The Impact of Digital Governance on Energy Efficiency: Evidence from E-Government Pilot City in China

Abstract

The digital economy plays a transformative role in enhancing energy efficiency and promoting sustainable development globally. As a key manifestation of digital governance, e-government has emerged as a vital instrument for accelerating the digital transformation of public administration and modernizing governance systems. This study examines the impact of digital governance on urban energy efficiency by analyzing China’s E-Government Pilot City (EPC) policy as a quasi-natural experiment. Using a Difference-in-Differences (DID) approach and balanced panel data from 282 prefecture-level cities (2006–2020), we find that the EPC policy significantly improves total factor energy efficiency (TFEE) by an average of 2.60%. Mechanism analyses reveal that digital governance enhances energy efficiency through industrial structure upgrading, green technology innovation, and foreign direct investment attraction. Heterogeneity analyses indicate that the policy’s benefits are more pronounced in larger, non-resource-based, and non-old industrial base cities, as well as in regions with stronger institutional environments and advanced digital infrastructure. However, spatial spillover effects suggest that while the EPC policy boosts local energy efficiency, it may inadvertently reduce efficiency in neighboring areas due to competitive dynamics and industrial relocation. These findings underscore the importance of tailored and coordinated policy designs to maximize the energy efficiency benefits of digital governance.

1. Introduction

Addressing energy efficiency is crucial in the global fight against climate change, enhancing economic resilience, and promoting sustainability. According to the International Energy Agency (IEA), achieving the net-zero emissions target in the energy sector by 2050—a goal crucial for keeping global temperature rise within 1.5 degrees Celsius as stipulated by the Paris Agreement—requires doubling the annual improvement rate of energy efficiency from 2% in 2022 to more than 4% annually from now until 2030. However, in 2023, the global energy intensity improvement was only 1.3% [1], significantly below the level needed to meet these efficiency goals. This challenge is globally recognized, with China, the world’s largest greenhouse gas emitter, at the forefront. The country’s rapid industrialization, urbanization, and growing energy demands exemplify a worldwide issue [2,3]. These factors underscore the need to enhance energy efficiency to reduce emissions, lessen dependence on imported fossil fuels, and mitigate environmental degradation [4]. Against this backdrop, identifying new policy instruments that can unlock energy efficiency gains—particularly those rooted in the ongoing wave of digital transformation—has become a question of both theoretical importance and pressing policy relevance.

Within energy efficiency research, the focus has traditionally been on identifying factors influencing energy efficiency outcomes, such as technological innovation [5], industrial structure optimization [6], and global market integration [7]. The pivotal role of digital economic development in enhancing energy efficiency has gained recognition [8]. Digital technologies, including big data analytics, the internet of things (IoT), artificial intelligence (AI), and blockchain, have become key enablers in the quest for improved energy efficiency [9]. Their incorporation into energy systems enables real-time monitoring, predictive maintenance, and optimized energy distribution and consumption, which is crucial for minimizing waste and enhancing energy utilization across sectors [10]. Yet, the bulk of this literature conceptualizes digitalization primarily at the level of firms and markets—focusing on “digital economy” dynamics such as smart production, digital finance, or platform-based trade—while paying relatively little attention to how the digital transformation of the state itself, through digital governance reforms, may reshape energy use and efficiency at the urban level.

The digital transformation era, driven by rapid progress in internet and information technologies, has ushered in a transformative phase of digital governance. Initiatives like the UK’s “Digital Government Strategy” in 2012 [11], Singapore’s “e-Government Masterplan 2011–2015” [12], and Denmark’s “Digital Strategy 2016–2020” [13] underscore a global trend towards digitizing government services, underlining the role of technology in improving public administration and citizen participation. This evolution, characterized by integrating advanced digital technologies into government operations, promises enhanced efficiency, transparency, and public engagement [14]. The economic ramifications of this digital shift have been thoroughly explored, highlighting its potential to transform the business landscape, augment investment efficiency, diminish corruption, and foster trust in public institutions, thereby invigorating urban economic vitality [15]. Nevertheless, despite the rapid diffusion of digital governance worldwide and its growing prominence in policy agendas, systematic evidence on whether and how digital governance itself contributes to energy efficiency—beyond its broader economic and governance effects—remains scarce. This disconnect between the prominence of digital government reforms in practice and their limited treatment in the energy efficiency literature constitutes a first core research gap that this paper addresses.

However, the academic consensus on the impact of the digital economy on urban energy efficiency remains elusive [10,16,17]. Research suggests that the digital economy could promote energy efficiency by improving resource allocation and information flows [18,19]. Digital enterprises, through the adoption of smart grid technologies, energy optimization analytics, and remote monitoring and control systems, play a pivotal role in driving sustainable development [16,20]. Nonetheless, concerns persist that the rapid expansion of the digital economy might reduce energy efficiency due to increased energy demands and potential adverse effects on resource allocation. In particular, despite their significant contributions to data processing and storage, data centers and cloud computing infrastructure consume high energy levels [10]. Further analysis indicates that while the digital economy may initially decrease energy efficiency, its maturation could improve it [4]. Moreover, while developed countries have significantly enhanced their energy efficiency through digital technologies, many developing nations need help to achieve comparable advancements [21]. Critically, the exploration of digital governance’s influence on energy efficiency is still nascent. While the contributions of public services to environmental protection are well-documented, showing both positive and potentially negative impacts on environmental quality—dependent on factors such as government size and fiscal levels—the nuanced effects of digital governance on energy efficiency demand further investigation [22].

In particular, existing studies rarely (i) distinguish clearly between market-oriented digitalization and state-led digital governance, (ii) identify the causal effect of digital governance on energy efficiency using credible quasi-experimental designs, or (iii) uncover the concrete mechanisms—such as industrial upgrading, green innovation, or international capital flows—through which digital governance may alter urban energy-use patterns. These limitations leave important conceptual and empirical questions unresolved and motivate the present study. This knowledge gap presents a substantial opportunity, as understanding the role of digital governance in promoting energy efficiency can provide strategic insights into utilizing technology for sustainable economic and environmental frameworks. This study aims to fill this gap by examining how digital governance acts as a catalyst for energy efficiency in China, offering insights with global relevance for discussions on sustainable development and digital transformation in governance.

A key example of this dynamic is the E-government Pilot City (EPC) policy introduced by the Chinese government in 2014, designed to integrate information technology into governmental functions [23]. This policy offers a valuable lens for assessing digital governance’s effects on energy efficiency through a Difference-in-Differences (DID) analysis. By treating the staggered rollout of the EPC policy across cities as a quasi-natural experiment, this paper is able to causally identify the impact of digital governance on urban energy efficiency rather than merely documenting correlations. This research confirms that the EPC policy significantly enhances energy efficiency in China, with an average increase of 2.60%. Our findings, upholding the assumption of parallel trends, demonstrate consistency across various robustness checks, including placebo tests and alternative specifications. The positive outcomes of the EPC policy are analyzed through three primary mechanisms: facilitating industrial structural upgrades (ISU), promoting green technology innovation (GTI), and attracting foreign direct investment (FDI). Heterogeneity analysis reveals that larger, non-resource-based, and non-old industrial base cities derive greater benefits from the EPC policy, attributed to their advanced infrastructure, diversified industrial structures, and adaptability to green innovations.

Moreover, the analysis highlights the critical roles of the local institutional environment and digital infrastructure in mediating digital governance’s impact on energy efficiency, with supportive institutional frameworks and strong digital foundations enhancing the EPC policy’s effectiveness. Even with this, our analysis also uncovers a potential drawback, as the EPC policy might inadvertently reduce energy efficiency in neighboring regions through competitive dynamics and the relocation of energy-intensive industries, emphasizing the importance of a balanced and comprehensive approach to digital governance and energy efficiency initiatives. Taken together, these results speak directly to the broader policy debate on how developing countries like China can leverage digital government reforms to reconcile economic development with their “dual carbon” and sustainable development goals, while managing possible negative spatial spillovers.

This study makes three main contributions to the existing literature by examining how the EPC policy promotes energy efficiency, thereby clarifying its importance and novelty. First, it introduces digital governance—captured through a nationwide e-government reform—as a novel and critical determinant of urban energy efficiency, thus extending the literature that previously focused predominantly on the digital economy, technological innovation, and market-based drivers [5,8,18,19,20]. By providing causal evidence that the EPC policy significantly improves energy efficiency, our findings help reconcile the mixed views on the digitalization–energy nexus and demonstrate that state-led digital transformation can be an effective instrument for energy conservation in developing countries.

Second, this paper advances the digital governance literature by shifting the focus from conventional outcomes such as administrative efficiency, public service quality, and economic growth [14,15] to a key environmental and resource-based outcome: energy efficiency. Through a DID framework, mechanism tests on industrial structural upgrades, green technology innovation, and foreign direct investment, and heterogeneity analyses across different types of cities and institutional environments, we uncover how and under what conditions digital governance can act as a potent catalyst for energy efficiency within governmental operations.

Third, our study broadens the international conversation by situating China’s EPC policy in the global context of digital government reforms and energy transitions. By documenting both beneficial effects and adverse spillovers on neighboring regions via the relocation of energy-intensive industries, the paper highlights the need for coordinated regional governance and offers actionable insights for policymakers worldwide who seek to design digital governance reforms that support sustainable energy management while avoiding unintended distributional and spatial consequences [21].

