Introduction

One of the fundamental requirements to achieve sustainable development encompassing economic, social, and environmental aspects is the reduction of carbon emissions and active engagement in combatting climate change (United Nations, 2015). However, global circumstances significantly impact the progress of worldwide sustainable development. For instance, the Russian-Ukrainian conflict has notably influenced the trajectory of sustainable development across nations. An illustrative example can be found in the World Energy Outlook 2022 (International Energy Agency, 2022) issued by the International Energy Agency (IEA) on October 13, 2022. That report concludes that for numerous countries, the mitigation of carbon emissions is no longer an immediate policy priority, signaling a shift in their short-term policy focus away from carbon emissions reduction. For instance, Germany has raised its coal-fired power generation and dropped its goal of attaining carbon neutrality by 2035. In addition, the conflict has escalated the unstable political environment, resulting in carbon emissions from machinery like airplanes, ships, tanks, and more. These incidents have heightened public apprehension about climate change. It remains crucial to explore and execute measures to curtail emissions in tackling the climate crisis. The present socio-economic system, driven by carbon-intensive economic growth and unsustainable resource use, has led to the emergence of a “greenhouse earth” (Speth and Zinn, 2008; Steffen et al. 2018). The digital economy offers a fresh approach to economic progress, warranting investigation for its potential to support sustainable, zero-carbon, and inclusive growth. Scholars and governments have begun to recognize the link between the digital economy and carbon emissions. Developing the digital economy establishes a closed-loop system for gathering and providing feedback on green development information, enhancing energy efficiency, and refining carbon emission monitoring and management (Wu and Zhu, 2022). Digitalization appears to be a central tool in decoupling economic growth from carbon emissions (Xu et al. 2022). Yet, there’s evidence that the digital economy can result in substantial carbon emissions. According to calculations by the Open Data Center Committee, China’s data centers consumed 93.9 billion kilowatt hours of energy and emitted 64.64 million tons of carbon in 2020 (ODCC, 2022). Projections suggest that by 2030, China’s data center’s total energy consumption could skyrocket to about 380 billion kilowatt hours, with carbon emissions surging by over 300% (ODCC, 2022).

Academics have extensively and somewhat successfully investigated how the digital economy influences carbon emissions. However, there are still the following shortcomings: the academic community holds varying conclusions regarding the relationship between the digital economy and carbon emissions (Romm, 2002; Alam and Murad, 2020; Ma et al. 2022; Hao et al. 2022; Wang et al. 2022b; Wu et al. 2023; Zhang et al. 2022; Cheng et al. 2023; Li and Wang, 2022; Xiang et al. 2022). Second, scholars mainly measure the impact of the digital economy on carbon emissions from three aspects: total carbon emissions (Hao et al. 2022; Wang et al. 2022b; Wu et al. 2023; Zhang et al. 2022), intensity (Wang et al. 2022a; Wang and Zhong, 2023; Yan et al. 2022; Zhong et al. 2022), and efficiency (Lyu et al. 2023), but few kinds of literature have studied the digital economy and carbon emissions decoupling based on decoupling theory. Third, existing research on carbon decoupling and the digital economy uses provincial data in China (Lu and Chen, 2022; Zhong et al. 2022), while ignoring urban data. Existing research has ignored the division of decoupling indices in good and bad order (Lu and Chen, 2022; Yu et al. 2022). Fourth, they ignored the spatial effects of the digital economy on carbon decoupling (Lu and Chen, 2022; Yu et al. 2022; Zhong et al. 2022). Hence, this study utilizes data from 281 cities in China spanning from 2012 to 2019 to establish a comprehensive indicator evaluation system for assessing the level of digital economy development. Building upon this, we used ArcGIS software and analyzed the spatial distribution and agglomeration of the digital economy and carbon emissions decoupling. In addition, we examined the direct effects, heterogeneity, and spatial impacts of the digital economy on carbon emissions decoupling. Excessive carbon emissions resulting from human activities exemplify negative external economies. The digital economy, with its technological and knowledge spillovers, serves as a prime instance of spillover effects. Hence, this article investigates the spatial characteristics and connection of the digital economy and decoupling carbon emissions, adopting a fresh spatial perspective that unveils their intrinsic relationship more effectively. Investigating this matter can offer a fresh perspective and valuable insights for tackling carbon emissions.

This paper contributes in several ways. Firstly, we utilize Tapio’s decoupling theory to categorize carbon emissions decoupling into different levels. This method enhances accuracy in assessing the digital economy’s impact on carbon emissions and economic growth, improving our understanding of their sustainable relationship. Secondly, we analyze data from 281 Chinese cities between 2012 and 2019 to investigate the link between the digital economy and carbon emissions decoupling, a topic not previously studied. Our city-level data provides more reliable insights into urban structures compared to provincial data. Given that cities contribute to 75% of anthropogenic greenhouse gas emissions, their role in achieving carbon peaking and neutrality is crucial. Thus, examining factors influencing urban emission reduction becomes practically significant. Thirdly, by applying Tobler’s first law of geography, we examine the spatial distribution and clustering of the digital economy and carbon emissions decoupling. We also explore how the digital economy affects nearby regions in terms of carbon decoupling, identifying whether it has a siphon or spillover effect. This approach yields unique insights into the spatial patterns of digital economy-driven carbon emissions decoupling, which aids discussions on sustainable urbanization.

