Introduction

Given the dual imperatives of energy security and achieving the “double carbon” goals, improving China’s energy efficiency has become a pressing issue in the country’s economic development agenda (Xu & Gu, 2024). The Chinese government has elevated energy efficiency as a priority and committed to “comprehensively improving resource utilization” during the 14th Five-Year Plan. This initiative is critical for promoting the harmonious coexistence of humanity and nature and addresses the major strategic need for sustainable economic growth.

Globally, digital finance exhibits divergent trajectories: Sub-Saharan Africa’s mobile money adoption surged to 55%, India’s UPI processed 10 billion monthly transactions, and China’s ecosystem-driven models (e.g., Alipay) achieved 93% payment penetration, contrasting with advanced economies’ focus on institutional digitization. Concurrently, industrial upgrading is critical to avoiding ‘premature deindustrialization’ (Fujiwara & Matsuyama, 2024). It has driven energy efficiency gains in South Korea, with an annual reduction of 2.1%, and in Germany, achieving a 15% productivity boost (Fraunhofer Institute, 2022). However, industrial upgrading has stalled in resource-dependent economies such as Brazil. However, energy efficiency improvements face systemic barriers: Jevons’ paradox persists (e.g., 8% U.S. energy use rise despite 20% efficiency gains (Sorrell et al., 2020), while outdated infrastructure limits India’s coal plants (62% efficiency vs. Japan’s 85%), and China’s regional disparities (e.g., Guangdong’s 40% higher productivity than Shanxi) highlight policy and technology gaps.

Digital finance, a new financial service model that integrates digital technology with financial innovation, represents a key direction for the financial industry’s future development. This is driven by the disruptive impact of the Internet revolution on traditional finance, alongside the adoption of digital innovation technologies such as big data and information technology (Gomber et al., 2017; Kong & Loubere, 2021; Liu et al., 2023). A pertinent question arises: how can digital finance influence industrial structure upgrading from the perspective of energy efficiency? Could a targeted digital finance strategy boost energy efficiency improvements by prioritizing investments in high-return industries that adopt clean technologies? This paper seeks to explore precisely these questions. As China transitions into the digital economy era, digital finance offers multiple benefits. It can lower corporate financing costs, broaden financing channels, increase financial resource availability, and alleviate capital mismatches. These advantages can mitigate corporate financing challenges, foster innovation, promote industrial structure upgrading, and ultimately drive a shift towards cleaner industries and green development (Agrawal et al., 2024). Additionally, digital finance creates a more efficient digital information-sharing platform. This platform raises public awareness of energy and environmental protection issues, which in turn promotes energy efficiency improvements.

From the perspective of energy efficiency, improvements in energy efficiency can drive industrial structure upgrading, which further stimulates financial development (Shen & Ren, 2023). Conversely, a decline in energy efficiency constrains national economic growth (Deichmann et al., 2019), thereby affecting financial development. Therefore, the impact of digital finance development and industrial structure upgrading on energy efficiency must be analyzed from both perspectives. On one hand, digital finance supports industrial structure optimization by providing diversified financial services, reducing transaction costs, addressing corporate financing constraints, and improving operational efficiency. On the other hand, energy efficiency improvements are closely tied to digital finance development, as the latter supports the transformation of the industrial structure from predominantly secondary industries to tertiary sectors, such as finance. This shift reduces the economy’s dependence on energy, improves energy efficiency, and mitigates environmental challenges (Rubino, 2017). The combined effects of digital finance development and industrial structure upgrading play a critical role in enhancing energy efficiency.

Based on the above analysis, this paper distinguishes itself from previous studies, which typically focused on the relationship between financial development and energy efficiency or financial development and industrial structure upgrading (Shen & Ren, 2023; Wu et al., 2024). Instead, it explores how digital finance affects energy efficiency through industrial structure upgrading, providing a new perspective on the relationship among these three factors. This study employs total factor productivity (TFP) to measure industrial structure upgrading, offering a novel approach to evaluating industrial transformation. The research process is divided into three stages: first, the paper incorporates digital finance into the model to assess its impact on energy efficiency. Second, it examines the relationship between industrial structure upgrading and digital finance. Finally, digital finance is used as the core explanatory variable, with industrial structure upgrading as the mediating variable, to verify the relationship between energy efficiency, digital finance, and industrial structure upgrading.

Theoretical analysis indicates that development of digital finance enhances the breadth, depth, and effectiveness of financial services (Misati et al., 2024), which better satisfies the financial needs for technological upgrades and industrial restructuring. Digital finance also has a guiding effect on resource allocation (Corrado & Corrado, 2017), offering an improved information-sharing platform and efficient information dissemination mechanisms, thereby influencing energy efficiency. This study selects annual data from 30 provincial-level administrative regions in China (including provinces and municipalities) from 2011 to 2022 as the research sample. Based on these data, an empirical test is conducted on the proposed theoretical hypotheses, aiming to explore the mechanisms through which digital financial development promotes energy efficiency in China.

