The nonlinear impact of digital finance on carbon emission efficiency: evidence from Chinese cities

Amid persistent growth in global carbon outputs, enhancing carbon emission efficiency (CEE) has emerged as a pivotal strategy for synchronizing economic and ecological objectives. Digital finance (DF), leveraging its distinctive technological capabilities and inclusive financial features, plays a catalytic role in CEE optimization. This research methodically investigates the effect of DF on CEE using panel data (282 Chinese cities, 2011–2022) and several panel econometric estimation techniques. The findings reveal that: 1. The influence of DF on CEE follows a U-shaped trend, first suppressing and then promoting. Heterogeneity analysis reveals that the U-shaped relationship is more pronounced across two dimensions—coverage breadth and usage depth—as well as in resource-based cities and key environmental protection cities. 2. Mechanism analysis demonstrates that DF exerts a U-shaped effect on green technology innovation, human capital accumulation, and public environmental awareness, thereby indirectly shaping CEE. 3. Policy synergy test reveals that the “Carbon Emission Trading” and “Low-Carbon City” pilot policies significantly amplify the U-shaped relationship between DF and CEE. In contrast, the effect of digitalization and infrastructure-oriented policies, exemplified by the “Smart City” and “Broadband China” pilot policies, is statistically insignificant. This study not only enriches the theoretical understanding of DF’s environmental effects but also provides policy insights for leveraging DF to accelerate green transformation.

1 Introduction

Global climate change’s increasing effects pose significant challenges to sustainable development, primarily driven by the intensifying greenhouse effect resulting from anthropogenic CO₂ emissions (Li et al., 2025; Nusa et al., 2025). In this context, advancing ecological modernization and mitigating climate-altering emissions have garnered widespread international consensus. Carbon emission efficiency (CEE), a major metric of low-carbon economic performance, is an important tool for balancing environmental sustainability with economic growth (Sun and Huang, 2020). Beyond offering transitional flexibility for conventional industries, CEE also supports the gradual dissociation between GDP growth and carbon output (Wang and Su, 2020). Global CO₂ emissions increased marginally by 0.8% year-on-year in 2024, while economic growth reached 3.2%². This indicates a further weakening of the linkage between carbon emissions and economic development and marking preliminary progress in global low-carbon transition. However, significant disparities in CEE persist across nations. Developed countries, leveraging clean energy technologies and industrial restructuring, have achieved notable emission reductions, whereas developing countries face greater challenges due to energy structure constraints and technological gaps (Napp et al., 2014). Thus, identifying effective pathways to enhance CEE has become an urgent priority, with decisive implications for global climate governance.

Recently, the rapid development of digital finance has provided new technological tools and innovations in financial models to address this challenge. As reported by China Research Network, the global DF market surpassed $8.5 trillion in 2024, with a compound annual growth rate of 18.7% from 2019 to 2024, reflecting its robust expansion (Figure 1). Digital finance, leveraging its advantages of low cost, high penetration, and intelligence, effectively breaks the spatial and temporal constraints of traditional financial services (Wang et al., 2024). Furthermore, AI technologies enhance its efficiency in data mining, risk pricing, and resource allocation. This enables more precise identification of green projects, tracking of carbon footprints, and optimization of green capital allocation (Wu et al., 2024), thereby creating new potential pathways for improving carbon emission efficiency (CEE). However, there is significant theoretical debate regarding the impact of digital finance on CEE. Some scholars emphasize its “green empowerment” effect, contending that it channels capital toward low-carbon sectors (Ding et al., 2023). Others, however, warn of a potential “carbon rebound” effect—for instance, where convenient credit stimulates high-carbon consumption, or more efficient capital allocation indirectly expands energy-intensive industries—which could suppress improvements in CEE (Le et al., 2020). Existing research has yet to adequately reconcile these opposing views, particularly lacking systematic empirical examination of the nonlinear mechanisms and complex transmission pathways through which digital finance influences CEE. Therefore, clarifying this debate and identifying the conditions under which digital finance serves as a “green accelerator” rather than a “carbon lock-in” tool is of significant theoretical and practical importance.

Figure 1

Figure 1. Distribution maps of DF.

Meanwhile, China, as the world’s largest carbon emitter and one of the economies with the most extensive application of digital finance, is experiencing intensive dual pressures. These pressures arise from the rapid development of digital finance on one hand, and the rigid constraints of carbon reduction on the other (as shown in Figures 1, 2). This context amplifies the interaction between DF and carbon emission efficiency, thereby enhancing the feasibility of empirically identifying their nonlinear characteristics and complex impact mechanisms. Simultaneously, the coordinated advancement of the national-level “Dual Carbon” goals and the “Digital China” strategy provides a unique research window. This window is characterized by clear policy structures and accelerating processes, which enables the observation of the long-term dynamic evolution of both DF and green transformation. Consequently, this window not only enhances the observability of their relationship but also provides a crucial real-world scenario for analyzing the underlying causal pathways and synergistic mechanisms in depth. Against this backdrop, this study focuses on Chinese cities to explore three core questions: (1) What is the overall impact of digital finance on carbon emission efficiency? (2) Through what key mechanisms does this impact occur? (3) How does the effect vary across regions with different characteristics and policy environments? The findings aim to provide a theoretical and empirical foundation for synergistically advancing digital finance innovation and low-carbon transition. They also hold significant reference value for other emerging economies facing similar challenges.

Figure 2

Figure 2. Distribution maps of CEE.

Marginal contributions of this study: 1. Research perspective: Unlike the prevalent linear analysis approaches adopted in existing literature (Ding et al., 2023; Zhang and Liu, 2022; Zhang et al., 2025), this paper establishes a nonlinear analytical framework that captures DF’s stage-dependent developmental dynamics, uncovering its heterogeneous effects on CEE across growth phases. 2. Mechanism framework: A tripartite “technology supply–factor allocation–social demand” framework is proposed to systematically explain how DF enhances CEE through green innovation, human capital (HC) accumulation, and public environmental awareness (PEA). 3. Policy dimension: Unlike studies focusing solely on direct policy effects (Hou et al., 2024; Yu and Zhang, 2021; Zhang et al., 2022a), this research demonstrates the synergistic role of national policies in moderating the DF–CEE relationship, offering empirical evidence for policy optimization.

2 Literature review

2.1 Research on measurement and influencing factors of CEE

The concept of CEE was first proposed by Kaya and Yokobori (1997). Early studies on its measurement employed single-factor indicators (e.g., CO₂ emissions per unit GDP) to assess the environmental impact of economic growth (Sun, 2005). While such intensity indicators offered computational simplicity (Vujović et al., 2018), they only reflected the proportional relationship between output and carbon emissions and failed to comprehensively consider the synergistic effects of factors like capital, labor, and technology. Consequently, scholars have shifted to efficiency measurement methods under a total-factor framework. These methods evaluate technical inefficiency by comparing actual input-output combinations with the production frontier (Farrell, 1957). Among them, the slack-based measure (SBM) model (Tone, 2001) enables more precise identification of input excesses and output shortfalls, thereby improving the reliability and scientific validity of CEE assessments.

Previous research on the drivers of carbon emission efficiency (CEE) formed two complementary strands: internal capacity building and external environmental intervention. The first strand focused on technological innovation as the core engine for internal capacity building. Broad technological progress (Xie et al., 2021) and green innovation (Fang et al., 2022) were identified as key factors in enhancing efficiency. Meanwhile, Zhao et al. (2023) examined the unique role of digital technological innovation and suggested a potential complex, U-shaped nonlinear relationship with CEE. Research on external environmental intervention primarily explored the role of policies and digitalization. Command-and-control policies such as the “Low-Carbon City” pilot policy (LCCPP) and the “Carbon Emission Trading” pilot policy (CETPP), and market-incentive policies such as carbon trading, were shown to effectively set emission reduction constraints and incentives, thereby improving CEE (Hou et al., 2024; Yu and Zhang, 2021; Zhang et al., 2022a; Li Z. et al., 2022). Conversely, studies examining digitalization-driven improvements in CEE often treated digitalization (e.g., the digital economy, digital infrastructure) as a static regional endowment or industrial characteristic. These studies investigated mechanisms such as optimizing industrial and energy structures, promoting technological progress and green innovation (Han and Jiang, 2022; Wang et al., 2022), and enhancing public environmental awareness and urban innovation capabilities (Du et al., 2023). However, the existing literature predominantly examined the direct impact of individual policies on CEE. Thus, given the practical need for coordinated policy development, how external policy shocks influence the relationship between digital finance and CEE remained systematically underexplored.

