Analysis of the evolution of the marginal effect of human capital structure in the process of industrial structure evolution

Abstract

The human capital structure changes as the industrial structure evolves. In order to achieve the matching development of human capital structure and industrial structure, this study, based on the theoretical framework of the dynamic evolution of the human capital structure, selects the panel data of 284 prefecture-level and above cities in China from 2003 to 2019 to explore the impact of China’s human capital structure on the upgrading of the industrial structure and the dynamic evolution of the marginal effect of the human capital structure. The results of the study reveal that the optimization of human capital structure can significantly promote the upgrading of the industrial structure, but this promotion is not significant during the industrialization stage and in small cities. The results of the quantile regression show that as the adjustment of the industrial structure deepens, the marginal effect of the human capital structure gradually increases. The mechanism test shows that the human capital structure can promote the upgrading of the industrial structure by facilitating technological progress, increasing total factor productivity, and boosting technology transactions. In the further examination, this study measures the human capital structure from the dimension of skill training, and finds that there is an obvious substitution effect between training and education. This substitution effect gradually weakens as the industrial structure upgrades. Therefore, it is recommended that local governments formulate policies for talent cultivation and introduction in combination with the current development situation of the industrial structure, strike a proper balance between talent introduction policies and industrial structure adjustment plans, and promote the coordinated development of human capital structure and industrial structure.

Introduction

In 2013, the tertiary industry in China surpassed the primary and secondary industries in added value, employment, and growth rate for the first time. The added value of the tertiary industry to the national economy was 46.88%, marking a 2.73 percentage point increase over the secondary industry. Furthermore, employment within the tertiary industry stood at 38.43%, a notably higher figure by 8.1 percentage points compared to the secondary sector. The expansion of the tertiary industry has been continuous, gradually assuming a dominant position in the national economy, thereby steering the economy towards a service-oriented industrial structure. The shift in the industrial structure towards a service economy signifies a transition from investment-driven growth to innovation-driven advancement, highlighting the growing importance of investing in human capital. Human capital is now recognized not merely as a production factor but as a crucial driver of innovation, facilitating the upgrading of industrial structures. Consequently, cities are now engaged in a fierce competition to attract talented individuals.

However, history warns us that human capital may not always be effective. South Korea’s experience before 1960 demonstrates that a substantial investment in higher education did not always align with societal demands, resulting in increased unemployment and dwindling labor remuneration (Temple 1999). In fact, during this period, the industrial structure of South Korea was dominated by agriculture, and the development of industry and the service sector was relatively backward. The excessive investment in higher education naturally led to the over-allocation of human capital. From this, it can be seen that the rationality of the “fierce talent competition” is open to question.

In the fierce competition for talents, cities often focus only on attracting talents while neglecting the adaptability between the talents and the local industrial structure.In theory, different stages of industrial development require a matching human capital structure (Huang et al. 2021). As industries with low added value and low technological content gradually transform into industries with high added value and high technological content, human capital also continuously shifts from low skills to high skills, achieving dynamic evolution. Different from the total amount of human capital or the level of human capital, the human capital structure focuses on the composition of human capital with different internal skills and its dynamic changes. The human capital structure needs to dynamically evolve in line with the industrial structure at different stages of economic development in order to effectively promote the adjustment of the industrial structure and avoid hindering economic growth.

In this process, the marginal effect of the human capital structure will change along with the adjustment of the industrial structure, and the marginal effect can reflect the dynamic changes in the relationship between input and output at a specific stage. Therefore, clarifying the impact of the human capital structure on the industrial structure and analyzing the marginal effect of the human capital structure during the evolution of the industrial structure can more accurately capture the changes in the industrial structure. It is beneficial for cities to formulate talent introduction policies in combination with the current development status of their industrial structures. Balancing talent introduction policies and industrial structure adjustment plans has important theoretical and practical significance.

This research is aimed at delving into three pivotal issues against the aforementioned backdrop. First and foremost, it endeavors to ascertain whether enhancing the human capital structure can effectively facilitate the upgrading of the industrial structure. Secondly, it focuses on investigating the dynamic evolution of the marginal effect of the human capital structure in tandem with the development of the industrial structure. Thirdly, it explores the underlying mechanism through which the human capital structure exerts an influence on the industrial structure.

To achieve these research objectives, this study constructs a theoretical analysis framework grounded in the endogenous theory of human capital. This framework is utilized to analyze the evolutionary process of the human capital structure and to explore how the human capital structure propels the upgrading of the industrial structure. The research makes use of panel data from 284 cities at and above the prefecture-level in China, covering the time span from 2003 to 2019. A fixed-effects model is adopted to explore the average impact of the optimization of the human capital structure on the upgrading of China’s industrial structure. Subsequently, a panel quantile model is employed to study the dynamic evolution of the marginal effect of the human capital structure. Mechanistic examinations are conducted from three perspectives: technological progress, total factor productivity, technology transactions. In the further examination, from the perspective of skills training, this paper re-measures the human capital structure and discovers that skills training serves as a certain substitute for education. As the industrial structure undergoes further deepening, this substitution effect gradually diminishes.

The main contributions of this study are in three aspects. First, this study constructs a theory based on the endogenous theory of human capital, describes the evolutionary characteristics of the human capital structure, emphasizes the importance of the human capital structure and its construction, and clarifies the important driving role of the human capital structure in the evolution process of the industrial structure. Second, at the indicator level, on the one hand, this study improves the indicators for measuring the industrial structure. It distinguishes between low-end technology manufacturing, mid-end technology manufacturing, high-end technology manufacturing, non-productive services, and productive services, and assigns different weights, which can more reasonably reflect the upgrading of the industrial structure. On the other hand, it measures the human capital structure from the perspective of skills training and discovers the substitution effect between education and training. Third, this study uses a panel quantile regression model to examine the dynamic evolution of the marginal effect of the human capital structure during the evolution process of China’s industrial structure, supplements the dynamic research on the marginal effect of the human capital structure in the existing literature, and provides relevant insights for local governments in making decisions on attracting talents and adjusting the industrial structure.

The remaining structure of this study is as follows. The second part is the literature review. In the third part, a mathematical model is used to illustrate the importance of optimizing the human capital structure in the development of the industrial structure. Theoretical analysis will be conducted and hypotheses will be put forward. The fourth part explains the model settings and variable specifications. The fifth part presents the empirical results and analysis of the fixed effects. The sixth part conducts a dynamic analysis using quantile regression. The seventh part is the mechanism test and analysis. The eighth part is a further test of the trained human capital structure. The final part consists of the conclusions and suggestions.

Literature review

Early research generally posits that human capital plays a crucial role in advancing industrial structure upgrades. High-level human capital can drive technological innovation (Li and Ma 2021), and regions with abundant human capital stock also tend to attract other productive factors, resulting in a comparative advantage and facilitating the transfer and allocation of resources among industries (Zhang et al. 2011). Subsequently, scholars have found limitations in the ability of human capital to drive the advancement of industrial structure. Two main considerations have emerged among scholars who hold this viewpoint.

The first group of scholars acknowledges the diversity of human capital and recognize that the impact on industrial structure varies depending on the types of human capital. Scholars generally believe that high-skilled human capital can promote the upgrading of the industrial structure through technological innovation (Wu and Liu 2021), income improvement, and urbanization (Wang and Li 2013). It can also better adapt to the evolving market demands and technological changes, and flow and transfer between different sectors, thus being better able to cope with changes in the industrial structure and economic shocks (Fusillo et al. 2022). The impact of low-skilled human capital on the upgrading of the industrial structure is uncertain (Wang and Li 2013), and it may even have a negative impact (Deng and Ke 2020). In this kind of view, in order to promote the upgrading of the industrial structure, scholars generally support increasing investment in high-skilled human capital, but pay little attention to low-skilled human capital. On the one hand, there exists a skill complementarity relationship between high-skilled human capital and low-skilled human capital, and both are crucial for the improvement of labor productivity (Liang and Lu 2019). On the other hand, the foundation for the improvement of the marginal effect of human capital is the increase in the total amount of human capital (Liu and Yuan 2023). Investment in human capital should respect the evolutionary laws of the human capital structure. It is not advisable to merely attach importance to higher education and increase the stock of high-skilled human capital. Instead, equal importance should be given to the development of basic education to facilitate the gradual evolution of low-skilled human capital into high-skilled human capital (Cheng et al. 2019), which reflects the dynamic changes within human capital. An advanced human capital structure can play a more significant role (Ma et al. 2023). Therefore, the human capital structure is of vital importance.

