The Effects of Urban Sprawl on Electricity Consumption: Empirical Evidence from 283 Prefecture-Level Cities in China

Abstract

Under the urban development trend of sprawl, improving energy use efficiency is a proper way to promote green and low-carbon construction in cities. This paper uses panel data from 283 prefecture-level and above cities in China from 2008 to 2019 to measure the urban sprawl index, and analyze the spatial-temporal evolution law of urban sprawl and electricity consumption. The relationship between urban sprawl and electricity consumption is empirically examined, and the differential effect of urban sprawl on electricity consumption is analyzed. Finally, the impact of urban sprawl on electricity consumption based on a spatial perspective is explored in depth by establishing a spatial error model. We found the following: (1) The levels of urban sprawl and urban electricity consumption are on the rise. The spatial distribution of urban sprawl is more dispersed, and cities with high electricity consumption levels are mostly concentrated in the eastern coastal areas. (2) Urban sprawl exacerbates electricity consumption, and this conclusion is still robust after a series of robustness tests were conducted and endogeneity issues were taken into account. In terms of the influence mechanism, urban sprawl mainly affects electricity consumption by changing the allocation of land resources, increasing the dependence on private transportation, and inhibiting green technology innovation. (3) The incremental effect of urban sprawl on electricity consumption is more pronounced in cities with high sprawl levels, weak environmental regulations, and low green innovation levels, as well as in west cities. (4) Urban sprawl and electricity consumption both have a significant positive spatial correlation. Electricity consumption of cities is not only related to their own regions but also influenced by the adjacent regions, and the spatial correlation is mainly reflected in the random error term. This paper deepens the understanding of the basic laws of urban sprawl affecting urban low-carbon development, which also has implications for new urbanization strategies and green development.

1. Introduction

Cities are spatial carriers where capital, labor, and infrastructure are highly aggregated, and they are the main battlegrounds for economic activities such as the allocation of resources and trade transactions in human societies [1]. The development and construction of cities is an important task for many countries, and the urbanization process of each country is advancing by leaps and bounds [2,3]. According to the World Cities Report 2022 published by UN-Habitat, the world’s urbanization rate has reached 56% in 2021. By 2050, this figure will rise to 68%, with an increase of about 2.2 billion urban dwellers, and a significant increase in urbanization levels is expected in all regions of the world [4]. Similarly, China’s urban development has been dramatic, with the country’s urbanization level growing at an unprecedented rate, from 17.92% in 1978 to 64.7% in 2022 [5]. Promoting urbanization is a powerful engine and an important strategy for China’s socio-economic development [6]. Urbanization is regarded as a necessary path to modernization, and the 20th National Congress of the Communist Party of China further rationalized the idea of a new urbanization strategy. Rational urbanization promotes the modernization of industrial structures and also helps improve energy efficiency and environmental quality [7,8].

However, in reality, disorderly expansion of urban space and reckless development of land resources have become common problems in the urbanization process of countries around the world [9]. Land element-based urbanization promotes land development and utilization, but land expansion that is detached from the actual needs of the population and economic development may lead to the phenomenon of a mismatch in land development in China [10], which is called urban sprawl [11]. The impacts of urban sprawl on ecological quality persist, and the impacts have considerable variability and complexity [12,13,14]. At the same time, the problem of climate change caused by large amounts of carbon emissions is extremely serious [15], and energy consumption is the main “culprit” affecting carbon emissions. The consumption of basic resources such as energy is exacerbated during rapid urbanization [16], and cities have become an important carbon source for carbon emissions [17]. The Chinese government pays more attention to the dynamic change in total energy consumption and makes a solemn commitment to energy conservation and emission reduction, and reducing energy consumption is a must for sustainable development [18]. Electricity is an important energy source to promote the sustainable development of China’s economy, and it has a fundamental role in the national economy [19]. However, as China’s electricity consumption continues to grow, there has not been an effective shift in the structure of electricity production, which is dominated by “thermal power” [20]. This type of power generation, which is based on the consumption of coal, a disposable energy source, generates large amounts of carbon emissions during the production process [21]. China faces energy constraints from urbanization [22] and carbon reduction targets [23], and there is an urgent need to clarify the ways in which the complex urbanization system affects electricity consumption [24]. Meanwhile, China is the world’s second-largest economy and the world’s largest CO₂ emitter [25], and rationally promoting urbanization and reducing energy consumption are important elements of the visionary blueprint for socialist modernization.

The issue of urban land use has also attracted widespread academic attention. However, most of the existing studies have focused on examining the impact of urban sprawl on economic performance and haze pollution, while less consideration has been given to its impact on energy consumption. How to promote urban green and low-carbon construction remains a key component of China’s sustained high-quality development. In view of this, in the context of rapid urbanization, we try to elaborate on the impact and mechanism of urban sprawl on electricity consumption. Firstly, the urban sprawl index is measured with 283 prefecture-level and above cities in China from 2008 to 2019. Meanwhile, the mechanism of urban sprawl affecting electricity consumption is explained at the theoretical level. Secondly, a series of robustness tests, such as instrumental variables, are used to identify the impact of urban sprawl on electricity consumption as accurately as possible and unfold the heterogeneity. Finally, the spatial econometric model is utilized to expand the analysis of the spatial spillover effects of urban sprawl and electricity consumption. And, in order to provide intellectual support for urban spatial layout as well as green and low-carbon development.

The marginal contribution of this paper may be reflected in three aspects. Firstly, in terms of research perspective, it combines urban sprawl with electricity consumption. It also examines the intrinsic mechanism of urban sprawl affecting electricity consumption from land resources, private transportation, and green technology innovation. It expands the study of the environmental effects of urban sprawl in the context of urbanization and construction in China. Secondly, the impact of urban sprawl on electricity consumption is examined in detail. Based on the spatial structure of cities, the strength of environmental regulation, the level of technological innovation, and the geographic location, this paper further analyzes the differentiated impacts of urban sprawl on electricity consumption so as to put forward more targeted recommendations. This paper analyzes the spillover effects of urban sprawl and electricity consumption from the perspective of spatial correlation. It makes up for the lack of spatial effects in the current research perspective and further enriches the research in the field of urban sprawl.

The rest of the study is structured as follows: Section 2 is a literature review. Section 3 attempts to elucidate the mechanism of action of urban sprawl in affecting electricity consumption. Section 4 is the model setting and variable descriptions. Section 5 is the empirical results, including the benchmark regression, robustness test, endogeneity problem test, mechanism of action, heterogeneity test, and extended analytical analysis. Section 6 is the discussion and analysis. Section 7 is the conclusions of research and policy recommendations.

2. Literature Review

The research in this paper deals with urban sprawl and electricity consumption. In order to organize the research progress in these two areas in recent years, the literature closely related to the research topic of this paper is divided into the following three categories for review.

2.1. Connotation and Measurement of Urban Sprawl

Early on, the concept of “urban sprawl” was characterized by the continuous expansion of urban boundaries and a shift to urban land at the expense of rural land [26]. In contrast to “compact” cities, urban sprawl tends to be a haphazard and decentralized, single-use and inefficient pattern of urban spatial expansion [27]. Urban sprawl refers to the expansion of land beyond the needs of population growth, the gradual expansion of urban economic activities to the periphery of the city, and the increasing decentralization of the urban form [28]. Although there is no specific and unified definition of urban sprawl, there is a consensus on its characteristics. And it is generally recognized that urban sprawl is characterized by inefficient, low-density, and “frog-hopping” expansion [29,30,31]. Conceptual uncertainty has led to a diversity of approaches to measuring urban sprawl. Urban sprawl measurement is categorized into single-indicator and multi-indicator methods. Among the single indicators, employment density [32,33], residential density [34], and population density [35] are used to assess the extent of urban sprawl based on the “low-density” nature of urban sprawl. Unlike the above literature, the urban sprawl index is also derived by comparing the growth rate of urban areas with the growth rate of the urban population [36,37]. On this basis, the degree of urban sprawl is assessed from the dimensions of urban land expansion and urban population expansion with the help of lighting data [38,39,40], thereby further broadening the scope of the study. In addition, unlike single-indicator measures, urban sprawl measures have been extended to multi-indicator measures [41]. Based on the characteristics of urban sprawl, density, utilization mix, and centrality are usually important dimensions of multi-indicator measures of urban sprawl [42,43]. Frenkel and Ashkenazi [44] used factor analysis to measure the three dimensions of density, scatter, and land use mix with a total of 13 indicators to finally obtain a composite index for assessing urban sprawl. Taking into account the local context and available data, the dimensions of urban sprawl, urban compactness, and urban form have also been used to measure the degree of urban sprawl in Shanghai [45].

