Digital economy development and carbon emission intensity—mechanisms and evidence from 72 countries

Abstract

This paper employs two-way fixed effects models, mediation models and spatial panel models to examine the impact of digital economy development on carbon emission intensity in 72 countries from 2013 to 2020 using TIMG index as the proxy variable of digital economic development. The results of this paper show that (1) there is an inverted U-shaped relationship between the development of the digital economy and the carbon emissions in various countries. The development of digital economy first has a positive effect on carbon emissions, and then, its impact turns negative after it reaches a designated inflection point. (2) The results of the mechanism analysis show that the development of the digital economy can reduce carbon emissions by promoting industrial upgrading, while it could have an inverted U-shaped nonlinear effect on carbon emissions through energy efficiency. (3) The development of a country’s digital economy may have a significantly inverted U-shaped spatial spillover effect on neighboring countries. This paper provides reference for countries to formulate carbon emission reduction policies and promote the coordinated development of digital economy and environmental protection.

Introduction

Greenhouse effects and climate warming are important challenges that all countries worldwide currently face. In November 2023, the United Nations Environment Program released the report “Emissions Gap Report 2023: Breaking Records—Temperatures hit new highs, the world fails to meet emissions reduction targets”, which shows that the global temperature is expected to increase by 2.5–2.9 °C in 2023. Even in the most optimistic scenario, the probability of the temperature increase being within 1.5% is less than 15%. Climate warming may lead to a series of serious consequences, such as extreme weather, crop failure and sea level rise. Environmental issues stemming from climate change transcend national borders, with one country’s environmental challenges potentially spilling over to others. Therefore, addressing the global climate problem necessitates the concerted efforts and collaboration among governments worldwide. Governments around the world have reached a consensus to take effective measures to reduce carbon emissions, strive for carbon neutrality (zero net carbon emissions) and realize sustainable economic development in our society.

As a new economic form, the digital economy uses data as the core factor of production and has the characteristics of digitalization, networking and intelligence; additionally, the effective use of communication technology is an important driving force for efficiency improvement and economic structure optimization. In recent years, the digital economy has developed rapidly in various countries, and the proportion that contributes to the global economy has increased rapidly; thus, the digital economy has become an important driving force for the economic growth of various countries. According to the statistics of the World Bank, the proportion of the digital economy in global GDP has reached more than 15%, and its growth rate in the past 10 years has been more than 2.5 times that of global GDP. Considering the strategic importance of the digital economy, countries have introduced relevant measures to boost the prosperity of their own digital economy. In the future, the digital economy will continue to develop at a high speed. This emerging economic form has the potential to profoundly impact the environmental conditions of various countries.

On the one hand, the development of the digital economy will foster advancements in technology, aid in enhancing the efficiency of resource utilization and promote the utilization of renewable energy, thereby reducing carbon emissions¹. On the other hand, during the development of the digital economy itself, a significant amount of energy may be required for the construction of digital infrastructure. Additionally, the operation of technology companies’ data centers necessitates a substantial consumption of energy, thus increasing carbon emissions. Thus, we posit that the digital economy also exerts a nonlinear, influence on carbon emission intensity across countries.

Although there are plenty of studies that discuss the relationship between digital economy and carbon emission, most of them are based on the city-level data or province-level data^2,3. Many emerging countries are in the initial stages of digital economy development. Clarifying the relationship between digital economy development and carbon emission intensity from a cross-national perspective holds significant reference value for emerging countries in formulating economic policies. Carbon emission reduction is a global issue, not confined to a single country. A cross-national approach allows for a better understanding of the impact of digital economy development on carbon emissions in various countries. Simultaneously, it provides a reference for formulating cooperative policies among countries and promoting the achievement of global carbon neutrality goals.

The relationship between digital economy and carbon emission across countries might be much different from that using province level data or city level data in China. Current studies on the relationship between digital economy development and carbon emission using cross-country data is rare, and the studies about the mechanism and spatial spillover effect has been rare.

In light of this, We use panel data of 72 countries from 2013 to 2020 and the TIMG index as the proxy variable for the degree of digital economy development in each country. This paper empirically studies the impact of each country’s digital economy development on its carbon dioxide emissions and further analyses it from the perspectives of heterogeneity analysis, mechanism testing and spatial spillover effects. We mainly focus on the following questions: Firstly, does the development of digital economy in various countries exert an inverted U-shaped nonlinear impact on carbon emission intensity? Among the multiple sub-dimensions of digital economy development, which dimension has a more significant influence on carbon emission intensity? Secondly, through which channel does the development of digital economy in different countries promote or inhibit carbon emission intensity? Thirdly, does the development of digital economy in one country could have a spatial spillover effect on carbon emission intensity in neighboring countries? Clarifying these issues is of great significance for developing country governments to formulate reasonable digital economy development strategies and achieve environmental cooperation among nations.

Through a fixed effect panel regression model, it is found that there is an inverted U-shaped relationship between digital economy development and carbon dioxide emissions. We also carried out heterogeneity analysis for four aspects in terms of digital technology (T), digital infrastructure (I), the digital market (M) and digital governance (G), which supplemented the literature to some extent. Second, this paper adopts the intermediary effect model to study how the digital economy affects carbon emissions. We find that the digital economy could decrease carbon emissions through industrial structure upgrading. We further find that the digital economy could have an inverse U-shaped relationship with energy consumption intensity, which has a positive impact on energy consumption intensity. Third, this paper uses the spatial panel model to investigate the spatial spillover effects of digital economy development on carbon dioxide emissions in neighboring countries. Carbon dioxide emissions are a common challenge faced by all countries, and the development of a digital economy in one country may also have spillover effects on neighboring countries. As a country’s digital economy evolves, it is highly probable that it will positively impact the environmental conditions of neighboring nations through information and technological innovations. However, there is also the potential for negative impact via the “Siphon” effect. Consequently, the development of a country’s digital economy may also generate an inverted U-shaped spatial spillover effect on its neighboring countries. This paper shows that the impact of digital economy development in neighboring countries on carbon emission intensity also has an inverted U-shaped relationship.

Our major contributions are as follows:

Firstly, we delve into the specific channels and mechanisms through which the development of digital economy impacts carbon emission intensity across countries. Notably, existing empirical research at the transnational level has not conducted targeted analyses in this regard^4,5,6,7. By adopting the intermediary effect model, we demonstrate that the digital economy influences carbon emissions primarily through industrial structure upgrading and reductions in energy intensity. This approach offers valuable insights to governments worldwide, enabling them to better comprehend the environmental implications of digital economy and to facilitate the synergistic development of digital economy and sustainable economy.