The second group of scholars considers the difficulty of making human capital work in certain countries or regions with relatively backward industrial structures. The role of human capital in developing countries and developed countries varies (Matousek and Tzeremes 2021). Given that public resources are limited in some countries or regions, the dilemma of whether to allocate resources to the development of industries or human capital accumulation arises. The industrial structure is a crucial factor affecting the marginal return of human capital investment. In the absence of a solid industrial foundation, it is difficult to generate effective demand for human capital, so priority should be given to establishing an industrial foundation rather than investing in human capital (Liu and Yuan 2023; Liao et al. 2023). Zhu et al. (2024) argue that premature de-industrialization can trigger the migration of high-skilled labor towards low-end sectors. Even with external technological advancements, this phenomenon may fail to foster the increased influx of high-skilled labor into higher-end sectors. Consequently, the economy may struggle to optimize its industrial framework, leading to a dearth of sustainable growth momentum. If educational policies are solely utilized to enhance the quality of the workforce, an incongruous industrial composition can result in an overabundance of human capital that does not align with the developmental stage. Therefore, in middle-income nations, the balanced distribution and accumulation of human capital across industries are paramount. The misallocation of high-skilled human capital to low-skilled industrial domains engenders an imbalance in human capital utilization (Marchiori and Franco 2020; Huang et al. 2023). By synthesizing these two viewpoints, it is not difficult to find that human capital and the industrial structure need to develop in a matched manner. The mismatch between the industrial structure and human capital will distort the ratio of production technology to human capital investment and reduce the labor productivity of industries (Bairoliya and Miller 2021). Investment in human capital through higher education can be divided into two situations: underinvestment and overinvestment (Liu and Yuan 2023). The over-allocation of human capital will be restricted by low-level production factors (Zheng 2021), and it will also lead to the crowding effect, reducing the innovation ability of talents (Zhao et al. 2023). The insufficient allocation of human capital in emerging technology industries will hinder technological innovation and the improvement of labor productivity, and reduce the overall productivity (Zhao et al. 2023; Yian 2019).

These two viewpoints also reflect a fact from the side: Human capital investment should depend on the specific stage of economic development (Duan et al. 2022; Wang and Tang 2019). According to Rostow’s (2001) theory of the evolution of economic stages, in the initial stage of economic development, industries are mostly labor-intensive. Human capital investment tends to meet the basic production needs, forming a human capital structure mainly composed of low-skilled labor. In the mature stage of economic development, the proportion of knowledge-intensive industries and technology-intensive industries increases, while the proportion of labor-intensive industries decreases. The development of industries urgently requires highly educated and skilled labor forces, and the required human capital is mainly high-skilled, triggering the endogenous accumulation of human capital (Liao et al. 2023). This dynamic adjustment promotes the gradual evolution of low-skilled human capital into high-skilled human capital, optimizes the human capital structure, and thus matches it with the industrial structure. Evidently, as the industrial structure continues to evolve, the corresponding matching human capital structure also evolves accordingly.

Based on the above analysis of the existing literature, this paper finds that most scholars conduct research on the impact of human capital on the industrial structure from a static perspective. They have overlooked the evolutionary processes of both the human capital structure and the industrial structure, and lack an understanding of the dynamic changes. In contrast, this paper focuses on a dynamic perspective, exploring the influence of the human capital structure on the industrial structure and its underlying mechanism. Additionally, it examines the evolution of the marginal effect of the human capital structure during the process of industrial structure evolution, aiming to provide references for local governments in making decisions regarding talent introduction and industrial structure adjustment.

Theoretical model and research hypotheses

Theoretical model

In endogenous growth theory, Lucas and Romer assumed human capital as a determining factor of technical progress, endogenizing it in the growth model. Human capital not only enters the production function as a production factor, but also influences technological progress. To investigate the relationship between human capital structure and industrial composition, this study develops a model following the insights of Chen and Gong (2005). The model in this study, distinct from that of Chen and Gong (2005), incorporates two key elements: Firstly, consider and calculate the critical value of the learning time cost, at which point the labor force in the agricultural sector no longer transfers to the non-agricultural sector, and the industrial structure remains in a stagnant state. Secondly, introduce the human capital structure into the model, demonstrate the dynamic evolution process of low-skilled human capital transforming into high-skilled human capital within the model, and explain the significance of optimizing the human capital structure.

In order to facilitate the analysis, the research sectors are simplified into the agricultural sector $x$ and non-agricultural sector $y$. The technology level is ${A}_{x}$ and ${A}_{y}$. There is no technological decline. The population is assumed to be constant in size and fully engaged in labor, standardized to 1. It is also assumed that in each period, land ownership is shared by the whole society. Land rents are equal and fixed for all individuals, and disregarding them does not impact the analysis results. Total expenditures to pay for consumer goods and services are derived from wage income.

Human capital is assumed to determine technological progress, as it is believed to be closely related, ${{dA}}_{i}/{{dh}}_{i} > 0$. The $x$ denotes the agricultural sector, and $y$ denotes the non-agricultural sector. Assuming that the elasticity of output of human capital is constant between sectors, the production function for the two sectors can be expressed as follows.

$${Y}_{i}={A}_{i}F(i)={A}_{i}{K}^{\upsilon }{({h}_{i}\times i)}^{\lambda },i=x,y;0 < \upsilon ,\lambda < 1$$

(3.1)

In Eq. (3.1), ${Y}_{i}$ denotes the production function of sector $i$, ${A}_{i}$ denotes the technology level in sector $i$, and ${h}_{i}$ indicates the human capital level in the sector $i$. $F\left(i\right)$ satisfying the first-order derivative is greater than zero and the second-order derivative is less than zero, physical capital is not considered in the latter for ease of analysis.

In the initial stages of the economy, the level of human capital in both sectors is ${h}_{0}$. Due to the developmental needs of economic growth, the manufacturing sector is given the opportunity to introduce technology and rapidly increase its total output value, and the high wages in manufacturing attract the labor force from the agricultural sector to move rapidly to the non-agricultural sector. Human capital aggregation externalities in the non-agricultural sector contribute to the overall increase in human capital levels. Consequently, the developmental requirements of the non-agricultural sector lead to a higher accumulation of human capital compared to the agricultural sector. From a static point of view, in order to facilitate the analysis, this study will set the level of human capital in the agricultural sector to be ${h}_{x}$. Because human capital in the agricultural sector lacks endogenous power, in the absence of external incentives, the level of human capital in the agricultural sector is always constant and equal to ${h}_{0}$. The level of human capital in the agricultural sector is always equal to the level of human capital in the agricultural sector ${h}_{y}$ and ${h}_{y}=h$. Due to the lack of endogenous motivation among human capital in the agricultural sector, the labor force is mainly engaged in low-skill production roles, reflecting the limited level of technology present in this sector. In contrast, the non-agricultural sector offers opportunities for higher-skilled production jobs. Importantly, policies do not restrict the transfer of labor between sectors, enabling low-skilled agricultural workers to transition to the non-agricultural sector through learning and investing in their human capital to enhance their skill levels. As a result of economic growth and increasing national incomes, households experience improved living conditions and find it easier to finance education, thereby creating a stronger incentive for individuals to self-finance human capital development and expedite the learning process required to enhance their skills. So assuming that the human capital investment function or the cost function of learning is affected by economic growth and is reflected in the time spent learning, the human capital investment function $v=v\left(h\right)-v\left({h}_{0}\right)=h* {dv}/{dh},\,v\left({h}_{0}\right)=0,$ ${dv}/{dh}$ is a constant greater than zero, $0 < v < 1$.

If viewed from a dynamic perspective, the non-agricultural sector in the $t$ period has obtained a higher level of human capital than the agricultural sector $h\left(t\right),\,{h}_{y}\left(t\right)=h\left(t\right)$. The human capital investment function is the following $v\left(t\right)=v\left(h\left(t\right)\right)-v\left({h}_{0}\right)=h\left(t\right)* {dv}\left(t\right)/{dh}\left(t\right)$, $v\left({h}_{0}\right)=0$, ${dv}\left(t\right)/{dh}\left(t\right)$ is a constant greater than zero, $0 < v\left(t\right) < 1$.