2.2. Study on the Effects of Urban Sprawl

Much of the research on the effects of urban sprawl has focused on assessing the socioeconomic and environmental effects of urban sprawl. In terms of socioeconomic effects, urban sprawl negatively affects the frequency of community interactions and class upward mobility [46,47,48]. In addition, there is a threshold effect of urban sprawl on economic development, with appropriate urban sprawl promoting economic development and excessive urban sprawl inhibiting economic development [49]. Regarding the environmental effects of urban sprawl, existing views generally agree that urban sprawl reduces the quality of the ecological environment [50,51] and significantly and negatively affects urban green total factor productivity [52]. Urban sprawl exacerbates PM_2.5 emissions [53], and fiscal decentralization also significantly strengthens the contribution of urban sprawl to PM_2.5 [54]. The expansion of urban space and the surge in urban population both have a negative impact on CO₂ [55]. And urban sprawl mainly increases carbon emissions from transportation, construction, and industrial sectors, which in turn raises the total urban carbon emissions [56]. Similarly, urban sprawl increases urban air pollutant concentrations through increased energy consumption and higher industrial production [57,58]. However, there are differences in the impacts of different dimensions of urban morphology on air quality, and while the urban scale has a significant impact on air quality, it is urban fragmentation that is the most important factor contributing to the deterioration of urban air quality [59]. In addition to affecting air quality, urbanization with a lower percentage of developed land use per capita tends to have higher concentrations of water pollution in watersheds [60]. Overall, urban sprawl has had a negative impact on sustainable urban development [61]. Unlike the above literature, another viewpoint claims that urban sprawl can optimize land resource use efficiency and positively affect regional ecology and that low sprawl can reduce per capita NOx and PM_2.5 emissions from roads [62].

2.3. Study of the Factors Influencing Electricity Consumption

In a study on the influencing factors of electricity consumption, scholars discuss the driving factors of electricity consumption at the micro level and macro level. At the micro-individual level, personal moral constraints and positive expected emotions are important factors influencing residents’ electricity consumption behavior [63]. Habits are more important factors influencing residents’ electricity consumption behavior, and residents with electricity-saving habits use an average of 15.54 kWh more electricity per month compared to residents without electricity-saving habits [64]. Household appliances and household size also contribute to household electricity consumption, and the positive effect is greater for urban households [65]. In addition, household income also positively affects electricity use, and electricity use behavior varies across age groups [66]. At the macro level, the level of financial development [67,68] and the digital economy [69,70] both affect electricity consumption. Electricity price is the most direct factor affecting the behavior of residential electricity consumption, and an increase in electricity price can reduce residential electricity consumption [71], and a marginal electricity price increase of about 40% leads to a decrease in electricity consumption of about 35% [72]. Electricity consumption changes in response to climate change, with higher cooling demand in the summer and higher heating demand in the winter leading to an increase in electricity consumption [73], and total electricity consumption is more sensitive to warming than residential electricity consumption [74]. The urban form is also one of the important factors affecting electricity consumption, with neighborhood density negatively correlating with single-family residential summer electricity consumption [75]. The urban built-up area of Chongqing is the center of urban electricity consumption, and the electricity consumption in the urban built-up area accounts for 34.34–45.69% of the electricity consumption in the urban administrative area [76]. The urbanization process promotes residential electricity consumption, which is heterogeneous at different stages of urbanization, and has a more pronounced impact on rural residential electricity consumption compared to urban residential electricity consumption [22].

By combing through the literature, the current results on urban sprawl and electricity consumption are relatively rich, laying the foundation for further research in this paper. However, taking a comprehensive view, there is still some room for improvement in the research of the existing literature. In terms of research content, most studies focus on assessing the impact of urban sprawl on the ecological environment, PM_2.5, etc., and less on exploring the impact of urban sprawl on electricity consumption, which is crucial for urbanization and low-carbon development. Therefore, this paper focuses on the relationship between urban sprawl and electricity consumption and its mechanism. In addition, in terms of analyzing the differences in impacts, existing literature mostly analyzes the impacts from the perspective of city size, etc., while this paper will meticulously examine the differences in the impacts of urban sprawl on electricity consumption from the perspective of urban spatial structure, the level of environmental regulation, the level of technological innovation, and the geographic location of the city. Moreover, the related literature focuses on assessing the linear or nonlinear relationship of urban sprawl on environmental pollution. Additionally, this paper will further explore the spatial spillover effects of urban sprawl and electricity consumption from the perspective of spatial correlation.

3. Research Hypothesis

3.1. The Impact of Urban Sprawl on Electricity Consumption

The impact of urban sprawl on electricity consumption can be categorized into two dimensions: production and living. Firstly, urban sprawl affects the way land resources are allocated, bringing about a shift in production activities and an adjustment of the industrial structure, while there are obvious differences in the electricity demand of different industries. At the present stage, government intervention and determination of spatial allocation of land resources are still relatively common. In pursuit of economic construction, local governments restrictively provide residential land, provide more industrial land, and create industrial parks to track economic growth targets. The allocation of land resources in favor of large-scale industrial land grants has led to the relocation of energy-consuming and low-end industries, restricting the development of the service sector and the supply of service products, and hindering the improvement of energy resource efficiency. Urban sprawl is inherently characterized by inefficient land resource allocation, and crude production patterns contribute to the increase in industrial electricity consumption. Secondly, urban sprawl leads to low-density expansion, resulting in urban spatial disorder, which makes residents more demanding in terms of mobility and so on. Urban sprawl increases residents’ transportation demand, and the supply of public transportation is difficult to meet residents’ emergent requirements in a short period of time, leading to a rapid increase in private car ownership. Urban sprawl increases the number of cars in the city and also increases electricity consumption. Finally, excessive urban sprawl is not conducive to the formation of agglomeration effects and inhibits green technological innovation. Technological innovation has obvious spatial agglomeration characteristics. Sharing, matching, and learning mechanisms are important for the diffusion and spillover of innovative knowledge [77]. The urban spatial structure with over-expansion of land weakens knowledge learning and technology spillover among firms, leading to inertia in industrial green transformation and thus adversely affecting energy-efficient production. In summary, the hypothesis was proposed as follows. The theoretical mechanism analysis diagram is shown in Figure 1.

Hypothesis 1.

Urban sprawl can exacerbate electricity consumption.

3.2. Urban Sprawl, Land Resource Allocation, and Electricity Consumption

Land resources in the process of urban sprawl are allocated to land developers and bidders who can bring the greatest profit returns, leading to changes in the scale and efficiency of urban industries and affecting electricity consumption. First, local governments may enhance land supply incentives and intensify urban sprawl under the influence of multiple factors such as economic growth patterns, fiscal and land systems, and performance appraisals. By controlling the primary market of land, local governments cheaply use sprawl land for industrial use on a large scale in a variety of ways such as agreement transfer [78]. They lead the development of the industry and accelerate the process of industrialization in order to achieve rapid economic growth in the short term and increase employment, thus forming a “land for development” economic growth model [79]. Therefore, the mismatch of land resources centered on the large-scale granting of industrial land and the unsaturated supply of commercial and residential land has led to the rapid development of energy-consuming and low-end industries, which is not conducive to the transformation of the industrial structure. In addition, the formation of a crude economic growth pattern may eventually lead to increasing electricity intensity. Second, the mismatch of land resources brought about by urban sprawl leads to the inertia of the industrial system to reduce power intensity. Local governments have, to a certain extent, acquiesced in the negative impact on the environment brought about by the introduction of energy-intensive industries in their development, which has contributed to their inertia in improving the efficiency of energy utilization. This has led to high industrial electricity consumption in some areas. Moreover, the lack of pressure and incentives for industrial firms that have acquired cheap land to reduce industrial electricity consumption may also create negative incentives to reduce electricity consumption. On the basis of this, the following hypothesis is proposed:

Hypothesis 2.

Urban sprawl exacerbates electricity consumption by altering land resource allocation.

3.3. Urban Sprawl, Private Transportation, and Electricity Consumption

Urban sprawl changes the accessibility of urban space, affecting the lives of residents and the way they travel on a daily basis. Urban sprawl leads to spatial decentralization of the population and separation of jobs and housing. As the land area in the central city is limited, the “sprawl” development model drives real estate development in the suburbs, making central city housing prices much higher than those in the suburbs. As a result, many residents who work in the central city choose to live in the suburbs, which are relatively remote and inexpensive, creating a spatial mismatch between their place of residence and their place of work. Under the living pattern of work–life separation, commuting distance and commuting time increase, and people are highly dependent on transportation. Transportation is also an important area of energy consumption, and the increasing number of road vehicles increases electricity consumption. On the one hand, private transportation involves the construction and maintenance of infrastructure such as roads, and this process inevitably increases electricity consumption. On the other hand, in the short term, public transportation cannot match the expanding and dispersed urban space. And due to the accessibility of private cars in time and space, residents’ dependence on private cars for travel is becoming stronger. Urban sprawl affects the mode of transportation used by residents for their daily trips and increases their dependence on private transportation. And, the increase in private automobiles on the consumer side invariably increases electricity consumption on the production side of the vehicle. With the implementation of new energy vehicle development strategies, private transportation might evolve into “driving with electricity”, further increasing electricity consumption [80]. Based on this, the following hypothesis is proposed:

Hypothesis 3.

