An examination of the decoupling effect and influential mechanisms of energy consumption and economic growth in Chinese urban areas

Abstract

Under the “dual-carbon” target and the shift towards a high-quality economy, achieving a decoupling of energy consumption from economic growth is crucial. This study investigates the decoupling effect and the spatial-temporal dynamics of energy use and economic development across 282 prefecture-level cities in China from 2006 to 2021, employing the Tapio decoupling model. To thoroughly assess the factors influencing energy decoupling, a geographically weighted Durbin model was constructed based on an extended STIRPAT model. The findings reveal that:① There are notable disparities and variations in the decoupling status across regions, and from the 11th Five-Year Plan period to the 14th Five-Year Plan period, the number of cities exhibiting a weak decoupling state has diminished, with numerous cities demonstrating a negative growth decoupling state. ② The decoupling state of the majority of cities is rather steady; nevertheless, a few cities continue to show limited decoupling stability, while the overall decoupling exhibits a more stable state. ③ The geographical main effect indicates that energy structure and energy intensity exert a considerable positive influence on the energy consumption decoupling index, but GDP per capita and population density have a notable negative impact on the energy consumption decoupling index. ④The spatial spillover effect indicates that a city’s energy intensity adversely affects the energy consumption decoupling index of all adjacent regions, while the degree of openness to external influences positively impacts the energy consumption decoupling index of neighboring areas. This paper presents policy recommendations in four areas: differentiated energy management, specialized financial and technical support, enhancement of energy efficiency, and reinforcement of regional cooperation and exchange to achieve the decoupling of energy consumption from economic growth.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-16262-4.

Introduction

Research background

Energy underpins economic and social growth, and the quest for low-carbon green development has emerged as a worldwide consensus¹. Since the Industrial Revolution, the extensive development and use of fossil fuels have significantly advanced human civilization, although it has also resulted in issues such as resource depletion and climate change². As the foremost carbon emitter, responsible for approximately 30% of worldwide carbon emissions, China bears a responsibility to assume a pivotal role in mitigating carbon emissions and enhancing its response to global climate change³. In 2020, China made a formal pledge to the international community to attain carbon peaking by 2030 and carbon neutrality by 2060. This ambitious objective has been incorporated into national economic planning, propelling the transition to a sustainable economic development model⁴.

The National Bureau of Statistics of China reports that China’s GDP increased by approximately 448.81% from 2006 to 2021, total energy consumption rose by about 113.01% during the same timeframe, underscoring the strong relationship between economic growth and energy consumption, which has also resulted in numerous environmental pollution issues. These issues present a significant challenge to natural ecology and the human habitat, while also imposing limitations on the sustainable growth of energy, economy, and environment⁵. China, as a leading global energy consumer, confronts the dual problem of ensuring economic growth while regulating carbon emissions⁶. This serves as a crucial indicator for assessing the implementation efficacy of high-quality development and is also a vital metric for achieving the 2030 temperature control objective of the Paris Agreement.

In response to this problem, the Chinese government has persistently enacted a range of environmental protection initiatives while fostering economic growth, aiming to diminish energy consumption and attain sustainable development while preserving economic vitality⁷. Sinopec’s study estimation indicates that China’s total primary energy consumption would reach its zenith between 2030 and 2035⁸, at which point high-end manufacturing and high value-added services will emerge as the new catalysts for China’s economic growth. By 2035, China’s energy consumption is expected to diverge from its economic growth trajectory. Beyond this point, a full decoupling of energy use and economic expansion is anticipated, ushering in a new era of ecological civilization characterized by robust, sustainable development⁹.

Nonetheless, owing to the disparities in economic development stages and resource endowments among Chinese cities, reducing the reliance of economic growth on energy consumption and achieving a green low-carbon transformation has emerged as a shared objective and a significant challenge¹⁰. This process needs policy guidance, technical assistance, and the collaborative involvement of all societal sectors. This study examines the decoupling of energy consumption from economic growth in Chinese cities, aiming to offer practical solutions through a comprehensive examination of the decoupling state and its affecting elements. A comprehensive examination of the decoupling dynamics between energy consumption and economic growth in Chinese cities, analyzed temporally and spatially, along with the influencing factors, will provide clarity on the decoupling status across various regions in China. This establishes a scientific foundation for governments at all tiers to devise differentiated emission reduction policies, facilitating Chinese cities in implementing more effective and targeted actions to cultivate a low-carbon economy, thereby promoting the decoupling process.

Review of literature

The intricate link between energy consumption and economic growth has long been a central subject of investigation in both economics and energy policy. In 1956, Robert Solow introduced the seminal Solow growth model, emphasizing energy consumption as a crucial driver of economic expansion and proposing a lasting cointegrated relationship between these variables¹¹. Subsequently, researchers have further examined the relationship between energy consumption and economic growth by including additional factors, such as capital¹² and labor¹³. In contrast to Robert Solow’s findings, Mohsin et al. demonstrate that the impact of renewable energy consumption on economic growth evolves from initially negative to positive over the long term. This dynamic relationship is substantially influenced by factors including energy consumption intensity and the pace of energy transition¹⁴. Furthermore, Topolewski, in his analysis of European nations, discovered that short-term economic expansion substantially elevates energy consumption, and this effect remains considerable and favorable in the long run¹⁵. Nevertheless, certain research indicate that heightened energy consumption does not inherently stimulate quick economic growth. A complex, non-linear relationship between energy use and economic growth has been established¹⁶, underscoring significant implications for future research.

The notion of decoupling in economics is utilized to assess the relationship between economic growth and environmental challenges, facilitating the development of decoupling indicators for diverse economic activity¹⁷. In 2002, the OECD introduced the inaugural index for measuring decoupling, utilized to examine the correlation between environmental pressures (EP) and economic conditions (DF) across time. Nonetheless, the OECD has exclusively classified the stage of decoupling into relative and absolute categories¹⁸. Vehmas et al. subsequently presented a decoupling technique grounded in integrated analysis of change magnitude, categorizing the decoupling relationships into six classifications¹⁹; nevertheless, the elucidation of these relationships remains rather vague. Subsequently, Tapio further categorized the decoupling stages into eight distinct classifications and introduced the renowned Tapio decoupling mode²⁰. The Tapio decoupling model particularly highlights the separation of environmental impact from economic growth, emphasizing the importance of mitigating environmental harm while maintaining economic expansion. This is explicitly demonstrated by the decrease of emissions via improved efficiency, the adoption of cleaner technology, or the execution of structural changes in economic activity²¹.

While previous research has primarily concentrated on the decoupling relationship between carbon emissions and economic growth, studies examining the decoupling link between energy consumption and economic development have been less prevalent. Gao et al. employed the Tapio decoupling model to assess the status of carbon emissions in relation to economic development at the provincial level in China. Their findings indicate that most provinces are in a state of weak decoupling²². Alongside Gao et al.‘s research, several experts have extensively investigated the decoupling of carbon emissions from economic development. Freire-González et al. employed a segmented sample regression methodology to examine the correlation between carbon emissions and economic growth across 164 nations. The findings revealed notable regional disparities, with the majority of countries in Europe and Oceania demonstrating decoupling, whereas nations in Africa, the Americas, and Asia predominantly did not²³. Yang et al. examined the decoupling of global carbon emissions from economic growth and discovered that Europe and North America significantly contributed to this global decoupling process²⁴. Moreover, certain studies indicate substantial disparities in the enhancement of decoupling among areas and sectors at varying phases of economic growth²⁵, with the metrics of decoupling initiatives differing throughout regions and industries²⁶.

An increasing number of scholars have transitioned their focus from examining the relationship between the environment and economic growth to investigating the decoupling of energy consumption from economic growth, a shift underpinned by a robust theoretical framework. Energy use is the primary source of carbon emissions, and a significant causal relationship exists between them. Energy-related activities constitute the primary source of CO2 emissions in China, representing a significant portion of overall emissions²⁷. Secondly, from the standpoint of policy implementation, the regulation of energy usage is more immediate and actionable. The dual-control policy on energy consumption in China directly addresses both the total volume and intensity of energy use, whereas carbon emissions control is more indirect and must be achieved through the reconfiguration of the energy system and the economy²⁸. This study primarily investigates the current state and trajectory of energy consumption and economic growth, employing econometric analysis to elucidate their interconnections²⁹. Zhong H et al. discovered that China’s energy usage exerts a nonlinear influence on economic growth³⁰. Numerous studies conducted globally have explored the decoupling relationship between energy consumption and economic growth across different nations. For instance, Jones et al. investigated the progress of various states in the United States, revealing that several had made significant strides in enhancing energy efficiency³¹. Kumar and Singh noted in their study in India that fast industrialization has established a significant link between energy consumption and economic growth; nevertheless, certain places have exhibited preliminary signs of decoupling through technical advances³². The configuration of energy consumption is a crucial determinant influencing the decoupling of energy consumption from economic growth³³. As the share of low-carbon energy sources in the energy consumption framework rises, the limitations of energy consumption on economic growth progressively lessen, facilitating the decoupling of the two³⁴.