Literature review

Influencing factors of energy efficiency

After reviewing the relevant literature, industrial structure upgrading is a crucial factor affecting energy efficiency. Zhu et al. (2024) conducted an empirical analysis based on panel data from 30 provinces in China and found that the optimization of industrial structure (i.e., industrial structure upgrading) significantly improves energy efficiency. However, other factors also play a significant role in this regard. For instance, Lin and Wu (2020) found that countries’ energy productivity rises slower during the economic take-off phase than economic growth. However, it stabilizes or declines as income levels increase and economies enter the mass consumption stage. Zhang and Adom (2018) also showed that human capital development, foreign direct investment, and urbanization promote energy efficiency. According to Yang and Wei (2019), technological change harms all-factor energy efficiency, while economic structure, level of openness to the outside world, and government fiscal spending positively affect it. Wu et al. (2023) incorporated green technology innovation and R&D investment into their model to analyze the mechanism through which digital finance affects energy efficiency. Finally, Wang et al. (2022) found that the convergence of environmental regulation and diversified agglomeration has a robust positive effect on promoting energy efficiency.

Research on the impact of industrial structure upgrading on energy efficiency

The relationship between industrial structure upgrading and energy efficiency has been widely studied by both domestic and international scholars, yet the conclusions remain varied. Two primary perspectives emerge from the literature. The first suggests that industrial structure upgrading promotes energy efficiency. For example, Yu (2020) used an enhanced Super-SBM method to measure total factor energy efficiency in China. The study found that adjustments in inter-industry service structures, intra-industry productivity growth, and increased technological innovation can significantly improve total factor energy efficiency. Similarly, Xing et al. (2021) found that industrial structure upgrading plays a mediating role in improving energy efficiency, and market-oriented reforms can enhance energy efficiency by promoting industrial structure upgrading. Zheng et al. (2023) analyzed the impact of industrial structure on energy efficiency in China using a spatial econometric model based on panel data from 30 provinces between 2004 and 2019. Their findings indicate that while industrial structure positively influences provincial energy efficiency, it also generates significant negative spatial spillover effects.

Conversely, the second perspective contends that industrial structure upgrading has little or even a negative impact on energy efficiency. Tao et al. (2024) challenge the assumption that industrial restructuring influences energy intensity, asserting that structural adjustments within industries have negligible effects on energy intensity. In another study, Zheng (2021) used the DEA method to evaluate China’s energy efficiency from 1978 to 2007, concluding that the growing share of the tertiary sector actually reduces energy’s technical efficiency.

Some scholars emphasize that the scale of industrial restructuring does not necessarily correspond to improvements in energy efficiency. Zhao et al. (2022) highlights a significant spatial correlation and interaction between the magnitude and quality of industrial restructuring and energy efficiency. The study argues that it is the quality, rather than the extent, of restructuring that serves as the key driver of energy efficiency improvements. While most studies affirm industrial upgrading's energy efficiency benefits, disagreements center on transmission mechanisms.

Research on the development of digital finance affecting the upgrading of industrial structure

The interplay among digital finance, industrial structure upgrading, and energy efficiency is anchored in three theoretical perspectives. First, the Information Asymmetry Reduction Theory (Stiglitz & Weiss, 1981) posits that digital finance alleviates financing constraints for SMEs through big data and AI-driven credit assessments (Gomber et al., 2017). This enables firms to invest in green technologies (e.g., Ant Group reduced loan rejection rates for micro-enterprises by 30%. Second, the Technology Spillover Effect (Grossman & Helpman, 1991) emphasizes how digital payment platforms integrate supply chain management to accelerate the diffusion of energy-efficient practices in manufacturing (Liu et al., 2022). This process helps facilitate the transition from energy-intensive industries to service-oriented sectors. Third, Institutional Evolution Theory (North, 1990) highlights digital finance as an institutional innovation that enhances transparency, curbs rent-seeking, and incentivizes energy efficiency prioritization (e.g., Guangdong adopted clean energy policies 40% faster than Shanxi). Together, these theories frame digital finance as both a technological enabler and institutional catalyst, driving sustainable structural transformation.

Based on the literature reviewed, digital finance promotes industrial structure upgrading both directly and indirectly. The direct pathway, rooted in Financial Constraint Theory (Beck et al., 2005), posits that digital finance alleviates corporate financing constraints by reducing information asymmetry and transaction costs. This, in turn, enables firms to invest in innovation and pursue structural upgrades. For example, in China’s eastern provinces, digital finance has lowered the financing costs of SMEs by 15%, facilitating their adoption of energy-saving technologies and participation in high-value-added industries. Empirical evidence from Lu et al. (2023) shows that an increase in digital finance penetration reduces industrial energy intensity through this channel. The indirect pathway, grounded in Endogenous Growth Theory (Romer, 1990), suggests that digital finance drives industrial upgrading by directing capital to innovation-intensive sectors (e.g., green tech). This upgrading, in turn, lowers energy consumption per unit of GDP. For instance, fintech-driven regions in eastern China exhibit higher energy productivity due to patent-led innovation spillovers (Tao et al., 2024). However, these mechanisms are less effective in resource-dependent regions like Shanxi, where structural inertia and coal reliance persist. Shen et al. (2024) conducted an empirical study using data from 1385 counties in China from 2014 to 2020 and found that the development of digital finance significantly promotes local industrial structure upgrading. Similarly, Ren et al. (2023) analyzed prefecture-level cities in China and reached a comparable conclusion, highlighting that digital finance development has a significant positive effect on regional industrial structure optimization. Jiao et al. (2024) note that digital inclusive finance has significantly accelerated the upgrading of industrial structures, and factor mobility has played a positive moderating role in the impact of digital inclusive finance on promoting industrial structure upgrading. Li and Li (2022) found that the rapid development of inclusive digital finance plays a substantial role in optimizing industrial structures, with technological innovation serving as a critical intermediary mechanism. Furthermore, Tchidi and Zhang (2024) suggest that digital finance affects industrial structure upgrading both directly and indirectly, with economic development acting as a mediating factor. In summary, despite variations in research methodologies, the literature consistently supports the positive impact of digital finance on industrial structure upgrading.

