The Impact of Green Technological Innovation on Industrial Structural Optimization Under Dual-Carbon Targets: The Role of the Moderating Effect of Carbon Emission Efficiency

Abstract

According to the “dual-carbon” goal, the solution to achieving balanced regional development and industrial structural optimization while promoting sustainable development goals lies in the synergistic evolution mechanism of carbon emission efficiency and green technological innovation. Using provincial panel data from China from 2000 to 2022, this study calculates industrial structural optimization coefficients for the “advanced,” “rationalization,” and “ecology” dimensions. The impact of green technological innovation on industrial structural optimization is experimentally explored using panel regression, threshold effect, and mediating effect methodologies, based on the constraint perspective of carbon emission efficiency. The findings show that (1) the optimization of the regional industrial structure is successfully driven by both carbon emission efficiency and green technological innovation; (2) the impact of green technological innovation on industrial structural optimization is positively regulated by carbon emission efficiency, which also helps to achieve sustainable economic transformation; and (3) this moderating effect exhibits significant regional heterogeneity and U-shaped nonlinear characteristics, in the order of “central > west > east”. This study reveals how green technological innovation affects industrial structural optimization under the constraint of carbon emission efficiency. It offers reference recommendations for the creation of sustainable development policies in the future.

1. Introduction

In the fast-developing economy, the problems of resource depletion and environmental degradation are becoming more significant [1,2,3,4] and the conventional economic growth model is no longer sustainable. Faced with the combined difficulties of economic downturn and environmental problems, establishing carbon neutrality and supporting sustainable development have become the shared aims of all countries [5,6,7].

Technological innovation has always been an effective tool for accelerating China’s economic transition and upgrading. Green technological innovation, as an emerging technical tool that can minimize resource consumption and pollution [8], strikes a balance between economic growth and environmental protection, paving the way for long-term development. Traditional industries with significant pollution and energy consumption are no longer meeting the demands of modern economic development [9]. Green technical innovation provides a fresh approach to industrial structural optimization. It gives a new impetus for transforming traditional industries. Advanced green technologies enable the transformation of conventional sectors into low-carbon operations. These technologies contribute to the development of environmentally friendly and high-efficiency industrial systems. This dual effect drives industrial upgrading and structural optimization. Such transformations significantly contribute to China’s high-quality economic development [10]. By optimizing industrial structures, green technological innovation lays the foundation for building a more resilient and sustainable economic future.

Aggregate carbon emissions have become a key indicator of environmental health under China’s dual-carbon targets. The carbon trading mechanism serves as a market-based solution to those emissions. The data from China’s Ministry of Ecology and Environment shows that the national carbon market covers over 4.5 billion tons of aggregate CO₂ emissions annually. Reducing carbon emissions, improving carbon emission efficiency, and realizing carbon neutrality will become the main themes of China’s economic development in the foreseeable future. The inevitable and continuous interaction between carbon emission reduction and green technological innovation makes it impossible to ignore the potential dynamic impact of carbon emission efficiency.

Accordingly, this study addresses the following research questions: (1) How do green technological innovation and carbon emission efficiency impact industrial structural optimization? (2) How does carbon emission efficiency moderate this impact? (3) What are the regional differences in the performance of green technological innovation and industrial upgrading? These questions guide the empirical analysis and theoretical exploration of this study.

The research goal of this paper is to explore the synergistic mechanism between green technological innovation and carbon emission efficiency in driving industrial structural optimization under the dual-carbon framework. The study aims to (1) theoretically and empirically identify the impact pathways of green technological innovation and carbon emission efficiency on industrial structural optimization; (2) uncover the moderating role of carbon emission efficiency in the relationship between green technological innovation and industrial upgrading, with a focus on its nonlinear and regional heterogeneous characteristics; and (3) provide a comprehensive evaluation of regional disparities in the effectiveness of green technological innovation for industrial optimization, thereby offering targeted policy implications for regional sustainable development.

This paper explores the relationship between green technological innovation, carbon emission efficiency, and the industrial structure through discussion and answers related questions. It explores a new means of industrial structural optimization, which is of great practical significance for realizing the dual-carbon goal and high-quality economic development. The synergy between green technological innovation and carbon emission efficiency provides replicable solutions for building environmentally friendly industrial systems. This systemic transformation drives China’s green, high-quality economic development. More importantly, it represents a critical pathway for achieving global sustainability goals.

This paper makes the following three key contributions.

Firstly, this study expands the theoretical understanding of how green technological innovation drives industrial structural optimization under the dual-carbon targets. We integrate the carbon emission efficiency and green innovation into a unified framework. This work enriches research on the mechanisms of China’s green transition.

Secondly, this paper adopts a moderating-effect model and a multi-threshold panel model. We reveal green technological innovation’s nonlinear and stage-dependent impacts on industrial structural optimization. This approach offers greater explanatory power than traditional linear methods.

Thirdly, we develop a composite indicator for industrial structural optimization based on three dimensions: advancement, rationalization, and ecologicalization. This new measure overcomes the single-indicator limitation in existing studies. It improves both the completeness and the accuracy of our assessment.

This paper is organized as follows: Section 2 systematically reviews the existing literature on green technological innovation, carbon emission efficiency, and industrial structural optimization. Section 3 establishes the theoretical mechanisms and discusses the research assumptions. Section 4 details the econometric model design, variable definitions, and data sources. Section 5 empirically tests the hypotheses using the model estimates, showing the empirical results and discussion. Section 6 synthesizes the main conclusions and provides actionable policy recommendations for achieving the dual-carbon targets. Section 7 analyzes the limitations of this study and prospects for future research.

2. Literature Review

To clarify the existing literature on the relationships between green technological innovation, carbon emission efficiency, and industrial structural optimization, this study conducts a literature review on three key aspects. Exploring these relationships is fundamental for advancing sustainability, as they directly influence resource utilization, environmental impact, and long-term economic viability.

2.1. Technological Innovation and Industrial Structural Optimization

The existing literature has conducted in-depth research on technological innovation and industrial structural optimization issues. These investigations have produced substantial results. Current studies have agreed on the role of technological innovation in promoting industrial structural optimization based on different perspectives. As sustainability becomes a global priority, the focus has increasingly shifted to green technological innovation, which aims to balance economic growth with environmental protection. An increasing number of scholars have examined whether green technological innovation can drive the optimization and upgrading of industrial structures. Existing studies have acknowledged its positive impact from various research perspectives, indicating its potential to create more resource-efficient and environmentally friendly industrial systems, key elements of the sustainability goals.

New Growth Theory identifies technological innovation as a core driver of sustained economic growth [11]. This theoretical framework further posits that dynamic technological progress explains fluctuations in total factor productivity over the short to medium term [12]. Klepper [13] shows that innovation, especially technological innovation, drives the development of emerging industries. By reshaping the patterns of market competition, such innovation facilitates the transformation and optimization of the industrial structure. As Asia’s largest economy, China is working towards “Made in China 2025”, an initiative that calls for innovation-driven enhancement of industrial structural optimization [14].

Regarding technological choice, Atkinson and Stiglitz [15] argue that localized technological innovation has more potent effects on industrial structural optimization and economic growth. Their analysis emphasizes the critical role of context-specific technological adaptation. Caselli and Coleman [16] suggest that developing countries can stimulate indigenous technological innovation by adopting advanced technologies. This process effectively facilitates industrial upgrading. In addition, Liao et al. [17] systematically examine China’s innovation patterns. Their research shows that coordinating technology adoption with secondary innovation creates optimal conditions for industrial structural optimization.

Regarding pathways and mechanisms, some scholars argue that a technological innovation-driven industrial chain is an evolutionary pathway for industrial structural optimization [18,19]. Breschi et al. [20] propose that technological innovation optimizes the industrial structure through labor productivity. Peneder [21] finds that technological innovation significantly impacts industrial structures by changing demand structures. Gustafsson et al. [22] point out that changes in industrial cycles are the effective pathways through which technological innovation affects industrial structural optimization.

Furthermore, as the concept of sustainable development evolves, more scholars examine whether green technological innovation can drive the optimization and upgrading of industrial structures. Existing studies acknowledge its positive impact from various research perspectives.

From the dual perspective of industrial structural rationalization and upgrading, Wang et al. [23] empirically find that, unlike foreign direct investment (FDI), green technological innovation has a positive promotional effect on industrial structural upgrading. From a regional perspective, Wang et al. [24] use stochastic frontier analysis to show that green technological innovation promotes the green transformation of regional industrial structures by improving local green total factor productivity. Through a comparative study of different industrial sectors in China, Xie and Teo [25] empirically demonstrate (based on threshold panel data) that green technological innovation is an effective driver of clean industrial structural upgrading, except for low-value-added sectors and clean-type (LC) sectors.

From a micro perspective, on the one hand, green innovation is often associated with a significant increase in pollution control costs. This cost increase crowds out firms’ resource investments in productive business activities, thereby suppressing their overall performance [26]. In addition, firms bear the costs of developing green technologies. However, their positive externality characteristics lead to a contradiction between socialized benefits and privatized costs. This reduces the willingness of firms to invest in R&D for green technological innovation [27].