We have found disparities in the distribution of the digital economy and urban carbon decoupling across China. Both exhibit positive spatial correlation and clustering. Moreover, we have noted that the influence of the digital economy on urban carbon decoupling hinges on the city’s types of resources and geographical location. Strikingly, our study has revealed that the digital economy exacerbates the urban carbon decoupling situation in neighboring regions through the siphoning effect.

The structure is as follows: Part two comprises a literature review, and part three presents a research hypothesis. Part four includes the dataset, sample set, and model set. Part five analyzes the empirical results while part six concludes and discusses the findings.

Literature review

Literature review of the digital economy and carbon emission

Scholars’ research on the relationship between the digital economy and carbon emissions mainly focuses on the following content: First, some scholars believe that the digital industry belongs to the industry with low energy density and low energy consumption increment (Romm, 2002). The dematerialization effect, decarbonization, and demobilization of the digital economy inhibit consumption-based carbon emissions (Alam and Murad, 2020; Ma et al. 2022). The digital economy reduces carbon emissions through industrial progress and energy consumption optimization, environmental governance, and innovation (Hao et al. 2022; Wang et al. 2022b; Wu et al. 2023; Zhang et al. 2022); it reduces the intensity of carbon emissions by improving innovation, increasing the proportion of clean energy consumption, upgrading industrial structure, improving energy utilization efficiency, and improving environmental regulation efficiency, and total factor productivity (Wang et al. 2022a; Wang and Zhong, 2023; Yan et al. 2022; Zhong et al. 2022); improves carbon emission efficiency by reducing energy consumption (Lyu et al. 2023), suppressing factor mismatch (Ge et al. 2022), technological innovation, reducing energy consumption intensity, and improving urban productivity (Zhao et al. 2022). The digital economy can reduce carbon emissions through spatial effects (Wang et al. 2022; Yi et al. 2022). Second, some scholars believe the digital economy increases carbon emissions (Dong et al. 2022; Zhang et al. 2022). The main reasons are that the digital economy has increased the demand for energy-intensive computers (Ma et al. 2022), and their use and disposal have increased the pressure on the environment (Houghton, 2015; Mickoleit, 2010). The surge in the number of digital centers has increased energy consumption and pollution (Jahangir et al. 2021). Third, some scholars also believe the digital economy and carbon emissions present an inverted U-shaped (Cheng et al. 2023; Li and Wang, 2022; Xiang et al. 2022), N-shaped (Zhang et al. 2022), inverted N-shaped (Hao et al. 2022), and other nonlinear relations (Wu et al. 2022).

In summary, scholars currently evaluate the impact of the digital economy on carbon emissions primarily through total emissions, intensity, and efficiency. However, a consensus on this issue has not been reached yet. There is a lack of literature utilizing decoupling theory to examine the relationship between the digital economy and carbon emissions. Previous research on carbon decoupling and the digital economy has predominantly employed provincial data from China (Lu and Chen, 2022; Zhong et al. 2022) and has neglected urban data. No literature has explored the spatial effects of the digital economy on carbon emissions decoupling. Using decoupling theory, this article analyzes the direct and spatial impact of the digital economy on carbon emissions decoupling in 281 Chinese cities, addressing the limitations of previous research.

Literature review of the decoupling theory

The concept of decoupling originates from physics, where it signifies the loss of correlation between two or more physical quantities. The OECD first introduced the decoupling theory to describe the disconnection between economic growth and resource consumption or environmental pollution, i.e., the decoupling of economic growth from these factors. In 2002, the OECD devised the decoupling index and decoupling factor to compare changes between the final and base years to clarify whether there was any decoupling in the economic system during that period. If the decoupling factor is positive and nearly 1, it is absolute decoupling, depicting environmental variables associated with economic development remaining stable or decreasing. If the decoupling factor is positive and near 0, it is relative decoupling, where both economic growth and environmental variables change positively, but the rate of change of the latter is lower. If the decoupling factor is 0 or negative, there is no decoupling. Based on the OECD measurement method, Hasan Rüstemoğlu concluded that Germany achieved an absolute decoupling of its GDP and carbon emissions from 1990 to 2015 (Rüstemoğlu, 2019). However, Tapio argues that this classification oversimplifies complex changes and tends to overstate some weaker ones. To address this limitation, Tapio introduced Tapio’s decoupling indicator system, which classifies decoupling into three main categories and eight subcategories, illustrated in Fig. 1. Tapio’s decoupling indicator is valued for its comprehensive nature and its independence from a base period selection. Consequently, scholars extensively employ this indicator to analyze the carbon emissions decoupling (de Freitas and Kaneko, 2011; Jiang and Li, 2017; Chen et al. 2023; Gao et al. 2021; Huang and Guo, 2022; Huang et al. 2022; Huo et al. 2021; Wang and Han, 2021; Zhao and Li, 2018). Previous researchers have utilized Tapio’s decoupling indicator to examine the decoupling of carbon emissions in various regions and industries (Amir et al. 2023; Chen et al. 2023; Dong et al. 2021; Hua et al. 2023; Jiang et al. 2018; Huo et al. 2021; Liu et al. 2022; Wei et al. 2022). They have also analyzed the impact of factors such as financial development, household income, Sino-US trade, urbanization, carbon emission trading policies, labor and investment, urbanization, industrialization, and trade openness on carbon emissions decoupling (Huo et al. 2021, Wang and Han, 2021; Duan et al. 2022; Huang and Guo, 2022; Lyu et al. 2022; Wang and Jiang, 2020; Wang et al. 2018; Wang et al. 2023; Wang and Zhang, 2021).

