Impact of Digital Economy on Energy Consumption and Energy Efficiency

Abstract

Driven by innovations in digital technologies, the digital economy is reshaping societal production and consumption patterns, exerting systematic effects on the energy system through the digital transformation of both the supply and demand sides of energy. Based on empirical analysis using provincial panel data from China between 2011 and 2022, this study demonstrates that the development of the digital economy significantly suppresses the scale of energy consumption while simultaneously improving energy utilization efficiency. After applying the instrumental variable method (with the interaction term of fixed-line telephones and information technology service revenue as the IV) and conducting multiple robustness checks (including lagged explanatory variables, variable substitution, sample trimming, and additional control variables), the core conclusion remains statistically significant. Mechanism tests reveal that the collaborative effects of green technological innovation, the upgrading of industrial structure, and digital inclusive finance form the key transmission path. Finally, heterogeneity analysis shows that the impact of the digital economy on energy consumption and energy efficiency is particularly pronounced in western regions, demonstrating significant regional heterogeneity.

1. Introduction

With the continuous growth of global energy demand and the increasing scarcity of resources, optimizing the energy structure, reducing energy consumption, and improving energy efficiency are critical challenges worldwide. As the world’s largest developing country, China’s energy consumption issue is especially prominent, bearing directly on both its sustainable economic development and global energy security and environmental protection [1,2,3,4]. Concurrently, the digital economy, defined as an economic model driven by digital information and knowledge through modern information networks [5,6], has experienced rapid development. The deep integration of the digital economy with the energy sector is opening new pathways for addressing the energy crisis [7,8], potentially reshaping energy systems through technologies like intelligent sensing and AI decision-making [9,10,11].

However, the net impact of the digital economy on the energy system is theoretically complex and ambiguous, presenting a significant research gap. On one hand, its development creates new energy demands through the construction of digital infrastructure and the operation of data centers, leading to a “rebound effect” that may promote energy consumption [12,13,14]. On the other hand, it holds the potential to suppress energy consumption and improve efficiency by optimizing energy allocation and production processes [15,16,17]. This duality reveals a clear gap in the literature: while existing studies acknowledge both possibilities, they often focus on a single perspective, leaving the net effect an unresolved empirical question. More critically, the specific mechanisms—the pathways through which the digital economy ultimately influences energy outcomes—remain underexplored and lack rigorous empirical testing. Furthermore, the potential for significant regional heterogeneity in these effects within a vast and diverse country like China has been largely overlooked.

Given this background, the core questions this study aims to address are as follows: What is the net impact of digital economy development on energy consumption and energy efficiency in China? And what are the specific underlying mechanisms and regional variations? To answer these questions, we focus on provincial panel data from 30 provinces in China between 2011 and 2022. The marginal contributions of this paper are threefold. First, it provides a robust empirical assessment of the net effect, moving beyond the theoretical dichotomy. Second, it explicitly investigates the transmission mechanisms by testing the intermediary roles of green technological innovation, industrial upgrading, and digital inclusive finance. Finally, it analyzes the regional heterogeneity of the impact, offering nuanced insights for targeted policy-making.

2. Literature Review and Research Hypotheses

2.1. Literature Review

2.1.1. The Impact and Measurement of the Digital Economy

Tapscott (1996) was one of the first to explore the concept of the digital economy, defining it as economic activities based on digital information and communication technologies, encompassing e-commerce, digital services, and various other fields [18]. In 2014, the European Commission defined the digital economy as a series of economic activities driven by digital knowledge and information as key production factors, operating through modern information networks, with the effective use of information and communication technologies (ICT) acting as a crucial driver for improving efficiency and optimizing economic structures. While the definition of the digital economy is still not universally settled, these early conceptualizations have laid the groundwork for subsequent research, clearly outlining the scope of study.

As a new, more advanced, and sustainable economic form, the digital economy has become increasingly important, necessitating theoretical explanations and analyses of the new economic forms and models associated with it. This would enhance the capacity of economic theories to support and guide the digital transformation of the economy and high-quality development in the new phase of growth. In recent years, research on the digital economy has deepened, covering various aspects from tools and frameworks to risks and challenges [19,20]. Notably, the development of the digital economy has brought many benefits. Some studies indicate that the digital economy has a significant role in promoting economic growth. Erik and Michael (2000), through research on e-commerce, found that the digital economy can reduce transaction costs, improve resource allocation efficiency, and drive economic growth [21]. Wang (2023) also found that the digital economy plays an essential role in promoting economic growth, both from the demand and supply sides [22]. Moreover, as an emerging economic model, the digital economy has had a profound impact on people’s work and life [23]. The development of the digital economy has created new jobs in fields such as software development, data analysis, and digital marketing. David (2015) pointed out that the development of digital technology increases the demand for high-skilled labor, giving rise to new professions [24]. The digital economy has also enhanced social inclusivity, enabling people in remote areas to access more educational and healthcare resources [25]. In the face of complex and volatile global economic situations, the digital economy has demonstrated remarkable resilience [26]. For example, during the COVID-19 pandemic, digital services such as telemedicine, online education, and collaborative work were widely applied, profoundly changing people’s work and lifestyle [27]. Despite the convenience brought about by the development of the digital economy, measuring its level of development remains a significant challenge.

2.1.2. Energy Consumption

Energy consumption occupies a central role in global economic and social development, involving issues related to resource utilization, environmental impacts, and economic sustainability, which has attracted extensive academic research. The International Energy Agency (IEA) has continuously tracked global energy consumption trends in its annual reports. According to data from the IEA in 2022, global energy consumption showed a steady increase, with renewable energy use growing, though fossil fuels still dominate the energy consumption structure. There are significant differences in energy consumption across regions, with developed countries generally consuming more energy per capita than developing countries.

Jakob (2012) confirmed through empirical research that there has not been an effective decoupling between economic growth and energy consumption, as both remain closely linked [28]. Many studies have used econometric models for analysis, such as Stern (2000), who established a Vector Autoregression (VAR) model and found that economic growth is a significant driver of energy consumption, with economic expansion typically leading to an increase in energy demand [29]. Industrial structure also has a far-reaching impact on energy consumption. Cleveland et al. (1984) noted that the industrial sector, due to its high-energy production characteristics, leads to higher total energy consumption in regions or countries where it occupies a larger share of the economy [30]. Technological progress has a dual regulatory effect on energy systems, both reducing energy intensity per unit of output through energy efficiency innovation and generating a rebound in energy demand due to economic expansion. Popp (2002) argued that technological innovation has a significant positive effect on improving energy efficiency, but it requires supporting policy tools to realize its potential benefits [31]. Zhou et al. (2022) further pointed out that technological iteration is a key variable in reshaping energy consumption patterns [32]. Based on structural decomposition analysis of energy consumption, technological factors can explain more than 50% of changes in energy intensity, a conclusion cross-validated in the empirical studies of Huang (1993), Sinton, and Levine (1994) [33,34].