On the other hand, when firms allocate limited resources to green technological innovation activities, it can reduce energy consumption, resource efficiency, and pollutant emissions [28]. This can help to achieve a better environmental performance and competitive advantages [29]. It also serves as an essential signal for evaluating corporate social responsibility and governance standards. Through the joint integration of digital technology and green finance [30], green technological enterprises can achieve financing cost advantages and resource agglomeration effects [31]. In contrast, high-pollution enterprises face increased financing constraints due to environmental risk premiums [32]. This increases the competitiveness of enterprises in the market, thereby steering the market toward a low-carbon economy and promoting industrial structural transformation [33,34].

From a macro perspective, Sarkar [35] shows that ecological technological innovation significantly promotes the upgrading of industrial structures towards a technology-intensive sustainable development model. This is achieved by promoting the green transformation of traditional industries and the integration of emerging ecological industrial chains. Ghisetti and Quatraro [36] propose that green technologies not only promote the greening of their industries but also significantly enhance the transformation of regional industrial structures toward a more environmentally optimized model through vertical collaborative innovation along upstream and downstream industrial chains. Kheder and Zugravu [37] conduct a study from the dual perspective of theoretical modeling and empirical testing. They find that green technological innovation, as a core driver, accelerates the green upgrading of the host country’s industrial structure under the incentive of stringent environmental regulations. Ghisetti et al. [38] find that after overcoming financing constraints, green technological innovation drives the transformation of manufacturing structures to a green model through cleaner production and an improved technical efficiency.

2.2. Carbon Emission Efficiency and Industrial Structural Optimization

Although the existing literature has extensively discussed the relationship between carbon emissions and industrial structural optimization, most studies focus on the role of total carbon emissions [39,40,41] and carbon trading [42,43,44] in industrial structural development. This narrow focus overlooks the critical role of carbon emission efficiency, a key metric for evaluating environmental sustainability. There is still a gap in direct research on “carbon emission efficiency and industrial structural optimization,” which is essential to fill in order to understand how industries can reduce their ecological footprint while maintaining economic competitiveness. There remains a gap in direct research on “carbon emission efficiency and industrial structural optimization”. Based on the existing literature, a research framework can be constructed from the following dimensions, and potential research directions can be explored.

From the perspective of coordination relationships, Zhou et al. [45] in their study based on China’s provinces, find a significant dynamic imbalance between carbon emission efficiency and industrial structural upgrading. Specifically, the carbon emission efficiency is much higher than the industrial structural level in the same region, and the coupling coordination degree between the two is low. Zhu and Zhang [46] demonstrate from a regional perspective (using the Yangtze River Delta as an example) that the development and carbon emissions of the same industry show differences within a single region. However, the cross-regional relocation of a single sector can reduce such differences and improve coordination between the industrial structure and carbon emission efficiency.

From the perspective of regression relationships, Wang et al. [47] use regression analysis to examine China’s carbon emission patterns. Their results indicate a positive correlation between carbon emission intensity and secondary industry value added. In contrast, Zhang et al. [39] find a significant negative correlation between carbon emission intensity and the tertiary industrial share. This finding highlights sectoral differences in emission–structure relationships. He and Wang [48] build on the Environmental Kuznets Curve framework. They systematically examine regression relationships between the industrial structure and carbon emission intensity by treating the industrial structure as a mediating variable.

Current research focuses mainly on how overall carbon emissions and carbon trading mechanisms affect the industrial structure. These studies consistently confirm their positive role in promoting industrial upgrading. However, systematic investigations of how the carbon emission efficiency itself affects industrial structural optimization remain scarce. Existing analytical perspectives must also be diversified to capture this emerging relationship fully.

2.3. Constraint Mechanisms of Technological Innovation Driving Industrial Structural Optimization

Scholars have investigated the constraint mechanisms of technological innovation in promoting industrial structural optimization. Their research argues that this driving effect evolves in response to changes in external factors. Understanding these constraints is vital for formulating policies that enhance the synergy between technological innovation, industrial development, and sustainability. Researchers can identify strategies to overcome barriers and foster more sustainable industrial transformation by exploring how external factors influence the relationship.

Regarding non-institutional factors, Xu et al. [49] discover that marketization and economic openness introduce threshold effects into the relationship between technological innovation and industrial upgrading. The mechanisms of action are similar: the impact of technological innovation on industrial structural optimization is limited below the threshold, while exceeding the threshold significantly enhances the effect. Gao et al. [50] argue that the export product quality effectively moderates the role of technological innovation in upgrading industrial value chains. Specifically, technological innovation promotes manufacturing value chain upgrading when the export product quality is below the threshold, but this driving effect becomes negligible when the quality exceeds the threshold. Yang et al. [51] examine the Beijing–Tianjin–Hebei region, showing that the digital infrastructure improves industrial eco-efficiency by promoting technological innovation, a process exhibiting a threshold effect tied to regional collaborative innovation (RCI). The greening impact of technological innovation on the industrial structure strengthens significantly only when RCI surpasses a critical level. Wang et al. [52] employ a spatial Durbin model in the Yangtze River Delta, demonstrating that regional integration enhances the synergistic effects between technological innovation and the digital economy, thereby driving industrial structural upgrading.

Regarding institutional factors, Jaffe and Palmer [53] propose a dual institutional effect. Their analysis suggests that institutional improvements directly affect technological innovation while shaping its structural optimization effects. Murmann [54] develops an evolutionary co-dependency framework. Technological innovation and institutional change interact dynamically, jointly driving industrial structural optimization and upgrading. Wang et al. [23] distinguish institutional dimensions. They empirically show that both formal institutional factors (e.g., regulatory frameworks) and informal institutional factors (e.g., cultural norms) moderate how technological capabilities influence industrial supply chains. Khattak et al. [55] focus on government regulatory interventions. Their results show that environmental regulations fundamentally alter the dynamic relationship between technological innovation and industrial upgrading. Technological change is recognized as a major driver of the energy transition [56]. Ge et al. [57] focus on renewable energy technological innovation (RETI). RETI significantly promotes industrial upgrading, especially in regions with developed green finance systems. This promotional effect intensifies as the level of green finance increases.

When discussing the constraining mechanisms of technological innovation in driving industrial structural optimization, existing research focuses on a relatively limited set of environmental moderating variables. Studies rarely concentrate their research perspectives on green technological innovation. The literature seldom explores factors that facilitate or hinder the effects of green technological innovation on industrial structural optimization. Few studies incorporate carbon emission efficiency into analytical frameworks assessing how green technological innovation influences industrial optimization and upgrading. Researchers have also largely overlooked the heterogeneous relationships between these variables. These research perspectives require further deepening and expansion.

3. Theoretical Analysis and Hypotheses Development

Green innovation integrates green technologies into all stages of social production. This process generates dual effects: technological progress and environmental spillovers [24]. Through enterprise-level and industry-level transformations, green innovation drives the optimization of regional industrial structures, laying a foundation for achieving sustainable development through industrial ecology and low-carbon transitions.

New products and processes developed by enterprises will exert technological application. They generate diffusion effects across the entire industry, forming industrial agglomerations, which optimizes the industrial structure.

Against high-quality economic development, traditional high-pollution, high-energy-consuming industries face constraints and elimination pressures. To meet low-carbon and green standards, these industries need to invest substantial human resources, capital, and technology to research, develop, and adopt new products and processes, thereby driving their green transformation. The resulting backward pressure effect continuously stimulates the optimization and upgrading of traditional industries. On the other hand, continuous innovation and development of green technologies will further create and cultivate more environmental protection industries while providing more development opportunities for low-carbon sectors such as service and high-tech industries [58]. New sectors will increasingly attract innovative talents and surplus labor, creating a talent agglomeration effect that optimizes labor employment structures and facilitates industrial upgrading [59]. Green technological innovation with low-carbon goals improves the efficiency of existing sectors and provides an additional impetus for economic growth [10]. This low-carbon-oriented innovation fosters a synergistic development pattern between economic growth and environmental protection, directly contributing to global sustainability goals. For example, technological advances in energy industries, such as power grids and storage, have developed new industrial chains [60,61].

In addition, under the influence of the demonstration effect, more social capital will withdraw from high-polluting and high-energy-consuming industries and shift to energy-saving and environmental protection fields, thus guiding the development of green industries [32]. This change in investment structure will further provide resources for optimizing and upgrading the industrial structure [62], ultimately contributing to establishing a sustainable industrial system that balances economic growth and ecological protection.

Based on this, we propose the following:

H1. 

Green technological innovation has a positive driving effect on industrial structural optimization.

The Chinese government formally announced its “dual-carbon” goals in 2020 to reduce greenhouse gas emissions and achieve sustainable development. These goals aim to reach carbon peaking by 2030 and achieve carbon neutrality by 2060. However, the path to carbon neutrality faces several significant challenges, including technological bottlenecks, economic cost pressures, and the difficulties of social transition [63].

In response, the Chinese government has actively promoted energy transformation and carbon reduction. Improving carbon emission efficiency has been identified as a key strategic focus. In this context, enhanced carbon efficiency is reflected not only in lower carbon emissions per unit of output. It also signals a broader shift toward more efficient resource allocation, cleaner production technologies, and more sustainable patterns of industrial development.

