The Non-Linear Impact of the Digital Economy on Carbon Emissions Based on a Mediated Effects Model

Abstract

In an increasingly digital age, studying how the digital economy affects carbon emissions is important for China’s dual carbon goals. Based on the panel data of 30 provinces and regions in China from 2012 to 2020, the impact of the digital economy on carbon emissions in China and the mechanism of its effect are empirically analyzed. This study reveals that the digital economy has an inverted U-shaped impact on China’s carbon emissions. Initially, it promotes carbon emissions but later inhibits them. The carbon emission reduction effect is more significant after the digital economy reaches the inflection point of the inverted U-shape in the more economically developed regions. Energy structure and energy use efficiency are the two key factors through which the digital economy affects China’s carbon emissions. Among them, the digital economy shows an inverted U-shaped effect on energy structure, which is first positive and then negative, and a positive U-shaped effect on energy use efficiency, which is first negative and then positive. Based on the above findings, this paper suggests that: First, to achieve peak carbon and carbon neutrality, the digital economy needs to be strengthened and the foundations of the digital economy need to be consolidated. Second, the digital transformation of the energy sector should be accelerated, and digitalization should lead to the low-carbon energy transformation. Finally, in the process of developing the digital economy, attention should be paid to the rebound in energy consumption caused by a large number of basic digital facilities, and the low-carbon integration of the digital economy and traditional industries is of great significance in reducing carbon emissions.

1. Introduction

In recent years, the issue of climate change has become a pressing concern for the survival and development of our planet. There has been a global consensus to protect our shared home, and as the world’s largest carbon emitter, China’s efforts in energy conservation and emission reduction are crucial to environmental governance and climate protection. The approaching “carbon peaking and carbon neutrality” has greatly accelerated the transformation of production and lifestyles. Carbon peaking and carbon neutrality is not only China’s commitment to writing green and protecting ecology but also the environmental constraint to achieve high-quality economic development [1]. How to achieve high-quality and stable economic development under the constraint of a “double carbon target” is a difficult problem for China to solve in the future.

The rapidly evolving digital economy offers a possible optimal solution to this dilemma. With its strong penetration and wide integration, digital technology has profoundly changed social production methods and industrial structures and has become a new driving force for economic development. Information and communication technologies (ICT), including cloud computing, AI technology, 5G, and so on, have great potential for low-carbon industrial transformation, and the positive interaction between digital industry and low-carbon emission reduction can help achieve the dual carbon goal as soon as possible [2,3]. However, the digital economy is not synonymous with a low-carbon economy; the productivity brought about by the technological revolution has led to more energy demand, and the expansion of digital infrastructure has led to huge electricity consumption [4]. The carbon emissions of the digital economy itself are growing rapidly as it expands.

The digital economy is a new form of an economy with data resources as the key element, modern information networks as the main carrier, and the integration and application of information and communication technologies and digital transformation of all factors as the important driving force to promote greater unity of equity and efficiency [5]. Researchers, institutions, and others quantify the digital economy based on its connotation, including four methods: the GDP accounting framework system, the study on measuring the value added of the digital economy, the construction of an index system to compile indices, and the preparation of satellite accounts for measurement. In general, the economic output belonging to the digital sector is easier to account for, but the scope of the digital economy created by the unique permeability characteristics of digital technology cannot be accurately described.

Existing studies show that the carbon reduction effect of the digital economy is mainly reflected in the energy mix and energy use efficiency [6]. On the one hand, the digital economy has a suppressive effect on energy consumption. This is directly reflected in the fact that it strongly promotes the development of clean energy and reduces the share of traditional energy consumption [7]. For example, the development of the information industry in the US has had a significant impact on reducing energy consumption in traditional industries [8]. In addition, the digital economy has had a relatively positive impact on reducing the use of fossil energy and thus carbon emissions [9]. On the other hand, the digital economy has a positive impact on the promotion of energy use and plays an important role in reducing consumption and improving efficiency in the energy sector, and the deep integration of the two forces companies to innovate in decarbonization technologies. The digital economy can foster technological progress, thereby improving the efficiency of energy use and reducing the pressure on carbon emissions. For example, increased ICT investment can promote process innovation [10], and new technologies lead to the lower energy consumption of products and lower energy consumption, which improves the efficiency of energy use and thus effectively reduces carbon emissions. Some scholars have also focused on the technological effects of the digital economy itself, with May et al. [11] arguing that the mere application of ICT in the automation of production processes can reduce energy consumption and improve the efficiency of energy use. At the same time, the technological progress embedded in ICT itself promotes energy saving and emission reduction in other parts of the enterprise through spillover effects, eliminating the unreasonable parts of the production process and promoting efficient energy use [12,13].

