The Digital Economy and Total Factor Productivity of the Manufacturing Industry: From the Perspective of Subdivided Manufacturing Sectors

Abstract

Global economic change and technological development have made the digital economy a key force for economic growth. With its innovation, penetration, and integration, the digital economy has penetrated into all aspects of the manufacturing industry, bringing opportunities and challenges to the manufacturing industry. Based on the development level of China’s digital economy and manufacturing industry data from 2001 to 2023, using DEA–BCC–Malmquist, this paper studies the impact of the digital economy on the total factor productivity of the manufacturing industry. It is found that the development of the digital economy impacted the total factor productivity of the manufacturing industry, in which cost and innovation research and development played an intermediary effect, and the digital economy has shown a more positive push in technology-intensive and labor-intensive industries than in capital-intensive ones. Based on the findings, this research is designed to empower the manufacturing sector to capitalize on emerging digital economy opportunities while advancing sustainable development goals.

1. Introduction

In the current era of profound changes in the global economic landscape and rapid technological advancements, the digital economy has emerged as a pivotal force driving global economic growth. It exerts a far-reaching influence on the development models, competitive dynamics, and even the global economic order across various industries [1]. With the rapid evolution of information technology, the digital economy, characterized by its innovativeness, penetrability, and integrativeness, has swiftly permeated every facet of manufacturing [2]. From product design and production to supply chain management and marketing, the wave of digital transformation is reshaping the manufacturing sector at an unprecedented pace. Big data analysis enables enterprises to precisely grasp market demands and achieve personalized customized production; cloud computing technology provides efficient storage and computing support, reducing operational costs; the deep integration of artificial intelligence and automation technologies is propelling the production process towards intelligence and flexibility, significantly enhancing production efficiency and product quality [3].

Total factor productivity has attracted widespread attention in the academic community that originated in Solow’s 1957 study of technological change and aggregate production function, which has now become an important indicator for analyzing the sources of economic growth and evaluating the quality of economic growth. The study proposes that total factor productivity refers to a variety of factors of production inputs, such as capital, labor, energy, natural resources, etc., outside of technological progress, technical efficiency, management, innovation, the socio-economic system, and other factors. Additionally, total factor productivity is the ratio of inputs to outputs, measuring the level of output per unit of inputs, innovation, management, the socio-economic system, and other factors that lead to an increase in output. The service sector advances faster digitally than industry/agriculture because focused digital efforts here boost growth more effectively [4]. Manufacturing, as the cornerstone of a nation’s real economy, plays a critical role in strengthening national economic power and elevating international competitiveness through the enhancement of total factor productivity [5]. In the context of the vigorous development of the digital economy, manufacturing faces unprecedented opportunities and challenges. The digital economy is a new driver of total factor productivity. It can increase total factor productivity in general and contribute to the sustainable development of the manufacturing industry. It has also contributed to TFP improvement in textile and specialized equipment manufacturing but has had little impact in pharmaceutical manufacturing [6]. On the one hand, the digital economy offers new pathways and means for manufacturing to overcome traditional development bottlenecks and achieve transformation and upgrading [7]; on the other hand, effectively leveraging the advantages of the digital economy to boost total factor productivity has become a key issue that manufacturing enterprises urgently need to address [8].

At present, research on the relationship between the level of the digital economy and the total factor productivity of manufacturing is still evolving, with many issues remaining to be explored. How can we measure the contribution of the digital economy to the total factor productivity of manufacturing? Through which specific mechanisms does the digital economy impact the total factor productivity of manufacturing? Are there differences in the impact on total factor productivity among different types of manufacturing during digital economic development? Addressing these questions holds significant theoretical and practical importance for understanding the intrinsic connection between the digital economy and manufacturing development and guiding manufacturing enterprises to rationally utilize digital technologies to enhance total factor productivity. This research focuses on conducting in-depth research into the impact of the digital economy on the total factor productivity of manufacturing, aiming to provide valuable references for promoting the high-quality development of manufacturing.

2. Theoretical Analyses and Research Hypotheses

2.1. The Digital Economy and Total Factor Productivity in the Manufacturing Industry

The digital economy is integrating into manufacturing with an unprecedented momentum, emerging as a key force facilitating the enhancement of its total factor productivity. There is a significant inverted U-shaped relationship between digital technology and total factor productivity. The critical value of high technology manufacturing is greater than that of medium and low technology manufacturing [9]. In the current wave of digitization, the digital economy is impacting the manufacturing industry [10]. Supply, production, and sales, as the core processes of manufacturing, have exploited the technological and model advantages of the digital economy to achieve efficiency increments, cost reductions, and strengthened value creation capabilities. These processes are closely interlinked and mutually reinforcing, jointly constructing a potent system that propels the total factor productivity of manufacturing forward [11].

In the supply chain, big data and cloud computing technologies have proffered robust support for supply chain management in manufacturing [12]. By collecting and analyzing a large amount of market data, supplier data, and logistics information, it is possible to forecast raw material demand, plan production strategies in advance, effectively avoid industry shutdowns due to shortages, and ensure the continuity of production activities [6]. This endows the entire manufacturing industry with greater stability in raw material supply and significantly mitigates the production efficiency losses attributed to unstable supply. Digital platforms have further expanded the avenues for manufacturing to scout for suppliers, transcending geographical constraints and enabling the selection of high-quality suppliers globally, thereby obtaining high-quality raw materials and components at more reasonable prices and effectively lowering the overall procurement costs of the industry. The digital collaborative mechanism established with suppliers enables real-time information sharing, allowing both parties to coordinate production plans and delivery arrangements more efficaciously, conspicuously enhancing the response speed and flexibility of the supply chain. This all-round improvement in supply chain efficiency has laid a solid foundation for the elevation of total factor productivity in the manufacturing industry [13].

In the production stage, the extensive adoption of intelligent manufacturing technologies has enabled the automation and intelligence of the production process. Robots and automated production lines can precisely and efficiently execute repetitive tasks, achieving both high production speed and consistent product quality, thereby significantly increasing output per unit time. The industrial internet interconnects production equipment, sensors, and control systems, facilitating real-time data sharing and communication among devices. Enterprises can monitor equipment status in real time, conduct predictive maintenance through data analysis, detect and address potential equipment failures proactively, reduce downtime, enhance equipment utilization, and increase effective production time, thus boosting industry production efficiency. Digital technologies have further facilitated the optimization and reconfiguration of production processes. Leveraging virtual simulation technology, manufacturers can simulate and optimize the entire production process during the design and planning stages, identifying issues early and making adjustments to avoid errors and rework, thereby shortening the production cycle. Through digital management systems, manufacturers can achieve precise management of the production process, monitor production progress and quality status in real time, make timely decisions and adjustments, ensure efficient and orderly operations, and improve total factor productivity [14].

In the sales stage, e-commerce platforms and social media have become critical channels for the manufacturing industry to promote products and reach customers [15]. These platforms enable the rapid dissemination of product information at a relatively low cost, expanding market coverage and increasing the size of the potential customer base. Compared with traditional sales channels, digital marketing significantly enhances the efficiency and precision of information dissemination, attracting more customer attention and driving product sales growth. Through in-depth analysis of customer big data, manufacturers gain deeper insights into customer needs and behavioral preferences. Based on these insights, they can undertake precise market segmentation and product positioning, develop products and services that better meet market demands, and enhance customer satisfaction and loyalty. The digital customer feedback mechanism allows for the prompt collection of customer opinions and suggestions, enabling rapid product improvements and optimizations, thereby strengthening market competitiveness. Improved sales efficiency and expanded market share enable manufacturers to fully realize production potential, achieve economies of scale, and ultimately drive the enhancement of total factor productivity.

Based on the above, this paper proposes Hypothesis H1: The digital economy can impact the total factor productivity of the manufacturing industry.

2.2. The Indirect Influence of the Digital Economy on the Total Factor Productivity in the Manufacturing Sector

• Cost Reduction

The digital economy offers multiple effective pathways for the manufacturing sector to reduce costs and expenses [16], thereby positively influencing total factor productivity. In the procurement phase, big data analytics enable manufacturing enterprises to accurately track raw material market dynamics, clearly understand price trends and supply-demand variations, and optimize purchasing decisions to minimize raw material procurement costs. Digital supply chain management systems broaden the supplier selection pool, allowing enterprises to comprehensively evaluate and select high-quality, low-cost suppliers. These systems also facilitate real-time monitoring of supply conditions, ensuring supply stability and minimizing potential losses [17]. Regarding inventory management, the integration of the Internet of Things (IoT) and big data enables enterprises to achieve precise, real-time inventory monitoring, implement demand-driven replenishment strategies, avoid overstocking, and reduce inventory holding costs and warehouse space utilization costs. The reduction in these costs allows enterprises to allocate more resources to production expansion or other value-added activities, thereby enhancing total factor productivity [18].