2.1.3. Digital Economy and Energy Consumption

In the context of global energy transition and carbon neutrality strategies, the digital economy, as a new economic form following agricultural and industrial economies, is reshaping the operation logic of energy systems through three pathways: the reorganization of production factors, technological paradigm shifts, and industrial ecosystem reconstruction. Driven by the digital wave, researchers are increasingly interested in the relationship between digitalization and energy consumption [35]. Current theoretical research often focuses on a single transmission path: neoclassical economics emphasizes the energy-saving potential of technological innovation [36], institutional economics focuses on the digital regulation reshaping energy markets [37], while evolutionary economics emphasizes path dependency in technological trajectories [38]. The reconstruction of energy systems by the digital economy is essentially a “creative destruction” process initiated by the penetration of digital technologies into the energy sector [39]. Based on the technological-economic paradigm theory, the digital economy has three core characteristics. First, the marginal return of data elements increases [40], breaking the traditional linear growth model of energy input by optimizing energy allocation efficiency. Additionally, there is the general-purpose nature of digital technologies [7] and the network externalities of platform economies [41], which together form the theoretical foundation for the impact of digital technologies on energy systems. From a positive dimension, the “digital dividend” formed by optimizing energy allocation efficiency through digital technologies has gained widespread academic recognition. From a negative dimension, however, the expansion of energy demand caused by new digital infrastructure, such as data centers and computing clusters, has created the typical paradox of “efficiency improvement coexisting with energy consumption growth.”

Current academic research on the relationship between the digital economy and energy consumption shows significant divergence. Scholars supporting the “energy consumption promotion theory” argue that the expansion of the digital economy exhibits a significant energy dependency characteristic. Their core arguments include the following: first, the large-scale construction of ICT infrastructure generates enormous energy sunk costs in the early stages [42]; second, the energy demand of global digital computing and communication networks is increasing exponentially, with energy consumption in data centers growing at an annual rate of 6% [43], reflecting the reemergence of Jevons’ Paradox in the iteration of digital technologies [44,45]; third, the ICT industry exhibits significant implicit energy consumption multiplier effects, such as China’s ICT sector’s implicit energy consumption in 2018 being three times the direct energy consumption, and the efficiency improvements brought by technological innovations are often offset by the rebound effects of economic growth [46].

In contrast, “energy consumption suppression theory” emphasizes the systematic energy-saving potential of digital technologies. Theoretically, digital technologies affect energy systems through three main pathways: improving energy conversion efficiency [47], optimizing energy consumption structure [48], and reducing energy intensity per unit of GDP [49]. Empirical studies show that a 1% increase in ICT investment can reduce energy consumption by 0.155% in Japan [50], and provincial-level data from China confirms the significant negative spatial spillover effect of internet development on electricity intensity [35]. Cross-national studies further verify the dual suppression effect of digitalization on both total energy consumption and intensity, promoting the transition of energy structures toward cleaner forms [51], with such optimization effects being particularly prominent in energy storage, distribution, and security [46,52].

2.1.4. The Policy-Driven Digital Economy in China

Unlike the organic, market-led digital evolution in some Western economies, China’s digital economy development has been significantly shaped by a series of proactive and top-down national strategies. These policies have not only accelerated the construction of digital infrastructure but also directed the integration of digital technologies into traditional sectors, including energy. Several key milestones within our study’s timeframe (2011–2022) are particularly noteworthy:

The “Broadband China” Strategy: This policy aimed to universalize broadband access and accelerate network upgrades. It directly led to a massive expansion of fiber-optic networks and an increase in internet penetration rates, which are core components of our digital infrastructure indicator. This enhanced connectivity is a prerequisite for smart grid management, remote energy monitoring, and the deployment of Internet of Things (IoT) devices in the energy sector.

The Commercialization of 4G (2013–2014): The issuance of 4G licenses in 2013 marked a leap in mobile communication technology. The high speed and low latency of 4G networks enabled real-time data transmission, which is critical for the operation of distributed energy resources, dynamic electricity pricing models, and mobile energy management applications. This technological leap facilitated the deep penetration of digital solutions into energy consumption and efficiency management.

The “14th Five-Year Plan” (2021–2025): Although our data extends to 2022, the early implementation phase of this plan firmly positioned the digital economy as a core driver of high-quality development. Its goals, such as promoting industrial digitization and achieving a “Digital China,” provided a sustained policy impetus. This ensured that the trends of digital-energy integration observed in our later sample years were backed by strong institutional support.

These policies collectively created a unique “national experiment” environment. They systematically reduced the costs and barriers to digital technology adoption, thereby amplifying the mechanisms through which the digital economy influences energy systems—namely, by accelerating green technological innovation, forcing industrial upgrading, and scaling up digital inclusive finance. Therefore, our empirical analysis based on provincial data during this period effectively captures the impact of this policy-driven digital transformation.

2.1.5. China’s Institutional Context: Differentiating the Digital-Energy Interaction

The impact of the digital economy on energy systems in China cannot be fully understood without considering the country’s unique institutional characteristics, which fundamentally differentiate its pathway from that of market-driven economies. China can be characterized as a state-led developmental state [1], where the government plays a proactive and directive role in economic structuring and technological catch-up. This institutional framework shapes the digital-energy interaction in several key ways that distinguish the Chinese model:

Government-Led Digital Infrastructure Investment: Unlike in market-driven economies where infrastructure rollout is primarily profit-motivated and can lead to digital divides, China’s “Broadband China” and “Digital Village” strategies represent nationwide, state-orchestrated initiatives. The government mobilizes public resources to ensure rapid, ubiquitous, and standardized digital infrastructure deployment, even in less profitable western and rural regions. This state-mandated universality ensures that the energy efficiency benefits of digital technologies (e.g., smart grids, industrial IoT) are not confined to economically advanced coastal provinces but can also be realized in less developed, often more energy-intensive interior regions. This mitigates the market failure of under-investment and accelerates the national-scale penetration of energy-saving digital applications.

Industrial Policy and Administrative Guidance: China’s industrial policies, such as those outlined in the “Made in China 2025” and the “14th Five-Year Plan,” explicitly guide the integration of digital and green goals. The state uses a combination of subsidies, tax incentives, and administrative targets to directly channel resources towards strategic sectors, including green technology innovation and the digital transformation of traditional industries. This directed technological change [53] reduces the uncertainty and upfront costs for firms, compelling a faster adoption of energy-efficient digital solutions than a purely market-driven process might achieve. For instance, mandates for industrial energy efficiency are now coupled with support for enterprise digitalization, creating a synergistic push.

The Role of State-Owned Enterprises (SOEs): Large SOEs, particularly in the energy and telecommunications sectors, act as key instruments of state policy. They are often the first movers in implementing large-scale digital-energy integration projects, such as smart power grids and national energy big data platforms. Their compliance with national strategic directives ensures the rapid scaling of pilot projects and creates a stable, large-scale demand for related digital technologies, further driving down costs and fostering ecosystem development.

In contrast, in market-driven economies, the digital-energy interaction is more emergent and fragmented, relying on corporate innovation, venture capital, and consumer choice, often leading to the slower and more uneven diffusion of technologies. Therefore, the Chinese institutional context does not merely accelerate the processes described in general theory; it actively structures them. It ensures that the development of the digital economy and its subsequent impact on energy consumption and efficiency are deeply intertwined with national strategic objectives, leading to a more rapid and spatially comprehensive effect than might be observed elsewhere. Our empirical analysis, set in this specific institutional context, thus captures the outcome of this distinct, state-facilitated model.

2.2. Research Hypotheses

The impact of the digital economy on energy systems is not direct but operates through a fundamental restructuring of production factors, technological paradigms, and institutional arrangements. Drawing on the technology–economic paradigm theory [46] and Schumpeter’s (1976) “creative destruction” concept, we posit that the digital economy acts as a general-purpose technology that disrupts incumbent energy-intensive regimes [39]. This disruption manifests through three interdependent, core channels derived from established economic theories: the direct efficiency effect (rooted in neoclassical growth theory), the structural transformation effect (informed by evolutionary economics and the Environmental Kuznets Curve), and the financial intermediation effect (anchored in financial development and transaction cost theories). The following hypotheses delineate these specific pathways.