Firstly, from the perspective of Endogenous Growth Theory, improving the carbon emission efficiency enhances the use of energy and resources. This raises total factor productivity and drives the transformation of industries from extensive, resource-intensive models to more technology-intensive and service-oriented structures. Furthermore, local governments have tightened carbon emission regulations under strong policy guidance. Firms’ green innovation activities are now increasingly influenced by regional carbon levels and environmental policies. In this regulatory context, firms are under greater pressure to internalize carbon-related costs.

Secondly, from an environmental cost perspective, profit-maximizing firms are more likely to invest in developing and adopting green technologies when the cost of carbon emissions exceeds that of green innovation. By integrating these technologies into products and production processes, firms can improve their environmental performance [64] and accelerate their low-carbon transition [65]. This “push effect” helps eliminate outdated capacity and fosters industrial upgrading, crucial for establishing a sustainable industrial system that reduces ecological footprints.

Thirdly, according to the Input–Output Theory, emerging industries can drive the synergistic upgrading of traditional industries in terms of technology, process, and management through the vertical extension and horizontal expansion of the industrial chain. This will realize the structural adjustment of the regional industrial structure. Improving the carbon emission efficiency also leads to the rapid development of strategic new industries such as green manufacturing and new energy equipment. These industries not only have the characteristics of a high technology content and low environmental load but can also drive the optimization and upgrading of traditional industries through the upstream and downstream transmission of the industrial chain. Under the combined effect of government regulation and market orientation, the drive to improve the carbon emission efficiency among regions is becoming a significant force in promoting industrial structural optimization.

Based on this, we propose the following:

H2. 

Carbon emission efficiency has a positive driving effect on industrial structural optimization.

Under the “dual-carbon” strategy, green technological innovation has become a key driver of industrial structural optimization. However, this effect is neither linear nor static [55]. The impact often depends heavily on the level of carbon emission efficiency. Specifically, carbon emission efficiency plays a crucial moderating role in the relationship between green innovation and industrial upgrading.

Firstly, from the perspective of financial constraints, firms or regions with a higher carbon emission efficiency are more likely to receive green credit and policy support. Financial institutions tend to allocate funding to environmentally friendly enterprises. This strengthens their financial capacity to invest in green R&D and apply new technologies. In contrast, highly polluting firms face tighter financing conditions and higher capital pressure. Such financial constraints can profoundly affect their capacity and willingness to engage in technological innovation [66,67,68].

Secondly, according to Porter’s hypothesis, strict environmental standards can trigger an innovation compensation effect. This not only forces traditional high-energy industries to upgrade but also encourages the emergence of new industries. Firms generally have stronger environmental compliance in regions with a high carbon emission efficiency. As a result, they are better positioned to promote the diffusion of green technologies. These technologies can spread horizontally across firms through imitation, collaboration, and competition. This broadens the application of green innovations and improves the coordination of industrial restructuring, ultimately fostering a sustainable industrial ecosystem that balances economic prosperity with environmental protection.

Thirdly, from the perspective of institutional quality, regions with a high carbon efficiency usually have more developed environmental governance, better green policy frameworks, and stronger industrial foundations. These conditions provide fertile grounds for green technologies to take root. In such environments, firms are more likely to adopt green innovations in their product design and production processes. This leads to more effective and direct improvements in the industrial structure. Robust institutional support ensures that industrial upgrades align with long-term sustainable development objectives. These include carbon neutrality and resource circularity.

Based on this, we propose the following:

H3. 

Carbon emission efficiency positively moderates the effect of green technological innovation on industrial structural optimization.

4. Empirical Design

4.1. Empirical Model

To test hypothesis 1, we create the benchmark regression model as follows:

$${ISO}_{i,t}={\beta }_0+{\beta }_1{GTI}_{i,t}+\theta {control}_{i,t}+{ϵ}_{i,t}$$

(1)

where i denotes the region and t denotes the year; ${ISO}_{i,t}$ represents the level of industrial structural optimization in province i during period t; ${gti}_{i,t}$ represents green technological innovation in province i during period t; and ${control}_{i,t}$ denotes the control variables, while ${\beta }_0$ and ${ϵ}_{i,t}$ denote the intercept term and the disturbance term.

To examine whether the carbon emission efficiency plays a moderating role between green technological innovation and industrial structural optimization, we include an interaction term between green technological innovation and carbon emission efficiency, as follows:

$${ISO}_{i,t}={\beta }_0+{\beta }_1{GTI}_{i,t}+{\beta }_2{CEE}_{i,t}+{\beta }_3{{GTI}_{i,t}\times CEE}_{i,t}+\theta {control}_{i,t}+{ϵ}_{i,t}$$

(2)

where ${CEE}_{i,t}$ denotes the natural logarithm of carbon emission efficiency in province i during period t; ${{\mathrm{a}\mathrm{n}\mathrm{d}GTI}_{i,t}\times CEE}_{i,t}$ represents the interaction term between green technological innovation and the logarithm of carbon emission efficiency.

This study examines whether the impact of green technological innovation on industrial structural improvement shows significant differences under varying levels of carbon emission efficiency. We develop a multi-threshold regression model using carbon emission efficiency as the key threshold indicator. We create the three-threshold model as follows:

$$\begin{array}{c}{GTI}_{i,t}={\beta }_0+{\beta }_1{GTI}_{i,t}\times I\left(CEE\le {\omega }_1\right)+{\beta }_2{GTI}_{i,t}\times I\left({\omega }_1<CEE\le {\omega }_2\right)+\\ {\beta }_3{GTI}_{i,t}\times I\left({\omega }_2<CEE\le {\omega }_3\right)+{\beta }_4{GTI}_{i,t}\times I\left({\omega }_3<CEE\right)+\theta {control}_{i,t}+{\epsilon }_{i,t}\end{array}$$

(3)

where ${\omega }_1<{\omega }_2<{\omega }_3$ defines three thresholds. These thresholds divide the observations into four different intervals.

4.2. Data and Variables

4.2.1. Explained Variables

Industrial structural optimization. Following Yuan et al. [69], this paper uses the entropy weight method to construct a comprehensive indicator of industrial structural optimization combining three dimensions: industrial structural advancement (Ais), rationalization (Thile), and ecologization (Eco). According to Wang et al. [70], industrial structure advancement is measured by the ratio of tertiary industry value added to secondary industry value added. According to Hao et al. [71], industrial structure rationalization is measured by an improved Thile index, which reflects a coordination among industries and efficient resource utilization [72,73] (Ernst, 1998; Kaplinsky and Readman, 2005). According to Ang [74], industrial structural ecologization is measured by the energy–GDP ratio, which is the ratio of total energy consumption in the national economy to GDP.

The calculation formula for the improved Thiel index is

$$T_{i,t}={\sum }_{m=1}^3{y}_{i,m,t}\ln \left({\displaystyle \frac{y_{i,m,t}}{e_{i,m,t}}}\right),m=1,2,3$$

(4)

where i denotes the region, t denotes the year, and m denotes the industry. $y_{i,m,t}$ represents the proportion of the gross production value of industry m in province i during period t to the regional gross production value. $e_{i,m,t}$ represents the proportion of employees in industry m in province i during period t to the total employees. The Theil index measures the rationalization of the industrial structure. The index approaches zero when the output is evenly distributed across sectors, indicating high structural rationalization. Conversely, if specific sectors disproportionately dominate the output, the index rises, reflecting low rationalization.

We use the entropy method to construct a comprehensive evaluation index for industrial structural optimization.

This index systematically quantifies progress in three dimensions: advanced industrial upgrading, rational resource allocation, and ecological transformation.

$$A=\left[\begin{array}{ccc}{a}_{11}& \dots & a_{1n}\\ ⋮& \ddots & ⋮\\ a_{n1}& \dots & a_{nn}\end{array}\right]$$

(5)

$$A_i={\sum }_{j=1}^n\begin{array}{c}\left|a_{i,j}\right|-1,i=1,2,3,\dots ,n\end{array}$$

(6)

where $A_i$ represents the total influence of the i indicator on the other (n − 1) indicators. A larger $A_i$ value indicates a higher degree of importance of the indicator to the entire system, requiring a higher weight coefficient. Therefore, normalizing $A_i$ yields the weight coefficients for each indicator:

$$w={\displaystyle \frac{A_i}{{\sum }_{i=1}^n{A}_i}},i=1,2,3,\dots ,n$$

(7)

This paper’s comprehensive measurement index of industrial structural optimization includes three indicators: industrial structural progress, rationalization, and ecologization. The observations for each indicator are denoted as $x_{ij}$. Due to differences in the characteristics of the indicators, different indicators have positive or negative effects on the measurement of the comprehensive industrial structural index. For positive indicators, their order degree is positively correlated with the overall system order degree; for negative indicators, their order degree is negatively correlated with the overall system order degree. Industrial structural progress is a positive indicator, while industrial structural rationalization and ecologization are negative.

