Digital Economy, Green Finance, and Carbon Emissions: Evidence from China

Abstract

This paper investigates the role of the digital economy in reducing carbon emissions, with a particular focus on the moderating and threshold effects of green finance. An analysis of data from 30 Chinese provinces shows that the digital economy significantly reduces carbon emission intensity by restructuring energy consumption and promoting green technological innovation. Green finance plays a crucial moderating role by alleviating financial barriers to digital transformation and supporting the implementation of emission-reducing technologies. The study reveals a nonlinear relationship, with green finance exhibiting a “strong initial, weak subsequent” threshold effect. At the same time, the digital economy’s impact on carbon reduction strengthens over time as technological development progresses. These findings contribute to understanding how digitalisation and green finance can work synergistically to drive sustainable low-carbon development.

1. Introduction

Since the Industrial Revolution, the utilisation of fossil fuels, including coal and oil, has surged significantly. The acceleration of industrialisation has resulted in significant carbon dioxide emissions, contributing to global warming, a critical climatic issue of worldwide concern that severely impacts diverse geographical locations and levels. Recent reports from the United Nations Environment Programme (UNEP) and the International Energy Agency (IEA) indicate that global carbon dioxide emissions attained a record high in 2023, primarily due to the persistent reliance on fossil fuels and inadequate progress in clean energy technology development. This occurrence has garnered significant global interest. China is considered a leader among middle-income countries, and reducing its carbon emissions is important to promote the sustainable growth of its domestic economy and an important step in tackling global climate change [1]. To realise China’s objective of reaching peak carbon dioxide emissions before 2030 and pursuing carbon neutrality by 2060, as articulated during the 75th session of the United Nations General Assembly, sustained efforts remain essential. Promoting carbon reduction protects the environment, promotes green ecological development, reduces energy consumption, improves resource efficiency, and promotes sustainable economic development [2].

The digital economy, which enhances energy efficiency and fosters technological innovation, is primarily considered as a crucial response to climate change. This novel economic model prioritises data resources as the fundamental production factor and leverages innovation-driven advancements in digital technology to progressively improve societal levels of digitalisation, networking, and intelligence, thereby effectively tackling the challenges presented by climate change [3,4]. Its advantages include effective information distribution, data generation, and sharing, which may enhance organisations’ productivity and operational efficiency, hence decreasing energy consumption and carbon emissions. In recent years, the digital economy has increasingly become a driving force in addressing environmental degradation and achieving carbon emission peaks and neutrality goals. However, its development has also faced challenges such as the massive energy consumption of data centres and insufficient funding for low-carbon technology R&D. Nevertheless, the digital economy promotes the growth of traditional industries by reducing carbon emissions and enhancing efficiency [5].

Presently, the majority of studies emphasise the beneficial effects of the digital economy on carbon reduction, hence offering a research avenue for this article to explore the influence of green finance on the digital economy’s modulation of carbon emissions. Chen et al. discussed how to improve the effect of carbon reduction through economic mechanisms, stimulate technological innovation, and promote sustainable development considering the supply side, consumer demand, and environment [6]. Green finance encompasses policies and instruments that facilitate environmental conservation and sustainable development initiatives via financial resources to direct capital towards low-carbon and ecological sectors, thereby fostering economic transformation and advancing a green economy. As a crucial objective for high-quality economic advancement in China’s digital era [7], green finance effectively disrupts the detrimental cycle of elevated costs associated with low-carbon transformation, challenges in validating emission reduction benefits, and minimal green premiums within the digital economy. This is achieved through targeted credit mechanisms for prioritised investment in clean computing infrastructure, the dynamic quantification of digital emission reduction efficacy via carbon accounting systems, and rigorous oversight of implicit carbon emissions in the supply chain through environmental information disclosure regulations. Consequently, a fundamental inquiry of this study is to investigate whether green finance, as a moderating variable, might amplify the carbon reduction impact of the digital economy.

This paper’s contributions are mostly evident in two areas. Initially, it conducts a comprehensive examination of the moderating influence of green financing on the impact of carbon reduction on the digital economy. The current literature predominantly analyses the effects of the digital economy or green finance on carbon emissions from a singular viewpoint. While prior research has demonstrated that both the digital economy and green finance independently contribute to carbon reduction, few studies have integrated the two into a unified analytical framework to investigate the potential synergistic impact, where “1 + 1 > 2.” This study seeks to examine the influence of the digital economy on carbon reduction, with specific emphasis on the dynamic moderating role and nonlinear threshold effect of green financing in this context. This article aims to address the following problems by developing an interactive analytical framework including the digital economy, green financing, and carbon emission intensity: How does green finance augment the carbon reduction impact of the digital economy via capital allocation, governmental incentives, and technological support? Is there a differential phase in its moderating function?

This paper is structured as follows: Section 2 delineates the theoretical analysis and hypotheses; Section 3 outlines the model design employed in the study; Section 4 details the data sources and presents descriptive statistics; Section 5 examines the empirical results and further explicates the moderating and threshold effects of green finance; and the concluding section summarises the study and provides recommendations for future research.

2. Literature Review and Research Hypotheses

2.1. Literature Review

2.1.1. Digital Economy

In recent years, with the broad application of the internet, big data, artificial intelligence, and cloud computing, the digital economy has become a key force driving global economic development, promoting high-quality economic development, and positively impacting environmental resources. Countries have strengthened their strategic planning for the digital economy as an important driving force for economic development and social progress. Some studies have directly examined the relationship between the digital economy and carbon emissions. Carbon Emission Performance (CEP) has become a key strategy for reducing carbon emissions, and the digital economy has been recognised as an important tool to support CEP [8]. Other studies have found that the emergence of new industries, such as e-commerce, mobile payments, and cloud computing, has led to a steady annual increase in the share of digital economy output, opening up new avenues for carbon emission reductions (CERs) [9]. At the same time, the digital economy can significantly improve efficiency and reduce carbon emissions by reducing energy intensity, lowering energy consumption, and advancing technological progress [10]. Studies by Chang et al. (2023) [11] and Wang & Kang (2024) [12] show that the digital economy reduces the intensity of carbon emissions by optimising the structure of industries and analysing the effect of the level of economic development and the degree of market openness on the moderating effect of this relationship. In addition, digital infrastructure can reduce carbon emissions by reducing resource mismatch and pollution. Simultaneously, digital infrastructure can diminish carbon emissions by mitigating resource misallocation and accounting for the diversity of urban resource endowment, thus facilitating carbon reduction via enhanced energy efficiency. The link between the digital economy and carbon emissions is intricate and constantly changing. The initial growth of the digital economy and technological advancement resulted in heightened carbon emissions; however, it simultaneously mitigated further increases in emissions, demonstrating an inverted “U”-shaped correlation with a spatial spillover effect [13,14].

2.1.2. Green Finance

Green finance, an innovative financial tool integrating finance and environmental protection, has become a crucial method for reducing carbon. It has a systematic capital allocation effect that positively impacts the low-carbon economy and indirectly contributes to it through guiding economic development and encouraging technological innovation [15]. Some scholars have explored the role of green finance by analysing elements such as green credit and bonds, noting that green credit and venture capital can reduce carbon emissions by promoting green technological innovation in enterprises [16]. In China, relevant scholars view green finance as a macroeconomic policy to optimise the commodity structure to better align resources with energy conservation and emission reduction [17].

Additionally, green finance is seen as a green business model that can effectively promote the development of a low-carbon economy. Furthermore, several experts have established an extensive green finance index system that includes various aspects such as green credit, green insurance, and green investment. Their findings indicate that although green financing can diminish carbon dioxide emissions, it may concurrently hinder the expansion of renewable energy utilisation and the influx of investments into green initiatives, thus obstructing the advancement of green finance itself [18].

The financial sector’s involvement in carbon reduction is garnering increased attention in academic circles. Jeucken (2001) [19] examined the relationship between financial institutions and sustainable development, highlighting the crucial role of banks in mitigating carbon emissions. Altaghlibi et al. (2022) [20] asserted that by regulating monetary policy, credit markets, and the overall financial system, banks and financial regulatory authorities can foster the advancement of green finance, direct capital towards low-carbon sectors, and thereby expedite the economy’s transition to low carbon. These investigations have generated increased scholarly interest in the nexus between money and the environment. Additional research has explicitly examined the correlation between green money and carbon emissions. Pretis et al. (2017) [21] and Khan et al. (2019) [22] examined the carbon-reducing impacts of green financing across multiple nations by developing diverse measures of green finance, concluding that it positively contributes to carbon reduction, with policy maturity serving as a moderating factor. Liu (2025) [23] used the geographical Durbin model to find that green funding’s ability to cut carbon emissions has a “neighbouring effect”. Lu and Xia (2025) [24] also used this model and found that this effect might make regional differences in emission reduction more serious. As a policy instrument, green finance may lower the financing costs of green bonds and enhance the success rate of funding for high-carbon projects, consequently augmenting returns on investment for clean technology initiatives and facilitating reductions in carbon emissions [25,26].

Current research has examined the mechanics of low-carbon transition from the dual viewpoints of the digital economy and green finance. The digital economy has been verified to diminish carbon emissions through technology integration and industrial enhancement; nonetheless, it poses the dangers of a “green paradox” and energy rebound effects. Research on green finance has concentrated on the policy impacts of instruments like green credit and bonds, revealing that they can mitigate carbon emissions by lowering technological expenses and enhancing financing frameworks; however, challenges such as regional disparities and institutional limitations persist in low- and middle-income nations. Nonetheless, current research possesses two limitations. First, the dynamic and varied characteristics of the moderating influence are inadequately represented. Most current work uses static models, overlooking the temporal lag impact of green financing regulations and the nonlinear “threshold dependence.” In addition, the method through which green financing enables the digital economy to mitigate carbon emissions is not well defined. The standardised moderating mechanism between green finance and the digital economy remains underdeveloped and the impacts of policy combinations, such as “digital infrastructure investment + green finance subsidies + carbon market linkage”, are unassessed, complicating the formulation of effective policy packages.