The labor force in the agricultural sector can choose to increase their human capital level to ${h}_{y}\left(t+1\right)$ through investment in human capital $v\left(t\right)$. Then in conjunction with the positive externality of learning, the level of human capital in the non-agricultural sector can be increased by investment in human capital $v\left(t\right)$ and the amount of labor transferred $l\left(t\right)$ to raise it. The equation for the evolution of human capital level is as follows.

$${h}_{y}(t+1)={h}_{y}(t)\{1+g{[v(t)l(t)]}^{\lambda }\}$$

(3.2)

To simplify analysis, time stamps $t$ are omitted when not misleading. Equations (3.3) and (3.4) are obtained from the human capital level evolution equation and the human capital investment function in the non-agricultural sector.

$$\frac{{\dot{h}}_{y}}{{h}_{y}}=g{(vl)}^{\lambda }$$

(3.3)

$$\dot{v}=\frac{dv}{d{h}_{y}}{\dot{h}}_{y}$$

(3.4)

Taking the non-agricultural sector product as the unit of valuation, the price of the non-agricultural sector product relative to the agricultural sector product is ${p}_{x}.$ The consumer’s budget constraint is ${p}_{x}{c}_{a}+{c}_{m}=I$, and the consumer faces a utility maximization problem.

$$\begin{array}{c}\max U({c}_{a},{c}_{m})=\frac{{({c}_{a}^{\alpha }{c}_{m}^{\beta })}^{1-\sigma }}{1-\sigma }\\ s.t.{p}_{x}{c}_{a}+{c}_{m}=I\end{array}$$

(3.5)

The no-arbitrage condition for the laborer’s position indicates that the production function of the agricultural sector in equilibrium is Eq. (3.6).

$${F}_{x}={\left[\frac{\alpha {h}_{x}^{\alpha }\,\,{{F}_{x}}^{{\prime} }}{\beta {h}_{y}^{\beta }\,{{F}_{y}}^{{\prime} }\,{F}_{y}^{\beta -1}(1-v)}\right]}^{\frac{1}{1-\alpha }}$$

(3.6)

Taking logarithms on both sides of Eq. (3.6) simultaneously and taking time $t$ derivative, we get Eq. (3.7).

$$nl=\left(\frac{v}{1-v}-\beta\right)\frac{{\dot{h}}_{y}}{{h}_{y}}$$

(3.7)

Among them, $n=\left((\alpha -1)\frac{{F^{\prime} }_{x}}{{F}_{x}}+\frac{{F^{\prime\prime} }_{x}}{{F^{\prime} }_{x}}+\frac{{F^{\prime\prime} }_{y}}{{F^{\prime} }_{y}}+(\beta -1)\frac{{F^{\prime} }_{y}}{{F}_{y}}\right)\, < \,0$_.

Integrating Eqs. (3.4)–(3.7) yields the growth rate of the level of human capital in the non-agricultural sector.

$$\frac{{\dot{h}}_{y}}{{h}_{y}}={g}^{\frac{1}{1-\lambda }}{\left[\frac{v}{n}\left(\frac{{h}_{y}}{1-v}\frac{dv}{d{h}_{y}}-\beta\right)\right]}^{\frac{\lambda }{1-\lambda }}$$

(3.8)

It can be seen that the growth rate of human capital levels in the non-agricultural sector $\dot{{h}_{y}}/{h}_{y}$ is a monotonically decreasing function of human capital level ${h}_{y}$. At the point when the growth rate becomes zero due to the time cost of learning reaching Eq. (3.9), the level of human capital in the non-agricultural sector stops increasing. Consequently, the production function of each sector stabilizes, leading to a state where the volume of employment remains constant. This causes the optimization process of the industrial structure to come to a standstill.

$$v=1-\frac{{h}_{y}}{\beta }\frac{dv}{d{h}_{y}}$$

(3.9)

The driving force behind the transfer of labor from the agricultural sector to the non-agricultural sector is the difference the level of human capital between the two sectors. However, this difference cannot expand indefinitely. When the time cost of learning exceeds a critical value, it will cause the cessation of improvement in human capital levels in non-agricultural sectors and a stagnation in the optimization of industrial structure. To break this state, measures need to be taken. First, the time cost of learning should be reduced. Second, the level of human capital in the agricultural sector needs to be enhanced. Emphasis should be placed on the popularization of basic education, narrowing the gap in human capital levels between the agricultural and non-agricultural sectors, promoting the transition from low-skilled human capital to high-skilled human capital, and accelerating the transfer of labor to non-agricultural sectors. It can be seen that, in order to promote the upgrading of the industrial structure, it is necessary to pay attention to the improvement of the level of human capital in society as a whole, to strengthen basic education to lay the foundation for higher education, to promote the transfer of human capital to the non-agricultural sector, especially knowledge-intensive or technology-intensive industries, and to achieve a decline in the total amount of low-skilled human capital in all sectors and an increase in the total amount of high-skilled human capital, which is precisely the process of optimizing the structure of the human capital of the society as a whole.

To demonstrate this dynamic trend, this paper sets the formula for the human capital structure as ${{SH}=h}_{y}y/\left({h}_{y}y+{h}_{x}x\right)$. Among them, ${h}_{x}x$ represents the total amount of low-skilled human capital in the agricultural sector, and ${h}_{y}y$ represents the total amount of high-skilled human capital in the non-agricultural sector. The increase in the human capital level ${h}_{x}$ of the agricultural sector leads to a decrease in the time cost of learning, which more rapidly drives the speed of labor force transfer. As a result, the total amount of low-skilled human capital ${h}_{x}x$ decreases. The improvement of the human capital level in the agricultural sector and the inflow of labor force contribute to the increase in the total amount of high-skilled human capital. Therefore, the improvement of ${SH}$ represents the optimization of the human capital structure.

The structure of output can be expressed as Eq. (3.10).

$${\kappa }_{y}=\frac{{Y}_{y}}{{Y}_{x}+{Y}_{y}}=\frac{1}{\frac{{Y}_{y}}{{Y}_{x}}+1}=\frac{1}{\frac{{A}_{y}}{{A}_{x}}{\left(\frac{{h}_{y}y}{{h}_{x}x}\right)}^{\lambda }+1}$$

(3.10)

Optimizing the human capital structure boosts industrial output share in the non-agricultural sector while reducing the share of agricultural sector output in society. This leads to the following proposition.

At the stage of human capital endogenization, it is essential to consider the development of both high-skilled and low-skilled human capital simultaneously. This approach, rather than solely relying on high-skilled human capital, is crucial for advancing the industrial structure. By prioritizing the evolution of human capital across the whole societal spectrum to evolve from low-skill to high-skilled, the optimization of the human capital structure can indeed facilitate the evolution of the industrial structure.

Research hypotheses

Human Capital Structure and Industrial Structure

The theoretical model mentioned above indicates that the optimization of the human capital structure can promote the upgrading of the industrial structure. During the continuous evolution of the industrial structure, human capital plays different roles. In the period when the industrial structure is mainly dominated by agriculture, the agricultural production mode mainly relies on the physical strength of the labor force. At this time, the role of human capital is equivalent to that of the labor force, and it enters the production function as a production factor. When the industrial structure enters the stage of servitization, especially when the service industry is mainly knowledge-intensive and technology-intensive, high-skilled human capital plays a huge role. At this moment, human capital no longer merely enters the production function as a production factor, but can also significantly improve the level of innovation.

It can be seen that high-skilled human capital has a relatively small impact on labor-intensive industries, but it has a significant impact on knowledge-intensive and technology-intensive industries. With the upgrading of the industrial structure, the marginal contribution of high-skilled human capital often shows an increasing trend, while the marginal contribution of low-skilled human capital will be suppressed. The optimization process of the human capital structure is not only about an increase in the proportion of high-skilled human capital, but also about the improvement of the overall human capital level of the whole society. Low-skilled human capital also evolves into high-skilled human capital through continuous learning. As industries gradually become dominated by knowledge-intensive and technology-intensive sectors, the positive role that optimizing the human capital structure can play is further enhanced.

Therefore, Hypothesis 1 is derived: The human capital structure can not only promote the upgrading of the industrial structure, but also the marginal effect of the human capital structure will continuously increase during the continuous evolution of the industrial structure.

From the perspective of theoretical analysis, the academic discussion on the reasons for the upgrading of the industrial structure focuses on Western economics and institutional economics.

Western economics holds that the root cause of industrial structure upgrading lies in innovation. Innovation can be regarded as the introduction of a new production function, which brings about technological progress in various industries. It can also create new market demands, stimulating the expansion or contraction of industries. The income elasticities of demand for the three sectors of industry increase successively. In the agricultural sector, the mechanization resulting from technological progress has a substitution effect on the labor force, leading to an increase in labor productivity. Meanwhile, the demand for agricultural products decreases, and the labor force migrates outward. The decreasing returns to scale and the income elasticity of demand being less than 1 will jointly cause a relative decline in the proportion of the output value and the labor force in the agricultural sector. Technological progress in the non-agricultural sector is often more robust. Moreover, the income elasticity of demand for the service industry is even higher than that of the manufacturing industry. The changes in productivity and demand also cause the proportion of employment in the manufacturing industry to rise first and then decline, while the proportion of employment in the service industry keeps increasing.

Institutional economics, on the other hand, emphasizes the importance of the influence of institutions on the upgrading of the industrial structure. Economists who support the “institutional determinism” view institutions as the main factor influencing changes in the economic structure. Different from the “technological determinism” perspective in Western economics, they believe that institutions are the ultimate cause affecting economic growth, and technological innovation is a manifestation of growth itself. It is the progress of institutions that promotes the development of technology. Institutions minimize the bargaining of people’s economic behaviors and mutual relationships, and reduce transaction costs. Proponents of “institutional determinism” hold that the income elasticity of demand is the result of the expansion of the market scale, and the differences in productivity among industries are due to changes in the degree of labor division. The shortcoming of explaining the driving forces behind the upgrading of the industrial structure in terms of the income elasticity of demand and productivity differences lies in the fact that it ignores the exponential growth of transaction costs as the labor division expands. Therefore, the specialized efficiency of the division of labor is the basis for changes in the industrial structure, and institutions that reduce transaction costs directly affect the upgrading of the industrial structure.

