Green finance is a concept derived from integrated considerations of the coordinated development of resources, environment, and finance. According to An (2008), green finance is a macroeconomic control policy that promotes energy saving, emissions reduction, and the coordinated development of economic resources and the environment. The “On Building a Green Financial System Guiding Opinions,” released by seven Chinese ministries and commissions on 31 August 2016, clearly states that “green finance” is the economic activity to support environmental enhancement, climate change response, and efficient utilization of resources. According to Zhang et al. (2016), in developing nations such as China, all green investment that can reduce the use of fossil fuels and lower energy consumption per unit is considered green finance. Thus, it may be considered that the development of green finance is vital for advancing green transformation and lowering energy intensity.

The theory of externalities is important in understanding green finance (An and Zi, 2020). Externality refers to the non-market impact of an economic agent conducting market activities on another economic agent. In the early stages of economic development, China’s rapid economic growth brought about excessive consumption of resources and environmental damage (Wang et al., 2020b). Economic actors produced goods and services to benefit themselves while seriously polluting the environment, but they did not pay the corresponding price, leading to negative externalities of environmental pollution. Externalities can result in market failures and inefficient resource allocation. By raising the investment and financing costs of traditional energy-consuming and high-polluting enterprises, green finance incorporates the negative externalities of corporate behavior into production costs, effectively helping to internalize the negative externality costs of environmental pollution. This influences corporate production decisions and development planning, reduces production energy consumption to some extent, and stimulates the development of green environmental projects and enterprises.

As the largest energy user and importer in the world, China is highly dependent on other nations for its energy needs, which in some ways restrains the country’s ability to expand economically. In terms of energy consumption, China is still mainly using fossil fuels. As a non-renewable and competitive commodity, an increase in the production and consumption of fossil fuels leads to a decrease in quantity. The vigorous development of green finance may successfully respond to the competitive challenges brought about by the decreasing supply of fossil fuels (Shu and Gu, 2020). On the one hand, green finance guides the flow of capital to renewable energy, energy conservation, and emission reduction through financial means and promotes the development and utilization of non-fossil energy. As a replacement for fossil fuels, the development of non-fossil fuels can effectively change the energy consumption structure and reduce the consumption of fossil fuels. On the other hand, green finance can force the reduction of investment and financing of high-energy-consuming and heavy-pollution projects and enterprises, reduce the consumption of fossil fuels, including coal, oil, and natural gas, and promote the green and low-carbon transformation of energy (Zhang and Wang, 2019).

Based on the analysis of green finance and energy intensity, hypothesis H1 is proposed.

H1:

The development of green finance can effectively reduce energy intensity.

In this study, we construct the transmission mechanism of green finance development impact energy intensity (see Figure 1). Green finance may guide enterprises to choose energy-saving industries and adjust the conventional industrial structure to make it “green.” Xu et al., 2018 found a correlation between the development of green credit and the upgrading of industrial structures through a gray correlation analysis. Furthermore, they analyzed the differences in the effects of green credit on industrial restructuring in each region. From the perspective of new institutional economics, Shao and Fang (2021) pointed out that regional green finance development and industrial structure have a coupled and coordinated relationship. Huang (2022) argued that green finance helps Vietnam change its industrial structure, which can lower its energy intensity by raising the share of the tertiary sector in the national economy.

The path of green finance to promoting industrial restructuring includes three main aspects: capital orientation, industrial integration, and information dissemination. First, by controlling the credit cost, financing efficiency, and return on investment of different projects, green finance directs the flow of funds from high-energy-consuming and high-polluting industries to energy-saving and environmental protection industries. It encourages the growth of environmental protection industries and facilitates the transformation of the industrial structure to green and rationalization. Second, green finance promotes the flow of resources to green industries, realizes the reallocation of production factors between industries, forms a new capital system, and promotes industrial structure integration and optimization (Feng et al., 2022). Third, producers and investors can adjust their respective production plans and investment strategies based on market prices for green products and enterprise disclosure of environmental information, thereby accelerating the growth of green industries and optimizing industrial structures.

Industrial upgrading is critical in addressing climate change (Wang et al., 2022). Luan et al. (2021) pointed out that upgrading the industrial structure can validly reduce energy intensity. Given the differences in each industry’s nature, there are remarkable differences in the energy intensity of various sectors. The secondary industry, primarily industrial, has the highest energy intensity. If the share of industries with higher energy intensity is greater and rises faster, then the overall energy intensity increases. Li and Lin (2014) analyzed the non-linear effect of industrial structure on energy intensity in China and found that when the share of secondary industry in GDP is less than 40.435%, industrial structure upgrading leads to a decrease in energy intensity. Guo (2016) examined the effect of industrial structure on the convergence of energy intensity in China and discovered that improving the industrial structure dramatically accelerates the rate of energy intensity decline and accelerates the convergence process of energy intensity between regions. Adjusting the industrial structure is the way to achieve stable economic growth and effectively reduce energy intensity (Wang S. et al., 2021).

