Unraveling the Spatial Effects of Fintech on Urban Energy Efficiency in China

Abstract

Improving urban energy efficiency is essential for addressing energy shortages and environmental pollution, thereby facilitating a win–win outcome for both the economy and the environment. As an emerging financial force, fintech is essential for facilitating energy saving, reducing emissions, and advancing modernization. Using panel data of 278 cities in China from 2011 to 2022 to construct a spatial Durbin model for investigating how fintech affects energy efficiency, the following results were found: (1) Energy efficiency shows positive spatial dependence features, and the enhancement of energy efficiency in this location positively influences the energy efficiency of spatially connected regions. (2) Fintech improves local energy efficiency and has notable positive geographical spillover effects on surrounding regions’ energy efficiency. (3) Three mediating pathways are identified: upgrading industrial structure, promoting green innovation, and driving green finance evolution. (4) The regulatory mechanism suggests that environmental regulations can help strengthen fintech’s geographical spillover benefits for the energy efficiency of neighboring areas. The impact of fintech on energy efficiency exhibits heterogeneity due to differences in urban resources and digital infrastructure. These insights offer important theoretical contributions and practical significance for policy-makers in advancing fintech development and urban energy efficiency.

1. Introduction

China, recognized as the largest developing nation globally, stands as a significant importer and consumer of energy. The BP 2024 Statistical Review of World Energy indicates that China’s total energy consumption constitutes 28% of global consumption, ranking first in the world, and it is about 1.84 times higher than the United States, the largest economy worldwide. Energy is a crucial foundation for socioeconomic development, as it is essential for manufacturing nearly all goods and services. However, China’s “rich coal, poor oil, poor gas” resource structure and ongoing industrialization and urbanization have led to sustained growth in energy demand and arduous emission mitigation tasks [1]. In 2024, the primary energy consumption of China rose by 4.3%, making it the country with the highest growth rate globally; its carbon emissions reached 12 billion tons, an increase of 0.2% compared to 2023, comprising 31.5% of all emissions globally. It is evident that while China’s economy is continuously developing, it has paid a relatively high ecological cost. The significant increase in energy consumption jeopardizes China’s energy security and generates severe air pollution, exerting a profound impact on economic and social advancement, together with individual life and health. The tasks of energy saving and emission mitigation remain arduous. Consequently, in light of the economy’s new normal, it is extremely necessary to study how to alleviate the constraints of energy resources, improve energy efficiency, and mitigate environmental pollution.

Recently, advances in cutting-edge technologies like AI, cloud computing, big data, and IoT have spurred the rapid development of fintech in China [2]. As a convergence of technology and finance, fintech fundamentally embodies a technology-driven financial innovation paradigm that catalyzes substantial changes within the financial services sector [3], which makes a country’s financial system work better, subsequently affecting economic activity and energy demand. The integration of traditional finance and digital technology enables fintech to provide more efficient and accessible financial services, making it a potential powerful tool for fostering environmentally friendly and high-quality economic development [4]. Therefore, actively building a new financial technology system that facilitates the sustainable and low-carbon evolution of resources and improves energy efficiency, along with studying how fintech affects energy use and energy efficiency, is not only an urgent need to break the economic growth deadlock within the limitations of resources and the environment but also a long-term strategy that provides significant reference for formulating energy policies and energy saving and emission mitigation policies. Against this backdrop, this study systematically investigates the impact of fintech on urban energy efficiency and its underlying mechanisms.

The advancement of China’s fintech is becoming more and more important for the national economy. The level of its development in China has continued to deepen and its reach has progressively expanded. Now, the development of fintech has been elevated to a national strategy, and financial technology has gradually integrated into every corner of social and economic production and life, which has emerged as a significant catalyst for promoting economic transformation and growth backdrop, reshaping the industrial structure, and changing human production and lifestyle. Fintech has evolved from a by-product into a new factor of production that is transforming traditional value creation. Fintech, utilizing digital technologies like 5G and big data, has expanded the reach and depth of use of traditional finance [5]. The improvement of information transmission efficiency brought about by fintech has changed the way the market operates and the structure of production organization [6], and fintech has become a profound force in reshaping the structure of the global economy and constructing a new development pattern. A growing number of experts are concentrating on the environmental effects of fintech. On the one hand, relying on powerful network platforms and mobile terminals, fintech services such as network credit, digital payment, digital currency, etc., have green attributes in themselves and can substitute for high-carbon energy resources such as coal, which can help to mitigate carbon emissions and drive the sustainable and low-carbon energy transition [7]. On the other hand, fintech is able to maximize the distribution of financial resources [8] through the construction of a convenient ecological network service platform, guiding credit resources to enterprises that actively carry out research, development, and upgrading of green technology, as well as environmentally friendly industries [9], gradually phasing-out inefficient and resource-wasting production links and sectors and accelerating the realization of reduced-carbon or carbon-free new energy sources and new technologies to replace traditional high-carbon energy sources and high-carbon industries and production methods to improve energy and environmental performance. It can be seen that fintech is the fulcrum for transforming the traditional economy and empowering conventional industries’ optimization and transition, and fintech will become a new engine in economic development [10]. However, fintech improves the utilization of the Internet, mobile terminals, etc., while the production and processing of these digital devices increase resource usage with exacerbating carbon dioxide emissions, which are not conducive to urban ecological environment governance [11].

Relevant studies on energy efficiency and its determinants are undertaken mainly from the following angles: One is the perspective of structural optimization. First of all, industrial structure upgrading is a significant way for improving energy efficiency [12], proactively advancing low-carbon sectors, and directing the transition and enhancement of high-pollution, high-resource-consumption traditional firms can fully release the structural dividend in energy saving and emission mitigation and improve energy efficiency [13]. Second, China’s urbanization has moved into a new phase in comprehensively improving the quality of development. The upgrading and adjustment of the population structure can contribute to the reduction in cities’ marginal cost of energy saving and emission mitigation through enhancing the public transportation utilization rate, the efficacy of resource allocation, and the sharing of facilities for pollution control and emission reduction, thus forming an energy conservation and emission mitigation intensification, which will improve environment quality and enhance the performance of carbon emissions [14]. Ultimately, optimizing the coal-centric energy consumption framework can expedite the progressive switching from conventional fossil fuels to renewable and green sources, promoting the formation of a safe, effective, reduced-carbon, and green energy ecology for enhancing energy efficiency [15]. The second is the standpoint of technological progress. Green innovation can foster a new era of digital energy development technologies and low-carbon energy-saving solutions. The profound integration and application of these innovations within the traditional high-carbon industrial chain can infuse novel energy sources into clean and reduced-carbon transition of enterprises, thereby enhancing energy efficiency [16]. Moreover, the extensive implementation of digital technology in energy utilization offers technical support for establishing a digital energy management platform, which improves the accuracy of environmental regulations, promotes the intensive use of resources use, and then promotes economic clean and reduced-carbon transition [17]. The third perspective is environmental regulation. A carbon emissions trading strategy can facilitate carbon emission mitigation and enhance energy efficiency within pilot areas [18]. Still, the energy rebound effect diminishes the emissions trading system’s efficacy in terms of achieving substantial energy saving and emission mitigation outcomes [19].

A growing body of the literature has recently emerged, specifically examining the nexus between fintech and environmental outcomes, particularly energy efficiency and pollution emissions. Firstly, in rural sub-Saharan Africa, fintech was combined with the intrinsic advantages of community-based organizations to enhance small-scale renewable resource funding availability [20]. What is important is that the strong beneficial association between financial technology development and renewable energy utilization was indicated [21]. Since then, there has been a steady stream of research on fintech and energy efficiency. On the one hand, fintech growth, renewable energy usage, government effectiveness, and FDI positively influence natural resource management, while urbanization was determined to have a detrimental impact on natural resource management [22]. Moreover, financial technology may directly enhance energy efficiency in OECD nations [23], which is a good and small-scale success case. On the other hand, fintech began to be associated with Environmental, Social, and Governance (ESG). In developed nations, fintech and ESG serve as net transmitters, whereas renewable energy functions as a net receiver before and during wartime [24], and fintech, globalization, and environmental taxation considerably and beneficially influenced energy transformation over the full Belt-and-Road sample [25]. Furthermore, fintech positively influences the energy transition in middle-income nations, reaffirming the substantial potential impact from foreign direct investment, urbanization, industrial value addition, and economic rise on this transformation [26]. The linear relationship between natural resources, environmental restrictions, and fintech exposed that fintech and EPS affect climate change in an uneven manner [27].

The international literature has also extensively explored the nexus between digital technology and environmental sustainability. A recent bibliometric analysis highlights that artificial intelligence (AI) is becoming a central force in the renewable energy transition globally, creating a smart green tide [28], a finding that resonates with the recent focus on fintech’s role in China. This comparative lens underscores a global trend of technological convergence for green goals [29]. However, the Chinese experience exhibits distinct particularities, as its unique energy structure and regulatory environment provide a powerful and unique context for rapid fintech expansion [30].

But the constraints of spillover distance make it difficult for the energy efficiency in more distant areas to enjoy the benefits of fintech spillovers [31]. Specifically, although fintech has weakened the impact of geographic distance to a certain extent, as the level of spillover distance increases, the communication of effective financial information continues to decrease, the greater the possibility of interpretation errors in the transmitted information, leading to geographic decay characteristics potentially appearing in the spatial spillover effect of fintech on energy efficiency [32]. Some enterprises tend to relocate the high-pollution, high-energy consumption part of the industry chain through re-location due to the pressure of transformation costs and energy resource constraints [33], thus modeling the increase in pollution emissions in the surrounding areas. Moreover, incentives such as competition for growth and economic catch-up strategies lead to significant market segmentation, which limits the spatial spillovers of fintech on energy efficiency [34]. Under the current competitive pattern of China’s “administrative area economy” [35], local governments may abuse their administrative power to implement local protection and market segmentation, which, to a certain extent, increases the transaction barriers and friction costs of cross-regional investment by fintech institutions and hinders the free movement and diffusion of inter-regional capital, technology, and resources, thus leading to the formation of the fintech spatial spillover boundary.

Despite the existing literature, a significant research gap remains in understanding the full impact of fintech on urban energy efficiency. Specifically, three critical limitations are evident. First, the spatial dimension is largely overlooked. While extant studies have examined the direct link between fintech and environmental outcomes, they predominantly fail to consider the spatial spillover effects, neglecting how fintech development in one city influences the energy efficiency of its neighbors. Second, there is a lack of in-depth mechanistic analysis. The current literature’s search direction predominantly concentrates on the analysis of the results, while less studies are analyzed from the aspect of the mechanism, lacking detailed theoretical guidance. The mediating roles of industrial structure upgrading, green innovation, and green finance development, though often mentioned, lack robust empirical verification within a spatial framework. Third, the existing literature largely ignores the effects of environmental regulations. The potential for environmental policies to strengthen or shape fintech’s impact on energy efficiency remains an unexplored area, creating a gap in understanding this relationship.

