Research on the impact of green finance on regional carbon emission reduction and its role mechanisms

Abstract

As a crucial instrument, green finance policy facilitates the transition toward a green and low-carbon regional economic structure, which is essential for realizing the target of “double carbon” and the harmonious coexistence of nature and humans. Therefore, taking the green financial reform and innovation pilot zone as a quasi-natural experiment, we select 270 cities from 2010 to 2021 as research samples and empirically assess the effects of the green finance policy on reducing regional carbon emissions through the double debiased machine learning (DDML) model. This study demonstrates that (1) green finance policy plays a significant role in promoting regional carbon emission reduction, and this conclusion remains valid after a variety of robustness tests; (2) the mechanism of action indicates that green finance policy contributes to regional carbon emission reduction by supporting green technological innovation and promoting the optimization of the industrial structure; (3) the analysis of heterogeneity reveals that green finance policy has a more pronounced effect on carbon emission reduction in the eastern region and in non-resource-based cities than in the central and western regions and in resource-dependent cities; and (4) the pilot policy of “Broadband China”, the pilot policy of information consumption, and the comprehensive experimental zone of big data has a synergistic effect on carbon reduction and emission reduction with green finance policy. The findings of this study not only contribute to deepening the understanding of the effects of the green finance pilot policy on regional carbon emission reduction but also provide policy support for local governments to explore green technology comprehensively, grasp new opportunities for green development, and expand the space for sustainable economic development with the assistance of the green finance pilot policy.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-02481-2.

Introduction

As global environmental problems become increasingly serious, sustainable economic development has become a common goal¹. Since the reform and opening up, China has achieved unprecedented levels of economic development, while energy and resource consumption have continued to grow, and carbon dioxide emissions have increased rapidly. As the largest carbon emitter globally, China has gradually strengthened its top-level design and constructed and implemented the “1 + N” carbon-neutral policy system, contributing to Chinese wisdom, solutions, and power to realize a beautiful vision of the carbon peak and neutrality on a global scale. Among them, the report of the 20th Party Congress put forward the strategic decision of “actively and steadily promoting carbon peak carbon neutrality”, which parallels the two paths of development and carbon emission reduction and realizes the goal of reducing carbon emissions without reducing productivity. As a crucial pillar of economic development, green finance has positively contributed to facilitating the transition toward green and low-carbon patterns, mitigating carbon emissions, and addressing global climate change². On this basis, in June 2017, the State Council’s executive meeting deliberated and approved the designation of green finance reform pilot zones, each with distinct emphases, across five provinces and autonomous regions: Zhejiang, Jiangxi, Guangdong, Guizhou, and Xinjiang. This study aimed to explore institutional mechanisms that can be replicated and scaled up. The establishment of green financial reform pilot regions (hereinafter pilot regions) serves to advance the development of a green financial framework and facilitate the green and low-carbon transformation of the broader economy and society. Studies have examined the factors influencing regional carbon emissions from the external environment, such as green credit, digital economy, industrial agglomeration, and economic growth. However, there is a lack of research on the mechanism of regional carbon emission reduction and the synergy between green finance policy and digital economy policies on regional carbon emission reduction from the perspective of green finance policy. Therefore, this paper selects the panel data of 270 prefecture-level cities in China from 2010 to 2021, empirically examines the carbon emission reduction effect of a green finance policy by using double machine learning, identifies the paths of its contribution to urban carbon emission reduction, further explores the synergistic effects of the “Broadband China” policy, the big data pilot zone, and the information consumption policy on carbon emission with green financial policy, provides for the effective implementation of the policy combination and explores the win‒win sustainable development mode of the economy and environment. The empirical results show that (1) a green finance policy can significantly promote regional carbon emission reduction; (2) a green finance policy can promote regional carbon emission reduction by increasing R&D investment, improving the efficiency of green technological innovation, guiding the flow of capital to the tertiary industry, and promoting the optimization of the industrial structure; (3) the heterogeneity shows that the green finance policy is better able to promote carbon emission reduction in the eastern region and resource cities, but for the central and western regions and non-resource cities, the green finance policy is not as effective; (4) digital economy policies such as the “Broadband China” policy, the big data pilot zone, the information consumption policy and the green finance policy can produce a synergistic effect on carbon reduction and emission reduction.