2.2. Mechanism of the Impact of Digital Economy on Carbon Emissions

2.2.1. Direct Effect of Carbon Emission Reduction in the Digital Economy

The effect of the digital economy on carbon emissions is primarily seen in three dimensions. The extensive implementation of digital technology in energy production, transportation, and sales has enhanced the efficient utilisation of energy resources, established an effective, clean, and low-carbon digital energy network, facilitated the adoption of clean energy and the advancement of renewable energy, and diminished urban carbon emissions [27]. Secondly, the establishment of digital infrastructure offers a technical foundation for the dynamic monitoring and intelligent analysis of carbon data services. Properly tracking energy consumption data throughout energy production and consumption helps organisations to find energy-saving possibilities and optimise carbon emission governance pathways. Meanwhile, a cross-departmental data-sharing platform built on blockchain technology and cloud computing has accomplished real-time collaboration in carbon emission and carbon sink accounting. This not only improves the targeting of government regulation and the transparency of the carbon trading market, but also advances the transition from broad to specific carbon governance [28]. Third, the promotion of digital platforms and the inclusiveness of digital finance have increased the convenience of network consumption for consumers’ green environmental protection awareness, promoted the upgrading of consumers’ lifestyles, reduced the occupation and consumption of traditional physical resources, and, thus, reduced carbon emissions. Consequently, the following research hypothesis is posited:

Hypothesis 1.

The digital economy has a direct negative impact on carbon emission intensity.

2.2.2. Indirect Effects of Carbon Emission Reduction in the Digital Economy

Economic development remains largely dependent on energy consumption, particularly the extensive use of fossil fuels and non-renewable energy sources. This leads to increased pollutant emissions and keeps energy intensity at a high level [29,30]. Such a development model exerts significant pressure on the environment and poses a severe challenge to sustainable economic development. Reducing the use of fossil fuels and energy intensity is considered as a primary goal for member countries of the Organization for Economic Cooperation and Development (OECD) to achieve sustainable development targets [31]. In the context of the digital economy, the functionality of digital devices and processes can enhance the energy efficiency of industries [32,33]. Digital technologies can shorten the research and development (R&D) cycle for clean energy and improve R&D efficiency. Integrating traditional energy companies with digital energy firms can significantly boost the operational efficiency of oil and gas enterprises [34]. Integrating advanced information technology with digital procedures in the energy sector may establish a new energy ecosystem. This strategy diminishes production expenses, innovates collaborative frameworks, enhances energy generation and use configuration, and expedites the energy transition.

Furthermore, implementing and advocating for digital technology in the renewable energy industry can markedly improve the efficiency of clean energy consumption. Through real-time data processing, big data analytics establishes an intelligent decision-making centre for the global energy internet. This enhances the efficiency of cross-regional clean energy allocation efficiency and diminishes reliance on fossil fuels through data-driven consumer response mechanisms, thereby systematically decreasing carbon reduction costs [35,36] and subsequently reducing carbon emission intensity. Traditional energy corporations may leverage advanced information technology to amalgamate energy and digital methodologies, establishing a novel energy ecosystem. This alters energy generation and utilisation modalities, expedites the energy transition, and enhances carbon emission efficacy [37,38]. Consequently, the following hypothesis is put forth:

Hypothesis 2a.

Changing the structure of energy consumption mediates between the digital economy and the reduction in carbon intensity.

The internal growth of the digital economy has facilitated advancements in green technology [39]. The innovation and transformation of the digital economy can facilitate knowledge sharing and resource integration, optimise the utilisation of existing funds, advance the development of digital infrastructure, mitigate temporal and spatial limitations, and create a flexible and open environment conducive to green innovation [40,41]. Simultaneously, firms can utilise digital technology to establish a collaborative network that facilitates information dissemination, broadens the scope of green innovation, and supplies both internal and external resources for green innovation across the spatial and temporal dimensions [42,43]. Simultaneously, green technologies may markedly diminish greenhouse gas emissions by advancing and implementing clean energy technologies, enhancing energy efficiency, and optimising manufacturing processes [24]. The following assumptions are established:

Hypothesis 2b.

Green technology innovation mediates between the digital economy and the reduction in carbon intensity.

2.2.3. The Moderating Effect of Green Finance on Carbon Emission Reduction in Digital Economy

Green financing is a key factor affecting carbon emissions and is essential to digital economic activity. Meeting carbon reduction objectives requires addressing the numerous ecological and environmental challenges inherited from the industrial economic era. Green finance is an essential funding source for advancing the digital economy while simultaneously strengthening environmental regulations and green assessment systems, thereby greatly enhancing the ability to tackle environmental issues. Furthermore, green financing alleviates the knowledge asymmetry between lenders and borrowers in financial markets, providing essential financial support to pollution-intensive companies [44]. Initiatives in the digital economy supported by green financing have an apparent inclination towards adopting and innovating new technology, therefore significantly aiding in mitigating environmental pollution [45,46]. The development of green financial markets provides economic support to digital economy initiatives, promoting effective market competition and price formation mechanisms and facilitating pathways for corporate low-carbon transitions [47].

Hypothesis 3.

Green finance enhances the carbon-reducing effectiveness of the digital economy by increasing investment in digital technology via its financial support.

2.2.4. Threshold Effect of Carbon Emission Reduction in Digital Economy

The nascent digital economy necessitates establishing infrastructure and technological apparatus from the research and development phase to the real investing phase. The investment phase mainly includes manufacturing operations that produce significant carbon emissions [48]. The initial degree of industrial agglomeration is minimal, hindering the attainment of economies of scale. As the digital economy evolves, industries converge and gain advantages from centralised access to infrastructure data and resource sharing. This can foster knowledge spillovers and economies of scale, augmenting a region’s innovation potential and enhancing carbon emission efficiency [49]. The advancement of green finance, alongside the implementation of environmental performance criteria and incentive systems, motivates firms to embrace more sustainable production processes and technology, thus driving the whole sector towards a low-carbon transition. The reallocation of capital, driven by green finance, facilitates the efficient management of carbon emissions and establishes a framework for sustainable growth in the digital economy era. The current literature has established that the amalgamation of financial “digitisation” and “greening” has a dual-threshold property of “U-shaped” and “inverted U-shaped” effects on carbon reduction [50]. When green development is inadequate, green finance fails to deliver adequate funding, and the inefficient allocation of resources leads to a negligible marginal impact of the digital economy on carbon emissions via green finance. As green finance development advances, governments and markets implement more robust green evaluation systems and risk management mechanisms, thereby mitigating the investment risks associated with green projects, broadening funding sources for these initiatives, and enhancing the digital economy through green finance, ultimately fostering the growth of a low-carbon economy [51]. Consequently, the following hypotheses are posited:

Hypothesis 4a.

The influence of digital economy advancement on carbon emission intensity demonstrates nonlinear traits contingent upon its developmental stage.

Hypothesis 4b.

The influence of digital economy advancement on carbon emission intensity demonstrates nonlinear traits contingent upon the degree of green finance development.

2.3. Assumptions and Variables

In this study, we propose several hypotheses to guide our analysis of the digital economy’s impact on carbon emission intensity and the role of green finance in this relationship. Our first hypothesis posits that the digital economy exerts a direct negative effect on carbon emission intensity. This hypothesis is directly linked to our primary independent variable, the digital economy, and our dependent variable, carbon emission intensity.

Hypotheses 2a and 2b explore potential mechanisms for the digital economy to influence carbon emission intensity. Specifically, Hypothesis 2a suggests that the digital economy facilitates a restructuring of energy consumption patterns, reducing carbon emission intensity. Meanwhile, Hypothesis 2b proposes that the digital economy promotes innovation in green technologies, reducing carbon emission intensity. Both hypotheses are associated with the following mediating variables within our analysis: energy consumption structure and green technological innovation.

Hypothesis 3 focuses on the moderating role of green finance—particularly green financial support—in enhancing the effectiveness of the digital economy’s contribution to reducing carbon emissions. This hypothesis relates to our moderator variable, green finance, suggesting that it can amplify the positive impact of the digital economy on carbon reduction by increasing investments in digital technology.

Finally, Hypotheses 4a and 4b address the potential nonlinear effects stemming from developments within the digital economy relating to carbon emission intensity, specifically contingent upon both its developmental stage and levels of green finance development. These hypotheses pertain to our examination of the threshold effects within the relationship between digital economy advancements and carbon emission intensity changes.

3. Research Design

3.1. Sample and Data

This article utilises panel data from 30 provinces in China, excluding Hong Kong, Macao, Taiwan, and Tibet, covering the period from 2010 to 2022. Carbon dioxide emissions statistics are mostly sourced from the CGER (Center for Global Environmental Research) database. The data utilised in formulating the evaluation indicators for the digital economy and green finance primarily derives from the professional statistical Yearbook of the China Statistical Yearbook, the China Industrial Statistical Yearbook, the China Energy Statistical Yearbook, and the statistical yearbooks of various provinces and municipalities.

3.2. Variable Definitions

3.2.1. Dependent Variable

Dong et al. (2018) [52] asserted that CO₂ emissions per unit of GDP serve as a superior metric and more accurately represent a nation’s energy and economic performance. Carbon emission intensity is determined by the ratio of carbon dioxide emissions to regional GDP. The carbon emission intensity index is the logarithmic value of carbon emission intensity, owing to the significant disparity between the index and other metrics.

3.2.2. Independent Variable

This study expands on Zhao et al. (2024) by including additional indicators for digital infrastructure, digital industry development, and digital financial inclusion while retaining key elements like internet broadband access ports and the number of domain names [53]. The digital economy is evaluated via the following three dimensions: digital economy infrastructure, digital industry development, and inclusive digital finance. A thorough assessment framework for the digital economy is established, incorporating 13 secondary indicators, such as the quantity of domain names, the count of information-based firms, and the digitalisation degree index.

Table 1. Index evaluation system for digital economy development level.

Primary Indicators	Secondary Indicators	Third-Level Indicators
The development level of the digital economy	Digital infrastructure	Number of domain names
		Number of IPv4 websites
		Number of internet broadband access ports
		Mobile phone penetration rate
		Cable length per unit area
	Digital industry development	Number of information enterprises
		Number of websites per 100 enterprises
		The proportion of enterprises with e-commerce transactions
		E-commerce sales volume
		Software business revenue
	Digital financial inclusion	Digital degree index
		The depth index is used
		Coverage breadth index

provides a specific description of the evaluation system for digital economy development level.

