Impact of Digital Technology Application on the Development of Low-Carbon Economic Transition: The Mediating Role of Green Investment

Abstract

Against the backdrop of in-depth integration between the digital economy and green low-carbon development, exploring how digital technologies facilitate the systematic low-carbon transition of economy and society bears profound theoretical and practical implications for accomplishing China’s “Dual Carbon” goals. Based on provincial-level panel data covering 31 Chinese provinces over the period from 2015 to 2024, this paper adopts two-way fixed-effect specification, instrumental variable approach and Bootstrap-based mediation test to empirically identify the causal impact, underlying mechanisms and heterogeneous boundary conditions of digital technology adoption on low-carbon economic transition. The empirical results demonstrate three core findings. First, digital technology applications exert a statistically significant positive effect on low-carbon economic transition, and this benchmark result remains robust after a battery of robustness tests and endogenous bias corrections. Second, the mechanism estimation uncovers a sophisticated transmission pathway: digital technologies directly accelerate low-carbon transition, yet generate an adverse indirect impact via the green investment channel, which jointly forms a suppressing effect in the mediation framework. Third, the enabling effect of digital technologies on decarbonization presents striking regional imbalance, with significant promotional effects concentrated exclusively in eastern provinces and regions featuring well-developed marketization, which highlights the indispensable moderating role of regional endowments and institutional environments. This study contributes novel empirical evidence to unpack the intricate nexus between digital advancement and green transition, and delivers actionable policy references for designing differentiated and coordinated strategies to integrate digital upgrading with low-carbon development.

1. Introduction

Against the ongoing convergence of the digital economy and green low-carbon development, the pervasive adoption of digital technologies is reshaping the developmental paradigm of economies and societies. Since the start of the twenty-first century, digital innovations including big data, artificial intelligence, the Internet of Things and blockchain have penetrated economic and social spheres at an unprecedented scale and depth, fueling transformations in global productive forces and production relations. As a new type of production factor, data is triggering a profound technoeconomic paradigm shift, rendering the digital economy a pivotal pillar of contemporary economic development [1]. Beyond its conventional role as an efficiency-enhancing instrument, digital technology has evolved into a core enabler that systematically restructures industrial ecosystems, optimizes the allocation of factor resources and pioneers’ innovative modes of social governance [2]. China has maintained robust momentum amid the global boom of the digital economy. Benefiting from its enormous market size, well-established infrastructure networks and supportive policy frameworks, China has achieved remarkable progress in emerging digital infrastructure such as 5G base stations, large-scale data centers and industrial internet platforms, emerging as a key global player in the adoption and innovation of digital technologies. By 2024, China had deployed the world’s largest and most technically sophisticated 5G network, alongside steadily rising volumes of connected devices on industrial internet platforms, laying a solid foundation for sustained digital economic expansion. Nevertheless, the exponential proliferation of digital technologies, while facilitating socioeconomic advancement, comes with substantial energy consumption and carbon emission concerns. The high power usage of hyper-scale data centers, resource depletion in hardware manufacturing and the rapid accumulation of electronic waste collectively form the carbon footprint of digitalization, rendering green digital transition an urgent contemporary policy priority [3]. Promoting the low-carbon transition of economy and society represents a global consensus to tackle climate change, as well as an indispensable pathway for China to accomplish high-quality development [4]. The formulation of China’s carbon peaking and carbon neutrality goals marks a new phase of the country’s green and low-carbon development. Nevertheless, such a transition is confronted with multi-faceted obstacles. From a technological perspective, entrenched path dependence on conventional technologies leads to prohibitive switching costs. Financially, insufficient investment and prominent risks restrain green innovation. Institutionally, the high costs and inherent complexity of environmental governance hamper the effectiveness of relevant policies [5].

The existing literature mostly explores low-carbon transition pathways from macro dimensions including environmental regulation, energy mix restructuring and industrial upgrading, while systematic investigations into digital technology as an emerging determinant remain scarce. Although a growing body of the literature has examined the carbon abatement effects of the digital economy, most analyses concentrate merely on total or direct impacts without thorough insights into underlying transmission mechanisms. Core functional advantages of digital technologies create feasible solutions to dilemmas restricting green investment. Equipped with robust capabilities in information processing, interconnection and information verification, digital tools substantially mitigate information asymmetry and transaction costs across financing and investment activities [6]. Big data and artificial intelligence improve the precision of environmental risk assessment; the Internet of Things together with sensor technologies enables the real-time monitoring of energy consumption and pollutant emissions; blockchain strengthens the transparency and credibility of environmental information. Jointly, these digital instruments boost the allocation efficiency of green capital and channel more social capital toward low-carbon sectors [7].

Against such a backdrop, this paper takes green investment as the pivotal transmission channel and establishes a theoretical framework of “digital technology adoption–green investment optimization–low-carbon economic transition” to explore the underlying mechanisms through which digital technologies enable low-carbon shift. Specifically, three core research questions are addressed: What is the net impact of digital technology adoption on low-carbon transition? Does green investment serve as a mediating variable in such nexus? Are there evident regional disparities within the identified impacts? By systematically answering these questions, this study theoretically unpacks the internal channels linking digital technologies to decarbonization, empirically verifies such relationships via rigorous quantitative analysis, and practically delivers evidence for targeted policy design. This research combines theoretical deduction with empirical evaluation. After developing the theoretical framework to elaborate interconnections among core variables, we conduct quantitative estimations based on China’s provincial panel data spanning 2015 to 2024 by deploying fixed-effect specifications, mediation models and heterogeneous effect analysis. Our empirical focus falls on the direct effect of digital technologies on the low-carbon transition, the mediating role of green investment, as well as heterogeneous regional patterns across provinces.

This paper’s marginal contributions are summarized as follows. First, drawing on the logic of the suppressing effect, this study disentangles two contrasting transmission paths: digital technologies positively promote low-carbon transition directly, whereas they hinder such progress indirectly through the green investment channel, which supplements empirical evidence for the existing transmission logic of “digital technology–green investment–low-carbon transition”. Second, this research proposes region-targeted policy recommendations to help local authorities coordinate digital infrastructure development and green investment for accelerated low-carbon transformation.

2. Literature Review

2.1. Digital Technology Application and Low-Carbon Economic Transition: Direct Effects and Theoretical Foundations

Digital technology application is a multidimensional and composite concept encompassing digital infrastructure, digital industrialization, and industrial digitalization. Its core lies in utilizing new-generation information technologies such as big data, cloud computing, artificial intelligence, the Internet of Things, and blockchain to manage data as a key production factor throughout its entire lifecycle, thereby achieving production process optimization, business model reshaping, and governance innovation [8]. Existing research has systematically explored the profound impacts of digital technologies on economic and social development from macro, meso, and micro levels, particularly their critical role in promoting low-carbon economic transition.

At the macro level, digital technologies significantly enhance total factor productivity through information integration and intelligent decision-making. They not only optimize the allocation efficiency of traditional production factors but also exert a profound restructuring effect on industrial structure. Based on Schumpeterian endogenous growth theory, Tian and Li (2022) demonstrated that the deep integration of digital technologies and the real economy will promote long-term industrial structure optimization and upgrading, injecting new momentum into high-quality economic development by facilitating industrial integration and fostering new business forms and models [9]. From a comparative perspective of the three major industries, Xu and Dong (2024) further confirmed that digital technologies have a significant promoting effect on the low-carbon transition of the secondary and tertiary industries, mainly through mechanisms such as stimulating green technological innovation, expanding the scale effect of economic agglomeration, enhancing public environmental awareness, and accelerating human capital accumulation, while their impact on the primary industry is currently insignificant [10]. Studies focusing on specific regions have also found that the digital economy can significantly promote green and low-carbon development in key areas such as the Yangtze River Economic Belt, and this effect remains valid after a series of robustness tests including replacing the explained variable and adjusting the sample interval [11]. Cai et al. (2026) further pointed out that the synergistic effect between the digital economy and government governance capacity can amplify its low-carbon empowerment effect by more than three times, highlighting the importance of the institutional environment in enabling digital technologies to exert their effects [12]. Focusing on the agricultural sector, Xu et al. (2026) found that the digital economy can not only improve grain production efficiency but also significantly reduce carbon emissions from grain production through the mediating role of green finance, providing a new path for agricultural low-carbon transition [13]. Fan et al. (2024) confirmed that the digital economy significantly improves provincial-level carbon productivity in China by optimizing the energy consumption structure and promoting green technological innovation [14]. A study of E7 economies by Zheng et al. (2025) also found that while digital economy development may increase carbon emissions in the short term, it can achieve low-carbon transition in the long run by promoting green technological innovation [15].

At the meso–industry level, the low-carbon empowerment effect of digital technologies exhibits significant industry heterogeneity. Through bibliometric analysis, Valizadeh et al. (2026) found that the application of digital twin technology in energy management first began in 2019, with manufacturing being the earliest industry to adopt this technology, while construction has now become the most active application field [16]. Digital twins significantly improve energy efficiency and reduce carbon footprints by connecting virtual intelligence with physical systems [16]. Focusing on platform-based organizations, Lim et al. (2026) pointed out that despite enterprises’ huge investments in digital technologies such as artificial intelligence, big data, and blockchain, digital technologies themselves have no direct impact on platform circular economy performance [17]. Instead, they exert their effects through the governance mechanism of digital platform ecosystem orchestration, which is further strengthened by platform structural assurance [17]. A study of manufacturing clusters by Rahnama et al. (2026) showed that although digital transformation offers great potential for sustainable manufacturing development, enterprises still face obstacles such as stakeholder resistance, high initial costs, and fragmented data utilization [18]. Based on survey data of farmers in Jiangxi Province, Deng et al. (2024) found that the application of plant protection drones can reduce pesticide use intensity in rice production by 24.9%, significantly promoting the green and low-carbon transition of agriculture [19]. A study of ginseng growers in Jilin Province by Chen and Zhang (2026) further confirmed that digital literacy significantly promotes the adoption of green agricultural technologies by enhancing farmers’ multidimensional technological cognition [20]. Shen and Jiang (2026) demonstrated that digital technology innovation significantly enhances forestry economic resilience, providing a new path for the low-carbon transition of forestry, a traditional high-carbon sink industry [21]. Chi et al. (2023) comprehensively reviewed the current research status and future directions of the digital technology-driven circular economy, identifying big data, the Internet of Things, BIM, artificial intelligence, and digital twins as the most promising technologies [22].

