The Impact of Digital Economy on Total Factor Energy Efficiency from the Perspective of Biased Technological Progress

Abstract

Enhancing Total Factor Energy Efficiency (TFEE) is pivotal for achieving China’s “dual carbon” goals and navigating the global challenge of sustainable development. The Digital Economy (DE) serves as a significant driver of TFEE improvement. However, China’s rapid industrialization has exacerbated energy insecurity and environmental degradation, highlighting the need to explore how the DE can address these challenges through biased technological progress. Building on panel data from 282 prefecture-level cities in China (2011–2022), this study employs the theory of biased technological progress to empirically examine the impact of the DE on TFEE from dual perspectives: skill-biased versus task-biased technological change. The findings reveal that the DE significantly enhances TFEE, a conclusion robust to rigorous testing and endogeneity controls; the DE primarily promotes TFEE through facilitating human capital and industrial transformation; the positive effect of the DE on TFEE is particularly pronounced in eastern and western regions, as well as in areas exhibiting moderate energy dependence; and the DE not only elevates local TFEE but also generates positive spatial spillover effects that significantly improve TFEE in neighboring regions. This study provides a framework for leveraging digitalization to enhance TFEE, with implications for policy design in developing countries pursuing sustainable transitions.

1. Introduction

Since China’s reform and opening-up, its industrial sector has achieved remarkable growth, with industrial value-added surged from CNY 162.14 billion in 1978 to CNY 40.02 trillion in 2022—an increase of over 245 times in just over four decades—demonstrating an average annual growth rate of 14.10%. By 2022, large-scale industrial profits reached CNY 8.41 trillion, significantly boosting economic expansion. However, rapid industrialization intensified fossil fuel shortages and environmental degradation. China dominated global energy consumption (24.32%) and CO₂ emissions (28.80%) in 2019, heightening the challenge of achieving its 2030 carbon peak and 2060 neutrality goals amid coal dependency and unabated emissions. Enhancing total factor energy efficiency (TFEE) is thus critical for the green transition. Developing the digital economy (DE) is essential to empower industry transformation, optimize the industrial structure, and facilitate green transitions in China. In 2022, the DE accounted for 41.50% of the GDP, an increase of 2.40 percentage points compared to the previous year. Digital technologies like 5G and big data increasingly empower industrial transformation, structural optimization, and green development—establishing DE advancement as a pivotal pathway toward sustainable growth [1].

At present, research on the DE, biased technological progress, and TFEE mainly focuses on four aspects: (1) The impact of the DE on biased technological progress. Some studies have pointed out that in the relationship between capital and labor factors, artificial intelligence, the Internet, and other technologies are conducive to promoting labor-biased technological progress [2]. With the deepening of research on the DE, other studies have subdivided labor into skilled labor and unskilled labor, and found that the DE will promote technological progress biased towards skilled labor [3]. Furthermore, the discussion on the direction of technological progress has been extended to the environmental field. Kruse-Andersen (2023) emphasized the necessity of directing technological change towards environmentally friendly technologies for sustainable growth [4]. Building on this, Wang et al. (2024) empirically tested that DE-driven technological progress exhibits a significant skill-biased characteristic, which is closely associated with improvements in environmental quality [5]. (2) The impact of biased technological progress on TFEE. Related studies have found that biased technological progress can significantly promote TFEE improvement [1]. Further research has suggested that there is a correlation between the effect of biased technological progress on TFEE and the elasticity of substitution between factors, while the impact on TFEE in different industries is heterogeneous [6]. However, a systematic empirical analysis that decomposes the impact of the DE on TFEE into distinct pathways of skill-biased and task-biased technological progress remains scarce. (3) The influencing factors of the DE. A substantial body of research in academia examines the factors influencing TFEE. These factors primarily include the environmental regulations [7,8], industrial structure [9], energy prices [10], marketization level [11], technological progress [12], foreign direct investment [13], green finance [14], and other related factors. (4) The impact of the DE on TFEE. Scholars have found that the DE acts as a catalyst for enhancing green TFEE [1,15]. From a micro perspective, scholars have elucidated how information and communication technology (ICT) can elevate the digitalization level of communication systems and enhance TFEE [16]. While ICT may reduce energy consumption and promote the development of new energy sources, it can also cause corresponding declines in energy prices, thus affecting the market demand for new energy sources with limited net economic benefits [17]. However, the positive impacts the DE brings, such as advancements in technology innovation and TFEE, have the potential to offset the underlying negative effects completely [18,19]. Li et al. (2023) further found that the capital-biased technological progress brought about by digital finance is beneficial for improving TFEE and transforming energy structure [16].