Given the shortcomings of current research, the potential marginal contributions in this research are as follows: First, from the spatial perspective, spatial-related factors are incorporated into the econometric model, and the design of the spatial econometric model empirically investigates fintech’s geographical spillover impact upon energy efficiency, hence broadening the research area of fintech and its influence on energy efficiency. Second, in terms of mechanism analysis, it provides a comprehensive examination of the probable mechanisms by which fintech influences energy efficiency from three aspects, namely, industrial structure, green innovation, and green finance, which provides evidence to support the optimization path of financial services for energy efficiency. Third, in terms of the moderating effect, this study expands the discussion about the moderating mechanism of fintech development affecting energy efficiency, which helps clarify fintech’s influence upon the collaborative enhancement of energy efficiency in local and surrounding areas from the spatial dimension, makes up for the regret of insufficient analysis of moderating mechanisms in existing studies, and provides a new realistic path choice for the coordinated promotion of energy conservation and emission reduction at the regional level.

The rest of this paper is as follows: After reviewing papers on fintech and energy efficiency, Section 2 formulates four hypotheses. Section 3 provides a data overview and analytical methodology, with its initial interpretations. Section 4 outlines the process of demonstrating spatial correlation. Section 5 and Section 6 provide the mechanism analysis and moderating effect analysis, respectively. Section 7 summarizes the conclusions, implications, and limitations of the paper.

2. Theoretical Analysis and Research Hypotheses

2.1. Fintech’s Influnce on Energy Efficiency and Spatial Spillovers Effects

The emergence of fintech has given rise to the advancement of economic digitization and intelligence and set off a wave of digital revolution. In the digital era, data has become a key economic input. Combined with traditional factors, it reconfigures production systems and input–output relationships [36]. This improves the marginal compensation of traditional factors and reduces energy efficiency losses from resource mismatch, thereby enhancing overall energy efficiency. Fintech is able to use digital technology to meet the diverse funding needs of new low-carbon industries and new energy enterprises and to meticulously control project credit investment in strong-energy-usage and high-emission companies [37], which helps encourage enterprises to actively engage in comprehensive energy-saving renovations, expedite the development of a novel energy system, and realize clean and efficient energy utilization, thereby reducing fossil-fuel-based energy consumption and emissions of pollutants and improving energy efficiency. In addition, fintech facilitates matching supply and demand in creation aspects by means of its data-based and intelligent technological advantages. By integrating innovation factors into the whole industrial chain of regional development, it greatly enhances the scientific nature of innovation decision-making and the efficiency of financial resource usage within the innovation market, and it then boosts the upgrading of traditional factors of production and ultimately improves the efficiency of energy usage.

With the continuous penetration of financial technology, the service coverage of fintech is more extensive and has a spillover effect. Fintech exerts a spatial spillover impact on energy efficiency mainly through diffusion effects and pollution transfer effects. On the one hand, fintech’s development will promote capital and method spillover, forming a diffusion effect [38], which can provide financial and technological guarantees for the enterprises in the surrounding areas to actively carry out green technology study, development, and upgrading transformation, thus increasing the energy efficiency of surrounding areas. On the other hand, the high-speed dissemination of information brought about by the utilization of digital technology in fintech has strengthened industrial synergy, expanded the geospatial scope of industrial agglomeration, and led to the interconnected development of industrial entities. The digital platform’s open ecosystem formed by fintech can integrate social reproduction links, including production, exchange, distribution, and consumption. The industrial agglomeration formed at the virtual space level transcends the limitations of spatial and temporal constraints, bringing about unlimited amplification of the scale effect and realizing the incremental increase in scale remuneration, enhancing economic efficiency and improving energy efficiency [39]. At the same time, fintech can carry out global planning and integrated development of energy utilization through the network effect formed by the digital economy, which contributes to the central supervision of the regional green development process and the expansion of its demonstration effect. Specifically, through the use of ecological environment management informatization means with a variety of modern information processing technology, it can strengthen the synergy of various departments, realize the real-time monitoring and timely supervision of resource utilization, as well as compress spatial and temporal distances to enhance inter-regional economic linkage, decrease the expenses associated with information storage and processing, enhance the efficiency of data sharing, and contribute to the realization of the regional collaborative prevention and joint control of environmental pollution and energy utilization [40], prompting enterprises in the region to mitigate energy usage with enhanced energy efficiency. Therefore, the first Hypothesis H1 is proposed.

H1. 

Fintech’s development contributes significantly to energy efficiency, exhibiting notable spatial spillover effects.

2.2. Mechanisms of Fintech’s Spatial Effect over Energy Efficiency

Based on the above discussions, we propose a conceptual framework, with Figure 1 illustrating the mechanisms through which fintech influences urban energy efficiency. The framework posits that fintech not only enhances local energy efficiency directly but also generates positive spatial spillovers to neighboring areas. These effects are primarily transmitted through three mediating channels: industrial structure upgrading, green innovation, and green finance development. The detailed hypotheses H2–H4 derived from this framework are elaborated in the following subsections.

First is the industrial structure optimization effect. Fintech accelerates the large-scale coverage of newest-generation information technologies, including digital network platforms and mobile terminals, providing sufficient technical support for low-carbon process intelligence in traditional yields and digital transformation in the whole chain of industries [41], promoting the realization of advanced industrial structures and traditional industries effectively solving the environmental constraints of high overall regional energy usage and high carbon emissions. At the same time, fintech can conveniently and efficiently guide financial capital, technology, labor, and more resources to the new energy sector, and the industry can flow through the Internet in an environmentally friendly way [42], gradually eliminating the backwards production methods characterized by excessive energy consumption, pollution, and emissions while realizing industrial restructuring and business model innovation, which, in turn, thereby diminishes energy waste and reduces carbon emissions intensity and improves energy efficiency. In addition, fintech can also use artificial intelligence and big data to break down traditional financial services’ “space–time barriers” for enhancing the upgrading of the industrial structure of surrounding areas, subsequently affecting energy efficiency. Specifically, fintech can provide cross-time and cross-regional financial resource rationing [43], improving the availability of financing for new low-carbon industries in the peripheral region, bringing development dividends to enterprises in peripheral regions that are actively optimizing and adjusting their industrial structure, and then realizing the lightening of the real industrial economy and green development, reducing high-carbon energy usage and carbon emissions, and improving energy efficiency. Consequently, this research suggests the second Hypothesis H2.

H2. 

Fintech can enhance energy efficiency by industrial structure optimization.

The second is the green innovation effect. Fintech can use digital technology to efficiently and conveniently realize the accurate docking between capital and green innovation thematic financing needs [44], guiding credit resources to implement green innovation research and development initiatives that alleviate financing constraints for small companies in green technologies, enhance the accessibility and efficiency of financial resources, and subsequently invigorate market participants’ enthusiasm for green innovation [45]. Enterprises promote the extensive application of clean energy, the low consumption of fossil fuels, and the methodical execution of efficiency and energy conservation by tackling core technologies such as green transformation, new energy, the coal chemical industry, and resource recycling so as to realize the double enhancement of ecological and economic benefits, thus improving energy efficiency. At the same time, financial technology has eased the liquidity constraints of residents and raised their income level [46], and the accumulation of wealth has made residents more aware of environmental protection and readiness to pay a premium for eco-friendly products [47], which will force enterprises to innovate the green production system continuously and provide more products and services with high technological content, low environmental costs, and high resource utilization efficiency, thus reducing high-carbon energy usage and carbon emissions while improving energy efficiency. In addition, fintech can optimize the spatial and temporal allocation efficiency of resources including knowledge, capital, and technology by virtue of information technology such as the Internet. It can also strengthen the positive interaction between enterprises in advanced technology and management styles between regions so as to drive the enterprises in neighboring regions to implement green advancements and promote the transformation of green and low-carbon innovations, reduce carbon emissions during production process, and thus improve energy efficiency. Consequently, this research proposes the third Hypothesis H3.

H3. 

Fintech enhances energy efficiency via driving green innovation.

The third pertains to green finance development’s influence. Fintech can enhance credit allocation, broaden funding sources, and assist green industries and projects by facilitating the development of green finance, ultimately improving environmental quality and enhancing energy efficiency [48,49]. As a key driving force for the green development of financial innovation, green finance promotes enterprises to conduct R&D and implement environmental protection technologies through offering financial support and incentives [50], which can mitigate greenhouse gas emissions [51]. With the help of green finance, fintech has guided capital to flow to environmentally friendly industries and traditional industries that support green transformation, promoted the optimization and advancement of enterprise production methods, and facilitated improvements in energy efficiency [52]. Fintech can optimize the framework of energy usage with green finance development [53] and by supporting renewable energy projects and improving technologies associated with energy saving and emission mitigation, hence further enhancing energy efficiency. Simultaneously, green finance effectively alleviates the information asymmetry associated with green initiatives, diminishes the uncertainty around technical innovation, and bolsters the region’s ability to bear risks in order to support the region’s continued green innovation [54]. In addition, green finance also enhances the environmental compliance requirements and production cost pressures of polluting enterprises, forcing them to improve their production technologies to avoid environmental risks [55] and thus improve energy efficiency. Consequently, this research proposes the fourth Hypothesis H4.

H4. 

Fintech enhances energy efficiency by driving green finance.

2.3. The Moderating Mechanism of Fintech Affecting Energy Efficiency

The aforementioned mechanisms do not operate in a vacuum; their effectiveness is likely shaped by the external institutional environment. Among various factors, environmental regulation (ER) stands out as a critical governmental policy tool that can potentially intensify the impact of fintech on energy efficiency [56]. Environmental regulation reduces the emission of pollutants from the source by restricting the establishment of some high-pollution projects., and it affects regional environmental decision-making, forces cities to strengthen pollution reduction technology innovation, and reduces pollutant emissions through terminal treatment, thus promoting the elimination and upgrading of industries with elevated energy usage and significant emissions and accelerating the industrial structure’s adjustment [57]. Due to the existence of many regulatory means, scholars have separated the categories of formal and informal environmental regulation according to different implementing subjects [58]. Under formal environmental regulations, the innovation compensation theory believes that a scientifically designed environmental regulation system can motivate regions to pursue technological creation and advance green product development. This innovation capability brings additional net production income to the region, improves energy efficiency, and offsets the price increase caused by environmental regulations [59]. At the same time, strict environmental supervision standards help attract high-quality foreign investment, introduce advanced production management technology, and help energy saving and pollution mitigation [60]. In addition, informal environmental regulation is also important. Strict environmental regulations can enhance public environmental awareness and green consumption tendency, increase the demand for green products, force cities to reach the value chain’s higher peak, and promote its green transformation and upgrading of traditional industrial structures. Stronger environmental policies create a compelling “push” factor for firms to seek out efficiency gains, while fintech provides the “pull” factor by reducing the information and transaction costs of finding these solutions [61]. Consequently, this research proposes the fifth Hypothesis H5.

H5. 

Environmental regulation helps to strengthen the positive impact of fintech upon energy efficiency.

3. Research Design

3.1. Spatial Autocorrelation Test

Analyzing the spatial characteristics of fintech requires testing its spatial relationship first to determine whether there is spatial connection. Moran’s index is used to study the spatial and temporal correlation characteristics and spatial aggregation changes in fintech development level, and the Stata 17 software is used to calculate the global and local Moran’s index, respectively. The global one quantifies the average degree of correlation of all spatial units with the neighboring areas across the entire region, while the local one identifies specific spatial aggregation areas and outliers.

