Is the Digital Divide Inhibiting Urban Energy Transitions?—Evidence from China

Abstract

The swift advancement of information technology has significantly impacted the energy transition. Being the largest energy consumer globally, China’s acceleration of the urban energy transition will promote sustainable development and pave the way for future development. This study examines the impact of the digital divide between cities regarding the energy transition by using panel data for 271 Chinese cities from 2006 to 2021. We found the following results. (1) The digital divide has an inhibitory effect on the energy transition of cities, hindering their development towards green and low-carbon development. (2) Further analyses found that the negative impact of the digital divide on energy transition can be effectively mitigated by strengthening urban infrastructure construction, promoting emerging technological innovation, and cultivating and attracting talent in the digital industry. (3) The results of the subsample analyses show that the inhibitory effect of the digital divide on energy transition is more significant in densely populated cities, economically and technologically developed cities, and non-resource-based cities. The above findings hold significant practical implications for overcoming the digital divide and the stagnation of energy transition, and for the scientific implementation of China’s “Carbon Neutrality” initiative.

1. Introduction

The 2024 Emissions Gap Report published by the United Nations Environment Programme highlights that global climate change is escalating into a matter of grave concern. In 2023, annual global greenhouse gas emissions increased by 1.9%, reaching a record-high of 57.1 gigatonnes of carbon dioxide equivalent. Nations must implement more robust measures in the new round of Nationally Determined Contributions, or the Paris Agreement’s 1.5 °C target will become unattainable within a few years, resulting in catastrophic consequences for humanity. The majority of global carbon dioxide emissions originate from energy activities, and there is a general consensus within the international community that global climate change must be addressed by actively developing and utilizing clean energy and controlling greenhouse gas emissions. Under such circumstances, the traditional economic growth model propelled by high-carbon energy has become unsustainable. Thus, it is imperative for the global energy sector to expedite the shift towards carbon-neutral energy [1].

The energy transition has garnered significant attention in the academic community as a crucial approach to achieve green and sustainable development [2,3,4]. Yu Yang and colleagues [5] suggested that the energy transition is not simply a straightforward substitution of one energy source for another; rather, it represents a profound transformation of the core components of the energy system into a novel arrangement of energy services. This process unfolds across an extended chain and intricate system, encompassing diverse facets of energy production, storage, transmission, and consumption, along with advancements in energy technology, management practices, and energy security. In a study by Fan Ying and Yi Bowen [6], it was also emphasized that energy transition implies a comprehensive and profound fundamental change within the energy system. Furthermore, the three key drivers of energy transition (increasing the penetration of renewable energy in the energy supply mix, electrification, and improving energy storage) are all conducive to environmental sustainability. A study by Kocaarslan and Soytas [7] based on empirical evidence revealed that cleaner energy sources play a more crucial role in curbing greenhouse gas emissions and ensuring environmental sustainability. The International Energy Agency has emphasized that electrification is among the most vital strategies for cutting down CO₂ emissions within the energy sector. Additionally, the increasing reliance on variable renewable energy sources will require energy storage to play a pivotal role in balancing generation and consumption patterns [8].

As the largest producer and consumer of energy globally, China exerts a significant influence on global climate change and environmental governance. The country’s fossil fuel endowment, characterized by an abundance of coal and a scarcity of oil and gas, shapes its coal-centric energy production and consumption pattern, posing challenges to the attainment of the carbon neutrality goal. This reality urgently requires China to transform traditional economic growth and productivity development paths and accelerate the shift to a new model based on renewable energy [9].

At the same time, as the global Internet experiences exponential growth, humanity has stepped into a new era of informatization. Amid this transformation, the issue of digital development disparities is coming to the fore. The Chinese government’s ‘14th Five-Year Plan for Digital Economy Development’ highlights that China’s digital economy faces notable issues of uneven, inadequate, and erratic development. Additionally, the digital divide across different industries, regions, and groups remains unresolved, leading to considerable regional imbalances in the development of the digital economy. The digital divide among regions, resulting from uneven development in the previous stage, is widening [10], and this situation is likely to exert a far-reaching influence on the energy transition. Specifically, the digital divide may impact the equitable distribution and efficiency of energy use. For example, some cities are able to make better use of digital technologies to optimize energy management and improve energy use efficiency due to the high penetration and application of ICTs, while other cities may face inefficient energy use and the inequitable distribution of energy due to the insufficient application of ICTs.

In light of the aforementioned realities, this study raises the following questions. Does the digital divide inhibit the realization of energy transition goals? If so, can appropriate actions be implemented to narrow the digital divide and facilitate a seamless energy transition? Answering the above questions will help us to accurately grasp the relationship between the impact of the digital divide on energy transformation, the transmission mechanism and the law of action, which is vital for accelerating the green and low-carbon transformation of China’s energy structure, realizing the absolute ‘decoupling’ of economic growth and carbon emissions, and achieving the goal of carbon neutral development.

The potential marginal contributions of this study are evident in the following three aspects: (1) In the context of digitalization and low-carbon development, this study introduces the digital divide into the research framework of energy transition in an innovative manner. To some extent, it broadens the research scope of the field of energy and environmental economics. Moreover, it provides practical directions for cities to formulate energy transition strategies and overcome the transition barriers brought about by the digital divide. (2) An evaluation system for urban energy transition is constructed from three dimensions: energy consumption, industrial structure, and energy efficiency. The entropy method is adopted to calculate and analyze the degree of energy transition in Chinese cities. Compared with previous studies, the energy transition evaluation system has a wider coverage and a more reasonable evaluation logic, which can more comprehensively and objectively reflect the actual situation of urban energy transition. (3) Commencing from three perspectives: reinforcing urban infrastructure development, fostering emerging technological innovation, and nurturing and attracting talent in the digital industry, this study further reveals some economic mechanisms that can effectively mitigate the negative impact of the digital divide on energy transition, providing solid theoretical support for subsequent related research and policy-making.

The remaining content is structured as follows: Section 2 offers a review of the relevant literature. In Section 3, the research hypotheses are proposed. Section 4 details the research methods and data sources. Section 5 carries out an empirical analysis. Finally, Section 6 presents the research conclusions, points out the limitations, and provides policy recommendations.