Secondly, we investigate the spatial spillover effect of digital economy development in one country on neighboring countries. Currently, as far as we know, all studies in cross-country level only focus on the relationship between digital economy and carbon emission intensity in its own country^6,7,8,9, none of them have considered the spatial spillover effect. The global warming is a global issue for each country. As the development of globalization, the connections between countries in economics, society and environment have been strengthened. Thus a country’ digital economy development has a significant spatial spillover effect. We demonstrate that a country’s digital economy development can have significant spillover effects on the carbon emission intensity in neighboring nations. Notably, it reveals that the impact of digital economy development in neighboring countries on carbon emission intensity also follows an inverted U-shaped relationship. These findings offer valuable insights for fostering global cooperation in policy-making to address climate change effectively.

Finally, we introduce the TIMG index as a proxy variable for measuring the degree of digital economy development in each country. This index, constructed using components such as digital technology, infrastructure, market, and governance, offers a more comprehensive and more suitable measure for assessing digital economy development in developing countries^10,11.

The structure of this paper is as follows: Part 2 reviews the past literature, and Part 3 presents the empirical methods and findings. Conclusions and policy recommendations are presented in Part 4.

Literature review

This paper reviews two strands of relevant literature: the effect of the digital economy on the environment and the influencing factors of carbon emission intensity for each country.

The multifaceted impact of the digital economy on the environment

The development of the digital economy has had an enormous impact on various aspects of the economy and society. Many scholars have considered whether the development of the digital economy has an environmental improvement effect from the perspective of carbon emission reduction. These studies were conducted mostly based on province-level or city-level data in one nation.

Generally, a negative relationship between digital economy development and carbon emissions is found in these studies. In terms of the underlying mechanisms, the development of digital economy can reduce carbon emission intensity by promoting the upgrading of the industrial structure, reducing the consumption of traditional energy such as coal, promoting scientific and technological innovation and other methods, that is, the development of a digital economy has an environmental improvement effect^{2,10,11,12,13}. However, the development of the digital economy may also have a negative impact on the environment due to the increased demand for energy consumption. Thus, the relationship between digital economy development and carbon emission intensity might be nonlinear. Under the nonlinear framework, most scholars find evidence that there is an inverted U-shaped relationship between the development of the digital economy and carbon emission intensity, that is, the development of the digital economy will increase carbon emission intensity in the early stage, while it will have a negative effect on carbon emission intensity later^9,14,15. In recent years, some studies also have shown that relationship between digital economy development and carbon emission intensity could be N-shaped^10,11.Similar results could also be seen in the studies by using green total factor productivity (GTFP) as the dependent variable, which stems from calculating the total factor productivity when considering the effect of environment^16,17,18.

Most current research makes conclusions from empirical results using city-level or province-level data instead of cross country data, which could lose some important information. Global warming and environment problem is a global issue, Research at the transnational level enables governments to draw lessons from the development experiences of other countries. Additionally, it provides valuable reference for countries to engage in cooperation regarding digital economy development.

Factors influencing a country’s carbon emission intensity

Global warming caused by carbon emissions has been recognized as a threat to public health and welfare for all countries. Therefore, many scholars have studied the influencing factors of carbon emission intensity from a cross-country perspective. According to the STIRPAT model, the impact of human activity on the environment depends mainly on three factors, P (population), A (affluence) and T (technology); thus, each country’s carbon intensity is affected by all these factors. The above three factors are decomposable¹⁹. Other factors may also influence carbon emission intensity. Based on the above model, researchers have found that the important factors influencing the carbon emission intensity of various countries include per capita income, urbanization, aging, trade openness, population density, etc.^6,7,20. In addition, other studies have shown that scientific and technological innovation²¹, and income inequality⁷, economy and policy uncertainty^22,23,24, nature resources²⁵, economic complexity^1,26, environmental taxes²⁷ also significantly impact the carbon emission intensity of various countries.

In recent years, considering the rapid growth of the proportion of the digital economy accounting for the global economy, some scholars have also analyzed the relationship between the digital economy and CO2 emissions in a global context. Dong et al.⁸ adopted per capita CO2 emissions and CO2 emission per unit of GDP as proxy variables of CO2 emissions and found that the development of a digital economy could reduce CO2 emission intensity. Besides, Li et al.^4,5 and Wang et al.^6,7 further demonstrates that the effect of digital economy development in a country could have nonlinear effect on carbon emission intensity in one country. However, none of them have dig into the mechanism about the nonlinear effect of digital economy on carbon emissions. None of them have considered about the spatial effect of the digital economy development on carbon emission in neighboring countries in their research. This has led to inadequate research on the impact of digital economy on carbon emission intensity from a transnational perspective.

In summary, scholars have conducted empirical analysis based on the panel data at the province or city level to investigate the relationship between digital economy development and carbon emission intensity, providing a useful reference for the analysis of the relationship between them two. However, few studies have analyzed the impact of a country’s digital economy development on its carbon emission intensity from a transnational perspective. In the limited research, they have not dig into the mechanisms and have overlook the spatial spillovers in the existing literature, we aim to bridge this gap and enrich the field by conducting a comprehensive analysis. Specifically, the TIMG index is employed as a proxy variable for the level of digital economy development in each country, leveraging panel data from 72 nations spanning the period from 2013 to 2020, we investigate the intricate relationship between a country’s digital economy development and carbon emission intensity, we also utilize the intermediary effect model to investigate the mechanisms. Furthermore, a panel spatial model is incorporated to examine the spatial spillover effects of digital economy on carbon emission intensity in neighboring countries.

Theoretical analysis and research hypothesis

How the digital economy affects carbon emission intensity

The Environmental Kuznets Curve (EKC) suggests that economic growth and environmental degradation initially follow a positive correlation, but eventually, as income rises above a certain threshold, the environmental condition decreases. This theory provides a foundation for understanding the potential relationship between the digital economy and carbon emission intensity. In the early stage of the development of the digital economy in various countries, the positive effect of digital economy on environment has not yet been reflected. However, the construction of digital infrastructure and the emergence of new digital industries often require significant energy consumption and can lead to an increase in carbon emissions. Thus, during these early stages, the development of the digital economy is positively correlated with carbon emission intensity.

As the digital economy gradually grows matures and reaches a certain level of development, several factors suggest that its impact on carbon emission intensity will turn negative. Firstly, digitalization leads to increased efficiency in various sectors, including energy production and consumption. Data becomes a key factor of production, enabling smarter resource allocation and optimization. Secondly, digital technologies enable the development of clean energy sources and more efficient energy systems. This, in turn, reduces the carbon intensity of energy production and consumption. Finally, due to “network effect”, the efficiency of economy system tends to improve, which encourages the more efficient utilization of the existing infrastructure, leading to a rise in output per unit of input, including energy. Thus, the combination of these factors suggests that, after a certain threshold of digital economy development is reached, its impact on carbon emission intensity will shift from positive to negative. Therefore, Hypothesis 1 is proposed:

H1: The development of a country’s digital economy has a first increases and then decreases nonlinear effect on the carbon emissions intensity.