At the micro–enterprise level, digital technologies demonstrate powerful empowerment effects on innovation activities and sustainable development performance. Ma and Gu (2025) pointed out that digital technologies effectively drive corporate green innovation through three mechanisms: data–labor synergy, data–capital synergy, and data–technology synergy [23]. Wang and Kang (2023) further showed that enterprise digital transformation simultaneously improves financial and environmental performance by promoting green product innovation, green process innovation, and green management innovation, providing strong support for enterprise sustainable development [24]. Based on the Stimulus–Organism–Response framework and resource orchestration perspective, Alabdali and Yaqub (2026) found that green digital transformational leadership significantly enhances supply chain resilience and sustainability by promoting the adoption of green disruptive technologies and fostering green digital congruence [25]. Using panel data of Chinese listed companies from 2011 to 2023, Zhao et al. (2026) found that climate risk significantly increases the level of corporate digital technology usage, with transition risk and acute physical risk having particularly pronounced effects [26]. Entrepreneurial spirit and dynamic capabilities strengthen this positive relationship, while unabsorbed slack resources weaken it [26].

Existing research has fully confirmed the positive direct impact of digital technology application on low-carbon economic transition from macro, meso, and micro dimensions. This impact manifests as industrial structure optimization, resource allocation efficiency improvement, and regional coordinated development at the macro level; differentiated low-carbon empowerment and circular economy practices across different industries at the meso level; and enhanced corporate green innovation capabilities, improved sustainable development performance, and increased climate risk response capabilities at the micro level. Meanwhile, the low-carbon effect of digital technologies is regulated by various factors such as the institutional environment, enterprise resources and capabilities, and regional development levels. However, the specific transmission paths of digital technologies’ low-carbon empowerment effect still need further in-depth exploration, and, especially, the mechanisms through which they indirectly affect low-carbon transition via intermediate variables have not yet been systematically elucidated.

2.2. The Core Role of Green Investment in Low-Carbon Economic Transition: Driving Factors and Economic Consequences

Green investment has both narrow and broad definitions in academic research. In the narrow sense, it specifically refers to capital directly invested in environmental improvement fields such as clean energy, pollution control, ecological restoration, energy-saving and environmental protection technologies, and infrastructure [27]. In the broad sense, its connotation has expanded to the ESG investment philosophy, which systematically incorporates environmental, social, and corporate governance factors into the entire investment decision-making process to achieve the dual goals of financial returns and positive environmental and social impacts [28]. As a key link connecting financial capital with green development, green investment is widely recognized as one of the core drivers of low-carbon economic transition.

At the macro level, green investment directly reduces carbon emissions and improves energy efficiency through large-scale capital investment supporting renewable energy development, the green transformation of traditional industries, and low-carbon infrastructure construction. Ren et al. (2022) showed that green investment significantly inhibits environmental pollution, but its effect depends not only on investment scale but also critically on the quality of the institutional environment [29]. Using panel data from 293 Chinese cities, Tang et al. (2025) found that the synergistic effect of green finance and the digital economy can significantly promote urban low-carbon transition through mechanisms including promoting green technological innovation, industrial upgrading, and energy efficiency improvement [30]. Moreover, this synergistic effect has a positive spatial spillover effect that can drive low-carbon development in surrounding areas [30]. Wu et al. (2025) further revealed a U-shaped relationship between green finance and carbon efficiency, and found that the stage of digital economy development has a significant threshold effect on this relationship [31]. Only when digital economy development reaches a certain level can the promoting effect of green finance on carbon efficiency be fully exerted [31].

At the micro–enterprise level, green investment can effectively improve corporate environmental performance, thereby enhancing their ESG ratings and market valuations. Using an ordered Probit model, Zhang et al. (2025) found that green investment significantly improves corporate ESG ratings, and this positive effect is strengthened by government subsidies, technological innovation, media supervision, and environmental information disclosure [32]. Heterogeneity analysis showed that green investment has a more significant effect on improving environmental performance for state-owned enterprises, large enterprises, and enterprises with weaker financing constraints [32]. Based on Italian enterprises, Quatraro and Ricci (2025) further revealed the heterogeneous impacts of green investment, finding that overall green investment increases firms’ sales per capita and average wages, but the effects vary significantly across different types of green investment, with circular economy and energy efficiency investments showing the most prominent economic performance [33]. Using data from 4375 non-financial enterprises in 74 countries, Rabbani et al. (2025) found that green investment significantly reduces firms’ climate risk exposure, and good ESG performance further amplifies this effect, enhancing firms’ resilience to regulatory and environmental shocks [34].

Regarding the driving factors of green investment, existing research mainly expands from two dimensions: external policies and internal capabilities. Externally, strict environmental regulations and clear low-carbon policy signals are the fundamental forces stimulating green investment demand, providing a stable policy framework for enterprises and investors by setting clear environmental standards and transition expectations [35]. Systematic green finance policies, such as central bank structural monetary policy tools and mandatory environmental information disclosure requirements, directly incentivize green capital supply by reducing financing costs and improving the risk-return structure of green projects [36]. Internally, enterprises’ technological innovation capabilities and digitalization levels also have important impacts on green investment decisions. Meanwhile, investors’ and consumers’ green preferences have gradually become important market forces driving corporate green investment [37].

Green investment plays an irreplaceable core role in low-carbon economic transition. It is both a bridge connecting financial resources with green projects and an engine driving technological innovation and industrial upgrading. Existing research has thoroughly explored the economic consequences and driving factors of green investment, confirmed its emission reduction effect at the macro level and performance improvement effect at the micro level, and revealed the multiple influences of policy, market, and internal enterprise factors on green investment. Meanwhile, studies have found that the effect of green investment exhibits significant heterogeneity, regulated by various factors such as the stage of digital economy development, institutional environment, and enterprise characteristics. However, research on the mechanisms through which digital technologies affect the low-carbon transition by influencing green investment is still relatively scarce, failing to place green investment within the complete logical chain of digital technologies and low-carbon transition.

2.3. Digital Technology Application and Green Investment: Correlation Mechanisms and Research Progress

With the rapid development of the digital economy, the relationship between digital technologies and green investment has received increasing academic attention. Existing research has initially revealed the multidimensional impact mechanisms of digital technologies on green investment, which together constitute the theoretical foundation for digital technologies empowering green investment.

First, digital technologies can effectively alleviate information asymmetry in green investment. Green investment is information-intensive, and the environmental benefits and investment risks of projects are often difficult to accurately assess, leading to prominent adverse selection and moral hazard problems. Through big data analysis and artificial intelligence algorithms, investors can obtain more comprehensive and accurate corporate environmental information and evaluate the true value and risks of green projects, thereby improving the scientific nature of investment decisions. Xie et al. (2026) showed that the development of banking fintech significantly improves the efficiency of corporate green investment, with the core mechanisms being alleviating financing constraints and agency problems [38]. Banks can more accurately identify high-quality green projects and reduce credit risks by applying technologies such as big data and artificial intelligence [38]. Zhu and Li (2026) also found that artificial intelligence development enhances market incentives for corporate green investment by increasing investor and media attention, thereby expanding the scale of green investment [39]. However, Ahmed et al. (2025) also pointed out that higher levels of digital technology disclosure may increase investors’ risk perception, leading to higher capital costs, and ESG performance plays a mediating role in this relationship [40]. This indicates that the impact of digital technologies on green investment is complex: they may promote investment by alleviating information asymmetry, but may also have an inhibitory effect by increasing risk perception.

Second, digital technologies optimize the efficiency of green resource allocation. Digital platforms can break geographical and industry barriers, achieve precise matching between green capital and high-quality projects, and improve capital utilization efficiency. Yang et al. (2025) found that digital economy development significantly expands the scale of corporate green investment by optimizing resource allocation and alleviating information asymmetry [41]. Using a multi-period difference-in-differences model, Jia et al. (2026) found that enterprise digital platforms significantly promote green technology co-innovation by strengthening consumer preferences, improving resource allocation efficiency, and reducing cooperation barriers, and the improvement of green technology innovation further drives the increase in corporate green investment [37]. Tang et al. (2025) further showed that the synergistic effect of green finance and the digital economy can jointly promote the urban low-carbon transition by promoting green technological innovation, industrial upgrading, and energy efficiency improvement [30]. Wu et al. (2025) found that the stage of digital economy development has a significant threshold effect on the relationship between green finance and carbon efficiency [31]. Only when digital economy development reaches a certain level can the resource allocation function of green finance be fully exerted [31]. In addition, digital technologies can broaden green financing channels by promoting green financial product innovation. For example, blockchain technology enables the traceability and verifiability of green assets, providing technical support for the development of innovative products such as green bonds and carbon financial derivatives.

Third, digital technologies indirectly promote green investment by enhancing corporate green innovation capabilities. Green innovation is an important prerequisite for green investment; only with advanced green technologies can enterprises carry out effective green investment activities. Digital technologies provide powerful tools and platforms for corporate green innovation, accelerating the research, development, and application of green technologies. Zhu and Li (2026) confirmed that artificial intelligence development significantly enhances the level of green investment in heavily polluting industries through mechanisms including reducing pollution emissions, alleviating financing constraints, and increasing investor and media attention [39]. A study of OECD countries by Kiran et al. (2026) also found that both green technological innovation and the green energy transition have significant promoting effects on digital economy development, forming a virtuous cycle of “digital technology–green innovation–green investment” [42]. Meanwhile, the integrated innovation of digital technologies and green finance has also provided new tools and channels for green investment, such as green digital credit and carbon financial derivatives, further broadening green financing sources.

However, some studies have pointed out that the impact of digital technologies on green investment exhibits heterogeneity. For example, in regions with imperfect digital infrastructure and low digital literacy, the green investment promotion effect of digital technologies may not be fully exerted. In addition, the application of digital technologies itself may bring new environmental problems, such as the high energy consumption of data centers, which need attention during development.

Existing research has initially confirmed the positive impact of digital technologies on green investment, with mechanisms including alleviating information asymmetry, optimizing resource allocation, and enhancing green innovation capabilities. Meanwhile, the impact of digital technologies on green investment exhibits significant heterogeneity, regulated by various factors such as regional digital development levels, enterprise characteristics, and institutional environments. However, most of these studies remain at the level of exploring the direct relationship between the two, failing to systematically analyze green investment within the complete chain of digital technologies and low-carbon economic transition.

The above literature reveals that although academia has conducted extensive research on the three pairwise relationships between digital technology, green investment, and the low-carbon transition across multiple levels and contexts, three critical research gaps remain. First, an integrated analytical framework is missing. Most studies examine digital technology or green investment’s impacts on the low-carbon transition separately, failing to systematically analyze all three in a unified model or rigorously reveal the complete action chain from digital technology to low-carbon transition via green investment. Second, green investment’s mediating role remains under-tested. Few studies have rigorously verified this transmission mechanism. Third, regional heterogeneity is under-explored. Existing research fails to adequately explain the significant cross-regional variations in digital technologies’ low-carbon empowerment effects, and lacks empirical evidence for formulating differentiated digital and green development policies tailored to local conditions.