Urban sprawl exacerbates electricity consumption by increasing reliance on private transportation.

3.4. Urban Sprawl, Green Technology Innovation, and Electricity Consumption

Urban sprawl has led to a gradual diffusion of economic activities concentrated in urban centers to the surrounding areas. Compared with peri-urban areas, urban centers have external economic and spillover effects such as specialized labor and capital, which give them an advantage in the knowledge economy [81,82]. Urban sprawl promotes the relocation of industries to peri-urban areas, impedes the flow and matching of factors such as knowledge, and weakens the agglomeration effect of industries. However, industrial agglomeration is an environment for technological innovation [83]. On the one hand, industrial agglomeration promotes inter-firm competition and incentivizes technological transformation, upgrading and renewal [84]. On the other hand, industrial agglomeration facilitates complementarities among firms, which is conducive to the enhancement of regional innovation capacity [85]. Compared with other activities, technological innovation is more dependent on spillover effects and has spatial externalities. Urban sprawl lengthens the spatial distance within cities, which is not conducive to the diffusion and sharing of knowledge, technological information and human capital, and inhibits green technological innovation. Green technology innovation follows the principle of ecological economy and is an important means to improve energy utilization efficiency. Green technological innovation can be used in the production side of the industry to promote clean production, enhance energy saving, and reduce industrial electricity consumption. At the same time, green technology innovation can be used in the field of new energy by reduce the use of traditional fossil energy, helping to promote the transformation of the energy structure, and reducing electricity consumption [86]. In general, urban sprawl is not conducive to knowledge diffusion and technological innovation, which creates great resistance to the green transformation of industrial enterprises and becomes a shackle to reduce electricity consumption. Accordingly, the following hypothesis is proposed:

Hypothesis 4.

Urban sprawl exacerbates electricity consumption by inhibiting green technology innovation.

4. Materials and Methods

4.1. Model Setting

4.1.1. Baseline Regression Modeling

Based on the mechanism analysis, in order to test the impact of urban sprawl on electricity consumption, the following benchmark model is constructed. The specific form of the setup is shown in Equation (1).

$${\mathrm{lnElec}}_{\mathrm{it}}=\mathsf{\alpha }+\mathsf{\beta }{\mathrm{Sprwal}}_{\mathrm{it}}+\mathsf{\delta }{\mathrm{X}}_{\mathrm{it}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\gamma }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(1)

Where $\mathrm{i}$ denotes city and $\mathrm{t}$ denotes year. ${\mathrm{lnElec}}_{\mathrm{it}}$ denotes the logarithmic value of total electricity consumption of the $\mathrm{i}$ city in $\mathrm{t}$ year, ${\mathrm{Sprwal}}_{\mathrm{it}}$ denotes urban sprawl index of the $\mathrm{i}$ city in $\mathrm{t}$ year, ${\mathrm{X}}_{\mathrm{it}}$ is a control variable that may affect electricity consumption, ${\mathsf{\mu }}_{\mathrm{i}}$ is city fixed effects, ${\mathsf{\gamma }}_{\mathrm{t}}$ is the time fixed effects, and ${\mathsf{\epsilon }}_{\mathrm{it}}$ is random error terms. The main coefficient of interest in this paper is the $\mathsf{\beta }$, whose meaning is the effect of urban sprawl on electricity consumption after controlling for regional characteristics.

4.1.2. Spatial Correlation Analysis

The global spatial autocorrelation index (Global Moran’s I) was used to determine whether urban sprawl and electricity consumption are spatially correlated between regions, as shown in Equation (2).

$$\mathrm{Global} \mathrm{Moran}\prime \mathrm{s} \mathrm{I}=\frac{\mathrm{n}\sum _{\mathrm{i}=1}^{\mathrm{n}}\sum _{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{W}}_{\mathrm{ij}}({\mathrm{x}}_{\mathrm{i}}-\stackrel{\mathrm{-}}{\mathrm{x}})({\mathrm{x}}_{\mathrm{j}}-\stackrel{\mathrm{-}}{\mathrm{x}})}{\sum _{\mathrm{i}=1}^{\mathrm{n}}\sum _{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{W}}_{\mathrm{ij}}\sum _{\mathrm{i}=1}^{\mathrm{n}}{({\mathrm{x}}_{\mathrm{i}}-\stackrel{\mathrm{-}}{\mathrm{x}})}^2}$$

(2)

Where $\mathrm{n}$ is the sample size, ${\mathrm{x}}_{\mathrm{i}}$ and ${\mathrm{x}}_{\mathrm{j}}$ are the urban sprawl index (electricity consumption) in $\mathrm{i}$ city and $\mathrm{j}$ city, respectively, and $\stackrel{\mathrm{-}}{\mathrm{x}}$ is the mean value. Global Moran’s I is in the range of [−1, 1]. When Global Moran’s I > 0, it means the spatial distribution is positively correlated. Conversely, when Global Moran’s I < 0, it means that the spatial distribution is negatively correlated. And, when Global Moran’s I = 0, it means that there is no spatial correlation [87]. ${\mathrm{W}}_{\mathrm{ij}}$ is the inverse distance spatial weight matrix.

The Local Moran’s I index scatter plot can well characterize the spatial agglomeration of urban sprawl (electricity consumption), and the Local Moran’s I measurement is shown in Equation (3).

$$\mathrm{Local} \mathrm{Moran}\prime \mathrm{s} \mathrm{I}=\frac{\mathrm{n}({\mathrm{x}}_{\mathrm{i}}-\stackrel{\mathrm{-}}{\mathrm{x}})\sum _{\mathrm{j}=1,\mathrm{j}\ne \mathrm{i}}^{\mathrm{n}}{\mathrm{W}}_{\mathrm{ij}}({\mathrm{x}}_{\mathrm{j}}-\stackrel{\mathrm{-}}{\mathrm{x}})}{\sum _{\mathrm{i}=1}^{\mathrm{n}}{({\mathrm{x}}_{\mathrm{i}}-\stackrel{\mathrm{-}}{\mathrm{x}})}^2}$$

(3)

the variables are defined as shown in Equation (2).

4.1.3. Spatial Econometric Models

Compared with traditional regression methods, spatial econometric models consider the spatial dependence and spatial correlation of samples. And common spatial econometric models are classified as spatial autoregressive model (SAR), spatial error model (SEM), and spatial Durbin model (SDM), and the model settings are shown in Equations (4)–(6), respectively [88].

$${\mathrm{lnElec}}_{\mathrm{it}}=\mathsf{\rho }{\mathrm{W}}_{\mathrm{ij}}{\mathrm{lnElec}}_{\mathrm{it}}+{\mathsf{\beta }}_1{\mathrm{Sprwal}}_{\mathrm{it}}+{\mathsf{\beta }}_2{\mathrm{X}}_{\mathrm{it}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\eta }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(4)

$${\mathrm{lnElec}}_{\mathrm{it}}={\mathsf{\beta }}_1{\mathrm{Sprwal}}_{\mathrm{it}}+{\mathsf{\beta }}_2{\mathrm{X}}_{\mathrm{it}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\eta }}_{\mathrm{t}}+{\mathsf{\phi }}_{\mathrm{it}} ,{ \mathsf{\phi }}_{\mathrm{it}}={\mathsf{\lambda }{\mathrm{W}}_{\mathrm{ij}}\mathsf{\phi }}_{\mathrm{it}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(5)

$${\mathrm{lnElec}}_{\mathrm{it}}=\mathsf{\rho }{\mathrm{W}}_{\mathrm{i}\mathrm{j}}{\mathrm{lnElec}}_{\mathrm{it}}+{\mathsf{\beta }}_1{\mathrm{Sprwal}}_{\mathrm{it}}+{\mathsf{\beta }}_2{\mathrm{X}}_{\mathrm{it}}+{\mathsf{\beta }}_3{\mathrm{W}}_{\mathrm{i}\mathrm{j}}{\mathrm{Sprwal}}_{\mathrm{it}}+{\mathsf{\beta }}_4{{\mathrm{W}}_{\mathrm{i}\mathrm{j}}\mathrm{X}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\eta }}_{\mathrm{t}}+{\mathsf{\phi }}_{\mathrm{i}\mathrm{t}}$$

(6)

Where $\mathrm{i}$ denotes city and $\mathrm{t}$ denotes year. ${\mathrm{lnElec}}_{\mathrm{it}}$ denotes the logarithmic value of total electricity consumption of the $\mathrm{i}$ city in $\mathrm{t}$ year, ${\mathrm{Sprwal}}_{\mathrm{it}}$ denotes the urban sprawl index of the $\mathrm{i}$ city in $\mathrm{t}$ year, ${\mathrm{X}}_{\mathrm{it}}$ is a control variable that may affect electricity consumption, ${\mathsf{\mu }}_{\mathrm{I}}$ is city fixed effects, ${\mathsf{\gamma }}_{\mathrm{t}}$ is the time fixed effects, and ${\mathsf{\epsilon }}_{\mathrm{it}}$ is random error terms. ${\mathrm{W}}_{\mathrm{ij}}$ is the inverse distance spatial weight matrix. ${\mathsf{\beta }}_3$, $\mathsf{\rho }$, and ${\mathsf{\beta }}_4$ are the spatial lag terms for urban sprawl, electricity consumption, and other control variables, respectively.

