Introduction

The haze of the COVID-19 epidemic has not faded, with local military conflicts one after another, increasing geopolitical tensions and increasing energy demand in 2022. The world is caught in an energy crisis, and the energy crisis is getting worse and worse. Since the reform and opening up, China’s economy has created a miracle in the world, with sustained and rapid growth. However, with the economic growth, the total amount of energy consumption is also increasing day by day, and the negative impact of ecological and environmental problems caused by energy consumption on economic development is becoming more and more obvious. According to the data of Bei**g Hengruihang Information Consulting Co., ltd (2021) released by British Petroleum Company (BP), since 2009, the total energy consumption of China has surpassed that of the United States, ranking first in the world and becoming the largest energy consumer in the world. In 2009–2019, the total energy consumption of China increased from 33.3 billion tons of standard coal to 48.3 billion tons of standard coal, with an average annual growth rate of 3.79%, almost twice the growth rate of global energy consumption; According to the proportion of total energy consumption in the world, the top four countries in 2019 are China, the USA, Russia, and India. Therefore, how to improve energy utilization efficiency, realize the transformation of energy consumption mode and reduce energy intensity has become an important theoretical and practical problem to be solved urgently in China. The Report to the 20th National Congress of the Communist Party of China (CPC) clearly points out that Reaching peak carbon emissions and achieving carbon neutrality will mean a broad and profound systemic socio-economic transformation. Based on China’s energy and resource endowment, China will advance initiatives to reach peak carbon emissions in a well-planned and phased way in line with the principle of building the new before discarding the old. China will exercise better control over the amount and intensity of energy consumption, particularly of fossil fuels, and transition gradually toward controlling both the amount and intensity of carbon emissions. Undoubtedly, promoting low-carbon energy transformation, effectively reducing energy intensity and actively and steadily promoting peak carbon emissions and achieving carbon neutrality are important measures to promote the green development of China. The Fourth Plenary Session of the 19th CPC Central Committee put forward for the first time that data should be used as production factors to participate in distribution according to their contributions, which marked the beginning of the era of large-scale release of dividends in the digital economy (Du et al. 2020). Digital financial inclusion applies a variety of digital and intelligent tools to promote the financial innovation of inclusive finance through digital financial services. It can reach a wider range of service groups and provide more abundant financial service products. It has the characteristics of low cost and wide range and provides an effective solution to the “last mile” problem in inclusive financial services. Digital financial inclusion has greatly expanded the depth and breadth of application in inclusive finance and laid a good foundation for industrial structure adjustment and industrial employment mobility in basic fields with its convenience of carrying the Internet and its essential characteristics of “popularization” and “benefit.” Meanwhile, digital financial inclusion has also broken the ‘”two-eight principle” existing in traditional finance, providing financial services for groups that have long been excluded from the scope of traditional financial services, thus effectively improving the living environment of traditional financially disadvantaged groups (Wei and Chen 2022). In March 2021, the 14th Five-Year Plan for National Economic and Social Development of the People’s Republic of China and the Outline of the 2035 Vision Goals proposed reducing energy efficiency consumption per unit of GDP by 13.5%. The National Development and Reform Commission released the Dual Control Plan for Improving the Intensity and Total Amount of Energy Consumption (Development and Reform Investment 2021), which aims to encourage energy conservation and efficiency improvements in different areas. This plan is very important to China’s current “double carbon” goal, which is to reduce carbon emissions by twice as much as they are now. This paper will use spatial econometric method to explore the spatial effect of digital inclusive finance on energy intensity in China, and combine industrial share with employment share to form industrial structure distortion system, which will be used as an intermediary variable to reveal the transmission mechanism of industrial structure distortion in the relationship between digital inclusive finance and energy intensity, and then from the perspective of digital inclusive finance, give some policy suggestions on energy conservation and emission reduction in China.

Literature review

As an indispensable resource, energy plays a key role in supporting the orderly operation of society and realizing sustainable development. Meanwhile, the financial industry, as the core sector of economic development, has a profound and extensive impact on production and life. Many scholars have done a lot of research on the relationship between the development of digital inclusive finance and energy intensity, and achieved fruitful results.