Existing research on digital finance, industrial structure upgrading, and energy efficiency suffers from three critical gaps. First, most studies treat industrial upgrading as an isolated outcome (Cheng et al., 2023) or a control variable (Chen & Zhang, 2021), overlooking its mediating role between digital finance and energy efficiency. For instance, Jiao et al. (2024) focus on digital finance’s direct impact on industrial optimization but fail to link it to energy efficiency gains. Similarly, (Cheng et al., 2023) examines digital finance’s role in industrial upgrading but ignores its environmental implications. Second, methodological limitations persist: cross-sectional analyses dominate the field (Tao et al., 2024), often overlooking panel data challenges like heteroskedasticity and endogeneity. Zheng (2021) use DEA models to assess energy efficiency but neglects autocorrelation, risking biased estimates. Additionally, studies like Zhang & Adom (2018) rely on single-factor energy efficiency metrics, which fail to capture total factor productivity dynamics (Yu et al., 2023). Third, the literature lacks a unified theoretical framework to explain the mechanisms linking digital finance to energy efficiency. While some studies emphasize resource reallocation (Gomber et al., 2017), others focus on innovation spillovers (Wu et al., 2024), but none integrate these pathways. For example, Li and Li (2022) show that digital finance raises the tertiary sector’s share, but they do not connect this to energy efficiency improvements.

Based on this, this study makes the following improvements, forming its main contributions: First, while analyzing the relationship between digital finance and energy efficiency, this study incorporates industrial structure upgrading into the framework, establishing a complete research model of “digital finance→industrial structure upgrading→energy efficiency.” This design not only focuses on the direct impact of digital finance on energy efficiency but also reveals the underlying mechanisms through which industrial structure upgrading shapes this relationship. It offers new insights into how digital financial innovation can indirectly promote sustainable development goals through economic structural adjustments. Second, this study employs total factor productivity (TFP) as a novel approach to measure industrial structure upgrading. By examining efficiency improvements in multi-industry and multi-factor input-output processes, the TFP metric avoids the issue of “misleading elevation” caused by solely relying on output ratios to assess industrial structure upgrading. This provides empirical support for the selection of a more accurate measurement method for structural transformation. Third, incorporating multiple control variables, such as industrial added value, marketization index, and regional industrial density, to account for potential factors influencing energy efficiency beyond digital finance and industrial structure upgrading. Simultaneously, utilizing provincial panel data from China for the period 2011–2022, the empirical analysis thoroughly considers potential endogeneity issues in the data. It also applies robustness tests and model adjustments (e.g., introducing province and year fixed effects, employing instrumental variables) to reduce estimation bias and ensure the robustness and credibility of the results.

These improvements offer a deeper understanding of the impact of digital finance development and industrial structure upgrading on energy efficiency and have significant implications for China’s green transformation and the enhancement of environmental quality.

Study design

Model design

This study uses theoretical analysis to construct a mediating effect model based on the method of Baron & Kenny (1986). The selection of this model is primarily based on the complex and multi-level relationships among digital finance, industrial structure upgrading, and energy efficiency. Compared to simple regression analysis, the mediation effect model can effectively distinguish between direct and indirect effects. It also reveals the specific transmission pathways between variables and clarifies the mechanistic role of the mediating variable in the overall influencing process. This allows for a more accurate test of theoretical hypotheses and a clearer understanding of the transmission mechanisms between variables. As follows, three models are established consecutively to investigate the relationship between digital finance, industrial structure upgrading, and energy efficiency.

Firstly, a model without mediating variables is constructed to test the direct effect of digital finance development on energy efficiency improvement.

$$\begin{array}{l}{{EE}}_{{it}}={\alpha }_{0}+{\alpha }_{1}{{DFL}}_{it}+{\alpha }_{2}{{IVA}}_{{it}}+{\alpha }_{3}{{FE}S}_{{it}}+{\alpha }_{4}{{RII}}_{{it}}+{\alpha }_{5}{{PC}}_{{it}}\\\qquad\quad\;+\,{\alpha }_{6}{{PC}}_{{it}}^{2} +{\alpha }_{7}{{RI}}_{{it}}+{\alpha }_{8}{{LFI}}_{{it}}++{\mu }_{{it}}+{\phi }_{{it}}+{\varepsilon }_{{it}}\end{array}$$
(1)

Secondly, a model is established to investigate whether the development of digital finance promotes upgrading industrial structure.

$$\begin{array}{l}{{OIS}}_{{it}}={\beta }_{0}+{\beta }_{1}{{DFL}}_{it}+{\beta }_{2}{{FES}}_{{it}}{+\beta }_{3}{{LFI}}_{{it}}+{\beta }_{4}{{FDL}}_{{it}}\\\qquad\quad\;\;+\,{\mu }_{{it}}+{\phi }_{{it}}+{\varepsilon }_{{it}}\end{array}$$
(2)

Lastly, the mediating effect of industrial structure upgrading in the relationship between digital finance development and energy efficiency improvement is examined by incorporating industrial structure upgrading into the model that tests the direct effect of digital finance development on energy efficiency improvement.

