Energy Implications of Urban Shrinkage in China: Pathways of Population Dilution, Industrial Restructuring, and Consumption Inertia

Abstract

The structural responsiveness of urban energy systems has emerged as a central challenge in the governance of shrinking cities. Urban shrinkage entails more than a redistribution of resources—it reflects deep tensions embedded in population spatial configuration, functional redundancy, and institutional inertia. To investigate the evolutionary trajectory and driving mechanisms of urban energy consumption (UEC) under the context of urban shrinkage, this study focuses on China, a country undergoing rapid internal regional transformation. Drawing on panel data from 281 cities between 2008 and 2021, the study integrates two-way fixed effects (TWFE) models, mediation analysis, and spatial econometric models to ensure the scientific rigor and robustness of the quantitative analysis. Contrary to the intuitive assumption that declining population leads to reduced energy loads, the results reveal a non-linear and asymmetric trajectory wherein per capita energy consumption increases alongside continued demographic decline. Mechanism decomposition further shows that declines in population density and the share of secondary industry suppress UEC through spatial dispersal and the retreat of energy-intensive sectors, respectively. In contrast, fiscal contraction and institutional path dependency collectively elevate the share of traditional energy consumption, reinforcing the structural inertia of UEC. This study illuminates the non-linear dynamics of energy system evolution under urban shrinkage and argues for a shift away from linear and target-driven governance paradigms toward governance frameworks that emphasize structural adaptation, distributive equity, and systemic resilience—thereby offering a theoretical and empirical foundation for multi-objective sustainable urban transitions.

1. Introduction

Cities are emblematic of modern civilization and serve as key operational units in addressing climate change, advancing low-carbon transitions, and supporting sustainable economic development. As energy lies at the heart of carbon neutrality efforts, urban areas—where economic activity concentrates—account for more than 60% of global energy consumption [1]. Under the combined effects of extended infrastructure lifespans, rigid energy service expectations, and capital lock-in, urban energy consumption (UEC) continues to rise, and short-term reductions through efficiency improvements remain difficult to achieve [2]. This underscores the urgency of translating cities’ inherent policy flexibility into tangible energy-reduction strategies in support of the Sustainable Development Goals (SDGs).

In China, urban areas account for roughly four-fifths of national energy consumption—significantly higher than the global average of 78% [3]—highlighting the structural inertia that hinders energy-saving transitions at the city level. This inertia is closely linked to the persistent demand from energy-intensive sectors, and can be largely attributed to land expansion [4], population agglomeration [5], and industrial cost structures [6]. Given the slow penetration of renewable energy in China’s energy mix, reducing overall energy consumption has become a more immediate priority than energy source substitution. In pursuit of locally adaptive SDG implementation, China has launched a series of policy instruments to reconcile resource-intensive development models with energy-saving imperatives. Since the 11th Five-Year Plan, China has made energy reduction a binding target across three subsequent national plans. The current 15th Five-Year Plan places even greater emphasis on cleaner and more efficient use of fossil energy on the demand side. Despite these efforts, a definitive turning point in UEC trends has yet to be observed [7].

Amid the accelerating differentiation of the global urban system and intensified population redistribution, urban shrinkage has increasingly evolved into a form of structural lock-in within regional development trajectories. As active participants in microeconomic processes, populations are deeply embedded in the functioning of cities and directly determine urban scale. Compared with economic or ecological indicators, population dynamics are regarded as a more typical and immediate manifestation of urban shrinkage [8]. The role of urban shrinkage in shaping sustainable development outcomes has emerged as a new anchor for causal identification, facilitating empirical evaluations of its implications for socioeconomic equity [9], land-use patterns [10], and greenhouse gas emission efficiency [11]. According to the smart shrinkage perspective, for cities experiencing contraction amid regional competition, it is neither scientifically sound nor practically feasible to adopt policy strategies derived from urban resilience paradigms developed during financial crises or pandemic disruptions, with the aim of restoring growth [12]. This reasoning is particularly relevant in the Chinese context, where shrinkage often coincides with severe negative externalities. China is now in the latter half of its rapid urbanization process, with both population and economic activity becoming increasingly concentrated in major cities and metropolitan clusters. In this dynamic equilibrium, some cities have become shrinking cities due to sustained net population outflows—making China a natural testing ground for extending the analytical frontier of urban shrinkage research [13].

Existing analytical frameworks for UEC are predominantly rooted in scale-effect assumptions under urban expansion, with both linear and non-linear effects of urban growth on UEC extensively documented [14]. However, the status and drivers of UEC under shrinkage conditions remain underexplored. This gap obscures the feedback mechanisms of energy use under the joint influence of population distribution, industrial restructuring, and consumption inertia. Moreover, shrinkage is often intuitively assumed to reduce energy demand—a view that, when adopted from a total-demand perspective, may result in a mismatch between policy objectives and actual energy performance. There is thus an urgent need to incorporate per capita resource burdens into the urban transformation discourse. To address these gaps, this study evaluates the implications of urban shrinkage for UEC using Chinese cities as empirical cases. Drawing on a balanced panel dataset covering 281 prefecture-level cities from 2008 to 2021, we integrate a two-way fixed effects (TWFE) model with mediation analysis to assess both the direct and indirect effects of shrinkage on UEC, with particular attention paid to the roles of population dilution, industrial restructuring, and consumption inertia. This research underscores the urgency of integrating the shrinkage–energy nexus into the broader urban sustainability agenda and provides empirical insights into the nuanced responses of UEC under post-growth conditions.

2. Literature Review and Research Hypotheses

2.1. Theoretical Conceptions and Measurement of Urban Shrinkage

Urban shrinkage first emerged as a prominent phenomenon in major cities across Europe and North America during the latter half of the 20th century, typically characterized by population decline and housing vacancy [15]. To avoid the pejorative connotations of terms like “urban crisis”, researchers formally introduced the more neutral concept of “urban shrinkage” in studies of population migration in German cities, defining it as a sustained and nearly irreversible loss of urban population [16]. Since then, urban shrinkage has proliferated rapidly: by the early 21st century, three-quarters of Eastern European cities had experienced some form of shrinkage [17], with parallel reports emerging from Japan, Australia, and Russia [18]. The drivers of urban shrinkage are heterogeneous and have been interpreted through diverse theoretical lenses. Early explanations drew on multi-stage development models, urban life cycle theories, and the rise in suburbanization [19]. Later approaches emphasized capital-driven dynamics and the functional reorganization of urban space, framing shrinkage as a process of spatial deagglomeration [20]. Meanwhile, demographic frameworks have attributed shrinkage to falling birth rates and increasing life expectancy [21].

