The Impact of Manufacturing Intelligence on Green Development Efficiency: A Study Based on Chinese Data

Abstract

This study used provincial panel data from 2011 to 2020 to empirically analyze the impact of improvements in the manufacturing intelligence level on the efficiency of green development. The results show the following: improvements in manufacturing intelligence significantly increased the efficiency of green development. Following the “Guidelines for the Construction of the National Intelligent Manufacturing Standard System” proposed by China as a quasi-natural experiment, the double-difference method was used to prove the following: the promotional effect is greater in eastern regions after the policy was implemented. An improvement in manufacturing intelligence increases green development efficiency by improving the efficiency of technological innovation and energy use. The impact of the manufacturing intelligence level on green development efficiency shows a threshold effect. The promotion effect is more pronounced in regions where the level of technological innovation and the strength of the government’s role cross this threshold. Therefore, the government should vigorously promote the intelligent transformation of the manufacturing industry and improve the efficiency of China’s green development.

1. Introduction

After more than 40 years of reform and opening up, China has become the world’s largest manufacturing country. It is large, but not strong. There are still problems, such as low-value-added products, high environmental pollution and insufficient energy utilization [1,2]. At present, the industrialization process is the unprecedented change of the century. A new industrial revolution is emerging, and a new round of scientific and technological revolution is ascendant, empowering China’s green development. The “14th Five-Year Plan for the Development of Intelligent Manufacturing” [3] proposes that China should take the in-depth integration of new-generation information technology and advanced manufacturing technology as the main line and promote the digital transformation and intelligent upgrading of the manufacturing industry, which provides new ideas for green development.

Manufacturing intelligence is the manufacturing practice of using advanced information and manufacturing technologies to increase the flexibility of manufacturing processes with the aim of optimizing the production and trading of products. It is able to respond to dynamic changes in the global marketplace [4,5]. In terms of practical progress, intelligent manufacturing in China has shown good development. Data on manufacturing intelligence for 2022 released by China’s Ministry of Industry and Information Technology show that the number of smart manufacturing pilots in China grew from 46 in 2015 to 342 in 2021.

Big data and smart technologies can be used to improve the designs, production, management, operations and services of the manufacturing industry. This is important for promoting the win–win situation of improved economic and environmental performance. Specifically, in design, intelligent manufacturing promotes networking and flexibility in industrial R&D, which helps to identify market demand. In production, it promotes the automation and intelligence of manufacturing processes, reduces the problem of overproduction, and also realises the green transformation of production methods, energy saving and consumption reduction. Lv et al. (2019) [6] examined the intelligent manufacturing platform of Haier Group. They found that intelligent production could achieve a 60% reduction in the material consumption of welding problems and a 60% increase in production capacity. In management, manufacturing intelligence promotes the digitization and visualization of the management process, and can also alleviate information asymmetry. In operations, it promotes innovation in marketing and the expansion of marketing channels in the manufacturing industry. In services, manufacturing intelligence is conducive to innovation, the improvement of value-added service models and the improvement of resource allocation efficiency. However, the overall impact of intelligent development in the manufacturing industry on the efficiency of green development in China and the underlying mechanisms require further study. This paper intends to explore this issue. This has important theoretical value and practical significance.

With the popularization of the concept of green development, the study of green development has received increasingly widespread attention from academics. The research literature on the influencing factors of green development efficiency [7,8], the efficiency differences between sub-sectors [9,10] and their spatial and temporal differences [11,12] is also increasing. With the development of digital technologies such as the Internet and artificial intelligence, scholars have begun to explore the efficiency of green development from an intelligence perspective. The main studies can be summarized as three aspects. The first aspect is the impact of artificial intelligence on economic development efficiency. Some scholars focus on the negative effects of artificial intelligence. Huang et al. (2019) [13], Gasteiger et al. (2017) [14] and Acemoglu et al. (2018) [15] proposed that human–machine competition will drive down wage rates, and the mismatch between workers’ own skills and automation skills will hinder total factor productivity. Therefore, the size of the AI sector needs to be controlled. However, some scholars have affirmed AI’s positive role in enhancing economic development efficiency. Brynjolfsson et al. (2000) [16], Lin et al. (2020) [17] and Mao et al. (2022) [18] argue that AI enables the upgrading of hardware and software to promote technological progress and energy efficiency, which in turn drives economic development efficiency. Graetz et al. (2018) [19], Chen et al. (2019) [20] and Yang et al. (2020) [21] found that automation and industrial robots, which are emerging production methods, can drive improvements in total factor productivity and promote high-quality economic development. Lv et al. (2020) [22] demonstrated that intelligence can enhance the competitiveness of enterprises, thus boosting productivity and further promoting China’s participation in global value chains. The second impact is artificial intelligence’s influence on green development. Shi (2020) [23] argued that AI has a positive impact on the scientific management of energy consumption, promoting green development, energy conservation and emission reduction. Waltersmann et al. (2021) [24] proposed that the application of AI can improve the efficiency of resource applications, which lead to improvements in the ecological performance of enterprises. Han et al. (2022) [25] believed that digital intelligence can achieve the full range of empowerment for green development through innovation. Third, AI improves the transmission path to green development. AI can significantly improve the efficiency of green development based on the transmission paths of human capital effect [26], structural effect [27,28], innovation effect [29] and cost effect [30]. Thus, a positive externality can be obtained for green development.

In summary, scholars have empirically analyzed the efficiency of green development at the national, regional and industry levels. They have conducted in depth studies on the effects of smart technology. These studies reveal that traditional industrial intelligence impacts the efficiency of economic development. In the intelligent process of traditional industries, although factors such as scale expansion, competition and mismatches of workers’ abilities can trigger negative effects, intelligent technology can improve economic development by enhancing productivity and promoting technological progress. Scholars have also begun to focus on the factor analysis of green development efficiency in the context of intelligence. AI is shown to be conducive to improving resource utilization efficiency and reducing energy consumption, thus promoting green development. However, few articles have studied the impact on green development efficiency from the perspective of manufacturing intelligence. This study aimed to explore the mechanisms and pathways by which manufacturing intelligence impacts green development efficiency. The aim was to answer the following questions: Does increased manufacturing intelligence promote green development efficiency? Existing studies have concluded that an increase in technological innovation and energy efficiency will increase green development efficiency. However, does increased manufacturing intelligence affect technological innovation and energy efficiency? What is the transmission mechanism for this? Under which circumstances will the impact be the greatest? This study used a fixed-effects model to explore the effect of manufacturing intelligence on green development efficiency. The transmission mechanism was explored using a mediating model based on the double-difference method, and the non-linear effects were explored using a threshold model.