Therefore, this paper further analyzes the mechanism through which the human capital structure influences the industrial structure from two aspects: industry innovation and institutional arrangements.

Human Capital Structure, Industry Innovation and Industrial Structure Upgrading

On the one hand, the optimization of the human capital structure can significantly promote the level of industry innovation. In terms of technological progress, the endogenous growth theory proposes that human capital is the core element driving technological progress (Lucas 1988). High-skilled human capital usually has received better education and training.Individuals with such human capital can skillfully use advanced technologies and tools, and they are also capable of driving technological progress (Wang et al. 2021) and optimizing production processes. Especially in knowledge-intensive or technology-intensive industries, human capital can give full play to its own role. At the same time, human capital is the carrier of knowledge, and the knowledge mastered by high-skilled human capital can be disseminated and diffused through externalities and expansibility. The spillover effect of knowledge enables peer enterprises or departments to imitate and learn new technologies, increasing social benefits. When the stock of knowledge accumulates to a certain extent, a leap from “quantitative change” to “qualitative change” can be achieved. The productive capacity contained in knowledge shows a multiplicative expansion trend, realizing technological innovation. When the proportion of high-skilled human capital in the labor force is higher, the characteristic of increasing marginal returns of human capital enables it to play a greater role in technological progress. With the spillover effect of knowledge, the technological level of the entire team or enterprise is improved, and technology advances on a larger scale.

In terms of factor efficiency, the optimization of the human capital structure can improve total factor productivity. Firstly, human capital can enhance the production efficiency of itself and other workers (Lucas 1988). High-skilled human capital is equipped with superior knowledge and technical capabilities, has a stronger ability to adapt to new environments and new technologies, and is more proficient in handling complex and precise tasks, making its production efficiency several times higher than that of low-skilled workers under the same conditions. Through the spillover effect of knowledge, other workers can imitate and learn, thereby improving their own production efficiency. Secondly, human capital can improve X-efficiency. The organization and operation of various production factors within an industry require human organization and management. High-skilled human capital has a stronger sense of responsibility and self-discipline, which can improve the management level of enterprises. By establishing effective incentive mechanisms through scientific management methods and optimizing the internal governance structure, resource idleness can be reduced. Thirdly, human capital can improve the utilization efficiency of physical capital and increase the marginal product of physical capital. Human capital is the carrier of new technologies, and the use of human capital and physical capital needs to be matched. High-skilled human capital can effectively manage and utilize production tools, equipment, and other physical capital. At the same time, the nature of increasing marginal returns of human capital represents a strong substitution effect. A small amount of human capital can replace a large amount of physical capital, achieving higher productivity. Therefore, when the proportion of high-skilled human capital in the team increases and the human capital structure is optimized, total factor productivity can be further enhanced.

On the other hand, the improvement of the level of industry innovation will also drive the upgrading of the industrial structure. According to the industrial structure theory, when production factors shift to the service industry with higher efficiency, especially the knowledge-intensive or technology-intensive producer service industry, the industrial structure is optimized and upgraded. Under the condition of free flow of factors, factors such as the knowledge spillover effect of technological progress and high factor efficiency will attract more factor resources such as high-skilled human capital, physical capital, and advanced technology to further flow into the service industry (Dai et al. 2023), thus forming a comparative advantage (Guan et al. 2018), improving the efficiency of resource use, optimizing resource allocation, and further highlighting the importance of the service industry. The technological breakthroughs brought about by the improvement of the level of industry innovation can transform traditional industries and improve the production efficiency of traditional industries. The integrated use of different technologies can break down the barriers between traditional industries and promote industrial integration. At the same time, industry innovation will also create new consumer demands, give birth to emerging industries, create new economic growth points, drive the development of upstream and downstream industries, improve the industrial chain, and optimize the industrial structure.

Therefore, Hypothesis 2 is derived: The human capital structure can promote the upgrading of the industrial structure by driving technological progress and increasing total factor productivity.

Human Capital Structure, Institutional Arrangements and Industrial Structure Upgrading

On the one hand, the optimization of the human capital structure can improve institutional arrangements. As an institutional arrangement, the technology trading market is an important platform for realizing the market-oriented allocation of resources and the industrialization of innovation achievements. It connects the entire process from technological innovation to industrialization (Kong et al. 2023). By formulating trading rules and improving systems such as intellectual property protection, it ensures the fairness of transactions. Industries with a more optimized human capital structure often possess strong technological capabilities, which directly drive technological progress and accelerate the diffusion and application of technologies. High-skilled human capital can enhance the professionalism of the technology trading market. During the technology trading process, it can accurately assess the value and risks of technologies, reduce transaction costs, improve the efficiency of technology transactions, increase market attractiveness, and make the technology trading market more active.

On the other hand, the improvement of the institutional system is also an inevitable condition for the upgrading of the industrial structure. The perfection of the technology trading market implies more transparent and open information, more standardized transaction processes, higher activity levels in the technology market, and a greater variety and quantity of technological resources available for trading in the market (Xu and Liu 2024). This can improve the matching degree between the supply and demand of technologies, reduce transaction costs such as search costs, negotiation costs, and procedural costs, efficiently regulate the resource distribution between the supply and demand sides, accelerate the application, transfer, and diffusion of new technologies, and increase industrial efficiency and added value. The increase in the activity level of technology trading also enables innovation entities to more accurately grasp market demands, guide the direction of innovation, and drive the flow of more factor resources such as capital from traditional industries to emerging industries, thus optimizing the industrial structure.

Therefore, Hypothesis 3 is derived: The human capital structure can promote the upgrading of the industrial structure by facilitating technology transactions.

Research design

Econometric modeling

To test the theoretical model from the previous chapter, physical capital is introduced into the production function using Eq. (3.1) to establish a technologically neutral C-D production function.

$$Y=A({{hL})}^{\alpha }{K}^{\beta }=A{H}^{\alpha }{K}^{\beta }$$

(4.1)

In this production function, the variables are denoted as follows. $Y$ denotes the output of the economy, $A$ denotes the level of technology, $K$ denotes physical capital inputs, $h$ denotes the average years of schooling, $L$ denotes the number of persons, and $H$ denotes the human capital input. Additionally, $\alpha$ and $\beta$ denote the output elasticity of human capital input and physical capital input, respectively, and $0 < \alpha < 1,$ $0 < \beta < 1,\alpha +\beta =1$.

Equation (3.10) show that human capital structure directly affects industry structure, which is then combined with Eq. (4.1) to introduce time $i$ region $t.$ The indicators of industry structure of the explanatory variables are denoted by ${STR}.$ The indicators of the core explanatory variable human capital structure are denoted by ${SH}.$ The control variables that determine the change of industrial structure ($X$), including physical capital stock ($K$), the degree of economic openness (${FDI}$), the level of informatization ($Inf$), economic policy support (${Gov}$), infrastructure development (${Road}$), Environmental Regulation (ER). Logarithms are taken separately for each variable and controlled for time fixed effects (${\xi }_{t}$) and city fixed effects (${\theta }_{i}$). ${\varepsilon }_{{it}}$ is the random error term, and the collation yields the econometric Model I:

$${\mathrm{ln}}\,ST{R}_{it}=\alpha lnS{H}_{it}+\beta ln{K}_{it}+\lambda ln{X}_{it}+{\xi }_{t}+{\theta }_{i}+{\varepsilon }_{it}$$

(4.2)

This study first chooses to use a fixed effects model to explore the average impact of optimizing human capital structure on China’s industrial structure upgrading, while also analyzing the heterogeneity of this impact. Next, to better ascertain the marginal effect of human capital structure during the evolution of industrial structure, the study conducts further tests using a panel quantile model. This approach allows for a more detailed observation of the role played by human capital structure across different developmental stages of industrial structure.