2.2 Research on DF and the environment

Regarding the environmental effects of digital finance, existing research confirmed its multi-dimensional impacts, such as promoting green technology innovation (Feng et al., 2022), mitigating environmental inequality (Li et al., 2022b), fostering green development (Zhuang et al., 2022), curbing corporate environmental violations (Khalid et al., 2024), and improving energy utilization efficiency (Yang and Masron, 2022). Against this broad backdrop, studies further focused on its core role in carbon emissions. Most scholars argued that digital finance could significantly reduce carbon emission intensity (Lu et al., 2023), primarily through pathways such as the promotion of green technology innovation and industrial structure upgrading (Ding et al., 2022; Shao et al., 2022). At the same time, research also noted the complexity of its effects, including potential spillover effects such as carbon leakage (Yan et al., 2023) or urban-rural inequality (Feng S. et al., 2024). Additionally, the research perspective deepened, shifting from “carbon emission quantity” to the more comprehensive “carbon emission efficiency”. A limited number of pioneering studies began to explore the direct impact of digital finance on carbon emission efficiency. They found that it could enhance efficiency by strengthening environmental regulations (Ding et al., 2023) or synergizing with green innovation (Zhang et al., 2025). However, most literature adopted a linear analytical framework, whereas the impact of digital finance on carbon emission efficiency might exhibit nonlinear dynamic characteristics—a topic that remained relatively understudied. Furthermore, existing research perspectives were largely confined to technological innovation and policy regulatory dimensions, and failed to systematically examine the potential transmission role of social forces such as public environmental awareness.

3 Theoretical mechanism and research hypotheses

3.1 Nonlinear relationship

In the nascent development phase of DF, its environmental impact is predominantly characterized by the expansion of scale. Firstly, from the perspective of the technology adoption lifecycle, the construction and operation of digital financial infrastructure are accompanied by significant energy consumption (Uddin and Rahman, 2012). Because the application of the technology is not yet mature and system optimization remains insufficient, the carbon emission intensity per unit of financial output remains at a relatively high level. This observation aligns with the technological innovation diffusion theory, which posits that efficiency losses are common during the early stages of technology adoption. According to this theory, emerging technologies often struggle to achieve optimal energy efficiency in their initial commercialization phase. Secondly, within the theoretical framework of information economics, digital financial platforms in their early development stage face severe information asymmetry issues (Akerlof, 1978). The absence of unified environmental information disclosure standards and a reliable green assessment system prevents market mechanisms from effectively identifying and pricing environmental externalities. This institutional deficiency leads to a dual market failure: On one hand, high-carbon projects may secure financing due to their observable short-term financial advantages (Le et al., 2020). On the other hand, investments in clean technologies, which carry positive long-term environmental externalities, are often hindered by the underdeveloped evaluation framework. This distortion in resource allocation ultimately manifests as a decline in overall CEE.

However, as information technology becomes widely adopted and DF reaches a more advanced stage, its environmental impact shifts towards significantly enhancing carbon emission efficiency (CEE), primarily by reducing resource consumption and transaction costs. This process is closely linked to the structural and technological effects described by the Environmental Kuznets Curve (EKC). From the perspective of the structural effect, the integration of DF with technologies such as artificial intelligence and blockchain effectively lowers the financing costs for green projects (Wu and Huang, 2022). This directs capital toward low-carbon sectors like clean energy and carbon capture. Consequently, it facilitates the transition to industrial structures characterized by lower pollution and emissions, generates economies of scale, and accelerates the green transformation of the economic framework. In terms of the technological effect, DF leverages online financial service systems and data-driven business models to substantially cut energy and resource consumption in financial operations. This is achieved through paperless transaction processes, intelligent credit approval algorithms, and optimized allocation of distributed computing resources. These advancements enhance total-factor CEE (Zhou et al., 2024), demonstrating a direct contribution of technology-enabled solutions to emission reduction. Furthermore, DF stimulates the development of green financial products and services such as green credit, green securities, and sustainable investments (Javeed et al., 2024). It also fosters the creation of market-based instruments, including carbon futures and the securitization of carbon assets. By generating accurate price signals, these innovations incentivize businesses to adopt clean technologies and optimize production processes, thereby synergistically improving CEE through both structural and technological pathways.

Hypothesis 1

Hypothesis 1. The impact of DF on CEE follows a U-shaped pattern, first suppressing and then promoting.

3.2 Mechanisms

Existing research on DF and CEE primarily focuses on policy and technological pathways such as environmental regulation, resource allocation, and green innovation, while the mechanisms in social dimensions remain underexplored. To address this gap, this study constructs a systematic analytical framework based on the Multi-Level Perspective (MLP) theory (Geels, 2005) to explain the mechanism through which DF affects CEE from three levels: technological niches (micro), socio-technical regimes (meso), and macro-environment (macro). At the micro level, technological niches serve as nurturing spaces for emerging technologies, aligning with the development needs of GTI. At the meso level, the rules, skills, and infrastructure contained in the socio-technical regime collectively manifest as the accumulation and transformation of HC. At the macro level, PEA embodies cultural values in the social dimension. Therefore, this study selects three key mechanisms—GTI, HC, and PEA—to systematically examine the impact mechanism of DF on CEE.

3.2.1 GTI mechanism

GTI faces financing constraints due to its long investment cycles, high risks, and information asymmetry (Hottenrott and Peters, 2012). On the one hand, the development of digital infrastructure requires substantial concentrated funding (Xie et al., 2024), which directly diverts financial resources that might otherwise have been allocated to green technology research and development. This competition for resources creates a “crowding-out effect” on green innovation projects in the capital markets. On the other hand, in the early stages of DF development, its risk assessment models primarily focus on traditional financial metrics (Lin and Ma, 2022), relatively overlooking the long-term environmental benefits and potential value of GTI. As a result, financial support for such projects remains relatively limited. However, as digital financial systems mature, their role in promoting green innovation becomes increasingly evident, generating significant “network effects” through improved resource allocation efficiency (Sun et al., 2024). Firstly, digital technology-based risk assessment systems mitigate information asymmetry (Lin and Ma, 2022), directing funds precisely to green innovation fields and reducing resource misallocation risks. Secondly, by leveraging big data analytics capabilities, DF achieves efficient matching between investor risk preferences and the characteristics of green projects (Feng et al., 2022). This directly enhances the identification and pricing of green innovation initiatives and significantly improves the conversion efficiency of innovative outcomes. Finally, through technological support such as blockchain, DF facilitates intellectual property transactions, directly reducing the commercialization risks associated with green technologies.

Moreover, green innovation directly reduces pollution and carbon emissions by improving production processes, reducing coal dependence, and increasing resource utilization efficiency (Zhao et al., 2024). Furthermore, it drives industrial restructuring toward cleaner and higher value-added directions (Xu et al., 2021), enhancing CEE at the source.

Hypothesis 2

Hypothesis 2. DF has a U-shaped impact (first suppressing then promoting) on GTI, thereby indirectly affecting CEE.

3.2.2 HC mechanism

During the initial stage of DF development, its technological substitution effect directly impacts the job structure within traditional financial business models (Balsmeier and Woerter, 2019), leading to unemployment risks and declining incomes for laborers with traditional skills, thereby temporarily suppressing the overall level of HC. However, as DF matures, its HC enhancement effects gradually emerge. Firstly, DF platforms integrate big data analytics and AI to establish intelligent talent assessment and matching systems, which improves the precision of human resource allocation across industries and roles. Secondly, the advancement of DF fosters diversified education and training systems, including online education resources and expanded career development pathways, effectively upgrading workforce skills and adaptability to industrial restructuring (Feng R. et al., 2024). Thirdly, DF provides robust support for entrepreneurial activities by lowering the barriers to startup financing and creating new types of job opportunities (Liu et al., 2022). These enterprises are more willing and capable of increasing investment in training employees’ environmental protection skills and capabilities in applying green technologies, thereby systematically enhancing HC quality.

Additionally, high-quality HC, which possesses professional knowledge, technical application capabilities, rapid learning ability, and systemic coordination skills, significantly enhances CEE (Zhen et al., 2024). Skilled labor also demonstrates stronger adaptability to digital technologies, bridging application gaps and further improving carbon efficiency (Li et al., 2023).

Hypothesis 3

Hypothesis 3. DF has a U-shaped impact (first suppressing then promoting) on HC, thereby indirectly affecting CEE.

3.2.3 PEA mechanism

In the early stages of DF, its instrumental characteristics, like lowering consumption thresholds and improving borrowing convenience (Li et al., 2020), may trigger irrational consumption behaviors (e.g., fast fashion, frequent electronics upgrades). This reinforces an instant gratification psychology, which conflicts with the long-term benefits of environmental actions and thereby suppresses PEA. As digital financial systems mature, their role in promoting environmental awareness becomes evident. Firstly, innovative products like digital carbon accounts and “Ant Forest” visually quantify low-carbon behaviors (Shao and Xu, 2023), enhancing public perception of environmental actions. DF utilizes technological means to directly embed environmental awareness into the moment of consumer decision-making. For example, payment tools can integrate features that display a product’s estimated carbon footprint or a company’s environmental rating when users scan QR codes to pay. This provides consumers with immediate, quantifiable information on environmental costs, transforming abstract eco-friendly concepts into tangible parameters for purchasing decisions and guiding users toward more environmentally conscious choices. Secondly, incentive policies implemented by DF platforms, such as Shenzhen’s “Carbon Coin” system, convert environmentally friendly behaviors into economic benefits. By linking green consumption actions to broader point, discount, or coupon systems, these policies deliver immediate positive feedback at the point of payment, directly enhancing public motivation to participate in low-carbon activities (Li F. et al., 2024). Lastly, technologies inherent to DF, such as blockchain, offer immutability and traceability, significantly enhancing the credibility of environmental data. This technology-enabled trust forms the foundation for effectively mobilizing and sustaining PEA.