The contributions of this study are as follows. (1) Research perspective innovation. This study incorporates green finance policy and regional carbon emission reduction into the same research framework, enriches the research results concerning the effects of green finance on carbon emission reduction, and analyzes the impact mechanism of green technological innovation and industrial structure upgrading, thus providing a theoretical basis for exploring the carbon emission reduction effect of green finance policy, and is an important addition to the existing research. (2) Research method innovation. This study overcomes the limitations of the traditional double-difference model and synthetic control method, and with the advantage of being machine learning data-driven, the double machine learning model is applied in research on the policy effect of green financial reform so that the research conclusions are more scientific and reasonable. (3) Practical application innovation. This paper empirically examines the synergistic emission reduction effect of digital economy policies such as the “Broadband China” policy, the big data pilot zone and the information consumption policy and examines the heterogeneous impact of green finance policy on regional carbon emission reduction from the perspective of regional location and resource endowment differences. The results of this study provide empirical evidence for government departments to optimize the design of policy portfolios, accurately assess the effectiveness of the pilot policy, and further improve and expand the scope of the pilot policy.

The remainder of this paper is structured as follows: The second section presents a review of the related literature. The third section provides an analysis of the theoretical framework, elucidates the mechanism of the effects of a green finance policy on regional carbon emission reduction, heterogeneity, and synergistic policy effects, and subsequently presents the research hypotheses of this study. section delineates the model construction, variable descriptions, data sources, and processing methodologies. The fifth section encompasses baseline regression and mechanism analysis. The sixth section comprises an expansive analysis, including an examination of heterogeneity and policy synergies. The seventh section presents the results, a discussion, and policy recommendations. The appendix presents the results of the robustness test. The Technology roadmap is shown in Fig. 1.

Fig. 1.

The technology roadmap in this study.

Literature review

Domestic and international studies on carbon emissions are common at the academic level. Zhang et al.³, Wang et al.⁴, Chen et al.⁵, Zhang et al.⁶, Zhang et al.⁷, Liang et al.⁸, Peng et al.⁹, Liu et al.¹⁰ and Amin et al.¹¹ demonstrated that the digital economy, economic growth, green finance, financial technology, industrial agglomeration, industrial structure, industrial investment, science and technology, and eco-innovation are all key drivers of carbon emissions.

To research the relationship between green finance policy and carbon emissions, some scholars at the regional level have used the dynamic DID model, a spatial measurement model, to accurately assess the impact of green finance policy on carbon emissions at the prefecture and municipal levels, and to test the dynamic impact effect of policy implementation. They noted that a green finance policy can not only substantially reduce the carbon emissions of a city, but also have a significant spatial spillover effect, leading neighboring cities to synergistically reduce their emissions^12–15. Furthermore, in terms of the mechanism, some scholars have verified that green technology progress, industrial structure optimization, and the alleviation of financing constraints are the main ways for green finance to promote carbon emission reduction^5,16. Some scholars have demonstrated the moderating role of digital economic development and social capital in green finance for sustainable development^17,18. In terms of heterogeneity, Hao et al. verified that a green finance policy affects the green transformation of two types of enterprises and reduces carbon emissions in the pilot area by incentivizing innovation inputs from non-heavily polluting enterprises and inhibiting innovation inputs from heavily polluting enterprises through the construction of a triple-difference model¹⁹. Cui et al. proved that a green finance policy contributes to a decline in the carbon emissions of heavy polluters by curbing production and operational activities²⁰. Fan et al. reported that the impact of green finance on emissions reduction is more significant in non-state-owned and non-listed companies²¹. The specific literature analysis is presented in Table 1.

Table 1.

Literature analysis.

