics, retail services, and after-sales services, while $P_{wit}$ represents the same in China’s manufacturing sector i. In China’s manufacturing sector i, $P_{cit}^{UD}$ indicates the number of upstream and downstream investment projects in year t, whereas $P_{cit}$ represents the overall number of real investment projects in the sector. In order to prevent sample loss due to the absence of greenfield investment in certain years, this study applies a logarithmic transformation to the value chain functional upgrading variable in the regression analysis.

(2)

The Decomposition approach

In this study, the updated Occupational Skills Database by Zaveri et al. (2023)., in conjunction with the Asian Development Bank’s (ADB) Input-Output Database, is used to measure functional upgrading indicators at sub-functional dimensions [50]. First, we categorize the production function within the value chain based on workers’ occupations. Using the ISCO-88 Occupational Classification and the Production Functions Comparison Table provided in the Appendix of Kordalska and Olczyk, we classify the production functions into four categories: (1) R&D function, which corresponds to occupations in engineering, health, teaching, etc.; (2) Management function, primarily covering occupations such as legislators and senior managers; (3) Fabrication function, which includes practical occupations such as technicians, machine operators, and transportation equipment operators; (4) Marketing function, involving occupations such as clerical workers and salespersons.

Second, the degree of specialization in the functional division of labor is measured with reference to the export comparative advantage approach proposed by Timmer et al. [5]. Specifically, the functional division of labor of a country’s industry i in function k is measured as follows:

$${FS}_i^k={\displaystyle \frac{f_i^k/{\sum }_k{f}_i^k}{{\sum }_i{f}_i^k/{\sum }_i{\sum }_k{f}_i^k}}$$

Finally, with the help of the input-output table, total exports are decomposed and labor compensation in each functional dimension is identified. Drawing on the value-added decomposition method proposed by [51], the compensation of laborers in different occupations in the value chain is calculated to measure the functional specialization division of labor in the industry. The specific calculation formula is:

$$f=W{d}^E$$

In the formula, the K × G matrix W comprises entries w sub k g, representing the proportionate earnings share for workers in function k in industry g. Given that the distribution of functions across industries in the global production process evolves annually, this study controls for the average distribution of each production function. Based on this, a centralized functional specialization index is computed to act as a proxy variable for functional upgrading.

3.2.2. Explanatory Variable

The explanatory variable in this study is the embeddedness of FIEs, which refers to the local production, purchasing and selling activities of FIEs in the host country. This study adopts the value-added decomposition method, which is common in the literature, to identify and decompose the different production and trading activities of FIEs in the host country’s industry by tracing all the value-added creation activities involved in MNCs along the value chain (measured in percentage terms), with reference to the study by Wang et al. [51].

3.2.3. Control Variable

As for control variables, based on existing empirical studies on functional upgrading [52], this study also controls for the following industry-level variables: (1) The level of economic development (LnGDPpc), measured using the natural logarithm of per capita GDP at the industry level; (2) industry scale (Lnscale), measured using the natural logarithm of output at the industry level; (3) capital intensity (Lncapital), measured using the total fixed asset formation of each industry in the manufacturing sector; (4) trade openness (Open), measured using the industry-level total import and export as a share of GDP; (5) State capital (Lnstate), measured as the natural logarithm of industry-level state-owned capital to control for government involvement; (6) Technological intensity (TechIntensity), which classifies industries into low- and high-technology groups according to their R&D intensity.

3.3. Data Sources and Descriptive Statistics

The data for measuring functional upgrading using the induction approach of enterprise business activity are sourced from fDi Markets. For the decomposition approach of labor compensation structure, data are obtained from the ADB-MRIO and the corresponding data from the International Labor Organization’s Occupational Skills Database. The explanatory variable regarding the embeddedness of FIEs is derived from the most recent version of the OECD’s AMNE Dataset, released in 2024. Data for the explanatory variable the embeddedness of FIEs come from the latest version of the OECD’s AMNE Dataset, released in 2024, which is an extension of the Input-Output Tables in the dimensions of firm ownership (DOEs vs. FIEs), ensuring consistency with the unit of analysis of the Functional Upgrading. Compared to the previous version, the statistical criteria for FIEs in China have been upwardly adjusted in the 2024 dataset, changing the minimum shareholding ratio for FIEs from 25% to 50%. This adjustment prevents the overestimation of foreign firms and the role of FIEs in China. Data on control variables are obtained from OECD database.

