Green Finance Policies, Urban Green Energy Efficiency and Regional Relative Disparities—Causality Tests Based on Dual Machine Learning

Abstract

China’s sustained economic growth and industrialisation have led to increasingly serious problems of resource consumption and environmental pressure, making green development an inevitable choice for the country’s transformation and development. Green finance policies are becoming an increasingly important tool for increasing the use of green energy in cities. Using a dual machine learning (DML) model, this paper assesses the specific impact of green finance policies on green energy efficiency in Chinese cities, the mechanism of action, and regional disparities. The analysis is based on objective and scientific measurement of the level of green finance policies and green energy efficiency in 282 Chinese cities at prefecture level and above from 2006 to 2022. Benchmark regression results show that green finance policies significantly promote green energy efficiency in Chinese cities, passing a rigorous robustness test. Green bond policies are found to have the greatest promotional effect, whereas green support policies are found to have no significant effect. The results of the heterogeneity analysis suggest that green finance policies are more effective in promoting green energy efficiency in resource-based cities, cities with established industrial bases, and more developed cities. The results of the impact mechanism suggest that green finance policies can promote green energy efficiency by allocating the three internal urban factors of labour, capital and technology. The results of the analysis of regional disparities demonstrate that green finance policies effectively reduce disparities in urban green energy efficiency at the national level, between the north and south, and between coastal and inland regions. However, they also widen the disparities between central and peripheral cities within each province, hindering balanced regional development. This paper makes relevant policy recommendations based on this.

1. Introduction

Since the “dual carbon” target was defined in 2020, prefecture-level cities across China are facing severe challenges in energy transformation and green development [1]. Resource-based cities like Taiyuan in Shanxi Province and Shenyang in Liaoning Province, with their heavy industrial structures, show much lower green energy efficiency compared to coastal cities like Shenzhen and Shanghai [2]. In addition, by the end of 2023, the balance of domestic and foreign currency green loans in China has reached approximately 4.22 trillion US dollars, up 36.5 percent year on year, but far lower than the level of developed countries [3]. This stark contrast highlights the substantial gap between green finance support and urban energy transition needs. In-depth research on how green finance policies impact urban energy efficiency not only clarifies the mechanisms of financial instruments in energy transformation but also provides scientific evidence for optimizing policy design and narrowing regional development gaps. This holds profound practical significance for driving China’s urban green and low-carbon transitions and achieving high-quality development.

Cities are the most densely populated areas, the most intense economic activity and the greatest sources of energy consumption and carbon emissions. Cities play a crucial role in achieving the green transition and promoting green finance policies. Many green technologies and environmental projects are often initially piloted and implemented at city level [4]. Therefore, policies and their effects at city level can provide valuable experience and reference material for other regions. Urban Green Energy Efficiency (UGEE) is a key indicator of the efficiency with which energy is used and the environmental impact of urban economic activities. UGEE typically involves maximising economic output while minimising energy consumption and environmental costs [5]. This efficiency depends not only on technological innovation and industrial upgrading, but also on the efficient allocation of resources. The rational combination of the three elements of labour, capital and technology is a key way of optimising the allocation of resources and improving green energy efficiency in cities. Currently, academics commonly use data envelopment analysis (DEA) or stochastic frontier analysis (SFA) to measure green energy efficiency [6,7]. These methods take multi-dimensional indicators such as energy inputs, economic outputs, and environmental impacts into account, thus providing a quantitative basis for policy formulation. As China has the second largest economy in the world, improving green energy efficiency in cities is an important initiative for promoting high-quality economic development. It is also directly related to the country’s ability to achieve carbon neutrality.

In this context, green finance policies are an important tool for encouraging the development of a green economy. Wang et al. [8] identified seven forms of green finance policies: green credit, green investment, green insurance, green bond, green support, green fund and green equity. These policies promote the rational allocation of resources and technological innovation by guiding capital flows, thereby improving energy efficiency and reducing environmental pollution. In terms of labour factor allocation, Jiang and Jiang [9] state that green finance policies promote the growth of green employment opportunities indirectly by providing green industries and projects with low-cost financial support. The rapid growth of the green industry has led to a significant increase in demand for labour on green projects, providing employment opportunities for a large number of workers in related fields. This promotes the development of workers’ green skills, facilitates the transformation of the labour market, and accelerates the learning and application of green technologies, thereby increasing cities’ green energy efficiency. In the green energy sector, in particular, labour transfers and optimisation of employment structures have increased the green technological capacity of the workforce. This has resulted in more efficient energy use, reduced waste, and the promotion of a green economy. Additionally, Li et al. [10] stated that green credit support facilitates financing for green projects and provides financial security for clean energy industry development. Green credit helps the clean energy sector to scale up and promotes energy efficiency in cities by changing labour mobility patterns and employment structures. Transferring labour from traditional energy sectors to green industries optimises the employment structure and promotes the widespread use of green technologies. This improves energy efficiency and promotes green energy efficiency in cities. In terms of allocating the scale factor, Zhang [11] pointed out that green finance policies achieve optimal fund allocation by increasing the financing scale of green enterprises while restricting that of polluting enterprises. This policy initiative has enabled green energy projects to progress from the initial stage of small-scale, inefficient pilot schemes to a phase of large-scale promotion and application with financial support. As financing increases, green businesses are able to speed up the development of technology and construction of facilities, thereby making green energy more efficient. In the field of clean energy in particular, larger-scale capital provides sufficient guarantees for technological innovation and industrial upgrading, thereby promoting overall energy efficiency improvements. Wang et al. [12] also highlighted that financing constraints are an effective means through which green finance policies can reduce the carbon emissions of major polluters by limiting their access to finance. Green finance policies curb the scale of financing for polluting enterprises, forcing them to accelerate their transformation. This pushes them to reduce investment in high-pollution, high-energy-consumption projects and promotes their development in the direction of more environmentally friendly, low-carbon technologies. This financial allocation mechanism promotes the green transformation of polluting enterprises and improves the energy efficiency of society as a whole, thus contributing to green energy efficiency improvements in cities at a macro level. In terms of the allocation of technological factors, Lee and He [13] state that green finance policies indirectly impact land ecological security by promoting the development of agricultural technologies, as well as providing crucial support for the innovation and implementation of green energy technologies. The implementation of green finance policies supports the transformation of traditional agriculture and creates conditions for the development and promotion of green, energy-related technologies. This policy enables cities to gradually introduce more advanced and efficient green technology innovations in energy production and use, thereby improving overall energy efficiency. Green finance policies encourage the use of green energy technologies by supporting technological development. This increases energy efficiency and reduces waste of resources. Meanwhile, Shi and Zhang [14] also emphasised that setting up pilot zones for green finance policies had a significant impact on the quantity and quality of enterprises’ green technological innovations. These pilot zones create a favourable environment for businesses to innovate in the area of green technologies by providing dedicated financial support and policy incentives. These pilot zones offer enterprises more financing opportunities for researching and developing new low-carbon and energy-saving technologies, as well as promoting the popularisation and application of green technologies. Green finance policies enhance technological innovation, increasing the efficiency of green energy and accelerating the green transformation of cities. Thanks to accelerated technological innovation, cities have become more efficient and sustainable in their energy management, utilisation and production, which has driven an overall increase in green energy efficiency.

However, despite existing literature revealing the positive impact of green finance policies on green energy efficiency, controversy and disagreement still exist. Due to differences in the implementation of local policies, the effectiveness of green finance policies may fluctuate in different regions, making their actual impact somewhat uncertain. Meng et al. [15] pointed out that green finance policies have a more significant impact on private enterprises and in areas with higher levels of financial development and stricter environmental regulations. In addition, China has a vast territory, and the economic development level of cities in the eastern region is much higher than that of cities in the western region, while the resource endowment of cities in the western region is higher than that of the eastern region [16]. These differences may affect the promoting effect of green finance policies on the green energy efficiency of cities. Specifically, Ren et al. [17] point out that, nationally, green finance policies are more effective in eastern coastal cities (e.g., Suzhou in Jiangsu Province and Hangzhou in Zhejiang Province) than in inland cities (e.g., Lanzhou in Gansu Province and Guiyang in Guizhou Province). This is closely related to the more developed market mechanisms and higher acceptance of green technologies in eastern cities. Li et al. [18] point out that, in terms of north–south differences, green finance policies are more likely to improve energy efficiency in southern cities due to their mild climate and lighter industrial structure. In contrast, the effects of these policies are often limited in northern cities due to their high heating demand and heavy industry. Tang et al. [19] point out that, in terms of coastal versus inland differences, coastal cities benefit from an open economy and international green financial cooperation. The use of green bond and green fund is also more advanced in these cities, resulting in a more significant policy effect. Deng et al. [20] point out that, within provinces, central cities (e.g., Guangzhou and Chengdu) implement green finance policies far more effectively than smaller, peripheral cities due to their greater access to resources and stronger policy implementation. These regional disparities indicate that green finance policies must be adapted to local circumstances, taking into account urban resources and regional characteristics.

Although the relationship between green finance policies and urban green energy efficiency has received extensive attention from academics, existing research still suffers from the following shortcomings. On the one hand, the construction of green finance policies indicators lacks specificity. Current studies typically represent green finance policies using dummy variables for a specific policy [21,22], but this approach does not fully reflect the actual role of green finance policies. The diversity and complexity of green finance policies should be assessed using more specific indicators. As a result, existing studies have not made full use of the coefficient of variation when creating specific indicators for green finance policies. This has led to an inadequate assessment of policy effectiveness. On the other hand, although some studies have begun to focus on the effects of green finance policies in different regions [23,24], the impact of such policies on regional disparities, particularly in reducing the gap between the north and south, coastal and inland areas, and provincial centres and peripheral cities, has been less frequently considered. This remains unclear.