In conclusion, prior research has established a thorough analytical framework regarding the interplay between the environment and economic growth, thoroughly examined the intricate relationship between these two elements, and offered a robust theoretical and methodological basis for this paper. Nevertheless, there exists potential for growth. The existing research on the decoupling of energy consumption from economic growth predominantly concentrates on the provincial level, making it challenging to thoroughly illustrate the decoupling phenomenon in Chinese cities within the framework of the “dual-carbon” objective. This study systematically quantifies the decoupling of energy consumption in 282 Chinese cities from 2006 to 2021 and performs thorough analyses across temporal and spatial dimensions to elucidate the decoupling phenomenon in Chinese cities at various developmental stages. Secondly, current research predominantly emphasizes static analysis and neglects the dynamic evolution of the decoupling impact across time and location. This study analyzes data from 2006 to 2021, elucidating the peculiarities of the regional and temporal evolution of the decoupling effect and addressing the research gap in this domain. Third, current research has concentrated on the determinants of the energy consumption decoupling index, frequently overlooking the geographical connection of these determinants. This work integrates the STIRPAT model with the spatial Durbin model to conduct a comprehensive analysis of the geographic interactions between the policies and economic behaviors of nearby cities, therefore enhancing the knowledge of the spatial significance of these aspects.

Consequently, we emphasize China’s necessity for “economic structural transformation,” “structural adjustment of energy consumption,” and other pragmatic concerns, grounded on “sustainable economic and social development” and “ecological civilization construction”³⁵. This study investigates the decoupling dynamics and spatiotemporal evolution between energy consumption and economic growth across 282 prefecture-level cities in China from 2006 to 2021, employing the Tapio decoupling model. By applying an extended STIRPAT model and constructing a spatial Durbin model, the analysis identifies key factors influencing the decoupling index. These findings provide a robust scientific foundation for advancing low-carbon economic growth and promoting the decoupling process within China.

Incremental contribution

(1) The literature lacks comprehensive research on the decoupling effects and influencing factors of energy consumption and economic growth in Chinese cities. This study evaluates 282 cities across China to investigate the decoupling relationship between energy consumption and economic growth, thereby contributing to the promotion of green, high-quality, and sustainable urban development. (2) This paper introduces key variables—namely energy structure, energy intensity, and the level of openness—as indicators to conduct a comprehensive analysis of the factors influencing energy consumption decoupling. This methodology aids in discerning both the driving forces and barriers to decoupling, ultimately fostering sustained decoupling and advancing high-quality development that harmonizes economic growth with reduced energy consumption. (3) Utilizing the Tapio decoupling model alongside the spatial Durbin model derived from the extended STIRPAT model, this paper enables a detailed examination of regional characteristics across various developmental stages. This approach establishes a scientific basis for reducing the dependence between energy consumption and economic growth, thereby supporting regional low-carbon development objectives.

Research hypotheses

The theory of regional disparities posits that variations in economic development, industrial composition, resource availability, and policy frameworks among regions can result in divergent development paths and attributes during socio-economic advancement³⁶. In examining the correlation between energy consumption and economic growth, regional disparities may be evident in notable variations in decoupling status, the extent of decoupling, and its stability across different regions. The notion of spatial correlation highlights the reciprocal influence and spillover effects of neighboring regions in socio-economic growth³⁷. The technology diffusion theory posits that improvements in energy technology disseminate among geographically proximate areas, potentially influencing regional energy consumption efficiency and decoupling levels³⁸. Simultaneously, regional economic integration and industrial transformations foster interconnected economic activities among cities, culminating in spatially correlated energy consumption patterns³⁹. Consequently, spatial correlation and spillover effects may exist regarding the decoupling of urban energy use from economic growth.

The STIRPAT model offers a theoretical framework for examining the impacts of demographic, economic, and technical variables on environmental stress⁴⁰. The energy structure reflects the composition of urban energy consumption, influencing its cleanliness and sustainability⁴¹. Energy intensity, an indicator of energy use efficiency, affects the correlation between energy consumption and economic growth⁴². Additionally, socio-economic factors such as population density, economic development level, and openness to external influences contribute to the decoupling of energy consumption by impacting energy demand and technology diffusion⁴³. This study posits the following fundamental hypotheses based on the aforementioned theoretical analysis and known research findings:

H1: There are notable regional disparities in the decoupling status of energy consumption from economic growth across various Chinese cities.

H2: A spatial association exists between the decoupling impact of energy consumption and economic growth in Chinese cities.

H3: Factors like energy structure, energy intensity, and population density significantly affect the energy consumption decoupling index.

Research methodology and data sources

Tapio decoupling model

The Tapio index quantifies the correlation between the rate of resource consumption change and economic growth, first employed to evaluate the extent of divergence between economic development and natural resource use. This study utilizes the Tapio index to investigate the decoupling relationship between energy consumption and economic growth, specifically assessing whether energy consumption is relatively independent of economic growth⁴⁴. This study selects the Tapio decoupling model as the primary technique to assess the decoupling relationship between energy consumption and economic growth, primarily due to the following considerations: The Tapio model effectively illustrates the intricate and varied relationship between energy consumption and economic growth, aligning well with the substantial developmental disparities among Chinese cities. The Tapio model not only emphasizes the direction of decoupling but also evaluates its intensity and the changing status of economy and energy, providing a multi-dimensional assessment that facilitates a comprehensive understanding of the dynamic characteristics of decoupling. The Tapio model is extensively utilized in decoupling research and has significant applicability.

The Tapio index may be measured in two forms, one of which is a chained form derived from data of adjacent years, the other is based on the first year of data calculation. This research employs the two-stage rolling calculation approach of the Tapio index to investigate the precise decoupling connection between energy consumption and economic development in Chinese cities⁴⁵. Subsequently, in accordance with China’s five-year strategic planning, the decoupling indices of the 11th Five-Year Plan Period (2006–2010), the 12th Five-Year Plan Period (2011–2015), the 13th Five-Year Plan Period (2016–2020), and the 14th Five-Year Plan Period (2021) are individually assessed, focusing on the decoupling indices of energy consumption and economic growth to analyze their relationship during these four planning periods. The decoupling stability coefficient is integrated with other methods for further investigation. The particular models are enumerated as follows:

1

In formula (1), represents the Tapio decoupling index of energy consumption and gross regional product for city i in year t. signifies energy consumption, encompassing all forms of energy utilized within a specified timeframe, including electric power (with a discounted standard coal coefficient of 1.229 tons of standard coal per 10,000 kilowatts), petroleum (with a discounted standard coal coefficient of 1.7143 tons of standard coal per ton), and artificial gas and natural gas (with an equivalent standard coal coefficient of 13.3 tons of standard coal per million cubic meters). The GB/T 2589 − 2008 energy statistics index system provides the aforementioned energy conversion coefficients. This study’s energy consumption analysis excludes other energy sources, including renewable and nuclear energy. We compute the overall energy consumption in tons of standard coal by summing the energy consumption of each source and adjusting it by their respective conversion factors for discounted standard coal. represents the Gross Regional Product, whereas 0 indicates the base period. Decoupling happens when the growth rates of and , which denote the rate of change in energy consumption and real output, respectively, are not parallel. When the economic growth rate exceeds the energy consumption growth rate and the decoupling index is lower, the decoupling is more pronounced. Conversely, as the disparity between the two growth rates diminishes and the decoupling index increases, the decoupling becomes less pronounced.

This paper classifies decoupling states into eight categories based on Tapio’s research findings²⁰. The specific types of decoupling states and their corresponding index values are presented in Fig. 1. The thresholds of 0.8 and 1.2 are established as the criteria for decoupling status, representing the empirical values prevalent in most decoupling studies. These thresholds accurately reflect the elasticity range between energy consumption and economic growth during industrialization and urbanization across various regions of China. They effectively distinguish between strong and weak decoupling states while aligning with existing decoupling research within the Chinese context^46,47. Strong decoupling is the optimal state for attaining a low-carbon economy and sustainable development, whereas strong negative decoupling is the most detrimental state. While the other states are considered unsustainable, weak decoupling is a comparatively more favorable condition⁴⁸. The hierarchical classification of decoupling states, from optimal to suboptimal, is as follows: strong decoupling, weak decoupling, recessionary decoupling, growth linkage, recessionary linkage, weak negative decoupling, strong negative decoupling, and negative growth decoupling.

Fig. 1.

Classification of decoupling states.

