Digitalization, Carbon Productivity and Technological Innovation in Manufacturing—Evidence from China

Abstract

Manufacturing is one of the vital carbon emission sources in China; its carbon emission reduction and carbon productivity improvement are the keys to the successful realization of “carbon peaking and carbon neutrality”. This paper investigates the impact of regional manufacturing digitalization on carbon productivity based on a panel data set covering 30 provinces in China over time from 2013 to 2020. We applied the mixed Ordinary Least-Squares (OLS) regression effect model and instrumental variable method, using a mediation effect model, and identified that technological innovation is the potential transmission channel of manufacturing digitalization affecting carbon productivity. The empirical results show that: (1) Digitalization of manufacturing can effectively contribute to increased carbon productivity. (2) Technological innovation plays a partial intermediary role in the impact of carbon productivity through the digitalization of manufacturing, and there is still much room for improvement. (3) Promoting the digitalization of manufacturing will promote technological innovation, which is an important influencing factor for manufacturing enterprises to enhance technological innovation. This research provides theoretical support for achieving carbon productivity in manufacturing in the context of digital development.

1. Introduction

Since its reform and opening up, China has made significant achievements. However, the long-term rough development model has not only led to low efficiency in resource allocation and weak technology in China but also caused the economy to face severe challenges in changing to a green and sustainable development model [1]. To control carbon emissions, China has been committed to promoting the implementation of carbon emission reductions in recent years and put forward the strategic goal of “carbon peaking and carbon neutrality” in September 2020. Under the current development situation, the only way for China to achieve these dual goals is to improve carbon productivity [2,3]. It is undoubtedly a critical step to ensure the smooth realization of “carbon peaking and carbon neutrality”. The “Made in China 2025” strategy puts forward “green development”, which requires manufacturers to take the sustainable development path of energy conservation and emission reduction. The development mode of manufacturing must be transformed from factor expansion to efficiency enhancement [4]. The 2021 Digital Carbon Neutral White Paper states that digital technologies are driving key industries’ digital and green transformation, and the digital economy has great potential for carbon emission reduction [5]. At present, the technical methods and cost-efficiency advantages of the digital economy have become a vital force in promoting low-carbon transformation and play an indispensable role in the implementation of carbon emission reduction targets, especially in the process of economic recoveries; technologies such as 5G, blockchain, and industrial internet have emerged in large numbers and rapidly penetrated the real economy, accelerating integration with the manufacturing industry, and the digitalization manufacturing is constantly improving [6].

Policy systems are an important guarantee to achieve “carbon peaking and carbon neutrality”. Tan [7] analyzed China’s low-carbon policies since the Eleventh Five-Year Plan, delineated the policy stages of “carbon peaking and carbon neutrality”, and analyzed the technical support of related policies. Dong [8] sorted the commitments made by China for emission reduction at international conferences, analyzed the efforts needed to be made by various industries under the target, and put forward policy recommendations for optimizing the energy structure, promoting technological innovation, and strengthening market guidance. Fu [9] analyzed China’s carbon emission reduction policies from four major emission areas: energy, industry, construction, and transportation, and made suggestions for optimizing China’s carbon emission reduction policies. Dubash [10] examined the systematic framework of national climate institutions and found that countries’ climate policies and their ability to implement them are closely linked to their political base. Newell [11] examined the development of carbon emission trading markets in various countries and proposed an optimization method for carbon emission trading systems. These studies analyzed the existing carbon reduction policies and proposed countermeasures to address the problems. For China, “carbon peaking and carbon neutrality”, as a strategic way to achieve high-quality development, is not only a national policy to address climate change, but also an inherent need to transform and upgrade the economic structure towards sustainable development.

In the context of “carbon peaking and carbon neutrality”, we expect that the optimal economic development approach is to keep the economy moving in the direction of sound and sustainable growth, while at the same time promoting a shift toward environmentally friendly and sustainable development. A sustainable development strategy is conducive to promoting the unity of ecological, economic, and social benefits, is conducive to promoting the transformation of the economic growth mode from rough and loose to intensive, so that economic development is coordinated with population, resources, and environment. It is also conducive to the sustained, stable, and healthy development of the national economy and the improvement of people’s living standards and quality. The vital way for China to achieve “carbon peaking and carbon neutrality” and sustainable economic growth is to improve carbon productivity. Currently, digitalization is expected to be the main driving force for these points. On the one hand, digital development takes information networks and communication technology as carriers. With its strong resource allocation efficiency, high level of technology innovation, low marginal cost loss and efficient real-time transmission, and other characteristics, it breaks the spacetime limitation of traditional economic development. It also provides the possibility for high efficiency and high-quality development of manufacturing [12]. On the other hand, the proposal of “carbon peaking and carbon neutrality” brings new opportunities for manufacturing in high-quality and environmentally friendly development modes, but also challenges the development efficiency and speed of manufacturing.

Does the digitalization of manufacturing improve carbon productivity? What is their exact relationship, and what is the specific conduction path? Considering the above problems, this paper will conduct a systematic study on the relationship between the digitalization and carbon productivity of manufacturing from the theoretical and empirical perspectives to understand the correlation between the two more clearly, and promote the digital construction actively on this basis, to provide beneficial policy inspiration for the realization of “carbon peaking and carbon neutrality”. Previous studies have focused on the digitalization of manufacturing or carbon emission reduction in manufacturing. However, there is a lack of research on the relationship between the two, especially on the role pathway between the digitalization of manufacturing and carbon productivity, and so this study hopes to fill the gap in this area and provide a new academic perspective on achieving “carbon peaking and carbon neutrality”.

The scientific novelty of this study lies in the comprehensive analysis of digitalization in China’s manufacturing, as well as the adoption of scientific empirical approaches to study the driving effect of digitalization in manufacturing on improving carbon productivity, and the differences in the driving effect in different regions. In addition, this study also innovatively analyzes the conduction path of manufacturing digitalization to improve carbon productivity by driving the development of technological innovation and tries to explore new development directions for improving carbon productivity in China’s manufacturing.

The main contributions of this study can be summarized as follows: first, this study creatively combines the two research fields of the digitalization of manufacturing and the carbon emissions of manufacturing, filling the gap between digitalization and the enhancement of carbon productivity. We find a correlation between the digitalization of manufacturing and carbon productivity, and explore the conduction path between the two. Second, this paper studies the impact effect of manufacturing digitalization from the perspective of green and sustainable development, revealing the direct impact of manufacturing digitalization on carbon productivity from a theoretical perspective and verifying it through empirical methods, as well as introducing technological innovation to study the mediating effect. Third, this study provides insights into improving carbon productivity and has explored a positive conduction path to enhance carbon productivity by promoting the digitalization of manufacturing.

The rest of this study is organized as follows: Section 2 reviews the relevant literature. Section 3 summarizes the impact mechanism of digitalization in manufacturing on carbon productivity and proposes relevant hypotheses. Section 4 conducts the research design, and interprets each indicator and calculation method in detail. Section 5 provides an analysis and discussion of the empirical results, and finally, Section 6 summarizes the conclusions and implications of this study.

2. Literature Review

The extant collaboration literature on manufacturing and carbon productivity has achieved fruitful results. This study reviews the literature in terms of the following aspects: the digitalization of manufacturing, the carbon productivity of manufacturing, and the relationship between digitalization and carbon productivity.