Based on this, this paper takes 282 prefecture-level and above cities in China from 2006 to 2022 as its research subject. It combines theoretical and empirical analyses and uses DML regression analysis to systematically assess the impact of green finance policies on urban green energy efficiency. It also explores the role of these policies and regional disparities, verifying their positive impact on urban green energy efficiency. Therefore, there are three possible contributions to the margin: first, the objective and scientific measurement of green finance policies indicators. Rather than simply using dummy variables for a particular policy, this study synthesises more specific green finance policies indicators through coefficients of variation. This approach can more accurately quantify the actual effects of green finance policies by combining theory and empirical evidence. This helps policymakers to gain a deeper understanding of the policies’ multidimensional impacts and provides them with a more precise basis for decision-making. Secondly, the data were processed using a DML approach. Unlike the traditional difference-in-differences (DID) method, this study employed this approach. This method can effectively solve problems of endogeneity and sample selection bias. It can also capture nonlinear effects between green finance policies and urban green energy efficiency. This enhances the accuracy and reliability of research results. This innovative method of analysis offers a fresh perspective for future related research and, finally, considers regional differences from multiple angles. This study focuses on the heterogeneity of green finance policies across different regions. It considers the differences between northern and southern regions, coastal and inland areas, and provincial cities. The study also systematically analyses the role of green finance policies in reducing the disparity in green energy efficiency development between regional cities. An in-depth discussion of the effects of policies in different regions makes it possible to provide more targeted and differentiated recommendations for optimising green finance policies and promoting coordinated regional development.

The subsequent sections of the paper are organised as follows: Section 2 focuses on the theoretical analysis and research hypotheses; Section 3 is on model construction, variable setting and data sources; Section 4 is on benchmark regression results, robustness tests, heterogeneity tests and mechanism tests; Section 5 further discusses the regional disparities; and Section 6 is on the conclusions, policy recommendations and research gaps.

2. Theoretical Analysis and Research Hypothesis

Green finance policies are an important tool for achieving sustainable development. It has become one of the key instruments for promoting the transformation of the national economic structure and optimising resource allocation [15]. In China, green finance policies are an important way to promote green development and achieve low-carbon goals in the face of environmental pressures and energy shortages. They are also an effective way to improve energy efficiency and enhance the utilisation of green energy [16]. In this section, we will analyse the potential impact of green finance policies on green energy efficiency in cities and develop relevant research hypotheses.

What impact do green finance policies have on urban green energy efficiency? This paper focuses on the following two theories. One is the theory of resource allocation efficiency proposed by Alfred Marshall, an economist in the 19th century. The efficiency of resource allocation is crucial to economic growth and social welfare. Liu and Li [25] highlighted that green finance policies can optimise the allocation of resources and encourage the swift growth of low-carbon and environmentally friendly industries by offering market incentives. In China, where environmental protection requirements are becoming increasingly stringent and green development goals are gradually being clarified, green finance policies (e.g., green bond, green fund, and green credit) are becoming one of the key tools for promoting energy conservation, reducing emissions, and encouraging green technological innovation. Meng et al. [26] also state that China’s green credit policies significantly reduce the cost of capital for green projects by providing adequate financial support in the form of lower interest rates and longer loan terms. This policy provides support to help cities access more funding when implementing energy-efficient and emission-reduction technologies. It reduces their reliance on traditional energy sources and promotes the widespread use of renewable energy. Therefore, green finance policies optimise the energy structure and enhance the efficiency of energy use by promoting green technology. Additionally, Liu and Yang [27] noted that green finance policies promote the interactive role of the market mechanism and policy guidance via the government-market synergy effect. Through green finance policies, the government signals to the market, guiding the flow of capital to the low-carbon economy. It ensures that green projects are adequately financed through regulations and standards, while promoting the prioritisation of projects with high green benefits. At the same time, the competitive nature of the market encourages innovation and cost reduction in green technologies, promoting their widespread adoption and enhancing energy efficiency. In summary, green finance policies promote balance between economic development and environmental protection in cities by optimising resource allocation. They also provide solid theoretical and practical support for enhancing urban green energy efficiency.

The other is the theory of externalities proposed by Marshall in 1890. This theory emphasises that external costs and benefits resulting from market behaviour cannot usually be resolved by market mechanisms alone. Therefore, government policy interventions are required in order to address market failures. Xie [28] pointed out that traditional energy use is often accompanied by negative externalities, such as pollution, resource waste, and ecological damage. These not only deteriorate the environment but also increase the overall burden on society. Green finance policies reduce negative externalities, helping capital to flow to green industries and improving the efficiency of energy use. Environmental pollution is a growing problem in China, with increasing carbon and air pollutant emissions from energy consumption and production leading to external costs. Therefore, the introduction of green finance policies not only aims to promote the research and development of green technologies, but also to reduce negative externalities and improve the efficiency of green energy. Jamali et al. [29] stated that the government promotes research into renewable energy and energy-saving and emission-reduction technologies, as well as providing green financing and guiding the flow of capital to green projects. The flow of capital accelerates the commercialisation of green technologies and effectively mitigates the negative externalities associated with traditional energy sources. It reduces pollution emissions and improves the efficiency of green energy use in cities. The effectiveness of green finance policies has gradually become apparent in practice. Xing et al. [30] noted that the rapid growth of the green bond market has provided substantial financial support for renewable energy projects, contributing to the accelerated development of green initiatives, including wind and solar energy projects. Thanks to green finance policies, capital has gradually flowed into the low-carbon economy, promoting the rapid development of green energy and enhancing energy efficiency. Figure 1 below shows the specific flowchart. Among them, an upward arrow means an increase, while a downward arrow means a decrease.

In summary, green finance policies can greatly enhance the efficiency of urban green energy usage by optimising resource allocation and minimising negative externalities. In light of the above theoretical analysis, we propose the following research hypotheses for testing:

Hypothesis 1. 

Green finance policies enhance the efficiency of urban green energy.

In what ways do green finance policies impact the efficiency of green energy in cities? In this paper, we begin with the configuration of the three main production factors: labour, scale, and technology. Labour, capital, and technology are key variables in the production function, and the way in which these factors are configured and utilised will directly impact the level of energy efficiency [31]. The essence of green finance policies lies in their ability to guide and regulate market behaviours by greening resource allocation. This influences how these factors are invested and their efficiency performance. Firstly, green finance policies promote the optimal allocation of labour and encourage green transformation. Against the backdrop of China’s current “carbon peak and carbon neutral” goal, its green finance sector provides credit and financing facilities for green industries, promoting the development of green projects and enterprises. This process has created conditions for the optimal allocation of labour factors by directing capital towards the green industry. A large number of labourers have been transferred from traditional, high-pollution, high-energy industries to green, clean-energy industries, thus accelerating the upgrading of the industrial and employment structures. Xu and Liu [32] point out that the rapid development of emerging green sectors, such as green buildings, green transportation, and renewable energy, has attracted a large number of highly qualified and skilled labourers. This improves labour efficiency and increases labour productivity per unit of energy output in these industries. In addition, green finance policies promote the development of green industries and indirectly improve the quality of human resources by supporting green vocational training and popularising environmental protection skills. Li et al. [33] mentioned that green finance provides financial support for environmental skills training and helps improve labourers’ green skills, leading to better human resource matching and technical support capacity in the green energy process in cities. Therefore, green finance policies promote the optimal allocation of labour factors, facilitating the development of green industries and the green transformation of the labour market, while improving the efficiency of urban green energy use.

Secondly, in the context of the current global green transition, green finance policies provide strong financial support. They also provide policy guidance for green projects and industry clusters at the city level. This promotes the formation of “green economies of scale”. Gazi et al. [34] state that green finance enables these projects to be realised on a large scale by encouraging and supporting the construction of major green infrastructure projects, such as new energy industrial parks and green transportation hubs. Expanding the scale of energy use not only reduces the cost per unit but also stimulates technological innovation and knowledge spillover through agglomeration, thereby enhancing overall green energy utilisation efficiency. The key to this process is that green finance enables relevant industries to gain a financial advantage. This leads to large-scale production, the promotion of technological synergies and experience-sharing, and enhanced resource allocation efficiency. For example, the initial results from green finance pilot zones in several Chinese cities (including Guangzhou in Guangdong and Huzhou in Zhejiang) demonstrate the positive effects of scaled-up green financing on green energy efficiency. These regions promote the clustering of green projects and enhance the efficiency of overall green energy use by pooling green financial resources and optimising resource allocation. In addition, Mu [35] pointed out that the continuous expansion of the green industry’s scale can attract more private and foreign capital to the sector, optimise the allocation of capital and resources further, and amplify the synergistic advantages of green energy utilisation. Green finance policies have effectively enhanced urban green energy efficiency, promoted the sustainable development of the green economy, and encouraged the clustering and scaling of green industries.

Finally, green finance policies may experience a delay in fulfilling their potential at a technological level. In some cases, they may even hinder green technological innovation. Zeng et al. [36] pointed out that, in practice, green finance policies tend to favour supporting green projects with mature business models and lower risk, while being relatively cautious about cutting-edge green technology innovations at a high-risk, uncertain stage. This financial bias means that green finance is more likely to be invested in the replication and scaling up of existing technologies than in supporting technological innovation and breakthroughs. This shift towards green finance could result in certain uncommercialised green technologies (e.g., hydrogen energy, bioenergy and carbon capture and sequestration) receiving inadequate financial support. Xu [37] also says that when banks and investment companies check green credit and green bond, they often ask for projects to be very profitable and to give a short-term return. This makes it hard for many new green technologies that could be very important to obtain enough money. In the short term, this tendency towards “market stabilization” may hinder original breakthroughs in green technologies, thereby slowing down the rate of improvement in green energy efficiency. Although applying existing green technologies can improve energy efficiency, relying solely on scaling up these technologies may not significantly improve energy efficiency in the long term, particularly given technological advances and breakthrough innovations in green energy. Therefore, although green finance policies can promote the development of green industries in the short term by encouraging technological innovation, more needs to be done to support innovative technologies in order to accelerate the improvement of green energy efficiency. Figure 2 below shows the specific path of action.

In summary, green finance policies significantly enhance urban green energy efficiency by promoting the green transformation and efficient allocation of labour resources, as well as the large-scale development of green industries. However, it has certain technological limitations, particularly with regard to high-risk, long-cycle green technology innovation. It still needs to be coupled with government R&D subsidies, technology insurance, venture capital, and other tools for linked governance. The following research hypotheses are proposed for testing based on this analysis:

Hypothesis 2. 

Green finance policies can improve the efficiency of green energy by influencing three key elements: urban labour, scale, and technology.

Green finance policies play a vital role in enhancing urban green energy efficiency and could have a significant impact on the disparity in green energy efficiency between different regions. The disparities in green energy efficiency between China’s northern and southern regions, coastal and inland regions, and cities in the centre and periphery of provinces are particularly pronounced due to significant differences in economic, financial, technological and resource conditions. Although green finance policies can reduce the green energy efficiency gap between regions to some extent by providing financial support and facilitating technology transfer, they may also exacerbate the disparity between certain regions, creating an imbalance in resource allocation.