The remainder of this paper is organized as follows. Section 2 delineates the background and formulates the theoretical hypotheses underpinning the study. Section 3 discusses the data collection process and the methodological framework employed in the analysis. Section 4 presents the baseline regression results, conducts robustness tests, explores the mechanisms at play, and examines potential heterogeneity within the data. Section 5 is dedicated to additional studies that extend the primary analysis. Finally, Section 6 summarizes the findings, discusses their implications, and offers conclusions.

2. Policy Background and Research Hypothesis

2.1. The EPC Policy

In a landmark effort to leverage these technologies for governance modernization and to forge a path towards a digital China, the National Development and Reform Commission (NDRC), in collaboration with 11 other departments, issued the “Notice on Accelerating the Implementation of Information Benefiting the People Project” in January 2014 [24]. This initiative was bolstered in June of the same year by issuing the “Notice on Approving Shenzhen and 79 Other Cities as National Pilot Cities for Information Benefiting the People”, colloquially known as the EPC policy [25]. This policy earmarked Shenzhen, among 80 cities, as a pilot site for the EPC initiative. These cities were charged with addressing critical challenges in public services within the traditional governance framework, as illustrated in Figure 1, which maps their spatial distribution.

The EPC policy showcases the Digital Governance model, offering marked improvements over traditional governance systems. This model integrates isolated governmental systems into a unified information-sharing platform through “Internet Plus Government Services,” enabling efficient data sharing and dismantling the traditional “information silos.” Services like “One-Net Handling” exemplify this new integration, promoting inter-departmental communication. Meanwhile, the quality of government services has significantly improved. The digitalization and intelligence of government operations have been enhanced, facilitating the provision of services that overcome geographical and temporal barriers. The “data traveling more than the people” initiative minimizes public service wait times, thus boosting service quality. Furthermore, EPC policy makes governance become more diversified and inclusive. Digital platforms have made government operations more transparent and accessible, directly communicating policies at the grassroots level and removing informational barriers. Enhanced online communication methods, such as public email services and official social media accounts, have broadened public participation in governance, empowering grassroots communities. This transformation improves efficiency and fosters a more engaged and informed citizenry.

Figure 2 elucidates a discernible upward trajectory in the informatization of governmental services across China during the decade spanning 2010 to 2020 [26], which can be attributed, in part, to the enactment of the EPC policy in 2014. Before the policy’s implementation, China’s informatization index stood at 0.545, marginally surpassing the global median of 0.471. The subsequent period witnessed a marked enhancement in this metric; by 2016, the index had escalated to 0.607, surpassing the global average of 0.492. This ascending pattern persisted, with the index reaching 0.681 in 2018—significantly outstripping the global mean of 0.549. The culmination of this trend was observed in 2020 when China’s index soared to 0.795, a stark contrast to its 2010 benchmark and substantially above the global average of 0.599. This progression was accompanied by a notable ascent in China’s global standing, improving by 33 positions from 78th in 2012 to 45th in 2020. Such advancements underscore China’s dedication to enhancing its e-government capabilities and illuminate the EPC policy’s palpable influence. In summation, the EPC policy emerges as a pivotal catalyst in the digital transformation journey of government governance, aimed at contemporizing the national governance apparatus and its competencies. Hence, the EPC policy not only acts as a crucial catalyst for enhancing the digital governance quotient but also provides a quasi-experimental environment for the empirical analysis of the impacts of digital governance [23].

2.2. Research Hypothesis

2.2.1. Industrial Structure Upgrades

Public services—including infrastructure, education, and healthcare—play a crucial role in upgrading the industrial structure. Guided by government strategy, their expansion drives major changes in the industrial framework. Notably, the government tends to privatize basic public services, which opens up new markets for service-based companies and draws more social investment into these areas, thereby advancing industrial restructuring [27]. The advent of digital governance serves to amplify this dynamic. With the government as a principal consumer, a pronounced demand for sophisticated, quality-enhanced digital solutions emerges. This demand not only elevates the market visibility of digital products but also compels providers to elevate their offerings in quality [28]. This demand and supply modification cycle fosters technological innovation, propelling industrial evolution and elevation, much like the U.S. semiconductor industry’s growth spurred by governmental procurement strategies [29].

Moreover, the consistent and clear procurement signals from the government act as a catalyst, drawing investment towards sectors prioritized by governmental spending, thereby charting a definitive course for their future growth. This phenomenon, coupled with the concept of “increasing returns,” positions these sectors as attractive hubs for labor, further driving industrial enhancement and optimization [30]. Digital governance transcends mere procurement; it extends to refining governmental functions, setting higher regulatory benchmarks, and fostering a more orderly market environment conducive to industrial reconfiguration [31].

The nexus between industrial structural optimization and energy efficiency is profound [6,32]. Viewing through the prism of resource allocation and the “structural dividend,” it is evident that as industries morph, the resultant alteration in their scale and composition—each characterized by distinct energy requirements and conservation potentials—directly influences energy consumption paradigms. The migration from the energy-intensive secondary sectors to the more benign tertiary sectors inherently curtails energy demand, optimizing resource deployment and enhancing energy efficiency [32]. Consequently, we propose the following hypothesis:

Hypothesis 1.

Digital governance fosters energy efficiency by facilitating upgrades in the industrial structure.

2.2.2. Green Technology Innovation

Digital governance is a pivotal enabler for green technology innovation by streamlining administrative frameworks and dismantling bureaucratic barriers, fostering an agile environment conducive to sustainable advancements [33]. This simplification of processes not only eases the operational load on businesses but also propels them toward embracing and developing sustainable technologies. The establishment of unified government service platforms promotes resource sharing and transparency, bolstering business confidence and incentivizing investments in green innovation projects that, despite their high risk, promise substantial rewards [34]. Additionally, enhancing public services through digital governance, including education, healthcare, and cultural programs, is instrumental in attracting a highly skilled workforce essential for pioneering innovative solutions [15].

Mirroring Porter’s observations on the nexus between innovation and sustainable development, green technologies significantly curtail energy consumption throughout production and utilization, elevating energy efficiency [35]. Green innovations, transcending sectoral boundaries, advocate for not only cleaner production techniques but also the conservation of energy and the embracement of sustainable consumption practices [36]. Moreover, advancements in green technologies are instrumental in modernizing energy infrastructure, reducing energy dissipation during storage and transmission, and diminishing the energy expenditure per production unit by incorporating automated and intelligent systems. Consequently, the digital overhaul of governmental governance acts as a catalyst for promoting green technology innovations, directly contributing to an enhancement in energy efficiency [37]. In light of these considerations, we propose the following hypothesis:

Hypothesis 2.

Digital governance enhances energy efficiency by fostering the innovation of green technologies.

2.2.3. Foreign Direct Investment

The digital transformation of government governance emerges as a crucial catalyst in enhancing energy efficiency by attracting FDI. Grounded in new public service theory principles, digitalization efforts aim to narrow the divide between domestic capabilities and global economic resources, fostering an environment conducive to FDI [38]. Governments streamline the investment process by establishing comprehensive online platforms, enabling foreign investors to navigate project exploration, agreement finalization, and negotiations efficiently [39]. This digital approach mitigates the informational gap and alleviates the bureaucratic burden on investors, thus rendering the investment journey more fluid and effective. The strategic deployment of technology and data analytics on these platforms optimizes the alignment of supply with demand, offering tailored policy support to foreign investors and consequently elevating the attractiveness of the business ecosystem [40]. Such enhancements diminish operational costs and investment risks, positioning the host nation as a more favorable destination for FDI [41].

The influx of FDI, particularly in domains like manufacturing and Research and Development (R&D), introduces cutting-edge technologies, managerial acumen, and sophisticated production methodologies [42]. This transference engenders a demonstration effect, prompting local businesses to embrace more efficient, eco-friendly production technologies, thereby advancing energy efficiency within the host nation [43]. Furthermore, the industrial linkage effect, which integrates local firms into the global value chains of foreign corporations, necessitates technological enhancements and augments the quality of intermediate goods and services [44]. This interplay not only elevates the competitive edge of local enterprises but also amplifies the overall energy productivity as companies adopt more efficient technologies to meet the standards of their international partners [45]. Considering these dynamics, we posit the following hypothesis:

Hypothesis 3.

Digital governance enhances energy efficiency by attracting foreign direct investment.