It should be emphasized that during the sample processing, it is necessary to match the manufacturing industry classification criteria of several databases, including the OECD-AMNE database, the fDi Markets database, and ADB-MRIO. Specifically, the industry classification standard of the OECD-AMNE database follows the International Standard Industrial Classification (ISIC Rev4), which contains 17 manufacturing sub-sectors. fDi Markets database adopts the industry classification standard of the General Industrial Classification of Economic Activities of the European Community (NACE Rev2.0), which specifically includes 24 manufacturing sub-sectors. In contrast, the industry classification standard in ADB-MRIO corresponds to the International Standard Industrial Classification (ISIC Re 3). Based on prior research, this study uses ISIC Rev.3 as the benchmark to align NACE Rev.2.0 and ISIC Rev.4.0, resulting in 14 manufacturing subsectors. The descriptive statistics of variables are shown in

Table 1. Descriptive statistics.

Variable	N	Mean	Sd	Min	Max
FS_Total	252	0.709	0.400	0	1.946
FS_FAB	224	0.825	1.632	−1.813	6.875
FS_RD	224	−0.577	0.604	−2.452	1.273
FS_MGT	224	−0.600	0.503	−2.525	0.457
FS_MAR	224	0.153	1.013	−2.130	2.727
FIEs	252	13.27	6.129	3.150	29.73
LnGDPpc	252	4.186	0.880	1.724	6.402
Lnscale	252	12.98	0.960	10.23	14.45
Lncapital	252	10.55	1.301	6.261	12.65
Open	252	0.281	0.299	0.0467	1.435
Lnstate	252	5.126	2.708	0	8.469
TechIntensity	252	0.643	0.480	0	1

.

4. Results

4.1. Benchmark Regression Results

Benchmark model results are in

Table 2. Benchmark regression results.

	(1)	(2)	(3)	(4)	(5)
Variable	FS_Total	FS_FAB	FS_RD	FS_MGT	FS_MAR
FIEs	0.046 ***	−0.118 **	0.043 **	0.033 ***	−0.011
	(3.577)	(−2.518)	(2.450)	(3.017)	(−0.445)
LnGDPpc	0.168 *	−0.815 **	0.232 *	−0.129	0.028
	(1.674)	(−2.164)	(1.664)	(−1.479)	(0.142)
Lnscale	−0.521 ***	2.107 ***	−0.723 ***	−0.069	0.096
	(−3.813)	(4.196)	(−3.888)	(−0.592)	(0.368)
Lncapital	0.143	−1.897 ***	−0.121	−0.213 **	−0.115
	(1.423)	(−5.356)	(−0.921)	(−2.590)	(−0.627)
Open	−0.095	1.052 *	−0.548 ***	−0.452 ***	0.074
	(−0.640)	(1.894)	(−2.664)	(−3.503)	(0.255)
Lnstate	0.006	0.066	0.067 ***	0.031 **	0.045
	(0.360)	(1.183)	(3.212)	(2.370)	(1.545)
TechInt	0.000	0.000	0.000	0.000	0.000
	(.)	(.)	(.)	(.)	(.)
Constant	4.653 ***	−2.247	8.326 ***	2.588 *	−0.091
	(2.806)	(−0.371)	(3.712)	(1.841)	(−0.029)
Industry FE	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES
Observations	252	224	224	224	224
R-squared	0.866	0.904	0.903	0.945	0.933

Note: *, **, *** indicate significance at the 10%, 5%, and 1% levels, respectively.