How the digital economy affects carbon emission intensity

The digital economy can indirectly affect the carbon emission level of a country by promoting the upgrading of its industrial structure. With the development of the digital economy, a large number of new economic forms such as the sharing economy have emerged, besides, the domestic information technology industry has developed and a large number of information technology enterprises have emerged, the economic structure has gradually shifted from resource-intensive and energy-intensive to technology-intensive, and the industrial structure has improved².

In addition, Industrial structure upgrading is often accompanied by a reduction in carbon emissions. Compared to the industrial sector, the service industry requires less energy consumption and has a lower carbon emission intensity per unit of GDP. Therefore, as the economy transforms from an industrial to service-oriented, the carbon emission intensity per unit of GDP will tend to decrease. During the process of industrial structure upgrading, a significant number of high-polluting enterprises and energy-intensive industries are phased out, leading to improved energy utilization efficiency in the production process. This results in a reduction in energy consumption and pollution emissions per unit of GDP, thereby fostering the development of a low-carbon and energy-efficient economy¹⁷. Wen et al.²⁸ find that industrial digitalization significantly reduces the pollution emission intensity of manufacturing enterprises and has a significant positive impact on their environmental performance. This indicates that industrial structure upgrading can effectively reduce the pollution emissions of enterprises. Therefore, the following hypothesis is proposed in this paper:

H2: The development of a country’s digital economy can reduce its carbon emission intensity by promoting the upgrading of its industrial structure.

In addition, the effect of the digital economy on energy consumption intensity could also be an important intermediary mechanism through which the development of a country’s digital economy affects carbon emission efficiency. Digital economy development usually has an inverted U-shaped relationship with energy consumption, and energy consumption is usually accompanied by higher carbon emission intensity^4,5. First, the operation and development of digital information industries require a large amount of energy, especially for electricity. For example, the construction of electronic raw devices, data center storage equipment, and related infrastructure all have a high demand for electricity. Moreover, during the process of digitization, enterprises tend to increase their investment in or replace production equipment, which increases energy consumption and leads to an increase in energy consumption per unit of GDP. All the above analyses could lead to an increase in energy consumption and carbon emission intensity at the early stage of the development of the digital economy. On the other hand, with the help of data elements, the digital economy can effectively overcome temporal and spatial limitations and promote new industries such as smart logistics, smart transportation and energy. These industries usually have high energy efficiency, so they can promote the energy efficiency of the whole country and thus reduce carbon emissions. Second, the digital economy can promote the transformation of traditional enterprises to intelligent and digital economies, improving production efficiency to reduce energy consumption per unit of output. In summary, the development of the digital economy in various countries has a nonlinear relationship, first positive and then negative effects on energy use intensity, which in turn has a nonlinear impact on carbon emission intensity. Therefore, the following propositions can be obtained:

H3: The development of a digital economy can have a first promoting and then inhibiting nonlinear effect on the carbon emissions of various countries through energy consumption intensity.

The spatial spillover effect of the digital economy on carbon emissions in neighboring countries

Digital economic development can decrease regional barriers across countries, while increasing economic ties between different regions. As a result, the exchange of resources and goods between different countries will be easier. The development of a digital economy in one country may have both promoting and inhibiting effects on carbon emission reduction in neighboring areas. On the one hand, digitization promotes the cross-regional flow of production factors among different regions and countries, and the technical information related to a country’s digital economy can be shared with other neighboring countries, thereby reducing the carbon emission intensity of the surrounding countries through the technology spillover effect. On the other hand, due to the siphon effect, countries with a high level of digital economy development often cause high-tech enterprises and related talent from neighboring countries to move to that country, which is not beneficial to industrial structure upgrading or energy efficiency improving in neighboring countries and thus has an inhibiting effect on carbon emission reduction activities in that country³. Therefore, Hypothesis 4 is proposed in this paper:

H4: The development of the digital economy can produce an inverted U-shaped nonlinear effect on the carbon emission intensity of neighboring countries.

Empirical results

Variable selection and data explanation

This paper selects annual data from 72 countries from 2013 to 2020 for research. The sample was selected mainly on the basis of the data available from the TIMG index and other variables. The main explanatory variable is the level of digital economy development in each country, while the explained variable is the carbon dioxide emission intensity of each country. The indicators of relevant variables are selected and explained as follows:

Dependent variables

Carbon intensity: In the existing literature, Carbon emission intensity (cei) and carbon dioxide emissions per capita(pce) are often used as proxy variables for carbon emissions⁸. Following Dong et al.⁸, we use carbon emission intensity (cei) as the proxy variable for carbon emission in the benchmark analysis. This variable is measured in units of carbon emissions per unit of GDP. The higher the value of cei, the greater the greenhouse gas emissions at a given level of economic development, thereby exerting more adverse impacts on the environment. Furthermore, to ensure the robustness of our findings, we also utilize per capita carbon emissions as an alternative explanatory variable in the robustness checks section. To mitigate the issue of heteroscedasticity, we apply a logarithmic transformation to the carbon emissions data.

Independent variables

Digital economy development degree (TIMG): This paper adopts the TIMG index calculated by Wang et al.²⁹ through multiple sub indexes as the proxy variable for the degree of digital economy development in various countries. Compared with previous indicators, this index is more comprehensive and based on the digital economy resource endowment and institutional environment of each country. It measures the development level of the digital economy from the four dimensions of digital technology, digital infrastructure, the digital market and digital governance. The digital technology dimension can be evaluated from aspects such as research output, human capital, and innovation capabilities. The digital infrastructure dimension primarily focuses on measuring coverage breadth, construction quality, and usage cost. The digital market dimension primarily concerns the overall size, segmented markets, and the status of digital trade. The digital governance dimension reflects the institutional foundation and environment for the development of a country’s digital economy. A higher numerical value of this indicator signifies a higher level of digital economic development for a country. Given its comprehensiveness and inclusiveness in measuring a nation’s digital development status, we have selected it as the core explanatory variable in this study.

Control variables

With reference to Dong et al.⁸, we select control variables from the aspects of energy structure, urbanization level, trade openness and economic growth rate of each country. The indicators of each variable are selected and explained as follows:

Trade openness (trade_open): We use the proportion of imports and exports to GDP to represent the degree of trade openness of a country. The data are derived from the WDI database. Trade openness affects the consumption and production of intermediate and final goods in a certain region and is usually positively correlated with CO2 emissions²⁰.

Urbanization level (urban): a country’s urbanization level is often positively correlated with its carbon emission intensity. In this paper, the proportion of the urban population to the total population is selected as the proxy variable of urbanization level in each country. The data source is the WDI database. A country’s urbanization level may increase the degree of population agglomeration, accelerate the process of industrialization, and thus increase carbon emission activities. Therefore, a country’s urbanization level is often positively correlated with its carbon emission intensity.