Digital financial inclusion and energy intensity

There is a plethora of literature and research findings on the link between economic progress and energy intensity. However, the existing related studies have not reached a consistent conclusion. Scholars use different research methods to find that there are two opposite effects between them. As a combination of digital technology and financial innovation, digital financial inclusion relies on modern technological innovation to break through geographical restrictions and expand the breadth of financial services while promoting the flow of information (King and Levine 1993; Fu et al. 2021; Akbar et al. 2022). Sadorsky (2010), and Chang (2015), think that digital financial inclusion can lead to an increase in energy intensity. The reason is probably that digital financial inclusion can bring cheaper credit services to demanders, prompting them to buy large household appliances with high energy consumption, such as televisions, refrigerators, and washing machines, thus increasing the consumption of traditional energy.

From different perspectives, Ouyang and Li (2018) draw different conclusions and found that the development of digital financial inclusion promotes the increase of energy utilization efficiency, thus effectively reducing energy intensity. Zhang et al. (2020) think that with the expansion of the financial development scale and the enhancement of corporate social responsibility awareness, financial institutions will increase their support for the clean energy industry out of consideration of system, risk, reputation, and other aspects, resulting in a high correlation between enterprises’ access to financing and the reduction of energy intensity. Digital financial inclusion has the characteristics of convenience, low transaction cost, and universal access, which can ease household credit constraints and promote household consumption (** to improve energy efficiency and reduce energy intensity. Inclusive finance can decrease the threshold for obtaining financial services, improve the quality of financial services, make it possible to promote regional innovation, and at the same time help to promote entrepreneurial activities. As the product of the organic combination of digital technology and finance, the healthy development of digital financial inclusion cannot be separated from the technical support provided by innovation and entrepreneurship (Zhang and Li 2022). With the improvement of innovation and entrepreneurship, digital finance can further rely on scientific and technological innovation, management innovation, product innovation, etc. to enhance the effect of energy efficiency. The stronger the regional innovation ability and the more active the entrepreneurial activities, the more conducive it is to the research and development of energy-saving technologies and the development of energy-saving industries, and the more it can improve energy efficiency and reduce energy intensity. Accordingly, this paper puts forward.

  • Hypothesis 1: Digital financial inclusion promotes energy efficiency and its influence has a spatial spillover effect.

Distortion of industrial structure due to digital financial inclusion

Digital financial inclusion can quickly convey the information about financial services by using digital technology, alleviate the financial exclusion caused by geographical location, lower the threshold of financial services, guide the transfer of funds to enterprises with core technologies and high production efficiency, and promote the adjustment of logistics, people flow, and information flow among industries, thus promoting the adjustment of employment and industrial structure. Digital financial inclusion can provide channels for innovation and entrepreneurship in economic growth, optimize social assets, increase income, narrow the income gap, promote reasonable and fair distribution, and achieve inclusive growth. With the continuous development and improvement of the financial system, financial services can effectively alleviate information asymmetry and transaction costs, thereby stimulating capital accumulation and technological innovation, and promoting industrial structure optimization and economic development. The digital development of inclusive finance optimizes the information asymmetry within and across industries, reduces the external financing cost of enterprises, and significantly affects the employment structure and industrial growth from the “horizontal effect” and “structural effect” channels. Digital financial inclusion can quickly and accurately match the demand ends of the industrial chain, provide funds for industrial development in time and effectively, provide in-depth digital financial service support, and further promote the rational and efficient allocation of financial resources, as well as promote industrial optimization and adjustment (Tang et al. 2022).When the industrial share of a department is not equal to or deviates from the employment share, industrial distortion will occur inevitably. Digital financial inclusion has obvious advantages in optimizing and upgrading industrial structures, but with the scarcity of resources, regional cooperation and competition never stop, and the spillover effect of spatial association also affects the relationship between digital financial inclusion and industrial structure distortion. Accordingly, this paper puts forward:

  • Hypothesis 2: Digital financial inclusion can effectively reduce industrial structure distortion, and its influence has a spatial spillover effect.

The intermediary transmission mechanism of the industrial structure distortion between digital financial inclusion and energy intensity