Green finance policies could help close the green energy efficiency gap between the Global North and South, as well as between coastal and inland regions, by improving the mobility of finance and technology. Chen et al. [38] highlight that regional economic disparities frequently arise from the uneven distribution of resources, including capital, technology, labour, and infrastructure. Green finance policies can effectively compensate for the shortage of capital and lack of technology in the utilisation of green energy in regions that are relatively lagging behind in terms of technology by directing the flow of capital to green industries. At the same time, when promoting green finance policies, the country tends to prioritise the construction of green energy infrastructure and the support of technological innovation in underdeveloped regions. Green finance policies accelerate the green transformation of landlocked and southern economies by providing low-cost green financing and improving the efficiency of their green energy use.

However, the implementation of green finance policies could exacerbate the green energy efficiency gap between different regions, posing some potential risks. Xiao et al. [39] noted that central cities (e.g., Beijing, Shanghai, and Guangzhou) are better able to attract green financial resources due to their stronger economic foundations and more comprehensive financial systems. These cities typically boast stronger technological innovation capabilities, superior market mechanisms and a greater pool of high-end talent. Consequently, they are better positioned to reap the benefits of green finance policies, which will further enhance the efficiency of their green energy usage. By way of comparison, peripheral cities and areas with weaker economies often struggle to secure financing and are unable to make full use of the resources and support offered by green finance policies, due to their lower level of financial market development and technological innovation capacity. This concentration of resources may further widen the green energy efficiency gap between cities in the centre of the province and those on the periphery. Figure 3 below shows the effects of specific area roles. Among them, an upward arrow means an increase, while a downward arrow means a decrease.

In summary, although green finance policies can generally promote the efficient use of green energy, they do not narrow the gap between different regions. In fact, they may exacerbate the difference in green energy efficiency between regions. The following research hypotheses are proposed for testing based on the appeal analysis:

Hypothesis 3. 

Green finance policies will reduce the green energy efficiency gap between cities in different regions of the country, such as the north and south, and between coastal and inland cities. However, they will increase the green energy efficiency gap between cities in the centre of provinces and those on the periphery.

3. Study Design

3.1. Model Construction

This paper uses DML to study the impact of green finance policies on urban green energy efficiency. This approach can provide a more effective framework for policy analysis than other regression models and double-difference methods. There are three main advantages to this approach.

First is the handling of high-dimensional data and nonlinear relationships: When studying the impact of green finance policies on urban green energy efficiency, numerous control variables are involved, which may exhibit complex nonlinear relationships. Traditional regression models struggle to capture these relationships and are prone to multicollinearity. DML, combined with machine learning algorithms, can automatically select relevant features and effectively capture complex relationships, thereby providing more accurate estimation results.

Second is reducing endogeneity issues: Green finance policy research often faces endogeneity problems, such as unobserved factors influencing policy selection. Traditional regression analysis relies on specific assumptions to address endogeneity but may be affected by assumption validity and model bias. DML, through a phased machine learning approach, can effectively reduce endogeneity issues, accurately estimate policy effects, and make causal inferences more reliable.

Third is improving estimation efficiency and accuracy: The impact of green finance policies involves multiple factors and their interactions. Traditional methods assume identical trends between experimental and control groups before policy intervention—a hypothesis that is difficult to validate and may lead to errors. DML outperforms traditional methods by handling multiple factors and selecting optimal models through adaptive algorithms, enhancing estimation precision. Additionally, DML does not rely on strict assumptions and can flexibly adjust according to data characteristics, avoiding biases caused by incorrect model settings.

Therefore, this paper uses DML to study the impact of green finance policies on urban green energy efficiency. First, the constructed partial linear model is shown below, referencing the approaches of existing scholars [40,41].

$$Ge{e}_{i,t}=\alpha Gf{p}_{i,t}+g(X_{i,t})+U_{i,t}$$

(1)

$$E(U_{i,t}{\left|Gf{p}_{i,t},X\right.}_{i,t})=0$$

(2)

In Equations (1) and (2) above, Gee_i,t represents the green energy efficiency of city i in year t and Gfp_i,t represents the level of green finance policies in city i in year t. The disposal coefficient, α, is the focus of this paper. Its positive or negative sign, size, and significance represent the disposal effect of green finance policies. X_i,t is the set of high-dimensional control variables. The specific functional form, g(X_i,t), may contain mixed variables that affect Gee_i,t and Gfp_i,t. These variables are estimated using a machine learning algorithm. U_i,t is the error term, which has a conditional mean of zero.

Additionally, to avoid the estimated coefficients becoming regularized estimators and introducing bias in the case of finite samples, we should avoid directly solving g(X_i,t) using machine learning algorithms. Therefore, it is important to consider the potential problems posed by regularization methods. Machine learning algorithms control the complexity of models by introducing regularization terms, which reduce the risk of overfitting. However, when the sample size is small, the regularization process may introduce systematic bias into the parameter estimates, particularly when the sample size is too small to accurately represent the data. This bias may affect both the accuracy of the model and the validity of the subsequent analysis by causing the estimates to deviate from the true parameters. Therefore, the following auxiliary regression is constructed:

$$Gf{p}_{i,t}=m(X_{i,t})+V_{i,t}$$

(3)

$$E(V_{i,t}{\left|X\right.}_{i,t})=0$$

(4)

In Equations (3) and (4), the function m(X_i,t) is the regression function of the disposition variable with respect to the high-dimensional control variable. V_i,t represents the error term, which has a conditional mean of zero. In this paper, we use a machine learning algorithm to estimate the specific functional form $\stackrel{\wedge }{m}(X_{i,t})$, and construct the residual estimate, as shown in Equation (5) below.

$${\stackrel{\wedge }{V}}_{i,t}=Gf{p}_{i,t}-\stackrel{\wedge }{m}(X_{i,t})$$

(5)

At the same time, the same machine learning algorithm is used to estimate $\stackrel{\wedge }{g}\left(X_{i,t}\right)$, obtain $Ge{e}_{i,t}-\stackrel{\wedge }{g}\left(X_{i,t}\right)=\alpha Gf{p}_{i,t}+U_{i,t}$, and regress ${\stackrel{\wedge }{V}}_{i,t}$ as an instrumental variable for Gfp_i,t. The resulting coefficient estimates are shown in Equation (6) below:

$$\stackrel{\vee }{\alpha }={({\displaystyle \frac{1}{n}}{\displaystyle {\sum }_{i\in I, t\in T}{\stackrel{\wedge }{V}}_{i,t}Gf{p}_{i,t}})}^{-1}{\displaystyle \frac{1}{n}}{\displaystyle {\sum }_{i\in I, t\in T}{\stackrel{\wedge }{V}}_{i,t}(Ge{e}_{i,t}-\stackrel{\wedge }{g}(X_{i,t}))}$$

(6)

At this stage, the rate at which the coefficient estimates converge depends on how quickly $\stackrel{\wedge }{g}\left(X_{i,t}\right)$ and $\stackrel{\wedge }{m}(X_{i,t})$ converge to g(X_i,t) and m(X_i,t), respectively. This implies that model accuracy depends not only on variable selection, but also on the efficiency with which model parameters are tuned. Using machine learning twice helps exclude the influence of a high-dimensional set of control variables on green finance policies. This improves the model’s stability and accuracy in complex environments and speeds up the convergence of the coefficient estimators. It also obtains more accurate results with limited samples.

Finally, based on the work of Zhang et al. [42], this paper employs a 5-fold cross-validation method to enhance the stability and reliability of the model results when dealing with regression samples. This method divides the dataset into five subsets and selects one subset at a time as the validation set. The remaining four subsets are used as the training set. This process is repeated several times to reduce overfitting and improve the model’s generalization ability. Additionally, to further improve the robustness of the estimation results, this paper calculates the model’s estimation results after each sampling. This is done by repeating the sampling 101 times and selecting the median of all the repeated sampling results as the final estimation results. The median-adjusted approach effectively minimizes the influence of random or extreme values on the results, ensuring more accurate and reliable final estimates. Therefore, adopting this method can effectively address potential heterogeneity in sample data and improve consistency in estimation results across multiple, repetitive evaluations. This provides more solid data support and a stronger theoretical basis for analyzing the impact of green finance policies on urban green energy efficiency.

3.2. Variable Settings

3.2.1. Explained Variables

This paper examines the impact of green finance policies on green energy efficiency (GEE) in cities. In this context, GEE is defined as the efficiency with which cities utilize energy while ensuring economic growth and reducing negative environmental impacts. To quantify this efficiency, this paper refers to study by Hong and Li [2], which considered the impact of undesired outputs on green energy efficiency while using labor, capital, and energy as input variables. In this case, labor is represented by the total number of people employed in the primary, secondary, and tertiary sectors. This variable reflects the overall allocation of labor and the impact of different industries on green energy efficiency. The impact of labor allocation and industry structure on green energy efficiency is significant, especially in cities where the tertiary industry dominates and contributes more to green production. Capital is an important factor in driving a city’s economic growth, particularly in green energy. Investment can effectively promote green technology research and development and green industry expansion. This paper measures urban fixed asset investment, which reflects the level of urban infrastructure construction. This construction directly affects energy use efficiency and the improvement of environmental protection facilities. Energy is one of the core indicators of green energy efficiency. Efficient energy use promotes economic growth and reduces environmental burden. This paper reflects through the city’s energy consumption, the rational control and optimization of energy consumption is a key factor to achieve green energy efficiency improvement, this indicator is directly related to the effectiveness of urban energy use. In addition, this paper selects the gross city product as the desired output, reflecting the city’s overall economic efficiency. Meanwhile, industrial sulfur dioxide, smoke and dust, and wastewater emissions are considered undesirable outputs. These emissions characterize the environmental burden of urban industrialization. Reducing these emissions is an important symbol of improving green energy efficiency.