Existing research provides a reference for understanding the relationship between the DE, biased technological progress, and TFEE. However, there is no consensus on the impact of the DE and TFEE, as well as the role and path of implementation. This may be related to the neglect of differences in the direction of technological progress in the DE. The theory of biased technological progress internalizes technological progress [20], extending it from focusing on the proportional allocation of capital and labor factors to analyzing the relevant mechanisms through the differences in output increases of different types of skilled labor and developing two perspectives: skill-biased technological progress and task-biased technological progress [21]. The former emphasizes that technological progress requires an increase in labor skills and human capital may affect the relationship between the DE and TFEE. The latter points out that technological progress will change production tasks and promote the migration of low- and medium-skilled labor to unconventional tasks. The progress of DE technology also has a certain directionality, either towards skill-oriented or task-oriented areas, fundamentally affecting labor market bias. However, existing research lacks an explanation for the differential impact of TFEE from the perspectives of technological progress, namely human capital and industrial structure transformation. Based on this, this study explains the reasons for the differential impact of TFEE from the two perspectives of biased technological progress theory, proposes two different paths to achieve this impact, and provides theoretical references for improving TFEE.

The possible marginal contribution of this article lies in the following: (1) From a theoretical perspective, we integrate the DE, biased technological progress, and TFEE into a unified analytical framework. Unlike prior studies that often treat technological progress as neutral, we dissect its dual nature—skill-biased versus task-biased—thereby providing a nuanced explanation for how the DE influences TFEE through human capital accumulation and industrial structure transformation, respectively. (2) Methodologically, we leverage a historical instrumental variable (the number of landlines in 1984) to strengthen causal inference regarding the DE-TFEE relationship, addressing endogeneity concerns more robustly than many correlational studies in the field. (3) This article takes into account the spatial spillover effects of the DE and reveals the dynamic characteristics of its impact on TFEE from a spatiotemporal perspective. We identify an inverted U-shaped temporal pattern in the impact of the DE on TFEE and a precise spatial attenuation boundary of approximately 400 km for these spillover effects. (4) Considering the regional and energy-dependency differences, this paper analyzes the heterogeneous impact of the DE on TFEE. Our analysis reveals that the DE exerts a more substantial influence on advancing TFEE in eastern and western regions. The level of the DE has a positive effect on TFEE in regions with medium energy dependence, whereas regions with high energy dependence experience a negative effect. The conclusions presented herein can offer valuable policy insights for developing nations seeking to optimize their energy consumption patterns and devise pertinent measures aimed at energy conservation and emissions reduction.

2. Theoretical Analysis and Research Hypotheses

2.1. Direct Effects of Digital Economy on Total Factor Energy Efficiency

As an important driving force for China’s current economic and social development, the DE can significantly improve the production efficiency of traditional industries on the basis of the digital technologies [22,23]. Firstly, from the perspective of industrial optimization, ICT can improve resource allocation efficiency in companies, thereby improving TFEE. From the market supply-and-demand perspective, informatization has reduced the asymmetry between market supply and demand, thus making the coordination between the two more effective [24]. The DE facilitates the amalgamation and distribution of digital technology with conventional economic resources, thereby improving TFEE [25]. The DE effectively lowers marginal costs for enterprises and facilitates the adjustment of industrial structure, promoting a profound integration between the digital and industrial economies [26]. Secondly, from the perspective of production methods, the DE’s core lies in information technology’s development and innovation. This progress promotes the amalgamation of the DE with traditional manufacturing, fostering process reengineering in the manufacturing sector and promoting environmentally friendly and low-carbon development [22]. The advancement of the DE accelerates technological innovation for enterprises, improves production efficiency, reduces energy loss, and enhances TFEE [27]. Furthermore, developing the DE enhances international digitalization and informatization, facilitating cross-border innovation and emission reduction technology sharing [28]. Based on this, the following hypotheses are proposed in this study:

H1. 

The DE can promote TFEE.

2.2. The Impact of Digital Economy and Human Capital on Total Factor Energy Efficiency

The skill-biased nature of DE technologies implies that their effective adoption and diffusion necessitate a workforce equipped with advanced skills, thereby increasing the demand for and returns to human capital [29]. Firstly, it is the penetration of digital technology and the choice of enterprise technology. Faced with the wave of the DE, enterprises pursuing profit maximization have the motivation to introduce advanced technologies, such as big data and artificial intelligence, to optimize production processes. However, the effective operation, maintenance, and secondary development of these complex technologies heavily rely on a workforce with corresponding knowledge reserves and skill levels [7]. Therefore, the technology adoption decisions of enterprises includes the demand for high-skilled human capital.