Global Moran’s index is calculated as follows:

$$\begin{array}{c}\mathrm{Global} {\mathrm{Moran}}^{\prime }\mathrm{s} \mathrm{I}={\displaystyle \frac{n}{{\sum }_{i=1}^n{\sum }_{j=1}^n{W}_{ij}}}·{\displaystyle \frac{{\sum }_{i=1}^n{\sum }_{j=1}^n{W}_{ij}\left(y_i-\overline{y}\right)\left(y_i-\overline{y}\right)}{{\sum }_{i=1}^n{\left(y_i-\overline{y}\right)}^2}}\end{array}$$

(1)

Local Moran’s index is calculated as follows:

$$\begin{array}{c}\mathrm{Local} {\mathrm{Moran}}^{\prime }\mathrm{s} \mathrm{I}={\displaystyle \frac{n\left(y_i-\overline{y}\right)}{{\sum }_{i=1}^n{\left(y_i-\overline{y}\right)}^2}}·{\displaystyle \frac{{\sum }_{j=1}^n{W}_{ij}\left(y_i-\overline{y}\right)}{{\sum }_{i=1}^n{\left(y_i-\overline{y}\right)}^2}}\end{array}$$

(2)

where i and j signify cities, $W_{ij}$ denotes spatial weight matrixes, n denotes the prefecture number of the study area, and y indicates the observed value of fintech development within the spatial unit.

3.2. Model Establishment and Variable Definition

3.2.1. Spatial Durbin Model

As described in the mechanism section, fintech will not only exert an influence on the region’s energy efficiency but also break through the spatial limitation, affecting the neighboring regions’ energy efficiency. Since inter-regional energy consumption, carbon emission performance, etc., have obvious spatial spillover effects [40], it is inferred that energy efficiency, which simultaneously takes the elements of labor, energy, capital, carbon dioxide emissions, and economic output into account, may similarly exhibit spatial dependence characteristics. Based on the significant regional correlation characteristics of fintech and energy efficiency from the spatial dimension, to fully reveal the spatial spillover influences of fintech affecting energy efficiency and at the same time to consider the utility check of the spatial econometric model, this study constructs the spatial Durbin model for the benchmark regression:

$$\begin{array}{c}{EE}_{it}={\alpha }_0+{\rho }_1{\displaystyle \sum _{j=1}^n}{W}_{ij}{EE}_{jt}+{\alpha }_1{fintech}_{it}+{\rho }_2{\displaystyle \sum _{j=1}^n}{W}_{ij}{fintech}_{jt}\\ +{\alpha }_2{Control}_{it}+{\rho }_3{\displaystyle \sum _{j=1}^n}{W}_{ij}{Control}_{jt}+{\mu }_i+{\eta }_t+{\epsilon }_{it}\end{array}$$

(3)

where i and j denote cities; t means year; ${\alpha }_0$ denotes a constant term; EE represents the explanatory variable energy efficiency; fintech denotes the core explanatory variable financial technology development level; ${Control}_{it}$ is the collection of control variables affecting energy efficiency; $W_{ij}$ denotes the spatial weight matrixes representing spatial relationship of areas; ${\alpha }_1$ and ${\alpha }_2$ denote the parameters to be estimated; ${\mu }_i$ and ${\eta }_t$ denote region and year fixed effects; ${\epsilon }_{it}$ defines the random disturbing factor; ${\rho }_1$ represents spatial autoregressive coefficient, and ${\rho }_2$,${\rho }_3$ signify coefficients of spatial lag terms for fintech as well as the control variables. When ${\rho }_1=0$ and ${\rho }_2, {\rho }_3 \ne 0$, Equation (3) is the spatial lag of X (SLX); when ${\rho }_1\ne 0$ and ${\rho }_2, {\rho }_3 =0$, Equation (3) represents the spatial autoregressive model (SAR).

3.2.2. Spatial Mechanism Test Model

To further investigate whether fintech can act on energy efficiency by industrial structure upgrading [62], green innovation, and finance economy development [63], a spatial mediation effect model was constructed:

$$\begin{array}{c}{M}_{it}={\alpha }_0+{\rho }_1{\displaystyle \sum _{j=1}^n}{W}_{ij}{M}_{jt}+{\alpha }_1{fintech}_{it}+{\rho }_2{\displaystyle \sum _{j=1}^n}{W}_{ij}{fintech}_{jt}\\ +{\alpha }_2{Control}_{it}+{\rho }_3{\displaystyle \sum _{j=1}^n}{W}_{ij}{Control}_{jt}+{\mu }_i+{\eta }_t+{\epsilon }_{it}\end{array}$$

(4)

$$\begin{array}{c}{EE}_{it}={\alpha }_0+{\rho }_1{\displaystyle \sum _{j=1}^n}{W}_{ij}{EE}_{jt}+{\alpha }_1{fintech}_{it}+{\rho }_2{\displaystyle \sum _{j=1}^n}{W}_{ij}{fintech}_{jt}+{\alpha }_2{M}_{it}+{\rho }_3{\displaystyle \sum _{j=1}^n}{W}_{ij}{M}_{jt}\\ +{\alpha }_3{Control}_{it}+{\rho }_4{\displaystyle \sum _{j=1}^n}{W}_{ij}{Control}_{jt}+{\mu }_i+{\eta }_t+{\epsilon }_{it}\end{array}$$

(5)

where $M_{it}$ denotes the mediating factors industrial structure optimization (IND), green innovation (TECH), and green finance development (GF); the core explanatory variable fintech, the set of control variables ${Control}_{it}$, and other parts of the model are the same as before.

3.3. Data and Variables

Considering data availability and completeness, this study takes the face-plate data of 278 cities in China from 2011 to 2022 as the research sample (Tibet and Hong Kong, Macao, and Taiwan are excluded for missing data). Data mainly comes from the Green Patent Database, Digital Inclusive Finance Research Center of Peking University, EPS database, China Urban Statistical Yearbook, China Energy Statistical Yearbook, and China Regional Statistical Yearbook, along with statistical yearbooks and statistical bulletins from every province, with missing values filled in through linear interpolation.

3.3.1. Independent Variable: Financial Technology (Fintech)

Currently, there are three main academic ways of assessing fintech development: First, the quantity of regional fintech enterprises serves as an indicator of the degree of regional fintech development [64]. Second, the China Digital Inclusive Finance Index from Digital Finance Research Center at Peking University is used [65]. The index employs microdata from Ant Financial Services Group, quantifying regional fintech in three dimensions: coverage, usage depth, and Internet financial services digitization extent. Third, a text-mining approach is employed to evaluate the level of fintech development with fintech keyword, news, patents, and job recruitment information retrieval results [66].

Firstly, regional the number of fintech companies is used to assess fintech. “Fintech”, “Cloud Computing”, “Big Data”, “Blockchain”, “AI”, “IoT”, and other fintech-related keywords were searched for on the website of Tianyancha to yield all pertinent samples’ commercial registration information. Only samples of companies with fintech keywords appearing in the company name or scope of business and with the operating status of “existing” or “in operation” and with an operating time of not less than 1 year were retained. Next, according to the Basel Committee on Banking Supervision’s categorization of fintech business models, we used regular expressions to execute fuzzy matching on fintech-related keywords, including “finance”, “insurance”, “credit”, “wealth management”, and “funds” in the business scope of the company, retaining samples that achieved successful matches as fintech companies. In addition, the data were deeply cleaned, and irrelevant information was filtered through layers of screening and filtering; ultimately, the annual tally of fintech companies in prefecture-level municipalities across each province was conducted to assess the degree of local fintech development.

Secondly, this study employs the already compiled Peking University Digital Financial Inclusion Index to assess regional fintech development for robustness testing. This index utilizes underlying information from Ant Gold Service’s trading accounts to portray the extent of China’s fintech development from multiple dimensions. Ant Financial Services is a prominent financial institution in China, and both the most widely used Alipay and the world’s largest money fund, Balance, are the company’s products, so the company’s data can serve as a good reflection of China’s fintech development level. Since the earliest index can only be traced back to 2011, this paper selects 2011–2022 as the sample period.

Third, we utilized text-mining technology to construct a new, quality-oriented fintech index that complements quantitative measurements. This approach aims to capture the technical complexity and innovative capabilities of the local fintech sector, rather than merely its scale. The construction method is as follows: (1) The titles and full text of all news articles from 2011 to 2022 for each city from the China Core Newspapers Full-text Database as well as recruitment data for fintech positions from major online platforms were collected. (2) We employed the Python 3.12 crawler technology to retrieve fintech-related keywords across four dimensions: basic technology, payment and clearing, intermediary service, and direct name. The specific keyword library is shown in

Table 1. Fintech keyword thesaurus.

Dimension	Keyword
Basic technology	Big data, cloud computing, artificial intelligence, blockchain, biometrics, Internet of Things
Payment and clearing	Online payment, mobile payment, third party payment, QR code payment, mobile payment, online payment
Intermediary service	Internet lending, Internet banking, e-banking, Internet insurance, Internet wealth management, insurance finance, mobile banking, direct banking, intelligent customer service
Direct name	Internet finance, financial technology, fintech

, enabling the identification of activities associated with high-quality or cutting-edge fintech. (3) The word frequency is summarized by year, and the entropy method objectively assigns weights to individual indicators and calculates the weighted sum to obtain a comprehensive indicator of financial technology development [67].

It is crucial to note that our measure of fintech is constructed exclusively based on terms related to digital and information technologies (e.g., AI, blockchain). It deliberately excludes any keywords related to financial purposes or environmental outcomes (e.g., green, credit, insurance, investment). This ensures that the variable captures technological capacity itself and is conceptually and empirically distinct from the mediating variables it is hypothesized to affect, such as industrial structure (IND), green innovation (TECH), and particularly green finance (GF), thereby precluding any issues of circular logic [68].

3.3.2. Dependent Variable: Energy Efficiency (EE)

Energy efficiency is assessed by the SBM–Malmquist–Luenberger method. Using 2011 as the base year, labor, capital, and energy are designated as input factors. The overall urban gross domestic product, GDP, is selected as the desirable output, while the urban emissions of sulfur dioxide, fumes (dust), and wastewater from industry are selected as unwanted outputs [69]. Specifically, the variables are measured as follows.

(1)

Labor input, measured using the total quantity of employees across all units, utilizing data from the China Urban Statistical Yearbook.

(2)

Capital input, assessed by the perpetual inventory technique with a depreciation rate of 10.96% [70].

(3)

Energy input, measured by using data on energy usage published in the China Energy Statistical Yearbook to assess regional energy input. Due to the large number of energy components, the unit of energy consumption was uniformly converted to tons of standard coal for convenience of calculation. Since prefectural cities do not publish their energy consumption, night-light data is used to decompose the energy consumption of each province to each prefectural city [71]. This method is founded on the strong empirical correlation between the intensity of artificial nighttime lighting and a variety of socioeconomic variables, including GDP, electricity consumption, and total energy consumption [72,73]. The underlying assumption is that the distribution of energy consumption within a province is proportional to the distribution of nighttime light emissions. Specifically, we calculate the share of each city’s average nighttime light luminosity within the total luminosity of its province. The city-level energy consumption is then estimated by multiplying the provincial total energy consumption by this share. This approach provides a reasonable and widely adopted approximation for sub-provincial energy use where official data is lacking [74].

(4)

We used real GDP as the desirable output, using 2011 as the base year and the GDP deflator to adjust nominal GDP for each city.