To measure the green energy efficiency of each prefecture-level city, this paper adopts the SBM Malmquist-Luenberger index method, as referenced in Liu et al. [7]. This approach has the advantage of assessing economic efficiency while taking into account factors such as resource utilization and environmental protection. This provides a more comprehensive and realistic assessment of green energy efficiency. The SBM Malmquist-Luenberger index offers several advantages over traditional efficiency assessment methods, such as DEA and SFA, particularly when it comes to environmental and energy efficiency. The SBM method’s ability to handle asymmetric input and output data is particularly important in green energy efficiency studies because energy consumption and pollution emissions are usually not linearly related, making their reduction process more difficult. Additionally, the SBM Malmquist-Luenberger index measures changes in efficiency and incorporates the time dimension into the analysis. This allows it to assess the long-term impact of policy changes on green energy efficiency. Many scholars have demonstrated the value of the SBM method for measuring green energy efficiency. Therefore, this paper uses the SBM Malmquist-Luenberger index method to measure green energy efficiency, taking advantage of its ability to consider environmental pollution and resource utilization. This method not only comprehensively reflects economic benefits, but also effectively addresses environmental externalities in green energy efficiency. It provides more precise theoretical support for assessing the effects of green finance policies. Therefore, this paper’s study of the impact of green finance policies on urban green energy efficiency through this approach is of great practical significance and academic value.

3.2.2. Explanatory Variables

This paper selects green finance policies (Gfp) as the explanatory variable. There are seven main categories of green finance policies: green credit, green investment, green insurance, green bond, green support, green fund, and green equity. This paper refers to Li [43], which uses the coefficient of variation method (coefficient of variation = (standard deviation/mean value) × 100%) to objectively weigh and sum the above seven policy categories. This method has three advantages over other synthetic methods.

First, the coefficient of variation is an objective statistical tool that is not subject to subjective judgment. Unlike the Analytic Hierarchy Process (AHP) or the weighted average method, which often rely on subjective expert assessments for weight assignment, the coefficient of variation (Cov) method objectively quantifies policy fluctuations by calculating the ratio of the standard deviation to the mean. This approach avoids human error and provides a more reliable, data-driven weight assignment scheme.

Secondly, the coefficient of variation method effectively addresses the differences and uncertainties among various green finance policies. The implementation effects of different policy instruments in green finance policies may be affected by various factors, resulting in large volatility. Calculating the coefficient of variation enables us to determine the relative volatility of each policy type and assign different weights according to their volatility level. This reflects the actual role of each policy more accurately.

Additionally, the coefficient of variation method is ideal for synthesizing multidimensional indicators. Because green finance policies are complex and involve multiple policy instruments, using the coefficient of variation method clearly reveals which policies influence the green financial system the most. This approach offers a systematic method for quantitatively assessing policy effects, avoiding the issue of improper weight assignment that can arise in traditional methods.

Therefore, the coefficient of variation method has become an important tool for this paper due to its objectivity, precision, and effective treatment of volatility. It can scientifically assess the impact of each policy under multiple policy systems and provides a reliable analytical framework for studying the impact of green finance policies on urban green energy efficiency. The weights shown in

Table 1. Weighting results.

Variables	Average Value	Standard Deviation	Coefficient of Variation	Weight (%)
Green credit policies	0.047	0.018	0.391	12.885
Green insurance policies	0.021	0.008	0.397	13.079
Green investment policies	0.012	0.005	0.411	13.532
Green bond policies	0.007	0.003	0.441	14.515
Green support policies	0.007	0.004	0.535	17.614
Green fund policies	0.047	0.018	0.394	12.972
Green equity policies	0.023	0.011	0.468	15.403

below were obtained by analyzing the data of seven green finance policies of 282 Chinese cities at the prefecture level and above from 2006 to 2022.

Table 1 shows that green support policies account for 17.614% of the total weight, indicating that government policy support is the primary driving force for promoting green finance and green economic transformation in the green financial system. Financial instruments such as green credit and green insurance play an important role in green investment. However, the effectiveness of these instruments is often greatly affected by the market environment and policy framework. Green support policies, such as financial subsidies, tax incentives, and guidance from laws and regulations, can effectively reduce the financing costs and risks of green projects, boost market confidence, and thus promote the flow of more funds to green projects. Therefore, the higher policy weighting reflects the central role of governments in establishing an environment that enables market participation and promotes the development of green technologies and industries.

3.2.3. Control Variables

Following the practices of existing scholars [44,45], this paper selects the following control variables:

Gross domestic product per capita (ten thousand dollars, Adgp): This indicator measures the level of urban economic development. Typically, a higher GDP per capita indicates a more economically developed city that can provide more financial support for implementing green finance policies and promoting green energy investment and innovation.

The natural growth rate (‰, Ngr): This is the difference between the birth and death rates, reflecting the rate of population growth. Higher rates of natural increase imply continued population growth, which increases energy demand and may put pressure on green energy efficiency. This requires green finance policies to guide it.

Number of urban private and self-employed persons (ten thousand people, Npi): The total number of private and self-employed workers in cities and towns is a reflection of the flexibility and diversity of the urban economy. An increase in private enterprises and self-employed individuals implies more opportunities for green investment and is likely to encourage the adoption of green technologies.

Number of people working in scientific research, technical services, and geological surveys (ten thousand people, Eua): The total number of people working in scientific research, technical services and geological surveys, reflecting the level of activity in science, technology, and innovation. A larger number of researchers usually implies the potential for green technological innovation and the promotion of green energy efficiency.

Average Employee Salary (millions of yuan, Nse): The sum of all employees’ salaries in the city divided by the number of employees. Salary levels are correlated with a society’s overall productivity level, and higher wages may promote higher spending power, thus supporting the diffusion of green technologies and green finance.

Population density (thousands of people/km², Pd): The total number of people per square kilometer. Higher population densities may lead to more concentrated energy demand, thus requiring more efficient green energy solutions.

Gross Industrial Output Value of Domestic Enterprises (billion yuan, Tiod): The sum of the output value of all locally invested enterprises. This value reflects the dynamism of the local economy, as well as the adoption of environmental protection measures and green technologies, which may affect green energy efficiency improvements.

Gross Industrial Output Value of Foreign-Invested Enterprises (billion yuan, Tiof): The gross industrial output value of foreign-invested enterprises, which usually have more advanced green technology and green energy investments, has a positive impact on improving energy efficiency.

Total profit (billion yuan, Tp): The sum of the net profits of all city firms. A higher total profit indicates that city firms have more money to invest in green technology and research and development to promote green energy efficiency.

Science expenditures (billion yuan, Se): The total amount of government or business investment in science and technology research and development. Science expenditures are the basis for green energy technology development. Higher science expenditures can promote innovation in green technology and green energy efficiency.

Number of University Students per 10,000 People (a hundred people, Ncs): The number of university students per 10,000 people. Improving the level of education will help cultivate more talent with knowledge of green technology, which will indirectly promote the improvement of green energy efficiency.

R&D personnel (thousands of people, Rdp): The number of R&D personnel is the core force of technological innovation. An increase in R&D personnel can improve green energy technology research, development, and application capabilities, as well as promote green energy efficiency.

R&D Internal Expenditure (billions, Rdi): Internal R&D expenditure directly affects the innovation and diffusion of green technologies, as well as enhancing green energy efficiency.

Highway mileage (thousands of kilometers, Me): Indicates the total mileage of urban highways. It affects transportation efficiency and is closely related to the popularity of green transportation modes, such as electric vehicles. The construction of an efficient transportation system affects the overall efficiency of green energy use.

These variables measure cities’ green energy efficiency in terms of economic development, technological innovation, energy demand, and transportation efficiency. These indicators help analyze the mechanisms of action and interrelationships of different factors when studying the impact of green finance policies on energy efficiency. Meanwhile, this paper adds the quadratic terms of each variable to the regression model to improve its goodness of fit.

3.2.4. Mechanism Variables

This paper is based on a theoretical analysis and aims to reveal how green finance policies affect urban green energy efficiency based on the three major factors of production: labor, capital, and technology.

The labor level of a city is mainly measured by the city’s share of employees in the primary sector (%, Ppe) and the share of employees in the tertiary sector (%, Pte) [32]. Ppe reflects the proportion of people employed in the primary sector (e.g., agriculture, mining, etc.) of a region’s economy. Primary industries are usually more energy intensive. Therefore, the proportion of its labor force may be inversely related to energy efficiency. Promoting green finance policies could encourage the primary sector to adopt more efficient, environmentally friendly technologies, reducing its energy consumption and improving overall energy efficiency. On the other hand, the tertiary sector mainly involves the service sector, which usually relies less on energy consumption than the primary sector. Green finance policies support the green transformation of the service sector by promoting green buildings and financial services. Thus, an increase in the labor force’s share of this sector may drive overall green energy efficiency. The growth of the service sector is often accompanied by digitalization and technological innovations that improve energy efficiency.

The level of capitalization of a city is mainly measured by the value added of the primary sector as a share of GDP (%, Ppi) and the value added of the tertiary sector as a share of GDP (%, Pti) [34]. Ppi measures the percentage contribution of the primary sector to a region’s GDP. Because the primary sector is typically more resource-intensive, its higher value-added share of GDP may be associated with lower energy efficiency. The primary sector is typically more resource-intensive, so its higher value added as a share of GDP may be associated with lower energy efficiency. However, implementing green finance policies may facilitate the transformation of the primary sector by encouraging green investments, which would enhance energy efficiency. The percentage contribution of the tertiary sector to GDP is usually closely related to improvements in energy use efficiency. Supported by green finance policies, the share of tertiary industries (e.g., green finance and technology services) has increased. This growth has the potential to promote further energy efficiency through technological innovation and service optimization.

The level of technology in a city is determined by the number of patents granted (pieces, Npa) and the number of invention patents granted (pieces, Nai) [36]. The number of patents is an important indicator of technological innovation, particularly in green technology. The increase in the number of patents granted may indicate that the region is making greater progress in green energy efficiency improvements, as green finance policies drive investment and R&D in environmentally friendly and energy efficient technologies. Invention patents, on the other hand, tend to be more innovative and involve more advanced technological breakthroughs. Green finance policies may direct funds toward research and development, which can facilitate the generation of new patents for green inventions. These technological innovations reduce energy consumption and increase the efficiency of green energy.

3.3. Data Sources and Descriptive Statistics

The research data in this paper was primarily obtained from the Wind database, the China Urban Statistical Yearbook, and provincial statistical yearbooks. These sources provide detailed information on major cities in China’s economy, energy consumption, and green finance policies. Considering the lack of experimental data required in 2023 and 2024, as well as the global impact of the COVID-19 pandemic, which may affect the stability of experimental results, this paper selects panel data from 282 cities between 2006 and 2022 as the research sample. Panel data collected over this time span can effectively capture the dynamics of green energy efficiency in cities as they implement green finance policies.