$$\begin{array}{l}{{EE}}_{{it}}={\gamma }_{0}+{\gamma }_{1}{{DFL}}_{it}+{{\gamma }_{2}{{OIS}}_{{it}}+\gamma }_{3}{{IVA}}_{{it}}+{\gamma }_{4}{{FE}S}_{{it}}+{\gamma }_{5}{{RII}}_{{it}}\\\qquad\quad\;\,+\,{\gamma }_{6}{{PC}}_{{it}}+ {\gamma }_{7}{{PC}}_{{it}}^{2}+{\gamma }_{8}{{RI}}_{{it}}+{\gamma }_{9}{{LFI}}_{{it}}+{\mu }_{{it}}+{\phi }_{{it}}+{\varepsilon }_{{it}}\end{array}$$
(3)

Variables and data description

The sample for this study comprises annual data from 2011 to 2022, covering 30 provinces and municipalities directly governed by the central government in China. The variables included in the analysis are energy efficiency, digital finance, industrial structure upgrading, industrial value-added, marketization level, regional industrial density, population density, regional innovation levels, and foreign investment. The Tibet Autonomous Region is excluded due to insufficient data availability.

To investigate the relationship among digital finance, industrial structure upgrading, and energy efficiency, this study draws on multiple data sources. These include the China National Bureau of Statistics, the China Statistical Yearbook (2011–2022), the China Energy Statistical Yearbook (2011–2022), and various provincial statistical yearbooks (2011–2022). The digital financial inclusion index, which serves as a measure of digital finance, is obtained from the Digital Finance Research Center at Peking University. Since China does not officially release capital stock data, this paper estimates the capital stock using the method adopted by Guariglia and Poncet (2008), with 2000 as the baseline year and a depreciation rate of 10.96%. Missing data are addressed through multiple imputation analysis and linear interpolation. The specific configurations of each variable are detailed as follows:

Dependent variable: energy efficiency (EE)

The dependent variable in this study is energy efficiency (EE), measured using a single-factor energy productivity index as proposed by (Yu et al., 2023). This method has two key advantages: first, it directly reflects the low levels of energy efficiency in terms of socioeconomic benefits. Second, it is easy to calculate and facilitates international comparisons. Energy productivity is defined as the ratio of total GDP to total energy consumption. Accordingly, this paper defines energy efficiency as GDP divided by total energy consumption, where an increase in output per unit of energy consumption indicates an improvement in energy efficiency. A higher value represents greater output per unit of energy consumed, signaling higher energy efficiency.

Core explanatory variable: digital finance (DFL)

The key explanatory variable in this study is digital finance (DFL), measured using the digital inclusive finance index compiled by the Digital Inclusive Finance Research Center at Peking University. This index, employed by Lu et al. (2023) to assess digital finance across provinces, is utilized in this study with a lagged period to mitigate the influence of endogeneity. Inclusive finance has evolved into a comprehensive financial service system, encompassing both payment and credit services (Ren et al., 2023). In China, the practice of inclusive finance is closely linked to the rise of innovative digital finance. The index captures both the breadth and depth of financial inclusion, ensuring balance and comparability across regions, making it a widely adopted measure of digital finance development.

Mediating variable: industrial structural upgrading (OIS)

The mediating variable in this study is industrial structure upgrading (OIS), which reflects the advancement of the industrial structure. The commonly used indicator for measuring industrial structure upgrading is the ratio of tertiary industry output to secondary industry output. However, some scholars (Jiang et al., 2020) have questioned this measurement method. They argue that, although industrial-structure upgrading appears to measure the share relationship among different industries, it is, in essence, a measure of productivity. Simply using the output value ratio to assess industrial structure upgrading fails to capture its fundamental nature and may result in “misleading elevation.” Changes in the proportion of industrial output value and technological structure proportions are defined as nominal industrial structure upgrading. In contrast, structural changes that effectively enhance productivity while improving distributional and environmental outcomes are considered real industrial-structure upgrading (Wu & Liu, 2021). Therefore, industrial structure upgrading consists of two essential dimensions: first, the evolution of proportional relationships, representing the quantitative aspect. Second, improvements in total factor productivity, reflecting the qualitative aspect. Given that “misleading elevation” is widely used—and that this measurement approach may carry inherent limitations and biases—this paper adopts the method proposed by Cheong and Wu (2014). This approach more accurately captures structural changes that raise productivity and generate positive distributional and environmental effects. The formula used to calculate OIS in this study is as follows:

$${OIS}={\left[\mathop{\prod }\limits_{j=1}^{m}\left(\mathop{\sum }\limits_{i=1}^{n}\frac{{Y}_{i}}{Y}\times \frac{{Y}_{i}}{{F}_{i,j}}\right)\right]}^{\frac{1}{m}}$$
(4)

In Eq. (4), Y is the output value; F is the factor of production; i is the i-th industry; j is the j-th factor of production; m is the number of types of factors of production; n is the total number of industries. Due to data availability, the expansion of production factors F in this paper is limited to the case of m = 2 and F = K, L, where K is capital and L is labor.

Control variables

In addressing issues of non-correlation and mitigating multicollinearity, this paper selects six control variables: industrial value added (IVA), marketization index (MI), regional industrial intensity (RII), population density (PC), regional innovation (RI), and foreign investment (LFI).