This study makes three key marginal contributions. First, based on existing theoretical paradigms, this paper disentangles the coexisting direct and indirect effects of digital technologies on the low-carbon transition with green investment acting as the mediator, thus supplementing empirical evidence for relevant studies at the intersection of digital economy and low-carbon development. Second, relying on up-to-date provincial panel data and multiple econometric specifications, this paper quantifies the overall influence of digital technologies and explores regional heterogeneity, offering empirical support for the divergent cross-region outcomes. Third, the empirical results facilitate differentiated policy design: eastern provinces can further integrate digital advancement with green investment, whereas central and western regions need to prioritize digital infrastructure construction. The conclusions also help enterprises formulate coordinated digital–green strategies to lift the sustainability of green investment.

3. Research Hypotheses

Based on innovation diffusion theory, transaction cost theory and information economics, the impacts of digital technology adoption on the low-carbon economic transition unfold along three dimensions: efficiency improvement, structural optimization and governance upgrading. First, regarding efficiency improvement, according to innovation diffusion theory, the popularization of new technologies breaks the shackles of original production technical paradigms and optimizes factor allocation. Relying on IoT sensing, big data analytics and artificial intelligence, digital technologies enable the real-time monitoring and precise regulation of energy consumption, material input and pollutant emissions [43]. Within industrial Internet and smart grid systems, algorithms dynamically streamline production arrangements and energy dispatching, improve resource efficiency and cut carbon emission intensity per unit output at the source to achieve technological emission reduction [44]. Second, from the perspective of structural optimization, according to innovation diffusion theory, traditional industrial systems are constrained by technical barriers and locked industrial boundaries, hindering the spontaneous emergence and evolution of new low-carbon business formats. As a general-purpose technology, digital technology penetrates across industries and fosters innovative low-carbon forms and business models [45]. Meanwhile, based on transaction cost theory, digitalization alleviates market information asymmetry, cuts the costs of factor search and contract conclusion, activates idle stock resources, boosts the development of sharing and platform economies, and reduces idled physical assets and overproduction [46]. Digital industrialization enables low-energy, high-value-added knowledge-intensive industries to replace high-energy-consuming sectors, while industrial digitalization forces traditional high-carbon industries toward intelligent and service-oriented transformation. Jointly, they shift the economic structure to a leaner, low-carbon model dominated by knowledge and services, and deliver structural emission reduction [47]. Finally, in terms of governance transformation, information economics suggests that distorted environmental information and information asymmetry constitute the core drivers of ineffective environmental governance and obstacles to the low-carbon transition. Technologies such as blockchain and big data improve the credibility, transparency and traceability of environmental data, laying a technical foundation for establishing more efficient green financial systems, environmental regulations and market trading mechanisms [48]. Blockchain guarantees the authenticity of carbon asset transactions, and big data improves the quality of corporate environmental disclosure and environmental risk pricing. Accordingly, financial capital is guided into green projects more precisely, strengthening market and policy incentives as well as restraints for low-carbon behaviors to realize governance-driven emission reduction [49]. In summary, digital technology application can systematically reduce the carbon emission intensity of economic and social activities by improving energy and resource efficiency, optimizing the industrial and economic structure, and enabling the green governance system, providing new pathways to overcome the cost and technical constraints of low-carbon transition. It is worth noting that digital technologies are general-purpose and neutral in nature, whose green benefits depend on specific application contexts. In the short run, without sound institutional regulations capital misallocation and the crowding-out of green investment may emerge as partial adverse disturbances, so their low-carbon dividends cannot follow a perfectly linear path. This highlights the necessity of establishing institutional arrangements to steer digital development toward a green economy. Nevertheless, such temporary and partial adverse impacts are dominated by the overall long-run positive effects of digitalization on the low-carbon transition, leading to an aggregate significantly positive outcome. Therefore, this paper proposes the core research hypothesis H1:

H1. 

Digital technology application is significantly correlated with advances in low-carbon economic transition.

Digital technology adoption affects low-carbon economic transition via green investment, yet such transmission manifests a complex mix of promotional and inhibitory effects. On the one hand, its underlying mechanisms operate through three core channels: lowering investment thresholds, optimizing capital allocation and strengthening risk management. First, digital technologies mitigate information asymmetry and substantially lower the thresholds for identifying and evaluating green investment projects. Big data and artificial intelligence enable multidimensional and dynamic assessments of corporate environmental performance and abatement potential, while blockchain guarantees the immutability and traceability of environmental and carbon emission data [50]. Collectively, these improvements raise the transparency and credibility of information on green projects, enabling investors to accurately identify genuine green assets and channel more capital into this sector [51]. Second, digital platforms and fintech optimize the allocation efficiency of green capital. From the perspective of transaction cost theory, cumbersome formalities and excessive contracting costs in the matching of traditional green investment supply and demand impede the efficient flow of capital. Digitalized green financial product trading platforms (e.g., green asset digital trading centers) efficiently match capital supply and demand [52]; big-data-based credit evaluation models can more accurately price the risk of green SMEs, broadening their financing channels; and smart contracts and other technologies can automate the execution of investment and financing terms linked to green performance, ensuring that capital flow and use meet predetermined environmental goals [53]. Third, digital technology enhances the risk management and pricing capabilities of green investments. From information economics theory, incomplete information prevents outside investors from accurately identifying and quantifying the ecological benefits and potential risks of green projects, thereby hindering the development and implementation of green financial products. By using IoT, remote sensing, and other technologies for the real-time environmental benefit monitoring of green projects (e.g., wind power, forestry carbon sinks), combined with climate risk models, investors can more quantitatively assess project physical and transition risks [54], thereby developing richer green financial derivatives to hedge risks, increasing market depth and liquidity. Thus, digital technology does not directly replace capital but rather fundamentally improves the market failures faced by green investments by reshaping information structures, transaction models, and risk management paradigms [55]. It unlocks green capital demand previously suppressed by information opacity, high evaluation costs, and risk uncertainty [56], activates supply, and thereby leverages, guides, and expands the scale and efficiency of society’s capital investment in low-carbon technologies and projects. Green investment, as the core hub converting financial resources into physical green assets and technologies, directly accelerates the pace of clean energy substitution, energy-efficient technological innovation, and low-carbon infrastructure construction through its increased scale and improved efficiency, ultimately powerfully driving the overall economic low-carbon transition. Therefore, green investment plays an indispensable transmission channel role between digital technology and low-carbon transition.

On the other hand, digital technologies may also exert inhibitory impacts on green investment from three dimensions. First is the capital crowding-out effect. In line with innovation diffusion theory, newly introduced digital technologies deliver relatively high returns in their initial deployment stage. Characterized by capital intensity and high profitability, booming digital industries draw substantial financial resources toward core digital sectors including internet platforms, artificial intelligence and data centers, crowding out long-term investment in green projects featuring long payback cycles and relatively lower returns such as clean energy and energy-saving retrofitting. The second dimension refers to the efficiency substitution effect. By boosting energy efficiency and resource allocation efficiency in conventional industries, digital technologies cut the energy consumption and emission intensity per unit GDP, which reduces the marginal demand for incremental investment in end-of-pipe pollution abatement and emission-reduction facilities and slows the expansion of overall green investment. The third dimension is the path dependence and lock-in effect. Innovation diffusion theory highlights inherent path dependence during technology diffusion. In regions dominated by high-carbon traditional industries, digital tools are preferentially deployed to upgrade existing high-carbon production procedures such as intelligent mining and smart petrochemical processing. Such practice consolidates technological and capital lock-in for high-carbon sectors in the short run rather than diverting funds toward thorough green transition. Moreover, without matched regulatory improvements, fast-expanding digital finance may facilitate capital inflows into pseudo-green or carbon-intensive projects and aggravate the misallocation of green capital.

Further analysis reveals that green investment does not always deliver the expected emission reduction outcomes [57]. Its efficacy in facilitating low-carbon transition largely depends on capital allocation directions and operational efficiency [58]. In practice, such investment is disproportionately concentrated in fields that yield quick results, such as pollution abatement and energy conservation retrofitting, while long-term low-carbon initiatives including clean technology innovation and energy restructuring receive insufficient funding [59]. This tendency merely achieves partial emission reductions and thereby hinders the sustained advancement of low-carbon transition. Meanwhile, under information asymmetry and imperfect institutional constraints, firms tend to conduct symbolic environmental investments to obtain policy incentives and financing advantages [60]. They launch projects that formally meet green standards yet generate limited real emission reductions. In this case, green investment deviates from its original objectives and leads to deteriorated resource allocation efficiency. Furthermore, continuous investment locked into existing technological and industrial pathways may reinforce the conventional development model, creating barriers for emerging low-carbon technologies with great decarbonization potential and slowing down the overall low-carbon transition.

Accordingly, the mediating effect of green investment between digital technology and low-carbon transition is not unidirectional. On the one hand, green investment serves as a vital transmission channel to accelerate low-carbon transformation. On the other hand, distorted investment structures and low allocation efficiency may weaken its positive effects or even exert inhibitory impacts on low-carbon transition. Accordingly, this study proposes Hypothesis 2:

H2. 

Green investment plays a mediating role in the impact of digital technology adoption on the low-carbon economic transition.

4. Research Design

4.1. Benchmark Model

This paper aims to test the impact of digital technology applications on the development of the low-carbon economic transition and its mediating mechanism. To this end, the following panel data models are employed for empirical analysis. To test research hypothesis H1 (digital technology application has a direct promoting effect on low-carbon economic transition development), this paper first constructs a two-way fixed effects model. Its basic form is as follows:

$${\mathit{LCTD}}_{\mathit{it}}={\alpha }_0+{\alpha }_1{\mathit{DIG}}_{\mathit{it}}+\alpha X_{\mathit{it}}+{\mu }_{\mathit{it}}+{\delta }_{\mathit{it}}+{\epsilon }_{\mathit{it}}$$

(1)

where i and t represent province and year, respectively. The dependent variable ${\mathit{LCTD}}_{\mathit{it}}$ is the index of low-carbon economic transition development level; the core independent variable ${\mathit{DIG}}_{\mathit{it}}$ is the digital technology application level; $X_{\mathit{it}}$ represents a set of control variables that may affect low-carbon transition; ${\mu }_{\mathit{it}}$ represents province fixed effects, controlling for time-invariant regional heterogeneity; ${\delta }_{\mathit{it}}$ represents year fixed effects, controlling for time trends and macro shocks; and ${\epsilon }_{\mathit{it}}$ is the random error term.

To test research hypothesis H2 (green investment plays a mediating role in the process of digital technology applications affecting the low-carbon economic transition), this paper constructs the following mediation effect models, drawing on the stepwise testing method of Wen et al. [52]:

$${\mathit{GI}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{DIG}}_{\mathit{it}}+\beta X_{\mathit{it}}+{\mu }_{\mathit{it}}+{\delta }_{\mathit{it}}+{\epsilon }_{\mathit{it}}$$

(2)

$${\mathit{LCTD}}_{\mathit{it}}={\beta }_0+{\beta }_1{\mathit{DIG}}_{\mathit{it}}+{\beta }_2{\mathit{GI}}_{\mathit{it}}+\beta X_{\mathit{it}}+{\mu }_{\mathit{it}}+{\delta }_{\mathit{it}}+{\epsilon }_{\mathit{it}}$$

(3)

where ${\mathit{GI}}_{\mathit{it}}$ is the mediating variable, representing the green investment level.