4.2. Variable Measurement and Descriptive Analysis

4.2.1. Independent Variable

The urban sprawl index is the independent variable of this paper. Currently, there are generally single-indicator measures and multiple-indicator measures for urban sprawl. The multiple-indicator measures weaken the essential feature of excessive spatial growth within urban sprawl. And, comparing with the data and technical processing of multiple indicators, the single-indicator measure is more suitable for the analysis of econometric models [31]. However, measuring urban sprawl using only methods such as the share of construction land is too unitary. For this reason, based on existing studies [46,89] and considering land expansion and population elements, this paper utilizes the ratio of the growth rate of the built-up area of a city to the growth rate of the population in the urban area to portray urban sprawl characteristics. The calculation method is shown in Equation (7).

$${\mathrm{Sprwal}}_{\mathrm{it}}=\frac{({\mathrm{LUR}}_{\mathrm{it}}/{\mathrm{LUR}}_{\mathrm{i}0})}{({\mathrm{POP}}_{\mathrm{it}}/{\mathrm{POP}}_{\mathrm{i}0})}$$

(7)

In the formula, the ${\mathrm{LUR}}_{\mathrm{it}}$ is the built-up area of the $\mathrm{i}$ city $\mathrm{t}$ year, and ${\mathrm{LUR}}_{\mathrm{i}0}$ is the built-up area of the base period. The ${\mathrm{POP}}_{\mathrm{it}}$ is the population of urban area of the $\mathrm{i}$ city $\mathrm{t}$ year, and ${\mathrm{POP}}_{\mathrm{i}0}$ is the population of the base period. The base period was chosen to be 2008. When the $\mathrm{Sprwal}>1$, it means that the growth rate of the built-up area of the city is greater than the growth rate of the population in the urban area. This indicates that the phenomenon of urban sprawl, and the more the value deviates from 1 indicates that the higher the degree of urban sprawl. When the $0<\mathrm{Sprwal}\ll 1$, it means that the growth rate of the built-up area of the city is smaller than the growth rate of the population in the urban area, and there is no urban sprawl phenomenon.

4.2.2. Dependent Variable

The dependent variable in this paper is total urban electricity consumption, which is logarithmized. With regard to city electricity consumption data, the caliber of city-level statistics changed in 2017. For this reason, the urban electricity consumption data used in this paper comes from the results of lighting data [90]. This data is a 1 km × 1 km resolution electricity consumption raster data. This paper further utilizes ArcGIS to parse this gridded data into panel data of urban electricity consumption in China.

4.2.3. Control Variables

In addition to the above core explanatory variables, this paper also selects a series of control variables to control other factors that may affect urban electricity consumption. Based on the findings of the existing literature, six indicators, namely, the level of economic development, the degree of government intervention, the level of educational input, industrial structure, the level of opening up to the outside world, and the level of urbanization, are selected as the control variables of the model. The specific descriptions are as follows.

(1) Level of economic development (Lnpgdp). Economic growth is usually an important factor contributing to the growth of electricity consumption [91,92]. And in this case, the logarithm of real GDP per capita is used as a proxy variable for the level of economic development.

(2) Degree of government intervention (Govern). The government may financially invest in environmental protection programs and intervene in environmental quality objectives, thereby negatively affecting electricity consumption [93]. In contrast, inappropriate interventions can cause resource allocation distortions and efficiency losses, thereby increasing electricity consumption. This paper uses the share of fiscal expenditure in GDP for measurement.

(3) Level of educational input (Educate). Educational input is conducive to improving the education level and energy-saving concepts of residents, affecting their energy-saving behavior, which in turn has a negative impact on electricity consumption. In this paper, it is expressed as the proportion of regional education expenditure to financial expenditure.

(4) Industrial structure (Instru). Generally speaking, a larger proportion of secondary industry greatly increases urban electricity consumption. In this paper, the ratio of the output value of the tertiary industry to the output value of the secondary industry is used to characterize the transformation of the advanced industrial structure [94].

(5) Opening-up level (Openne). Increasing the level of openness to the outside world may generate technological spillover effects and improve the efficiency of energy use. But, it may also affect the local industrial structure, such as increasing the proportion of local polluting industries and increasing electricity consumption [95]. In this paper, we characterize the opening-up level by the share of actual FDI in GDP.

(6) Level of urbanization (Urbani). With the increase in the urbanization rate of the population, lifestyle changes have prompted greater demand for basic domestic energy consumption, such as household appliances and private cars, and population urbanization has become an important driver of the increase in electricity consumption [76]. In this paper, the level of urbanization is expressed as the share of the non-farm population in the total population.

4.2.4. Mechanism Variables

According to the relevant theoretical analysis and existing literature, this paper selects land resource allocation and private transportation dependence as mechanism analysis variables. (1) Land resource allocation (Landra): The spatial expansion brought by urban sprawl may be more reflected in the large-scale use of industrial land. The expansion of industrial land will inevitably squeeze commercial and residential land and cultivated forest land and increase electricity consumption. For this reason, this paper chooses the ratio of industrial land to urban construction land to measure the mismatch of land resources. Among them, the area of industrial land is the sum of the area of industrial land and storage land [96,97]. (2) Private transportation dependence (Ptrans): Urban sprawl leads to a spatial decentralization of the population and a separation of jobs and residences. Under the living pattern of job–residence separation, due to the increase in commuting distance and commuting time, residents’ dependence on motor vehicles for travelling becomes stronger, thereby increasing electricity consumption. For this reason, this paper chooses the logarithmic value of private car ownership to characterize the degree of private transportation dependence [98]. (3) Green technology innovation (Pgpatt): Urban sprawl lengthens the spatial distance within cities and weakens the agglomeration effect. It is not conducive to the diffusion and sharing of knowledge, technical information, and human capital, which inhibit green technological innovation and thus become a shackle to reduce electricity consumption. Therefore, this paper chooses the per capita green patent acquisition to characterize green technology innovation [99,100].

4.3. Data Source

The sample data in this paper is a panel dataset based on Chinese cities at the prefecture level and above, covering 283 cities (due to missing data, cities with more missing values such as Bijie and Tongren are excluded, and Tibet, Hong Kong, Macao, and Taiwan are not involved). Due to the availability of data, the sample data spans from 2008–2019. Built-up area, population of urban area, urban construction land area, and industrial and warehousing land area were obtained from the “China Urban Construction Statistical Yearbook”. The total urban electricity consumption was obtained from lighting data. The control variables were obtained from the “China City Statistical Yearbook”. The data on private car ownership were obtained from the CEIC database (https://www.ceicdata.com/zh-hans (accessed on 1 June 2023)), and the number of green patents acquired by each city was derived from the China Innovation Patent Research Database (CIRD) of the China Research Data Service Platform (CNRDS). Among them, the real GDP was deflated to eliminate the price effect with 2008 as the base period, and the missing data were filled in using the interpolation method. The variable selection and descriptive statistics are shown in

Table 1. Descriptive description of variables.

Variables	Obs	Mean	SD	P25	P75
lnElec	3396	22.9006	0.8026	22.3090	23.4516
Sprwal	3396	1.2060	0.2941	1.0022	1.3188
lnPgdp	3396	10.4790	0.6275	10.0673	10.9088
Govern	3396	0.1879	0.0937	0.1248	0.2233
Educat	3396	0.1800	0.0416	0.1513	0.2067
Instru	3396	0.9295	0.5152	0.6230	1.0689
Openne	3396	0.0173	0.0174	0.0041	0.0246
Urbani	3396	52.6688	15.3768	41.4365	62.3275

.

5. Results

5.1. Trends in Spatial-Temporal Evolution of Urban Sprawl and Electricity Consumption

In order to visualize the spatial and temporal evolution trend of urban sprawl and electricity consumption, this paper uses ArcGIS to map the spatial distribution of urban sprawl and electricity consumption in 2009 and 2019, respectively. As shown in Figure 2 and Figure 3.

