Digital Economy, Green Dual Innovation and Carbon Emissions

Abstract

The digital economy serves as a pivotal catalyst for sustainable and eco-friendly development. This study employs a suite of advanced econometric models, including the fixed effects, mediation, threshold and moderation model, to elucidate the intricate dynamics by which the digital economy influences carbon emissions through the lens of green innovation. Building on the existing research on digital economy, green technology innovation and carbon emissions, this paper takes a dual-innovation perspective and divides green technology innovation into disruptive green technology innovation and incremental green technology innovation. And from the government and the public level, it explores how social concerns affect the effect of digital economy on carbon emissions. The analysis is grounded in a comprehensive dataset encompassing a decade of provincial-level data from 2011 to 2021 across China’s 30 provinces. The benchmark regression outcomes indicate the digital economy’s ability to substantially cut down carbon emissions; the threshold effect and mediating effect models demonstrate that a single-threshold effect exists and that disruptive and progressive green technological innovations mediate such ability. Additional research reveals that the digital economy’s impact on carbon emissions could be positively moderated by public and governmental attention. Eastern and western regions in China, as well as those with high levels of foreign investment and low levels of technological transaction activity, are more affected by the digital economy in terms of carbon emission reduction. Our conclusions offer practical recommendations for digital economy’s coordinated advancement and carbon emissions mitigation, and guide local governments to achieve sustainable development goals (SDGs).

1. Introduction

Air pollution, water pollution and other environmental issues have emerged as a result of rapid industrial growth. Consequently, global attention has shifted to sustainable development. Since the unanimous adoption of the 2030 Agenda for Sustainable Development by all UN Member States in 2015, the 17 sustainable development goals (SDGs), such as “Affordable and clean energy”, “Industry, Innovation and Infrastructure” and “Climate action”, have set the trajectory for sustainable development globally and catalyzed a worldwide movement for change. Nevertheless, escalating geopolitical tensions and increasingly severe climate change have posed significant challenges to the attainment of the sustainable development goals. In June 2023, BP released the “BP Statistical Review of World Energy 2022”, highlighting that the energy market was plunged into turmoil following the Russian–Ukrainian conflict, which has once again triggered profound thinking about affordable and sustainable modern energy in all countries. Meanwhile, gas and coal prices in Europe and Asia have surged to record levels. Despite this, global carbon dioxide emissions grew by 0.8 percent in 2022, while mainland China’s carbon emissions fell back slightly by 0.1 percent in 2022. In response, China is earnestly exploring integration strategies that harmonize the digital economy with initiatives aimed at curtailing carbon emissions. China values and recognizes the digital eco-civilization’s significance for green and low-carbon growth. Adopting carbon account and carbon credit policies, for instance, accelerates the synergistic transformation of digitization and decarbonization. These policies promote green and low-carbon production and lifestyles by creating positive incentives, aiming to “benefit the people in a green way”. The term “digital economy” refers to new types of business partnerships caused by Internet technology growth [1]. The digital economy is increasingly prevalent, underscoring its growth urgency in various areas. First, the digital economy serves as an emerging engine of economic growth, significantly supporting high-quality growth by overcoming the temporal and spatial constraints of the traditional economic models [2,3,4]. Second, it encourages industrial upgrading, stabilizes economic cycle and facilitates the construction of a modernized economic structure [5,6,7]. Third, the digital economy is instrumental in fostering environmentally sustainable development [8,9,10]. Consequently, examining the potential for China to leverage its burgeoning digital economy to achieve reductions in carbon emissions carries substantial implications for both theoretical discourse and practical policy-making.

Innovation in technology stands as a pivotal approach to combat climate change and to advocate for the diminution of carbon emissions [11,12], and is also an important driving force for achieving the sustainable development goals. Green technology innovation aims to further integrate low-carbon environmental protection considerations into technological innovation. With the implementation and practice of sustainable development goals, people worldwide have recognized the significant practical importance of green technology innovation. Compared to other types of technological innovation, the role of green technological advancements in curbing carbon emissions is notably pronounced. Augmenting green technological innovation within urban conglomerates can expedite economic expansion and foster industry growth in a clustered, interconnected and environmentally sound manner [13]. With global environmental problems becoming more prominent, green technology innovation is imperative [14,15]. A multitude of academicians have substantiated the mitigating influence of green technological innovation on carbon emissions [16,17], a pattern that is evident across the N-11 (N-11 countries include: Bangladesh, Egypt, Indonesia, Iran, Korea, Mexico, Nigeria, Pakistan, Philippines, Turkey and Vietnam) [18] nations, G7 countries (the G7 includes the United States, Britain, France, Germany, Japan, Italy and Canada) [19], Singapore [20] and BRICS nations [21]. However, there is a dearth of research that delves into green technological innovation from a dual innovation perspective. The theory of dual innovation is rooted in the dual characteristics of organizations. According to the different degrees of innovation, scholars divide dual innovation into two types: breakthrough innovation and progressive innovation [22]. Following this logic, green technology innovation can also be bifurcated into disruptive and progressive types, each distinguished by its unique attributes [23]. Disruptive green technology innovation, or “from zero to one” innovation, involves assimilating new knowledge and applying new technology to achieve innovation. This represents a profound revolution of the original technology, requiring high levels of innovation conception and resource integration. Progressive green technology innovation, or “adding bricks and mortar”, involves the improvement and growth of existing green technology. This type of innovation is less demanding in terms of the capabilities and resources of the innovator [24]. The integration of both radical and incremental green technological innovations is essential for refining the energy matrix and effecting a comprehensive societal shift towards low carbon footprints. Simultaneously, the growth of the digital economy and carbon emissions are strongly linked to public and government attention [25,26]. In Europe and the US, public governance has increasingly been incorporated into environmental governance models due to the growing severity of environmental issues and rising public awareness of environmental protection. In China, this model is still in its initial stage. Questions such as “why carry out environmental pluralistic governance” and “at which level should the center of gravity of environmental pluralistic governance be placed” continue to emerge.

Based on the above, considering that China has been launching a pilot carbon emissions trading program since 2011, we take the data from 30 provinces in China from 2011 to 2021 as a research sample to study the relationship between the digital economy and carbon emissions. Possible marginal contributions of this research: (1) Building upon the current literature on the correlation among “digital economy, green technology innovation and carbon emission”, we categorize green technology innovation into two distinct types: disruptive and progressive. Subsequently, we delve into an analysis of the underlying mechanisms that contribute to variations between the digital economy and carbon emissions. (2) Our investigation aims to ascertain whether there exists a threshold effect within the context of a dual green innovation transmission mechanism. (3) The study also explores the influence of societal focus, encompassing both governmental and public realms, on the interplay between these two factors. (4) We assess the diverse effects of foreign investment intensity and technology transaction activity based on existing research.

2. Theoretical Framework and Hypotheses

2.1. Digital Economy and Carbon Emissions

Existing studies on the electronic economic situation on carbon emissions remains in the ascendant. Some scholars think that the digital economic situation might create a rise in carbon exhausts [27,28], while this point of view holds that as economic degrees rise, high environmental quality might initially be experienced, yet is expected to rebound and boost when a certain financial criteria is gone beyond, supporting the Environmental Kuznets Curve (EKC) theory [29]. Along with the above perspectives, some scholars have argued for the carbon emission reduction duty of the digital economic climate from different perspectives. The digital economy itself is a low-carbon economy. On the one hand, as a fundamental part of the digital economy, industrial digitization minimizes carbon exhausts by integrating emerging information technology, enhancing development and technical inputs and upgrading the ability of industrial development. Additionally, the increase in industrial digitization in numerous areas can stimulate local commercial development, magnifying the emission-reducing impact of a commercial makeover [30]. On the other hand, Danish K found that the impact applied by economic expansion and metropolitan development on CO₂ emissions within specific industries has proven to be statistically negligible, suggesting that integrating brand-new forms such as digitization is more effective in decreasing carbon discharges and advertising financial development [31]. Digital finance, a vital part of the more comprehensive digital economy, has the capability to produce a variety of work opportunities, catalyze the increase in labor and stimulate market activity. It accomplishes this by fostering the growth of micro, tiny and medium-sized business (MSMEs) [32], and the building and construction of ecommerce [33], which subsequently cultivates low-carbon and adaptable cities. Combined with other existing studies, it can be summarized that the digital economy mainly affects carbon emissions through the scale effect, synergy effect and technology spillover effect.