Furthermore, heightened PEA strengthens social supervision, pressuring polluting firms to improve environmental governance (Zhang et al., 2022b), while accelerating market penetration of energy-saving products (Jakučionytė-Skodienė and Liobikienė, 2023), collectively enhancing CEE from both production and consumption perspectives.

Hypothesis 4

Hypothesis 4. DF has a U-shaped impact (first suppressing then promoting) on PEA, thereby indirectly affecting CEE.

4 Data sources and model construction

4.1 Dependent variable

DF: This study primarily follows the approach of Li Z. et al. (2024) by utilizing the Peking University Digital Financial Inclusion Index, divided by 100 to mitigate heteroscedasticity. This index, jointly compiled by the Institute of DF at Peking University and Ant Group Research Institute, is authoritative and widely recognized, providing a comprehensive measure of DF development across various levels in China (see Supplementary Appendix S8).

4.2 Independent variable

CEE: Building on Gao et al. (2021) methodology, we adopt the Super-SBM model incorporating undesirable outputs to quantify CEE (Equation 1). Unlike unidimensional metrics like aggregate emissions or carbon intensity (Vujović et al., 2018), this input-output based efficiency measure offers a comprehensive evaluation of regional low-carbon transition effectiveness (see Supplementary Appendix S9). The input factors include: (1) Capital input, deflated to the base year 2006 using the fixed asset investment index and calculated via the perpetual inventory method. (2) Labor input, measured by the number of employed persons at the prefecture level (including urban units and private sector workers). (3) Energy input, estimated based on total electricity consumption, natural gas supply, and liquefied petroleum gas supply, converted into standard coal equivalents. The desirable output is real GDP deflated to 2006 constant prices, while the undesirable output is the total energy-related carbon emissions at the city level.

${\rho }_{kt}=\mathit{min}\frac{\frac{1}{m}{\displaystyle {\sum }_{i=1}^m}\frac{\overline{x_i}}{x_{ik}}}{\frac{1}{s_1+s_2}\left({\displaystyle {\sum }_{r=1}^{s_1}}\frac{\overline{y^g}}{y_{rk}^g}+{\displaystyle {\sum }_{q=1}^{s_2}}\frac{\overline{y^b}}{y_{qk}^b}\right)},s.t.\left\{\begin{array}{c}\overline{x}\ge {\displaystyle \sum _{j=1,\ne k}^n}{\theta }_j{x}_{ij}\\ \overline{y^g}\le {\displaystyle \sum _{j=1,\ne k}^n}{\theta }_j{y}_{rj}^g\\ \overline{y^b}\ge {\displaystyle \sum _{j=1,\ne k}^n}{\theta }_j{y}_{qj}^b\\ \overline{x}\ge x_k,\overline{y^g}\le \overline{y^b},\overline{y^g}\ge 0,\theta \ge 0\end{array}\right.(1)$

In Equation 1, ${\rho }_{kt}$ represents the CEE of cities, $n$ denotes the number of DMUs, each DMU has $m$ inputs, $S_1$ desirable outputs, and $S_2$ undesirable outputs, while $y_{rk}^g$ , $y_{qk}^b,\hspace{0.5em}{x}_{ik}$ and ${\theta }_j$ are the correlation coefficients of the expected output of $r^{th}$ , the unexpected output of $q^{th}$ , the input of $i^{th}$ , and the linear combination of $j^{th}$ .

4.3 Control variables

The following control variables are chosen to take into consideration other possible impacting factors. (1) Economic development (ED): The logarithm of GDP per capita (Sun and Huang, 2020), capturing its direct impact on resource allocation efficiency. Regional economic development determines resource endowments, industrial structure, and the capacity for green technology investment, while also profoundly affecting the penetration depth and development model of DF. It serves as a key macro-level condition influencing CEE. (2) Population density (PD): The logarithm of population per administrative area (Lu et al., 2023). Agglomeration effects can not only reduce the marginal cost of digital financial service provision and promote its scalable development but may also influence regional energy consumption patterns and CEE through knowledge spillovers and concentrated transportation demand. (3) Industrial development level (IDL): The proportion of secondary industry value-added in GDP. The industrial sector is typically a major source of energy consumption and carbon emissions, and its scale directly determines the baseline emission intensity of a region. Additionally, the industrial structure influences the demand for both traditional and digital financial services. (4) Financial development level (FDL): The ratio of year-end financial institution loan balances to GDP (Ma et al., 2025). A mature financial system serves as an important foundation for the development of DF, and the two may exhibit either synergistic or competitive relationships. Meanwhile, the overall level of financial development itself affects GTI and polluting investment decisions through capital allocation efficiency. (5) Openness: The proportion of total imports and exports to GDP (Yi et al., 2022). International trade influences the environmental performance of host countries through multiple channels such as technology spillovers, competition effects, and the “pollution haven” effect. Simultaneously, an open economy also shapes emerging sectors such as digital financial services.

4.4 Mediating variables

(1) GTI: Following Li et al. (2022a), this study measures it by summing the number of green invention patent applications and green utility model patent applications. Patent data objectively reflect regional green technology R&D capabilities and comprehensively capture innovation activities by including both invention and utility model patents. (2) HC level: Drawing on Yuan and Zhu (2018), it is measured by the proportion of regular higher education institution enrollments to the registered population. College students represent the core source of future high-quality labor, making this indicator an effective proxy for long-term regional HC reserves. The higher education community possesses not only professional expertise and learning capabilities but also serves as a critical vehicle for disseminating green technologies and low-carbon concepts. A higher proportion of this demographic indicates a richer reserve of high-quality talent and a stronger foundation for innovation capacity in a region’s transition towards a green economy. (3) PEA: Following Zhou and Ding (2023), it is measured by aggregating Baidu search indices (PC and mobile) using “environmental pollution” as the keyword (PEA1). As China’s most mainstream and widely used search engine, Baidu broadly reflects real-time public attention and genuine information needs. Its search data exhibit notable timeliness and representativeness, effectively overcoming the limitations of traditional survey methods in terms of timeliness and potential subjective bias. The Baidu search index not only captures proactive public attention and information-seeking behavior regarding environmental pollution issues but also reflects societal-level environmental anxiety and potential engagement intentions, serving as a valid proxy for social environmental awareness and latent pressure for green governance. Furthermore, following Hu et al. (2023), this study constructs a second measure (PEA2) by collecting annual public message data from the environmental protection section of the People’s Daily Online Leadership Message Board and applying a logarithmic transformation. This measure aims to characterize public behavior in expressing environmental concerns through formal channels. Together, PEA1 and PEA2 measure public environmental concern from the dimensions of daily, latent attention and formal, active expression, respectively, thereby enhancing the robustness of the findings.

4.5 Data sources and descriptive statistics

Given that around 2011, third-party mobile payment platforms represented by Alipay and WeChat Pay entered a phase of rapid commercialization and widespread adoption, with transaction volumes experiencing explosive growth, this period marks a transition from initial exploration to the new stage of scaled and ecosystem-driven development in digital finance. Therefore, selecting 2011 as the sample start year effectively captures the complete developmental trajectory of digital finance as a core explanatory variable, from its inception to its deepening impact. This approach avoids including a large number of “zero-value” or “low-variation” observations that would arise from an earlier sample period, thereby enabling a more precise identification of its true impact on carbon emission efficiency. The dataset used in this study primarily consists of panel data from 282 prefecture-level cities in China covering the period from 2011 to 2022. Data were mainly sourced from the National Intellectual Property Administration, China City Statistical Yearbook, China Urban-Rural Construction Statistical Yearbook, the official website of the National Bureau of Statistics, and the Institute of DF at Peking University. Data for both PEA proxies were manually collected: the Baidu search index (PEA1) and the People’s Daily Online Leadership Message Board (PEA2). Table 1 presents the descriptive statistics of the variables.

Table 1

Table 1. Descriptive statistics.

4.6 Models

4.6.1 Baseline model

Drawing on the classic EKC theoretical framework in environmental economics and the research model of Li Z. et al. (2024), we selected DF as the core explanatory variable and constructed a two-way fixed effects model incorporating both linear and quadratic terms of DF:

${\text{CEE}}_{\text{it}}={\mathrm{\alpha }}_0+{\mathrm{\alpha }}_1{\text{DF}}_{\text{it}}+{\mathrm{\alpha }}_2{\text{DF}}_{\text{it}}^2+{\mathrm{\alpha }}_3{C}_{it}+{\mathrm{\mu }}_{\mathrm{i}}+{\mathrm{\lambda }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\text{it}}(2)$

In Equation 2, $i$ denotes the city and $t$ denotes the year, $C_{it}$ is a vector of control variables, ${\mathrm{\mu }}_{\mathrm{i}}$ captures city fixed effects, ${\mathrm{\lambda }}_{\mathrm{t}}$ captures year fixed effects, ${\mathrm{\epsilon }}_{\text{it}}$ is the random error term, If ${\mathrm{\alpha }}_1$ is negative and ${\mathrm{\alpha }}_2$ is positive, this would confirm Hypothesis 1.