This paper is organized as follows: the Introduction section briefly introduces the background of this study as well as a literature review. Section 2 analyzes the mechanism of the impact of manufacturing intelligence on green development efficiency. Section 3 measures the level of manufacturing intelligence and green development efficiency. Section 4 introduces the empirical model and variable selection. Section 5 describes and explains all the empirical results. Section 6 discusses the research findings, policy recommendations and the limitations of this study.

2. Mechanistic Analysis and Research Hypothesis

Since intelligent technology has the characteristics of substitution, synergy and penetration, the intelligent development of the manufacturing industry can inject a new impetus to increase the efficiency of green development. This paper will analyze the internal mechanism of the impact that improvements in the manufacturing intelligence level have on green development efficiency, focusing on three different aspects: direct impact, indirect impact and non-linear impact. Then, we put forward our research hypotheses.

2.1. Direct Impact

With the development of technologies such as big data, artificial intelligence and cloud computing, the degree of integration between intelligence and manufacturing in terms of elements, organization and production is deepening. This integration’s impact can be summarized into four aspects. First, intelligence can alleviate the problem of information asymmetry, improve the efficiency of resource allocation, and thus enhance the efficiency of green development. Production intelligence can be used to build a product life cycle system that includes information sharing, covering the whole process of product design, production and transportation. This can improve the accuracy of information, and Toorajipour et al. (2021) make a similar point [31]. It can also break down the barriers to information dissemination, leading to a smoother dissemination of information within and between enterprises and thus enabling enterprises to more appropriately allocate resources such as labor, capital and technology. Secondly, manufacturing intelligence helps enterprises to better identify the real and explicit needs of customers, and then carry out flexible production and personalized customization R&D, reducing the problem of overproduction. This is in line with Zhong et al. (2017) [32]. For example, the Saint Angelo Company realized the individualized and unique clothing customization needs of customers through its intelligent manufacturing system, which increased production efficiency by 50%, shortened the average production cycle from the original 15 days to 7 days, and increased the quality qualification rate to 99.6% (Data from Saint Angelo Company http://www.baoxiniao.com.cn/article/show/id/1596.html (accessed on 25 June 2022)). At the same time, based on intelligence, enterprises can exclusively design product recycling and reuse according to customer preferences to better realize resource recycling. Thirdly, some traditional production factors are replaced by smart devices. Driven by digital technologies such as big data and the Internet of Things, intelligent technologies have replaced some of the traditional factors of production, such as labor, capital and raw materials, and the dependence on traditional factors of production has decreased. This is consistent with Frank et al. (2019) [33]. Intelligent production is more synergistic and organized, and each production chain can maximize the value of each element, ultimately achieving an increased output with the same inputs and an improved production efficiency. As far as the labor factor is concerned, intelligence has a crowding-out effect on low-skilled jobs. Machines replace labor for mass production, automating complicated tasks and improving production efficiency. For example, Dongfang Electric Group achieved a 24 h, unmanned, continuous processing through intelligent technology and was able to autonomously provide raw materials within 40 s, finally achieving a 99% quality compliance rate and a 650% increase in per capita efficiency [34]. At the same time, the production intelligence leads to higher requirements for worker skills, and its push-back mechanism is conducive to the improvement in workers’ overall skills, thus leading to further improvements in production efficiency. Fourth, intelligence can automatically identify high-energy-consuming and high-pollution-production links, as confirmed by Nishant et al. (2020) [35]. Manufacturing intelligence can enable the analysis of information and the proposal of solutions in real time, which enhances enterprises’ production organization and substantially improves synergy. Therefore, this paper proposes research Hypothesis 1:

H1. 

Intelligent manufacturing development promotes green development efficiency.

2.2. Indirect Effects

During the process of manufacturing intelligence, the level of technological innovation and the efficiency of energy use can be promoted, which in turn affects the efficiency of green development.

The core of manufacturing intelligence is innovation. Manufacturing intelligence can cluster innovation elements and pressure firms to innovate technologically, as follows. First, intelligence gathers the innovation factors. The development of manufacturing intelligence means that enterprises need to invest more in R&D funds and equip corresponding high-skilled talents, which will help to obtain economic and talent elements that are required for enterprises to realize technological innovation. At the same time, the improvement in technological innovation is also conducive to the development of enterprise intelligence. The use of new equipment and technology provides the necessary technical support for enterprises to achieve a higher level of intelligence. There is a two-way link between intelligence and technological innovation. Second, manufacturing intelligence gives rise to new forms and models. The higher technical requirements for the product development sector prompt companies to improve their innovation abilities. For example, Foton Motor established an intelligent manufacturing system, thus shortening the product development cycle by approximately 30% and increasing production efficiency by 18% [36]. Due to the knowledge spillover effect of technological innovation, during the innovation process, enterprises can more quickly obtain market and resource information, which is conducive to mutual learning and imitation, and improves the adaptability of technology and production operation links. The process of technological innovation is accompanied by a reduction in environmental pollution, optimization of resource allocation and improvement in production level, and the efficiency of green development is enhanced. Therefore, this paper proposes research Hypothesis 2:

H2. 

Manufacturing intelligence can contribute to green development efficiency by increasing the level of technological innovation.

Manufacturing intelligence extends enterprises’ production possibilities, so that enterprises are no longer constrained by the scope of the economy, long-tail effect and economy of scale, as they are under the traditional production model. Accordingly, they do not excessively consume resources when improving enterprises’ production efficiency and promote the efficiency of green development through the energy-saving and emission reduction effect brought about by accelerating the structural adjustment of factors.