4.2. Variable Definition and Measurement

4.2.1. Independent Variable: Digital Technology Application (DIG)

Digital technology adoption constitutes a multidimensional construct encompassing infrastructure construction, technological integration and socioeconomic empowerment. To systematically quantify its development level, this paper follows the statistical framework specified in the Statistical Classification of Digital Economy and Its Core Industries (2021) issued by China’s National Bureau of Statistics and refers to established empirical practices from the authoritative literature [53,54]. A provincial-level comprehensive evaluation indicator system for digital technology adoption is established across three dimensions: digital infrastructure, digital industrialization and industrial digitalization. Theoretically, the development of digital technology is reflected not only in the popularization and access of hardware infrastructure, but also in the output of digital industrial forms and the digital empowerment of traditional industries. Specifically, indicators such as the number of web pages and registered domain names characterize the agglomeration of regional digital network resources and the penetration of internet access, serving as typical proxies for the spatial coverage and technological penetration of digital infrastructure. The express delivery volume and output value of digital industries capture the commercial application and activity of digital technology, which reflect the scale of digital industrialization. In addition, digital investment in traditional industries and the output generated by online integration measure the depth of integration between digital technology and the real economy [61]. The above indicators collectively form a complete transmission chain of digital technology, namely infrastructure access–industrial output–real economy empowerment, which conforms to the theoretical hierarchy of digital economy development. Therefore, the selection of indicators is fully supported by economic theories and mechanisms. Detailed indicator definitions are listed in

Table 1. Comprehensive evaluation indicator system for digital technology application.

Primary Indicator	Secondary Indicator	Specific Definition/Calculation Explanation	Weight
Digital Infrastructure	Internet Broadband Access Rate	Number of regional internet broadband access ports/Permanent resident population	0.0622
	Internet Broadband Penetration Rate	Number of regional internet broadband access users/Permanent resident population	0.0585
	Mobile Telephone Exchange Capacity	Mobile telephone exchange capacity (lines)	0.0531
	Long-distance Optical Cable Line Length	Length of long-distance optical cable lines (km)	0.0497
	Number of Web Pages	Number of web pages	0.0556
	Number of Domain Names	Number of domain names	0.0526
Digital Industrialization	Per Capita Total Telecom Business	Total regional telecom business/Permanent resident population	0.0572
	Mobile Telephone Penetration Rate	Number of mobile telephone users per hundred people (subscribers/100 people)	0.0483
	Number of Corporate Units in Information Transmission, Software, and IT Services	Number of corporate units in this industry	0.0568
	Share of Employment in Information Software Industry	Employment in information transmission, software, and IT services/Total urban unit employment	0.0496
	Number of Domestic Patents Granted	Number of domestic patents granted	0.0597
	Number of Domestic Patent Applications Received	Number of domestic patent applications received	0.0558
Industrial Digitalization	Digital Inclusive Finance Index	Peking University Digital Inclusive Finance Index	0.0589
	Proportion of Enterprises with E-commerce Transaction Activities	Proportion of enterprises with e-commerce transaction activities (%)	0.0475
	E-commerce Sales	Sales revenue achieved by enterprises through e-commerce (10,000 RMB)	0.0618
	Number of Websites per Hundred Enterprises	Number of websites owned per hundred enterprises	0.0452
	Value Added of Secondary and Tertiary Industries	Sum of value added of secondary and tertiary industries (100 million RMB)	0.0533
	Science and Technology Innovation Investment	Internal expenditure on R&D by industrial enterprises above designated size (10,000 RMB)	0.0594
	Express Delivery Volume	Express delivery volume (10,000 items)	0.0608

. The entropy weight method is adopted for composite index calculation. Raw data are first standardized via range normalization, followed by the successive computation of information entropy, difference coefficients and objective weights for each indicator. Provincial annual composite scores of digital technology adoption are finally obtained through weighted summation. Since indicator weights are endogenously determined by the discrete characteristics of raw data, this method effectively eliminates biases induced by subjective weighting. It should be noted that the entropy weight method assigns weights according to the discrete differences in raw data. It automatically imposes lower weights on highly collinear indicators and higher weights on core indicators with abundant differentiated information. Therefore, this method inherently mitigates the interference of multicollinearity, making the prior elimination of collinear variables unnecessary [62]. Meanwhile, the indicator system adopted in this study strictly follows the official national statistical framework and well-established paradigms in the existing literature. It features clear hierarchical dimensions and definite theoretical boundaries. Its structural stability and reliability have been validated in numerous empirical studies, which effectively avoids biases arising from arbitrary indicator selection and index synthesis.

4.2.2. Dependent Variable: Low-Carbon Economic Transition Development (LCTD)

Low-carbon economic transition represents a systematic evolutionary process featuring coordinated advancement across economic, social, environmental and technological dimensions. Conceptually, it differs from carbon abatement and green development: carbon abatement centers merely on the single environmental outcome of aggregate carbon emission reduction, while green development covers extensive fields ranging from ecological governance and social welfare to economic sustainability. By contrast, the low-carbon economic transition defined in this study refers to the comprehensive transformation of economic systems toward low-carbon operation driven by energy restructuring, industrial upgrading and policy regulation. To comprehensively and objectively quantify regional low-carbon economic transition performance, this study constructs a composite evaluation system rooted in the Pressure–State–Response (PSR) framework covering three dimensional clusters: low-carbon output (Pressure), low-carbon consumption and resources (State), as well as low-carbon policies and environmental governance (Response). The finalized indicator system consists of 14 fundamental indicators, as presented in

Table 2. Indicator system for low-carbon economic transition development.

Primary Indicator	Secondary Indicator	Specific Definition	Attribute	Weight
Low-Carbon Output	Carbon Productivity	GDP/Carbon emissions	Positive	0.1215
	Energy Processing and Conversion Efficiency	Energy output/Energy input	Positive	0.0971
Low-Carbon Consumption	Household Consumption Carbon Emissions	Carbon emissions/Household consumption expenditure	Negative	0.0996
	Government Consumption Carbon Emissions	Carbon emissions/Government consumption expenditure	Negative	0.0796
Low-Carbon Resources	Share of Zero-Carbon Energy	Zero-carbon energy consumption/Total energy consumption	Positive	0.0928
	Energy Carbon Emission Coefficient	Carbon emissions/Total energy consumption	Negative	0.0785
	Carbon Sink Density	Carbon sink volume/Administrative area	Positive	0.0684
Low-Carbon Policy	Low-Carbon Economy Development Plan	Whether a specific plan is formulated (Yes = 1, No = 0)	Positive	0.0427
	Carbon Emission Monitoring, Statistics, and Supervision System	Whether established (Yes = 1, No = 0)	Positive	0.0492
	Public Awareness of Low-Carbon Economy Knowledge	Awareness rate or questionnaire score	Positive	0.0589
	Environmental Protection and Energy Saving Standard Compliance Rate	Proportion of compliant enterprises or verification pass rate	Positive	0.0563
	Carbon Tax/Carbon Trading Policy	Whether implemented (Yes = 1, No = 0)	Positive	0.0450
Low-Carbon Environment	Waste Carbon Emission Intensity	Waste carbon emissions/Waste generation	Negative	0.0591
	Industrial “Three Wastes” Treatment Index	(Wastewater treatment rate + Solid waste disposal rate + Waste gas treatment rate)/3	Positive	0.0513

. The selection of indicators mainly refers to official documents issued by the National Development and Reform Commission and the Ministry of Ecology and Environment of China, including the Low-carbon Development Indicator System, Annual Report on Citizens’ Ecological and Environmental Behavior Survey and China Urban Low-carbon Development Evaluation Report, together with authoritative existing academic literature [55,63,64], guaranteeing the representativeness and academic recognition of the selected indicators. For policy-related indicators such as low-carbon development plans and carbon trading pilots, this study adopts 0–1 binary dummy variables for assignment. This approach is a widely accepted classic method for quantifying macroeconomic policies [65], and its adoption is based on three major considerations. First, in terms of indicator connotation, the official issuance of provincial low-carbon development plans and inclusion in the carbon trading pilot program represent a fundamental transition in regional low-carbon governance frameworks. Whether a policy is implemented serves as a core criterion to distinguish regional low-carbon responses, with clear and objective identification rules. Second, regarding data availability, there are no uniformly standardized micro-level statistics on the implementation depth, enforcement intensity and progress of low-carbon policies across Chinese provinces. In addition, policy provisions and assessment criteria vary substantially across regions. Subjective scoring or graded assignment would inevitably introduce artificial bias and violate the principle of objective measurement. Third, numerous existing studies have confirmed that policy implementation status can effectively capture fundamental disparities in low-carbon governance across regions, which makes it suitable for constructing long-term comprehensive indices at the provincial level. Given divergent measurement units and unknown preliminary weights across sub-indicators, the entropy weight method is employed for objective weighting to eliminate biases originating from subjective assignment. This method determines indicator dispersion degrees by calculating corresponding information entropy; indicators with higher dispersion (namely those delivering richer informative content) are assigned larger weights, rendering the approach particularly suitable for evaluating the composite development of sophisticated multidimensional systems. Compared with PCA and factor analysis, which tend to compress the economic connotation of original indicators and lose the practical meaning of sub-dimensions, the entropy weighting method can fully retain the economic attributes of each indicator and better fit the construction of our multidimensional composite index.

4.2.3. Mediating Variable: Green Investment (GI)

Synthesizing existing studies [56,66] to more comprehensively measure the practical scope of green investment under the new development concept, this paper constructs a multi-indicator provincial green investment index. It should be noted that green investment is defined broadly and narrowly in the existing literature. In the broad sense, it covers green financial investments within the ESG framework, including green credit, green bonds and green funds, as well as physical green investment [28]. In the narrow sense, it specifically refers to physical fixed asset investment targeting ecological governance, pollution abatement and the upgrading of green industries [27]. In line with the research logic of this study, the core pathway through which the integration of digital and real economies empowers low-carbon economic transformation lies in optimizing resource allocation in the real economy and encouraging firms and governments to conduct environmental governance and green production investment. This study focuses on physical green investment activities rather than green capital financing in financial markets. Accordingly, this study adopts the narrow definition of physical green investment for empirical analysis. The index includes the following three key dimensions: Environmental Protection Investment, including industrial pollution control investment and urban environmental infrastructure construction investment, representing the traditional core areas of green investment; Ecological Protection and Restoration Investment, incorporating forestry investment and water conservancy construction investment, reflecting investment in natural capital and maintenance of ecosystem services; and Green Industry Investment, using fixed asset investment in the energy conservation and environmental protection industry as a proxy variable to capture investment in emerging strategic industries such as green technology and clean energy. For the measurement method, the three types of investment (at current year prices) are first summed to obtain the nominal total green investment for the year. Second, to eliminate price fluctuation effects, the total is deflated to a base-year comparable price total using the fixed asset investment price index. Lastly, to mitigate heteroskedasticity stemming from scale discrepancies and improve the normality of dataset distribution, this study adopts the natural logarithm of deflated real total green investment as the proxy for green investment (GI).