4.6.2 Mechanism analysis model

To examine the transmission channels through which DF affects CEE, we followed Jiang (2022) and established the following mediation model:

${\mathrm{M}}_{\text{it}}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1{\text{DF}}_{\text{it}}+{\mathrm{\beta }}_2{\text{DF}}_{\text{it}}^2+{\mathrm{\beta }}_3{\mathrm{C}}_{\text{it}}+{\mathrm{\mu }}_{\mathrm{i}}+{\mathrm{\lambda }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\text{it}}(3)$

In Equation 3, ${\mathrm{M}}_{\text{it}}$ represents the mediating variables, which specifically include GTI, HC level, and PEA. All other variables remain consistent with those in Model (2).

4.6.3 Moderating effect model

To further examine the potential moderating role of external policy shocks in the process through which DF affects CEE, the following moderating effect model is established:

${\text{CEE}}_{\text{it}}={\mathrm{\delta }}_0+{\mathrm{\delta }}_1{\text{DF}}_{\text{it}}+{\mathrm{\delta }}_2{\text{DF}}_{\text{it}}^2+{\mathrm{\delta }}_3\text{policy}{+{\mathrm{\delta }}_4\text{policy}·{\text{DF}}_{\text{it}}+{\mathrm{\delta }}_5\text{policy}·{\text{DF}}_{\text{it}}^2+{\mathrm{\delta }}_6\mathrm{C}}_{\text{it}}+{\mathrm{\mu }}_{\mathrm{i}}+{\mathrm{\lambda }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\text{it}}(4)$

In Equation 4, $policy$ represents the policy variable, $policy·{DF}_{it}$ and $policy·{DF}_{it}^2$ denote the interaction terms between the policy and the linear and quadratic terms of DF, respectively. All other variables remain consistent with those in Model (2).

5 Empirical results analysis

Based on preliminary fitting results, a “ U-shaped” relationship is observed between DF and CEE (Figure 3). To draw more robust conclusions, it is necessary to further validate this relationship through rigorous regression analysis.

Figure 3

Figure 3. DF and CEE fitting line.

5.1 Benchmark regression results analysis

Table 2 reveals statistically significant coefficients for DF (negative, p < 0.01) and DF² (positive, p < 0.01), confirming a U-shaped association between DF and CEE. This nonlinear relationship remains robust as additional control variables are gradually introduced. This finding, building upon the linear relationship identified by Zhang and Liu (2022), further reveals a dynamic non-linear effect between DF and CEE characterized by initial inhibition followed by subsequent promotion, thereby deepening the understanding of the complex mechanism between the two. This pattern arises because, in the early stages of DF, services primarily cater to traditional economic activities, and the environmental benefits of technological applications are not yet fully realized. Initial business models and product innovations focus on conventional economic activities, such as retail consumption and routine transfers (Wang et al., 2024). Digital credit primarily supports the financing needs of enterprises in traditional manufacturing and service sectors, with relatively little consideration for environmental benefits. As DF matures, it leverages big data to precisely match green credit allocations, optimizing resource allocation efficiency (He et al., 2022). DF institutions can identify firms with strong environmental performance and green development potential, offering favorable credit terms and greater financing support to green projects (Javeed et al., 2024), thereby promoting corporate green development and effectively enhancing CEE.

Table 2

Table 2. Benchmark regression results.

Regarding the control variables, economic development: The coefficient is significantly negative, as local officials tend to allocate resources to visible economic growth metrics to demonstrate political achievements quickly, while neglecting less tangible environmental performance metrics like CEE (Shen et al., 2021). Openness: The coefficient is significantly negative, likely due to foreign firms relocating high-pollution, high-energy-consumption industries to China (Repkine and Min, 2020), suppressing CEE and creating a “pollution haven” effect. Population density: The coefficient is significantly negative, reflecting that elevated population concentration intensifies both economic operations and residential consumption, consequently amplifying energy requirements and environmental pollutants (Lu et al., 2023). Financial development: The coefficient is positive, as financial development improves capital allocation efficiency, driving the transition to technology-intensive industries (Yang and Ni, 2022) and enhancing energy use efficiency while reducing emission intensity. Industrial development level: The coefficient exhibits a negative association, reflecting the industrial sector’s substantial role in carbon emissions (Lu et al., 2023).

5.2 Robustness tests

To ensure the robustness of the findings, several tests were conducted: 1. Alternative dependent variable: Following (Charnes et al., 1978), CEE was recalculated using the CCR model. 2. Exclusion of municipalities: Given the unique economic, infrastructural, and social characteristics of municipalities, the four direct-administered municipalities were excluded³. 3. Winsorization: To mitigate outlier effects, a 1% two-sided winsorization was applied to the panel data. 4. Lagged control variables: To address potential endogeneity, all five control variables were lagged by one period (Li F. et al., 2024). 5. Logarithmic transformation: A logarithmic transformation was applied to the openness indicator to mitigate its fluctuations. Table 3 Column (1)–(5) results remain consistent with the benchmark results.

Table 3

Table 3. Robustness test results.

5.3 Endogeneity test

Firstly, following Guo et al. (2023) and Yang et al. (2024), we used the geographic distance from sample cities to Hangzhou as an instrumental variable. Hangzhou is the birthplace and global headquarters of Ant Group and its predecessor, Alipay. As the pioneer of China’s mobile payment and digital financial services, Alipay’s development history is nearly synchronous with the evolution of DF in China. Consequently, Hangzhou has gathered the nation’s most concentrated elements of DF technology, talent, data, and capital, forming a powerful industrial cluster with significant spillover effects. The geographical distance from each city to Hangzhou intuitively measures the convenience and cost of accessing technological spillovers and market influences from this core radiation source. Shorter distances are associated with higher efficiency, earlier penetration, and deeper adoption of digital financial technologies, which ensures the relevance of the instrumental variable to the level of urban digital financial development. Meanwhile, this instrumental variable satisfies the exogeneity condition. A city’s geographic location is historically determined and thus strictly exogenous, unaffected by contemporary economic fluctuations or CEE changes during the sample period. Moreover, geographic distance is unlikely to directly affect contemporary CEE; its influence should primarily operate through DF penetration. This satisfies the exclusion restriction for a valid IV. Considering the cross-sectional nature of distance data, we constructed an interaction term between this distance variable and the logarithm of the previous year’s national internet penetration rate. Secondly, drawing on Liu (2023), we used the one-period lag of DF as an instrumental variable. Under controlled variables and fixed effects, the lagged term of DF is uncorrelated with current CEE while maintaining time-series autocorrelation, satisfying both the relevance and exogeneity conditions for instrumental variables.

Table 4 columns (1)–(2) present results using distance to Hangzhou as the instrumental variable, while columns (3)–(4) use the one-period lag of DF as the instrumental variable. The results indicate that after accounting for endogeneity, DF still exhibits a significant U-shaped relationship with CEE. The Kleibergen-Paap rk Wald F-statistics exceed the 10% critical values of the Stock-Yogo weak identification test, suggesting no significant weak instrument problem. This collectively demonstrates the appropriateness of both instrumental variable selections.

Table 4

Table 4. IV tests results.

5.4 Heterogeneity analysis

Considering the significant disparities in the development levels of different dimensions of DF across cities, this study focuses on three key aspects—coverage breadth, usage depth, and digitization level—to analyze the heterogeneous impact of various dimensions of DF on CEE (Table 5 Column (1)–(3)). In terms of coverage breadth, DF exhibits a significant U-shaped relationship with CEE (coefficients: −0.324 and 0.064). This indicates that in the initial expansion phase, the digitization of traditional economic activities may not immediately lead to emission reductions and could even be accompanied by increased energy consumption. However, once a certain threshold of coverage is achieved, network effects and scalable green services come into play, significantly enhancing CEE. Regarding usage depth, a similarly significant U-shaped relationship is observed between the application depth of DF and CEE (coefficients: −0.177 and 0.054). This suggests that early applications were largely concentrated in non-green areas such as consumer credit, offering limited emission reduction benefits. As application scenarios extend into green finance, carbon trading, and energy management, DF contributes markedly to improving CEE by optimizing resource allocation and facilitating technological innovation.

Table 5

Table 5. Heterogeneity test results.