$$y_{ij}=\left\{\begin{array}{c}{\displaystyle \frac{x_{ij}-{min}_{ij}}{{max}_{ij}-{min}_{ij}}},\mathrm{when}{x}_{ij}\mathrm{is}\mathrm{a}\mathrm{positive}\mathrm{indicator}.\\ {\displaystyle \frac{{max}_{ij}-x_{ij}}{{max}_{ij}-{min}_{ij}}},\mathrm{when}{x}_{ij}\mathrm{is}\mathrm{a}\mathrm{negative}\mathrm{indicator}.\end{array}\right.$$

(8)

Calculate the ratio of the i-th sample value under the j-th indicator to the sum of all samples of that indicator.

$$p_{ij}={\displaystyle \frac{y_{ij}}{{\sum }_{i=1}^m{y}_{ij}}}$$

(9)

Calculate the information entropy of the j-th indicator.

$$e_j=-k{\sum }_{i=1}^m{p}_{ij}ln{p}_{ij},k>0,0\le e_{ij}\le 1$$

(10)

Calculate the redundancy of information entropy.

$$g_i=1-e_{ij}$$

(11)

Calculate the weights of each evaluation indicator.

$$w_j={\displaystyle \frac{g_j}{{\sum }_{j=1}^n{g}_j}}$$

(12)

4.2.2. Core Explanatory Variable

Green technological innovation. Based on the green patent standards of the Green Technology Patent Classification System issued by the National Intellectual Property Administration of China, this paper retrieves the number of green patent applications (invention patents and utility model patents) in each province to measure green technological innovation. Following the research of Abdullah et al. [75], the classification system is made internationally comparable by adopting the International Patent Classification Green Inventory (IPC Green Inventory 2023) of the World Intellectual Property Organization (WIPO). While focusing on China’s national conditions, development needs, and “double-carbon” target strategy, the classification system maintains the relative independence and integrity of green and low-carbon technological branches. The level of green technological innovation in each province is measured by the number of green patent applications per 10,000 people. In addition, to reduce the impact of heteroscedasticity in the regression results, this indicator is subjected to mean normalization processing.

4.2.3. Threshold Variable

Carbon emission efficiency. This paper references the study by Zhang et al. [76], taking the labor force, capital stock, and total energy consumption as input variables, regional GDP as the desirable output, and carbon dioxide emissions as the undesirable output. A super-SBM model is constructed using the input–output table to measure the regional carbon emission efficiency. Carbon emission-related data are calculated using the carbon accounting methodology of the Intergovernmental Panel on Climate Change (IPCC).

Table 1. Input–output table.

Variable Type	Variable Name	Variable Description	Data Source
Input Variables	Labor	Number of Employees	China Statistical Yearbook and Provincial Yearbooks
	Capital	Fixed Asset Investment
	Energy	Total Energy Consumption
Expected Output	GDP	Gross Domestic Product
Non-expected Output	Carbon Emissions	Carbon Dioxide Emissions	China Emission Accounts and Datasets (CEADs)

summarizes the input-output indicator system for carbon emission efficiency measurement.

The spatial distribution of the carbon emission efficiency in 2005, 2010, 2015, and 2022 is shown in Figure 1. Using ArcGIS 10.8.2 software, the Natural Breaks (Jenks) method was employed to divide the carbon emission efficiency of each province into five stages, from low to high. The spatial evolution maps present carbon emission efficiency distributions at five-year intervals (2005, 2010, and 2015) to capture long-term trends, with 2022 as the terminal year of our study period (2000–2022) to provide a contemporary snapshot.

Spatially, China’s carbon emission efficiency exhibits specific regional heterogeneity, presenting a distribution pattern of higher efficiency in the east and lower efficiency in the west. Eastern coastal regions (such as the Yangtze River Delta and Pearl River Delta) have a significantly higher carbon emission efficiency due to the advanced nature of their economic structures and high green technological intensity; the central region has moderate efficiency values influenced by the high proportion of heavy industry and the energy structure; and the western region has lower efficiency values due to the strong resource dependence and low clean technology penetration rates.

Temporally, China’s provincial carbon emission efficiency generally shows an upward trend. In 2005, only three provinces and cities had a carbon emission efficiency in the 0.24–0.36 interval, accounting for 10% of the total sample. By 2022, the number of provinces and cities in the 0.24–0.36 interval increased to 24, accounting for 80% of the total sample; and the number of provinces and cities in the 0.36–0.49 interval increased to 10, accounting for 30.3% of the total sample.

4.2.4. Control Variables

Considering that other factors may potentially impact the optimization of the industrial structure, this paper introduces the following relevant variables to support the empirical test: The size of the government (GOV) is measured by the ratio of government public financial expenditures to GDP. The level of economic development (GDP) is measured by GDP per capita. Referring to Wang et al. [77], the level of human capital is expressed in college students per 10,000 persons. Referring to the study of Inekwe [78], research and development (RD) investment intensity is measured by the ratio of RD investment to GDP. The degree of openness to the outside world (OPEN) is measured by the ratio of the amount of foreign direct investment utilized to regional GDP. Referring to the study of Lin et al. [79], the intensity of intellectual property protection (IPR) is expressed by the ratio of the volume of technology market transactions to regional gross domestic product, where the larger the value, the more active the technology market transactions in the region and the better the effect of indexed IPR protection.

Specific variable selection and calculation methods are shown in

Table 2. Key variables and variable descriptions.

Variable Type	Variable Name	Variable Description
Explained Variable	Industrial Structure Optimization (ISO)	Composite Index of Industrial Structural optimization
Core Explanatory Variable	Green Technology Innovation (GTI)	Number of Green Patent Applications per 10,000 Persons
Threshold Variable	Carbon Emission Efficiency (CEE)	The Logarithm of Carbon Emissions Efficiency
Control Variables	Government Size (GOV)	Government Fiscal Expenditure/GDP
	Economic Development Level (GDP)	The Logarithm of GDP per Capita
	Human Capital Level (HC)	Number of College Students per 10,000 Persons
	R&D Intensity (RD)	R&D Expenditure/GDP
	Openness Level (OPEN)	Foreign Direct Investment (FDI)/GDP
	Intellectual Property Protection (IPR)	Technology Market Transactions/GDP

.

4.3. Data Description and Descriptive Statistics

This study uses panel data from 30 provincial-level regions in the Chinese Mainland (excluding Tibet and Hong Kong, Macao, and Taiwan regions) from 2000 to 2022. The data on industrial structural advancement, rationalization, and ecologicalization are sourced from the China Statistical Yearbook and provincial statistical yearbooks. Green technological innovation data are obtained from the China Research Data Services Platform (CNRDS). Carbon emission data are shown in Table 1. Government size, economic development level, and human capital data are sourced from the China Statistical Yearbook, the National Bureau of Statistics, and provincial statistical yearbooks. R&D investment intensity data are derived from the China Science and Technology Statistical Yearbook. Openness data are obtained from the Wind database, provincial statistical yearbooks, and statistical bulletins. Intellectual property rights protection data are sourced from the China Statistical Yearbook. Partially missing data are supplemented using interpolation methods.

The descriptive statistical findings for the key variables in this paper are shown in

Table 3. Descriptive statistics.

Variable	Observation	Mean	Standard	Min.	Max.
ISO	690	0.243	0.123	0.080	0.997
GTI	690	6.666	1.974	1.099	10.93
CEE	690	0.214	0.0750	0.101	0.738
GOV	690	0.218	0.105	0.069	0.758
GDP	690	10.23	0.878	7.923	12.15
HC	690	0.017	0.008	0.002	0.044
RD	690	1.480	1.105	0.146	6.845
OPEN	690	0.025	0.024	0	0.163
IPR	690	0.013	0.025	0	0.191

.

Table 3 shows the descriptive statistics of key variables, including observations and mean, standard deviation, maximum, and minimum values. Due to the significant differences in the original values of the level of economic development (GDP), direct use could introduce bias; the maximum value of carbon emission efficiency (CEE) is far above the mean, with a right-skewed distribution indicating potential outliers with high carbon emissions. Therefore, natural log transformations are applied to these variables.

Specifically, industrial structural optimization (ISO) has a mean of 0.243, a standard deviation of 0.123, and a range of 0.080–0.997, indicating significant regional variation in the industrial structure. Green technological innovation (GTI) has a mean of 6.666, a standard deviation of 1.974, and a range of 1.099–10.93, indicating a high dispersion in green technological innovation. The high standard deviation reflects an uneven technological development across regions: the maximum value of 10.93 likely corresponds to technologically advanced areas, while the minimum value of 1.099 may represent regions with weak technological foundations.

5. Empirical Results and Discussion

Before conducting the empirical analysis, auxiliary tests are conducted to ensure the rationality of the econometric model and the reliability of the research results.

Results of the multicollinearity test are shown in

Table 4. Tests for multicollinearity.

Variable	Average Variance Inflation Factor	1/VIF
GTI	6.520	0.153
CEE	2.430	0.411
GOV	2.040	0.490
GDP	8.990	0.111
HC	3.920	0.255
RD	4.750	0.211
OPEN	1.410	0.711
IPR	3.060	0.327
Mean	VIF	4.140

. The calculation shows that the average variance inflation factor (VIF) value of the research variables in this paper is 4.140, with VIF values ranging from 1.410 to 8.990, all below the critical threshold of 10 and within an acceptable range. This indicates that the multicollinearity problems in the econometric model are weak and will not significantly affect the estimation results in the subsequent analysis.