Literature type	References	Main findings
Factors affecting carbon emission	Zhang et al.3, Wang et al.4	The digital economy; economic growth
	Chen et al.5, Zhang et al.6	Green finance, financial technology
	Zhang et al.7, Liang et al.8, Peng et al.9	Industrial agglomeration, structure and investment
	Liu et al.10, Amin et al.11	Science, technology and eco-innovation
Effect of green finance policy on carbon emission	Zhang et al.12,Wang et al.13,Yang et al.14,Chen et al.15	Green finance policy can reduce the carbon emissions, and have a spatial spillover effect
Path of green finance policy to carbon emission	Chen et al.5,Dai et al.16	Green technology progress, industrial structure optimization and alleviation of financing constraints have mediating effects
	Liu et al.17,Solangi et al.18	Digital economy development and social capital have moderating effects
Heterogeneity of green finance policy on carbon emission	Hao et al.19,Cui et al.20	Reducing carbon emissions from heavy polluters
	Fan et al.21	Reducing carbon emissions from non-state-owned and non-listed companies

In summary, while the literature offers insights and methodologies for empirical research on the impact of a green finance pilot policy on carbon emissions, it also has certain limitations. First, the focus has focused primarily on the effects of green credit on carbon emissions, with limited exploration of the potential pathways through which green finance policy contributes to carbon reduction. Second, most studies have relied on traditional causal inference models, such as difference-in-differences and synthetic control methods, to evaluate policy impacts. However, these OLS regression-based approaches can suffer from the “curse of dimensionality” because of redundant control variables, and they may also introduce bias in the model specification, leading to unresolved endogeneity issues and decreased estimator accuracy. Therefore, this study contributes to existing research by empirically testing the positive effect of a green finance policy on carbon emission reduction through dual machine learning and further exploring the mediating role of green technology innovation and industrial structure upgrading.

Theoretical analysis and research hypotheses

Impact of the green finance policy on carbon emissions

General Secretary Xi Jinping has consistently highlighted the pivotal function of green finance in fostering low-carbon growth. He emphasized that it is necessary to “do a good job in the five articles of science and technology finance, green finance, inclusive finance, pension finance, and digital finance”, “improve economic policies for green and low-carbon development and strengthen financial support”, and “continuously optimizes the economic policy toolbox to support green and low-carbon development and give full play to the pulling effect of green finance”. Hence, green finance is essential for ensuring the achievement of green development objectives and “dual carbon” targets. By enacting green financial measures, the dual effectiveness of governmental steering and market modulation can be fully leveraged, fostering deep integration between the green and financial sectors, and ultimately achieving the mutual benefit of sustainable economic growth and environmental preservation²². First, green finance provides financial support for low-carbon projects. Green finance policies encourage financial institutions to set up special funds and preferential loans to low-carbon areas to provide financial protection for environmental protection enterprises, prioritizing the financing needs of key areas, such as clean energy, emission reduction and energy savings, and ecological protection, which provide development opportunities for emerging green industries and projects²³. Second, green finance is conducive to reducing the financing cost constraints for low-carbon and green enterprises. On the one hand, lower financing costs provide small and medium–sized enterprises with the opportunity to expand and flourish, help environmental protection enterprises overcome the high cost disadvantage of the pre-entrepreneurial period, and then access the delayed returns from environmental protection investments, creating conditions for the continued environmental protection operations of enterprises and forming a virtuous cycle of capital flow. On the other hand, higher financing costs lead polluting enterprises in policy-constrained industries to stop expanding in the short term or even downsize the existing scale, forcing polluting enterprises to carry out low-carbon and green transformation⁵. Third, green financial measures and environmental regulations work in tandem to steer the economy toward a green transition. The additional costs imposed by stringent environmental standards compel enterprises to engage in meaningful green technological advancements, leading to “offsetting benefits”²⁴. The incremental costs of disciplinary environmental regulations force enterprises to undertake substantial green technological innovations, resulting in “compensatory gains”. Therefore, an increase in environmental cost pressure encourages enterprises to actively avoid high-pollution industries and projects, thereby facilitating green transformation and upgrading the societal economy²⁵. Fourth, green financial measures encourage financial institutions to develop innovative products. By launching financial products such as green bonds, green credit, green insurance, and green funds, and exploring new green financial models such as carbon finance and climate finance, financial institutions can provide funding sources for low-carbon projects, broaden financing channels, and ensure successful docking between green industries and financial institutions²⁰. Accordingly, this study proposes the following hypothesis:

H1: Green finance policy can significantly contribute to carbon emission reduction in pilot cities.