In terms of trends in temporal evolution, the level of urban sprawl increases significantly over time during the period 2009–2019. A small number of cities were in the not-yet-spreading stage, 63.60% of the cities had a sprawl index greater than 1, and the growth rate of the built-up area of cities was smaller than the growth rate of the urban population in 2009. Up until 2019, the majority of Chinese cities showed the urban sprawl phenomenon and were dominated by low sprawl types, 89.40% of cities had a sprawl index greater than 1, and the sprawl indexes were in the range of 1.00–2.00. A small number of cities have medium sprawl status. No city reached a high state of sprawl, i.e., an urban sprawl index of 3.00 and above. The proportion of cities with different types of urban sprawl is not balanced, and the spatial distribution is relatively decentralized. The overall situation is characterized by “sporadic high-spread”. On a smaller administrative scale, a significant increase in the degree of sprawl could be found in some cities such as Erdos, Ya’an, and Anshun during the study period. Overall, urban sprawl is increasing, but it is not out of control in China. This may be influenced by the price of land urban construction gradually migrating from the central area to the periphery. Moreover, investment in land development is an important source of funding for renovation and new town development. A combination of factors has led to an increase in overall urban sprawl. At the same time, however, urban sprawl is not yet uncontrollable, due to China’s two-way control of the usage and quantity of land for urban construction.

In terms of general evolutionary trends, during the study period, the total electricity consumption of cities has been increasing year by year. The electricity consumption of most cities was less than 20 billion Kw·h in 2009, and urban electricity consumption increased significantly in 2019. China’s average urban electricity consumption grew from 9.89 billion Kw·h to 14.17 billion Kw·h from 2009 to 2019, with an overall growth rate of 43.30% and an average annual growth rate of 4.33%. Cities with high electricity consumption levels are mostly concentrated in the eastern region, such as Beijing and Shanghai, and are clearly and progressively consuming more than 80 billion Kw·h of electricity. And individual cities in the western region are also increasing their electricity consumption, but overall urban electricity consumption is less than 60 billion Kw·h. The reason for this phenomenon may be that electricity is also an essential production factor in industrial production and is vulnerable to economic conditions and other factors. And industrial electricity consumption occupies a large proportion of the city’s total electricity consumption. The eastern coastal cities are economically ahead of the inland cities, which makes the eastern coastal cities become the main force of electricity consumption.

5.2. Baseline Regression Results

Based on the econometric model constructed above, this paper examines the study of the effect of urban sprawl on electricity consumption, and the results are presented in

Table 2. Benchmark regression results.

Variables	(1) lnElec	(2) lnElec	(3) lnElec	(4) lnElec	(5) lnElec	(6) lnElec	(7) lnElec
Sprwal	0.0232 ***	0.0210 ***	0.0208 ***	0.0207 ***	0.0207 ***	0.0215 ***	0.0207 ***
	(0.0050)	(0.0046)	(0.0046)	(0.0046)	(0.0046)	(0.0045)	(0.0042)
lnPgdp		0.0917 ***	0.0903 ***	0.0880 ***	0.0841 ***	0.0791 ***	0.0731 ***
		(0.0145)	(0.0146)	(0.0146)	(0.0145)	(0.0144)	(0.0147)
Govern			−0.0396 *	−0.0570 ***	−0.0362	−0.0435 *	−0.0404 *
			(0.0205)	(0.0211)	(0.0230)	(0.0230)	(0.0227)
Educat				−0.0619	−0.0472	−0.0454	−0.0441
				(0.0501)	(0.0509)	(0.0499)	(0.0496)
Instru					−0.0094 **	−0.0085 **	−0.0077 **
					(0.0037)	(0.0037)	(0.0036)
Openne						0.2538 ***	0.2119 **
						(0.0926)	(0.0948)
Urbani							0.0008 **
							(0.0003)
_cons	22.5999 ***	21.6913 ***	21.7111 ***	21.7485 ***	21.7884 ***	21.8325 ***	21.8570 ***
	(0.0059)	(0.1441)	(0.1459)	(0.1460)	(0.1455)	(0.1435)	(0.1443)
Year	Yes	Yes	Yes	Yes	Yes	Yes	Yes
City	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	3396	3396	3396	3396	3396	3396	3396
R2	0.9730	0.9756	0.9757	0.9757	0.9759	0.9761	0.9763

Note: *, **, and *** denote 10%, 5%, and 1% significance levels, respectively. Values in parentheses are standard errors.

. Column (1) of Table 2 shows the estimation results with only time and individual fixed effects and without the inclusion of control variables, and columns (2) to (7) gradually include city-level control variables. The regression results show that the estimated coefficient of the effect of urban sprawl on electricity consumption without the inclusion of control variables is 0.0232 and is significant at the 1% level. This result remains largely stable with the gradual addition of city-level control variables. The result of column (7), which includes the full set of control variables, shows that the sign of the regression coefficient for the core explanatory variable, urban sprawl, is significantly positive at the 1% level. The results of the baseline model estimation tentatively indicate that urban sprawl significantly increases urban electricity consumption, validating Hypothesis 1.

As far as the control variables are concerned, the level of economic development significantly increases the consumption of electricity. As the main energy substance in production and life, electricity is highly relevant to modern economic and social development. In the secondary industry-dominated economic model to promote economic development, the industry is still the most important sector of electricity consumption. Therefore, economic development is the most important factor to increase electricity consumption, which is similar to the finding of Song et al. [101]. The government is the main body of “emission reduction and energy saving”, and the government’s policies and behaviors will have an important impact on electricity consumption. At this stage, with the promotion of ecological civilization construction, local governments begin to pay attention to environmental protection, promulgate policies to guide the development of energy-saving and emission-reduction technologies, and force enterprises to green transformation and technological innovation, which has a negative impact on electricity consumption. An advanced industrial structure significantly reduces urban electricity consumption. Compared with the traditional manufacturing industry and other secondary industries, the tertiary industry’s power demand is smaller. Moreover, the upgrading of the industrial structure is essentially the transfer of resource elements from inefficient to efficient sectors, which is conducive to the improvement of urban energy efficiency and thus reduces electricity consumption, and the result is basically consistent with the finding of Guang et al. [102]. Openness to the outside world significantly increases electricity consumption. This may be due to the fact that the increase in the level of openness to the outside world is accompanied by a large inflow of FDI, which leads to the agglomeration of pollution-intensive industries and promotes electricity consumption. The level of urbanization has a significant positive relationship with electricity consumption, which means that the higher the level of regional urbanization, the higher the regional electricity consumption. Rural residents consume less electricity for living and travelling than urban residents. Therefore, the increase in the urbanization rate leads to an increase in electricity consumption.

5.3. Robustness and Endogeneity Tests

5.3.1. Replacing the Measurement of the Independent Variable

Urban sprawl is typically characterized by rapid land expansion leading to a decentralized distribution of population, which is reflected in low density and decentralization. Higher urban population densities may indicate a greater likelihood of agglomeration of economic activities and a lower tendency to urban sprawl. In contrast, in low population density areas, the spatial structure tends to be decentralized and the tendency of urban sprawl is higher [103]. This means that urban population density and the level of urban sprawl are roughly inversely quantitatively related [104]. Therefore, the logarithmic value of urban population density (lnDens) is used to replace the urban sprawl index in the benchmark regression as the independent variable. The results of the model estimation are shown in column (1) of

Table 3. Robustness tests: changing the measurement of variables and removing policy-related disturbances.

Variables	(1) lnElec	(2) lnElec	(3) lnElec	(4) lnElec	(5) lnElec
lnDens	−0.0278 ***
	(0.0065)
Sprwal		0.1969 ***	0.0200 ***	0.0238 ***	0.0203 ***
		(0.0641)	(0.0047)	(0.0066)	(0.0046)
_cons	21.8717 ***	5.4305 ***	21.8421 ***	21.8728 ***	21.8224 ***
	(0.1454)	(1.8667)	(0.1568)	(0.1935)	(0.1659)
Controls	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes
City	Yes	Yes	Yes	Yes	Yes
N	3396	3297	3060	2437	2863
R2	0.9762	0.7257	0.9744	0.9730	0.9737

Note: *** denote 1% significance levels. Values in parentheses are standard errors.

, where the coefficient of urban population density is significantly negative. This means that higher population densities can reduce electricity consumption. And, it indicates that urban sprawl promotes electricity consumption.

5.3.2. Replacing the Measurement of the Dependent Variable

Considering the possible bias of electricity consumption measured from lighting data, the regression is further conducted using the whole society’s electricity consumption data published in the statistical yearbook as the dependent variable [105]. The regression results, as shown in column (2) of Table 3, show that the coefficient of urban sprawl is significantly positive at the 1% level, again indicating that the conclusion of the baseline regression is robust.

5.3.3. Removing Interference from Relevant Policies

Another challenge to the regression results is that in verifying the impact of urban sprawl on electricity consumption, it may be interfered with by policies such as New Energy Demonstration Cities, Low-Carbon Pilot Cities, and Broadband China. In order to exclude the interference of the above policies and ensure the accuracy of the benchmark regression, the samples of policy implementation are excluded from the regression, and the results are shown in columns (3)–(5) of Table 3, respectively. Compared to the baseline results, the significance level of the urban sprawl coefficient does not change after taking into account the relevant policy disturbances. This also indicates that the conclusion that urban sprawl increases electricity consumption remains robust.