From the perspective of scale effects, as the new round of technological revolution continues to develop, the knowledge flow and innovation concentration effects brought about by the bipolar transnational flow of inventive talent are becoming increasingly important [34]. The digital economy has triggered a shift in comparative advantage from traditional production factors to digital factors, and the international division of labor based on the new comparative advantage of digital resources will make inter-country cooperation and coordination more convenient, thereby triggering scale effects to improve the efficiency of the global value chain division of labor and achieve efficient allocation and utilization of energy, thereby achieving carbon emission reduction. From the perspective of synergy effects, Sun has empirically proven the dynamic effect of jointly promoting growth and carbon reduction through random shocks in digital infrastructure, digital industrialization and industrial digitalization by constructing a three-sector DSGE model for residents, manufacturers and governments [35]. Meanwhile, the digital economy eliminates some time and space limitations, which is conducive to breaking down information barriers between enterprises and industries and building a collaborative management model for industrial value chains. In this way, different carbon emission subjects can share emission reduction technologies and experience, promote collaborative emission reduction and jointly push forward the green and low-carbon transformation. From the perspective of the spillover effect of technology, the unique advantage of the digital economy lies in its ability to quickly compile and emulate complex advanced knowledge across regions or industries. This characteristic triggers the complementary fusion of diverse application scenarios for existing technologies, promoting the generation and circulation of non-native knowledge spillover effects and knowledge exchange across industries. It also restructures the existing local technology knowledge system and promotes the diversified development of new technologies [36]. The innovation and application of carbon emission reduction technologies can further promote the clean production of enterprises and drive CO₂ emission reduction. Therefore, Hypothesis 1 is proposed.

Hypothesis 1.

Digital economy growth can curb carbon emissions.

2.2. The Mediating Role of Green Dual Innovation

Expansion within the digital economy is instrumental in nurturing green technological advancements [37,38], with the understanding that its influence may vary between disruptive and progressive forms of green innovation. Existing research suggests several mechanisms through which the digital economy fosters technological innovation that is green: Firstly, for disruptive green technology innovation, the digital economy’s innovation and transformation, primarily evident through its integration with cutting-edge information technologies like big data and cloud computing, are capable of significantly advancing the development of digital infrastructure [39]. This alleviates the time and space constraints, laying a foundation for relevant units to pursue green technological innovation [40] and providing more opportunities for innovation entities to achieve disruptive green innovation. Secondly, for progressive green technology innovation, green technologies enhance the international competitiveness of enterprises [41]. In the digital era, green technologies have gradually become key tools for companies to benefit and expand their market share [42]. Therefore, while the digital economy significantly increases the total number of joint patent applications by enterprises, it also produces an obvious “catfish effect”: in order to maintain or increase market share in the competition, enterprises must constantly seek innovation to distinguish themselves from powerful rivals and improve their competitiveness. Competitors’ investment and breakthrough in green technology have not only enriched the variety and quantity of green products on the market, but also gradually improved consumers’ awareness and acceptance of green products. This change in consumer preferences has further stimulated green innovation by companies to meet market demand. At the same time, while the green technology breakthroughs made by powerful competitors have a demonstration effect on other enterprises, their innovation achievements and market performance also convey a clear signal to the market that green technology innovation has huge market potential, so that enterprises can further choose and upgrade green technology under the incentive of green innovation [43] and increase the output of progressive innovation. Additionally, digitization can augment green technology innovation levels by attracting government subsidies [44], thus significantly reducing the cost and providing positive incentives for green technology innovation behavior. Finally, digital transformation promotes knowledge sharing and resource integration, making full use of existing assets [45] and further promoting the green transformation of enterprises through scale effects, providing a flexible and open innovation environment for green innovation [46]. For most regions, resources are scarce compared to the demand for green technology innovation. Digitalization helps the rational allocation of limited resources, effectively alleviating insufficient allocation of talents, funds and technology in green technology innovation [47], thereby reaping greater economic and social benefits. Overall, the resource elements required for disruptive green technology innovation are more comprehensive, the capacity requirements for innovation subjects are higher and the risks faced in research and growth are greater. Consequently, the stimulative impact of factor mobility, propelled by the digital economy, could be more pronounced in the realm of disruptive green technology innovation compared to its progressive counterpart. In conjunction with the above, we present Hypotheses 2a and 2b.

Hypothesis 2a.

Digital economy growth helps cut down carbon emissions through disruptive green technology innovation.

Hypothesis 2b.

Digital economy growth helps cut down carbon emissions through progressive green technology innovation.

2.3. The Moderating Role of Social Concerns

Stakeholder theory indicates that the government and the public are not only the direct beneficiaries of the digital economy, but also an indispensable force in environmental governance.

When considering the government’s role, heightened environmental vigilance can compel authorities to craft and refine policies with greater efficacy and to enhance regulatory measures. This allows for the more efficient monitoring and assessment of social carbon emissions and ensures the implementation and enforcement of relevant policies. On one hand, by actively utilizing data elements, the government can broaden information communication channels and bolster the development and enforcement of environmental conservation strategies, thus enhancing the strength of environmental regulation policies and improving air pollution [48]. On the other hand, the growth in government attention promotes the growth and use of environmental data and information. This breaks the environmental information cocoon, forms a complete ecological environmental data information system and improves the transparency of local government environmental protection efforts [49], thereby promoting the comprehensive decarbonization of society.

From a public attention perspective, the digital economy has eliminated the asymmetry of environmental information resources and provided the public with more platforms to participate in environmental supervision and voice their concerns. By analyzing Internet search data, we can track netizens’ behavior in real time and gain insights into public preferences and behavioral intentions. Public attention to environmental issues reflects their environmental preferences and willingness to participate in governance, helping to narrow the urban–rural pollution gap [50]. Public attention to environmental issues directly affects low-carbon decisions in individuals’ daily lives. On one hand, public environmental demands can influence local governments’ environmental governance behaviors [51], improving the environmental efficiency of unsustainable cities [52]. On the other hand, the public’s concerns about the environment affect the sustainable behavior of enterprises. By enhancing market demand for green products, enterprises are guided to produce green products, further reducing regional carbon emissions.

Comprehensively, the digital economy’s role on carbon reduction is much more complex, showing a multi-path, multi-dimensional comprehensive impact [53]. Government attention is mainly reflected in its mandatory and authoritative nature, but it is difficult to cover all periods, fields and levels of carbon emission reduction comprehensively. Moreover, it takes time to complete the relevant resource allocation, relatively limiting its scope. Public attention tends to be diverse and stable. As long as environmental problems exist, public concern will continue and play a role. However, since it is mainly exerted through public opinion of the relevant object, its intensity is relatively limited. Nevertheless, as both government attention and public attention increase, entities responsible for carbon emission will be more inclined to use digital means to cut down carbon emissions by applying the technology for collecting, storing, analyzing and sharing data for energy conservation.

In conjunction with the above, we present Hypotheses 3a and 3b:

Hypothesis 3a.

Government attention moderates the relationship between digital economy growth and carbon emissions.

Hypothesis 3b.

Public attention positively moderates the relationship between digital economy growth and carbon emissions.

Drawing on the research findings and underlying assumptions, a conceptual framework illustrating the influence of the digital economy on carbon emissions is developed (Figure 1).

3. Empirical Analysis of the Impact of the Digital Economy on Carbon Emissions

3.1. Model Construction

To test the above research hypotheses, the following benchmark model is constructed to analyze the direct transmission mechanism:

$${ce}_{it}={\alpha }_0+{\alpha }_1{Dige}_{it}+{\gamma control}_{it}+{\mu }_{it}+{\lambda }_{it}+{\epsilon }_{it}$$

(1)

Equation (1) encapsulates the comprehensive effect of the economy on carbon emissions. Here, i denotes province and t denotes time. The dependent variable ce_it represents the carbon emissions of region i in period t. The key explanatory variable Dige_it indicates the growth level of the digital economy in region i in period t. control_it encompasses a suite of additional control variables that are known to influence carbon emissions. α₀ is the constant term, and α₁ represents the total level of them; μ_it is the region fixed effect, λ_it is the year fixed effect and ε_it is a stochastic error term.