In this study, we develop an indicator that characterizes the decoupled state’s level of stability as follows:

2

In formula (2), T represents the study year, whereas I denotes the overall study duration. Due to the influence of the specific decoupling index on the value, a definitive range for decoupling stability indicators remains ambiguous. A smaller -value generally signifies greater stability in the decoupling state, whereas a larger -consequence suggests reduced decoupling security.

Augmented STIRPAT and Spatial Durbin models

Holdren and Ehrlich initially employed the IPAT model to examine the effects of population, economics, and technological advancement on environmental pressures⁴⁹, whereas Waggoner and Ausubel subsequently introduced the ImPACT model, which differentiates the technology level () into T and the consumption of unit output ()⁵⁰. The IPAT and ImPACT models can only analyze changes in influencing factors in equal proportions, with the elasticity of I to the model variables fixed at a constant value of 1, resulting in a limited applicability. Building on this foundation, Dietz and Rosa subsequently introduced the STIRPAT model, illustrated in formula (3), whereby the environmental pressure () and its influencing elements no longer exhibit equal proportional fluctuations but rather assume a nonlinear shape⁴⁰.

3

By applying logarithms to both sides of formula (3), the model is transformed:

4

In formula (3) and formula (4): represents environmental pressure, denotes population, signifies the economy, and indicates the technical level, which serves as the model’s intercept term. The coefficients of each influencing component are 、and, while represents the error term. The enhanced STIRPAT model can incorporate additional variables to examine their influence on environmental pressure. This paper incorporates the influencing factors of energy consumption, including energy structure, industrial structure, urbanization level, degree of openness, and forest coverage, which may affect the decoupling of energy consumption, as referenced in the pertinent literature^51,52. The model is subsequently expanded based on formula (4) as follows:

5

Given that the decoupling dynamics of each city may exhibit geographical interactions, spatial elements are incorporated, and a spatial panel model is employed to study the determinants influencing the decoupling of each city. Building upon formula (5), the spatial Durbin model of the energy consumption decoupling index is developed to thoroughly examine the causes causing geographical variations in the energy consumption decoupling index.

6

In formula (5) and formula (6): serves as the explanatory variable, representing the energy consumption decoupling index for city in year . is the energy consumption decoupling index of city , a neighboring city of city , in year . represents the spatial autoregressive coefficient of the explanatory variables, signifying the direction and magnitude of the geographic spillover impact of the decoupling index of adjacent cities on the decoupling index itself. denotes the spatial weight matrix; represents the explanatory variables; signifies the spatial effect coefficient of the explanatory variables; and denote the temporal fixed effect and spatial fixed effect, respectively; shows the random error term.

Variable selection and data source

Explained variable

The energy consumption decoupling index , indicative of city ‘s energy consumption decoupling in year , accurately reflects the variations in the link between urban energy consumption and economic growth. Nevertheless, the decoupling index possesses both positive and negative values, rendering it unsuitable for straight logarithmic transformation. This study, using Dong et al.‘s research, initially identifies the integer of the smallest decoupling index among 282 cities, subsequently subtracts this integer from the decoupling index of each city for each year, and ultimately computes the logarithm of the results⁵³. Modifying the raw data to conform to the logarithmic model addresses the issue of negative data values, hence enhancing the stability and interpretability of the model analysis results⁵⁴.

Explanatory variables

The augmented STIRPAT model incorporates the effects of demographic, technological, and economic development factors on environmental pressures. Drawing from existing literature on energy consumption impact factors^55,56, this paper introduces variables such as energy structure, energy intensity, and openness level to conduct a more detailed analysis of the determinants of the decoupling index of energy consumption. Specifically, when economic growth escalates, industrial activity, business demand, and urban living standards would correspondingly rise, resulting in increased energy consumption⁵⁷. This research considers the level of economic development as a determinant of energy consumption decoupling, quantified by the per capita GDP () indicator. To guarantee the precision of the analysis, all GDP (pgdp) statistics are anchored to 2006, and real price adjustments have been implemented via the GDP deflator for each region to mitigate the influence of inflationary pressures and enhance the comparability of economic data between years. The concentration of people stimulates increased economic activity and living requirements, hence resulting in higher energy consumption⁴³. In this study, population serves as a determinant for decoupling energy consumption, represented by the population density () indicator, defined as the ratio of the resident population of a region to its urban area. Technological advancement is crucial for fostering sustainable development, enabling a balance between economic expansion and environmental conservation⁵⁸. Consequently, this article considers technological levels as a determinant of energy consumption decoupling. The technology level variable () encompasses both performance-based and input-based components. The performance-based technology level is represented by the energy intensity () indicator, defined as the ratio of total energy consumption to nominal GDP, serving as a measure of energy use efficiency⁵⁹. The input-based technology level variable is represented by the R&D input intensity () indicator, which is the ratio of internal funding expenditure to nominal GDP, signifying the extent of investment in technological innovation inside the city⁶⁰.The energy structure () denotes the composition and proportional distribution of urban energy consumption, significantly influencing the cleanliness and sustainability of urban energy usage⁴¹. Consequently, this article utilizes energy structure () as a determinant for energy consumption decoupling, defined as the ratio of coal use to total energy consumption. Foreign direct investment (FDI) typically introduces sophisticated managerial expertise and technology, influencing local energy efficiency and infrastructure⁶¹. This research identifies the degree of openness to external influences () as a significant determinant in energy consumption decoupling, quantified as the ratio of total foreign investment to regional GDP.

Sources of data

In light of data availability and comparability, this analysis selects 282 prefecture-level cities in China, omitting Hong Kong, Macao, and Taiwan. This report examines the period from 2006 to 2021, reflecting the substantial modification of China’s dual-control strategy on energy usage initiated in 2006.The energy consumption decoupling index () is derived from the Tapio decoupling model, utilizing data on per capita GDP (), population density (), the ratio of total foreign investment to GDP (), total energy consumption (), energy intensity (), and energy structure () sourced from the “Statistical Yearbook of China’s Cities,” the “Statistical Yearbook of Energy in China,” and the “Statistical Yearbook of the Environment” for each prefecture-level city. Environmental Statistics Yearbook, Statistical Yearbook of Prefectural Municipalities, Statistical Bulletin of National Economic and Social Development of Prefectural Municipalities, and more government statistics. Linear interpolation supplements a minimal quantity of absent data.

Results

Prolonged examination of the decoupling of energy consumption

The decoupling index of energy consumption and economic development for Chinese cities from 2007 to 2021 is computed annually using formula (1) with 2006 as the base year. This paper analyzes Chinese cities by categorizing them into four regions: east, central, west, and northeast. This classification is based on the study by Pang et al.⁶² and is informed by the “Opinions of the Central Committee of the Communist Party of China and the State Council on the Promotion of the Rise of the Central Region,” the “State Council Issued the Implementing Opinions on a Number of Policies and Measures on the Development of the Western Region,” and the regional division standards established by the National Bureau of Statistics, acknowledging the significant disparities in development, resource endowment, and energy consumption across different regions in China. The West and Northeast regions are analyzed. Table 1 illustrates that the eastern region comprises cities like Beijing, Tianjin, and Shanghai, which exhibit elevated levels of economic development, industrialization, and urbanization. The center region has cities like Anqing and Datong, which possess abundant resources, with coal predominating energy usage. The western region comprises cities like Chengdu, Chongqing, and Kunming, characterized by abundant but unevenly distributed resources, a complex economic development landscape, and elevated energy consumption levels. The northeastern region include Anshan City, Fushun City, and other municipalities. It possesses a robust conventional industrial foundation and abundant resources; nonetheless, it experiences significant demand for transformation in its economic development, with energy consumption being relatively moderate.

Table 1.

Statistical tables on regional economic development and energy consumption in China.

Indicator Type	Statistic	Eastern Region	Central Region	Western Region	Northeast Region
Maximum Value	Economic Development Level (Yuan)	218,118	205,941	187,415	157,914
	Energy Consumption (tons of standard coal)	41,239,987	13,645,692	34,642,262	22,166,696
Minimum Value	Economic Development Level (Yuan)	19,915	20,974	25,279	28,932
	Energy Consumption (tons of standard coal)	473488.7	358395.3	340028.3	422590.9
Mean Value	Economic Development Level (Yuan)	68380.93	69633.54	78990.84	74816.27
	Energy Consumption (tons of standard coal)	4,057,394	3,413,081	4,709,239	3,991,785

The spatial distribution of the decoupled state of energy consumption in Chinese cities is visualized with the help of ArcGIS 10.8 software, as shown in Fig. 2. In the eastern region, cities such as Suzhou, Wuxi, and Nantong demonstrated predominantly weak decoupling from 2006 to 2021. Here, economic growth slightly outpaced energy consumption, which, while increasing, did so at a diminished rate—suggesting an improvement in economic efficiency. Conversely, cities like Jinan, Qingdao, and Yantai have shifted from a weak decoupling state to one characterized by negative growth decoupling, reflecting a pronounced dependence on energy consumption for economic expansion, particularly in energy-intensive sectors like industry and infrastructure. In contrast, Dongying, Weifang, and Zibo primarily exhibit a negative growth decoupling state, where economic growth is heavily reliant on increased energy consumption, raising concerns regarding sustainable development. Notably, Hangzhou and Ningbo transitioned from weak decoupling to a state of growth linkage during the study period, with energy consumption and economic growth rising at nearly identical rates. This pattern indicates a lack of decoupling and underscores the urgent need for further optimization of energy structures and enhancement of energy efficiency.