5.1. Baseline Results

Table 5. The baseline results.

	(1)	(2)	(3)	(4)	(5)	(6)
	ISO (OLS)	ISO (OLS)	ISO (RE)	ISO (RE)	ISO (FE)	ISO (FE)
GTI	0.024 ***	−0.012 ***	0.028 ***	0.016 ***	0.028 ***	0.031 ***
	(0.002)	(0.003)	(0.001)	(0.004)	(0.001)	(0.005)
Controls	No	Yes	No	Yes	No	Yes
Cons	0.080 ***	−0.352 ***	0.053 ***	0.074	0.053 ***	0.226 ***
	(0.015)	(0.062)	(0.021)	(0.081)	(0.008)	(0.087)
N	690.000	690.000	690.000	690.000	690.000	690.000
Adj. R2	0.150	0.691			0.442	0.615

Note: Standard errors are reported in parentheses. The symbols *** denote significance at the 1%, level.

reports the benchmark regression results of green technological innovation on industrial structural optimization. The Hausman test yields a p-value of 0, which allows us to firmly reject the null hypothesis. Therefore, this paper uses a fixed-effects model to test the benchmark regression relationship between green technological innovation and industrial structural optimization.

In the fixed-effects model, the coefficient of green technological innovation is positive at the 1% significance level regardless of whether control variables are included, indicating that green technological innovation significantly promotes industrial structural optimization. As shown in column (5), without control variables, the coefficient of green technological innovation on industrial structural optimization is 0.028 at the 1% significance level. In column (6), when control variables are included, the coefficient of green technological innovation further increases to 0.031, which is significant at the 1% level. Economically, a 1% increase in green technological innovation levels effectively increases the regional industrial structural optimization coefficient by 0.031%, which validates Hypothesis 1.

This study adds the variable of carbon emission efficiency to the benchmark regression model to test Hypothesis 2. The specific empirical results are presented in

Table 6. The baseline results after introducing carbon emission efficiency.

	(1)	(2)	(3)	(4)
	ISO	ISO	ISO	ISO
GTI			0.023 ***	0.029 ***
			(0.001)	(0.005)
CEE	0.642 ***	0.209 ***	0.368 ***	0.190 ***
	(0.039)	(0.038)	(0.035)	(0.037)
Controls	No	Yes	No	Yes
Cons	0.105 ***	−0.116 *	0.009	0.273 ***
	(0.009)	(0.059)	(0.009)	(0.085)
N	690.000	690.000	690.000	690.000
Adj. R2	0.255	0.608	0.523	0.629

Note: Standard errors are reported in parentheses. The symbols *** and * denote significance at the 1% and 10% levels, respectively.

.

Table 6 reports the regression results of green technological innovation and carbon emission efficiency on industrial structural optimization after introducing the carbon emission efficiency variable. The results in columns (1) and (2) show that in regressions with and without control variables, the impact coefficient of carbon emission efficiency on industrial structural optimization is significantly positive, and both are significant at the 1% level. Columns (3) and (4) show the regression results of the fixed-effects regression model with the carbon emission efficiency variable added to the benchmark regression. fThe estimated coefficients of green technological innovation and carbon emission efficiency are significantly positive, indicating a positive correlation between carbon emission efficiency and industrial structural optimization. Regarding economic significance, a 1% increase in carbon emission efficiency will promote a 0.190% improvement in the regional industrial structural optimization level. The results show that green technological innovation and carbon emission efficiency significantly promote industrial structural optimization, further verifying Hypothesis 1 and tentatively verifying Hypothesis 2.

5.2. Moderating Effect Results

To further verify whether the carbon emission efficiency positively impacts industrial structural optimization, this study introduces an interaction term TJ (GTI* CEE) between green technological innovation and carbon emission efficiency to construct a moderation-effect model. The regression results of the moderation effect are shown in

Table 7. The moderating effect results.

	(1)	(2)	(3)
	ISO (FE)	ISO (Moderating Effect)	ISO (Moderation Effect After Centralization)
GTI	0.029 ***	0.015 ***
	(0.005)	(0.005)
CEE	0.190 ***	−0.289 ***
	(0.037)	(0.079)
TJ (GTI* CEE)		0.084 ***
		(0.012)
GTI_c			0.033 ***
			(0.005)
CEE_c			0.271 ***
			(0.038)
TJz (GTI_c* CEE_c)			0.084 ***
			(0.012)
Controls	Yes	Yes	Yes
Cons	0.273 ***	0.557 ***	0.714 ***
	(0.085)	(0.093)	(0.112)
N	690.000	690.000	690.000
Adj. R2	0.629	0.653	0.653

Note: Standard errors are reported in parentheses. The symbols *** denote significance at the 1%, level.

.

As shown in column (2), after adding the interaction term of green technological innovation and carbon emission efficiency TJ (GTI* CEE) to the benchmark regression model, green technological innovation still has a positive driving effect on industrial structural optimization at the 1% significance level. Under the moderation of carbon emission efficiency, a 1% increase in green technological innovation will increase the coefficient of industrial structural optimization by 0.015%. Although the coefficient of carbon emission efficiency is negative, the coefficient of the interaction term is significantly positive (0.084 ***). Due to a certain degree of multicollinearity between the original variables and the interaction term (VIF = 9.520), this study adopts the centralized moderation effect test to reduce the interference of multicollinearity, and the results are shown in column (3). After centralizing the variables, the VIF value drops to 4.200, indicating that centralization effectively reduces multicollinearity between the original variables and the interaction term. The coefficients of green technological innovation, carbon emission efficiency, and their interaction term are all significantly positive at the 1% level. Compared to the uncentralized moderation results, the coefficients of green technological innovation and carbon emission efficiency increase. This indicates that improvements in green technological innovation and carbon emission efficiency are independent drivers of industrial structural optimization and reinforce this process through their synergistic effects.

5.3. Threshold Regression Results

To test Hypothesis 3, this study constructs a multiple threshold dynamic model by treating carbon emission efficiency as the threshold variable. This study employs a grid search (grid 300) and bootstrap sampling (bs 300) to test the thresholds and obtains asymptotic values of F-statistics, p-values, threshold values, etc., through the bootstrap test for threshold effects. In addition, a random seed (Seed = 12,345) is set in the threshold regression analysis to ensure the reproducibility of the research results.

As shown in

Table 8. The threshold effect results.

Threshold Variable	Threshold Type	F-Value	10% Critical Value	5% Critical Value	1% Critical Value
Carbon Emission Efficiency	Single Threshold	87.2	17.443	21.202	29.670
	Double Threshold	41.88	16.110	62.184	96.051
	Triple Threshold	23.62	22.551	26.806	64.412

and

Table 9. Threshold estimates and confidence intervals.

Threshold Variable	Threshold Type	Threshold Value	p-Value	95% Confidence Interval	BS Times
Carbon Emission Efficiency	Single Threshold	0.1033	0.000	(0.1033, 0.1086)	300
	Double Threshold	0.3592	0.067	(0.3531, 0.3811)	300
	Triple Threshold	0.4052	0.090	(0.3919, 0.4207)	300

and Figure 2, the carbon emission efficiency has a significant triple threshold effect: the first threshold is 0.1033, the second is 0.3592, and the third is 0.4052. Therefore, this paper selects the triple threshold effect regression to estimate the relationship between green technological innovation and industrial structural optimization. The results are shown in Table 9.

Figure 2 shows the threshold estimates and confidence intervals for carbon emission efficiency (CEE). In the first threshold test, the likelihood ratio (LR) statistic shows a significant spike near the threshold, indicating that the moderating effect of carbon emission efficiency on the relationship between green technological innovation and industrial structural optimization has a breakpoint. In the second threshold test, the LR statistic again shows a significant jump at a higher threshold, confirming that the threshold effect has multiple stages. The LR statistic corresponding to the third threshold remains above the critical value. The numerous significant peaks of the LR statistic confirm that the threshold effect of carbon emission efficiency is not a single threshold effect but a significant triple threshold effect.

Table 10. The threshold regression results.

	(1)
	ISO
Core Explanatory Variable	GTI
Threshold Variable	CEE
GTI1 (CEE ≤ 0.1033)	0.096 ***
	(0.013)
GTI2 (0.1033 < CEE ≤ 0.3592)	0.032 ***
	(0.004)
GTI3 (0.3592 < CEE ≤ 0.4052)	0.038 ***
	(0.005)
GTI4 (0.4052 < CEE)	0.047 ***
	(0.005)
Controls	Yes
Cons	0.373 ***
	(0.080)
N	690.000
Adj. R2	0.682

Note: Standard errors are reported in parentheses. The symbols *** denote significance at the 1% level.

shows three sequential threshold values for carbon emission efficiency. These three threshold values can divide the model into four threshold intervals with different estimated coefficients.

When the carbon emission efficiency is below the first threshold of 0.1033, the coefficient of green technological innovation is significantly positive at 0.096 and passes the 1% significance test. This suggests that under an extremely low carbon emission efficiency, green technological innovation still positively promotes industrial structural optimization, but with a relatively weak intensity. This may be because an excessively low carbon emission efficiency is often accompanied by stringent environmental regulations that require firms to incur additional costs to meet the standards. While this may incentivize technological upgrading, it may also suppress the full release of industrial transformation dividends in the short run.