5.3.4. Excluding Macro-Systemic Differences

(i) Incorporating multidimensional interaction fixed effects. Urban sprawl has different development trends in different regions. Therefore, this paper controls for the unobservable effects at the provincial level over time by controlling for the joint “province-year” fixed effects. As shown in column (1) of

Table 4. Robustness tests: excluding macro-systematic differences and removing outliers.

Variables	(1) lnElec	(2) lnElec	(3) lnElec	(4) lnElec
Sprwal	0.0130 ***	0.0207 ***	0.0259 ***	0.0200 ***
	(0.0043)	(0.0045)	(0.0061)	(0.0044)
_cons	24.0708 ***	21.8570 ***	22.0697 ***	21.7217 ***
	(0.1924)	(0.2156)	(0.1807)	(0.1609)
Controls	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
City	Yes	Yes	Yes	Yes
Pro#Year	Yes	No	No	No
N	3396	3396	3396	3036
R2	0.9841	0.9763	0.9629	0.9751

Note: *** denote 1% significance levels. Values in parentheses are standard errors.

, the Sprawl coefficient is still positive at the 1% significance level. It remains consistent with the baseline regression results. (ii) Replacing the standard error clustering hierarchy. The baseline regression refers to the general practice of standard error clustering at the same level (city level) of the study object. However, this clustering approach ignores the fact that there are often strong correlations (e.g., infrastructure, environmental regulation intensity, etc.) among cities in the same province. We therefore re-clustered the standard errors to the province level and report the estimates as a robustness test in column (2) of Table 4. As can be seen from the results, there is no obvious change in the size or significance of the coefficients.

5.3.5. Removal of Outliers

(i) Excluding extreme value interference. In order to avoid possible outliers in the data from influencing the benchmark regression results, this paper applies a bilateral 1% shrinkage to all continuous variables. The results are shown in column (3) of Table 4, and the core explanatory variable urban sprawl remains consistent with the benchmark regression results. (ii) Excluding municipalities and provincial capital cities. On the one hand, municipalities directly under the central government and provincial capital cities may be affected by policy favoritism and may have a relatively large advantage in energy-saving technology and high-quality human capital compared with general prefecture-level cities. This makes the energy use efficiency of these cities stronger. On the other hand, the governments of municipalities and provincial capitals generally pay more attention to environmental issues. Tougher environmental constraints and more rational planning of the urban sprawl process may also lead firms to favor green technological innovation. Therefore, failure to exclude these factors may have some impact on the stability of our estimation results. The results in column (4) of Table 4 show that the results remain robust after excluding municipalities and provincial capitals.

5.3.6. Endogeneity Test

To mitigate the possible endogeneity problems of mutual causality and omission of unobservable variables in the baseline regression, this paper adopts the instrumental variable test to regress. Specifically, the relief degree of land surface is selected as an instrumental variable for urban sprawl [98]. The greater the relief degree of land surface, the more segregated it presents geographic features and the more dispersed population distribution. This implies that the possibility of population density and industrial agglomeration is lower. And the urban spatial pattern tends to develop in a more disorderly and decentralized spreading pattern. Therefore, the relief degree of land surface theoretically positively affects urban sprawl, and the relief degree of land surface does not directly affect urban power consumption. It meets the requirement of exogeneity. However, since the relief degree of land surface is a non-temporal variable, it is further constructed a cross-multiplier by the relief degree of land surface and lagged period of the urban sprawl index. It is ultimately used as an instrumental variable indicator (Tsrdls) of urban sprawl [106]. From the regression results in column (1) of

Table 5. Instrumental variables test.

Variables	First Stage	Second Stage
	(1) Sprwal	(2) lnElec
Tsrdls	0.3923 ***
	(0.0411)
Sprwal		0.0217 **
		(0.0093)
Controls	Yes	Yes
Year	Yes	Yes
City	Yes	Yes
Observations	3113	3113
Cragg–Donald Wald F statistic	724.609
Kleibergen–Paap Wald rk F statistic	91.278
Stock–Yogo weak ID test critical values (10%)	16.38

Note: ** and *** denote 5%, and 1% significance levels, respectively. Values in parentheses are standard errors.

in the first stage, it can be seen that there is a significant positive correlation between the instrumental variable (Tsrdls) and urban sprawl (Sprwal). Also, the Cragg–Donald Wald F-statistic and Kleibergen–Paap Wald rk F-statistic, are both greater than the Stock–Yogo critical value of 16.38 under the Stock–Yogo weak ID test critical values (10%), thus ruling out the possibility of a weak instrumental variable. The regression results from the second stage column (2) show that urban sprawl significantly exacerbates urban electricity consumption. The above analysis shows that the baseline regression conclusions of this paper still hold after considering potential endogeneity issues using the instrumental variables approach.

5.4. Mechanism Testing

Based on the theoretical analysis, this paper further identifies and tests the mechanism by which urban sprawl affects urban electricity consumption through land resource allocation, private transportation and green technological innovation. The specific estimation results are shown in

Table 6. Mechanism test results.

Variables	(1) Landra	(2) Landra	(3) Ptrans	(4) Ptrans	(5) Pgpatt	(6) Pgpatt
Sprwal	0.0200 ***	0.0206 ***	0.0887 **	0.0869 **	−0.3316 *	−0.2816 **
	(0.0076)	(0.0074)	(0.0436)	(0.0422)	(0.1748)	(0.1429)
_cons	0.2193 ***	0.2829	4.3972 ***	2.0753	0.4735 ***	27.5717 ***
	(0.0083)	(0.2131)	(0.0475)	(1.3583)	(0.1493)	(6.0903)
Controls	No	Yes	No	Yes	No	Yes
Year	Yes	Yes	Yes	Yes	Yes	Yes
City	Yes	Yes	Yes	Yes	Yes	Yes
N	3371	3371	2935	2935	3396	3396
R2	0.0465	0.0505	0.9139	0.9163	0.2365	0.3470

Note: *, **, and *** denote 10%, 5%, and 1% significance levels, respectively. Values in parentheses are standard errors.

. From the results in column (2) of Table 6, it can be seen that the coefficient value of sprawl is significantly positive at the 1% level. It indicates that urban sprawl changes the way of land resource allocation, increases the proportion of industrial land, and indirectly promotes urban electricity consumption, verifying Hypothesis 2. Columns (3)–(4) of Table 6 show the regression results for private transportation. It can be seen that urban sprawl increases private transportation trips, and the number of private cars increases, which promotes energy consumption [107], testing Hypothesis 3. From the results in column (6) of Table 6, it can be seen that the coefficient of Sprwal is significantly negative at the 5% level. This indicates that urban sprawl inhibits urban green technological innovation, which is detrimental to the green transformation of the industry and indirectly promotes urban electricity consumption, validating Hypothesis 4.

5.5. Heterogeneity Analysis

China is a vast country, with different spatial plans for different regions, different environmental goal constraints, and different bases for innovation. The stage of economic development also varies greatly, which may lead to a differentiation of the effects of urban sprawl on electricity consumption. Therefore, this paper focuses on the four dimensions of urban spatial structure, environmental regulation intensity, urban innovation base, and urban geographic location to analyze heterogeneity.

5.5.1. Heterogeneity Analysis of Urban Spatial Structure

Columns (2) and (4) of

Table 7. Heterogeneity of urban spatial structure.

Variables	High-Spread		Low-Spread
	(1) lnElec	(2) lnElec	(3) lnElec	(4) lnElec
Sprwal	0.0201 ***	0.0177 ***	−0.0047	−0.0060
	(0.0052)	(0.0039)	(0.0122)	(0.0105)
_cons	22.5709 ***	21.7036 ***	22.6479 ***	22.0757 ***
	(0.0090)	(0.2299)	(0.0121)	(0.1656)
Controls	No	Yes	No	Yes
Year	Yes	Yes	Yes	Yes
City	Yes	Yes	Yes	Yes
N	1254	1254	2142	2142
R2	0.9689	0.9731	0.9761	0.9785

Note: *** denote 1% significance levels. Values in parentheses are standard errors.

report the heterogeneous effects of urban sprawl on electricity consumption at different levels of sprawl. The paper categorizes the sample cities into high and low sprawl groups according to whether the level of sprawl exceeds the national average for cities. Among them, the regression results for the high sprawl group are significantly positive, while the regression results for the low sprawl group are insignificant. This indicates that the higher the level of urban sprawl, the more significant the marginal incremental effect of urban sprawl on electricity consumption. The higher the level of urban sprawl in the region, the longer the spatial distance between enterprises, colleges and universities, research institutes, and other innovation subjects becomes, and the agglomeration economic effect of innovation factors is weakened. This is not conducive to the positive effect of agglomeration on innovation through risk diversification and knowledge spillover, and it hinders the realization of industrial green transformation and promotes electricity consumption. In addition, the higher the level of urban sprawl, the greater the dependence of residents on private cars, and the more obvious the promotion of electricity consumption.