3. Materials and Methods

3.1. Data Sources

This study employs a balanced panel dataset spanning 282 prefecture-level cities in China from 2006 to 2020 to scrutinize the EPC policy’s efficacy in augmenting energy efficiency. The choice of the endpoint in 2020 was dictated primarily by the availability of comprehensive and verified datasets necessary to assess the EPC policy’s efficacy in augmenting energy efficiency. Most recent data beyond 2020 have yet to undergo the rigorous validation processes required for academic research, particularly in the context of energy and economic indicators. Moreover, the year 2020 marks a pivotal boundary as China embarked on expansive digital governance efforts in 2021, culminating in the release of the “Guiding Opinions of the State Council on Strengthening the Building of a Digital Government” in 2022 [46], which aimed to innovate and advance digital governance across the nation. By isolating the impact of the EPC policy prior to these substantial national developments, this temporal delineation ensures the reliability of our findings, minimizing the influence of confounding variables associated with the subsequent widespread policy implementation. The dataset amalgamates energy consumption figures from the China Energy Statistical Yearbook (National Bureau of Statistics of China) with economic indicators such as GDP, adjusted to 2006 price levels to mitigate inflationary distortions. Innovatively, green patent data were curated based on the International Patent Classification Green Inventory [47], facilitating a nuanced analysis of technological innovation at the municipal level. Complementary datasets were extracted from the Express Professional Superior (EPS) database [48], alongside the China Statistical Yearbook and the China City Statistical Yearbook, ensuring a multidimensional understanding of the urban economic landscape. The dataset was meticulously refined to exclude cities affected by administrative restructuring or significant data gaps, culminating in a robust sample of 4230 observations.

3.2. Variable Selection

In this analysis, energy efficiency is posited as the pivotal dependent variable, congruent with the methodological frameworks established by Zhou et al. [49] and Hong et al. [50]. This approach integrates the principal inputs of labor, capital, and energy against the backdrop of GDP as the output of interest while accounting for undesirable outputs such as SO₂, smoke, and effluents. Capital input is quantified through the total fixed asset investment, adjusted to 2006 values via a 9.6% depreciation rate, employing the perpetual inventory method and the fixed asset price index for calibration. Labor input is gauged by the year-end employee count in urban sectors, whereas energy consumption is inferred through the innovative use of nighttime light intensity as a surrogate measure. Given that our proxy for energy input, nighttime light data, may contain noise in cities dominated by the service sector or in those that have completed light emitting diode renovation, we assess its validity by examining its correlation with electricity consumption statistics and find a significant positive relationship between the two at the one percent level. To ascertain the total factor energy efficiency across various prefecture-level cities, we utilize the SBM–Malmquist–Luenberger index method. An uptick in the index value is emblematic of an augmentation in energy efficiency, underscoring the method’s sensitivity to improvements in the efficient utilization of resources.

To present the variations and trends in energy efficiency across different regions and over time more effectively, Figure 3 utilizes a topographical representation. In this figure, cities excluded from the sample are marked in white. In contrast, the color intensity of other cities denotes energy efficiency, with dark blue indicating the highest efficiency levels. Due to space constraints, data from only four selected years—2006, 2010, 2015, and 2020—are displayed. A noticeable increase in dark blue regions from 2006 to 2020 suggests a general upward trend in China’s energy efficiency. Moreover, a spatial analysis reveals that the eastern coastal regions exhibit significantly higher energy efficiency than the central and western regions. This disparity is likely attributable to the more developed industrial base and technologically advanced production facilities in the eastern coastal areas. Additionally, these regions were among the first in China to implement digital governance and informatization, which have facilitated more effective resource utilization and reduced energy consumption. In contrast, the central and western regions demonstrate lower energy efficiency due to their relatively lower economic development and lesser degrees of industrialization and informatization.

At the heart of our analysis lies the EPC policy, serving as an emblem of the digital transformation within government governance [23]. This transformation is epitomized by the inception of EPCs, signifying a leap toward the digitization of governmental operations. Consequently, the EPC policy is utilized as a benchmark for this digital evolution, with cities recognized as E-government pilot cities from a specified year onwards being attributed a value of 1 and all others a value of 0. To rigorously assess the causal impact of digital governance on energy efficiency, our study leverages the implementation of the EPC policy as a quasi-natural experiment. Utilizing the DID methodology, this research probes how digital governance affects energy efficiency.

Before evaluating the EPC policy’s direct outcomes on energy efficiency, it is imperative to establish whether the policy substantively augments the digital governance capabilities of regional governments. To this end, our analysis employs two pivotal indicators: website performance and digital governance focus. In our study, government websites are identified as pivotal digital platforms that facilitate services to the public and businesses, thereby serving as effective indicators of local governments’ digital governance capabilities. To gauge the efficacy of digital governance across regional governments, we employ the “Government Website Performance Evaluation Report” issued by the China Software Testing Center (CSTC) [51]. Since 2014, only the top 100 cities based on website performance scores have been publicly disclosed. To maintain sample integrity without losing valuable data, we created a binary variable for website performance. Cities ranked in the top 100 are assigned a value of 1, while those not in the top 100 receive a value of 0.

Furthermore, we evaluate the level of digital governance by analyzing the frequency of related keywords in government work reports from 2006 to 2020. The keyword dictionary includes big data, blockchain, artificial intelligence, communication technology, the Internet of Things, cloud computing, smart city, smart village, and e-government platform. The proportion of these keywords to the total word count is used to construct the variable representing the focus on digital governance. The regression analyses, presented in Appendix A,

Table A1. The EPC Policy’s impact on website performance, digital governance focus, and base-station counts.

	(1)	(2)	(3)
Variables	Website Performance	Digital Governance Focus	Base-Station Counts
EPC	0.381 ***	0.000 ***	0.013 ***
	(0.029)	(0.000)	(0.001)
City FE	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
Constant	−1.891 *	−0.001 **	−0.497 ***
	(0.986)	(0.001)	(0.096)
R2	0.455	0.4220	0.525
N	4230	4023	4230

Note: The significance levels of 1%, 5%, and 10% are denoted by ***, **, and *, respectively. Robust standard errors are reported in parentheses.

, columns (1)–(3), show that the EPC policy has a statistically significant and positive effect on improving website performance, strengthening the prioritization of digital governance within local administrations, and increasing the number of telecommunication base stations. This empirical evidence provides a solid foundation for the subsequent analysis of how the EPC policy contributes to improvements in energy efficiency.

To ensure a thorough and precise evaluation of digital governance’s influence on energy efficiency, our model integrates a suite of control variables, as informed by extant literature [20,50,52,53]. These variables encompass GDP, Fiscal, Openness, Human capital, Finance, Industrialization, and Population density. Including these controls is strategic and aimed at neutralizing any extraneous effects these variables might exert on energy efficiency outcomes. The methodology and detailed computations for these variables are systematically outlined in

Table 1. Variable definitions and descriptive statistics.

Types	Variables	Definitions	Mean	S.D.
Dependent variable	TFEE	Total factor energy efficiency measured by the SBM–Malmquist–Luenberger index method	0.313	0.122
Core independent variable	EPC	=1 if the prefecture-level city implemented the EPC policy, 0 otherwise	0.120	0.119
Control variables	GDP	Gross domestic product/Total population, logarithm	10.452	0.723
	Fiscal	Budget revenues/Budget expenditures	0.467	0.228
	Openness	Total exports and imports of goods, logarithm	13.985	2.134
	Human capital	Number of university students enrolled per 10,000 persons	0.054	0.014
	Finance	Balance of loans from financial institutions, logarithm	16.089	1.283
	Industrialization	Number of industrial enterprises above designated size, logarithm	6.554	1.118
	Population density	Total population/Size of administrative area	5.735	0.919
Mechanism variables	Structure	Value added of the tertiary sector/Value added of the primary and secondary sector	0.723	0.391
	GTI	Green technology patents granted per 10,000 persons	0.121	0.321
	FDI	Total foreign direct investment	0.506	1.063

, providing a comprehensive framework for our analysis.

3.3. Empirical Strategy

3.3.1. Baseline Model: DID Approach

The launch of the first batch of EPCs in 2014 provides a discrete policy shock that can be treated as a quasi-natural experiment. Cities selected into the 2014 EPCs constitute the treatment group, while non-selected cities in the same period form the control group. Because we observe both treated and control cities before and after the 2014 policy shock, a DID design is more appropriate than either a simple before–after comparison for treated cities or a single cross-sectional regression in a specific year.

We choose the DID procedure for three main reasons. First, it allows us to difference out time-invariant city characteristics, such as historical industrial structure, geographic endowments, and long-standing institutional factors, that may simultaneously influence both the probability of being selected as a pilot city and the level of energy efficiency. Second, by incorporating year fixed effects, the DID framework controls for common macro shocks, including nationwide energy-saving campaigns, business cycle fluctuations, and other contemporaneous policy changes that affect all cities in a given year. Third, given that the EPC policy is introduced only once in 2014 for a subset of cities, the standard two-way fixed effects DID model offers a transparent and tractable way to identify the average treatment effect of digital governance on urban energy efficiency under a clear and testable parallel-trend assumption. Together, these features make DID a theoretically grounded and empirically robust strategy for our setting. The following regression model is constructed:

$$TFE{E}_{i,t}={\alpha }_0+{\alpha }_{1}EP{C}_{i,t}+{\alpha }_n{\displaystyle \sum X_{i,t}+{\mu }_i+{\delta }_t+{\epsilon }_{i,t}}$$

(1)

In Equation (1), TFEE_i,t is the dependent variable indicating the energy efficiency of city i in year t. EPC_i,t represents the EPC policy, which is set to 1 for cities designated as E-government pilot cities in and after the year of designation and 0 otherwise. X_i,t encompasses a range of control variables such as GDP, Fiscal, and other relevant factors that might influence energy efficiency. The model includes μ_i and δ_t to account for city-fixed effects and year-fixed effects, along with a random error term ε_i,t. Our study primarily focuses on the sign and significance of the coefficient α₁.