During the production process, the digital economy plays a pivotal role in cost reduction. IoT technology facilitates equipment interconnectivity, enabling real-time monitoring of equipment status and predictive maintenance strategies to pre-emptively identify and address potential faults, thus preventing production disruptions and reducing maintenance costs. Additionally, the widespread adoption of digital communication tools and online office platforms has eliminated traditional time and space constraints, significantly lowering internal communication costs, travel expenses, and other administrative overheads. The comprehensive reduction in costs and expenses enhances resource utilization efficiency, enabling enterprises to achieve higher output without increasing input factors, which strongly promotes the improvement of total factor productivity in the manufacturing industry.

Based on this analysis, this paper proposes Hypothesis H2: The digital economy can enhance the total factor productivity of the manufacturing industry by reducing costs, with cost reductions serving as a mediating factor in this process.

ii

Promote Innovation

The digital economy has created an unprecedentedly favorable environment for manufacturing innovation [19], thereby emerging as a powerful driver for enhancing total factor productivity. On the one hand, digital technologies themselves are rich sources of innovation, providing manufacturing with novel tools and methodologies [6]. Big data analytics functions as a sophisticated market detector, capable of precisely extracting latent value from vast datasets, assisting manufacturing enterprises in identifying new market demand signals, clarifying product improvement directions, and making R&D more targeted. Artificial intelligence algorithms serve as efficient R&D assistants, applicable in the product design and simulation testing phases, rapidly iterating design schemes and accelerating the R&D process while reducing R&D risks and unnecessary resource investment [20]. The application of these digital technologies enables enterprises to introduce innovative products more swiftly, meet diverse market demands, increase output without additional input, and thereby enhance total factor productivity.

On the other hand, the digital economy has revolutionized the innovation model of manufacturing, vigorously promoting the development of open innovation [3]. Through Internet platforms, manufacturing enterprises have transcended traditional innovation boundaries, facilitating extensive and in-depth collaborative exchanges with research institutions, universities, suppliers, and users worldwide [21]. This cross-domain and cross-regional cooperation network allows for the full integration of innovation resources from all parties, forming a robust innovation synergy. Enterprises can collaborate with research institutions to solve key technological challenges, jointly cultivate innovative talents with universities, develop new materials and processes with suppliers, and directly incorporate user feedback into product innovation [13]. Through open innovation, enterprises can continuously explore new technological domains and business models, create more high-value-added products and services, gain a competitive advantage in the market, achieve significant output growth without additional input, and inject continuous impetus into the enhancement of manufacturing total factor productivity [22].

Based on this, this paper proposes Hypothesis H3: The digital economy can enhance the total factor productivity of manufacturing by fostering innovation, that is, innovation and R&D play a mediating role in the process of the digital economy improving the total factor productivity of manufacturing.

3. Research Design

3.1. Sample Selection and Data Sources

This paper selects the sample data of China’s national manufacturing industry from 2001 to 2023. The specific data sources are as follows: (1) Macro data such as the level of the digital economy, industrial output value, and foreign investment are derived from the National Bureau of Statistics. (2) Data related to the manufacturing industry, such as the main business cost, operating expenses, R&D expenses, the number of scientific research employees, and the net value of fixed assets, are obtained from research reports of the China Manufacturing Association, the Institute of Industrial Economics of the Chinese Academy of Social Sciences, etc., and are also manually collected and statistically processed by gathering manufacturing-related data from databases such as CSMAR and WIND. Based on the collection of the raw data, this paper eliminates the observations with missing variables in the research samples and conducts a Winsorize tail-trimming at the upper and lower 1% for all research sample data to mitigate the influence of extreme values in the research samples. The upper and lower 1% Winsorize not only suppresses outliers but also ensures the comparability of results by correcting the symmetry of the distribution, aligning the statistical criteria [23].

According to research [24], the sample was divided into 31 sub-industries based on the National Economic Industry Classification (National Standard GB/T4754-2017), under the manufacturing sector in the National standards for the classification of national economic sectors issued by the National Bureau of Statistics of China (GB/T4754-2017). Since the name and scope of four sub-industries changed during the period, the difference sample was excluded. Additionally, according to the industry characteristics, it is divided into three categories: labor-intensive manufacturing, capital-intensive manufacturing, and technology-intensive manufacturing. Finally, this paper ultimately acquires 621 research samples, encompassing 27 sub-manufacturing industries with a total of 23 years of data from 2001 to 2023. The statistical software Stata 16.0 is employed to conduct research and analysis on the relevant data based on the sample data. The detailed classification is presented in

Table 1. Classification of China’s manufacturing sectors.

Industry Type	Industry Name
Labor-intensive manufacturing	Agricultural and Sideline Food Processing Industry, Food Manufacturing Industry, Wine, Beverage and Refined Tea Manufacturing Industry, Tobacco Product Manufacturing Industry, Textile Industry, Textile, Garment and Apparel Manufacturing Industry, Leather, Fur, Feather and Their Products and Footwear Manufacturing Industry, Wood Processing and Wood, Bamboo, Rattan, Palm and Grass Products Manufacturing Industry, Furniture Manufacturing Industry, Paper and Paper Products Manufacturing Industry, Printing and Recorded Media Reproduction Industry, Cultural, Educational, Arts and Crafts, Sports and Entertainment Goods Manufacturing Industry, Other Manufacturing Industry
Capital-intensive manufacturing	Petroleum, Coal and Other Fuel Processing Industry, Chemical Fiber Manufacturing Industry, Rubber and Plastic Products Industry, Non-metallic Mineral Products Industry, Ferrous Metal Smelting and Rolling Processing Industry, Non-ferrous Metal Smelting and Rolling Processing Industry, Metal Products Industry, Metal Products, Machinery and Equipment Repair Industry
Technology-intensive manufacturing	Chemical Raw Materials and Chemical Products Manufacturing, Pharmaceutical Manufacturing, General Equipment Manufacturing, Special Equipment Manufacturing, Automobile Manufacturing, Railway, Shipbuilding, Aerospace and Other Transportation Equipment Manufacturing, Electrical Machinery and Equipment Manufacturing, Computer, Communication and Other Electronic Equipment Manufacturing, Instrument and Meter Manufacturing, Comprehensive Utilization of Waste Resources Industry

.

3.2. Definition of Variables

(a)

Explanatory Variable

The explanatory variable in this study is the level of digital economy (DIGT). The DIGT index measures the extent to which a region or industry has developed in the digital economy, reflecting the penetration and application of digital technologies in economic activities. This includes aspects such as digital infrastructure development, digital technology innovation capacity, the scale of digital industries, and the integration of the digital economy with traditional industries [25]. Drawing on the work [10], this study uses the ratio of the added value of digital industries (such as electronic information manufacturing and software and information technology services) to GDP as a proxy for DIGT, thereby capturing the scale and contribution of these industries to the overall economy.

(b)

The Explained Variable

The explained variable in this study is the total factor productivity (TFP) of the manufacturing sector. TFP measures the overall efficiency with which inputs are converted into outputs in the manufacturing production process. It not only encompasses traditional input factors such as labor and capital but also incorporates the effects of technological progress, management innovation, and scale economies on production efficiency [25]. Specifically, TFP reflects the level of production technology, resource utilization efficiency, and overall competitiveness of the manufacturing sector within a given period.

Drawing on the methodologies proposed by researches [26,27], labor input (e.g., the number of employees in the manufacturing industry) and capital input (e.g., net fixed assets, total current assets, etc.) are selected as input indicators, and industrial output value and main business revenue are selected as output indicators. The number of employees as a proxy variable for labor inputs is intuitive and commonly found in firms’ financial statements and statistics, facilitating comparisons across time [28]. Capital inputs need to reflect the scale of resources actually occupied by the firm. Net fixed assets measure long-term production capacity, and current assets reflect short-term working capital; the combination of the two can avoid single-indicator bias [29]. The total amount of fixed assets and current assets is directly related to the actual resources of the enterprise such as production equipment, inventory, accounts receivable, etc., which can reflect the degree of reliance of manufacturing enterprises on tangible resources [30]. This research employs the DEA–BCC–Malmquist approach to calculate TFP. DEA (Data Envelopment Analysis) is a non-parametric method based on linear programming that evaluates the relative efficiency of decision-making units (DMUs) with multiple inputs and outputs. In this context, different regions or enterprises within the manufacturing sector are treated as DMUs. By constructing a DEA model, the efficiency scores of each DMU can be obtained, allowing for the determination of whether they lie on the production frontier.