First of all, the intelligent process of manufacturing is the deep integration of manufacturing and digital technologies such as the Internet of Things, which is conducive to reducing energy consumption during the production and transportation process and improving energy use efficiency. Take 147 enterprises in Banan District of Chongqing City as an example. These enterprises have prompted a 12% increase in energy utilization and 87% increase in production efficiency through intelligent transformation [37]. Through intelligent technology, enterprises are able to select the most appropriate energy sources during the production process, increase the use of new green materials and reduce pollution. Intelligence combines smart detection and sensors to accurately identify under-utilized energy and effectively improve energy use efficiency by improving production processes and technologies. This can reduce undesired outputs in green development. Improved energy use efficiency means that less energy is consumed for the same level of output, and the reduction in energy consumption during enterprise production equates to a reduction in input; that is, production intelligence makes it possible to reduce the input required for the same level of output. Second, intelligence reduces the probability of wasting energy due to equipment failure. Equipment failure will lead to forced production suspension or increased defective products, and intelligent production overcomes the previous problems of manual troubleshooting and overhauling, which can consume a lot of energy and manpower, as well as the problem of low production efficiency. It automatically selects the equipment on the production line that is running well to continue the operation, greatly reducing the possibility of production stagnation, which, in turn, reduces the consumption of energy when production is stalled and maximizes the use of energy. For example, the Mustang Battery Company has increased energy usage by more than 10%, reduced defective production products by more than 10%, and reduced product scrap rates to less than 1.5 per 1000 by combining intelligence with line manufacturing [38]. Finally, the emergence of intelligent technology reduces the energy demands of enterprise production. In the context of intelligence, enterprise production uses more intelligent machines for production, which reduces the demand for traditional energy and the pollution caused by traditional energy production, which, in turn, improves the efficiency of green development. Enterprises are also focusing more on the development of intelligent technologies and the upgrading and iteration of machines and equipment, rather than limiting themselves to the study of energy inputs. With the development of intelligence, the manufacturing value chain is constantly being reconstructed, and existing production models are being upgraded and transformed in a more rational and advanced direction. A green production system has been established, and low-pollution and high-efficiency enterprises are gradually replacing high-pollution and high-energy-consuming enterprises. Industrial added value is improved, further enhancing the efficiency of green development. Therefore, this paper puts forward research Hypothesis 3:

H3. 

Manufacturing intelligence can contribute to green development efficiency by improving energy efficiency.

2.3. Non-Linear Effects

The extent to which the level of manufacturing intelligence affects the efficiency of green development is influenced by the capacity for technological innovation and the strength of the government’s role.

Generally speaking, when technological innovation in a particular region is at a high level, the region has a better innovation environment and greater technological spillover, which will also have a positive impact on the improvement in green development efficiency. With further improvements in the level of technological innovation, the operational efficiency of intelligent equipment and the utilization rate of the energy that is used will be further improved, and the role that intelligence plays in driving green development will be more obvious. The improvement in operational efficiency means that the utilization and recycling rates of the production factors will be improved, and the improvement in energy utilization means that the pollution and resource waste brought by production will be reduced, and the efficiency of the green development of enterprises will be improved. Second, after technological innovation reaches a certain level, the capital investment of intelligent enterprises is no longer limited to research and development; there is an increase in the available capacity to focus on green development. Although there is an increasing call for sustainable development in the market, only the R&D funds are used to improve the development efficiency, which is not enough to meet the market challenges. Only by using intelligent means to improve the efficiency of green development can the competitiveness of enterprises be truly enhanced so that intelligent enterprises can better use their advantages, and green development efficiency will be further improved. Third, the continuous improvement in technological innovation will promote the concentration of innovation elements, which will be conducive to promoting the improvement in green development efficiency. When technological innovation reaches a certain level, it will accelerate the overflow of innovation results and attract more high-skilled and high-quality talents. This agglomeration of highly skilled workers will enable the greater utilization of intelligent equipment and improve production efficiency. At the same time, high-quality workers will pay more attention to green production, forming a new growth point for enterprises and advocating a win–win situation for both economic and environmental benefits, thus promoting green development efficiency.

The government plays an important role in the improvement in manufacturing intelligence retarding the efficiency of green development. Firstly, through supporting policy measures, the government can provide sufficient funds for R&D investment, promote enterprises’ digital technological innovation [39], realize the deep integration of intelligence and the real economy, and have a positive impact on green development. The government can also increase fiscal spending through the national public finance “multiplier effect”. Regarding intelligent equipment, environmental protection equipment, etc., subsidies and other ways to guide the development of manufacturing intelligence will further promote the efficiency of green development. Secondly, the government can signal to the enterprises and market by encouraging the intelligent transformation and green development policies of the manufacturing industry, so that more social capital will flow into the production of intelligent and green production. Through the formulation of relevant policies and the establishment of a flexible governance signal of intelligent demonstration enterprises focusing on green development, the government will guide more enterprises to develop intelligently and benignly, improve traditional production models, focus on green production, and improve the efficiency of green development through intelligent technology.

Thirdly, since ecological environment is a public product, it requires the joint action of the “invisible hand” of the market and the “visible hand” of the government. Since public goods are non-competitive and non-exclusive, the “visible hand” of the government can reduce the market failure caused by the consumption characteristics of public goods and help to improve the efficiency of resource allocation through the government’s regulation of the market. Therefore, this paper proposes research Hypothesis 4:

H4. 

Technological innovation and the government’s role have a threshold effect on the impact of increased manufacturing intelligence on the efficiency of green development.

3. Measurement of Manufacturing Intelligence and Green Development Efficiency

3.1. Measurement of Manufacturing Intelligence

This paper drew on the ideas and methods of Sun et al. (2019) [40] and Liu et al. (2021) [41] to construct a manufacturing intelligence evaluation index system with three levels—basic inputs, production applications and market benefits—including three primary indicators and ten secondary indicators, as shown in

Table 1. Manufacturing intelligence index system.

Level 1 Indicators	Secondary Indicators	Metrics
Basic input layer	R&D funding input	High-tech manufacturing R&D expenditure
	Smart device input	Fixed asset investment in information transmission, software and information technology services
	Internet infrastructure investment	Fibre optic cable length/provincial area
	Staff input	Number of employees in high-tech manufacturing
		Number of people in information transmission, software and information technology services
Production application layer	Software development and service situation	Software business revenue
	The degree of industrialization of intelligent technology	High-tech manufacturing new product output value
		High-tech manufacturing valid invention patent
Market benefit layer	Smart device market profit	Total profit of high-tech manufacturing industry
	Smart device market efficiency	High-tech manufacturing main business income/employees
	Environmental improvement	Industrial solid waste generation/GDP
	Energy intensity	Energy consumption/GDP

.