This growing body of research has facilitated the development of practical approaches to measure urban shrinkage, most commonly by tracking population changes and their consequences. Early empirical work often relied on single-dimensional, external indicators to qualitatively assess whether shrinkage was occurring. However, the multi-dimensional nature of urban shrinkage necessitates a pluralistic approach to quantification. Economic decline, land-use transformation, and fluctuations in habitat quality often accompany population loss and form the basis for different measurement strategies. Beyond demographic indicators, economic metrics—such as GDP per capita, industrial structure, and tax revenues—are among the most frequently employed [22]. Given that population loss within a defined geographic space inevitably increases per capita land consumption, indicators such as vacancy rates of residential or office buildings and the accessibility of public infrastructure are also commonly used [23]. From a landscape perspective, shrinkage has been measured using indices such as the largest patch index, Shannon diversity index, and leaf area index [24]. Among these, population change remains the most foundational and widely adopted metric, owing to its accessibility and representativeness. Increasingly, scholars tend to conceptualize urban shrinkage in a narrow sense—as a specific empirical phenomenon or stage—indicating a growing global consensus on its significance.

Despite the expanding scope of empirical studies, limited understanding of the overall shrinkage process continues to fuel misconceptions. In response, Haase et al. [15] developed and tested a conceptual model of urban shrinkage that captures the full process, including its antecedents, impacts, and feedback mechanisms. They emphasize the importance of contextual sensitivity and caution against imposing normative assumptions when analyzing shrinkage. Notably, they argue against constructing universal governance frameworks, warning that such efforts may overestimate the transferability of so-called “best practices”. Their work provides critical guidance for future studies grounded in diverse regional contexts.

2.2. Integrating Urban Shrinkage into the Green Development Agenda

Urban shrinkage has given rise to a series of adverse socioeconomic consequences. The net out-migration of the labor force—particularly young, highly skilled individuals—has led to demographic aging and weakened local capacities for technological innovation. Simultaneously, both population loss and declining spatial density have resulted in rising vacancy rates across urban space, which in turn have eroded fiscal revenues and raised operating costs for local governments [25]. Shrinking cities also face heightened social instability, often reflected in increased unemployment and crime rates [26]. In the current era, green development has emerged as a critical pathway toward high-quality urban transformation. Shrinking cities will not simply vanish from the global urban landscape; rather, accelerated population aging and prolonged economic stagnation—or even absolute decline—will continue to reinforce the trajectory of shrinkage. As such, urban shrinkage must be understood as an embedded and long-term component of regional green development agendas.

The critical issue is not to further explain how shrinkage occurs but to reframe it as a structural premise for environmental transitions. Existing urban environmental studies largely focus on the implications of expansion—highlighting the scale and agglomeration effects of urban growth. Numerous reports have documented environmental degradation caused by the high-intensity discharge of pollutants during the urbanization process [27], while others suggest that talent agglomeration and technological spillovers may help improve resource-use efficiency [5]. Echoing these findings on expansion, the environmental consequences of urban shrinkage are likewise dual in nature. Xiao et al. [28], for example, found that over 50 Chinese cities exhibited divergent carbon emission trajectories: while growing cities peaked in carbon emissions before 2014, shrinking cities experienced continued increases. Importantly, some of the environmental harms linked to urban shrinkage stem not from shrinkage per se but from its secondary impacts—namely fiscal stress and underutilized infrastructure. At the same time, shrinkage may offer opportunities to improve environmental quality. Studies have shown that many shrinking cities in the U.S. Rust Belt exhibit higher levels of vegetative cover, albeit largely due to unmanaged regrowth in vacant lots [29]. More intuitively, population loss can lead to reductions in overall resource consumption, municipal waste generation, and urban heat island intensity [30]. In the case of Berlin, empirical research indicates that ecosystem service capacity and quality are significantly enhanced under scenarios of shrinkage rather than urban expansion [31]. It is also crucial to recognize that even for a single factor, shrinkage may exert contradictory effects on related environmental indicators. For instance, studies have reported conflicting impacts on carbon emission intensity versus carbon efficiency [8,11]. To date, the dynamics and underlying mechanisms of energy use—particularly per capita energy burden—have received insufficient attention in the context of urban contraction. Expanding the research frontier in this area is therefore both necessary and timely.

2.3. Research Hypotheses

Beyond estimating the overall effect, identifying the key indirect mechanisms through which population outflow influences UEC is essential for a comprehensive understanding of how urban shrinkage affects energy consumption. This study structures the mechanism analysis around a population distribution–industrial restructuring–consumption inertia framework. First, the population-related consequences of urban shrinkage extend beyond a mere reduction in aggregate size, as such a reduction often initiates a fundamental shift in population distribution patterns. The contraction of urban populations weakens the economies of scale of centralized infrastructure systems, potentially raising unit energy costs in the short term. More importantly, declining population density—driven by sustained out-migration—substantially reduces aggregate demand for public services within the city. Under conditions where the expansion of new infrastructure is constrained, this density dilution effect significantly lowers energy consumption per unit area and may partially or fully offset the marginal increase in energy intensity caused by efficiency losses. As a result, the net effect at the per capita level may manifest as a reduction in UEC [32]. Second, industrial restructuring represents a critical dimension of urban contraction. In shrinking cities—especially those historically reliant on energy-intensive industries—labor outflows and looming environmental constraints increasingly render high-energy enterprises unsustainable under the new operating conditions. Many such enterprises either downscale or relocate to economically more dynamic regions. This spatial shift in energy demand transforms the industrial composition of shrinking cities, creating opportunities for cleaner service-oriented sectors to consolidate or emerge. Third, the inertia embedded in energy consumption patterns must be considered. The extensive distribution networks and ease of use of traditional fossil fuels make it difficult for cities to rapidly transition toward cleaner energy sources. Compared to other cities, shrinking cities often face severe fiscal limitations and underinvestment in energy infrastructure, which in turn reduce their capacity and motivation for energy structure upgrading. Taken together, the analysis posits three interrelated mediating pathways—population dilution, industrial restructuring, and consumption inertia—through which urban shrinkage may affect UEC dynamics.