The BCC model (Banker–Charnes–Cooper model) extends DEA by considering variable returns to scale (VRS). It introduces the convexity assumption to decompose technical efficiency into pure technical efficiency and scale efficiency, i.e., technical efficiency = pure technical efficiency × scale efficiency. This decomposition allows for a more detailed analysis of the sources of inefficiency, distinguishing between technical limitations and scale-related issues.

The Malmquist index is used to measure changes in TFP over time. It calculates the TFP index by comparing the distances between production frontiers at two distinct points in time. A Malmquist index greater than 1 indicates an increase in TFP; equal to 1 signifies no change; and less than 1 suggests a decline in TFP. The BCC–Malmquist method within the DEA framework integrates the efficiency decomposition provided by the BCC model with the dynamic measurement of TFP changes offered by the Malmquist index, enabling a comprehensive and dynamic analysis of TFP variations and their underlying factors in the manufacturing sector.

(c)

Mediating Variables

Cost–Expense Ratio (COST): The cost–expense ratio reflects the level of costs and expenses incurred by manufacturing enterprises for each unit of output generated during their production and operation processes. This variable comprehensively captures the enterprises’ cost control capability and operational efficiency in both production and sales, thereby exerting a significant influence on enterprise productivity [31]. In this study, the cost–expense ratio is calculated as the sum of main business costs and operating expenses divided by the total industry output value. A lower cost–expense ratio indicates that an enterprise can achieve higher output with relatively less cost input, reflecting superior cost management and higher economic benefits. Conversely, a higher cost–expense ratio suggests poor cost control and may lead to compressed profit margins.

Innovation and R&D (INNO): Innovation and R&D are critical pathways influencing total factor productivity (TFP). Higher levels of innovation generally result in higher TFP [32]. In this study, the ratio of R&D expenditure to the total industry output value is used to measure innovation and R&D activities. The proportion of R&D expenditure gauges the intensity and emphasis placed on R&D investment within various manufacturing industries. R&D expenditure encompasses funds allocated for research and development activities, including expenses for new product development, technological improvements, and exploration of new processes, such as salaries for R&D personnel, purchases of R&D equipment, and costs of experimental materials. A higher proportion of R&D expenditure signifies greater investment in technological innovation, indicating a stronger focus on enhancing product competitiveness, exploring new markets, and promoting industry-wide technological progress. Conversely, a lower proportion of R&D expenditure may suggest insufficient impetus for technological innovation, potentially affecting the industry’s long-term sustainable development and market competitiveness.

(d)

Control Variables

Manufacturing Scale (SIZE): This variable reflects the average production scale of each manufacturing sector, indicating the overall production capacity and market concentration within the industry. The scale of the industry significantly influences total factor productivity [33]. In this study, it is calculated by dividing the total output value of each manufacturing sector by the number of enterprises in that sector. A higher value of this indicator suggests a larger average production scale per enterprise in the industry.

Ownership Structure (SOE): This variable describes the proportion of state-owned and state-controlled enterprises in the industrial economy, reflecting the status and influence of the state-owned economy in the industrial sector and the relative pattern of different ownership structures within the industrial economy. Different ownership types face varying financing conditions and costs, which have distinct impacts on total factor productivity [34]. In this study, it is measured by dividing the total output value of state-owned and state-controlled enterprises by the total industrial output value. A higher ratio indicates a greater share of state-owned and state-controlled enterprises in the industrial economy.

Human Capital (HUM): This variable measures the proportion of personnel with scientific and technological capabilities among all employed personnel in each manufacturing sector, reflecting the quality and innovation potential of human resources within the industry. High-level human capital can lead to high levels of total factor productivity [35]. In this study, it is determined by dividing the number of scientific and technological activity personnel in each sector by the total number of employed personnel in that sector. Scientific and technological activity personnel include those directly engaged in R&D activities and those specifically involved in managing and supporting these activities. The total number of employed personnel refers to all labor forces actually engaged in production and operation activities in enterprises within the sector. A higher ratio indicates better human capital quality and greater innovation potential in the sector.

Capital Intensity (FIN): Capital intensity reflects the net fixed asset value per employee in the manufacturing sector, indicating the capital intensity of the industry, i.e., the degree of reliance on capital input for production. Different levels of capital intensity affect enterprise operations, and lower capital density can be more conducive to improving total factor productivity [18]. In this study, it is measured by dividing the net fixed asset value of the industry by the total number of employees in the manufacturing sector. The net fixed asset value is the original value of fixed assets minus accumulated depreciation, reflecting the current value of fixed assets. A higher value indicates greater capital intensity and a higher value of fixed assets per employee.

Foreign Direct Investment (FDI): This variable reflects the participation and influence of foreign capital in each manufacturing sector, indicating the level of openness and international favorability towards the sector. It can significantly impact total factor productivity [36]. In this study, it is calculated by dividing the industrial output value of foreign-funded enterprises by the total output value of the sector. Foreign-funded enterprises include Sino-foreign joint ventures, Sino-foreign cooperative enterprises, and wholly foreign-owned enterprises. A higher ratio indicates a larger proportion of foreign direct investment in the sector and a higher degree of internationalization.

The definitions of all the variables are presented in

Table 2. Definition of variables.

Type	Name	Code	Interpretation
Dependent Variable	Total factor productivity	TFP	DEA–BCC–Malmquist index was computed
Independent Variable	Level of digital economy	DIGT	The proportion of the added value of the digital industry to GDP
Mediator Variable	Cost and expense ratio	COST	The main business cost of manufacturing + operating expenses)/industrial output value
	Proportion of R&D expenses	INNO	R&D expenditure of each industry/total output value of the industry
Control Variable	Scale of the manufacturing industry	SIZE	The gross product value of each industry in the manufacturing sector/the number of enterprises in the industry
	Ownership structure	SOE	The total output value of state-owned and state-controlled enterprises/the total output value of industry
	Human capital	HUM	The number of scientific and technological activity personnel in each industry/the total number of employed personnel in the corresponding industry
	Capital intensity	FIN	The net value of fixed assets in the industry/the total number of all employees in the manufacturing industry
	Foreign direct investment	FDI	The industrial output value of the three forms of foreign investment enterprises/rhe total industrial output value of the industry

.

3.3. Model Specification

In order to understand why the digital economy has an impact on total factor productivity, an extended interpretation of the Solow growth model is introduced. An extended interpretation of Solow’s growth model is introduced: digital technologies are viewed as “augmented technological advances (A)” whose effects depend on complementarities with traditional factors [37]. For example, big data optimization of the supply chain can be understood as improving capital turnover efficiency (K/A synergy), while artificial intelligence scheduling of the workforce corresponds to improved labor allocation efficiency (L/A synergy). However, the limitation of Solow’s model is that he is unable to explain the endogenous sources of technological progress and thus needs to be complemented by invoking endogenous growth theory when analyzing the impact of digitization on TFP. Endogenous growth theory mediates that technological progress and knowledge accumulation are truly endogenous to the economy, citing) theory [38] of directed technological change, which suggests that digital technology may facilitate knowledge spillovers by lowering the marginal cost of innovation, which is consistent with the endogenous growth framework. These additions provide a clearer theoretical anchor for the hypotheses that follow.

This article draws on the research methods [29,39] and employs Model 1 to assess the influence of the digital economy level on the total factor productivity of the manufacturing industry:

TFP = β₀ + β₁DIGT + β₂Controls + ∑Ind + ∑Year + μ

(1)

In the above equation, TFP represents the total factor productivity of the manufacturing industry, DIGT denotes the level of digital economy development, Controls refers to the control variables, and Ind and Year indicate that the fixed effects of sub-industries and years have been controlled. μ is the random disturbance term. If the coefficient of DIGT is significantly positive, it suggests that a higher level of digital economy development can markedly enhance the total factor productivity of the manufacturing industry.

This article employs Models 2 and 3 to examine the mediating effect of the cost and expense ratio in the influence of the digital economy level on the total factor productivity of the manufacturing industry:

COST = β₀ + β₁DIGT + β₂Controls + ∑Ind + ∑Year + μ

(2)

TFP = β₀ + β₁COST + β₂DIGT + β₃Controls + ∑Ind + ∑Year + μ

(3)

In the aforesaid equation, COST represents the cost and expense ratio, TFP represents the total factor productivity of the manufacturing industry, DIGT denotes the level of digital economy development, Controls refers to the control variables, and Ind and Year indicate that the fixed effects of sub-industries and years have been controlled. μ is the random disturbance term. If the coefficients of DIGT and COST in Formula (2) are significantly negative, it suggests that the extent of digital economic development can significantly lower the cost and expense ratio. If the coefficient of COST and TFP in Formula (3) is significantly negative, it implies that the cost and expense ratio will significantly reduce the total factor productivity of the manufacturing industry. The combination of the two indicates that the degree of digital economic development can significantly increase the total factor productivity of the manufacturing industry by exerting the mechanism of reducing the cost and expense ratio.