Generally speaking, the energy consumption of the primary industry is low. The secondary industry is dominated by industries with high energy consumption, so the demand for energy is naturally high. The demand for energy in the tertiary industry is generally small. With the continuous improvement of the economic development level, the industrial structure is gradually upgraded, and the secondary industry develops rapidly, taking the lead in surpassing the primary industry in proportion. The total energy demand is constantly increasing, and the energy intensity is increased accordingly. With the development of industrialization to a certain level, the industrial structure is advanced, and the tertiary industry accounts for the largest proportion. At this time, the total energy consumption decreases and the energy intensity continues to decline. Inclusive finance can make it more efficient to use money and other factors of production, which can lead to a change in the way industries are set up. The characteristics of digital financial inclusion such as easy availability, low cost, convenience, and diversification can stimulate the growth and upgrading of consumption, and the diversification of consumer demand will drive the development of modern service industries and emerging industries in the region, thus promoting the upgrading of industrial structure. Digital financial inclusion can improve the financing environment of innovative and entrepreneurial groups, stimulate small and medium-sized micro-enterprises and low-income groups to release the vitality of innovation and entrepreneurship, and promote the adjustment of industrial structure (Shao 2022). Moreover, with the help of information technology, digital financial inclusion can accurately screen and realize the efficient allocation of financial resources among industries, thus promoting the upgrading of industrial structures. With the continuous optimization of industrial structures, the proportion of high-end manufacturing industries and tertiary industries has increased relatively. Both industrial upgrading and industrial structure upgrading will affect the energy intensity, and the level of energy consumption within each industry depends on the energy consumption level of each industry within the industry and its proportion within the industry, or the proportion of energy-consuming industries within the industry. The evolution of industrial structure is a process in which output and employment share are transferred between departments. Once the output share of an industry or sector deviates from the employment share, the industrial structure will be distorted. This kind of structural distortion will inevitably have an adverse impact on the efficiency of resource allocation, and it is likely to hinder the due decline of energy intensity. Accordingly, this paper puts forward:

  • Hypothesis 3: The distortion of industrial structures enhances energy intensity and its influence has a spatial spillover effect.

  • Hypothesis 4: Digital financial inclusion restrains energy intensity by reducing the distortion of industrial structure and has a spatial spillover effect.

Research design

Space weight matrix setting

Considering that the adjacent weight matrix, geographical distance matrix, and economic distance matrix are symmetric matrices, the construction of these three spatial weight matrices may not conform to the actual situation of asymmetric spatial relations among provinces with different levels of economic development. Therefore, nested weights are constructed based on adjacent relations and economic distances to better reflect the spatial relations among provinces. Specifically, based on the traffic distance and economic distance between regions, spatial weight matrices Wd and We are constructed respectively. Among them, the element wkjd of Wd is the reciprocal of the shortest expressway kilometers in two provincial capitals, and the diagonal element is set to 0. The element wkje of We is the absolute reciprocal of the difference of GDP per capita between provinces from 2011 to 2020, where the diagonal element is set to 0. However, there may be some deviations in measuring the spatial association between regions by any distance standard alone. For this reason, referring to the practice of (Shao et al. 2022), this paper constructs a geo-economic nested spatial weight matrix Wn = ωWd + (1 − ω)We which consider both geographical distance and economic distance. Through simulation, it is found that the fitting degree of the model reaches the optimal state when ω is 0.3. The definition and calculation method of matrix are shown in Table 1. The above three matrices are standardized when used for parameter estimation.

Table 1 The definition of weight matrix
Full size table

Selection of variables and data sources

Energy intensity is selected as the explained variable in this paper. Energy intensity is expressed by the ratio of total energy consumption to actual GDP (ton of standard coal/ten thousand yuan). The Index of Digital Financial Inclusion is the development level of digital financial inclusion. It is selected as the core explanatory variable in this study, and the digital financial inclusion index by Peking University is used to measure the development level of digital financial inclusion. Industrial structure distortion. Industrial structure distortion is chosen as the variable that shows how digital financial inclusion affects energy intensity. It is measured by the industrial structure distortion index, which is the square root of the difference between the employment share and output share of three industries in an economy.

Based on previous research, the control variables in this paper are urbanization, technological progress, energy price, and degree of openness. Urbanization brings the concentration of production factors such as population, capital, and technology. Urbanization requires a lot of resources to invest in infrastructure construction and maintenance, which is likely to cause the change of energy intensity. Urbanization is measured by the proportion of urban resident population to the total population. It is the main way for patent research institutions to express innovative output, and it is also the concrete representation of technological progress. This research chooses the number of patent grants as the variable of technological progress. Energy prices are measured by the retail price index of fuel commodities. The degree of opening to the outside world refers to the degree to which a country or region’s economy is open to the outside world, which is embodied in the degree to which the market is open. This paper selects the proportion of total import and export to GDP, that is, foreign trade dependence represents the degree of opening to the outside world.