2. Literature Review

2.1. The Digital Divide

The term “Digital Divide”, often called the information divide, first emerged in 1989 in an article titled “Digital Divide”, published in The Educational Supplement of the United Kingdom. In 1999, the United States National Telecommunications and Information Administration (NTIA) also brought up the concept in its report titled ‘Being Left Behind on the Net: Defining the Digital Divide’. The digital divide was once considered a phenomenon in the field of information technology (IT), including IT development and IT utilization, and in particular, considered the reality of the economic and social gap due to the uneven development of network technologies [11,12,13]. With a deepening understanding of the digital divide, it has gradually developed from the initial binary concept into a complex, multi-level, and multi-form dynamic phenomenon [14]. Currently, academics consider the digital divide from three main perspectives: the ICT access gap, the ICT use and skills gap, and the gap in the benefits derived from ICT use, i.e., the gap of accessibility, the gap of usage, and gap of output [13,15,16]. As Internet penetration increases, the digital divide at the access level has continued to narrow, and the digital divide has shifted to a second digital divide, that is, the ICT usage gap [17,18]. Regarding the causes of the digital divide, scholars commonly concur that it stems from the interplay of macro and micro elements. On the macro level, economic, geographical, social, cultural, and policy factors come into play. At the micro level, an individual’s skills, educational attainment, income, gender, age, and similar factors contribute to the digital divide [19,20,21]. Although relevant research has made progress in bridging the digital divide, it still prevails and has become an inescapable problem in modern society [22]. If the digital divide cannot be effectively addressed, it may become a new source of development imbalance between and within countries [23].

Many studies have focused on the range of social inequalities caused by the digital divide. Yan Hui [24] defined the scope of digital poverty, pointing out the performance of digitally impoverished groups in digital material entities, digital services, digital psychology, digital capabilities, digital endeavors, digital social norms, digital support, and digital impacts. He Zongyue and Song Xuguang [25] utilized the matched data from the China Digital Inclusive Financial Development Index and the China Family Panel Studies to investigate the influence of digital economic development on multidimensional household poverty. A study by Zhang Xun et al. [26] showed that the digital divide has brought about new inequalities in opportunities, making it impossible for residents to enjoy the dividends from the rapid development of the Internet industry among themselves, which may further lead to the widening of the gap between the rich and the poor and an increase in the incidence of poverty. Luo Tingjin and Cha Hongwang [27], on the other hand, emphasized reductions in poverty by narrowing the digital divide in order to break the vicious circle leading to the digital divide and poverty. Peng Bo and Yan Feng [12] argued that the existence and widening of the digital divide would make China’s Internet geographically show a four-level difference. In addition, under the urban–rural dichotomy, rural development is gradually lagging behind in the digital wave because of insufficient network coverage, slow Internet speeds, and the low penetration rates of computers and other terminal devices in remote rural areas [28]. With the aging of Chinese society as a whole, new structural inequalities such as the digital divide in old age will develop [29,30].

2.2. Factors Affecting Energy Transition

Many scholars have also conducted a large number of in-depth studies on the energy transition, covering the measurement of the transition indicators and factors affecting the energy transition. In terms of energy transition indicator measurement, academics mainly use a single indicator, including the proportion of fossil energy consumption and energy consumption per unit of GDP for measurement [31,32], or use a multidimensional indicator system that includes comprehensive indicator empowerment for measurement [33,34]. When it comes to the driving forces behind energy transition, Fischer et al. [35] indicated that energy transition is a complex, protracted and irreversible process. It is influenced by multiple economic and social elements and has distinct policy-driven traits. Belaïd et al. [36] used a socio-technical systems model to qualitatively analyze Saudi Arabia’s energy transition, and found that the government’s institutions and policies are the main constraints. Mo Jianlei et al. [37] affirmed the role of renewable energy subsidy policies in controlling carbon emissions by predicting the future evolutionary trend of carbon emissions and energy consumption in China. Overall, two kinds of specific policy tools have propelled the energy transition process. The first category consists of incentive policies for new energy development, like direct subsidies and tax reductions. The second category involves regulatory policies for traditional fossil energy consumption, such as carbon quotas and carbon taxes.

The energy transition is also influenced by factors such as regional resource endowment and technological innovation. Zhao et al. [38] analyzed the energy transition using resource endowment, policy, and technological conditions as explanatory variables and found that resource endowment plays an important role in urban energy transition. Yang et al. [39] showed that renewable energy innovations can significantly improve the structure and efficiency of sustainable energy transition. Some scholars delved into the trends, socio-economic driving forces, and environmental effects of Beijing’s energy transition process over the past few decades. Their findings indicated that energy structure diversification, characterized by the trends of reducing coal use, promoting clean energy, and restructuring the economy, has promoted the energy transition [40].

2.3. The Impact of Digitization on Energy Transition

The rapid development of digital technologies has provided unprecedented opportunities to reshape the energy landscape. Artificial intelligence, big data, cloud computing, blockchain and other digital technologies possess features like high-level innovation, strong penetration abilities, and wide coverage. These technologies exert a profound influence on the processes of energy production, transportation, and consumption [41,42]. For example, different levels and segments, such as new energy exploration, capacity forecasting, load scheduling, P2P energy trading platforms, carbon emission monitoring, and carbon reduction design, can all be optimized through digital technologies, and move towards high-end intelligence [43]. Lee et al. [44] also found that AI will also have a positive impact on energy transformation in their recent study.

On the one hand, while some scholars have noted the social inequalities caused by the digital divide and some have explored the impacts of digital technologies (or the Internet) on the energy transition, there is little literature that correlates the two and explores the combined impacts of ICT access (the access divide), ICT use (the usage divide), and ICT skills (the output divide) on the energy transition in different regions. On the other hand, considering that cities are the main consumers of energy, they are bound to be the main battlegrounds and key places for the country’s energy transition. Therefore, this study uses the panel data of 271 Chinese cities from 2006 to 2021 to examine the impact of the digital divide on the urban energy transition and to extend the analysis of moderating effects and regional heterogeneity.

3. Theoretical Analysis and Research Hypothesis

Utilizing digital technology in energy production, management, transmission, and consumption can improve quality, enhance efficiency, and accelerate efforts towards energy conservation and emission reduction [45]. Some scholars also showed that the introduction of ICT applications into the manufacturing process will, in the short term, promote technological progress in enterprise production or changes in the use of factors, thereby reducing energy consumption. The digital divide creates differences in individuals’ access to and use of information resources among groups with different connections, hence creating ‘a gap between the information rich and the information poor’ [46,47,48]. Cities at the digital disadvantage end of the spectrum lack foresight and innovation in energy system upgrading and transformation planning due to insufficient network coverage, the inadequate construction of digital platforms, and the inability to keep abreast of cutting-edge technologies such as new types of energy storage and smart grid optimization solutions. Those cities still rely on the traditional energy model, which hinders their progress towards a clean energy system. Particularly within the energy production chain, cities on the disadvantaged side of the digital divide lack the ability to utilize digital technology for real-time monitoring and automated management. This situation impedes the transformation of energy production towards greater intelligence and efficiency, ultimately resulting in inefficiencies. In the consumption chain, due to the existence of the gap between cities, it is difficult to comprehensively use new digital technology such as blockchain to achieve the efficiency and transparency of energy transactions. It can be seen that the digital divide is not conducive to achieving the accurate allocation and efficient use of energy resources. In light of this, we put forward the following hypothesis:

Hypothesis 1. 