3.2. Variable Description and Data Source

3.2.1. Variable Description

(1) Dependent variable

Carbon emissions (ce): This term here is defined as “carbon dioxide emissions”. The reduction depends mainly on technological progress and improving energy efficiency. In view of the fact that both fossil energy and the cement production process produce carbon dioxide, we draw on the calculation method of Zhu (2018) [54] and adopt the following formula to calculate carbon emissions:

$$ce={\displaystyle \sum _{i=1}^7{E}_i}\times {NCV}_i\times {CC}_i\times {COF}_i\times 44/12+k\times Q$$

(2)

where E signifies the aggregate energy consumption for the i-th province, with NCV_i denoting the calorific value of the i-th energy type, and CCi referring to the carbon content associated with that energy source. COF_i is indicative of the oxidation factor for the i-th energy source. The ratio 44/12 is the molecular weight ratio of carbon dioxide to carbon, and the emission factor for carbon dioxide from the i-th energy source can be derived from this formulaic relationship: ${NCV}_i\times {CC}_i\times {COF}_i\times 44/12$. Q is the volume of cement produced, and k denotes the carbon dioxide emission factor inherent in the cement manufacturing process. Specifically, fossil energy consumption can be subdivided into coal consumption, coke consumption, oil consumption (including gasoline, kerosene, diesel, fuel oil) and natural gas consumption. And the emission factors for seven types of fossil fuels—coke, coal, kerosene, diesel, gasoline, fuel oil and natural gas—as well as for cement production, are 2.848, 1.647, 3.174, 3.150, 3.045, 3.064, 21.670 and 0.527, respectively. In the process of primary energy consumption, a considerable part is used for power generation and heating; although the electricity and heat generated by this part of the energy consumed may not be used in the province, the carbon dioxide generated is left in the province. Therefore, the energy consumption data in this paper are the sum of three types of consumption: terminal energy consumption, power generation energy consumption and heating energy consumption. CO₂ emissions from cement production only count CO₂ emissions from chemical reactions and do not include CO₂ from the burning of fossil fuels during production. In addition, according to the IPCC and the China National Climate Change Coordination Group Office and the Energy Research Institute of the National Development and Reform Commission, in addition to fossil energy combustion, cement, lime, calcium carbide and other industrial production processes, due to chemical and physical reactions, will also produce carbon dioxide, but the carbon dioxide emitted in the industrial production process is mainly concentrated in cement production. Up to now, the provincial data of lime and calcium carbide have been difficult to obtain and the emission proportion has been relatively small, so they are not included in the calculation in this paper. Data on fossil fuel consumption were sourced from the “China Energy Statistical Yearbook” covering the years 2011 to 2021, while cement production figures were obtained from the CSMAR database.

(2) Key explanatory variable

Digital Economy Growth Level (Dige): Digital economy evaluation should not only consider the basic factors such as the Internet but also introduce the index system of digital transactions to reflect the new economic factors. Therefore, we draw on the practices of Zhao (2020) [55] and Yi (2022) [56] and measure the growth level of the digital economy from three aspects, digital industrialization, digital finance and digital infrastructure, through the index construction method.

In particular, digital industrialization forms the fundamental component of the digital economy. It represents a burgeoning sector that stems from data elements and is characterized by its high level of technological sophistication and innovation. Key industries encompass the Internet, software, information services and telecommunications. This study evaluates digital industrialization through an input–output perspective, considering data availability as measured by the number of employees engaged in Internet-related activities and output. The specific indicators include the following: the proportion of employment in urban units involved in information transmission, software and information technology services; and the percentage of total telecommunications business relative to the permanent resident population at year-end. On the other hand, digital finance plays an exceedingly crucial role within the digital economy by not only providing essential financial support and resource allocation but also fostering high-quality economic development through technological innovation and robust data analysis capabilities. Referring to Guo’s (2020) [57] measurement of the inclusive development of digital finance, we use “Peking University Digital Inclusive Financial Index” as a proxy indicator, which covers the coverage breadth, depth of use and degree of digitalization of digital-inclusive finance, and can represent the development level of digital finance in a more comprehensive way. Digital infrastructure is the basic guarantee and important driving force for the development of the digital economy, assessed by metrics such as Internet and mobile phone penetration rates. Ultimately, the entropy weight TOPSIS approach is employed to normalize the aforementioned indices, condense the data into a singular construct and derive a comprehensive index reflecting the growth of China’s provincial digital economies. A detailed representation of the indices and their respective measurement methodologies is presented in

Table 1. Comprehensive growth level of Chinese urban digital economy.

Primary Index	Secondary Index	Concrete Measure Index (Attribute)	Measurement Method
Digital Economy Growth Level	digital industrialization	Number of Internet-related employees (+)	Percentage of employees in computer services and software
		Internet-related output (+)	Total telecommunications services per capita
	digital finance	Inclusive growth of digital finance (+)	The Peking University Digital Financial Inclusion Index of China (PKU-DFIIC)
	digital infrastructure	Number of mobile Internet users (+)	Number of mobile phone users per 100 people
		Internet penetration (+)	Density of Internet access terminals

.

(3) Intermediary variable

Green Dual Innovation (gti): Based on the above definitions of disruptive green technology innovation and progressive green technology innovation, combined with the studies of Atuahene-Gima (2005) [58] and Qian (2022) [59], we selected the number of green invention patent applications in each province as the proxy variable of disruptive green technology innovation (gti-o). The number of green utility model patent applications was selected as the proxy variable of progressive green technology innovation (gti-p). The application data of invention patent and utility model patent are from the Chinese Research Data Service Platform (CNRDS), and the number of green patent applications can be screened by identifying whether the patent IPC classification number is the “Green Patent IPC Classification Number List”. Additionally, an invention patent represents a novel technical concept encompassing products, methods, or enhancements that can enhance or create processes, methods, materials, components, products, or their combinations. The protection period for such patents is 20 years. In contrast, the utility model patent exhibits a lower creativity and technical level compared to the invention patent. It pertains to the improvement or modification of specific product structures in terms of shape, structure, or combination by proposing practical new technical solutions with a protection period of 10 years. Therefore, the application number of green invention patents and green utility model patents as proxy variables to distinguish the difference between the two green innovation behaviors has good availability and high fit. Figure 2 shows the measure of disruptive green technology innovation versus progressive green technology innovation.

(4) Moderator variable

Social attention (att): Social attention to the environment can consider stakeholders’ concerns, including governmental and public perspectives. Governmental focus on environmental issues is predominantly evident through the quantity of local environmental statutes and regulatory frameworks dedicated to ecological conservation. Based on the data in the Laws & Regulations Database—Chinalawinfo, we calculate this number and take its natural logarithm to form a proxy indicator of government attention (att-g). With Internet advancing, search engines can often form relevant search indices through “click volume” and “search volume” to record the public’s search for specific events, thereby better reflecting public attention. Compared with other environmental topics, the severity of haze weather can be directly felt by the public, thus affecting the public’s environmental perception and triggering search behavior. Baidu, as the largest Chinese search engine, has a wide coverage and easy access to data. By analyzing the search frequency and geographical distribution, we can have an in-depth understanding of relevant data in different regions of China. Therefore, referring to the studies of Guo (2020) [52] and Wu (2022) [60], we adopt the Baidu haze search index as the proxy indicator of public concern (att-s), which is the weighted sum of the PC search index and mobile search index.

(5) Control variables

To specifically gauge the impact of the electronic economic situation on carbon emissions and reduce the predisposition that may arise from omitted variables, we have actually integrated economic development signs known to influence carbon emissions at the provincial level, while also representing year- and province-specific effects. The in-depth variable setups are laid out listed below:

Marketization Extent (mark): Increased marketization can promote competition in energy markets, thereby reducing energy production and energy consumption as a cost. For example, after the introduction of market competition mechanisms in the energy sector, enterprises will actively seek ways to reduce costs, including improving energy efficiency and adopting cleaner energy technologies, in order to reduce carbon emissions and meet environmental protection requirements. Drawing on the jobs of Shi (2017) [61] and Zhao (2019) [62], the total marketization index is utilized to evaluate the level of marketization, with data acquired from the pertinent marketization index records.

Foreign Investment Inflow (fdi): On the one hand, enterprises from developed countries are more inclined to transfer polluting or sunset industries to developing countries with relatively lower environmental regulations and standards through foreign direct investment in order to reduce the high environmental compensation costs in their home country. This exacerbates domestic environmental pollution in the host country. On the other hand, foreign-invested enterprises in their home countries are subject to stringent environmental regulations, leading to the development of a mature set of environmental management principles that positively influence environmental protection in the host country through technology spillover and knowledge diffusion. Therefore, inflows of foreign direct investment can bring advanced production technologies, clean technologies and sophisticated management expertise that enhance domestic production efficiency in the host country and consequently reduce its domestic environmental pollution. This is evaluated using the natural logarithm of the real Foreign Direct Financial investment (FDI) utilized.