Industrial Value Added (IVA): IVA is calculated as (current year’s industrial output value/last year’s industrial output value)−1. This indicator not only helps assess the growth of regional industrial economic scale but may also influence energy utilization efficiency. On the one hand, industrial economic development can drive technological innovation and improve production efficiency, thereby optimizing energy use efficiency. On the other hand, regions that overly rely on industrial growth may experience lower overall energy efficiency due to a higher proportion of energy-intensive industries. Marketization Index (MI): This study adopts the China Marketization Index compiled by (Wang Xiaolu et al., 2021) and (Wang Xiaolu et al., 2025) to measure the level of marketization across regions and the extent of government intervention in the market (Ren & Ren, 2024). A higher MI value indicates a higher degree of marketization, meaning that the market plays a greater role in resource allocation, which is more conducive to improving energy efficiency. Regional Industrial Intensity (RII): RII is calculated as the ratio of total industrial output to total GDP. While industrial intensity significantly influences energy efficiency, its impact is ambiguous. On one hand, industrial clustering near market centers can reduce transaction and transportation costs, thus improving energy efficiency. Technological spillovers also contribute to efficiency. On the other hand, industrial agglomeration driven by “policy rent” may hinder technological diffusion and exacerbate regional competition, leading to resource misallocation, overinvestment, and reduced energy efficiency (Wei & Wu, 2021). Population Density (PC): PC is defined as the number of residents at year-end divided by the administrative area of the region. Urbanization tends to generate a “crowding effect,” suggesting a non-linear relationship between population density and energy efficiency. Initially, an increase in population improves infrastructure efficiency, thereby enhancing energy efficiency. However, excessive population density can have the opposite effect. To capture this dynamic, the squared term of population density is included in the model. Regional Innovation (RI): This variable is measured by the number of patents applied per 10,000 people in a region. Technological innovation plays a crucial role in improving energy efficiency (Tao et al., 2024). Foreign Investment (LFI): This is measured by the ratio of foreign direct investment (FDI) to total GDP. FDI promotes energy efficiency by facilitating the spillover of advanced technologies and production techniques (Shinwari et al., 2024). These control variables provide a comprehensive framework for analyzing the factors influencing energy efficiency within a region.

Descriptive statistics

Table 1 presents the definitions, measurement methods, and descriptive statistics for each variable. Notably, the digital finance variable shows a high standard deviation and a significant gap between its maximum and minimum values. This discrepancy can be explained by the rapid growth of digital finance in certain provinces, particularly during the period from 2012 to 2013. Moreover, there is considerable variation in digital finance across provinces, although these differences have been gradually narrowing in recent years. The mediating variable, industrial structure upgrading, has a maximum value of 5.475, a minimum value of 1.784, and a standard deviation of 0.635. Since both the explanatory and mediating variables are ratio-based indicators, their values show a relatively stable data trend, resulting in a modest standard deviation. The descriptive statistics align closely with previous studies using similar indicators, supporting the reliability of the data.

Table 1 Descriptive statistics of variables.
Full size table

Results and discussion

Analysis of overall empirical results without mediating variables

This section examines the direct effect of digital finance on energy efficiency through an initial regression analysis, excluding the mediating variable of industrial structure upgrading. The dependent variable is energy efficiency, while the independent variable is digital finance.

Given that the data sample comes from different provinces and consists of panel data, this study employs a two-way fixed-effects model with province and year fixed effects to control for regional differences and time effects. This approach accounts for the varying conditions across provinces, as well as the expected impacts of policies, technologies, and other factors on energy efficiency across regions. The subsequent empirical analysis also follows this two-way fixed-effects model.

The regression results indicate a positive correlation between digital finance and energy efficiency, significant at the 1% level, with an estimated coefficient of 0.0198 for digital finance (DFL). Column (2) of the analysis includes control variables, and the positive correlation remains significant at the 1% level, though the coefficient decreases to 0.0135. Notably, the constant term is no longer significant, suggesting that omitted variable bias is adequately controlled. The findings presented in Table 2 confirm that digital finance development promotes improvements in energy efficiency. This outcome aligns with theoretical expectations. Digital financial technologies function as an inclusive financial model that leverages artificial intelligence, big data, and other innovations (Wang et al., 2021). These technologies have transformed access and operational efficiency in sectors such as finance, healthcare, education, and financial products (Shahzad et al., 2017). These advancements contribute to more efficient capital and energy use, ultimately enhancing energy efficiency.

Table 2 Validation of the relationship between digital finance without mediating variables and energy efficiency improvement.
Full size table

The impact of digital finance development on industrial structure upgrading

Column (1) of Table 2 reports the results without including control. Since the development of digital finance has a significant positive effect on energy efficiency upgrading, this study further explores the test of mediating effect. Specifically, this section examines the effect of digital finance on industrial structural upgrading, with industrial structural upgrading as the explanatory variable and digital finance as the explanatory variable.

Based on the existing literature, this paper selects the control variables that influence the transformation and upgrading of industrial structure as follows: marketization index (MI), The marketization index measures the degree of market orientation across different regions. A higher level of marketization enables the market to play a greater role in resource allocation, promotes fair competition, and facilitates technological innovation and industrial structure upgrading. The level of foreign investment (LFI), the inflow of foreign capital will bring advanced technology and abundant capital, which will help improve productivity and promote the upgrading of industrial structure. The level of financial development (FDL), the development of financial industry is conducive to the improvement of resource allocation efficiency, promote industrial transformation and upgrading, and promote economic development. First, heteroscedasticity and autocorrelation tests were conducted. The results showed that both data sets had heteroscedasticity between groups and autocorrelation within groups, and the original hypotheses of “homoscedasticity” and “no autocorrelation” were firmly rejected at the 1% significance level. Two-way fixed effects, controlling for province- and year-fixed effects, were continued to control for these effects.