2.1. Digitalization of Manufacturing

With the rapid development of the new generation of information technology, such as big data, cloud computing, and artificial intelligence, the digital economy is profoundly changing the basic concept of manufacturing and the logic of value creation. It has become a new driving force leading economic growth [13]. Specifically, the value dimension of digitalization in manufacturing is reflected in four aspects: driving the improvement of industrial efficiency, promoting industrial cross-border integration, reconstructing the competitive model of industrial organizations, and empowering industrial upgrading [14]. In the existing research related to the digitization of manufacturing, scholars have pointed out from the perspectives of innovation driving [15], efficiency improvement [16], and structure optimization [17] that the process of digitalization in manufacturing changes the industrial form of manufacturing by innovating products and technologies with digital technology; manufacturing intelligence refers to optimizing the production process through advanced manufacturing technology to make it flexible and thus achieve a more efficient transaction process; and digital transformation is a complex process of using digital technology to reconfigure products, services, organizational structures, and even business models.

Based on the background of digitalization in manufacturing, Beier [18], Bueno [19], and Bag [20] also proposed that “Industry 4.0” aims to apply the principles and technologies of the Internet of Things to manufacturing and that intelligent manufacturing is the core of “Industry 4.0”. Bonvillian [21] and others have pointed out that the advanced manufacturing program can play a role in restructuring manufacturing, expanding the entrepreneurial model, and reversing the severe social chaos caused by the manufacturing recession. Foreign scholars’ research on “Industry 4.0” and “Advanced Manufacturing Leadership Strategy” has put forward the trend of future development of manufacturing, i.e., intelligence and digitalization, which is similar to the current strategy of intelligent and digital transformation of manufacturing in China.

In addition, at the macro level, the digital economy positively impacts the quality of manufacturing development, and the innovation environment plays a positive regulatory role [22], which will be amplified after a specific scale limit is breached [23]. At the meso level, digital technologies can significantly contribute to manufacturing growth and improve manufacturing quality by reducing costs, reducing the number of workers, and improving profitability [24]. At the micro level, digital technology can improve the efficiency and performance of manufacturing enterprises in product development [25], manufacturing, and operation management. It can also improve the service of manufacturing enterprises [26], linking them closely with consumers and providing them with more diversified products and services, thus helping to transform and upgrade the manufacturing industry into a service industry [27].

2.2. Carbon Productivity of Manufacturing

Low-carbon sustainable development is closely related to global development; it is a development concept that can adapt to the needs of contemporary society and will not bring harm to future generations, so as to achieve the sustainable use of resources and the sustainability of the ecological environment [28]. Xie and Zhu [29] constructed a conditional process effect model based on binary legitimacy theory to investigate in depth the mechanisms of action and boundary conditions of green innovation in manufacturing enterprises affecting their sustainable development. Carbon productivity is the gross domestic product (GDP) ratio over a specific time period to carbon dioxide emissions over the same period [30], with higher figures indicating more significant green development efficiency per unit [31]. Unlike traditional productivity, carbon productivity is a “balanced” indicator that effectively integrates economic development and environmental protection, seeking to harmonize the two. It can accurately measure the degree of decoupling between economic development and carbon dioxide emissions and judge the progress of “carbon peaking and carbon neutrality” policies [32]. Against the goal of “carbon peaking and carbon neutrality”, improving carbon productivity is crucial to constructing China’s ecological civilization and high-quality economic development. Presently, research on carbon productivity has become the focus of attention of domestic scholars, especially on the factors influencing carbon productivity. Established research suggests that technology spillovers [33], industrial restructuring [34], technological progress [35], technological innovation performance [36], and global value chain embeddedness [37] are effective in increasing carbon productivity. At the same time, export commodity structure [38] and energy mix [39] reduce carbon productivity.

Some scholars have also studied the characteristics of carbon productivity trends at different levels: from a country perspective, carbon productivity, in general, shows an increasing trend in all countries in the world, with relatively high carbon productivity in high-income countries and relatively low carbon productivity in China, which lies below the world average [40]. From China’s regional perspective, the carbon productivity of China’s provinces generally tends to increase year on year. However, the eastern region is growing significantly faster than the central and western regions [41]. In the industry, China’s manufacturing sector also shows an increasing trend in carbon productivity, with higher technology industries having higher carbon productivity [6]. Based on the trends and characteristics of carbon productivity, many scholars have conducted empirical studies on its influencing factors.

2.3. The Relationship between Digitalization and Carbon Productivity

Along with the booming digital economy and the “carbon peaking and carbon neutrality” policies proposed, a growing body of literature focuses on the energy-saving and emission-reduction effects of digital transformation. Deng and Zhang [42] used the data from 285 cities in China from 2012 to 2018 as a sample to verify that the development of digital finance can promote energy conservation and emission reduction through economic growth, industrial restructuring, and technological innovation. Yi [43] and Wang’s [44] research shows that the digital economy can reduce carbon emissions through technological innovation and energy mix optimization, and that the digital economy has a spatial spillover effect on carbon emission reduction. Xie [45] used provincial data to find that the development of the digital economy can significantly reduce regional carbon emissions through biased technological advances. Peng [46] demonstrated that corporate IT capability can promote green technological innovation. Zhang C and Liu C [47] conducted an empirical study on the carbon reduction potential of the digital technology industry or digital economy. They found that the ICT industry significantly contributes to CO₂ reduction in China, and the impact of CO₂ reduction is more significant in the central region than in the eastern region. Bhujabal P [48] argued that increased investment in digital infrastructure also has a significant effect on reducing carbon emissions. Asongu [49] argued that Internet penetration in Africa expands trade openness and reduces carbon emissions. Khan [50] and Tsaurai [51] argued that in emerging market countries, IT accelerates financial development and significantly reduces carbon emissions. Salahuddin [52] found that a 1% increase in internet usage in EU countries is associated with a 0.16% increase in carbon emissions. Xu [53] found that the digital economy’s infrastructure, structural optimization, technological innovation, and resource allocation effects significantly improved urban carbon emissions. Wu [54] argued that the direct and indirect effects of the digital economy on the digitalization of industries by improving the efficiency of resource and energy use reduced carbon emissions.

In addition, it has been shown that digital development impacts carbon productivity, mainly by increasing the efficiency of resource allocation and promoting technological innovation. From the perspective of resource allocation efficiency, Gao [55] and Ali’s [56] studies have shown that the digital economy has contributed to the integration and allocation of digital technology and traditional economic resources in all countries, thus reducing energy consumption in all countries. Ding [57] believed that the digital economy could effectively reduce the marginal costs of enterprises, adjust the industrial structure, promote the integration of the digital economy and the industrial economy, and improve the resource allocation efficiency of enterprises while promoting high-quality economic development, thereby enhancing carbon productivity. From the perspective of market supply and demand, Yang [58] found that the development of information technology in the digital economy can reduce the degree of information asymmetry between market supply and demand, achieve efficient matching between market supply and demand, further enhance the efficiency of market allocation of resources, and reduce the inefficient consumption of resources.

From the perspective of technological innovation, Hsiao [59] pointed out that the development of the digital economy not only accelerates the speed of technological innovation in enterprises but also deepens and improves the industrial structure of the digital economy, which dramatically improves the productivity of enterprises and reduces their energy losses, thus generating positive externalities on the carbon productivity of enterprises. Bukht and Heeks [60] argued that the digital economy is consistent with the traditional economy in the evolution process and evolution principle. Its core is the development and innovation of information technology, and that innovation in information and technology can effectively facilitate the integration of the digital economy with traditional manufacturing industries, realize process reengineering in manufacturing, and help promote the formation of green and low-carbon business models. Curran [61] and Brynjolfsson’s [62] research showed that the development of the digital economy has accelerated the pace of innovation in global digitalization and informatization and accelerated the cross-border innovation and sharing of carbon reduction technologies. Zhu and Ma’s [63] research showed that the development of digital finance in the digital economy is conducive to realizing technological innovation, thereby increasing the green total factor productivity and reducing carbon emissions. Throughout the existing literature, although more and more scholars are focusing on the relationship between China’s digital transformation and the achievement of the “carbon peaking and carbon neutrality” target, most of them have only conducted macro-level analyses or analyses based on regional data, and there is a lack of sector-specific discussions. However, by improving carbon productivity, all industries are practitioners of energy saving and emission reduction, and clarifying the factors that influence energy saving and emission reduction in each industry is the basis for scientific decision-making [64]. Because of this, this paper, based on the theoretical analysis, used an empirical approach to investigate the impact mechanism of the digital transformation of China’s manufacturing on its carbon productivity and to explore the internal and external factors that enhance carbon productivity to provide more comprehensive empirical evidence for the research on the effect of energy conservation and emission reduction of China’s manufacturing industry from the micro level.