Analysis of impact mechanisms

Drawing on the goals and attributes of green finance initiatives, this study expounds on the operational framework of green finance pilot zones in diminishing urban carbon emissions through advancements in green technology and industrial structure optimization.

1. Green technological innovation. First, the implementation of a green finance policy provides a suitable policy environment for enterprises’ green technological innovation, and effectively reduces the R&D risks and costs of green technologies. For green enterprises, green finance policy optimizes resource allocation through financial instruments and system designs, guide social capital flow to low-carbon and eco-friendly fields, help provide sufficient financial support for green enterprises to carry out green technologies, avoid the problem of a lack of funds in the process of green technologies, and stimulate the vitality of enterprises’ green technological innovation⁵. In addition, for “two high and one leftover” enterprises, the implementation of a green finance policy increases the difficulty of financing in the short term, and it is not realistic to carry out large-scale R&D investment. Therefore, to meet market demand for green products, these enterprises tend to improve the efficiency of green technological innovation and realize their own sustainable development²⁶. On the other hand, green technology innovation refers to innovation activities that follow ecological principles and ecological economic laws to save resources and energy, which are among the key factors in promoting carbon emission reduction²⁷. First, the innovation economics theory holds that technological innovation is an important driving force for economic growth and industrial transformation²⁸. Green technological innovation also has an innovation-driven effect, acting mainly on the production side, reducing the dependence of energy-consuming industries on traditional fossil fuels, and reducing greenhouse gas emissions in the production process by promoting the development and application of low-carbon technologies, such as clean energy, energy conservation and emission reduction, and carbon capture and storage²⁹. Second, green technology innovation can improve the production efficiency of products and effectively reduce the market price of green products, thereby narrowing the price gap between green products and traditional products, thereby helping to improve the recognition of green products among the public and promoting the public’s inclination toward green consumption and the formation of a low-carbon lifestyle³⁰. Finally, according to the theory of externalities, green technological innovation has technological spillover effects, which can be categorized into two types: MAR and Jacobs externalities. Regardless of the externalities, green technology innovation can effectively promote the synergistic cooperation of enterprises within the same industrial chain and between different industrial chains; promote the sharing and circulation of production factors, infrastructure, and green technology among enterprises; and simultaneously foster complementary strengths in low-carbon technology R&D, thereby enabling the green transformation and upgrading of the entire industrial chain^29,30.

2. Enhancing the industrial structure. The industrial structure depicts the distribution of resources among various industrial sectors in a given region, reflecting the proportionality between the levels of industrial development³¹. Industrial structure upgrading refers to the upgrading of industries along the industrial ladder³². According to Mathieu Clark’s law, as economic development progresses, the workforce initially shifts from primary industry to secondary industry and subsequently migrates to tertiary industry with an increase in per capita national income. Consequently, the upgrading of the industrial structure reflects the process of continuous optimization of the national economy and the direction of optimization and development of the regional economic structure. On one hand, a green finance policy can promote the upgrading of the industrial structure in a region. As a key way to achieve high-quality economic development within the constraints of environmental protection, a green finance policy can promote upgrading of the industrial structure by guiding the flow of capital and promoting industrial integration³³. In terms of guiding capital flow, rich green financial products can improve consumers’ green deposit capacity, increase green consumption behaviors, form synergies between the demand and supply sides, attract more social capital to low-carbon and green industries, and jointly promote the optimization and enhancement of the industrial structure³⁴. In promoting industrial integration, green finance policy has effectively supported the development of environmentally friendly enterprises through the issuance of green bonds, the establishment of green funds and other preferential measures, restricting the development of “two-high and one-remaining” enterprises, and incentivizing high-polluting enterprises to carry out self-sustaining technological innovation and accelerating the transformation of the industrial structure³⁵. On the other hand, industrial structure upgrading plays a key role in realizing carbon emission reduction³⁶, which influences the level of carbon emissions by increasing the efficiency of resource allocation and adjusting the composition of industrial structures³⁷. Initially, the optimization of the industrial structure enhances the interindustry resource allocation efficiency, thereby decreasing carbon dioxide emissions per unit of output. This promotes profound integration and synergy among diverse industries, ultimately spurring accelerated growth of the tertiary sector and effectively contributing to the reduction in regional carbon emission levels³⁸. Second, the direction of evolution of the industrial structure plays a decisive role in determining the trend of change in energy consumption and is also a key element in determining the change in incremental carbon emissions, with strong convergence among the three. In the current industrialization process, secondary industry comprises most of the high-energy-consuming sectors, and the carbon emissions it produces account for the largest proportion among the three types of industries. An advanced industrial structure can trigger structural effects through market returns and industrial policies, accelerate the transformation of the industrial structure from an industry-dominated industry to a service-dominated industry, achieve the goal of a “service-oriented economy”, reduce the dependence of economic development on energy and resource consumption, and effectively control and reduce the level of carbon emissions^39,40. Accordingly, this study proposes the following hypothesis:

H2: Green finance policy can promote carbon emission reduction in pilot cities through green technology innovation.

H3: Green finance policy can promote carbon emission reduction in pilot cities through industrial structure upgrading.

Heterogeneous characteristics

Considering the differences in the development characteristics of cities²², this paper further explores the heterogeneous effects of a green finance policy on regional carbon emission reduction.

First, the eastern region has a more developed economy and society, a higher degree of financial marketization, a higher level of green financial development, a richer range of green financial products, and a more standardized green credit review process; thus, the relevant financial institutions in the eastern region are able to implement the undertaking and incentives of the green finance policy more effectively and promote the effective allocation of green financial resources⁴¹. The eastern region tends to have greater awareness of low-carbon environmental protection, which is more conducive to the implementation of green finance policy to achieve more significant carbon emission reduction effects¹. Second, in terms of resource endowment heterogeneity, there was a difference between the carbon emission reduction effects of resource-based cities and those of other cities⁴². Resource-based cities, with their rich natural resources, present greater potential and advantages in terms of green industrial development and economic transformation, and their resource endowment provides strong conditions for the implementation of a green finance policy. By contrast, non-resource cities face greater challenges in the process of green transformation because of the relative lack of natural resources and a single or backward industrial structure. This inherent difference makes the implementation of the green finance pilot policy in non-resource cities relatively limited, especially in realizing the goals of pollution and carbon reduction, which are often less significant than those in resource cities⁴³. Accordingly, this study proposes the following hypothesis:

H4: Regional heterogeneity and resource endowment heterogeneity exist in green finance policy for regional carbon emission reduction.

Policy synergies

The digital economy provides new opportunities for the development of green finance through e-commerce, blockchain, mobile payments, and other electronic means, helping to reduce regional carbon emissions. On the one hand, from an investor’s perspective, the digital economy directly connects individuals and enterprises by building an open network investment platform. The platform is based on an information-sharing mechanism, which can reduce the cost of market information search, reduce the risk of information asymmetry, and improve the matching efficiency of investment and financing. In addition, digital means have released more financial resources, lowered the investment threshold for green financial products, and enabled groups with limited financial knowledge to participate in green investment through inclusive financial services, providing a human basis for the implementation of green finance policy. On the other hand, the digital economy, with the help of data analysis technology, effectively identifies and evaluates market risk; dynamically monitors the market performance of green financial products, investment returns, and corporate green credit ratings; and provides investors with a more scientific and accurate basis for decision-making⁴⁴. From the perspective of financial institutions, big data technology improves the quality of green resources acquired by financial institutions by cleaning and classifying massive information, solves the information asymmetry problem between potential green investors and financial institutions, urges financial institutions to create tiered and graded financial products that satisfy different demand groups, improves the design of financial products, realizes the continuous innovation of green products, and reduces the level of carbon emissions; blockchain technology will credit products and services efficiently embedded in the supply chain and industrial chain related to green financial support, thus improving the efficiency and transparency of bond business; artificial intelligence technology screens customers, identifies potential risk levels, and provides environmental risk early warning information to financial institutions through environmental risk assessment⁴⁵. Therefore, digital technology provides important support for the implementation of green finance policies and further enhances regional carbon emission reduction. The theoretical framework is illustrated in Fig. 2.

Fig. 2.

The hypothetical framework in this study.

H5: The synergistic effect of green finance and digital economy policies can significantly contribute to regional carbon emission reduction.