When the carbon emission efficiency exceeds the first threshold of 0.1033 but remains below the second threshold of 0.3592, the coefficient of green technological innovation decreases to 0.032. Still, it remains significant at the 1% level. This indicates that after the carbon emission efficiency exceeds the first threshold, the marginal effect of green technological innovation weakens. At this stage, the pressure of environmental regulations and technological innovation incentives tend to balance, and industrial optimization and upgrading gradually enter a stable adjustment period.

When the carbon emission efficiency exceeds the second threshold of 0.3592 but remains below the third threshold of 0.4052, the coefficient of green technological innovation returns to 0.038. This reflects that after the carbon emission efficiency exceeds the second threshold value, the industrial optimization effect of green technological innovation regains strength. A possible reason is that the moderate carbon emission space provides a buffer for technological innovation, allowing enterprises to balance the cost of emission reduction and long-term benefits, forming a virtuous cycle.

When the carbon emission efficiency exceeds the third threshold of 0.4052, the coefficient of green technological innovation further increases to 0.047. This means that at the highest interval of carbon emission efficiency, the effect of green technological innovation on industrial optimization and upgrading is most significant. At this point, the carbon emission efficiency has overcome the bottleneck of environmental constraints, fully released the dividend of technological innovation, and driven the industrial structure toward green upgrading.

In summary, the carbon emission efficiency level dynamically affects how green technological innovation impacts industrial structural optimization. Under different carbon emission efficiency threshold levels, green technological innovation’s coefficient values and significance levels show specific differences. When the carbon emission efficiency (CEE) is less than the initial threshold (0.1033), green technological innovation (GTI) has the greatest positive influence on industrial structural optimization (ISO), with a coefficient of 0.096. As the CEE rises to the middle range (0.1033–0.3592), GTI’s promotional effect on ISO declines, with a minimum coefficient of 0.032. After reaching the medium efficiency range (0.3592–0.4052), the coefficient rises marginally to 0.038. When the CEE reaches the third threshold (0.4052), the GTI coefficient strengthens to 0.047. Clearly, when the carbon emission efficiency improves, the optimization of the industrial structure through green technological innovation shows a U-shaped tendency, followed by an upward trend.

That is, an excessively low or high carbon emission efficiency may weaken the role of green technological innovation. In contrast, moderately lenient (0.3592 < CEE ≤ 0.4052) and highly efficient paths (CEE > 0.4052) are more conducive to further stimulating the potential of technological innovation for industrial structural optimization and upgrading.

5.4. Heterogeneity Analysis

Due to natural, geographical, and social factors, there are differences in carbon emissions and green technological innovation levels among different regions in China. To further investigate the differentiated moderating role of carbon emission efficiency constraints in the different areas, this study divides 30 provinces on the Chinese Mainland (excluding Tibet, Hong Kong, Macao, and Taiwan) into three major regions—eastern, central, and western—for heterogeneity testing. The results of the heterogeneity analysis for the benchmark regression model and the moderation effect model are shown in

Table 11. Heterogeneity analysis of benchmark regression.

	(1)	(2)	(3)	(4)
	ISO (National)	ISO (Eastern)	ISO (Central)	ISO (Western)
GTI	0.029 ***	0.024 ***	0.033 ***	0.025 ***
	(0.005)	(0.008)	(0.008)	(0.007)
CEE	0.190 ***	0.385 ***	0.032	−0.021
	(0.037)	(0.055)	(0.048)	(0.138)
Controls	Yes	Yes	Yes	Yes
Cons	0.273 ***	0.050	0.817 ***	0.166
	(0.085)	(0.150)	(0.140)	(0.133)
N	690.000	276.000	207.000	207.000
Adj. R2	0.629	0.826	0.590	0.589

Note: Standard errors are reported in parentheses. The symbols *** denote significance at the 1% level.

and

Table 12. Heterogeneity analysis of moderation effect.

	(1)	(2)	(3)	(4)
	ISO (National)	ISO (Eastern)	ISO (Central)	ISO (Western)
GTI	0.015 ***	0.015 *	0.010	0.012
	(0.005)	(0.008)	(0.011)	(0.010)
CEE	−0.289 ***	0.019	−0.379 **	−0.597 *
	(0.079)	(0.160)	(0.155)	(0.353)
TJ (GTI* CEE)	0.084 ***	0.049 **	0.104 ***	0.072 *
	(0.012)	(0.020)	(0.037)	(0.040)
Controls	Yes	Yes	Yes	Yes
Cons	0.557 ***	0.262	1.038 ***	0.196
	(0.093)	(0.173)	(0.159)	(0.134)
N	690.000	276.000	207.000	207.000
Adj. R2	0.653	0.829	0.604	0.593

Note: Standard errors are reported in parentheses. The symbols ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

, respectively.

As shown in Table 11, green technological innovation in China’s eastern, central, and western regions universally promotes the optimization of the industrial structure, in the order of “central region > western region > eastern region”.

In the eastern region, green technological innovation significantly and positively impacts industrial structural optimization at the 99% confidence level. However, its coefficient is lower than the national average. A possible reason is that green technological innovation in the eastern region has entered a mature stage, with technological applications tending to stabilize and marginal returns decreasing.

The central region has the highest and most significant coefficient of green technological innovation, at 0.33%. This may indicate that the effect of green technological innovation on industrial optimization and upgrading is more pronounced in the central region due to technology transfer from the eastern region and policy preference. Meanwhile, the high proportion of traditional manufacturing in the central area, such as low-carbon transformation in steel and chemical industries, notably impacts the overall industrial structural optimization.

The coefficient of green technological innovation in the western region is also lower than the national average but slightly higher than in the eastern region. This can be attributed to the lagging infrastructure and path dependence on the resource-based economy in the western region.

The carbon emission efficiency coefficient in the eastern region is 0.385 and passes the 1% significance test, which is much higher than the national level. This indicates that its carbon emission efficiency has a powerful positive moderating effect on green technological innovation. A possible reason is that the eastern region has highly efficient emission reduction technologies, which forms a positive cycle between carbon emission efficiency and technological innovation. Policy instruments such as carbon markets and green finance effectively incentivize enterprises to improve their carbon emission efficiency, reinforcing technological innovation’s effect on industrial optimization and upgrading.

The regression coefficient of carbon emission efficiency on industrial structural optimization in the central region is 0.032. Still, it is insignificant, indicating that the moderating role of carbon emission efficiency has not been fully realized. This may be because heavy industries such as coal and metallurgy still account for a large proportion of the central region. Although green technologies have been introduced, improvements in carbon emission efficiency are limited in the short term due to energy structural constraints.

The carbon emission efficiency coefficient in the western region is negative and insignificant. The high cost of low-carbon transformation in the western region will likely crowd out industrial resources and instead inhibit structural optimization in the short term.

Table 12 presents the results of the heterogeneity analysis of the moderation effect. As shown in the table, after introducing the interaction term between green technological innovation and carbon emission efficiency, the interaction coefficients for China’s eastern, central, and western regions are positive at different significance levels.

For the eastern region, the moderation effect is significantly positive at 0.049, which is under the 5% significance level but lower than the national average and the levels in the central and western regions. This indicates weaker synergistic effects, possibly because technological applications in the eastern region are approaching saturation, green technologies are already relatively mature, and improvements in carbon emission efficiency yield limited marginal gains for technological innovation. Meanwhile, as high-carbon industries in the eastern region have shifted to central and western areas, the local industrial structure is dominated by service and high-tech industries, so synergistic effects rely more on optimizing existing stock technologies.

The central region shows the most substantial moderating effect, with the most significant synergy between green technological innovation and carbon emission efficiency, much higher than in other nationwide regions. This can be attributed to the low-carbon transformation policy under the Central China Rise Strategy, which promotes the green retrofitting of high-energy-consuming industries such as steel and chemicals, further strengthening the synergy between technology and carbon emission efficiency. In addition, the high demand for the transformation of high-carbon sectors (with a high proportion of heavy industries) in the central region drives enterprises to urgently reduce the cost of emission reduction through technological innovation, forming a closed loop of “technological upgrading–efficiency improvement–structural optimization”.

In the western region, the moderation effect is positive but only significant at the 10% level, indicating that the synergistic effects have not been fully released. This is probably due to the weak technological base in the western region: the lagging application of green technological innovation makes it difficult to form an effective interaction with carbon emission efficiency, thereby suppressing the moderating effect.

Table 13. Heterogeneity analysis of threshold regression.