The corresponding hypotheses derived from this theoretical framework are as follows:

Hypothesis 1.

Urban shrinkage tends to increase UEC, thereby exacerbating per capita energy burdens.

Hypothesis 2.

Urban shrinkage reduces UEC by diluting population density.

Hypothesis 3.

The contraction of the industrial sector mitigates upward pressure on UEC under shrinking conditions.

Hypothesis 4.

Amplified inertia in fossil energy consumption constitutes a key mediating channel through which urban shrinkage increases UEC.

3. Materials and Methods

3.1. Study Area

This study focuses on 278 prefecture-level cities across 30 provincial-level administrative regions in China, excluding Hong Kong, Macau, Taiwan, and Tibet. The selected cities span the eastern, central, and western regions and encompass all major regional hubs, ensuring broad representativeness in both geographical distribution and urban typology (Figure 1a). The study period spans from 2008 to 2021. The choice of 2008 as the starting year is grounded in three considerations: First, it marks the launch of China’s CNY 4 trillion stimulus package in response to the global financial crisis. The resulting overexpansion in infrastructure and real estate fundamentally altered urban development trajectories, planting institutional seeds for subsequent urban shrinkage. Second, in response to growing international pressure for climate governance, China began formalizing its national energy accounting system around the same time, which indirectly improved the quality and availability of city-level data. Third, 2008 represents a critical inflection point at which early signs of urban shrinkage began to emerge in China—characterized by net population outflows, industrial hollowing, and mounting fiscal stress in small- and medium-sized cities. Together, the sampled cities account for over 90% of China’s total energy consumption and resident population (Figure 1b), affirming the robustness and representativeness of the dataset.

3.2. Research Design

The construction of a multi-source database to quantify urban shrinkage and UEC serves as a foundational component of this study. During this stage, data on the permanent urban population, energy balance sheets, and related indicators are compiled and integrated through a series of formulaic transformations to generate the core variable panel spanning 2008 to 2021. The resulting spatiotemporal dataset provides qualitative insight into the relationship between urban shrinkage and UEC. Spatial visualizations are produced using ArcGIS (version 10.6). In the baseline empirical analysis, a TWFE model is employed to estimate the overall impact of urban shrinkage on UEC. Subsequently, the potential mediation pathways are tested using population density (PD), the share of secondary industry (SSI), and the share of traditional energy (STE) as mediating variables. Mediation analysis is conducted using both a stepwise regression approach and a Bootstrap procedure to ensure the reliability of mechanism identification. Given the spatial interdependence among cities, spatial econometric models are introduced as robustness checks. All descriptive statistics, specification tests, and model estimations are implemented using Stata (version 17) with official command packages, ensuring analytical rigor. Interpretation of the results is grounded in both the visualized spatial patterns and estimated parameters.

3.3. Methodology

3.3.1. Linear Econometric Model

The STIRPAT framework is widely used to identify the driving factors of environmental indicators and energy, with a particular emphasis on the effects of population, wealth, and technology [5]. As an extension of the IPAT model, STIRPAT effectively mitigates the quantification bias caused by the oversimplification of the framework by introducing random error terms and control variables. The basic form of the IPAT equation is as follows:

$$E_{it}=\alpha P_{it}^{\beta }{A}_{it}^{\omega }{T}_{it}^{\phi }{\epsilon }_{it}$$

(1)

where E represents the dependent variable; P, A, and T correspond to the population, wealth, and technology factors, respectively. α is the constant term, and ε is the random error term. β, ω, and φ are the elasticity coefficients for each factor, and i and t represent the city and year codes, respectively. The logarithmic form of the IPAT equation is thus as follows:

$$\ln E_{it}=\ln \alpha +\beta \ln P_{it}+\omega \ln A_{it}+\phi \ln T_{it}+{\epsilon }_{it}$$

(2)

In this study, the dependent variable corresponds to UEC, and the population factor corresponds to urban shrinkage. The equation is further derived as follows:

$$Urban energy consumptio{n}_{it}={\beta }_0+{\beta }_1{\left(Urban shrinkage\right)}_{it}+{\displaystyle \sum _{k=1}^m{\rho }_k{C}_{it}^k}+{\epsilon }_{it}$$

(3)

where C encapsulates the wealth factor, technology factor, and other control variables. β₁ and ρ_k represent the estimated coefficients for urban shrinkage and variable k, respectively.

3.3.2. Spatial Econometric Model

It has been reported that regional energy consumption in China exhibits spatial network correlation [33]. Under this premise, neglecting this characteristic in the quantitative framework poses a potential risk of amplifying estimation bias. After baseline regression, this study employs a spatial econometric model as a robustness check tool, incorporating the spatial correlation and heterogeneity of UEC into the modeling process. The Spatial Error Model (SEM) defines the spatial error term as a precursor to the spatial characteristics of the dependent variable. The equation is as follows:

$$\begin{array}{c}\hfill Urban energy consumptio{n}_{it}={\beta }_1{\left(Urban shinkage\right)}_{it}+{\displaystyle \sum _{j=1}^m{\rho }_j{C}_{it}^j}+{\gamma }_{it}\hfill \\ \hfill {\gamma }_{it}=\lambda {\displaystyle \sum _{j=1}^N{w}_{ij}{\gamma }_{jt}}+{\epsilon }_{it}\hfill \end{array}$$

(4)

where γ represents the spatial autocorrelation error term, and λ is the spatial error coefficient; w is the spatial matrix element. The Spatial Lag Model (SLM) emphasizes the necessity of spatial lag terms, assuming that the development trend of a city’s indicators is influenced by the status of the same indicator in neighboring cities. The equation is as follows:

$$\begin{array}{c}Urban energy consumptio{n}_{it}=\mu {\displaystyle \sum _{j=1}^N{w}_{ij}Urban energy consumptio{n}_{jt}}\hfill \\ \hfill \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} +{\beta }_1{\left(Urban shinkage\right)}_{it}+{\displaystyle \sum _{k=1}^m{\rho }_k{C}_{it}^k}+{\epsilon }_{it}\end{array}$$