2.2.1. The Impact of the Digital Economy on Energy Consumption and Energy Efficiency

The rapid development of the digital economy is reshaping the evolution trajectory of the global energy system, with its essence lying in the technological innovation that reconstructs the operational paradigm of the energy value chain. This transformation goes beyond the superficial application of digital tools and, in essence, creates a systemic mechanism for performance enhancement through the pervasive penetration of informational elements. The fundamental logic is to break the traditional economic reliance on physical resources, constructing an innovative framework where data flow dominates energy flow, thus reducing energy consumption and improving energy efficiency [54,55,56].

First, the digital economy triggers the substitution effect of production factors, promoting the optimization and upgrading of energy consumption structures and driving the low-carbon restructuring of industrial structures. Traditional industrial economies are constrained by a linear input model of material factors, creating a rigid correlation between energy consumption and output scale. In contrast, digital technologies, through the virtualization and reinvention of physical elements, establish new pathways for the substitution of production factors. Simultaneously, the rise of the sharing economy breaks down information barriers and integrates fragmented supply and demand into dynamic interactive networks, activating and utilizing idle resources. In the manufacturing sector, digital prototyping technology replaces physical testing processes, drastically reducing resource consumption during the research and development phase. This substitution effect permeates all stages of the industrial chain, promoting the formation of flexible collaboration networks that compress overall system energy consumption by eliminating redundant nodes [56].

Second, as the core platform for the development of the internet, digital technologies build service platforms and conversion systems that allow key elements such as knowledge, information, and data to be efficiently integrated into production activities, thereby profoundly transforming the utilization patterns and efficiency of traditional energy factors. For example, in the transportation sector, precise traffic information services provided by digital technologies optimize transportation routes and make travel and logistics modes more rational, effectively reducing energy consumption in the transportation process and improving energy efficiency [57]. The International Energy Agency predicts that the full application of digital technologies in transportation could cut energy consumption by approximately 20%. From an industrial perspective, as the key support for the digital industry, advancements in digital technologies not only drive the expansion of the digital industry but also promote the optimization and upgrading of industrial structures. In this process, the share of traditional industrial sectors decreases, and energy consumption consequently reduces [58]. At the same time, digital technologies empower traditional energy development technologies, significantly lowering the cost of clean energy development and enhancing its utilization, driving the gradual transition from a coal-dominated energy structure to a green, low-carbon one. The decline in the share of coal consumption is significant for improving energy efficiency. Moreover, digital technologies facilitate the deep integration of industries, especially the collaborative development of digital industries and traditional industries, fostering new business models [5,6]. These new business models reshape the structure and utilization patterns of energy demand, opening new pathways for efficient energy use and fully promoting the sustainable development of the energy sector. Based on the theoretical deductions above, we derive the following research hypothesis:

Hypothesis H1.

The digital economy will reduce energy consumption and improve energy efficiency.

2.2.2. The Impact of Industrial Upgrading

The theoretical logic behind the reduction of energy consumption through industrial upgrading driven by the digital economy is rooted in the dual pathway of digital technologies transforming traditional industries and fostering new business models. On the one hand, the general-purpose nature of digital technologies [7] drives the migration of production factors from high-energy-consuming industries to technology-intensive sectors: traditional manufacturing utilizes industrial internet technologies for equipment interconnection and intelligent production scheduling, precisely controlling production processes. On the other hand, according to the IEA’s 2020 report, the energy consumption per unit of added value in the ICT sector is only a quarter of that in traditional manufacturing. The expansion of the ICT sector directly reduces the energy intensity of the economy, while the increasing share of emerging services such as e-commerce and digital finance further compresses the overall energy consumption. The deeper mechanism lies in the network externalities of the platform economy [41], which restructure industrial organization. This structural transformation follows the evolutionary pattern of the Environmental Kuznets Curve and confirms the applicability of the directed technological change theory [53] in the digital era. Digital platforms strengthen market signals through real-time data flows, guiding innovation resources to concentrate on clean technology fields, thus forming a positive feedback loop of “structural upgrading–technological leap–energy consumption reduction.” Based on the above theoretical reasoning, we derive Hypothesis H2.

Hypothesis H2.

The digital economy reduces energy consumption through industrial upgrading.

2.2.3. The Impact of Green Innovation

The non-competitive characteristic of data elements [40] lowers the costs of research, development, and diffusion of green technologies. For instance, blockchain technology has improved the lifecycle traceability efficiency of power batteries by 300%, and energy consumption in recycling resource utilization has been reduced by 80%. Artificial intelligence-driven material simulations have compressed the development cycle of clean technologies by 60%, overcoming the traditional “valley of death” in innovation [59]. Digital twin technologies replace physical testing with virtual simulations, significantly reducing the energy consumption of steel companies per ton of steel produced, validating the realization of Schumpeter’s “creative destruction” mechanism in energy efficiency improvements. On a deeper level, the network effects of the platform economy [41] reshape the green innovation ecosystem. Industrial internet platforms connect 1200 companies to share energy-saving processes, with the overall energy intensity decreasing year by year. This “technology–market–institution” co-evolution essentially encodes discrete green knowledge as reusable digital assets, achieving a systemic leap in energy efficiency through a diffusion model with near-zero marginal costs, while avoiding the rebound effect found in Jevons’ Paradox. Thus, we propose Hypothesis H3.

Hypothesis H3.

The digital economy increases energy efficiency through green innovation.

2.2.4. The Impact of Digital Inclusive Finance

The hypothesis that digital inclusive finance (DF) serves as a transmission channel is grounded in the synergistic application of financial intermediation theory and transaction cost economics [60]. The digital economy, characterized by big data, blockchain, and platform technologies, directly revolutionizes the traditional financial system by mitigating core frictions. It profoundly reduces information asymmetry through dynamic, data-driven credit scoring, breaking the reliance on physical collateral [61]. Simultaneously, it drastically lowers transaction costs by simplifying processes and achieving economies of scale, allowing financial institutions to serve a massive client base via digital platforms at minimal marginal cost. This technological empowerment leads to the rise of digital inclusive finance—a more efficient, transparent, and accessible financial ecosystem.

This reconfigured financial system subsequently impacts energy consumption and efficiency through a structured pathway. Digital inclusive finance builds a synergistic mechanism of “technological innovation–capital flow–market expansion,” systematically reshaping the logic of energy resource allocation. On the supply side, decentralized platforms and smart contracts break traditional credit constraints, significantly reducing the financing costs for clean energy projects and improving the liquidity of energy assets. On the demand side, mobile payment technologies and data-driven credit models lower barriers for consumers and small enterprises, increasing the penetration of distributed energy and energy-efficient solutions in grassroots markets.

The development of digital inclusive finance diversifies resource acquisition channels and creates a strong impetus for competition and technological reform. This “capital allocation–technology penetration–behavioral incentives” multidimensional linkage reconstructs the operational paradigm of energy economics. By enhancing price discovery and risk pricing functions, it guides production factors to continuously cluster in high-efficiency, low-carbon sectors. Its internal logic aligns with Schumpeter’s innovation theory on the catalytic role of financial intermediaries in technological revolutions, forming a deeply coupled evolutionary path of “digital finance empowering technological decarbonization [39].”