	(1)	(2)	(3)	(4)
	ISO (National)	ISO (Eastern)	ISO (Central)	ISO (Western)
Core Explanatory Variable	GTI	GTI	GTI	GTI
Threshold Variable	CEE	CEE	CEE	CEE
GTI1 (CEE ≤ 0.1033)	0.096 ***	0.026 ***	0.043 ***	0.064 ***
	(0.013)	(0.007)	(0.008)	(0.013)
GTI2 (0.1033 < CEE ≤ 0.3592)	0.032 ***	0.031 ***	0.053 ***	0.026 ***
	(0.004)	(0.007)	(0.010)	(0.006)
GTI3 (0.3592 < CEE ≤ 0.4052)	0.038 ***	0.034 ***	0.039 ***	0.030 ***
	(0.005)	(0.007)	(0.008)	(0.006)
GTI4 (0.4052 < CEE)	0.047 ***	0.042 ***	0.045 ***	0.027 ***
	(0.005)	(0.008)	(0.007)	(0.006)
Controls	Yes	Yes	Yes	Yes
Cons	0.373 ***	0.082	0.991 ***	0.181
	(0.080)	(0.143)	(0.129)	(0.124)
N	690.000	276.000	207.000	207.000
Adj. R2	0.682	0.847	0.652	0.638

Note: Standard errors are reported in parentheses. The symbols *** denote significance at the 1% level.

shows the results of the heterogeneity analysis of the threshold regression. Column (1) shows the threshold effect regression results for the national sample, as previously discussed, which exhibits a U-shaped nonlinear characteristic. Columns (2), (3), and (4), respectively, display the threshold effect regression results for the eastern, central, and western regions.

In the eastern region, the coefficients of green technological innovation (GTI) across threshold intervals stabilize between 0.026 and 0.042. While all intervals are significantly positive, the coefficients are lower than at the national level and show no significant increase. The possible reason is that the eastern region has already achieved the widespread adoption of green technologies. Although the marginal effect of green technological innovation overall is low, it is still in a slow upward phase. The central region displays an accelerated breakthrough pattern, peaking at 0.053 within medium-efficiency thresholds (0.1033 < CEE ≤ 0.3592), significantly exceeding both national benchmarks and its performance in other ranges. This surge reflects synergistic industrial capacity absorption during a critical development phase. The western region shows its strongest GTI impact (0.064 ***) at the low-efficiency threshold (CEE ≤ 0.1033), but coefficients decline thereafter. This pattern indicates a widespread inability to cross the initial efficiency threshold, constraining the structural optimization potential despite the early-stage promise. Together, these findings show how changes in the stage of CEE development reshape the national U-shaped curve of green technological innovation, driving industrial structural optimization.

Figure 3 shows that there is spatial heterogeneity in the coefficients of green technological innovation (GTI) on industrial structural optimization (ISO) at carbon emission efficiency (CEE) thresholds. The national sample (Figure 3a) shows a clear U-shaped trend: the GTI coefficient peaks at 0.096 when CEE ≤ 0.1033, then falls within the medium-efficiency range (0.1033 < CEE ≤ 0.4052) before rising again in the high-efficiency zone (CEE > 0.4052). Regional differentiation is distinct: In the eastern region (Figure 3b), the coefficient of influence of GTI on ISO is stable at 0.026–0.042, with a slow increase at low levels. In the central region, the coefficient of influence of GTI on ISO shows an inverted U-shaped peak of 0.053 in the interval (0.1033 < CEE ≤ 0.3592), with an overall fluctuating trend. In the western region, the coefficient of influence of GTI on ISO continues to decay from 0.064 in the low-efficiency zone to 0.027 in the high-efficiency zone, with an initial threshold dilemma.

To examine regional variations in green technological innovation (GTI), we divided 30 provinces (excluding Tibet, Hong Kong, Macao, and Taiwan) into high- and low-GTI groups.

Table 14. Heterogeneity analysis by green technological innovation level.

	(1)	(2)	(3)	(4)	(5)	(6)
	ISO (National)	ISO (High GTI)	ISO (Low GTI)	ISO (National)	ISO (High GTI)	ISO (Low GTI)
GTI	0.031 ***	0.020 ***	0.019 ***	0.015 ***	0.026 **	0.002
	(0.005)	(0.007)	(0.006)	(0.005)	(0.011)	(0.007)
CEE				−0.289 ***	0.472	−0.404 ***
				(0.079)	(0.366)	(0.120)
TJ (GTI* CEE)				0.084 ***	0.012	0.107 ***
				(0.012)	(0.036)	(0.028)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Cons	0.226 ***	−0.418 **	0.148	0.557 ***	0.112	0.252 **
	(0.087)	(0.164)	(0.097)	(0.093)	(0.166)	(0.102)
N	690.000	345.000	345.000	690.000	345.000	345.000
Adj. R2	0.615	0.748	0.153	0.653	0.788	0.187

Note: Standard errors are reported in parentheses. The symbols *** and ** denote significance at the 1% and 5% levels, respectively.

presents the heterogeneity results for baseline and moderating effect models.

Column (1) presents the baseline regression model for the national sample. Column (2) shows the baseline regression model for regions with high green technological innovation levels. Column (3) displays the baseline regression model for regions with low green technological innovation levels. In regions with high green technological innovation levels, the coefficient of Gti is 0.020. Although it is significantly positive, its value is lower than the national sample. This indicates that in regions with high innovation levels, the promoting effect of green technological innovation on Ind is weaker than the national average. This may be because regions with high innovation have reached the stage of diminishing marginal effects or because there are other constraints. In regions with low green technological innovation levels, the coefficient of Gti is 0.019. Its value is close to but slightly lower than in regions with high green technological innovation levels. A possible reason is the high intra-group heterogeneity or the existence of unobserved interfering factors. The “potential space” for green technological innovation in these regions is ample, but other conditions, such as supporting policies and insufficient funds, may restrict the actual effect.

Column (4) shows the moderating effect model for the national sample. Column (5) presents the moderating effect model for regions with high green technological innovation levels. Column (6) displays the moderating effect model for regions with low green technological innovation levels. In regions with high green technological innovation levels, the coefficient of GTI is 0.026. It is significantly positive, and its value is higher than the national sample. The coefficient of TJ (GTI* CEE) is 0.012. It is positive but not significant. This indicates that CEE fails to produce a significant moderating effect in regions with high innovation levels. This may be because the carbon emission efficiency in regions with high green technological innovation levels has reached a saturated state. In regions with low green technological innovation levels, the coefficient of GTI is 0.002. The coefficient of TJ (GTI* CEE) is 0.107. It is positive and significant. This shows that in regions with low green technological innovation levels, the moderating effect of CEE is powerful. Improving the carbon emission efficiency will significantly amplify the promoting impact of green technological innovation in optimizing the industrial structure. It is a key path for regions with low innovation levels to break through development bottlenecks.

5.5. Robustness Test

To ensure the reliability of the research results, this study conducts the following robustness tests.

First, the core explanatory variable green technological innovation is replaced. The explanatory variable changes from the number of green patent applications (GTI) to the number of green patent approvals (GTI1). The benchmark model and the moderation effect model above are re-estimated. The results of the robustness regression with the replaced core explanatory variable are shown in

Table 15. Results of replacing the core explanatory variable.

	(1)	(2)	(3)
	ISO (Replacing Core Explanatory Variable)	ISO (with CEE)	ISO (Moderating Effect)
GTI1	0.029 ***	0.025 ***	0.014 ***
	(0.005)	(0.005)	(0.005)
CEE		0.173 ***	−0.192 ***
		(0.038)	(0.073)
TJ1 (GTI1* CEE)			0.069 ***
			(0.012)
Controls	Yes	Yes	Yes
Cons	0.205 **	0.219 **	0.440 ***
	(0.087)	(0.085)	(0.092)
N	690.000	690.000	690.000
Adj. R2	0.613	0.624	0.642

Note: Standard errors are reported in parentheses. The symbols *** and ** denote significance at the 1% and 5% levels, respectively.

.

Columns (1) and (2) present the benchmark regression results with the replaced explanatory variable. Column (3) also introduces the moderation effect model with the interaction term between the substituted explanatory variable and carbon emission efficiency. Overall, green technological innovation has a significant promoting effect on industrial structural optimization. After replacing the explanatory variable, the interaction term also shows a significant moderating effect on industrial structural optimization. These results are consistent with the previous regression results, demonstrating the robustness of the above conclusions.

Secondly, we conduct robustness tests using lagged terms of the core explanatory variable. We conduct a lag effect test by introducing the core explanatory variable’s first-order and second-order lagged terms into the regression models. We re-run the benchmark model and the moderating effect model for this purpose. The robustness regression results for the lag effects are shown in

Table 16. Robustness test results of lag effects.

	(1)	(2)	(3)	(4)	(5)
	ISO	ISO (L. GTI)	ISO (L2. GTI)	ISO (L. GTI)	ISO (L2. GTI)
GTI	0.031 ***	0.006	0.008	−0.008	−0.005
	(0.005)	(0.007)	(0.006)	(0.007)	(0.007)
CEE				−0.375 ***	−0.423 ***
				(0.110)	(0.154)
TJ (GTI* CEE)				0.087 ***	0.093 ***
				(0.015)	(0.018)
Controls	Yes	Yes	Yes	Yes	Yes
L. GTI		0.034 ***		0.031 ***
		(0.007)		(0.007)
L2. GTI			0.035 ***		0.031 ***
			(0.006)		(0.006)
Controls	Yes	Yes	Yes	Yes	Yes
Cons	0.226 ***	0.366 ***	0.401 ***	0.664 ***	0.707 ***
	(0.087)	(0.091)	(0.094)	(0.097)	(0.099)
N	690.000	660.000	630.000	660.000	630.000
Adj. R2	0.615	0.645	0.653	0.672	0.682

Note: Standard errors are reported in parentheses. The symbols *** denote significance at the 1% level.