Fig. 1: Spatial distribution.
figure 1

a Spatial distribution of urban digital economy in 2019; b The spatial distribution of urban carbon decoupling in 2019.

Full size image

In summary, there’s limited literature on the link between the digital economy and carbon emissions decoupling. Some scholars, based on provincial data in China, propose that the digital economy initially aids decoupling, yet this effect weakens over time due to industrial structure mediation and positive regulation from network centrality. However, existing studies lack quantity, depth, and comprehensiveness, and often omit city-level data. They also overlook Tapio’s decoupling indicator classification. The spatial impact of the digital economy on carbon decoupling is underexplored, despite the rise of new economic geography. Carbon emission involves externality, necessitating spatial considerations. Scholars employ spatial econometric models to study specific factors’ spatial correlation with carbon emissions (Li and Fan, 2023; Lin et al. 2023; Liu et al. 2023; Long et al. 2020). Nonetheless, the spatial impact of the digital economy on carbon emissions decoupling lacks research.

Research hypothesis

Digital economy and carbon decoupling

This article believes that the digital economy will promote the decoupling of urban economic growth and carbon emissions from four aspects. From a technological progress standpoint, the digital economy has the potential to drive green technological advancements and enhance pollution control measures (Wurlod and Noailly, 2018). Following Davidow’s law, initial-generation products can automatically capture 50% of the market share upon entering the market. To expand their market presence, enterprises increase their investment in innovation, leading to higher innovation output. Notably, green innovation and environmental innovation are crucial outcomes of innovation and play a significant role in achieving emission reduction (Liu et al. 2022; Zhang et al. 2021; Zhang et al. 2017). Hence, this study asserts that the digital economy can effectively reduce emissions and facilitate the decoupling of urban carbon emissions by fostering advancements in green innovation.

From an energy standpoint, the digital economy has the potential to lower energy intensity, decrease the costs associated with renewable energy utilization (Moyer and Hughes, 2012), increase the share of non-fossil energy sources, and optimize industrial structure (Wang et al. 2021). Furthermore, it can enhance the efficiency of energy allocation across the upstream and downstream sectors (Yan et al. 2016), reduce energy intensity, and foster a conducive environment for green growth (Ulucak, 2020). As a result, this study posits that the digital economy can effectively reduce the intensity of urban carbon emissions and facilitate decoupling by curtailing energy consumption, improving energy efficiency, and promoting the transformation and upgrade of energy structures.

From an economic growth standpoint, scholars argue that the advancement of the digital economy facilitates the alignment of product and service supply and demand, lowers transaction costs (Romm, 2002), and can result in more efficient production methods and enhanced productivity (Moyer and Hughes, 2012). In addition, the digital economy promotes dematerialization effects (Liang, 2021), thereby reducing material production and consumption and curbing the demand for high-carbon-emitting products, both directly and indirectly (Akimoto et al. 2022). We think that the digital economy has improved production efficiency and reduced information asymmetry through big data. Through accurate prediction of market demand, enterprises have made accurate production and sales and improved the matching efficiency between supply and demand. Meanwhile, the digital economy promotes the growth rate of marginal returns of various factors by adding new data as the production factor and reforming and restructuring the traditional production factors, which amplifies and superimposes economic growth. Therefore, the digital economy promotes urban carbon decoupling through economic growth effect and efficient use of resources.

From a regulation perspective, smartphone applications can enhance the accountability of public agricultural extension services by enabling real-time remote supervision. Consequently, this leads to a reduction in both the cost and time needed for regulation (Namyenya et al. 2022). Therefore, digital technology can integrate various social entities, including platforms, businesses, consumers, and governments, into the carbon emission governance system. This expansion enhances the scope, methods, and scenarios of environmental oversight, thereby contributing to the development of a collaborative, multi-agent framework for green governance. Simultaneously, the digital economy aids governments in establishing integrated ecological environment perception systems and expediting the implementation of intelligent monitoring systems for carbon emissions.

Based on the above analysis, we put forward the research hypothesis:

H1: The digital economy can promote the decoupling of urban economic growth and carbon emissions.

Digital economy, space effect and carbon decoupling

Spatial effects include the spillover effect and the siphoning effect. The spillover effect means something drives the development of things around it. In economics, the siphoning effect refers to a city with an advantageous position that can attract various resources from a city with a relatively inferior position. The siphoning effect may cause cities with advantages to be more competitive while cities with relative disadvantages are more disadvantaged. Digitalization and carbon emissions have significant spatial spillover effects (Fang et al. 2022; Liu et al. 2021). However, no scholars have examined whether the digital economy affects the decoupling of neighborhood carbon emissions from economic growth. This article argues that the digital economy can have both spillover and siphon effects on carbon emissions decoupling in neighborhoods.

First, the digital economy facilitates decoupling through spillover effects. It helps establish cross-regional industrial chains and industry-university research platforms (Yang et al. 2021; Fang et al. 2022). Technological spillovers from the industrial chain promote the green transformation of neighborhoods (Handfield et al. 2005). In addition, industry-university research platforms contribute to the adoption of green technology across regions (Fang et al. 2022). Thus, this article concludes that the digital economy generates technology and learning spillovers, actively promoting the decoupling of carbon emissions in neighborhoods.