(5)

where μ represents the spatial lag coefficient. By adding the spatial error term to the SLM, which accounts for both the intrinsic development inertia of UEC and the surrounding spillover effects, this study constructs the following Spatial Durbin Model (SDM):

$$\begin{array}{c}Urban energy consumptio{n}_{it}=\mu {\displaystyle \sum _{j=1}^N{w}_{ij}Urban energy consumptio{n}_{jt}}\hfill \\ \hfill \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} +{\beta }_1{\left(Urban shinkage\right)}_{it}+\theta {\displaystyle \sum _{j=1}^N{w}_{ij}}{\left(Urban shinkage\right)}_{jt}\\ \hfill \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em} +{\displaystyle \sum _{k=1}^m{\rho }_k{C}_{it}^k}+{\displaystyle \sum _{k=1}^m\left(\eta {\displaystyle \sum _{j=1}^N{w}_{ij}{C}_{jt}^k}\right)}+{\epsilon }_{it}\hfill \end{array}$$

(6)

where θ and η are used to represent the effects of the explanatory variables of neighboring cities on the UEC of the current city. Furthermore, the LM and robust LM tests are used to assess the practical applicability of the three spatial econometric models outlined above.

3.4. Variable Selection and Data

3.4.1. Core Variables

In the process of regional reconstruction triggered by urban shrinkage, per capita energy indicators possess strong identificatory power, as they reveal the structural risks inherent in individual loads. In this study, per capita energy consumption (PEC) is defined as the representative variable for UEC, a strategy that not only retains the informational foundation of urban energy utilization but also incorporates the real mapping of population dynamics on the energy system’s pressure. Additionally, PEC has been widely used in studies on regional energy performance evaluation, energy accessibility analysis, and sustainability boundary determination [34,35], further enhancing its practical relevance as a core variable in this paper. A comprehensive review of existing sources confirms the lack of official long-panel data on urban energy consumption in China. Therefore, aggregated and disaggregated energy consumption data are derived from national and provincial energy balance tables in the China Energy Statistical Yearbook, combined with sectoral value-added and resident population to construct weighted factors, employing a top-down proportional scaling method [36]. This study references BP’s Statistical Review of World Energy to validate the accuracy of the indicator data [37]. Comparative analysis of city-level aggregates reveals that the generated data on PEC, as well as STE and renewable energy (SRE), exhibit similar trends and relatively small absolute deviations compared to BP’s reported figures (Figure 2). Furthermore, given the strong correlation between traditional energy consumption and CO₂ emissions, we calculated the bivariate correlation coefficients between total energy consumption and city-level CO₂ emissions reported by the China Emission Accounts and Datasets [38]. The annual correlation coefficients were all statistically significant at the 1% level, ranging from 0.661 to 0.814. Collectively, these results affirm the high consistency of the constructed urban energy consumption dataset with existing records.

The identification of shrinking cities is mainly based on three core perspectives: population dynamics, economic resilience, and a composite indicator system. Among these, population change is the most fundamental and widely used information source for identifying whether a city is in an expansion or shrinkage phase. Referring to existing research paradigms [8], this study constructs the urban shrinkage index (USI) based on the annual population change rate, with the following formula:

$$US inde{x}_{it}=-100\times {\displaystyle \frac{R{p}_{it}-R{p}_{i,\left(t-1\right)}}{R{p}_{i,\left(t-1\right)}}}$$

(7)

where Rp represents the urban resident population. Given that this study empirically focuses on urban shrinkage rather than the broader evolution of urban scale, weight coefficients are used for numerical conversion to both align with the research topic and adjust the data width. Positive and negative values of the index indicate that the target city is in a state of shrinkage and expansion, respectively.

3.4.2. Mediating Variables

This study selects typical macro variables to explore the potential pathways through which urban shrinkage reshapes UEC. In response to the spatial restructuring of populations caused by urban shrinkage, we introduce PD as a measure to test the population dilution pathway. It is expected that a decrease in population density will help alleviate per capita energy load, reflecting the passive compression of energy demand due to population migration and spatial dispersal under conditions of urban shrinkage.

At the same time, UEC is heavily influenced by the composition of the industrial system. To identify the energy-saving effects of the industrial restructuring pathway, we select SSI as a representative variable. In traditional development models, the secondary sector typically exhibits high-energy consumption characteristics. During urban shrinkage, phenomena such as industrial hollowing and deindustrialization are frequent, and a decline in the share of the secondary sector can reflect the structural suppression of UEC due to the reduction in the proportion of high-energy-consuming industries.

Furthermore, UEC is also modulated by the lock-in effect of the energy structure. This study uses STE to examine the validity of the consumption inertia pathway. Under the dual constraints of industrial decline and fiscal tightening, shrinking cities often face institutional inertia and capital bottlenecks in energy system upgrades. This exacerbates dependence on traditional energy sources, which becomes a key pathway for the continued rise in UEC.

3.4.3. Control Variables

To improve the accuracy of net effect identification, the benchmark model incorporates control variables from various dimensions, including economic factors, policy, and external capital participation [39,40,41,42]. Detailed information on these variables is presented in

Table 1. Selection and definition of control variables.

Variable	Calculation	Symbol
Economic endowment	Ratio of regional GDP to resident population	EE
Low-carbon policy	Quantified score of low-carbon policy intensity	LP
Technological innovation	Share of scientific expenditure in regional GDP	TI
Urbanization level	Proportion of urban population in total population	UL
Foreign trade	Share of foreign direct investment in regional GDP	FT

.

3.4.4. Data Collection and Processing

Resident population data mainly come from the China City Statistical Yearbook and the National Population Census dataset, supplemented by city development bulletins. Data for control variables are primarily collected from the China City Statistical Yearbook and the China Urban and Rural Construction Statistical Yearbook; energy policy data are identified using NLP algorithms applied to government texts [43]. Missing data are supplemented using trend functions and hotspot imputation methods. Descriptive statistics of the variables indicate that the differences between the mean values and data ranges of the variables are small (

Table 2. Descriptive statistics of variables.