In summary, the digital economy is a significant driver of low-carbon development and an enabler for achieving our carbon peak and carbon-neutral goals [14,15]. However, the digital economy does not mean a low-carbon economy, and the rapid development of the digital economy itself can lead to huge carbon emissions. Steffen Lange et al. [16] argue that instead of saving energy, digitalization brings additional energy consumption and increases carbon emissions. Through the above literature review, the existing studies have explored the two-way relationship between the digital economy and carbon emissions and energy consumption more deeply, but there are also some shortcomings. (1) The impact of the digital economy on carbon emissions is complex and has two sides. The existing literature focuses more on the positive carbon emission reduction effect of the digital economy but ignores the high carbon emission of the digital economy itself and fails to pay attention to the non-linear characteristics of the impact of the digital economy on carbon emissions. (2) Under the assumption that the digital economy has a non-linear effect on carbon emissions, the path of the non-linear effect has not received sufficient attention. To address the above problems, this paper tries to make the following extensions based on existing studies: First, to summarize the connotation of the digital economy from four dimensions: digital infrastructure, digital industrialization, industrial digitalization, and digital economic development, and to quantify the level of digital economy development in each provincial and regional area in China. Second, under the non-linear assumption of carbon emission reduction in the digital economy, the interrelationship between the digital economy and carbon emission as well as the mechanism of action is theoretically analyzed. The panel data of 30 provinces and regions (excluding Tibet) in China from 2012 to 2020 are selected to empirically demonstrate the impact of the digital economy on carbon emission, and the contributions of energy structure and energy use efficiency in carbon emission reduction are sequentially examined. Finally, the accuracy of the above empirical results is verified by comparing different digital economy calculation dimensions based on the entropy value method and the distance between superior and inferior solutions method, which strongly ensures the robustness of the empirical results.

2. Theoretical Analysis and Research Hypothesis

2.1. Digital Economy and Carbon Emissions

The development of the digital economy based on data production factors is having a disruptive effect on economic consumption. Paperless offices, network communication, and tram travel have reduced the consumption of traditional fossil energy to a certain extent, and energy consumption is gradually becoming low-carbon and green. Digital technology reduces energy losses on the supply, production, and consumption sides of energy. In addition, the digital economy can promote technological progress, thereby improving the efficiency of energy use and reducing the pressure on carbon emissions [17]. However, there is a downside to carbon reduction. The rapid development of the digital economy is based on a large amount of infrastructure, and its energy consumption is growing exponentially. Studies show that in a worst-case scenario, communication technologies (CTs) could consume 51% of the world’s electricity by 2030, and their carbon emissions from electricity use could account for 23% of global greenhouse gas emissions [18]. The authors use the Life Cycle Assessment (LCA) method to collect data on the emissions and resource consumption associated with a product or service. Second, energy consumption is estimated by measuring indicators such as user-directed usage time, charging time, power consumption, and battery charging time. The production of electronic chips, devices, etc., increases energy consumption, and electricity consumption has a positive and significant impact on CO₂ emissions. Secondly, the digital economy has led to the creation of many new businesses, and the high production has triggered an induced demand for energy, resulting in the phenomenon of energy rebound [19]. Only when the digital economy is combined with green and low-carbon development, which truly takes into account the simultaneous improvement in productivity of the digital transformation of low-carbon industries and the efficiency of carbon emission reduction, can digital technology become a powerful tool for energy saving and emission reduction in China, and effectively reduce carbon dioxide emissions. Based on this, the first hypothesis of the impact of the digital economy on carbon emissions is proposed.

H1: 

The impact of digital economy development on carbon emissions follows an inverted U-shaped non-linear relationship.

2.2. Digital Economy, Energy Mix, and Carbon Emissions

The traditional energy structure and consumption pattern, which is highly dependent on fossil energy, is a deep ditch and high barrier on the double carbon road. China is a country where coal is the main source of energy. Burning large amounts of coal causes large amounts of carbon emissions, and the sulfur dioxide in the exhaust gas further aggravates acid rain pollution. The coal-based energy structure makes environmental problems more and more serious. Therefore, energy structure adjustment is one of the important tasks for China to achieve the double carbon goal. Adjusting China’s energy structure means reducing the consumption of fossil energy resources and vigorously developing new and renewable energy resources. The digital economy, based on digital technology and information networks, creates the conditions for energy transformation and the popular application of renewable energy [20]. Digital technology is crucial to the development of the smart grid in terms of large-scale grid integration of new energy, safe and controllable grid load, intelligent use of energy to flexibly meet people’s changing needs, solving the problems of new energy power transmission, consumption, and storage, and providing technical support for the widespread application of new energy. While digital technology has great potential for the development and use of renewable energy, it has also significantly changed the traditional pattern of energy consumption [21]. In addition, with the rapid development of the digital economy, energy consumption is gradually adjusting to new industries, which to some extent reduces the proportion of traditional high-consumption industries and optimizes the energy structure. In short, the digital economy promotes the development and use of new energy sources [22], reduces traditional fossil energy consumption, and improves the energy structure.