In conclusion, the theoretical contribution of this paper is to reveal the mechanism of the effect of digitalization for manufacturing on the improvement of carbon productivity, which not only enriches the relevant theoretical foundation in the field of the digital economy and carbon emission reduction but also cross-analyses the two theories to find the direct influence between the two variables and explore the indirect influence of technological innovation on both. The practical significance lies in finding the amplifier between the digitalization of manufacturing and the enhancement of carbon productivity, revealing the mediating role of technological innovation on the two core variables, and exploring a positive path for the high-quality and low-carbon development of manufacturing.

3. Research Hypotheses

Under the vision of achieving the “carbon peaking and carbon neutrality” target, the issue of carbon reduction in the manufacturing sector is currently a topic of concern for scholars. The manufacturing sector is responsible for completing the carbon reduction task. Clarifying the relevant theories and drivers will help to explore the pathway to “carbon peaking and carbon neutrality”. Based on the above discussion, we show that scholars have explored the driving factors of carbon emission reduction in manufacturing. Different from previous studies, this paper tries to clarify the theoretical mechanism that the digitalization of manufacturing affects its carbon productivity from the perspective of technological innovation.

3.1. Direct Effect of Manufacturing Digitalization on Carbon Productivity

The effect of manufacturing digitalization on its carbon productivity is achieved in two main ways: firstly, a high level of digitalization makes industries support environmentally friendly characteristics, and the negative impact on the environment will be reduced. Regions with a higher level of digital transformation in manufacturing also have a higher green level than other regions. In addition, as their economic strength is also more robust, they tend to pay more attention to environmental benefits, and their environmental regulations are stricter. The degree of importance local governments attach to environmental protection will affect the motivation of enterprises to save energy and reduce emissions. As an essential indicator of the degree of importance local governments attach to environmental protection, the strength of environmental regulations will have a specific impact on the energy-saving and emission-reduction effects of the digital transformation of enterprises [65]. For enterprises in areas with stronger environmental regulations, the contribution of digital transformation to their energy saving and emission reduction is more significant. This is because environmental regulations have an innovation compensation effect, i.e., under environmental regulations, enterprises will improve their long-term competitiveness by improving their technology level and calming down short-term costs [66]. Therefore, under solid environmental regulations, enterprises are more aware of environmental protection and pay more attention to green and sustainable development at the production and operation levels. At this time, digital transformation has a more significant role in improving green technology innovation, which can better promote energy saving and emission reduction [67].

Secondly, as the industrial foundation of the digital economy, the digital industry can promote manufacturing to reduce carbon emissions. The digital industry can transform and upgrade traditional manufacturing using the penetration and derivation of digital technology, promote the development of manufacturing towards intelligence and greening, and at the same time increase the added value of the industry, effectively improve the allocation efficiency of resources and reduce energy consumption and carbon emissions [68], and energy flow and resource flow. In terms of energy flow, digital transformation is conducive to the introduction of lean management concepts in the management of enterprises, enabling enterprises to arrange and adjust manufacturing plans and optimize inventories promptly based on data collection and feedback in the manufacturing process to achieve effective coordination between the internal “human and financial resources” of enterprises and external demand, thereby reducing overall energy consumption, improving energy use, and optimizing the energy structure [69]. In terms of resource flow, digital transformation helps to promote the development of the production process towards flexibility, networking, platform, and miniaturization [70], and to reconfigure the production process through the establishment of the Industrial Internet of Things to optimize the production process and improve the efficiency of resource use, thereby reducing the intensity of pollution emissions [71].

In summary, the higher the digitalization of manufacturing, the more it helps to increase carbon productivity. Therefore, our research hypothesis is as follows:

H1: 

Improving the digitalization of manufacturing can significantly increase its carbon productivity.

3.2. Direct Effect of Manufacturing Technological Innovation on Carbon Productivity

In 1912, Schumpeter proposed a theory of innovation to examine the impact of technological innovation on the economy and society. Only with the support of scientific and technological innovation is China expected to achieve the “carbon peaking and carbon neutrality” goal as scheduled [72]. Jiang [73] found that R&D intensity could reduce pollutant emissions from manufacturing enterprises in China. Xing [74] showed that sustainable exploration of innovation and sustainable use of innovation positively impacted the pollution emissions of manufacturing enterprises. Using a sample of heavily polluting industries, Zhang [75] found that corporate process innovation had no significant effect on environmental performance and that product innovation could improve environmental performance. Using a sample of 285 cities in China, Chen [76] confirmed that technological innovation could reduce environmental pollution, but that there is a threshold value for this effect. In summary, regarding the relationship between technological innovation and carbon emissions, most scholars believe that technological innovation positively contributes to reducing carbon emissions.

In 1995, Hart proposed a basic view of natural resources based on the resource-based view, arguing that the pro-environmental behavior of firms can bring them sustainable competitive advantage and that firms gain environmental sustainability through technological innovation, which in turn leads to their green competitive advantage. In addition, technological innovation depends mainly on the level of R&D investment, and the impact of R&D investment on corporate environmental governance can be considered in two ways [77]. (i) R&D investment enables companies to improve resource efficiency and reduce CO₂ emissions with minimum energy consumption [78]. (ii) R&D investment helps to improve clean energy technologies, and these new technologies help to improve energy efficiency [79].

In summary, R&D investment in manufacturing helps increase energy consumption and productivity, which is an essential tool for manufacturing to enhance carbon productivity. Therefore, our research hypothesis is as follows:

H2: 

Improving technological innovation in manufacturing can significantly increase carbon productivity.

3.3. The Mediating Role of Technological Innovation between Digitalization and Carbon Productivity for Manufacturing

Digitalization of manufacturing boosts carbon productivity by increasing technological innovation. This is because technologies such as big data, cloud computing, and artificial intelligence in the digital economy can radiate to different industry sectors, significantly improve technological innovation within the relevant fields, drive their element upgrade, bring new opportunities for implementing energy conservation and emission reduction, and explore more environmentally friendly and green innovative development models, thus promoting carbon neutrality and carbon emission reduction in the industry. The technological advances and digital communication technology applications embedded in digital transformation can not only improve the efficiency of production factors and energy efficiency [80,81], but also upgrade through production process automation, promote enterprises to material input, product manufacturing, and sales process of “fine” management, prompting enterprises in the precision of the production process at the same time, the control of each link of energy consumption of real-time monitoring, reduce the speed of energy consumption, reduce the waste in production, and produce positive feedback for enterprise energy conservation and emissions reduction [82]. By transforming and upgrading the production methods of traditional manufacturing, operational efficiency and energy efficiency can be significantly improved [83]. Currently, China’s energy consumption intensity is still higher than the world average. If digital transformation is used to improve the industry’s energy efficiency, it can effectively promote the reduction of carbon intensity per unit of GDP. Moreover, the improvement of energy and resource-use efficiency plays a huge role in promoting carbon emission reduction.