The digital divide inhibits the energy transition of cities.

At the regional level, the digital divide is mainly reflected in the regions that lack human, material and financial resources for the construction of digital infrastructure in the early stage, resulting in an environment with poor information access and a gradually expanding access gap; secondly, due to the different levels of Internet penetration, regional economic endowment and other factors brought about by the differences in digital literacy, the use gap and output gap are gradually forming [14]. At the same time, the digital divide will be profoundly changed by the combined influence of a variety of factors such as technological innovation, policy orientation, and socio-economic structural changes. The construction of infrastructure has the potential to boost the accessibility and utilization of information and communication technologies. This, in turn, can enhance the overall digital proficiency of the local population and effectively mitigate the issue of the digital divide [49,50,51]. As a result, it can reduce the negative impact that the digital divide has on the energy transition in cities. Therefore, in the complex association between the digital divide and urban energy transition, the role of infrastructure cannot be ignored. The following hypothesis is proposed in this study:

Hypothesis 2. 

Strengthening infrastructure inhibits the expansion of digital divide and its negative impact on energy transition.

In terms of energy production, cities can rely on advanced automation control and other technologies to promote the fine monitoring and regulation of energy production and promote the intelligent transformation of the production model [45]. In terms of the energy consumption structure, emerging technologies open up new avenues for the development and utilization of renewable energy sources. They lower the barriers and costs associated with using new energy, which can help cities raise their proportion of non-fossil energy, optimize their consumption structure, and decrease their reliance on fossil energy [52]. In terms of industrial structure upgrading, emerging technologies bring about new industries and drive the digital and intelligent transformation of traditional industries. This, in turn, further advances the deepening of the industrial division of labor and promotes industrial development [53]. For cities with a digital divide, new technologies and processes promote the transformation of traditional high-energy-consuming industries to low-energy-consuming, high-value-added emerging industries. In addition, technological innovation can improve product quality, optimize service processes, and reduce rework and product obsolescence rates, thus reducing production costs per unit of product and enhancing energy efficiency [54]. Emerging technological innovation can reduce the inhibiting effect of the digital divide on urban energy transition. Based on the above content, this study proposes the following hypothesis:

Hypothesis 3. 

Emerging technological innovation inhibits the expansion of the digital divide and its negative impact on energy transition.

Moreover, human talent, being a fundamental aspect of enterprise digital transformation [55], exerts a significant influence on the process of urban energy transformation that cannot be overlooked. On the one hand, the knowledge externalities generated by talent can accelerate the spread of digital technologies in the energy sector [56] and promote energy transformation. Specifically, digital talent can effectively optimize resource allocation and achieve complementary resource advantages through continuous diffusion and circulation in the core agglomeration and peripheral areas [57]. This talent space spillover effect can effectively bridge the digital divide [58], and thus, in energy development, use, management, and other aspects, digital technology can achieve a qualitative leap. On the other hand, the economic effect of digital talent concentration can significantly reduce the cost of technology application, providing an economically viable basis for energy transformation. It can be seen that cities actively cultivating or introducing digital industry-related talents significantly weaken the negative obstacles imposed by the digital divide on the energy transition. Hence, the following hypothesis is put forward in this study:

Hypothesis 4. 

Cultivating or introducing digital industry-related talents will effectively inhibit the digital divide and its negative impact on energy transition.

4. Research Design

4.1. Econometric Model Setting

To study the impact of the digital divide on energy transition, this study constructed the following econometric model:

$${\mathrm{ET}}_{\mathrm{it}}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1{\mathrm{DD}}_{\mathrm{it}}+{\mathrm{\beta }}_2{\mathrm{Control}}_{\mathrm{it}}+{\mathrm{\gamma }}_{\mathrm{i}}+{\mathrm{\delta }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{it}}$$

(1)

In this equation,$\mathrm{i}$ denotes the city, $\mathrm{t}$ denotes the year, ${\mathrm{ET}}_{\mathrm{it}}$ represents the level of energy transition of city $\mathrm{i}$ in year $\mathrm{t}$, ${\mathrm{DD}}_{\mathrm{it}}$ denotes the digital divide between cities, and the regression coefficient ${\mathrm{\beta }}_1$ denotes the impact of the level of digital divide on the energy transition of the city. ${\mathrm{Control}}_{\mathrm{it}}$ denotes a set of control variables including government disposable power (${\mathrm{GCP}}_{\mathrm{it}}$), the strength of investment in science, technology and innovation (${\mathrm{ISTI}}_{\mathrm{it}}$), and the per capita gross regional product (${\mathrm{PRGDP}}_{\mathrm{it}}$). In addition, taking into account the exclusion of realities due to differences in resource endowments and other aspects of individual cities, this study includes individual fixed effects ${\mathrm{\gamma }}_{\mathrm{i}}$ in the model. At the same time, seasonal price fluctuations, domestic and foreign macroeconomic policy changes and other unobservable factors are generated, and we also include the year fixed effects in the model, denoted as ${\mathrm{\delta }}_{\mathrm{t}}$. Then, ${\mathrm{\epsilon }}_{\mathrm{it}}$ is a random perturbation term.