Moreover, digital inclusive finance breaks the spatial and temporal boundaries of traditional financial services, establishing a precise resource allocation system. Big data analysis enables credit evaluation without heavy reliance on collateral, directing financial capital into clean energy development and energy efficiency transformations. The traceable financing system built on blockchain technology ensures the transparency of fund flows, preventing misallocation into high-energy-consuming projects. Dynamic risk assessment models monitor project energy consumption in real-time, using financial levers to guide market participants toward energy-saving solutions, forming a positive feedback loop between capital allocation and energy efficiency. Based on this theoretical reasoning, which posits that the digital economy fosters a unique financial ecosystem that systematically lowers barriers and directs capital towards energy-efficient and low-carbon activities, we propose Hypothesis H4:

Hypothesis H4.

The digital economy reduces energy consumption and improves energy efficiency through the development of digital inclusive finance.

2.2.5. Regional Heterogeneity Analysis

China’s regions exhibit significant spatial differences in natural resource distribution, ecological environmental characteristics, and climatic conditions. This coupling of geographical endowments with economic and social development profoundly influences the transmission path of digital technologies’ energy regulation effects.

There are essential differences between the western and non-western regions in terms of economic development models and energy structures. The western regions are dominated by traditional resource-based industries, with a strong reliance on energy, while they also possess abundant renewable energy resources. However, they face structural challenges in improving energy utilization efficiency due to limited technological absorption capacity and infrastructure development. In contrast, non-western regions have more developed service and advanced manufacturing sectors, with deeper integration between the digital economy and the real economy, focusing more on optimizing the operation efficiency of existing energy systems through intelligent methods.

Moreover, the differentiated policy environment and market mechanisms have shaped distinct technological penetration pathways. The western regions benefit from national strategic plans that combine digital economy and energy transition policies, with digital technologies often deeply aligned with renewable energy development and inter-regional energy allocation under ecological protection constraints. In contrast, non-western regions, with mature market mechanisms and innovation ecosystems, see digital technologies more easily driving energy-saving retrofits in traditional industries and energy management upgrades through market competition. Therefore, regional heterogeneity likely exists between the western and non-western regions. Based on existing literature [32,62], this paper divides the 30 provinces into western and non-western regions for heterogeneity analysis and proposes Hypothesis H5.

Hypothesis H5.

The impact of the digital economy on energy consumption and energy efficiency varies significantly between the western and non-western regions.

3. Data, Variables, and Models

3.1. Data Sources

The original data for this study are derived from the following sources: the “China City Statistical Yearbook,” the “China Statistical Yearbook,” the “China Electronic Information Industry Statistical Yearbook,” the “China Science and Technology Yearbook,” the National Bureau of Statistics of China, the China Digital Inclusive Finance Development Database, and the Economic Prediction System (EPS) Database. During the data processing, it was observed that data related to Hong Kong, Macau, Taiwan, and Tibet were severely incomplete. As a result, these four regions were excluded from the sample. Additionally, in order to mitigate the impact of extreme values, the sample data were trimmed to the 1st and 99th percentiles.

3.2. Variables

3.2.1. Explanatory Variables

Based on the studies by Yong (2024) [63], Duan (2024) [64], and incorporating the actual development of the digital economy, we selected 20 basic indicators from three dimensions—digital infrastructure, digital industrialization, and industrial digitization—to construct the indicator system for digital economy development level (

Table 1. Digital economy indicator system.

Primary Indicator	Secondary Indicator	Tertiary Indicator	Indicator Definition	Attributes
Digital Economy	Digital Infrastructure	Internet Broadband Access Rate	Number of broadband internet access ports/total population	+
		Internet Broadband Penetration Rate	Number of broadband internet users/total population	+
		Mobile Telephone Infrastructure Scale	Capacity of mobile telephone switches	+
		Long-Distance Optical Cable Length	Length of long-distance optical cable	+
		Number of Websites	Direct data	+
		Number of Domain Names	Direct data	+
		Per Capita Telecom Business Volume	Total telecom business volume/local population	+
		Mobile Telephone Penetration Rate	Direct data	+
		Number of Legal Entities in Information Transmission, Software, and IT Services	Direct data	+
	Digital Industrialization	Proportion of Employment in Information Software	Employment in information transmission, software, and IT services in urban units/total urban unit employment	+
		Number of Domestic Patent Applications Authorized	Direct data	+
		Number of Domestic Patent Applications Accepted	Direct data	+
		Peking University Digital Inclusive Finance Index	Direct data	+
		Proportion of Enterprises Engaged in E-commerce	Direct data	+
		E-commerce Sales Volume	Direct data	+
		Number of Websites per 100 Enterprises	Direct data	+
	Industrial Digitalization	Value Added of Secondary and Tertiary Industries	Value added of secondary industries + tertiary industries	+
		Investment in Technological Innovation	R&D expenditure of above-scale industrial enterprises	+
		Express Delivery Volume	Direct data	+

The “+” in the “Attributes” column denotes a positive indicator, meaning that a higher value of the indicator reflects a greater contribution to or higher level of development of the digital economy.

). The entropy method was used to determine the final digital economy development level for 30 provinces between 2011 and 2022. We note that while emerging indicators such as 5G base station density and data element transaction volume are conceptually important, systematic provincial panel data for these metrics are not yet publicly available for our study period. Their inclusion remains a valuable direction for future research as data becomes more accessible.

3.2.2. Dependent Variables

This study uses energy consumption and energy efficiency as the dependent variables. Energy consumption is represented by the total coal consumption, while energy efficiency is measured by the reciprocal of energy intensity, calculated using the formula: GDP/Total energy consumption.

3.2.3. Mediating Variables

There are three mediating variables in this study: the level of green innovation, industrial upgrading, and digital inclusive finance. The specific data indices for these variables are provided in

Table 2. Variable definitions.

Variable		Symbol	Definition
Dependent Variable	Energy Consumption	TEC	Total Standard Coal Consumption/10,000
	Energy Efficiency	EUE	GDP/Total Coal Consumption
Independent Variable	Digital Economy Development Level	DED	Obtained by the Entropy Method
Intervening Variable	Green Innovation Level	GT	Ln (Number of Green Invention Patent Applications + 1)
	Industrial Upgrading	AS	Value Added of Tertiary Industry/Value Added of Secondary Industry
	Digital Inclusive Finance	DF	Peking University Digital Inclusive Finance Index:
Control Variables	Education Level	EDU	Number of People with Primary Education × 6 + Number of People with Secondary Education × 9 + Number of People with High School and Vocational Education × 12 + Number of People with College and Above Education × 16)/Total Population Aged 6 and Above
	Industrial Structure	IS	Proportion of the Tertiary Industry in GDP
	Urbanization Rate	UR	Urban Population Residing for More Than Six Months/Total Population × 100%
	GDP	GDP	GDP of Each Province
	Population Size	PS	Resident Population of Each Province
	Energy Consumption Structure	ECS	Electricity Consumption/Total Energy Consumption

.

3.3. Model

To examine the relationship between the digital economy, energy consumption, and energy efficiency, the following model was employed:

TECi,t = α0 + α1DED + α2EDUi,t + α3ISi,t + α4URi,t + α5GDPi,t + α6PSi,t + α7ECSi,t + ΣYear + ΣProvince + εi,t

(1)

EUEi,t = β0 + β1DED + β2EDUi,t + β3ISi,t + β4URi,t + β5GDPi,t + β6PSi,t + β7ECSi,t + ΣYear + ΣProvince + εi,t

(2)

In this model, i and t represent the province and the year, respectively. DED denotes the level of digitalization, TEC represents energy consumption, and EUE represents energy efficiency. Year and Province indicate the fixed effects for year and province, and ε is a random disturbance term. The baseline model is then extended as follows to investigate the moderating effects of green innovation level (GT), industrial upgrading (AS), and digital inclusive finance (DF) on the relationship between digitalization and energy consumption/energy efficiency. M is used to represent the green innovation level (GT), industrial upgrading (AS), and digital inclusive finance (DF).