Industrialization Level (indus): During the process of industrial production, energy plays a crucial role. The advancement of industrialization leads to increased resource consumption in both production and daily life, resulting in a direct rise in carbon emissions. Over 70% of China’s carbon dioxide emissions stem from industrial production or generative sources. Introducing digital and green technologies into the industrial sector is essential for reducing carbon emissions in China. This is evaluated by the percentage of industrial value added in relationship to the gross domestic product (GDP).

Financial Sector Development (fin): On the one hand, the level of financial development can promote the adjustment of industrial structure and realize carbon emission reduction through the full flow of financial resources and the optimization of resource allocation within the region. On the other hand, carbon dioxide emission intensity can be indirectly affected by promoting economic growth and innovation ability. This is indicated by the ratio of total deposits and loans of financial institutions to the GDP.

Economic Development Level (el): The relationship between economic development and carbon emissions is complex. On one hand, economic growth drives heavy industry development and improves living standards, leading to increased carbon emissions. On the other hand, economic progress fosters high-tech industries and clean energy development, which helps mitigate carbon emissions. This is stood for by the natural logarithm of GDP per capita.

In conclusion, the variables chosen for the analysis and their respective measurement methods exist in

Table 2. Selection and measurement of variables.

Variable Type	Variable Name	Measurement Method
Dependent variable	Carbon emissions (ce)	Calculate the natural logarithm of carbon dioxide emissions
Key explanatory variable	Digital economy growth level (Dige)	Digital economy comprehensive growth index
Intermediary variable	Disruptive green technology innovation (gti-o)	Number of green invention patents granted (tens of thousands)
	Progressive green technology innovation (gti-p)	Number of green utility model patents granted (tens of thousands)
Moderator variable	Government attention (att-g)	The natural logarithm of number of local environmental regulations
	Public attention (att-s)	The natural logarithm of Baidu haze search index
Control variables	Degree of marketization (mark)	Marketization index
	Level of foreign investment (fdi)	The natural logarithm of FDI actually utilized
	Level of industrialization (indus)	Ratio of industrial value added to GDP
	Financial growth level (fin)	The ratio of total deposits and loans of financial institutions to the GDP
	Level of economic growth (el)	Natural logarithm of gross domestic product per capita
	Year	The value of a virtual variable is 1 in a certain year, otherwise 0
	Province	The value of a virtual variable that belongs to a certain province is 1, otherwise it is 0

.

3.2.2. Data Source and Description

This research leverages economic growth figures from the period spanning 2011 to 2021, encompassing data across China’s 30 provinces (excluding Hong Kong, Macao, Taiwan and Tibet Autonomous Region). Since 2011, China has implemented a pilot carbon emission trading policy, a tradable quota system, which fully utilizes digital resources to form new financial innovative products and financial activities, serving as an important market mechanism for energy conservation and emission reduction. Therefore, our sample period started from 2011, resulting in 330 provincial panel data. Unless indicated otherwise, the variable datasets are derived from the China Statistical Yearbook and the respective provincial statistical yearbooks. Missing data for individual years in some provinces were supplemented using a linear interpolation method according to the variable characteristics.

As

Table 3. Descriptive statistics of variables.

Variable	N	Mean	p50	SD	Min	Max
ce	330	10.335	10.370	0.778	8.353	12.217
Dige	330	0.217	0.172	0.154	0.007	0.960
gti-o	330	0.112	0.048	0.166	0	1.001
gti-p	330	0.400	0.186	0.614	0	4.579
att-g	330	4.726	4.732	0.777	2.485	6.652
att-s	330	0.195	0.153	0.171	0.001	1.118
fdi	330	0.019	0.017	0.015	0.000	0.080
mark	330	6.715	6.705	2.369	0.010	12.480
fin	330	3.276	3.051	1.150	1.518	8.131
indus	330	0.351	0.363	0.088	0.097	0.530
el	330	10.866	10.833	0.459	8.531	12.122

shows, the mean value of the explained variable carbon emission (ce) is 10.335, slightly lower than the median of 10.370. The maximum value is 12.217 and the minimum value 8.353; the former one is about 1.46 times of the latter one, with a standard deviation of 0.778. The data highlight the variability in carbon emissions across different provinces in China, even as they collectively exhibit elevated levels. The key explanatory variable, representing the growth level of the digital economy (Dige), has an average value of 0.217, exceeding the median value of 0.048. It reaches a peak of 0.960 and a trough of 0.007, indicating that the maximum exceeds the minimum by a factor of approximately 137.14. This disparity highlights the unbalanced growth of China’s digital economy, likely due to differences in digital infrastructure, talent pool and fund allocation among provinces. For green dual innovation, the mean value of disruptive green technology innovation (gti-o) is 0.112, with a maximum value of 1.001, which is much lower than that of progressive green technology innovation (gti-p), indicating that China’s green technology innovation is mainly progressive. The mean values of government attention (att-g) and public attention (att-s) are close to their respective medians, suggesting consistent social concern for environmental protection. The standard deviation for government concern is 0.777, while for public concern it is 0.171, indicating that government concern fluctuates more across different periods and regions compared to public concern. Among the control variables, except for marketization degree (mark) and financial growth level (fin), the standard deviation of other variables is lower than 1, showing a small range of change. At the same time, a VIF test was adopted in this paper to avoid the high autocorrelation of explanatory variables. VIF values in the test results were all less than 3 and the mean value was 2.18, so there was no serious multiple collinearity in the selected variables.

3.3. Regression Analysis of Benchmark Model

We employ a two-directional fixed-effects model for our regression analysis, with the outcomes delineated in

Table 4. Baseline regression of the impact of the digital economy on carbon emissions.

Variable	(1)	(2)	(3)
	Reference Regression	Lag Phase 1	Lag Phase 2
Dige	−0.904 ***
	(0.249)
L.Dige		−0.863 ***
		(0.231)
L2.Dige			−1.094 ***
			(0.245)
fdi	0.804	0.408	0.254
	(0.756)	(0.709)	(0.646)
fin	−0.014	−0.021	−0.012
	(0.023)	(0.021)	(0.017)
indus	0.965 ***	1.094 ***	1.076 ***
	(0.218)	(0.198)	(0.170)
el	0.014	0.020	0.025
	(0.041)	(0.035)	(0.029)
mark	−0.013	−0.018	−0.025 *
	(0.014)	(0.014)	(0.013)
_cons	9.845 ***	9.846 ***	9.885 ***
	(0.456)	(0.402)	(0.343)
FE-YEAR	YES	YES	YES
FE-PROVINCE	YES	YES	YES
N	330	300	270
r2	0.224	0.241	0.310

*** and * are significant at the level of 1% and 10% respectively.

. The columns (1), (2) and (3) sequentially illustrate the outcomes for the base model and the digital economy growth variable with one and two periods of lag, respectively. In column (1), the regression coefficient for the digital economy growth is −0.904, which is statistically significant at the 1% level, signifying that the expansion of the digital economy is associated with a reduction in carbon emissions. The advancement of the digital economy drives technological progress. As nations address climate change, there is a growing emphasis on integrating technology research and development with carbon reduction methods, leading to advancements in green technology innovation. This not only enhances energy efficiency but also reduces energy consumption and effectively lowers carbon emissions. Additionally, the fusion of the digital economy with daily life has heightened public awareness of low-carbon environmental protection, encouraging the adoption of eco-friendly products and facilitating a shift towards low-carbon, clean and green energy consumption structures, ultimately reducing regional carbon emissions.

Upon incorporating the hysteretic effect in columns (2) and (3), the regression coefficient for the digital economy growth variable peaks with a two-period lag at −1.094, also attaining significance at the 1% level. This implies that the influence of the digital economy on lowering carbon emissions is not immediate but occurs with a delay. Consequently, Hypothesis 1 is substantiated: digital economy growth can curb carbon emissions. The following mechanism analysis will explore the green dual innovation mechanism in depth.