3.2.3. Intermediate Variable

Green patents exemplify the inventive utilisation of eco-friendly technology focused on resource conservation and environmental preservation, while also inherently indicating the extent and magnitude of regional green innovation. Utilising the research of Dong et al. (2020) [54] as a reference, the logarithm of green patent applications serves to characterise the extent of regional green technological innovation.

Addressing environmental degradation, particularly carbon emissions, and progressively advocating for the use of renewable and clean energy as alternatives to conventional fossil fuels are essential for realising a green transformation of China’s energy consumption framework. According to the research conducted by Feng et al. (2009) [55], the energy consumption structure is assessed by the ratio of total coal use to total energy consumption.

3.2.4. Adjusting Variable

Based on the research of Huang et al. (2022) [56], seven factors such as green credit, green investment and green insurance are selected and a green finance assessment system is created. The calculation technique of each indicator is provided

Table 2. Evaluation system of green finance development level index.

Primary Indicators	Secondary Indicators	Third-Level Indicators
Development level of green finance	Green Credit	Total credit for environmental protection projects/Total credit of the entire province
	Green Investment	Investment in environmental pollution control/GDP
	Green Insurance	Environmental pollution liability insurance income/Total premium income
	Green Bond	Total issuance of green bonds/Total issuance of all bonds
	Green Support	Fiscal environmental protection expenditure/Fiscal general budget expenditure
	Green Fund	Total market value of green funds/Total market value of all funds
	Green Benefits	The total amount of carbon trading, energy consumption rights trading, pollution discharge rights trading/Equity market trading

.

In the transmission mechanism model, green fund support adopts green credit (FS) as the measurement indicator, while digital technology investment (DTI) selects the proportion of the added value of the digital technology industry to GDP as the measurement indicator.

3.2.5. Control Variables

To address endogeneity concerns stemming from missing variables, the model adds many control variables, including government intervention, urbanisation level, and human capital level. Individual sample missing values are imputed by linear interpolation. The intensity of research and development (RES), measured by the ratio of internal R&D expenditure to regional GDP, is linked to the innovation capacity of the digital economy. The level of urbanisation (URB), expressed as the ratio of urban population to total population, relates to agglomeration effects and resource allocation efficiency in the digital economy. Foreign trade openness (OPE), evaluated by the ratio of total imports and exports to regional GDP, connects with the internationalisation of the digital economy and low-carbon technology transfer. Human capital level (HUM) is reflected by the ratio of higher education students to the total population, relating to innovation and application within the digital economy. Government intervention degree (GOV) is indicated by government fiscal expenditure relative to regional GDP, influencing carbon emission policies.

3.3. Methodology

3.3.1. Mediation Effect Model

Based on the statistical software Stata 14, the correlation between the digital economy and carbon emissions is analysed. A direct effect model is then constructed to examine the possible pathways via which the digital economy may influence carbon emission intensity, as shown in Figure 1. Consequently, green technological innovation is incorporated as a mediating variable, which results in the formulation of the following model:

$${\mathrm{m}\mathrm{e}\mathrm{d}}_{\mathrm{i},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{D}\mathrm{I}\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_{\mathrm{j}}{\mathrm{x}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{c}}+{\mathsf{\sigma }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(1)

$${\mathrm{C}\mathrm{E}\mathrm{R}}_{\mathrm{i},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{D}\mathrm{I}\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_2{\mathrm{m}\mathrm{e}\mathrm{d}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_{\mathrm{j}}{\mathrm{x}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{c}}+{\mathsf{\sigma }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(2)

Equation (1) is a basic regression model to test the direct effect of the digital economy (DE) and green finance (GF) on carbon emission intensity (CER) and whether green finance moderates the effect of the digital economy on CER. By introducing an interaction term (DE × GF), the moderating role of green finance in the relationship between the digital economy and carbon intensity can be assessed. Equation (2) is a mediated effect model to test the indirect effect of the digital economy on carbon emission intensity through mediating variables. By introducing the mediating variable and its interaction term with the digital economy, the mechanism of the mediating variable’s role in the relationship between the digital economy and carbon emission intensity can be assessed. ${\mathrm{C}\mathrm{E}\mathrm{R}}_{\mathrm{i},\mathrm{t}}$ and ${\mathrm{D}\mathrm{I}\mathrm{G}}_{\mathrm{i},\mathrm{t}}$ represent the city and year, respectively, and denote the carbon emission intensity and level of digital economic development, respectively; ${\mathrm{m}\mathrm{e}\mathrm{d}}_{\mathrm{i},\mathrm{t}}$ represents the mediator variable; ${\mathrm{x}}_{\mathrm{i},\mathrm{t}}$ are control variables, representing other possible characteristic variables that may affect carbon emission reduction; and ${\mathsf{\sigma }}_{\mathrm{t}}$ is the intercept term. We apply an elasticity coefficient to the relationship between digital economy and carbon emission intensity. If ${\mathsf{\alpha }}_1$ < 0 in a significant manner, this indicates that the digital economy can effectively suppress carbon emission intensity; conversely, the digital economy cannot effectively suppress carbon emissions. ${\mathsf{\alpha }}_{\mathrm{j}}$ represents the degree of influence of the control variable on carbon emissions. ${\mathsf{\mu }}_{\mathrm{c}}$ and ${\mathsf{\sigma }}_{\mathrm{t}}$ are unobservable individual- and time-fixed effects. ${\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$ is the random disturbance term.

3.3.2. Adjustment Effect Model

To verify the regulatory role of green finance, the following model is constructed:

$${\mathrm{C}\mathrm{E}\mathrm{R}}_{\mathrm{i},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{D}\mathrm{I}\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_2{\mathrm{G}\mathrm{F}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_3{\mathrm{D}\mathrm{I}\mathrm{G}}_{\mathrm{i},\mathrm{t}}\ast {\mathrm{G}\mathrm{F}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_4{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{c}}+{\mathsf{\sigma }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(3)

Among them, ${\mathrm{D}\mathrm{I}\mathrm{G}}_{\mathrm{i},\mathrm{t}}\ast {\mathrm{G}\mathrm{F}}_{\mathrm{i},\mathrm{t}}$ represents the interaction term of the digital economy and green finance; $\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}$ represents the control variable; and the meaning of the other variables remains unchanged.

3.3.3. Threshold Effect Model

With the advancement of green finance, the influence of the digital economy on carbon emissions will vary. This research uses the Hansen threshold regression model to establish the subsequent panel threshold model for the purpose of validating this result.

$${\mathrm{C}\mathrm{E}\mathrm{R}}_{\mathrm{i},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{G}\mathrm{F}}_{\mathrm{i}\mathrm{t}}\mathrm{I}({\mathrm{q}}_{\mathrm{i}\mathrm{t}}\le \mathsf{\gamma }1)+{\mathsf{\alpha }}_2{\mathrm{G}\mathrm{F}}_{\mathrm{i}\mathrm{t}}\mathrm{I}\left({{\mathrm{q}}_{\mathrm{i}\mathrm{t}}>\mathsf{\gamma }}_1\right)+{\mathsf{\alpha }}_4{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{c}}+{\mathsf{\sigma }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(4)

where ${\mathsf{\gamma }}_{\mathrm{i}}$ represents the $\mathrm{i}$-th threshold value and $\mathrm{I}$ (・) is the schematic function. The condition in parentheses is 1, or otherwise 0. The threshold variable is indicated, and the other variables have the same meaning.

3.3.4. Conduction Mechanism Model

Furthermore, this paper constructs a moderating model to verify the conductivity of digital technology investment. The specific settings are as follows:

$${\mathrm{D}\mathrm{T}\mathrm{I}}_{\mathrm{i},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{D}\mathrm{I}\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_2{\mathrm{G}\mathrm{F}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_{\mathrm{j}}{\mathrm{x}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{c}}+{\mathsf{\sigma }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(5)

$${\mathrm{D}\mathrm{T}\mathrm{I}}_{\mathrm{i},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{D}\mathrm{I}\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_2{\mathrm{G}\mathrm{F}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_3{\mathrm{D}\mathrm{I}\mathrm{G}}_{\mathrm{i},\mathrm{t}}\ast {\mathrm{G}\mathrm{F}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_{\mathrm{j}}{\mathrm{x}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{c}}+{\mathsf{\sigma }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(6)

Model (5) is without intersection and multiplication terms, while Model (6) is with intersection and multiplication terms. Among them, ${\mathrm{D}\mathrm{T}\mathrm{I}}_{\mathrm{i},\mathrm{t}}$ is the income variable of conduction. The explanations of the other variables are consistent with those of the benchmark model and will not be elaborated here.

4. Empirical Results

4.1. Benchmark Regression

Table 3. Results of the benchmark regression.

Variable	(1)	(2)	(3)
	Base Model	Model with Control Variables	Full Model with Fixed Effects
$\mathrm{D}\mathrm{I}\mathrm{G}$	−2.937 ***	−3.270 ***	−1.798 ***
	(−12.56)	(−11.31)	(−6.44)
$\mathrm{R}\mathrm{E}\mathrm{S}$		8.367 *** (6.12)	8.015 *** (4.38)
$\mathrm{U}\mathrm{R}\mathrm{B}$		1.484 *** (3.46)	2.485 *** (4.34)
$\mathrm{O}\mathrm{P}\mathrm{E}$		−0.191 (−1.36)	−1.057 *** (−5.37)
$\mathrm{H}\mathrm{U}\mathrm{M}$		−25.452 *** (−4.29)	−23.076 *** (−3.93)
$\mathrm{G}\mathrm{O}\mathrm{V}$		1.129 *** (3.63)	1.275 *** (3.44)
$\_\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}$		7.956 *** (51.05)	7.852 *** (35.77)
Control Variable	N	Y	Y
Fixed Year	N	N	Y
Area Fixed	N	N	Y
Sample Size	390	390	390
$\mathrm{Adjustment} \mathrm{of} {\mathrm{R}}^2$	0.300	0.404	0.440

Note: *** represent the significant level of 1%; the value in brackets below the coefficient is the robust standard error t value.

presents the results of the benchmark regression, providing a foundational analysis of the impact of the digital economy on carbon emission intensity (CER). The base Model (1) includes only the primary explanatory variable, the digital economy development index, which has a coefficient of −2.937 (t-statistic = −12.56). This indicates a strong negative correlation between the digital economy and carbon emission intensity, suggesting that higher levels of digital economy development are associated with a lower carbon emission intensity, even without controlling for other factors. In Model (2), after incorporating control variables, the coefficient remains negative at −3.270 (t-statistic = −11.31) and statistically significant, indicating that the negative impact of the digital economy on carbon emission intensity is robust, even after accounting for other variables. In Model (3), with the addition of fixed effects, the coefficient is −1.798 (t-statistic = −6.44), showing that the digital economy still has a significant negative impact on carbon emission intensity after controlling for all observable and unobservable factors. This demonstrates that the digital economy is an effective tool for reducing carbon emissions.