Fig. 2.

Progression of energy consumption decoupling status in Chinese cities. This map is based on the standard map (Review No. GS(2019)1825) from the Standard Map Service website of the Ministry of Natural Resources of China. The base map boundaries have not been modified. http://bzdt.ch.mnr.gov.cn.

Cities such as Anqing and Datong in the central region exhibit a predominantly weak decoupling state, indicating that while their economies are expanding, the rate of growth in energy consumption is relatively slower, reflecting some advancements in energy efficiency. Conversely, cities like Bengbu, Chuzhou, and Jiujiang are transitioning from weak decoupling to a negative growth decoupling state, which signifies an increasing reliance on energy consumption as their economies advance. In contrast, Suizhou, Tongling, and Wuhan predominantly remain in a negative growth decoupling state, suggesting that their economic expansion is largely contingent on heightened energy consumption, highlighting the challenges these cities face in achieving sustainable development. Additionally, cities such as Lu’an and Loudi have shifted from a weak decoupling state to a growth linkage state during the study period, indicating a synchronized pattern between energy consumption and economic growth, thereby failing to decouple the two and necessitating urgent measures to address this challenge. Meanwhile, cities like Changde, Xinzhou, and Zhangjiajie initially demonstrated strong decoupling but have since transitioned to other decoupling stages.

In the western region, cities such as Chengdu, Chongqing, and Kunming predominantly display a weak decoupling state from 2006 to 2021, marked by economic growth rates that outpaced energy consumption. In contrast, cities like Lanzhou, Xining, and Yinchuan shifted from a weak decoupling state to one characterized by negative growth decoupling. Urumqi, Guiyang, and Xi’an, among others, predominantly exhibited a negative growth decoupling state, indicating that economic expansion was heavily dependent on increased energy usage. Conversely, cities such as Lhasa and Hohhot transitioned from a weak decoupling state to a growth linkage state over the study period. In the Northeast, cities including Anshan, Fushun, and Dandong also moved from a weak decoupling state to a negative growth decoupling state, reflecting an unsustainable economic model reliant on escalating energy consumption. Meanwhile, cities like Benxi and Qitaihe progressed from a weak decoupling state to a growth linkage state, demonstrating parallel growth rates.

There are notable disparities and variations in the decoupling status across the eastern, western, central, and northeastern areas, indicative of the distinct features of each region concerning economic development and energy use. Certain cities in the eastern area can regulate the growing rate of energy consumption alongside economic expansion, but other cities exhibit economic growth that is significantly reliant on heightened energy consumption. Numerous cities in the western region are witnessing rapid increases in energy consumption due to economic expansion and must address the issues of sustainable development. Variations in the decoupling status across cities in the central region indicate the reliance of distinct cities on energy consumption throughout industrialization and urbanization. Although several communities in the northeastern area have successfully decoupled energy consumption from economic growth, others continue to see energy consumption outpacing economic growth.

To ascertain the statistical significance of the differences in decoupling status among regions, one-way ANOVA and post hoc multiple comparison tests were used. Table 2 presents the descriptive statistics and ANOVA findings for the decoupling indices across the eastern, central, western, and northeastern regions. The descriptive statistics indicate a considerable disparity in the mean values of the decoupling index across the four regions. The western region exhibits the highest mean value of the decoupling index, signifying the most unfavorable condition of decoupling; conversely, the northeastern region has the lowest mean value of the decoupling index, reflecting the comparatively optimal state of decoupling; the eastern and central regions occupy an intermediate position. The standard deviation of the western region is markedly higher than that of the other three regions, signifying that the disparities in decoupling status across cities in the western region are more pronounced, hence highlighting the issue of developmental imbalance. The ANOVA results indicate a highly significant variation in the decoupling index across the four regions (F = 14.63, p < 0.001), so establishing a statistical foundation for the formulation of regionally diversified energy strategies.

Table 2.

ANOVA results for the decoupling index by region.

Region	Sample Size	Standard Deviation	F-value	P-value
Eastern Region	1290	0.218	14.63	0.000***
Central Region	1200	0.216
Western Region	1215	0.340
Northeast Region	495	0.225

***, **, * denote significance at the 1%, 5%, and 10% levels, respectively.

The two-by-two regional differences were further examined using Tukey HSD post hoc testing, with results presented in Table 3. The test results indicate distinct patterns of variation in decoupling status among locations. The western region exhibits the most pronounced disparities compared to other regions, with a markedly elevated decoupling index relative to the eastern, central, and northeastern regions, hence affirming the assertion that the western region has a comparatively deficient decoupling status. The western and northeastern regions exhibit the most significant disparities (mean difference = −0.081, p < 0.001), potentially attributable to variations in industrial organization and resource endowment between the two areas. The Northeast region exhibits superior decoupling, evidenced by a markedly lower decoupling index compared to the East, albeit with a little degree of difference. This outcome may be associated with the deceleration of economic growth and the modernization and enhancement of conventional heavy industries in the Northeast. The disparity between the eastern and central regions, as well as that between the northeastern and central regions, is minimal, indicating that the decoupling status of these areas is relatively aligned and may warrant the implementation of analogous energy policy instruments.

Table 3.

Tukey HSD Post-hoc multiple comparison Results.

Regional Comparison	Mean Difference	p-value
Central-Eastern	−0.010	0.794 (0.010)
Western-Eastern	0.043	0.000*** (0.010)
Northeast-Eastern	−0.037	0.033* (0.014)
Western-Central	0.053	0.000*** (0.011)
Northeast-Central	−0.028	0.188 (0.014)
Northeast-Western	−0.081	0.000*** (0.014)

***, **, * denote significance at the 1%, 5%, and 10% levels, respectively.

Analysis of short-cycle energy consumption decoupling

The decoupling index of energy consumption from economic growth throughout the 11th Five-Year Plan Period (2006–2010), the12th Five-Year Plan Period (2011–2015), the13th Five-Year Plan Period (2016–2020), and the14th Five-Year Plan Period (2020–2021) is calculated using the first year of each planning period as the reference year. This paper examines the decoupling of energy consumption from economic growth in China from 2006 to 2021, referencing the study by Pang et al.⁶², which categorizes this decoupling into four distinct five-year planning periods in light of the specific economic and development objectives outlined in China’s five-year plans, including targets for energy consumption, conservation, and emission reduction.

With the help of ArcGIS 10.8 software to visualize the decoupling state of energy consumption and economy in each city during the four five-year planning periods in China, as shown in Fig. 3. During the 11th Five-Year Plan period, several cities, including Beijing, Baoding, Binzhou, and Anqing, experienced a weak decoupling state, characterized by a modest increase in energy consumption relative to economic growth. This trend can be largely attributed to the binding target set by the State Council, which mandated a reduction of energy consumption per unit of GDP by approximately 20%. The robust implementation of carbon emission reduction policies across various regions further contributed to this phenomenon. In contrast, certain cities, such as Baishan and Benxi, exhibited a negative decoupling state, indicating that energy consumption grew at a rate surpassing economic development during this expansion.

Fig. 3.

Progression of the decoupling of energy use from the economy per municipality throughout four five-year planning intervals. This map is based on the standard map (Review No. GS(2019)1825) from the Standard Map Service website of the Ministry of Natural Resources of China. The base map boundaries have not been modified. http://bzdt.ch.mnr.gov.cn.

Comparatively, the 12th Five-Year Plan period saw a decrease in the number of cities, including Beijing, Anqing, and Anyang, that maintained a weak decoupling state. However, the number of cities experiencing negative decoupling growth increased, notably in Baoding, Baicheng, and Bengbu, reflecting intensified energy consumption amidst industrialization and urbanization. By the 13th Five-Year Plan period, fewer cities demonstrated weak decoupling, while others, such as Beijing, Baoding, and Baicheng, faced negative growth decoupling, with their economic expansion increasingly dependent on elevated energy consumption.