Based on the analysis of the transmission mechanism of green finance impact energy intensity, hypothesis H2 is proposed.

H2:

Industrial structure upgrading plays a mediating role in reducing energy intensity with green finance.

A combination of changes in GDP and energy consumption influences the changes in energy intensity. A certain amount of energy that produces more economic benefits or a certain amount of GDP using less energy can reduce energy intensity (Lin and Sun, 2011). Similar to traditional finance, the development of green finance can contribute significantly to economic growth (Zhou et al., 2020). Green finance has the power to increase energy efficiency and reduce energy consumption (Hong et al., 2022). Green finance can directly reduce energy intensity by promoting economic growth and reducing energy consumption. However, there is a scale effect on the impact of green finance on energy intensity. Green finance promotes economic growth, and the expansion of economic output forces enterprises to increase their energy demand and consumption. Green finance may lessen the inhibitory effect on energy intensity through the scale effect. Therefore, the influence of green finance on energy intensity is not always fixed under the combined effects of changes in GDP and energy consumption.

Simultaneously, there are variations in the effects at each stage of the green finance development process. At the early stage of green finance development, pertinent policies are not yet perfect, and only a limited amount of financial resources can be mobilized. Additionally, the green industry’s investment cycle is lengthy, expensive, and slow, and the time lag of the policy prevents green finance from playing its role in time (Shi and Shi, 2022). With the gradually increasing level of green finance development and the accumulation of financial resources, green finance can strongly encourage the green upgrading and transformation of traditional industries and the sprouting and growth of green environmental industries. When the degree of green finance development is too high, the regional disparity in green finance development expands, and the implementation of green finance policies by local governments may be highly divergent, which can reduce the effectiveness of green finance (Huang and Chen, 2022).

Based on the above analysis, we propose hypothesis H3.

H3:

There is a threshold effect for green finance development to affect energy intensity.

Green finance is an essential innovation of conventional environmental regulation; ignoring its spillover effects would underestimate its impact effects (Ouyang et al., 2022). Scholars have gradually focused on the spatial spillover effects of green finance with the rise of the spatial economics theory. Dong and Fu (2018) examined the characteristics of the geographical distribution of green finance development and focused on the spatial spillover impact of green finance concentration areas. Wang et al. (2021b) also pointed out that the development of green finance fueled the growth of local clean energy and encouraged the production and utilization of clean energy in the surrounding provinces. The spatial spillover effect of green finance is mainly reflected in two aspects: industrial agglomeration and demonstration effect.

On the one hand, in green finance development, regions with location and policy advantages can draw green financial resources, including capital, market, technology, and labor, forming the agglomeration effect of green industries. The agglomeration of green industries can drive the development of the upstream and downstream sectors, promote the integration of the green finance industry chain, improve the production efficiency of the entire industry, and promote energy conservation. On the other hand, through experience exchange, information sharing, and resource interchange, regions with a high level of green finance development can play a good demonstration role and promote green finance development in neighboring regions. Therefore, green finance has an apparent positive externality, which drives the development of local energy-saving and environmentally friendly industries, and also has a radiation diffusion effect on the surrounding areas and promotes interregional synergistic emission reduction.

Based on the analysis of the spatial spillover effects of green finance, hypothesis H4 is proposed.

H4:

There is a spatial spillover effect of green finance development on energy intensity.

This study selected panel data from 30 provinces (excluding Tibet, Hong Kong, Macao, and Taiwan) in China from 2007 to 2019 as the research sample, using data mainly from Wind, CEADs, the National Bureau of Statistics, and provincial statistical yearbooks, with individual missing data supplemented by interpolation.

The descriptive statistics for each variable are shown in Table 1. The mean energy intensity (EI) value is 0.981, indicating that energy intensity is still high in China and that most provinces have more room for reduction. The energy intensity ranges from 0.208 to 3.507 and varies greatly among Chinese provinces. The average green finance index (GF) is only 0.350, and overall, the level of green finance development in China is low. Meanwhile, the green finance index of different regions ranged from 0.168 to 0.943, indicating the uneven development of green finance in China. Table 1 also shows significant differences in industrial structure (IS), energy price index (P), R&D intensity (RD), foreign trade intensity (Trade), and FDI intensity (FDI) in different provinces.