4.2.4. Control Variables

To control for other key factors that may affect low-carbon economic transition and improve the accuracy of model estimation, drawing on relevant research [67,68], this paper selects the following five control variables. (1) The economic development level, proxied by the natural logarithm of regional gross domestic product (lnGDP), controls the stage-dependent impacts of economic scale on energy demand and technology adoption. (2) Industrial structure, measured as the ratio of tertiary industry-added value to secondary industry-added value (IND2), reflects the relative scale advantage of the service sector versus manufacturing within regional economies. (3) The regional electricity consumption share, denoted by the proportion of provincial electricity consumption in the national total electricity usage (COAL), captures each province’s relative position in the nationwide electricity consumption layout; a larger value generally implies a higher reliance of local economic activities on electricity or a greater concentration of power-intensive industries, which may shape the local trajectory of low-carbon transition. (4) Environmental regulation stringency, expressed as the ratio of industrial pollution abatement investment to regional GDP (ER), identifies the direct effect of governmental environmental governance on economic transition. (5) The urbanization level, measured by the share of urban population in total resident population (URBAN), accounts for systematic influences originating from population agglomeration, shifts in lifestyle and infrastructure construction demands.

4.3. Data Source

This paper uses panel data from 31 provinces, autonomous regions, and municipalities directly under the central government in China (excluding Hong Kong, Macau, and Taiwan) from 2015 to 2024 as the research sample. Considering the availability and comparability of indicator data, the data for the selected variables are mainly sourced from the China Statistical Yearbook, China Energy Statistical Yearbook, China Environmental Statistical Yearbook, China Information and Communication Statistical Yearbook, China Financial Yearbook, the Peking University Digital Inclusive Finance Index, and the statistical yearbooks of various provinces (autonomous regions, municipalities). To address missing data issues, this study uses linear interpolation to reasonably supplement missing values for individual years.

4.4. Descriptive Statistics

Table 3. Descriptive statistics results.

VarName	Obs	Mean	SD	Min	Median	Max
lctd	310	0.521	0.291	0.083	0.547	1.065
dig	310	0.157	0.121	0.033	0.119	0.747
gi	310	3.700	1.137	1.972	3.405	8.164
er	310	0.003	0.003	0.000	0.002	0.031
lngdp	310	9.326	0.465	8.660	9.187	10.807
ind2	310	1.423	0.746	0.665	1.268	5.283
urban	310	0.603	0.125	0.239	0.597	0.938
coal	310	0.032	0.023	0.001	0.026	0.094

presents descriptive statistics for the core variables employed in this study. For the key explained variable, low-carbon economic transition development (LCTD) has a mean value of 0.521, a standard deviation of 0.291, with minimum and maximum values of 0.083 and 1.065, respectively, implying substantial cross-provincial disparities in low-carbon transition progress. The core explanatory variable digital technology adoption (DIG) yields a mean of 0.157 and a standard deviation of 0.121, ranging from 0.033 to 0.747, which reflects pronounced regional unevenness in the advancement of digital technologies. Regarding the mediating variable green investment (GI), its average value stands at 3.700 alongside a standard deviation of 1.137, spanning between 1.972 and 8.164, indicative of considerable provincial divergence in green investment intensity. In terms of control variables, the standard deviation of economic development level (lnGDP) is 0.465 with a value range of [8.660, 10.807], consistent with the actual distribution of provincial economic output ranging from hundreds of billions to several trillion RMB and revealing prominent regional economic gaps. Industrial structure (IND2) has a mean of 1.423 and a standard deviation of 0.746, varying between 0.665 and 5.283, which suggests provinces are located at disparate developmental stages shifting from industry-dominated to service-oriented economies and exhibit substantial structural heterogeneity across regions. The regional electricity consumption share (COAL) averages at 0.032 with a standard deviation of 0.023 and an interval of [0.001, 0.094], implying divergent provincial standings in the national electricity consumption landscape and uneven spatial patterns of regional electricity reliance and energy consumption features. Environmental regulation stringency (ER) yields a mean and standard deviation both equal to 0.003, demonstrating that environmental governance investment remains modest across the full sample period with limited cross-provincial volatility. Urbanization level (URBAN) has a mean value of 0.603 and a standard deviation of 0.125, ranging from 0.239 to 0.938, covering the full spectrum from incipient urbanization to highly urbanized economies and well capturing unbalanced urbanization progress across China’s provinces. All variables are logarithmically transformed or standardized prior to regression to improve data stationarity and strengthen the comparability and robustness of empirical estimations.

5. Empirical Results Analysis

5.1. Multicollinearity Test

To detect potential multicollinearity within the multivariate regression specification and prevent biased coefficient estimates induced by high pairwise correlations among explanatory variables, the variance inflation factor (VIF) test is implemented across all regressors. As a standard diagnostic metric for linear dependence among independent variables, larger VIF values signal more severe multicollinearity. A common empirical cutoff of ten is conventionally adopted; values exceeding this threshold indicate severe multicollinearity that necessitates model or variable revision. Corresponding test statistics are summarized in

Table 4. Multicollinearity test results.

Variable	VIF	1/VIF
dig	3.80	0.26
er	1.23	0.81
lngdp	4.60	0.22
ind2	2.52	0.40
urban	3.18	0.31
coal	3.04	0.33
Mean VIF	3.06

. All explanatory variables yield VIF values well below the critical threshold of 10, with the maximum value of 4.60 corresponding to economic development level (lnGDP) and the minimum of 1.23 for environmental regulation stringency (ER), alongside an average VIF of 3.06 for the full set of regressors. These diagnostic outcomes confirm that multicollinearity remains within an acceptable range and does not materially impair the stability or validity of parameter estimation. Accordingly, the baseline empirical specification is statistically valid, and the ensuing regression results enjoy satisfactory statistical reliability.

5.2. Benchmark Regression Analysis

This paper uses a two-way fixed effects model controlling for both individual and time effects for benchmark regression analysis to test the direct impact of digital technology applications on low-carbon economic transition developments.

Table 5. Benchmark regression results.

	(1)	(2)
dig	0.633 ***	0.318 ***
	(11.685)	(4.191)
er		0.257
		(0.205)
lngdp		0.050
		(1.214)
ind2		0.070 ***
		(3.662)
urban		0.301 **
		(2.418)
coal		−3.438 **
		(−2.002)
_cons	0.422 ***	−0.164
	(46.950)	(−0.455)
N	310.000	310.000
Provincial fixed effects	YES	YES
Year fixed effects	YES	YES
r2	0.329	0.403
F	136.531	30.771

Note: ** and *** indicate significance at the 10% and 5% levels, respectively, with t-values in parentheses.

reports the benchmark regression results. Specifically, Column (1) includes only the core independent variable digital technology application (DIG); its regression coefficient is 0.633, highly significant at the 1% level, preliminarily verifying the positive promoting effect of digital technology applications on low-carbon economic transition developments. In Column (2), after further including control variables such as economic development level, industrial structure, energy structure, environmental regulation intensity, and urbanization level, the coefficient of DIG is 0.318, still significantly positive at the 1% level. This result indicates that after controlling for other key factors that may affect low-carbon transition, digital technology application can still significantly promote low-carbon economic transition development, robustly supporting hypothesis H1. Economically, the coefficient of 0.318 suggests that, holding other conditions constant, a one-unit increase in digital technology application is associated with an average 0.318-unit improvement in the level of low-carbon economic transition. This tangible magnitude indicates that the promotional effect of digital technology is not only statistically significant but also economically substantive, confirming the real contribution of digital empowerment to green low-carbon transformation. The underlying mechanism is that digital technology systematically promotes the economic and social low-carbon transition by enabling the intelligent management of energy and resources to improve efficiency, driving structural optimization towards a knowledge- and service-led lighter direction, and leveraging information transparency and data governance to optimize the green investment and policy environment.

5.3. Robustness Testing

First, an alternative measurement is adopted for the core explanatory variable. To mitigate the potential measurement bias inherent to composite index construction and following the empirical practice of textual analysis for technology orientation quantification [69], this study employs Python 3.11-based web crawling to conduct full-text content analysis on provincial annual Government Work Reports spanning 2015–2024. The total annual word frequencies of digital technology-related keywords, including internet, big data, artificial intelligence, blockchain, cloud computing and digital economy, are compiled and logarithmically transformed to construct a textual proxy for digital technology development (denoted as DIG2), which captures local governments’ strategic emphasis and policy inclination toward digital industries. This approach constructs a new proxy variable from the policy perspective. Given its independent calculation logic relative to the original composite index, it effectively mitigates measurement bias arising from a single indicator system and represents a classic strategy for testing the robustness of variable measurement. This indicator effectively reflects the institutional soft environment and policy backing for digital technology deployment. As presented in Column (1) of

Table 6. Robustness testing results.

	(1)	(2)	(3)	(4)
	Replacing Explanatory Variable	Replacing Dependent Variable	Excluding Municipalities	Shortening Time Interval
dig2	0.004 ***
	(6.114)
dig		0.050 **	0.283 ***	0.309 **
		(2.379)	(3.608)	(2.012)
er	1.444	0.757 **	0.428	0.628
	(1.174)	(2.177)	(0.321)	(0.173)
lngdp	0.049	−0.027 **	0.029	0.153 *
	(1.233)	(−2.384)	(0.607)	(1.950)
ind2	0.060 ***	0.015 ***	0.047 *	0.015
	(3.269)	(2.844)	(1.960)	(0.426)
urban	0.137	0.293 ***	0.400 **	0.562 **
	(1.091)	(8.502)	(2.550)	(2.604)
coal	−3.898 **	0.445	−2.862 *	−3.249
	(−2.364)	(0.936)	(−1.659)	(−0.934)
_cons	−0.086	0.270 ***	−0.015	−1.211 *
	(−0.247)	(2.707)	(−0.036)	(−1.732)
Provincial fixed effects	YES	YES	YES	YES
Year fixed effects	YES	YES	YES	YES
N	310.000	310.000	270.000	186.000
r2	0.442	0.516	0.370	0.191
F	35.972	48.487	23.192	5.855

Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively, with t-values in parentheses.

, the estimated coefficient of DIG2 equals 0.004 and is statistically significant at the 1% significance level (t = 6.114). The consistent positive estimate verifies the robust promotional effect of digital technology adoption on low-carbon economic transition after switching to a completely different policy-text-based measurement.