3.3.2. Mediating Effect Model

To assess the potential mechanisms through which the EPC policy affects energy efficiency, we construct a set of mechanism variables. In applied economic research, mediation-effect (or causal mediation) models are a common tool for mechanism analysis. However, in our context these models face two important challenges. First, mediating variables such as technological innovation, industrial upgrading or factor allocation are typically jointly determined with energy efficiency and are likely to be correlated with unobserved city-level shocks. This correlation makes the mediator endogenous to the error term, leading to biased estimates of both the mediation effect and the direct policy effect, which undermines the reliability of standard mediation tests [54]. Second, conventional mediation frameworks often require strong functional-form and exclusion restrictions (for example within a structural equation model) that are difficult to justify in a setting with multiple channels and policy spillovers.

In light of these concerns, and to remain consistent with our difference-in-differences identification strategy in the baseline analysis, we adopt a reduced-form “intermediate outcome” approach: we directly treat each mechanism variable as the dependent variable in a DID-type regression. This procedure allows us to test whether the EPC policy significantly changes the proposed mechanism variables, under the same identifying assumptions as the baseline model. So the mechanism testing model is depicted in Equation (2):

$$Mechanis{m}_{i,t}={\beta }_0+{\beta }_{1}EP{C}_{i,t}+{\beta }_n{\displaystyle \sum X_{i,t}+{\mu }_i}+{\delta }_t+{\epsilon }_{i,t}$$

(2)

In Equation (2), Mechanism_i,t represents the mechanism variable for city i in year t, and the remaining symbols retain their meanings consistent with the baseline regression. β₁ is the coefficient of the core explanatory variable in our mechanism analysis.

4. Results

4.1. Baseline Regression

Table 2. Baseline regression results: E-government pilot city policy and total factor energy efficiency.

	(1)	(2)	(3)	(4)	(5)
Variables	TFEE	TFEE	TFEE	TFEE	TFEE
EPC	0.026 ***	0.029 ***	0.027 ***	0.026 ***	0.026 ***
	(0.006)	(0.006)	(0.006)	(0.005)	(0.005)
GDP		0.030 ***	0.032 ***	0.047 ***	0.047 ***
		(0.008)	(0.008)	(0.012)	(0.012)
Fiscal		0.023	0.022	0.029 *	0.030 *
		(0.016)	(0.016)	(0.016)	(0.016)
Openness			0.004	0.005	0.005
			(0.004)	(0.004)	(0.004)
Human capital			0.673 ***	0.643 ***	0.630 ***
			(0.156)	(0.155)	(0.159)
Finance				−0.001	−0.001
				(0.009)	(0.009)
Industrialization				−0.025 ***	−0.026 ***
				(0.006)	(0.006)
Population density					0.036
					(0.049)
City FE	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes
Constant	0.309 ***	−0.015	−0.118	−0.117	−0.315
	(0.001)	(0.083)	(0.096)	(0.153)	(0.293)
R2	0.689	0.692	0.693	0.695	0.695
N	4230	4230	4230	4230	4230

Note: The significance levels of 1% and 10% are denoted by *** and *, respectively. Robust standard errors are reported in parentheses.

delineates the regression outcomes, delineating the EPC policy’s efficacy in enhancing energy efficiency. Column (1) initially isolates the analysis to year and regional fixed effects, aiming to curtail the socio-economic developmental factors’ sway at the municipal tier on the empirical insights. Progressively, Columns (2) through (5) augment this foundational model by sequentially incorporating a suite of control variables, culminating in Column (5), which amalgamates all control variables alongside the stipulated fixed effects. The uniformly positive and statistically significant coefficients associated with the EPC policy, spanning Columns (1) through (5), corroborate the hypothesis that the EPC policy framework catalyzes energy efficiency advancements within pilot cities.

Particularly, the analysis in Column (5) reveals a discernible increment in energy efficiency within EPCs by an average of 2.60% relative to their Non-EPCs, illustrating the tangible impact of digital governance transformation on energy efficiency. In other words, after controlling for confounding factors, the introduction of digital governance generates a non-trivial, economically meaningful improvement in urban energy efficiency, suggesting that digital transformation within the public sector can be as powerful as more traditional, market-based policy instruments in shaping energy-use outcomes.

Theoretically, the improvement in energy efficiency under the EPC policy aligns with the principles of information asymmetry reduction in economics, which posits that better access to information can lead to more efficient resource allocation and usage [55]. By reducing information asymmetries, digital governance allows for more informed decision-making and efficient operations, contributing to increased energy efficiency. Additionally, this theory is complemented by the institutional theory [56], which suggests that institutions, like the frameworks established under the EPC policy, play a critical role in reducing transaction costs and promoting more efficient use of resources through improved regulations and governance structures. Our findings thus provide empirical support for the view that digital governance can function as an “institutionalized information infrastructure” that simultaneously relaxes information constraints and strengthens formal rules, thereby improving energy efficiency even though the policy itself is not explicitly framed as an energy or environmental regulation.

This study’s insights resonate with preceding research [2,3,8], which collectively highlight the digital economy’s propitious role in fostering energy efficiency. Analogous policy evaluations, such as Wen et al. [52] exploration of the Broadband China Strategy, have reported a 2.61% surge in green total-factor energy efficiency. Concurrently, references [50] and [57] have documented significant upticks in urban and total-factor energy efficiency, attributed to the Low-Carbon City policy and carbon emissions trading pilot policy, respectively. Viewed against this backdrop, the 2.60% improvement associated with the EPC policy is broadly comparable to the effect of Broadband China and somewhat smaller than the gains observed for explicitly environmental policies such as Low-Carbon City and carbon trading. This pattern is intuitive: while Low-Carbon City and carbon trading policies directly target emissions and energy use, EPC is a digital governance reform whose energy-efficiency effects are indirect, operating through improved information flows, regulatory quality, and administrative efficiency. The similarity in magnitude to Broadband China also suggests that digital infrastructure and digital governance are complementary levers: the former provides the “pipes” of digital connectivity, whereas the latter reshapes how governments and firms actually utilize those digital capabilities in day-to-day decision-making.

While the effect sizes in our investigation may not eclipse those in other studies, they unequivocally signal the EPC policy’s substantial contribution to energy efficiency, particularly when viewed against the backdrop of broader digital governance initiatives. The comparative analysis of these effect magnitudes with those from alternative policy assessments accentuates the criticality of embedding digital strategies within policy frameworks to materialize sustainable energy paradigms. Relative to earlier work that largely focused on the “digital economy–energy efficiency” nexus, our results show that embedding digital technologies in government processes can deliver energy-efficiency improvements that are on par with major digital infrastructure initiatives and constitute a meaningful complement to targeted environmental regulations. In essence, our research furnishes empirical substantiation that digital transformation, epitomized by the EPC policy, is instrumental in augmenting energy efficiency, thereby underscoring the strategic import of digital integration in governmental policy-making for sustainable development. By situating the EPC policy alongside Broadband China, Low-Carbon City, and carbon trading experiences, this study offers a more nuanced understanding of how different types of digital and environmental interventions interact in practice, highlighting digital governance as a distinct yet underexplored pillar in the broader policy portfolio for improving energy efficiency.

4.2. Robustness Tests

We conduct robustness checks using alternative specifications, including parallel trend and placebo tests.

4.2.1. Parallel Trend Tests

Ensuring the validity of causal claims in DID analysis hinges critically on the treatment and control groups displaying parallel trends or no marked systematic discrepancies before the policy’s enactment. This necessitates that, in the context of our study on energy efficiency, the cities under examination should manifest stable trends before the EPC policy’s deployment, thereby fulfilling the parallel trends prerequisite. To rigorously assess this foundational premise, we leveraged event study techniques to execute parallel trend assessments and scrutinize the policy’s dynamic ramifications.

Central to our analysis was the designation of 2014—the pilot year of the EPC policy—as the reference point. We conducted distinct DID regressions for the policy variable spanning eight years preceding and six years following this pivotal year. The outcomes, as depicted in Figure 4, substantiate the parallel trends assumption: before the pilot cities’ selection, the coefficients of EPC were statistically negligible, with the regression coefficients oscillating around the null value. This pattern underscores a fundamental similarity in energy efficiency trajectories between pilot and non-pilot cities before the policy’s introduction.

Furthermore, the analysis of dynamic effects post-policy implementation unveils that the digital transformation spurred by the EPC policy has markedly bolstered energy efficiency. Notably, the policy’s influence did not manifest instantaneously but exhibited a temporal delay, with the coefficient becoming significantly positive in the third year post-implementation and displaying an upward trajectory in subsequent years. This gradual intensification of the policy’s impact underscores a cumulative effect, hinting at an inherent time lag between the policy’s announcement, its operational execution, and the maturation of government information and technology projects.

4.2.2. Applications of Synthetic DID

In our examination, while the application of parallel trend tests served to affirm the validity of our results, we acknowledge the intrinsic limitation associated with this method: the existence of parallel trends before the implementation of a policy does not guarantee their perpetuation following the intervention. To address this methodological challenge, we employed the innovative Synthetic Difference-in-Differences (SDID) methodology as articulated by Arkhangelsky et al., (2021) [58]. The SDID estimator represents a methodological advancement by amalgamating the traditional DID approach with the Synthetic Control method, thereby enhancing the robustness and accuracy of the estimations.