The Malmquist–Luenberger productivity index (ML index) method was employed to assess the green total factor energy efficiency of each prefecture-level city, which can deal with the problem of undesirable output more scientifically and take dynamics of production efficiency into account. The ML index exceeds one if the production efficiency of the decision unit currently surpasses that of the previous. The ML index is composed of a technical efficiency index, which refers to the movement of production factor mix efficiency relative to the frontier, and a technical progress index, which refers to the movement of the production frontier.

(5)

Regarding undesirable outputs, following the common practice on eco-efficiency measurement in China, all undesirable outputs are assigned equal weights in the SBM-ML model [75]. This assignment is based on two considerations. First, from a policy perspective, China’s environmental protection system addresses these pollutants with comparable urgency and stringency [76], as reflected in the ‘Ten Environmental Protection Measures’ and other key policies that target coordinated control of multiple pollutants. Second, methodologically, in the absence of explicit market prices or social costs to differentiate the severity of different pollutants, the assumption of equal weights is a common and neutral benchmark that avoids subjective arbitrariness [77]. This approach ensures that the model captures the joint production of economic growth and environmental pollutants without prioritizing one pollutant over another.

3.3.3. Control Variables

To mitigate the effects of missing variables during estimation, drawing on existing paper, this study selects the following variables that significantly influence urban energy efficiency as control variables [78].

(1)

Population density (POP) was expressed using the ratio of population to administrative area, which can portray agglomeration activities’ impacts.

(2)

Economic development (PGDP) was expressed using GDP per capita, which reflects economic expansion’s influence on energy efficiency.

(3)

Science expenditure (SCI) was quantified as the ratio of science expenditure to regional GDP.

(4)

Government intervention (GOV) was characterized by the ratio of local general budget expenditures to regional GDP. Market mechanisms cannot solve the problem of researching, developing, and promoting energy conversation and emission mitigation technologies, so government intervention is needed. Reasonable government intervention can help solve market failures and enforce various energy saving and emission mitigation measures while enhancing energy governance.

(5)

Urbanization rate (UR) was expressed as the ratio of the urban population to the city’s to measure urbanization evolution’s impact.

(6)

Infrastructure construction (IC) was defined by the ratio of postal and telecommunications services to the area GDP. Highly developed postal and telecommunication services provide urban residents with convenient channels for information exchange and real-time interaction. However, the construction and operation of this communication infrastructure require a large amount of energy support, which has an effect on energy efficiency.

3.3.4. Mediating Variables

The mediating variables include industrial structure, green innovation, and green finance development.

(1)

Industrial structure (IND) was defined by the ratio of tertiary industry’s added value to GDP, reflecting the influence of industrial scale characteristics on energy efficiency.

(2)

As for green technology innovation (TECH), green invention patents can reflect a superior degree of innovation, and the volume of patent applications can better reflect green innovation’s actual extent than the quantity of patents authorized. Therefore, green invention patent applications per 100 individuals in each city are used as a measurement [8].

(3)

Green finance development (GF), calculated using the entropy method, was measured by constructing a system of assessment indicators that included four first-level green indicators, namely bonds, credit, funds, and insurance, as well as fourteen second-level indicators, like green investment, support, rights, and interests [79].

3.3.5. Moderating Variable

Environmental regulations (ER) were measured by using text analysis. The government work report was analyzed by word division, and the frequency of terms about environmental regulation from government statements of prefecture-level cities was counted. We calculated the overall word frequency among complete government reports to measure the influence of the external environmental management system, and the total amount of environmental protection penalties in each city was used to measure environmental regulation for cross-validation, which directly reflects the strictness of local environmental regulations and the intensity of their enforcement [80].

Indicator definitions of above variables are shown in

Table 2. Indicator definitions.

Variable	Name	Symbol	Indicator Measure	Unit
Dependent variable	Energy efficiency	EE	Calculated by SBM–Malmquist–Luenberger	-
Independent variable	Financial technology	Fintech	Number of regional fintech companies	home
Mediating variables	Industrial structure	IND	Ratio of tertiary value added to GDP	%
	Green innovation	TECH	Patent applications for green inventions per 100 people	Pieces /100 persons
	Green finance	GF	Entropy method	-
Control variables	Population density	POP	Ratio of population to administrative area	Persons /km2
	Economic development	PGDP	GDP per capita	Yuan /person
	Science expenditure	SCI	Ratio of local science expenditure to GDP	%
	Government intervention	GOV	Ratio of local general budget expenditures to GDP	%
	Urbanization rate	UR	Ratio of urban population to total city population	%
	Infrastructure construction	IC	Ratio of total post and telecommunications business to GDP	%
Moderating variable	Environmental regulations	ER	Government work reports and environmental protection penalties	%

. Descriptive statistics after processing the data are shown in

Table 3. Descriptive statistics of variables.

Variables	Obs	Mean	S.D.	Min	Max
EE	3336	34.8240	14.4196	10.2558	138.7942
Fintech	3336	193.9909	76.0084	19.5300	361.0663
IND	3336	4296.1720	1009.0150	1436.0000	8387.0000
TECH	3336	4.2953	7.6892	0.0000	93.8556
GF	3336	7.2843	0.9919	4.3730	9.8854
POP	3336	446.9832	346.1415	5.0000	2712.0000
PGDP	3336	56,050.1100	32,514.9600	97.2000	203,489.0000
SCI	3336	28.5563	25.6646	1.2826	229.1220
GOV	3336	199.0323	94.5009	35.2607	744.2022
UR	3336	569.6020	149.3328	181.5000	1177.9000
IC	3336	3.3369	0.8167	1.0495	6.2810
ER	3336	454.0808	29.2367	350.5970	700.3065

.

3.4. Spatial Weighting Matrixes Setting

Considering the spatial effects of geographic distance and given the fact that there are regional radiation effects of economic factors, the following economic–geographic nesting matrix is constructed.

$$\begin{array}{c}{W}_{eco-geo}=\left\{\begin{array}{cc}\hfill \tau W_1+(1-\tau )W_2,& i\ne j\hfill \\ \hfill 0,& i\ne j\hfill \end{array}\right.\end{array}$$

(6)

where $W_1$ represents the inverse distance matrix; when $i\ne j$, $W_1=1/d_{ij}$, $d_{ij}$ denotes the distance between city i and city j; when $i=j$, $W_1=0$. $W_2$ represents the economic distance matrix; when $i\ne j$, $W_2=1/\left|\overline{Y_i}-\overline{Y_j}\right|$, and $\overline{Y_i}$ is the average value of the real PGDP in city i during 2011–2022; when $i=j$, $W_2=0$. τ lies between 0 and 1, indicating the inverse distance matrix share, and in this paper, we take $\tau =0.5$. It should be noted the spatial econometric model is processed with adjacency, geographic distance weight, economic geography nested weight matrix, and transportation distance matrix, respectively, to ensure the regression results’ robustness. We label above four matrixes as Adj Matrix, Geo-dis Matrix, Eco-geo Matrix, and Trans Matrix for ease of reference. The adjacency matrix is constructed using the Queen’s method, and the matrix element is 1 when the cities are close together and 0 otherwise. The geographic distance weight matrix is generated by the distance between two regions’ latitude and longitude coordinates. The transportation distance matrix calculates the actual shortest high-speed rail commuting distance between the centers of various cities in China using map software and defines the matrix elements using reciprocals [81].

4. Empirical Analysis

4.1. Spatial Correlation Test

4.1.1. Global Spatial Autocorrelation

We begin by assessing the spatial dependence of urban energy efficiency using the global Moran’s I index. The greater the absolute value of Moran’s I, the more robust the spatial association. A positive value indicates positive spatial autocorrelation, a negative value indicates negative spatial autocorrelation, and a value near zero suggests the absence of spatial correlation [82].

Table 4. Energy efficiency Moran’s I of 278 cities in China (2011–2022).

Year	Adj Matrix		Geo-Dis Matrix		Eco-Geo Matrix
	Moran’s I	p Value	Moran’s I	p Value	Moran’s I	p Value
2011	0.2331	0.0000	0.0517	0.0000	0.1165	0.0000
2012	0.1918	0.0000	0.0403	0.0000	0.0888	0.0000
2013	0.1261	0.0011	0.0287	0.0000	0.1025	0.0000
2014	0.1560	0.0001	0.0328	0.0000	0.1193	0.0000
2015	0.1657	0.0000	0.0401	0.0000	0.1286	0.0000
2016	0.1590	0.0001	0.0287	0.0000	0.0912	0.0000
2017	0.1885	0.0000	0.0370	0.0000	0.1028	0.0000
2018	0.1707	0.0000	0.0337	0.0000	0.0869	0.0000
2019	0.2033	0.0000	0.0356	0.0000	0.1020	0.0000
2020	0.1893	0.0000	0.0372	0.0000	0.1186	0.0000
2021	0.1708	0.0000	0.0361	0.0000	0.1144	0.0000
2022	0.1704	0.0000	0.0360	0.0000	0.1142	0.0000

gives Moran’s I for the energy efficiency of 278 cities in China from 2011–2022 to determine whether energy efficiency is spatially correlated. The results indicate that the indices are all considerably positive at 1%, showing that energy efficiency is not randomly distributed spatially but presents positive spatial dependence characteristics.

4.1.2. Local Spatial Autocorrelation

Moran’s function is relatively limited. Global spatial autocorrelation solely indicates the general state of spatial autocorrelation regarding energy efficiency in cities. If there is spatial heterogeneity, the accuracy of the test will decline. To further investigate the distributional properties of the variables to demonstrate their aggregation status, we conducted a local autocorrelation test and drew the Moran’s I scatter plot. First, the far point in the scatter plot indicates this year’s global Moran’s I, while the distance from sample points to this far point on the way represents the level of aggregation significance. There are four patterns based on energy efficiency. The “high–high” pattern indicates that the area has strong spatial distribution and is encircled by areas with elevated levels. The low–low pattern indicates that the area with weak spatial distribution is clustered by regions with low levels. The low–high pattern implies that it has a low level of spatial distribution on its own and is encircled by several areas exhibiting elevated values. The high–low pattern indicates a region characterized by significant spatial distribution and that is encircled by several areas exhibiting low levels. That is, quadrants one and three indicate beneficial spatial correlation, while quadrants two and four indicate negative autocorrelation. If it is evenly distributed across all four quadrants, it indicates the absence of spatial autocorrelation among the regions.

In this paper, we selected the first year and the last year to draw the Moran’s I scatter plot for energy efficiency, as shown in Figure 2 and Figure 3 (using a hyphen for minus signs). It is not difficult to see that, in comparison to the Moran’s scatter plot of 2011, the quantity of districts clustered at the origin in 2022 is smaller, which indicates that the spatial clustering effect of energy efficiency of China’s prefectural-level cities is more obvious. On the whole, most of the regions are distributed in quadrants one or three, reflecting that China’s energy efficiency exhibits features of either strong value with strong value aggregation or weak value with weak value aggregation, as well as the existence of positive spatial autocorrelation. While the visual clustering is not extremely tight, the consistent statistical significance of the global Moran’s I justifies the employment of spatial econometric models to control for and estimate these spatial dependencies.

4.2. Benchmark Regression Results

Prior to estimating the spatial econometric model, we first established a foundational bivariate relationship between the core variables. As shown in Appendix

Table A1. Difference in energy efficiency between high- and low-fintech groups.

Group	N	Mean of EE	Std. Dev.	t-Statistic	p-Value
High-Fintech	1668	37.51	12.07	−10.93	0.000
Low-Fintech	1668	32.14	15.99

Notes: Groups are divided by the median value of fintech.