Economic development level (gdp_per): In this paper, the per capita GDP of a country is used to represent the level of economic development of a country. Generally, the higher the level of economic development, the lower the carbon emission intensity.

Population aging (old): We use the proportion of people aged 65 years and older among the total population as the proxy variable for population aging. The higher the level of aging, the more likely a country is to recognize the importance of the environment, thereby negatively impacting carbon emission intensity³⁰.

Mediating variables

Advanced degree of Industrial structure (IST): The development of a digital economy may have an impact on a country’s carbon emissions by promoting industrial structure upgrading. With reference to Yao et al.¹⁸ and³¹, we use the industrial upgrading index (IST) to represent the industrial development level of a country, and the definition is as follows:

$$IST_{it} = R_{it1} \times 1 + R_{it2} \times 2 + R_{it3} \times 3$$

(1)

Among them, $R_{it1}$, $R_{it2}$ and $R_{it3}$ represent the proportion of a country’s first, second and tertiary industry added value to GDP, respectively.The primary industry primarily comprises agriculture, the secondary industry encompasses industry and manufacturing, and the tertiary industry refers to the service sector. Generally speaking, the tertiary industry offers the highest value-added, followed by the secondary industry, and the primary industry has the lowest. Under this measurement approach,; the larger the value of $IST_{it}$ is, the more advanced the industrial structure of the country.

Energy consumption intensity (energy_int): According to the theoretical analysis, the development of the digital economy might have a nonlinear impact on the carbon emission efficiency of a country through energy consumption intensity. In this paper, the energy intensity level of primary energy per GDP (energy_int) is used to characterize the energy intensity of each country. This indicator represents the amount of energy consumed by an economy to produce one unit of GDP, thereby depicting both energy consumption intensity and efficiency. The data source is the WDI database. The higher the value is, the greater the energy consumption per unit of GDP of the country.

Variable description and data source are summarized in Table 1. Also, the descriptive statistical results of each variable are shown in Table 2. As shown in the table, there are significant differences in the degree of digital economy development across countries, with the minimum value being only 0.107 and the maximum value being 0.944. The degree of digital economy development may have an enormous impact on the carbon emission reduction activities of each country.

Table 1 Data description and data source.

Table 2 Description statistics.

Model principles

Baseline regression model

As shown in Eqs. (2) and (3), we use a two-way fixed effect regression model to investigate the impact of digital economy development on carbon dioxide emissions in various countries. In addition to panel regression, existing studies also commonly employ panel time series models, such as the panel ARDL model for analysis^23,24,32.These models can account for both the cross-sectional dependence and dynamic relationships among variables. Our rationale for choosing a static model is multifaceted: firstly, based on the theoretical underpinnings of panel time series models, methods like panel ARDL are best suited for scenarios with a large T (time periods) and a small N (number of entities), whereas our sample comprises over 70 economies, which does not fulfill this criterion. Secondly, due to the large sample size and the inclusion of both linear and quadratic terms of the digital economy in our equations, delving into slope heterogeneity would significantly complicate the analysis. Consequently, in the heterogeneity analysis section, we have opted to compare and contrast the differences between developed and emerging economies. Thirdly, adopting a static panel regression model also facilitates subsequent research endeavors, enabling us to conduct mechanism tests and spatial panel regression analyses with ease. To reduce the problem of missing variables and mitigate endogeneity, we add the bilateral fixed effects of countries and years to the model. In the baseline regression, Eq. (2) is first estimated to study the linear impact of the degree of digital economy development on carbon emission intensity. In addition, as shown in Eq. (3), we also add the square term of the TIMG index of each country to the regression to investigate the nonlinear influence of the digital economy development level on the carbon emission intensity of each country.

$${\text{lncei}}_{it} = \beta_{0} + \beta_{1} dige_{it} + \beta_{3} {\text{X}}_{it} + \lambda_{i} + \mu_{t} + {\upvarepsilon }_{it}$$

(2)

$${\text{lncei}}_{it} = \beta_{0} + \beta_{1} dige_{it} + \beta_{2} dige_{it}^{2} + \beta_{3} {\text{X}}_{it} + \lambda_{i} + \mu_{t} + {\upvarepsilon }_{it}$$

(3)

where ${\text{lncei}}_{it}$ is the logarithm of the carbon emission intensity of each country and TIMG denotes the development degree of the digital economy of each country. ${\text{X}}_{it}$ represents the control variable, $\lambda_{i}$ and $\mu_{t}$ are the fixed effects that do not change with time and country, respectively, and ${\upvarepsilon }_{it}$ is the error term of the regression. In Eq. (3), the marginal effect of the digital economy on carbon emission intensity is $\beta_{1} + 2\beta_{2} \times dige_{it}$, and the marginal effect of the development degree of the digital economy on carbon emission intensity may be related to the development level of the digital economy.

Intermediary effect model

In addition to affecting carbon emission intensity directly, the development of a digital economy could also indirectly affect carbon emissions by promoting industrial upgrading and affecting energy consumption intensity within a country. To test the mechanism through which digital economy development impacts the carbon emission intensity of various countries, this paper uses the intermediary effect model to test whether the industrial structure and energy consumption intensity are intermediary variables through which digital economy development affects carbon emission reduction.

$${\text{mid}}_{it} = \beta_{0} + \beta_{1} dige_{it} + \beta_{2} dige_{it}^{2} + \beta_{3} {\text{X}}_{it} + \lambda_{i} + \mu_{t} + {\upvarepsilon }_{it}$$

(4)

$${\text{lncei}}_{it} = \beta_{0} + \beta_{1} dige_{it} + \beta_{2} dige_{it}^{2} + \beta_{3} {\text{X}}_{it} + \beta_{4} {\text{mid}}_{it} + \lambda_{i} + \mu_{t} + {\upvarepsilon }_{it}$$

(5)

where ${\text{mid}}_{it} { }$ denotes the corresponding intermediary variable and the remaining variables are consistent with Formula (3).

Spatial panel model

To study the effect of digital economy development on carbon emissions, this paper employs a spatial econometric model to study the spatial spillover effect between model variables. As shown in Eqs. (6) and (7):

$$Y = \alpha + \rho WY + X\beta + WX\delta + \mu$$

(6)

$$\mu = \lambda W\mu + \varepsilon$$

(7)

where $Y$ is the explained variable of concern in this paper, which refers to the carbon dioxide emissions in the country of interest, and $X$ is a matrix composed of all the explanatory variables, including the first and square terms of the TIMG index and other control variables; $WY$ and $WX$ denote the spatial interaction effects of the independent variables and dependent variables. $\rho$, $\beta$, $\delta$, $\alpha$, and $\lambda$ are the relevant regression coefficients. $\rho$ and $\lambda$ represent the spatial autocorrelation coefficients. $\mu$ represents the residual sequence matrix, and $W\mu$ represents the interaction effect of the residual terms.