Panel quantile regression adopts the fundamental concept of treating explanatory variables as a functional distribution. This method estimates the impact of these variables at quantile points, conditioned on them being explanatory variables. This estimation is achieved through minimizing the sum of weighted residual absolute values. By incorporating fixed effects into the quantile regression method, the system enhances robustness against the influence of extreme values and heteroskedasticity. This, in turn, assures the indivisibility of the random perturbation term. Consequently, distinct regression equations can be derived for different quantile points $\tau$($0 < \tau < 1$). The intricate structure of Model II is outlined as follows.

$${\mathrm{ln}}\,ST{R}_{it|\tau }={\alpha }_{\tau }lnS{H}_{it|\tau }+{\lambda }_{\tau }ln{X}_{it|\tau }+{\xi }_{t|\tau }+{\theta }_{i|\tau }+{\varepsilon }_{it|\tau }$$

(4.3)

Variable setting and data sources

The explanatory variable is the index of industrial structure upgrading (${STR})$. Xu (2008) believes that the industrial structure hierarchy coefficient can better reflect the trend of industrial structure upgrading. Considering that different types of manufacturing and service industries contribute differently to the upgrading of industrial structure, this paper improves the industrial structure hierarchy coefficient on the basis of Xu (2008). The manufacturing industry is divided into low-end technology manufacturing industry, middle-end technology manufacturing industry and high-end technology manufacturing industry, and the service industry is divided into productive service industry and non-productive service industry, and different weights are given to construct indicators. The measurement formula is ${\sum }_{i=1}^{n}{{SC}}_{i}\times i$. Where $i$ ranges from 1 to 6, ${{SC}}_{i}$ denote the share of GDP accounted for by the primary industry, low-end technology manufacturing, mid-range technology manufacturing, high-end technology manufacturing, non-productive services, and productive services, respectively.

Among them, the high-end technology manufacturing industry includes the manufacturing of chemicals and pharmaceuticals, general equipment, special equipment, automobiles, railroad, shipbuilding, aerospace and other transportation equipment, electrical machinery and equipment, computer, communications and other electronic equipment manufacturing, and instruments and meters. The mid-range technology manufacturing industry includes petroleum, coal and other fuel processing industry, rubber, plastics, non-metallic mineral products, ferrous and non-ferrous metal smelting and processing, metal products and other manufacturing industries. Low-end technology manufacturing industries include agricultural and food processing, foodstuffs, alcohol, beverages, tobacco, textiles, clothing, leather and feathers, timber, furniture, papermaking, printing, literature and education, and other manufacturing industries. Productive services include transportation, storage and postal services, information transmission, computer services and software, finance, leasing and business services, and scientific research, technical services and geological survey (Dai et al. 2023). Given the limitations of data availability, the substitutability of manufacturing main business revenue to output value, and the fact that the employment share and the output value share move in the same trend, this paper uses the share of manufacturing main business revenue and the employment share of the service sector to measure the shares of different types of manufacturing and service sectors in GDP at the city level, respectively.

The explanatory variable is human capital structure (${SH}$). According to the theoretical model described above, the human capital structure is expressed as ${{SH}=h}_{2}{l}_{2}/\left({h}_{1}{l}_{2}+{h}_{2}{l}_{2}\right)$.In this study, the education indicator method is chosen, The stock of human capital is the product of the average years of education and the number of the labor force. The human capital structure refers to the proportion of the stock of high-skilled human capital in the total stock of human capital. The measurement formula of years of education per capita is expressed as $\mathop{\sum }\nolimits_{j=1}^{n}{E}_{j}{W}_{j}$, ${E}_{j}$ represents the years of education for the $j$-type cultural level, ${W}_{j}$ represents the weight of the labor force that accepts the cultural level of $j$. The weights of the labor force with j = 1, 2, 3, 4, 5 denote the employed people with no schooling, elementary school literacy, middle school literacy, high school literacy (including secondary vocational education), and college and above literacy (including higher vocational education), respectively. The corresponding years of education are 3, 6, 9, 12, and 16 years, respectively. This study distinguishes between individuals with high school education or below as low-skilled human capital and those with college education or above as high-skilled human capital. Therefore, through calculation, the formula for the human capital structure can be simplified as ${SH}={E}_{5}{W}_{5}/\mathop{\sum }\nolimits_{j=1}^{n}{E}_{j}{W}_{j}$.

The control variables are as follows. (1) Physical capital stock ($K$), which is selected with reference to the method used by Shan (2008) and calculated using the perpetual inventory method, ${K}_{t}=\left. (1-\delta \right){K}_{t-1}+\Delta {I}_{t}/{P}_{t},$ $\delta$ for a depreciation rate of 10.6%. Due to the lack of data on gross fixed capital formation and its deflator at the city level, this study refers to the methodology used by Xu (2017), and replaces it with the amount of investment in fixed assets and its corresponding deflator of investment in fixed assets, ${I}_{t}$ represents the fixed assets investment amount for period $t$, ${P}_{t}$ is the price level for period $t$. To mitigate potential biases, the year 1992 is selected as the base period in this study. Initially, the provincial fixed capital stock data for 1992 (obtained from Shan (2008)) is adjusted to the municipal level. This adjustment is based on the relative proportions of fixed asset investments in each prefecture-level city compared to the total fixed asset investment across the entire province in the same year. This process allows for the determination of the fixed capital stock of the prefecture-level cities at the base period. Subsequently, the fixed asset investment deflator, which comprises provincial data, is utilized. The fixed asset investment amount for the years 2017–2019 was calculated based on the growth rate. (2) The degree of openness to the outside world (${FDI}$), expressed as the proportion of the total amount of foreign capital actually utilized by each region in that year to the regional GDP of that year after being converted into RMB at the exchange rate. (3) Level of informationization (${Inf}$), expressed as the number of international Internet users per 10,000 people. (4) Economic policy support ($G{ov}$), expressed as the proportion of fiscal expenditure of each regional government to the GDP of the year. (5) Financial support (${Fin}$), expressed as total loans as a share of GDP for the year in each region. (6) Infrastructure development (${Road}$), expressed as the ratio of the area of real urban roads to the area of urban areas at the end of the year. (7) Environmental Regulation (${ER}$), expressed as the proportion of industrial pollution treatment investment to industrial added value.

In this study, panel data from 284 prefecture-level and higher cities in China from 2003 to 2019 are selected. The data include information from various sources such as the China Urban Statistical Yearbook, the China Labor Statistical Yearbook (which provided labor force weights for different educational levels at the provincial level), as well as the China Statistical Yearbook and statistical yearbooks from different provinces and cities in previous years. However, data from specific regions, including Chaohu City in Anhui Province, Laiwu City in Shandong Province, Sansha City and Danzhou City in Hainan Province, Bijie City and Tongren City in Guizhou Province, Turpan City and Hami City in Xinjiang Province, Tibet, Hong Kong, China, Taiwan, China, and Macao, China are excluded due to data unavailability. To address missing data for certain regions and years, interpolation is employed. Moreover, variables relate to price factors are standardized to 2003 values, with fluctuations in price levels adjusted using relevant indices. Additionally, all variables are log-transformed for analysis purposes.

The descriptive statistics of the main variables are shown in Table 1 below. During the sample period, the coefficient of industrial structure upgrading exhibits variations, with a mean value of 3.67, a minimum value of 2.73, and a maximum value of 5.21. This indicates notable differences in the industrial structure over time. Moreover, the mean value for human capital structure stands at 19.86, revealing a current need for enhancements in the level of human capital structure.

Table 1 Descriptive statistics of the main scalars.

Empirical analysis

Baseline regression

According to the results of Hausman test, the original hypothesis that there is no systematic difference between the fixed effects model and random effects is rejected, thus this study uses the empirical method of fixed effects to verify the impact of human capital structure on industrial structure. Regression Model I is empirically analyzed, and its results are shown in Table 2.

Table 2 Results of the basic regression on the impact of human capital structure on industrial structure.

Table 2 demonstrates the results of the basic regression of the impact of human capital structure on industrial structure. The column (1) indicates the results without the inclusion of control variables, while columns (2)–(4) display the outcomes after incorporating these control variables. Irrespective of the presence of control variables, the results consistently reveal significantly positive coefficients of the core explanatory variable human capital structure significantly positive. The research highlights that an increase in the proportion of high-skilled human capital is associated with a substantial enhancement in industrial structure.

As the theoretical analysis indicates, with the gradual evolution of low-skilled human capital into high-skilled human capital, the human capital structure is optimized. In this process, production continuously develops towards automation and intelligence, and the original technology transforms into automated technology. The adjustment pattern of the human capital structure dominated by high-skilled human capital will continuously promote the digestion, absorption, and application of various technologies and trigger innovation, thus driving the industry to gradually shift from labor-intensive industries to knowledge-intensive industries (Liu et al. 2018). In similar studies, Acemoglu (2003) have investigated how a rise in the proportion of skilled labor contributes to industrial development by highlighting the faster growth in the marginal productivity of skilled labor compared to unskilled labor. Similarly, the absence of skilled labor within the industrial structure could hinder the development of local industrial structures and stifle economic returns (Teixeira and Queirós 2016).