In this paper, the entropy value method was used to assign values to the indicators. To ensure the comparability of data among regions, this paper obtained the weights of each indicator through national data, and then used the weight values to measure the data of 30 provinces, including cities and autonomous regions (the lack of relevant regional data from Tibet was serious and beyond research scope of this paper). The estimated value of the manufacturing intelligence level of each region was obtained from 2011 to 2020. The level of manufacturing intelligence in 2011 and 2020 is shown in Figure 1.

3.2. Measurement of Green Development Efficiency

Recent studies mostly used the directional distance function DDF model, the data envelopment analysis DEA model, and the extended SBM-DEA model. The undesirable outputs were considered and combined with the Malmquist–Luenberger index to measure green development efficiency at the national, provincial, and city levels. This paper uses the true value of the green total factor productivity to measure green development efficiency. Drawing on the models proposed by Chung et al. (1997) [42] and Tone (2001) [43], which include undesirable outputs, a super-efficient SBM model (super-SBM) was constructed. Compared to the non-super-efficient SBM model, the super-SBM compensates for the tendency for multiple decision units to ensure efficiency is achieved at the same time in the standard efficiency model, allowing for further judgement of the relative efficiency values when multiple efficient decision units exist. The expression for the super-efficient SBM model is as follows.

$$\mathsf{\rho }=\mathrm{m}\mathrm{i}\mathrm{n}\frac{\frac{1}{\mathrm{n}}\sum _{\mathrm{i}=1}^{\mathrm{n}}\frac{\stackrel{-}{\mathrm{x}}}{{\mathrm{x}}_{\mathrm{i}\mathrm{k}}}}{\frac{1}{{\mathrm{a}}_1+{\mathrm{a}}_2}(\sum _{\mathrm{l}=1}^{{\mathrm{a}}_1}\frac{\stackrel{-}{{\mathrm{y}}^{\mathrm{d}}}}{{\mathrm{y}}_{\mathrm{l}\mathrm{k}}^{\mathrm{d}}}+\sum _{\mathrm{s}=1}^{{\mathrm{a}}_2}\frac{\stackrel{-}{{\mathrm{y}}^{\mathrm{u}}}}{{\mathrm{y}}_{\mathrm{s}\mathrm{k}}^{\mathrm{u}}})}$$

(1)

$$\mathrm{s}.\mathrm{t}\left\{\begin{array}{c}\stackrel{-}{\mathrm{x}}\ge \sum _{\mathrm{j}=1,\mathrm{j}\ne \mathrm{k}}^{\mathrm{m}}{\mathrm{x}}_{\mathrm{i}\mathrm{j}}{\mathsf{\lambda }}_{\mathrm{j}};\stackrel{-}{{\mathrm{y}}^{\mathrm{d}}}\le \sum _{\mathrm{j}=1,\mathrm{j}\ne \mathrm{k}}^{\mathrm{m}}{\mathrm{y}}_{\mathrm{l}\mathrm{j}}^{\mathrm{d}}{\mathsf{\lambda }}_{\mathrm{j}};\stackrel{-}{{\mathrm{y}}^{\mathrm{u}}}\ge \sum _{\mathrm{j}=1,\mathrm{j}\ne \mathrm{k}}^{\mathrm{m}}{\mathrm{y}}_{\mathrm{s}\mathrm{j}}^{\mathrm{u}}{\mathsf{\lambda }}_{\mathrm{j}}\\ \stackrel{-}{\mathrm{x}}\ge {\mathrm{x}}_{\mathrm{k}};\stackrel{-}{{\mathrm{y}}^{\mathrm{d}}}\le {\mathrm{y}}_{\mathrm{k}}^{\mathrm{d}};\stackrel{-}{{\mathrm{y}}^{\mathrm{u}}}\le {\mathrm{y}}_{\mathrm{k}}^{\mathrm{u}}\\ {\mathsf{\lambda }}_{\mathrm{j}}\ge 0,\mathrm{i}=\mathrm{1,2},\dots ,\mathrm{n};\mathrm{j}=\mathrm{1,2},\dots ,\mathrm{m};\mathrm{l}=\mathrm{1,2},\dots ,{\mathrm{a}}_1;\mathrm{s}=\mathrm{1,2},\dots ,{\mathrm{a}}_2\end{array}\right.$$

(2)

where it is assumed that there are m decision units, each consisting of n inputs, ${\mathrm{a}}_1$ desired output and ${\mathrm{a}}_2$ non-desired output. x, ${\mathrm{y}}^{\mathrm{d}}$, ${\mathrm{y}}^{\mathrm{u}}$ are the elements in the input matrix, desired output matrix and non-desired output matrix, respectively, and $\mathsf{\rho }$ is green development efficiency values.

The input indicators set in this paper included labor, capital, and energy indicators. The labor input was measured by the number of employees at the end of the year in each province (city and autonomous region). The capital input was measured by the perpetual inventory method of Zhang (2004) [44]. The energy input was measured by the energy consumption at the end of the year in each province (city and autonomous region). Expected output in output indicators was measured by the total GDP of each region at constant 2000 prices. Undesirable output was constructed using the entropy method to construct a comprehensive indicator of pollution emissions, considering carbon emissions, industrial sulfur dioxide, wastewater and industrial solid waste emissions.

As the super-SBM model provides a static description of green development efficiency, the Malmquist index is needed to dynamically analyze the change in efficiency values over the intervening years. A GML index was constructed based on the global set of production technologies, which allows for effective inter-period comparisons and avoids the linear programming non-solution problem. Therefore, the SBM-GML method was used to measure green development efficiency. We referred to Qiu (2008) [45] to calculate the real value of green development efficiency, assuming that green development efficiency in 2011 was 1; then, green development efficiency in 2012 provides the base period value of the previous year 1 multiplied by the ML index of that year, etc. Green development efficiency in 2012 and 2020 is shown in Figure 2.