5.5.2. Heterogeneity Analysis of Urban Environmental Regulation Intensity

This paper further examines whether there are differences in the impact of urban sprawl on electricity consumption across different urban environmental regulatory intensities. The strong environmental regulation group and the weak environmental regulation group are divided according to whether the urban environmental regulation intensity is greater than the national average for cities of environmental regulation intensity. As can be seen from columns (2) and (4) of

Table 8. Heterogeneity in urban environmental regulation.

Variables	Strong Environmental Regulation		Weak Environment Regulation
	(1) lnElec	(2) lnElec	(3) lnElec	(4) lnElec
Sprwal	0.0226	0.0211	0.0237 ***	0.0205 ***
	(0.0171)	(0.0158)	(0.0051)	(0.0043)
_cons	22.3373 ***	22.0652 ***	22.6435 ***	21.8146 ***
	(0.0194)	(0.3288)	(0.0062)	(0.1545)
Controls	No	Yes	No	Yes
Year	Yes	Yes	Yes	Yes
City	Yes	Yes	Yes	Yes
N	488	488	2908	2908
R2	0.9713	0.9735	0.9729	0.9765

Note: *** denote 1% significance levels. Values in parentheses are standard errors.

, the regression results for the strong environmental regulation group are not significant, while urban sprawl has a significant contribution to electricity consumption in the urban phase of weaker environmental regulation. This suggests that higher environmental regulation strengthens the emission reduction concept and environmental regulation behavior of local governments, raises the environmental target constraints on local governments, and restricts land grants to high energy-consuming enterprises. In addition, it can also force high-pollution and high energy-consumption industries to accelerate the pace of green transformation. Environmental regulation is used to prompt enterprises to participate in technological innovation, especially in areas involving energy conservation and emission reduction, and to increase the innovation power of industrial green transformation. This has to some extent weakened the role of urban sprawl in promoting electricity consumption. When environmental regulations are weaker, localities face less environmental pressure and less incentive to reduce land concessions to energy-consuming enterprises and urban sprawl has a stronger effect on electricity consumption.

5.5.3. Heterogeneity in the Level of Innovation Base

Differences in the green innovation base of cities may lead to differences in production efficiency, which in turn leads to the spread of cities with heterogeneous electricity consumption effects. In this paper, we use the total amount of green patents granted to measure the city innovation base. And the sample is divided into a high green innovation level group and a low green innovation level group based on the mean value of green patent grants. As shown in columns (2) and (4) of

Table 9. Heterogeneity in the level of innovation base.

Variables	High Green Innovation Level		Low Green Innovation Level
	(1) lnElec	(2) lnElec	(3) lnElec	(4) lnElec
Sprwal	0.0154	0.0045	0.0229 ***	0.0228 ***
	(0.0111)	(0.0098)	(0.0054)	(0.0046)
_cons	23.6332 ***	22.4690 ***	22.3423 ***	21.8181 ***
	(0.0118)	(0.4200)	(0.0066)	(0.1547)
Controls	No	Yes	No	Yes
Year	Yes	Yes	Yes	Yes
City	Yes	Yes	Yes	Yes
N	683	683	2713	2713
R2	0.9779	0.9847	0.9731	0.9754

Note: *** denote 1% significance levels. Values in parentheses are standard errors.

, the regression coefficients of urban sprawl are significantly positive at the 1% level for the low green innovation level group and insignificant for the high green innovation level group. It indicates that the effect of urban sprawl on electricity consumption shows very significant heterogeneity in the characteristics of the urban green innovation base. Green technological innovation is a key factor in improving energy efficiency, which helps industrial enterprises to clean up and decarbonize their production, thus reducing electricity consumption. The higher the level of urban innovation, the more it mitigates to some extent the contribution of urban sprawl to electricity consumption.

5.5.4. Heterogeneity of Geographic Location

Cities in different regions have certain differences in economic development and industrial structure. Therefore, the impact of urban sprawl on electricity consumption is also likely to be heterogeneous, and it is necessary to conduct a comparative analysis by region. Additionally, 283 cities at the prefecture level and above are categorized into three regions: east, central, and west. From the regression results in

Table 10. Heterogeneity of geographic location.

Variables	East	Central	West
	(1) lnElec	(2) lnElec	(3) lnElec
Sprwal	0.0083	0.0131	0.0332 ***
	(0.0053)	(0.0126)	(0.0069)
_cons	21.8188 ***	22.0268 ***	21.4169 ***
	(0.2533)	(0.2002)	(0.2489)
Controls	Yes	Yes	No
Year	Yes	Yes	Yes
City	Yes	Yes	Yes
N	1200	1200	996
R2	0.9843	0.9734	0.9755

Note: *** denote 1% significance levels. Values in parentheses are standard errors.

, it can be seen that, compared with the east and center, the more obvious the promotion effect of urban sprawl on electricity consumption is in the western region. The reason may be that, since the implementation of the Western Development Strategy in 2000, with the policy advantages and the absorption of a large number of foreign production factors, the size of the cities in the western region has increased rapidly. And, the National New Urbanization Pilot Work Program explicitly requires that new pilot cities should be tilted to the central and western regions. The cities in the central and west regions accounted for 57% of the total number of selected cities in 2014 and were as high as 61% in both 2015 and 2016. The phenomenon of urban sprawl is obvious, leading to a more serious mismatch of land resources and a significant rise in electricity consumption. However, the level of economic development of cities in the east and central regions is leading the country, which has produced a strong “siphon effect” on the resource elements of other regions, with a large number of foreign laborers, high-quality capital, and other agglomerations. Based on that fact, human capital and other characteristics of agglomeration ensure the innovation effect of human capital on the driving role of industrial green transformation. This, to a certain extent, weakens the negative effect of urban sprawl on electricity consumption.

5.6. Extensibility Analysis

5.6.1. Spatial Autocorrelation Analysis of Urban Sprawl and Electricity Consumption

OLS does not take into account the effects of spatial interactions between cities. This may lead to biased results and conclusions that inherently lack spatial implications [108]. Therefore, the spatial relevance of urban sprawl is further considered to investigate whether there is a significant difference in the effect of urban sprawl on electricity consumption on its own and in neighboring cities. Before conducting the spatial econometric analysis, the existence of the spatial effects of urban sprawl and electricity consumption are examined separately. To this end, a spatial autocorrelation test was conducted using the global Moran’s I index method, and the results are shown in

Table 11. Global Moran’s I value.

Year	lnElect			Sprwal
	Moran’s I	Z	sd(I)	Moran’s I	Z	sd(I)
2008	0.111 ***	22.184	0.005	−0.000	0.607	0.005
2009	0.111 ***	22.189	0.005	0.018 ***	4.369	0.005
2010	0.112 ***	22.425	0.005	0.001	0.840	0.005
2011	0.113 ***	22.645	0.005	0.009 ***	2.550	0.005
2012	0.113 ***	22.648	0.005	0.007 **	2.112	0.005
2013	0.110 ***	22.025	0.005	0.005 **	1.736	0.005
2014	0.110 ***	22.024	0.005	0.005 **	1.713	0.005
2015	0.110 ***	21.946	0.005	0.006 **	1.822	0.005
2016	0.109 ***	21.911	0.005	0.007 **	2.015	0.005
2017	0.109 ***	21.893	0.005	0.012 ***	3.057	0.005
2018	0.109 ***	21.857	0.005	0.016 ***	3.825	0.005
2019	0.109 ***	21.842	0.005	0.012 ***	2.950	0.005

Note: **, and *** denote 5%, and 1% significance levels, respectively.

. The global Moran’s I value for urban sprawl are all positive and pass at least the 5% significance level test overall. There are individual cases of non-significance, but basically it can be assumed that there is spatial correlation of urban sprawl. The global Moran’s I value for electricity consumption are concentrated at 0.109–0.113, and all pass the 1% significance level test, that is, it indicates that electricity consumption has strong spatial autocorrelation. This phenomenon may be attributed to the fact that there is an economic competition effect between cities, which makes the industrial structure tend to be homogeneous. This results in the increase of electricity consumption in the region, which significantly increases the electricity consumption of other cities. And, local governments may have formed vicious competition in land allocation, resulting in a significant positive spatial spillover effect of urban sprawl.