Formula (6) represents different models depending on the value of the parameter λ. Specifically, if λ = 0, then the model is called the spatial Durbin model (SDM), which indicates that the dependent variable of a specific country is subject to the spatial spillover effect of both independent variables and dependent variables of neighboring countries. If λ = 0 and δ = 0, then the model is called the spatial autoregression model (SAR), which indicates that the dependent variable of a country is only subject to the spatial spillover effects of the dependent variables of other countries in the surrounding region. If $\rho = 0$ and δ = 0, then the model is called the spatial error model (SEM), which indicates that the spatial spillover effect comes from the influence of missing variables in neighboring regions.

Under the above model setting, the change in the independent variable in a specific region will not only affect the value of the dependent variable in that region but also affect the value of the dependent variable in other regions. Among them, the former is called the “direct effect”, while the latter is called the “indirect effect” or spatial spillover effect. Due to the autocorrelation coefficient, the regression result does not represent an estimate of the direct and indirect effects, but it can be calculated from the regression coefficient of the spatial spillover model.

$$\begin{aligned} \left[ {\frac{\partial E\left( Y \right)}{{\partial x_{1k} }},\frac{\partial E\left( Y \right)}{{\partial x_{2k} }}, \cdots \frac{\partial E\left( Y \right)}{{\partial x_{Nk} }}} \right] & = \left( {I - \rho W} \right)^{ - 1} \left( {\beta_{k} I + \delta_{k} W} \right) \\ & = \left( {I - \rho W} \right)^{ - 1} \left[ {\begin{array}{*{20}c} {\begin{array}{*{20}c} {\beta_{k} } & {w_{12} \delta_{k} } \\ {w_{21} \delta_{k} } & {\beta_{k} } \\ \end{array} } & {\begin{array}{*{20}c} \ldots & {w_{1N} \delta_{k} } \\ \ldots & {w_{2N} \delta_{k} } \\ \end{array} } \\ {\begin{array}{*{20}c} \ldots & \ldots \\ {w_{N1} \delta_{k} } & {w_{N2} \delta_{k} } \\ \end{array} } & {\begin{array}{*{20}c} \ldots & \ldots \\ 0 & {\beta_{k} } \\ \end{array} } \\ \end{array} } \right] \\ \end{aligned}$$

(8)

In the above formula, diagonal and nondiagonal elements represent direct and indirect effects, respectively. As shown in Eq. (9), for a SAR model, the direct effect of the SAR model is the diagonal element, and the “indirect effect” is the average of the nondiagonal elements of $\left( {I - \rho W} \right)^{ - 1} \beta$

$$\left[ {\frac{\partial E\left( Y \right)}{{\partial x_{1k} }},\frac{\partial E\left( Y \right)}{{\partial x_{2k} }}, \cdots \frac{\partial E\left( Y \right)}{{\partial x_{Nk} }}} \right] = \left( {I - \rho W} \right)^{ - 1} \beta$$

(9)

For the SEM, $\delta_{k} = 0$ and $\rho = 0$; therefore, the direct effect is $\beta_{k}$, and the indirect effect is 0. Among the three models, the spatial Durbin model is most suitable for studying the “direct effect” and “indirect effect” between the independent variable and the dependent variable. In the following analysis, we also use the LR model to determine which model should be used as the main model for the study of spatial spillover effects. Specifically, we use the null hypothesis that “the SDM can be degraded to the SAR model” and “the SDM can be degraded to the SEM” to carry out the research.

When estimating the spatial econometric model, it is necessary to set the spatial distance matrix first. Because the vast majority of sample countries in this paper are not adjacent, we do not use the 0–1 spatial distance matrix, which is constructed according to whether each country is adjacent. Instead, following Yao et al.¹⁸, the distance between the capitals of two countries is selected to represent the geographical distance between two countries. We then take the reciprocal value of the square term of the geographical distance between two countries as the corresponding element of the nondiagonal element in the spatial measurement matrix and set all diagonal elements as 1. The data source for the geographical distance between the capitals of the two countries is the CEPII database. In the following section, the spatial measurement matrix is normalized, so the spillover effect measures the weighted average of the carbon emission intensity of the region, representing the digital economic development of all neighboring regions.

Panel regression analysis

Baseline regression results

According to the principle in Eq. (3), the linear term and square term of the TIMG index in each country is taken as the explanatory variable, the carbon emission intensity is taken as the explained variable, and control variables are also added to carry out the panel regression analysis. As shown in the first column of Table 3, when the TIMG index is used as the only independent variable of regression, the sign of its coefficient is positive, while it is not significant at the 10% level, indicating that the development degree of the digital economy has a weak positive correlation with the overall carbon emission intensity of countries. However, the relationship between these two parameters changes with the value of TIMG index changes. As shown in Columns 2 and 4 of Table 3, when the square term of the TIMG index is added to the regression, the coefficient of the primary term of the TIMG index is significantly positive, while that of the square term is significantly negative, indicating that the marginal impact is negative at first; however, as the level of digital economy development increases, the magnitude of the impact gradually decreases and finally even decreases to a negative value. Taking the regression results in Column 4 as an example, the marginal influence of the TIMG index on carbon emission intensity is 4.19–6.02*TIMG. When the TIMG index is greater than 0.696, digital economy development will have a significantly negative impact on carbon emission intensity. When the TIMG index is less than 0.696, digital economy development has a positive relationship with the carbon emission intensity of a country. In the early stage of digital economy development, the operation of information enterprises and the construction of infrastructure need to consume energy such as electricity, while the improvement in energy efficiency brought about by the development of the digital economy has not yet been reflected; thus, digital economy development might lead to an increase in carbon emissions. When the development of the digital economy reaches a high level, the digital economy will be deeply integrated with the development of manufacturing companies, and the development of the digital economy will reduce carbon emissions per unit GDP and promote the process of carbon emission reduction. Thus, Hypothesis 1 is supported by these results. In the current research examining the relationship between digital economy and carbon emission intensity at the transnational level, Dong et al.⁸ have discovered a negative impact of digital economy on carbon emission intensity. Meanwhile, Li et al.^4,5 and Wang et al.³³ have delved into potential nonlinear relationships, revealing an inverted U-shaped effect where digital economy initially promotes and then inhibits carbon emission intensity in various countries. This finding aligns with the conclusion of previous research, further validating that the nonlinear relationship between digital economy and carbon emission holds true even in transnational research contexts. Among the relevant control variables, per capita income is negatively correlated with carbon emissions, while urbanization and trade openness increase carbon emission intensity.

Table 3 Baseline Regression of digital economy development on carbon emission level.

Robustness test

• (A) Regression coefficient test.