4. Materials and Methods

4.1. Model Setting

The following theoretical model was constructed:

$${\mathrm{Y}}_{\mathrm{i}\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{X}}_{\mathrm{i}\mathrm{t}}+{{\mathsf{\beta }}_2\mathrm{Z}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(3)

Y is the explained variable, indicating the efficiency of green development. X is the core explanatory variable of this paper, indicating the level of intelligence in the manufacturing industry. Z is a group of control variables. i represents the province (city and autonomous region), t represents the year, ${\mathsf{\beta }}_0$ is a constant item, ${\mathsf{\beta }}_1$ and ${\mathsf{\beta }}_2$ are the coefficients, and ${\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$ is the random error term.

In December 2015, China proposed the “Guidelines for the Construction of the National Intelligent Manufacturing Standard System” [46]. The document pointed out that it is necessary to promote the intelligentization of the production process, accelerate the integration and development of the new generation of information technology and manufacturing technology, accelerate the green transformation and upgrading of the manufacturing industry, and build an efficient, clean, low-carbon, circular green manufacturing system. Due to the exogenous nature of the time at which the policy was proposed, this paper refers to the practice of Li et al. (2019) [47]. The eastern region and the central and western regions were used as the experimental group and the control group, respectively, and the double-difference model was used to test the intelligent manufacturing effect policy. The model is as follows:

$${\mathrm{Y}}_{\mathrm{i}\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{D}\mathrm{i}\mathrm{d}}_{\mathrm{i}}\times {\mathrm{T}}_{\mathrm{t}}+{\mathsf{\alpha }}_2{\mathrm{Z}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\delta }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(4)

Did represents a regional dummy variable (1 in the eastern region and 0 in the central and western region). T is a time dummy variable, which is 0 before the “Guidelines for the Construction of National Intelligent Manufacturing Standard System” in 2015, and 1 in 2015 and afterward. ${\mathsf{\mu }}_{\mathrm{i}}$ indicates the regional fixed effect, ${\mathsf{\delta }}_{\mathrm{t}}$ expresses the year fixed effect, and ${\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$ is a random disturbance item. The definitions of other variables are the same as those in Formula (3).

In order to test the mechanism of the level of intelligence in the manufacturing industry on the efficiency of green development, a recursive model was added on the basis of Formula (4) to construct a mediation effect model:

$${\mathrm{M}}_{\mathrm{i}\mathrm{t}}={\mathsf{\lambda }}_0+{\mathsf{\lambda }}_1{\mathrm{D}\mathrm{i}\mathrm{d}}_{\mathrm{i}}\times {\mathrm{T}}_{\mathrm{t}}+{{\mathsf{\lambda }}_2\mathrm{Z}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\delta }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(5)

$${\mathrm{Y}}_{\mathrm{i}\mathrm{t}}={\mathsf{\phi }}_0+{\mathsf{\phi }}_1{\mathrm{D}\mathrm{i}\mathrm{d}}_{\mathrm{i}}\times {\mathrm{T}}_{\mathrm{t}}+{\mathsf{\phi }}_2{\mathrm{M}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\phi }}_3{\mathrm{Z}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\delta }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(6)

M is an intermediary variable: the level of technological innovation and energy efficiency, respectively.

In order to test the non-linear relationship between manufacturing intelligence and green development efficiency, the following threshold model was constructed:

$${\mathrm{Y}}_{\mathrm{i}\mathrm{t}}={\mathsf{\gamma }}_0+{\mathsf{\gamma }}_1{\mathrm{X}}_{\mathrm{i}\mathrm{t}}\times \mathrm{I}(\mathrm{H}\le {\mathrm{r}}_1)+{\mathsf{\gamma }}_2{\mathrm{X}}_{\mathrm{i}\mathrm{t}}\times \mathrm{I}(\mathrm{H}>{\mathrm{r}}_1)+{\mathsf{\gamma }}_3{\mathrm{Z}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}\mathrm{t}}$$

(7)

H represents the threshold variable; technological innovation and the government role are taken as the threshold variable. ${\mathrm{r}}_1$ is the threshold value. $\mathrm{I}(·)$ is a indicative function, and it takes 1 when the condition in brackets is satisfied and 0 otherwise.

4.2. Variable Description and Data Source

4.2.1. Variables Selection

Manufacturing intelligence level and green development efficiency have been measured in the early part.

In order to control for the influence of other factors on green total factor productivity, the control variables selected in this paper included the following: (1) the human capital level, measured by average years of education; (2) GDP per capita, measured by the per capita real GDP of each province (city and autonomous region); (3) openness to the outside, measured by the ratio of FDI to GDP in each region; (4) industrial structure, measured by the ratio of the added value of the tertiary industry to the added value of the secondary industry; (5) the telecommunications development level, measured by the long-distance optical cable mileage.

The intermediary and threshold variables are as follows: (1) energy efficiency, measured by the ratio of regional industrial added value to energy consumption. (2) Technological innovation. China’s technological innovation mainly includes invention patents, utility model patents and design patents. Among them, invention patents refer to new technical solutions proposed for products, methods or their improvements. Compared with utility model and design patents, invention patents have a higher technical level and are not easy to imitate. Referring to Yao et al. (2017) [48], this is measured by the ratio of the number of invention patents granted to the total amount of patents granted. (3) The government role, measured by the ratio of government fiscal expenditure to GDP of each province (city and autonomous region).