. The embeddedness of FIEs and functional upgrading at the industry-wide level, determined by the induction approach, explain column (1). The regression relationship between FIE embeddedness and functional upgrading is highly positive at the 1% significance level. This finding supports H2 while disproving H1 and shows that FIEs’ embeddedness helps China’s manufacturing industry upgrade functionally inside the value chain. The value chain’s functional upgrading process includes the dynamic evolution of functional specialization indicators, including fabrication, R&D, management, and marketing. This study examines how FIE embedding affects specialization across dimensions. Columns (2) to (5) explain China’s manufacturing industry’s value chain’s functional specialization in fabrication, R&D, management, and marketing. All models use industry and year fixed effects as control variables. At the 1% statistical significance level, the regression analysis shows a negative association between FIE embeddedness and China’s value chain’s “Fabrication” functional specialization. FIEs’ embeddedness reduces demand for pure “fabrication” and drives the Chinese manufacturing industry away from low-end “fabrication” through technology spillovers. In columns (3)–(5) of Table 2, the embeddedness of FIEs shows a statistically significant positive correlation with R&D functional specialization at the 5% significance level and an even stronger positive correlation with management at the 1% significance level, while its effect on the upgrading of marketing functions is not significant. This suggests that the embeddedness of FIEs prompts China to get rid of the low-end “Fabrication” link, prompting China’s industrial chain toward R&D and management, but has no effect on the Marketing function.

4.2. Robustness Tests

The empirical results in the previous part show that the embeddedness of FIEs has a facilitating effect on the functional upgrading of Chinese manufacturing industry. In order to further demonstrate the reliability of the results, this study conducts a series of robustness tests.

First, the estimation method is replaced. Due to the use of industry-level panel data, problems such as between-group heteroskedasticity, within-group autocorrelation, and between-group contemporaneous correlation may occur. In order to solve these problems, this study uses the PCSE method to re-estimate the regression. The results are shown in column (1) of

Table 3. Robustness Tests: Lagged Explanatory Variables & Alternative Estimation Methods.

	(1)	(2)
	FS_Total	FS_Total
FIEs	0.041 ***
	(3.992)
L.FIEs		0.046 ***
		(3.530)
Control variables	YES	YES
Constant	0.000	4.527 **
	(.)	(2.577)
Observations	252	238
R-squared	0.779	0.871

Note: **, *** indicate significance at the 5%, and 1% levels, respectively.

, which shows that the coefficient of the main explanatory variable “Fabrication” function is still significantly negative. The results of the other explanatory variables have not changed substantially, which indicates that the embeddedness of FIEs in the upgrading of manufacturing function still exists.

Second, the explanatory variable is lagged by one period. Considering the possible time lag in the impact of the embeddedness of FIEs on industrial function upgrading, this study lags the explanatory variables by one period and re-examines the relationship. The results are shown in column (2) of Table 3, and it can be seen that the main conclusions of this study are still right even if the explanatory variables are lagged by one period.

4.3. Mechanism Analysis

As the theoretical analysis suggests, this upgrading effect stems from the productivity and creation effects of FIE embeddedness, which creates a skill premium by creating a demand for high skills, thus contributing to the functional upgrading of the value chain. To summarise, the creation and productivity effects on high-skill activities are the core mechanisms through which the embeddedness of FIEs contributes to the functional upgrading of value chains. In order to prove the existence of this mechanism, the following two aspects are explored in this study.

First, this study examines the creation effect of FIE embeddedness on upgrading functions in the value chain. Theoretically, if the embeddedness of FIEs can create new occupations, increase the demand and application of high-skilled labour—especially the demand for skills corresponding to the functions of R&D, management and marketing—and change the structure of the functional division of labour, then the production chain of skills corresponding to functions such as R&D, management and marketing will be extended, and the production length of the overall value chain will increase accordingly [53]. In order to test the existence of the creation effect, this study first adopts the length of the value chain as an indicator of the creation effect of FIE embeddedness, using Wang et al.’s [35] method to measure the length of the value chain. Then, according to the mediation effect test procedure by Baron and Kenny (1986) [54], the following mediation effect models (3) and (4) are constructed:

$${\mathrm{L}\mathrm{e}\mathrm{n}\mathrm{g}\mathrm{t}\mathrm{h}}_{\mathrm{j},\mathrm{t}}={\beta }_0+{{\beta }_1\mathrm{L}\mathrm{e}\mathrm{n}\mathrm{g}\mathrm{t}\mathrm{h}}_{\mathrm{j},\mathrm{t}}+X_{\mathrm{j},\mathrm{t}}^{\prime }\mathsf{\theta }+{\mu }_{\mathrm{t}}+{\eta }_{\mathrm{j}}+{\epsilon }_{\mathrm{j},\mathrm{t}}$$