Variable	Unit	Mean	Max	Min	SD	Kurtosis
PEC	tons/person	2.596	40.942	0.392	2.692	85.905
USI	—	−0.295	30.869	−93.225	3.359	216.122
EE	CNY ten thousand	4.941	22.495	0.360	3.177	6.732
LP	—	65.528	195.250	0.000	44.055	2.570
TI	%	0.260	6.310	0.013	0.263	94.393
UL	%	54.157	100.000	13.140	15.995	2.991
FT	%	1.642	13.164	0.000	1.702	6.778
PD	persons/km2	5.742	9.089	1.653	0.979	4.258
SSI	%	46.530	90.972	11.317	11.145	3.605
STE	%	93.318	99.952	73.567	5.441	3.104

), and logarithmic transformations are not required.

4. Results and Analysis

4.1. Spatiotemporal Patterns of UEC and Urban Shrinkage

Figure 3 illustrates the temporal trajectories and spatial configurations of UEC and urban shrinkage, revealing a co-evolutionary pattern across time and space. Overall, UEC exhibits a slow upward trend, with two notable periods of acceleration—prior to 2012 and after 2016—corresponding, respectively, to key phases of industrial restructuring and institutional shifts in the energy system. In parallel, the degree of urban shrinkage experienced two pronounced surges before stabilizing at a relatively high level, suggesting that population-driven structural reconfigurations of the urban system have become increasingly path-dependent. The mean curves of both indicators show synchronized movements during multiple pivotal years, offering preliminary evidence of an interlinked dynamic between shrinkage and UEC. Spatially, UEC hotspots evolved from early-stage fragmentation toward a concentrated clustering in northern interior regions and resource-dependent cities across Central and Western China. The spatial distribution of urban shrinkage underwent similar restructuring. By the end of the study period, both phenomena exhibited strong spatial overlap in regions such as the Bohai Rim, Northeast China, and the central hilly belt. This spatial co-location delineates a non-symmetric response pattern—characterized by population decline coupled with increasing energy demand—which empirically maps the so-called “urban energy paradox” in shrinking cities. These spatiotemporal linkages provide an empirical starting point for subsequent causal identification and pathway decomposition, while also underscoring the analytical imperative of shifting from aggregate energy consumption metrics to per capita performance indicators when examining energy transitions in the context of urban contraction.

4.2. Driving Impacts of Urban Shrinkage on UEC

Table 3. Estimated effects of urban shrinkage on UEC based on linear econometric models.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
USI	0.0223 ***	0.0129 ***	0.0119 ***	0.0105 ***	0.0212 ***	0.0167 ***
	(0.0037)	(0.0034)	(0.0034)	(0.0033)	(0.0037)	(0.0037)
EE			0.1658 ***	0.1109 ***	0.1301 ***	0.0942 ***
			(0.0084)	(0.0110)	(0.0056)	(0.0071)
LP			0.0007 ***	0.0024 ***	0.0008 ***	0.0012 ***
			(0.0002)	(0.0007)	(0.0002)	(0.0004)
TI			−0.3052 ***	−0.3330 ***	−0.4653 ***	−0.4984 ***
			(0.0609)	(0.0604)	(0.0578)	(0.0571)
UL			0.0049 **	−0.0059 **	0.0108 ***	0.0047 ***
			(0.0019)	(0.0025)	(0.0012)	(0.0015)
FT			0.0003	0.0085	−0.0024	−0.0001
			(0.0110)	(0.0110)	(0.0072)	(0.0072)
constant	2.6025 ***	2.0318 ***	1.5493 ***	1.9800 ***	1.3356 ***	1.5769 ***
	(0.0109)	(0.0373)	(0.0965)	(0.1248)	(0.0609)	(0.0783)
City-FE	Yes	Yes	Yes	Yes	Yes	Yes
Year-FE	No	Yes	No	Yes	No	Yes
Winsorization	No	No	No	No	Yes	Yes
Obs.	3892	3892	3892	3892	3892	3892
R2	0.0101	0.1791	0.1847	0.2086	0.3085	0.3388

Note: Robust standard errors in parentheses; ***, and ** denote statistical significance at the 1% and 5% levels, respectively.

presents the estimated effects of urban shrinkage on UEC under a linear econometric framework. Prior to formal regression analysis, we conducted panel stationarity tests using the ADF, Hadri LM, and IPS methods to confirm the stability of the variables [44,45]. The Hausman test further supports the appropriateness of employing fixed effects specification. The six model variants represent a diversified empirical strategy, integrating robust design considerations across three dimensions: model complexity, the treatment of fixed effects, and the mitigation of outlier influence. Across all specifications, urban shrinkage consistently exhibits a statistically significant positive effect on UEC at the 1% level. Across all four fully specified models, the estimated coefficients indicate that a one-unit increase in the shrinkage index corresponds to an average increase in per capita energy consumption of approximately 0.01 to 0.02 units, thereby supporting Hypothesis 1. Interpretations of the control variables are based on the full-model results, summarized following a majority-rule principle. Economic endowment and urbanization level both exert significantly positive effects on UEC, suggesting that advanced stages of development are typically accompanied by higher energy consumption intensity. In contrast, technological innovation exerts a mitigating effect, in line with theoretical expectations that improved efficiency can drive structural energy optimization. Interestingly, the policy variable capturing low-carbon initiatives is also statistically significant but exhibits a positive coefficient. This counterintuitive result may reflect the fact that many city-level green policies in China still prioritize energy security and economic stability and thus have yet to translate into substantive energy reductions. Lastly, the degree of external openness shows no significant relationship with UEC across all specifications, suggesting that openness is not a primary driver of urban energy dynamics in this context.

4.3. Mechanisms Underlying the Link Between Urban Shrinkage and UEC

A two-stage mediation model is employed to disentangle the structural pathways through which urban shrinkage affects UEC (

Table 4. Mechanism test based on TWFE models.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	Y = PD	Y = PEC	Y = SSI	Y = PEC	Y = STE	Y = PEC
USI	−0.0045 ***	0.0120 ***	−0.0584 ***	0.0123 ***	0.0163 **	0.0094 ***
	(0.0004)	(0.0034)	(0.0218)	(0.0033)	(0.0073)	(0.0033)
PD		0.3350 **
		(0.1569)
SSI				0.0313 ***
				(0.0025)
STE						0.0670 ***
						(0.0075)
constant	5.6841 ***	0.0757	43.8502 ***	0.6079 ***	92.9970 ***	−4.2529 ***
	(0.0133)	(0.9008)	(0.8172)	(0.1640)	(0.2747)	(0.7079)
control variables	Yes	Yes	Yes	Yes	Yes	Yes
Obs.	3892	3892	3892	3892	3892	3892
R2	0.1508	0.2096	0.6130	0.2418	0.7420	0.2258

Note: Robust standard errors in parentheses; ***, and ** denote statistical significance at the 1% and 5% levels, respectively.