In addition, it is often overlooked in the existing literature that the digital economy can also hinder the decoupling of carbon emissions in neighborhoods through the siphoning effect. The digital economy facilitates the construction of digital infrastructure and urban management, attracting high-quality resources from neighboring areas (Xu and Sun, 2021). According to Colin Clark, income differences primarily determine population mobility. In the context discussed in this article, the digital economy’s ability to bring efficient production methods and productivity improvements attracts high-quality talent to regions with robust digital economy development (Moyer and Hughes, 2012). As a result, the digital economy intensifies the concentration and convergence of high-quality resources (Luo et al. 2022), thereby enhancing the factor structure of cities (Duggal et al. 2007). However, it is precisely because regions with strong digital economy development lure high-quality resources from neighboring areas that a considerable amount of these valuable resources is depleted within the neighborhoods. This, in turn, hampers the sustainable development of the neighborhood’s economy and leads to increased carbon emissions. Thus, this article argues that the development of the digital economy can also suppress the decoupling of carbon emissions in neighborhoods through the siphoning effect.

Therefore, this article proposes the following competitive hypothesis:

H2: The digital economy deteriorates the surrounding carbon decoupling state through the siphoning effect.

H3: The digital economy improves the surrounding carbon decoupling through spillover effects.

Methods

Benchmark model setting

We set up the following linear regression model to test the relationship between the digital economy and carbon decoupling:

$$tapioc_{it} = \alpha _0 + \alpha _1dec_{it} + {\sum} {\alpha X_{it}} + u_i + v_t + \mu _{it}$$
(1)

In the above formula, i represents the city; t represents the year; tapiocit represents the state of urban carbon decoupling; decit represents the urban digital economy;α0, X, u, v and μ represent constant term, control variables, individual effect, time effect, and random error. To ensure the reliability of the results, we control the city-fixed and year-fixed effects and adopt robust standard error to mitigate the disturbance.

The setting of the spatial measurement model

We have established a spatial econometric economic model to explore the spatial effect of the digital economy on carbon decoupling:

$$\begin{array}{l}tapioc_{it} = \alpha _0 + \rho W{{{\mathrm{tapioc}}}}_{it} + \alpha _1dec_{it} + \varphi _1Wdec_{it} + \alpha _2X_{it}\\ + {\sum} {\varphi WX_{it}} + u_i + v_t + \mu _i\mu = \lambda W\mu + \varepsilon \end{array}$$
(2)

Among them, ρ and W represent the spatial autocorrelation coefficient and spatial weight matrix. φ1, φ, and λ are the coefficient of the interaction term between the core explanatory variable, control variable, random disturbance term, and the spatial weight matrix. Equation (2) is the spatial Dubin model (SDM) to examine the spatial effects. When the spatial Dubin model meets certain conditions, it can be simplified into the spatial lag model (SLM) and spatial error model (SEM).

Variable selection

Explained variable

Firstly, calculating total carbon emissions of cities (CO2), which can be divided into three components:

  1. (1)

    Carbon emissions from urban electricity (c1):

    Calculate these emissions by multiplying social electricity consumption by the baseline emission factor of the regional power grid.

  2. (2)

    Carbon emissions from urban thermal energy (c2):

    Urban thermal energy relies on steam and hot water heating, often from raw coal. Calculate emissions by multiplying total urban steam and hot water heating by the raw coal emission coefficient and thermal efficiency value, then dividing by the average low calorific value of raw coal. A thermal efficiency value of 70% is used.

  3. (3)

    Carbon emissions from urban manufactured gas, natural gas, and liquefied petroleum (c3):

$$C_3 = petroleum \times M + natural\_gas \times Q$$
(3)

In this case, petroleum and M represent the supply volume and carbon emission coefficient of artificial manufactured gas and natural gas, respectively, while natural_gas and Q represent the supply volume and carbon emission coefficient of liquefied petroleum, respectively. Calculating carbon emissions for each component allows us to determine the total urban carbon emissions (CO2).

Secondly, calculate the carbon decoupling index for each city using Tapio’s (2005) decoupling index system (Tapio, 2005).

$$T_{it} = \frac{{\Delta co_2^{it}/co_2^{i,t - 1}}}{{\Delta gdp^{it}/gdp^{i,t - 1}}} = \frac{{\left( {co_2^{it} - co_2^{i,t - 1}} \right)/co_2^{i,t - 1}}}{{\left( {gdp^{it} - gdp^{i,t - 1}} \right)/gdp^{i,t - 1}}}$$
(4)

In the formula provided, i represents city, and t represents time. According to Tapio’s classification, the decoupling status includes strong growth decoupling, weak growth decoupling, strong recession decoupling, weak recession decoupling, growth non-decoupling, recession non-decoupling, growth coupling, and recession coupling.