4.2. Variable Measures

4.2.1. Explained Variable

China’s energy transition aims to achieve peak carbon and carbon-neutral development goals, focusing on three main aspects: energy consumption, industrial structure, and energy efficiency. Referring to the study by Lin and Cheung [59], the Energy Transition Index (ET) adopts the three aspects of energy efficiency, energy consumption (EC), and industrial structure upgrading to construct a multidimensional quantitative analysis framework to measure urban energy transition. Concurrently, taking into account the practicality and availability of data, this study selects the green total factor energy efficiency (EE) as an indicator to represent the energy use efficiency in China. Regarding the precise measurement of the green total factor energy efficiency, with reference to the existing relevant studies [60,61], the labor, capital, and energy consumption of the city are selected as input factors, the gross regional product is used as the desired output, and industrial sulfur dioxide, industrial soot, and industrial wastewater emissions are used as undesirable outputs, being quantified using the SBM–Malmquist–Luenberger index approach to assess the green total factor energy efficiency of each prefecture-level city. The final energy consumption of each city was measured using the dataset of Yang Guanglei et al. [62], which provided comprehensive information on the final energy consumption of 30 fossil fuels and four clean energy sources (i.e., nuclear energy, hydroelectricity, wind energy, and solar energy) in seven industrial sectors in 331 cities in China from 2005 to 2021. For the measurement of industrial structural upgrading (IS), the different degrees of impact generated by the three major industries on energy transition are considered comprehensively and the value is obtained according to Equation (2):

$${\mathrm{IS}}_{\mathrm{it}}={\mathrm{AGR}}_{\mathrm{it}}+2{\mathrm{IND}}_{\mathrm{it}}+3{\mathrm{SER}}_{\mathrm{it}}$$

(2)

In this equation, ${\mathrm{AGR}}_{\mathrm{it}}$ represents the proportion of the added value of the primary industry to the city’s GDP in that year, ${\mathrm{IND}}_{\mathrm{it}}$ represents the proportion of the added value of the secondary industry to the GDP, and ${ \mathrm{SER}}_{\mathrm{it}}$ represents the proportion of the added value of the tertiary industry to the GDP. This study standardized the three aforementioned indicators, and then employed the entropy weight method to calculate the energy transition index. Based on the results, a map depicting the evolution of the spatial pattern of energy transition in Chinese cities was created (see Figure 1, which only shows the years 2006, 2011, 2016, and 2021 due to space limitations).

4.2.2. Explanatory Variable

The core explanatory variable in this study is the digital divide (DD). Based on the definition of the ICT development index provided by the International Telecommunication Union, the index system of the informatization development index formulated by the National Bureau of Statistics of China, and on relevant studies [58,63], the index system for measuring the level of ICT development is constructed from the three dimensions of ICT access, use and skills, including three secondary indicators and seven tertiary indicators, as shown in

Table 1. Indicator system for measuring the development level of information access technology.

Primary Indicator	Secondary Indicator	Tertiary Indicator	Description of Indicators
Information and Communication Technology Development Index	Information and Communication Technology Acquisition (Gap in Access)	Fixed Telephone Penetration Rate	Fixed telephone users at the end of the year as a ratio to the total population
		Mobile Telephone Penetration Rate	Mobile telephone users at the end of the year as a ratio to the total population
	Information and Communication Technology Usage (Gap in Use)	Internet Penetration Rate	Internet users as a ratio to the total population
		Telecommunications Business Volume	——
		Per Capita Regional Gross Product	——
	Information and Communication Technology Skills (Gap in Output)	Secondary Education Penetration Rate	Number of students in general secondary schools as a ratio to the total population
		Higher Education Penetration Rate	Number of students in general colleges and universities as a ratio to the total population

. The comprehensive evaluation level of ICT development is not only related to specific indicators but also to the weighting of each indicator. Therefore, the ICT development level of each city in this study was measured using the entropy value method in the objective assignment method.

For the measurement of the digital divide, referring to the study by Duan Jiran et al. [58], the locational entropy method was used to measure the digital divide (DD) of cities by comparing the ICT development level of each city with the national average, and this value was used as the independent variable in this study. The formula for calculation is as follows:

$${\mathrm{DD}}_{\mathrm{it}}= {\displaystyle \frac{{\mathrm{IDI}}_{\mathrm{it}}}{{\mathrm{IDI}}_{\mathrm{t}}}}$$

(3)

In this equation,${\mathrm{IDI}}_{\mathrm{it}}$ denotes the level of information access technology development of city $\mathrm{i}$ in year $\mathrm{t}$, ${\mathrm{IDI}}_{\mathrm{t}}$ denotes the average level of information access technology development of the entire country in year $\mathrm{t}$, and ${\mathrm{DG}}_{\mathrm{it}}$ denotes the urban digital divide. At the same time, our study also draws a map of the evolution of the spatial pattern of the digital divide in Chinese cities (see Figure 2, which only shows the years 2006, 2011, 2016 and 2021 due to space limitations).

4.2.3. Moderating Variables

We use moderating effect analysis to investigate the intermediate mechanisms through which the digital divide affects energy transition. The following three moderating variables will be tested in this study [64]. ① The level of infrastructure construction (INF). Infrastructure development is a crucial factor in the dissemination and adoption of digital technologies. INF is measured by the ratio of the urban road area in the region to the total population. ② The level of emerging technology innovation (NTI). NTI reflects a region’s ability to generate and adopt new digital-related technologies. A higher level of innovation means that more digital-based new solutions can be developed to address the digital divide and promote urban energy transition. NTI is measured by the number of invention patents granted. ③ The number of employees in the digital industry (NUM). A large number of practitioners in the information transmission, computer services, and software industries can drive the development and application of digital technologies, thereby regulating the relationship between the digital divide and urban energy transition. NUM is measured by the number of employees in the information transmission, computer services, and software industries.

4.2.4. Control Variables

To further ensure the accuracy of the research, referring to the relevant studies [65], this study incorporates a series of control variables: ① Per capita regional gross domestic product (PRGDP). PRGDP is calculated by dividing the total regional gross domestic product of the city by the total population at the end of the year. Incorporating PRGDP will control the influence of economic output factors on the research results. ② Total industrial output value (TIOV). TIOV reflects the scale and activity of industry, which affects the energy demand. In this paper, TIOV is indicated by the ratio of the current total industrial output value of industrial enterprises above a designated size in the city to the total output. ③ Industrial sulfur dioxide emissions (ISDEs). Controlling this variable can help analyze the impact of the digital divide on energy transformation under different pollution levels. ④ Human capital level (HC). HC is calculated by the number of full-time undergraduate and junior college students in ordinary institutions of higher learning per 10,000 people in each city. ⑤ Advanced of Industrial Structure (AIS). The transformation of the industrial structure can alter the demand and composition of energy consumption. AIS is evaluated by comparing the proportion of the tertiary industry’s added value to GDP with that of the secondary industry’s added value to GDP. ⑥ Government financial control (GCP) uses a proxy of the ratio of the government financial expenditure of each city to the total output. ⑦ Investment effort in education (IE) is gauged by the proportion of the city’s total education expenditure to the total output. ⑧ The Intensity of Investment in Scientific and Technological Innovation (ISTI) is represented by the ratio of the total scientific and technological expenditure of the city to the total output.