Secondly, it is the accumulation of human capital and the upgrading of TFEE management. On the one hand, regions with higher levels of DE development can create a “siphon effect” on high-end talent by providing more comprehensive information platforms [30] and public infrastructure [31,32]. The gathering of technology talent can effectively enhance the promotion effect of the DE on urban innovation capability, which is mainly reflected in capital-intensive and intelligence-intensive high-tech industries [33]. On the other hand, in order to fill the skills gap, companies will increase their internal investment in digital training for existing employees, while externally utilizing talent introduction policies from other regions [34]. This high-quality talent team is the core driver of technological innovation [35]. They are able to diagnose energy waste points in the production process more accurately, design and execute more optimized energy scheduling algorithms, and promote the transformation of energy management from extensive experiential to data-driven and real-time optimized refined models.

Ultimately, this manifests as an improvement in energy allocation and utilization efficiency. The direct result of process innovation and management optimization driven by digital technology and human capital is more effective allocation and utilization of energy factors in the production process, namely the improvement of TFEE [36]. Based on this, the following hypothesis is proposed:

H2. 

The DE can promote TFEE through human capital.

2.3. The Impact of Digital Economy and Industrial Structure Transformation on Total Factor Energy Efficiency

The DE influences TFEE not only through skill-biased channels but also through task-biased technological progress. This bias manifests as a re-organization of production processes, where technology substitutes for labor in routine, codifiable tasks, often leading to the automation of middle-skilled jobs and a reallocation of labor towards non-routine cognitive and manual tasks [21].

The development of the DE is a catalyst for this transition. Technologies such as AI and robotics automate manufacturing and administrative routines, reducing the relative demand for medium-skilled labor in traditional industrial roles. Simultaneously, the DE itself spawns new service industries and increases the demand for high-skilled service labor. This leads to a macro-economic structural shift—a rise in the share of the tertiary sector in the economy.

Industrial structure upgrading enhances TFEE through two primary pathways. On the one hand, Intrinsic Efficiency of the Service Sector. The tertiary sector is typically less energy-intensive than the industrial sector. Therefore, a shift in economic weight towards services inherently lowers the energy consumption per unit of economic output [37].

On the other hand, Efficiency Spillovers. A more advanced service sector, particularly producer services like R&D and logistics, provides knowledge-intensive inputs that help the entire economy, including the remaining industrial base, to optimize production processes, manage supply chains more efficiently, and reduce energy waste. Based on this, the following hypothesis is proposed:

H3. 

The DE can promote TFEE through industrial structural transformation.

2.4. Spatial Spillover Effects

Based on the networked attributes of the DE and its reshaping effect on factor flows, its technological spillovers may transcend geographical boundaries to form cross-regional synergistic effects. Specifically, the network penetration and data-sharing characteristics of the DE drive the cross-regional transmission of innovation factors, enabling core regions’ technological innovations and management models to radiate to surrounding areas through digital platforms [38]. This spillover mechanism not only optimizes local energy resource allocation efficiency but also triggers chain reactions in neighboring regions’ energy systems through industrial linkages and technology demonstration pathways [39]. The interconnectivity of digital infrastructure accelerates the diffusion of energy-saving technologies, while cloud-based collaborative management platforms reduce transaction costs in cross-regional energy dispatch, generating synergistic efficiency gains [40]. Based on this, the following hypothesis is proposed:

H4. 

The DE can enhance TFEE in neighboring regions through spatial spillover effects.

The research framework diagram is shown in Figure 1.

3. Econometric Model and Data

3.1. Model Construction

3.1.1. Benchmark Regression Model

This paper designs a benchmark regression model to test the impact of the DE on TFEE [41]. The model is as follows:

$$TFE{E}_{it}={\beta }_0+{\beta }_{1}D{E}_{it}+{\beta }_2{X}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(1)

TFEE represents the level of TFEE, DE represents the level of the DE, and x_ijt represents the control variables that may affect TFEE. ${\mu }_{it}$ represents a time-fixed effect, ${\lambda }_{it}$ represents a region-fixed effect, and ${\epsilon }_{it}$ represents a random error term.

3.1.2. Mechanism Verification Model

To test the mediating effects of human capital and industrial structure, the following regression model is constructed [40]:

$$M_{it}={\beta }_0+{\beta }_{1}D{E}_{it}+{\beta }_2{X}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(2)

$$TFE{E}_{it}={\beta }_0+{\beta }_{1}D{E}_{it}+{\beta }_2{M}_{it}+{\beta }_3{X}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(3)

M represents the mechanism variable.