After adding all control variables in the regression model, it is found that every 1% optimization of human capital structure leads to a 0.0202% promotion in industrial structure upgrading. The results from column (4) show that the level of information technology and financial support exhibit significant positive relationships with industrial structure upgrading. This suggests that both information technology and financial support play crucial roles in driving industrial structure upgrading. Information technology, with its positive externality, not only boosts the science and technology sector but also reduces processing cost in various industries, ultimately enhancing technology levels and productivity. Meanwhile, an appropriate financial structure and scale can offer the necessary financial support for industrial transformation and address financing requirements. Although the coefficients of physical capital stock and degree of openness to the outside world are positive, they are not statistically significant. This could indicate that their specific effects vary across different economic development stages. Conversely, the coefficient for economic policy support is significantly negative, while the coefficient for infrastructure construction, although not significant, is negative. The negative coefficient of economic policy support may be due to the mismatch between the implementation of policies and the regional resource endowments. There are differences in regional resource endowments. When formulating the direction of industrial development, the implementation of policies may not be based on the advantages of regional resource endowments, and instead guides resources to flow into industries that do not match the regional resource endowments. The implementation of policies also requires flexibility. Once it lags behind the changes in resource endowments, it will lead to the distortion of resource allocation and a decline in social efficiency (Zhang et al. 2011). The negative coefficient of infrastructure may be because the road infrastructure construction in various cities has been relatively complete. Further increasing investment in infrastructure may reduce the output elasticity. Road infrastructure has a greater impact on traditional manufacturing industries, while high-tech industries rely more on infrastructure such as high-speed communication networks. If the local infrastructure does not conform to the resource endowments and the demands of the leading industries, it will be unable to effectively promote the upgrading of the industrial structure and may even have a negative impact.

Robustness tests

To assess the robustness of the impact of human capital structure on the upgrading of industrial structure, the Model I underwent robustness testing. The details are as follows. (i) Replace the core explanatory variable. This study regards employees with an education level of a college degree and above in each region as human capital. Therefore, the proportion of this segment of the population is considered the human capital structure (${Sh}$) and replaces the core explanatory variables (${SH})$. The results of the regression analysis are shown in Table 3, paragraph (1), with the regression outcomes depicted in column (1) of Table 3, indicating robust results. (ii) Replace the explained variables. First, the index of industrial structure upgrading is remeasured by considering the technology intensity of different industries. In constructing the index, the original weights are replaced with the intensity of R&D investment funding across different industries, and a new index of industrial structure upgrading (${str}$) is calculated. The regression results are shown in column (2) of Table 3, indicating robust results. Second, drawing on the setting of Fu (2010), an index known as industrial structure supererogation (${TS}$) is established. This index characterizes the systematic evolution of industrial structure, where the proportion of the three industries progresses in the order of the first, second and third industries. The specific settings are as follows: set the proportion of each of the three industries in GDP as a component of the space vector ${X}_{0}=\left({x}_{\mathrm{1,0}},{x}_{\mathrm{2,0}},{x}_{\mathrm{3,0}}\right)$, calculate its vectors with the arrangement of industries from low level to high level respectively X₁ = (1, 0, 0), X₂ = (0, 1, 0), X₃ = (0, 0, 1), the angle of the vectors are respectively θ₁, θ₂, θ₃:

$${\theta }_{j}=arccos\left(\frac{\mathop{\sum }\limits_{i=1}^{3}{x}_{i,j}\cdot {x}_{i,0}}{\left(\mathop{\sum }\limits_{i=1}^{3}{\left({x}_{i,j}^{2}\right)}^{\frac{1}{2}}\cdot \mathop{\sum }\limits_{i=1}^{3}{\left({x}_{i,0}^{2}\right)}^{\frac{1}{2}}\right)}\right),j=1,2,3$$

(5.1)

$$TS=\mathop{\sum }\limits_{k=1}^{3}\mathop{\sum }\limits_{j=1}^{k}{\theta }_{j}$$

(5.2)

Table 3 Robustness test.

The larger the ${TS}$, the higher the level of advanced industrial structure. The regression results are shown in column (3) of Table 3, indicating robust results. (iii) Shrinking treatment. Regressing again after shrinking all variables up and down by 1%, assigning a value of 1% to numbers less than 1% and 99% to numbers greater than 99%, while keeping the sample size unchanged, yielded the results are presented in column (4) of Table 3. Analysis of these results reveals that they are robust, indicating that the original regression results remain unaffected by extreme values. (iv) Excluding 2008 data. In order to prevent the influence of the global financial crisis that took place in 2008 on the regression results, this study excludes the data from that year and re-conducted the regression analysis. Therefore, the results presented in column (5) of Table 3 demonstrate the robustness of the results.

Endogeneity test

The relationship between human capital structure and industrial structure may exhibit endogeneity problems due to the complex nature of factors affecting both structures. While this study has made efforts to include as many factors affecting industrial structure as possible in the model, the intricate relationship between human capital and industrial structure could lead to omitted variables that create correlations between explanatory variables and residual. Moreover, the intricate and potentially reciprocal causation between human capital structure and industrial structure further complicates the issue. To address these concerns and mitigate endogeneity problems stemming from omitted variables and reverse causation, this study utilized the instrumental variable method.

(i) This paper draws on the approaches of Bartik (2009), Yi and Zhou (2018) to construct the Bartik instrument, which is represented by the product of the human capital structure at the first lag order ${{SH}}_{j,t-1}$ and the temporal difference of the national human capital structure, ${\Delta{SH}}_{t,t-1}$. The reasons are as follows: On the one hand, since the national human capital structure will not be significantly affected by the industrial structure of a certain prefecture-level city, the change in the national human capital structure is relatively exogenous to a specific prefecture-level city. On the other hand, demand shocks at the prefecture-level city level, other than the human capital structure, may also lead to estimation bias. However, as long as a single prefecture-level city is not so important that its internal demand shock is significantly correlated with the national human capital structure, the Bartik instrument is effective. The results of two-stage Least Squares (2SLS) regressions are shown in column (1) and (2) of Table 4. The stage 1 F-statistics yielded values of 283.23, instrumental variables are significantly demonstrated at the 1% level, confirming the validity of the instrumental variables. Moreover, the estimated coefficients on human capital structure in stage 2 remain significantly positive, even after accounting for endogeneity, thus reinforcing the robustness of the results.

Table 4 Endogeneity test.

(ii) The selection of the second instrumental variable comes from the article of Chen and Zhang (2016), which chooses the cross term of the number of colleges and universities in 1998 and the expansion scale of national colleges and universities with a lag of four years as the instrumental variable of human capital structure. University expansion in China began in 1999, leading to a rapid increase in human capital, but with significant regional differences. University expansion increases the proportion of high-skilled human capital in cities with universities but does not affect cities without higher education institutions. Therefore, the exogenous policy shock of university expansion has different impacts on the growth of high-skilled human capital across regions, and this regional heterogeneity can serve as an exogenous source. As shown in column (3) and (4) of Table 4, the instrumental variable is significant at the 1% level, with an F-statistic of 233.75, confirming its validity. The estimated coefficient for human capital structure in the second stage remains significantly positive, indicating the robustness of the results.

Heterogeneity test

The relationship between human capital structure and industrial structure is complex, and the mechanism of action between the two varies in different regions or at different stages of economic growth. As a result, this study conducts the test of time heterogeneity and the test of city size heterogeneity, aiming to analyze how these variables interact and influence each other over time and across different urban sizes.

(i) Temporal heterogeneity test. In 2013, China transitioned into the stage of industrial structure servicization, as identified by Xia (2010) and the determination criteria of industrial structure servicization. The roles played by human capital structure differ between the stages of industrialization and industrial structure servicization. To further explore this phenomenon, a re-regression analysis is conducted on data spanning from 2003 to 2012 and from 2013 to 2019, with the corresponding results present in columns (1) and (2) of Table 5. Interestingly, during the period of 2003 to 2012, the impact of human capital structure on industrial structure is found to be statistically insignificant. This suggests that during the industrialization stage, the relationship between human capital structure optimization and industrial structure upgrade is not apparent. However, a shift is observed in the subsequent stage of industrial structure servicization from 2013 to 2019. During this period, optimizing human capital structure is found to significantly promote the upgrading of industrial structure. Specifically, for every 1% improvement in human capital structure optimization, industrial structure upgrade is promoted by 0.0285%. This estimated coefficient is notably higher than that of the whole sample, pointing to a marked difference in the impact of human capital structure at various stages of industrial structure evolution, in alignment with theoretical analysis. This also confirms the theoretical model proposed by Wang and Tang (2019) that optimal human capital is stage-specific and changes with the industrial structure.

Table 5 Heterogeneity test.