4.2. Robustness Testing

First, the robustness of the model is tested by substituting explanatory variables. When the natural logarithm plus one is used to replace the digital economy index, the coefficient in the first column of

Table 4. Results of the robustness test.

Variable	(1)	(2)	(3)	(4)
	Replace Explanatory Variables	Excluding Outliers	Delete the Municipality	Add Control Variables
$\mathrm{D}\mathrm{I}\mathrm{G}$	−0.932 ***	−0.613 ***	−0.625 ***	−1.915 ***
	(−4.90)	(−4.74)	(−4.83)	(0.324)
$\mathrm{R}\mathrm{E}\mathrm{S}$	−2.008 *** (−2.50)	−2.142 *** (−2.72)	−1.377 (−1.50)	4.648 ** (1.847)
$\mathrm{U}\mathrm{R}\mathrm{B}$	−2.391 *** (−8.33)	−2.397 *** (−8.24)	−3.012 *** (−7.36)	2.695 *** (0.505)
$\mathrm{O}\mathrm{P}\mathrm{E}$	0.152 *** (2.81)	0.146 *** (2.67)	0.072 (1.00)	−0.739 *** (0.214)
$\mathrm{H}\mathrm{U}\mathrm{M}$	0.526 (0.18)	0.793 (0.27)	1.830 (0.55)	−12.43 * (6.751)
$\mathrm{G}\mathrm{O}\mathrm{V}$	1.426 *** (8.51)	1.434 *** (8.42)	1.482 *** (8.02)	2.094 *** (0.301)
$\mathrm{I}\mathrm{N}\mathrm{D}$				1.781 *** (−4.35)
$\_\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}$	9.506 *** (77.21)	9.493 *** (76.87)	9.736 *** (55.54)	6.161 *** (0.264)
Control Variable	Y	Y	Y	Y
Fixed Year	Y	Y	Y	Y
Area Fixed	Y	Y	Y	Y
Sample Size	390	390	338	390
$\mathrm{Adjustment} \mathrm{of} {\mathrm{R}}^2$	0.993	0.993	0.993	0.993

Note: *, **, *** represent the significant level of 10%, 5%, and 1% respectively; the value in brackets below the coefficient is the robust standard error t value.

remains negative (−0.932, t-statistic = −4.90). This indicates that even after adjusting the explanatory variables, the negative impact of the digital economy on carbon emission intensity persists. This substitution confirms the robustness of the initial model results. Second, the stability of the results is verified by reducing the impact of outliers. After applying a 2% Winsorization to all model variables, the coefficient for the second column data remains negative (−0.613, t-statistic = −4.74). This indicates that the results are not driven by outliers, further confirming the significant negative impact of the digital economy on carbon emission intensity. Thirdly, robustness is assessed by excluding municipalities directly under the central government (Beijing, Tianjin, Shanghai, and Chongqing). Given the significant disparities in economic scale and administrative level between these municipalities and other provinces, their potential biasing effect on the results is eliminated. Upon exclusion, the third column coefficient remains negative (−0.625, t-statistic = −4.83), signifying that the results are not influenced by these special regions and the negative impact of the digital economy on carbon emission intensity remains stable. Lastly, the robustness is reinforced by incorporating additional control variables. Considering the strong association between industrialisation and carbon emission intensity, the degree of industrialisation is added as a control variable. The fourth column coefficient remains negative (−1.915, t-statistic = 0.324), further substantiating the robust negative impact of the digital economy on carbon emission intensity, which is not affected by other significant variables.

4.3. Endogeneity Test

Table 5. The endogeneity test results.

Variable	(1)	(2)	(3)
	$2\mathbf{S}\mathbf{L}\mathbf{S}$	Explained Variables Lagged One Period	Explanatory Variables Lagged One Period
$\mathrm{D}\mathrm{I}\mathrm{G}$	−0.004 ***	−0.759 ***	−0.712 ***
	(−6.61)	(−5.53)	(−5.02)
$\mathrm{R}\mathrm{E}\mathrm{S}$	2.995 (1.39)	−2.525 *** (−2.98)	−1.888 ** (−2.47)
$\mathrm{U}\mathrm{R}\mathrm{B}$	0.580 (1.16)	−2.313 *** (−7.24)	−2.327 *** (−7.60)
$\mathrm{H}\mathrm{U}\mathrm{M}$	−21.450 *** (−2.77)	−2.508 (−0.87)	2.014 (0.68)
$\mathrm{G}\mathrm{O}\mathrm{V}$	1.229 *** (3.88)	1.410 *** (8.50)	1.414 *** (8.20)
$\_\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}$	8.191 *** (42.54)	9.554 *** (68.99)	9.363 *** (73.34)
Control Variable	Y	Y	Y
Fixed Year	Y	Y	Y
Area Fixed	Y	Y	Y
Sample Size	390	360	360
$\mathrm{Adjustment} \mathrm{of} {\mathrm{R}}^2$	0.245	0.993	0.994
$\mathrm{K}\mathrm{P}\mathrm{r}\mathrm{k}\mathrm{L}\mathrm{M}$	0.889 ***
$\mathrm{K}\mathrm{P}\mathrm{r}\mathrm{k}\mathrm{W}\mathrm{a}\mathrm{l}\mathrm{d}\mathrm{F}$	89.300

Note: **, *** represent the significant level of 5%, and 1% respectively; the value in brackets below the coefficient is the robust standard error t value.

uses the instrumental variable approach to address the possible endogeneity between the digital economy index and carbon emission intensity. Specifically, the one-period lagged carbon emission intensity and the one-period lagged digital economy index are used as instrumental variables to minimise the effects of reverse causality and omitted variable bias on the estimation results. We use the two-stage least squares method to solve the endogeneity problem, which achieves this effectively by introducing instrumental variables, fitting the endogenous explanatory variables to the instrumental variables in the first-stage regression, and obtaining the fitted values. Replacing the endogenous explanatory variables in the original model with the fitted values in the second-stage regression effectively solves the endogeneity problem and makes the estimation results more accurate. The results show that the coefficient of the digital economy index is −0.004 and statistically significant (t-statistic = −6.61) when using the one-period lagged carbon emission intensity as an instrumental variable. This suggests that even after accounting for endogeneity, the development of the digital economy still has a significant negative impact on carbon emission intensity. Similarly, when the one-period lagged digital economy index is used as an instrumental variable, the coefficient is −0.759 (t-statistic = −5.53), further confirming that the negative impact of the digital economy on carbon emission intensity is statistically robust. In addition, the coefficient of the digital economy index is −0.712 (t-statistic = −5.02), which remains negative and statistically significant when the one-period lagged values of both explanatory and interpreted variables are considered. This further suggests that the negative effect of the digital economy on carbon emission intensity still holds after controlling for dynamic correlation in the time series. These results consistently show that the development of the digital economy contributes significantly to the reduction in carbon emission intensity, and this finding remains robust after accounting for endogeneity issues. In addition, Table 5 reports the results of the relevant statistical tests, the Kleibergen–Paap rk LM statistic and the Cragg–Donald Wald F statistic, which all indicate that the chosen instrumental variables are valid. This further enhances the credibility and ensures the reliability of the findings.

4.4. Mediation Effect Test

We analyse the effects of the digital economy’s growth on carbon emission intensity trajectories. The framework of energy consumption and green technology innovation is incorporated as an intermediary variable, with the results of the mechanism of action delineated in

Table 6. Results of mediation effect.

Variable	$\left(1\right)$	$\left(2\right)$	(3)	(4)	(5)	(6)
	$\mathbf{C}\mathbf{E}\mathbf{R}$	$\mathbf{E}\mathbf{C}\mathbf{S}$	$\mathbf{C}\mathbf{E}\mathbf{R}$	$\mathbf{C}\mathbf{E}\mathbf{R}$	$\mathbf{G}\mathbf{I}$	$\mathbf{C}\mathbf{E}\mathbf{R}$
$\mathrm{D}\mathrm{I}\mathrm{G}$	−3.221 ***	−0.574 *	−2.873 ***
	(−10.95)	(−1.86)	(−12.64)
$\mathrm{E}\mathrm{C}\mathrm{S}$			0.606 ***
			(19.73)
$\mathrm{D}\mathrm{I}\mathrm{G}\_1$				−3.221 ***	2.099 ***	−2.949 ***
				(−10.95)	(3.84)	(−10.20)
$\mathrm{G}\mathrm{I}$						−0.130 ***
						(−3.31)
$\mathrm{R}\mathrm{E}\mathrm{S}$	5.232 *** (3.45)	13.234 *** (6.81)	−2.782 * (−1.92)	5.232 *** (3.45)	5.054 * (1.80)	5.887 *** (4.01)
$\mathrm{U}\mathrm{R}\mathrm{B}$	1.396 *** (3.22)	2.933 *** (5.85)	−0.380 (−1.12)	1.396 *** (3.22)	0.727 (1.45)	1.490 *** (3.40)
$\mathrm{O}\mathrm{P}\mathrm{E}$	−1.105 (−0.74)	−1.107 *** (−5.80)	0.566 *** (5.08)	−1.105 (−0.74)	0.692 *** (2.97)	−0.015 (−0.10)
$\mathrm{H}\mathrm{U}\mathrm{M}$	−21.503 *** (−3.55)	−50.172 *** (−6.92)	8.881 ** (2.02)	−21.503 *** (−3.55)	20.277 *** (2.70)	−18.876 *** (−3.17)
$\mathrm{G}\mathrm{O}\mathrm{V}$	0.168 *** (3.68)	−5.291 *** (−18.11)	4.372 *** (15.77)	0.168 *** (3.68)	2.677 *** (9.00)	1.514 *** (4.05)
$\_\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}$	7.953 *** (50.88)	18.851 *** (103.28)	−3.462 *** (−5.85)	7.953 *** (50.88)	−0.798 *** (−3.77)	7.850 *** (47.36)
Control Variable	Y	Y	Y	Y	Y	Y
Fixed Year	Y	Y	Y	Y	Y	Y
Area Fixed	Y	Y	Y	Y	Y	Y
Sample Size	390	390	390	390	390	390
$\mathrm{Adjustment} \mathrm{of} {\mathrm{R}}^2$	0.392	0.483	0.674	0.392	0.456	0.405

Note: *, **, *** represent the significant level of 10%, 5%, and 1% respectively; the value in brackets below the coefficient is the robust standard error t value.