3.1.3. Spatial Durbin Model

To verify the spatial effect of the DE on TFEE, a spatial Durbin model is constructed [42], which has the following formula:

$$TFE{E}_{it}=\rho {\displaystyle \sum _{j=1}^n{W}_{ij}TFE{E}_{jt}}+{\beta }_{1}D{E}_{it}+{\beta }_2{X}_{it}+{\theta }_1{\displaystyle \sum _{j=1}^n{W}_{ij}D{E}_{jt}+}{\theta }_2{\displaystyle \sum _{j=1}^n{W}_{ij}{X}_{jt}+}{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(4)

$X_{it}$ is a vector of control variables. $W_{ij}$ is the element of the spatial weights matrix W, representing the spatial connection between city i and city j. $\rho$ is the spatial autoregressive coefficient, capturing the impact of neighboring cities’ TFEE on the local city’s TFEE. The selection of the spatial weight matrix is directly related to the robustness of the spatial econometric model estimation results and the construction of the spatial geographic distance matrix W (geo). In reality, the spatial connection between two cities may be influenced by the level of economic development and operational mode. In order to better characterize the spatial variation characteristics of variables with increasing economic distance, this paper constructs an economic geographic distance matrix W (eco-geo).

3.2. Variable Measurement

3.2.1. Explained Variable

This study uses a super-efficient EBM model that includes unexpected outputs to measure the TFEE of various prefecture-level cities in China. The specific calculation method is as follows:

$${\rho }^{\ast }=min{\displaystyle \frac{\theta -{\epsilon }_x{\displaystyle {\sum }_{i=1}^m\frac{w_i^{-}{s}_i^{-}}{x_{ik}}}}{\varphi +{\epsilon }_y^g{\displaystyle {\sum }_{T=1}^s\frac{w_r^g{s}_r^{g+}}{y_{rk}^g}+{\epsilon }_y^b{\displaystyle {\sum }_{p=1}^h\frac{w_p^b{s}_p^{b-}}{y_{pk}^b}}}}}$$

(5)

$$s.t.\left\{\begin{array}{l}{\displaystyle \sum _{j=1}^n{\lambda }_j{X}_{ij}}-s_i^{-}\le \theta x_{ik},i=1,...,m\\ {\displaystyle \sum _{j=1}^n{y}_{rj}^g{\lambda }_j}+s_r^{g+}\ge \phi y_{rk}^g,r=1,\cdot \cdot \cdot ,s\\ {\displaystyle \sum _{j=1}^n{y}_{pj}^b{\lambda }_j}-s_p^{b-}=\phi y_{pk}^b,p=1,\cdot \cdot \cdot ,h\\ {\lambda }_j,s_i^{-},s_r^{g+},s_p^{b-}\ge 0,\theta \le 1,\phi \ge 1\end{array}\right.$$

(6)

where ${\rho }^{*}$ represents the TFEE and $\theta$ and $\phi$ are the radial parameters for inputs and outputs, respectively. ${\lambda }_j$ is an intensity variable for linear combination. $s_i^{-}$ is an input slack variable. $s_r^{g+}$ is a desirable output slack variable. $s_p^{b-}$ is an undesirable output slack variable.

(1)

Input Factors:

①

Energy Input: The total energy consumption of each prefecture-level city is used as the input indicator.

②

Capital Input: Fixed capital is a significant input factor in the production process. However, when calculating the actual value of fixed assets, the depreciation component must be excluded. Under the perpetual inventory method, the calculation does not consider depreciation and scrapping issues. The specific formula is as follows:

$$K_t=K_{t-1}\cdot (1-{\delta }_t)+I_t$$

(7)

Here, the base-year capital stock is taken as 2000, and the depreciation rate ${\delta }_t$ is set at 10.96%. $I_t$ represents the annual investment amount.

③

Labor Input: Labor input is a crucial factor in the production process. Labor input involves not only the quantity of labor but also factors such as labor time and labor quality. This study measures labor input by using the number of employed persons per unit at the year’s end.

(2)

Expected Output: Regional gross domestic product (GDP) is an important indicator for measuring the economic development of a region.

(3)

Unexpected Output: There is currently no consensus on the measurement indicators for unexpected output. However, environmental indicators are primarily used as substitutes, including emissions such as exhaust gases, wastewater, solid waste, particulate matter, and CO₂ and SO₂ emissions. Among these, sulfur dioxide (SO₂) is the main direct pollutant emitted during energy consumption. This study chooses sulfur dioxide (SO₂) as the measure of unexpected output. The research focuses on 282 cities in China, and MaxDEA8.0 software is utilized to solve the super-efficiency EBM model for unexpected output.