Digital transformation in the production and operation process may promote enterprise technological innovation in the following two ways: First, digital transformation transforms and upgrades the original production and operation mode by applying cutting-edge information and communication technologies, such as big data, cloud computing, and artificial intelligence, etc. When this advanced digital technology is used, it contains the characteristics of technological progress. Some scholars have found that introducing technologies such as “ABCD” (Artificial Intelligence, Block Chain, Cloud Computing, and Big Data) into various businesses, such as manufacturing and organization management through digital transformation, will improve enterprises’ production technology innovation capability [81]. Second, digital technology can complement other production factors and enhance enterprise technological innovation capability. According to Schumpeter’s innovation theory, enterprise innovation includes the recombination of production factors, such as adding a new combination of production factors and conditions that have never been seen before to the production system [84]. Specific to the enterprise production operation, integrating digital technology containing technological progress with other factors of production can promote enterprise production, supply chain, sales, and so on. Each link of reform and an effective connection can more directly realize the optimal configuration of production factors, speed up information circulation, and improve the enterprise technology innovation ability [85].

In summary, the increase of digitalization in manufacturing can effectively lead to the improvement of technological innovation to produce new products and improve production processes (see Figure 1). Innovation in products and processes can help to improve carbon productivity. Therefore, our research hypothesis is as follows:

H3: 

Technological innovation in manufacturing mediates between digitization and carbon productivity.

H1 is to test whether the increase in digitalization of manufacturing has a positive effect on carbon productivity. H2 is to test whether the increase in technology innovation of manufacturing has a positive effect on carbon productivity. H3 is to test whether technological innovation mediates between digitalization and carbon productivity.

4. Methodology and Data

4.1. Test Method for Mediation Effects

Digitization of manufacturing is an effective way to increase carbon productivity, and technological innovation also has an effect in this conduction path. According to the existing research results and related theoretical analysis, we construct a mediating effect model to analyze the relationship between the digitalization of manufacturing and carbon productivity, and explore whether there is a mediating effect between technological innovation and the two variables.

If the explanatory variable influences the explained variable through a specific variable, the variable is the intermediary variable. In this study, technological innovation in manufacturing (SIT) is the mediating variable, digitalization of manufacturing (Dig) is the explanatory variable, and manufacturing carbon productivity (Cp) is the explained variable. To study the mediating effect of technological innovation is to better understand the mechanism underlying the impact of manufacturing’s efforts in digital transformation on enhancing carbon productivity. Based on the previous discussion, digitalization in manufacturing acts on its carbon productivity through direct and indirect paths, where technological innovation plays a mediating role in the effect of digitalization in manufacturing on carbon productivity. In this study, we construct a mediating effect model (Figure 2) to examine the direct and indirect effects of manufacturing digitalization on carbon productivity in manufacturing with technological innovation as a mediating variable. Based on the previous discussion, digitalization in manufacturing acts on its carbon productivity through direct and indirect paths, where technological innovation plays a mediating role in the effect of digitalization in manufacturing on carbon productivity.

The stepwise regression coefficient test proposed by Baron and Kenny [86] and Wen and Ye [87] is the classic method of testing for mediating effects. This paper draws on their method and constructs models (1) to (3) to test the relationship between the variables. The specific test steps are as follows.

Step 1: Test the effect of Dig on Cp.

$$\mathrm{C}\mathrm{p}=\mathsf{\alpha }\mathrm{D}\mathrm{i}\mathrm{g}+\mathsf{\epsilon }1(\alpha \ne 0)$$

(1)

If α is not significant, it indicates that Dig may not have a significant effect on Cp. If α is significant, proceed to further tests.

Step 2: Test the effect of Dig on SIT, Dig and SIT together on Cp, and observe the significance of β, γ, and α′.

$$\mathrm{S}\mathrm{I}\mathrm{T}=\mathsf{\beta }\mathrm{D}\mathrm{i}\mathrm{g}+\mathsf{\epsilon }2(\beta \ne 0)$$

(2)

$$\mathrm{C}\mathrm{p}=\mathsf{\alpha }\mathrm{\prime }\mathrm{D}\mathrm{i}\mathrm{g}+\mathsf{\gamma }\mathrm{S}\mathrm{I}\mathrm{T}+\mathsf{\epsilon }3(\mathsf{\alpha }\mathrm{\prime }\ne 0; \mathsf{\gamma }\ne 0)$$

(3)

Step 3: Test the mediation effect of βγ according to Bootstrap method and observe whether β*γ is significant at 95% confidence interval.

The following may occur when performing Step 2 and Step 3:

I.

If β and γ are significant, then test coefficient α′; if α′ is not significant, this indicates that the direct effect is not significant and there is a full mediation effect.

II.

If β and γ are significant, then test coefficient α′; if α′ is significant, this indicates a significant direct effect. Compare the sign of βγ with α′; if the signs are the same, this indicates the existence of partial mediation, and if the signs are different, this indicates a possible masking effect.

III.

If at least one of β and γ is not significant, observe the significance of βγ. If βγ is not significant, this indicates that there is no mediating effect.

IV.

If at least one of β and γ is not significant, observe the significance of βγ. If βγ is significant, observe the significance of α′; if α′ is significant, compare the signs of βγ and α′. If the signs are the same, this indicates a partial mediation effect; if the signs are different, this indicates a possible masking effect.

V.

If at least one of β and γ is not significant, observe the significance of βγ. If βγ is significant, observe the significance of α′; if α′ is not significant, this indicates that the direct effect is not significant and there is a full mediation effect.

The coefficient α in Equation (1) is the total effect of the explanatory variable on the explained variable, the coefficient β in (2) is the effect of the explanatory variable on the mediating variable, the coefficient α′ in (3) is the direct effect of the explanatory variable on the explained variable after controlling for the effect of the mediating variable on the explained variable, and ε1, ε2, and ε3 are the regression residuals.

The mediation effect is the product of the coefficient βγ, and the relationship between the mediation effect, the direct effect, and the total effect are as follows:

$$\mathsf{\alpha }=\mathsf{\alpha }\prime +\mathsf{\beta }\mathsf{\gamma }$$

(4)

4.2. Econometric Model Building

According to the theoretical analysis above, this paper draws on Wen’s [87] method, constructing models (5) to (8) to test the hypotheses presented in the previous section:

$${\mathrm{C}\mathrm{p}}_{\mathrm{i},\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{D}\mathrm{i}\mathrm{g}}_{\mathrm{i},\mathrm{t}}+\mathsf{\gamma }\sum {\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{t}}+\sum \mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(5)

$${\mathrm{C}\mathrm{p}}_{\mathrm{i},\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{S}\mathrm{I}\mathrm{T}}_{\mathrm{i},\mathrm{t}}+\mathsf{\gamma }\sum {\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{t}}+\sum \mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(6)

$${\mathrm{S}\mathrm{I}\mathrm{T}}_{\mathrm{i},\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{D}\mathrm{i}\mathrm{g}}_{\mathrm{i},\mathrm{t}}+\mathsf{\gamma }\sum {\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{t}}+\sum \mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(7)

$${\mathrm{C}\mathrm{p}}_{\mathrm{i},\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{D}\mathrm{i}\mathrm{g}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_2{\mathrm{S}\mathrm{I}\mathrm{T}}_{\mathrm{i},\mathrm{t}}+\mathsf{\gamma }\sum {\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{t}}+\sum \mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(8)

In (5) to (8), i represents the region, t represents the year, Controls_i,t is the control variable, ε_i,t is the random disturbance term, the explanatory variable Cp is manufacturing carbon productivity, the core explanatory variable Dig is the level of manufacturing digitalization, and the mediating variable SIT is the level of manufacturing technological innovation. In addition, models (5) and (6) test the effects of Dig and SIT on Cp, respectively. The stepwise regression method was used to construct models (5), (7), and (8) to test the mediating effect of SIT between Cp and Dig, with the following steps: (i) if the effect of Dig on Cp in model (5) is significant, (ii) the effect of Dig on SIT in model (7) is significant, and (iii) if SIT passes the significance test in model (8), then there is a mediating effect.