4.2.2. Data Sources

The data in this paper were obtained from statistical yearbooks for the period 2011–2020, which were published by China’s National Bureau of Statistics. For manufacturing intelligence, the R&D funding input, staff input, degree of industrialization of intelligent technology, smart device market profit and smart device market efficiency were obtained from the “China Statistical Yearbook on High Technology Industry” (“China Statistical Yearbook on High Technology Industry” https://data.cnki.net/v3/Trade/yearbook/single/N2022010268?zcode=Z018 (accessed on 25 July 2022)). Smart device input, internet infrastructure investment, staff input, software development and service situation, and environmental improvement are from “China Statistical Yearbook” (“China Statistical Yearbook” http://www.stats.gov.cn/tjsj./ndsj/ (accessed on 25 July 2022)). Energy intensity was obtained from “China Energy Statistical Yearbook” (“China Energy Statistical Yearbook” https://data.cnki.net/v3/trade/Yearbook/Single/N2016120537?zcode=Z025 (accessed on 25 July 2022)). For green development efficiency, labor input was obtained from the “China Labor Statistical Yearbook” (“China Labor Statistical Yearbook” https://data.oversea.cnki.net/chn/area/yearbook/Single/N2022020102?dcode=D29 (accessed on 25 July 2022)). Capital input and desired output were obtained from the “China Statistical Yearbook”. Energy input and non-desired output were obtained from the “China Energy Statistical Yearbook”. For the remaining variables, human capital level was obtained from “China Human Capital Index Report 2021” (“China Human Capital Index Report 2021” http://humancapital.cufe.edu.cn/rlzbzsxm/zgrlzbzsxm2021/zgrlzbzsbgqw_zw_.htm (accessed on 27 July 2022)), published by the China Centre for Human Capital and Labor Market Research. GDP per capita, openness to the outside, industrial structure, telecommunications development level, technological innovation and government role were obtained from the “China Statistical Yearbook”. Energy efficiency was obtained from the “China Energy Statistical Yearbook”.

To reduce the heteroskedasticity and the influence of covariance among variables, the natural logarithm was taken for the control variables GDP per capita and telecommunication development level. The descriptive statistics of the main variables are shown in

Table 2. Descriptive statistics for variables.

Variables	Sample Size	Average	Standard Deviation	Minimum Value	Maximum Value
Green development efficiency	300	1.1689	0.2241	0.8958	2.5642
Intelligent level of manufacturing	300	0.0949	0.1233	0.0023	0.8794
Human capital level	300	9.2529	0.9169	7.4739	12.7820
GDP per capita	300	10.8406	0.4361	9.7058	12.0130
Openness to the outside	300	0.0192	0.0150	0.0004	0.0796
Industrial structure	300	1.2187	0.6960	0.5180	5.2968
Telecommunications development level	300	13.4669	0.8961	10.8325	15.1993

.

5. Empirical Results and Discussion

5.1. Benchmark Regression Results

The Effect of Manufacturing Intelligence on Green Development Efficiency

Table 3. Benchmark regression and robustness test.

Variables	Fixed-Effects Model	Random-Effects Model	Replacement of Core Explanatory Variables	Regression with Reduced Tails	Exclude Provinces
	(1)	(2)	(3)	(4)	(5)
Intelligent level of manufacturing	0.7312 *** (0.1592)	0.3577 *** (0.1291)	0.1035 *** (0.0244)	1.0497 *** (0.1847)	0.5479 *** (0.1180)
Human capital level	0.0464 * (0.0261)	0.0215 (0.0229)	0.0472 * (0.0262)	0.0480 * (0.0255)	0.0107 (0.0196)
GDP per capita	0.1819 *** (0.0574)	0.1364 *** (0.0445)	0.2067 *** (0.0563)	0.1439 ** (0.0575)	0.0047 (0.0462)
Openness to the outside	−2.6027 *** (0.7159)	−2.4490 *** (0.7050)	−2.6601 *** (0.7191)	−2.5140 *** (0.7025)	−0.8082 (0.6509)
Industrial structure	0.3752 *** (0.0331)	0.2246 *** (0.0244)	0.3799 *** (0.0333)	0.3667 *** (0.0326)	0.1172 *** (0.0319)
Telecommunications development level	−0.0780 *** (0.0282)	0.0410 *** (0.0156)	−0.0805 *** (0.0283)	−0.0739 *** (0.0276)	0.1009 *** (0.0254)
Constant term	−0.6578 (0.4301)	−1.3211 *** (0.3696)	−0.8983 ** (0.4151)	−0.3367 (0.4326)	−0.5537 * (0.3255)
Ra2	0.4803	0.5778	0.4680	0.4820	0.4300
F value	121.65		119.82	128.15	93.89
Wald value		576.46
Fixed effect	YES		YES	YES	YES
Observations	300	300	300	300	260
Hausman test	0.0000

Note: *, **, and *** mean significant at the significance levels of 10%, 5% and 1%, respectively; the numbers in brackets are the standard errors; the same is shown below.

shows the benchmark regression results of the impact of manufacturing intelligence level on green development efficiency. Since the p-value of the Hausman test is significant at 0, the fixed-effects model was used for analysis. Column (1) of Table 3 shows that the coefficient of the core explanatory variable manufacturing intelligence level is positive and significant, indicating that the improvement in the level of intelligence in the manufacturing industry can promote the improvement in total green factor productivity. Among the control variables, the coefficients of human capital level, GDP per capita and industrial structure are positive and significant, indicating that the improvement in human capital level, and economic development, and the optimization of the industrial structure, are conducive to the efficiency of green development improvements. The coefficients of the degree of openness to the outside and the level of telecommunications development are significantly negative, which means that China’s extensive trade processing, long-distance optical cables and other infrastructure construction may have an adverse impact on the efficiency of green development.

5.2. Endogeneity and Robustness Testing

5.2.1. Replacement of Core Explanatory Variables

Replacing core explanatory variables is a common method for robustness testing. This paper used principal component analysis to re-calculate the level of manufacturing intelligence in each region. Column (3) in Table 3 shows that improving the level of intelligence can significantly promote the efficiency of green development. The sign of the regression coefficient of each variable is consistent with the result obtained from the previous benchmark regression, which proves that the result is robust.

5.2.2. Regression with Reduced Tails

The results of the reduced-tail regression of the core explanatory variables at the 1% level, column (4) of Table 3, show that the regression results, significance and the previous section for the core explanatory variables and control variables are generally consistent, proving that the results are robust.

5.2.3. Exclude Provinces

Since the level of green development efficiency varies among provinces, and green development efficiency of some provinces is far ahead than that of other provinces, this paper excludes the top four regions in terms of green development efficiency, namely Beijing, Shanghai, Tianjin and Chongqing. The results of column (5) of Table 3 show that this improvement in the intelligent level of the core explanatory variable manufacturing industry promotes the efficiency of green development, which proves that the results are robust.