). Columns (1) and (2), which test Hypothesis 2, highlight a sustainable development pathway whereby shrinkage indirectly reduces UEC through the dilution of population density. Unlike conventional views in which high density is equated with superior energy efficiency, the decline in density within shrinking cities does not necessarily signal a loss of technical efficiency in infrastructure. Instead, the energy-saving effect is largely driven by the systemic contraction of energy demand that accompanies demographic depletion. Lower density manifests as looser spatial structure and functional downgrading [46]; once population support wanes, service facilities and industrial systems scale back or even cease operations, sharply reducing utilization rates. At the same time, population outflow from core areas and higher land/vacancy ratios curb the operating frequency of energy-intensive installations, lowering energy use per unit area and per capita. Thus, density reduction produces an indirect, structurally induced energy-saving pathway that conventional studies often overlook.

The secondary sector has long been viewed as the structural load-bearer of urban energy systems, with even marginal shifts in its share capable of triggering system-wide turbulence [47]. Regression results demonstrate that urban shrinkage significantly reduces the share of the secondary industry, thereby mitigating UEC, confirming Hypothesis 3. This pathway implies that the withdrawal of energy-intensive industrial segments has become a key mechanism for relieving the energy burden in shrinking cities. Unlike proactive green transitions, this adjustment is not an endogenous outcome of environmental regulation or technological substitution but rather a structural adaptation to dwindling factor endowments and declining locational attractiveness. Contraction in labor, land, and capital jointly undermines the agglomeration logic of high-consumption industries, pushing the industrial mix toward less energy-intensive activities and generating an unintended reduction in energy intensity.

Shrinking cities do not automatically evolve toward more refined or low-carbon energy systems. Our analysis reveals that continued contraction increases the share of traditional energy in end-use consumption, and this rise is positively and significantly correlated with UEC. The validation of Hypothesis 4 underscores the lack of synergy between urban shrinkage and the clean energy transition within the energy system. Instead, it exposes pervasive path dependence and lock-in within urban energy systems. As modern energy infrastructure loses economic and demographic support, shrinking municipal budgets and weakening institutional capacity hinder the maintenance and expansion of renewable alternatives. By contrast, conventional fuels—owing to their stable supply chains—often become the default option for sustaining minimal urban functions [48]. The energy structure thus gravitates toward the existing high-carbon trajectory and, through the interplay of technology, institutions, and user practices, solidifies into a difficult-to-reverse, high-consumption mode. These findings caution that urban shrinkage can induce a regressive reconfiguration of energy use, and that structural inertia must be explicitly addressed in any putative green-transition agenda.

Considering the latent risk of effect bias in identifying mechanism paths with stepwise regression [49], this study further introduces the Bootstrap resampling technique to identify the direct, indirect, and total effects of urban shrinkage on UEC under each pathway (Figure 4). The results show that the confidence intervals for the indirect effects corresponding to the three pathways do not include the zero-reference line, indicating that the population dilution, industrial restructuring, and consumption inertia pathways are statistically significant and form important transmission channels for the impact of urban shrinkage on UEC.

4.4. Robustness Checks

To further validate the robustness of the UEC response effect, this study incorporates spatial effects into the identification framework to control for potential spatial dependence and external influence transmission among cities. Specifically, we construct two types of weight matrices—geographical inverse distance matrix and economic–geographical nested matrix—to measure spatial effects [50,51] and embed them into spatial econometric models. The results in

Table 5. Spatial effect test.

Matrix	Test	Statistic	p Value
Geographical inverse distance	Moran’s I	161.265	0.000
	LM-spatial error	3197.436	0.000
	robust LM-spatial error	3742.457	0.000
	LM-spatial lag	40.981	0.000
	robust LM-spatial lag	586.002	0.000
Economic–geographical nested	Moran’s I	194.098	0.000
	LM-spatial error	4759.524	0.000
	robust LM-spatial error	5375.588	0.000
	LM-spatial lag	40.843	0.000
	robust LM-spatial lag	656.907	0.000

show that the LM and its robust version’s test statistics significantly support the existence of spatial effects.

To maintain comparability with the baseline estimation, this study estimates SEM, SLM, and SDM based on the TWFE setup (

Table 6. Robustness test based on spatial econometric models.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	SEM	SLM	SDM	SEM	SLM	SDM
USI	0.0087 ***	0.0094 ***	0.0071 **	0.0061 **	0.0075 **	0.0074 **
	(0.0031)	(0.0031)	(0.0030)	(0.0028)	(0.0029)	(0.0030)
control variables	Yes	Yes	Yes	Yes	Yes	Yes
γ	2.3143 ***			5.1878 ***
	(0.0031)			(0.0731)
μ		2.0813 ***	2.2004 ***		5.2859 ***	2.1291 ***
		(0.1164)	(0.0875)		(0.0692)	(0.0548)
Matrix	Geographical inverse distance			Economic–geographical nested
Obs.	3892	3892	3892	3892	3892	3892
R2	0.1334	0.0983	0.0892	0.1231	0.0178	0.0651

Note: Robust standard errors in parentheses; ***, and ** denote statistical significance at the 1% and 5% levels, respectively.

). Although spatial diagnostic indicators suggest that SDM is the optimal estimation framework [52], the positive effect of urban shrinkage on UEC remains stable across various spatial models. The results from this part indicate that the strengthening trend of UEC is not caused by model specification but rather forms a robust response under the multiple structural constraints in shrinking cities. This further corroborates the structural association between urban shrinkage and UEC enhancement proposed in this study.