The SDID methodology is distinguished by its dual-weighting system, incorporating both individual and temporal weights. This system preferentially selects, from the control group, those units that most closely resemble the treatment group in terms of pre-treatment characteristics and aligns temporal periods to reflect post-intervention conditions. Such a weighting scheme ensures that the weighted average outcome of the control group parallels the mean outcome trajectory of the treated group for each period preceding the intervention. Furthermore, it guarantees that for each control unit, the pre-intervention weighted average outcome consistently differs by a constant from the outcome in the post-intervention period. The integration of individual fixed effects within the SDID framework further refines the analysis, accommodating individual-specific temporal variations.

Utilizing the SDID estimator to evaluate the impact of the EPC policy on energy efficiency, we observed a significant average treatment effect of 0.030, which is statistically significant at the 1% level. This finding not only reaffirms the positive influence of the EPC policy on enhancing energy efficiency but also underscores its pivotal role in advancing the broader objective of sustainable urban development. The adoption of the SDID approach in our analysis not only corroborates our initial observations but also elevates the methodological sophistication of our study, providing a more detailed and dependable evaluation of the EPC policy’s contribution to energy efficiency improvements.

4.2.3. Placebo Test

To further insulate our analysis from the potential confounding effects of unobserved variables on the selection of EPCs, we implemented a placebo test. This methodological safeguard bolsters the credibility of our primary conclusions, attributing observed enhancements in energy efficiency unequivocally to the EPC policy. The placebo test involved conducting 1000 iterations of random sampling across all 282 prefecture-level cities. In each iteration, 80 cities were arbitrarily designated as the dummy treatment group, with the remaining 202 serving as the control group. Subsequent regression analyses, adhering to the specifications of our baseline model, were performed for each randomized grouping.

Theoretically, should the observed improvements in energy efficiency be solely attributable to the EPC policy, the explanatory variable pertaining to these artificially constructed treatment groups would not exert a significant influence on the regression outcomes. Figure 5 succinctly encapsulates the results of this exercise, where the bulk of the estimated coefficients gravitate towards zero, and the majority of p-values surpass the 0.1 threshold. The authentic estimated coefficient of 0.026, demarcated by a dashed line in the figure, emerges as an outlier within this distribution. This anomaly within the distribution underscores the improbability of our empirical findings being artifacts of random variation. Consequently, this placebo test substantiates the inference that the EPC policy exerts a definitive, positive impact on energy efficiency, untainted by the latent effects of other unaccounted for variables.

4.2.4. Additional Robustness Checks

In our quest to fortify the cogency of the key findings presented in

Table 3. Robustness checks with alternative models and specifications.

Line	Test Methods	Coefficients	Standard Error	N
	Panel A: Controlling disturbing policies
(1)	Energy Saving and Emission Reduction Fiscal Policy	0.026 ***	0.005	4230
(2)	Air Pollutant Emission Standards Policy	0.021 ***	0.005	4230
(3)	Low Carbon City Policy	0.025 ***	0.005	4230
(4)	Carbon Emissions Trading Pilot Policy	0.026 ***	0.005	4230
(5)	All policies	0.021 ***	0.005	4230
	Panel B: PSM-DID model
(6)	Kernel matching	0.036 ***	0.007	2918
(7)	Radius matching	0.033 ***	0.007	2897
(8)	Nearest neighbor matching	0.035 ***	0.007	2913
	Panel C: Adjustment of sample range and core variables
(9)	Adding province-year fixed effects	0.031 ***	0.006	4155
(10)	Replacement core variables	−0.044 ***	0.010	4230
(11)	Adding predicate variables	0.026 ***	0.006	4230
(12)	Excluding the municipality sample	0.023 ***	0.006	4170
	Panel D: Other tests
(13)	Excluding samples in 2020	0.020 ***	0.005	3948
(14)	Considering lag effects	0.015 ***	0.005	3948
(15)	Considering the expectation effect	−0.012	0.010	4230
(16)	Considering two-way clustering (city, year) and a wild-cluster bootstrap check.	0.026 ***	0.012	4230

Note: The dependent variable is TFEE. The significance levels of 1% is denoted by ***. Robust standard errors are reported. In line (15), the core explanatory variable is Pre 1.

, we embarked on a comprehensive array of robustness checks. These checks are instrumental in validating the integrity of our primary regression outcomes, thereby solidifying the foundation of our conclusions.

Firstly, acknowledging the shift towards green, low-carbon growth amidst escalating environmental challenges, we scrutinized the distinct impact of the EPC policy amidst a landscape populated with concurrent energy-oriented policies. To delineate the EPC policy’s unique contribution from other national energy initiatives like energy conservation and structural optimization strategies, we meticulously identified four pivotal policies detailed in national directives during our study’s timeline, namely the Energy Saving and Emission Reduction Fiscal Policy (2011), Air Pollutant Emission Standards Policy (2013), Low Carbon City Policy (2010), and Carbon Emissions Trading Pilot Policy (2007). These were operationalized as binary variables in our model, post-implementation in pilot cities. Adjusting for these concurrent policy interventions (Table 3, Panel A), we noted a slight reduction in the EPC policy’s coefficient magnitude, yet it preserved its statistically significant positive influence. This observation implies that, despite potential initial over-estimations, the EPC policy’s role in augmenting energy efficiency is both significant and distinct.

Secondly, to address potential selection bias stemming from variances in economic characteristics across cities, we employed the Propensity Score Matching (PSM) technique, specifically the PSM-DID approach, to evaluate the EPC policy’s environmental ramifications. We applied three matching algorithms—nearest-neighbor, kernel, and caliper matching—to ensure the robustness of our sample selection. The logistic regression model, predicated on city-level control variables, facilitated the matching process. The post-matching balance tests revealed no significant disparities between the treatment and control groups, evidenced by negligible t-statistics for covariates and a standardized bias reduction below 10%. The DID estimations, derived from these matched cohorts (Table 3, Panel B), consistently showed a significantly positive coefficient for the variable EPC, corroborating the initial regression outcomes and affirming the soundness of our primary insights.

Thirdly, although our baseline model incorporates city and time fixed effects, we acknowledge that provincial-level, time-varying factors may also influence energy efficiency. To address this potential source of bias, we include province-year interaction fixed effects in the model. In addition, we perform a robustness test by redefining the dependent variable as the ratio of total coal consumption to GDP (tons per yuan), providing an alternative measure of energy efficiency. Since the selection of EPC cities may have been affected by characteristics such as representativeness and resource endowment, selection bias could exist. To mitigate this concern, we include city-level selection criteria and their interactions with time in the regression analysis. These criteria encompass topographical characteristics, river density, designation as a key environmental protection city, and the average levels of fiscal dependence and digital infrastructure prior to EPC implementation. Furthermore, to ensure that the results are not driven by the unique economic structures of China’s four major municipalities—Beijing, Shanghai, Tianjin, and Chongqing—we exclude them from the sample and re-estimate the model (Table 3, Panel C). The findings continue to indicate a statistically significant positive effect of the EPC policy, thereby reinforcing the robustness and credibility of our empirical conclusions.

Fourthly, recognizing the potential influence of the COVID-19 pandemic in 2020—which may have led to emission reductions due to containment measures and thus biased energy-related data—we exclude the 2020 samples from the regression analysis. Moreover, considering the possible lagged effects of the EPC policy on energy efficiency, we employ the lagged energy efficiency as the new dependent variable to capture delayed policy impacts. In addition, local governments may have anticipated the implementation of the EPC policy and undertaken preparatory actions to meet application requirements, which could generate an anticipation effect and bias the estimation results. To control for this possibility, we introduce a dummy variable Pre 1 for the year preceding the actual implementation of the EPC policy. Finally, for statistical inference, we apply two-way clustering at the city and year levels and employ the wild cluster bootstrap method to account for serial correlation and spatial dependence. The estimation results, reported in Table 3, Panel D, show no significant deviation from the baseline findings, confirming the robustness and reliability of our estimates.

4.3. Mechanism Tests

4.3.1. Facilitating Industrial Structure Upgrades

In the theoretical framework posited earlier, integrating digital governance mechanisms is identified as a critical driver for augmenting energy efficiency, predominantly by catalyzing industrial transformation. Our study adopts the ratio of value-added in the tertiary sector to that in the primary and secondary sectors as an analytical lens to examine the industrial landscape. The empirical evidence presented in

Table 4. Mechanism test: Industrial structure upgrades, green technology innovation and foreign direct investment.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	Structure	GTI	FDI	L.Structure	L.GTI	L.FDI
EPC	0.048 ***	0.180 ***	0.257 ***	0.046 ***	0.162 ***	0.028 ***
	(0.010)	(0.014)	(0.044)	(0.009)	(0.013)	(0.043)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.909	0.730	0.857	0.913	0.743	0.865
N	4230	4230	4230	3948	3948	3948

Note: The significance levels of 1% is denoted by ***. Robust standard errors are reported in parentheses.