In conclusion, during the 11th to 14th Five-Year Plan periods, a growing number of cities have experienced a state of negative growth decoupling, characterized by an increasing dependence of economic growth on energy consumption. This trend poses significant challenges to achieving sustainable development. In contrast, several towns have successfully maintained a weak decoupling state across these periods, effectively managing energy consumption alongside economic growth. This suggests a notable improvement in energy efficiency.

One-way ANOVA study of the decoupling index over the 11th, 12th, 13th, and 14th Five-Year Plan periods reveals a highly significant variation in the decoupling index among these eras (F = 276.07, p < 0.001), as illustrated in Table 4. The statistical significance of this result surpasses that of the regional variance analysis, suggesting that temporal changes exert a more substantial influence on the decoupling status.

Table 4.

ANOVA results of decoupling indices by period.

Period	Sample Size	Standard Error	F-value	P-value
11th Five-Year Plan (2006–2010)	1120	0.203	276.07	0.000***
12th Five-Year Plan (2011–2015)	1400	0.120
13th Five-Year Plan (2016–2020)	1400	0.317
14th Five-Year Plan (2021)	280	0.335

***, **, * denote significance at the 1%, 5%, and 10% levels, respectively.

The Tukey HSD post hoc test illustrates the temporal progression of decoupling, as demonstrated in Table 5. The decoupling status between the 11th and 12th Five-Year Plan periods is relatively stable, with an insignificant difference (p = 0.671), suggesting that the decoupling status at the onset of the 12th Five-Year Plan period perpetuates the positive trend established during the 11th Five-Year Plan period, and the impact of the dual-control policy on energy remains intact. During the 13th Five-Year Plan period, a decline occurred, with the decoupling index markedly exceeding that of the 11th and 12th Five-Year Plan periods, signifying a crucial turning point in the decoupling process. This may pertain to the structural adjustments associated with the new economic normal and the resurgence of heavy chemical industry in certain places during that time. The trend of decline persisted during the 14th Five-Year Plan period, as the decoupling index exceeded that of the preceding period, indicating an exacerbation of the decoupling challenge in recent times.

Table 5.

Tukey HSD Post-hoc multiple comparison Results.

Period Comparison	Mean Difference	p-value
12th vs. 11th	−0.011	0.671 (0.010)
13th vs. 11th	0.200	0.000*** (0.010)
14th vs. 11th	0.251	0.000*** (0.016)
13th vs. 12th	0.211	0.000*** (0.010)
14th vs. 12th	0.262	0.000*** (0.016)
14th vs. 13th	0.051	0.006** (0.016)

***, **, * denote significance at the 1%, 5%, and 10% levels, respectively.

Analysis of the stability of energy consumption decoupling

Employing formula (2) to get the absolute value of the annual rate of change of the city’s decoupling index in relation to the preceding year and subsequently calculating the mean of these absolute values to derive the decoupling intensity coefficient as shown in Table 6. The examination of decoupling stability indicates that the decoupling intensity coefficient for the majority of cities are low, suggesting that the decoupling status in these areas is comparatively stable. The Northeast region’s average decoupling intensity coefficient is 0.73, indicating a very consistent decoupling state. Daqing possesses a decoupling intensity coefficient of 0.498, indicating a more stable decoupling state than the norm. The average decoupling intensity coefficient for the eastern region is around 0.82, just above that of the northeastern region; however, it demonstrates superior decoupling stability overall. The maximum value of the decoupling intensity coefficient for specific cities is 19.18, suggesting that while the majority of cities in the eastern region exhibit greater stability in their decoupling status, a minority display less stability, and these outliers can significantly influence the overall decoupling stability of the region. The central region’s condition parallels that of the eastern region, exhibiting an average decoupling strength coefficient of 1.01; however, its decoupling stability is inferior to that of the eastern and northeastern regions. Furthermore, the peak value of the decoupling intensity coefficient for specific cities in the central region attains 20.56, the highest among all regions, indicating a notably weak decoupling stability. The average decoupling intensity coefficient in the western region is 1.61, indicating relatively weak decoupling stability; however, the total average increases due to elevated decoupling intensity coefficients in several cities within the region. In conclusion, while the majority of cities exhibit low decoupling intensity coefficients and a generally stable decoupling status, certain regions present extreme values that indicate weak decoupling stability, necessitating focused analysis and appropriate interventions. The Northeast region has the highest overall performance, while the Western region encounters more significant obstacles compared to the rest of the nation. The Central and Eastern regions occupy a median position, with several cities requiring attention.

Table 6.

Stability of energy consumption decoupling.

Region	Count	Mean	Std	Min	25%	50%	75%	Max
Northeast region	33	0.731	0.564	0.169	0.351	0.509	0.882	2.844
Eastern region	86	0.821	2.257	0.079	0.211	0.347	0.551	19.179
Central region	80	1.014	2.518	0.097	0.240	0.393	0.653	20.557
Western region	81	1.610	2.716	0.104	0.403	0.645	1.484	19.138
All	280	1.094	2.370	0.079	0.267	0.465	0.879	20.557

Examination of determinants influencing the decoupling of energy consumption

Formal evaluation of the model configuration

This research conducted a series of standardized tests using the research of some scholars^63,64. to determine the appropriate model configuration, the results are shown in Table 7. The LM test is employed to ascertain the appropriateness of a panel model incorporating spatial interaction effects. The findings indicate the presence of both spatial error effects and spatial lag effects, necessitating the utilization of the spatial Durbin model. The Hausman test indicates that the null hypothesis is rejected at the 1% significance level, leading this study to select the fixed effects model over the random effects model. The LR test and Wald test were employed to determine if the spatial Durbin model would reduce to a spatial error model or a spatial lag model, and the findings indicated that the spatial Durbin model remained an appropriate model form. The LR test and the assessment of the goodness of fit R2 were employed to determine the appropriate fixed effects, ultimately resulting in the selection of the two-way fixed effects spatial Durbin model.

Table 7.

Primary outcomes of the formal evaluation of the model configuration.

	Statistic	P
LM-lag	1434.182	0.000***
LM-error	1708.738	0.000***
RLM-lag	367.005	0.000***
RLM-error	641.561	0.000***
rWald_spatial_lag	37.80	0.000***
Wald_spatial_erro	24.80	0.000***
LR-sar	37.28	0.000***
LR-sem	24.37	0.000***
Hausman	105.14	0.000***
LR-both-time	36.15	0.000***
LR-both-ind	7266.55	0.000***

***, **, * signify significance at the 1%, 5%, and 10% levels, respectively.

Spatial correlation assessment

This paper constructs three distinct spatial weight matrices to enhance the credibility of the estimation results derived from the spatial measurement model, which are influenced by the choice of these matrices. Drawing on established spatial weight selection methodologies from existing literature, we develop the economic distance weight matrix⁶⁵, the binary neighborhood weight matrix⁶⁶, and the geographical distance decay weight matrix⁶⁷. The analysis results derived from the economic distance weight matrix serve as the baseline, whereas those from the binary neighborhood weight matrix and the geographical distance decay weight matrix function as robustness tests. The significance of all three matrices demonstrates strong robustness. This work initially examines whether the energy consumption decoupling index of each city exhibits geographic association, employing the Moran index, a method often utilized in the literature^68,69. This document presents the annual worldwide Moran index findings at the center, utilizing the economic distance weight matrix. Table 8 illustrates that the energy consumption decoupling index for each city exhibits spatial correlation overall, with the correlation index demonstrating a general rising trajectory over time. The Moran’s I index did not attain statistical significance in 2007, 2014, 2015, and 2016, maybe indicating the particularities of China’s energy policy and economic context during these years. 2007 marked the initial phase of the 11th Five-Year Plan’s implementation, during which the synergistic effects of regional energy policies had yet to materialize; the years 2014–2016 represented a transitional period in China’s new normal economic landscape and the nascent stage of supply-side structural reform, resulting in transient fluctuations in energy consumption patterns due to regional economic restructuring. While individual years exhibit negligible spatial correlation, the entire study period (2006–2021) reveals that most years demonstrate strong spatial connection, particularly post-2017, when the correlation becomes more pronounced and meaningful. This indicates that the spatial correlation of the decoupling condition of energy consumption among cities has progressively intensified with the more profound execution of the regional coordinated development strategy and the alignment of energy policies. This study use the spatial Durbin model on the complete sample to more thoroughly capture spatial dependence while ensuring model consistency and continuity. Furthermore, the non-significant yet positive Moran’s I index for the non-significant years indicates a degree of spatial correlation, and the LM test findings endorse the application of the spatial model.

Table 8.

Global Spatial correlation analysis for energy consumption decoupling indices.