(ii) City size heterogeneity test. According to the “Notice on Adjusting the Standard for the Division of City Size”, this study classifies cities with a population of less than 500,000 as small cities and those with a population of over 500,000 as large cities. The results of the heterogeneity testing are presented in columns (3) and (4) of Table 5. The results show that the effect of human capital structure on industrial structure is insignificant for small cities. However, a positive and statistically significant estimated coefficient is observed for large cities. Specifically, the coefficient value of 0.0326 in large cities exceeds the 0.0124 coefficient for the whole sample, suggesting that optimizing human capital structure has a more pronounced effect on industrial structure upgrading in larger cities. This study analyzes the reasons behind this result by examining relevant literature. The findings suggest two main factors. First, in large cities with relatively developed economies, the tertiary industries is prominent, especially in emerging industries. However, there is often a shortage of high-skilled labor in these emerging industries (Zhao et al. 2023), which could benefit from better matching with high-skilled labor to create more job opportunities and increase industrial value-added. Second, urbanization efforts aimed at population growth have raised the overall level of human capital accumulation (Zhao et al. 2023). In cities with higher levels of human capital, high-skilled labor can play a more significant role and become a key driver of industrial structural upgrading.

Dynamic analysis based on quantile regression

Analysis of the evolution of the marginal effects of human capital structure

The fixed effects model, while able to depict the average impact of human capital structure on industrial structure, lacks the ability to demonstrate the dynamic role of human capital structure in the evolution of industrial structure. The heterogeneity test conducted provides further evidence of this limitation. To avoid falling into the “evidence trap” of the average impact coefficients, empirical analysis is conducted using regression Model II, with the results are shown in Table 6.

Table 6 Analysis of the evolution of the marginal effects of human capital structure.

Table 6 reveals a significant difference in the marginal effect of human capital structure on industrial structure across different levels of industrial structure. The estimated coefficients for human capital structure are significantly positive, ranging from 0.3 to 0.9, but they are not statistically significant in the 0.1–0.2 decile. The estimated coefficients increase with increasing quantile points. This shows that the impact of human capital structure on industrial structure becomes more apparent as the industrial structure matures.

The increase in the marginal effect of human capital structure indicates that with the transformation of industries from labor-intensive to technology-intensive and knowledge-intensive, the industrial structure is more and more dependent on the optimization of human capital structure. Initially, in the phase of industrial structure backwardness, labor-intensive industries hold a relative advantage, making low-skilled human capital sufficient for industrial structure development, thus minimizing the effect of optimizing human capital structure. The rise of emerging industries requires a large number of talents with cutting-edge technology and strong innovation ability, and the transformation of traditional industries also requires a large number of high-skilled human capital. Optimization of human capital structure can promote knowledge flow and technology diffusion among industries, strengthen inter-industry collaboration, and promote industrial synergy and transformation. With higher-skilled labor being more sought-after in advanced stages of industrial structure development, while the need for low-skilled human capital decreases.

Liu and Yuan (2023) and Liao et al. (2023) have discussed the relationship between industrial structure and human capital at the theoretical level. They argue that without a certain industrial structure, it is difficult to generate higher demand for skilled labor. This study empirically validates this conclusion by focusing on the marginal effects of human capital structure, providing stronger evidence of the evolution of human capital structure marginal effects during the process of industrial structural evolution.

Similarly, the results of this study echo those of Wu and Zhu (2023) as well as Wang and Li (2023). Wu and Zhu (2023) argue that if a significant portion of the labor force is comprised of uneducated or unskilled workers during the transition to a more advanced industrial structure, it may result in a human capital deficiency, hindering the alignment with the evolving industrial framework. Their argument revolves around the idea that the composition of educational inputs can shape the quality of economic progression by influencing the human capital arrangement and its synchronization with the industrial landscape. Contrarily, our study delves deeper into the marginal effect of human capital structure during the process of deepening industrial structural adjustments. Meanwhile, Wang and Li (2023) argue that for human capital structure to effectively promote industrial structural enhancement, it must be congruent with upgrading the prevailing industrial configuration. Should the human capital structure deviate from the industrial framework, its ability to propel industrial growth may be compromised, potentially leading to an inhibitory effect. In contrast to our investigation, Wang and Li (2023) focus on scrutinizing the disparity between human capital structure and industrial composition, while our study concentrates on elucidating the marginal effects of human capital structure, offering valuable insights for policymakers in different developmental stages of industrial restructuring.

In terms of control variables, the estimated coefficients decrease with the increase of the quantile for information technology level, with all coefficients being significant at the 0.1–0.8 quantile. This suggests that information technology support could yield a higher rate of return during stages of relatively backward development in the industrial structure. Conversely, the estimated coefficients of financial support increase with the quantile for financial support, which are also all significant at the 0.1–0.9 quantile. This indicates that financial support may offer a higher rate of return during stages of relatively advanced industrial structure development. The issue of resource allocation distortion and efficiency loss in the society is not alleviated by economic policy support. Other control variables are not show significance, potentially due to variations in resource endowments across different regions.

To further verify the robustness of the results, this paper replaces the core explanatory variable ${SH}$ with ${Sh}$ and conducts quantile regression again. As shown in the first row of Table 7, the results are robust.

Table 7 Analysis of the evolution of the marginal effects of human capital structure.

Analysis of the evolution of the marginal effect of human capital structure from the perspective of different city sizes

Table 7 presents the empirical results regarding the heterogeneity of the marginal effect of human capital structure on the impact of industrial structure across various quartiles in cities of different sizes. Two indicators have been chosen to gauge the central explanatory variable of human capital structure for the analysis through quantile regression. In small cities, the estimated coefficient of human capital structure is not achieve statistical significance. Conversely, in large cities, the estimated coefficients of human capital structure exhibited significance with the 0.5–0.9 decile range, with the marginal effect of human capital showing a progressive enhancement. This trend may be attributed to the advanced development of the tertiary industry in large cities, while small cities remain predominantly reliant on the secondary industry. In large cities, high-skilled human capital enjoys superior employment prospects, contrasting with the more suitable environment for low-skilled human capital in small cities. For the human capital structure to have a meaningful impact, it needs to align with the prevailing industrial structure in the locality. This observation suggests that the efficacy of human capital structure become evident only when the industrial structure has sufficiently progressed, at which point its influence grows progressively stronger.

Mechanism testing

The empirical evidence proves that the optimization of human capital structure is shown to facilitate the upgrading of industrial structure. In the previous theoretical analysis, promoting technological progress, increasing total factor productivity, facilitating technology transactions are the three aspects for realizing the upgrading of the industrial structure. If all variables are included in the same empirical model, the problem of two-way causality may become more prominent. In order to avoid more complex endogeneity problems, this paper conducts a mechanism test by identifying the impact of the core explanatory variable on the mechanism variable (Jiang 2022). The integrity of the transmission path is supplemented by the logical path of the impact of the mechanism variable on the explained variable expounded in the previous theoretical analysis, as well as the further demonstration of the relevant conclusions in the reference literature, so as to comprehensively prove the validity of the mechanism variable. The results are shown in Table 8.

Table 8 Mechanism testing.

(ⅰ) Technological progress (${RD}$). The previous theoretical analysis shows that technological innovation is an important factor in promoting the upgrading of the industrial structure, and existing research generally supports this view. It is proposed by Wang and Li (2013) that technological progress serves as the initial point for industrial structure upgrading. Technological progress plays a particularly prominent role in the secondary and tertiary industries. It increases the marginal revenue of these two industries, attracting a large number of labor forces to transfer. Technological progress will also further promote the deepening of the division of labor and open up new industrial sectors, thereby enhancing the ability to absorb agricultural labor forces. Ye et al. (2022) proposed that technological innovation can promote knowledge exchange among industries and drive the development of the industrial structure. Under the effect of interconnected knowledge spillover, there is a significant positive correlation between the two.

Referring to the mechanism testing method of Jiang (2022), if the optimization of the human capital structure can significantly promote technological progress, it can also be indirectly demonstrated that technological progress is an important influencing mechanism by which the human capital structure promotes the upgrading of the industrial structure. Therefore, this paper selects research and development expenditure as the measurement indicator of technological progress, and the regression results of the mechanism test are shown in Column 1 of Table 8. The results show that the coefficient of ${SH}$ is 0.2144 and is significant at the 5% confidence level, indicating that the human capital structure can significantly promote technological progress, thereby optimizing the upgrading of the industrial structure.

(ⅱ) Total Factor Productivity (${TFP}$). As the theoretical analysis shows, under the condition of free flow of factors, factor resources will be transferred to industries with higher efficiency. Industries with high Total Factor Productivity (TFP) will attract more factor resources such as high-skilled human capital, physical capital, and advanced technologies to further flow in, thus realizing the continuous upgrading of the industrial structure. In relevant research, Yu et al. (2022) proposed that TFP can analyze the performance of industries themselves, such as industrial production efficiency and technological changes, and further identify industries in the rapid development stage. Therefore, improving TFP is the foundation of industrial structure adjustment and also a new way for the optimization of the industrial structure.

In this paper, the DEA-Malmquist method is used to measure the Total Factor Productivity. The regression results of the mechanism test are shown in Column 2 of Table 8. The results show that the coefficient of ${SH}$ is 0.0604 and is significant at the 1% confidence level, indicating that the human capital structure can significantly increase TFP. It also further shows that TFP is an important influencing mechanism by which the human capital structure promotes the upgrading of the industrial structure.