, specifically in Column (1), emphatically highlights the effectiveness of the EPC policy in fostering industrial restructuring within the pilot cities. The coefficient associated with the variable EPC is significantly positive at the 1% level, illustrating the instrumental role of the EPC policy in precipitating a shift towards a more service-oriented and, by extension, more energy-efficient economic configuration. This shift is critical as the tertiary sector typically exhibits higher energy efficiency due to less energy-intensive activities than manufacturing and agriculture.

Moreover, considering potential lag effects, we include a one-period lag of the mechanism variable, as reported in Column (4), which provides further empirical evidence that the EPC policy fosters industrial restructuring. This restructuring process often entails the adoption of advanced digital tools that enhance resource and operational management, thereby minimizing waste and reducing energy consumption. In addition, industrial upgrading not only reshapes the economic structure but also accelerates the integration of innovative technologies and practices that substantially decrease energy input per unit of output in the services sector. This transformation is supported by the deployment of technologies such as smart grids and the Internet of Things, which optimize energy utilization and improve operational efficiency. These technological advancements not only enhance resource allocation but also encourage the transition from traditional, inefficient energy systems to more sustainable alternatives, effectively mitigating the environmental footprint of industrial activities. Consequently, these findings provide strong validation for Hypothesis 1 of our study.

4.3.2. Promoting Green Technology Innovation

To further elucidate the impact of digitized governance on energy efficiency, our study introduces an innovative metric: the incidence of green technology patents per 10,000 individuals. This serves as a proxy for green technology innovation, enabling an examination of how the digital transformation of government services can catalyze advancements in environmentally sustainable technologies, thereby enhancing energy efficiency. The empirical findings, detailed in Table 4, Column (2), provide compelling evidence of the positive impact of EPC policy on fostering green technology innovation. Specifically, when we replace the simple patent count with a quality-adjusted measure, the proportion of green patents among total patents, the results remain consistent. The EPC policy continues to exhibit a significant and positive impact on green technology innovation, reinforcing the robustness of our findings. Introducing such innovations is crucial, as they often lead to developing and adopting more energy-efficient processes and sustainable production techniques. The data reveals a significant positive correlation between the variable EPC and cities’ green technology innovation level.

Moreover, the analysis incorporates a one-period lag of the mechanism variable, as shown in Column (5). The significantly positive coefficient confirms the sustained effectiveness of the EPC policy in driving green technology innovation. These advancements extend beyond mere reduction in energy consumption; they also involve integrating renewable energy systems, such as solar and wind power, which decrease reliance on non-renewable resources. Enhanced waste management protocols include adopting more sophisticated recycling processes and utilizing waste-to-energy technologies that transform municipal waste into electricity and heat. Meanwhile, smarter urban planning incorporating eco-friendly infrastructure, like green buildings and energy-efficient public transport systems, significantly reduces the urban carbon footprint. This multifaceted approach supports immediate reductions in energy use and sets a foundation for long-term sustainability, aligning with broader environmental goals [36,37]. In summation, the findings of this research compellingly validate hypothesis 2 of our paper.

4.3.3. Attracting Foreign Direct Investment

Our empirical investigation, particularly the insights derived from Column (3) of Table 4, delves into the impact of the EPC policy on the inflows of FDI. Unfortunately, given data constraints at the prefecture-city level over our study period, we are unable to systematically separate FDI into greenfield versus merger-and-acquisition components or to construct a consistent sectoral breakdown for all cities. The statistical evidence delineates a significant positive correlation between the EPC policy and FDI volumes, with the relationship being statistically significant at the 1% level. This finding underscores a strong nexus between integrating digital governance frameworks and enhancing a city’s attractiveness to international investors. More specifically, implementing the EPC policy is correlated with a notable increase of 25.72% in foreign investment inflows into cities, highlighting the pivotal role of digital governance in attracting global capital. Increased FDI brings in additional capital and often introduces advanced, more energy-efficient technologies and practices. Further empirical analysis, incorporating a one-period lag of the mechanism variable in Column (6), reinforces this conclusion by confirming the positive and persistent effect of the EPC policy on foreign investment and energy efficiency.

Importantly, the influx of FDI often brings advanced technologies and management practices that significantly enhance energy efficiency. This includes adopting cutting-edge manufacturing equipment, green building technologies, and renewable energy systems that align with international environmental standards. These technologies facilitate more efficient use of energy resources and promote sustainability by reducing emissions and lowering the carbon footprint of urban environments [42,43]. Moreover, FDI typically stimulates competitive improvements in local industries, encouraging them to adopt more energy-efficient technologies and processes to remain competitive. This enhances operational efficiencies and drives broader economic reforms that prioritize sustainability. Therefore, our analysis lends empirical support to hypothesis 3.

4.4. Heterogeneity Analysis

4.4.1. City Characteristics

Firstly, significant differences exist between large and small-to-medium-sized cities in terms of infrastructure and other facilities, leading to potentially distinct outcomes from the implementation of EPC policy. For this analysis, cities are classified as “large” or “small” based on the median number of permanent residents, a straightforward, empirical criterion reflecting the urban scale and associated administrative capacities [20]. Regression analyses differentiate between these two categories. Results from columns (1) and (2) of

Table 5. Heterogeneity effects: City characteristics.

	Population Size		Resource-Based City		Old Industrial Bases
	Small	Big	No	Yes	No	Yes
	(1)	(2)	(3)	(4)	(5)	(6)
Variables	TFEE	TFEE	TFEE	TFEE	TFEE	TFEE
EPC	−0.003 (0.009)	0.041 *** (0.007)	0.026 *** (0.007)	−0.004 (0.008)	0.040 *** (0.007)	−0.017 ** (0.008)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.685	0.724	0.688	0.692	0.663	0.776
N	2141	2086	2565	1665	2805	1425

Note: The significance levels of 1% and 5% are denoted by *** and **, respectively. Robust standard errors are reported in parentheses.

reveal that EPC policy significantly improves energy utilization efficiency in large cities, but the effect is insignificant in smaller cities. The disparity may be attributed to the greater mobility of factors in large cities, which have stronger capabilities in absorbing capital and labor and more intensive and efficient business operations and pollution control equipment compared to smaller cities. This facilitates the agglomeration effects of public services and knowledge spillovers, thereby effectively controlling pollution emissions from industrial expansion and, more noticeably, improving environmental quality.

Secondly, there is a strong correlation between a city’s resource endowment and its industrial and energy structures. Compared to non-resource-based cities, resource-based cities often have a heavier industrial structure, higher energy consumption per unit of GDP, and a weaker foundation for green development. Classification into resource-based and non-resource-based cities follows the guidelines of the “National Plan for Sustainable Development of Resource-based Cities (2013–2020)” issued by the State Council, offering a government-validated framework for analyzing the impact of resource endowments on urban economic and environmental strategies [59,60]. From this perspective, we examine the heterogeneous effects of EPC policy on energy efficiency. Observing the estimation results in columns (3) and (4) of Table 5, the EPC policy’s coefficient for non-resource-based cities is 0.026, which is significant at the 1% level. However, the impact on resource-based cities is not significant. The extensive economic development model centered around resource extraction and processing in resource-based cities has created a dependency, with a high-energy-consuming and polluting industrial structure facing a “low-end lock-in” dilemma, making the effects of EPC policy in promoting the transition from old to new energies quite limited.

Lastly, we identify cities with different functional orientations based on the “National Old Industrial Base Adjustment and Renovation Plan (2013–2022)” by the National Development and Reform Commission, which provides a governmental framework to categorize cities as “old industrial base” and “non-old industrial base” [61,62]. This classification is instrumental in understanding the variable impacts of the EPC policy on cities with distinct industrial legacies and development trajectories. Observing the estimation results in columns (1) and (2), it is evident that implementing the EPC policy significantly positively impacts the energy efficiency of non-old industrial base cities but not old industrial base cities. Old industrial base cities typically have intensive energy use and an extensive development model, which limits the potential for short-term energy efficiency improvements. In contrast, non-old industrial base cities, with higher marketization levels and environmental quality demands, respond more positively when EPC policies are effectively enforced. This creates a strong “warning effect” that compels enterprises to innovate in green technology, thereby leading to industrial structure optimization and enhanced energy utilization efficiency.

4.4.2. Institutional Environment

The intricate interplay between institutional frameworks and digital government governance enhances energy efficiency. Our paper delves into the nuanced impacts of formal and informal institutional dimensions on the synergy between digital governance and energy efficiency. For the analysis of formal institutions, we employ two key metrics: the Marketization level and Intellectual Property Protection (IPP) level [63,64]. Marketization level is quantified using the Fan Gang Marketization Index, applicable to prefecture-level cities [65]. To assess the level of intellectual property protection in prefecture-level cities, we use a ratio that compares the city’s I.P. trial settlements to its GDP, normalized against the national ratio of I.P. trial settlements to national GDP. These metrics are crucial as they reflect the regulatory and policy landscape’s capacity to foster innovation and efficient market operations. In the realm of informal institutions, the intensity of clan culture is scrutinized, as delineated by [66]. Clan culture is represented by the percentage of people with the top three surnames. The prevalence of these surnames is interpreted as an indicator of clan influence, which can significantly impact the local governance dynamics and the acceptance and effectiveness of new policies, including those related to digital governance and energy efficiency. As presented in

Table 6. Heterogeneity effects: Institutional environment.