This article employs Models 4 and 5 to examine the mediating effect of innovation and research and development in the influence of the digital economy level on the total factor productivity of the manufacturing industry:

INNO = β₀ + β₁DIGT + β₂Controls + ∑Ind + ∑Year + μ

(4)

TFP = β₀ + β₁INNO + β₂DIGT + β₃Controls + ∑Ind + ∑Year + μ

(5)

In the above equation, INNO denotes the ratio of R&D expenditure, TFP represents the total factor productivity of the manufacturing industry, DIGT denotes the level of digital economy development, Controls refers to the control variables, and Ind and Year indicate that the fixed effects of sub-industries and years have been controlled. μ is the random disturbance term. If the coefficient of DIGT and INNO in Formula (4) is significantly positive, it indicates that the level of digital economic development can substantially promote innovation and R&D activities. If the coefficient of INNO and TFP in Formula (5) is significantly positive, it suggests that increased innovation and R&D can significantly enhance the total factor productivity (TFP) of the manufacturing industry. Taken together, these findings imply that the level of digital economic development can significantly boost the TFP of the manufacturing industry by enhancing innovation and R&D.

4. Empirical Examination

4.1. Descriptive Statistical Analysis

Table 3. Descriptive statistics.

Variable	Observation	Mean	Standard Deviation	Minimum	Maximum
TFP	621	1.042	0.084	0.837	1.366
DIGT	621	22.952	11.365	7.115	43.886
COST	621	1.366	0.947	0.001	3.484
INNO	621	0.954	0.67	0.138	3.374
SIZE	621	0.789	1.881	−1.542	8.607
SOE	621	0.201	0.23	0.006	0.994
HUM	621	0.036	0.028	0.002	0.119
FIN	621	23.356	22.282	2.204	118.682
FDI	621	0.26	0.151	0.001	0.774

presents the descriptive statistics of the variables employed in the regression analysis of this paper. The following can be observed from the table: Concerning the independent and dependent variables, the mean of total factor productivity (TFP) is 1.042, reflecting the average level of total factor productivity in the manufacturing sector. The relatively small standard deviation of 0.084 suggests that the total factor productivity of each sample is relatively concentrated around the mean, and the differences in production efficiency among different samples are not substantial. The narrow gap between the mean and the minimum value of 0.837 and the maximum value of 1.366 corroborates that the distribution of total factor productivity in the samples utilized in this study is relatively concentrated and the overall development is relatively balanced. The mean of the digital economy level (DIGT) is 22.952, representing the average level of digital economy development in China. The relatively large standard deviation of 11.365 indicates that the dispersion degree of the digital economy level in the samples is high, and there exist considerable differences in the degree of digital economy development among different years. The substantial range between the minimum value of 7.115 and the maximum value of 43.886 implies that the digital economy level has undergone significant changes within the sample time interval.

Among the mediating variables, the mean cost and expense ratio (COST) is 1.366, signifying that, on average for the entire manufacturing sector, the sum of the main business costs and operating expenses of manufacturing enterprises constitutes 1.366% of the industry’s output value. This relatively high value reflects the heavy overall cost and expense burden of the manufacturing industry, which might exert pressure on profits. The standard deviation of 0.947 indicates a moderate degree of dispersion, suggesting that there are certain disparities in cost control levels among different industries. The minimum value of 0.001 and the maximum value of 3.484 demonstrate that the distribution range of the cost and expense ratio within the sample is relatively broad, with extreme cases of both exceptional and extremely poor cost control. The mean proportion of R&D expenses (INNO) is 0.954, indicating that the average proportion of R&D expenses in each industry’s total output value approaches 1%. This ratio reflects the industry’s emphasis on and investment level in R&D. Overall, the R&D investment is at a certain level, but there remains room for enhancement. The standard deviation of 0.67 indicates a moderate degree of dispersion, suggesting that the differences in R&D investment among different industries might be notable. The considerable gap between the minimum value of 0.138 and the maximum value of 3.374 highlights the imbalance in R&D investment across various industries.

In terms of the control variables, the mean value of the manufacturing scale, which is 0.789, represents the gross production value of various manufacturing industries. The relatively large standard deviation of 1.881 indicates obvious differences in manufacturing scale among different industries, suggesting the existence of both large-scale and small-scale industrial groups. The minimum value of −1.542 and the maximum value of 8.607 imply that the distribution of manufacturing scale is rather dispersed. The negative minimum value may hint at certain exceptional circumstances or production write-offs. The mean value of the ownership structure (SOE) is 0.201, suggesting that the total output value of state-owned and state-controlled enterprises accounts for an average of 20.1% of the total industrial output value, indicating that the state-owned economy holds a certain share in the industrial economy. The standard deviation of 0.23 indicates a moderate degree of dispersion in the ownership structure among the samples, with certain differences in the proportion of the state-owned economy in different industries. The significant gap between the minimum value of 0.006 and the maximum value of 0.994 reflects significant disparities in the status of state-owned and state-controlled enterprises across industries. The mean value of human capital (HUM) is 0.036, suggesting that the average number of scientific and technological personnel in each industry constitutes only 3.6% of the total employment in that industry, indicating a relatively low investment in scientific and technological talents across the entire industry. The relatively small standard deviation of 0.028 indicates that the proportion of scientific and technological personnel varies minimally among different industries, with a relatively concentrated distribution. The small gap between the minimum value of 0.002 and the maximum value of 0.119 further validates the relatively uniform distribution of human capital across industries and the overall low level. The mean value of capital intensity (FIN) is 23.356, reflecting the overall capital intensity of the industry. The relatively large standard deviation of 22.282 indicates significant differences in capital intensity among different industries. Some industries may be capital-intensive, while others are relatively less so. The significant gap between the minimum value of 2.204 and the maximum value of 118.682 highlights the high degree of dispersion of capital intensity in the samples. The mean value of foreign direct investment (FDI) is 0.26, meaning that the industrial output value of foreign-funded enterprises accounts for an average of 26% of the total industry output value, suggesting that foreign direct investment exerts a certain influence in the industry. The standard deviation of 0.151 indicates a moderate degree of dispersion in the proportion of foreign direct investment among the samples, with certain differences in the ability to attract foreign investment among different industries. The large gap between the minimum value of 0.001 and the maximum value of 0.774 reflects the unbalanced distribution of foreign direct investment in different samples.

4.2. Correlation Analysis

To analyze the correlation between the digital economy level and the total factor productivity (TFP) of the manufacturing industry, this study carried out a correlation analysis on the employed variables. The specific analysis results are presented in

Table 4. Correlation analysis.

	TFP	DIGT	COST	INNO	SIZE	SOE	HUM	FIN	FDI
TFP	1
DIGT	0.193 ***	1
COST	−0.270 ***	−0.626 ***	1
INNO	0.069 *	0.312 ***	−0.099 **	1
SIZE	0.138 ***	0.186 ***	−0.496 ***	−0.083 **	1
SOE	0.093 **	−0.139 ***	−0.079 **	−0.021	0.749 ***	1
HUM	0.095 **	0.638 ***	−0.374 ***	0.783 ***	0.198 ***	0.106 ***	1
FIN	0.193 ***	0.456 ***	−0.465 ***	−0.025	0.645 ***	0.486 ***	0.359 ***	1
FDI	−0.098 **	−0.371 ***	0.273 ***	0.078 *	−0.351 ***	−0.431 ***	−0.185 ***	−0.491 ***	1

Standard errors in brackets, * p < 0.1, ** p < 0.05, *** p < 0.01.