2.3. Digital Economy, Energy Efficiency, and Carbon Emissions

The carbon-reducing effect of the digital economy on energy efficiency occurs in three main ways. First, the digital economy increases the flow of traditional production factors and promotes the flow of supply and demand information across industries, sectors, and regions, thus promoting the rational allocation of resources and improving the efficiency of energy use [23]. With the help of digital technologies such as big data and cloud computing, enterprises can keenly grasp the behavioral preferences of consumers and make more rational production plans. In addition, the digital application of the production process can accurately allocate resources in each link and optimize the production process, thus improving the efficiency of energy use and reducing carbon emissions. Second, the digital economy promotes technological innovation, supports research and development of low-carbon technologies, and improves the efficiency of energy use [24]. On the one hand, the economic impact of the digital economy has gradually emerged, and the digital financial platform has realized multi-channel innovation financing, which provides stable financial support for technological innovation and promotes technological innovation. On the other hand, technological advances in ICT have promoted energy saving and emission reduction in other industries through spillover effects [25], and new technologies have led to lower energy consumption of products and effectively reduced carbon emissions. Third, the digital economy saves production costs and reduces energy waste in production [26]. Digital technology has changed the past “personal judgment” and “empiricism” production model, and data-driven precision products have improved production quality and effectively reduced the redundant aspects of the production process that waste energy. In summary, two pathways are proposed for the carbon reduction impact of the digital economy.

H2: 

The development of the digital economy can contribute to carbon reduction by optimising the energy mix.

H3: 

The development of the digital economy can contribute to carbon reduction by improving the efficiency of energy use.

3. Materials and Methods

3.1. Benchmark Multiple Regression Model

To examine the non-linear effect of the digital economy on carbon emissions, a panel regression model was constructed based on the study by Yunfei Xie [27] by adding the squared term of the digital economy as follows:

$$lnca{r}_{i,t}=a_0+a_{1}di{g}_{i,t}+a_{2}di{g}_{i,t}^2+\phi X_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(1)

where $lnca{r}_{i,t}$ represents the carbon emitted by the province i in a year t, and $di{g}_{i,t}$ denotes the level of digital economy development in the province i in a year t, $di{g}_{i,t}^2$ denotes the square of the digital economy in the province i in a year t, and $X_{i,t}$ is a control variable that includes the level of economic development, demographic factors, industrial technology, and energy consumption. ${\mu }_i$, ${\lambda }_t$, and ${\epsilon }_{i,t}$ denote province fixed effects, time effects, and error terms, respectively.

3.2. Intermediate Mechanism Test

Theoretical analysis suggests that the digital economy can reduce carbon emissions by improving energy use efficiency $ln{ee}_{i,t}$ and optimizing the energy mix ${lnem}_{i,t}$.The model is as follows [28]:

$$M_{i,t}={\beta }_0+{\beta }_{1}lndi{g}_{i,t}+{\beta }_{2}di{g}_{i,t}^2+\rho X_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(2)

$$lnca{r}_{i,t}={\gamma }_0+{\gamma }_{1}lndi{g}_{i,t}+{\gamma }_{2}di{g}_{i,t}^2+{\gamma }_3{M}_{i,t}+\tau X_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(3)

where $M_{i,t}$ denotes each mediating variable, namely energy use efficiency, energy structure, and industrial structure. First, the baseline regression of equation (1) is performed, if $a_1$ is significant and $a_2$ is negative, it means that the digital economy has an inverted U-shaped effect on carbon emissions; second, Equation (2) is applied to test the effect of the digital economy on the mediating effect variables, and if ${\beta }_1$ passes the significance test, then proceed to the next step; finally, Equation (3) is applied to test the effect of the mediating effect variables on carbon emissions through ${\gamma }_1$ and ${\gamma }_3$. The significance of the mediating effect is determined by whether it is a partial or full mediating effect.

3.3. Variable Selection and Data Sources

3.3.1. Explained Variables

Carbon emissions ($lnca{r}_{i,t}$): measured by the total annual carbon emissions of each province, taking the natural logarithm data from the China Stock Market and Accounting Research Database (CSMAR).