To ensure the accuracy of the conclusions of the benchmark regression in this paper, the robustness test of the regression coefficient is carried out next. First, to ensure the accuracy of the statistical inference of regression coefficients, this paper first used the method of Driscoll and Kraay (1998) modified covariance matrix to obtain the robust standard error, and overcame the problems of sequence correlation and cross-sectional correlation to some extent. Second, the explanatory variable is replaced by the logarithm of per capita CO2 emissions (lnpce). Finally, the main explanatory variables, control variables and explained variables were winsorized at the 5% and 95% percentiles, the regression in Eq. (2) was conducted again. As shown in Table 4, the regression coefficient of the first term of the TIMG index is still significantly positive, while the coefficient of the square term is significantly negative, indicating that after changing settings, the TIMG index and carbon emission intensity of various countries still show an inverted U-shaped relationship, while the sign of the regression coefficient for the control variable has not changed, indicating that the results of the benchmark regression in this paper are reliable.

Table 4 Robustness test of the impact of the digital economy on carbon emissions.

• (B) Panel quantile regression.

In the benchmark analysis, this paper examines the overall relationship between the degree of digital economy development and carbon emission intensity in various countries through static panel regression. However, the relationship between digital economy development and carbon emission intensity may differ under different carbon emission intensity values. Next, following Doğan et al.³⁴,we use a panel quantile regression model to investigate the relationship between the digital economy and the carbon emission intensity of various countries at different carbon dioxide emission intensities; the results are shown in Table 5. As shown in Table 5, under the different quantiles of CO2 emission intensity, the coefficients of the primary term of the TIMG index are significantly positive, while the coefficients of the secondary term are significantly negative, indicating that the TIMG index and carbon emission intensity of various countries exhibit an inverted U-shaped relationship. The marginal effect of the increase in the TIMG index on carbon emission intensity is positive at first and becomes negative after reaching a critical node, which is consistent with the results of the baseline regression and once again demonstrates the accuracy of Hypothesis 1. In addition, the turning point of the inverted U-shaped relationship becomes more pronounced as the carbon emission intensity increases. Therefore, the marginal impact of the digital economy on carbon emissions becomes negative at a higher speed when the carbon emission intensity is low, indicating that the carbon emission reduction effect is stronger at this point.

Table 5 Marginal impact of digital economy development on carbon emissions at different subpoints.

Mechanism inspection

To test the hypothesis proposed in Sect. 3.2, this paper uses industrial structure and energy consumption intensity as intermediary variables and explores the mechanism of the impact of digital economy development on carbon emission intensity through the intermediary effect model. The regression results for industrial structure and energy consumption intensity used as intermediary variables are shown in Tables 6 and 7, respectively. As shown in Column (2) of Table 6, after controlling for the time effect and individual effect, the coefficient of the primary term of the TIMG index was significantly positive, indicating that it has a significant positive impact on the degree of advancement of the industrial structure (IST). With the development of the digital economy, the industrial structure will increase.

Table 6 Intermediate effects test of the upgrading of the industrial structure.

Table 7 Intermediate effects test of energy consumption.

As shown by the regression results in Columns (3) and (4) of Table 6, there is an inverse U-shaped relationship between the IST index and carbon emission intensity, and the marginal impact of the IST index on carbon emission intensity is 2.11–1.16*IST. Considering that the minimum IST in the sample countries is 1.266, the corresponding maximum marginal impact is − 0.316. Therefore, the marginal influence of the IST index on carbon emission intensity is negative for all sample values, indicating that industrial structure upgrading helps to reduce the carbon emission intensity of countries. In addition, compared with those of the regression results in Column (1), the regression coefficients of the first and second terms of the TIMG index in the regression results in Column (3) decrease in absolute terms, indicating that when industrial structure transformation is taken into account, the negative impact of digital economy development on carbon emission efficiency decreases, which also indicates that industrial structure transformation is an important channel through which digital economy development promotes carbon emission reduction. Therefore, Hypothesis 2 is strongly supported.

The development of a digital economy might have a nonlinear effect on carbon emission intensity by influencing energy consumption intensity in each country. As shown in Column (3) of Table 7, when energy use intensity is taken as the independent variable, the regression coefficient of the primary term of the TIMG index is significantly positive, while the coefficient of the square term is significantly negative, indicating that there is an inverted U-shaped relationship between the development degree of the digital economy and the energy consumption of a country, and the corresponding turning point of the TIMG index is 0.665. In the early stage of the development of the digital economy, a large amount of infrastructure needs to be built; in addition, the development of information enterprises consumes a large amount of electricity, both of which lead to an increase in energy consumption. However, when the level of digital economy development is high enough, the energy use efficiency of related enterprises will improve, so the energy consumption per unit GDP of the manufacturing industry will decrease, and the total energy consumption will decrease. Under the two opposite effects of the digital economy on energy use intensity, its development first has a positive effect on energy use intensity, followed by a negative effect.

In addition, from the results in Column (5) in Table 7, the marginal impact of energy use intensity on the carbon emission intensity of each country is 0.353–0.025*energy_int, and the corresponding inflection point for energy consumption intensity is 14.14. According to the results of the descriptive analysis in Table 2, the energy consumption intensity of most countries is lower than that value, thus, there is a positive relationship between energy consumption and carbon emission intensity for most countries. A comparison of the regression coefficients of the TIMG index and its squared term in Columns (1) and (5) in Table 7 reveals that after controlling for energy consumption intensity, the absolute values of the coefficients of the first and square terms of the TIMG index significantly decrease, indicating that the development of the digital economy first has a positive and then a negative impact on the carbon emission intensity of countries by affecting energy consumption. In the existing research, while Li et al.^4,5 and Wang et al.³³ have confirmed an inverted U-shaped relationship between digital economy and carbon emission intensity, they have not delved into the underlying mechanisms responsible for this relationship. We uncover that variations in energy consumption intensity serve as a significant factor contributing to this inverted U-shaped pattern. Furthermore, through the analysis presented in Table 6, it is validated that industrial upgrading plays a unidirectional role in the influence of digital economy on carbon emission intensity. Specifically, the digital economy exerts a one-way inhibitory effect on carbon emission intensity solely through the upgrading of industrial structure, without any stimulating effect. This finding aligns with the conclusion drawn by Dong et al.⁸ on the unidirectional suppression of digital economy on carbon emission intensity.

Heterogeneity analysis

Regression results for the TIMG subdivision indicators

In the analysis above, we use the TIMG composite index as a proxy variable for the level of digital economy development in various countries and find that the overall relationship between the digital economy development and the carbon emission intensity of various countries is inverted U-shaped. However, the TIMG index represents only the overall level of digital economy development and cannot account for the impact of the development of various dimensions of the digital economy on carbon emissions.