In the process of data processing, this paper took the existence of missing data in individual cities into account and used the moving average method to supplement the missing values. This method preserves the time series characteristics of the data better and reduces the bias caused by missing data. Additionally, to eliminate the potential impact of extreme outliers on the analytical results, this paper employs 1% and 99% quantile tailing during data processing. This practice aims to exclude extreme values, ensuring a more robust analysis and enhancing the reliability and validity of the study’s conclusions [40]. This paper ensures the completeness and consistency of the data through the above data processing measures. The specific descriptive statistics are shown in

Table 2. Descriptive statistics of the main variables.

Variables	(1)	(2)	(3)	(4)	(5)
	N	Mean	Sd	Min	Max
Gee	4794	0.3222	0.1333	0.0214	1.1770
Gfp	4794	0.0223	0.0083	0.0031	0.0545
Adgp	4794	4.7867	3.4923	0.0099	46.7749
Ngr	4794	5.6509	5.1355	−16.6400	39.1800
Npi	4794	62.3240	99.3106	0.4935	1481.2030
Eua	4794	1.2652	4.2382	0.0100	71.7150
Nse	4794	5.2429	2.7371	0.4958	22.1498
Pd	4794	0.4346	0.3395	0.0050	2.7120
Tiod	4794	2721.4110	3494.8490	0.7630	41,529.8100
Tiof	4794	519.7509	1693.7170	0.0000	48,311.31
Tp	4794	207.6485	348.5847	−186.7393	5161.2910
Se	4794	10.1128	35.3973	0.0034	554.9817
Ncs	4794	1.8152	2.0809	0.0200	12.9400
Rdp	4794	20.6878	45.7220	0.0100	1152.495
Rdi	4794	58.7791	163.7289	0.0003	3358.1130
Me	4794	0.4863	0.3438	0.0290	8.2620
Ppe	4794	0.0259	0.0633	0.0000	0.7397
Pte	4794	0.5452	0.1359	0.0991	0.9942
Ppi	4794	0.1277	0.0816	0.0003	0.4989
Pti	4794	0.4081	0.1018	0.0858	0.8387
Npa	4794	6.3258	16.5071	0.0000	279.1770
Nai	4794	1.2586	4.2686	0.0000	88.1270

below.

4. Empirical Analysis

4.1. Benchmark Regression Estimation Results

This paper uses a DML approach combined with an advanced algorithmic model of relevant data to thoroughly analyze the impact of green finance policies on urban green energy efficiency. Specifically, the samples were split at a ratio of 1:4 to ensure diversity in the dataset and the robustness of the results. The Random Forest algorithm was chosen for model selection in this study due to its ability to handle large-scale, high-dimensional data for prediction and solution. The Random Forest algorithm effectively captures the complex, nonlinear relationship between urban green energy efficiency and green finance policies, making high-precision predictions possible. The specific results are shown in

Table 3. Benchmark regression estimates.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	Gee	Gee	Gee	Gee	Gee	Gee	Gee	Gee	Gee
Gfp	0.9142 ***	1.1363 ***
	(4.01)	(4.86)
Green credit policies			0.2566 ***
			(3.04)
Green insurance policies				0.9464 ***
				(3.49)
Green investment policies					0.4975 ***
					(2.78)
Green bond policies						1.1706 ***
						(2.75)
Green support policies							0.4503
							(1.32)
Green fund policies								0.4056 ***
								(4.67)
Green equity policies									0.5800 ***
									(4.32)
Control variable with one term in the hierarchy	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Quadratic term of the control variable	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Fixed effects of time	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
City fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	4794	4794	4794	4794	4794	4794	4794	4794	4794

Note: *, **, and *** indicate significant at the 10%, 5%, and 1% levels, respectively, with t-values in parentheses.

.

Column (1) of Table 3 controls for the primary term of the city variable of interest, as well as time and city fixed effects. These controls effectively eliminate possible bias, ensuring the reliability and accuracy of the estimation results. The results show that, at the 1% significance level, a 1% increase in green finance policies leads to a 0.9142% increase in urban green energy efficiency. Based on column (1), column (2) controls the quadratic term of the relevant urban variables. The result still shows a significant promotional effect at the 1% level of significance, which further proves the positive promotional effect of green finance policies on urban green energy efficiency. This finding preliminarily affirms research hypothesis 1. Columns 3 through 9 examine the impact of seven green finance policies—green credit, green insurance, green investment, green bond, green support, green fund, and green equity—on urban green energy efficiency. Through individual testing of each policy, the impact of different policy tools on promoting urban green energy efficiency is discussed. The role of these policy tools in promoting urban green energy efficiency is explored through the individual testing of each green finance policy.

Except for the green support policies, all of the other six green finance policies demonstrate significant promotional effects at the 1% level of significance. Each of these policies has its own characteristics. Green credit policies, for example, reduce financing costs by providing low-interest loans for green projects. It also encourages enterprises and the government to invest in green energy facilities and improve energy efficiency. Green insurance policies provide risk protection for green projects, alleviating investors’ concerns and encouraging the implementation of more projects. Green investment policies attract capital to technologies that protect the environment and save energy. It promotes the growth of green industries and improves energy efficiency. Green bond policies enable companies and governments to raise funds by issuing green bond, which are exclusively used for green projects. This provides stable financial support for green energy projects, accelerating their implementation. Green fund policies support green technology research and development, as well as project investments, by establishing special funds to promote the long-term sustainable development of green projects. Green equity policies encourage green enterprises to expand their markets through equity financing, thereby promoting the commercialisation and application of green technologies.

Secondly, the facilitating effect of green bond policies is the most prominent of all six green finance policies. This is primarily due to its ability to mobilize large-scale funds through the capital market and provide stable, long-term financial support for green energy projects [30]. Green bond financing is ideal for capital-intensive green infrastructure projects because it has lower costs and longer repayment terms than green credit. Its market-oriented character attracts a wide range of investors, which enhances the liquidity and transparency of funds. This accelerates the implementation of green projects. The environmental credentials and clear usage guidelines of green bond have given investors more confidence in them, further facilitating the inflow of funds into the green energy sector. Other policies, such as green credit and green fund, are limited by funding and risk management constraints, which makes it difficult to provide support on a large scale. As a result, green bond policies have contributed significantly to improving green energy efficiency in cities by providing large-scale, low-cost financial support and capital market incentives.

Ultimately, the green support policies did not reach the required level of statistical significance, despite presenting a facilitating effect. This is mainly due to the heterogeneity of policy implementation and the lag in market response [46]. Because the economic structures, resource endowments, and policy implementation efforts of different cities vary considerably, the actual effects of policies differ from region to region. Additionally, the impact of green support policies typically takes time to materialize due to factors such as market adjustments and capital flows. At the same time, green support policies may not be targeted precisely enough to reach the areas that need support the most, which affects their overall effectiveness. In the short term, the external economic environment, such as energy price volatility, may have a greater impact on green energy efficiency, thereby masking the effects of green support policies. Therefore, the significant impact of green support policies on green energy efficiency in cities is limited by these factors.

4.2. Robustness Tests

This paper presents robustness tests conducted by replacing explanatory variables, adjusting the study sample, removing outliers, considering province–time interaction fixed effects, resetting the model, and considering lags.

First, we need to be able to handle cases where the green energy efficiency value exceeds 1. This allows for a more detailed ranking of the assessed units and distinguishes between cities with better performance. This is important for identifying cities with more effective green finance policies implementation. In this paper, we refer to Cheng et al. [47] and use the super-efficiency CCR model for robustness testing. Based on the original CCR model, the Super-efficiency CCR model provides a more stable and reliable efficiency assessment when data is more intensive or differences in efficacy are large. Thus, it enhances the robustness and credibility of the analysis results. The specific estimation results are shown in Column 1 of

Table 4. Robustness tests 1.

Variables	(1) Replacement of Gee	(2) Adjustment of the Study Sample		(3) Removing Exceptions		(4) Considering Province–Time Interaction Fixed Effects
		Adjustment of Cities	Adjustment Year	1% Indentation	5% Downsizing
X	1.2183 ***	0.7315 ***	0.6397 **	0.8454 ***	0.6929 ***	1.4658 ***
	(3.93)	(3.18)	(1.99)	(4.34)	(4.72)	(5.34)
Control variable with one term in the hierarchy	Yes	Yes	Yes	Yes	Yes	Yes
Quadratic term of the control variable	Yes	Yes	Yes	Yes	Yes	Yes
Fixed effects of time	Yes	Yes	Yes	Yes	Yes	Yes
City fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Province–Time Fixed Effects	No	No	No	No	No	Yes
sample size	4794	4539	2538	4698	4314	4794

Note: *, **, and *** indicate significant at the 10%, 5%, and 1% levels, respectively, with t-values in parentheses.

. The regression coefficients of green finance policies on urban green energy efficiency increased and remained positive at the 1% significance level.

Second, this paper adjusts the research sample for robustness testing. Large cities such as Beijing, Shanghai, and Tianjin are deleted, along with cities in the Inner Mongolia and Xinjiang Uygur autonomous regions. These cities have large economies, and the impacts of green finance policies may lead to an over-concentration effect, resulting in extreme or high energy efficiency performance. In regions such as Inner Mongolia and Xinjiang, which rely more heavily on traditional energy sources, the effects of green finance policies may be delayed or incomplete. Conversely, deleting data from 2006 to 2010 and from 2020 to 2022 helps eliminate the effects of abrupt changes in policy and data quality. From 2006 to 2010, green finance policies were still in their infancy. The period from 2020 to 2022 may be affected by special factors, such as epidemics and data fluctuations. Data from these years may not accurately reflect the long-term impact of green finance policies. Deleting these periods improves the reliability of the robustness test and ensures that the analytical results are more stable and scientific. This helps avoid biased conclusions due to abnormal data or short-term fluctuations. The specific estimation results are shown in column (2) of Table 4. Even when accounting for differences in city size or years, the green finance policies continue to be a significant factor in promoting green energy efficiency.