Second, the explained variable is substituted for robustness check. To alleviate measurement errors arising from composite index construction, green total factor productivity (LCTD2) calculated via the undesirable-output SBM-DEA model is adopted as an alternative proxy for low-carbon economic transition [70]. The SBM-DEA model is a mainstream tool for measuring low-carbon and green efficiency. Its input–output indicator system has been widely validated by existing studies, featuring a well-established paradigm and high academic recognition. Replacing the composite evaluation index with efficiency indicators enables conclusion verification from different perspectives. Meanwhile, the mature model specification guarantees objective and reliable test results, so there is no need to construct alternative input–output frameworks. From the input–output efficiency perspective, this metric quantifies green technical efficiency associated with decoupling economic expansion from carbon emissions. The regression results in Column (2) reports that the coefficient of core explanatory variable DIG is 0.050 and statistically significant at the 5% level (t = 2.379). Such evidence demonstrates that digital technology adoption not only elevates the composite level of low-carbon economy but also prominently improves green total factor productivity, thereby validating the robustness of baseline findings from an efficiency perspective.

Thirdly, the sample coverage is adjusted. Given that municipalities directly under the central government (Beijing, Tianjin, Shanghai and Chongqing) possess distinctive administrative ranks, economic structures and privileged policy resources that may bias benchmark estimations [71], these four municipalities are excluded from the original sample for an additional regression. Subsample regression excluding special samples offers the advantage of intuitively identifying whether the core findings are driven by a small number of outliers, and it effectively examines the external validity and generalizability of research conclusions. Column (3) tabulates the corresponding estimation results after sample exclusion. The coefficient of the core explanatory variable DIG is 0.283 and statistically significant at the 1% level (t = 3.608). This outcome confirms that the baseline finding is not driven by a small set of special observations and remains valid for conventional provincial samples.

Finally, the sample period is shortened. To concentrate on the in-depth development of the digital economy during the 14th Five-Year Plan period and eliminate potential structural discrepancies inherent to early-stage observations, the research window is restricted to 2019–2024 to identify the contemporary impacts of digital technology adoption [72]. Narrowing the research time window is a common method to test the temporal stability of conclusions. It helps identify whether the core effect fluctuates along with changes in the macroeconomic environment and development stages, and fully verifies the temporal persistence of the findings. As shown in Column (4), the coefficient of DIG equals 0.309 and is statistically significant at the 5% level (t = 2.012) within the subsample spanning 2019–2024. This finding reveals that the low-carbon enabling effect of digital technology persists significantly across the latest six years amid the rising strategic priority of the digital economy, verifying favorable temporal stability of the baseline conclusion.

Overall, the four robustness checks verify the reliability of baseline regression results from four dimensions: alternative measurement of explanatory variables, alternative measurement of dependent variables, sample representativeness and temporal consistency. All empirical results consistently corroborate the core conclusion that digital technology adoption exerts a statistically significant promoting effect on low-carbon economic transition. Such corroborative evidence solidifies the robustness and credibility of this study’s empirical outcomes.

5.4. Endogeneity Test

Although provincial and year fixed effects are incorporated into the baseline specification to mitigate the omitted variable bias stemming from time-invariant regional heterogeneity and common macroeconomic shocks, potential endogeneity bias arising from reverse causality and unobserved omitted variables remains a concern. To address such endogeneity, the two-stage least-squares (2SLS) estimator is employed for re-estimation. Following conventional empirical practices in the extant literature, the one-period lag of digital technology adoption (L.DIG) is chosen as the instrumental variable (IV) for the core explanatory variable DIG [73]. The rationality is reflected in the following aspects. First, relevance. The regional digital technology level in the previous period is strongly correlated with its current level. This is verified by the highly significant coefficient of L.DIG (0.883, p < 0.01) and a sufficiently large Wald F statistic of 426.42 in the first-stage regression. This value far exceeds the critical value of 16.38 corresponding to the 10% maximal bias level, suggesting no weak instrument problem. Second, this approach alleviates reverse causality. The one-period lagged value of digital technology reflects objective conditions formed in the past, so current low-carbon transition activities cannot retroactively alter the historical digital technology foundation. This largely mitigates the bias arising from bidirectional causality from the perspective of temporal logic. Third, it addresses omitted variable bias and measurement errors. On the one hand, provincial time-varying contextual factors and industrial policies show strong continuity. The lagged digital technology level can indirectly capture regional policy characteristics in previous periods. Combined with two-way fixed effects, this strategy absorbs most impacts of both time-varying and time-invariant omitted variables. On the other hand, the DIG indicator in this study is constructed based on official statistics and a well-established evaluation system, leading to minor systematic measurement errors. Since the random errors of lagged and current variables are mutually independent, the lagged term can effectively isolate the disturbance caused by random measurement errors. Collectively, the selected instrument satisfies the identifying assumptions required for valid IV estimation. The 2SLS estimation outcomes are documented in

Table 7. Endogeneity test results.

	(1)	(2)
	First	Second
Variables	dig	lctd
L.dig	0.883 ***
	(20.650)
dig		0.312 ***
		(2.776)
er	0.099	−0.339
	(0.220)	(−0.218)
lngdp	−0.026 *	0.057
	(−1.908)	(1.311)
ind2	−0.003	0.049 **
	(−0.374)	(2.219)
urban	0.015	0.390 ***
	(0.331)	(2.728)
coal	−0.264	−4.362 **
	(−0.375)	(−2.098)
Provincial fixed effects	YES	YES
Year fixed effects	YES	YES
Cragg-Donald Wald F statistic	426.42
LM statistic (p-value)		19.178 (0.000)

Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively, with t-values in parentheses.

. Column (1) presents the first-stage results, confirming strong predictive power of L.DIG for current-period DIG. The second-stage estimates in Column (2) reveal that after correcting for endogeneity, the coefficient on DIG equals 0.312 and is statistically significant at the 1% level (z = 2.78). This estimate carries the same sign and comparable magnitude relative to the baseline coefficient of 0.398. Meanwhile, the under-identification LM statistic is significant at the 1% level, rejecting the null hypothesis of under-identification for the instrumental variable.

Collectively, the 2SLS results confirm that the positive promoting effect of digital technology adoption on low-carbon economic transition remains robust after addressing potential endogeneity bias via the instrumental variable approach. This evidence validates the core findings from the baseline regression and strengthens the credibility of the study’s causal inferences.

5.5. Mediation Effect Test

To identify the mediating role of green investment in the linkage between digital technology adoption and low-carbon economic transition, sequential regression estimations based on specifications (2) and (3) are implemented, with the corresponding results reported in

Table 8. Mediation effect test results.

	(1)	(2)
	gi	lctd
dig	2.594 ***	0.254 ***
	(6.401)	(3.145)
gi		0.025 **
		(2.188)
er	19.368 ***	−0.220
	(2.887)	(−0.174)
lngdp	−1.091 ***	0.077 *
	(−4.973)	(1.802)
ind2	0.226 **	0.064 ***
	(2.216)	(3.364)
urban	2.600 ***	0.237 *
	(3.914)	(1.865)
coal	−12.708	−3.125 *
	(−1.385)	(−1.826)
_cons	11.930 ***	−0.457
	(6.200)	(−1.197)
Provincial fixed effects	YES	YES
Year fixed effects	YES	YES
N	310.000	310.000
r2	0.438	0.414
F	35.521	27.425

Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively, with t-values in parentheses.

. Given that the research sample consists of long provincial panel data, two-way fixed effects for provinces and years are incorporated into the stepwise regression equations to avoid model specification bias, thereby controlling for region-specific inherent characteristics and time-varying macroeconomic shocks. As shown in Column (1), the coefficient of digital technology adoption on green investment is 2.594 and statistically significant at the 1% level, implying that digital technology development significantly stimulates green investment expansion. Economically, the coefficient of 2.594 indicates that each one-unit increase in digital technology adoption significantly expands the scale of green investment by 2.594 units, confirming that digital development effectively stimulates green capital accumulation. In Column (2), after incorporating green investment into the regression specification, the direct coefficient of digital technology adoption stands at 0.254 and remains significant at the 1% level; meanwhile, the coefficient on green investment is positive at 0.025 and statistically significant at the 5% level.

Given the statistical limitations inherent to the stepwise regression approach, this study further re-examines the mediating effect using the more rigorous bias-corrected bootstrap method with 1000 repeated sampling iterations. The bootstrap procedure circumvents the normality assumption for the product of regression coefficients and yields more reliable statistical inferences.

As documented in

Table 9. Mediation effect results based on bootstrap estimation.

Effect Type	Effect Value	Bootstrap SE	95% Percentile Confidence Interval	95% Bias-Corrected (BC) CI	Significance Judgment
Indirect Effect	−0.199	0.061	[−0.318, −0.089]	[−0.333, −0.096]	Significant
Direct Effect	0.661	0.173	[0.344, 1.020]	[−0.333, −0.096]	Significant
Total Effect	0.462

, digital technology adoption exerts a significantly positive direct effect on low-carbon transition (effect = 0.661, bias-corrected 95% CI = [0.365, 1.029]). From an economic perspective, each additional unit of digital technology adoption directly lifts the low-carbon transition level by 0.661 units. By contrast, the indirect effect transmitted via green investment is statistically significant yet negative (effect = −0.199, bias-corrected 95% CI = [−0.333, −0.096]). In economic magnitude, the green investment channel offsets a 0.199-unit improvement in low-carbon transition per unit increase in digital technology adoption due to capital crowding-out. The direct effect and indirect effect have opposite signs, and none of the bias-corrected confidence intervals contain zero, indicating a typical suppressing effect. A suppressing effect suggests that if the mediating variable (green investment) is omitted, the total effect of the core explanatory variable (digital technology) on the explained variable (low-carbon transition) will be underestimated. In this study, the results reveal that green investment does not exert a simple positive mediating effect as hypothesized in H2. Instead, this transmission path imposes an inhibiting impact on low-carbon transition and partially offsets the direct promoting effect of digital technology. This finding demonstrates the complexity of the transmission mechanism running from digital technology through green investment to low-carbon transition. The total effect is a combination of the positive direct effect and the negative indirect effect. A plausible explanation is that the rapid development of digital technology attracts substantial capital in the short run, thereby creating a crowding-out effect on certain long-term, capital-intensive green projects. Alternatively, the overall improvement in economic efficiency driven by digital technology reduces the marginal demand for green investment targeting end-of-pipe governance.

5.6. Heterogeneity Analysis

This heterogeneity analysis examines group differences in the total direct effect of digital technology on low-carbon transition, which represents a common paradigm for heterogeneity research based on macro panel data in China. Accordingly, separate mediation effect tests are not repeated for each subsample.