5. Discussion

5.1. Structural Paradox Between Urban Shrinkage and Energy Transition

The relationship between urban shrinkage and energy transition is not a harmonious synergy but rather a profound structural tension. Under the multiple constraints of population loss, industrial shrinkage, and fiscal tightening, the basic load carried by urban energy systems has not decreased synchronously; on the contrary, there has been a persistent increase in per capita energy consumption. The asymmetric pattern of population reduction and energy increase identified in this study reveals a hidden energy structural paradox in shrinking cities: the contraction of quantity does not naturally lead to qualitative transitions, and resource depletion does not spontaneously trigger system optimization. This paradox highlights the failure of conventional energy system logic in the context of urban shrinkage.

A widespread institutional collusion centered on the logic of growth exists between contemporary urban planning and energy allocation [53,54]. The scale of infrastructure and technological choices often rely on linear extrapolations of population, industry, and fiscal size, resulting in large-scale, supply oriented, centralized control, and a path dependence on long-term stable demand [55]. In expanding cities, this logic forms a self-reinforcing loop of population agglomeration, service concentration, and energy efficiency improvement. Once the city enters a shrinkage trajectory, this loop rapidly disintegrates. Although the continuous decline in population density weakens the economies of scale in infrastructure and leads to reduced utilization of supporting facilities and passive shrinkage of redundant energy systems, the external expansion of energy consumption is effectively suppressed during this overall decline in usage frequency and energy demand. This phenomenon indicates that while low-density states may not enhance unit energy efficiency, they can achieve energy-saving effects by compressing systemic load and reducing marginal energy consumption demand. This provides a theoretical angle for understanding the non-efficiency-driven energy-saving mechanisms in urban shrinkage.

More critically, the sunk costs and technological rigidity of energy systems make them one of the most difficult structural elements to optimize in shrinking cities. On the one hand, the optimal cost realization of centralized power grids, district heating, and traditional fuel chains depends on scale and density [56]; once imbalanced, these systems fall into operational difficulties with low marginal returns. On the other hand, the tightening of urban fiscal resources further restricts pre-investment in clean energy infrastructure, leading to resistance to new energy penetration due to institutional inertia and friction. The dual lock-in development inertia exposes shrinking cities to high vulnerability in energy transition, making it difficult to replicate the structural benefits brought by green growth.

Within this tension, the existing paradigm of “smart shrinkage” struggles to adequately address energy issues. This paradigm focuses on land reclamation, service integration, and institutional adaptation to fiscal constraints [57], while paying insufficient attention to the rigidity of energy infrastructure and the transformation mechanisms of energy consumption structures. Energy issues should be placed at the core of urban shrinkage research, rather than merely as an additional dimension in the post-development phase. This paper advocates for introducing the concept of energy-resilient shrinkage into urban shrinkage research, emphasizing the flexible adaptation of energy infrastructure, structural re-adjustment of energy consumption, and institutional guarantees for energy equity in the context of multi-dimensional shrinkage. The core focus is not simply improving energy efficiency but rather constructing an urban governance model capable of adapting to shrinkage. This research also resonates with a key viewpoint in the sociology of infrastructure, namely, that energy systems are not only material networks but also institutional presences that shape urban space and everyday life.

In conclusion, urban shrinkage and energy transition do not form a naturally coupled win–win path but are a complex reconstruction process accompanied by tensions and institutional constraints. The key to understanding this process lies in shifting from population decline to systemic coupling disintegration, from the rationality of reduction to the adjustability of structure, and from apparent energy savings to the redistribution of load. This shift concerns the re-setting of both technology and institutions and reconstructs the ethical foundation of energy justice in the context of urban decline.

5.2. Policy Implications and Strategic Shifts

If regional policies remain confined to replicating established empirical pathways, they will not only fail to effectively address the structural shocks induced by urban shrinkage but may also exacerbate energy system misalignment and regional inequalities. Based on the three identified causal mechanisms, this section proposes a set of governance strategies that are both structurally sensitive and ethically embedded. By integrating energy subsystem functions with principles of energy justice, these strategies aim to provide actionable policy pathways for the transformational governance of shrinking cities.

The sustained dilution of population density poses the most direct challenge to the energy system—not primarily through changes in aggregate consumption but via structural degradation in urban spatial efficiency. Given the substantial contraction in internal urban demand, part of UEC is compressed during the shrinkage process. This indicates that energy governance should not attempt to reverse population dispersal trends but rather focus on constructing adaptive infrastructure systems. In this context, the National Development and Reform Commission (NDRC) should take the lead in guiding subordinate local agencies to embed dual-core orientations of population elasticity and energy elasticity into urban energy planning. In practice, efforts must be made to prevent a new wave of energy deprivation resulting from redundant capacity cutbacks. Specifically, the deployment of modular combined heat and power (CHP) units and distributed renewable microgrids in low-density areas should be promoted, alongside establishing minimum energy service guarantees supported by fiscal mechanisms to ensure comprehensive energy accessibility in peripheral regions. The positive role of secondary industry withdrawal in alleviating UEC pressures must be fully acknowledged. However, due to absent complementary institutional guidance, such industrial adjustments risk triggering fiscal stagnation, employment contraction, and industrial heritage vacancy—thereby accumulating new governance risks [58]. Energy-intensive enterprises often exhibit structural stickiness in most shrinking cities, and their exit pathways require careful institutional design. It is recommended that the Ministry of Ecology and Environment jointly with the Ministry of Industry and Information Technology establish pilot “Green Industrial Transformation Zones” to integrate resource recycling and low-carbon manufacturing strategies into existing industrial land, thereby fostering industry–energy co-evolution pathways. Concurrently, fiscal subsidies should incentivize the green exit of high-energy firms, including but not limited to equipment depreciation buybacks, energy-saving retrofitting grants, and targeted incentives for participation in green supply chains. To prevent the formation of extensive energy “black holes” following industrial exit, retired land parcels should be prioritized for energy storage facility deployment within an integrated energy–land–capital governance framework. Additionally, electric vehicle infrastructure should serve as critical nodes for accommodating micro energy loads to support the stable operation of distributed energy systems [59]. This initiative aligns with recent systematic analyses of the spatial evolution of charging infrastructure [60]. It is important to remain vigilant about path dependency within energy systems, which renders shrinking cities highly vulnerable during the clean energy transition. The technological trajectories of centralized energy supply, fiscal capacity shrinkage at the local level, and barriers to new energy integration constitute multiple traps—especially under a carbon market regime that fails to effectively redistribute resources to disadvantaged regions—exposing governance structures to systemic risks [61]. Accordingly, it is recommended that the National Energy Administration establish an energy vulnerability identification mechanism incorporating a composite indicator system encompassing infrastructure aging, energy service continuity, and residents’ payment capacity, thereby providing policy safety nets for high-risk areas. Practical measures include priority allocation of carbon trading quotas, enhanced fiscal transfers, and streamlined approval processes for new energy projects. Furthermore, policy design should be embedded within an energy justice framework to ensure a fair transition. This entails equitable quota distribution guaranteeing adequate energy access for vulnerable communities, preferential deployment of new energy infrastructure in aging neighborhoods and peripheral areas, and safeguarding residents’ rights to information and participation throughout the process.