3.3.2. Core Explanatory Variables

Digital Economy Development Index ($lndig$): This paper follows the State Council’s definition of the connotation of the digital economy and calculates the digital economy development index by referring to the China Digital Economy Development Report and the White Paper on China Digital Economy Development by the China Communications Institute in previous years. The specific indices are selected as the number of web pages, domain names, Internet broadband access ports, software business revenue, software product revenue, urban unit employment in information transmission, the software and information technology service industry, the product sales cost of the electronic information manufacturing industry, the number of enterprises in the electronic information manufacturing industry, length of optical fibre cable, mobile phone penetration rate, total telecommunications business, R&D expenditure of industrial enterprises above the scale, the number of enterprises with websites, the number of enterprises with e-commerce transaction activities, the proportion of enterprises with e-commerce transaction activities, fixed investment in information transmission, software and information technology service industries, fixed investment in scientific research and technology service industries, the number of patent applications, the number of R&D institutions, and the use of the entropy value method to obtain the comprehensive index value.

3.3.3. Intermediate Variables [29]

Energy mix (${lnem}_{i,t}$): The share of coal consumption in total energy consumption is chosen to measure the energy structure. Both coal consumption and total energy consumption are taken from China Statistical Yearbook and the China Energy Statistical Yearbook in previous years. Energy use efficiency ($ln{ee}_{i,t}$): The energy consumption of CNY 10,000 GDP is chosen to measure the energy use efficiency, i.e., the lower the energy consumption of CNY 10,000 GDP, the higher the energy use efficiency. The data are taken from the China Statistical Yearbook.

3.3.4. Control Variables

Based on existing research, relevant control variables are introduced, including the level of economic development ($lnpgd{p}_{i,t}$), demographic factors ($ln{p}_{i,t}$), energy consumption ($lne{s}_{i,t}$), and industrial technology ($ln{ti}_{i,t}$). The level of economic development is expressed by GDP per capita, the population level is measured by the year-end resident population of the province [29], the energy consumption is expressed by the total annual energy consumption of the province [30], and the industrial technology is calculated by the number of patent applications of industrial enterprises above the scale [31]. The variables $ln{inst}_{i,t}$, $ln{ip}_{i,t}$, $ln{cg}_{i,t}$, and $ln{dd}_{i,t}$ will be described in more detail when they appear later in this article. The above data are taken from the China Statistical Yearbook and the CSMAR database.

All variables are taken as natural logarithms. Descriptive statistics for the main variables are presented in

Table 1. Descriptive statistics of variables.

Variable	Observations	Average	Standard	Minimum	Maximum	Data Source	Data Duration
lncar	270	10.437	0.736	8.617	11.930	CSMAR	2012–2020
lndig	270	6.748	0.895	4.664	8.989	NSBC
lndig2	270	46.311	12.341	21.661	80.803	NSBC
lnpgdp	270	10.831	0.424	9.849	12.009	NSBC
lnp	270	8.208	0.741	6.347	9.444	NSBC
lnti	270	9.267	1.428	5.371	12.630	CSMAR
lnes	270	9.472	0.766	7.399	13.448	NSBC
lnee	270	−0.439	0.852	−3.684	4.077	NSBC
lnem	270	0.6384	0.265	0.009	2.734	NSBC
lnpeo	270	6.603	0.896	3.910	8.266	NSBC
lninst	270	0.826	0.232	0.477	1.832	CSMAR
lnip	270	−1.225	0.589	−5.001	−0.613	CSMAR
lncg	270	1.901	0.837	−0.916	3.962	CSMAR
lndd	270	0.102	0.099	0.120	0.563	NSBC

.

4. Results and Discussion

4.1. Benchmark Multiple Regression Analysis

Given the results of the Hausman test (p value of nil) for cross-sectional correlation and heteroskedasticity, the fixed effects model is chosen to estimate Equation (1) in this paper, and the regression results are presented in

Table 2. Benchmark multiple regression results.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Lncar					Lndig	Lndig
lndig	0.444 ***	0.461 ***	0.535 ***	0.470 ***	0.478 ***
	(0.115)	(0.115)	(0.118)	(0.106)	(0.109)
lndig2	−0.033 ***	−0.033 ***	−0.036 ***	−0.032 ***	−0.032 ***		0.069 ***
	(0.008)	(0.008)	(0.008)	(0.007)	(0.007)		(0.001)
lnpgdp		−0.131	−0.448 **	−0.505 **	−0.480 **	1.575 ***	0.097 ***
		(0.086)	(0.140)	(0.167)	(0.156)	(0.137)	(0.029)
lnp			1.460 ***	1.438 ***	1.423 ***
			(0.206)	(0.210)	(0.194)
lnti				0.065 **	0.054 **
				(0.022)	(0.021)
lnes					0.029 *
					(0.014)
Fixed time	Control	Control	Control	Control	Control	Control	Control
Province Fixed	Control	Control	Control	Control	Control	Control	Control
N	270	270	270	270	270	270	270

Note: t-statistics are shown in parentheses. Significance: * p < 0.1, ** p < 0.05, and *** p < 0.01.