Next, when we consider the inclusion of the eight decoupling states in the explained variables, the econometric model becomes complex, and these states present pros and cons (Mi and Zhao, 2022). In growth decoupling, a smaller growth rate in carbon emissions signifies a more favorable urban decoupling status. Conversely, in recession-type decoupling, a greater reduction in carbon emissions corresponds to an improved city decoupling state. The most optimal scenario arises from robust urban economic growth coupled with decreased carbon emissions. On the contrary, recession-strong-decoupling indicates the least desirable circumstances. Thus, the ranking of decoupling states aligns as follows: weak growth decoupling > non-decoupling in growth > growth coupling, weak decline decoupling > non-decoupling in decline > decline coupling. With the increase (decrease) in positive (negative) economic growth rates, carbon emission growth rates decrease, forming a symmetrical optimization path for recession and growth decoupling. In conclusion, according to the optimal sequencing of decoupling, this paper re-induces five states and constructs the interpreted variable (tapioc). We divide the urban carbon decoupling state into the ideal state (strong growth decoupling, tapioc = 4), weak decoupling state (weak growth decoupling, recession weak decoupling, tapioc = 3), non-decoupling state (growth non-decoupling, recession non-decoupling, tapioc = 2), coupling state (growth coupling, recession coupling, tapioc = 1), and bad state (strong recession decoupling, tapioc = 0). To mitigate the influence of base period changes on results, this research employs data from the previous year as the base period for urban carbon decoupling calculation. Table 1 illustrates the diverse urban carbon emissions decoupling types and the structure of dependent variables in this study.

Table 1 Types of urban carbon emission decoupling.
Full size table

Core explanatory variables

The measurement methods of the digital economy mainly include the national economic accounting method, satellite account, index method, etc. The index method is real-time, flexible, innovative, and can systematically reflect the national or regional digital economy. So, we study the impact of the digital economy on urban carbon decoupling, which is more suitable for using the index method. Regarding Zhao et al. (2020), we measure the urban digital economy from the proportion of Internet users, mobile phone users, information transmission and technology services employment, the total number of telecommunications services per capita, and the urban digital inclusive financial index.

Control variables

Some scholars believe that technological innovation in the information industry will raise carbon emission intensity (Wang et al. 2021). Technology expenditure reflects the level of emphasis and support for technology, and general technology expenditure is positively correlated with regional technology levels. The IPAT model, STIRPAT extended model, and LMDI model all acknowledge the influence of population and technology on the environment (Dietz and Rosa, 1997). Hence, this article includes technology expenditure and regional population size as control variables.

Government expenditure can impact urban carbon emissions. According to some scholars, an increase in total government spending in the short term can result in higher carbon dioxide emissions (Galinato and Galinato, 2016). The overall impact of government expenditure on total carbon emissions is positive, as the negative direct impact is offset by positive indirect effects (Adewuyi, 2016). Therefore, this article selects government expenditure as a control variable.

Human capital has the potential to influence urban carbon emissions. Some scholars argue that enhancing human capital can contribute to the reduction of carbon emissions (Bano et al. 2018; Sheraz et al. 2021). ICT capital enhances carbon emission efficiency by facilitating the accumulation of human capital (Xu et al. 2022). In addition, the impact of human capital on carbon emissions differs before and after the turning point of the Environmental Kuznets Curve (EKC) (Wang et al. 2023). Hence, this article considers human capital as a control variable.

Finance can indeed impact the levels of urban carbon emissions (Ali et al. 2022; Zhang et al. 2022). Considering the influence of urban finance on the extent of decoupling urban carbon emissions, this article considers urban financial development as the control variable.

Based on this, the control variables selected in this paper are as follows: ① Science and technology expenditure(tech). This paper adopts the proportion of scientific expenditure in the general budget of local finance to measure urban scientific and technological expenditure. ② Government intervention (gov). This paper uses the proportion of expenditure in the general budget of local finance to GDP as a measure of government intervention. ③ Human capital (human). Measure human capital in cities by the proportion of education expenditure to GDP. ④ Population (lnp). We use the logarithm of the population at the end of the year to measure the size of the urban population. ⑤ Financial development (fa). We use the ratio of the balance of loans from financial institutions at the end of the year to the city’s current GDP to measure the financial development.

Data source

Social electricity consumption, artificially manufactured gas and natural gas supply, liquefied petroleum gas, the proportion of internet users, mobile phone users, information transmission and technology services deployment, the total number of telecommunications services per capita, science and technology expenditure, government intervention, human capital, population, financial development, etc., are obtained from the China Urban Statistical Yearbook. Heat data is sourced from the China Urban Construction Statistical Yearbook. The baseline emission factors of the regional power grid are obtained from the official website of the Ministry of Ecology and Environment. The average low calorific value and carbon emission coefficient of raw coal are derived from the Guidelines for the Preparation of Provincial Greenhouse Gas Inventories and the 2006 IPCC Guidelines for National Greenhouse Gas Inventories. We have chosen the time frame spanning from 2012 to 2019 for our analysis. Some incomplete data are processed by interpolation. Table 2 presents the descriptive statistics of all variables.

Table 2 Descriptive statistics of variables.
Full size table

Result

Spatial distribution and agglomeration

Space distribution

Figure 1a illustrates the predominant distribution of the low-level digital economy in the northeast and northwest regions, while the high-level digital economy is concentrated in the Southeast Coast and municipalities directly under the Central Government. In Fig. 1b, which displays the spatial arrangement of urban carbon decoupling in 2019 with 2018 as the base year, it becomes apparent that cities with unfavorable carbon decoupling status are primarily situated in Northeast China and North China. It is notable that the majority of cities currently do not exhibit decoupling.