Mi,t = γ0 + γ1DED + γ2EDUi,t + γ3ISi,t + γ4URi,t + γ5GDPi,t + γ6PSi,t + γ7ECSi,t + ΣYear + ΣProvince + εi,t

(3)

TECi,t = ν0 + ν1DED + ν2Mi,t + ν3ISi,t + ν4URi,t + ν5GDPi,t + ν6PSi,t + ν7ECSi,t + ν8EDUi,t + ΣYear + ΣProvince + εi,t

(4)

EUEi,t = δ0 + δ1DED + δ2Mi,t + δ3ISi,t + δ4URi,t + δ5GDPi,t + δ6PSi,t + δ7ECSi,t + δ8EDUi,t + ΣYear + ΣProvince + εi,t

(5)

In estimating the aforementioned models, we employ an ordinary least squares (OLS) estimator with two-way (year and province) fixed effects. This approach effectively controls for unobserved time-invariant heterogeneity across provinces and time-specific shocks common to all provinces. We acknowledge the potential for heteroscedasticity in provincial-level panel data, which, if present, could lead to biased standard errors and affect the reliability of statistical inference. To mitigate this concern, all regression models are estimated using robust standard errors clustered at the provincial level. This method produces heteroscedasticity-consistent standard errors, which remain valid even in the presence of heteroscedasticity and within-province serial correlation, thereby ensuring the robustness of our hypothesis testing.

3.4. Descriptive Statistics

Table 3. Descriptive statistics.

Variable	N	Mean	Std. Dev.	Min	Max
TEC	360	1.548	0.915	0.182	4.139
EUE	360	1.729	0.905	0.455	4.816
DED	360	0.130	0.107	0.018	0.584
GI	360	7.190	1.406	2.197	10.089
AS	360	1.357	0.744	0.527	5.244
DF	360	243.928	107.640	18.330	460.691
EDU	360	9.313	0.912	7.474	12.782
IS	360	49.421	9.133	32.600	83.900
UR	360	0.601	0.121	0.350	0.896
GDP	360	27,178.470	22,996.160	1370.400	129,513.600
PS	360	4614.570	2858.810	568.150	12,684.000
ECS	360	0.371	0.149	0.006	0.687

presents the descriptive statistics. Our sample consists of 360 observations from 30 provinces between 2011 and 2022. The average value of digital economy development is 0.1304, with a minimum of 0.0177 and a maximum of 0.5838, indicating significant disparities in the level of digital economy development across the provinces.

Control Variables

To reduce measurement errors caused by model specification and obtain more accurate results, other variables that may influence energy consumption are also controlled for. When energy consumption and energy efficiency are used as dependent variables, the control variables include education level, industrial structure, urbanization rate, GDP, population size, and energy consumption structure. The specific definitions of these variables are provided in Table 2.

4. Results

4.1. Baseline Regression Results

According to the regression results in

Table 4. Baseline regression.

	(1)	(2)
VARIABLES	TEC	EUE
DED	−1.356 ***	2.561 ***
	(−2.60)	(2.95)
PS	0.000 ***	−0.000 ***
	(4.17)	(−2.92)
IS	−0.011 ***	−0.006
	(−2.70)	(−0.92)
ECS	0.677 ***	−1.838 ***
	(3.94)	(−5.22)
GDP	0.000 ***	0.000 ***
	(3.22)	(2.74)
UR	0.094	−2.187 **
	(0.19)	(−2.53)
EDU	−0.082 *	0.359 ***
	(−1.76)	(4.31)
Constant	1.430 *	1.250
	(1.82)	(0.93)
Year	YES	YES
Province	YES	YES
Observations	360	360
R-squared	0.987	0.963

Robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

, Model (1), with energy consumption as the dependent variable, shows that after controlling for variables and incorporating fixed effects for year and province, the coefficient for the digital economy development index is significantly negative. This indicates that the development of the digital economy has a significant energy-saving effect. This result supports the notion that the application of digital technologies has not triggered a significant rebound effect. The core mechanism lies in the direct reduction of energy input requirements by improving energy efficiency in enterprises. The regression results in Table 3, Model (2), where energy efficiency is the dependent variable, also show a significantly negative association with the digital economy development index. This reveals that the digital economy effectively enhances energy conversion efficiency by restructuring the allocation of production factors. As a new economic form centered on digital knowledge and operating through information networks, the digital economy, through the deep penetration of information and communication technologies, increases energy efficiency while maintaining economic growth. This finding confirms that the digital economy can reduce energy consumption and increase energy efficiency, supporting Hypothesis H1.

4.2. Endogeneity Treatment

Although controlling for variables and fixed effects has partially mitigated endogeneity bias, unobservable factors may still interfere with the validity of the estimation results. To address potential endogeneity issues arising from omitted variables, an instrumental variable (IV) approach was employed for robustness testing. The interaction term between the number of fixed-line telephones per 100 people and the previous year’s national information technology service revenue was selected as the instrument [64,65]. A valid instrument needs to satisfy both relevance and exogeneity conditions. Regarding relevance, the development of the digital economy is related to the development of historical telecommunication infrastructure, which meets the relevance condition for IV. Regarding exogeneity, the number of telephones should not directly affect regional energy consumption. The results in

Table 5. Instrumental variable method.

	IV1 First-Stage (1)	IV1 Second-Stage (2)	IV1 Second-Stage (3)
VARIABLES	DED	TEC	EUE
DED		−6.472 ***	15.695 ***
		(−3.48)	(4.24)
PS	0.000 ***	0.001 ***	−0.001 ***
	(8.00)	(5.02)	(−4.33)
IS	0.000	−0.012 ***	−0.001
	(0.25)	(−3.65)	(−0.19)
ECS	−0.008	0.537 **	−1.478 ***
	(−0.35)	(2.57)	(−3.55)
GDP	0.000 ***	0.000 ***	−0.000 **
	(17.59)	(3.78)	(−2.55)
UR	−0.384 ***	−3.199 **	6.266 **
	(−4.16)	(−2.52)	(2.47)
EDU	0.005	−0.048	0.271 ***
	(0.78)	(−0.96)	(2.72)
IV2	0.000 ***
	(2.92)
Constant	0.113	3.349 ***	−3.675 *
	(1.17)	(3.32)	(−1.83)
Year	YES	YES	YES
Province	YES	YES	YES
Observations	360	360	360
R-squared	0.987	0.982	0.929

z-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

show that the negative impact of the digital economy development index on energy consumption (coefficient = −6.472 ***) and its positive effect on energy efficiency (coefficient = 15.695 ***) remain statistically significant, and the direction is consistent with the baseline regression. This indicates that the endogeneity issue does not substantially alter the unbiased estimation of the core variables, and the baseline model exhibits high result robustness. The instrument validity test shows that in the first-stage regression, the instrument is significantly correlated with the endogenous variable (F-value = 24.8384 > 10), meeting the weak instrument test criterion, further confirming the scientificity of the estimation method and validating the reliability of Hypothesis H1.

4.3. Robustness Check

4.3.1. Lagged Explanatory Variables for One and Two Periods

To control for the interaction effects of contemporaneous variables, the baseline model introduces lagged explanatory variables. In

Table 6. Lagged explanatory variables.