In order to better reflect the characteristics of spatial agglomeration, two representative years, 2009 and 2019, are selected to produce Moran’s I index scatterplots of urban sprawl and electricity consumption. The results are shown in Figure 4 and Figure 5. As can be seen in Figure 4, the local Moran’s I index of urban sprawl is mostly clustered in the third quadrant, showing obvious Low-Low clustering characteristics in 2009. However, after a decade of development, the distribution of urban sprawl scatter points is more decentralized in 2019, and the low aggregation gradually moves to the first quadrant, tending to High-High aggregation. This indicates that the level of urban sprawl is increasing, but there are differences in the strategic orientation, development planning, and economic foundation of each city. This leads to obvious regional differences in the level of urban sprawl in China, and the spatial correlation of urban sprawl is weakened. As can be seen in Figure 5, the change in the scatter distribution of Moran’s I index of urban electricity consumption between 2009 and 2019 is not obvious. The data points are more evenly scattered in the four quadrants, and the local Moran’s I index is significantly positive at the 1% level. This suggests that urban electricity consumption is strongly influenced by neighboring regions, that is, the coexistence of “High-High”, “High-Low”, “Low-High”, and “Low-Low”. However, there are relatively more data points distributed in the first and third quadrants, and the phenomena of “High-High” and “Low-Low” aggregation are more obvious. This suggests that there is both spatial dependence and heterogeneity in urban electricity consumption. It is the result of a combination of the cities’ industrial structure and ecological emphasis, and the implementation of the concept of synergistic regional governance.

5.6.2. Results of the Spatial Effects Test

Due to the existence of spatial correlations, this paper further analyzes them using spatial econometric models. Firstly, the inverse distance spatial weight matrix is used to estimate the spatial error model (SEM), spatial autoregressive model (SAR), and spatial Durbin model (SDM). The appropriate spatial econometric model was selected using the LM test, etc., and the specific results are shown in

Table 12. Spatial measurement model estimation.

Variables	SEM	SAR	SDM
	(1) lnElec	(2) lnElec	(3) lnElec
Sprwal	0.0168 ***	0.0189 ***	0.0185 ***
	(0.0019)	(0.0020)	(0.0020)
λ	3.4318 ***
	(0.0603)
$\mathsf{\rho }$		2.5635 ***	2.5983 ***
		(0.0361)	(0.0369)
W × Sprwal			0.0675 **
			(0.0324)
LM Test	1.247
LM-error	1211.472 ***
Robust LM-error	1212.660 ***
LM-lag	0.676
Robust LM-lag	1.864
lrtest both ind,df(10)	395.21 ***
lrtest both time,df(10)	23170.50 ***
Controls	Yes	Yes	Yes
Year	Yes	Yes	Yes
City	Yes	Yes	Yes
N	3396	3396	3396
R2	0.2893	0.0086	0.1244

Note: ** and *** denote 5%, and 1% significance levels. Values in parentheses are standard errors.

. The regression results show that the LM statistic and robust LM statistic of the spatial error model passed the 1% significance level test, but the LM statistic and robust LM statistic of the spatial lag model failed the significance test. In addition, based on the results of the test, the spatial error model with double fixation of time–area was chosen as a way to explain the spatial impact of urban sprawl on electricity consumption. From the estimation results of the spatial error model, the spatial coefficient λ of the spatial error term is 3.4318 and passes the 1% significance level test. It indicates that there is a significant spatial agglomeration phenomenon and spatial dependence on urban electricity consumption. That is, the electricity consumption of a city is not only related to its own factors but also affected by the neighboring cities, and the spatial correlation is mainly reflected in the random error term.

6. Discussion

This study confirms that urban sprawl exacerbates the impact of electricity consumption in China. Similarly, Spain faces rapid urban sprawl, with greater urban sprawl associated with higher electricity consumption [109]. Living in detached dwellings boosts electricity consumption, and urban sprawl may be an important driver of future surges in electricity consumption [110]. It is of great significance to grasp the interaction between urban spatial patterns and energy consumption, and then promote the construction of new urbanization and high-quality urban development. However, there are still some shortcomings in this paper. Firstly, this paper mainly utilizes statistical yearbook data to measure the urban sprawl index, while the statistical data “urban built-up area” may include some parts that do not belong to “cities” in the strict sense. This may have a certain impact on the results of urban sprawl measurement. Therefore, with the help of new data and methods, such as lighting data, extracting urban area and population data to measure urban sprawl, which is a breakthrough point in the future. Secondly, this paper describes and examines the mechanism of urban sprawl on electricity consumption from the aspects of land resource allocation, private transportation, and green technology innovation, as well as deepens the understanding of the relationship between urban sprawl and electricity consumption. But, the impact mechanisms of urban sprawl on electricity consumption is more complex. In the future, the impact mechanisms of urban sprawl on electricity consumption will be analyzed from more dimensions. Third, due to data limitations, this paper takes prefecture-level and above cities in China from 2008 to 2019 as research samples to explore the relationship between urban spatial patterns and energy consumption. The iteration of practice and strategic planning leaves the timeline of the study to be improved. Meanwhile, the promotion of urbanization with counties as important carriers is a new focus in China. Therefore, a more microscopic and detailed study on the relationship between spatial patterns and energy consumption in counties are possible further research.

7. Conclusions

This paper empirically examines the impact and mechanism of urban sprawl on electricity consumption based on the measurement of the urban sprawl index by processing the panel data of 283 prefectural-level and above cities in China from 2008 to 2019. Secondly, the heterogeneity of urban sprawl on electricity consumption is analyzed from four dimensions: urban spatial structure, environmental regulation intensity, urban innovation base, and urban geographic location. The SEM model is further used to investigate the impact of urban sprawl on electricity consumption, and the main conclusions are as follows: (1) In general, the levels of urban sprawl and urban electricity consumption show an upward trend. There are obvious differences between urban sprawl and electricity consumption in China. And, the spatial distribution of urban sprawl is relatively decentralized, while the cities with high electricity consumption levels are mostly concentrated in the eastern coastal areas. (2) Urban sprawl exacerbates electricity consumption. The conclusion still remains after the robustness test and the consideration of endogeneity issues. Urban sprawl exacerbates electricity consumption by changing the allocation of land resources, increasing the dependence on private transportation and inhibiting green technology innovation. (3) The effect of urban sprawl on electricity consumption is more pronounced in cities with high levels of sprawl, weak environmental regulations, and low levels of green innovation, as well as in west cities. (4) The Moran index shows that urban sprawl and electricity consumption both have a significant positive spatial correlation. The estimation of the SEM model shows that the electricity consumption of cities is not only related to their own regions but also influenced by the neighboring regions, and the spatial correlation is mainly reflected in the random error term.

Based on the above research conclusions, the following suggestions can be obtained. (1) Reasonable planning of urban spatial patterns, focusing on the improvement of energy utilization efficiency. In the process of new urbanization, we should comply with the reasonable demand of population density and economic growth and scientifically plan the urban development boundary. And, weighing the proportion of urban industrial land, residential land, and agricultural land. Correcting the short-sighted behavior of local governments in “seeking development with land” and reduce the degree of mismatch of urban land resources. Activating idle industrial land and optimize the spatial layout of industries; Promoting the advanced industrial structure in a gradual and orderly manner. And, paying attention to the integration of industry and city, the balance of business and residential development, and functional composites in the planning and construction of industrial parks. In addition, we will increase the compactness of urban spatial layout and reduce unnecessary commuting for seeking living services. Emphasis will be placed on the development of high-capacity green public transportation to reduce the proportion of private car trips. At the same time, promoting industrial specialization and diversified agglomeration and facilitates knowledge flow, matching, and sharing. (2) Based on regional development characteristics and advantages, differentiated territorial spatial planning should be introduced. Cities should strengthen the forcing effect of environmental regulation policies and continuously promote the improvement and innovation of environmental regulations. The demand for low-carbon development should be reflected in national spatial planning. With the help of stringent environmental regulation policy tools, the efficiency of land utilization can be improved. At the same time, it promotes intensive spatial layout and clustered industrial development mode and gives full play to the knowledge spillover effect to improve the cultivation of green innovation activities and help the green transformation of the industry. Urban sprawl in the eastern and central regions has not significantly exacerbated power consumption, and the optimal level of agglomeration and optimal city size should be maintained. In the western region, it is necessary to make full use of factor pooling, knowledge sharing, and technological spillover to promote the green development and transformation of cities and maintain a moderate spreading trend. (3) Regional alliances and establishing synergistic mechanisms for low-carbon development should be strengthened. Urban sprawl and electricity consumption have a positive spatial spillover effect. Therefore, sustainable urban development must abandon the “beggar-thy-neighbor” mode of thinking and break down urban administrative boundaries. Regional cooperation and establish a series of cooperative mechanisms should be strengthened, such as information sharing and joint law enforcement. The exchange of experience in land space planning between cities and the matching of measures to improve energy efficiency should be promoted, thus effectively mitigating the boosting effect of land sprawl on electricity consumption. We should deepen the reform of officials’ performance appraisal system and continue to weaken the share of GDP growth rate in the performance appraisal of local officials. We should seek focus points for the emerging competitiveness of cities, reduce energy consumption and change the development model, and rationally promote the construction of new urbanization.