Next, this paper uses the subdivision index of the TIMG index to explore the impact of digital economy development on carbon emission reduction in four dimensions: digital technology, digital infrastructure, the digital market and digital governance. The main explanatory variables in the above regression were replaced by the subdivision indicators of the TIMG index, and panel analysis was carried out. As shown in Table 8, first, the primary regression coefficient of each subindex is significantly positive at the 1% significance level, while the square term is significantly negative at the 1% level, indicating that the impact of digital economy development on carbon emission intensity is valid in multiple dimensions. From the inflection point value, we can see that in addition to the digital technology index and digital infrastructure, the inflection point value of the digital market index is relatively low, meaning that it has a strong negative effect on carbon emission intensity. The larger the scale of the digital market is, the greater the energy consumption per unit of output value. The more people use the internet, the easier it is to create new output through mobile phones and other communication devices. Moreover, with the expansion of the digital market, the development of emerging businesses such as e-commerce can effectively reduce energy consumption and logistics costs while operating traditional retail enterprises, thus promoting carbon emission reduction. Although the impact of digital governance on carbon emission intensity remains in an inverted U-shaped pattern, its negative effect is notably weaker compared to other sub-indicators.

Table 8 Heterogeneity analysis of subdivision index of TIMG index.

Regression results for different groups of countries

For countries with different income levels, the impact of digital economy development on carbon emissions might be different. We then divide countries in our sample into three groups: high-income countries, upper-middle-income countries, low-income countries and lower-middle-income countries. We carry out heterogeneity analysis and study the effects of the digital economy and carbon emissions in three groups of countries separately; the results are shown in Table 9. For different groups of countries, an inverted U-shaped relationship still exists for the two variables. The coefficient of the linear term of the TIMG index is significantly positive, while the coefficient of the quadratic term is significantly negative; however, the significance level is different for different country groups. In addition, the inflection points of the TIMG index were 0.639, 0.857 and 0.833. For high-income countries, the negative effect of digital economy development on carbon emissions is much stronger than that for other groups. The industrial structure in these countries is more advanced than that in other regions, and tertiary industry makes a major contribution to economic development in these countries. Therefore, when energy use is more efficient, industrial structure upgrading and energy consumption reduction effects are stronger, the negative impact of the development of the digital economy on carbon emission intensity is more obvious. For the other two groups of countries, the development of the digital economy is in the early stage, added value of secondary industry as a proportion of the whole economy is relatively high, and much digital infrastructure is still being built, which consumes a large amount of energy. In addition, the positive effect of digital economy development on industrial upgrading and energy efficiency has not been fully reflected. Therefore, the development of a digital economy will ultimately increase carbon emission intensity in most countries.

Table 9 Heterogeneity analysis of the TIMG index and carbon emission intensity in different country groups.

Spatial spillover effects of digital economy development on carbon emissions

Spatial autocorrelation test

Before examining the spatial spillover effect of digital economy development on carbon emission intensity, we test the spatial autocorrelation of lncei and TIMG carbon emissions through Moran’s index to examine whether carbon emission intensity and digital economy development in one country are affected by those in neighboring countries. Table 10 lists the average global Moran’s index for carbon emission intensity and the TIMG index for each year. It can be seen that in each year of the sample period, the global Moran’s indices are both positive and significant at the 1% level, which implies that carbon emissions and digital economy development between neighboring countries have strong similarities. By comparing Moran’s index in each year, we can see that the global Moran’s index for two countries shows an upward trend as time progresses, indicating that the average degree of spatial autocorrelation gradually increases. For example, in 2013, the global Moran’s index of carbon emission efficiency was 0.108, while in 2020, the value rose to 0.172. Moreover, Moran’s index has significantly improved, indicating that we should give attention to the spatial spillover relationships among variables in different countries. In the following analysis, we utilize the spatial panel regression model to study the relationship between two variables and investigate the impact of digital economy development on carbon emission reduction in neighboring countries. Although spatial auto-correlation among countries’ carbon emission intensities could be seen in current studies (Khezri et al. 2021), to the best of our knowledge, there are no research that have considered the spatial auto-correlation of digital economy development across countries. Our finding has filled this gap in existing literature.

Table 10 Global Moran’s index of carbon emission intensity and the TIMG index.

Spatial panel regression model results

The development of the digital economy in a specific country will not only have an impact on the level of carbon emissions in that region but might also affect the carbon emissions in neighboring regions. We use a spatial panel regression model to investigate this potential effect.

First, an LR test is conducted to select the appropriate spatial panel regression model. As shown in Table 11, at the significance level of 1%, the null hypothesis that the spatial Durbin model (SDM) could degenerate into the SEM and SAR models could be rejected (the two LR test values are 56.77 and 56.94, respectively, with corresponding p values both less than 0.001). Therefore, we choose the SDM to study the spatial spillover effect of digital economy development in one country on carbon emission intensity in another country.

Table 11 Likelihood ratio test for spatial model selection.

The marginal impact of the digital economy on carbon emission intensity might not be fully reflected in the estimated coefficient of the SDM³⁵(Chen et al. 2020). In addition to the regression coefficients estimated by the SDM, Table 11 summarizes the direct, indirect and total impacts of each explanatory variable on CO2 emission efficiency. “Indirect effects” represent the effects of explanatory variables in neighboring countries on CO2 emissions. The regression coefficients of the first and square terms of the TIMG index are quite different from those of the baseline regression in Table 2. In the spatial econometric model, the spatial correlation between variables in different countries is taken into account, and the relationship between variables is more accurately described.

As shown in Table 12, for the primary term of the TIMG index, the coefficients for direct effects, indirect effects and total effects are all significantly positive at the 1% level, while the square terms are all significantly negative, indicating that the digital economic development of neighboring countries has an inverted U-shaped nonlinear spillover effect on their own carbon emission intensity. On the one hand, digital economy development in one country could reduce the carbon emission intensity of neighboring countries through information sharing and technology spillover effects; on the other hand, due to the “siphon effect”, high-tech industries and related talent might transfer to other countries, which inhibits industrial structure upgrading in neighboring countries and the improvement of their carbon emission intensity. This proves the accuracy of Hypothesis 4 proposed in Sect. 3.This conclusion underscores that the development of digital economy in one country not only impacts its own carbon emission intensity but also extends its influence to neighboring countries. While numerous scholars have conducted studies on the spatial spillover effects of digital economy within a country’s boundaries, using provincial or prefecture-level data³, to our knowledge, there has been a lack of research examining the spatial spillover effects of digital economy development for one country on the environmental conditions of neighboring nations. This paper bridges this gap. In addition, it points to a potential research direction for future investigations on digital economy at the transnational level, urging scholars to focus on the spatial spillover effects of digital economy development among nations.

Table 12 Estimation results of spatial econometric models for the full sample.