	Marketization Level		IPP Level		Clan Culture
	Low	High	Low	High	Low	High
	(1)	(2)	(3)	(4)	(5)	(6)
Variables	TFEE	TFEE	TFEE	TFEE	TFEE	TFEE
EPC	0.011 (0.007)	0.047 *** (0.010)	−0.003 (0.010)	0.041 *** (0.009)	0.013 (0.009)	0.036 *** (0.006)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.714	0.714	0.760	0.732	0.698	0.700
N	2031	2157	2105	2109	2115	2115

Note: The significance levels of 1% is denoted by ***. Robust standard errors are reported in parentheses.

, empirical evidence reveals a distinct pattern through median-based group regressions: the core variable underpinning EPC policy exhibits significant positivity exclusively within cohorts characterized by higher marketization levels, robust intellectual property regimes, and pronounced clan cultures. This pattern underscores the premise that the efficacy of digital governance in augmenting energy efficiency is inherently contingent upon the prevailing institutional milieu. On the formal institutional front, an environment conducive to market operations and intellectual property rights fortifies the foundation for innovation, particularly in the green technology sector. Moreover, the robustness of clan cultures can synergize with formal institutional mechanisms, offering an ancillary stratum of social impetus for adherence to energy efficiency mandates.

4.4.3. Infrastructure Development

Our paper posits that the caliber of local infrastructure significantly influences the efficacy of EPC policy in enhancing energy efficiency. To substantiate this hypothesis, this study meticulously selected three pivotal variables as proxies for local infrastructure levels: Infrastructure investment, digital infrastructure, and internet coverage [67]. Infrastructure investment refers to the total investment in fixed assets per capita. Digital infrastructure denotes the total postal and telecommunication operations per capita. Internet coverage is measured by the number of internet users per 10,000 people. These proxies are specifically chosen because they encapsulate different aspects of infrastructure critical to the deployment and efficacy of digital governance. Employing a median split regression analysis, this investigation endeavors to delineate the differential impacts of EPC policy across diverse infrastructural landscapes. The empirical findings, encapsulated in

Table 7. Heterogeneity effects: Infrastructure development.

	Infrastructure Investment		Digital Infrastructure		Internet Coverage
	Less	More	General	Good	General	Good
	(1)	(2)	(3)	(4)	(5)	(6)
Variables	TFEE	TFEE	TFEE	TFEE	TFEE	TFEE
EPC	−0.006 (0.010)	0.038 *** (0.009)	−0.025 *** (0.007)	0.039 *** (0.008)	−0.015 * (0.008)	0.046 *** (0.010)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.762	0.730	0.712	0.709	0.781	0.722
N	2109	2107	2101	2094	2104	2114

Note: The significance levels of 1% and 10% are denoted by *** and *, respectively. Robust standard errors are reported in parentheses.

, unveil a pronounced dichotomy in the influence of EPC policy contingent upon the infrastructure’s robustness. In cities endowed with sophisticated infrastructure, the variable EPC emerges as significantly positive, underscoring the symbiotic relationship between advanced digital infrastructure and the successful implementation of digital governance measures. Conversely, in locales characterized by suboptimal digital infrastructure, the coefficient of EPC transitions to a significant negative, revealing an adverse effect on energy efficiency. At the forefront is the misallocation of resources in cities with insufficient infrastructure. This leads to a situation where investments in digital innovations are nullified by the existing infrastructure’s inability to support advanced technological frameworks. This mismatch increases energy demand without corresponding efficiency improvements, as emerging digital solutions are forced to function alongside outdated systems, thereby elevating energy consumption without achieving optimization. The deficiencies in digital infrastructure further exacerbate the challenges related to cybersecurity and data management, diminishing the potential benefits of digital governance in enhancing energy efficiency.

5. Further Studies: Analysis of Spatial Spillover Effects

The advent of digital governance, coupled with the pervasive influence of digital technology, suggests that policies like the EPC policy might extend their impact beyond local boundaries, potentially affecting energy efficiency in neighboring areas. Acknowledging these complex interactions, assessing the EPC policy’s broader impacts is crucial, considering their potential spillover effects on the energy efficiency of neighboring cities [68,69].

In determining the most suitable spatial econometric model, we conducted a series of tests—including L.R., LM, Wald, and Hausman tests—which identified the Spatial Durbin Model (SDM) with two-way fixed effects as the most appropriate for our study. To ensure the robustness of our findings, we employed four distinct spatial matrices in our analysis: the Adjacency matrix, Inverse distance matrix, Economic matrix, and Economic-geographical matrix, thereby providing a thorough examination of the EPC policy’s impact on energy efficiency from multiple spatial perspectives.

Table 8. Spatial Durbin Model (SDM) estimation of the EPC policy’s direct and indirect effects.

	Adjacency Matrix	Inverse Distance Matrix	Economic Matrix	Economic-Geographical Matrix
	(1)	(2)	(3)	(4)
Variables	TFEE	TFEE	TFEE	TFEE
EPC	0.026 *** (0.005)	0.024 *** (0.005)	0.022 *** (0.005)	0.020 *** (0.005)
W × EPC	−0.033 *** (0.010)	−0.180 * (0.093)	−0.031 ** (0.013)	−0.037 *** (0.012)
sigma2_e	0.004 *** (0.000)	0.004 *** (0.000)	0.005 *** (0.000)	0.004 *** (0.000)
Direct effect	0.026 *** (0.005)	0.024 *** (0.005)	0.022 *** (0.005)	0.020 *** (0.005)
Indirect effect	−0.032 *** (0.010)	−0.217 * (0.125)	−0.030 ** (0.013)	−0.037 *** (0.013)
R2	0.011	0.215	0.187	0.187
N	4230	4230	4230	4230

Note: The significance levels of 1%, 5%, and 10% are denoted by ***, **, and *, respectively. Robust standard errors are reported in parentheses.

reports the spatial effects of EPC policy on energy efficiency. Based on SDM regression with four types of spatial weight matrices, the results show that the spatial interaction terms (W × EPC) coefficients are significantly negative, indicating exogenous digital governance interaction effects exist in the sample regions. Further decomposition reveals that while EPC policy promotes energy efficiency in local regions, they have a negative spatial spillover effect, potentially hindering improvements in surrounding areas.

This negative spillover may be understood through competitive regional dynamics and resource redistribution. The digital infrastructure and governance improvements in pilot cities can attract a disproportionate share of resources, including human capital and investment, leading to a concentration effect. As detailed in Appendix B,

Table A2. Pathways of the EPC policy: Labor mobility and industrial emission effects.

	(1)	(2)	(3)	(4)	(5)
Variables	University Student	Technical Staff	Industrial Wastewater	Industrial SO2	Industrial NOx
EPC	0.026 *** (0.006)	0.003 *** (0.001)	−0.120 *** (0.032)	−1.758 *** (0.237)	−2.839 ** (1.131)
City FE	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes
Constant	0.096 (0.164)	0.044 (0.027)	2.384 *** (0.913)	−19.289 ** (8.484)	56.809 (37.457)
R2	0.905	0.592	0.805	0.771	0.358
N	4230	4035	4230	4230	4230

Note: The significance levels of 1% and 5% are denoted by *** and **, respectively. Robust standard errors are reported in parentheses. University student is represented by the ratio per ten thousand people. Technical staff refers to the proportion of information and software industry employees relative to the total workforce. Industrial wastewater, Industrial SO₂, and Industrial NO_x are each represented by their respective emission quantities.

of our study, implementing EPC policy significantly draws university graduates and high-technology professionals to these regions. This centralization can exacerbate regional disparities in digital capabilities and energy efficiency measures as surrounding regions struggle to compete with the more digitally advanced EPCs [70]. The theory of agglomeration economies suggests that as resources cluster in technologically advanced areas, peripheral areas may suffer from reduced investments and a drain in skilled labor, which can negatively impact their energy efficiency and economic resilience [71].

The negative spatial spillover effect could also stem from the relocation of energy-intensive industries to adjacent areas with less stringent environmental regulations or digital governance standards. As EPC policy implements more rigorous energy efficiency and pollution control measures, industries unable or unwilling to comply may migrate to neighboring regions, thereby increasing the energy demand and environmental burden in those areas [72]. Further econometric results presented in Appendix B, Table A2 underscore that EPC policy implementation contributes to significant reductions in industrial wastewater, sulfur dioxide, and nitrogen oxide emissions, facilitating the transition of energy-intensive industries. This phenomenon aligns with the pollution haven hypothesis, which posits that strict environmental regulations can migrate “dirty” industries to regions with laxer standards, thus redistributing rather than reducing environmental impacts [73].

6. Conclusions and Implications

6.1. Research Conclusions

Utilizing a balanced panel dataset from 282 prefecture-level cities in China spanning 2006 to 2020, this study empirically investigates the impact of digital governance on energy efficiency by leveraging the quasi-natural experiment of the EPC policy. Our DID model estimates reveal that the implementation of the EPC policy significantly enhances urban energy efficiency by an average of 2.60%. This core finding remains robust after a series of tests, including parallel trend assessments, placebo tests, and alternative model specifications. Mechanism analyses confirm that the EPC policy bolsters energy efficiency primarily through three channels: facilitating industrial structural upgrading, fostering green technology innovation, and attracting higher levels of foreign direct investment. In comparison with existing policy evaluations, such as those on Broadband China, Low-Carbon City, and carbon emissions trading [50,57], our estimated effect size is of a similar order of magnitude to digital infrastructure policies and somewhat smaller than targeted environmental regulations, suggesting that digital governance operates as an indirect yet powerful driver of energy efficiency rather than a direct command-and-control tool.