4.3. Data Description and Processing

China’s energy consumption is mainly concentrated in cities, and a vast majority of air pollutant emissions are associated with urban energy activities. This study uses city-level panel data in China for the period 2006–2021, including the level of ICT in cities, input–output data on the green total factor energy efficiency, and data on the overall upgrading of industries in cities. A series of control variables are obtained from the National Bureau of Statistics of China, the Urban Statistical Yearbook of past years and the Flush iFinD database. Additionally, this study logarithmically treats variables such as the GDP per capita and adopts linear interpolation for a small number of missing values in the samples. Descriptive statistical analyses of the key variables are detailed in

Table 2. Variable definitions and descriptive statistics.

Variable	Variable Name	Observations	Mean	Standard Deviation	Minimum	Maximum
ET	Energy Transition Index	4336	0.1537	0.1039	0.0302	0.9072
EC	Energy Consumption	4336	0.0021	1.0012	−0.8612	7.6990
IS	Industrial Structural Upgrading	4336	0.0010	1.0016	−3.0295	3.8092
EE	Green Total Factor Energy Efficiency	4336	0.0049	0.9986	−2.2851	6.6236
DD	Digital Divide	4336	1.0012	0.3933	0.3757	3.3017
INF	The Level of Infrastructure Construction	4336	2.2388	0.0895	0.3242	2.7590
NTI	The Level of Emerging Technology Innovation	4336	4.4795	1.9584	0.0000	11.2797
NUM	The Number of Employees in the Digital Industry	3794	0.4254	0.4987	0.0198	4.4649
GCP	Government Financial Control	4336	0.1838	0.1021	0.0426	1.4852
IE	Investment Effort in Education	4336	0.0003	0.0005	0.0000	0.0088
ISTI	Intensity of Investment in Scientific and Technological Innovation	4336	0.0024	0.0024	0.0000	0.0415
TIOV	Total Industrial Output Value	4336	2.1216	0.0805	1.2635	3.2048
ISDE	Industrial Sulfur Dioxide Emissions	4336	12.6916	0.1403	10.9671	13.7840
HC	Human Capital Level	4336	4.5676	1.2138	0.0000	7.2437
PRGDP	Per Capita Regional Gross Domestic Product	4336	10.4899	0.7255	4.5951	13.0557
AIS	Advanced of Industrial Structure	4336	0.9788	0.5547	0.0943	5.3482

, and the results clearly indicate the presence of a marked regional digital divide in China. The degree of energy transformation is also widely different, which is in line with the reality of the imbalance and insufficiency in China’s development. Specifically, the Energy Transition Index (ET) has 4336 observations, with a mean value of 0.1537, a standard deviation of 0.1039, a minimum value of 0.0302, and a maximum value of 0.9072. These figures suggest that there is a certain degree of diversity in the process of energy transition among cities. The Digital Divide (DD) variable also has 4336 observations, with a mean value of 1.0012 and a relatively large standard deviation of 0.3933, which suggests significant differences in the level of digital divide among cities in China. The minimum value of DD is 0.3757, and the maximum value is 3.3017, further highlighting the existence of a pronounced regional digital divide in China.

5. Empirical Results Analysis

5.1. Benchmark Regression

Before conducting the regression, a multicollinearity test was first conducted. In the test results, the maximum VIF was 3.29 and the average VIF was 1.95, which was less than 10, indicating that there was no significant multicollinearity between variables. The baseline regression results for model (1) are shown in

Table 3. Benchmark regression results.

	(1)	(2)	(3)
DD	−0.0315 ***	−0.0156 ***	−0.0242 ***
	(0.0024)	(0.0023)	(0.0024)
GCP		−0.0010	−0.0770 ***
		(0.0088)	(0.0094)
IE		4.6367 ***	5.4902 ***
		(1.0053)	(0.9694)
ISTI		1.4500 ***	0.5355 **
		(0.2785)	(0.2703)
TIOV		−0.0312 ***	−0.0197 **
		(0.0084)	(0.0081)
ISDE		−0.0705 ***	−0.0547 ***
		(0.0046)	(0.0048)
HC		−0.0018 **	−0.0032 ***
		(0.0008)	(0.0008)
PRGDP		0.0390 ***	0.0024
		(0.0013)	(0.0024)
AIS		0.0069 ***	0.0027
		(0.0015)	(0.0017)
Individual Effects	yes	yes	yes
Time Effects	yes	no	yes
Control Variables	no	yes	yes
_cons	0.1853 ***	0.7187 ***	0.9112 ***
	(0.0024)	(0.0637)	(0.0631)
N	4336	4336	4336
r2	0.9483	0.9464	0.9520
F	178.4769	452.7970	55.8431

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

. In column (1), without control variables, the regression coefficient of the digital divide with regard to urban energy transition is 0.0315, passing the 1% significance test. This indicates that there is a significant negative correlation between the digital divide and urban energy transition. Column (2) considers time fixed effects and adds all control variables. The result remains significantly negative. Further, column (3) builds on the basis of column (2) by considering the impact of city-level unobserved effects that do not vary over time. The negative and significant effect of the digital divide on urban energy transition remains. Intuitively, for every one percentage point increase in the regional digital divide, on average, the urban energy transition will be delayed by about 0.0242 percentage points; that is, the larger the digital divide is, the more the urban energy transition is hindered and the more difficult it is to realize an effective shift towards cleaner and more efficient energy use. Therefore, Hypothesis 1 is validated.

Meanwhile, as shown in column (2) of Table 3, under individual fixed effects, the coefficients of the control variables TIOV, ISDE, and HC are significantly negative, indicating that an increase in the total industrial output value of the city, an increase in industrial sulfur dioxide emissions, and an improvement in the level of human capital all hinder energy transition; the coefficients of IE, ISTI, PRGDP, and AIS are positive and significant, meaning that strengthening the intensity of educational investment, enhancing the intensity of scientific and technological innovation investment, increasing the per-capita regional gross domestic product, and optimizing the advancement of the industrial structure contribute to promoting urban energy transition. Under two-way fixed effects, the results are shown in column (3) of Table 3. It is evident that augmenting government regulation, raising the total industrial output value of the city, increasing industrial sulfur dioxide emissions, and elevating the level of human capital pose obstacles to energy transition. In contrast, intensifying educational investment and strengthening investment in scientific and technological innovation contribute to the promotion of urban energy transition.