Table 3 presents the regression results with and without control variables. In column (1), digital finance exhibits a positive coefficient of 0.0197 at the 1% significance level, indicating a direct positive effect on industrial structure upgrading. In column (2), after controlling for marketization index, foreign investment level, and financial development level, the coefficient slightly increases to 0.0208 while remaining significant at the 1% level. This suggests that development of digital finance still plays a vital role in promoting industrial structure upgrading even after accounting for other factors. This result may be since all businesses in all stages of digital financial services are transacted online through the network. The optimization and upgrading of the industrial structure are promoted through the channels of industrial digitization and digital industrialization, enabling the transfer of production factors from inefficient to highly efficient sectors. This business model of digitization, centralization, and online services has a solid green attribute. By extending the boundaries of financial services, modern information technology mitigates the high-risk premiums and operating costs created by information asymmetry, thereby promoting industrial-structure upgrading (Ren et al., 2023; Shen & Ren, 2023).

Table 3 Impact of digital finance development on industrial structure upgrading.
Full size table

Testing the mediating effect of industrial structure upgrading on energy efficiency

Since digital finance significantly affects industrial structure upgrading, this paper continues to test for mediating effects. In this paper, the overall measurement results will be tested and analyzed under the condition of adding the mediating variable of industrial structure upgrading, with energy efficiency as the explanatory variable and digital finance as the explanatory variable. The results in Tables 2 and 3 are then combined to investigate the mediating effect of industrial structure upgrading on energy efficiency.

Table 4 integrates the results from previous analyses (columns 1–4) and specifically presents the regression results for testing the mediation effect (columns 5–6). Column (5) and column (6) display the regression results without and with control variables, respectively. The coefficients of the core explanatory variable, digital finance (DFL), and the mediating variable, industrial structure upgrading (OIS), remain significantly positive at least at the 5% level in both regressions. Compared to the baseline model without the mediating variable (columns 1–2), the coefficient of digital finance remains significantly positive after including industrial structure upgrading but decreases in magnitude. This pattern confirms that industrial-structure upgrading serves as a partial mediator in the relationship.

Table 4 Mediation effect test of industrial structure upgrading affecting energy efficiency.
Full size table

The mediation analysis in Section 4.3 (Table 4) demonstrates that industrial structure upgrading (OIS) serves as a partial mediator in the relationship between digital finance and energy efficiency. When OIS is included in the model, the direct effect of digital finance on energy efficiency decreases from 0.014 (p < 0.01) to 0.010 (p < 0.05), indicating that 37% of the total effect operates indirectly through structural upgrading. This aligns with Financial Constraint Theory (Beck et al., 2005): digital finance alleviates capital misallocation, enabling firms in high-value sectors (e.g., green tech) to scale operations while phasing out energy-intensive industries. Regional heterogeneity further validates this mechanism: in the eastern region, OIS mediates 52% of the effect (OIS coefficient: 0.502, p < 0.10), driven by advanced digital infrastructure and innovation spillovers (Tao et al., 2024), whereas structural inertia in central/western regions (e.g., Shanxi’s coal dependency) stifles mediation. These findings underscore that digital finance’s environmental benefits are contingent on institutional and structural readiness, offering policymakers a granular framework to prioritize context-specific interventions.

Robustness test

To enhance the robustness of the prior findings, this paper extends the analysis by incorporating digital finance as the primary explanatory variable and industrial structure upgrading as a mediating variable within the model. Table 5 presents the results of this analysis. The findings confirm that the positive effect of the development of digital finance on energy efficiency remains robust even when controlling for industrial structure upgrading. The coefficients of both the explanatory and mediating variables are positive and statistically significant at the 1% level. These results further substantiate the role of the development of digital finance in promoting energy efficiency through the mediating effect of industrial structure upgrading. The methodology is outlined as follows:

Table 5 Robustness tests.
Full size table

Inclusion of control variables. To further analyze the relationship between digital finance, industrial structure upgrading, and energy efficiency, two control variables are incorporated: financial development level (FDL) and environmental regulation (ER). The financial development level is included as it enhances resource allocation efficiency, facilitates industrial transformation and upgrading, and supports improvements in energy efficiency. Following Yan et al. (2023), the financial development level is measured as the ratio of regional financial institutions’ loan balance to regional GDP. Environmental regulation (ER) is measured as the investment in industrial pollution control divided by the value added by the secondary industry. According to Porter’s hypothesis, moderate environmental regulation stimulates technological innovation and improves energy efficiency, whereas inefficient regulation hampers it. The regression results incorporating these variables are presented in column (1) of Table 5.

Adjusting the time frame. Given that 2013 is widely regarded as the starting point for China’s digital financial development, this paper excludes pre-2013 data to avoid structural changes in the sample. The sample period is thus shortened to 2013–2022 for the regression analysis. The corresponding results are shown in column (2) of Table 5.

Exclusion of municipalities directly under central government control. Significant differences exist across Chinese regions in resource endowment, national policies, and economic development. To address potential bias arising from these differences, this paper excludes the four municipalities directly under central government control—Shanghai, Beijing, Tianjin, and Chongqing—in line with the methodology of Du et al. (2023). The results of this robustness check are presented in column (3) of Table 5.