(Appendix A), a t-test for the difference in means indicates that cities with an above-median fintech development level exhibit a significantly higher level of energy efficiency (p < 0.01) compared to cities with a below-median level. These preliminary results provide initial support for a positive association between the two variables, justifying the subsequent investigation into their relationship within a more rigorous spatial econometric framework.

After the preliminary spatial correlation test of fintech and energy efficiency, it was discovered there is a strong spatial effect between them. Based on this, before proceeding to parameter estimation, this paper first conducts an applicability test of the spatial panel model utilizing the LM, Wald, and LR tests in

Table 5. Results of LM, Wald, and LR tests.

Test	Adj Matrix	Geo-Dis Matrix	Eco-Geo Matrix
Moran’s I	16.636 ***	25.708 ***	20.957 ***
LM-error	262.005 ***	549.255 ***	423.473 ***
Robust LM-error	146.911 ***	244.890 ***	205.940 ***
LM-lag	136.173 ***	328.275 ***	258.089 ***
Robust LM-lag	21.078 ***	23.910 ***	40.555 ***
Wald_spatial-error	237.62 ***	183.24 ***	184.66 ***
Wald_spatial-lag	258.07 ***	273.03 ***	267.98 ***
LR_spatial-error	229.74 ***	239.37 ***	197.19 ***
LR_spatial-lag	248.30 ***	273.2 ***	260.42 ***
Hausman	56.69 ***	1404.52 ***	−105.64
LR_SDM_ind	95.77 ***	132.24 ***	73.61 ***
LR_SDM_time	3340.24 ***	3338.09 ***	3418.84 ***

Notes: * p < 0.1, ** p < 0.05, *** p < 0.01.

, which show that the LM-lag, LM-error, Robust LM-lag, and Robust LM-error tests under three weight matrices are all significant. In addition, both the LR and Wald test results refuted the initial hypothesis that SDM can be degraded into SAR or SEM, suggesting that there is a double spillover of both independent and dependent variables in urban energy efficiency. Hausman’s test values for both the adjacency matrix and geographic distance weight matrix significantly refused the hypothesis of random effects. The correlation coefficients of the LR_SDM_ind and LR_SDM_time tests under three matrices are statistically significant.

The SDM is also a general form that nests the spatial autoregressive (SAR) and spatial error (SEM) models as special cases [83]. Unlike the SAR model, which only captures spillovers through the dependent variable, and the SEM, which treats spatial dependence as a nuisance in the error term, the SDM allows us to explicitly model spillover effects originating from both the dependent variable and the independent variables in neighboring locations [84]. This is crucial for our study, as it enables us to directly test our core hypothesis that financial technology development in one city can exert significant spillover effects on the energy efficiency of its neighbors—a mechanism that would be obscured in an SAR or SEM specification. Taken together, this paper identifies the SDM with region–year double fixed effects as the optimal model.

The city, year, and city–year fixed effects of SDM were selected for regression estimation, and the regression results are presented in

Table 6. Spatial regression results of fintech on energy efficiency.

Variable	Adj Matrix			Eco-Geo Matrix
	City	Year	Both	City	Year	Both
Fintech	−0.0439 ***	0.0480 **	0.1036 ***	0.0950 ***	0.0456 *	0.1001 ***
	(0.0129)	(0.0236)	(0.0275)	(0.0315)	(0.0261)	(0.0314)
W×Fintech	0.0513 ***	−0.0079	0.0825 ***	−0.1089 ***	0.0079	0.1723 **
	(0.0138)	(0.0163)	(0.0143)	(0.0326)	(0.0674)	(0.0720)
Spatial rho	0.1590 ***	0.2961 ***	0.1060 ***	0.3821 ***	0.4970 ***	0.1873 ***
	(0.0242)	(0.0222)	(0.0250)	(0.0411)	(0.0394)	(0.0487)
sigma2_e	57.1628 ***	149.6729 ***	55.7030 ***	55.7787 ***	153.0659 ***	54.7955 ***
	(1.4032)	(3.7176)	(1.3650)	(1.3698)	(3.8165)	(1.3401)

Notes: Standard errors in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

. The significantly positive spatial autoregressive coefficient (ρ) confirms the existence of positive spillover effects in urban energy efficiency. This result implies that an improvement in one city’s energy efficiency considerably enhances that of its neighboring cities through geographic and economic linkages. This result reaffirms the necessity of adding spatial factors when analyzing energy efficiency. Therefore, in the endeavor to enhance energy efficiency, relevant policies should not be formulated only for the target cities but should focus on the linkage and cooperation between cities to mitigate pollution and carbon emissions. After adding the spatial factor, the regression coefficient of fintech under the region–year fixed effects of SDM are 0.1036 and 0.1001 at a significance level of 1%, again proving fintech’s positive influence on energy efficiency. The spatial lag term’s coefficient of fintech (W × fintech) is notably positive, showing that fintech exerts beneficial geographical spillover effects, and increases in fintech levels within the local city significantly increase the energy efficiency of neighboring areas.

In Table 6, showing the city–year fixed model, the spatial lag term for fintech is significantly non-zero, suggesting that the regression coefficients of the explanatory variables do not effectively reflect the spatial spillovers of this variable on the explanatory variables. Thus, it is essential to continue to separate the spatial effects into three categories, namely total, indirect, and direct, by partial differentiation to avoid the bias that results from deriving spatial spillovers based on the point estimates. The outcomes are presented in

Table 7. SDM effect decomposition of fintech on energy efficiency.

Variables	Adj Matrix			Geo-Dis Matrix			Eco-Geo Matrix
	Direct	Indirect	Total	Direct	Indirect	Total	Direct	Indirect	Total
Fintech	0.107 ***	0.100 ***	0.207 ***	0.106 ***	0.752 ***	0.858 ***	0.103 ***	0.229 ***	0.332 ***
	(0.028)	(0.015)	(0.032)	(0.031)	(0.172)	(0.162)	(0.032)	(0.084)	(0.074)
POP	0.029 ***	−0.054 ***	−0.024 **	0.029 ***	−0.218 ***	−0.189 ***	0.026 ***	−0.063 ***	−0.037 ***
	(0.004)	(0.011)	(0.011)	(0.004)	(0.055)	(0.054)	(0.004)	(0.012)	(0.013)
PGDP	0.000 ***	−0.000 ***	0.000	0.000 ***	−0.001 ***	−0.001 ***	0.000 ***	−0.000 ***	−0.000
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
UR	0.005	0.012 **	0.017 ***	0.002	0.066 **	0.068 **	0.006	0.035 ***	0.040 ***
	(0.003)	(0.006)	(0.006)	(0.003)	(0.030)	(0.029)	(0.003)	(0.013)	(0.012)
SCI	−0.026 **	−0.013	−0.040 *	−0.023 **	0.275 **	0.252 *	−0.031 ***	0.045	0.013
	(0.011)	(0.022)	(0.022)	(0.011)	(0.133)	(0.131)	(0.011)	(0.043)	(0.042)
GOV	−0.022 ***	0.059 ***	0.037 ***	−0.026 ***	0.299 ***	0.273 ***	−0.027 ***	0.158 ***	0.131 ***
	(0.005)	(0.007)	(0.007)	(0.005)	(0.048)	(0.047)	(0.005)	(0.016)	(0.015)
IC	−1.702 ***	1.202 **	−0.499	−1.476 ***	5.882 **	4.406	−1.187 ***	0.804	−0.382
	(0.424)	(0.581)	(0.460)	(0.381)	(2.880)	(2.698)	(0.401)	(1.122)	(0.951)

Notes: Standard errors in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

. The direct effect reveals the driving effect of fintech and the control variables on a city’s energy efficiency, the indirect effect demonstrates their influence on neighboring cities, and the total effect describes the impact on all sample cities. It can be seen that the values of the direct, indirect, and total effects of fintech under three weight matrices are all beneficial, passing at 1% significance, showing that fintech not only enhances this region’s energy efficiency but also significantly improves that of neighboring cities through the spatial spillover effect.

As for control variables, population density (POP)’s direct effect is notably positive, aligning with the expectation that agglomeration in larger cities enables more efficient energy allocation and utilization [85]. PGDP has positive direct and total effects because economic growth raises incomes. This allows people to afford more energy, reducing resource strain, and people are also able to make better use of energy, which increases regional and national energy efficiency as a whole. Urbanization rate (UR)’s direct, indirect, and total influences are all positive, indicating that increased urban concentration promotes overall energy efficiency, likely through heightened demand and more efficient resource sharing across interconnected regions. Scientific expenditure (SCI) has a favorable total effect but a negative direct effect, which reflects a short-term trade-off. Initial R&D energy inputs might temporarily lower local efficiency, but eventual technological breakthroughs and spillovers yield net positive long-term gains [86]. The significantly negative direct effect coupled with a positive spillover effect of government intervention (GOV) reveals a strategic dilemma. Local governments, pressured by short-term economic assessments, may inadvertently protect high-energy-consuming industries, thereby initially inhibiting energy efficiency [87]. Conversely, stringent environmental governance in neighboring regions exerts positive competitive pressure, compelling broader regional improvements. The direct impact of infrastructure construction (IC) is markedly negative, and the indirect effect is favorable, suggesting that the construction of transportation and information infrastructure in the city needs to consume substantial resources, and the development of infrastructure in neighboring areas can enhance regional connectivity and resource mobility, partly offsetting local resource consumption pressures.

In summary, the results of the benchmark regressions clearly support our Hypothesis H1. As shown in Table 6 and Table 7, the direct, indirect, and total effects of fintech are all significantly positive at the 1% level, both under the adjacency matrix and the economic–geographic nested matrix (e.g., under the Eco-geo matrix, direct effect = 0.103, p < 0.01; indirect effect = 0.229, p < 0.01). This suggests that the development of fintech not only significantly improves local energy efficiency but also significantly improves energy efficiency in neighboring regions through spatial spillover effects. Therefore, we accept Hypothesis H1.

4.3. Robustness Tests

Although the previous section’s analysis of the estimation model has initially confirmed the reliability of the estimation results, on this basis, this robustness test study attempts to further replace the sample interval, shrink the tail 1% treatment, lag the explanatory variable, fintech, by one period, and change the weight matrix and the alternative fintech measurement indicator.

4.3.1. Transforming the Sample Interval

In June 2013, Ant Financial launched Yu’ebao, so 2013 is considered as the first year of digital finance development and a milestone in fintech evolution. This work modifies the research sample interval to 2013–2022 to re-estimate the SDM, with results shown in columns (1) to (3) of

Table 8. Robustness test.