	(1)	(2)	(3)	(4)
VARIABLES	TEC	TEC	EUE	EUE
L.DED	−1.226 **		2.334 ***
	(−2.36)		(2.60)
L2.DED		−1.078 **		2.270 ***
		(−2.21)		(2.64)
PS	0.000 ***	0.000 ***	−0.000 **	−0.000 **
	(4.42)	(4.71)	(−2.44)	(−2.32)
IS	−0.013 ***	−0.015 ***	0.000	0.006
	(−3.22)	(−3.49)	(0.01)	(1.08)
ECS	0.641 ***	0.459 **	−1.525 ***	−0.850 **
	(3.37)	(2.25)	(−3.72)	(−1.99)
GDP	0.000 ***	0.000 ***	0.000 ***	0.000 ***
	(3.10)	(3.56)	(2.96)	(3.13)
UR	0.045	−0.034	−1.984 **	−1.164
	(0.09)	(−0.06)	(−2.05)	(−1.05)
EDU	−0.063	−0.034	0.333 ***	0.270 ***
	(−1.38)	(−0.73)	(3.82)	(3.10)
Constant	1.442 *	1.692 *	1.005	0.841
	(1.87)	(1.95)	(0.75)	(0.57)
Year	YES	YES	YES	YES
Province	YES	YES	YES	YES
Observations	330	300	330	300
R-squared	0.989	0.990	0.966	0.971

Robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

, columns (1) and (3) present the regression results for one-period lag (L.DED), while columns (2) and (4) display the results for two-period lag (L2.DED). The empirical results show that the digital economy index continues to exert a statistically significant negative effect on energy consumption (coefficient = −1.226 **) and a positive effect on energy efficiency (coefficient = 2.334 ***) even under lag conditions. Although the two-period lag coefficients decay to −1.078 ** and 2.270 ***, their significance does not undergo a substantial change. This result confirms the robustness of Hypothesis H1.

4.3.2. Replacement of Explanatory Variables

To ensure the robustness of the baseline model results, this study conducts a replacement test of the explanatory variables by reconstructing the measurement index system. Based on five core dimensions—internet penetration rate, workforce size, industrial output scale, mobile terminal penetration rate, and digital finance coverage [66,67]—the digital economy development index is rebuilt using the objectively weighted entropy method. The empirical results in

Table 7. Replacement of explanatory variables.

	(1)	(2)
VARIABLES	TEC	EUE
DED_i	−0.926 ***	1.988 ***
	(−4.51)	(6.01)
PS	0.000 ***	−0.000
	(2.91)	(−1.38)
IS	−0.009 **	−0.008
	(−2.52)	(−1.39)
ECS	0.779 ***	−2.046 ***
	(4.66)	(−5.82)
GDP	0.000 *	0.000 ***
	(1.66)	(5.61)
UR	1.626 ***	−5.250 ***
	(4.64)	(−5.56)
EDU	−0.086 *	0.364 ***
	(−1.82)	(4.26)
Constant	0.864	2.334
	(1.17)	(1.64)
Year	YES	YES
Province	YES	YES
Observations	360	360
R-squared	0.988	0.965

Robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.

show that after replacing the variables, the digital economy continues to exert a statistically significant negative effect on energy consumption (coefficient = −0.926 ***) and a positive effect on energy efficiency (coefficient = 1.988 ***) at the 1% significance level, with the coefficient directions fully consistent with the baseline model. This further confirms the reliability of Hypothesis H1.

4.3.3. Replacement of Dependent Variables

Given the intensive reliance of digital economic activities on electricity resources, total electricity consumption (GC) is chosen as a proxy for energy consumption. At the same time, to eliminate the impact of monopoly pricing on energy costs, carbon emission intensity per unit of GDP (CEI) is constructed to measure energy efficiency, with the formula: GDP/Total carbon emissions of each province. The empirical results in

Table 8. Replacement of dependent variables.

	(1)	(2)
VARIABLES	GC	CEI
DED	−7.134 ***	6.064 ***
	(−5.00)	(4.24)
PS	0.001 **	−0.001 ***
	(2.47)	(−3.82)
IS	−0.029 **	−0.014 ***
	(−2.44)	(−2.82)
ECS	0.351	−1.049 ***
	(0.66)	(−4.18)
GDP	0.000 ***	−0.000 **
	(6.72)	(−2.12)
UR	1.786	−1.250 **
	(1.29)	(−2.08)
EDU	−0.192	0.092
	(−1.58)	(1.14)
Constant	0.964	2.857 **
	(0.45)	(2.10)
Year	YES	YES
Province	YES	YES
Observations	360	360
R-squared	0.961	0.937

Robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05.

show that the digital economy index (DED) exhibits a significant negative association with GC (coefficient = −7.134, *** p < 0.01), and a positive coefficient of 6.064 for CEI (*** p < 0.01). The experimental results are consistent with Hypothesis H1.

4.3.4. Shortened Time Frame

To further ensure the robustness of our baseline model, we shorten the study period to 2013–2022. This adjustment is theoretically grounded, as 2013 marks a critical juncture in China’s policy-driven digital transformation, initiated by the launch of the “Broadband China” strategy and the commercial rollout of 4G networks. These policies dramatically accelerated the construction of digital infrastructure and the penetration of digital technologies into the real economy, including the energy sector. The empirical results in

Table 9. Shortened time frame.

	(1)	(2)
VARIABLES	TEC	EUE
DED	−1.497 ***	1.970 **
	(−2.85)	(2.02)
PS	0.000 ***	−0.000 **
	(4.71)	(−1.98)
IS	−0.014 ***	0.006
	(−3.53)	(1.06)
ECS	0.456 **	−0.880 **
	(2.30)	(−2.11)
GDP	0.000 ***	0.000 **
	(3.87)	(2.55)
UR	−0.189	−1.554
	(−0.35)	(−1.47)
EDU	−0.042	0.283 ***
	(−0.92)	(3.24)
Constant	1.972 **	0.879
	(2.23)	(0.58)
Year	YES	YES
Province	YES	YES
Observations	300	300
R-squared	0.990	0.971

Robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05.

show that, even within this period of intensive digitalization, the digital economy continues to exert a significant negative effect on energy consumption (coefficient = −1.497 *) and a positive effect on energy efficiency (coefficient = 1.970). The consistency in the signs and significance of the coefficients with our baseline results confirms that the core conclusion—the digital economy’s significant role in suppressing energy use and improving efficiency (H1)—is robust, even when focusing specifically on a phase defined by pivotal national policies.

4.3.5. Adding Control Variables

In the baseline regression model, the degree of government intervention and the level of openness to external markets were included as control variables, given their structural impact on the energy-economic system and their potential interactive mechanisms with the digital economy. Government intervention is embedded within a dual-theory framework. On one hand, according to the “Porter Hypothesis,” appropriate environmental regulations can incentivize companies to adopt digital energy-saving technologies through innovation compensation effects [68]. On the other hand, direct government intervention may trigger “soft budget constraints,” leading to the persistence of inefficient capacity [69]. Such institutional friction could suppress the Pareto improvement that the digital economy might otherwise bring to the energy system.

The level of openness to external markets influences energy consumption through three theoretical pathways. First, industry transfer based on comparative advantage may lead to a “pollution haven” effect, in which energy-intensive industries are relocated from high-regulation areas to low-regulation areas [70]. This structural shift may obscure the energy-saving effects generated by the digital economy through the reorganization of production factors. Second, the technology diffusion brought by international trade may encounter a “threshold effect,” meaning that foreign direct investment (FDI) can only effectively enhance energy efficiency when the host country’s human capital and infrastructure reach a certain threshold [71]. This necessitates controlling for the level of openness to separate the marginal contribution of digital technology to energy efficiency. Third, the global pricing mechanism for energy commodities generates price transmission through trade channels, where fluctuations in international oil prices may distort local energy substitution elasticity [72]. This type of exogenous shock needs to be captured through the openness variable. The degree of government intervention is measured by the ratio of local government general budget revenue to GDP, and the level of openness is represented by the total value of goods imports and exports (in billions of yuan). The experimental results are shown in

Table 10. Adding control variables.