Space agglomeration

Following the first law of Geo-economics, cities in closer proximity exhibit heightened mutual influence. Thus, we employ a normalized inverse distance spatial weight matrix to scrutinize spatial correlations. By utilizing Matlab-R2017a, we compute the inverse distance spatial weight matrix following travel standardization. As displayed in Table 3, the global Moran’s I index for urban carbon decoupling utilizing the inverse distance matrix exceeds 0 and is significant during 2013–2018. That signifies a positive spatial correlation in urban carbon decoupling across most years. In addition, Table 3 demonstrates that the global Moran’s I index for the urban digital economy surpasses 0 and is significant, indicating a favorable spatial correlation. Consequently, spatial analysis becomes crucial for examining the interaction between the digital economy and urban carbon decoupling.

Table 3 Global Moran Index.
Full size table

Next, this paper uses the LISA agglomeration diagram to analyze the agglomeration types of the digital economy and urban carbon decoupling. These patterns encompass four types: high-high (H-H), high-low (H-L), low-high hollow (L-H), and low-low (L-L) types (Wu et al. 2023).

Figure 2A shows that the H-H type concentration area of the digital economy mainly includes the Southeast Coast. The L-L type is distributed in Northeast China, Gansu, and Yunnan. Figure 2B shows the LISA cluster diagram of China’s urban carbon decoupling in 2019. The results show that the L-H type of urban carbon decoupling is mainly distributed in Gansu and Shanxi of China, and the L-L type is mainly concentrated in Shandong, Hebei, and Inner Mongolia Autonomous Region.

Fig. 2: LISA chart.
figure 2

A LISA chart of digital economy in 2019; B LISA chart of urban carbon decoupling in 2019.

Full size image

Benchmark regression results

In Column (1) of Table 4, the coefficient of the urban digital economy is 0.083 and significant at the 1% level. Columns (2) to (4) are the benchmark regression results after adding the control variables. In column (4), the coefficient of the digital economy is 0.074, which is statistically significant at the 1% level. Table 4 shows that the digital economy has improved urban carbon decoupling. Therefore, this paper verifies hypothesis 1 and posits that the digital economy has improved urban carbon decoupling by reducing information asymmetry, precision production, and sales, increasing output, promoting urban green development, and strengthening environmental governance.

Table 4 Benchmark regression results.
Full size table

Robust test

Add missing variables

We further test the robustness of the benchmark regression by increasing the logarithm of actual output (lnrgdp), foreign investment (fdi), urbanization level (urb), energy structure (es), and industrialization level (ind) that may affect urban carbon decoupling as control variables in the benchmark regression model. In Column (1) of Table 5, the coefficient of the digital economy is 0.076 and significant after adding control variables. The result means that the digital economy still improves the carbon decoupling status.

Table 5 Robustness test.
Full size table

Interaction fixed effect

The interactive fixed effects model fully considers multidimensional shocks and different reactions, which can better reflect the authenticity of specific problems. Therefore, we next use the interaction fixed effect of individuals and time to test robustness. After adding the interactive fixed effect, column (2) in Table 5 shows the coefficient of the digital economy is 0.069 and significant at the 5% level. This finding suggests that the digital economy can significantly improve the urban carbon decoupling status, which verifies hypothesis 1.

Excluding municipalities directly under the Central Government

Municipalities under the Central Government generally have large built-up areas, a large population, and an important position in politics, economy, science, and culture. Therefore, it’s political particularity may impact the digital economy and urban carbon decoupling differently. So, after deleting Beijing, Shanghai, Chongqing, and Tianjin, we find the digital economy is 0.073 and significant at the 1% level. The finding further verifies hypothesis 1.

Eliminate relevant policies affecting carbon emissions

We conduct robustness tests by excluding provinces and cities involved in carbon trading. Table 5, column (5) shows that the digital economy is 0.082, significant at the significance level of 1%. The result means that the digital economy improves the urban carbon decoupling status. The robustness test verifies hypothesis 1.

Shrinkage treatment

We conducted a 1% tail reduction on the digital economy to prevent the impact of outliers on the research results. In Table 5, Column (5) shows that the digital economy after tail reduction is 0.087, significant at the significance level of 5%. The result means that the digital economy improves the urban carbon decoupling status. The robustness test verifies hypothesis 1.

Treatment of endogenous problems

We use the number of telephones per 100 people in 1995 as an instrumental variable estimation for endogenous analysis. This paper believes the number of telephones in 1995 is historical data, which will not affect the current carbon decoupling status and meet exogenous conditions. From the perspective of relevance, the telephone is an important communication device. The number of telephones provides resources for the digital economy. So, the number of telephones per 100 people in 1995 is related to the development of the digital economy. However, the digital economy evolved while the number of telephones per 100 people in 1995 remained constant. Therefore, we select the Interaction between the number of telephones per 100 people in 1995 and Internet users as the instrumental variable (iv). In Table 6, the first column shows that the is positively correlated with the digital economy. The result is consistent with the expectation.

Table 6 Treatment of endogenous problems.
Full size table

The second column of Table 6 reports that the F statistic of the weak instrumental variable test with heteroscedasticity in the first stage is >10. Week identification test (Cragg-Donald Wald F statistic) = 366.252 is >16.38 at the 10% level. The P value of the Underidentification test is significant at the 1% level. Meanwhile, the digital economy is 0.152 and significant at 5%. Therefore, after considering the endogenous analysis, the digital economy can still improve the carbon decoupling status, confirming hypothesis 1.