(3)

$${\mathrm{F}\mathrm{S}\_\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}}_{\mathrm{i}\mathrm{j},\mathrm{t}}={\beta }_0+{{\beta }_1 \mathrm{F}\mathrm{I}\mathrm{E}\mathrm{s}}_{\mathrm{j},\mathrm{t}}+{{\beta }_2\mathrm{L}\mathrm{e}\mathrm{n}\mathrm{g}\mathrm{t}\mathrm{h}}_{\mathrm{j},\mathrm{t}}+X_{\mathrm{j},\mathrm{t}}^{\prime }\mathsf{\theta }+{\mu }_{\mathrm{t}}+{\eta }_{\mathrm{j}}+{\epsilon }_{\mathrm{j},\mathrm{t}}$$

(4)

Specific regression findings are given in

Table 4. Mediation analysis: creation effect.

	(1)	(2)
	Length	FS_Total
FIEs	0.062 ***	0.033 **
	(6.353)	(2.349)
Length		0.212 **
		(2.294)
Control variables	Yes	Yes
Constant	6.318 ***	3.602 **
	(5.098)	(2.099)
Sobel Z	2.158 **
Bootstrap	[0.004, 0.025]
Industry FE	YES	YES
Year FE	YES	YES
N	234	234
R2	0.970	0.880

Note: **, *** indicate significance at the 5%, and 1% levels, respectively.

. As shown in column (1), the coefficient of the embeddedness of FIEs is significantly positive. In column (2), both the embeddedness of FIEs and the value chain length remain significantly positive, suggesting that FIE embeddedness stimulates the creation of new occupations and extends the production chain—clear evidence of the creation effect. The Sobel test showed that the Z-value statistic of 2.158 is statistically significant at 5%. Additionally, a bootstrap sampling test (1000 iterations) shows that the mediating effect’s 95% confidence interval, [0.004, 0.025], does not include zero, showing statistical significance. The above results indicate that the embeddedness of FIEs has a creation effect; i.e., the embeddedness of FIEs has an impact on functional upgrading mainly through creating demand for skills related to R&D and management functions.

Second, we analysed the productivity effect of the embeddedness of FIEs. Theoretically, the technology spillover effect of FIEs can not only decrease production and operation costs but also introduce a productivity effect by expanding production scale and increasing output per unit of labour [55]. We use industry-level total factor productivity (TFP) to assess the mediating effect of the embeddedness of FIEs. Following the existing literature [56,57], we employ two approaches to examine the underlying mechanism: the stochastic frontier analysis method (Productivity1) and the non-parametric estimation method (Productivity2).

The regression results in

Table 5. Mediation analysis: Productivity effect.

	(1)	(2)	(3)	(4)
	Productivity1	FS_Total	Productivity2	FS_Total
FIEs	0.019 ***	0.038 ***	0.025 ***	0.037 ***
	(3.071)	(3.032)	(2.913)	(2.917)
Productivity1		0.394 ***
		(3.530)
Productivity2				0.357 ***
				(4.041)
Control variables	Yes	Yes	Yes	Yes
Constant	1.807	3.941 **	6.310 ***	3.644 **
	(1.613)	(2.420)	(5.107)	(2.128)
Sobel Z	1.98 **		2.11 **
Bootstrap	[0.002, 0.018]		[0.003, 0.020]
Industry FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
N	252	252	252	252
R2	0.803	0.873	0.647	0.876

Note: **, *** indicate significance at the 5%, and 1% levels, respectively.

show that the coefficients of FIEs in columns (1) and (3) are significantly positive. In columns (2) and (4), both the embeddedness of FIEs and total factor productivity (TFP) remain significantly positive as well. The Sobel test indicates that the Z-values are 1.98 and 2.11, respectively, both significant at the 5% level. Furthermore, the results of the 1000-time bootstrap sampling test show that the 95% confidence intervals of the mediating effects are [0.002, 0.018] and [0.003, 0.020], respectively, neither of which includes zero. These results suggest that the embeddedness of FIEs exerts a productivity effect—namely, that it influences functional upgrading primarily by enhancing productivity related to R&D and management functions. This finding is consistent with the theoretical inference that the embeddedness of FIEs promotes productivity-driven functional upgrades.