	(1)	(2)	(3)	(4)
VARIABLES	EUE	TEC	EUE	TEC
DED	2.945 ***	−1.638 ***	2.887 ***	−1.554 ***
	(3.45)	(−3.18)	(3.55)	(−3.21)
PS	−0.000 ***	0.000 ***	−0.000 ***	0.000 ***
	(−2.88)	(4.25)	(−3.04)	(4.35)
IS	−0.003	−0.013 ***	−0.004	−0.012 ***
	(−0.51)	(−3.34)	(−0.58)	(−2.96)
ECS	−1.719 ***	0.590 ***	−1.711 ***	0.600 ***
	(−5.23)	(3.43)	(−4.59)	(3.26)
GDP	0.000 ***	0.000 ***	0.000 ***	0.000
	(2.59)	(3.33)	(3.32)	(1.49)
UR	−1.706 *	−0.259	−2.344 ***	0.190
	(−1.82)	(−0.49)	(−2.63)	(0.37)
EDU	0.372 ***	−0.092 **	0.351 ***	−0.077
	(4.40)	(−1.98)	(4.15)	(−1.64)
DGI	−2.602 **	1.911 ***
	(−2.34)	(3.72)
DOOW			−0.000 *	0.000
			(−1.75)	(1.62)
Constant	0.958	1.645 **	1.723	1.142
	(0.71)	(2.13)	(1.26)	(1.39)
Year	YES	YES	YES	YES
Province	YES	YES	YES	YES
Observations	360	360	360	360
R-squared	0.963	0.988	0.963	0.988

Robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

.

After introducing the level of openness and the degree of government intervention as control variables, the digital economy development index (DED) continues to show a statistically significant positive effect on energy efficiency (EUE) (column 1 coefficient = 2.945 ***, column 3 coefficient = 2.887 ***), and a statistically significant negative impact on energy consumption (TEC) (column 2 coefficient = −1.638 ***, column 4 coefficient = −1.554 ***), further validating Hypothesis H1.

5. Mechanism Analysis

We have established the relationship between the digital economy, energy consumption, and energy efficiency. But what are the mechanisms behind this relationship? In the theoretical analysis section, we identified that the digital economy may primarily reduce energy consumption and improve energy efficiency through green innovation, industrial upgrading, and digital inclusive finance. Based on this, we conduct a mechanism analysis using a stepwise approach to detect the mediating effects.

According to columns (1) and (3) in

Table 11. Individual mediating effects.

	(1)	(2)	(3)	(4)
VARIABLES	GI	EUE	AS	TEC
GI		0.090 **
		(1.98)
AS				−0.170 ***
				(−2.85)
DED	2.317 **	2.352 ***	2.759 ***	−0.888 *
	(2.39)	(2.99)	(6.40)	(−1.84)
PS	0.000	−0.000 ***	−0.000 **	0.000 ***
	(1.47)	(−3.09)	(−2.49)	(4.91)
IS	−0.014 **	−0.004	0.037 ***	−0.004
	(−2.20)	(−0.83)	(13.26)	(−1.14)
ECS	0.130	−1.850 ***	0.382 **	0.742 ***
	(0.33)	(−5.89)	(2.20)	(4.04)
GDP	−0.000 ***	0.000 ***	−0.000 ***	0.000 **
	(−2.64)	(3.26)	(−5.78)	(2.31)
UR	4.113 ***	−2.558 ***	−2.029 ***	−0.250
	(3.97)	(−2.99)	(−4.40)	(−0.50)
EDU	−0.094	0.367 ***	0.129 ***	−0.060
	(−1.00)	(4.88)	(3.10)	(−1.36)
Constant	5.210 ***	0.780	0.161	1.457 **
	(3.60)	(0.66)	(0.25)	(2.15)
Year	YES	YES	YES	YES
Province	YES	YES	YES	YES
Observations	360	360	360	360
R-squared	0.976	0.963	0.983	0.988

Robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

and column (1) in

Table 12. Joint mediating effects.

	(1)	(2)	(3)
VARIABLES	DF	TEC	EUE
DF		−0.003 **	0.012 ***
		(−2.58)	(8.51)
DED	95.015 ***	−1.087 **	1.447 *
	(3.92)	(−2.17)	(1.79)
PS	−0.008	0.000 ***	−0.000 **
	(−1.48)	(3.72)	(−2.24)
IS	0.068	−0.010 ***	−0.006
	(0.43)	(−2.72)	(−1.16)
ECS	−19.559 **	0.622 ***	−1.609 ***
	(−2.09)	(3.66)	(−4.87)
GDP	0.000 ***	0.000 ***	0.000 **
	(2.77)	(3.48)	(1.98)
UR	−65.794 *	−0.091	−1.416 *
	(−1.76)	(−0.18)	(−1.90)
EDU	2.793	−0.074	0.326 ***
	(1.12)	(−1.61)	(4.33)
Constant	133.142 ***	1.806 **	−0.311
	(3.50)	(2.30)	(−0.25)
Year	YES	YES	YES
Province	YES	YES	YES
Observations	360	360	360
R-squared	0.997	0.988	0.968

Robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

, it can be observed that the digital economy significantly improves green innovation (coefficient = 2.317 **), industrial upgrading (coefficient = 2.759 ***), and digital inclusive finance (coefficient = 95.015 ***). Furthermore, column (3) in Table 10 shows that the coefficient for industrial upgrading is −0.170 ***, indicating that industrial upgrading can reduce energy consumption, thus validating Hypothesis H2. Column (2) in Table 10 shows that the coefficient for green innovation is 0.090 **, indicating that the conclusion of Hypothesis H3 is reliable, and improving green innovation levels can enhance energy efficiency. Columns (2) and (3) in Table 11 show that the digital finance index has coefficients of −0.003 ** for energy consumption and 0.012 *** for energy efficiency, further supporting the validity of Hypothesis H4. The impact of the digital economy on energy consumption and energy efficiency is realized through three core pathways: green innovation, industrial upgrading, and digital inclusive finance.

6. Heterogeneity Analysis

The value of studying regional heterogeneity in the effects of the digital economy on energy consumption and energy efficiency lies in the significant spatial variation of intrinsic geographical endowments within China. Due to inherent differences in natural resource distribution, ecological carrying capacity, and climatic conditions across regions, combined with the varying development stages of economic growth and industrial structure, the effects of digital technologies on energy system transformation exhibit spatial differentiation patterns. Existing research indicates a non-linear relationship between regional development levels and the energy regulation capacity of the digital economy [73]. This characteristic is especially prominent in the comparison between the western and non-western regions—where the former is constrained by traditional energy dependence and the matching of emerging technologies, while the latter benefits from the aggregation effects of market mechanisms and innovation factors. Following the official classification by China’s National Bureau of Statistics (NBS) and existing literature [32,62], we divide the 30 sample provinces into western and non-western regions. The western region includes 12 provinces: Chongqing, Sichuan, Guizhou, Yunnan, Tibet, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang, Guangxi, and Inner Mongolia. The remaining 18 provinces constitute the non-western region.

Columns (1) and (2) in

Table 13. Heterogeneity analysis.