.

4.4.1. Energy Consumption Structure

The energy consumption structure denotes the ratio and allocation of various energy types within the total energy consumption of a country or area within a certain timeframe. In column (1), the predicted coefficient for the amount of digital economy growth on the energy consumption structure is −3.221, significant at the 1% level. This suggests that the advancement of the digital economy may significantly mitigate the energy consumption structure, hence decreasing the share of coal in overall energy consumption. Upon integrating the energy consumption structure as a mediating variable into the model, the findings indicate that the estimated coefficients for the energy consumption structure and the level of digital economy development on carbon emission intensity are 0.606 and −2.873, respectively, both of which meet the significance criterion. Specifically, controlling for other variables, a one-unit rise in digital economy development would indirectly decrease regional carbon emissions by 0.880 units (−1.452 × 0.606 = −0.880), resulting in a total effect of −3.753 (−2.873 − 0.880 = −3.753). The digital economy may attain carbon reduction via both direct and indirect impacts, with the direct effect being around thrice that of the indirect effect. Hypothesis H2a is validated.

4.4.2. Green Technology Innovation

In column (5), the calculated coefficient for the level of digital economy growth on green technological innovation is 2.099, which meets the significance criterion. This suggests that the advancement of the digital economy may significantly promote green technological innovation, hence elevating the innovation capacity of urban areas. Upon integrating green technological innovation as a mediating variable into the regression model, the estimated coefficients for green technological innovation and the level of digital economy development on carbon emission intensity are −0.130 and −2.949, respectively, both of which meet the significance criterion. Specifically, with all other parameters held constant, a one-unit rise in digital economy growth would indirectly decrease regional carbon emissions by 0.273 units (−0.130 × 2.099 = −0.273), resulting in a total effect of −3.222. The digital economy can facilitate carbon reduction via both direct and indirect impacts, but the indirect effect is far less apparent than the direct one. Consequently, Hypothesis H2b is affirmed.

4.5. Analysis of Heterogeneity

This section delves into the heterogeneity of the impact of the digital economy on carbon intensity, covering time variation, the level of development of the digital economy, and differences across different regions of China. The analysis aims to reveal how these factors affect the effectiveness of the digital economy in reducing carbon emissions.

In 2010, China issued the Decision on Accelerating the Cultivation and Development of Strategic Emerging Industries, listing the new generation of the information technology industry as one of the seven strategic emerging industries, which promoted the development of industries related to the digital economy. In 2014, China first proposed the “Internet Plus” action plan, which promoted the deep integration of the internet with traditional industries, injecting new impetus for the development of the digital economy, promoting the rapid development of the digital economy, and giving rise to a series of new industries and business models. In 2018, China issued the Opinions of the CPC Central Committee and State Council on Deepening Reforms of Institutional Mechanisms and Accelerating the Implementation of the Innovation-Driven Development Strategy, which explicitly called for accelerating the development of the digital economy and promoting the deep integration of the internet, big data, and artificial intelligence. Then, in 2022, China issued the 14th Five-Year Plan for the Development of the Digital Economy, which clearly defined the objectives, tasks, and safeguards for the development of the digital economy during the 14th Five-Year Plan period, and proposed that the added value of core industries in the digital economy would account for 10% of China’s GDP by 2025 and that the share of these core industries in the digital economy would account for 10% of China’s GDP. Due to the space problem, it is impossible to show the development of the digital economy each year, so we choose these four points in time to reflect the development of the digital economy.

In four maps, as shown in Figure 2, the colour intensity is strongly connected with the degree of digital economy development. As shown in Figure 2a, places exhibiting extremely low levels of the digital economy are broadly dispersed, with emerging areas of low levels mostly centred in some coastal regions and reasonably established economic zones. In contrast, regions with medium and high levels have not yet formed significantly in 2010. From Figure 2b,c, the number of very-low-level regions decreases, the distribution of low-level regions expands, medium-level regions appear in specific locations, and high-level regions begin to concentrate in certain core areas of the economy. In 2018, the momentum of development accelerates markedly, with a further decrease in the number of low-level regions and a significant increase in the number of medium-level regions, as well as the emergence of consecutive high-level regions, marking a shift towards a more concentrated pattern of development in the digital economy. By 2022, regions with extremely low levels have significantly diminished, persisting only in a few isolated areas; low-level regions have further shrunk, medium-level regions have expanded their coverage, and high-level regions exhibit a more stable distribution in key economic zones, resulting in a notable overall enhancement in the development of the digital economy in Figure 2d. This signifies that China’s digital economy has evolved over twelve years from initial endeavours to moderate expansion and consolidation. The geographical pattern of this has consistently developed, with developmental emphasis progressively transitioning towards medium- and high-level locations, illustrating the disparate advancement of China’s digital economy across various areas and its swift upgrading trajectory over time.

In 2014, China released the National Plan for Addressing Climate Change (2014–2020), which clarified the overall requirements, main objectives, and key tasks of China’s response to climate change by 2020 and proposed the target of reducing carbon dioxide emissions per unit of GDP by 18% compared to 2015. The 2018 “13th Five-Year Plan” mentioned a carbon emission intensity reduction target, while the 2022 policy further strengthened the dual control of carbon emissions and the carbon peak target. Due to space constraints, it is not possible to show the carbon emission intensity for each year, so these four points in time have been chosen to reflect the fluctuations in carbon emission intensity.

The intensity of the colours in Figure 3 visually represents the degree of carbon intensity, with darker shades indicating higher levels. In 2010, areas with a reduced carbon emission intensity are primarily located in southern China, whilst those with an elevated intensity are generally situated in northern provinces in Figure 3a. Between 2010 and 2018, the emission intensity of the southern provinces, as illustrated in Figure 3a–c, generally shows an upward trend, indicated by darker colours, while the emission intensity of the northern provinces decreases, indicated by lighter colours. By 2022, as shown in Figure 3d, the pattern of carbon emission intensity has grown increasingly intricate. Certain coastal provinces see an escalation in emission intensity, shown by deeper hues, whilst interior provinces display diverse alterations, with some regions becoming lighter and others darker. A subsequent study indicates that from 2010 to 2018, the regional distribution of carbon emission intensity in China generally adhered to a pattern of “eastern < central < western” and “northern < southern.” From 2018 to 2022, several coastal provinces saw a resurgence in carbon emission intensity. Since 2010, the persistent reduction in carbon emission intensity in China’s inland regions indicates the enforcement of stringent environmental policies, the rapid integration of renewable energy, and industrial enhancement propelled by centralised decarbonisation mandates and interregional technology transfer under the national green development strategy. The resurgence in carbon emission intensity in China’s coastal provinces from 2018 to 2022 may be ascribed to the emphasis on economic recovery and industrial production during that timeframe, revitalising fossil-fuel-dependent businesses. Moreover, local administrations’ attempts to reconcile economic growth objectives with decarbonisation mandates resulted in a provisional easing of emission regulations in export-driven areas.

According to Figure 2 and Figure 3, this paper compares and discusses at the provincial level. The changes in the digital economy and the intensity of carbon emissions in the eastern region are more complicated. In 2010, the eastern region had a better digital economy foundation, but the carbon emission intensity was higher. In 2014, the digital economy developed rapidly, and the intensity of carbon emissions decreased. In 2018, the digital economy developed further, and the carbon emission intensity decreased. In 2022, the digital economy entered a mature stage. However, the intensity of carbon emissions in some coastal provinces increased, which may have been related to economic recovery and the recovery of energy-intensive industries. In the central region, the changes in the digital economy and carbon emission intensity are more obvious. In 2010, the foundation of the digital economy in the central region was weak, and the carbon emission intensity was high. In 2014, the digital economy began to develop, and the carbon emission intensity declined. In 2018, the digital economy developed rapidly, and the carbon emission intensity declined significantly. In 2022, the digital economy developed further, and the carbon emission intensity continued to decline, which shows that the digital economy has a more significant optimisation effect on carbon emission intensity. In the western region, the digital economy and carbon emission intensity change more slowly. In 2010, the foundation of the digital economy in the western region was weak, and the carbon emission intensity was high. In 2014, the digital economy began to develop, and the carbon emission intensity declined. In 2018, the digital economy developed further, and the intensity of carbon emissions declined. In 2022, the digital economy entered a mature stage, and the carbon emission intensity declined further. However, the overall level here is still higher than that of the eastern and central regions. In the northeastern region, changes in the digital economy and carbon emission intensity are slower. In 2010, the northeastern region had a weak digital economy foundation and a high carbon emission intensity. In 2014, the digital economy began to develop, and the carbon emission intensity declined. In 2018, the digital economy further developed, and the carbon emission intensity declined. In 2022, the digital economy entered a mature stage. However, a decline in carbon emission intensity is not apparent, which may be related to the pressure of economic transformation and the difficulty of industrial restructuring in the northeast region.