5.2. Robustness Tests

5.2.1. Replacement of Measures

This study conducted preliminary robustness tests by replacing the explained and explanatory variables. First, for the explanatory variables, the entropy weight method was replaced by calculating the coupled coordination degree of energy use efficiency, energy consumption, and industrial structure upgrading, and the degree of urban energy transition was obtained by re-measurement. Second, the location entropy method was used to calculate the Internet penetration gap (IPG) and the total telecommunication business gap (TPG) of each city and replace the original digital divide variables. In

Table 4. Results of the robustness test.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
DD	−0.0038 ***				−0.0186 ***	−0.0236 ***	0.0013
	(0.0006)				(0.0038)	(0.0023)	(0.0013)
IPG		−0.0365 ***
		(0.0116)
TPG			−0.1026 ***	−0.0100 ***
			(0.0153)	(0.0038)
Individual Effects	yes	yes	yes	yes	yes	yes	yes
Time Effects	yes	yes	yes	yes	yes	yes	yes
Control Variables	yes	yes	yes	yes	yes	yes	yes
_cons	0.0825 ***	1.0068 ***	1.0089 ***	0.0972 ***	0.5504 ***	0.6850 ***	0.9959 ***
	(0.0156)	(0.0633)	(0.0630)	(0.0155)	(0.0627)	(0.0632)	(0.0633)
N	4336	4336	4336	4336	3856	4250	4336
r2	0.4939	0.9509	0.9513	0.4896	0.9224	0.9408	0.9508
F	12.0129	44.6169	48.8988	8.1233	23.2057	45.2480	43.5301

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

, the regression results after substituting the explained variables are presented in column (1). Columns (2) and (3) display the regression results when the explanatory variables are substituted, and column (4) shows the regression results with the substitution of both the explanatory variables and the explanatory variable (TPG). As shown in regression results (1) to (4), the regression coefficients of the digital divide are negative, and all are highly significant at the 1% level. This clearly demonstrates that, even when alternative indicators are utilized, the digital divide still exerts a significant inhibitory effect on urban energy transition. This finding effectively verifies the robustness of the model adopted in our study.

5.2.2. Excluding Special Samples

Considering the possible existence of special samples, this study adopts two methods of bilateral truncation and excludes the sample observations of provincial capital cities and municipalities directly under the central government. (1) Bilateral truncation: Because there may be outliers in the data collection, which may distort the regression results, and because the distribution of the original data may be unsatisfactory, bilateral truncation can optimize the distribution and make the statistical analysis method more effective. The regression results are shown in column (5) of Table 4. (2) Excluding the data of provincial–capital cities and municipalities directly under the central government, provincial capital cities have special characteristics and representative differences in their politics, economy, culture, etc., and the advantage of their concentrated resources makes the data biased and not representative of ordinary cities. Meanwhile, since municipalities directly under the central government enjoy greater economic autonomy compared to ordinary prefecture-level cities, the samples of provincial capital cities and municipalities directly under the central government are removed, and the robustness test is conducted once again. The regression outcomes are shown in column (6) of Table 4. It can be seen that the inhibitory effect of the digital divide on urban energy transition is still significant after excluding the special samples. Specifically, with other variables held constant, after bilateral trimming, for every one-unit increase in the digital divide, the urban energy transition changes in the opposite direction by 0.0186 units. After omitting the data from provincial capitals and municipalities directly under the central government, a one-unit increase in the digital divide results in a change of 0.0236 units in the opposite direction for urban energy transition. These results indicate that the digital divide has a significant negative impact on urban energy transition, and that there are slight differences in the degree of this impact under different data-processing methods.

5.2.3. Counterfactual Test

To further verify the robustness of the previous regression results, this study conducted a counterfactual test. Specifically, the urban digital divide variable was randomly assigned such that it was not associated with the digital divide between actual cities. Based on this, the model with this random assignment incorporated was utilized once more for regression analysis. The regression outcomes are presented in column (7) of Table 4. At this time, the coefficient in the model was 0.0013 and did not reach statistical significance. This result shows that when the digital divide variable is detached from its actual connotation and characteristics, its influence on the evolution of urban energy transformation nearly disappears, confirming that the regression results obtained under the original model setting have good robustness.

This study conducts robustness tests through three testing methods: replacing measurement indicators, eliminating special samples, and counterfactual testing. In summary, a range of test results conclusively show that the suppressive influence of the digital divide on urban energy transition remains consistent across various conditions. This aligns with the conclusions drawn from the benchmark regression analysis and further reinforces the credibility of the research findings.

5.3. Endogeneity Test

The benchmark regression results show a significant negative relationship between the digital divide and energy transition in cities. Although this study integrates other factors affecting energy transition as well as unobservable factors that vary over time and individually, the estimation results may still be biased because of potential endogeneity problems. Firstly, reverse causality might exist. Specifically, the digital divide restricts urban energy transition, while urban energy transition, conversely, could intensify the digital divide. Secondly, the model may have issues with omitted variables. Therefore, to overcome the potential endogeneity problem, this study adopted an instrumental variable approach. Three instrumental variables were selected: lagging the core explanatory variable by one period, that is, assessing the impact of the city’s digital divide situation in the previous year on the level of energy transition in the current period; constructing an instrumental variable based on the exogenous variable of city distance; and constructing an instrumental variable based on historical data. Instrumental variables based on the lag period may alleviate the endogeneity problem caused by two-way causality, and instrumental variables based on distance and historical data are more exogenous, which is conducive to separating the confounding factors in the estimation and enhancing the validity of the estimation results.

5.3.1. Lagged One-Period Variables

Considering that there may be a time lag in the promotion of energy transition by the digital divide, this study first adds the explanatory variables in the lagged period (DDt) to the model; the regression findings are presented in column (1) of

Table 5. Results of the endogeneity test.

	(1)	(2)	(3)
DD		−0.1410 ***	−0.1521 ***
		(0.0081)	(0.0513)
DDt	−0.0224 ***
	(0.0025)
Individual Effects	yes	yes	yes
Time Effects	yes	yes	yes
Control Variables	yes	yes	yes
_cons	0.8419 ***
	(0.0640)
N	4065	4336	4336
F	39.2217	61.5155	27.1151

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

. The test results show that the regression coefficient of the digital divide in the t − 1 period is significantly negative at the 1% level. This indicates that the digital divide has the same significant inhibitory effect on the energy transition in the next period after removing the reverse causality problem, which is consistent with the benchmark regression results.

5.3.2. Distance Instrumental Variables

Considering that the digital divide is mainly caused by imbalances in the level of information and communication development among cities, referring to He Zongyue et al. [66] and Zhang Xun et al. [26], we use a geographic information system (GIS) to calculate the spherical distance between each city and its provincial capital city as a distance instrumental variable to obtain more reliable conclusions. In terms of relevance, the capital city is usually the economic center of a province and often the center of information and communication development, resulting in a large gap between it and the rest of the cities. Therefore, the spherical distance between a city and its provincial capital city strongly correlates with the digital divide. As far as exogeneity is concerned, while cities with closer proximity to their provincial capital cities may have more developed levels of information and communication, this proximity does not directly affect the city’s energy transition, and therefore, exogeneity is satisfied.