	No West (1)	West (2)	No West (3)	West (4)
VARIABLES	TEC	TEC	EUE	EUE
DED	−0.100	−14.626 ***	2.202 ***	13.933 ***
	(−0.23)	(−7.95)	(2.67)	(3.41)
PS	0.000 ***	0.000 **	−0.000 ***	0.000
	(3.70)	(2.00)	(−2.75)	(0.88)
IS	−0.007 ***	0.005	−0.020 ***	−0.016 **
	(−2.66)	(0.98)	(−2.62)	(−2.22)
ECS	0.217	0.408	−0.699	−1.948 ***
	(1.00)	(1.40)	(−1.53)	(−3.87)
GDP	0.000 ***	0.000 ***	0.000 **	0.000
	(2.86)	(3.89)	(2.31)	(0.83)
UR	1.573 ***	−5.210 ***	−3.065 ***	1.347
	(3.77)	(−3.60)	(−3.02)	(0.67)
EDU	−0.047	−0.005	0.318 ***	0.211 **
	(−1.20)	(−0.08)	(3.64)	(2.06)
Constant	−0.075	2.307 **	3.114 **	−0.857
	(−0.11)	(2.53)	(2.17)	(−0.84)
Year	YES	YES	YES	YES
Province	YES	YES	YES	YES
Observations	228	132	228	132
R-squared	0.994	0.975	0.970	0.968

Robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05.

show that in both the western and non-western regions, the digital economy significantly reduces total energy consumption. Additionally, columns (3) and (4) in Table 12 indicate that the digital economy significantly improves energy efficiency in both regions, with results remaining statistically significant at the 1% level. However, a comparison reveals that the coefficient for the impact of the digital economy on energy consumption is not significant in the non-western regions. While the coefficient for the digital economy’s impact on energy efficiency is significant in both the western and non-western regions at the 1% level, the magnitude of the values differs substantially.

The non-western region has a higher share of the service sector, and the potential for improving energy efficiency in traditional industries is relatively limited. In contrast, the western region has a larger share of heavy industry, where the application of digital technologies can significantly reduce energy consumption in high-energy-consuming equipment. Furthermore, the path dependence of energy system structures means that in the western region, renewable energy advantages can be leveraged to enhance dispatch efficiency through digital technologies, effectively overcoming energy absorption bottlenecks. In the non-western region, where fossil fuels dominate, technological improvements face physical limits. Additionally, the spatial differentiation of policy interventions has led the western region to form a synergistic mechanism for digital infrastructure and energy transition under specialized policy support, while the non-western region relies on market-driven, gradual optimization paths. These factors collectively shape the regional differentiation in the effects of the digital economy on energy systems.

7. Conclusions and Recommendations

This study, based on provincial panel data from 2011 to 2022 in China, systematically investigates the impact of the digital economy on energy consumption and energy efficiency, as well as its underlying mechanisms. The empirical results show that the development of the digital economy significantly reduces the scale of energy consumption and enhances energy efficiency through technological penetration and structural transformation. The use of the instrumental variable method (interaction term between fixed-line telephones and information technology service revenue) and multiple robustness tests (including lagged variables, indicator replacement, sample trimming, and additional control variables) further confirms the reliability of the core conclusions. The mechanism analysis reveals that the digital economy drives the optimization of the energy system through three synergistic paths: green technological innovation, industrial upgrading, and digital inclusive finance. Additionally, heterogeneity analysis indicates that the impact of the digital economy on reducing energy consumption and improving energy efficiency is significantly stronger in the western regions than in the non-western regions, primarily due to differences in industrial structure, policy environment, and market mechanisms. In conclusion, the development of the digital economy plays a key role in reducing energy consumption and improving energy efficiency.

Based on the above conclusions, this study proposes the following recommendations:

1.

Prioritize Differentiated Digital-Energy Integration Policies Based on Regional Heterogeneity. Our heterogeneity analysis confirms that the digital economy’s effect on reducing energy consumption and improving efficiency is significantly stronger in western regions. Therefore, policy support and investment in digital infrastructure (e.g., 5G, IoT) should be strategically prioritized in western provinces. The focus should be on deploying digital technologies to optimize their abundant renewable energy sources and retrofit energy-intensive traditional industries. In contrast, policy in eastern regions should emphasize leveraging digitalization for marginal efficiency gains in service sectors and advancing R&D in next-generation energy-saving technologies.

2.

Channel Policy Support through the Three Validated Mediating Pathways. Our mechanism tests confirm that green innovation, industrial upgrading, and digital finance are key transmission channels. Accordingly, for green innovation, establish government-backed R&D funds specifically for digital-enabled green technologies (e.g., AI for material science, blockchain for battery recycling), as our results show GT significantly improves energy efficiency. For industrial upgrading, develop targeted subsidies to help traditional manufacturing sectors (e.g., steel, chemicals) adopt industrial internet platforms and smart production systems, directly activating the AS mechanism that suppresses energy consumption. For Digital Finance, encourage financial institutions to develop “green digital loan” products via regulatory sandboxes, using big data for credit assessment to fund SME energy-efficient retrofits and thereby strengthen the DF pathway.

3.

Implement “Green by Design” Standards for Digital Infrastructure to Mitigate its Energy Footprint. This directly addresses the “digital energy paradox” noted in the introduction—the fact that the digital economy itself creates new energy demands. We recommend introducing mandatory energy efficiency standards (e.g., PUE thresholds) and green power consumption requirements for data centers and other digital infrastructure. Policies should further incentivize the construction of new data centers in western regions to leverage their renewable energy potential, thereby turning a core component of the digital economy’s own consumption into a driver for cleaner energy deployment.

This study also has some limitations, as shown below:

This study, while providing insights into the digital economy’s impact on energy in China, has certain limitations that also indicate directions for future research. First, the findings are context-specific, derived from China’s unique policy-driven model and institutional setting. The extent to which these results can be generalized to other market-driven economies may be limited. However, the methodological framework and the identified mechanisms (green innovation, industrial upgrading, digital finance) offer a valuable template for analyzing similar relationships in other countries. Second, our regional heterogeneity analysis, though informative, employs broad geographical classification. Future research could benefit from more granular groupings based on specific criteria such as resource endowment or industrial structure concentration to uncover more nuanced patterns.

Furthermore, while this study confirms the significant mediating roles of green technological innovation (GT), industrial upgrading (AS), and digital inclusive finance (DF), its analysis at the aggregate level presents avenues for future refinement. Subsequent research could delve deeper by (1) deconstructing green innovation into sub-dimensions (e.g., green product vs. process innovation) to pinpoint the digital economy’s precise catalytic pathways; (2) distinguishing between ‘inter-industry structural change’ and ‘intra-industry efficiency gains’ within industrial upgrading to clarify the fundamental source of energy savings; and (3) incorporating micro-level mechanisms such as ‘resident digital consumption’ habits and macro-level ‘cross-regional energy dispatch’ facilitated by digital platforms, which were mentioned in the introduction but warrant separate, focused investigation. Exploring these nuanced channels will provide an even more granular understanding of the digital economy’s impact on the energy system.

Finally, this study employs the reciprocal of energy intensity to measure energy efficiency. While this is a well-established metric in the literature, it does not distinguish between pure technical efficiency gains and shifts in industrial structure, nor does it account for differences in energy quality across provinces. Future research could adopt more nuanced measurements, such as calculating quality-weighted energy consumption or employing decomposition techniques like Structural Decomposition Analysis (SDA), to isolate the distinct effects of the digital economy on technical energy efficiency.