Effects	Transforming Sample Interval			Shrinking 1%			One Period Lagged
	Adj Matrix (1)	Geo-Dis Matrix (2)	Eco-Geo Matrix (3)	Adj Matrix (4)	Geo-Dis Matrix (5)	Eco-Geo Matrix (6)	Adj Matrix (7)	Geo-Dis Matrix (8)	Eco-Geo Matrix (9)
Direct	0.0683 **	0.1068 ***	0.1136 ***	0.0671 **	0.0579 *	0.0509 *	0.1079 ***	0.1262 ***	0.1197 ***
	(0.0331)	(0.0370)	(0.0371)	(0.0272)	(0.0297)	(0.0303)	(0.0291)	(0.0321)	(0.0328)
Indirect	0.1644 ***	0.9186 ***	0.2477 **	0.0842 ***	0.7325 ***	0.2759 ***	0.1170 ***	0.7287 ***	0.2196 ***
	(0.0207)	(0.2663)	(0.0971)	(0.0143)	(0.1657)	(0.0785)	(0.0160)	(0.1931)	(0.0850)
Total	0.2327 ***	1.0254 ***	0.3613 ***	0.1513 ***	0.7903 ***	0.3268 ***	0.2249 ***	0.8549 ***	0.3393 ***
	(0.0364)	(0.2550)	(0.0852)	(0.0302)	(0.1566)	(0.0693)	(0.0328)	(0.1839)	(0.0751)
R-squared	0.1378	0.0736	0.0893	0.1144	0.0515	0.0678	0.0931	0.0515	0.0747
Control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	2780	2780	2780	3336	3336	3336	3058	3058	3058

Notes: Standard errors in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

. Both direct and indirect effects of fintech on energy efficiency are significantly beneficial, indicating the robustness of the empirical results.

4.3.2. Shrinking 1% Treatment

To mitigate the influence of extremes, the sample data were Winsorized at the 1% threshold. Columns (4) to (6) of Table 8 show that the regression coefficients of the core explanatory variable, fintech, remain considerably beneficial, indicating the robustness of the empirical results.

4.3.3. Lagging the Fintech by One Period

For addressing the possible endogeneity problem due to bidirectional causality, fintech is treated with one lag period with reference to an existing study [88]. The estimation results based on three matrices are shown in columns (7) to (9) of Table 8. After being lagged by one period, both fintech’s direct and indirect influences on energy efficiency are markedly positive, demonstrating the robustness of the regression results.

4.3.4. Alternative Spatial Weight Matrix: Transportation Distance Matrix

The transportation distance matrix is used to accurately reflect the practical economic connectivity between cities. The results presented in

Table A2. Robustness results using the transportation distance matrix.

Variables	Fintech	POP	PGDP	UR	SCI	GOV	IC
Direct	0.084 **	0.030 ***	0.0002 ***	0.003	−0.022 **	−0.026 ***	−1.432 ***
Indirect	1.203 ***	−0.193 ***	−0.001 ***	0.087 ***	0.403 ***	0.340 ***	5.098
Total	1.286 ***	−0.163 ***	−0.001 ***	0.090 ***	0.381 **	0.314 ***	3.666

Notes: Standard errors in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

(Appendix A) show that the direct, indirect, and total effects of fintech on energy efficiency are statistically significant. The signs and significance of the control variables are also highly consistent with our baseline results. This confirms that our conclusions are robust to a more realistic specification of spatial interdependence based on transportation networks.

4.3.5. Alternative Fintech Measurement Indicator

Fintech is re-measured using text-mining methods and incorporated into the model analysis based on both quantitative and qualitative assessments. The results are shown in

Table A3. Robustness results using alternative fintech measurement indicator.

Variables	Adj Matrix	Geo-dis Matrix	Eco-geo Matrix	Trans Matrix
Direct	0.011 ***	0.011 ***	0.010 ***	0.008 ***
Indirect	0.010 ***	0.075 ***	0.023 ***	0.120 ***
Total	0.021 ***	0.086 ***	0.033 ***	0.129 ***

Notes: Standard errors in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

(Appendix A), which indicates that the direct, indirect, and total effects of fintech on energy efficiency are positive and statistically significant at the 1% level. More importantly, this confirms that technological quality of fintech, rather than merely its quantity, is also a key driver of energy efficiency improvements.

4.4. Endogeneity Test

4.4.1. Instrumental Variables Method

Though the fixed-effects model used in the benchmark regression can mitigate endogeneity to some degree and reduce its impact on estimation results, there may still be factors such as unobservable variables that give rise to the endogeneity problem. The first major source of endogeneity is omitted variables, i.e., important factors affecting energy efficiency may be overlooked. Therefore, instrumental variables are used in this paper to further deal with the endogeneity problem. The second main source leading to endogeneity is the reverse causality problem, i.e., fintech brings innovation resources to the region and creates a favorable innovation environment, which helps to achieve the improvement in energy efficiency. However, cities with improved energy efficiency may be more attractive to resource factors and more likely to promote fintech development, and thus, a reverse causality may exist, for which the explanatory variables are treated with two lags in this paper.

This paper constructs the cross-multiplier IV1 of distance from cities to Hangzhou and annual national Internet access and IV2 of the core explanatory variable fintech lagged by two periods as instrumental variables [89] for performing 2SLS. On the one hand, the fintech index is created from the big transaction data of Ant Financial, headquartered in Hangzhou, and the distance data between each city and Hangzhou almost hardly affects energy efficiency. On the other hand, digital technologies like the Internet are a technological basis for the development of fintech, so the quantity of Internet access is directly correlated with the current state of fintech development. In order to match the cross-sectional data with the panel data, this paper cross-multiplies the above cross-sectional variable with the time-varying variable, i.e., the quantity of Internet access nationwide, as an instrumental variable. In addition, the fintech of the two delayed periods is associated with the fintech in the first two periods and is less associated with the energy efficiency in the current period, so it satisfies the correlation and exogeneity assumptions.

The table below reports the results of the two-stage regressions of instrumental variables, with column (1) and (3) being the first-stage results and column (2) and (4) being the second-stage results. The first indicates that at the significant level of 1%, the regression coefficient of the cross-multiplier term between the distance and the quantity of national Internet access of that year is significantly negative, suggesting that the further one is from the center of fintech development, the lower the development of fintech. The estimated regression coefficient of the two-period lag is considerably beneficial, showing that the two-period lag of fintech is significantly correlated with the current fintech development level. The hypothesis of the correlation of instrumental variables is verified, and the premise that the instrumental variables are valid is satisfied.

Based on the second stage, this paper further examines its correlation of instrumental variables. The F-statistics of the first stage are 595.029 and 243.221, which are much higher than the critical value of 16.38 of Stock–Yogo at the level of 10%, suggesting the absence of a weak instrumental variable issue. The p-value of the KP-LM is less than 0.01, thus significantly rejecting the unidentifiable original setting. The above tests jointly illustrate the validity of the instrumental factors. From the estimation outputs, the coefficient of fintech is highly positive, suggesting that when the potential endogeneity is weakened, this research’s conclusion remains valid, i.e., the development of fintech substantially enhances the improvement in energy efficiency.

4.4.2. Exogenous Policy Shock Test: DID

Regions with advanced fintech capabilities typically possess larger asset scales and stronger competitiveness, potentially indicating inherently higher energy efficiency. Therefore, this paper employs a difference-in-differences (DID) approach to further mitigate endogeneity issues [90]. On 31 December 2015, the State Council issued the “Plan for Promoting Inclusive Finance Development (2016–2020),” encouraging financial institutions to leverage fintech in establishing Internet-based financial service platforms. Given that this policy was formulated at the central level, it can be regarded as an exogenous shock promoting fintech development in regions [91]. For instance, Shenzhen launched its pilot carbon emissions trading scheme (ETS) in 2013, transitioning from high-carbon industries to low-carbon, high-value-added sectors while developing a comprehensive green finance ecosystem [92]. Huzhou targeted small- and medium-sized enterprises and the agricultural sector by implementing a “green credit plus” model. This approach leverages digital platforms to assess small businesses’ environmental performance, offering preferential loan rates for low-carbon projects and significantly reducing carbon intensity [93]. However, due to varying levels of fintech development across regions, areas with weaker fintech capabilities experienced a relatively larger impact. This provides an opportunity for constructing a DID model to identify the causal relationship between regional fintech development levels and energy efficiency.

$$\begin{array}{c}{EE}_{it}={\alpha }_0+{\alpha }_1{Treat}_{it}\times {Post}_{it}+{\alpha }_2{Control}_{it}+{\mu }_i+{\eta }_t+{\epsilon }_{it}\end{array}$$

(7)

The sample is divided into high and low groups based on the median level of fintech development across regions at the end of 2015, corresponding to the control and treatment groups, respectively. These are labeled as $\mathrm{T}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}$, with the treatment group assigned a value of 1 and the control group assigned a value of 0 [94]. A time-varying variable $\mathrm{P}\mathrm{o}\mathrm{s}\mathrm{t}$ is also set, taking a value of 0 for 2015 and earlier years and 1 thereafter. Other symbols are defined as above. Column (5) of

Table 9. Endogeneity test.

Variables	IV1		IV2		DID	GMM
	Fintech (1)	EE (2)	Fintech (3)	EE (2)	Fintech (1)	EE (6)
Fintech		0.2598 **		0.3460 *		0.1830 *
		(0.1565)		(0.1904)		(0.1033)
IV_1	−0.0280 ***
	(0.0028)
IV_2			0.2889 ***
			(0.0616)
KP-LM		64.688 ***		32.411 ***
CD Wald-F		595.029 [16.38]		243.221 [16.38]
Treat × Post					2.4451 ***
					(1.1516)
L.EE						0.4147 ***
						(0.0694)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3336	3330	2780	2780	3336	2780
R-squared		0.1373		0.1252	0.1330
AR(1)						0.001
AR(2)						0.107
Hansen						0.260

Notes: Standard errors in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01; Stock–Yogo F-test at the 10% significance level in square brackets.

shows that the coefficient for $\mathrm{T}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}\times \mathrm{P}\mathrm{o}\mathrm{s}\mathrm{t}$ is 2.4451 and is statistically significant at the 5% level, indicating that fintech significantly enhances urban energy efficiency.

4.4.3. Dynamic Panel Analysis GMM

Due to historical inertia, energy efficiency may continue to attract fintech investment. The System GMM approach is employed to alleviate the bias of the lag dependent variable. Column (6) of Table 9 shows that AR(1) and AR(2) indicate first-order autocorrelation in the disturbance term without second-order autocorrelation, while the Hansen test confirms no instrumental variable over-identification issues. The coefficient for L.EE of 0.417 is significantly less than 1 at the 1% level, indicating that energy efficiency exhibits mean reversion rather than random walk behavior. The significant positive coefficient for Fintech of 0.1830 confirms that, even after controlling for L.EE and its resulting endogeneity, fintech continues to promote energy efficiency, thus validating the earlier conclusions.

5. Mechanism Analysis

Based on the theoretical mechanism described in the preceding section, fintech mainly affects energy efficiency via industrial structure optimization, fostering green innovation, and upgrading green finance development. To verify the basic hypothesis, we further explored the mechanism by which fintech influences energy efficiency.

Table 10. Mechanisms test results.