4.3. Variables Definition

Balanced panel data covering 30 provinces in China from 2013 to 2020 ere collected for this study. The variables are constructed as follows.

(1) Explained variable: manufacturing carbon productivity (Cp). Combined with the previous discussion, the measure of manufacturing carbon productivity was borrowed from the measure of Sun and Du [88] and expressed by using the ratio of manufacturing sales output to total manufacturing carbon emissions. The results are shown with Cartesian heat maps in Figure 3. Darker blue indicates lower carbon productivity and darker red indicates higher carbon productivity. We can find that Beijing, Guangdong, Zhejiang, Jiangsu, and Shanghai are at the top, while Xinjiang, Ningxia, Heilongjiang, Inner Mongolia, and Shanxi are at the bottom. Total manufacturing carbon emissions were calculated by referring to the IPCC Guidelines for National Greenhouse Gas Inventories (2006) and the International Energy Outlook viewpoint, using the products of raw coal, coke, gasoline, kerosene, diesel, and fuel oil; the product of seven types of energy consumption data; and the corresponding CO₂ emission factors to calculate the manufacturing CO₂ emissions.

(2) Explanatory variable: digitalization of manufacturing (Dig). In terms of the measurement of the integration of digitalization and manufacturing, this paper draws on the past measurement methods of the integration of informatization and industrialization for empirical investigation [89], and makes theoretical and data contributions to measuring China’s digitalization and manufacturing integration development. First, to select provincial digitalization indicators (

Table 1. Provincial digitalization indicators system.

Type	Dimension	Indicator
Digital Investment (A)	Talent Investment (Aa)	Number of employees in the software and information technology services industries (Aa1)
		Number of people working in the computer, communications, and other electronic equipment manufacturing industries (Aa2)
		The full-time equivalent of R&D personnel in manufacturing enterprises above the designated size (Aa3)
	Digital Infrastructure Investment (Ab)	Number of manufacturing enterprises in the electronic information industry (Ab1)
		The proportion of enterprises with e-commerce transaction activities (Ab2)
		R & D funding investment (Ab3)
		Number of computers used per every hundred people (Ab4)
		Every 100 enterprises have websites (Ab5)
Digital Effectiveness (B)	Technical and Economic Benefits (Ba)	The total output value of the computer, communications, and other electronic equipment manufacturing industry (Ba1)
		Software business revenue (Ba2)
		Income from the high-tech industries (Ba3)
		Telecom business volume (Ba4)
		Number of patent applications (Ba5)
		E-commerce sales volume (Ba6)
	Ecological Benefit (Bb)	Unit amount of energy consumption (Bb1)
		Investment in industrial pollution control was completed (Bb2)

), we considered two dimensions of input and output for selection. For input, drawing on Luo’s study [90], 8 indicators were selected for measurement from two perspectives: talent investment and digital infrastructure investment (Aa1–Ab5). For output, drawing on Huang’s study [91], 8 indicators were selected for measurement from two perspectives: technical and economic benefits and ecological benefit (Ba1–Bb2), and we used the entropy method for the indicator score (Figure 4). Second, the gray correlation method measured the integration degree of the digital economy and manufacturing industry in different regions. Finally, the degree of integration was multiplied by the level of regional digitalization to obtain the indicator value of the provincial manufacturing digitalization level, and the results are shown with Cartesian heat maps in Figure 5. Darker blue indicates lower digitalization of manufacturing and darker red indicates higher digitalization of manufacturing. We can find that Beijing, Shanghai, Guangdong, and Tianjin are at the top, while Xinjiang, Gansu, Yunnan, Shandong, and Inner Mongolia are at the bottom.

(3) Mediating variable: technological innovation of manufacturing (SIT). Technological innovation is a complex economic input and output process, which can be measured from two perspectives: input and output. In this paper, drawing on Li’s study [92], 3 indicators were selected from each perspective, and the entropy value method was applied to measure the regional manufacturing technology innovation level. The indicators are shown in

Table 2. Technological innovation indicators system.

Type	Indicator
Innovation Input (A)	Number of R & D active enterprises (Aa)
	R & D funds for manufacturing enterprises above the scale (Ab)
	Number of R & D projects of manufacturing enterprises above the scale (Ac)
Innovation Output (B)	Number of effective invention patents per unit of R&D expenditure (Ba)
	Percentage of new product sales revenue (Bb)
	High-tech industry production and operation (Bc)

, the scores for each indicator are shown in Figure 6, and the results of technological innovation level of manufacturing are shown with Cartesian heat maps in Figure 7. Darker blue indicates lower technological innovation of manufacturing and darker red indicates higher technological innovation of manufacturing. We can find that Beijing, Guangdong, Shanghai, Jiangsu, Zhejiang, and Tianjin are at the top, while Xinjiang, Gansu, Qinghai, Guangxi, and Jilin are at the bottom.

(4) Control variables (Controls): By analyzing previous research results, this paper selected 4 variables to control for the effects on carbon productivity in manufacturing, controlled for sample year fixed effects, and all variables are described in

Table 3. Definition and description of variables.

Variable		Symbol	Variable Description
Explained Variable	Manufacturing Carbon Productivity	Cp	Manufacturing industrial output/manufacturing carbon emissions
Core Explanatory Variables	Digitalization of Manufacturing	Dig	Digital integration degree of manufacturing industry and regional digital level multiplier
Mediating Variable	Technological Innovation in Manufacturing	STI	A combined score of the entropy method for six measures
Control Variables	Environmental Regulation	Envir	Main business income/total energy consumption of above-scale manufacturing enterprises
	Endowment Structure	Endow	Total assets of manufacturing enterprises above the scale/number of all employees
	Foreign Investment	FDI	Foreign-invested and Hong-Kong-, Macao-, and Taiwan-invested manufacturing enterprises’ main business income/above-scale manufacturing enterprises’ main business income
	firm scale	Scal	The main business income of manufacturing enterprises above the scale/number of enterprise units

.

Environmental Regulation (Envir): The cost constraint theory argues that environmental regulation imposes additional costs on firms, leading to increased burdens and reduced competitiveness [93]. But there are scholars who disagree; for example, Porter thinks that in the short term, environmental regulations will cause higher costs for companies. Yin [94] points out that the relationship between the intensity of environmental regulations and green total factor productivity in manufacturing is consistent with a “U”-shaped relationship. Zha [95] finds that environmental regulations bring positive effects for the economic performance of China’s industrial carbon emissions. Drawing on these studies, we considered the effect of environmental regulation on carbon productivity in manufacturing, choosing the ratio of main business income and total energy consumption of above-scale manufacturing enterprises as the measure of environmental regulation.