The statistically insignificant yet positive coefficient for digitization (0.035 and 0.0005) reflects not its ineffectiveness but the conditional and complex nature of its impact on CEE. This can be explained systematically. Firstly, measurement limitations. Common proxies (e.g., internet penetration) capture general infrastructure but not the targeted application of digital technologies for carbon reduction, likely leading to an underestimation of their true effect. For example, Hangzhou’s “City Brain” reduces emissions primarily through its precise “smart transportation” module, not general broadband coverage. Secondly, policy dependency. Digital technologies are value-neutral; their green impact depends critically on supportive regulations. In Zhangjiakou, a renewable energy zone, regulations mandate that data centers use local wind and solar power, ensuring digital construction absorbs green electricity. Without such “command-and-control” policies, digital infrastructure may instead connect to coal-heavy grids, increasing emissions. Thus, digitization’s benefits are not automatic but policy-contingent. Thirdly, developmental heterogeneity. Benefits realization involves significant time lags and varies by stage. During our sample, cities ranged from the high-energy phase of data center construction to reaping rewards from smart energy management. This heterogeneity in timing and intensity results in an unstable average effect in pooled regression, explaining the statistical insignificance.

From the resource endowment perspective, the research sample was stratified along two dimensions⁴: (a) resource-based cities contrasted with non-resource-based cities, and (b) key environmental protection cities and non-key environmental protection cities. Columns (4)–(7) show that in resource-based cities and key environmental protection cities, the U-shaped relationship between DF and CEE appears steeper. These cities generally suffer from the “resource curse” effect with more rigid traditional industrial structures, where DF introduction can more significantly break technological lock-in effects (Xu and Cai, 2024). After passing the development inflection point, DF significantly promotes GTI, particularly in key areas such as clean energy and pollution control. Moreover, given the more stringent environmental constraints and transformation pressures faced by resource-based and key environmental protection cities, local governments often actively integrate DF into their industrial green upgrading policy systems. Through targeted subsidies, guidance for green credit, and collaborative development of digital infrastructure, they enhance the synergistic effect between DF and the transformation of traditional industries, thereby significantly amplifying emission reduction outcomes.

5.5 Mechanism analysis

Table 6 reports test results for the mechanisms through which DF affects CEE. Columns 1–4 show that DF has U-shaped effects on GTI, HC level, and PEA, indirectly influencing CEE, thus confirming Hypotheses 2, 3, and 4.

Table 6

Table 6. Mechanism test.

First, in terms of GTI, the U-shaped impact of DF highly aligns with the evolution of China’s green financial policies and market maturity. In the early stage of development, although DF platforms (such as Alipay and WeChat Pay) became widespread rapidly, their business focus was on payment convenience and consumer credit expansion. Platform algorithms and risk control models did not incorporate environmental benefits as core indicators. In addition, early green projects were characterized by long investment cycles, difficult risk identification, and low short-term returns, which naturally mismatched with digital capital pursuing high liquidity. As a result, it failed to effectively direct resources toward green R&D. As the DF system gradually improved, it significantly promoted GTI by optimizing financing channels for green projects and reducing information asymmetry in innovation activities, effectively shortening R&D cycles (Lin and Ma, 2022). Large technology companies have begun to actively expand their presence: Ping An Bank has built an “integrated green financial services platform”, using AI to dynamically rate corporate carbon emission data and linking it to loan interest rates, accurately reducing financing costs for green technology companies. This model of “policy guidance + digital risk control” has effectively solved the financing constraints faced before green innovation and has become a key driving force pushing GTI over the inflection point of the U-shaped curve.

Secondly, in terms of HC, the early stage of DF development exhibited uneven technology penetration (Chen et al., 2023). Residents in remote areas and elderly groups were excluded from green skill training systems due to insufficient digital access capabilities, making it difficult for them to meet the demands of the new wave of industrial green upgrading, thereby constraining the overall improvement of HC quality. This phenomenon somewhat contradicts China’s strategic direction of promoting “common prosperity” and “inclusive finance.” However, with the improvement of digital infrastructure, DF platforms have demonstrated strong potential for educational empowerment. For example, Alibaba’s “Inclusive Education Initiative” provides vocational institutions in small and medium-sized cities with access to course resources essential for the green economy, such as artificial intelligence and big data analysis, through its cloud platform; WeBank, in collaboration with local governments, has carried out “Green Finance Rural Outreach” activities in Guizhou, Gansu, and other regions, using mobile apps to provide skill training and certification for rural laborers in areas such as photovoltaic operation and maintenance and energy efficiency management. These initiatives not only enhance the green skill reserves of the labor force but also, through the positive cycle of “digital capabilities—green employment—income increase,” provide large-scale and specialized HC support for the economic green transition.

Finally, in terms of public environmental attention (PEA1 and PEA2), in the initial stage, platforms primarily promoted promotional and entertainment content to pursue user engagement and transaction volume, causing environmental information to be drowned in a flood of commercial traffic. This made it difficult for the public to form effective low-carbon awareness. In addition, the low visibility of users’ personal carbon footprints and the lack of immediate feedback on carbon reduction actions hindered the activation of green consumption willingness. With the deep integration of platform strategies and the national “Green Consumption” initiative, significant changes have taken place. For example, Alipay’s “Ant Forest” translates users’ low-carbon behaviors (such as taking public transport and making online payments) into virtual “green energy” through digital financial accounts, which is then matched with real tree-planting ecological projects. This greatly enhances the sense of achievement and visibility of public participation. The project has attracted hundreds of millions of users since its launch, successfully transforming abstract environmental concepts into daily practices. In addition, UnionPay’s QuickPass has integrated with multiple municipal carbon-inclusive platforms, allowing users to accumulate carbon points through green consumption and exchange them for benefits. These designs leverage the account systems and incentive feedback mechanisms of DF to effectively shape public green preferences (Zhang and Liu, 2022) and drive a societal wave of low-carbon transformation on the consumption side.

5.6 Further discussion: Policy impacts

Although existing research has extensively discussed that DF may have a direct impact on CEE by improving capital allocation efficiency and supporting GTI, it has generally overlooked a critical issue: the emissions reduction effectiveness of DF is inevitably profoundly influenced by the external policy environment. At the current critical period when China is accelerating the coordinated implementation of the “Digital China” and “Dual Carbon” strategies, the Chinese government does not rely on a single policy but has constructed a multi-level, multi-dimensional policy experimentation system. This system includes policy instruments such as CETPP (launched successively since 2011), LCCPP (launched successively since 2010), “Smart City” Pilot Policy (SCPP) (implemented in multiple batches since 2012), and “Broadband China” pilot policy (BCPP) (launched successively since 2014). These policies provide support for the green and low-carbon transition from different dimensions: SCPP and BCPP focus on digital infrastructure upgrading and intelligent technology penetration, providing key technological empowerment for the emissions reduction process; CETPP and LCCPP emphasize the use of market-based carbon pricing mechanisms, institutional innovation, and capacity building. This study selects CETPP, LCCPP, SCPP, and BCPP as key moderating variables, focusing on analyzing their moderating roles in the process of DF affecting CEE.

In Table 7 columns (1)–(2), the interaction terms between pilot policies and DF² are significantly positive, indicating CETPP and LCCPP further strengthens the U-shaped relationship between DF and CEE, making the curve steeper. Specifically, after the inflection point, DF’s promoting effect on CEE is stronger in policy pilot areas. This is because in low-carbon policy pilot areas, digital financial institutions impose stricter credit constraints on high-pollution, high-energy-consumption enterprises during credit approval processes, while creating green approval channels and providing priority financing support for green enterprises, thereby strengthening DF’s green development attributes (Wang et al., 2023). Column (3) shows SCPP makes the U-shaped relationship between DF and CEE flatter but insignificant, reflecting that SCPP indirectly affects DF development by improving overall digital governance efficiency but lacks targeted guidance for the financial sector’s green transformation. Column (4) indicates BCPP makes the U-shaped relationship steeper but insignificant, possibly because digital infrastructure development created favorable conditions for DF growth but lacked supporting environmental regulation policies to effectively guide DF’s emission reduction role.

Table 7

Table 7. Policy adjustment results.

6 Main research findings and implications

6.1 Conclusions

DF has injected new vitality into China’s financial development and demonstrated unique advantages in coordinating economic development with environmental protection. Based on panel data from 282 Chinese cities (2011–2022) and employing econometric methods including panel fixed-effects models and instrumental variable approaches, this study systematically examines the impact of DF on CEE. The findings reveal: (1) The impact of DF on CEE exhibits a distinct “inhibit first, promote later” U-shaped characteristic. In the early stage of development, due to suboptimal resource allocation, immature technology application, and an incomplete regulatory framework, DF may somewhat inhibit the improvement of CEE. However, once its development level crosses the inflection point, DF significantly promotes the improvement of CEE by enhancing financial service efficiency, expanding green financing channels, and strengthening environmental risk management capabilities. This U-shaped relationship remains robust after addressing endogeneity and undergoing multiple rigorous tests, indicating high reliability of the conclusion. Heterogeneity analysis shows that the U-shaped relationship is more pronounced in the dimensions of coverage breadth and usage depth, as well as in resource-based cities and key environmental protection cities. (2) Mechanism analysis reveals that DF has a U-shaped effect on GTI, HC, and PEA, thereby indirectly influencing CEE. Initially, DF may fail to effectively drive the green transition due to insufficient technology spillover and limited public awareness. As it develops more deeply, DF significantly promotes the research, development, and application of green technologies, enhances the green skills of the labor force, and strengthens societal environmental awareness, thus indirectly yet crucially driving the improvement of CEE. (3) Tests on policy synergy effects show that the CETPP and LCCPP can effectively strengthen the U-shaped relationship between DF and CEE, indicating that combining DF with well-defined environmental policies can achieve better emission reduction outcomes. In contrast, digital and infrastructure-oriented policies, represented by SCPP and BCPP, did not exhibit significant effects, reflecting the differential impacts of various types of policies in synergistically promoting green and low-carbon development. This finding offers important insights for improving the green financial policy system in the future.