Alternative measurement of variables. The paper primarily uses total factor productivity to measure industrial structure upgrading. However, a commonly used alternative is the ratio of tertiary industry output to secondary industry output. To test the robustness of the findings, this alternative measure is employed as a proxy for industrial structure upgrading. The results of this approach are presented in column (4) of Table 5.

Mitigating the influence of endogenous factors. Energy efficiency upgrades and industrial structure improvements may exhibit a bidirectional causal relationship. Specifically, the enhancement of industrial structure can facilitate advancements in energy efficiency, while improvements in energy efficiency can, in turn, promote the upgrading of industrial structures. On one hand, a shift in industrial structure from heavy or light industry towards a service-oriented model is likely to reduce energy consumption within the national economic system and enhance overall energy efficiency. On the other hand, improvements in energy efficiency necessitate a transition in the mode of economic development, thereby fostering the continuous optimization and upgrading of industrial structures.

Moreover, issues such as interpolation methods in data processing, measurement errors, and the presence of omitted explanatory variables may contribute to endogeneity problems in the model. To address this, the present study employs an instrumental variables approach to mitigate the impact of endogenous factors on the results, as evidenced by the regression findings presented in column (5) and (6) of Table 5. This study follows the methodology of Han et al. (2021) by selecting two instrumental variables for industrial structure upgrading: the one-period lagged value of industrial structure upgrading and the average level of industrial structure upgrading in neighboring provinces. For the lagged one-period value of industrial structure upgrading, on the one hand, it is inherently related to the current level of industrial structure upgrading. On the other hand, it cannot directly affect the current energy efficiency. Therefore, the lagged value of industrial structure upgrading meets the requirements of relevance and exogeneity as an instrumental variable. As for the average industrial structure upgrading of neighboring provinces, since regional industrial policies and economic spillover effects may affect a province’s industrial structure upgrading but have a minimal direct impact on its energy efficiency. Consequently, this variable meets the relevance and exogeneity requirements of a valid instrumental variable.

The Durbin-Wu-Hausman (DWH) tests for both instrumental variables are insignificant, supporting the acceptance of the null hypothesis that “all explanatory variables are exogenous.” Therefore, based on the data, it can be inferred that industrial structure upgrading is not endogenous, although theoretical endogeneity concerns may still persist. Consequently, the findings in columns (5) and (6) remain relevant and valuable. Additionally, the non-robust F-statistics for both instrumental variables significantly exceed 10, indicating that the selected instrumental variables are not weak.

The results derived from the instrumental variable approach indicate that the coefficients for digital finance development and industrial structure upgrading maintain a positive correlation at the 1% significance level. This reaffirms the presence of a partial mediating effect of industrial structure upgrading.

Following a series of robustness tests, the study reveals that the coefficients of both the core explanatory variable—digital finance—and the mediating variable—industrial structure upgrading—remain significant. Moreover, their directional impacts show minimal variation across the various methodologies employed. These robustness tests encompassed the addition of control variables, alteration of the sample interval, removal of specific samples, modification of variable measurements, and consideration of endogeneity influences. Collectively, these results underscore the robustness of the research findings. They highlight the critical intermediary role of industrial structure upgrading in the relationship between digital finance development and the enhancement of energy efficiency.

Heterogeneity analysis

The regional heterogeneity in the relationship between digital finance development, industrial structure upgrading, and energy efficiency improvement varies significantly across different areas and groups. This section explores these variations. Table 6 presents the results of a heterogeneity analysis for the eastern, central, and western regions. Columns (1) – (3) show the stepwise regression results for the eastern region, (4) – (6) for the central region, and (7) – (9) for the western region.

Table 6 Analysis of heterogeneity in East, Central and West regions.
Full size table

In this study, following the classification method proposed by Li and Li. (2022), provincial-level data were divided into three regions—Eastern, Central, and Western—and stepwise regression analyses were conducted separately for each of these regions. The results are presented in Table 6.

The results from Table 6 indicate that in the eastern region, digital finance development exerts a statistically significant partial mediating effect on energy efficiency improvement. However, this mediating effect is insignificant in both the central and western regions. One possible explanation for the significant effect in the eastern region is that its advanced digital economy allows it to benefit from being an early adopter. The region has attracted a large pool of talent, technologies, and enterprises related to digital finance. This influx promotes rapid development in the tertiary sector, leading to the growth of green industries and the upgrading of the industrial structure, which in turn facilitates the reallocation of resources. Less energy-efficient and more polluting enterprises struggle to survive in this environment, while new, high-quality energy resources are introduced, encouraging the flow and accumulation of high-end energy inputs. This fosters the growth of industries with lower energy consumption and higher marginal output, further enhancing energy efficiency in the eastern region.

In contrast, the digital economy in the central and western regions is relatively underdeveloped. These regions face geographical and environmental challenges, which have hindered their external development compared to eastern cities. Additionally, the central and western regions serve as major energy bases in China, but their energy consumption has historically been reliant on coal and oil. The secondary industry has long dominated these regions, and a lack of innovation resources has further restricted the ability to upgrade industrial structures. Consequently, these regions face significant obstacles to leveraging digital finance to effectively promote industrial upgrading and improve energy efficiency.