4.5.1. Analysis of the Temporal Heterogeneity of the Whole Population

In 2015, China issued significant policy papers, including the “Guiding Opinions of the State Council on Actively Promoting the ‘Internet Plus’ Action” and “Made in China 2025”, signalling the commencement of industrial digital transformation. Implementing these measures offered assistance for the growth of the digital economy and considerably broadened its implications. Consequently, the research samples are categorised into the following two intervals: before and after 2015. In terms of time heterogeneity, before 2015, the coefficient of the digital economy is −0.129, statistically insignificant (t-statistic = −0.44), indicating that the impact of the digital economy on carbon emission intensity is not significant in this period, and its development has not yet significantly contributed to carbon emission reduction, whereas, after 2015, the coefficient of the digital economy becomes −0.354 and is statistically significant at the 5% level (t-statistic = −2.42), which indicates that with the further development of the digital economy and the promotion of relevant policies, its negative impact on carbon emission intensity begins to appear and gradually strengthens, indicating that the potential of the digital economy to promote carbon emission reduction is gradually unearthed and plays a more positive role in realising the goal of low-carbon development. The analytical results demonstrate that since 2015, the digital economy has exerted a substantial negative influence on reducing carbon emissions, whereas its impact was negligible before 2015. The fast advancement of the digital economy has facilitated the optimisation of industrial structures, accelerating the shift towards a low-carbon economy. The sophisticated nature of digital management enhances resource allocation efficiency, thereby decreasing carbon emissions. The utilisation of digital technology is crucial in energy management and monitoring, contributing to reducing carbon emissions. Before 2015, the absence of these requirements rendered the carbon-reducing impact of the digital economy negligible. Consequently, it may be asserted that the carbon-reducing impact of the digital economy has a nonlinear nature to some degree.

4.5.2. Heterogeneity of the Development Level of the Digital Economy

This section classifies 30 provinces into the following two categories based on their digital economy development levels in 2022: high-level regions for digital economy and low-level regions for digital economy. The high-level group includes 15 provinces such as Guangdong, Beijing, and Shanghai, while the low-level group comprises 15 provinces including Qinghai, Gansu, and Heilongjiang. From the viewpoint of the heterogeneity of the development level of the digital economy, the coefficient of provinces in the low-level digital economy group is −1.757 and statistically significant at the 1% level (t-statistic = −8.16), which indicates that in areas with a relatively weak digital economy foundation, the advancement of the digital economy is still able to produce a more significant negative impact on the intensity of carbon emissions despite its limited level of development, suggesting that the digital economy’s initial development is able to inhibit the growth of carbon emissions to a certain extent, providing a new impetus for the low-carbon transformation of these regions. In the provinces in the high-level digital economy group, the coefficient is further reduced to −14.288 and statistically significant at the 1% level (t-statistic = −9.00), which indicates that in regions with a more developed digital economy, with the in-depth application of digital technology and the digitalisation transformation of industries accelerated, its inhibitory effect on carbon emission intensity is stronger and the synergistic effect of the digital economy and low-carbon development is more prominent. These regions have greater advantages and potential in utilising the digital economy to realise green transformation, which can provide a demonstration of and reference for low-carbon development in the country.

4.5.3. Heterogeneity of the Regional Distribution

According to the National Bureau of Statistics, China is divided into the following four regions: east, central, west, and northeast. The eastern region includes 10 provinces such as Guangdong and Zhejiang; the central region consists of 6 provinces, including Hunan and Hubei; the western region comprises 12 provinces, such as Sichuan and Yunnan; and the northeast includes Liaoning, Jilin, and Heilongjiang.The results are shown in

Table 7. Heterogeneity results.

Variable	Time Heterogeneity		Digital Economy		Area Distribution
	Before 2015	After 2015	Low-Level Group	High-Level Group	East	Middle	West	Northeast
$\mathrm{D}\mathrm{I}\mathrm{G}$	$-0.129$	−0.354 **	−1.757 ***	−14.288 ***	0.163	−1.389 ***	−0.807 **	−0.900
	(−0.44)	(−2.42)	(−8.16)	(−9.00)	(1.51)	(−6.93)	(−2.36)	(−0.45)
$\mathrm{R}\mathrm{E}\mathrm{S}$	−1.925 ***	−1.997 ***	3.738	8.921 ***	−4.211 ***	−1.110	0.140	−46.022 **
	(−3.10)	(−5.05)	(1.65)	(4.52)	(−5.31)	(−0.88)	(0.14)	(−2.48)
$\mathrm{U}\mathrm{R}\mathrm{B}$	−0.004	−0.066	0.472	4.738 ***	−0.948 ***	−2.355 **	−1.885 ***	0.743
	(−0.03)	(−0.61)	(0.84)	(9.41)	(−2.99)	(−2.40)	(−4.19)	(0.56)
$\mathrm{O}\mathrm{P}\mathrm{E}$	2.907	2.844	0.250	−1.949 ***	0.145 **	0.718 ***	0.102	−0.400
	(0.31)	(0.88)	(1.60)	(−8.32)	(2.44)	(2.85)	(0.98)	(−0.97)
$\mathrm{H}\mathrm{U}\mathrm{M}$	0.934 **	0.901 ***	−11.962	−11.135	11.909 ***	22.714 ***	−18.619 ***	−4.422
	(2.28)	(3.92)	(−1.48)	(−1.25)	(3.02)	(3.74)	(−6.19)	(−0.52)
$\mathrm{G}\mathrm{O}\mathrm{V}$	−0.680	−1.032	1.783 ***	−0.416	0.905 ***	3.396 ***	1.082 ***	0.644 **
	(−0.74)	(−1.17)	(2.68)	(−1.25)	(2.64)	(7.61)	(7.40)	(2.31)
$\_\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}$	9.302 ***	9.160 ***	7.691 ***	7.526 ***	8.552 ***	8.720 ***	9.787 ***	9.455 ***
	(25.81)	(41.71)	(29.73)	(44.09)	(38.28)	24.99	50.52	17.02
Control Variable	Y	Y	Y	Y	Y	Y	Y	Y
Fixed Year	Y	Y	Y	Y	Y	Y	Y	Y
Area Fixed	Y	Y	Y	Y	Y	Y	Y	Y
Sample Size	180	240	195	195	130	78	143	39
$\mathrm{Adjustment} \mathrm{of} {\mathrm{R}}^2$	0.996	0.997	0.218	0.409	0.988	0.997	0.998	0.994

Note: **, *** represent the significant level of 5%, and 1% respectively; the value in brackets below the coefficient is the robust standard error t value.

, in terms of the heterogeneity of regional distribution, the coefficient of the digital economy in the eastern region is 0.163, statistically insignificant (t-statistic = 1.51), which may be due to the fact that the economy of the eastern region is more developed, its industrial structure has been relatively optimised, the marginal effect of further development of the digital economy on carbon emission reduction is relatively small, and its impact on the intensity of carbon emissions has not yet reached the level of statistical significance. The coefficient of the digital economy in the central region is −1.389, which is statistically insignificant. The economy coefficient is −1.389 and is statistically significant at the 1% level (t-statistic = −6.93), indicating that the carbon emission intensity in the central region has been significantly reduced under the impetus of the digital economy and that the digital economy plays an important role in carbon emission reduction in the region, which may be closely related to the measures of accelerating the construction of digital infrastructure in the central region in recent years and promoting the digital transformation of traditional industries. The integration and development of traditional industries effectively improve the efficiency of energy utilisation and reduce carbon emissions. The coefficient of the digital economy in the western region is −0.807 and is statistically significant at the 5% level (t-statistic = −2.36), which indicates that the development of the digital economy in the western region has also positively affected carbon emission reduction, and even though the degree of its influence is slightly smaller than that in the central region, it is still statistically significant, reflecting the role of the digital economy in promoting the greening of the western region and the reduction in carbon emissions. The digital economy plays an important role in promoting green development in the western region and narrowing the carbon emission gap with the eastern region. The coefficient of the digital economy in the northeastern region is −0.900, which is statistically insignificant (t-statistic = −0.45), which may be related to the pressure of economic transformation faced by the northeastern region and the difficulty of industrial structural adjustment, etc. The carbon emission reduction effect of the digital economy in the region has not yet been fully revealed. Further strengthening the interaction between digital technology and traditional industries is necessary. It is necessary to further strengthen the in-depth integration of digital technology and traditional industries and promote the optimisation and upgrading of economic structure to better utilise the digital economy’s potential in carbon emission reduction.

4.6. Regulatory Effect

Table 8. Results of the group regression.

Variable	(1)	(2)
	Low-Level Group	High-Level Group
$\mathrm{D}\mathrm{I}\mathrm{G}$	−0.138	−0.615 **
	(−0.76)	(−2.24)
$\mathrm{R}\mathrm{E}\mathrm{S}$	0.038 *** (2.72)	0.005 (0.35)
$\mathrm{U}\mathrm{R}\mathrm{B}$	−8.375 *** (−4.09)	6.396 (1.26)
$\mathrm{O}\mathrm{P}\mathrm{E}$	11.457 ** (2.35)	4.603 *** (2.70)
$\mathrm{H}\mathrm{U}\mathrm{M}$	1.495 (1.10)	2.631 * (1.81)
$\mathrm{G}\mathrm{O}\mathrm{V}$	−0.313 * (−1.94)	0.367 *** (1.61)
$\_\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}$	8.129 *** (50.75)	7.852 *** (35.77)
Control Variable	Y	Y
Fixed Year	Y	Y
Area Fixed	Y	Y
Sample Size	390	390
$\mathrm{Adjustment} \mathrm{of} {\mathrm{R}}^2$	0.404	0.440

Note: *, **, *** represent the significant level of 10%, 5%, and 1% respectively; the value in brackets below the coefficient is the robust standard error t value.

presents the results of a grouped regression analysis that examines the different impacts of the digital economy on carbon emission intensity based on the level of green finance development. This analysis helps us to understand how green finance acts as a moderating variable to influence the relationship between the digital economy and carbon emissions. The coefficient of the digital economy is −0.138, which is statistically insignificant (t-statistic = −0.76) in regions with a low level of green finance development. This indicates that the digital economy has less impact on the carbon emission intensity in regions with less developed green finance. A lack of green financial support may limit the effectiveness of digital technologies in reducing carbon emissions. On the contrary, the coefficient of digital economy is −0.615 and is statistically significant at the 5% level (t-statistic = −2.24) in regions with a higher level of green finance development. This indicates that the negative effect of the digital economy on carbon emission intensity is more significant in regions with more developed green finance. Abundant green financial resources may enhance the ability of firms to invest in more environmentally friendly digital technologies, thereby reducing carbon emissions more effectively. The control variables included in the model, such as government intervention, urbanisation level, and human capital level, also show different effects in the two groups. For example, in the high-level group, the coefficient of government intervention is 0.005, which is statistically insignificant (t-statistic = 0.35), suggesting that government intervention may have only a limited additional impact in regions where green finance is already more developed. In the low-level group, on the other hand, the coefficient of government intervention is 0.038 and is statistically significant at the 1% level (t-statistic = 2.72), suggesting that government intervention plays a more critical role in stimulating the carbon abatement effect of the digital economy in regions where green finance is less developed.