6.2 Policy recommendations

To enhance the role of digital finance in promoting carbon emission efficiency, it is essential to establish a multi-stakeholder, multi-mechanism collaborative policy framework. This framework should systematically advance institutional design, market incentives, and public participation.

1. Strengthen institutional design to build a “carbon-sensitive” digital financial infrastructure. A linkage mechanism between digital finance and carbon accounting should be established. Under the leadership of financial regulators, a “carbon efficiency label” system should be integrated into mainstream digital platforms. This system would assess the lifecycle carbon efficiency of financial products, publicly disclose their “carbon impact values,” and thereby steer capital toward low-carbon sectors. Moreover, intelligent coordination between digital finance and environmental regulations must be enhanced. Specifically, corporate green performance on digital platforms should be incorporated into environmental credit evaluation systems. This would enable preferential green credit channels for high-performing enterprises while restricting digital financial services for persistent high emitters without transition plans.

2. Incentivize corporate green transformation to amplify digital finance’s emission reduction impact. Expand digital financial tools to support green technologies. Blockchain technology can be leveraged to establish integrated platforms for green technology financing and intellectual property trading, streamlining the verification, valuation, and financing of green patents and other intangible assets to alleviate project financing constraints. Meanwhile, establish specialized digital credit mechanisms for low-carbon transitions. Financial institutions should be encouraged to develop differentiated credit products based on corporate carbon data, providing intelligent risk assessment and dynamic credit support for enterprises undertaking energy-saving upgrades and cleaner production. This would reduce the financial costs of green transformation.

3. Deepen public participation to extend digital finance to low-carbon consumption. Create incentive mechanisms linking personal carbon accounts with digital finance. Leverage the central bank digital currency pilot programs to establish personal carbon credit systems. Convert green behaviors such as low-carbon commuting and consumption into accumulative and tradable digital carbon credits, which can be exchanged for digital currency or used to access preferential financial services. Moreover, promote user-end low-carbon guidance tools such as “carbon filters”. Develop and promote “carbon efficiency filtering” features on platforms like Alipay and WeChat Pay, providing users with prioritized displays and recommendations for low-carbon products and services. This will enhance public awareness and willingness to participate in low-carbon choices.

6.3 Limitations

1. This study primarily focuses on analysis at the city level in China, and the universality of the revealed U-shaped relationship and its underlying formation mechanisms faces challenges. The conclusions drawn are rooted in China’s specific policy environment, financial structure, and industrialization process. Further exploration is needed to understand the relationship between DF and CEE on a global scale. Due to systematic differences among countries in financial structures, regulatory frameworks, green technology foundations, and energy mixes, whether the U-shaped relationship and its mechanisms identified in this study apply to other countries and regions requires further verification.

2. This study concentrates on the direct impact of DF on CEE but does not delve into the potential macroeconomic dynamic feedback that may arise after efficiency improvements, particularly the carbon rebound effect. Although improvements in CEE can help reduce carbon emissions per unit of output, they may also trigger a rebound effect due to factors such as lower energy costs and expanded consumer demand, potentially leading to an increase in total carbon emissions. Future research urgently needs to introduce a more macro and dynamic analytical framework that connects micro-level efficiency gains with macro-level total emissions changes, to assess the net emission reduction effect of DF—specifically, to examine whether, while improving efficiency, it may ultimately lead to a rebound in total carbon emissions due to economic scale expansion.

Data availability statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Author contributions

YC: Methodology, Visualization, Conceptualization, Funding acquisition, Software, Validation, Formal Analysis, Project administration, Investigation, Resources, Writing – original draft, Supervision, Writing – review and editing, Data curation.

Funding

The author(s) declared that financial support was received for this work and/or its publication. This project is supported by funding from the Henan Provincial Soft Science Research Program Project: “Research on the Integration Mechanism and Effectiveness Evaluation of Sci-Tech Finance in Promoting High-Quality Economic Development in Henan Province” (242400410220).

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fenvs.2026.1739323/full#supplementary-material

Footnotes

^{Abbreviations:} CEE, Carbon emission efficiency; DF, Digital finance; LCCPP, “Low-carbon city” pilot policy; CETPP, “Carbon emission trading” pilot policy; GTI, Green technology innovation; HC, Human capital; PEA, Public environmental awareness; SCPP, “Smart City” Pilot Policy; BCPP, “Broadband China” pilot policy.

References

Akerlof, G. A. (1978). “The market for “lemons”: quality uncertainty and the market mechanism,” in Uncertainty in economics (Academic Press), 235–251. doi:10.1016/B978-0-12-214850-7.50022-X

CrossRef Full Text | Google Scholar

Balsmeier, B., and Woerter, M. (2019). Is this time different? How digitalization influences job creation and destruction. Res. Policy 48 (8), 103765. doi:10.1016/j.respol.2019.03.010

CrossRef Full Text | Google Scholar

Charnes, A., Cooper, W. W., and Rhodes, E. (1978). Measuring the efficiency of decision making units. Eur. J. Oper. Res. 2 (6), 429–444. doi:10.1016/0377-2217(78)90138-8

CrossRef Full Text | Google Scholar

Chen, M., Chu, Q., Zhang, T., Li, Q., Liu, J., and Yamaka, W. (2023). The structural causes and trend evolution of imbalance and insufficiency of development of digital inclusive finance in China. Sustainability 15 (13), 10286. doi:10.3390/su151310286

CrossRef Full Text | Google Scholar

Ding, X., Gao, L., Wang, G., and Nie, Y. (2022). Can the development of digital financial inclusion curb carbon emissions? Empirical test from spatial perspective. Front. Environ. Sci. 10, 1045878. doi:10.3389/fenvs.2022.1045878

CrossRef Full Text | Google Scholar

Ding, K., Li, J., and Wang, Q. (2023). Digital finance, government intervention, and carbon emission efficiency in China. Environ. Sci. Pollut. Res. 30 (56), 119356–119371. doi:10.1007/s11356-023-30730-2

PubMed Abstract | CrossRef Full Text | Google Scholar

Du, Y., Zhou, J., Bai, J., and Cao, Y. (2023). Breaking the resource curse: the perspective of improving carbon emission efficiency based on digital infrastructure construction. Resour. Policy 85, 103842. doi:10.1016/j.resourpol.2023.103842

CrossRef Full Text | Google Scholar

Fang, G., Gao, Z., Wang, L., and Tian, L. (2022). How does green innovation drive urban carbon emission efficiency? —evidence from the yangtze river economic belt. J. Clean. Prod. 375, 134196. doi:10.1016/j.jclepro.2022.134196

CrossRef Full Text | Google Scholar

Farrell, M. J. (1957). The measurement of productive efficiency. J. R. Stat. Soc. Ser. A 120 (3), 253–290. doi:10.2307/2343100

CrossRef Full Text | Google Scholar

Feng, S., Zhang, R., and Li, G. (2022). Environmental decentralization, digital finance and green technology innovation. Struct. Change Econ. Dyn. 61, 70–83. doi:10.1016/j.strueco.2022.02.008

CrossRef Full Text | Google Scholar

Feng, R., Shen, C., and Guo, Y. (2024). Digital finance and labor demand of manufacturing enterprises: theoretical mechanism and heterogeneity analysis. Int. Rev. Econ. Finance 89, 17–32. doi:10.1016/j.iref.2023.07.065

CrossRef Full Text | Google Scholar

Feng, S., Liu, J., and Xu, D. (2024). Digital financial development and indirect household carbon emissions: empirical evidence from China. Environ. Dev. Sustain. 26 (9), 23401–23435. doi:10.1007/s10668-023-03603-4

CrossRef Full Text | Google Scholar

Gao, P., Yue, S., and Chen, H. (2021). Carbon emission efficiency of China’s industry sectors: from the perspective of embodied carbon emissions. J. Clean. Prod. 283, 124655. doi:10.1016/j.jclepro.2020.124655

CrossRef Full Text | Google Scholar

Geels, F. W. (2005). Processes and patterns in transitions and system innovations: refining the co-evolutionary multi-level perspective. Technol. Forecast. Soc. Change 72 (6), 681–696. doi:10.1016/j.techfore.2004.08.014

CrossRef Full Text | Google Scholar

Guo, D., Qi, F., Wang, R., and Li, L. (2023). How does digital inclusive finance affect the ecological environment? Evidence from Chinese prefecture-level cities. J. Environ. Manage. 342, 118158. doi:10.1016/j.jenvman.2023.118158