First, to explore regional heterogeneity in the impacts of digital technology adoption on low-carbon economic transition, the full sample is divided into eastern, central and western subgroups following the regional classification issued by the National Bureau of Statistics of China, and grouped regressions are conducted [74]. It should be noted that although the traditional division of eastern, central and western regions is based on administrative geography, it is highly correlated with regional endowments of digital infrastructure, industrial structure characteristics and basic conditions for green transition within China’s regional economic system. This classification presents distinct economic stratification and serves as a classic quantitative grouping framework for heterogeneity analysis using provincial macro data. As reported in

Table 10. Regional heterogeneity test results.

	(1)	(2)	(3)
	Eastern	Central	Western
dig	0.216 **	0.177	−0.047
	(2.180)	(0.525)	(−0.221)
er	1.293	−2.160	1.605
	(0.312)	(−0.625)	(1.233)
lngdp	0.013	0.084	−0.003
	(0.172)	(1.150)	(−0.029)
ind2	0.104 ***	0.040	0.026
	(3.311)	(0.845)	(0.740)
urban	0.647 ***	0.474	0.452 **
	(2.693)	(1.148)	(2.076)
coal	−6.497 **	0.736	0.306
	(−2.378)	(0.225)	(0.091)
_cons	0.257	−0.593	−0.018
	(0.343)	(−0.974)	(−0.024)
Provincial fixed effects	YES	YES	YES
Year fixed effects	YES	YES	YES
N	110.000	100.000	100.000
r2	0.562	0.352	0.225
F	19.907	7.595	4.067
Fisher’s Permutation test p-value	0.000

Note: ** and *** indicate significance at the 10% and 5% levels, respectively, with t-values in parentheses.

, the enabling effect of digital technology adoption exhibits pronounced regional disparities. Specifically, digital technology adoption yields a significantly positive promoting effect exclusively in the eastern region, with a coefficient of 0.216 that is significant at the 1% level. Economically, every one-unit increase in digital technology adoption in eastern China contributes to an average 0.216-unit improvement in local low-carbon economic transition, indicating a substantial economic contribution of digital empowerment, whereas its coefficients for central and western regions lack statistical significance. In economic terms, the empirical results imply that digital development in central and western regions cannot effectively generate tangible low-carbon transition benefits, and its low-carbon driving force has not yet been economically manifested. It is worth noting that merely comparing the significance of regression coefficients across groups is a descriptive analysis and cannot rigorously prove that the between-group differences are statistically significant. To further consolidate empirical support for regional heterogeneity, we implement Fisher’s cross-sample coefficient equality test for three regional subgroups. The corresponding p-value is less than 0.01, which confirms that the inter-group discrepancies are statistically significant rather than only numerical differences. This quantitative result confirms that the low-carbon enabling effect of digital technology differs not merely in magnitude but presents substantial structural heterogeneity across eastern, central and western regions. It provides solid empirical evidence for regional divergence from a mathematical perspective and effectively addresses the limitation that grouped regression can only conduct descriptive comparisons. Such evidence points to unbalanced low-carbon dividends from digital technologies across geographic zones. Two underlying mechanisms account for the documented heterogeneity. First, regions differ substantially in terms of their digital economy fundamentals and industry integration depth. The eastern area boasts mature digital infrastructure, diversified practical application scenarios and well-functioning data factor markets, enabling digital technologies to facilitate productivity improvement and industrial structural upgrading efficiently. Second, disparate foundational conditions for green transition contribute to the gap. Benefiting from a less energy-intensive industrial structure and sophisticated green financial systems, eastern regions are better positioned to translate digital advantages into tangible green investment and low-carbon output. By contrast, central and western areas are constrained by path dependence on high-pollution traditional industries and inadequate supporting facilities, restricting the full realization of digital low-carbon dividends.

Second, this paper further explores heterogeneity from the perspective of regional marketization to verify whether the low-carbon transition effect of digital technology adoption depends on local institutional environments. Drawing on the Report on China’s Provincial Marketization Index compiled by Wang et al. (2021) [54], the provincial marketization index is selected as the proxy variable reflecting regional institutional conditions and market development. Referring to common practices in the existing literature, this paper divides the full sample into high-marketization and low-marketization groups annually based on the median value of the index to conduct heterogeneous grouping tests. It should be noted that this study adopts median grouping to examine heterogeneity across marketization levels, instead of constructing continuous interaction terms or applying threshold regression models. This choice is rational and scientific given the research objectives. First, interaction terms can only identify the linear average moderating effect of marketization and capture merely monotonic co-variation between variables. They fail to characterize structural and stepwise boundary differences that emerge when marketization exceeds a certain level. Second, threshold regression models focus on estimating a single critical threshold value, which is suitable for studies targeting threshold identification. However, the primary focus of this paper is not to calculate the exact threshold of marketization, but to explore whether hierarchical differences in institutional marketization lead to systematic disparities in the low-carbon enabling effect of digital technology. In this context, the grouping comparison approach for structural testing is more appropriate.

The grouped regression results are presented in

Table 11. Heterogeneity results by marketization level.

	(1)	(2)
	Low-Marketization Group	High-Marketization Group
dig	−0.094	0.222 **
	(−0.439)	(2.151)
er	−0.373	−2.175
	(−0.293)	(−0.453)
lngdp	0.014	0.030
	(0.274)	(0.428)
ind2	0.053 **	0.089 ***
	(2.259)	(2.772)
urban	0.443 **	0.597 ***
	(2.283)	(3.027)
coal	−2.915	−5.546 **
	(−1.147)	(−2.135)
_cons	−0.023	0.062
	(−0.052)	(0.100)
Provincial fixed effects	YES	YES
Year fixed effects	YES	YES
N	150.000	160.000
r2	0.259	0.503
F	7.438	23.083
Fisher’s Permutation test p-value	0.000

Note: ** and *** indicate significance at the 10% and 5% levels, respectively, with t-values in parentheses.

. In the high-marketization subgroup, the coefficient of digital technology adoption is 0.222 and statistically significant at the 5% level. Economically, a one-unit increase in digital technology adoption contributes to a 0.222-unit improvement in low-carbon economic transition in highly marketized regions, confirming the economically meaningful promotional effect of digital empowerment under sound institutional conditions. This demonstrates that digital technologies can effectively accelerate low-carbon economic transition in regions with mature market mechanisms and superior institutional environments. By comparison, the coefficient is −0.094 and statistically insignificant for the low-marketization group. In economic terms, although the coefficient fails to pass the significance test, the negative value indicates that digital technology cannot generate effective low-carbon dividends in low-marketization areas and may even produce a slight inhibitory effect on low-carbon transformation. This indicates that the promoting effect does not emerge and may even exert a slight adverse impact in such regions. To further consolidate empirical evidence for market-oriented heterogeneity, we perform Fisher’s cross-sample coefficient equality test between high- and low-marketization subgroups, and the corresponding p-value is smaller than 0.05, which statistically confirms the coefficient discrepancy between the two subsamples is not only a simple numerical gap but has significant statistical meaning.

In highly marketized regions, clear property rights, reliable contract enforcement and equitable market competition reduce transaction costs and uncertainties during the integration of digital and green technologies and secure returns from innovation, thus fully stimulating the low-carbon dividends of digital technologies. In contrast, market segmentation, distorted factor allocation and institutional obstacles prevalent in low-marketization areas hinder the realization of positive externalities of digital technologies; resource misallocation may further induce unintended adverse outcomes.

5.7. Discussion

This section compares our baseline regression, mediation effect estimation and grouped heterogeneity regression results with relevant prior literature, summarizes consistent findings, identifies the marginal contributions of this research, and elaborates its theoretical and policy implications.

First, results from the baseline regression. The existing literature has verified the positive linkage between digital technology progress and low-carbon transition from macroeconomic, industrial and micro-firm perspectives, arguing that digital development facilitates decarbonization via energy restructuring and green innovation [13,14,15]. Consistent with these mainstream findings, our baseline estimation documents a statistically significant positive association between digital technology adoption and low-carbon economic development. This correlation survives multiple robustness checks including instrumental variable correction, alternative variable measurement and sample truncation, which further consolidates empirical evidence on the digital–low-carbon nexus. While prior studies predominantly concentrate on the aggregate net outcome of digitalization, this paper extends the research scope by further exploring underlying transmission mechanisms.

Second, evidence from the green investment-mediated mechanism. Prevailing studies generally propose a straightforward theoretical path: digital technologies mitigate information asymmetry and financing constraints through big data and artificial intelligence, expand green investment volume, and accordingly promote low-carbon advancement, which implies a unidirectional linear chain of “digitalization → rising green investment → low-carbon upgrading” [38,39]. However, our Bootstrap-based mediation test delivers distinct evidence: digital technology is positively correlated with low-carbon transition via a direct channel but negatively correlated through the green investment pathway, jointly generating a notable suppressing effect. In the short run, booming digital industries attract massive social capital and crowd out financing for long-cycle, capital-intensive green projects; meanwhile, productivity growth reduces marginal investment demand for end-of-pipe pollution abatement, resulting in an adverse indirect linkage via green investment. This empirical pattern echoes scattered evidence from Ahmed et al. (2025) [40] regarding the capital-cost-raising effect of digitalization. Marginal contribution compared with existing literature [40]: This study challenges the conventional consensus that green investment serves solely as a positive mediator. By identifying the suppressing effect within the complete “digital technology–green investment–low-carbon transition” framework, this paper reconciles conflicting empirical results across prior works, revises the oversimplified linear assumption, and enriches theories concerning green capital allocation under digital development.

Third, heterogeneous results across regions and marketization levels. Previous research has acknowledged heterogeneous low-carbon outcomes of digitalization and threshold characteristics of digital dividends [20,31], yet most analyses separately examine geographic disparity or single economic indicator without integrating regional location and institutional marketization into one empirical framework. Our subsample regressions reveal that the positive digital–low-carbon correlation only holds statistically for Eastern China and highly market-oriented regions, whereas no significant linkage emerges in central–western provinces and low-marketization areas. Well-defined property rights, unimpeded factor mobility and equitable market competition constitute essential institutional prerequisites to materialize digital decarbonization gains; conversely, path dependence on high-carbon traditional industries, market segmentation and insufficient institutional support hinder the release of low-carbon potential in less developed regions. Making a marginal contribution compared with the existing literature, this paper explains heterogeneous coefficient magnitudes from the perspective of institutional endowment and defines marketization as a critical contingent condition for digital decarbonization. It overcomes the limitation of earlier heterogeneity analyses that merely report significance without exploring underlying mechanisms, and expands research on the institutional boundary of digitally driven low-carbon development.