Year	Global Moran ‘s I index	Year	Global Moran ‘s I index
2007	−0.005	2015	−0.002
2008	0.008**	2016	0.004
2009	0.017***	2017	0.012***
2010	0.012***	2018	0.015***
2011	0.012***	2019	0.02***
2012	0.015***	2020	0.022***
2013	0.013***	2021	0.012***
2014	0.0045

***, **, * denote significance at the 1%, 5%, and 10% levels, respectively.

Examination of contributing variables

This research conducts a thorough diagnosis of multicollinearity for the explanatory variables prior to executing the spatial Durbin model regression. The variance inflation factor (VIF) served as the primary assessment, whilst tolerance was computed for cross-validation purposes. Table 9 indicates that the multicollinearity issues among all explanatory variables remain within acceptable thresholds. The VIF values for all variables are well below the conventional threshold of 10, with a maximum VIF of 2.23 and an average VIF of 1.49, suggesting a minimal correlation among the variables. The tolerance of all variables exceeds 0.4, much beyond the essential threshold of 0.1, hence further substantiating the absence of multicollinearity issues. Consequently, the spatial Durbin model regression analysis can be executed directly with the chosen explanatory variables, eliminating the necessity for procedures like variable elimination or principal component analysis.

Table 9.

Results of multicollinearity test for explanatory variables.

Variables	VIF	1/VIF
rd	2.23	0.449
pgdp	1.64	0.608
pi	1.61	0.620
open	1.27	0.787
ei	1.13	0.883
es	1.05	0.948
Mean VIF	1.49

Table 10 presents the outcomes of the spatial Durbin model regression utilizing three distinct weighting matrices, encompassing both main effects and spillover effects. The model’s resilience can be efficiently validated by the multi-dimensional spatial weight configuration. All three weighting matrices demonstrate a significantly positive spatial autocorrelation coefficient (ρ) at the 1% significance level, indicating substantial spatial correlation in energy decoupling among cities. A positive coefficient suggests that a modification in the energy decoupling index of one region will elicit a corresponding change in adjacent regions. Under the economic distance weight matrix, a 1% increase in a city’s energy decoupling index results in an average increase of 0.357% in the energy decoupling indices of its adjacent cities. The effect is 0.067% and 0.093% for the binary neighborhood and geographic distance decay weight matrices, respectively. The spatial autocorrelation coefficient associated with the economic distance weighting matrix is markedly higher than those of the other two weighting matrices, indicating that the influence of economic linkages between cities on energy decoupling is more pronounced than that of mere geographic distance. The variations in spatial correlation strength across different weight matrices illustrate the multifaceted nature of inter-city connections and offer an empirical foundation for comprehending the spatial synergy mechanism of regional energy consumption decoupling.

Table 10.

Primary outcomes of the Spatial Durbin model estimate.

VARIABLES	Economic Distance Weight Matrix		Binary Neighborhood Weight Matrix		Geographical Distance Decay Weight Matrix
	(1)	(2)	(1)	(2)	(1)	(2)
	Main effects	Spillover effects	Main effects	Spillover effects	Main effects	Spillover effects
lnrd	−0.002	−0.058	−0.001	−0.024***	−0.003	−0.012*
	(0.003)	(0.056)	(0.003)	(0.006)	(0.003)	(0.007)
lnei	0.372***	−0.237***	0.373***	−0.035***	0.372***	−0.042***
	(0.004)	(0.059)	(0.004)	(0.010)	(0.004)	(0.013)
lnes	0.119***	−0.260*	0.109***	0.008	0.112***	−0.004
	(0.011)	(0.146)	(0.011)	(0.020)	(0.011)	(0.026)
lnpGDP	−0.055***	−0.097	−0.058***	−0.013	−0.057***	−0.036**
	(0.010)	(0.078)	(0.010)	(0.013)	(0.010)	(0.017)
lnpi	−0.113***	−0.119	−0.090**	−0.128*	−0.123***	−0.035
	(0.042)	(0.290)	(0.041)	(0.072)	(0.040)	(0.082)
lnopen	−0.001	0.035**	0.000	0.001	−0.001	0.006*
	(0.002)	(0.016)	(0.002)	(0.003)	(0.002)	(0.004)
ρ	0.357*** (0.112)		0.067*** (0.021)		0.093*** (0.045)
R-squared	0.019		0.060		0.129

***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively, while Maineffects and Spillovereffects refer to main and spillover effects, respectively.

The model’s goodness of fit under the three weight matrices is 0.019, 0.060, and 0.129, respectively, indicating relatively low values. This is primarily attributable to the following factors: The decoupling of energy consumption from economic growth is a multifaceted process influenced by numerous factors, including technical advancements, energy price variations, and policy modifications, among others, and the model does not encompass all potential influencing elements. Third, substantial disparities exist in the developmental stages, resource endowments, and industrial structures of various regions, resulting in variable directions and intensities of influencing factors across these regions, hence impacting the overall model’s fitting efficacy. Nonetheless, the coefficients of the model’s independent variables are statistically significant, and their directional signs align with theoretical expectations. This suggests that the model can effectively elucidate the principal factors influencing energy consumption decoupling and its mechanisms, thereby offering valuable insights for policy development.

Main effects denote the direct impact of explanatory variables on the energy decoupling index within a certain region. Energy intensity is a primary component influencing the energy consumption decoupling index. The coefficients are 0.372, 0.373, and 0.372 across the three weighting matrices, all of which are statistically positive at the 1% level. This signifies that for each 1% rise in energy intensity, the energy consumption decoupling index will augment by approximately 0.37%. An elevation in energy consumption per unit of GDP will markedly diminish the extent of decoupling between energy consumption and economic growth, aligning with theoretical anticipations. Energy intensity serves as a crucial metric for assessing energy utilization efficiency, and its diminution signifies a decrease in the energy dependence of economic activities, hence facilitating the decoupling of energy consumption from economic growth. The energy structure demonstrates a notable beneficial impact, with coefficients of 0.119, 0.109, and 0.112 across the three weighting matrices, all significant at the 1% level. This indicates that for each 1% rise in coal’s proportion of energy consumption, the energy consumption decoupling index will grow by approximately 0.11%. An increase in coal’s share indicates a high level of carbonization in the energy system, which hinders the decoupling of energy consumption from economic growth. This outcome underscores the need of optimizing the energy framework and diminishing reliance on coal to facilitate the decoupling of energy usage.

The GDP per capita demonstrates a substantial negative impact on the energy consumption decoupling index, with coefficients of −0.055, −0.058, and − 0.057 across the three weighting matrices, all significant at the 1% level. This suggests that heightened economic development facilitates the decoupling of energy use from economic growth. As GDP per capita increases, the economic structure becomes optimized, resulting in a higher share of high value-added and low-energy-consuming industries, enhanced energy efficiency, and a less reliance on energy consumption. Population density demonstrates a notable negative effect on the energy consumption decoupling index, with coefficients of −0.113, −0.090, and − 0.123 across the three weighting matrices, all of which are statistically significant. This indicates that a rise in population density facilitates the decoupling of energy use from economic growth. Areas with high population density typically exhibit superior infrastructure and public services, benefiting from substantial economies of scale and agglomeration effects that enhance energy use efficiency and diminish per capita energy consumption. Additional variables, including the magnitude of capital investment and the degree of external openness, exert no substantial influence on the energy consumption decoupling index.

Spillover effects denote the influence of explanatory variables in one location on the energy decoupling index in adjacent regions. Regarding spillover effects, energy intensity significantly influences local energy consumption decoupling and exerts a notable negative spillover effect on adjacent regions. The coefficients for the geographical spillover effect are − 0.237, −0.035, and − 0.042 across the three weight matrices, all of which are significant at the 1% level. This indicates that a rise in energy intensity within a region impedes the decoupling of local energy consumption while simultaneously facilitating the decoupling of energy consumption in adjacent regions. This ostensibly paradoxical phenomenon may arise from interregional industrial relocation and resource redistribution, wherein energy-intensive industries migrate from energy-efficient areas to those with greater energy intensity, resulting in a disparity in the decoupling of interregional energy consumption. The spatial weight matrix of economic distance reveals a spatial spillover effect of energy structure is −0.260, which is statistically significant, showing that an increase in coal consumption share in one region facilitates the decoupling of energy consumption in adjacent regions. This may result from the concentration of coal usage in certain locations, prompting other regions to transition to cleaner energy sources, thereby enhancing the overall regional energy framework.