Simultaneously, to circumvent the issue of instrumental variables remaining constant over time in the fixed-effects model, we draw on the methodologies of Nunn and Qian [67], as well as Fang Fuqian et al. [68]. By constructing the interaction term between the spherical distance of each city from its provincial capital and the time-varying national Internet user count, two-stage least squares (2SLS) regressions are conducted. The data on the national Internet user count are sourced from the National Bureau of Statistics (NBS). The results of the instrumental distance variable regression are presented in column (2) of Table 5. It can be found that the digital divide still inhibits energy transition in cities at the one percent level of significance. Specifically, after controlling for other variables, for every one-unit increase in the digital divide, the urban energy transition will change in the opposite direction by 0.1410 units.

5.3.3. Historical Instrumental Variables

Drawing on the study of Huang Qunhui et al. [69], this study adopts the historical postal data of each city as a historical instrumental variable for the digital divide. Specifically, the number of post offices per million people in each city in 1984 was chosen to construct the instrumental variable.

The reason for this choice is that prior to the emergence of Internet technology, postal and telecommunications services were the main means of communication and information transmission. Cities with a high number of historical post offices are likely to have a higher level of digital technology. Moreover, the number of historical post offices has a strong path dependence. Areas with better postal and telecommunications services in the early years tend to remain at the forefront of the subsequent upgrading of communication technologies and the construction of digital infrastructure [70]. At the same time, the number of historical post offices is unlikely to have a direct impact on the city’s present-day energy transition. This satisfies the correlation and exogeneity principles of instrumental variables.

For the empirical application, the interaction term between the number of landline telephones per 100 people in 1984 and the number of Internet users in the country (which is time-dependent) is constructed for the 2SLS regression for each city. Column (3) of Table 5 shows the regression results.

After considering the endogeneity issue, the digital divide still exhibits a significant inhibitory effect on the energy transition. This result does not change from the results of the baseline regression, thus further corroborating the conclusion obtained from the baseline regression; this is that the widening of the digital divide indeed poses an obstacle to the urban energy transition.

5.4. Moderating Effects

This study further discusses the intrinsic mechanisms by which the urban digital divide affects energy transition by introducing a moderating variable into the results of the baseline regression. Column (1) of

Table 6. Results of the moderating effect analysis.

	(1)	(2)	(3)
DD	−0.2190 ***	−0.0697 ***	−0.0123 ***
	(0.0566)	(0.0058)	(0.0037)
DD*INF	0.0857 ***
	(0.0249)
DD*NTI		0.0068 ***
		(0.0008)
DD*NUM			0.0051 **
			(0.0021)
INF	−0.0807 ***
	(0.0246)
NTI		−0.0064 ***
		(0.0011)
NUM			0.0225 ***
			(0.0040)
Individual Effects	yes	yes	yes
Time Effects	yes	yes	yes
Control Variables	yes	yes	yes
_cons	1.0673 ***	0.8442 ***	0.8180 ***
	(0.0801)	(0.0630)	(0.0739)
N	4336	4336	3794
r2	0.9521	0.9529	0.9615
F	46.8896	53.2998	60.6264

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

shows the estimation results related to introducing the moderating variable infrastructure development level (INF). The estimated coefficients of the interaction term between the digital divide and infrastructure development level have the opposite sign of the coefficients of the digital divide variable, which indicates that the level of infrastructure development can effectively attenuate the negative effect of the digital divide on the city’s energy transition, i.e., the level of infrastructure development has an inverse moderating effect on the impact of the digital divide on the city’s energy transition. Hypothesis 2 is verified, which is also consistent with the conclusions of relevant studies [71,72].

Columns (2) and (3) in Table 6 introduce the level of innovation in emerging technologies and the number of people working in digital industries as moderating variables, and the estimation results are reported in columns (2) and (3) of Table 6, respectively. The results show that the regression coefficients of the interaction terms between them and the digital divide are 0.0068 and 0.0051, respectively. This indicates that in cities with a higher level of emerging technology innovation and a larger number of employees in the digital industry, the impact of alleviating the inhibitory effect of the digital divide on urban energy transition is more significant, further verifying Hypothesis 3 and Hypothesis 4 of this study.

5.5. Heterogeneity Test

Considering the diversity of Chinese cities, it remains to be explored whether the unique resource endowments and developmental characteristics of cities affect the effect of the digital divide on energy transition. Consequently, taking into account the disparities in population density, economic activities, economic development levels, regional coordinated development, and resource distribution, this research differentiates samples by dividing them into the eastern, central, and western regions, areas on either side of the Hu Huanyong Line and the Yangtze River Economic Belt, as well as resource-based and non-resource-based cities. It then conducts heterogeneity analyses, and the regression results are presented in

Table 7. Results of the heterogeneity test.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	Southeastern Side	Northwestern Side	Eastern Region	Central Region	Western Region	Yangtze River EB	Non- Yangtze River EB	Resource-based City	Non-resource-based City
DD	−0.0258 ***	−0.0112	−0.0264 ***	−0.0271 ***	−0.0158 ***	−0.0161 ***	−0.0265 ***	0.0101	−0.0219 ***
	(0.0025)	(0.0075)	(0.0038)	(0.0044)	(0.0041)	(0.0039)	(0.0030)	(0.0084)	(0.0025)
Individual Effects	yes	yes	yes	yes	yes	yes	yes	yes	yes
Time Effects	yes	yes	yes	yes	yes	yes	yes	yes	yes
Control Variables	yes	yes	yes	yes	yes	yes	yes	yes	yes
_cons	0.7834 ***	2.1115 ***	1.3017 ***	−0.1727	1.1048 ***	1.3237 ***	0.6892 ***	0.4758 ***	1.3633 ***
	(0.0648)	(0.2675)	(0.1115)	(0.1112)	(0.0994)	(0.1147)	(0.0782)	(0.0940)	(0.0837)
N	3952	384	1536	1536	1264	1648	2688	1760	2576
r2	0.9546	0.9327	0.9673	0.8940	0.9483	0.9590	0.9483	0.8966	0.9669
F	54.2398	13.5478	34.8548	14.1847	24.9044	26.5532	37.6142	7.9965	61.6309

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

.