5.3. A Mechanistic Test of Manufacturing Intelligence’s Effect on Green Development Efficiency

5.3.1. Policy Effects of the “Guidelines for the Construction of the National Intelligent Manufacturing Standard System”

The model controls for both annual effects and regional effects. As shown by the estimation results in

Table 4. Policy effects.

Variables	Double-Difference Model
Did × T	0.1597 *** (0.0256)
Human capital level	0.0218 (0.0221)
GDP per capita	0.2157 *** (0.0440)
Openness to the outside	−0.7994 (0.7096)
Industrial structure	0.2116 *** (0.0247)
Telecommunications development level	0.0494 *** (0.0164)
Constant term	−2.2047 *** (0.3796)
Ra2	0.6609
Wald value	673.74
Fixed effect	YES
Observations	300

Note: *** means significant at the significance level of 1%.

, the regression coefficient of the core explanatory variable manufacturing intelligence is significantly positive at the 1% level, which indicates that the implementation of the “Guidelines for the Construction of the National Intelligent Manufacturing Standard System” significantly contributed to the efficiency of green development in the eastern region, with a coefficient of 0.1597.

5.3.2. The Parallel Trend Test

Since the application of the double-difference model must satisfy the premise of the parallel trend assumption, that is, the experimental group and the control group have the same change trend before the implementation of the policy. In order to avoid the effect of multicollinearity, the fourth year before the policy implementation of the “Guidelines for the Construction of National Intelligent Manufacturing Standard System” is estimated as the reference group in this paper, and the test results are shown in Figure 3. Before the policy was implemented, the confidence interval fluctuation range contained 0. After the policy was implemented, the confidence interval fluctuation range was far from 0. It can be seen that before the policy was implemented in 2015, the coefficients of the core explanatory variables were not significant, indicating that there was no significant difference in green development efficiency between the experimental group and the control group, passing the parallel trend test. In addition, the estimated coefficients in the year when the policy was implemented were positive but not significant, and the estimated coefficients were all significantly positive after the policy was implemented, indicating that the policy of the “Guidelines for the Construction of the National Intelligent Manufacturing Standard System” has a lag period of approximately one year regarding the efficiency of green development, and has sustained positive effects.

5.3.3. The Transmission Path of Technological Innovation and Energy Efficiency

The previous section has confirmed that the improvement in manufacturing intelligence can promote the improvement in green development efficiency, but its mechanism of influence still needs to be further tested. The regression results of the mechanism test are shown in

Table 5. Mechanism test on the influence of manufacturing intelligence on the efficiency of green development.

Variables	Technological Innovation	Green Development Efficiency	Energy Efficiency	Green Development Efficiency
	(1)	(2)	(3)	(4)
Did × T	0.0281 *** (0.0100)	0.1023 *** (0.0228)	0.0543 *** (0.0122)	0.0817 *** (0.0225)
Energy efficiency				0.5130 *** (0.1095)
Technological innovation		0.2580 * (0.1386)
Human capital level	−0.0487 *** (0.0114)	0.0773 *** (0.0266)	−0.0184 (0.0140)	0.0742 *** (0.0250)
GDP per capita	0.0279 (0.0237)	0.2435 *** (0.0536)	0.5725 *** (0.0291)	−0.0430 (0.0812)
Openness to the outside	0.3154 (0.3307)	−1.8697 ** (0.7462)	0.6544 (0.4050)	−2.1240 *** (0.7239)
Industrial structure	−0.0070 (0.0149)	0.3534 *** (0.0336)	−0.1325 *** (0.0182)	0.4195 *** (0.0355)
Telecommunications development level	−0.0059 (0.0124)	−0.0890 *** (0.0278)	−0.1168 *** (0.0151)	−0.0306 (0.0298)
Constant term	0.3844 ** (0.1691)	−1.4464 *** (0.3847)	−3.7547 *** (0.2071)	0.5789 (0.5519)
Ra2	0.2194	0.4705	0.1858	0.5711
F value	6.31	107.01	89.61	117.05
Fixed effect	YES	YES	YES	YES
Observations	300	300	300	300

Note: *, **, and *** mean significant at the significance levels of 10%, 5% and 1%, respectively.

. Column (1) shows that manufacturing intelligence can significantly improve the level of technological innovation. Column (2) shows that the estimated coefficient of technological innovation level on green development efficiency is significantly positive, indicating a significant mediating effect of technological innovation; that is, the level of manufacturing intelligence can promote the improvement in green development efficiency by improving the level of technological innovation. The test results in columns (3) and (4) show that manufacturing intelligence can also improve green development efficiency by increasing energy use efficiency, and the energy use efficiency mechanism is verified.

5.4. The Threshold Effect of Manufacturing Intelligence on Green Development Efficiency

The results of the threshold number sampling test using Stata 15.0 software are shown in

Table 6. Threshold effect test results for manufacturing intelligence’s effect on green development efficiency.

Threshold Variable	Threshold	Threshold Value	F Value	p Value	10% Critical Value	5% Critical Value	1% Critical Value
Technological innovation	Single threshold	0.3472	56.71	0.0000 ***	24.1243	27.9323	36.6306
	Double sill	0.3472, 0.1136	10.54	0.3867	69.1699	84.3016	144.4616
Government’s role	Single threshold	0.1765	60.61	0.0000 ***	30.8952	37.3986	48.3053
	Double sill	0.1765, 0.1954	16.12	0.4167	26.8287	31.7205	47.9296

Note: *** means significant at the significance level of 1%.

. As can be seen from Table 6, when technological innovation and government role are used as threshold variables, the single-threshold test is significant at the 1% level and the double threshold is not significant with single-threshold values of 0.3472 and 0.1765, respectively. The regression analysis based on the single threshold is determined based on the threshold estimates and the corresponding confidence intervals, and the regression results are shown in

Table 7. Threshold regression estimation results for manufacturing intelligence’s effect on green development efficiency.