Heterogeneity analysis

Heterogeneity of urban resource types

Urban resource endowment affects carbon emissions. This paper classifies the cities into non-resource-based and resource-based cities to test the Heterogeneity. In Table 7, the first column shows that the digital economy of non-resource-based cities is 0.079, significant at 1%. In the second column, the digital economy of resource-based cities is −0.024, which is negative but not statistically significant. So, the digital economy can improve the carbon decoupling level of non-resource-based cities, not yet resource-based cities.

Table 7 Heterogeneity analysis.
Full size table

We believe that a resource-based city takes the mining and processing of natural resources as the leading industry. Its industrial development is highly dependent on resources, and its ability to gather talent and innovation is weak, resulting in its slow development of green transformation and poor carbon decoupling capacity.

Regional heterogeneity

This paper divides the cities into eastern, central, and western cities to test the Heterogeneity. The third column of Table 7 shows that the eastern digital economy (dec) is 0.106, significant at the 5% level. The fourth column indicates that the coefficient of the digital economy in the central region is 0.062, which is significant at the level of 10%. The fifth column suggests that the coefficient of the digital economy in Western China is −0.032, which is negative but not statistically significant. Therefore, the digital economy can improve the carbon decoupling level in eastern and central cities, but not yet in western cities. We believe that the digital economy in the eastern and central cities is developing better, enabling the development of traditional industries through digital technology and improving the urban green development and decoupling status. However, the digital economy in Western cities is poor, the industrial structure is unreasonable, and the dependence on resources is strong, which leads to the small impact of the digital economy on carbon decoupling in Western cities.

Research on spatial effect

Table 3 and Fig. 2 show that urban carbon decoupling has a positive spatial correlation and the characteristics of spatial aggregation. Therefore, it is necessary to use spatial econometric models to analyze the spatial effects of the digital economy on urban carbon decoupling.

Table 8 shows the results of the LM test, the Hausman test, the LR test, and the Wald test. Meanwhile, Elhorst (2010) suggested that if the results of the LM test support any one or two of the SEM and the SAR, we need to use the SDM to estimate (Elhorst, 2010). According to Table 8 and Elhorst (2010), we should use the spatial Durbin model to analyze the spatial relationship between the digital economy and carbon emissions decoupling (Elhorst, 2010).

Table 8 Space effect test.
Full size table

In Table 9, models (1), (2), and (3) are the regression results of the spatial weight matrix of inverse distance, economic distance, and the nested weight matrix of geographical and economic distance. After considering the spatial effect, the coefficients of the model (1)–model (3) digital economy (dec) are all positive and significant at the 5% level. The result shows that the digital economy can still improve the city’s carbon decoupling status after considering the spatial effect. Models (1)–(3) show that rho is positive and significant at a 1% significance level. Therefore, the carbon decoupling status of the surroundings positively impacts this city’s carbon decoupling status.

Table 9 Spatial econometric regression results.
Full size table

Meanwhile, we also found that the digital economy with close geographical distance, economic distance, and economic, geographical distance can reduce the carbon decoupling of the city. This result shows that the digital economy can worsen the carbon decoupling status of the surroundings through the siphoning effect. We believe that regions with a well-developed digital economy can attract high-quality resources from the surrounding, further enhancing the potential for local economic development and enhancing the emission reduction advantages. As a result, where the digital economy is developing well, it strengthens its carbon decoupling status but worsens the carbon decoupling status of the surroundings. Therefore, Hypothesis 2 holds.

Discussion

This study aims to examine the correlation between the digital economy and the decoupling of carbon emissions in 281 Chinese cities from 2012 to 2019. Our findings reveal an imbalanced distribution of digital economy development and carbon emissions decoupling among Chinese cities, showing a spatial agglomeration phenomenon. The conclusion aligns with previous research in the field. For example, Kang et al. (2016) observed a spatial imbalance in China’s carbon emissions (Kang et al. 2016). In addition, Tang et al. (2021) highlighted variations in the development level of China’s digital economy among different cities (Tang et al. 2021).

Our research indicates that the digital economy holds significant promise not only in advancing human progress, bridging the digital divide, and fostering social development but also in driving the decoupling of urban carbon emissions. Consequently, the digital economy plays a pivotal role in facilitating sustainable economic, social, and environmental growth. Its development aligns with the United Nations’ sustainable development goals and aids the Chinese government in achieving harmonized progress in digitization and environmental sustainability (Wu and Zhu, 2022). Policymakers can leverage the digital economy to reduce emissions and achieve high-quality economic development.

Furthermore, we also examined the role of urban heterogeneity in the digital economy and carbon emissions decoupling. Policymakers can develop targeted policies to foster the digital economy and decouple carbon emissions in these areas. For instance, there is a concentration of regions with inadequate decoupling of carbon emissions in Northeast and North China, with many of these areas also being L-L agglomeration zones. As a result, these locations represent crucial and challenging areas for carbon emission control. It is imperative to swiftly transform the existing, unsustainable industrial structure and introduce green development initiatives such as digital transformation to invigorate the economy. To further accelerate the decoupling of carbon emissions, we actively encourage non-resource-based cities and those in the Middle East to enhance their digital economy development. Our focus is on promoting digital technology and smart city solutions to attract investments, drive innovation, improve resource efficiency, and reduce carbon emissions.