The degree of openness to the external environment demonstrates a notable positive geographical spillover effect under the diminished spatial weight matrices of economic and geographic distance, with coefficients of 0.035 and 0.006, respectively. This indicates that heightened sensitivity to external influences in a region may impede the decoupling of energy use in adjacent areas. Industrial agglomeration and the expansion of economic activity resulting from liberalization may elevate overall regional energy demand, hence increasing surrounding regions’ dependence on energy consumption during economic development. The spatial spillover impact of population density, as measured by the binary neighboring spatial weight matrix, is −0.128, which is statistically significant, suggesting that an increase in population density in one location facilitates the decoupling of energy consumption in adjacent regions. This may result from the urban agglomeration effect in densely populated regions, which enhances overall regional energy use efficiency via industry specialization and resource sharing. Regarding the intensity of funding input, while its primary effect is not significant across all three weighting matrices, it demonstrates a significant negative spatial spillover effect under the binary adjacency and geographic distance attenuation spatial weighting matrices, with coefficients of −0.024 and − 0.012, respectively. This indicates that an augmentation of R&D inputs in a region facilitates the decoupling of energy consumption in adjacent areas, likely attributable to technological innovation and knowledge spillover effects that enhance total regional energy efficiency.

Endogeneity tests

The endogeneity issue emerges due to the potential bi-directional causality between energy intensity (ei) and the energy consumption decoupling index (T), meaning alterations in the decoupling index may potentially influence energy intensity. This research employs the instrumental variable methodology to tackle this issue. The energy intensity of each city lagged by one period (ei_lag1) is chosen as the instrumental variable for energy intensity. This variable is strongly correlated with the current period’s energy intensity but is theoretically uncorrelated with the error term of the decoupling index, thereby fulfilling the correlation and exogeneity criteria for instrumental variables. Table 11 displays the primary outcomes of the instrumental variable estimation. The initial regression results indicate a significant correlation between lagged energy intensity and current energy intensity at the 1% level, with the F-statistic substantially exceeding the empirical critical value of 10, implying the absence of a weak instrumental variable issue. This work employs two ways to test for endogeneity: the first is the augmented regression method, which incorporates the residuals from the first-stage regression into the original model. The second is a straightforward endogeneity assessment with the ivreg2 command. Both methodologies indicate a substantial endogeneity issue for energy intensity.

Table 11.

Endogeneity test results.

Variables	OLS	IV-2SLS
lnrd	−0.001	0.000
	(0.002)	(0.003)
lnei	0.372***	0.047***
	(0.004)	(0.006)
lnes	0.112***	0.166***
	(0.011)	(0.016)
lnpgdp	−0.057***	0.000
	(0.010)	(0.009)
lnpi	−0.113***	−0.037***
	(0.042)	(0.005)
lnopen	−0.001	−0.206***
	(0.002)	(0.018)
First stage F-statistic		16582.010***
Chi-sq		402.830***
N	3920	3920
R-squared	0.868	0.335

***, **, * denote significance at the 1%, 5%, and 10% levels, respectively.

The results of the instrumental variable two-stage least squares (2SLS) estimation are generally congruent with those of the baseline model presented in Table 4. Upon adjusting for endogeneity, the positive influence of energy intensity on the decoupling index persists significantly at 0.047, while the positive impact of energy structure at 0.166 and the negative effect of population density at −0.037 also remain substantial. The degree of openness to the external environment exhibits a notable negative impact on the decoupling index (−0.206), although the effects of GDP per capita and financing intensity are not significant. The results demonstrate that the principal conclusions of this study regarding the elements affecting the decoupling of energy consumption are steadfast and dependable, notwithstanding the endogeneity issue.

Robustness analysis

This section performs robustness tests from two perspectives—sample time period division and variable measurement methods—to validate the dependability of the empirical results presented in this study.

Robustness tests for sample time period segmentation

This study separates the sample into two sub-samples, 2006–2010 and 2011–2021, and conducts separate regression analysis, considering the implementation of China’s 12th Five-Year Plan in 2011 and the potential significant changes in energy policy and economic development methods. Table 12 illustrates variations in the regression outcomes across the two time periods; however, the influence of the key variables stays consistently stable. Energy intensity exerts a notable positive influence on the energy consumption decoupling index across both time periods, with the coefficient in the first period (0.461) marginally exceeding that of the second period (0.367), suggesting a diminished yet still significant effect of energy intensity on the decoupling index. The impact of GDP per capita transitions markedly from a substantial positive effect in the initial period (0.053) to a considerable negative effect in the subsequent period (−0.065), potentially indicative of the positive outcomes of the energy dual-control policy and the economic structural transformation policy during the 12th Five-Year Plan, as the elevation in economic development commences to facilitate decoupling. This alteration may indicate the positive impacts of the energy dual-control policy and the economic structural reform strategy implemented during the 12th Five-Year Plan, with the rise in economic development levels commencing to facilitate decoupling. Notwithstanding fluctuations in the significance of individual variables, the principal conclusions remain consistent during the sample period, thereby affirming their reliability.

Table 12.

Regression results for different time periods.

Variable	2006–2010		2011–2021
	p-value	t-value	p-value	t-value
lnrd	−0.073*	−1.85	0.008	0.82
lnei	0.461***	13.55	0.367***	20.84
lnes	0.034	0.42	0.026	0.62
lnpgdp	0.053*	1.82	−0.065***	−3.07
lnpi	0.088	0.34	0.041	0.56
lnopen	0.005	0.68	0.002	0.58
constant	2.549*	1.83	2.856***	6.91
within R²	0.513		0.873
N	1120		3080

***, **, * denote significance at the 1%, 5%, and 10% levels, respectively.

Robustness tests for the way variables are measured

The study re-evaluates the model to assess the sensitivity of the results based on variable measurement, utilizing newly afforested land (green) as a proxy for energy intensity (ei) and industrial structure (struc) as a proxy for per capita GDP (pgdp). Table 13 presents the regression outcomes following variable substitution, indicating a decline in model fit; however, the directional effects of the important variables remain aligned with the baseline model. The positive impact of newly afforested areas on the decoupling index aligns with the original energy intensity variable. The optimization of industrial structure greatly aids in decoupling, aligning with the theoretical expectation that transitioning to service and high-tech manufacturing diminishes energy reliance. The effect coefficient of energy structure (es) in the proxy variable model (0.229) exceeds that of the baseline model (0.084), demonstrating the robustness of the energy structure’s role when accounting for various variable combinations. Energy intensity and energy structure are the principal elements affecting the decoupling of energy consumption from economic growth, in both temporal segmentation and variable replacement, hence affirming the robustness of the study’s conclusions.

Table 13.

Regression results after variable substitution.

Variable	Baseline Model		Substitution Variable Model
	p-value	t-value	p-value	t-value
lnrd	−0.007**	−2.33	0.035***	6.32
lnei	0.352***	109.38
lngreen			0.108***	6.61
lnes	0.084***	7.72	0.229***	11.05
lnpgdp	0.009**	2.00
lnsturc			−0.460***	−20.98
lnpi	−0.052	−1.28	0.107	1.43
lnopen	−0.002	−0.98	−0.020***	−5.62
constant	2.744***	12.32	1.903***	4.29
within R²	0.796		0.241
N	4200		4200

***, **, * denote significance at the 1%, 5%, and 10% levels, respectively.

Examination of findings and suggestions

Discussion

This article explores the decoupling of energy consumption from economic growth in Chinese cities, emphasizing the transition towards green, low-carbon practices. It identifies a robust correlation between energy use and economic development, while also revealing a non-linear relationship between these two variables. Economic expansion does not inherently result in increased energy consumption in the near term; nonetheless, a positive link may exist between the two in the long run. This aligns with the findings of the majority of current investigations by Khan⁷⁰. Simultaneously, many cities may exhibit disparate decoupling statuses despite comparable developmental stages, with much of this variance likely arising from the variability of the city’s economic structure. This aligns with the current findings of Shan et al.⁷¹. Cities characterized by high-tech and modern service sectors, such as Shenzhen and Hangzhou, typically exhibit superior decoupling, evidenced by reduced energy consumption per unit of GDP and diminished reliance on energy consumption for economic growth. Conversely, cities characterized by substantial, energy-intensive industries (e.g., Anshan, Benxi, etc.) encounter more challenges in attaining successful decoupling, with their economic expansion frequently coinciding with or outpacing increases in energy consumption. Secondly, regarding the classification of decoupling states, this study utilizes the Tapio decoupling model to categorize these states into eight distinct types: strong decoupling, weak decoupling, and recessionary decoupling, among others. This classification aligns with the research model established by Dong et al.⁷².

Further, regarding the exploration of influencing factors, this paper and some scholars have focused on the influence of energy intensity, energy structure, and level of economic development on energy consumption decoupling, and concluded that these factors have a positive or negative influence on decoupling^26,73. Within a specified spatial weight matrix, energy structure, funding intensity, GDP per capita, and population density significantly negatively impact the energy consumption decoupling index of neighboring regions, while the degree of external openness exhibits a significant positive effect. This study highlights the city level in China, distinguishing itself from the majority of previous research by Xia et al.⁷⁴, which concentrates on the provincial or national level. City-level analysis can clarify developmental discrepancies among areas and provide local governments with insights for creating more focused decoupling initiatives. This study examines energy consumption decoupling not only in its static state but also in terms of its spatio-temporal evolution, elucidating the dynamic variations of the decoupling state across different periods and regions. This parallels the findings of Li et al.³⁴, who examined the spatio-temporal patterns of energy consumption among urban inhabitants in China and the mechanisms influencing this phenomenon.