.

The regression results in columns (1)–(5) of the table all show that the digital economy has a non-linear effect on provincial carbon emissions in China. Looking at the regression results in column (5), the coefficient of the quadratic term of the digital economy is significantly negative and the coefficient of the primary term is positive, both of which are significant at the 1% level. The other variables in column (5) are explained as follows:

First, higher levels of economic development can reduce carbon emissions: the coefficient of $lnpgd{p}_{i,t}$ is significantly negative at the 5% level. On the one hand, the digital economy promotes higher per capita income, which has a positive effect on economic development, as shown by the regression results in columns (6) and (7). The higher the level of economic development, the more residents will increase their demand for intangible culture, thus reducing material demand and lowering carbon emissions; in addition, the increase in per capita income will correspondingly increase residents’ environmental awareness, and various clean and low-carbon technologies will become popular, and they will continuously practice low-carbon behavior.

Second, the coefficient for the effect of the population factor on carbon emissions is significantly positive, and in terms of the magnitude of the coefficient, this is the most significant factor in increasing carbon emissions. The larger the population, the more carbon emissions it produces. The population factor is also a key factor in China’s high carbon emissions, which is in line with our national conditions.

Finally, the coefficient for industrial technology and energy consumption is significantly positive. This is because progress in industrialization is generally associated with increased energy consumption and industrial expansion, while the continuous improvement of industrial technology has facilitated industrial development and increased carbon emissions.

4.2. Robustness Analysis

The robustness test, which examines the robustness of the evaluation method and the explanatory power of the indicators, is performed by changing certain parameters and conducting repeated experiments to see if the empirical results change with the change in parameter settings. First, the data sample is reduced. Considering the inadequate digital infrastructure in low-income provinces and the low level of both digital economy development and carbon emissions, which may affect the impact of the digital economy on overall carbon emissions, four provinces with the lowest level of economic development are excluded from the data sample, and the regression results are shown in

Table 3. Robustness and endogeneity analysis.

	(1)	(2)	(3)	(4)	(5)	(6)
lndig	0.577 **	0.450 ***	0.504 **	0.423 ***	0.888 ***
	(0.173)	(0.090)	(0.225)	(0.129)	(0.262)
lndig2	−0.039 ***	−0.032 ***	−0.036 **	−0.028 ***	−0.060 ***
	(0.011)	(0.007)	(0.016)	(0.009)	(0.018)
lndd						−0.146 ***
						(0.034)
lndd2						−0.030 ***
						(0.007)
lnpgdp	−0.469 ***	−0.280 **	−0.366 *	−0.48 **	−0.517 ***	−0.536 ***
	(0.126)	(0.128)	(0.197)	(0.024)	(0.125)	(0.126)
lnp	1.288 ***	1.340 ***	1.361 ***	1.29 ***	1.487 ***	1.263 ***
	(0.188)	(0.227)	(0.437)	(0.000)	(0.226)	(0.226)
lnti	0.059 **	0.054 **	0.141 ***	0.06 **	0.038	0.054 **
	(0.025)	(0.021)	(0.046)	(0.013)	(0.027)	(0.027)
lnes	0.026 *	0.022 **	0.002	0.03 *	0.029 ***	0.029 ***
	(0.014)	(0.009)	(0.016)	(0.070)	(0.011)	(0.010)
lninst				−0.250 *
				(0.142)
lncg				−0.086 **
				(0.040)
Fixed time	Control	Control	Control	Control	Control	Control
Province Fixed	Control	Control	Control	Control	Control	Control
N	234	270	270	270	270	270

Note: t-statistics are shown in parentheses. Significance: * p < 0.1, ** p < 0.05, and *** p < 0.01.

, column (1). Second, the regression method is substituted. Panel PCSE estimation and FGLS estimation are used, and the regression results are shown in Table 3, columns (2) and (3). Third, control variables have been added. The carbon sink value of urban green space in urban ecology plays a non-negligible role in mitigating global warming and actively responding to climate change [32]. Urban green space is used to characterize the environmental benefits of the digital economy on carbon emissions. In addition, the industrial structure will be continuously upgraded to meet the new development needs led by the digital economy [33]. On the one hand, based on the strong permeability of digital technology, traditional production factors and data factors are integrated and circulated more efficiently within the industry, which promotes the efficient upgrading of the industry and effectively reduces industrial energy consumption. On the other hand, digitalization is integrated into all aspects of the industry to optimize resource allocation with accurate data. Therefore, the digital economy promotes the high-efficiency operation of industry and advanced industrialization, thus promoting the reduction of carbon emissions and increasing the control variables of an urban green area ($lnc{g}_{i,t}$) and an advanced industrial structure ($lnins{t}_{i,t}$); the advanced industrial structure is expressed by the ratio of the secondary industry value added to the tertiary industry value. The regression results are shown in Table 3 column (4). It can be seen that the quadratic term of the digital economy is always significantly negative, regardless of whether the sample is dropped, the method is changed, or control variables are added. The quadratic term of the digital economy is always significantly negative, which proves that the conclusions of this paper are robust.