The data used in this paper are annual, and the research samples are drawn from the Chinese mainland’s other 31 provinces (autonomous regions and municipalities), excluding Hong Kong, Macau, Taiwan Province, and Tibet, and span the years 2011–2020. Data from the China Statistical Yearbook, the China Science and Technology Statistical Yearbook, the Peking University Digital Financial Inclusion Index (2011–2020), and all provinces, autonomous regions, and municipalities directly under the Central Government were used. For a few missing data, it is supplemented and improved by means of mean method, analogy method, interpolation method, or other estimation methods. To eliminate the influence of price factors, the local GDP, fuel price index, and foreign direct investment were all reduced in 2011. In order to avoid the influence of heteroscedasticity, this paper makes logarithmic processing on the basis of flattening the data. Table 2 reports the statistical description of related variables.

Table 2 The descriptive statistics of variables
Full size table

Model setting

Referring to Yang et al. (2022), this paper combines the analysis method of intermediary effect with the spatial Durbin model (SDM) to investigate the intermediary effect of the industrial structure distortion between digital financial inclusion and energy intensity. The spatial Durbin model can better measure the spatial spillover effect than other spatial lag models (SAR and SEM). The model measures three parts in total: The correlation between local explained variables and local core explained variables and controlled variables is the first; the spatial autocorrelation between local explained variables and nearby explained variables is the second; and the spatial autocorrelation of local explanatory variables with core explanatory variables and control variables in neighboring areas is the third. This paper constructs the following model:

$$\mathrm{ln}\;{e}_{\mathrm{it}}^{i}={\alpha }_{0}+{\alpha }_{1}{\mathrm{ln\;df}}_{it}+{\alpha }_{2 }X \;{\mathrm{control}}_{\mathrm{it}}+{\alpha }_{3} W\;{\mathrm{ln\;e}}_{\mathrm{it}}^{i}+{\alpha }_{4} W\;\mathrm{ln}\;{df}_{\mathrm{it}}+{\alpha }_{5} WX\; {\mathrm{control}}_{it}+{\mu }_{i}+{v}_{t}+{\varepsilon }_{it}$$
(1)
$$\mathrm{ln\;{sd}_{\mathrm{it}}}={\beta }_{0}+{\beta }_{1}\mathrm{ ln}\;{df}_{it}+{\beta }_{2}\mathrm{ X }\;{\mathrm{control}}_{it}+{\beta }_{3} W\;\mathrm{ ln }\;{sd}_{\mathrm{it}}+{\beta }_{4} W\;\mathrm{ ln }\;{df}_{\mathrm{it}}+\beta 5\; WX {\mathrm{control}}_{it}+{\mu }_{i}+{v}_{t}+{\varepsilon }_{it}$$
(2)
$$\mathrm{ln}\;{e}_{\mathrm{it}}^{i}={\mathrm{\alpha }}_{0}^{^{\prime}}+{\mathrm{\alpha }}_{1}^{\mathrm{^{\prime}}}\mathrm{ln}\;{df}_{it}+{\mathrm{c}}_{1}\mathrm{ ln}\;{ sd}_{\mathrm{it}}+{\mathrm{\alpha }}_{2}^{\mathrm{^{\prime}}} X\; {\mathrm{control}}_{it}+{\mathrm{\alpha }}_{3}^{\mathrm{^{\prime}}}W\;\mathrm{ln}\;{e}_{it}^{i}+{\mathrm{\alpha }}_{4}^{\mathrm{^{\prime}}}W\;\mathrm{ln}\;{df}_{it}+{\mathrm{c}}_{2} W\;\mathrm{ln}\;{sd}_{it}+{\mathrm{\alpha }}_{5}^{\mathrm{^{\prime}}} WX {\mathrm{control}}_{it}+{\mu }_{i}^{\mathrm{^{\prime}}}+{v}_{t}^{\mathrm{^{\prime}}}+{\varepsilon }_{it}^{\mathrm{^{\prime}}}$$
(3)

where lneiit is the dependent variable, indicating the energy intensity; lndfit is the independent variable, indicating the digital financial inclusion index; lnsdit is the intermediary variable, indicating the distortion degree of industrial structure; Xcontrolit represents other control variables; W is the spatial weight matrix; μi and μi′ represent individual effects; vt and vt′ represent time effects; εit and εit′ represent random disturbance terms. According to the test steps of the intermediary effect, if the regression coefficients α1, β1, and c1 in Formulas (1), (2) and (3) are all significant, it indicates that the intermediary effect is significant. On this basis, if the regression coefficient α1′ in Formula (3) is significant, it indicates that some intermediary effects exist. If α1′ is not significant, it indicates that there is a complete intermediary effect.

Empirical analysis

Spatial correlation test

Testing the spatial correlation of economic variables is the key step to establishing a spatial econometric model. The Moran’s I index reflects the similarity of attribute values of spatial adjacent or adjacent regional units, and the spatial autocorrelation coefficient is used to quantitatively describe the spatial correlation of things, which is usually tested by global autocorrelation and local autocorrelation.