Explanatory Variables	Threshold Variable	Explanatory Variables	Threshold Variable
	Technological Innovation		Government’s Role
Intelligent level of manufacturing (Technological innovation ≤ η)	0.6008 *** (0.1470)	Intelligent level of manufacturing(Government’s role ≤ η)	0.7047 *** (0.1451)
Intelligent level of manufacturing (Technological innovation > η)	2.7770 *** (0.3204)	Intelligent level of manufacturing(Government’s role > η)	2.6851 *** (0.3008)
Human capital level	0.0431 * (0.0239)	Human capital level	0.0328 (0.0238)
GDP per capita	0.1264 ** (0.0531)	GDP per capita	0.0772 (0.0541)
Openness to the outside	−3.2908 *** (0.6630)	Openness to the outside	−3.1918 *** (0.6572)
Industrial structure	0.2599 *** (0.0344)	Industrial structure	0.2808 *** (0.0328)
Telecommunications development level	−0.0044 (0.0278)	Telecommunications development level	−0.0508 * (0.0259)
Constant term	−0.8637 ** (0.3951)	_cons	0.2735 (0.4115)
Ra2	0.5352	Ra2	0.5545
F value	131.53	F value	133.44
Observations	300	Observations	300

Note: *, **, and *** mean significant at the significance levels of 10%, 5% and 1%, respectively.

.

When technological innovation is the threshold variable, as shown in Table 7, the coefficient of technological innovation is 0.6008 when technological innovation is below the threshold value of 0.3472. When technological innovation is above the threshold value, the coefficient of technological innovation is 2.7770. This indicates that when technological innovation is improved to a certain degree, its contribution to green development efficiency increases in a leapfrog manner. When the government’s role is used as the threshold variable, the coefficient of the government’s role is 0.7047 when the government’s role is lower than the threshold value of 0.1765. When the government’s role is higher than the threshold value, the coefficient of the government’s role is 2.6851. This indicates that, when the government’s role of the region crosses the threshold value, the promotion of green development efficiency is more obvious when the level of manufacturing intelligence is improved.

6. Conclusions

Based on the theoretical analysis of the impact that the manufacturing intelligence level has on green development efficiency, this study used panel data of 30 sample provinces (cities and autonomous regions) from 2011 to 2020 to conduct an empirical analysis. It used a fixed-effects model, a mediator model based on the double-difference method and a threshold model. The main conclusions of this study are as follows: first, an improvement in the level of intelligence in the manufacturing industry significantly increases the efficiency of green development. Second, China’s “Guidelines for the Construction of the National Intelligent Manufacturing Standard System” were used for a quasi-natural experiment. We applied the double-difference method to demonstrate that the improvement in manufacturing intelligence can significantly promote green development efficiency and reveal its heterogeneity, i.e., the promotional effect is greater in eastern regions. Third, from the mechanistic perspective, the level of manufacturing intelligence can improve green development efficiency through promoting technological innovation and improving energy efficiency. Fourth, the promotion effect of the manufacturing intelligence level on green development efficiency is non-linear, and there is a threshold effect for the technological innovation level and the government’s role in the influence of the manufacturing intelligence level on green development efficiency. In regions where the level of technological innovation and the government’s role exceed the threshold, the promotion effect is more obvious.

The first contribution of this study is revealing that the level of manufacturing intelligence continued to increase over the period 2011–2020. There is regional heterogeneity. In the process of increasing the level of manufacturing intelligence, there is a positive effect on the green development efficiency. Secondly, not only was the direct effect of manufacturing intelligence on China’s green development efficiency investigated, but the indirect effect was also explored, i.e., the transmission effects of the level of technological innovation and energy use efficiency. Finally, this work revealed that, when the level of technological innovation and the strength of the government’s role cross the thresholds of 0.3472 and 0.7047, respectively, the promotion effect is more pronounced. It provides useful ideas with which for China and other countries to improve their green development efficiency through enhancing intelligence. It also enriches the research on manufacturing intelligence and green development efficiency.

These empirical results may provide some implications for the government. First, as manufacturing intelligence can significantly contribute to the efficiency of green development, efforts should be made to increase manufacturing intelligence. This can be achieved through the construction of product cloud platforms and the use of advanced digital technologies. Second, based on heterogeneity, it can be seen that increased intelligence has a greater effect on the efficiency of green development in the eastern region. Therefore, differentiated policies should be implemented according to regional characteristics. AI relies on information technology and digital platforms, and there is a certain spatial spillover effect. Therefore, the east, middle and west should, based on their respective advantages and characteristics, strengthen the synergistic development between regions. The flow of advanced technology and talents should be guided from the eastern region to the central and western regions, thus narrowing the intelligence gap between regions. The radiation of intelligence should be fully considered, and different regions should work together to realize the efficiency of manufacturing intelligent, empowered green development. Third, the role that the transmission mechanism plays in technological innovation and energy use efficiency should be fully explored. Firstly, investment in R&D should be increased in key technology areas. This provides financial security when upgrading manufacturing intelligence and promotes the extension of enterprises to the front end of the innovation chain. Secondly, enterprises should improve their technical innovation ability, focus on overcoming technical difficulties and realize the “butterfly change” upgrade. The integration of intelligent manufacturing industry, academia and research should be deepened and highly skilled people should be encouraged to participate. This would generate more independent innovation and disruptive innovation achievements. Fourth, the threshold effect shows that the role of manufacturing intelligence in promoting green development efficiency is more obvious when the strength of the government role crosses the threshold. Therefore, the government should increase its market-oriented reforms and give full play to the role of government guidance and services. This would release signals to the market by formulating policies and setting up intelligent demonstration enterprises so that more social capital flows to production intelligence and guides enterprises to empower their green development through intelligent manufacturing. The government should strengthen incentives and assessment guidance, play an integrated and coordinating role, and encourage more enterprises to form a mature, replicable and replicable new model of intelligent manufacturing. This would enhance the efficiency of green development. Financial support for smart manufacturing and environmental protection should also be increased, giving play to the “multiplier effect” of public finance. Adequate R&D investment funds and policies such as R&D expense deduction could lay the foundation for enterprises’ digital technology innovation, and also promote their intelligent transformation. Green project subsidies and subsidies provide a guarantee for green production, thus promoting the efficient and green development of enterprises.

This study also has some limitations. Firstly, this study was only conducted at the level of Chinese provinces (cities and autonomous regions). Although the existing results are robust enough, the micro-enterprise level was not considered. In the future, data from listed companies may be considered. Secondly, this study only conducted an in-depth analysis of the manufacturing sector. Based on industry differences, continued research on other industries is planned.