Study design

Modeling

Double machine learning models

Compared with the traditional multi-temporal double difference method, double machine learning is more suitable for this study in terms of model estimation and variable selection. However, the level of regional carbon emissions is influenced by various factors, and in view of the accuracy of empirical assessment, the model should incorporate these interfering factors as much as possible. Nonetheless, challenges such as multicollinearity and the “curse of dimensionality encountered when managing high-dimensional variables can detract from the estimation accuracy⁴⁶. Double machine learning reduces the estimation bias stemming from the “curse of dimensionality” by applying a regularization algorithm to a preselected group of highly predictive, high-dimensional control variables⁴⁷. On the other hand, a nonlinear relationship between the regional carbon emission level and the influencing factors is very likely to exist, and the conventional multiple linear regression model and the threshold effect model have obvious shortcomings in the process of setting up, which can easily lead to bias in the estimation results; however, dual machine learning, by virtue of its algorithmic advantages, can effectively avoid potential model selection bias when dealing with nonlinear relationships⁴⁸.

Drawing on the methodology of Chernozhukov et al.⁴⁹ this study employs a dual-machine learning model to investigate the influence and underlying mechanisms of green finance policies on reducing regional carbon emissions. The specific model settings are as follows:

1

where i is the city; t is the year; denotes the explanatory variable, which is the regional carbon emission level of city i in year t; is the core explanatory variable; if city i is a green financial reform pilot zone in year t; otherwise; the coefficient of disposition, denoted as , is the focus of this study;is the set of high-dimensional control variables that affect the explanatory variables and core explanatory variables through and ; and andare random error terms. and are the conditional expectations for and , respectively.

To estimate the treatment effect, , is first estimated via the machine learning model, and model (1) becomes:

2

To eliminate the estimation bias of obtained by the machine learning model, an orthogonal treatment is performed on , that is, the effect of on is eliminated, and the orthogonal regression variable is obtained: . where is a machine learning algorithm. At this point, the orthogonal regression variable can be regarded as an instrumental variable for , which is obtained by combining Model (2):

3

In the specific application of dual-machine learning models, because the specific functions of and are unknown, it is necessary to select appropriate machine learning algorithms to measure their estimators and . Compared with other machine learning algorithms, such as neural networks, ridge regression, and decision trees, the support vector machine (SVM) algorithm uses the principle of minimizing structural risk in its model selection process, that is, to minimize the generalization error while ensuring that the training error is minimized. This principle can effectively avoid overfitting, which gives the SVM algorithm strong generalization ability⁵⁰. Therefore, in this study, a support vector machine was used as an estimation algorithm for machine learning, with reference to Zhang and Li⁵¹. The test and training sets were divided at a 1:4 ratio, with all remaining unspecified parameters adopting the settings from the study by Chernozhukov et al.⁴⁹.

Multitemporal double difference models

When conducting a policy effect assessment, dual machine learning still needs to satisfy the basic assumptions of traditional double differencing, such as the parallel trend assumption and irreversibility assumption. Therefore, we construct the following DID model in the benchmark regression:

4

where is the city fixed effect, is the time fixed effect, is the random error term, and the remaining variables are consistent with Model (1).

To further examine the dynamic effects of green finance policy and conduct parallel trend tests, this study refers to Sun Weizeng et al.⁵², taking the current period of the policy as the base period and constructing the following model:

5

where indicates that city i is the treatment group, and indicates that city i is the control group; is the indicator function; is the current period of the establishment of the test zone; and the relative time from the establishment of the test zone is the reference system (), where is the base period, and the rest of the variables are the same as those in Model (2). The model focuses mainly on the regression coefficients and ; If these two coefficients are not significantly different from zero, it indicates that the DID model constructed in this study meets the parallel trend test⁵².

Variable settings

This study first draws on existing research designs to construct explanatory variables, core explanatory variables, and mediating variables, and then selects relevant characteristic variables from the economic, governmental, and social levels and provides detailed explanations to ensure the scientificity of variable selection in this study.