In urban geography studies, differences in population density in China are usually delineated by the Hu Huanyong Line (HHL). The HHL line, proposed as a demarcation stretching from Heihe in Heilongjiang to Tengchong in Yunnan, highlights the significant spatial variation in China’s population distribution: the southeastern side of the HHL is small in area but densely populated, while the northwestern side of the HHL is large in area but sparsely populated [73]. Columns (1) and (2) report the regression results for the southeastern and northwestern sides of the HHL. The results show that the regression coefficient for the area on the southeast side of the HHL is −0.0258, showing a significant negative effect, whereas the regression coefficient for the area on the northwest side is negative but not significant. The southeast side of the HHL is densely populated, with relatively high energy consumption and an urgent need for energy transition. The existence of the digital divide in this region limits the intelligent and efficient upgrading of the energy system, which in turn affects the effectiveness of the overall energy transition. The population density in the northwestern region is comparatively low, and its economic development lags to a certain degree. The energy industry in this area predominantly relies on resource-based development approaches. Although the digital divide exists, it has not yet had as strong an impact on the energy transition as in the southeastern region due to its more relaxed energy industry situation.

From the perspective of economic development levels, columns (3)–(5) in Table 7 show that the regression coefficients of the eastern, central, and western regions are −0.0264, −0.0271, and −0.0158, respectively. All of these coefficients are notably negative, suggesting that in the eastern, central, and western regions, the digital divide exerts a negative and inhibitory influence on urban energy transition. When it comes to the extent of this influence, there are differences among the three regions. The inhibitory effect is relatively more pronounced in the eastern and central regions, but less so in the western region. This could be attributed to the fact that, despite the eastern region having a higher level of digitization, the implementation of digital technology within the energy sector remains uneven, thereby hindering the progress of the energy transition. In the central region, as a result of the transfer of industries, energy demand is growing rapidly in the process of economic emergence, but the digital divide is limiting the absorption and application of advanced digital technologies, such as digital energy monitoring equipment in the field of industrial energy production and consumption, which affects the popularization of the process, thus slowing down the process of energy transition. In the western region, owing to geographic and economic factors, the development of digital technology is relatively lagging, while the structure of its energy industry is relatively homogeneous. Some areas are still dominated by the development of traditional energy sources; therefore, the negative impact of the digital divide is weaker. With the continued promotion of the Western Development Strategy and the rising demand for energy transformation, the impact of the digital divide may gradually come to the fore.

The Yangtze River Economic Belt serves as a crucial strategic area for the coordinated regional development of China. There is close collaboration among the cities along the belt in aspects such as the economy, transportation, and energy. In contrast, the non-Yangtze Economic Belt regions are relatively weak in terms of regional synergy. This difference may affect the cross-regional application and synergistic effects of digital technologies on the energy transition. According to the regression results in columns (6)–(7) of Table 7, the negative effect of the digital divide on the urban energy evolution transition in the Yangtze River Economic Belt region is relatively small, and the negative effect in the non-YREB region is relatively large. As an important economic corridor in China, the Yangtze River Economic Belt has certain synergistic advantages in terms of the digital economy and energy transition under policy support and regional synergistic development. Despite the digital divide, information exchange and technical cooperation within the region alleviate the impact of the digital divide on energy transition to a certain extent.

Columns (8) and (9) of Table 7 report the estimation results for resource-based and non-resource-based cities. The results show that the coefficients for non-resource cities are still significantly negative, whereas the regression coefficients for resource cities are insignificant, indicating that the inhibiting effect of the digital divide on energy transition exists only in non-resource-based cities. This may be due to the fact that the energy industry in resource-based cities is dominated by resource development and primary processing. These cities are facing the challenges of resource dependence and industrial restructuring. However, as resource-based cities transition to a green and sustainable energy development model, the digital divide may gradually become an important influencing factor.

6. Conclusions, Limitations, and Policy Suggestions

Employing the panel data of 271 Chinese cities from 2006 to 2021, this study analyzes the effect of the digital divide on energy transition and reaches the following primary conclusions: (1) The digital divide has an inhibitory effect on urban energy transition and hinders the transition to green and low-carbon development. Even after conducting the robustness test and taking endogeneity into account, the results remain valid. (2) Upon further analyzing the moderating effect, we discovered that enhancing urban infrastructure construction, driving innovations in emerging technologies, as well as fostering or bringing in digital industry talents can efficiently alleviate the adverse influence of the digital divide on energy transition. (3) There is some regional heterogeneity in the digital divide, inhibiting urban energy transition. Among them, the inhibitory effect of the digital divide on energy transition is more significant in the densely populated areas southeast of the HHL, in eastern and central cities, and in non-resource-based cities.

Based on the above findings, combined with the current situation of digital divide and energy transformation in China, this study puts forward the following three policy recommendations: first, the government needs to pay attention to the inhibiting effect of the digital divide on energy transition. It should develop a digital divide monitoring and assessment system and regularly publish regional digital divide index reports, thereby providing a basis for precise policymaking. At the same time, local governments should adjust the policy direction in a timely manner, continuously monitor the effectiveness of promoting digital technology within the energy sector, and strive to solve the series of problems resulting from the digital divide and its impact on the energy transition, thereby forming a dynamic optimization of the policy closed loop. Secondly, the ‘digital divide’ should be bridged by strengthening urban infrastructure construction, promoting emerging technology innovation, and cultivating or attracting digital industry talents, so as to empower energy transformation. Specifically, to strengthen urban infrastructure construction, the government can provide targeted support to areas with weak digital infrastructure to ensure the accessibility of digital services. For example, by increasing subsidies for the construction of 5G base stations in the central and western regions, local energy enterprises and residents can access high-speed networks at a low cost, overcome the bottleneck of information lag, and create a favorable digital environment for energy transformation. It should promote emerging technological innovation and create a digital energy co-innovation platform to gather the wisdom of all parties. To cultivate or attract talent, the government can build an online and offline network platform for talent exchange and organize regular training to help talent grow. Third, energy resource endowment varies significantly from region to region, and transformation strategies should be formulated based on local realities. For example, Shanxi and other places should continue to promote the clean and efficient use of coal, reduce pollutant emissions, and steadily promote green energy transformation.

Although this study has carried out a more in-depth analysis of the impact of the digital divide on energy transformation and put forward corresponding policy recommendations, there are still some limitations. For example, although this study considers the moderating variables of infrastructure development, technological innovation and human resources, the actual economic and social system is more complex and there may still be potentially important factors that can effectively mitigate the negative impacts of the digital divide on energy transition, which can be explored in depth in future research.