Variables	Ind		Tech		GF
	(1)	(2)	(3)	(4)	(5)	(6)
Fintech	−2.4286 **	0.1002 ***	0.0828 ***	0.0765 **	0.0080 ***	0.1115 ***
	(1.1588)	(0.0304)	(0.0118)	(0.0305)	(0.0007)	(0.0306)
W × Fintech	31.4317 ***	0.8471 ***	−0.0642	0.7326 ***	−0.0127 ***	0.6275 ***
	(4.6154)	(0.1350)	(0.0469)	(0.1242)	(0.0026)	(0.1242)
Ind		0.0010 **
		(0.0005)
W × Ind		−0.0155 ***
		(0.0042)
Tech				0.3221 ***
				(0.0455)
W × Tech				−0.7613 **
				(0.3725)
GF						0.0084 ***
						(0.0019)
W × GF						0.1671 ***
						(0.0413)
control variables	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3336	3336	3336	3336	3336	3336
R-squared	0.1816	0.0418	0.2027	0.0445	0.2106	0.0520

Notes: Standard errors in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

shows the mechanisms test outcomes under the geographic distance weight matrix. Columns (1) and (2) are the mechanisms tests for industrial structure optimization. In column (1), fintech’s spatial spillover effects are considerably favorable at the 1% significant level, demonstrating that fintech can facilitate industrial structure optimization in the surrounding cities. Column (2) shows that the coefficient for industrial structure optimization (Ind) is considerably beneficial, suggesting that industrial structure optimization can enhance energy efficiency and perform a conduction role in fintech promoting energy efficiency. So, fintech can improve energy efficiency via promoting industrial structure optimization. The reason for that is that fintech promotes industrial structure optimization by improving resource allocation efficiency, increasing total factor productivity, and exploring new products and services [95], which effectively improves the economic production efficiency and energy usage structure, accelerates backward production capacity’s elimination, raises the share of low-energy-using and low-emission companies, and reduces both unitary energy consumption and carbon emission intensity [96], so that energy efficiency can be improved. The results of the mediation effect test support Hypothesis H2. Column (1) of Table 10 shows that there is a significant spatial spillover effect of fintech on IND, and column (2) shows that the coefficient of IND itself is significantly positive at the 5% level (β = 0.0010, p < 0.05). This suggests that fintech can indeed enhance energy efficiency through the path of optimizing industrial structure.

Columns (3) and (4) show the mechanisms test outcomes of green innovation. Fintech’s influence from promotion in local green innovation (Tech) is significantly positive, which means that fintech can enhance energy efficiency by green innovation. Science and technology constitute the primary productive forces, with green technology serving as principal catalyst in increasing the intensity of energy utilization, and it is a significant influencing factor for energy efficiency increases. As gathering places of economic vitality and innovation ability, cities can effectively have an effect of driving urban innovation and can release the dividends of advanced technology into traditional industries by improving traditional technology and introducing new energy-saving technology, thus promoting the conversion of traditional industries to intelligent intensification, reducing environmental pollution and energy consumption, and improving energy efficiency [97]. The above results validate Hypothesis H3. Column (3) shows that fintech has a significant contribution to local green innovation (β = 0.0828, p < 0.01), and column (4) shows that the coefficient of green innovation on energy efficiency is significantly positive at the 1% level (β = 0.3221, p < 0.01). Therefore, green innovation is an effective mediating channel for fintech to influence energy efficiency.

Columns (5) and (6) show the findings of the mechanisms test regarding green finance, which indicates that fintech can considerably enhance green finance development in the area and with an increase in green finance development and energy efficiency of its neighboring regions being significant at the 1% level. As a reminder, the fintech variable measures technological adoption, while the GF variable measures financial outcomes. Their distinct operationalizations ensure that the observed mediation effect is not an artifact of measurement overlap. This is because by providing financial support, reshaping resource allocation, and guiding capital flow to green industries, green finance provides economic incentives for the green transformation of enterprises, encourages regional enterprises to implement more efficient energy utilization technologies and equipment, supports them in technology upgrading, equipment optimization, and production process improvement, directly reduces energy consumption per output unit, and thus, improves energy efficiency [51]. The results of the mechanism analysis support Hypothesis H4. As shown in column (5), fintech has a significant positive effect on local green finance development (β = 1.0312, p < 0.01). Column (6) shows that the coefficient of green finance is also significantly positive (β = 0.0084, p < 0.01). This confirms that the promotion of green finance development is an important mechanism for fintech to enhance energy efficiency.

6. Further Analysis

6.1. The Moderating Effect of Environmental Regulations

The development of fintech contributes to enhancing urban energy efficiency, but several challenges remain. First, fintech’s beneficial effect on energy efficiency has not been fully released, as in some cities, this influence is relatively weak. Second, the advancement of fintech can sometimes negatively affect the energy efficiency of neighboring regions, resulting in a situation where enhancements in energy efficiency in one area may hinder progress in others, meaning they can be caught in a campaign governance predicament known as the “beggar-thy-neighbor” effect. To address these issues and promote both fintech development and energy efficiency, this paper suggests implementing stronger environmental regulations as a viable strategy.

Environmental regulations, as an important government policy instrument aimed at safeguarding the environment and fostering sustainable development, play an indispensable and vital role in mitigating climate change and addressing environmental pollution and other global challenges.

A significant reason why the advancement of fintech may result in diminished energy efficiency in neighboring regions sometimes is that polluting firms are relocated closer to the region due to the penalty effect of investment and funding, increasing the share of polluting industries in neighboring regions. As a matter of fact, the “voting with feet” behavior of polluting enterprises essentially stems from the lack of environmental regulations [98], which gives polluting enterprises an opportunity to carry out strategic emissions reduction or campaign-style management. Therefore, the key to reversing the negative spatial spillovers induced by fintech is to strengthen regional environmental regulations, such as adopting strong command-oriented environmental regulations, market-oriented environmental regulation, or even public environmental regulation. To further explore the impact of environmental regulations, this research develops the moderating effect model as follows.

$$\begin{array}{c}{EE}_{it}={\alpha }_0+{\rho }_1{\displaystyle \sum _{j=1}^n}{W}_{ij}{EE}_{jt}+{\alpha }_1{fintech}_{it}+{\rho }_2{\displaystyle \sum _{j=1}^n}{W}_{ij}{fintech}_{jt}+{\alpha }_2{MER}_{it}+{\rho }_3{\displaystyle \sum _{j=1}^n}{W}_{ij}{ER}_{jt}\\ +{\alpha }_3{Control}_{it}+{\rho }_4{\displaystyle \sum _{j=1}^n}{W}_{ij}{Control}_{jt}+{\mu }_i+{\eta }_t+{\epsilon }_{it}\end{array}$$

(8)

where ${ER}_{it}$ denotes the moderator variable environmental regulations (ER); the core explanatory variable fintech, ${Control}_{it}$, and the remainder of the model retain the same definitions as before.

Table 11. Results of moderating mechanism.

Variables	EE (1)	EE (2)	EE (3)	EE (4)
Fintech	0.0916 ***	0.0983 ***	0.0945 ***	0.0981 ***
	(0.0307)	(0.0305)	(0.0305)	(0.0305)
W × Fintech	0.7184 ***	0.6977 ***	0.6252 ***	0.6135 ***
	(0.1246)	(0.1241)	(0.1259)	(0.1259)
ER	−0.0144 *		−0.0086 *
	(0.0087)		(0.0047)
W × ER	−0.1512		−0.1347 *
	(0.1128)		(0.0814)
ER × Fintech	0.0048 *	0.0014	0.0520 ***	0.0494 ***
	(0.0027)	(0.0018)	(0.0100)	(0.0097)
W × ER × Fintexh	0.0949 **	0.0609 **	0.7164 ***	0.6627 ***
	(0.0380)	(0.0282)	(0.2229)	(0.2201)
control variables	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Observations	3336	3336	3336	3336
R-squared	0.0455	0.0478	0.0436	0.0499

Notes: Standard errors in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

shows the moderating mechanisms’ results in terms of how fintech further affects urban energy efficiency. In column (1), the coefficient for environmental regulations (ER) is negative (−0.0144) but not very significant. This suggests that environmental regulations may result in some negative effects on energy efficiency, but these effects are relatively weak. The reason is that such regulations can increase compliance costs for firms, which may negatively impact their economic performance in the short run. However, over the long run, interaction between fintech and environmental regulations (ER × Fintech) shows a significantly positive coefficient. This indicates that fintech promotes energy efficiency more effectively when environmental regulations are stricter [99]. Stringent environmental regulations can create a favorable policy environment that encourages enterprises to implement more sustainable technologies and production processes. As a result, they are likely to implement energy conservation and emissions reduction strategies, ultimately transforming traditional energy use and production processes and accelerating improvements in energy efficiency.

Notably, from the spatial lag of the above interaction terms, interaction term’s estimated output between fintech and environmental regulation is considerably beneficial, showing that environmental regulations also help to strengthen the beneficial geographical spillover effect of fintech on neighboring regions. Focusing on the cross-multiplier term’s influence between fintech and environmental regulations on energy efficiency, column (2) shows that the cross-multiplier coefficient is considerably favorable, corroborating the conclusion in column (1). So, fintech’s influence on energy efficiency constrained by environmental regulations will be further released, which suggests that environmental regulations enforcement can have a complementary effect with fintech development on energy efficiency. The results of the moderated effects test support Hypothesis H5. As shown in column (1), the coefficient on the interaction term between fintech and environmental regulation (ER × fintech) is significantly positive at the 10% level (β = 0.0048, p < 0.1), and the coefficient on its spatial lagged term (W × ER × fintech) is significantly positive at the 5% level (β = 0.0949, p < 0.05). This suggests that strengthening environmental regulations can indeed reinforce the positive impact of fintech on energy efficiency in the local and neighboring areas. The results of environmental protection penalties in columns (3) and (4) show that the coefficient of the interaction term ER × fintech (0.0520) is still significantly positive at the level of 1%, which further confirms the core findings of this paper. Therefore, Hypothesis H5 is valid.

6.2. Heterogeneity Analysis

6.2.1. Resource Dependence Heterogeneity

Resource-based cities face unique challenges, such as single-industry structures and significant pressure for green transformation. Based on the list of resource-based cities by The State Council, we divided the sample into resource-based and non-resource-based cities to capture their internal heterogeneity. As shown in columns (1) and (2) of

Table 12. Heterogeneity analysis.

Variables	Resource-Based (1)	Non-Resource-Based (2)	Low Infrastructure (3)	High Infrastructure (4)
Fintech	−0.0669	0.1282 *	−0.0301	0.1696 **
	(0.0469)	(0.0696)	(0.0404)	(0.0716)
Control variables	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Observations	1668	1668	1668	1668
R-squared	0.1140	0.0346	0.1300	0.1383

Notes: Standard errors in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

, the impact of fintech is significant in non-resource-based cities but not in resource-based cities. This stems from the close correlation between energy efficiency and a city’s resource endowment. Leveraging their abundant resource reserves, resource-based cities generally exhibit relatively low energy conservation awareness, leading most to operate inefficiently. In contrast, non-resource-based cities possess fewer resources and face urgent transformation needs, creating the strongest practical motivation to enhance energy efficiency [100].

6.2.2. Digital Infrastructure Heterogeneity

The application of fintech fundamentally relies on underlying digital infrastructure, with its positive impacts being more pronounced in cities possessing advanced digital foundations. According to the Report on the Development of Digital Economy in Chinese Cities, the sample was divided into two groups based on the coverage rate of the comprehensive digital infrastructure index: high- and low-digital-infrastructure groups. The coefficients for fintech in columns (3) and (4) of Table 12 are statistically significant in the high-digital-infrastructure group but not in the low group. This is because advanced digital infrastructure reduces the relative cost of energy usage and promotes the research and development of advanced technologies, indicating that digital infrastructure and fintech are complementary. Superior infrastructure enhances the effectiveness of fintech in improving energy efficiency.

6.3. Potential Negative Effects: The Rebound Effect

While our empirical results robustly demonstrate a positive effect of fintech on energy efficiency, it is prudent to acknowledge and discuss potential countervailing mechanisms. Chief among these is the energy rebound effect, also known as the Jevons Paradox [101].