Based on H1, to verify the impact of green finance on energy intensity, the following baseline model was constructed: E I i t = α 0 + α 1 G F i t + α n c o n t r o l i t + v i + ε i t (1) where i represents province, t represents the year, E I i t represents the energy intensity of province i in year t, G F i t represents the green financial index of province i in year t, c o n t r o l i t is a set of control variables, v i represents the individual effect of province i that does not change over time, and ε i t represents the random disturbance term.

Based on H2, we verified whether industrial structure upgrading is a mediating variable for green finance that affects energy intensity. This study drew on Wen and Ye (2014) to conduct a mediating effect test. First, we tested whether the effect of green finance on the industrial structure is remarkable according to Eq. 2. The industrial structure index is introduced in Eq. 3 to test the effects of green finance and industrial structure on energy intensity. Finally, the validity of the mediation effect of the industrial structure was judged based on the regression coefficients of Eqs 2, 3, respectively. The specific form of the mediating effect model was set as follows: I S i t = β 0 + β 1 G F i t + β n c o n t r o l i t + v i + ε i t (2) E I i t = γ 0 + γ 1 G F i t + γ 2 I S i t + γ n c o n t r o l i t + v i + ε i t (3) where I S i t represents the industrial structure index of province i in year t. If the coefficient of energy intensity β 1 and coefficient of industrial structure γ 2 are remarkable, the mediating effect is significant. β 1 γ 2 represents the degree of influence of green finance on energy intensity through the industrial structure.

Based on H3, we chose the panel threshold model to explore the non-linear effects of different stages of green finance development on energy intensity. The threshold model is expressed as follows: E I i t = β 0 + β 1 G F i t · I G F i t ≤ γ + β 2 G F i t · I G F i t > γ + β n c o n t r o l i t + v i + ε i t (4) where green finance (GF) is both a threshold variable and a threshold effect variable, γ represents the threshold value, I · is an indicator function that takes the value of 0 or 1 and satisfies the condition in parentheses as 1; otherwise, it is 0. Eq. 4 is a single-threshold model that can be expanded to a double- or three-threshold model, according to the actual situation.

Based on H4, the spatial spillover effect of green finance is further discussed. Before conducting spatial econometric analysis, it is necessary to determine whether the explained variables are spatially correlated. In this study, we used the global Moran’s I index to test the spatial autocorrelation of energy intensity, and the Moran’s I index is defined as follows: M o r a n ’ s   I = ∑ i = 1 n ∑ j = 1 n w i j y i − y ¯ y j − y ¯ S 2 ∑ i = 1 n ∑ j = 1 n w i j (5) where S 2 = ∑ i = 1 n y i − y ¯ 2 n , y ¯ = ∑ i = 1 n y i n , n is the total number of regions, y i represents the observed value of province i, and w i j is the spatial weight. Moran’s I take values ranging from −1 to 1, where 0 indicates no spatial autocorrelation, greater than 0 indicates the presence of positive spatial autocorrelation, and less than 0 indicates the presence of negative spatial autocorrelation.

To explore the spatial spillover effect of green finance on energy intensity, the spatial interaction terms of green finance, energy intensity, and control variables are introduced in Eq. 1, which is further extended into a spatial Durbin model as follows: E I i t = α 0 + ρ W E I i t + α 1 G F i t + α n c o n t r o l i t + θ 1 W G F i t + θ n W c o n t r o l i t + v i + ε i t (6) where ρ is the spatial autoregressive coefficient; W is the spatial weight matrix; and θ 1 and θ n are the elasticity coefficients of the spatial interaction terms of the explanatory and control variables, respectively.

Considering that the distance between two regions may impact the spatial spillover effect of green finance, we used a geographic distance matrix as the spatial weight matrix in this study. The specific form is as follows: w i j = 1 d i j 2 , i ≠ j 0 , i = j (7) where d i j represents the geographical distance between places i and j.

We first used the Hausman test to select fixed and random effects, and the test result is p = 0.0002; therefore, we used the fixed effects model in this study. The effect of green finance on energy intensity was examined using the baseline model. As shown in Columns (1) and (2) of Table 2, the regression coefficients for green finance are all remarkably negative at the 1% level. When control variables are added, the energy intensity decreases by an average of 1.030 units for every 1 unit increase in the green finance index, demonstrating that green finance development validly reduces energy intensity and supports H1.