3.4. Robustness Test

To affirm the dependability of our findings, this study undertakes a series of effectiveness, looking at the effect of the digital economic climate on carbon discharges, using four unique methods:

(1)

Replace the key explanatory variable: Complying with the technique of Guo (2020) [57], the Digital Financial Addition Index (DFI) put together by the Digital Money Research Center of Peking College was used to change the digital economy growth level, the core explanatory variable, and the regression was carried out again.

(2)

Replace the dependent variable: Referring to the practice of Dong (2022) [53], the ratio of total carbon emission to population was used as the proxy variable of per capita carbon emission to replace the carbon emission of the explained variable.

(3)

Omitting special years: The COVID-19 pandemic considerably influenced provincial development in 2020–2021. We recalculated the model leaving out data from these years to mitigate the possible skewing impacts of these phenomenal periods on our research outcomes.

(4)

Exclusion of centrally administered municipalities: Given the unique economic profiles of municipalities directly governed by the Central Government, the dynamics of carbon emission reduction attributable to the digital economy in these areas could be markedly different from other provinces. To reduce the potential variability that these cities might bring to the findings, the data from Beijing, Shanghai, Tianjin and Chongqing were omitted, and the analysis was reconducted.

Table 5. Robustness test of carbon emissions affected by digital economy.

Variable	(1)	(2)	(3)	(4)
	Replace the Key Explanatory Variable	Replace the Dependent Variable	Excluding Special Years	Excluding Municipalities Directly under the Central Government
DFI	−0.004 ***
	(0.001)
Dige		−3.431 **	−0.794 ***	−0.790 **
		(1.467)	(0.289)	(0.316)
fdi	0.262	−9.030 **	0.927	−0.155
	(0.703)	(4.450)	(0.696)	(0.943)
fin	−0.015	−0.070	−0.037	−0.029
	(0.021)	(0.134)	(0.031)	(0.026)
indus	0.857 ***	−3.039 **	0.693 ***	0.861 ***
	(0.203)	(1.285)	(0.206)	(0.233)
el	−0.010	−0.288	−0.291 ***	−0.059
	(0.039)	(0.239)	(0.100)	(0.041)
mark	0.015	−0.061	−0.003	−0.008
	(0.013)	(0.084)	(0.015)	(0.016)
_cons	11.971 ***	10.303 ***	13.963 ***	11.364 ***
	(0.502)	(0.508)	(1.172)	(0.513)
FE-YEAR	YES	YES	YES	YES
FE-PROVINCE	YES	YES	YES	YES
N	330	330	270	286
r2	0.327	0.224	0.233	0.334

*** and ** are significant at the level of 1% and 5% respectively.

presents the outcomes of the robustness tests employing the aforementioned methods. Across each method, the coefficient for the growth of the digital economy remains significantly negative, corroborating the initial research outcomes and thus substantiating the robustness and reliability of the conclusions drawn.

3.5. Endogenous Analysis

There may be endogeneity problems in the direct transmission mechanism between digital economy growth and carbon emissions. In this study, we use the 2sls instrumental variable method and SYS-GMM model to address these endogeneity problems:

(1) 2sls instrumental variable method

Drawing on prior scholarly work, the 1984 telephone count serves as an instrumental variable for the digital economy. Given that this instrumental variable is derived from cross-sectional data and is not immediately panel-data compatible, the technique recommended by Nunn (2014) [63] is implemented. This involves multiplying the previous year’s provincial Internet user count by the per capita telephone numbers from 1984, which helps to mitigate the issue of reverse causality.

Table 6. Endogenous analysis of the impact of digital economy growth on carbon emissions.

Variable	2sls Instrumental Variable Method		SYS-GMM Model
	Phase 1	Phase 2
	Dige	ce	ce
IV	0.049 ***
	(0.010)
Dige		−1.858 **	−0.220 *
		(0.805)	(0.123)
Controls	YES	YES	YES
FE-YEAR	YES	YES	YES
FE-PROVINCE	YES	YES	YES
N	330	330	300
Underidentification test	LM statistic = 25.748; p = 0.0000
Weak identification test	Cragg–Donald Wald F statistic = 26.664
	10% maximal IV size = 16.38
AR (1)			0.001
AR (2)			0.665
Hansen			0.764

***, ** and * are significant at the level of 1%, 5% and 10% respectively.

illustrates the endogeneity test outcomes. In the initial phase, the instrumental variable’s regression coefficient is 0.049, achieving significance at the 1% level; in the subsequent phase, it is −1.858, significant at the 5% level. The effect’s significance and orientation are in line with the baseline regression, affirming the validity of the paper’s conclusions.

In addition, for the test of “insufficient identification of instrumental variables”, the P values of the LM statistics of Kleibergen–Paap rk are all 0.0000, which significantly rejects the null hypothesis. In the weak instrumental variable recognition test, the Wald F statistic of Kleibergen–Paap rk is greater than the critical value of the Stock–Yogo weak recognition test at the 10% level. This shows that the instrumental variables selected in this paper do not have problems of insufficient identification and weak instrumental variables, and the instrumental variables show good effectiveness characteristics. It is reasonable to choose the number of Internet users in each province multiplied by the number of telephones per capita in 1984 as the instrumental variable of the comprehensive development index of the digital economy.

(2) SYS-GMM model

The SYS-GMM approach is adept at rectifying issues of endogeneity associated with dynamic panel-data models, such as simultaneous equation bias and omitted variable bias. The model, as depicted in Equation (3), accounts for these complexities:

$${ce}_{it}={\psi }_0+{\psi }_1{ce}_{it}{-}_1+{\psi }_2{Dige}_{it}+{\gamma control}_{it}+{\mu }_{it}+{\lambda }_{it}+{\epsilon }_{it}$$

(3)

In this formulation, _{ceit − 1} represents the reliant variable with a one-period lag, with all the various other variables retaining their definitions from Formula (1). The searches provided in Table 6 suggest that the regression coefficient related to the growth of the electronic economic situation is −0.220, accomplishing statistical value at the 10% degree. The impact’s importance and orientation line up with the baseline regression, consequently corroborating the research study’s objective. The AR(1) test’s p value is below 0.1, while the AR(2) examination’s p value surpasses 0.1, pointing to the absence of autocorrelation in the error term. Additionally, the Hansen examination’s p value goes beyond 0.1, affirming the legitimacy of the selected crucial variables.

The effectiveness examination outcomes emphasize that the fundamental final thought—improved development of the digital economy situation assists in a decrease in carbon exhausts—holds true after regulating for endogeneity. This highlights the substantial relevance of increasing digital economy growth for the present phase of carbon emission reduction initiatives.

4. Mechanisms of the Digital Economy’s Role in Carbon Emissions

4.1. Model Construction for Mechanism Testing

To examine the potential mechanisms by which the digital economy affects carbon emissions, we initially examine the role of disruptive and progressive green technology innovations as mediators. The methodology is as follows: expanding upon Equation (1) from the bidirectional fixed effects model that relates the digital economy growth index (Dige) to carbon emissions (ce), an extended model is formulated:

$${gti}_{it}={\beta }_0+{\beta }_1{Dige}_{it}+{\gamma control}_{it}+{\mu }_{it}+{\lambda }_{it}+{\epsilon }_{it}$$

(4)

$${ce}_{it}={\pi }_0+{\pi }_1{Dige}_{it}+{\pi }_2{gti}_{it}+{\gamma control}_{it}+{\mu }_{it}+{\lambda }_{it}+{\epsilon }_{it}$$

(5)

Here, gti_it denotes the mediating variable, encompassing both disruptive green technology innovation (gti-o) and progressive green technology innovation (gti-p), with all other variables consistent with their definitions in Equation (1). Mediation is assessed using a sequential regression approach, determining the presence of a mediating effect by evaluating the statistical significance of coefficients including β₁, π₁ and π₂.

Furthermore, to evaluate the moderating impact of societal concern on the interplay between the digital economy and carbon emissions, this study incorporates measures of societal concern and its interaction terms with digital economy growth into the model, as per Equation (1). The refined model is presented below:

$${ce}_{it}={\kappa }_0+{\kappa }_1{Dige}_{it}+{\kappa }_2{att}_{it}+{\kappa control}_{it}+{\mu }_{it}+{\lambda }_{it}+{\epsilon }_{it}$$

(6)

$$\begin{array}{c}{ce}_{it}={\sigma }_0+{\sigma }_1{Dige}_{it}+{\sigma }_2{att}_{it}+{\sigma }_3{Dige}_{it}\times {att}_{it}\\ +{\sigma control}_{it}+{\mu }_{it}+{\lambda }_{it}+{\epsilon }_{it}\end{array}$$

(7)

In Formula (6), att_it is the regulating variable, including government attention (att-g) and public attention (att-s); in Formula (7), Dige_it × att_it represents the government attention’s interaction term with digital economy growth, and the public attention’s interaction term with digital economy growth; the meanings of the other variables are the same as above.