(ⅲ) Technology transaction (${TT}$). The preceding theoretical analysis indicates that an enhanced level of activity in technology transactions not only enables a better alignment of the needs between suppliers and demanders, thereby reducing transaction costs, but also stimulates technology dissemination and the conversion of research outcomes. It guides the direction of innovation, propels the flow of more factor resources like capital from traditional industries to emerging ones, and consequently optimizes the industrial structure. Prevailing research commonly acknowledges that the vibrancy of the technology transaction market is a pivotal factor driving the upgrading of the industrial structure. For instance, Xu and Liu (2024) posited that the technology market can expedite the transformation of technology commodities by supplying complementary technological resources. It effectively addresses the technological deficiencies within industries, enhances the efficiency of resource allocation, and fosters the upgrading of the industrial structure.

In this paper, the activity level of technology transactions is gauged by the proportion of the technology market’s transaction volume to the GDP. The regression results of the mechanism test are presented in Column 3 of Table 8. The findings reveal that the coefficient of ${SH}$ is 0.1824 and is statistically significant at the 5% confidence level. This suggests that the human capital structure can notably boost the activity level of technology transactions, which in turn propels the upgrading of the industrial structure.

A further examination of the structure of trained human capital

The previous discussion on the measurement of the human capital structure index was limited to the educational level. In order to demonstrate the influence of the multi-dimensional aspects of the human capital structure on the upgrading of the industrial structure, this section introduces the training human capital structure from the perspective of skill training. Vocational skill appraisal certificates serve as evidence of workers’ vocational skill levels and can be used to measure the level of human capital from the dimension of skill training. Vocational skill level certificates are classified into five categories: intermediate level, primary level, advanced level, technician level, and senior technician level. In this paper, the proportion of the number of people who have obtained advanced certificates or above is used as the measurement indicator for the training human capital structure (${TSH}$). For the convenience of distinction, in this section, ${SH}$ is referred to as the educational human capital structure.

In Column (1) of Table 9, the training human capital structure, the educational human capital structure, and the interaction term between the two are included in the regression without introducing control variables. The results show that although the coefficient of the training human capital structure is smaller than that of the educational human capital structure, the training human capital structure also has a significant positive impact on the industrial structure. The interaction term between the training human capital structure and the educational human capital structure is significantly negative, which indicates that there is an obvious substitution effect between training and education. When the training human capital structure is optimized, the influence of the educational human capital structure on the upgrading of the industrial structure is inhibited to a certain extent.

Table 9 Alternative role of training human capital structure.

The regression results with control variables introduced are shown in Column (2) of Table 9. The significance of the training human capital structure, the educational human capital structure, and the interaction term between the two remains unchanged, and the results are robust. To further demonstrate the robustness of the results, this paper replaces ${SH}$ with Sh and reruns the regression, as shown in Columns (3) and (4) of Table 9. The results are robust, further confirming that among the influencing factors promoting the upgrading of the industrial structure, the training human capital structure can substitute for the educational human capital structure.

In order to further analyze the changes in the substitution effect of the training human capital structure on the educational human capital structure during the process of promoting the upgrading of the industrial structure, this subsection conducts a quantile regression, and the results are shown in Table 10. The results show that the interaction term is only significant at the quantiles of 0.1–0.5, and it is not significant at the quantiles of 0.6–0.9, indicating that this substitution effect gradually decreases as the industrial structure deepens.

Table 10 Quantile regression of training human capital structure.

In regions dominated by traditional industries, through skill training, workers can quickly master advanced production technologies and improve production efficiency. For example, if agricultural practitioners receive training in modern planting and breeding technologies, the level of agricultural mechanization can be increased. As the industrial structure deepens, knowledge-intensive and technology-intensive industries develop, which requires a higher level of education for the labor force. Correspondingly, the demand for and dependence on vocational skill training are relatively low, and the marginal benefit of skill training is difficult to cover the training cost. Therefore, as the industrial structure deepens, the substitution effect of the training human capital structure on the educational human capital structure gradually decreases, and the upgrading of the industrial structure becomes more dependent on the educational human capital structure.

Conclusions and recommendations

Through theoretical and empirical analyses of panel data from 284 prefecture-level and above cities in China, this study delves into the impact of human capital structure on industrial structure upgrading and the dynamic evolution of human capital structure. The results of the study reveal that the optimization of human capital structure can significantly promote the upgrading of the industrial structure, but this promotion is not significant during the industrialization stage and in small cities. This allows us to discover the importance of human capital structure optimization, but also makes it clearer that the human capital structure needs to match the development of the industrial structure. Large cities are often the agglomeration of technological innovations and high-end industries, while many small cities are dominated by a single traditional or resource-based industry, and there are differences in the industrial structure of the two types of cities, which require naturally different industrial policies and talent policies.

The results of the quantile regression show that as the adjustment of the industrial structure deepens, the marginal effect of the human capital structure gradually increases. This conclusion proves that the process of deepening industrial structural adjustment depends more and more on the human capital structure, and that regions dominated by knowledge-intensive industries should pay more attention to optimizing the human capital structure and encouraging the flow of talents to key and emerging industries when formulating talent policies. Furthermore, further mechanism testing uncovers that Technological progress, total factor productivity, technology transactions are the mechanisms through which the human capital structure influences the industrial structure. Skill training plays a role in substituting for education, and this substitution effect gradually decreases as the industrial structure deepens. Consequently, these results yield pertinent policy recommendations.

First, in order to optimize the human capital structure within the service-oriented industrial structure, it is crucial to emphasize the quality of education at different levels. Human capital investment has progressively replaced material capital investment as the primary driver of economic growth. While investing in higher education is important for cultivating high-skilled human capital, it is equally essential to focus on basic education and ensure the seamless articulation between various education levels. Basic education serves as the cornerstone for higher education, allowing for the creation of a well-rounded talent pool and facilitating the evolution of human capital from low-skill to high-skill levels. By establishing a solid foundation in basic education, the human capital structure can be optimized effectively. In regions dominated by traditional industries such as agriculture and simple manufacturing, more emphasis should be paid to skill training. International experience can be drawn upon, such as Germany’s dual vocational education system which combines traditional apprenticeships with modern vocational education to cultivate substantial numbers of specialized technical professionals. Policy guidance can be strengthened. Special skill training funds can be set up in training programs for in-demand job types. Enterprises should be encouraged to strengthen cooperation with vocational colleges and training institutions, improve the training system, and regulate the order of the training market.

Second, local governments should tailor the characteristics of regional industries, formulate policies for the introduction of talents and industrial restructuring plans, and promote the coordinated development of human capital structure and industrial structure. For instance, significant differences exist between resource-based and technology-driven cities in terms of their industrial foundations and human capital endowments, necessitating the adoption of differentiated strategies. Resource-based cities exhibit over-reliance on traditional heavy industries, rigid industrial structures, and limited skill sets among traditional industry workers; these cities should leverage their industrial base advantages, prioritize coordinated development between strategic emerging industries and traditional manufacturing sectors, and cultivate technical and skilled talents aligned with occupational requirements. Illustratively, the old industrial bases in Northeast China are capitalizing on their resource advantages by formulating action plans for emerging industries, fostering innovative sectors and business models, implementing pilot reforms for dual-training programs of technical personnel during industrial transformation, and cultivating more “big country craftsmen” through systematic initiatives. Technology-driven cities, conversely, are characterized by dominance in high-tech and productive service industries, relying on human capital configurations to fulfill innovation-driven functions; these cities should enhance high-skilled human capital reserves and advance industrial upgrading through scientific and technological innovation. Illustratively, the Yangtze River Delta innovation clusters has tailored policies to support and safeguard high-skilled human capital, intensified the cultivation of strategic scientific capabilities through national laboratory development and allied initiatives, advanced technology transfer and commercialization of research outcomes, and expanded high-end service industry chains—thereby establishing a strategic talent hub and technological innovation epicenter. China demonstrates regional disparities in both human capital composition and industrial distribution patterns. Implementing differentiated strategies while emphasizing city-specific policy priorities constitutes a fundamental requirement for ensuring structural alignment between human capital structure and industrial architecture.

It is important to recognize the limitations of this study. The study covers data from 2003 to 2019, focusing on the impact of human capital structure on industrial structure before the epidemic, which can exclude the interference of sudden exogenous shocks, such as epidemics, so as to reveal the dynamic evolution of the marginal effect of human capital structure more clearly. Therefore, the findings of this study are more applicable to the normalized economic environment. The realistic relevance of the study can be enhanced in future studies by incorporating post-2020 data using appropriate research methods. In addition, this study assessed human capital structure mainly from the perspective of education, although in further tests, this study supplemented the measurement of human capital structure from the perspective of skills training, human capital contains multiple dimensions that should be further added in future studies.