3.2.2. Core Explanatory Variable

The core explanatory variable of this article is the DE. Regarding measuring the DE, as early as 2014 the European Union utilized five primary indicators, such as broadband access and digital technology applications, along with 31 corresponding secondary indicators. In 2018, the OECD used the satellite account approach to reflect the digital economic development status. Leveraging prior studies [43,44,45,46], this study selects four primary indicators of the DE—digital infrastructure, digital economic development potential, digital economic application capabilities, and the digital economic development environment—along with their corresponding secondary indicators, to measure the DE. The specific measurement indicators are presented in

Table 1. Index system for evaluating the level of the digital economy.

Indicator	Category Criteria	Description of Indicators
Digital Economy	Digital Economy Infrastructure	Ratio of Internet users to total population
		Ratio of mobile phone subscriptions to total population
		Length of long-distance fiber-optic cable lines
	Digital Economy Development Potential	Total number of patent applications and authorizations across various types
		Ratio of R&D expenditure to regional gross domestic product (GDP)
	Digital Economy Application Capability	Total sales revenue of electronic products, such as communication and computing devices
		Total revenue from software services
		Total revenue from telecommunications services
	Digital Economy Development Environment	Digital inclusive finance index
		Total number of employees in industries such as communication and computing

.

3.2.3. Intermediary Variables

Human capital (HUM) is expressed as the proportion of the population with a college degree or above per 100,000 people.

Industrial structure upgrading (IS) is represented by the ratio of the tertiary industry’s value added to GDP.

3.2.4. Control Variables

We select the following control variables based on existing research [15,47,48,49]. Energy Price (PRICE): We use the fuel and power purchase price index as a proxy variable, with a base period conversion to 100 in 2000. Market Segmentation (MS): The market segmentation index is calculated based on the variance of changes in relative prices of industrial products [50]. Foreign Direct Investment (FDI): The logarithm of actual utilized foreign direct investment in each prefecture-level city is used as a proxy variable. Green Finance (GF): The green finance index for each prefecture-level city is calculated using the entropy weighting method [51]. This paper selects panel data of 282 cities above the prefecture level in China from 2011 to 2022 to empirically study the impact and transmission mechanism of the DE on TFEE, as shown in

Table 2. Variable definitions.

Variable Type	Name	Symbol	Definition
Explained Variables	Total Factor Energy Efficiency	TFEE	Use the super-efficiency EBM model to calculate
Explanatory Variables	Digital Economy	DE	Comprehensive indicators constructed based on digital infrastructure, digital economic development potential, digital economic application capabilities, and the digital economic development environment
Intermediary Variables	Human capital	HUM	The proportion of the population with a college degree or above per 100,000 people
	Industrial structure upgrading	IS	The ratio of the tertiary industry’s value added to GDP
Control variables	Energy Price	PRICE	The fuel and power purchase price index
	Market Segmentation	MS	The variance of changes in relative prices of industrial products
	Foreign Direct Investment	FDI	The logarithm of actual utilized foreign direct investment
	Green Finance	GF	Green finance index

.

4. Empirical Findings and Discussion

4.1. Impact of the Digital Economy on Total Factor Energy Efficiency

Table 3. Benchmark regression results.

Variable	(1)	(2)	(3)	(4)
DE	0.772 ***	0.642 ***	0.059 ***	0.047 ***
	(0.058)	(0.122)	(0.011)	(0.007)
W × DE			0.155 ***	0.056 ***
			(0.059)	(0.020)
Control	NO	YES	YES	YES
variables
_cons	0.458 ***	−0.187 *	0.469 ***	0.240 ***
	(0.017)	(0.100)	(0.044)	(0.054)
City	YES	YES	YES	YES
Year	YES	YES	YES	YES
Wald test (SAR)			115.71 ***	52.08 ***
Wald test (SEM)			149.71 ***	68.31 ***
N	3384	3384	3384	3384

Notes: * p < 0.10, *** p < 0.01; numbers in parenthesis are robust standard error.

presents the estimation results of the influence of the DE on TFEE. Among them, columns (1) and (2) are the estimation results of non-spatial panel models. The empirical results demonstrate that the DE exerts a statistically significant positive effect on the enhancement of TFEE, aligning with theoretical expectations and thereby validating Hypothesis H1.

Columns (3) and (4) in Table 3 show the estimation results of the spatial Durbin model. Table 3 shows the regression results of SDM based on the geographic distance matrix and economic distance matrix under the dual fixed conditions of time and city area. Both the geographic distance matrix and economic distance matrix, as well as the spatial error model (SEM) and spatial lag model (SAR) have significant spatial effects. The spatial error model preliminarily supports the introduction of spatial econometric methods. Secondly, based on the Wald test, the Spatial Durbin Model (SDM) refuses to simplify to SEM or SAR models, indicating that SDM can more fully capture the local and spillover mechanisms between variables. The results showed that the DE coefficients in all SDM models were significantly positive, indicating that TFEE has obvious positive spatial interdependence This means that the penetration of digital technology in the local area not only optimizes its own energy utilization system but also promotes the improvement of TFEE in neighboring areas through mechanisms such as technology diffusion, experience sharing, and relevant institutional channels [52], which confirms hypothesis H4.