PubMed Abstract | CrossRef Full Text | Google Scholar

Han, J., and Jiang, T. (2022). Does the development of the digital economy improve carbon emission efficiency? Front. Ecol. Evol. 10, 1031722. doi:10.3389/fevo.2022.1031722

CrossRef Full Text | Google Scholar

He, F., Wang, M., and Zhou, P. (2022). Evaluation of market risk and resource allocation ability of green credit business by deep learning under internet of things. PLOS ONE 17 (4), e0266674. doi:10.1371/journal.pone.0266674

PubMed Abstract | CrossRef Full Text | Google Scholar

Hottenrott, H., and Peters, B. (2012). Innovative capability and financing constraints for innovation: more money, more innovation?. Rev. Econ. Stat. 94 (4), 1126–1142. doi:10.1162/REST_a_00227

CrossRef Full Text | Google Scholar

Hou, J., Shi, C., Fan, G., and Xu, H. (2024). Research on the impact and intermediary effect of carbon emission trading policy on carbon emission efficiency in China. Atmos. Pollut. Res. 15 (4), 102045. doi:10.1016/j.apr.2024.102045

CrossRef Full Text | Google Scholar

Hu, D., Wang, E., Ye, Q., Chen, S., and Gu, X. (2023). How do online public messages affect local government responsiveness in China? A multilevel analysis based on the message board for leaders. J. Chin. Polit. Sci., 1–32. doi:10.1007/s11366-023-09878-8

CrossRef Full Text | Google Scholar

Jakučionytė-Skodienė, M., and Liobikienė, G. (2023). Changes in energy consumption and CO2 emissions in the Lithuanian household sector caused by environmental awareness and climate change policy. Energy Policy 180, 113687. doi:10.1016/j.enpol.2023.113687

CrossRef Full Text | Google Scholar

Javeed, S. A., Latief, R., Cai, X., and Ong, T. S. (2024). Digital finance and corporate green investment: a perspective from institutional investors and environmental regulations. J. Clean. Prod. 446, 141367. doi:10.1016/j.jclepro.2024.141367

CrossRef Full Text | Google Scholar

Jiang, T. (2022). Mediating effects and moderating effects in empirical research on causal inference. China Ind. Econ. 39 (5), 100–120. doi:10.19581/j.cnki.ciejournal.2022.05.005

CrossRef Full Text | Google Scholar

Kaya, Y., and Yokobori, K. (1997). Environment, energy, and economy: strategies for sustainability, 4. Tokyo, Japan: United Nations University Press.

Google Scholar

Khalid, F., Sun, X., Akram, R., and Srivastava, M. (2024). Digital finance and corporate environmental violations. Finance Res. Lett. 66, 105674. doi:10.1016/j.frl.2024.105674

CrossRef Full Text | Google Scholar

Le, T.-H., Le, H.-C., and Taghizadeh-Hesary, F. (2020). Does financial inclusion impact CO2 emissions? Evidence from Asia. Finance Res. Lett. 34, 101451. doi:10.1016/j.frl.2020.101451

CrossRef Full Text | Google Scholar

Li, J., Wu, Y., and Xiao, J. J. (2020). The impact of digital finance on household consumption: evidence from China. Econ. Model. 86, 317–326. doi:10.1016/j.econmod.2019.09.027

CrossRef Full Text | Google Scholar

Li, G., Xue, Q., and Qin, J. (2022a). Environmental information disclosure and green technology innovation: empirical evidence from China. Technol. Forecast. Soc. Change 176, 121453. doi:10.1016/j.techfore.2021.121453

CrossRef Full Text | Google Scholar

Li, G., Zhang, R., Feng, S., and Wang, Y. (2022b). Digital finance and sustainable development: evidence from environmental inequality in China. Bus. Strategy Environ. 31 (7), 3574–3594. doi:10.1002/bse.3105

CrossRef Full Text | Google Scholar

Li, Z., Bai, T., and Tang, C. (2022). How does the low-carbon city pilot policy affect the synergistic governance efficiency of carbon and smog? quasi-Experimental evidence from China. J. Clean. Prod. 373, 133809. doi:10.1016/j.jclepro.2022.133809

CrossRef Full Text | Google Scholar

Li, G., Wu, H., Jiang, J., and Zong, Q. (2023). Digital finance and the low-carbon energy transition (LCET) from the perspective of capital-biased technical progress. Energy Econ. 120, 106623. doi:10.1016/j.eneco.2023.106623

CrossRef Full Text | Google Scholar

Li, F., Guo, Y., and Liu, B. (2024). Impact of government subsidies and carbon inclusion mechanism on carbon emission reduction and consumption willingness in low-carbon supply chain. J. Clean. Prod. 449, 141783. doi:10.1016/j.jclepro.2024.141783

CrossRef Full Text | Google Scholar

Li, Z., Yuan, B., Wang, Y., Qian, J., and Wu, H. (2024). The role of digital finance on the synergistic governance of pollution & carbon: evidence from Chinese cities. Sustain. Cities Soc. 115, 105812. doi:10.1016/j.scs.2024.105812

CrossRef Full Text | Google Scholar

Li, Z., Wang, Y., and Bai, T. (2025). International digital trade and synergetic control of pollution and carbon emissions: theory and evidence based on a nonlinear framework. J. Environ. Manage. 376, 124450. doi:10.1016/j.jenvman.2025.124450

PubMed Abstract | CrossRef Full Text | Google Scholar

Lin, B., and Ma, R. (2022). How does digital finance influence green technology innovation in China? Evidence from the financing constraints perspective. J. Environ. Manage. 320, 115833. doi:10.1016/j.jenvman.2022.115833

PubMed Abstract | CrossRef Full Text | Google Scholar

Liu, W. (2023). The digital economy and environmental pollution: new evidence based on the support of logistics development. J. Clean. Prod. 427, 139210. doi:10.1016/j.jclepro.2023.139210

CrossRef Full Text | Google Scholar

Liu, S., Koster, S., and Chen, X. (2022). Digital divide or dividend? The impact of digital finance on the migrants’ entrepreneurship in less developed regions of China. Cities 131, 103896. doi:10.1016/j.cities.2022.103896

CrossRef Full Text | Google Scholar

Lu, F., Li, Z., and Zhang, S. (2023). Does digital finance development affect carbon emission intensity: evidence from China. Int. Rev. Econ. Finance 88, 1272–1286. doi:10.1016/j.iref.2023.07.036

CrossRef Full Text | Google Scholar

Ma, Z., Xiao, H., Li, J., Chen, H., and Chen, W. (2025). Study on how the digital economy affects urban carbon emissions. Renew. Sustain. Energy Rev. 207, 114910. doi:10.1016/j.rser.2024.114910

CrossRef Full Text | Google Scholar

Napp, T. A., Gambhir, A., Hills, T. P., Florin, N., and Fennell, P. S. (2014). A review of the technologies, economics and policy instruments for decarbonising energy-intensive manufacturing industries. Renew. Sustain. Energy Rev. 30, 616–640. doi:10.1016/j.rser.2013.10.036

CrossRef Full Text | Google Scholar

Nusa, P. D., Buchori, I., Dewa, D. D., Putri, S. N. A. K., and Pangi, P. (2025). Impact of urbanization on carbon emissions and ecological quality in the semarang metropolitan region, Indonesia. Environ. Monit. Assess. 197 (6), 1–17. doi:10.1007/s10661-025-14057-8

PubMed Abstract | CrossRef Full Text | Google Scholar

Repkine, A., and Min, D. (2020). Foreign-funded enterprises and pollution halo hypothesis: a spatial econometric analysis of thirty Chinese regions. Sustainability 12 (12), 5048. doi:10.3390/su12125048

CrossRef Full Text | Google Scholar

Shao, Z., and Xu, Y. (2023). Moving towards carbon neutral lifestyle through FinTech social media platform: a case study of ant forest. Front. Environ. Sci. 11, 1160986. doi:10.3389/fenvs.2023.1160986

CrossRef Full Text | Google Scholar

Shao, H., Cheng, J., Wang, Y., and Li, X. (2022). Can digital finance promote comprehensive carbon emission performance? Evidence from Chinese cities. Int. J. Environ. Res. Public Health 19 (16), 10255. doi:10.3390/ijerph191610255

PubMed Abstract | CrossRef Full Text | Google Scholar

Shen, F., Liu, B., Luo, F., Wu, C., Chen, H., and Wei, W. (2021). The effect of economic growth target constraints on green technology innovation. J. Environ. Manage. 292, 112765. doi:10.1016/j.jenvman.2021.112765

PubMed Abstract | CrossRef Full Text | Google Scholar

Sun, J. W. (2005). The decrease of CO2 emission intensity is decarbonization at national and global levels. Energy Policy 33 (8), 975–978. doi:10.1016/j.enpol.2003.10.023