The efficiency gains and cost reductions brought by fintech could also theoretically stimulate increased energy consumption in other areas [102]. For instance, the pervasive convenience of digital payments may encourage higher consumption levels, thereby increasing the energy footprint associated with the production and logistics of goods and services. Improved access to capital via digital lending platforms could accelerate economic activities and expansion in energy-intensive sectors, potentially offsetting some of the efficiency savings at a macro level.

Our findings should therefore be interpreted as the net effect after considering such potential rebounds. The significant positive coefficient suggests that the positive forces from promoting green innovation and structural upgrading currently outweigh these negative counterforces in the Chinese context [30]. However, recognizing the rebound effect is crucial for policy-makers. Mitigation strategies could include implementing carbon pricing mechanisms or integrating green mandates into the algorithms of fintech platforms themselves to ensure that efficiency gains translate into absolute reductions in energy use and emissions [103] and not just intensified consumption.

7. Conclusions

7.1. Main Findings

Given the growing concern about emissions from energy consumption, this research systematically investigates the spatial effects of fintech and its influencing mechanisms of cities in China by analyzing data on energy efficiency. The following are the main findings.

First, energy efficiency displays a notable positive spatial dependence. When energy efficiency improves within one region, it positively influences the energy efficiency of adjacent areas. This means that an increase in energy efficiency in a specific area not only optimizes local energy consumption and utilization but also benefits surrounding areas by fostering advancements in their energy efficiency.

Second, enhancing the development of fintech in Chinese cities can improve local energy efficiency and has favorable positive spatial spillover effects on surrounding regions. This suggests that as the demands for energy efficiency increase, certain regions may benefit from fostering fintech to boost energy efficiency. Robustness and endogeneity tests again confirm this finding.

Third, fintech mainly improves energy efficiency through three paths: optimizing industrial structure, promoting green innovation, and driving green finance development. Fintech guides production factors towards environmentally friendly enterprises and provides cross-temporal and cross-regional financial resource allocation through big data technology and cloud computing platforms, promoting the transition of the traditional industrial structure towards reduced carbon emissions and digitalization. Simultaneously, fintech can invigorate local enterprises’ enthusiasm for green innovation and transformation through digital technology, achieving a dual improvement in ecological and economic benefits by tackling core technologies. Moreover, with the rise of green finance, fintech has opened up new funding channels for environmental protection companies and created a financing environment that promotes enterprises to obtain better green benefits. Additionally, fintech enhances the energy efficiency of both local and surrounding regions through these three paths.

Fourth, enhancing environmental regulations to strengthen financial supervision can further enhance the beneficial effect of fintech on local energy efficiency and the geographical spillover effects on neighboring regions. This indicates that as environmental restrictions intensify, the favorable moderating effect of fintech on energy efficiency will be further strengthened, and it can further promote the improvement in the energy efficiency of surrounding areas. At the same time, fintech has demonstrated more significant improvements in energy efficiency for non-resource-based cities and those with advanced digital infrastructure, indicating that fintech is more effective in cities with greater potential for enhancing energy efficiency.

7.2. Theoretical and Practical Implications

7.2.1. Theoretical Contributions

Firstly, the paper further expands the current theoretical framework of the interaction between fintech and energy efficiency. Through an in-depth investigation of cities in China, the study discovers energy efficiency’s spatial dependence features and emphasizes the spatial spillover effects of energy efficiency across various locations.

Secondly, by introducing the spatial Durbin model, this article systematically analyzes fintech’s impact on energy efficiency. We not only focus on fintech’s direct effect but also concentrate on its indirect influence of adjacent regions via spatial spillover effects. It is found that fintech not only enhances local regional energy efficiency but also exerts a favorable feedback effect on adjacent regions, revealing the complex relationship between fintech and energy efficiency. This establishes a novel theoretical system for understanding the diffusion effect of fintech and the interaction between regions.

Finally, this paper also reveals how different paths contribute to fintech’s mechanism of influencing energy efficiency. It indicates that fintech affects energy efficiency through three paths: optimizing industrial structure, promoting green innovation, and driving green finance development, which offers a new standpoint framework about paths to improve energy efficiency. Additionally, by incorporating the moderating effect of environmental regulations, this paper discovers the moderating effect of environmental regulations at the prefectural city level, giving a new theoretical structure for understanding the relationship between environmental regulations and energy efficiency, and it has enriched the heterogeneity analysis of the role of fintech in improving energy efficiency in different cities.

7.2.2. Practical Significance

This research has significant implications for policy-making, energy utilization, and promoting low-carbon and sustainable urban development.

Firstly, we find that the advancement of fintech in cities substantially enhances energy efficiency. This reminds policy-makers that while advocating for environmental protection and a low-carbon economy, they should also formulate targeted strategies to bolster local fintech ecosystems. To this end, governments should move beyond general support and implement precise fiscal incentives, such as offering corporate tax reductions or VAT rebates for fintech firms that develop and deploy specific green solutions, like blockchain-based platforms for green asset verification or AI-driven energy efficiency audit tools. This initiative aims to enhance ability of fintech to serve critical sectors, including clean energy, carbon reduction technologies, and energy conservation and environmental protection.

Secondly, fintech’s spatial spillover effects on energy efficiency reveal that fintech not only influences local areas but also impacts nearby cities. This result emphasizes that local governments should strengthen fintech cooperation among regions and promote coordinated regional development to deepen the radiating and driving effect of fintech on the energy efficiency of surrounding areas. A concrete mechanism to achieve this is the establishment of cross-regional “Green Fintech Pilot Zones” in national strategic areas (e.g., Beijing–Tianjin–Hebei, Yangtze River Delta). Within these zones, authorities should unify green certification standards, facilitate the secure sharing of energy and carbon data with fintech platforms, and encourage the development of innovative financial products based on this data. This would enable fintech to significantly contribute to improving the energy efficiency of surrounding regions. Then, fintech institutions should be encouraged to collaborate in building digital supervision platforms to ensure that the dividends of fintech development can better benefit enterprises in surrounding regions that are actively engaged in industrial chain optimization, green innovation, and green finance development. Lastly, local governments should gradually break the ideology and political logic about localism, eliminate the segmentation and multi-track operation of the factor market, promote the smooth flow of financial elements in a wider range, serve the cohesive national market’s establishment, and thereby broaden fintech’s spatial spillover channels.

Finally, industrial structure, green innovation, and green finance development are important mechanisms for fintech to enhance energy efficiency and should be central to the green development strategy. The government ought to direct the fintech sector to implement green technological innovation. This can be achieved by integrating “carbon performance” into the core of financial decision-making. Regulators could introduce mandatory disclosure requirements for major fintech firms, compelling them to publish the environmental impact (e.g., green-to-brown investment ratios) of their financed projects. This would promote the financial resource allocation of enterprises with low energy usage and low waste, reducing energy waste from the source. At the same time, commercial banks and other institutions should be supported to expand environmental rights, pledge credit business, and develop a multi-layered fintech product system, for instance, by using IoT and smart contracts to dynamically link loan interest rates to verified energy savings, thereby empowering energy efficiency [37]. On the one hand, fintech institutions can enlarge the supply of credit funds to enterprises implementing energy conservation and green innovation through syndicated loans and bank–insurance cooperation and assist enterprises in green transformation. On the other hand, commercial banks can guide and encourage key carbon-emitting industries such as chemicals, nonferrous metals, and building materials to upgrade and transform through measures such as establishing a priority approval mechanism for green projects, lowering the financing costs of green projects, and improving the multi-level fintech product system, achieving lightweight and green development of industries and empowering energy efficiency.

Moreover, given the regulatory role of environmental regulations on fintech, the government should establish robust inter-regional financial supervision mechanisms and collaborative early-warning systems based on the industry-specific insights enabled by fintech. When formulating energy policies, fintech-driven monitoring of energy waste and carbon emissions should be incorporated into the systemic financial risk monitoring framework. Projects that fail to comply with environmental protection norms, are highly polluting, or have backwards production capacity and technology should be firmly resisted, and no new credit support should be provided. The government must adhere to the principle of tailoring approaches to local conditions and implementing differentiated strategies, establishing long-term mechanisms to promote energy conservation and efficiency gains in resource-based cities and those with underdeveloped digital infrastructure. It should continuously accumulate supporting resources—including technologies and talent—required for green and low-carbon technology applications, proactively undertake industrial transfers from other regions, and maximize institutional dividends from investments in digital infrastructure.

This study provides key insights for policy-makers and related enterprises to strengthen fintech’s spatial spillover effects on energy efficiency and facilitate high-quality and low-carbon development layouts within the structure of sustainable and green development.

7.3. Limitations and Future Research

Despite valuable perspectives on attaining both environmental and low-carbon economic growth being offered, this study also has certain limitations.

Firstly, the data source has limitations, with severe data deficiencies for some areas. This exclusion may introduce a sample selection bias and limit the generalizability of our findings to the entire western region of China. In addition, the influencing factors of energy efficiency involve economic, social, and technological aspects, and this paper may not have considered some potential factors. In the future, the database ought to be enlarged to enhance the comprehensiveness of the indicator system and empirically demonstrate the rebound effect of energy consumption.

Secondly, although this study identifies statistically significant positive spatial spillover effects of fintech on energy efficiency, it is noteworthy that the strength of the spatial autocorrelation, as indicated by Moran’s I values, is moderate. This suggests that the spillover effects, while present and important, might be tempered by other regional-specific factors or barriers to diffusion. Future research could delve into identifying these mitigating factors to gain a more comprehensive understanding of the inter-regional dynamics of energy efficiency improvements. At the same time, while our study provides robust evidence of the spatial spillover effects, the policy recommendations derived are necessarily qualitative. Future research could integrate our estimated spillover coefficients into a scenario analysis model to quantify the potential energy efficiency gains from specific regional policies, such as establishing a unified ‘Green Fintech Pilot Zone’ in the Yangtze River Delta.

Thirdly, regarding the generalizability of our findings, an important consideration arises from the unique context of our study. While China is an emerging economy, its fintech sector is mature and widely adopted, with strong government intervention [104]. Consequently, the spatial spillover effects and mechanisms we identify may be more pronounced and observable than in countries with less developed digital infrastructures. For instance, India’s Unified Payments Interface (UPI) exemplifies a highly successful and widely adopted fintech innovation in the payment systems sector. However, its impact on overarching national goals, such as energy efficiency, may be attenuated in the absence of equally robust, complementary policy priorities, regulatory approaches, or infrastructure that deliberately steer financial flows and corporate behavior toward green outcomes [105]. Our findings can thus be seen as revealing the potential “best-case” outcomes of fintech development. They are most directly applicable to other regions or nations approaching a similar level of technological maturity. For emerging economies at an earlier stage of fintech development, our study provides a critical benchmark and a roadmap of what can be achieved through sustained investment in financial technology. Future research could explore the differential impacts of fintech across countries with varying levels of digital infrastructure to establish a more generalized theory.

Finally, while this paper concludes that energy efficiency is significantly impacted by China’s fintech development., it does not take into account the development status of fintech in other countries. The relationship between fintech among two countries can be further investigated in future studies, such as considering whether the development level of fintech in other countries is faster than that of China and how this distinction impacts the enhancement of spatial spillover and energy efficiency.