Global autocorrelation

To investigate the spatial effect of digital financial inclusion on energy intensity in China, the global Moran’s I index should be used to test the spatial correlation, and the calculation formula is as follows:

$$\mathrm{I}={\mathrm{n}}_{i}+1{\mathrm{n}}_{j}=1 {\mathrm{nw}}_{ji}^{i}=1{\mathrm{n}}_{\mathrm{j}}=1{\mathrm{nw}}_{ij}\left({\mathrm{X}}_{i}-{\mathrm{X}}_{i}\right)\left({\mathrm{X}}_{j}-{\mathrm{X}}_{j}\right)\mathrm{i}=1\mathrm{n}({\mathrm{X}}_{i}-{\mathrm{X}}_{i})$$
(4)

where wij is the weight matrix, it represents a nested weight matrix formed by weighting geographical distance and economic distance in the paper, ** and Xj are the variable values of province i and province j respectively, ** and Xj are the expected average of ** and Xj respectively, and n is the number of provinces. I ∈ [-1, 1], I > 0, indicating positive correlation; I < 0, indicating negative correlation. The greater the value of I, the greater the spatial distribution correlation; The smaller the value of I, the smaller the spatial distribution correlation; When I = 0, it means that the spatial distribution is random. Table 2 shows Moran’s I of the energy intensity and Digital financial inclusion. As can be seen from Table 2, Moran’s I of each variable is very positive and significant, indicating that there is a positive spatial dependence in every province Table 3.

Table 3 The Moran’s I of Energy Intensity and Digital Financial Inclusion
Full size table

Local autocorrelation

To further investigate the spatial autocorrelation degree of a specific province, the nested spatial weight matrix is also used to draw the scatter plot of local Moran’s I of energy intensity and digital financial inclusion in 2020. The local Moran’s I index is calculated as follows:

$$\begin{array}{cc}\mathrm{Local\; Moran\mathrm{^{\prime}}}\mathrm{s }\;{\mathrm{I}}_{i}={\mathrm{V}}_{ij}=1,\; \mathrm{j}\ne {i}_{n}\; {\mathrm{M}}_{ij}\;{\mathrm{V}}_{j}\end{array}$$
(5)

where Vi and Vj are the standardized values of variables of province i and province j respectively, and Mij is the standardized spatial weight. Figure 1 and 2 represent the Moran scatter plots of energy intensity and digital financial inclusion in 2011 and 2020, respectively. The first and third quadrants show positive spatial correlation, which are high-high agglomeration and low-low agglomeration, respectively. The second and fourth quadrants represent spatial negative correlation, representing low–high agglomeration and high-low agglomeration respectively. Most of the scatter points in Fig. 1 and 2 are located in the first and third quadrants, which further shows that there is a certain spatial agglomeration characteristics between them. Therefore, spatial factors should be considered when studying their relationship, otherwise it will affect the accuracy of estimation.

Fig. 1
figure 1

The Moran scatter plot of digital financial inclusion and energy intensity in 2011

Full size image
Fig. 2
figure 2

The Moran scatter plot of digital financial inclusion and energy intensity in 2022

Full size image

The local Moran indexes of energy intensity and digital financial inclusion are calculated. Due to the limited space, only the local Moran index distribution table of energy intensity and digital financial inclusion in 2020 is shown. From Table 4, it can be seen that most of the local Moran’s indexes of 31 provinces in mainland China are distributed in the first and third quadrants, indicating that there is a strong spatial correlation between local energy intensity and inclusive finance, which is consistent with the results of the global Moran’s index, that is, the influence of spatial factors should be considered in the study of the impact of digital financial inclusion and energy intensity.

Table 4 The local Moran’s index distribution of energy intensity and digital financial inclusion in 2020
Full size table

According to the test results of global and local spatial correlation, there is a significant spatial correlation between energy intensity and digital financial inclusion. Based on this, the spatial factor should be introduced into the model.

Analysis of spatial econometrics

Spatial econometric model selection and configuration

To obtain the best regression results, before analyzing the spatial econometric model, a series of tests are needed to determine the specific form of the spatial econometric model. The paper selects the space panel model according to the following steps. The first step is to conduct LM inspection. The geo-economic nested spatial weight matrix is used to test the SEM model and SAR model, and the OLS is tested for spatial correlation, including LM-Lag and robust LM-Lag tests, LM-Error and robust LM-Error tests. If the LM test passes on one or two models, it is necessary to use the Wald test to determine the choice of spatial model. If the test results reject both the original assumptions, the spatial Durbin model is selected.