Additionally, the regression coefficients of the control variables in Column (2) are all statistically significant at the 10% level, demonstrating that various factors influence changes in energy intensity. The coefficients of the control variables are generally consistent with those expected. The regression coefficients of both energy price and R&D intensity are remarkably negative, indicating that energy intensity decreases as energy price and R&D intensity increase. Additionally, the coefficient of foreign trade intensity is remarkably negative, demonstrating that regions with a higher degree of openness to the outside world can reduce energy intensity by introducing green production technologies. The coefficient of FDI is remarkably positive, indicating that the introduction of FDI increases China’s energy intensity, supporting the “pollution refuge” hypothesis.

As shown in Table 2, based on the verification in Column (2) that green finance can reduce energy intensity, Column (3) shows the link between green finance and industrial structure, where the regression coefficient of green finance is obviously positive, suggesting that the development of green finance validly promotes the industrial structure upgrading. Column (4) shows the effects of green finance and industrial structure on energy intensity, where the coefficient of industrial structure is obviously negative at the 1% level, revealing that upgrading industrial structure can reduce energy intensity. Meanwhile, the regression coefficient of green finance is still remarkably negative in Column (4), but the absolute value is small compared to Column (2). According to the test mechanism of the mediating effect, it can be seen that industrial structure upgrading plays a part of mediating role in the process of reducing energy intensity by green finance development, and the mediating effect accounts for 60.25% of the total effect, confirming H2.

The existence of the threshold effect was first tested and passed the double threshold test. Therefore, the model was set as a double-threshold model. The results of the double-threshold regressions in Table 3 show that the impact of green finance on energy intensity can be divided into three stages: when the green finance index is less than 0.199, green finance development has no statistically significant on energy intensity. When the green finance index is greater than 0.199 and less than 0.512, the green finance regression coefficient is significant, and the effect of green finance suppressing energy intensity is strong, and when the green finance index is greater than 0.512, the coefficient of green finance is also significant, but the effect of green finance on suppressing energy intensity decreases.

Overall, when the green finance index exceeds 0.199, the role of green finance in affecting energy intensity begins to emerge and shows the non-linear characteristic of the diminishing marginal effect, confirming H3. This may be because green finance is still immature at the early stage of development, and the mobilization effect on social capital is much lower than that of traditional finance, preventing it from fully fulfilling its function in resource allocation. When the green finance development index is too high, it occupies other advantageous resources and has a crowding-out effect on other industries’ technological upgrading and product development; thus, the impact on energy intensity decreases.

First, a spatial autocorrelation test of energy intensity was conducted to calculate Moran’s I index of energy intensity from 2007 to 2019, and the results are presented in Table 4. The results show that Moran’s I indices of energy intensity from 2007 to 2019 are all significantly greater than 0 at the 1% level, suggesting a significant positive spatial correlation between energy intensities among provinces. To better understand the spatial clustering characteristics of energy intensity, we drew the scatter plots of Moran’s I index of energy intensity in 2007, 2011, 2015, and 2019 (see Figure 2). It can be seen that most provinces fell in the first and third quadrants in 2007, 2011, 2015, and 2019, indicating that the spatial distribution of energy intensity has remarkable spatial clustering characteristics (i.e., high values clustered with high values and low values clustered with low-value clustering).

Spatial autocorrelation tests indicate the existence of spatial autocorrelation of regional energy intensity; therefore, a spatial econometric model can be used. The LM, Wald, and LR tests were first performed to analyze whether the spatial Durbin model could degenerate into a spatial error model or a spatial lag model. All the results passed the 10% significance test, so it was concluded that the fixed-effect spatial Durbin model should be selected. As there is a lot of interaction information in adjacent regions, the interpretation of the results based on regression coefficients alone would be biased; therefore, we decomposed the effect of green finance on energy intensity into direct, indirect, and total effects (Chen et al., 2016). Specifically, the effect of green finance in the region on the energy intensity of the region is a direct effect, the effect of green finance in neighboring regions on the energy intensity of the region is an indirect effect or spatial spillover effect, and the sum of the direct and indirect effects is the total effect.

Table 5 presents the direct, indirect, and total effects estimated using the spatial Durbin model. As shown in Table 5, the regression coefficients of the direct, indirect, and total effects of green finance are all significantly negative at the 5% level. This indicates that not only will the development of green finance in the region reduce the region’s energy intensity, but also the development of green finance in the neighboring regions will reduce the region’s energy intensity. In other words, the development of green finance reduces the local energy intensity and aids in lowering energy intensity in other areas; that is, green finance has a spatial spillover effect, confirming H4.