Conclusion and policy implications

Research findings

Improving energy efficiency is critical to advancing China’s green transformation in economic and social development. Digital finance plays a key role in this process by driving industrial structure upgrading, which in turn enhances energy efficiency. Improving energy efficiency is also a key strategy for ensuring national energy security, building a modern energy system, achieving the “double carbon” goal, and accelerating the development of a robust energy sector. This paper conducts a theoretical analysis of the mechanisms through which digital finance development, industrial structure upgrading, and energy efficiency improvements are interconnected, based on a review of relevant research literature. This study uses provincial panel data from 30 provinces and cities in China spanning 2011 to 2022. Based on this data, a mediating effect model is constructed with industrial structure upgrading as the mediating variable to explore the impact of digital finance on energy efficiency.

The findings reveal several important insights. First, the development of digital finance significantly contributes to improving energy efficiency. Digital finance enables businesses to transact online through a digital, networked, and intensive operational model that reduces transaction costs, accelerates the circulation of capital and energy, and directly enhances energy efficiency. Second, digital finance also plays a significant role in promoting the upgrading of the industrial structure. Through channels such as industrial digitalization and digital industrialization, digital finance facilitates the transfer of production factors from less efficient sectors to more efficient ones, optimizing the industrial structure. Third, the study establishes an intermediary effect model and finds that the impact of digital finance on energy efficiency decreases when industrial structure upgrading is included in the model. This indicates that industrial structure upgrading plays a crucial mediating role in the relationship between digital finance development and energy efficiency improvement. Fourth, by examining regional differences, the study finds that in China’s eastern region, industrial structure upgrading serves as a significant partial mediator in the effect of digital finance on energy efficiency. However, in the central and western regions, where digital finance development lags, digital finance has neither a direct impact on energy efficiency nor a significant effect on industrial structure upgrading. Robustness tests include adding variables, altering sample intervals, removing outliers, adjusting variable measurements, and employing instrumental variable methods. These tests confirm the stability of the core explanatory variables and mediating variables, as their coefficients remain significant and consistent in both direction and magnitude.

This research provides empirical support for China’s efforts to accelerate digital finance development, upgrade industrial structures, advance the energy transition, and improve energy efficiency. By explicitly modeling the mediation effect of industrial structure upgrading, this study reveals the mechanism through which digital finance indirectly enhances energy efficiency, contributing to a more comprehensive understanding of this relationship. Additionally, the study innovatively employs total factor productivity (TFP) to measure industrial structure upgrading, offering a more precise assessment of structural transformation and its sustainability implications. The findings also offer valuable insights and practical experiences from China for developing countries aiming to enhance their energy efficiency. Future research can extend this study by exploring the applicability of these findings in other regions or analyzing the dynamic relationship between digital finance development and energy efficiency over longer periods.

Policy implications

This research advances the understanding of the relationship between digital finance development, industrial upgrading, and energy efficiency, offering important policy implications.

Firstly, deep integration of high-value-added industries in eastern China. The empirical results demonstrate that industrial structure upgrading (OIS) mediates approximately 52% of digital finance’s (DFL) total effect on energy efficiency (EE) in eastern regions. To amplify this synergy, policymakers should prioritize channeling digital financial resources into innovation-intensive sectors such as green technology and digital services. For instance, provincial governments could establish a “Digital Finance-Industrial Upgrading Linkage Fund” to provide targeted low-interest loans to high-productivity firms in advanced manufacturing and tertiary industries. Loan rates could be dynamically adjusted based on firms’ total factor productivity (TFP) growth, thereby incentivizing structural transformation.

Secondly, addressing structural inertia in central/western China through direct efficiency gains. While the mediating role of industrial upgrading is insignificant in central/western regions, digital finance retains a direct positive impact on EE. Policy efforts should focus on bridging digital infrastructure gaps (e.g., accelerating 5 G deployment and promoting mobile payment platforms like Guizhou’s “Village Chain” initiative) to reduce transaction costs for SMEs and enhance energy allocation efficiency. For coal-dependent provinces (e.g., Shanxi, where coal accounts for 62% of energy use), a “Digital Finance + Clean Energy Transition” program could redirect capital from fossil fuels to renewables via digital credit tools. This initiative should be paired with binding, province-level targets for reducing annual energy intensity.

Thirdly, institutional coordination for long-term sustainability. Given that industrial upgrading mediates 37% of DFL’s total effect nationally, cross-regional institutional frameworks are critical. The central government should mandate a “Green Finance Disclosure Platform” to require digital finance providers (e.g., Ant Group) to publicly report borrowers’ energy efficiency metrics. These data would then be integrated into local officials’performance evaluations. Additionally, adopting Zhejiang’s “patent-collateralized financing” model—where energy-saving technology patents serve as collateral for digital loans—would accelerate technology spillovers.

Limitations and prospects

Although this study provides empirical evidence on the relationship between digital finance and energy efficiency, several limitations warrant attention. First, while provincial-level analysis can reveal regional patterns, it obscures critical intra-provincial disparities. As county-level and township-level data become more accessible, finer-grained analyses could enable more precise policy targeting. Second, although instrumental variables were employed to address endogeneity, incorporating quasi-natural experiments such as fintech regional pilot programs would help establish clearer causal identification strategies. Third, while focusing on macro-level energy performance, this study has not explored micro-level behavioral mechanisms such as household and corporate energy efficiency dynamics. Addressing these limitations would provide a more robust evidentiary foundation for sustainable financial policy design.