4.4. Endogeneity Analysis

Given the strong path dependence of industrial functional upgrading—where industries with higher functional levels tend to sustain upgrading capacity—and the potential bidirectional causality between the embeddedness of FIEs and upgrading, this study includes the lagged dependent variable and employs both system and difference GMM estimations. As shown in

Table 6. Endogeneity Analysis.

	(1)	(2)
	System-GMM	Difference GMM
L.Fs_Total	0.247 **	0.163 *
	(2.09)	(1.87)
FIEs	0.029 **	0.025 ***
	(1.95)	(2.95)
Control variables	YES	YES
AR (1)	0.010	0.015
AR (2)	0.107	0.155
Hansen	0.800	0.736
Observations	238	224

Note: *, **, *** indicate significance at the 10%, 5%, and 1% levels, respectively.

, the estimated coefficients of the embeddedness of FIEs are 0.029 and 0.025, which are significant at the 5% and 1% levels, confirming that its positive effect on functional upgrading remains robust after addressing endogeneity.

4.5. Heterogeneity Analysis

Up to now, the previous analysis focused on the overall level of the embeddedness of FIEs, but ignored the differential impact of the embeddedness of FIEs on the functional upgrading of manufacturing industries. Next, a more detailed heterogeneity analysis of the relationship is conducted in order to provide further empirical evidence for the benchmark regression results.

FIEs not only become factors of production within the industry, but also have an impact on the industry through upstream and downstream industry linkages. FIEs can act both as suppliers of intermediate goods in the upstream and as producers in the downstream. Functional upgrading not only relies on the changes in the structure of value-added income distribution within the industry level, but is also affected by the changes in the structure of inputs and outputs upstream and downstream of the industrial chain [58]. Therefore, differences in the embeddedness of FIEs approach may have different impacts on functional upgrading. Referring to Wang et al.’s study [35], this section classifies FIEs-related value chain activities into the following three types: (1) “FIEs-DOEs”, i.e., upstream value-added providers are FIEs and downstream product producers are DOEs; (2) “DOEs-FIEs “, i.e., upstream value added providers are DOEs and downstream product producers are FIEs; (3) “FIEs-FIEs”, i.e., both upstream value added providers and downstream product producers are FIEs.

In order to further distinguish the different impacts of heterogeneity of the embeddedness of FIEs mode on the functional upgrading of manufacturing industry, this study sets up the empirical model as follows:

$${\mathrm{F}\mathrm{s}}_{\mathrm{i}\mathrm{j},\mathrm{t}}={\alpha }_0+{{\beta }_0 \mathrm{D}\mathrm{O}\mathrm{E}\_\mathrm{F}\mathrm{I}\mathrm{E}}_{\mathrm{j},\mathrm{t}}+{{\beta }_1 \mathrm{F}\mathrm{I}\mathrm{E}\_\mathrm{D}\mathrm{O}\mathrm{E}}_{\mathrm{j},\mathrm{t}}+{{\beta }_2 \mathrm{F}\mathrm{I}\mathrm{E}\_\mathrm{F}\mathrm{I}\mathrm{E}}_{\mathrm{j},\mathrm{t}}+X_{\mathrm{j},\mathrm{t}}^{\prime }\mathsf{\theta }+{\mu }_{\mathrm{t}}+{\eta }_{\mathrm{j}}+{\epsilon }_{\mathrm{j},\mathrm{t}}$$

(5)

The results in column (1) of

Table 7. Heterogeneity analysis.