4.2. Analysis of Mediating Effects Based on Green Dual Innovation

This study explores how the expansion of the digital economy influences carbon emissions through the mediating role of green dual innovation. The regression results are outlined in

Table 7. Regression results of the mediating effect of green dual innovation.

Variable	Reference Regression	Disruptive Green Technology Innovation		Progressive Green Technology Innovation
	(1)	(2)	(3)	(4)	(5)
	ce	gti-o	ce	gti-p	ce
Dige	−0.904 ***	0.863 ***	−0.496 **	1.850 **	−0.780 ***
	(0.249)	(0.165)	(0.248)	(0.810)	(0.246)
gti-o			−0.474 ***
			(0.085)
gti-p					−0.067 ***
					(0.018)
fdi	0.804	0.730	1.149	2.346	0.962
	(0.756)	(0.500)	(0.722)	(2.455)	(0.740)
fin	−0.014	0.014	−0.007	0.136 *	−0.005
	(0.023)	(0.015)	(0.022)	(0.074)	(0.022)
indus	0.965 ***	0.266 *	1.091 ***	0.406	0.993 ***
	(0.218)	(0.144)	(0.209)	(0.709)	(0.213)
el	0.014	0.072 ***	0.049	0.389 ***	0.041
	(0.041)	(0.027)	(0.039)	(0.132)	(0.040)
mark	−0.013	0.035 ***	0.003	0.148 ***	−0.003
	(0.014)	(0.009)	(0.014)	(0.046)	(0.014)
_cons	9.845 ***	−1.140 ***	9.305 ***	−5.491 ***	9.474 ***
	(0.456)	(0.301)	(0.444)	(1.480)	(0.456)
FE-YEAR	YES	YES	YES	YES	YES
FE-PROVINCE	YES	YES	YES	YES	YES
N	330	330	330	330	330
r2	0.224	0.470	0.300	0.432	0.261

***, ** and * are significant at the level of 1%, 5% and 10% respectively.

. Columns (2) and (4) demonstrate that the growth of the digital economy significantly stimulates both disruptive and progressive green technology innovation. Specifically, the coefficient for the impact of the digital economy on disruptive green technology innovation is 0.863 at a significance level of 1%, while its effect on progressive green technology innovation stands at a coefficient of 1.850, significant at a level of 5%. Furthermore, the regression coefficients for the growth of the digital economy and green dual innovation in columns (3) and (5) are significantly negative, reaching the 1% level of significance. Notably, the inclusion of the mediating variable, green dual innovation, results in a decrease in the coefficients for the digital economy’s growth from −0.904 to −0.496 and −0.780, respectively. This suggests that both types of green technology innovation act to lessen the adverse impact of the digital economy’s growth on carbon emissions, thereby supporting Hypotheses 2a and 2b. The calculated mediating effect of disruptive green technology innovation is found to be 45.25%, and for progressive green technology innovation, it is 13.71%.

The difference in the role of the digital economy on green dual innovation may come from the following: progressive innovation is characterized by the enhancement of existing green technologies, whereas disruptive innovation typically demands substantial resource investment and is particularly responsive to shifts in the digital economy. The expansion of the digital economy aids in the distribution and optimization of green innovation resources, creating synergies that assist local governments in surmounting green innovation challenges, thus enabling a “qualitative leap” in green innovation and creating disruptive green technology innovations. At the same time, the empirical results show that there are differences in the mediating effects of green dual innovation between the digital economy and carbon emissions. The more pronounced mediating effect of disruptive innovation is likely due to the greater resource demands associated with green dual innovation. As societal emphasis on green development and the imperative to reduce carbon emissions intensify, regions are increasingly adopting stringent environmental regulations, particularly concerning carbon emissions. In light of China’s evolving green technology infrastructure, provinces are more likely to utilize the digital economy to foster new green technologies to achieve carbon emission reduction goals. With advancements in the green technology system and the pursuit of “dual carbon” objectives, regions can align their growth strategies with their resource capabilities, further enhancing green technology in both “quality” and “scale”.

4.3. Panel Threshold Analysis Based on Green Dual Innovation

Confirmation has been obtained that the digital economy indeed contributes to the reduction in carbon emissions through the mechanism of green dual innovation. Subsequently, an inquiry arises whether this influence presents a threshold effect—namely, if there exist critical values for disruptive and progressive green technology innovation that modify the digital economy’s effect on carbon emissions under different circumstances. To explore this, the study employs a panel threshold model to determine the equilibrium point where the digital economy’s impact via green dual innovation is observed. The ensuing panel threshold model is thus formulated:

$$\begin{array}{c}{ce}_{it}={\rho }_0+{\rho }_1{Dige}_{it}I ({gti}_{it}\le r_1)+{\rho }_2{Dige}_{it}I(r_1<{gti}_{it}\le r_2)\\ +\dots +{\rho }_n{Dige}_{it}I({gti}_{it}>r_n)+{\rho control}_{it}+{\mu }_{it}+{\lambda }_{it}+{\epsilon }_{it}\end{array}$$

(8)

Here, Dige is the core explanatory variable, gti is the limit variable, r is the threshold value and I is the indication function that takes the worth of 1 or 0. It is 1 if the problem is satisfied, and 0 otherwise.

After 300 repeated samples, the examination outcomes are shown in

Table 8. Test results of threshold effect of disruptive green technology innovation.

Variable	Threshold Number	Fstat	Prob	Confidence Interval			Threshold Value
				10%	5%	1%
gti-o	Single	28.49	0.0467	24.6388	29.6830	38.5710	0.1520
	Double	17.86	0.1533	20.1671	27.3137	40.4712
	Triple	21.69	0.3300	35.1271	39.5965	54.1837

and

Table 9. Test results of threshold effect of progressive green technology innovation.

Variable	Threshold Number	Fstat	Prob	Confidence Interval			Threshold Value
				10%	5%	1%
gti-p	Single	24.27	0.0700	22.0394	26.7789	34.0766	0.3357
	Double	14.37	0.1500	17.4612	20.8467	26.0678
	Triple	6.14	0.7433	19.4920	23.0390	30.4939

. According to the F-value and p-value, disruptive green technology innovation passes the solitary limit test at a significance degree of 5%, with a threshold worth of 0.1520. Progressive green technology innovation passes the single-limit examination at the significance degree of 10%, with a limit value of 0.3357. This shows that building a solitary threshold design is reasonable and that limit estimation can be conducted.

Table 10. Threshold estimation results of green binary innovation.

Variable	(1)	(2)
	Disruptive Green Technology Innovation	Progressive Green Technology Innovation
0_c	−0.643 ***	−0.709 ***
	(0.246)	(0.244)
1_c	−1.254 ***	−1.007 ***
	(0.228)	(0.242)
fdi	0.877	0.842
	(0.729)	(0.730)
fin	−0.013	−0.014
	(0.022)	(0.022)
indus	0.982 ***	0.957 ***
	(0.210)	(0.211)
el	0.047	0.040
	(0.040)	(0.040)
mark	−0.005	0.011
	(0.014)	(0.015)
_cons	9.429 ***	9.433 ***
	(0.448)	(0.449)
FE-YEAR	YES	YES
FE-PROVINCE	YES	YES
N	330	330
r2	0.282	0.279

*** are significant at the level of 1%.

shows that, when disruptive green technology innovation is below the threshold value 0.1520, each 1% increase in digital economy growth level can contribute to a 0.643% reduction in carbon emission. When the disruptive green technology innovation exceeds the threshold value of 0.1520, each 1% increase in the digital economy growth level can achieve a 1.254% reduction in carbon emission. For progressive green technology innovation, when it is below the threshold value of 0.3357, each 1% increase in the digital economy growth level helps cut down carbon emissions by 0.709%. When progressive green technology innovation exceeds the threshold value of 0.3357, each 1% increase in the digital economy growth level makes a 1.007% reduction in carbon emissions.