A significant shortcoming of this study must be emphasized: the omission of clean energy sources, including renewable and nuclear energy, from the energy consumption metrics. This decision, influenced by data availability, may lead to an underestimation of real decoupling progress regarding the “dual-carbon” aim and could impact the usefulness of policy recommendations. The swift advancement of renewable energy is a crucial factor in dissociating energy consumption from economic growth, and the omission of clean energy may not accurately represent the realities of the energy transition in many regions. Future study ought to incorporate renewable energy inside the analytical framework to establish a more thorough decoupling analysis method.

The findings indicate that energy has historically been a catalyst for economic expansion; however, it also incurs environmental costs and security difficulties. Addressing these difficulties necessitates examining the influence of economic structure on energy efficiency to reconcile economic development with energy consumption⁷⁵. Green policies are essential to encourage the utilization of non-fossil energy sources³⁶. Energy efficiency facilitates the separation of economic growth from energy consumption, while energy-related innovations incentivize the reduction of energy intensity. China, as the greatest developing nation, has seen rapid economic expansion, resulting in a significant increase in energy consumption that profoundly affects the global energy framework and environmental conservation⁷⁶. Decoupling the economy from energy to reform the economic growth paradigm and foster high-quality development has emerged as China’s current strategic priority⁷⁷.

This study initiates from the viewpoint of energy consumption to investigate the decoupling effect and the variables impacting urban energy consumption and economic growth. However, it is important to note that future research may be expanded in at least three distinct areas: ①Future research could implement advanced data processing techniques, such as machine learning algorithms for predicting missing data, and develop a more extensive database on urban energy consumption that encompasses renewable and nuclear energy, thereby enhancing the reliability and policy relevance of the study’s findings. ②Alternative research methodologies are available. A semi parametric global vector autoregressive model may be employed to investigate the effects of the decoupling of energy consumption from economic development in China. ③Regarding research drivers, various facets of the indicators are integrated into the analysis of influencing factors while employing more precise and sophisticated models to examine their effect on the energy consumption decoupling index, thereby elucidating the underlying dynamics and motivations influencing them.

Findings

This study employs the Tapio decoupling index model to evaluate the decoupling index of urban energy consumption across two temporal baseline periods. It further investigates the spatio-temporal evolution characteristics of urban energy consumption decoupling. Building on this analysis, we developed a spatial Durbin model within the expanded STIRPAT model to explore the determinants influencing the decoupling of energy consumption. The key findings are as follows:

1. The decoupling state of energy consumption exhibits variations and oscillations across various locations. Chinese cities often have a strong correlation between energy consumption and economic growth, with the tendency toward negative growth decoupling state becoming more pronounced. Numerous cities in the eastern and central regions predominantly exhibit weak decoupling. Conversely, several communities in the western region predominantly exhibit negative growth decoupling and exacerbate their energy reliance throughout economic expansion. Between the 11th and 14th Five-Year Plan period, a growing number of cities exhibited negative growth decoupling, while only a few cities managed to sustain weak decoupling at various intervals.

2. Regarding energy consumption decoupling stability, the majority of cities in China have modest decoupling intensity coefficients, indicating a generally stable decoupling state, but several regional extremes exist. The Northeast region excels overall, although the West has significant obstacles, and the Central and Eastern regions are average, including cities requiring attention, such as Zibo, Linfen, Lu’an, and Shiyan.

3. The spatial main effect indicates that energy intensity and energy structure positively influence the energy consumption decoupling index, while GDP per capita and population density negatively affect it. The significance of this main effect underscores the necessity of optimizing energy utilization and advancing the transformation of the energy structure.

4. Regarding spatial spillover effects, the city’s energy intensity, energy structure, GDP per capita, and population density exert a significant negative influence on the energy consumption decoupling index of adjacent regions across various weight matrices, whereas the degree of external openness demonstrates a notable positive effect under specific weight matrices. Moreover, the magnitude of financing inputs had a notable adverse effect on the energy consumption decoupling index of adjacent regions under specific weighting matrices.

Recommendations for policy

The examination of the decoupling effect between energy consumption and economic growth in Chinese cities is crucial for China’s development, and the subsequent remedies are offered based on the study’s findings:

Initially, to effectively address the regional disparities in the decoupling of energy consumption from economic growth across Chinese cities, it is crucial to develop targeted energy management strategies. The eastern and central regions should prioritize accelerating their decoupling efforts to minimize the strong correlations between energy consumption and economic growth. Meanwhile, the western region must focus on reducing energy dependency and enhancing the decoupling of energy use from economic activities. Additionally, the successful initiatives of cities that have consistently achieved effective decoupling should be documented as exemplary case studies, serving as models for replication and broader application.

Furthermore, dedicated finances and technological support mechanisms must be created. The acceleration of special funds for decoupling development is essential to assist cities in implementing energy conservation, emission reduction, clean energy initiatives, and energy structure transformation projects, thereby ensuring financial security for the decoupling process. Conversely, for cities exhibiting inadequate decoupling stability, such as Zibo and Linfen, a technical support platform will be created to consolidate advanced energy-saving and environmental protection technologies and equipment from both domestic and international sources, offering tailored technical solutions for these municipalities. The technical support platform will conduct activities like technical training and consulting services to enhance awareness of environmental protection in the region and develop professional and technical expertise, ensuring human resource support for the decoupling process.

Enhance energy efficiency within the current energy framework. Research and analysis of the spatial main effect indicate that altering the existing energy consumption structure of Chinese cities within a short timeframe is challenging; however, enhancing energy utilization efficiency can facilitate a reduction in carbon emissions during energy use. To enhance regional energy efficiency, firms must increase investments in emission reduction, while the government should implement suitable incentives and punishments to support the program. For firms with developmental potential, government agencies could provide preferred policies to mitigate the inefficiencies and poor output associated with initial emission reduction efforts. Enterprises with outdated production capacities must actively respond to government directives aimed at eliminating these inefficiencies and participate in supply-side structural reforms, thereby facilitating the transformation and optimization of their industrial frameworks. Furthermore, the government should acknowledge the negative impacts of GDP per capita and population density on the energy consumption decoupling index. By implementing strategic urban development planning, it can optimize population and industrial distribution, ultimately achieving effective decoupling of energy consumption from economic growth.

Fourth, it is crucial to strengthen frameworks for regional collaboration and exchange. Research on spatial spillover effects underscores the necessity for government entities to bolster interregional cooperation concerning energy intensity, energy structure, and population density. This collaborative effort is vital for advancing the decoupling of energy consumption across China’s cities, enhancing energy efficiency, and optimizing energy structures. Additionally, attention must be paid to the impact of external openness and capital investment on the decoupling index of energy consumption in neighboring regions. By promoting external openness and increasing capital investment, we can further facilitate the decoupling of energy consumption.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Author Contributions: Conceptualization, H.C., M.H.; methodology, H.C., M.H.; software, C.L.; validation, C.L.; formal analysis, C.L.; investigation, C.L.; resources, H.C.; data gathering, C.L.; writing—original draft preparation, C.L.; writing—review and editing, H.C., M.H.; visualization, C.L.; supervision, H.C., M.H.; project administration, H.C., M.H.; funding acquisition, H.C. All the authors have read and agreed to the published version of the manuscript.

Funding

Guangxi Philosophy and Social Science Planning Project "Mechanisms and Pathways of Data Elements Empowering Urban-Rural Integrated Development in Guangxi" (Grant number: 24JYB002);

Guangxi Philosophy and Social Science Planning Project "Research on Job Creation in Guangxi’s Counties and Prevention of Large-Scale Poverty Reversion" (Grant number: 24JYF001);Guangxi Graduate Education Innovation Program "Reform of Interdisciplinary and Collaborative Postgraduate Training Models in the Context of New Liberal Arts" (Grant number: JGY202223);

Innovation Project of Guangxi Graduate Education “Research on the Mechanisms and Pathways for Cultivating Innovative Abilities of Management Graduate Students in Guangxi under the Science and Education Empowerment Model”, (Grant number: JGY2024251);

Innovation Project of Guangxi Graduate Education “Reform and Practice of Artificial Intelligence Empowerment for High-Quality Development of Public Administration Graduate Education”, (Grant number: JGY2024261).

Data availability

All data generated or analysed during this study are included in this published article [and its supplementary information files].

Declarations

Competing interests

The authors declare no competing interests.