4.2. Endogeneity Analysis

Although benchmark regression has controlled for a range of variables and fixed individual and time effects, there may still be issues of reverse causality or omitted variables between the DE and TFEE. To alleviate potential endogeneity bias, this study used the instrumental variable method (IV) for estimation. This article draws on relevant research and selects the number of fixed telephones in each city in 1984 as an instrumental variable for the level of urban DE development [53]. This instrumental variable is effective for two reasons: firstly, historical communication infrastructure is the material foundation for the development of a contemporary DE. As an early form of ICT, the popularity of fixed telephone networks reflects the region’s acceptance and first mover advantage of communication technology. This advantage has path dependence, making areas with well-developed early communication facilities more advantageous in building modern digital infrastructure, such as broadband and mobile networks, meeting conditions related to instrumental and endogenous variables. Secondly, the number of fixed telephones in 1984 mainly indirectly affects the current TFEE by influencing the development of the contemporary DE and is unrelated to other unobservable factors that affect the current TFEE. As an early communication tool, the distribution of fixed telephones was mainly influenced by national administrative planning and geographical conditions during the early stages of reform and opening up, and was not directly related to the current market factors and environmental regulatory intensity that affect TFEE.

The results are presented in column (1) of

Table 4. Robustness test results.

Variable	(1)	(2)	(3)	(4)
DE_1	1.016 ***	−0.936 ***
	(0.126)	(0.228)
L.EI			0.891 *	0.909 **
			(0.050)	(0.066)
DE			0.096 ***	0.099 ***
			(0.104)	(0.087)
Kleibergen–Paap rk LM	23.487 ***
Kleibergen–Paap rk Wald F	119.443
Control	YES	YES	YES	YES
variables
City	YES	YES	YES	YES
Year	YES	YES	YES	YES
_cons	0.752 ***	0.289 **	−0.078	−0.070
	(0.136)	(0.135)	(0.067)	(0.047)
N	3384	3384	3384	3384

Notes: * p < 0.10, ** p < 0.05, *** p < 0.01; numbers in parenthesis are robust standard error.

. It can be seen from the results that the DE variables have passed the 1% significance test, and the variable coefficient is positive. This indicates that after considering endogeneity issues in the model, the positive impact of the DE on TFEE still has significance. Moreover, for the test of the null hypothesis “insufficient identification of instrumental variables”, the LM statistic of Kleibergen–Paap rk is 23.487, leading to a significant rejection of the null hypothesis through a 1% significance test. For the test of weak identification of instrumental variables, the Wald F statistic of Kleibergen–Paap rk is 119.443, which is greater than the critical value of 16.38 at the 10% level of the Stock Yogo weak identification test.

4.3. Robustness Analysis

This study employs three approaches to verify the robustness of the empirical findings obtained from the previous analysis. The results are presented in columns (2)–(4) of Table 4. Firstly, the calculation method of the core explanatory variable is altered to observe the changes in regression results. This study draws on previous research [54] and uses principal component analysis to reevaluate the measurement of the DE. Secondly, the regression is conducted using the system generalized method of moments (GMM) model. Thirdly, a two-step GMM approach is employed for the regression analysis. The findings in Table 3 reveal a consistent positive impact of the DE on TFEE. This indicates that the model specification in this study is robust and effective.

4.4. Heterogeneity Analysis

Considering the varying economic development levels, geographical locations, and resource endowments across different regions, the sample is stratified into three regions according to the official regional classification of the National Bureau of Statistics of China: East, Central, and West. The findings, outlined in columns (1) to (3) of

Table 5. Results of heterogeneity analysis.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	East	Central	West	High	Medium	Low
DE	1.001 ***	0.087	0.047 **	−0.698	0.767 ***	0.527 *
	(0.097)	(0.253)	(0.349)	(0.106)	(0.147)	(0.265)
Control	YES	YES	YES	YES	YES	YES
variables
City	YES	YES	YES	YES	YES	YES
Year	YES	YES	YES	YES	YES	YES
_cons	−0.057	−0.461 ***	−0.281	−0.191 *	−0.026	−0.068
	(0.107)	(0.169)	(0.186)	(0.102)	(0.143)	(0.170)
N	1296	984	1104	1404	780	1200

Notes: * p < 0.10, ** p < 0.05, *** p < 0.01; numbers in parenthesis are robust standard error.