As can be seen from Table 5, LM_test_Error and robust LM_test_Error pass the 1% significance test, while LM_test_Lag and robust LM_test_Lag also pass the 1% significance test, which indicates that the research object of this paper has the dual effects of spatial lag and spatial error autocorrelation, so the spatial Durbin model should be considered. The second step is to conduct the LR test and the Wald test. Whether the spatial Durbin model can be transformed into a spatial lag model and a spatial error model is tested by the LR test and Wald test, respectively. The test results show that the original assumption that the SDM model can degenerate into a SAR model and a SEM model is rejected, and the SDM model is accepted. Table 6 shows that LR test values are 34.19 and 34.22, respectively, and both of them reject the original hypothesis at a significant level of 1%, which indicates that the spatial Durbin model will not degenerate into the spatial lag model and the spatial error model, and the SDM model is the most suitable for empirical analysis. The third step is to conduct the Hausman test. After the SDM model is selected, a Hausman test is carried out on the SDM model to determine whether a fixed effect or random effect is selected for analysis. The test results show that the chi2(6) value is 45.27, and the P value is 0. With a significance level of 1%, the SDM model should be looked at using the fixed effect method.

Table 5 The spatial dependence test
Full size table
Table 6 The results of LR test and the Wald test
Full size table

Analysis of spatial econometric model results

To compare and test the robustness of parameter estimation for each variable, this paper also lists the estimation results of OLS, SAR, and SEM models with double fixed effects. Table 7 reports the measurement estimation results of the space panel model. Because the spatial lag term of energy intensity in SAR and SDM models is an endogenous variable, and the spatial lag term of error term is included in the SEM model, the maximum likelihood method is chosen to estimate these models to obtain consistent parameter estimation. In addition, this paper comprehensively judges and selects the best model among the above spatial models by statistics such as goodness of fit, log-likelihood logarithmic function value, etc. The results show that the SDM model is the best model in this empirical research, both from the log-likelihood value and goodness of fit (Han and **e 2017)[38]. Therefore, it is reasonable to choose the spatial Durbin model with fixed effect to analyze the impact of digital financial inclusion on energy intensity. According to the regression results of the SDM model, the coefficient of digital financial inclusion is -0.0020, and it has passed the 1% significance test, which shows that digital financial inclusion has a significant inhibitory effect on energy intensity, verifying hypothesis 1.

Table 7 The regression results of three models
Full size table

As the spatial Durbin model explains the spatial economic correlation among provinces, its parameter estimation results cannot directly reflect the real effect of direct effect and spatial spillover effect. Referring to the partial differential method proposed by Le Sage and Pace (2009) , the influence coefficients of each variable on energy intensity are decomposed into direct effect, indirect effect, and total effect.

It can be seen from Table 8 that the direct effect, indirect effect, and total effect of digital financial inclusion are all negative and pass the significance level of 1%, which indicates that digital financial inclusion can not only reduce the energy intensity in this area but also in other areas. From the regional level or the overall level, the promotion of digital financial inclusion can significantly curb energy intensity. From the standpoint of control variables, increased urbanization can reduce not only the energy intensity of the local area but also the energy intensity of China as a whole, because increased urbanization not only causes the agglomeration of production factors such as population, capital, and technology but also leads to an increase in total energy demand. Meanwhile, it can also produce economies of scale, improve the level of regional economic development, and promote the increase of production efficiency and the decrease of energy intensity. On the contrary, the improvement of technological progress and foreign trade dependence not only lead to the improvement of local energy intensity but also adversely affect the energy intensity of China as a whole. The reason is probably that technological progress is measured by patents while the patent conversion rate or utilization rate of all the provinces is not high. Over the years, according to the fact that many regions in China blindly pursue economic growth, the threshold for FDI entry is lowered, and the dependence on foreign trade and energy intensity change in the same direction, it is once again verified that China has indeed become the migratory place of polluting industries in developed countries. That is, the “pollution paradise hypothesis” has been established in China. As for energy prices, the moderate increase in energy prices in a region can indeed curb the local energy intensity and the overall energy intensity, but the spatial effect on energy intensity in other regions is not significant.