In conclusion, the threshold for disruptive green technology innovation is more readily attainable relative to that of progressive green technology innovation. Upon surpassing this threshold, the impact of the digital economy on carbon emission reduction is markedly amplified, highlighting the pivotal role and profound implications of disruptive green technology innovation. Despite the higher resource allocation and increased risks associated with disruptive innovation, it has the potential to revolutionize existing technologies, leading to significant advancements in energy efficiency and emission reduction, and to catalyze innovation across related technological domains.

This brings greater marginal benefits and further strengthens the digital economy’s capacity to cut down carbon emissions. In contrast, progressive green technology innovation involves the improvement and growth of existing technologies, with relatively mature outcomes. This results in slightly higher initial benefits in the first stage compared to disruptive green technology innovation. However, due to its low actual marginal benefit, more accumulation is needed to cross the threshold. Consequently, its effect on the relationship between the two is lower than that of disruptive green technology innovation.

4.4. Analysis of Moderating Effects Based on Social Concerns

To delve deeper into the regulatory dynamics through which the digital economy’s expansion influences carbon emissions, this study incorporates social focus from both governmental and public sectors as moderating variables. The outcomes of the regression analysis are presented in

Table 11. Regression results of the moderating effect of social concern.

Variable	Reference Regression	Government Attention		Public Attention
	(1)	(2)	(3)	(4)	(5)
Dige	−0.904 ***	−1.052 ***	−1.168 ***	−0.906 ***	−0.860 ***
	(0.249)	(0.252)	(0.245)	(0.249)	(0.247)
att-g		0.032 *	0.030 *
		(0.017)	(0.017)
Dige*att-g			−0.143 **
			(0.061)
att-s				0.024	0.004
				(0.029)	(0.029)
Dige*att-s					−0.146 ***
					(0.053)
fdi	0.804	0.780	0.927	0.789	0.740
	(0.756)	(0.755)	(0.760)	(0.756)	(0.748)
fin	−0.014	−0.021	−0.020	−0.014	−0.007
	(0.023)	(0.023)	(0.023)	(0.023)	(0.023)
indus	0.965 ***	0.965 ***	0.974 ***	0.982 ***	1.008 ***
	(0.218)	(0.218)	(0.218)	(0.219)	(0.217)
el	0.014	0.015	0.028	0.013	0.044
	(0.041)	(0.041)	(0.041)	(0.041)	(0.042)
mark	−0.013	−0.013	−0.014	−0.012	−0.011
	(0.014)	(0.014)	(0.014)	(0.014)	(0.014)
_cons	9.845 ***	9.769 ***	9.638 ***	9.753 ***	9.496 ***
	(0.456)	(0.459)	(0.466)	(0.468)	(0.472)
FE-YEAR	YES	YES	YES	YES	YES
FE-PROVINCE	YES	YES	YES	YES	YES
N	330	330	330	330	330
r2	0.224	0.228	0.235	0.226	0.246

***, ** and * are significant at the level of 1%, 5% and 10% respectively.

. Specifically, column (3) reveals that the interaction term coefficient for governmental focus and the growth of the digital economy is −0.143, which is statistically significant at the 5% level. Similarly, column (5) indicates that the interaction term coefficient for public focus and the growth of the digital economy is −0.146, reaching significance at the 1% level. These coefficients are in accordance with the direction of the regression coefficient attributed to the digital economy’s growth, suggesting a synergistic relationship between social attention and the economy’s digital transformation in reducing carbon emissions. This indicates that the moderating effects of government attention and public attention are present and positively moderated; thus, Hypothesis 3-a and 3-b are valid.

To be more precise, the influence of public scrutiny on the linkage between the digital economy’s expansion and carbon emissions exerts a more pronounced regulatory impact than the focus directed by governmental bodies. This difference can be attributed to the distinct nature of government and public attention. Government attention serves as the fundamental constraint, primarily reflecting the baseline requirements of environmental protection. For marginal effects beyond mandatory requirements, its impact is cut down, and there is often a delay before resources such as funds and human resources matched with policies can form an effective support system. As a result, the impact of the digital economy’s growth on carbon emission reduction is not immediately amplified.

Public attention is a vital complement to the government’s concern, reflecting people’s increasing demand for a better life, with the public being the direct beneficiary of environmental welfare. In addition to voluntarily adopting energy-saving and emission-reduction practices in their everyday routines, the heightened public focus also motivates businesses to more effectively discharge their societal duties, thus strengthening the digital economy growth’s inhibition effect on carbon emissions. Therefore, further enhancing public participation in environmental governance is crucial for addressing complex environmental problems.

5. Further Analysis

Given the disparities in resources, policies and market conditions across Chinese provinces, the extent of economic growth and the profiles of carbon emissions differ. To meticulously examine the varying effects of the digital economy’s expansion on carbon emissions among provinces with distinct characteristics, we categorize the provincial samples from various standpoints and perform an analysis that accounts for this heterogeneity.

5.1. Heterogeneity Analysis Based on Foreign Investment

Initially, the provincial samples are bifurcated into two categories based on the degree of foreign investment—those with low and those with high levels of foreign investment. This classification is based on the proportion of total foreign investment relative to the GDP. The specific method involves using the median of the sample to classify the provincial samples through the dichotomy. On the basis of Formula (1), we then assess the heterogeneity of the impact of the growth of digital technology on carbon emissions. The findings of this assessment are presented in

Table 12. Results of heterogeneity analysis of foreign investment level.

Variable	(1)	(2)
	Low Level of Foreign Investment	High Level of Foreign Investment
Dige	−0.710	−0.974 ***
	(0.543)	(0.184)
fdi	1.895	−0.516
	(1.640)	(0.676)
mark	0.064 *	0.041 ***
	(0.034)	(0.013)
fin	−0.041	0.072 ***
	(0.039)	(0.027)
indus	0.815 **	0.662 ***
	(0.347)	(0.189)
el	−0.046	0.280 ***
	(0.050)	(0.093)
_cons	10.310 ***	6.698 ***
	(0.579)	(1.077)
FE-YEAR	YES	YES
FE-PROVINCE	YES	YES
N	165	165
r2	0.392	0.324

***, ** and * are significant at the level of 1%, 5% and 10% respectively.

.

On the basis of the results, the digital economy growth’s inhibition effect on carbon emission in regions with high foreign investment levels is more significant than in regions with low foreign investment levels. The regions with an elevated level of foreign investment exhibit an influence coefficient of −0.974, which is deemed statistically significant at the 1% level. Conversely, in regions with a lower level of foreign investment, the impact coefficient of the digital economy’s growth on carbon emissions is −0.710 and fails to achieve statistical significance, highlighting the disparities between the two groups.

On one hand, foreign investment has brought about the “pollution paradise” effect. Companies from more advanced nations, in an effort to reduce the environmental costs incurred domestically, may be more inclined to relocate their pollution-intensive industries to developing countries with less stringent environmental regulations through foreign direct investment, exacerbating environmental degradation in the recipient countries. Nonetheless, the digital economy has the potential to foster the greening of these polluting industries and substantially reduce carbon emissions. On other hand, foreign investment can introduce a “pollution halo” effect. Such an investment is usually accompanied by advanced production technology and new environmental protection concepts, positively impacting local environmental protection through technology spillover and knowledge diffusion. Moreover, regions with high foreign investment levels often prioritize green investment, prompting entities responsible for carbon emissions to consciously utilize the advantages of the digital economy, develop green technology and practice environmental protection measures. Therefore, in provinces with a high level of foreign investment, the digital economy growth more significantly cuts down carbon emissions.

5.2. Heterogeneity Analysis Based on Technology Transactions

To further explore the digital economy growth’s impact on carbon emissions in provinces with different technology-trading characteristics, this study classifies provincial sample-based technology-trading activity. Technology-trading activity is measured by the proportion of a region’s technology trade volume relative to its GDP. The specific approach involves taking the median of the sample and classifying the provincial-level samples into two categories, low technology-trading activity and high technology-trading activity, through the binary method. Based on Formula (1), the heterogeneity of technology trading is tested, and the results are shown in

Table 13. Results of heterogeneity analysis of technical transaction activity.