CrossRef Full Text | Google Scholar

Sun, W., and Huang, C. (2020). How does urbanization affect carbon emission efficiency? Evidence from China. J. Clean. Prod. 272, 122828. doi:10.1016/j.jclepro.2020.122828

CrossRef Full Text | Google Scholar

Sun, H., Luo, Y., Liu, J., and Bhuiyan, M. A. (2024). Digital inclusive finance, R&D investment, and green technology innovation nexus. PLOS ONE 19 (1), e0297264. doi:10.1371/journal.pone.0297264

PubMed Abstract | CrossRef Full Text | Google Scholar

Tone, K. (2001). A slacks-based measure of efficiency in data envelopment analysis. Eur. J. Oper. Res. 130 (3), 498–509. doi:10.1016/S0377-2217(99)00407-5

CrossRef Full Text | Google Scholar

Uddin, M., and Rahman, A. A. (2012). Energy efficiency and low carbon enabler green IT framework for data centers considering green metrics. Renew. Sustain. Energy Rev. 16 (6), 4078–4094. doi:10.1016/j.rser.2012.03.014

CrossRef Full Text | Google Scholar

Vujović, T., Petković, Z., Pavlović, M., and Jović, S. (2018). Economic growth based in carbon dioxide emission intensity. Phys. A 506, 179–185. doi:10.1016/j.physa.2018.04.074

CrossRef Full Text | Google Scholar

Wang, Q., and Su, M. (2020). Drivers of decoupling economic growth from carbon emission – an empirical analysis of 192 countries using decoupling model and decomposition method. Environ. Impact Assess. Rev. 81, 106356. doi:10.1016/j.eiar.2019.106356

CrossRef Full Text | Google Scholar

Wang, J., Dong, K., Sha, Y., and Yan, C. (2022). Envisaging the carbon emissions efficiency of digitalization: the case of the internet economy for China. Technol. Forecast. Soc. Change 184, 121965. doi:10.1016/j.techfore.2022.121965

CrossRef Full Text | Google Scholar

Wang, Z., Liang, F., Li, C., Xiong, W., Chen, Y., and Xie, F. (2023). Does China’s low-carbon city pilot policy promote green development? Evidence from the digital industry. J. Innov. Knowl. 8 (2), 100339. doi:10.1016/j.jik.2023.100339

CrossRef Full Text | Google Scholar

Wang, Y., Jiang, A., Zhang, S., and Chen, W. (2024). Traditional finance, digital finance, and financial efficiency: an empirical analysis based on 19 urban agglomerations in China. Int. Rev. Financ. Anal. 96, 103603. doi:10.1016/j.irfa.2024.103603

CrossRef Full Text | Google Scholar

Wu, Y., and Huang, S. (2022). The effects of digital finance and financial constraint on financial performance: firm-level evidence from China’s new energy enterprises. Energy Econ. 112, 106158. doi:10.1016/j.eneco.2022.106158

CrossRef Full Text | Google Scholar

Wu, H., Wen, H., Li, G., Yin, Y., and Zhang, S. (2024). Unlocking a greener future: the role of digital finance in enhancing green total factor energy efficiency. J. Environ. Manage. 364, 121456. doi:10.1016/j.jenvman.2024.121456

PubMed Abstract | CrossRef Full Text | Google Scholar

Xie, Z., Wu, R., and Wang, S. (2021). How technological progress affects the carbon emission efficiency? Evidence from national panel quantile regression. J. Clean. Prod. 307, 127133. doi:10.1016/j.jclepro.2021.127133

CrossRef Full Text | Google Scholar

Xie, Q., Wang, D., and Bai, Q. (2024). “Cooperation” or “competition”: Digital finance enables green technology innovation—a new assessment from dynamic spatial spillover perspectives. Int. Rev. Econ. Finance 93, 587–601. doi:10.1016/j.iref.2024.04.040

CrossRef Full Text | Google Scholar

Xu, X., and Cai, H. (2024). The impacts on regional “resource curse” by digital economy: based on panel data analysis of 262 resource-based cities in China. Resour. Policy 95, 105152. doi:10.1016/j.resourpol.2024.105152

CrossRef Full Text | Google Scholar

Xu, L., Fan, M., Yang, L., and Shao, S. (2021). Heterogeneous green innovations and carbon emission performance: evidence at China’s city level. Energy Econ. 99, 105269. doi:10.1016/j.eneco.2021.105269

CrossRef Full Text | Google Scholar

Yan, B., Wang, F., Chen, T., Liu, S., and Bai, X. (2023). Digital finance, environmental regulation and emission reduction in manufacturing industry: new evidence incorporating dynamic spatial-temporal correlation and competition. Int. Rev. Econ. Finance 83, 750–763. doi:10.1016/j.iref.2022.10.025

CrossRef Full Text | Google Scholar

Yang, C., and Masron, T. A. (2022). Impact of digital finance on energy efficiency in the context of green sustainable development. Sustainability 14 (18), 11250. doi:10.3390/su141811250

CrossRef Full Text | Google Scholar

Yang, L., and Ni, M. (2022). Is financial development beneficial to improve the efficiency of green development? Evidence from the “Belt and Road” countries. Energy Econ. 105, 105734. doi:10.1016/j.eneco.2021.105734

CrossRef Full Text | Google Scholar

Yang, F., Chen, T., and Zhang, Z. (2024). Enhancing urban water efficiency through digital financial inclusion: evidence from China. J. Clean. Prod. 434, 140246. doi:10.1016/j.jclepro.2023.140246

CrossRef Full Text | Google Scholar

Yi, M., Liu, Y., Sheng, M. S., and Wen, L. (2022). Effects of digital economy on carbon emission reduction: new evidence from China. Energy Policy 171, 113271. doi:10.1016/j.enpol.2022.113271

CrossRef Full Text | Google Scholar

Yu, Y., and Zhang, N. (2021). Low-carbon city pilot and carbon emission efficiency: quasi-experimental evidence from China. Energy Econ. 96, 105125. doi:10.1016/j.eneco.2021.105125

CrossRef Full Text | Google Scholar

Yuan, H., and Zhu, C. (2018). Did national high-tech zones promote the transformation and upgrading of China’s industrial structure? China Ind. Econ. 8, 60–77. doi:10.19581/j.cnki.ciejournal.2018.08.004

CrossRef Full Text | Google Scholar

Zhang, M., and Liu, Y. (2022). Influence of digital finance and green technology innovation on China’s carbon emission efficiency: empirical analysis based on spatial metrology. Sci. Total Environ. 838, 156463. doi:10.1016/j.scitotenv.2022.156463

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhang, H., Feng, C., and Zhou, X. (2022a). Going carbon-neutral in China: does the low-carbon city pilot policy improve carbon emission efficiency? Sustain. Prod. Consum. 33, 312–329. doi:10.1016/j.spc.2022.07.002

CrossRef Full Text | Google Scholar

Zhang, H., Xu, T., and Feng, C. (2022b). Does public participation promote environmental efficiency? Evidence from a quasi-natural experiment of environmental information disclosure in China. Energy Econ. 108, 105871. doi:10.1016/j.eneco.2022.105871

CrossRef Full Text | Google Scholar

Zhang, X., Liang, D., and Zhang, S. (2025). The impact of digital–financial dual pilot policy on carbon emission efficiency: evidence from Chinese cities. Land 14 (4), 686. doi:10.3390/land14040686

CrossRef Full Text | Google Scholar

Zhao, X., Long, L., Yin, S., and Zhou, Y. (2023). How technological innovation influences carbon emission efficiency for sustainable development? Evidence from China. Resour. Environ. Sustain. 14, 100135. doi:10.1016/j.resenv.2023.100135

CrossRef Full Text | Google Scholar

Zhao, Q., Jiang, M., Zhao, Z., Liu, F., and Zhou, L. (2024). The impact of green innovation on carbon reduction efficiency in China: evidence from machine learning validation. Energy Econ. 133, 107525. doi:10.1016/j.eneco.2024.107525

CrossRef Full Text | Google Scholar

Zhen, Z., Cao, R., Gan, X., and Wang, C. (2024). The impact of digital finance on urban entrepreneurial activity: the moderating effect of advanced human capital structure. Finance Res. Lett. 64, 105414. doi:10.1016/j.frl.2024.105414

CrossRef Full Text | Google Scholar

Zhou, B., and Ding, H. (2023). How public attention drives corporate environmental protection: effects and channels. Technol. Forecast. Soc. Change 191, 122486. doi:10.1016/j.techfore.2023.122486

CrossRef Full Text | Google Scholar

Zhou, C., Zhou, J., Chen, G., and Xu, C. (2024). Digital finance and greener emissions: evidence from China. Appl. Econ. 56 (49), 5830–5844. doi:10.1080/00036846.2023.2266599

CrossRef Full Text | Google Scholar

Zhuang, R., Mi, K., Zhi, M., and Zhang, C. (2022). Digital finance and green development: characteristics, mechanisms, and empirical evidences. Int. J. Environ. Res. Public Health 19 (24), 16940. doi:10.3390/ijerph192416940

PubMed Abstract | CrossRef Full Text | Google Scholar