Robustness test

In the above analysis, we select the reciprocal of the square of the geographical distance between two countries as the constituent element of the spatial matrix. To conduct a robustness test, we select the reciprocal geographical distance between two countries as the element of the spatial matrix. The results are shown in Table 13. The TIMG index also has a nonlinear indirect effect on carbon emission intensity. This indicates that the development of the digital economy in a country will still have a first positive and then negative effect on the carbon emission intensity of neighboring countries, which is consistent with the result above and thus proves the accuracy of Hypothesis 4.

Table 13 Spatial econometric model robustness test.

Discussion and limitations

Using the TIMG index as a measure of the development level of digital economy in various countries, this article conducts panel regression models and finds an inverted U-shaped relationship between the development of digital economy and carbon emission intensity. In the analysis of cross-country panel data, Dong et al.⁸ found that the digital economy in each country negatively impacts carbon emission intensity but positively affects per capita carbon emissions. They further argue that the mechanisms of the digital economy’s impact on carbon emission intensity mainly include promoting economic growth, financial development, and industrial restructuring. However, this study indicates that regardless of whether using carbon emissions per unit of GDP or per capita carbon dioxide emissions as the dependent variable, the development of the digital economy exerts an inverted U-shaped influence on both, which aligns with the conclusions drawn by Li et al.^4,5 and Wang et al.²⁹. However, neither Li et al.⁵ nor Wang et al.²⁹ have analyzed the channels that lead to the inverted U-shaped effect of the digital economy on carbon emission intensity or its spatial spillover effect on neighboring regions. Based on the aforementioned literature, this paper finds that the digital economy primarily exerts an inverted U-shaped influence on carbon emission intensity by affecting energy efficiency, while it negatively impacts carbon emission intensity through industrial restructuring. Furthermore, using a spatial spillover model, we discover that the development of digital economy in various countries can also generate an inverted U-shaped spatial spillover effect on neighboring countries, thereby enriching research on the relationship between digital economy and carbon emission intensity at the cross-country level.

This study also has certain limitations. We mainly focus on the macro-level and does not take into account the industry heterogeneity. The effect of digital economy on carbon emission might be different across industries. A future avenue to study this topic could focus on the heterogeneity. Besides, given the role that the energy consumption play in this study, we could exploring the role of renewable energy consumption in this study.

Conclusions and policy recommendations

Climate change is a common challenge faced by all countries worldwide. Moreover, the proportion of the digital economy contributing to the global economy is increasing rapidly. Understanding the relationship between these two variables is helpful for governments in all countries to implement better carbon emission reduction policies. However, current empirical research examining the relationship between the two variables using cross-country data remains scarce. Among the limited studies, there is a dearth of thorough analyses delving into the mechanisms through which the development of digital economy in various countries impacts carbon emissions as well as the spatial spillover effects. To bridge this gap, we employ a cross-country panel dataset encompassing 72 nations from 2013 to 2020 to conduct a empirical research. Leveraging the mediation effect model and spatial spillover effect model, our analysis delves into the underlying mechanisms of how the digital economy influences carbon emissions, as well as the potential spatial spillover effects between these two variables.

We find that there is an inverted U-shaped relationship between the digital economy and carbon emissions across countries. Digital economy development has a positive effect on carbon emissions in the early stage, and then, its impact turns negative after it reaches a specific inflection point, which supports the ECK hypothesis. which is in line with the current study and ECK hypothesis.

In addition to the aforementioned, our study also yields the following novel findings compared to existing literature:

Firstly, we investigate the mechanism how digital economy can affect carbon emission intensity. We find that the development of digital economy can promote carbon emission reduction by promoting industrial structure upgrading. We also show that an important channel that digital economy could have an inverse U-shaped relationship with carbon emission intensity is through energy consumption intensity. This underscores the importance for governments worldwide to leverage digital technologies effectively to drive industrial upgrading while concurrently enhancing energy efficiency, thereby fostering more pronounced carbon emission reduction outcomes within their respective nations.

Secondly, we also use the panel spatial model to study the spatial spillover effect of a country’s digital economy development on neighboring countries. We find that the inverted U-shaped impact of the digital economy on carbon emission intensity not only exists in the country itself, but also in neighboring countries. Currently, there is no literature that verifies the spatial spillover of the digital economy on neighboring countries at the transnational level. This article fills this gap. At the same time, this discovery also reflects the importance of unity and cooperation among countries in developing the digital economy and environmental protection.

Thirdly, we make use of the TIMG index as a proxy variable for the level of digital economic development. This indicator includes multiple dimensions, which can more completely depict the development level of the digital economy. We conduct the heterogeneity analysis between the sub-dimensions of the digital economy, and finds that the development of digital technology, digital infrastructure and digital market have a greater effect on carbon emission reduction than digital governance. This finding helps to further understand the relationship between the digital economy and carbon emission intensity.

Finally, we investigated the relationship between the digital economy and carbon emission intensity for countries at different stages of development. As a result, it was found that digital economy has a more rapid negative effect on carbon emissions in high-income countries, while for most countries in middle-level and low-income countries, digital economy development and carbon emissions generally have positive effects on carbon emissions. This means that different types of countries should use different ways to adopt the digital economy to reduce carbon emission intensity.

The above conclusions have some policy implications. Firstly, Governments of various countries should adopt diverse strategies tailored to their respective levels of digital economy development, utilizing the digital economy as a tool to address environmental issues. This paper found that for less developed countries with a comparatively low level of digital economy development, the government should pay much attention to the balance between digital economy development and carbon emission reduction. For developed countries, especially those with a high level of digital economy development, they should fully focus on how to reduce carbon emissions with digital technology. Secondly, governments should encourage the integration of the digital economy and the real economy, use the digital economy to drive the upgrading of industrial structures in various countries, pay attention to the layout of low-pollution and low-energy industries and promote the coordinated development of green finance and the digital economy, which will help the economy shift towards a more environmentally friendly direction. Lastly, governments must intensify national cooperation with neighboring countries regarding digital economy development and carbon emission reduction. Our study reveals an inverted U-shaped spatial spillover effect of a county’s digital economy on neighboring nations. Hence, the development of digital economy in countries with a low level of digital development might hinder neighboring nations’ carbon reduction efforts. It is crucial that countries should make cooperation to reach carbon neutrality. Countries within the same region can establish a platform for sharing information of digital economy development and sustainable practices. By collaborating and learning from each other’s experiences, these nations can work towards a more integrated and sustainable development path, ensuring that their digital economies contribute positively to global climate goals.

This study also has certain limitations. We mainly focus on the macro-level and does not consider the heterogeneity across industry. The effect of digital economy on carbon emission might be different across different industries even in one country. The future study could focus on the heterogeneity using micro-level data. Besides, given the role that the energy consumption play in this study, we could exploring the role of renewable energy consumption in this study.