.

As indicated in the table, Columns (2) and (3) present the benchmark regression models with the first-order and second-order lagged green technological innovation, respectively. To eliminate dynamic specification bias, we re-estimate the models using the first-order and second-order lagged green technological innovation as the core explanatory variables. The coefficients of lagged GTI remain significantly positive at 0.034 and 0.035, respectively, consistent with the benchmark results. This suggests that the promoting effect of green technology on industrial structural optimization is not affected by temporal assumptions.

Columns (4) and (5) show the moderating effect models with the first-order and second-order lagged green technological innovation, respectively. The moderating variable TJ (GTI* CEE) coefficients remain significantly positive, at 0.087 and 0.093, respectively. This indicates that the moderating effect of carbon emission efficiency has a time lag. Green technological innovation and the carbon emission efficiency require time accumulation to generate synergistic effects and jointly promote industrial structural optimization.

Thirdly, the instrumental variable (IV) method is used. As there may be bidirectional causality between the study variables of green technological innovation and industrial structural optimization, this study selects the first-order and second-order lagged terms of green technological innovation as instrumental variables. It adopts the Two-Stage Least Squares (2SLS) method to test the regression model.

To verify the rationality, validity, and feasibility of the instrumental variables, the test results of the instrumental variables are shown in

Table 17. Instrumental variable test results.

	(1)	(2)	(3)
First-stage F-value	253.12	270.92	81.61
Kleibergen–Paap rk LM statistic	146.331	148.211	45.294
Kleibergen–Paap Wald rk F statistic	253.116	270.922	81.612

. The first-stage F-statistics of the three regression models are all greater than 10, indicating that the instrumental variables are rationally selected. The Kleibergen–Paap rk LM statistic tests the non-identifiability of the instrumental variables, and the results show that all p-values are 0, rejecting the null hypothesis and proving the identifiability of the instrumental variables. The Kleibergen–Paap rk Wald F-statistic tests the problem of weak instrumental variables, and the results show that both the benchmark regression model and the moderation-effect model exclude the problem of weak instrumental variables.

The results of the two-stage regression (2SLS) are shown in

Table 18. The endogeneity test results.

	(1)	(2)	(3)
	ISO	ISO	ISO
GTI	0.052 ***	0.050 ***	0.043 ***
	(0.007)	(0.007)	(0.010)
CEE		0.399 ***	0.084
		(0.092)	(0.251)
TJ (GTI* CEE)			0.040
			(0.028)
Controls	Yes	Yes	Yes
N	630.000	630.000	630.000
Adj. R2	0.622	0.648	0.656

Note: Standard errors are reported in parentheses. The symbols *** denote significance at the 1% level.

. Here, columns (1) and (2) present the benchmark regression results without and including the carbon emission efficiency variable. Column (3) further introduces the interaction term (TJ) between green technological innovation and carbon emission efficiency. Specifically, the following can be noted.

In both the benchmark regression model and the moderation-effect model, the coefficients of green technological innovation on industrial structural optimization are significantly positive at the 1% level, indicating that green technological innovation significantly promotes industrial structural optimization, further confirming the reliability of the conclusion.

Without the interaction term, the carbon emission efficiency significantly promotes efforts to optimize and upgrade the industrial structure at the 1% significance level. With the interaction term, the carbon emission efficiency still positively drives industrial structural optimization. Similarly, the interaction term between green technological innovation and the carbon emission efficiency positively affects industrial structural optimization. The conclusion remains robust.

6. Conclusions and Policy Recommendations

6.1. Main Conclusions

Against the “double-carbon” targets, enhancing green technological innovation capabilities is crucial for advancing China’s industrial structural optimization and green transformation. Based on panel data from 30 Chinese provinces from 2000 to 2022, this study empirically investigates the internal logical relationships among green technological innovation, carbon emission efficiency, and industrial structural optimization from the perspective of carbon emission efficiency constraints. The research findings are as follows:

First, green technological innovation can effectively promote regional industrial structural optimization. This effect remains valid after a series of robustness tests, demonstrating the strong impetus of green technological innovation for the green development of industries. Second, the carbon emission efficiency has a significant positive driving effect on industrial structural optimization. Third, the carbon emission efficiency positively moderates the impact of green technological innovation on industrial structural optimization, exhibiting a U-shaped nonlinear characteristic. Moreover, this moderating effect varies across regions, showing a “central region > western region > eastern region” pattern.

6.2. Policy Recommendations

To further stimulate the positive role of green technological innovation in industrial structural optimization under the double-carbon goals, this paper proposes the following policy recommendations.

At the policy level, construct a “technology–emission–industry” trinity framework. Integrate green technological innovation into the national medium- and long-term science and technology plan and establish special funds to support key technological research in new energy, carbon capture, hydrogen-based steelmaking, and other fields. Additionally, it will improve the accounting standards for carbon emission efficiency, establish a unified national carbon efficiency monitoring platform, and promote key high-carbon industries to formulate efficiency improvement roadmaps.

Implement differentiated coordination strategies at the regional level.

The eastern region should enhance high-end technology and market-driven development. Relying on the innovation hubs in the Yangtze River Delta and the Pearl River Delta, it should establish technology demonstration zones. The region should focus on breaking through key technologies, such as new energy vehicle battery recycling and energy storage in smart grids. It should also tackle technical challenges and promote industrial energy-saving solutions in semiconductors and high-end equipment manufacturing industries. These efforts aim to boost the international competitiveness of green technologies.

The central region should prioritize the green transformation of traditional industries and policy coordination. Special funds should be established to support clean production upgrades in steel, cement, and other traditional sectors. Particular emphasis should be placed on promoting hydrogen-based steelmaking through targeted policy incentives. A cross-provincial carbon compensation mechanism should be implemented. This should be complemented by flexible subsidy policies aligned with industry decarbonization needs. Major heavy industrial provinces should be encouraged to secure transition funding through carbon trading. Policy coordination must be strengthened to ensure green technological implementation. This requires establishing a regulatory framework for technological applications. Additionally, training programs should be provided for industry practitioners.

The western region should prioritize the layout of clean infrastructure and ecological value conversion. It should leverage its natural resource advantages to build large-scale new energy bases, focusing on developing new energy storage technologies such as photovoltaic hydrogen production and vanadium flow batteries. The region should radiate industrial demand to the eastern and central regions and promote decarbonization in high-carbon industries such as steel and chemicals. This can be achieved by piloting green electricity substitution in electrolytic aluminum plants and applying CCUS technology in coal chemical industries.

7. Insufficient Research and Prospects

Despite the multidimensional empirical analysis of the internal logic among green technological innovation, carbon emission efficiency, and industrial structural optimization, this study still has several limitations that should be addressed in future research.

Firstly, this study uses provincial-level data. It limits the ability to capture more detailed spatial, sectoral, and firm-level heterogeneity. From a spatial perspective, this study is based on provincial data and does not fully explore sub-regional and city-level differences. Future sub-regional studies should analyze how regional characteristics moderate the identified relationships. For instance, resource-based cities in Shanxi and export-oriented cities in Guangdong vary significantly regarding green technology, the policy response, and industrial bases. At the city level, future research could use higher-resolution data and case studies to analyze different city types. This can clarify the impact of green technological innovation, improving the practical value of the conclusions. From an industry perspective, this study does not examine how different sectors respond differently to low-carbon policies and transformation pressures. Industries already covered by China’s carbon trading system face stronger policy constraints and incentives than sectors not yet included. Future research should assess how industries adapt to the carbon market and explore their transformation paths. From an enterprise perspective, this study does not analyze how firms, especially under current economic pressures, balance carbon reduction goals with cost control. In reality, firms’ responses to green policies differ depending on their size, ownership, and sector. Their decisions about adopting green technologies often involve complex trade-offs between emission targets and productivity. Future research should use firm-level data to investigate different types of enterprises’ behavioral and strategic responses.

Secondly, this study uses quantitative indicators of carbon emission efficiency but does not consider other important factors, such as financing constraints and environmental regulations. Additionally, this study overlooks the approach of quantifying the economic–environmental co-benefits of green technologies within a cost–benefit analysis framework, such as focusing on calculating the cost of carbon reduction per unit of GDP. In the future, we can incorporate quantitative indicators of cost–benefit synergy into regional policy evaluation models to provide more precise guidance for low-carbon development strategies.

Finally, the model assumes that the moderating role of carbon emission efficiency is relatively stable. However, it does not account for external shocks and policy shifts, such as U.S. climate policy rollbacks, trade wars, tariff changes, or sudden international policy adjustments, which may alter these dynamics. Future studies should apply event-based analysis to test the effects of such shocks and improve the model’s capacity to handle policy uncertainty. This would enhance the timeliness and applicability of the research.

By addressing these limitations and research prospects, we hope to offer new perspectives for scholars studying sustainable development under dual-carbon goals and provide policymakers with more targeted guidance promoting green and sustainable industrial transformation.