Endowment Structure (Endow): Capital and labor, as the most basic elements in the inputs of production activities in each industry, can influence the changes of the industry’s industrial structure, reflecting whether the industry is capital-intensive or labor-intensive [96]. The climbing capital–labor ratio mainly relies on the crude industrial scale expansion to achieve. The trend of heavy industry is obvious in recent years, which raises the capital intensity and leads to the deterioration of environmental quality, so the endowment structure will have a negative impact on carbon emission efficiency [97,98]. However, some scholars have different opinions; Wang [99] finds that the capital–labor ratio has a significant positive relationship with environmental total factor productivity. This is due to the high level of technology in capital-intensive firms, where technological advances offset their negative impact on carbon efficiency. Drawing on these studies, we considered the effect of endowment structure on carbon productivity in manufacturing, choosing the ratio of total assets of manufacturing enterprises above the scale and number of all employees as the measure of endowment structure.

Foreign Investment (FDI): Foreign-invested enterprises are an indispensable driving force in the development process of China’s manufacturing industry. There are different conclusions about the effect of foreign investment on carbon productivity in manufacturing. Hoffman’s [100] Granger causality test of foreign investment and CO₂ emissions shows that the inflow of foreign firms leads to an increase in carbon emissions in middle-income countries. Perkins’s [101] study finds that multinational firms increase competition in the market, which can reduce the profit margin of local firms. As a result, the ability or willingness of local firms to invest in advanced equipment is weakened, which is detrimental to carbon reduction. However, some scholars have different conclusions. He [102] points out that foreign investment has led to an increase in the absolute amount of industrial carbon emissions in China, but this does not necessarily indicate that foreign investment will inhibit the increase in carbon productivity. Therefore, we considered the effect of foreign investment on carbon productivity in manufacturing, choosing the ratio of foreign-invested and Hong-Kong-, Macao-, and Taiwan-invested manufacturing enterprises’ main business income and above-scale manufacturing enterprises’ main business income as the measure of foreign investment.

Firm Scale (Scal): Large enterprises have comparative advantages of economies of scale, risk sharing, and financing channels, which guarantee continuous R&D investment [103]. Most studies show that there is a positive relationship between firm scale and productivity. Cong [104] pointed out that the larger the company, the more resource advantage it has, which plays an important role in improving carbon productivity. However, some scholars have a different conclusion. Mansfield [105] finds that firm scale expansion inhibits firm technological innovation. We also considered the effect of firm scale on carbon productivity in manufacturing, choosing the ratio of the main business income of manufacturing enterprises above the scale and number of enterprise units as the measure of firm scale.

4.4. Sample Selection and Data Source

Combined with data availability, the study interval was determined to be 2013–2020. The data were mainly sourced from the China Statistical Yearbook, China Industrial Statistical Yearbook, China Energy Statistical Yearbook, China Science and Technology Statistical Yearbook, China Software and Information Service Industry Development Report, official websites of the National Bureau of Statistics, National Data and Ministry of Industry and Information Technology, and the statistical yearbooks of each region, and the data were standardized. Because of the more severe data deficiency in Tibet, Tibet was excluded from the measurement, while Hong Kong, Macao, and Taiwan were also not included.

Table 4. Descriptive statistics.

Variables	N	Mean	sd	Min	Max
Cp	240	0.976	0.671	0.163	3.678
SIT	240	0.285	0.146	0.0644	0.710
Dig	240	0.110	0.0572	0.0238	0.373
Envir	240	1.847	1.146	0.216	4.835
Endow	240	5.159	0.610	3.910	6.606
FDI	240	0.180	0.147	0.0112	0.618
Scal	240	3.374	1.014	1.536	7.504

presents the descriptive statistics of variables. As can be seen, the standard deviation of Cp is 0.671, which is large compared with other significant variables, indicating that the data are highly volatile, and the Cp varies from region to region, with the mean, minimum, and maximum values being 0.976, 0.163, and 3.678, respectively. The standard deviation of Dig is 0.146, and the mean, minimum, and maximum values are 0.110, 0.0238, and 0.373, respectively, indicating that there is also a significant difference in Dig among regions and the investment in digital upgrading of manufacturing in most regions still needs to be strengthened. The minimum and maximum values of SIT are 0.0644 and 0.710, respectively, indicating that there are also some differences in SIT across regions.

5. Results and Discussion

5.1. Correlation Analysis

A Pearson correlation coefficient test was conducted to verify the relationship between the variables, and the results are shown in

Table 5. Variable correlation coefficient matrix.

Variable	Cp	STI	Dig	Envir	Endow	FDI	Scal
Cp	1
SIT	0.785 ***	1
Dig	0.792 ***	0.851 ***	1
Envir	0.726 ***	0.508 ***	0.436 ***	1
Endow	−0.272 ***	−0.0700	0.0470	−0.714 ***	1
FDI	0.700 ***	0.577 ***	0.637 ***	0.562 ***	−0.263 ***	1
Scal	0.0670	0.0270	0.272 ***	−0.234 ***	0.594 ***	0.205 ***	1

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

. The relationship between Dig and Cp was significantly positive at the 1% level, initially indicating that digitalization in manufacturing can positively impact carbon productivity in manufacturing, providing an opportunity to study its intrinsic impact mechanism. The relationship between SIT and Cp was significantly positive at the 1% level, indicating that technological innovation SIT and Cp are significantly positive at the 1% level, indicating that technological innovation helps to improve carbon productivity in manufacturing. The above analysis initially confirms hypothesis H1 and hypothesis H2.

As the panel data contain both cross-sectional and time-series data, the model may have problems with highly correlated explanatory variables and unstable sequences. To avoid influencing the model results or causing pseudo-regression, this study used the variance inflation factor method and the LLC test to test the multicollinearity and smoothness of the model, respectively.

As shown in

Table 6. Results of unit root test and multicollinearity test for variable benchmark regression models.

Variable	Unit Root Test		Multiple Collinearity Test
	The Adjusted T-Statistic	p	The Coefficient of Variance Inflation
Cp	−3.578	0.006 ***
SIT	−3.994	0.001 ***	4.858
Dig	−4.803	0.000 ***	5.115
Envir	−5.309	0.000 ***	4.270
Endow	−2.974	0.021 **	4.828
Fdi	−2.573	0.066 **	2.355
Scal	−4.514	0.000 ***	2.463

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

, the test results show that the maximum VIF is 5.115, which is less than 10, so there is no need to worry about multicollinearity. All variables under the LLC test reject the original hypothesis of the “existence of unit root” at the 5% significance level, so the data in this study were considered stable.

5.2. Regression Analysis

This paper developed the analysis of econometric models (5) to (8) using mixed OLS regressions, and the results are presented in

Table 7. Results of regression analysis.

Variable	Model (5) Cp	Model (6) Cp	Model (7) STI	Model (8) Cp
Dig	5.434 ***		1.906 ***	3.884 ***
	(7.98)		(18.78)	(3.98)
SIT		1.889 ***		0.813 **
		(7.34)		(2.42)
Envir	0.318 ***	0.311 ***	0.034 ***	0.290 ***
	(11.40)	(10.22)	(4.23)	(12.10)
Endow	0.196 ***	0.178 ***	0.070 ***	0.139 ***
	(3.47)	(3.17)	(4.25)	(2.84)
FDI	0.733 ***	0.901 ***	0.090 *	0.660 ***
	(3.80)	(5.55)	(1.65)	(3.68)
Scal	−0.047	0.029	−0.044 ***	−0.011
	(−1.25)	(0.73)	(−6.65)	(−0.35)
Constant	−1.194 ***	−1.312 ***	−0.216 ***	−1.019 ***
	(−4.80)	(−4.33)	(−2.75)	(−4.40)
Year	Yes	Yes	Yes	Yes
N	240	240	240	240
R2	0.817	0.802	0.790	0.823
F	163.4	160.9	268.9	159.5

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

. Column 1 shows the results of Dig’s regression analysis on Cp. The regression value of digitization of manufacturing and carbon productivity is 5.434. The relationship between the two is significant at the 1% level, indicating that the higher digitization of the manufacturing, the higher carbon productivity, i.e., the better the carbon reduction effect, and the hypothesis H1 holds. Accelerating the digital transformation and upgrading manufacturing positively impact increasing its carbon productivity.