Heterogeneity analysis further reveals that the benefits of the EPC policy are more pronounced in larger cities, non-resource-based cities, and those without an old industrial base, attributable to their superior infrastructure, diversified economic structures, and greater adaptability. Moreover, the policy’s effectiveness is contingent on a supportive institutional environment (evidenced by higher marketization and robust intellectual property protection) and advanced digital infrastructure. Conversely, our spatial econometric analysis uncovers a potential drawback: while the EPC policy boosts local energy efficiency, it generates negative spatial spillover effects, inadvertently hampering energy efficiency in neighboring regions, likely due to competitive dynamics and the relocation of energy-intensive industries. These findings both corroborate and extend earlier studies on digitalization and environmental policy: consistent with prior evidence that digital development and regulatory policies can enhance energy efficiency [2,3,8], our results show that digital governance reforms embedded in public administration can deliver comparable gains, while also revealing a previously underexplored spatial trade-off in the form of adverse spillovers to neighboring jurisdictions. Taken together, the evidence yields a fresh insight: digital governance is not merely a vehicle for administrative modernization but also a structurally important, albeit indirectly targeted, instrument for promoting energy-efficient urban development in a way that interacts with local capacity, institutional quality, and inter-city competition.

6.2. Practical Implications

Our findings offer several actionable insights for policymakers and governments aiming to harness digital governance for sustainable development.

First, governments should strategically promote digital governance through continued prioritization and investment in the digital transformation of public administration. The proven efficacy of policies like the EPC in enhancing energy efficiency confirms the value of integrating digital tools into governance to achieve environmental and energy conservation goals. Compared with earlier policy experiences—such as Broadband China, Low-Carbon City, and carbon trading—our results imply that digital governance reforms can deliver energy-efficiency improvements of a comparable magnitude while simultaneously generating broader governance and service-delivery benefits, reinforcing the case for treating digital government as a core component of green development strategies rather than a purely administrative modernization agenda.

From a policy-prioritization perspective, this “co-benefit” nature suggests that, per unit of energy saved, digital governance reforms are likely to be at least as cost-effective as some dedicated energy-efficiency instruments that rely heavily on direct fiscal subsidies or command-and-control regulations, even though a full marginal abatement cost comparison is beyond the scope of this study and remains an important avenue for future research.

Second, given the heterogeneous effects of the EPC policy, governments ought to adopt a differentiated, place-based approach to policy design. In larger and more developed cities, policy can focus on leveraging agglomeration economies for green innovation. In contrast, resource-based or old industrial cities may require targeted support to first address foundational challenges, such as modernizing infrastructure and facilitating industrial transition, to create the necessary preconditions for digital governance to take effect.

Third, institutional and infrastructural foundations should be strengthened. The effectiveness of digital governance is highly dependent on the local context. Thus, policymakers should concurrently strengthen formal institutions (e.g., by deepening market reforms and enforcing intellectual property rights) and invest heavily in building robust digital infrastructure, including high-speed internet and data centers. This dual focus ensures that the digital governance framework operates on a solid and enabling foundation.

At the same time, the energy footprint of digitalization itself cannot be ignored: data centers, cloud-computing facilities, and IoT backhaul networks consume substantial amounts of electricity and, if supplied by carbon-intensive grids, may offset part of the gains from improved energy efficiency on the demand side. In addition, by lowering information and transaction costs, digital governance may induce behavioral and structural rebound effects—for instance, increased use of digital public services and associated infrastructure—that partially dilute the net energy and emissions savings. Policymakers should therefore align the expansion of digital governance with parallel efforts to decarbonize the power sector and to monitor potential rebound, so that the long-run balance between the benefits and the energy costs of digitalization remains favorable.

Fourth, to mitigate negative spatial spillover effects, it is essential to establish regional and national coordination mechanisms. Policymakers should design regional collaborative governance frameworks that discourage a “race to the bottom” and prevent the mere shifting of pollution. This could involve establishing regional energy efficiency targets, creating cross-jurisdictional environmental compensation schemes, and promoting the sharing of green technologies and best practices among neighboring regions. In light of our evidence that EPC-type reforms improve local energy efficiency but may reduce efficiency in surrounding cities, such coordinated approaches are particularly important to ensure that digital governance contributes to net regional and national gains rather than redistributing energy-intensive activities across space.

6.3. Theoretical Implications

This study makes several key contributions to the academic literature. First, it extends the digital governance discourse beyond its conventional focus on administrative efficiency and public service delivery by providing robust empirical evidence that digital governance is a critical and effective catalyst for environmental governance, specifically energy efficiency. This bridges the literature on digital government with that on sustainable development. Compared with prior studies that largely treat digitalization as a market-driven phenomenon within the “digital economy,” our analysis shows that state-led digital governance reforms can generate energy-efficiency benefits that are empirically comparable to both digital infrastructure programs and some environmental regulations, thereby filling an important conceptual and empirical gap. In doing so, it also points to the need for theoretical frameworks that jointly consider the emissions-reduction potential of digital governance and the energy costs of the underlying digital infrastructure, as well as possible rebound effects, rather than treating digitalization as an unambiguously energy-saving force.

Second, this study enhances the understanding of underlying influence mechanisms. By identifying and empirically validating three specific transmission channels—industrial structure, green innovation, and FDI—this research offers a more granular understanding of how digital governance influences energy efficiency, moving from a “black box” to a clearer causal pathway. Whereas much of the existing literature infers mechanisms indirectly from correlations between digitalization and aggregate outcomes, our mechanism testing provides direct evidence that digital governance reshapes the industrial mix, stimulates green innovation, and alters international capital flows in ways that jointly explain the observed efficiency gains, offering a more integrated explanation than earlier single-channel studies.

Third, the research highlights the contextual nature of digital governance outcomes. The heterogeneity findings underscore the importance of contextual factors, aligning with institutional theory. They demonstrate that the effectiveness of digital governance is not uniform but is significantly moderated by local city characteristics, institutional quality, and infrastructural readiness. This adds nuance to existing work on both digitalization and environmental policy, which often assumes fairly homogeneous treatment effects, by showing that digital governance reforms are most effective where institutional and infrastructural conditions are favorable, and less so in structurally constrained environments.

Fourth, this study introduces a spatial perspective into digital governance research. The discovery of negative spatial spillover effects introduces a critical spatial dimension to the study of digital governance policies. It suggests that the impact of such policies is not confined to jurisdictional boundaries and calls for theoretical models that can account for inter-regional dynamics and cross-border externalities. In contrast to prior studies that largely focus on within-city or within-region outcomes, our spatial analysis reveals that digital governance can generate both positive local and negative neighboring effects, pointing to a more complex geography of digital reform impacts and opening up a new research frontier at the intersection of digital governance, spatial economics, and environmental policy.

6.4. Limitations and Future Research Directions

Despite its contributions, this study has several limitations that pave the way for future research. These limitations mainly relate to issues of external validity, potential biases inherent in administrative data, and the broader macroeconomic and crisis context in which digital governance reforms unfold.

First, the generalizability of the findings is constrained by the study’s exclusive focus on China. Future research should examine whether similar results hold in other national contexts with differing political systems, governance models, and levels of digital development. Comparative studies involving countries such as those in the EU or Singapore would be particularly valuable. In addition, cross-country comparative designs could be used to assess whether the energy-efficiency effects of digital governance are robust across different regulatory traditions, levels of market development, and exposure to recent global shocks, thereby clarifying the extent to which our findings reflect China-specific conditions versus more general patterns.

Second, the reliance on administrative data may introduce specific forms of measurement bias, such as under-reporting or misclassification of energy use, inconsistencies in data collection across jurisdictions, or changes in reporting standards over time. While our use of fixed effects and robustness checks helps to mitigate some of these concerns, future research should strive to triangulate administrative records with independent data sources—such as enterprise surveys, smart-meter data, or satellite-based proxies—to assess the direction and magnitude of potential biases and to validate the stability of the estimated policy effects.

Third, the temporal scope of our analysis, which ends in 2020, implies that the full effects of recent crises—most notably the COVID-19 pandemic and subsequent financial and energy market disturbances—are only partially captured, if at all. These crises may have affected both energy efficiency (e.g., through industrial slowdowns, changes in sectoral energy demand, or increased residential consumption) and the implementation trajectory of EPC reforms (e.g., delays, shifts in priorities, or accelerated digitalization in specific domains). Although our specification includes time fixed effects that absorb common shocks, we cannot fully rule out that crisis-induced structural breaks or heterogeneous crisis responses across cities may influence the estimated impacts. Future research could explicitly model these crises as distinct episodes, compare pre-crisis and post-crisis policy effects, or use extended samples to assess whether the relationship between digital governance and energy efficiency is stable across turbulent and “normal” periods. Cross-country comparative work that examines how different states’ digital governance systems responded to the pandemic and other crises would be especially informative.