. The following can be seen from the table:

From the perspective of explanatory variables and mediating variables, the correlation coefficient between the digital economy level (DIGT) and TFP is 0.193, positively correlated and significant at the 1% level. This indicates that the enhancement of the digital economy level has a positive facilitating effect on total factor productivity. As the degree of digital economic development rises, it may drive the increase in manufacturing total factor productivity through promoting technological innovation and optimizing resource allocation. The correlation coefficient between the cost and expense ratio (COST) and TFP is −0.270, negatively correlated and significant at the 1% level. This implies that the higher the cost and expense ratio, the lower the total factor productivity. Excessive costs and expenses will constrict the resources available for technological investment and innovation, hampering the improvement of total factor productivity; conversely, industries with higher total factor productivity can control costs and expenses more effectively. The correlation coefficient between the proportion of research and development expenses (INNO) and TFP is 0.069, positively correlated and significant at the 10% level. Although the correlation is relatively weak, it still suggests that an increase in the proportion of research and development expenses has a certain positive influence on total factor productivity. Increasing research and development investment contributes to promoting technological progress and innovation, which is in line with the technological support required for the improvement of total factor productivity, thereby facilitating the enhancement of total factor productivity. The correlation coefficient between the digital economy level (DIGT) and the cost and expense ratio (COST) is −0.626, negatively correlated and significant at the 1% level. This indicates that there is a reverse relationship between the two, meaning that the higher the digital economy level, the lower the cost and expense ratio; conversely, the lower the digital economy level, the higher the cost and expense ratio. The application of digital technology can enhance the operational efficiency of enterprises and promote the rapid circulation of information and the efficient allocation of resources, enabling enterprises to better coordinate in procurement, production, and sales, reducing operating expenses and unnecessary losses in the main business costs. The correlation coefficient between the digital economy level (DIGT) and the proportion of research and development expenses (INNO) is 0.312, positive, and significant at the 1% level. This suggests that there is a significant positive correlation between the digital economy level and the proportion of research and development expenses, that is, the higher the degree of digital economic development, the higher the proportion of research and development expenses in the total output value of each industry. The digital economy brings new technologies and innovation models, improves the innovation ecological environment, and promotes innovation mechanisms such as industry–university–research cooperation, which is conducive to enterprises obtaining more research and development resources and support. At the same time, it broadens the market scope and intensifies market competition. In order to stand out in the fierce competition, enterprises need to continuously increase research and development efforts and launch more competitive products and services, thereby driving up the proportion of research and development expenses.

In terms of control variables, the correlation coefficient between manufacturing scale (SIZE) and TFP is 0.138, which is positive and significant at the 1% level. This indicates that larger manufacturing enterprises tend to have higher total factor productivity. Large-scale firms can achieve economies of scale, resulting in cost advantages across procurement, production, and sales. They also possess more resources for technological research and development, equipment upgrades, and management innovation, thereby enhancing overall production efficiency.

The correlation coefficient between ownership structure (SOE) and TFP is 0.093, significant at the 5% level. This suggests that an increase in the proportion of state-owned and state-controlled enterprises’ output in the total industrial output is associated with higher total factor productivity. State-owned enterprises often hold strategic positions in key industries and benefit from policy support and resource allocation. These advantages enable them to conduct large-scale R&D and infrastructure projects, promoting industrial upgrading and positively impacting total factor productivity.

The correlation coefficient between human capital (HUM) and TFP is 0.095, significant at the 5% level. This indicates that a higher proportion of scientific and technological personnel in an industry’s workforce is associated with higher total factor productivity. As the core drivers of innovation, these professionals contribute to technological and management advancements. High-quality human capital can effectively apply new technologies and processes, increasing the technological sophistication and intelligence of production, thus boosting total factor productivity.

The correlation coefficient between capital intensity (FIN) and TFP is 0.193, significant at the 1% level. This suggests that a higher ratio of net fixed assets to the number of employees in the manufacturing sector is associated with higher total factor productivity. Higher capital intensity implies greater investment in fixed assets, leading to improved automation and precision in production. This can also facilitate technology adoption and innovation, further enhancing total factor productivity.

The correlation coefficient between foreign direct investment (FDI) and TFP is −0.098, significant at the 5% level. This indicates that a higher proportion of foreign-funded enterprises’ output in the total industrial output is associated with lower total factor productivity. While FDI can bring advanced technologies and management practices, it may also have negative effects. For instance, foreign firms may prioritize short-term profits over long-term innovation and have limited impact on local technological advancement and industrial upgrading. Additionally, increased competition from foreign firms can negatively affect local enterprises, reducing their production efficiency and lowering the overall industry’s total factor productivity.

In summary, there are varying degrees and directions of correlations between each variable and total factor productivity (TFP). These correlations reflect the complex interplay of various factors influencing TFP and provide valuable insights for further research into the determinants and improvement paths of total factor productivity. It should be noted that these correlations represent only linear relationships between two variables. The specific multiple linear correlation relationships require empirical regression analysis for verification.

4.3. Multivariate Regression Analysis

Table 5. Results of the multiple regression analysis.

	Regression 1	Regression 2
	TFP	TFP
DIGT	0.002 ***	0.003 ***
	[0.000]	[0.001]
SIZE		−0.006 **
		[0.003]
SOE		0.090 ***
		[0.028]
HUM		−0.552 ***
		[0.159]
FIN		0.0001
		[0.000]
FDI		0.100 ***
		[0.028]
Cons	0.992 ***	0.929 ***
	[0.007]	[0.017]
Sigma_U	0.001	0.001
	[0.005]	[0.005]
Sigma_E	0.063 ***	0.062 ***
	[0.004]	[0.003]
N	621	621

Standard errors in brackets, ** p < 0.05, *** p < 0.01.

presents the outcomes of the multivariate regression. Considering that the majority of the explained variables are distributed within the range of 0 to 1, for the purpose of guaranteeing the validity and accuracy of the estimation results, referring to the research approach [40], this article employs the Tobit model for analysis. The Tobit model showcases distinctive advantages when addressing such issues of limited dependent variables, being capable of effectively circumventing the potential biases that might arise in traditional linear regression models under such circumstances, thereby offering more reliable result support for the research.

To conduct an in-depth analysis of the robustness of the influence of the digital economy level on the total factor productivity of the manufacturing sector, this paper undertakes two rounds of regression analyses. In Regression 1, the model incorporates only the core explanatory variable of the digital economy level, with the aim of initially exploring the fundamental correlation between the digital economy level and the total factor productivity of the manufacturing sector, serving as the basis for subsequent in-depth exploration, whereas Regression 2 is a further expansion based on Regression 1 by adding multiple control variables. In this way, it is possible to more comprehensively and precisely assess the combined effect of various factors on the total factor productivity of the manufacturing sector, making the research findings more practically explicable and credible.

After systematically analyzing the regression results, it was found that in both regression models, the coefficients of the digital economy level (DIGT) and total factor productivity (TFP) were significantly positive. Specifically, in Regression 1, the coefficient for DIGT was 0.002, and in Regression 2, it increased to 0.003, with both coefficients significant at the 1% level. This indicates that, holding other factors constant, an increase in the level of digital economy development exerts a robust positive impact on the total factor productivity of the manufacturing sector. The digital economy, leveraging its advanced technological innovation capabilities—particularly through the widespread application of big data, cloud computing, and artificial intelligence—has substantially optimized resource allocation efficiency and facilitated the transformation of production processes toward greater intelligence and automation, thereby significantly enhancing the total factor productivity of the manufacturing sector.

Regarding the control variables, the regression coefficients of the state-owned ratio (SOE) and foreign direct investment (FDI) in relation to total factor productivity (TFP) are both conspicuously positive. Among them, the regression coefficient of the state-owned ratio is 0.090 and that of foreign direct investment is 0.100, both being significant at the 1% level. This implies that under identical conditions, an elevation in the proportion of state-owned elements within the industry and an expansion of the scale of foreign direct investment can significantly facilitate the enhancement of total factor productivity in the manufacturing sector. State-owned enterprises, leveraging their substantial financial prowess and potent technological research and development capabilities, as well as comprehensive infrastructure construction capabilities, play a leading and exemplary role in the manufacturing industry, driving the entire industry to achieve technological advancement and efficiency improvement. The augmentation of foreign direct investment not only infuses ample funds into the manufacturing industry but also introduces advanced management experience and frontier technologies, bringing new vitality to the industry and forcefully promoting the improvement of total factor productivity.