Third, outliers may be caused by data entry or measurement errors, or by the unique economic conditions of a particular city. If these issues are not addressed, they may lead to a biased regression model estimation and affect accurate policy judgment. This paper discusses the implementation of 1% and 5% tailing, which involves removing the top and bottom 1% or 5% of extreme values in the data distribution. This effectively reduces the impact of outliers on the results, allowing for a more focused analysis of the core data and enhancing the robustness of the results. This approach helps avoid excessive volatility caused by a few extreme observations. Thus, it ensures that the study’s findings are more representative of the general effects of green finance policies in most cities. The specific estimation results are shown in column (3) of Table 4. The baseline regression results still hold when performing 1% or 5% up- and down-deflations. Green finance policies still promote urban green energy efficiency at the 1% significant level.

Fourth, this paper resets the DML model to perform robustness tests. Adjusting the sample partition ratio effectively optimizes the balance between the training and validation sets. This prevents overfitting or underfitting due to insufficient training data and ensures more generalizable results. Modifying the specific segmentation ratios to 1:2 and 1:7 increases the amount of data in the training set, enhancing the model’s ability to learn complex relationships and recognize the relationship between green finance policies and green energy efficiency. Gradient boosting and neural network algorithms are powerful nonlinear modeling tools that can capture complex nonlinear features difficult for traditional linear models to capture. These algorithms improve the adaptability of models when dealing with urban green energy efficiency and policy impacts. The gradient boosting algorithm effectively reduces prediction errors through iterative optimization. The neural network algorithm excels at fitting large-scale data, especially for analyzing complex economic policies. It provides more accurate predictions. The specific estimation results are shown in column (1) of

Table 5. Robustness tests 2.

Variables	(1) Resetting DML Models				(2) Hysteresis
	Sample Split Ratio 1:2	Sample Split Ratio 1:7	Gradient Boosting Algorithm	Neural Networks Algorithms	Lag 1	Lag 2
Gfp	1.1205 ***	1.0469 ***	1.9043 ***	0.0560 ***	0.8765 ***	0.8265 ***
	(4.82)	(4.70)	(8.21)	(2.80)	(3.72)	(3.55)
control variable with one term in the hierarchy	Yes	Yes	Yes	Yes	Yes	Yes
quadratic term of the control variable	Yes	Yes	Yes	Yes	Yes	Yes
fixed effects of time	Yes	Yes	Yes	Yes	Yes	Yes
city fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
sample size	4794	4794	4794	4794	4512	4230

Note: *, **, and *** indicate significant at the 10%, 5%, and 1% levels, respectively, with t-values in parentheses.

. Green finance policies promote urban green energy efficiency at the 1% significant level when the sample split ratio is 1:2 or 1:7 and either the gradient boosting or neural network algorithm is used.

Fifth, endogeneity problems usually stem from bidirectional causality. For example, green finance policies may affect energy efficiency, and changes in energy efficiency may affect policy formulation or adjustment. If such interactions are not accounted for, the model’s estimates may be biased, which could affect judgments about policy effects. Introducing lags captures the dynamic nature of changes in green energy efficiency. The effects of green finance policies are not immediately apparent; rather, they take time to manifest in increased energy efficiency. The study in this paper can identify the delayed effects of policies and avoid misclassification due to short-term fluctuations in independent variables by introducing lags. This approach improves the model’s robustness and applicability in different contexts. It also reveals the long-term cumulative effects of policies. The specific estimation results are shown in column 2 of Table 5. Both lag 1 and lag 2 of green finance policies promote urban green energy efficiency at the 1% significance level. The results of the robustness test further affirmed research hypothesis 1.

4.3. Heterogeneity Test

4.3.1. Resource Endowment

Based on the characteristics of the economic structure of cities, especially the degree of resource dependence, and in accordance with the Circular on the National Sustainable Development Plan for Resource-Based Cities (2013–2020) [48], this paper categorizes 282 Chinese cities as either resource-based or non-resource-based. Prefecture-level cities and those involving some county jurisdictions are classified as resource-based. These cities usually rely on traditional resource extraction and heavy industries and often face sustainable development problems, such as resource depletion and environmental pollution. Special attention must be paid to the transformation and sustainable development of these cities when promoting green energy efficiency and implementing green finance policies. In contrast, the economic structure of non-resource cities is more diversified and may focus more on services or high-tech industries. Thus, their green energy efficiency improvements differ from those of resource cities. This categorization helps develop more precise, targeted green finance policies for different types of cities. The goal is to maximize the effects of these policies and promote green transformation and energy efficiency in all types of cities. Specific results are shown in columns (1) and (2) of

Table 6. Results of heterogeneity test.

Variables	(1) Resource-Based	(2) Non-Resource-Based	(3) Old Industrial Base	(4) Non-Old Industrial Base	(5) Developed City	(6) More Developed Cities	(7) Less Developed City
Gfp	1.0781 **	0.5720 **	1.0441 ***	1.0340 ***	0.9183 **	1.5379 ***	0.1144
	(2.51)	(2.17)	(3.18)	(3.29)	(2.23)	(3.18)	(0.35)
Control variable with one term in the hierarchy	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Quadratic term of the control variable	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Fixed effects of time	Yes	Yes	Yes	Yes	Yes	Yes	Yes
City fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Sample size	1309	3485	1597	3197	1666	1664	1464

Note: *, **, and *** indicate significant at the 10%, 5%, and 1% levels, respectively, with t-values in parentheses.

. Green finance policies promote increased green energy efficiency at the 5% significance level in both 1309 resource-based and 3485 non-resource-based cities, but to a greater extent in the former than the latter. One possible reason for this difference is that cities based on resources face more urgent challenges and needs in the process of green transformation. Resource-based cities typically depend on traditional, high-polluting, energy-intensive industries. Their economic structures are relatively rigid and lack sufficient green investments and technological innovations. The guiding role of green finance policies is therefore particularly important. Non-resource-based cities have more diverse economic structures and are already more efficient in terms of green energy. However, the further role of green finance policies may be relatively insignificant.

4.3.2. Industrial Development Base

In accordance with the National Old Industrial Base Adjustment and Transformation Plan (2013–2020) [49], this paper classifies 282 Chinese cities as either old or non-old industrial base cities. Prefecture-level cities and those above involving some county jurisdictions are categorized as old industrial base cities. These cities have historically relied on traditional, heavy, and energy-intensive industries, such as iron and steel, coal, and power. These industries consume large amounts of energy and exert greater pressure on the environment. Implementing green finance policies is crucial for promoting the sustainable development of these cities. Cities with old industrial bases face the dual challenges of adjusting their industrial structures and undergoing green transformations. Green finance policies can help these cities accelerate technological innovation and optimize their energy structures. On the other hand, cities that are not based on old industries tend to have a more diversified economic structure and a relatively easier green transition, although they also need to promote green energy efficiency. Therefore, dividing cities into those with an old industrial base and those without can help formulate more accurate green finance policies for different regions and maximize the benefits of these policies. The specific results are shown in columns 3 and 4 of Table 6. Among the 1597 old industrial base cities and the 3197 non-old industrial base cities, green finance policies promote increased green energy efficiency at the 1% significant level. However, the promotion is slightly stronger in old industrial base cities. This discrepancy could be caused by a number of factors. First, the economic structure of cities with an old industrial base is traditionally dominated by industries with high energy consumption and emissions. These cities are particularly in need of green finance policies [26]. Green finance policies can provide financial support and incentives for technological innovation in cities with old industrial bases. These policies promote the transformation and upgrading of industries, thus accelerating the improvement of energy efficiency. Secondly, old industrial cities face more urgent pressure and demand during the green transformation process, making them more receptive to green finance policies. The effects of these policies are more apparent in these cities. In contrast, the economic structure of cities without an old industrial base is more diversified. The green energy efficiency of some of these cities is already relatively high, and the effect of promoting green finance policies is relatively weak [32]. Finally, cities without an old industrial base may face more external constraints in the process of policy implementation. These constraints include the strength of the local government’s policy support and the market’s acceptance of green financial products. This leads to a slightly slower rate of green energy efficiency improvement than in cities with an old industrial base [36].

4.3.3. Urban Hierarchy

Cities of different tiers differ significantly in terms of their economic structure, market size, industrial development, capacity for technological innovation, and capital investment. According to “2020 Chinese Cities Business Charm Ranking” [40], this paper categorizes first- and second-tier cities as developed, third-tier cities as more developed, and fourth- and fifth-tier cities as less developed. Developed and more developed cities usually have stronger economic strength and an advanced technological foundation, enabling them to better undertake green finance policies. These policies promote the improvement of green energy efficiency. Less developed cities have lower levels of economic development and lagging infrastructure. They also have limited acceptance and implementation of green finance policies, as well as weaker demand and capacity for green transformation. Therefore, categorizing cities according to their level of economic development helps analyze the mechanisms and effects of green finance policies in cities at different stages of development. This analysis also reveals differences in green energy efficiency improvement between different types of cities. The specific results are presented in columns 5, 6 and 7 of Table 6. Green finance policies promote a significant increase in green energy efficiency in both developed and more developed cities. However, the effect is stronger in more developed cities than in developed cities. Green finance policies do not significantly promote energy efficiency in less developed cities. There may be a number of reasons for this discrepancy. Firstly, developed cities and those that are more developed have better economic foundations, a more complete industrial structure, and stronger capabilities for technological innovation. Green finance policies can promote the use of green technologies and improve energy efficiency by providing enterprises with green financing, capital subsidies, tax incentives, and other forms of support [27]. Secondly, green finance policies may play a more prominent role in less developed cities than in more developed ones, despite the latter’s superior economic level, due to the former’s more urgent need for industrial transformation [30]. Finally, the infrastructure and market environment in less developed cities is relatively weaker, which makes it more difficult to localise and implement green finance policies [37]. These cities have relatively little market demand for green financial products. This means that there is insufficient incentive for companies to transition to green practices, and local governments may lack the financial and policy support needed to effectively promote green development. Additionally, the industrial structure of less developed cities may be more dependent on traditional, high-polluting, high-energy industries, making the potential and necessity for green transformation lower and the promotional effect of green finance policies less obvious. The results of the heterogeneity test still support research hypothesis 1.

4.4. Mechanism Testing

The above results suggest that green finance policies have a significant impact on improving green energy efficiency in the city. To explore in depth how green finance policies affect green energy efficiency through different channels, this section will analyse their mechanisms of action in more detail. According to the conclusions of the theoretical analysis, the impact of green finance policies can be explained by three main factors. First, policies can improve the urban labour market, enhancing labour productivity and green skills. Second, policies can expand the size of the city, optimising resource allocation and realising the economic scale effect. Third, policies can promote urban technological innovation, facilitating the application of green technologies and enhancing energy efficiency. These three mechanisms are intertwined and work together to promote green energy efficiency in cities. In this paper, we continue to use DML for causal mediation analysis, splitting the sample at a ratio of 1:4 in order to test the appeal of the three transmission mechanisms based on the random forest algorithm.