Table 8 The result of spatial effect decomposition
Full size table

Robustness test

To further enhance the reliability of the research conclusion, this paper tries to test the robustness by the following three methods: First, changing the weight matrix. The adjacent weight matrix is used to replace the nested matrix, and then the two-way fixed effect space Durbin model is used to re-estimate; second, considering the subjective error caused by sample selection, the sample data from 2013 to 2020 will be intercepted here for re-estimation; third, eliminating outliers. To avoid the influence of the possible extreme values in the research sample on the estimation results, all variables are trimmed with 1% bilateral tail reduction and regressed again. Table 9 shows that the symbols of explanatory variables are basically consistent with the previous results, so the spatial effect of digital financial inclusion on energy intensity is robust.

Table 9 The results of robustness test
Full size table

Intermediary effect test

Digital financial inclusion can curb energy intensity. This part uses the intermediary effect method to further test the transmission mechanism of hypothesis 4. The specific steps are as follows: Step 1 is to check whether the coefficient of digital financial inclusion in Formula (1) is significant. If it is significant, it indicates that the total effect of digital financial inclusion on energy intensity exists, and then goes to the next step; Step 2 is to check whether the coefficient of digital financial inclusion in Formula (2) and the coefficient of industrial structure distortion in Formula (3) are significant. If they are both significant, there is an indirect effect, and you go directly to step 4. If at least one of them is not significant, then you go to the next step. Step 3 is to conduct a Bootstrap test. If it is significant, there is an indirect effect, and the test proceeds to the next step. Otherwise, the test is terminated. Step 4 is to check whether the coefficient of digital financial inclusion in Formula (3) is significant. If not, it is a complete intermediary effect. If it is significant and is less than in Formula (1), it means that the distortion of the industrial structure plays a partial intermediary effect and the intermediary effect accounts for β1c1/(β1c1 + α1′) of the total effect. The specific test results are shown in Table 10.

Table 10 Intermediary effect test results
Full size table

The regression coefficient of digital financial inclusion in column (2) is significantly negative at the level of 1%, which indicates that digital financial inclusion can alleviate the distortion of industrial structure and verifies hypothesis 2. The regression coefficient of industrial structure distortion in column (3) is significantly positive at the level of 1%, indicating that industrial structure distortion is not conducive to the decrease of energy intensity, which verifies hypothesis 3; The regression coefficient of digital financial inclusion is significantly negative at the level of 1%, and its absolute value is smaller than the regression result of the benchmark model in column (1). According to the test process of intermediary effect, the distortion of industrial structure plays a part of the intermediary effect in the influence of digital financial inclusion on energy intensity, which proves that hypothesis 4 is valid. The total effect of inclusive finance on energy intensity is − 0.0020, which is significant. The direct effect of digital financial inclusion on energy intensity is − 0.0018, and the indirect effect of digital financial inclusion on energy intensity through industrial structure distortion is − 0.0002(= − 0.0113 × 0.0205), that is, the intermediary effect accounts for 11.40% of the total effect (= 0.0113 × 0.0205/0.0020).

Discussion

The empirical results of this paper show that digital financial inclusion has greatly expanded the depth and breadth of application in inclusive finance, and the development of digital financial inclusion can promote the decrease of energy intensity. Moreover, the distortion of industrial structure has a partial intermediary effect in the relationship between digital financial inclusion and energy intensity. This is basically consistent with the research conclusions of Shao (2022) on the relationship between finance and energy. At the same time, digital financial inclusion promotes the decrease of energy intensity and its influence has spatial spillover effect; When the sector’s industrial share is not equal to or deviates from the employment share, it will inevitably lead to the distortion of industrial structure, and the spillover effect of spatial correlation will also affect the relationship between digital financial inclusion and the distortion of industrial structure. Obviously, digital financial inclusion effectively curbed the distortion of industrial structure, and its influence had spatial spillover effect; Industrial structure distortion enhances energy intensity and its influence has spatial spillover effect. Digital financial inclusion can restrain energy intensity by reducing industrial structure distortion and has spatial spillover effect. Therefore, it is necessary to speed up the development of digital financial inclusion, attach importance to the structural effect of digital financial inclusion, increase the innovation of products and services of digital financial inclusion, strengthen the supervision of digital financial inclusion, develop green low-carbon industries, and reduce the distortion of industrial structure, so as to give full play to the role of digital finance in reducing energy intensity, accelerate the green transformation of development mode, and help green development and the smooth realization of “double carbon” goals.