However, the regression coefficients of manufacturing scale (SIZE) and human capital (HUM) with respect to total factor productivity (TFP) are significantly negative, with values of −0.006 and −0.552, respectively. This indicates that under the current model specification and sample data conditions, both manufacturing scale and human capital exert an inhibitory effect on the total factor productivity of the manufacturing sector. The inverse relationship between manufacturing scale and TFP may be attributed to several adverse effects associated with scale expansion. Firstly, an enlarged scale can lead to diseconomies of scale, resulting in a substantial increase in management and coordination costs, which offset the benefits of economies of scale [41]. Secondly, resource misallocation may occur, where some enterprises experience either excess or insufficient resources, thereby reducing overall efficiency. Scale expansion can also dampen competition, weakening the impetus for innovation and efficiency improvements. Moreover, if technological updates lag behind, it becomes challenging to overcome efficiency bottlenecks, ultimately leading to a decline in TFP as the manufacturing scale increases. Additionally, there may be a mismatch between the structure of human capital and the actual needs of the manufacturing industry. Highly educated talents might not be optimally allocated to positions within the manufacturing sector where they can best utilize their professional skills, leading to inefficiencies and wasted human resources. Furthermore, flaws in the industry’s internal talent cultivation and incentive mechanisms may prevent the full realization of human capital’s potential, causing human capital not only to fail in positively promoting TFP but also to exhibit an inhibitory trend.

4.4. Mechanism Analysis

To conduct a more comprehensive and in-depth exploration of the influence mechanism through which the level of the digital economy acts on the total factor productivity of the manufacturing sector, this study incorporates two crucial mediating variables, namely, the cost–expense ratio and innovation and R&D, and constructs a series of regression models for analysis. The analysis results are presented in

Table 6. Results of mediation mechanism analysis.

	Regression 1	Regression 2	Regression 3	Regression 4
	COST	TFP	INNO	TFP
COST		−0.015 ***
		[0.005]
INNO				0.029 ***
				[0.009]
DIGT	−0.009 ***	0.003 ***	0.014 ***	0.003 ***
	[0.003]	[0.001]	[0.002]	[0.001]
SIZE	−0.253 ***	−0.012 ***	−0.101 ***	−0.004
	[0.029]	[0.004]	[0.020]	[0.003]
SOE	1.334 ***	0.117 ***	0.409 ***	0.076 ***
	[0.223]	[0.030]	[0.141]	[0.028]
HUM	−2.190 *	−0.549 ***	6.492 ***	−1.160 ***
	[1.185]	[0.159]	[0.913]	[0.254]
FIN	−0.001	0.001	−0.006 ***	0.001
	[0.002]	[0.000]	[0.001]	[0.000]
FDI	0.123	0.103 ***	0.141	0.085 ***
	[0.231]	[0.028]	[0.161]	[0.028]
Cons	1.200 ***	0.961 ***	0.426 ***	0.923 ***
	[0.124]	[0.020]	[0.103]	[0.017]
Sigma_U	0.203 ***	0.001	0.343 ***	0.001
	[0.038]	[0.005]	[0.058]	[0.005]
Sigma_E	0.305 ***	0.062 ***	0.161 ***	0.061 ***
	[0.014]	[0.003]	[0.006]	[0.003]
N	621	621	621	621

Standard errors in brackets, * p < 0.1, *** p < 0.01.

.

Firstly, the regression scenario where the cost–expense ratio serves as the mediating variable is inspected. The outcomes of Regression 1 distinctly reveal that a significant negative correlation exists between the level of the digital economy and the factor cost of the industry. This implies that the vigorous advancement of the digital economy can effectively and tangibly reduce the factor cost that the industry demands during the production and operational procedures. The digital economy thoroughly leans on cutting-edge and advanced technologies such as big data and cloud computing to precisely manage the supply chain and realize efficient and rational allocation of resources [42]. In this manner, unnecessary intermediate links in the production process have been successfully curtailed, and a considerable amount of resource waste has been evaded, thereby fundamentally lowering the factor cost in all aspects of production and operation. The data from Regression 2 indicate that the escalation of factor cost exerts a marked inhibitory effect on total factor growth. Concurrently, although the regression coefficient of the digital economy level remains conspicuously positive, compared to the situation in Regression 1, its coefficient value has diminished. This alteration implies that the cost–expense ratio has played a partial mediating role throughout the entire process where the level of the digital economy influences total factor growth. That is to say, the impact of the digital economy on total factor growth has infused positive impetus through the indirect path of reducing the industry’s factor cost.

Next, the regression analysis with innovation and R&D serving as the mediating variable is inspected. The outcomes of Regression 3 explicitly suggest that the level of the digital economy significantly elevates the proportion of industry R&D expenditure, and a significant positive correlation exists between the level of the digital economy and the level of industry innovation. Through its distinctive advantages, the digital economy is conducive to fostering a vigorous and creative innovation ecological environment [43]. In this auspicious ecological environment, enterprises can access more diverse and abundant innovation resources and possess more platforms for display and communication. These favorable conditions markedly accelerate the dissemination and sharing speed of knowledge and technology within the industry, comprehensively stimulating the innovation vitality of the entire industry and compelling enterprises to continuously enhance their innovation levels. Regression 4 reveals that the innovation level exerts a significant facilitating effect on total factor growth. Simultaneously, although the regression coefficient of the digital economy level remains significantly positive, its value has decreased compared to Regression 3. This phenomenon amply demonstrates that innovation and R&D also play an irreplaceable partial mediating role in the process of the digital economy level influencing total factor growth. That is, the digital economy indirectly promotes total factor growth by enhancing the innovation level of the industry, injecting a continuous impetus into the development of the manufacturing industry.

In conclusion, through systematic regression analyses of the two mediating variables, namely, the cost and expense ratio and innovation and R&D, it is affirmed that in the process of the digital economy level influencing the total factor growth of the manufacturing industry, there indeed exist partial mediating effects of cost reduction and innovation improvement.

4.5. Robustness Tests

To ensure the reliability and stability of the research conclusions, this study employs three methods—replacing the explained variable, lagging the explanatory variable, and excluding the pandemic samples—to conduct robustness analysis on the main regression results. The analysis results are reported in

Table 7. Results of robustness test.

	Regression 1	Regression 2	Regression 3
	TFP_CCR	F.TFP	TFP
DIGT	0.004 **	0.003 ***	0.004 ***
	[0.002]	[0.001]	[0.001]
SIZE	0.041 *	−0.009 ***	−0.007 **
	[0.022]	[0.003]	[0.003]
SOE	0.174	0.080 ***	0.097 ***
	[0.112]	[0.029]	[0.030]
HUM	−1.067 *	−0.555 ***	−0.585 ***
	[0.574]	[0.163]	[0.167]
FIN	−0.004 ***	0.000 *	0.000
	[0.001]	[0.000]	[0.000]
FDI	0.210 *	0.102 ***	0.102 ***
	[0.123]	[0.029]	[0.029]
Cons	0.991 ***	0.935 ***	0.927 ***
	[0.065]	[0.018]	[0.018]
Sigma_U	0.001	0.001	0.001
	[0.009]	[0.006]	[0.006]
Sigma_E	0.114 ***	0.062 ***	0.063 ***
	[0.004]	[0.004]	[0.004]
N	621	594	594

Standard errors in brackets, * p < 0.1, ** p < 0.05, *** p < 0.01.

.

Replacement of the Explained Variable: In Regression 1, the original explained variable, total factor productivity (TFP), was replaced with total factor productivity under the assumption of constant returns to scale (TFP_CCR) in this paper. From the regression results, the regression coefficient of the digital economy level (DIGT) was 0.004 and was significantly positive at the 5% significance level. This indicates that even when the explained variable was changed, the digital economy level still had a significant positive promoting effect on the new explained variable, suggesting that the improvement of the digital economy level could drive the growth of economic indicators related to total factor productivity, preliminarily verifying the robustness of the original conclusion.

Lagging of the Explained Variable: In Regression 2, the explained variable was lagged by one period and set as F.TFP in this paper. At this time, the regression coefficient of the digital economy level (DIGT) was 0.003 and was significantly positive at the 1% significance level, once again demonstrating that the digital economy level had a significant promoting effect on total factor growth and further supporting the robustness of the original conclusion.

Exclusion of Epidemic Samples: In Regression 3, the epidemic samples were excluded and the regression analysis was conducted in this paper. At this time, the regression coefficient of the digital economy level (DIGT) was 0.004 and was significantly positive at the 1% significance level, still proving that the digital economy level had a significant promoting effect on total factor growth and supporting the robustness of the original conclusion.

Through the three robustness test methods of replacing the explained variable, lagging the explained variable, and excluding the epidemic samples, the regression coefficient of the digital economy level remained significantly positive under different model settings. This fully indicates that the conclusion drawn in the main regression analysis regarding the relationship between the digital economy level and the growth of total factor productivity in manufacturing is not affected by the selection of variables and model settings. The core conclusion that the digital economy level promotes the growth of total factor productivity in manufacturing has high credibility and stability.

4.6. Endogeneity Examination

In the investigation of the influence of the digital economy level on total factor productivity, to tackle the possible endogeneity problem, this research adopted the internet penetration rate [44] and the lagged cross-product term of the digital economy as instrumental variables and carried out a two-stage regression analysis. The analysis results are presented in

Table 8. Endogeneity examination: two-stage model.