The regression coefficient of Dig on Cp in model (5) is 5.434 and is highly significant at the 1% significance level; the regression result of model (6) shows that the regression value of SIT on Cp is 1.889 and the relationship between the two is significant at the 1% level, indicating that technological innovation in the manufacturing can effectively enhance its carbon productivity and achieve carbon emission reduction, and hypothesis H2 holds, which is generally consistent with the overall findings of the existing literature. The regression coefficients of Dig and SIT in model (7) are significant and greater than zero at the 1% level, indicating that digitalization in manufacturing can contribute to technological innovation in manufacturing. In model (8), we put SIT into the basic regression model, and compared with model (5), the regression coefficient of Dig is reduced to 3.884 at the 1% significance level. In comparison, the regression coefficient of SIT is significant at the 5% level, indicating a partial mediating effect of technological innovation between digitization of manufacturing and carbon productivity, and hypothesis H3 holds.

5.3. Heterogeneity Analysis

Considering the heterogeneity of the socioeconomic development environment of different regions, this paper further investigated the relationship between digitalization of manufacturing, technological innovation, and carbon productivity in the eastern, central, and western regions.

Table 8. Results of the regional model regression analysis.

Variable	Eastern			Central			Western
	Model (5) Cp	Model (7) STI	Model (8) Cp	Model (5) Cp	Model (7) STI	Model (8) Cp	Model (5) Cp	Model (7) STI	Model (8) Cp
Dig	9.658 ***	1.612 ***	7.809 ***	3.916 ***	2.531 ***	−2.099	1.583 **	1.741 ***	4.743 ***
	(10.54)	(7.79)	(6.50)	(2.71)	(8.20)	(−0.89)	(2.20)	(8.86)	(4.34)
SIT			1.147 ***			2.377 ***			−1.815 ***
			(3.15)			(3.69)			(−3.86)
Envir	0.317 ***	0.059 ***	0.249 ***	0.108	−0.036 ***	0.193 **	0.476 ***	0.048 **	0.563 ***
	(9.08)	(5.53)	(7.87)	(1.30)	(−2.78)	(2.57)	(5.21)	(2.19)	(6.93)
Endow	0.411 ***	0.173 ***	0.212 *	−0.207	0.064 ***	−0.358 ***	0.172 **	0.047 *	0.258 ***
	(3.71)	(4.20)	(1.96)	(−1.32)	(2.77)	(−2.70)	(2.00)	(1.92)	(3.02)
FDI	−0.511 *	−0.045	−0.459 *	2.178 *	−0.208	2.674 **	0.338	−0.019	0.303
	(−1.84)	(−0.56)	(−1.80)	(1.73)	(−0.95)	(2.70)	(0.80)	(−0.15)	(0.73)
Scal	−0.244 ***	−0.068 ***	−0.166 ***	−0.011	−0.071 ***	0.158 **	−0.036	−0.033 ***	−0.096 ***
	(−4.26)	(−4.98)	(−3.01)	(−0.15)	(−7.77)	(2.25)	(−1.14)	(−4.14)	(−2.93)
Constant	−1.684 ***	−0.603 ***	−0.993 **	0.981 *	0.029	0.913 *	−0.856 *	−0.128	−1.089 **
	(−4.01)	(−3.33)	(−2.37)	(1.79)	(0.30)	(1.84)	(−1.91)	(−1.05)	(−2.56)
Year	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	80	80	80	48	48	48	88	88	88
R2	0.875	0.763	0.885	0.731	0.893	0.784	0.712	0.826	0.760
F	121.5	83.75	114.3	113.4	77.93	69.11	94.35	149.5	54.57

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

shows the results of the heterogeneity analysis. It can be seen that the regression coefficients of model (5) Dig on Cp are all highly significant at the 1% significance level for each of the four geographically differentiated regions. The regression coefficients of model (7) Dig and SIT are also significant at the 1% level, which indicates that increased digitization in each region will lead to an improvement in both carbon productivity and technological innovation in that region. This also basically proves that hypotheses H1 and H2 hold at the heterogeneity. After adding model (8), the regression coefficients of digitization of manufacturing, technological innovation, and carbon productivity in the eastern and western regions are all significant at the 1% level, indicating that there is a partial mediating effect of technological innovation between digitalization of manufacturing and carbon productivity in these two regions. In contrast, the regression coefficients of manufacturing digitization and carbon productivity in the central regions are insignificant.

5.4. Robustness Analysis

(1) Controlling for endogeneity: Considering the possible time lag effect of digitization of the manufacturing and technological innovation on carbon productivity, the explanatory variables, mediating variables, and control variables were treated with a lag of 1 period to mitigate reverse causality bias. The findings are consistent with the previous discussion, further validating the reliability of the findings of this paper, and the results are shown in

Table 9. Results of the robustness analysis.

Variables	Model (5) Cp	Model (6) Cp	Model (7) STI	Model (8) Cp
Dig	6.627 ***		1.916 ***	5.128 ***
	(8.18)		(17.23)	(4.32)
STI		2.164***		0.782 *
		(7.15)		(1.95)
Envir	0.278 ***	0.275 ***	0.031 ***	0.254 ***
	(9.90)	(8.14)	(3.81)	(10.59)
Endow	0.132 **	0.108 *	0.068 ***	0.079
	(2.24)	(1.73)	(4.19)	(1.52)
FDI	0.699 ***	0.984 ***	0.091	0.627 ***
	(3.34)	(5.52)	(1.62)	(3.17)
Scal	−0.054	0.028	−0.040 ***	−0.022
	(−1.35)	(0.69)	(−5.94)	(−0.68)
Constant	−0.801 ***	−0.894 **	−0.216 ***	−0.632 **
	(−3.04)	(−2.53)	(−2.78)	(−2.51)
Year	Yes	Yes	Yes	Yes
N	210	210	210	210
R2	0.819	0.793	0.788	0.824
F	155.9	147.0	221.6	152.7

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

.

(2) Re-testing of the mediating effect: Because of the weakness of the stepwise regression test, we attempted the Bootstrap test, which determines the existence of mediating effects by the critical value of the confidence level. Setting the number of replicate samples at 5000, the results are shown in

Table 10. Results of the Bootstrap test for the mediation effect.

Impact	The Direct Effect of Dig on Cp
parameter	Effect	SE	LLCI	ULCI
figure	0.4495	0.0707	0.3102	0.5888
impact	The indirect effect of Dig on Cp
parameter	Effect	BootSE	BootLLCI	BootULCI
figure	0.3429	0.0858	0.1705	0.5093

. Both the indirect effect interval [0.1705, 0.5093] and the direct effect interval [0.3102, 0.5888] for the level of digitalization in manufacturing do not contain zero, further indicating that the level of technological innovation in manufacturing plays a partially mediating role in the relationship between the level of digitalization and carbon productivity, and hypothesis H3 is again verified.

5.5. Discussion

We find that manufacturing in China has had a digital divide for a considerable amount of time. Each province’s manufacturing has a digital development flaw. “High in the east and low in the west, high in the south and low in the north” continues to be the development level. However, when viewed in conjunction with the horizontal comparison of years, it is clear that each province’s manufacturing is becoming more digitally integrated. Jiangsu, Zhejiang, Guangdong, Beijing, and Shanghai have been leading the way and are well ahead of other provinces. We hypothesize that the significant differences in manufacturing digitalization in each province may be caused by differences in the technology and size of manufacturing firms, as well as the poor development of digital infrastructure, the low level of investment, and the failure of the development of the digital economy to effectively drive the improvement of manufacturing digitalization. As a result, to raise manufacturing’s digitalization, it is essential to identify the digital economy’s current weak point in this province and strongly push for its growth to realize the full digitalization of manufacturing.