The effectiveness of policy implementation may be significantly affected by differences in urban labour levels. To better measure this difference, this paper uses the proportion of employees in the primary and tertiary sectors as an indicator of the labour level in each city. The primary industry is usually closely related to resource extraction and agricultural production. It has lower labour productivity and a greater environmental impact. In contrast, the tertiary industry involves services and high technology. It has higher labour productivity and green innovations, and environmentally friendly technologies are more prevalent. Therefore, these two factors can reflect not only the city’s economic structure, but also its labour market’s ability to transition to a green economy. The specific estimation results are shown in

Table 7. Results of the urban labour mechanism test.

Variables	(1) Direct Effect	(2) Indirect Effect	(3) Total Effect	(4) Indirect Effect	(5) Total Effect
Gfp	1.1363 ***	0.0750	0.9449 ***	0.1599 **	0.8584 ***
	(4.86)	(1.18)	(4.06)	(0.98)	(3.57)
Control variable with one term in the hierarchy	Yes	Yes	Yes	Yes	Yes
Quadratic term of the control variable	Yes	Yes	Yes	Yes	Yes
Fixed effects of time	Yes	Yes	Yes	Yes	Yes
City fixed effects	Yes	Yes	Yes	Yes	Yes
Sample size	4794	4794	4794	4794	4794
Sobel Z value			0.553 *		0.535 *

Note: *, **, *** indicate significant at the 10 per cent, 5 per cent and 1 per cent levels, respectively, with t-values in parentheses.

.

Table 7 shows that green finance policies have a significant direct effect on green energy efficiency in cities. This phenomenon has emerged due to the financial and technological support provided by green finance policies. These policies have accelerated the adoption of green technologies, particularly in labour-intensive industries, and have improved energy efficiency. However, the results of the indirect effect show that, although green finance policies can increase the proportion of employees in both the primary and tertiary industries, this increase is not significant for the primary industry. This finding suggests that green finance policies have a slower impact on the transformation of resource-based industries because technological innovation and green transformation in these industries require more time and financial support. Promoting the tertiary industry is even more significant, as it suggests that financial support for the green transformation of service and high-tech industries is more likely and is quickly reflected in improved green energy efficiency [32]. The results of the total effect analysis show that green finance policies affect urban green energy efficiency by increasing the proportion of people employed in the primary and tertiary sectors. This, in turn, affects urban green energy efficiency. Specifically, green finance policies are better able to promote the green development of the tertiary sector. This optimises the industrial structure and further improves green energy efficiency in cities.

The effectiveness of policy implementation may be significantly affected by differences in city size levels. To measure this difference more accurately, this paper uses the proportion of added value from the primary and tertiary industries to GDP as representative indicators of city size. Cities with a high proportion of GDP accounted for by the primary industry tend to rely on traditional, resource-based industries with low labour productivity and limited potential for green transformation. In contrast, cities with a high proportion of GDP accounted for by the tertiary industry tend to have a strong service and technology sector, with high labour productivity and an industrial structure that is better suited to green transformation. Therefore, selecting these two weights as indicators of city size can reflect the city’s potential for green economic transformation and the applicability of green finance policies. The specific estimation results are shown in

Table 8. Results of tests of the city-size mechanism.

Variables	(1) Direct Effect	(2) Indirect Effect	(3) Total Effect	(4) Indirect Effect	(5) Total Effect
Gfp	1.1363 ***	0.5131 ***	0.6693 ***	0.6324 ***	0.9217 ***
	(4.86)	(6.40)	(2.81)	(5.08)	(3.99)
Control variable with one term in the hierarchy	Yes	Yes	Yes	Yes	Yes
Quadratic term of the control variable	Yes	Yes	Yes	Yes	Yes
Fixed effects of time	Yes	Yes	Yes	Yes	Yes
City fixed effects	Yes	Yes	Yes	Yes	Yes
Sample size	4794	4794	4794	4794	4794
Sobel Z value		0.564 *			0.550 *

Note: *, **, *** indicate significant at the 10 per cent, 5 per cent and 1 per cent levels, respectively, with t-values in parentheses.

.

As shown in Table 8 above, the results of the indirect effects indicate that green finance policies promote growth in the value added by the primary and tertiary industries as a percentage of GDP. The results also show that green finance policies have a greater promotional effect on the value added by the tertiary industry as a percentage of GDP (0.6324 ***). This phenomenon may be due to the fact that green finance policies provide direct financial support to high-tech, service, and innovative industries, which have a greater advantage in terms of technological innovation and green transformation [34]. Therefore, the increase in the value added by the tertiary sector as a proportion of GDP is a better reflection of the effect of green finance policies in promoting green technological advances and improving energy efficiency. The results of the total effect analysis show that green finance policies ultimately improve the urban green energy efficiency by increasing the value added by the primary or tertiary sectors in GDP. Notably, the value added by the tertiary sector has increased as a proportion of GDP, demonstrating the significant contribution of green finance policies to the green transformation of the service and high-tech industries, thereby promoting the urban green energy efficiency.

Differences in the level of urban technology can have a significant impact on the effectiveness of policy implementation. This paper uses the number of patents granted and the number of patents granted for inventions as indicators of cities’ technological levels. The number of patents and invention patents granted usually reflects a city’s capacity for technological innovation and level of R&D. An increase in the number of patents indicates a city’s activity in technological innovation and intellectual property protection. The number of patents for inventions, however, is a more direct reflection of the originality and high value of technological innovation. Therefore, the number of patents and invention patents granted can effectively reflect the impact of green finance policies on the city’s technological progress and improvement in green energy efficiency.

As shown in

Table 9. Results of the urban technology mechanism test.

Variables	(1) Direct Effect	(2) Indirect Effect	(3) Total Effect	(4) Indirect Effect	(5) Total Effect
Gfp	1.1363 ***	−0.3199	1.0043 ***	−1.2612	0.9767 ***
	(4.86)	(−0.02)	(4.35)	(−0.34)	(4.16)
Control variable with one term in the hierarchy	Yes	Yes	Yes	Yes	Yes
Quadratic term of the control variable	Yes	Yes	Yes	Yes	Yes
Fixed effects of time	Yes	Yes	Yes	Yes	Yes
City fixed effects	Yes	Yes	Yes	Yes	Yes
Sample size	4794	4794	4794	4794	4794
Sobel Z value			0.640 *		0.638 *

Note: *, **, *** indicate significant at the 10 per cent, 5 per cent and 1 per cent levels, respectively, with t-values in parentheses.

above, the results of the indirect effects indicate that, although green finance policies have a dampening effect on the increase in patents granted and patents granted for inventions, this effect is not significant. Although green finance policies may have facilitated the green transformation of traditional industries by reducing financing costs and providing green subsidies, the process may have temporarily hindered the technological progress of innovative firms, particularly with regard to the frequency of patent applications and capital investment. This may not fully reveal the long-term effects of technological innovation. The results of the total effects analysis show that green finance policies indirectly contribute to the urban green energy efficiency by inhibiting the growth in the number of patents granted for inventions. Green finance policies encourage the adoption of existing technologies by reducing the financing costs of green projects and offering green credit. This approach relies less on breakthroughs in innovative technologies. Financial resources tend to support mature technologies or those with application potential, which may slow the growth rate of technological innovation, particularly with regard to new inventions [36]. In summary, the above affirms research hypothesis 2.

5. Further Discussion

The above study revealed that green finance policies would promote green energy efficiency by influencing three key internal factors in cities: labour, scale, and technology. However, disparities between cities may affect the effectiveness of the implementation of green finance policies. For example, southern regions with developed economies and active technological innovation are better able to utilise policies that promote green energy efficiency. Coastal regions with open economies and strong green finance support are also better placed to do so. In contrast, inland regions face greater financial and technological constraints. Provincial capitals and central cities with abundant resources and strong policy support are more likely to promote green transformation. Peripheral cities, however, are relatively slow to promote green energy efficiency due to infrastructure and financial constraints. Therefore, this paper adopts Zhang and Li’s approach [40], in which the green energy efficiency indicator of the city is used as the denominator, while the largest city indicator in the corresponding region is used as the numerator. A smaller value indicates a smaller gap in green energy efficiency between cities within a given region. Relative disparities were mainly measured in the following four regional categories. The specific results are shown in

Table 10. Regression estimates of green finance policies on regional development gap.

Variables	(1) Relative Urban Disparities	(2) Relative Gap Between North and South	(3) Relative Disparities Between the Coast and the Interior	(4) Relative Gap Between the Periphery of the Centre
Gfp	−13.1697 ***	−13.4550 ***	−13.0944 ***	11.7911 ***
	(−4.30)	(−4.43)	(−4.27)	(5.37)
Control variable with one term in the hierarchy	Yes	Yes	Yes	Yes
Quadratic term of the control variable	Yes	Yes	Yes	Yes
Fixed effects of time	Yes	Yes	Yes	Yes
City fixed effects	Yes	Yes	Yes	Yes
sample size	4794	4794	4794	4794

Note: *, **, *** indicate significant at the 10 per cent, 5 per cent and 1 per cent levels, respectively, with t-values in parentheses.

below.

Columns (1), (2), and (3) of Table 10 show the results of the regression estimation of the effects of green finance policies on the relative green energy efficiency gaps between cities at the national level and in the north–south and coastal inland regions. All of these results are significant at the 1% level, which implies that the implementation of green finance policies effectively reduces the green energy efficiency gap between regions in China.

Firstly, at the national level, green finance policies have encouraged cities of all types, especially those in less economically developed areas, to increase their investment in green energy and improve their energy efficiency by promoting the flow of funds to green projects. These policies have lowered the barrier to entry for funds, enabling a more balanced distribution to cities at different stages of economic development. This enhances green energy efficiency and reduces regional disparities in the adoption of green technologies due to a lack of funding. In cities with a weak economic base, implementing green finance policies has helped them overcome the financial and technological challenges of the green transition, reducing the green energy efficiency gap with more economically developed cities [17].