There are significant regional variances in the degree of green finance development among provinces owing to factors such as geographic location, level of economic development, and degree of openness to the outside world (Zheng et al., 2020). Meanwhile, the energy structure is directly impacted by the resource endowment of various provinces, leading to differences in the energy intensity in each province (Wang et al., 2020a). In this study, based on economic development level and geographical location, the sample data were divided into three parts: eastern, central, and western for regional heterogeneity testing. The regression results are presented in Table 6.

The regression results in Table 6 show that the regression coefficients of green finance in the east, central, and west are all remarkably negative at the 5% level, which shows that green finance in all three regions can effectively reduce energy intensity. However, the effect of green finance is more pronounced in the central and western regions compared to the eastern regions. That is, there is regional heterogeneity in the effect of green finance on energy intensity. This may be because the eastern region is at the stage of green transformation and industrial upgrading, and various environmental regulation tools are relatively complete, so green finance cannot validly enhance the effectiveness of existing environmental policies. The central region is under the double pressure of economic development and green transformation, and the traditional environmental regulation policy has an issue of one-size-fits-all. Green finance can offer financial support for enterprise technological innovation to promote green transformation, which can supplement strict environmental regulation policies to a certain extent. The development of green finance in the western region is concentrated in a few cities, which works in harmony with and complements existing environmental regulation policies. It has a greater influence on energy intensity (Zhu et al., 2021).

To ensure the robustness of the baseline model regression results, we performed robustness tests, such as excluding singular values, replacing core explanatory variables, and using the instrumental variables method.

(1) Exclusion of singular values. The variables were subjected to a tail reduction below 1% and above 99% to eliminate the interference of outliers, and then the baseline model was regressed. The regression results are shown in Column (1) of Table 7. After excluding outliers, the effect of green finance on energy intensity is still remarkably negative at the 1% level, proving that the regression results of the baseline model are robust.

Based on the fact that green finance can validly reduce energy intensity, this study empirically analyzed the influence mechanism of green finance on energy intensity by using the mediating effect model, threshold model, and spatial Durbin model based on province-level data of China from 2007 to 2019, from the perspective of industrial structure upgrading as the entry point, and on the basis of constructing a green finance index. The main conclusions are as follows: First, the development of green finance validly reduces energy intensity, in which industrial structure upgrading plays a mediating role in the influence of green finance on energy intensity. The energy price index, R&D intensity, and foreign trade intensity reduce energy intensity, whereas FDI intensity positively affects energy intensity. Second, green finance has a non-linear effect on energy intensity. The energy intensity can be effectively suppressed only when the green finance index is greater than 0.199, while the suppression effect decreases when the green finance index exceeds 0.512. Third, green finance has a remarkable spatial spillover effect on energy intensity, and the development of green finance helps form a pattern of synergistic promotion of emission reduction among regions. Fourth, in the regional heterogeneity analysis, the inhibitory effect of green finance on energy intensity is more substantial in the central and western regions, which may be due to the different endowment advantages in the eastern, central, and western regions.

Based on theoretical and empirical analysis, we make the following recommendations.

First, the guidance role of green finance should be utilized. The government should accelerate the formulation of green finance development policies, complete incentive mechanisms and regulatory mechanisms, and address the problem of environmental externalities. Through the establishment of special funds for green development, financial subsidies, tax incentives, and other measures to guide more capital to flow into green industries, fully utilize the fiscal policy’s capacity to leverage capital. Differential credit policies should be strictly implemented to raise polluting enterprises’ financing costs and promote industries’ green transformation. Financial institutions should improve the green financial risk management system, provide more green financial products and services, and address the problems of asymmetric information and maturity mismatch of green investment and financing.

Second, green finance should be developed according to local conditions. The eastern region should further develop and improve the green finance system by upholding the existing green finance policies. It should take advantage of the developed financial system and a high degree of openness to the outside world; new green financial products should be created, and more advanced green financial development models should be aggressively explored. For the central and western regions, government subsidies, tax concessions, credit incentives, and other policies should be increased, and the entry threshold for green credit should be suitably lowered. Green financial resources such as capital, talent, and technology should be introduced to encourage the development of green finance. Green development evaluation and assessment should be conducted to enhance enterprises’ green awareness and mobilize local governments’ enthusiasm to develop local green finance.

Third, build a regional coordinated development mechanism. Given the spatial spillover effect of green finance, the demonstration and radiation role of green financial center cities should be considered. Coordination and cooperation among regions should be strengthened, and advanced achievements and experiences in green finance should be promoted and borrowed to drive the development of green finance in the region. With the help of modern network technology, specialized green finance platforms should be built to integrate and share resources. The interconnection of policies, technologies, and markets among regions should be promoted to facilitate the coordinated development of green finance and fully release the spatial contribution of green finance in reducing energy intensity.