4.3. Endogenous Discussion

Consider that there may be a two-way causal relationship between the digital economy and carbon emissions. On the one hand, there is a non-linear impact of the digital economy on carbon emissions, first promoting and then inhibiting. On the other hand, regions with higher carbon emissions may accelerate the digitalization process under the requirement of the dual carbon target. To avoid the endogeneity problem of bias in the regression results, this paper introduces Internet broadband users ($lnpe{o}_{i,t}$) as an instrumental variable. The core of the digital economy is informatization and data, and the network is the carrier of information transmission. According to Metcalfe’s law, the value of a network is equal to the square of its number of nodes, and the more the number of users in the network, the greater the total value of the network [34]. Therefore, regions with a developed digital economy are bound to be accompanied by a large group of Internet users. The specific regression results are shown in column (5) of Table 3. In addition, this paper uses the distance method of superior and inferior solutions instead of the entropy method to recalculate the level of digital economy development in each province and obtain a new digital economy ($lnd{d}_{i,t}$), where $lnd{d2}_{i,t}$ is its squared term, and the regression results are shown in Table 3 column (6). From the results in columns (5) and (6), the quadratic terms of the core explanatory variables of the digital economy are all significantly negative at the 1% level. After addressing the endogeneity issue, the conclusions of this paper still hold.

4.4. Intermediary Mechanism Test

The results of the empirical evidence so far show that the impact of the development of the digital economy on carbon emissions has an inverted U-shaped non-linear relationship of promotion and then suppression. However, the exact mechanism of the effect has not yet been determined. In this paper, we use stepwise regression to test the existence of the mediating effect, and the results are shown in

Table 4. Intermediate mechanism test.

	(1)	(2)	(3)	(4)	(5)	(6)
	Lncar	Lnem	Lncar	Lncar	Lnee	Lncar
lndig	0.453 ***	0.424 **	0.330 ***	0.453 ***	−5.169 *	0.436 ***
	(0.129)	(0.173)	(0.121)	(0.129)	(2.639)	(0.128)
lndig2	−0.031 ***	−0.030 **	−0.022 ***	−0.031 ***	0.342 *	−0.029 ***
	(0.009)	(0.012)	(0.008)	(0.009)	(0.186)	(0.009)
lnpgdp	−0.467 ***	−0.367 **	−0.360 ***	−0.467 ***	−0.046	−0.507 ***
	(0.121)	(0.162)	(0.113)	(0.121)	(2.470)	(0.122)
lnp	1.363 ***	0.050	1.349 ***	1.363 ***	−18.384 ***	1.279 ***
	(0.220)	(0.296)	(0.203)	(0.220)	(4.508)	(0.223)
lnti	0.068 **	0.043	0.055 **	0.068 **	−0.447	0.057 **
	(0.026)	(0.035)	(0.024)	(0.026)	(0.540)	(0.027)
lnes	0.027 ***	−0.179 ***	0.079 ***	0.027 ***	9.641 ***	0.123 **
	(0.011)	(0.014)	(0.013)	(0.011)	(0.215)	(0.052)
lncg	−0.069 *	−0.052	−0.054	−0.069 *	0.193	−0.081 **
	(0.039)	(0.052)	(0.036)	(0.039)	(0.792)	(0.039)
lnem			0.289 ***
			(0.046)
lnee						−0.132 *
						(0.070)
Fixed time	Control	Control	Control	Control	Control	Control
Province Fixed	Control	Control	Control	Control	Control	Control
N	270	270	270	270	270	270

Note: t-statistics are shown in parentheses. Significance: * p < 0.1, ** p < 0.05, and *** p < 0.01.

.

The results in columns (1) and (4) of Table 4 show that the inverted U-shaped effects of the digital economy on carbon emissions are significant at the 1% level. Columns (2) and (5) show that the digital economy also has a non-linear effect on the energy mix and the efficiency of energy use. The regressions in columns (3) and (6) prove that hypotheses 2 and 3 are valid and that energy mix and energy use efficiency have partial mediating effects.