Interestingly, our research reveals that the digital economy can suppress neighborhood carbon emissions decoupling due to the siphoning effect, a novel finding given the absence of spatial Durbin models in existing literature exploring the relationship between the digital economy and carbon emissions decoupling. That highlights the need for urban managers to address imbalanced development in the digital economy and reasonably plan its growth to ensure high-quality development while minimizing carbon emissions in neighborhoods.

Conclusion and recommendations

Scholars primarily examined the digital economy’s effects on total, intensity, and efficiency of carbon emissions. However, few have analyzed its impact on carbon emissions decoupling. This article innovatively utilizes Chinese urban data to explore both the spatial distribution and characteristics of the digital economy and carbon emission decoupling. It further investigates the direct impact, heterogeneity, and spatial effects of the digital economy on carbon emission decoupling. Key findings include:

  1. (1)

    The digital economy’s lower level mainly concentrates in the northeast and northwest regions, while the higher level is predominant in municipalities directly under the Central Government and the Southeast Coast.

  2. (2)

    Cities facing challenges in carbon decoupling are primarily found in Northeast China and North China, with most cities still lacking decoupling progress.

  3. (3)

    A positive spatial correlation exists between the digital economy and urban carbon decoupling.

  4. (4)

    The Southeast Coast stands out as the H-H concentration area for the digital economy. Conversely, the L-L category extends across Northeast China, Gansu, and Yunnan. Gansu and Shaanxi host the L-H type for urban carbon decoupling, while Shandong, Hebei, and the Inner Mongolia Autonomous Region witness significant L-L cluster concentrations.

  5. (5)

    Through empirical study, the digital economy’s positive impact on urban carbon decoupling is verified, with results consistently supported by robustness tests.

  6. (6)

    Heterogeneity analysis reveals significant enhancements in urban carbon decoupling by the digital economy in non-resource-based cities, particularly those located in the central and eastern regions.

  7. (7)

    The carbon decoupling status of surrounding areas positively affects this city’s decoupling progress. Moreover, our findings indicate that a closely aligned geographical, economic, and economic-geographical proximity of the digital economy can mitigate the city’s carbon decoupling. This dynamic implies a spatial spillover effect on urban carbon emissions decoupling. However, it’s important to note that the digital economy can also exacerbate carbon decoupling challenges in surrounding regions through the siphoning effect.

Based on these findings, we propose the following recommendations: To begin with, the uneven development of the digital economy across China poses a risk of widening the digital divide. To mitigate this, we recommend bolstering digital infrastructure and literacy in less-developed regions, such as the Northeast and West, to ensure equitable digital access nationwide. Secondly, carbon emission decoupling varies among Chinese cities, posing particular challenges in the northeast and north. Addressing this necessitates tailored emission reduction policies, technical assistance, and robust enforcement of stringent environmental measures. The positive agglomeration effects observed in both the digital economy and carbon decoupling underscore the potential for cross-regional economic policies and environmental interventions. Moreover, the digital economy not only drives economic growth but also enhances urban carbon decoupling. Our research underscores the need for tailored strategies to foster urban digital economies and curb carbon emissions, promoting a balanced development of digitalization and environmental sustainability for robust economic progress. Notably, in non-resource-based cities, especially in the central and eastern regions, the digital economy has played a pivotal role in advancing urban carbon decoupling. This phenomenon is attributed to China’s structural challenges. To address this, the paper advocates for comprehensive system reforms, the establishment of a unified market, seamless digital resource flow within the market, leveraging digital technology to boost traditional industries, nurturing emerging sectors, and ultimately fostering shared prosperity among residents. Furthermore, regions with mature digital economies exhibit a Matthew effect, where stronger areas attract premium resources, potentially impeding carbon emission decoupling in neighboring regions. To counteract this trend, we propose proactive measures by the Chinese government to channel high-quality human and material resources to underdeveloped digital economy areas. That can be achieved through initiatives like rural revitalization, facilitating new urbanization projects, bridging the digital divide, and ensuring widespread digital service accessibility in daily life and work.

Limitations and future directions

An innovative aspect of this study is its examination of the spatial distribution and traits of the digital economy and carbon emission decoupling. Furthermore, it explores the direct influence, diversity, and spatial impacts of the digital economy on carbon emission decoupling, utilizing data from Chinese cities. This investigation is a novel endeavor. Our findings illustrate that leveraging the digital economy not only spurs economic and societal transformation but also enhances urban carbon emission decoupling. Consequently, integrating the digital economy can foster economically, socially, and environmentally sustainable progress, aligning with the United Nations 2030 Agenda for Sustainable Development. Naturally, governments should remain attentive to the digital divide and the potential siphon effect associated with the digital economy.

This paper has limitations: Firstly, it does not account for the global outbreak of COVID-19 in 2020. The measures taken by China to combat the pandemic, such as restrictions on gatherings, work suspension, business shutdowns, and city closures, have impacted urban carbon emissions and may have affected the empirical results of this study. Future research should consider including data from the epidemic period to analyze its impact on the empirical findings. Secondly, this paper primarily focuses on the spatial distribution and agglomeration of decoupling between the digital economy and carbon emissions, as well as their relationship and spatial effects. However, it does not analyze the underlying mechanisms of their interaction. Therefore, further exploration is needed to understand the mediation and regulatory mechanisms between the two factors.