	(1)	(2)	(3)	(4)	(5)
	FS_Total	FS_FAB	FS_RD	FS_MGT	FS_MAR
DOE_FIE	0.039 *	−0.160 **	0.064 **	0.029	0.056
	(1.889)	(−2.140)	(2.307)	(1.645)	(1.442)
FIE_DOE	0.101 ***	−0.326 ***	0.087 *	0.030	−0.112 *
	(3.034)	(−2.605)	(1.875)	(0.998)	(−1.723)
FIE_FIE	0.021	0.042	−0.006	0.039 **	−0.025
	(1.052)	(0.567)	(−0.201)	(2.197)	(−0.649)
Control variables	YES	YES	YES	YES	YES
Constant	5.095 ***	−6.680	9.666 ***	2.420	0.375
	(3.011)	(−1.080)	(4.188)	(1.652)	(0.116)
Industry FE	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES
Observations	252	224	224	224	224
R-squared	0.868	0.907	0.906	0.945	0.934

Note: *, **, *** indicate significance at the 10%, 5%, and 1% levels, respectively.

show that the coefficient of the embeddedness of FIEs of the type “FIEs-DOEs” is significantly positive at the 1% level, suggesting that the embeddedness of FIEs in upstream industries has a significant contribution to functional upgrading at the overall level. The coefficient of the embeddedness of FIEs of the type “DOEs-FIEs” is significantly positive at the 10% level, indicating that the embeddedness of FIEs in downstream industries also positively affects the functional upgrading at the overall level. The coefficient of the embeddedness of FIEs of the type “FIEs-FIEs” is positive but not significant, indicating that when FIEs act as both upstream value-added providers and downstream producers, and the impact on the functional upgrading is relatively limited.

The regression results in column (2) of Table 7 show that both “FIEs-DOEs” and “DOEs-FIEs” can negatively affect the level of specialization in the “Fabrication” function, while “FIEs-FIEs” of value chain activities do not have a significant impact on the overall level of functional upgrading. This implies that both upstream value-added providers being FIEs and downstream producers being FIEs will have a disincentive for manufacturing industries to specialize in pure “Fabrication” activities. This may be due to the fact that the technology spillover from FIEs, force the Chinese manufacturing industry to move away from the low-end “Fabrication” segment through demonstration and competition effects (e.g., automation applications), thus reducing the demand for skills related to pure “Fabrication” function. The regression results in column (3) of Table 7 show that both the “DOEs–FIEs” and “FIEs–DOEs” modes of embeddedness have a significantly positive effect on the upgrading of the R&D function. This finding suggests that vertical linkages between domestic and foreign firms—whether upstream or downstream—are conducive to the enhancement of R&D capabilities within China’s manufacturing sector. The results in column (4) of Table 7 show that the type “FIEs-FIEs” of FIEs activities have a positive impact on the management function of China, which means that when both upstream value-added providers and downstream producers are FIEs, China can effectively transform the advanced management experience of FIEs into the advantage of domestic “management” skills. The results in column (5) of Table 7 show that FIEs activities of the type “FIEs-DOEs” have a negative impact on the domestic “Marketing” function. This is mainly because FIEs from upstream value-added suppliers are prone to disadvantage the country’s marketing function through market forces, such as the monopolization of sales channels by large international buyers.

5. Discussion

Based on the task-biased theory of technological progress, this study introduces the embeddedness of FIEs as a crucial factor influencing functional upgrading in value chains and establishes a coherent framework examining the relationship between FIE embeddedness and functional upgrades. Task-biased technological progress emphasises that technological change is not neutral—it systematically alters the relative demand for tasks and skills, favouring high-skill, non-routine functions such as R&D and management. Within this framework, the embeddedness of FIEs serves as a key transmission channel through which technological progress affects domestic industries. Through technology spillovers, management learning and capital–skill complementarities, FIEs reshape the skill composition and task structure of domestic production, thereby driving the transition from low-end fabrication toward higher-value-added functional specialisation. By integrating this mechanism into the analysis, the study bridges two strands of literature—one on FIE embeddedness, which focuses mainly on technology spillovers and productivity effects, and another on functional upgrading, which examines technological progress and globalisation but often neglects the role of FIEs. Using data from China’s manufacturing industries, this study empirically tests this theoretical chain, confirming that FIE embeddedness significantly promotes upgrading toward R&D and management functions.