6. Conclusions and Suggestions

6.1. Conclusions

Based on China’s provincial panel data from 2015 to 2024, this paper systematically investigates the correlation, transmission mechanism and boundary conditions between digital technology application and low-carbon economic transition. The main findings are as follows. First, digital technology application presents a stable positive correlation with the development of the low-carbon economic transition. The positive correlation remains statistically significant after a series of strict robustness tests addressing endogeneity, variable measurement, sample coverage and time span. It indicates that digital technologies provide reliable technical support for the systematic green and low-carbon transition of China’s economy and society through direct channels such as enabling intelligent energy and resource management, advancing knowledge-based and service-oriented industrial transformation, and improving the accuracy of environmental governance. Second, the enabling mechanism of digital technologies for the low-carbon transition is complex, and green investment acts as a suppressor instead of a connecting bridge. Mediation test results reveal that digital technology application has a positive direct correlation and a negative indirect correlation with the low-carbon transition, which jointly form a suppressing effect. Specifically, despite the favorable direct correlation, the booming digital economy may attract capital into high-return digital sectors in the short term, crowding out funding for long-cycle and capital-intensive green projects. Meanwhile, elevated overall operational efficiency cuts the marginal demand for end-of-pipe governance investment, thus generating mild restraining correlation with low-carbon transition via the green investment path. This finding verifies that the nexus of “digital technology–green investment–low-carbon transition” is not a simple linear transmission but a sophisticated process with inherent conflicts, revising the traditional opinion that green investment solely serves as a positive mediator. Third, the correlation between digital technology application and the low-carbon transition shows prominent regional heterogeneity and is subject to specific regional conditions and institutional environments. Heterogeneity analysis shows the positive correlation is only statistically significant in Eastern China and regions with high marketization. The low-carbon dividends of digitalization are not universally available and carry distinct threshold features: sound digital infrastructure, lightweight industrial structure and superior market-oriented institutions are preconditions necessary for realizing the effective synergy between digital advancement and the green transition. In regions with an underdeveloped digital economy, inefficient factor allocation and inadequate institutional guarantees, the low-carbon potential of digital technologies cannot be fully released, and resource distortion may even curb its potential green benefits. These results underline the critical moderating effect of regional initial endowments and institutional foundations.

6.2. Suggestions

The findings of this study offer novel theoretical insights into the driving mechanisms of green transition amid the digital economy era.

First, this research advances theoretical understandings of how digital technologies correlate with green transition. Empirical evidence documents a robust positive direct association between digital technologies and the low-carbon transition, alongside a negative indirect linkage transmitted via green investment, forming a typical suppressing effect. Accordingly, digitalization affects decarbonization not through a single linear pathway, but via dynamically balanced competing effects including facilitation and capital crowd-out. Such results contradict the oversimplified “digitalization → expanded green investment → low-carbon upgrading” assumption prevalent in the existing literature. Future theoretical modeling should accommodate mixed marginal effects instead of presuming universally beneficial technological impacts.

Second, this paper highlights the value of an institution–technology synergy perspective for green transition research. Heterogeneity results reveal the favorable low-carbon linkage of digital technologies only exists in regions with high marketization. Advanced technologies alone cannot guarantee desirable environmental outcomes, as their efficacy is embedded within local institutional arrangements. Sound market institutions covering property rights protection, unimpeded factor mobility and fair competition cut transaction costs of industrial integration and unlock the environmental dividends of digitalization. Hence, relevant theories ought to move beyond technological determinism and focus on the interplay, coordination and friction between technology and institutions, establishing an integrated analytical framework covering technology, institution and governance.

Third, this study provides fresh explanations for unbalanced regional green development. The insignificant decarbonization correlation in central and western China originates not merely from inadequate digital infrastructure but lagging market-oriented reforms. Underdeveloped regions are constrained by both digital access barriers and institutional defects, making it difficult to translate digital progress into tangible environmental gains. Accordingly, research on coordinated regional green development needs to expand beyond conventional drivers such as resource endowments and industrial composition toward emerging factors including digital capacity, data factor markets and business environment.

Based on the empirical results of provincial panel data, mediating mechanism analysis, as well as regional and institutional heterogeneity examinations, this study proposes policy recommendations from three dimensions: provincial overall planning, prefecture-level implementation and corporate practice. The empirical evidence reveals that digital technology exerts a significant and positive direct effect on the low-carbon transition. Nevertheless, green investment generates adverse impacts during the transmission process, and the effectiveness of digital technology varies substantially across regions and institutional environments. Accordingly, relevant policies should consolidate the direct emission reduction benefits of digital technology, address the inefficient transmission of green investment, and carry out targeted optimization in line with regional characteristics and institutional conditions.

First, the benchmark regression results show that a one-unit increase in digital technology (DIG) raises the level of low-carbon transition by an average of 0.318 units, and this finding remains statistically significant across multiple robustness tests. This demonstrates that digital technology delivers stable emission reduction performance. Instead of merely expanding investment scale, policies should prioritize improving practical application outcomes. From the provincial perspective, authorities shall tailor digital transformation initiatives to local industrial structures and focus on high-energy-consuming sectors including industry, power supply, construction and transportation. Key indicators such as carbon emission intensity and carbon emissions per unit of GDP shall be decomposed and assigned to prefecture-level cities. At the prefecture level, local governments shall leverage existing data infrastructure to improve the monitoring system for energy consumption and carbon emissions, and conduct dynamic supervision on major emission-intensive enterprises. Relevant governance targets can be further refined for industrial parks and specific industries. For enterprises, digital energy management tools should be adopted to optimize core production procedures, so as to materialize the emission reduction advantages of digital technology in daily operations.

Second, regarding the negative mediating effect of green investment, policymakers shall attach greater importance to capital allocation structure rather than investment volume. The empirical results indicate that green investment fails to amplify the emission reduction effects of digital technology, largely because capital is excessively allocated to short-term pollution control projects. Therefore, policy guidance should steer more capital toward clean technology research and development as well as energy structure adjustment. Provincially, governments shall improve the database for environmental and carbon emission information to enhance the identification accuracy of green projects, and adopt fiscal and credit policies to guide capital allocation structurally. Prefecture-level authorities shall formulate differentiated support schemes for various green projects based on local industrial features, and maintain a reasonable capital ratio between energy-saving renovation and technological innovation. Local financial institutions are encouraged to develop smart contract-based green financial products. With the support of digital technologies, the carbon reduction performance of green projects can be quantitatively evaluated, and credit lines as well as interest rates can be determined accordingly. This framework effectively prevents capital from being captured by pseudo-green projects. Enterprises are required to strengthen environmental information disclosure and reduce symbolic environmental investment aimed solely at obtaining policy benefits. Digital tools can be utilized to report authentic environmental performance and constrain opportunistic symbolic green investment.

Third, differentiated development strategies should be implemented under provincial coordination and prefecture-level delivery, in accordance with the quantified regional heterogeneity results. Heterogeneity analysis suggests that digital technology achieves prominent low-carbon effects in eastern China, while the corresponding coefficients are statistically insignificant in central and western regions, where the dividends of digitalization have not been fully unlocked. In view of such quantitative disparities, region-specific policies are formulated and extended to subordinate cities and market entities. For eastern provinces with solid digital foundations and sound market environments, provincial governments shall prioritize cutting-edge technological innovation and model promotion. More pilot autonomy can be granted to support the construction of digital low-carbon industrial parks and smart zero-carbon cities. Leading enterprises are encouraged to conduct research on deep decarbonization technologies and develop replicable collaborative models between governments and enterprises as well as among firms for nationwide promotion. Prefecture-level cities in eastern regions shall pursue innovation-driven development and promote in-depth integration between digital technology and low-carbon development in high-end manufacturing and modern service industries. For central provinces undergoing industrial transformation, provincial authorities shall set clear tasks and timetables for the digital and green upgrading of traditional industries. Industrial parks serve as key platforms for pilot practices to accelerate the transformation of high-carbon sectors such as iron and steel and chemical industry, and gradually unlock the decarbonization potential of digital technology. For western provinces endowed with distinctive ecological resources, the primary task for provincial governments is to remedy deficiencies in digital infrastructure at the county and rural levels by clarifying construction schedules and coverage. Prefecture-level cities shall capitalize on ecological advantages to foster characteristic business models, including eco-product value realization and intelligent operation of new energy facilities. This helps translate resource endowments into green economic advantages and cultivate the low-carbon effects of digital technology in a gradual manner.

Fourth, comprehensive institutional reforms should be deepened to fully exploit the low-carbon potential of digital technology. The grouping regression results on marketization show that digital technology presents significant positive effects in highly market-oriented regions, whereas the coefficients are negative and insignificant in regions with low marketization, and the between-group differences are statistically validated. In this context, tiered improvements to institutional frameworks and regulatory systems are necessary. Provincially, reforms on the marketization of data, labor, land and other production factors should be accelerated to eliminate regional market segmentation. Fundamental systems concerning property rights protection and fair competition need to be improved to remove barriers to factor mobility from the top-level design. At the prefecture level, a well-balanced and prudent regulatory regime for digital industries shall be established. Regtech-enabled intelligent inspection platforms can be built to set quantitative criteria and penalties for greenwashing and data fraud, with regular inspections to maintain the order of the green market. For enterprises, strengthened intellectual property protection can secure the revenue from innovations in digital and green technologies, and stabilize their long-term investment expectations. Multi-level institutional optimization alleviates factor allocation distortions, enabling less market-oriented regions to cross institutional thresholds and fully release the green externalities of digital technology.

6.3. Research Limitations and Future Directions

Although this study has yielded some meaningful findings, certain limitations remain that need to be addressed in future research. First, regarding research data and variable measurement. This study relies on provincial-level panel data. While such data can reflect overall trends from a macro perspective, the observational empirical design cannot fully eliminate confounding factors compared with quasi-experimental approaches. Accordingly, the empirical findings should be interpreted as robust correlational patterns rather than as definitive causal effects. Future research can integrate multi-source data and adopt alternative quantitative methods to optimize variable measurement and reduce measurement errors. Research can combine enterprise-level data to deeply examine the impact mechanism of digital technology on carbon emissions from specific production and operation activities from a micro perspective, or use city-level data to improve the precision and relevance of research conclusions. Second, regarding mechanism research, although this paper verifies the mediating role of green investment, there may be other important transmission channels through which digital technology promotes low-carbon transition, such as green technology innovation, industrial structure optimization, and changes in resident consumption behavior. Subsequent research can construct a more comprehensive mechanism analysis framework to explore the relative importance and interactions among multiple transmission paths. Third, regarding the research perspective, this study mainly focuses on the situation within China, while the impact of digital technology on low-carbon transition may vary across countries. Future international comparative studies could analyze differences in the role of digital technology in different development levels and institutional environments, providing references for countries to formulate digital–green synergistic development policies suitable for their own characteristics. In addition, the empirical analysis of this study is confined to a linear specification framework, without extended examinations such as threshold or interaction effect estimations, leaving the non-linear and stage-dependent features underlying the digital technology–low-carbon transition nexus unexplored. Future research may adopt threshold regression and moderating-effect models centered on institutional environment and green capital investment to identify critical threshold values and interactive mechanisms for digital decarbonization, thereby deepening the empirical rigor of relevant analyses. Additionally, as digital technology rapidly develops and application scenarios continuously expand, new research questions will continue to emerge, requiring the ongoing tracking of the latest developments in digital technology and their impact on sustainable development.