Additionally, it can be seen that the development trend of these two items exhibits the state of “high in the east and low in the west, high in the south and low in the north”, and that the gap between the development level of these two items in the eastern coastal provinces and other provinces is gradually closing. This is according to digitalization and technological innovation of manufacturing in each province as shown in the column chart.

Finally, it was discovered that there is a mediating influence of the digitalization of manufacturing on carbon productivity with technological innovation as a mediator by employing manufacturing technology innovation as a mediating variable in the empirical investigation. The digitization of manufacturing also significantly raises its technological innovation, indicating that manufacturing invests more in digital technology and spends more on R&D during the production and operation process. This process also has the effect of accelerating the high-quality development of manufacturing by enhancing technological innovation capability and raising manufacturing’s carbon productivity.

6. Conclusions and Implications

6.1. Conclusions

(1)

This paper identified the relationship between digitalization and carbon productivity in manufacturing. Unlike previous studies on carbon emissions, this paper has been concerned more with the effects of digitalization on the production efficiency of the manufacturing as a whole and entire districts, especially on the effects of carbon emission reduction brought by the direct improvement of carbon productivity. Therefore, we shifted the research object from single carbon emission to carbon productivity. On the one hand, this view is consistent with the research needs of digitalization on the production efficiency improvement of manufacturing. On the other hand, the view is also consistent with the green development demand for “carbon peaking and carbon neutrality”. It is one of the important research innovation points of this paper. Compared with previous studies that simply study carbon emissions [37,38,40,41,42,43,44,45,46], the results obtained in this paper are similar. We combined economic development and environmental protection effectively, and used carbon productivity to precisely measure the degree of decoupling between economic development and carbon emission. We also revealed the role of digitalization in improving carbon productivity for each link of manufacturing.

(2)

This paper is concerned more with the effects of technological innovation on the relationship between digitalization and carbon productivity. In contrast to previous studies that select innovation indicators, involving various aspects such as green innovation, technological innovation, and management innovation [40,44,53,54,55], we identified technological innovation indicators closely related to digitalization and examined the effects on carbon productivity. We have demonstrated the mediating role of technological innovation between digitalization and carbon productivity in the previous. The results indicated that technological innovation driven by digitalization development is a critical factor in carbon productivity improvement. Technological innovation is a vital link between digitalization and carbon productivity, and also a critical point to productivity improvement in manufacturing.

(3)

This paper took into account multiple synergistic aspects and considered overall integrated development rather than just concentrating on one specific issue. Another important conclusion of this paper is the necessity of synergy technological innovation and digitalization development as approaches to the issue of carbon reduction. In comparison to previous studies on carbon reduction from a single perspective [42,45,47,48,49,50,51,52], this perspective can more effectively integrate resource advantages, increase industrial value addition, enhance production efficiency, and optimize the production chain, all of which are significant for improving carbon productivity in manufacturing.

This paper shows a comprehensive analysis of digitalization in China’s manufacturing, as well as the adoption of scientific empirical approaches to study the driving effect of digitalization in manufacturing on improving carbon productivity, and the differences in the driving effect in different regions. In addition, this paper also innovatively analyzes the conduction path of manufacturing digitalization to improve carbon productivity by driving the development of technological innovation and exploring the new development direction for improving carbon productivity in China’s manufacturing.

The main contributions of this paper are summed up as follows: first, this paper creatively combines the two research fields of digitalization of manufacturing and carbon emissions of manufacturing, filling the gap between digitalization and the enhancement of carbon productivity. We find a correlation between the digitalization of manufacturing and carbon productivity, and explore the conduction path between the two. Second, this paper studies the impact effect of manufacturing digitalization from the perspective of green and sustainable development, revealing the direct impact of manufacturing digitalization on carbon productivity from a theoretical perspective and verifying it through empirical methods, as well as introducing technological innovation to study the mediating effect. Third, this paper provides insights into improving carbon productivity and has explored a positive conduction path to enhance carbon productivity by promoting the digitalization of manufacturing.

6.2. Implications

This paper proposes several management implications based on the above research and conclusions. First, under the goal of “carbon peaking and carbon neutrality”, the digitalization upgrading of manufacturing should be accelerated to play a role in promoting the improvement of carbon productivity. The digital economy has the dual efficacy of promoting economic development and enhancing carbon productivity. In contrast, China’s digitalization of manufacturing is still low. There is a big difference between the eastern and the western areas of the country, which could be more conducive to the coordinated development of China’s economy. Based on this, we should further increase the investment in the digital transformation of manufacturing, especially the investment in digital infrastructure construction of manufacturing in central and western China, to alleviate the problem of uncoordinated regional development in China and accelerate the pace of transformation of China’s manufacturing to a high-quality and sustainable development mode.

Second, technological innovation in manufacturing should be enhanced to play a mediating effect of improving carbon productivity in manufacturing. Achieving “carbon peaking and carbon neutrality” is arduous and requires strong support from science and technology. The increase in technology absorption capacity and R&D investment can improve the digitalization level of manufacturing. Technological innovation is the first driving force to lead development, technological innovation is a vital intermediary transmission mechanism between the digital transformation of manufacturing and improving carbon productivity, and manufacturing in each region should pay attention to the improvement of their technological innovation level to promote digital transformation and innovation capacity, to strengthen the research and development of energy technology and energy-saving technology to break through the essential core technology of pollution reduction and carbon reduction In this regard, we will play the role of the transmission mechanism of technological innovation and promote the positive impact of technological innovation on enhancing carbon productivity.

Third, the carbon neutrality policy guarantee mechanism should be improved and a sound enterprise carbon emission trading market established. Since the Kyoto Protocol in 1997, nearly 20 countries and regions have enacted laws to address climate change [59]. The current legal system in China has more energy-related provisions, but fewer provisions directly related to carbon reduction. There is still a gap in the law of “ carbon peaking and carbon neutrality “. It is suggested that on the basis of the existing policies, laws to cope with climate change and reduce carbon emissions should be formulated and promulgated as soon as possible, so as to provide a strong legal guarantee for achieving “carbon peaking and carbon neutrality”. In addition, the current carbon trading market has problems such as small scale, low carbon price, and poor activity, and the scope and strength of carbon emissions trading policy implementation needs to be expanded and strengthened. In terms of the carbon emissions trading system, it is necessary to further refine the carbon quota allocation mechanism, carbon verification rules, and formulate strong disciplinary measures to promote the formation of a market-oriented carbon price formation mechanism.

Fourth, the construction of new infrastructure should be accelerated, the deep integration and development of the real economy and digital technology promoted, and the transformation and upgrading of traditional manufacturing empowered. While integrating digital technology into traditional production and improving production efficiency, we should also focus on the research, development, and application of green and low-carbon technologies. We should give full play to the driving role of low-carbon industries, thus promoting the overall transformation of the industry to an intelligent and low-carbon state, and providing a viable path to achieve “carbon peaking and carbon neutrality”.

Finally, based on regional development differences, heterogeneous governance strategies should be implemented. Based on the differences in endowments of different regions and the differences in the impact of the digital economy on carbon emissions, the pace of digital economy development in each region should be adjusted, the industry barriers and geographical restrictions of new models and new business models broken, and the differences and synergy of digital economy governance in each region enhanced.