Further, this study highlights how the task-biased technological progress effect induced by FIE embeddedness fundamentally shapes the functional division of labour within China’s manufacturing value chain. In the Chinese context, FIEs not only transmit advanced technologies but also alter the structure of task demand, creating productivity and innovation-driven upgrading effects. The productivity effect enhances efficiency and profitability through technological diffusion, while the creation effect expands new high-skill tasks and extends the value chain length. By identifying these mechanisms, this study extends the application of task-biased technological progress theory to the field of value chain upgrading and clarifies how FIE embeddedness serves as both the carrier and catalyst for functional upgrading.

Finally, by adopting a systems perspective to deconstruct three types of FIE-related activities in the value chain, this study provides new perspectives for future empirical research through the analysis of their heterogeneity. Especially in the context of the increasing proportion of FIE-embedded value chain activities and the increasing diversification of FIEs in the host country’s value chain activities, this study helps to identify the characteristics of different types of FIE embeddedness in value chain activities and then provides strong support for the use of FIEs to promote the upgrading of the country’s value chain. This study argues that the inconsistency between the existing conclusions on FIEs and skill upgrading and skill inequality may be due to the heterogeneity of the embeddedness of FIEs [59,60]. This study identifies FIE value chain activities as ‘DOEs—FIEs’, ‘FIEs—DOEs’ and ‘FIEs—FIEs’, and it is found that the embeddedness of FIE activities has differentiated impacts on the functional upgrading of different value chains, which makes the analysis of the embeddedness of FIEs and value chain upgrading more comprehensive and in-depth.

6. Conclusions

This study examines how FIE embeddedness affects the functional upgrading of the value chain and its mechanism in 14 subsectors of China’s manufacturing industry from 2003 to 2020 using the OECD AMNE database in 2024. A summary of the primary research results is given as follows: First, FIE embeddedness positively affects functional upgrading across the manufacturing sector. FIE embeddedness facilitates the functional upgrading of R&D and management but not marketing at a sub-functional level. Mechanism analysis shows that FIE integration in China’s manufacturing value chain predominantly fosters functional upgrading through productivity-enhancing and innovation-driven benefits. Heterogeneity analysis shows that both “DOEs–FIEs” and “FIEs–DOEs” activities improve R&D, “FIEs–FIEs” activities improve management, and “FIEs–DOEs” activities inhibit marketing.

This study suggests the following policy changes based on its findings. First, the Chinese government must increase its efforts to attract FIEs, as these enterprises catalyse the functional upgrading of China’s industrial value chain. Second, to promote functional upgrading in China’s manufacturing value chain, the government must create incentive policies for the labour market that fully leverage both the productivity and creation effects of FIEs on high-skilled labour. Third, investment attraction strategies that promote functional upgrading should consider FIEs’ embeddedness to improve their efficacy and precision. Heterogeneity analysis shows that FIEs’ functional upgrading effects vary by embedding mode. Thus, FIE attraction strategies that promote functional upgrading must consider the relevance of these policies to upgrading across functional dimensions.

This study has limitations, but future research can address and enhance them. First, the current classification of functional division of labour is based on the first main category of ISCO-88 from the International Labour Organization, making it difficult to obtain more granular data. This study focuses on the four types of functional divisions of labour, which may limit real-world diversity. More data on the functional division of labour linked with occupational skills is expected to expand our understanding of functional upgrading, including functional specialisations related to value chain connections. Second, the study’s focus on FIE embeddedness and functional upgrading in China’s value chain limits its generalisability. Future study can be extended to other countries to permit cross-country comparisons and investigate FIEs’ favourable impact on functional upgrading in specific countries, laying the groundwork for targeted FIE policies. Third, while this study highlights FIEs’ embeddedness in value chain functional upgrading at the macro-industry level, it lacks micro-enterprise empirical evidence, limiting its analytical framework. Micro-enterprise data can be used to study FIEs’ economic effects on value chain functional upgrades.