	First-Stage	Second-Stage
	DIGT	TFP
DIGTIV	0.966 ***
	[0.014]
DIGT		0.002 ***
		[0.001]
SIZE	0.826 ***	−0.003
	[0.080]	[0.003]
SOE	−8.525 ***	0.071 **
	[0.702]	[0.032]
HUM	14.811 ***	−0.343 *
	[4.184]	[0.176]
FIN	0.008	0.001
	[0.006]	[0.000]
FDI	−5.566 ***	0.054 *
	[0.729]	[0.031]
Cons	14.279 ***	0.971 ***
	[0.334]	[0.019]
F	2939.492	6.07
R2_a	0.967	0.052
Idstat		529.405
Widstat		4810.9
N	594	594

Standard errors in brackets, * p < 0.1, ** p < 0.05, *** p < 0.01.

.

In the first-stage regression, the emphasis is placed on investigating the correlation between the instrumental variable and the endogenous explanatory variable (the digital economy level, DIGT). The outcomes reveal that the regression coefficient of the instrumental variable (DIGTIV) is 0.966, highly significant at the 1% significance level. This initially suggests a robust correlation between the chosen instrumental variable and the endogenous explanatory variable. Simultaneously, the F value of the model is as high as 2939.492, further powerfully attesting to a strong correlation between the instrumental variable and the endogenous explanatory variable, fulfilling the correlation condition of the instrumental variable and indicating that the instrumental variable is effective and potent.

The second-stage regression is primarily employed to test the validity of the instrumental variable and determine the influence of the digital economy level on total factor productivity. The Anderson canon. corr. LM value is 529.41, significant at the 1% level. This implies that at the 1% significance level, the model can reject the null hypothesis of “insufficient identification of instrumental variables”, further demonstrating a strong correlation between the instrumental variable and the explanatory variable, once again satisfying the correlation condition of the instrumental variable. The Cragg–Donald Wald F value is 4810.9, also significant at the 1% level. This finding indicates a very strong correlation between the instrumental variable and the endogenous explanatory variable, with no issue of weak instrumental variables, guaranteeing the reliability of the estimation results. At this juncture, the regression coefficient of the digital economy level (DIGT) is 0.002, significant at the 1% significance level, and this coefficient is essentially consistent with the basic regression. This indicates that after considering and addressing the endogeneity issue, the impact of the digital economy level on total factor productivity remains stable, and the conclusion previously drawn in this paper regarding the relationship between the digital economy level and total factor productivity is relatively reliable.

Through the aforementioned first-stage and second-stage regression analyses, the interaction term of the lagged one-period Internet penetration rate and digital economy level, selected as an instrumental variable in this study, satisfies the relevance condition for instrumental variables and demonstrates no evidence of weak instruments. Moreover, after addressing the endogeneity issue, the direction and significance of the impact of the digital economy level on total factor productivity remain largely consistent with the baseline regression results. This indicates that the main regression findings of this study exhibit robustness and reliability, effectively mitigating potential endogeneity concerns and providing a solid empirical foundation for further research and policy formulation.

This paper further utilizes the GMM model for endogeneity examination, and the regression outcomes are presented in

Table 9. Endogeneity examination: outcomes of the GMM model.

	(1)
	TFP
L.TFP	0.154 ***
	[0.042]
DIGT	0.004 ***
	[0.001]
SIZE	−0.004
	[0.005]
SOE	0.119 *
	[0.061]
HUM	−1.063
	[0.652]
FIN	0.001
	[0.000]
FDI	0.058
	[0.110]
Cons	0.803 ***
	[0.057]
N	594

Standard errors in brackets, * p < 0.1, *** p < 0.01.

. Judging from the regression results, the main regression results have no endogeneity problems, and specific analyses can be conducted from the following aspects. It can be observed from the regression results that the coefficient of L.TFP (the lagged one-period value of the explained variable, total factor productivity) is 0.154 ***, which is also highly significant at the 1% significance level. This reflects the dynamic lag effect of the explained variable itself, namely, the previous total factor productivity has a significant positive influence on the current period. Meanwhile, its standard error is [0.042], guaranteeing the estimation accuracy. This reasonable estimation of the lag effect further indicates the rationality of the model specification and variable relationships and indirectly reflects that the main regression results have not been severely distorted by endogenous factors. The coefficient of the core explanatory variable DIGT is 0.004 ***, which is significant at the 1% significance level. This indicates that, after controlling for other factors, DIGT has a significant positive impact on the explained variable. Furthermore, the standard error of this coefficient estimate is relatively small ([0.001]), suggesting that the estimation result is relatively precise. This outcome is consistent with the expected direction and significance of the impact of the DIGT variable in the main regression, and no coefficient estimation bias caused by endogeneity issues has emerged. This initially implies that the main regression results possess a certain degree of robustness and are not significantly disrupted by endogenous factors.

4.7. Heterogeneity Analysis

During the exploration of the influence of digital economy development on total factor productivity, considering that the characteristics of different industries may result in variations in the impact effects, this study conducted a meticulous division of the samples. Based on industry characteristics, they were classified into labor-intensive industries, capital-intensive industries, and technology-intensive industries. Regressions were re-performed for each industry sample with the aim of comprehensively dissecting the mechanism of the digital economy’s impact on total factor productivity in different industry contexts. The analysis results are reported in

Table 10. Results of heterogeneity analysis.

	Regression 1	Regression 2	Regression 3
	Labor-Intensive	Capital-Intensive	Technology-Intensive
DIGT	0.003 **	0.003	0.004 ***
	[0.001]	[0.002]	[0.001]
SIZE	−0.004	−0.008	−0.010 *
	[0.007]	[0.008]	[0.005]
SOE	0.065	0.042	0.133 ***
	[0.069]	[0.090]	[0.041]
HUM	−0.821	0.222	−0.549 **
	[0.733]	[0.771]	[0.231]
FIN	0.001	0.001	0.001
	[0.001]	[0.001]	[0.000]
FDI	0.115 *	0.067	0.112 ***
	[0.060]	[0.121]	[0.038]
Cons	0.936 ***	0.935 ***	0.910 ***
	[0.037]	[0.056]	[0.027]
Sigma_U	0.001	0.001	0.000 **
	[0.007]	[0.007]	[0.000]
Sigma_E	0.061 ***	0.063 ***	0.062 ***
	[0.006]	[0.008]	[0.005]
N	230	115	276

Standard errors in brackets, * p < 0.1, ** p < 0.05, *** p < 0.01.

.

Labor-Intensive Industries: The outcome of Regression 1 reveals that in labor-intensive industries, the regression coefficient of the digital economy level (DIGT) amounts to 0.003, which is significantly positive at the 5% significance level. This indicates that the development of the digital economy exerts a significant promotional effect on the total factor productivity of labor-intensive industries. The production activities within labor-intensive industries rely highly on a substantial number of labor forces, and the deep integration of the digital economy has brought forth numerous alterations [45]. With the aid of intelligent management systems, enterprises can precisely allocate labor resources based on real-time order information and production progress, rationally arrange the tasks and working hours for each worker, and effectively mitigate the phenomena of labor idleness and waste. The introduction of automated production equipment further elevates the degree of mechanization and automation in the production process, reduces errors and inefficiencies in manual operations, and renders the production process more seamless and efficient.

Capital-Intensive Industries: The results of Regression 2 suggest that in capital-intensive industries, the regression coefficient of the digital economy level is 0.003, which is not significant. This implies that, as of now, no significant influence of the development level of the digital economy on the total factor productivity of this industry has been identified. The production and operational characteristics of capital-intensive industries determine their heavy reliance on large-scale capital investment, with a considerable amount of funds being utilized for the procurement of advanced large-scale equipment and the construction of specialized large-scale infrastructure, etc. [46]. Once the production mode of such industries is established, it typically exhibits high stability and fixity. The low responsiveness of capital-intensive industries to the digital economy may be explained by the deep-seated contradiction between their structural characteristics and the diffusion path of digital technologies. First, high sunk costs and asset specialization create “technology lock-in”: the physical irreversibility of heavy industrial equipment (e.g., steelmaking and chemical industry) forces companies to extend the lifecycle of traditional technologies, and the marginal benefits of digital transformation take decades to cover the costs, much higher than the technology iteration cycle of the service sector. This is much higher than the technology iteration cycle in the service industry. Second, technology adaptation barriers exacerbate transformation resistance; digital twins, industrial Internet of Things, and other technology architectures are difficult to be compatible with closed industrial control systems; data integration requires customized middleware, significantly pushing up the cost of transformation. Third, organizational inertia and skills mismatch constitute institutional barriers, vertical management structure impedes cross-departmental collaboration, and the lack of employees’ digital skills and internal organizational resistance to automation further slows down the process. In addition, government policies may be a potential influencing factor. Existing tax incentives may not provide a dedicated compensation mechanism, leading to a lower willingness of enterprises to digitally transform; the actual access of heavy enterprises to dedicated funding for digital transformation is much lower than the expected and actual costs, thus also reducing the adoption of digital technologies.