, demonstrate that the DE significantly fosters TFEE in both the East and West regions, whereas its effect on TFEE in the Central region, though positive, lacks statistical significance. The swift economic progress in the eastern region has led to the maturity of technology, management practices, and various industries. The improvement in digital technology has effectively guided the enhancement of TFEE. In the West region, the DE has facilitated technology diffusion and efficient allocation of market factors, driving progress in energy technologies. Under the premise of utilizing energy resources efficiently, optimization of energy factor allocation has led to improved TFEE.

To further examine the heterogeneity in energy dependency, we construct a continuous measure of energy dependence (Energy_Dep) as the ratio of total energy consumption to GDP. We split the sample into three groups (high, medium, and low energy dependence) based on the tertiles of Energy_Dep. The results, shown in columns (4) to (6) of Table 5, demonstrate that the DE has a suppressive but nonsignificant effect on TFEE in regions with high energy dependency, whereas it significantly enhances TFEE in regions with medium and low energy dependency. One possible reason is that regions with medium and low energy dependencies are less developed and have inadequate infrastructure. Consequently, the impact of the DE on TFEE is more pronounced in these regions. For regions with high energy dependency, the development of the DE promotes technological advancements, leading to a decline in the transaction prices of energy factors. However, these regions struggle to reduce their energy dependency and consequently squeeze the utilization of other factors, decreasing TFEE.

4.5. Mechanism Testing

The benchmark regression analysis has confirmed that the DE will promote TFEE, but further exploration of the underlying mechanism between the two is still needed. This article explores the different impacts of human capital and industrial structure transformation on TFEE from skill-oriented and task-oriented perspectives. The results are shown in

Table 6. Mechanism inspection results.

Variable	(1)	(2)	(3)	(4)
	HUM	TFEE	IS	TFEE
DE	0.496 ***	1.208 ***	0.601 ***	0.780 ***
	(0.026)	(0.026)	(0.020)	(0.181)
HUM		0.623 **
		(0.312)
IS				0.852 ***
				(0.011)
Control	YES	YES	YES	YES
variables
City	YES	YES	YES	YES
Year	YES	YES	YES	YES
_cons	2.122 ***	5.248 ***	0.129	0.033 *
	(0.055)	(0.455)	(0.209)	(0.019)
N	3384	3384	3384	3384

Notes: * p < 0.10, ** p < 0.05, *** p < 0.01; numbers in parenthesis are robust standard error.

. Column (1) shows that the coefficient of the impact of the DE on human capital is positive and has passed the significance test at the 1% level. Column (2) shows that after adding human capital, the significance and sign direction remain unchanged, indicating that the DE has improved TFEE through human capital, verifying H2. Column (3) shows that the impact coefficient of the DE on industrial structure transformation is positive and has passed the significance test at the 1% level. Column (4) shows that after joining the industrial structure transformation, the significance and symbol direction have not changed, indicating that the DE has improved TFEE through industrial structure transformation, verifying H3.

4.6. Spatial Effect Attenuation Test

Existing research has demonstrated that the DE exerts a significant promoting effect on TFEE. However, this conclusion only reveals the global impact characteristics of the DE. To further analyze the spatiotemporal heterogeneity of its mechanism, this study constructs a dynamic lag model and a geographical attenuation weight matrix to conduct empirical tests from temporal persistence and spatial dimensions. The results are shown in

Table 7. Spatial effect attenuation test.

	Time Dimension		Spatial Dimension
1-period lag	0.209 ***	0–100 km	0.040 **
	(0.030)		(0.014)
2-period lag	0.228 ***	100–250 km	0.047 *
	(0.016)		(0.067)
3-period lag	0.200 **	250–400 km	0.035 *
	(0.021)		(0.016)
4-period lag	0.159 *	400–550 km	0.048
	(0.036)		(0.005)

Notes: * p < 0.10, ** p < 0.05, *** p < 0.01; numbers in parenthesis are robust standard error.

.

Regarding dynamic effect analysis, by introducing 1–5-period lag variables, we identified an inverted U-shaped evolutionary path with diminishing marginal effects in the DE’s promotion of TFEE. Specifically, the enhancement effect progressively strengthens during the 1–2-period lags, reaching peak significance at the 2-period lag. Subsequently, a gradual attenuation trend emerges after the 3-period lag, though statistical significance persists until the 4-period lag. This indicates that energy system transformations driven by digital technology penetration exert long-term cyclical impacts lasting over four years.

From the spatial perspective, DE’s spatial spillover effects on TFEE exhibit significant distance constraints. The spillover effects remain statistically significant within a 0–400 km radius, but dissipate beyond this critical threshold. This spatial boundedness suggests that DE’s spatial spillovers are constrained by practical factors such as factor mobility costs and institutional environment disparities.