Variable	(1)	(2)
	Low Tech-Trading Activity	High Tech-Trading Activity
Dige	−1.173 **	−0.321 *
	(0.474)	(0.182)
fdi	0.768	−0.175
	(1.551)	(0.596)
mark	0.026	0.036 ***
	(0.022)	(0.013)
fin	0.120 **	0.076 ***
	(0.055)	(0.026)
indus	0.727 **	0.451 ***
	(0.341)	(0.163)
el	−0.093 *	0.251 ***
	(0.051)	(0.093)
_cons	10.784 ***	6.956 ***
	(0.620)	(1.055)
FE-YEAR	YES	YES
FE-PROVINCE	YES	YES
N	165	165
r2	0.529	0.375

***, ** and * are significant at the level of 1%, 5% and 10% respectively.

.

The results indicate that the impact of digital economy development on carbon emissions is more pronounced in regions with low technology-trading activity compared to those with high technology-trading activity. Specifically, the coefficient for digital economy development’s impact on carbon emissions in regions with low technology-trading activity is −1.173, significant at a 5% level, while in regions with high technology-trading activity, the coefficient is −0.321, significant at a 10% level. To further test the difference between the groups, the Chow test p value of 0.0000 confirms the difference between the groups. This may be due to regions with low technology-trading activity having low flow efficiency for various technologies, including energy-saving and emission-reduction technologies, making a spillover effect more impossible, brought by from the technological advantages of model regions and enterprises in green growth. Meanwhile, digital economy growth can promote exchanges between enterprises and regions, addressing information asymmetry to some extent and compensating for the deficiency caused by the low activity of technology transactions. Therefore, in provinces with low technology-trading activity, digital economy growth more significantly cut down carbon emissions.

5.3. Heterogeneity Analysis Based on Geographic Location

Additionally, a more nuanced analysis was performed considering the geographical distribution. Utilizing the categorization framework provided by the National Bureau of Statistics of China, the provincial-level data were segmented into four distinct regions: the east, the center, the west and the northeast (the eastern provinces include Beijing, Fujian, Guangdong, Guangxi, Hainan, Hebei, Jiangsu, Shandong, Shanghai, Tianjin and Zhejiang; the central provinces include Anhui, Henan, Hubei, Hunan, Jiangxi, Inner Mongolia and Shanxi; the western provinces are Gansu, Guizhou, Ningxia, Qinghai, Shaanxi, Sichuan, Xinjiang, Yunnan and Chongqing; the northeastern provinces are Heilongjiang, Liaoning and Jilin.). The outcomes of the regression analysis are detailed in

Table 14. Results of geographical location heterogeneity analysis.

Variable	(1)	(2)	(3)	(4)
	Eastern	Central	Western	Northeast
Dige	−1.135 ***	−0.094	−1.526 **	−1.677
	(0.223)	(1.160)	(0.659)	(1.501)
indus	−0.142	0.766	1.811 ***	0.048
	(0.292)	(0.573)	(0.466)	(0.343)
fin	−0.024	0.010	−0.022	−0.375 ***
	(0.032)	(0.124)	(0.061)	(0.122)
fdi	0.694	1.169	7.816 **	−1.595
	(0.824)	(3.342)	(3.852)	(0.912)
el	0.117	−0.321	−0.079	−1.258 **
	(0.129)	(0.217)	(0.055)	(0.455)
mark	0.023	−0.048	0.031	−0.014
	(0.018)	(0.035)	(0.038)	(0.037)
_cons	9.240 ***	13.884 ***	10.035 ***	24.986 ***
	(1.478)	(2.193)	(0.633)	(4.997)
FE-YEAR	YES	YES	YES	YES
FE-PROVINCE	YES	YES	YES	YES
N	121	77	99	33
r2	0.526	0.620	0.424	0.862

*** and ** are significant at the level of 1% and 5% respectively.

.

The findings indicate that the expansion of the digital economy in the eastern and western regions exerts a notably suppressive influence on carbon emissions. Specifically, in the eastern region, the regression coefficient linking the growth of the digital economy to carbon emissions is −1.135, which is statistically significant at the 1% level. For the western region, this coefficient stands at −1.526, indicating significance at the 5% level. The Chow test’s P value, which is 0.0000, substantiates the variance among the regional groups. Conversely, the impact coefficients for the central and northeast regions, at −0.094 and −1.677, respectively, do not reach statistical significance. Several factors contribute to this outcome: (1) The energy consumption in the eastern and western regions continues to be predominantly coal-based. Digital economy growth is at the stage of increasing marginal returns, and the government strongly guides intelligent and low-carbon practices. This helps enterprises better respond to the implementation of the “dual carbon” policy, improving resource utilization efficiency, reducing carbon emissions and enhancing ecological environment quality. (2) In the central region, the digital infrastructure is more developed and efficiently utilized. Although the digital economy optimizes regional industrial structure and resource allocation, its marginal returns are low, making its inhibition effect on carbon emissions insignificant. (3) In the northeast region of China, the development of digital infrastructure remains relatively nascent. Carbon emissions are less sensitive to digital economy changes, and the cut-down effect of digital technology growth has not sufficiently offset the carbon emissions accumulated during the digital industrialization stage.

6. Conclusions and Policy Recommendations

The digital economy has become a crucial driver for sustainable growth, exerting a complex and diverse impact on carbon emissions through various channels. Combining the sustainable development goals of “Affordable and Clean Energy”, “Industry, Innovation and Infrastructure” and “Climate Action”, this study utilizes a comprehensive dataset covering China’s provinces from 2011 to 2021, employing a range of econometric models such as fixed effects, mediation, thresholds and moderation analyses to analyze the ways in which the digital economy contributes to reducing carbon emissions. The focus is particularly on the role of green dual innovation. In contrast to existing research, this study adopts a dual innovation perspective by categorizing green technology innovation into disruptive and incremental types. It explores how these two types of innovation mediate the relationship between the digital economy and carbon emissions, as well as their differences. Additionally, it examines how social attention influences the impact of the digital economy on carbon emissions by quantifying environmental legislative actions and internet searches for environment-related terms from both government and public perspectives. The findings reveal several key insights: (1) the advancement of the digital economy is significantly associated with reduced carbon emissions; (2) green dual innovation serves as a mediating factor, with disruptive green technologies exhibiting a more pronounced effect than incremental ones; (3) both governmental and public attention positively moderate the relationship between the digital economy and carbon emissions, with public concern demonstrating greater influence; (4) the effectiveness of the digital economy in reducing carbon emissions depends on foreign investment levels, technology trade intensity and geographical context. Notable effects are observed in areas with substantial foreign investments and limited technology trade, as well as in eastern and western regions. At the same time, this paper also has some limitations that could be addressed in future research. Firstly, the sample size only covers 30 provinces in China from 2011 to 2021, so further verification using samples from other countries or regions is needed before generalizing conclusions. Additionally, including small amounts of industrial production-related CO₂ emission data such as lime or calcium carbide processes would improve calculation accuracy. Furthermore, considering multi-party concerns including enterprises’ perspectives or media viewpoints should be explored when discussing social concerns.

Based on these conclusions, the study recommends several strategies: Firstly, integrating the digital economy into traditional production sectors is essential for fostering holistic development while leveraging digital technologies to promote carbon neutrality. Moreover, encouraging green innovation, especially disruptive forms, can capitalize on leading high-tech firms’ pioneering roles. Despite the positive externalities of green technology, such as improved resource-utilization efficiency and ecological environment quality, its research and growth investment are substantial and the income cycle is long. Therefore, the government should improve green technology standards, strengthen fiscal, tax and financial services, enhance intellectual property protection and stimulate innovation vitality by optimizing the policy environment to form synergy between the government and the market. Moreover, it should guide and promote a dual environmental governance model with public participation as the main force, utilize big data to obtain and disclose environmental data in a timely manner, strengthen supervision and implementation and optimize channels for public environmental participation. A variety of interactive platforms that encourage public participation in environmental conservation should also be developed, thereby cultivating a supportive societal framework that facilitates the reduction in carbon emissions. Furthermore, it is imperative for governments to facilitate the exchange of pertinent technologies, capital and skilled professionals through digital collaboration initiatives. This approach not only fosters the sharing of resources but also paves the way for mutually beneficial outcomes. All regions should base their strategies on regional growth differences, implementing heterogeneous governance strategies that combine the digital economy with their unique growth characteristics and advantages. Developed regions should leverage the favorable conditions and sound foundation of the digital economy in central cities, maximizing agglomeration and radiation effects to form new drivers of economic growth. Less-developed regions should identify the power points for local digital economy growth to enhance internal-growth momentum, and ensure the effective utilization of resource endowments and special functions, promoting tailored strategies that align with local conditions.