Specifically, at the early stage of development, the digital economy has built a large amount of infrastructure, which is accompanied by huge electricity consumption, while thermal power generation, which relies on coal resources, is still the mainstream mode of power generation in China, leading to increased consumption of coal resources, and the energy-increasing effect of the digital economy (direct impact and economic growth) is greater than the energy-reducing effect (energy efficiency improvement and second turn change). When the digital economy develops to a certain level, it not only provides the technical support for the widespread use of new energy sources but also significantly changes the traditional energy consumption pattern and begins to have a positive impact on optimizing the energy structure. As shown in column (3), the digital economy has an inverted U-shaped effect on the energy mix, first promoting it and then inhibiting it. The energy use efficiency is expressed by the energy consumption of CNY 10,000 GDP, and the smaller the value, the higher the energy use efficiency. When the digital economy is less developed, a large amount of infrastructure equipment will have higher energy consumption, and its energy-saving and efficiency-enhancing effect will be smaller. As the level of digitalization increases, the economic effects of the digital economy and the ability of technology to radiate gradually emerge, promoting the flow of supply and demand information across industries, fields, and regions, and the rapid flow of traditional factors of production, promoting a more rational allocation of resources and improving energy use efficiency. In addition, the digital application of the production process can accurately allocate the resources in each link and optimize the production process, thus improving energy use efficiency. As shown in column (6), the digital economy has a positive U-shaped effect on energy use efficiency, which is negative and then positive.

4.5. Discussion

We note that as the share of digital technologies in the economy increases the share of traditional production sectors (of which industry is the most important) in GDP decreases. This may also affect the results, as the industry is a major emitter of greenhouse gases. As this article uses an index-based approach to measure the digital economy, it is not possible to accurately determine the share of digital economy output in total GDP, so we have considered the share of the industrial economy in GDP for our analysis to see whether a decline in the share of industrial output affects the inverted U-shaped effect of the digital economy on carbon emissions. The results are presented in

Table 5. Impact of declining share of industrial production.

	(1)	(2)	(3)	(4)	(5)
lndig	0.416 ***	0.479 ***	0.380 ***	0.350 ***	0.448 ***
	(0.121)	(0.129)	(0.141)	(0.126)	(0.135)
lndig2	−0.031 ***	−0.032 ***	−0.026 **	−0.023 ***	−0.030 ***
	(0.008)	(0.009)	(0.010)	(0.009)	(0.009)
lnip	0.015 *	−0.001	0.022	−0.009	−0.012
	(0.007)	(0.010)	(0.022)	(0.016)	(0.017)
lnpgdp		−0.480 **	−0.527 ***	−0.359 ***	−0.538 ***
		(0.157)	(0.124)	(0.113)	(0.127)
lnp		1.424 ***	1.185 ***	1.372 ***	1.289 ***
		(0.173)	(0.243)	(0.208)	(0.230)
lnti		0.054 **	0.060 **	0.052 **	0.050 *
		(0.023)	(0.027)	(0.025)	(0.028)
lnes		0.029 *	0.162 **	0.080 ***	0.029 ***
		(0.014)	(0.066)	(0.013)	(0.011)
lnee			−0.187 **
			(0.090)
lncg			−0.076 *	−0.058	−0.092 **
			(0.039)	(0.037)	(0.041)
lnem				0.290 ***
				(0.046)
lninst					−0.262 *
					(0.143)
N	270	270	270	270	270

Note: t-statistics are shown in parentheses. Significance: * p < 0.1, ** p < 0.05, and *** p < 0.01.

. The coefficient on the quadratic term for the digital economy is significantly negative when no control variables are considered, and the regression results are shown in column (1) of Table 5. When the other control variables are considered together, as shown in column (2) of Table 5, the results still hold well, and the main explanatory variables are all significant at the 95% confidence level. We also examine the possible impact of the decline in the industrial share on the mediating mechanism. As shown in columns (3) and (4) of Table 5, energy use efficiency and energy mix still show partial mediating effects in the mechanism of the effect of the digital economy on carbon emissions, and the mediating effect of the digital economy on carbon emissions becomes progressively larger as the share of the digital economy gradually increases due to the decline in the industrial share. In order to further increase the robustness of the results, we also considered adding control variables and the regression results are presented in column (6) of Table 5. The results are compared with the robustness tests in Section 4.2, which show that the inverted U-shaped effect of the digital economy on carbon emissions still significantly holds.