4.7. Discussion

While this study confirms the overall positive impact of the DE on TFEE, it is imperative to acknowledge and address two significant challenges that may constrain its benefits. The first pertains to the digital divide, which risks exacerbating regional disparities. Our heterogeneity analysis indicates that the positive effects of the DE are more pronounced in the eastern regions, suggesting that areas with less advanced digital infrastructure and human capital may be left behind. This uneven distribution of benefits could lead to a “green divide,” where technologically advanced regions accelerate their energy transition while less developed areas become further locked into inefficient, conventional development paths.

Another critical consideration is the energy rebound effect, wherein efficiency gains from digitalization may stimulate increased energy consumption, thereby offsetting some of the potential savings. This risk appears particularly relevant in regions with high energy dependence, where our results show a less significant negative impact of the DE on TFEE. In such contexts, efficiency improvements might primarily reinforce energy-intensive production patterns rather than fostering structural transformation. Therefore, achieving meaningful energy conservation will require supplementing DE development with robust policy measures—such as carbon pricing, energy consumption caps, and targeted support for green transitions—to mitigate rebound effects and ensure that efficiency gains translate into absolute resource conservation.

Despite its contributions, this study has several limitations that warrant attention in future research. First, while our measurement of the DE draws on the established literature, it may not fully capture the rapidly evolving landscape of new digital formats and business models. Future studies could develop more dynamic and comprehensive indicator systems. Second, although we identified human capital and industrial structure as key pathways, the intrinsic mechanisms linking the DE to TFEE are likely more complex, potentially involving factors like technological innovation and shifts in consumer behavior, which merit deeper investigation. Lastly, the generalizability of our findings, which are based on data from Chinese cities, needs to be tested in other contexts, such as different developing countries or developed economies. Employing quasi-natural experiments, like policies promoting digital infrastructure, could also provide more robust causal evidence in future work.

5. Conclusions and Policy Implication

5.1. Conclusions

This study utilizes the theory of biased technological progress to explore the impact of the DE on TFEE through the dual channels of human capital (skill-biased) and industrial structure transformation (task-biased). The results indicate the following:

From the perspective of impact effect, the DE can promote TFEE and pass robustness and endogeneity tests. Secondly, from the perspective of the impact mechanism, the DE can promote TFEE by facilitating the transformation of human capital and industrial structure. Thirdly, in terms of heterogeneity, the impact of the DE on TFEE is significant in the eastern and western regions. The impact of the DE on TFEE has a positive effect on regions with moderate energy dependence, while it has a negative effect on regions with high energy dependence. Fourthly, in terms of time dimension, the DE has a long-term effect of improving TFEE and shows an inverted U-shaped trend. On the spatial dimension, the DE has a significant spatial spillover effect on the TFEE of surrounding areas, and this improvement effect exhibits distance attenuation characteristics.

5.2. Policy Implication

Based on the above findings, several policy recommendations are proposed as follows:

First, implement differentiated DE development strategies tailored to local conditions. In eastern regions, where the DE’s promotion of TFEE is most significant, policy should focus on fostering high-end integration and frontier innovation. This includes supporting the deep integration of AI, industrial internet, and big data with advanced manufacturing to create smart and ultra-efficient production systems. In western regions, policy efforts should concentrate on accelerating the deployment of 5G networks and data centers, while simultaneously implementing digital skill training programs to enhance the local workforce’s capacity to absorb and apply new technologies. For central regions, policies should aim to reduce market segmentation, improve the business environment for digital enterprises, and strengthen intellectual property protection to stimulate digital innovation that is effectively translated into TFEE gains.

Second, design precise policy tools for regions with varying energy dependency levels. For regions with medium and low energy dependency, the positive impact of the DE should be amplified. Governments should actively guide the DE to empower the service industry and high-tech manufacturing, and introduce preferential policies for corporate digital transformation and TFEE audits. For regions with high energy dependency, instead of simply pushing digitalization, the state should provide special transition funds to support these regions to develop alternative industries and support the R&D and application of digital technologies for the green transformation of their traditional energy industries.

Third, strengthen the dual-pillar support of talent and technology. Regarding the skill-biased path, the government should deepen collaboration between industries, universities, and research institutes to co-develop curricula and training programs that meet the demands of the DE. For enterprises, tax incentives can be offered to encourage them to establish and improve internal digital skill training systems and to invest in R&D in key areas of energy utilization. Regarding the task-biased path, policy should guide the DE to facilitate the “task reconfiguration” of traditional industries. This can be achieved by subsidizing the adoption of industrial robots and IoT platforms in key sectors and by building industrial internet platforms to enable cross-regional and cross-enterprise sharing of energy and production capacity, thereby fundamentally improving energy allocation efficiency.