To enhance the institutional enforceability of these policy recommendations, a three-tier responsibility system is proposed. At the central level, the NDRC and the National Energy Administration should jointly lead the establishment of an energy adaptability assessment system for shrinking cities, incorporating it into broader urban renewal plans and carbon budget targets. At the local level, cross-sectoral task forces should be formed to develop energy transition roadmaps with clearly defined milestones and responsibilities. At the market level, energy enterprises should restructure their profit models by developing product portfolios tailored for low-density supply and socially equitable pricing, while actively participating in regional clean energy platform construction under regulatory guidance. In summary, governance strategies addressing the profound restructuring of energy systems under urban shrinkage must balance structural adaptability with ethical justice. Institutionalized transitions should compensate for the apparent decline caused by population loss, ultimately advancing a new urban energy paradigm characterized by adaptability, resilience, and inclusiveness.

5.3. Limitations and Future Directions

This paper first systematically reveals the mechanisms through which urban shrinkage reshapes UEC via the three structural pathways of population, industry, and energy. This provides empirical evidence at the mechanistic level for understanding the energy consumption logic of shrinking cities. However, to gain a deeper understanding of the evolution of energy systems within the context of shrinkage, further expansion is needed in terms of behavioral characterization, dynamic modeling, and institutional dimensions.

While focusing on macro structural variables helps clarify the path mechanisms under institutional constraints, it does not address the individual decision-making dimensions in the evolution of UEC. During shrinkage, migration willingness, household energy preferences, and enterprise exit strategies collectively shape the urban energy demand landscape. Future research should introduce high-frequency microdata and behavioral modeling frameworks to capture the intrinsic layers of energy responses. Additionally, the current identification strategy, based on linear and mean effect assumptions, is unable to reveal the structural transitions and path lock-in characteristics of urban systems during the shrinkage process. Identifying non-linear turning points of UEC will depend on the incorporation of spatiotemporal interaction models, dynamic threshold identification, and feedback sensitivity tests. A more fundamental shift is required, where energy system adjustments should not be limited to efficiency improvements but must be embedded within an energy justice institutional framework. The elderly, low-income households, and residents of marginalized spaces in shrinking cities are more susceptible to energy poverty risks. The transition from resource allocation logic to rights protection mechanisms in energy governance will be the core issue for future research on deepening urban energy equity.

6. Conclusions

Based on panel data from 278 cities in China from 2008 to 2021, this study focuses on the development mechanisms of UEC under the context of urban shrinkage, constructing and validating three structural pathways: population dilution, industrial restructuring, and consumption inertia. Using a combination of two-way fixed effects models, path mediation decomposition, and spatial robustness checks, this study systematically depicts the structural logic of energy transition in shrinking cities. This paper challenges the conventional expectation that population reduction leads to energy reduction, asserting that there exists a paradoxical mechanism of population loss and energy increase in shrinking cities. The main findings include the following:

(1)

The spatiotemporal pattern of urban shrinkage and UEC shows a typical asymmetric response of population decline and increased energy burden. UEC shows a generally slow upward trend, with acceleration before 2012 and after 2016, highly coupled with the stage of urban shrinkage. Spatially, high UEC areas gradually concentrate in shrinking cities in the central and western regions and Northeastern China, exhibiting a high degree of spatial co-occurrence.

(2)

The decline in population density becomes a structural pathway for UEC reduction, challenging the traditional assumption that high density equals high efficiency. While high density tends to improve energy efficiency in expanding cities, in shrinking cities, population evacuation leads to spatial functional degradation and service demand convergence, which reduces the energy load carried at the individual scale and thus compresses per capita energy consumption. This pathway indicates that density is not a one-dimensional energy-saving variable but rather a bidirectional mechanism factor constrained by the urban life cycle context.

(3)

The decline in the share of secondary industries has a significant energy-saving effect, essentially forming a structural channel through industrial exit. The trend of industrial hollowing out and de-industrialization during urban shrinkage reduces the proportion of energy-intensive manufacturing, thereby weakening the basis for UEC enhancement. However, this process also implies employment structure shocks and fiscal contraction risks, suggesting that industrial transformation should proceed in tandem with energy reduction.

(4)

The rise in the share of traditional energy consumption constitutes a key channel for UEC reinforcement, reflecting institutional energy lock-in trends. Under the dual pressure of fiscal constraints and technological stagnation, shrinking cities tend to rely on existing fossil energy infrastructure, forming path dependency on energy structures, which further increases UEC. This pathway reveals the institutional risks of high-carbon inertia, indicating that energy reduction is not just a technical process but also a matter of institutional reconstruction.

Based on these conclusions, this paper proposes a resilience-oriented energy reduction strategy framework: First, implement a density-sensitive energy system layout, reshaping energy service units with population elasticity as the premise. Second, optimize the industrial substitution and re-embedding mechanisms, guiding the smooth exit of energy-intensive industries and the effective integration of low-carbon industries. Third, advocate for integrating energy justice into urban governance practices to address the energy vulnerabilities of shrinking areas, improving baseline service guarantees, procedural participation mechanisms, and spatial equity assessments. Urban shrinkage should not be viewed as an unmanageable process of decline, but rather as a real-world scenario for testing whether energy systems possess structural resilience and institutional adaptability. The ongoing energy transition calls for not only precise structural diagnostics but also the justice of institutional design. Only through organic synergy between structural understanding, behavioral response, and ethical frameworks can such synergy support a resilient low-carbon urban future.