The inclusion of fixed effects helps to control for unobserved heterogeneity across regions and over time. The results show that the fixed effects are statistically significant in both groups, confirming the importance of considering these factors in the analysis.

To deeply explore which factors in green finance can effectively enhance the carbon reduction effect of the digital economy, this content will focus the research on the key area of green fund support (FS) and select green credit indicators as its alternative variables for further analysis. In column (2), the interaction term between the digital economy and green finance remains significantly negative, suggesting that their combined effect strengthens carbon reduction. The coefficients and signs of other control variables are stable, confirming that the digital economy and green finance enhance carbon reduction. The research findings reveal that the coefficients for the digital economy and green finance for carbon emission intensity are −2.413 and −0.371, respectively, both significant at the 1% level in

Table 9. Moderating effect results.

Variable	(1)	(2)
	GF	$\mathbf{F}\mathbf{S}$
$\mathrm{D}\mathrm{I}\mathrm{G}$	−2.413 ***	−1.427 ***
	(−11.20)	(−2.78)
$\mathrm{G}\mathrm{F}$	−0.371 *** (−3.14)	−2.13 *** $(-8.11)$
$\mathrm{D}\mathrm{I}\mathrm{G}$$*GF/$ $\mathrm{D}\mathrm{I}\mathrm{G} \mathrm{*} \mathrm{F}\mathrm{S}$	−5.834 *** (−8.16)	−6.273 * $(-1.74)$
$\mathrm{R}\mathrm{E}\mathrm{S}$	0.049 ** (2.50)	0.051 ** (2.58)
$\mathrm{U}\mathrm{R}\mathrm{B}$	1.285 (0.24)	8.384 (1.53)
$\mathrm{O}\mathrm{P}\mathrm{E}$	74.747 *** (6.78)	72.081 *** (7.13)
$\mathrm{H}\mathrm{U}\mathrm{M}$	5.110 (1.27)	4.639 (1.05)
$\mathrm{G}\mathrm{O}\mathrm{V}$	−0.292 (−0.82)	−1.021 ** (−2.50)
$\_\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}$	9.620 *** (67.70)	8.824 *** $\left(30.67\right)$
Control Variable	Y	Y
Fixed Year	390	390
$\mathrm{Adjustment} \mathrm{of} {\mathrm{R}}^2$	0.454	0.454

Note: *, **, *** represent the significant level of 10%, 5%, and 1% respectively; the value in brackets below the coefficient is the robust standard error t value.

. The interaction term coefficient between the digital economy and green finance is −5.834, which is significant at the 1% level. This indicates that a beneficial synergistic effect may occur when the two are amalgamated.

4.7. Regulation Mechanism Test

Digital technology investment is introduced as a mediating variable to further investigate the role of green finance funding support in the relationship between the digital economy and carbon reduction.

In column (1) of

Table 10. Test of the regulatory mechanism.

Variable	(1)	(2)	$\mathbf{Variable}$	(3)
	$\mathbf{C}\mathbf{E}\mathbf{R}$	$\mathbf{D}\mathbf{T}\mathbf{I}$		$\mathbf{C}\mathbf{E}\mathbf{R}$
$\mathrm{D}\mathrm{I}\mathrm{G}$	−1.217 ***	1.241 ** (2.48)	$\mathrm{D}\mathrm{I}\mathrm{G}$	−1.427 ***
	(−9.23)			(−2.68)
$\mathrm{F}\mathrm{S}$	−1.035 *** (−4.42)	1.098 ** $\left(2.36\right)$	$\mathrm{D}\mathrm{T}\mathrm{I}$	−2.021 *** (−6.11)
$\mathrm{D}\mathrm{I}\mathrm{G}$$*\mathrm{F}\mathrm{S}$	−1.086 ** (−2.53)	1.151 *** $\left(5.31\right)$	$\mathrm{D}\mathrm{T}\mathrm{I}$$*\mathrm{F}\mathrm{S}$	−1.063 * (−1.84)
$\mathrm{R}\mathrm{E}\mathrm{S}$	0.049 ** (2.50)	8.921 *** (4.52)	$\mathrm{R}\mathrm{E}\mathrm{S}$	0.051 ** (2.58)
$\mathrm{U}\mathrm{R}\mathrm{B}$	1.285 (0.24)	4.738 *** (9.41)	$\mathrm{U}\mathrm{R}\mathrm{B}$	8.384 (1.53)
$\mathrm{O}\mathrm{P}\mathrm{E}$	74.747 *** (6.78)	−1.949 *** (−8.32)	$\mathrm{O}\mathrm{P}\mathrm{E}$	72.081 *** (7.13)
$\mathrm{H}\mathrm{U}\mathrm{M}$	5.110 (1.27)	6.157 (1.46)	$\mathrm{H}\mathrm{U}\mathrm{M}$	4.639 (1.05)
$\mathrm{G}\mathrm{O}\mathrm{V}$	−0.292 (−0.82)	−5.364 (1.48)	$\mathrm{G}\mathrm{O}\mathrm{V}$	−1.021 ** (−2.50)
$\_\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}$	9.620 *** (67.70)	10.647 *** $\left(20.31\right)$	$\_\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}$	8.824 *** (30.67)
Control variable	Y	Y	control variable	Y
Sample size	390	390	sample size	390
$\mathrm{Adjustment} \mathrm{of} {\mathrm{R}}^2$	0.454	0.455	$\mathrm{Adjustment} \mathrm{of} {\mathrm{R}}^2$	0.454

Note: *, **, *** represent the significant level of 10%, 5%, and 1% respectively; the value in brackets below the coefficient is the robust standard error t value.

, the estimated coefficients of digital technology, green finance support, and their interaction term are all negative and significant, showing that they can significantly reduce carbon emission intensity. In column (2), the coefficients of these variables are positive and significant at the 5% significance level, which indicates that the combination of the digital economy and green financial support can significantly enhance investment in digital technology. This suggests that financial assistance effectively alleviates the financial pressure on firms in digital transformation. In column (3), the interaction term coefficient between digital technology investment and green finance support is significantly negative for carbon emission intensity. This means that green finance support can enhance the carbon reduction effect of the digital economy through digital technology investment. It shows that green finance support directly promotes emission reduction and amplifies the digital economy’s carbon reduction benefits via digital technology investment. This creates an effective transmission path of “digital economy → digital technology investment → carbon emission reduction”, thereby significantly improving carbon reduction outcomes. The policy direction of financial assistance further motivates firms to concentrate their digital technology research and development on emission reduction scenarios, including energy management optimisation and real-time carbon emission monitoring. The extensive use of technologies such as the Internet of Things and blockchain can reduce energy consumption and carbon emissions per unit of production in industrial firms by 8% to 12%. Simultaneously, providing green insurance for technological iteration risks has elevated the technology adoption rate among firms by 24%. This sequence of effects not only expedites the transition of digital technology from concept validation to commercial application, but also perpetually enhances the carbon reduction potential of the digital economy through a positive feedback loop of “financing incentives—technology penetration—efficiency improvement.” Consequently, Hypothesis H3 is confirmed.

4.8. The Threshold Test

To experimentally investigate the nonlinear influence of the digital economy on carbon emission intensity, we utilise a panel threshold model. We apply Hansen’s methodology to assess threshold effects in the panel data. In the threshold effect model, the first threshold (Single) marks where the independent variable’s influence on the dependent variable begins to change, indicating an initial transformation in their relationship. The second threshold (Double) represents a significant change in this influence, reflecting a more profound relationship transformation. These thresholds illustrate the nonlinear relationships between variables and help us to understand different stages more accurately. The Bootstrap method is a statistical technique for estimating overall parameters by repeated sampling with returns from the original sample. The Bootstrap method with 400 repetitions means that 400 samples with put-backs are taken from the original sample, and each time, a self-help sample of the same size as the original sample is obtained. Statistics of interest, such as mean, standard deviation, etc., are computed for each self-help sample to obtain estimates of the 400 statistics. These estimates form an empirical distribution that can be used to infer the overall parameters, such as estimating the overall mean, constructing confidence intervals, and conducting hypothesis tests. This method applies to a wide range of data distributions, and the 400 replications strike a good balance between computational cost and estimation accuracy. Referring to Formula (4), after 400 Bootstrap sampling tests, the threshold effect results are shown in

Table 11. Results of the threshold model test.

Variable	Type	F-Number	P-Number	Significance Level			Threshold Value	Confidence Interval
				10%	5%	1%
$\mathrm{D}\mathrm{I}\mathrm{G}$	$\mathrm{S}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{l}\mathrm{e}$	33.55	0.050	26.209	33.504	49.831	0.025	[0.024, 0.026]
	$\mathrm{D}\mathrm{o}\mathrm{u}\mathrm{b}\mathrm{l}\mathrm{e}$	10.27	0.540	25.112	28.343	43.687	0.101	[0.094, 0.102]
$\mathrm{G}\mathrm{F}$	$\mathrm{S}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{l}\mathrm{e}$	23.48	0.063	20.369	24.238	35.380	0.660	[0.659, 0.661]
	$\mathrm{D}\mathrm{o}\mathrm{u}\mathrm{b}\mathrm{l}\mathrm{e}$	3.33	0.887	17.144	21.237	30.291	0.801	[0.797, 0.801]

. The digital economy and green finance pass the single-threshold test as threshold variables. However, in the double-threshold test, both p-values exceed 10%, failing this test.