Conclusions and policy recommendations

Conclusion

Reducing energy intensity is a powerful starting point to help achieve the goal of “double carbon.” The rapid development of science and technology promotes financial innovation, giving birth to a new financial development model of digital financial inclusion, and provides new kinetic energy for curbing energy intensity. Under this background, it is of great practical significance to analyze the relationship among digital financial inclusion, industrial structure distortion, and energy intensity. Based on the provincial panel data from 2011 to 2020, this paper empirically tests the influence of digital financial inclusion on energy intensity and the mediating effect of industrial structure distortion on the relationship between digital financial inclusion and energy intensity by using the spatial Durbin model and the mediating effect method. The research shows that digital financial inclusion can promote the decline of energy intensity and the distortion of industrial structure plays a partial intermediary effect in the influence of digital financial inclusion on energy intensity. That is, the influence of digital financial inclusion on energy intensity is partly realized through the distortion of industrial structure.

Policy recommendations

Accelerating the development of digital financial inclusion

First, it is necessary to strengthen the top-level design, introduce relevant support and incentive policies, optimize the business environment, relax market access restrictions, improve the risk prevention and control mechanisms, and build a good ecological pattern of benign interaction, complementary advantages, and win–win cooperation among relevant participants in inclusive finance. Second, there is a need to strengthen financial education, especially in remote rural areas; improve the financial knowledge level and financial literacy of rural financial consumers and other inclusive financial service groups; bring more economic entities into the financial service system; and enjoy the “dividend” of the development of digital financial inclusion. Third, it is urgent to consolidate the infrastructure of digital financial inclusion, especially to accelerate the construction of digital villages, promote the upgrading of rural information infrastructure, and let more people enjoy high-quality and low-price mobile network services. Fourth, regional cooperation needs to be strengthened, the flow of funds and talent in regional digital financial inclusion needs to be encouraged, regional cooperation needs to be strengthened, and everyone needs to work together to help energy intensity go down.

Increasing digital financial inclusion product and service innovation

There is a need to give full play to the advantages of technology and finance, integrate modern technology into digital financial inclusion, complete the integration of “energy” and “finance,” and realize the development mode of “energy + digital financial inclusion.” It is necessary to make full use of new technologies such as meta-universe, artificial intelligence, big data, blockchain, and cloud computing to carry out the process and business innovation; actively develop online finance; improve the efficiency of financial demand response, approval, and handling; break through the space and time constraints to supply financial services; effectively reduce the cost and risk of financial supply and demand docking and resource allocation; make it easier for energy to trade across time and space; and promote energy consumers to actively participate in market activities according to effective rules.

Strengthening the supervision of digital financial inclusion

Green finance and inclusive finance generally face the problem of information asymmetry, and most of the clients in inclusive finance lack information, credit, and guarantee. It is difficult for financial institutions to evaluate their credit, which makes the basic cost of inclusive finance high. Digital financial inclusion uses modern technology to improve the quality and efficiency of financial services, but it also brings about changes in financial risks. Therefore, it is necessary to persist in upholding integrity and innovation, promote the application of big data, Internet, and other technologies in the field of green finance, continue to build a good ecological environment for the development of digital green finance and digital financial inclusion, strengthen behavior supervision and function supervision, improve the financial consumer complaint and handling mechanism, effectively protect the security of consumer data and funds, and crack down on illegal activities such as financial fraud, illegal fund-raising, endangering data and network security, and infringing on personal privacy, so that digital financial inclusion can better serve economic development.

Reducing the distortion of the industrial structure

The distortion of industrial structure leads to the improvement of energy intensity, which plays a partial intermediary role in the spatial effect of digital financial inclusion in restraining energy intensity. Therefore, the government should eliminate the differences between urban and rural residents in employment, education, medical care, income, consumption, public investment, etc.; break down the institutional barriers of urban and rural population mobility; build an inclusive financial service system in urban and rural areas; crack the disadvantages of uneven capital supply; and form a new pattern of urban and rural development featuring mutual promotion between workers and peasants; mutual complementarity between urban and rural areas; comprehensive integration; and common prosperity. Second, the industrial structure distortion must be reduced by increasing government investment in all levels of education; improving the balanced allocation mechanism of public resources between urban and rural areas; improving the education level of the labor force; promoting the cross-regional and inter-departmental mobility of the labor force; improving the efficiency of departmental allocation of labor resources; changing the long-standing mismatch of public resources; and reducing the distortion of industrial structure. Third, there is a need to give full play to the restraining effect of industrial structure distortion on energy intensity by speeding up the replacement of traditional industries by high-tech industries, eliminating backward production capacity, breaking the flow barriers of production factors such as cross-regional, cross-industry, and cross-department people flow, logistics, and information flow, and optimizing resource allocation.