Secondly, the green energy efficiency gap between northern and southern regions is gradually narrowing, with green finance policies playing a pivotal role in this process. The south usually has more advanced green technologies and higher energy efficiency. In contrast, the north is held back by traditional industrial structures and technological lag. Green finance policies in this context have helped increase the introduction and application of green energy technologies in the northern region, especially in the transformation process of traditional industries such as coal, iron, and steel, which has played a positive role [18]. This has led to a gradual narrowing of the green energy efficiency gap between the northern and southern regions, with the north seeing effective improvements. This has further contributed to the coordinated development of the interregional economy.

Finally, green finance policies have also narrowed the gap between coastal and inland regions. While coastal regions have relatively high green energy efficiency, mainly due to their early green transition and abundant financial support, inland regions have got off to a late start and face challenges such as insufficient funding and slow technology adoption. Green finance policies promote the flow of green capital to inland areas, providing more green financial and technical support to inland cities, especially through policy instruments such as government green credit and green bond [19]. Improvements in green energy efficiency in these regions are gradually reducing the gap between coastal and inland areas, thereby promoting balanced inter-regional development.

Overall, green finance policies have helped cities in different regions achieve synergistic improvements in green energy efficiency, promoting the rational allocation of funds and accelerating the adoption of green technologies. This policy promotes a balanced green transition among cities and provides China with a solid foundation on which to compete in the global green economy. The implementation of green finance policies has, to a certain extent, alleviated regional imbalances and contributed to the smooth growth of regional economies and green energy efficiency.

However, the results in column 4 of Table 10 suggest that green finance policies exacerbate disparities in green energy efficiency between cities on the outskirts of provincial centres and those in the centre. This implies that the benefits of these policies are not distributed evenly across all cities in a given province. The purpose of green finance policies is to promote overall urban green energy efficiency by providing financial support, guidance on green technology, and measures to improve the environment. However, the provincial capitals are better able to make effective use of green financial resources thanks to their more developed economic bases, well-developed infrastructure and higher levels of marketisation. They have a strong absorptive capacity and are able to leverage policy funding to accelerate the promotion and application of green technologies, further enhancing green energy efficiency. By contrast, peripheral cities in the provinces tend to face greater structural challenges, such as limited access to finance, an industrial structure dependent on traditional energy sources and a weak green technology base. Although these cities have received some financial support through the green finance policies, their relatively low level of economic development and inefficient allocation of funds and resources have caused them to fall behind in improving green energy efficiency. This has prevented them from reaping the benefits of the policy in the same way as other cities [20]. Consequently, green finance policies further amplify the gap in green energy efficiency between central and peripheral cities. In summary, research hypothesis 3 is confirmed.

6. Conclusions, Policy Recommendations, and Research Gaps

6.1. Conclusions

In the context of the dual goals of sustainable development and a low-carbon economy, improving urban green energy efficiency has become a key issue for China. In recent years, green finance policies have attracted considerable interest from scholars, particularly with regard to improving the energy efficiency of green initiatives in urban areas. It is an important tool for promoting green transformation. Not only have green finance policies been shown to play a significant role in promoting environmental protection and sustainable development goals [25], but they have also been found to have a profound impact on energy efficiency through capital flows and market mechanisms [26]. However, although green finance policies have achieved significant results in many areas, existing research also highlights some of the challenges and limitations they face. The effects of green finance policies tend to vary depending on factors such as region, city size and level of economic development. For example, green finance policies tend to be implemented more slowly in resource-based cities and old industrial areas due to their historical dependence on energy structures. On the other hand, when local governments implement green finance policies, problems such as a lack of implementation and uneven distribution of funds can arise, limiting the effectiveness of the policies.

Furthermore, research on impact mechanisms demonstrates that green finance policies not only enhance labor input and channel funds toward green projects [31,34] but also boost energy efficiency by reducing financing costs for green technologies and promoting their widespread adoption [36]. However, in practical policy design and implementation, while these measures show potential to improve urban green energy efficiency, a critical challenge remains: addressing regional development disparities. Moreover, while green finance policies may narrow the green energy efficiency gap between regions, they could simultaneously widen the development divide between central cities and peripheral areas.

Using panel data from 282 Chinese cities at prefecture level and above between 2006 and 2022, this paper examines the specific effects and mechanisms of green finance policies on urban green energy efficiency. The benchmark regression results show that green finance policies significantly promote green energy efficiency in Chinese cities, passing the rigorous robustness test. Green bond policies are found to have the greatest promotional effect, whereas green support policies are found to have no significant promotional effect. The heterogeneity results show that green finance policies are more effective in promoting green energy efficiency in resource cities, old industrial bases and more developed cities. The results of the impact mechanism show that green finance policies promote green energy efficiency by allocating the three internal factors of urban labour, capital, and technology. The results of the relative regional gap show that implementing green finance policies effectively reduces the nationwide, northern and southern, and coastal and inland gaps in green energy efficiency, promoting balanced regional development. However, it will widen the gap between provincial capitals and peripheral cities.

6.2. Policy Recommendations

Given the significant impact of green finance policies on urban green energy efficiency, this paper makes three recommendations.

Firstly, the green finance policy system and incentive mechanism should be improved. Although green finance policies significantly enhance green energy efficiency, green support policies are ineffective. To this end, the system’s design must be optimised further to ensure the policy’s precision and efficiency. Firstly, nationally harmonised green finance standards should be developed. Firstly, the People’s Bank of China and the National Development and Reform Commission should take the lead in improving the criteria for classifying green finance projects, clarifying the scope of support (e.g., renewable energy and energy-saving technologies) and establishing a green project certification mechanism to ensure that funds flow to genuine green projects. Secondly, they should implement differentiated incentive policies. For resource cities, old industrial bases and developed cities, green finance tax incentives or low-interest loan policies should be implemented to encourage enterprises to invest in energy-saving and emission reduction projects. At the same time, credit guarantees should be provided for small and medium-sized green enterprises to lower their financing thresholds. Thirdly, supervision and performance evaluation should be strengthened. The Ministry of Finance and the Ministry of Ecology and Environment should establish a green finance monitoring platform to regularly assess the effectiveness of policy implementation, with a focus on monitoring the efficiency of fund utilisation and preventing the diversion or inefficient allocation of funds.

Secondly, we need to allocate resources precisely in order to transform key cities. Research has shown that green finance policies are more effective in cities with a strong focus on resources, such as old industrial bases and developed cities. Urban construction should prioritise key areas and optimise resource allocation. Firstly, promote energy structure transformation. In resource-based cities (e.g., Taiyuan in Shanxi) and old industrial bases (e.g., Changchun in Jilin), priority should be given to using green finance funds for coal substitution projects, such as constructing wind and solar power bases, or developing energy-saving industrial park transformations. Secondly, encourage the application of green technologies. In developed cities (e.g., Beijing and Hangzhou), green finance should support smart grids, green buildings, and low-carbon transport projects. This will encourage the participation of social capital and form a replicable, scalable green city construction model. Thirdly, enhance the implementation capacity of small and medium-sized cities. In response to the lack of significant effect of green support policies, these cities should receive green finance project management training and technical support to help them formulate and implement energy-saving and emission reduction project plans. This will improve the efficiency and effectiveness of the policies.

Finally, it promotes regional balance and reduces disparities between provinces. Although green finance policies have been effective in reducing national and regional green energy efficiency gaps, they have widened the divide between central and peripheral cities within provinces. The following measures are needed for urban applications. First, a regional green finance collaboration mechanism should be established. Secondly, relying on regional coordinated development strategies (e.g., the Guangdong–Hong Kong–Macao Greater Bay Area and the Chengdu–Chongqing Economic Circle), promote the sharing of green financial resources between central and peripheral cities and the joint development of cross-regional green projects (such as a regional clean energy network). Thirdly, provide special support for peripheral cities. A special green finance fund should support peripheral cities (e.g., Fuyang in Anhui Province and Yibin in Sichuan Province) in developing small-scale green projects (e.g., rural biogas or distributed photovoltaic systems) to ensure the policy benefits remote areas and promotes their green transformation. Thirdly, dynamic monitoring and policy optimisation should be conducted. Use the national energy big data platform to track green energy efficiency data in real time for each city, dynamically adjust the allocation of green financial funds, increase support for less efficient areas and promote balanced regional green development. Ensure that the policy is implemented and continuously optimised.

6.3. Limitations

Firstly, although research shows that green finance policies have a better promoting effect in resource-based cities, old industrial bases and more developed cities, it has not delved deeply into how green finance policies specifically affect enterprises of different industries and scales. For instance, the green bond policy may be more effective for the green transformation of large enterprises, but for small and medium-sized enterprises, especially those with relatively scarce capital and technology, they may face higher financing costs and thresholds for technology introduction. Therefore, future research should take into account the heterogeneity of different industries and enterprise types to further refine the implementation effect of green finance policies and explore how to solve the financing difficulties of small and medium-sized enterprises in green transformation through policy adjustments, thereby enhancing the overall green energy efficiency.

Secondly, there is still insufficient research into the impact mechanism. While research indicates that green finance policies encourage green energy efficiency by allocating labour, capital and technology more effectively, the precise manner in which this mechanism operates and its full impact remain unclear. For example, there is still a lack of in-depth quantitative analysis and model validation of how the innovation and diffusion of technology can be combined with the mobility of capital and the upgrading of labour force skills, particularly in relation to improving green energy efficiency. Future research should further refine the mechanism of green finance policies by adopting more complex econometric models and conducting in-depth explorations of the interactive effects of technological innovation and green investment. This will provide a more precise theoretical basis for policy design.

Finally, this study fails to adequately explore regional differences in the implementation of green finance policies, nor does it consider the variability in implementation by local governments. Although green finance policies have helped to reduce the green energy efficiency gap between northern and southern regions, as well as between coastal and inland areas, they may have widened the gap between provincial capital cities and their surrounding areas. This phenomenon may be influenced by multiple factors, such as the execution of policies by the local government, the local industrial structure, and the attractiveness of investment in green projects. However, existing studies have failed to analyse the specific impact of these factors on policy effectiveness in depth. This has resulted in the findings being difficult to apply to different cities and regions. Therefore, future research should pay more attention to differences in local governments’ implementation capacity, exploring ways to balance regional development disparities while ensuring the fairness and effectiveness of green finance policies in different regions. This can be achieved through adjusting local policies and designing green financial instruments at a local level.