Table 11 shows that as the advancement of the digital economy escalates, its influence on urban carbon reduction diverges. When the growth level of the digital economy is below the initial threshold, a 1% increase in its development results in a 3.324-unit rise in carbon emissions. This suggests that the initial phases of digital economy development do not facilitate decreased carbon emissions. Once the development level of the digital economy is above the initial threshold, the projected coefficient for carbon reduction is −0.945, indicating a substantial inhibitory influence on carbon emissions. The initial phase of the digital economy entails the manufacturing and deployment of new equipment, the construction of infrastructure, and digital transformation, all of which can result in increased energy consumption and carbon emissions, thereby hindering the complete realisation of the carbon reduction potential of the digital economy. As the digital economy advances, its intrinsic efficiency enhancements, optimal resource distribution, and empowerment of conventional sectors become evident. Utilising digital technology may optimise resource allocation, improve resource efficiency, and provide a synergistic optimisation effect, considerably enhancing the digital economy’s carbon reduction potential. Consequently, Hypothesis H4a is affirmed.

When the level of green finance development falls below the initial threshold value, its calculated coefficient is −1.320 and meets the significance criterion in

Table 12. The regression results of the threshold model.

Variable	(1)	(2)
	The Development Level of the Digital Economy	Level of Green Finance Development
$\mathrm{Threshold} \mathrm{Value} ({\mathsf{\gamma }}_1$)	0.025	0.800
$\mathrm{D}\mathrm{I}\mathrm{G}\mathrm{I}$ $({\mathrm{q}}_{\mathrm{i}\mathrm{t}}\le {\mathsf{\gamma }}_1$)	3.324 *** (4.39)	−1.320 *** (−9.54)
$\mathrm{D}\mathrm{I}\mathrm{G}\mathrm{I}$ $({\mathrm{q}}_{\mathrm{i}\mathrm{t}}>{\mathsf{\gamma }}_1$)	−0.945 *** (−9.38)	−1.074 *** (−9.77)
$\mathrm{R}\mathrm{E}\mathrm{S}$	−1.695 * (−1.85)	−1.382 (−1.49)
$\mathrm{U}\mathrm{R}\mathrm{B}$	−3.918 *** (−24.62)	−3.963 *** (−24.55)
$\mathrm{O}\mathrm{P}\mathrm{E}$	0.337 *** (6.37)	0.314 *** (5.77)
$\mathrm{H}\mathrm{U}\mathrm{M}$	4.873 ** (2.20)	2.897 (1.24)
$\mathrm{G}\mathrm{O}\mathrm{V}$	1.051 *** (8.39)	1.045 *** (8.21)
$\_\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}$	10.303 *** (151.58)	10.412 *** (162.56)
Control Variable	Y	Y
Sample Size	390	390
${\mathrm{R}}^2$	0.947	0.945

Note: *, **, *** represent the significant level of 10%, 5%, and 1% respectively; the value in brackets below the coefficient is the robust standard error t value.

. Upon surpassing the first barrier of the green finance development index, the calculated coefficient continues to be strongly negative, but with a diminished absolute value. In the preliminary phase of green financing, a heightened policy and market focus on green investments can readily induce a substitution effect for high-carbon emission projects, efficiently curtailing carbon emissions and enhancing the impact of carbon reduction. As green finance advances, the market progressively adjusts to and integrates its influence, reducing the marginal effect. Moreover, with the proliferation of green finance, there exists a potential for “greenwashing”, wherein specific projects may fail to deliver the anticipated substantial reductions in emissions, or the market’s “novelty premium” for green finance products diminishes, leading to a less significant deterrent effect on carbon emissions relative to the initial phase [57]. Consequently, Hypothesis H4b is validated.

In different studies exploring the relationship between the digital economy, green finance, and carbon emission intensity, various econometric methods have been adopted to adapt to their respective research objectives and data characteristics. Hong et al. [11] explored the inhibitory effect of the digital economy on carbon emission intensity through theoretical analyses and empirical tests on Chinese provincial panel data, using econometric models, spatial econometric analysis, and its mediating effect and spatial spillover effect through industrial structure upgrading, considering the dynamic changes in the time series, and also captured the interactions between regions through spatial econometric modelling. Wang et al. [57] investigated the impact of the digital economy on the carbon emission table and carbon emission intensity by using the double fixed effect model and mediated effect model. Zhao et al. [58], on the other hand, explored how the digital economy acts on the impact of green innovation on carbon emission performance using the system generalised method of moments, two-stage least squares, and panel quantile regression, with the choice of these methods reflecting the consideration of different model assumptions and data characteristics. Niu et al. [59] used a cross-regional input–output model to construct a carbon spillover feedback effect model to analyse the transfer of carbon emission pressure between regions from the perspective of spatial spillover. They assessed the fairness of carbon emissions from digital industries. This study mainly focused on the digital economy and green finance’s comprehensive impact on carbon emission intensity. It adopted a panel data model to analyse the direct impacts of the digital economy and green finance on carbon emission intensity and, at the same time, explored how green finance regulates the impact of the digital economy on carbon emission intensity through the moderating effect model and analysed the indirect impact of the digital economy on carbon emission intensity through the structural adjustment of energy consumption and green technological innovation by the intermediary effect model. E. Stoenoiu et al. [60] mainly used the generalised linear model (GLZ) to analyse the relationship between the information and communication technology (ICT) skills of employees trained by enterprises and the performance of enterprises through stepwise regression analysis. They analysed the interaction effects of factors such as the country, the year, and the proportion of ICT personnel. The interaction effects of factors such as country, year, and ICT personnel share were analysed. At the same time, this study used a panel data model and spatial econometric model, focusing on the relationship between the digital economy, industrial structure upgrading, and carbon emission intensity and its internal mechanism.

5. Conclusions and Discussion

5.1. Conclusions

This study examines the impact of the digital economy on carbon emission intensity in China and finds that the digital economy significantly reduces carbon emissions through both direct and indirect pathways. Specifically, the digital economy directly decreases carbon emission intensity by optimising resource allocation and enhancing energy efficiency, thereby supporting Hypothesis 1. It also indirectly reduces carbon emissions by restructuring the energy consumption mix (Hypothesis 2a) and promoting green technological innovation (Hypothesis 2b). Moreover, green finance acts as a vital moderator, amplifying the carbon-reducing effect of the digital economy and, thus, verifying Hypothesis 3. The research further reveals a nonlinear relationship, where the influence of the digital economy on carbon emission intensity changes with its development stage and the level of green finance development, confirming Hypotheses 4a and 4b. These results underscore the significance of integrating digital transformation with green financial strategies to attain sustainable low-carbon development.

These findings are consistent with the existing literature [61,62] and further confirm the importance of digital transformation in promoting a sustainable and eco-friendly economic development model. However, this study may have limitations, such as not considering spatial spillover effects [63] and considering green finance as a unified whole without distinguishing the specific impacts of different financial instruments. Future research needs to further explore how the digital economy and green finance interact with each other to affect carbon emissions and provide a more comprehensive analysis using spatial econometric methods. In addition, research should refine the components of green finance and analyse the differences and synergistic effects of different financial instruments in supporting the development of the digital economy and promoting carbon emission reduction. Future research directions include changing the experimental conditions to explore the interactions between digital development, green finance, and complementary policies. This study provides new insights into understanding how digitalisation and green finance can synergise to promote sustainable low-carbon development and provides valuable references for policymakers in promoting digital economy development and achieving carbon emission reduction targets. Future research should further deepen the understanding of the interaction mechanisms between the digital economy and environmental systems and explore more effective policy tools and strategies to promote the coordinated development of the economy, society, and the environment.

Notably, the study identifies green finance as an important moderating factor that enhances the environmental benefits associated with digital development. However, the empirical evidence suggests that the marginal effects of green finance diminish over time, reflecting potential saturation or inefficiencies in allocation. In contrast, the impact of carbon reduction on the digital economy appears to strengthen as technological and institutional capabilities mature. These findings underscore the evolving nature of digital and financial instruments in shaping carbon outcomes and highlight the interdependencies between technological systems and financial mechanisms in sustainable development.

The broader significance of this research lies in its implications for national and international climate governance. The results affirm that digital infrastructure development, when complemented by targeted green financing mechanisms, can support emission reduction goals without undermining economic performance. This has particular relevance for emerging economies seeking integrated approaches to climate policy and industrial modernisation.

5.2. Policy and Practical Implications

The findings of this study offer valuable insights for policymakers and practitioners aiming to leverage the digital economy and green finance to achieve sustainable low-carbon development. To maximise the carbon-reducing potential of the digital economy, it is essential to create a conducive policy environment that promotes the integration of digital technologies with sustainable finance. Regulatory frameworks should be designed to encourage innovation in green financial instruments while ensuring transparency, accountability, and long-term impact. For instance, policymakers could introduce targeted incentives for financial institutions that invest in digital infrastructure projects with proven environmental benefits, such as smart grids and energy-efficient data centres. Additionally, regulatory oversight should focus on preventing greenwashing practices and ensuring that green finance products contribute to emission reductions.

Moreover, tailored strategies that align digital economic growth with environmental objectives can serve as replicable models for other regions facing similar developmental and environmental challenges. Policymakers should consider developing comprehensive plans integrating digital transformation with green finance initiatives. These plans could include establishing digital innovation hubs focused on green technologies, providing tax breaks for companies that adopt digital solutions to reduce carbon emissions, and creating public–private partnerships to fund green digital projects. By fostering collaboration between the public and private sectors, policymakers can accelerate the adoption of digital technologies that drive sustainable development and reduce carbon emissions.

5.3. Limitations and Directions for Future Research

Despite its contributions, the study is subject to several limitations. First, it does not explicitly address spatial spillover effects, which may play a critical role in shaping the interregional dynamics of carbon emissions and policy diffusion. Incorporating spatial econometric approaches in future research would provide a more comprehensive understanding of these interactions.

Second, while green finance is treated as a unified construct, it encompasses heterogeneous instruments with potentially distinct environmental effects. Further disaggregation—examining the specific roles of green bonds, green credit, and carbon funds—would yield more precise policy recommendations.

Moreover, the analysis does not consider the broader institutional and regulatory landscape beyond green finance. Future studies should examine the interplay between digital development, green finance, and complementary policies such as carbon pricing, environmental regulation, and public support for clean technology adoption.

Lastly, incorporating additional moderating variables—such as regional economic development, industrial composition, energy mix, and cultural or institutional factors—could deepen our understanding of the conditional pathways through which digital and financial transformations affect environmental outcomes. These extensions would offer more nuanced and context-sensitive guidance for designing emission reduction strategies in diverse policy environments.