Existing studies have discussed the relationship between the digital economy and carbon emissions. For example, the digital economy affects carbon emissions through energy consumption, technological progress, and the modernization of industrial institutions. The relationship between the two is inverted U-shaped [35]. Some studies look at how carbon emissions can be affected under fiscal power by an industrial structure that has a U-shaped effect [36]. Given the correlation between traditional fossil energy consumption and environmental pollution, this article further explores how the digital economy has an inverted U-shaped effect on carbon emissions through energy under a dual carbon target, for example by considering the role of energy structure and energy use efficiency in this effect, thus extending research on the digital economy in the environmental field. In addition, this article considers the impact of different levels of digital economy development on carbon emissions in the empirical model, and the comparative results further expand research on the impact of the digital economy on low-carbon development.

We have tried to compare the results of the Chinese economy with other economies, both highly developed and developing. The level of development of the digital economy varies widely around the world today, and the difference between “hyper-digital” and “digitally weak” countries is becoming increasingly apparent. Some researchers have analyzed the impact of the digital economy on carbon emissions from a global perspective. Some scholars have shown that e-commerce harms carbon emissions in European economies by promoting the development of the digital economy, and that e-commerce has a significant carbon reduction effect in countries with medium to high carbon emissions [37]. In addition, studies have shown that economic growth and industrial structure play a more important mediating role between the digital economy and carbon emissions from a global perspective [29], while China relies more on the influence pathway in the energy sector. In particular, trade in digital products in the context of economic globalization is important for reducing carbon emissions, and carbon reduction are particularly effective in countries that develop digital trade and technologies and actively address environmental issues [38]. Overall, the development of the digital economy has significantly reduced carbon intensity and is an effective means of saving energy and reducing carbon emissions. Economically developed countries with more mature digital technologies contribute more to carbon emissions, which is consistent with the findings of this paper that the carbon reduction effect of the digital economy is stronger in provinces with higher economic levels after the inverted U-shaped inflection point is reached in the Chinese region.

5. Conclusions

5.1. Conclusions

This paper uses panel data from 30 provinces in China from 2012 to 2020 to construct a panel regression and mediated effects model to study the impact and mechanism of the digital economy on China’s carbon emissions. The main conclusions are as follows: First, the digital economy has an inverted U-shaped impact on China’s carbon emissions, which is first promoted and then suppressed. The digital economy can significantly reduce carbon emissions when it develops to a certain level, and the absolute value of the coefficient of the quadratic term of the digital economy becomes larger after removing some provinces with lower economic levels, indicating that the carbon emission reduction effect of the digital economy after the turning point is stronger in provinces with a higher economic level. Second, the digital economy can affect carbon emissions by affecting energy structure and energy efficiency, both of which have partial mediating effects. Specifically, the digital economy has an inverted U-shaped effect on energy structure, first promoting and then inhibiting, and a positive U-shaped effect on energy efficiency, first negative and then positive. When the digital economy develops to a certain level, the carbon emission reduction effect of the digital economy is realized through the optimization of energy structure and the improvement of energy use efficiency, and the emission reduction effect is obvious.

5.2. Policy Recommendations

Based on the above research results, the following suggestions are made: First, all provinces in China should accelerate the development and quality improvement of the digital economy, and regions with weak digital economy should pay special attention to infrastructure innovation and consolidate the digital technology base, to promote the digital economy to reach the inverted U-shaped inflection point as soon as possible and show a positive environmental effect of carbon emission reduction. Second, all provinces in China should strengthen the innovation of digital economy applications and focus on digital transformation in the energy sector. Energy use efficiency and energy structure are powerful ways for the digital economy to help achieve the dual carbon target. 5G, IT, AI, and other digital high technologies can effectively stimulate energy innovation and realize the green and low-carbon transformation of energy under digital leadership. Finally, due to the non-linear effect of the digital economy on carbon emissions, energy structure, and energy efficiency, when its development level is not high, the digital economy shows an inhibiting effect on carbon emission reduction. Therefore, while guiding and supervising the high-quality development of the digital economy, we should pay attention to the mutual integration of the digital economy and low-carbon development, and avoid the digital economy following the old path of “development first and governance later”. Therefore, while guiding and supervising the high-quality development of the digital economy, we should pay attention to the mutual integration of the digital economy and low-carbon development, avoiding the old path of “development before governance”.

5.3. Outlook

Although this is a study that examines how the digital economy affects carbon emissions, there are still some limiting gaps that could serve as a direction for future work. First, the digital economy covers a wide range of content, and the digital economy indicators in this paper suffer from shortcomings; therefore, a more comprehensive indicator system is needed. Second, this study only explores the impact of China’s digital economy on carbon emissions, and it is worthwhile to extend the study of this paper to other economies in the future, including countries with hyper-developed and average economic development. Third, this study only explores the impact of the digital economy on carbon emissions at the level of a single country, and it is a worthwhile direction to explore the heterogeneity of the impact of digital economy development on carbon emissions among international countries in the future.