Technology-Intensive Industries: The results of Regression 3 indicate that in technology-intensive industries, the regression coefficient of the digital economy level is 0.004, which is significantly positive at the 1% significance level, suggesting that the development of the digital economy has a highly significant positive impact on the total factor productivity of this industry. Technology-intensive industries consistently place technological innovation and R&D investment at the core, and the emergence of the digital economy has created exceptional development conditions for them [47]. Vast data resources are akin to treasures. Through big data analysis techniques, enterprises can accurately perceive the minute changes in market demand, proactively plan R&D directions, and develop new products and technologies that better align with market requirements. Advanced information technology has broken geographical constraints, accelerating global technological exchanges and cooperation. Enterprises can engage in real-time communication and collaboration with research institutions, universities, and counterparts worldwide, sharing the latest scientific research achievements and technical experiences. Efficient innovation platforms furnish convenient channels for resource integration, facilitating the collision of ideas and the integration of technologies, and expediting the process of transforming new technologies from R&D to application. Take a certain artificial intelligence chip R&D enterprise as an example. By leveraging the big data analysis provided by the digital economy, it can profoundly understand the demand trends for chip performance and functions in the market. Simultaneously, through online technical exchange platforms, it can collaborate with top international research teams to promptly overcome technical difficulties, shorten the R&D cycle, and enhance product performance and quality, thereby significantly elevating production efficiency and product added value and effectively driving the growth of total factor productivity.

From the results, the heterogeneity analysis based on industry characteristics discloses significant disparities in the impact of digital economy development on the total factor productivity of various types of industries. In technology-intensive industries, the digital economy demonstrates the most vigorous promotional effect, followed by labor-intensive industries, while in capital-intensive industries, this influence remains indistinct.

5. Summary

This paper employs the data on the development level of the digital economy and the manufacturing sector from 2001 to 2023 to investigate the influence of the digital economy on the total factor productivity of the manufacturing industry. The principal conclusions of this article are as follows: Firstly, the development of the digital economy can enhance the total factor productivity of the manufacturing industry. Secondly, costs and expenses, as well as innovation and research and development, exert mediating effects in the process through which the digital economy elevates the total factor productivity of the manufacturing industry. Thirdly, the enhancement effect of the digital economy on the total factor productivity of the manufacturing industry varies among different sub-sectors. The digital economy has shown a more positive push in technology-intensive and labor-intensive industries than in capital-intensive ones.

Based on the research findings, the following recommendations are proposed for the manufacturing industry:

Firstly, the manufacturing sector should fully attach great importance to the substantial driving force of the digital economy. It should actively increase investment in the domain of digital technologies, initiatively introduce cutting-edge technologies such as big data, cloud computing, and artificial intelligence, and deeply integrate them into all aspects of daily operations. By utilizing these technologies to precisely analyze market demands and its own operational status, it can further optimize resource allocation, reduce resource waste, and improve resource utilization efficiency. It should also promote the intelligent and automated upgrading and transformation of production processes by means of digital technology so as to conform to the development trend of the digital economy, fully exert the positive promotion effect of the digital economy on total factor productivity, enhance competitiveness, and achieve sustainable development.

Secondly, the manufacturing industry should make full use of digital economy measures to reduce costs, such as constructing a digital management system, precisely planning resource allocation, minimizing unnecessary expenditures, and optimizing supply chain management to lower procurement and logistics costs through information sharing and collaboration with partners. The ERP system coverage rate of China’s above-scale manufacturing enterprises is only 54% in 2023 (Ministry of Industry and Information Technology’s Intelligent Manufacturing Development Report), leading to inefficiency in production planning and inventory management. Supply Chain Cost Ratio: procurement and logistics costs account for 18.7% of revenue in the traditional model, higher than the level of similar enterprises in Germany (11.2%) and Japan (13.5%) (McKinsey 2023 Global Supply Chain Benchmark Study). Haier Group’s report in 2023 indicates that the networking rate of equipment through MES system has increased from 35% to 82%, and production costs have been reduced by 19%; enterprises applying blockchain technology have reduced procurement costs by 12–15% (e.g., Sany Heavy Industries reduced costs by 14.3% in 2022) and improved logistics responsiveness by 40%; Midea Group increased raw material utilization from 89% to 94% through its industrial internet platform, resulting in an annual cost savings of more than USD 700 million (Sustainability Report 2024). Simultaneously, it is necessary to vigorously enhance the innovation capability, increase R&D investment, cultivate and attract innovative talents, focus on carrying out innovation activities that integrate digital technologies with manufacturing operations, utilize big data to achieve precise services, and leverage artificial intelligence to improve operational efficiency. In this way, it can seize the opportunities presented by the digital economy, promote total factor growth, and achieve sustainable development.

Finally, labor-intensive industries can further enhance the intensity of digital transformation to fully leverage the advantages of the digital economy in optimizing labor allocation. The introduction of AGV+RFID system in Yagor’s textile factory has reduced labor in sewing by 23%, but the current industry-wide penetration rate of automated equipment is less than 15% (China Textile Industry Federation 2023). After digital transformation, DYG has reduced the unit cost to 24.7% through an intelligent scheduling system, which is better than competing countries such as Vietnam (28%) and Bangladesh (26%). Shenzhou International increased the automation rate of its sewing equipment from 8% to 35% in three years, which is expected to release 20% of redundant labor (about 3.8 million people) and increase labor productivity by 28%; technology-intensive industries should continuously draw on the innovative resources provided by the digital economy to strengthen their innovation capabilities; capital-intensive industries need to explore the digital economy integration models that are suitable for their own characteristics and gradually promote the transformation and upgrading of the production mode. Thus, it can assist different industries in seizing the development opportunities of the digital economy better and achieving the sustainable development goals. After Huawei Hesse adopted EDA cloud computing, the 5 nm chip design cycle was shortened from 18 months to 11 months, with a 39% increase in efficiency. After the application of a digital twin, Changdian Technology controlled the failure rate at 18%, saving more than CNY 1.2 billion in annual trial and error costs. Sanan Optoelectronics’ 2023 financial report disclosed that through the AI patent analysis platform, the industrialization cycle of core patents was compressed from 5.2 years to 3.1 years, and the efficiency of technology transformation was improved by 2.3 times.

6. Limitations and Future Research

This paper is a study on the development of the digital economy and total factor productivity in the manufacturing industry; it found that cost and innovation R&D play a mediating effect and that this enhancement varies across different types of industries. There are also limitations in this study, which are mainly reflected in the following aspects: First, although the study classifies manufacturing industries into technology-intensive, labor-intensive, and capital-intensive industries, this classification fails to fully take into account the differences and complexity within the industries, which may lead to a lack of universality and representativeness in the research results. Second, apart from the digital economy, cost and expenses, and innovation and R&D, other factors that may affect the total factor productivity of the manufacturing industry have not been considered. Third, there are differences in policy environments, laws and regulations, and digital economy infrastructures in different regions, which may affect the applicability of the research results. Moreover, this paper does not take into account potential negative spillovers, such as automation-induced job displacement or the digital divide between firms, diminishing returns or trade-offs.

Future research could adopt a more detailed industry categorization methodology to more accurately reveal the differences in the impact of the digital economy on different manufacturing industry segments. Cross-country comparative studies can be conducted to explore the similarities and differences in the impact of the digital economy on the manufacturing industry in different countries to provide a basis for formulating more targeted policy recommendations. In addition to the digital economy, costs and innovation, and research and development, other factors that may affect total factor productivity in the manufacturing sector, such as the policy environment, market demand, and international trade, could also be considered to build a more comprehensive analytical framework. In addition, the differences in the role of policy for three different types of manufacturing industries in terms of digital adoption can be further researched.

Moreover, due to the Malmquist Index fluctuations deviating from true efficiency changes, future research will incorporate stage analysis, control variables, and robustness tests to further enhance the reliability of causal inferences. In order to reflect the differences in numerical impacts between industries, further optimization will be made to study to highlight the differences in impact effects.