1. Explanatory variables. Following the methodology of Zhang et al., this study selected carbon emission intensity (carbon) as the explanatory variable, measured using the natural logarithm of the ratio of total regional carbon emissions to GDP¹². Compared with the indicator of total regional carbon emissions, carbon emission intensity focuses on evaluating the level of carbon emissions per unit of economic cost output, revealing the relationship between economic growth and carbon emissions; i.e., a reduction in carbon intensity better represents a reduction in the environmental costs caused by economic growth, which is more compatible with the goal of a “double carbon” economy.

2. Core explanatory variables. In this study, the green finance policy established in 2017 was treated as a quasi-natural experiment and DID was the core explanatory variable. If prefecture-level city i is a green finance reform pilot zone in year t, it is the treatment group, ; Otherwise, it is the control group, .

3. Mediating variables. Since patents are the best indicator of technological innovation [52], this study adopts the framework of Wang et al.³⁵ and Li et al.⁵³ Three indicators are used to assess the level of green technological innovation: the quality of green technological innovation (Greeinn_qua), the quantity of green technological innovation (Greeinn_num), and total green technological innovation (Greeinn_tot). Specifically, the quality of green technological innovation is quantified by the number of green invention patent applications, whereas quantity is measured by the number of green utility model patent applications. The total green technological innovation was calculated as the sum of these two categories. Additionally, owing to the high contribution of the industrial sector to carbon emissions, following Wang et al.,¹³ the level of industrial structure is measured by the proportion of the output value of tertiary industry to the output value of secondary industry in each region (Str).

4. Control variables. According to the studies of Zhang¹², Han⁵⁴, Li²², Zhao⁵⁵ et al., a city’s level of economic development, government attitude, and social development affect the city’s choice of a crude or refined economic development model, which further affects its level of carbon emissions. Therefore, this study selected a series of control variables from the three dimensions of economy, government, and society. In the economic dimension, the control variables include the industrialization level (Ind), measured as the ratio of total industrial output value to GDP; the financial development level (Fin), gauged by the proportion of financial institution loan balances to GDP at the year end in each region; the social consumption level (Eco), assessed using the ratio of total retail sales of consumer goods to GDP; and the urban economic density (DEco), quantified by the ratio of GDP to the administrative region’s land area. For the government dimension, the control variables are the degree of government intervention (gov), measured by the ratio of regional fiscal revenue to GDP, and government self-sufficiency (SGov), calculated as the ratio of the local general public budget revenue to the local general public budget expenditure. In the social dimension, the control variables are population density (Pop), which is determined by the ratio of the region’s total population to its administrative land area; education expenditure level (Edu), which is assessed through the ratio of education expenditure to local general public budget expenditure; and human capital level (Cap), which is measured by the ratio of students enrolled in general undergraduate and specialized schools to the region’s total population. Table 2 lists the specific variables.

Table 2.

Definition of variables.

Variables	Name	Measurement	Unit	References
Explanatory	Carbon	Ln (total regional carbon emissions/GDP)	billion tons/million dollars	Zhang et al.12
Core explanatory	DID	0–1 variable, assigned a value of 1 if the city is a pilot area, in 2017 and beyond, and 0 otherwise	–	Li et al.22
Intermediary	Greeinn_qua	Ln (number of patent applications for green inventions + 1)	–	Wang et al.35
	Greeinn_num	Ln (number of green utility patent applications + 1)	–
	Greeinn_tot	Ln(number of green invention patent applications + number of green utility model patent applications + 1)	–
	Str	Tertiary sector output/secondary sector output by region	–	Wang et al.13
Controls	Ind	Gross industrial output/GDP	–	Zhang12
	Fin	Year-end loan balances of financial institutions/GDP by region	–	Zhao55
	Eco	Total retail sales of consumer goods/GDP	–	Li et al.22
	DEco	GDP/land area of administrative regions	million/km2	Zhang12
	Gov	Regional revenues/GDP	–	Han54
	SGov	Local general public budget revenue/local general public budget expenditure	–	Zhang12
	Pop	Total population of the region/land area of the administrative region	persons/km 2	Li et al.22
	Edu	Number of students enrolled in general undergraduate programs/total population of the region	–	Zhang12

Data sources and processing

Based on the principles of data accessibility and research consistency, this study meticulously selected 270 prefecture-level cities in China spanning 2010–2021 as its focal research entities. The selection process excluded the samples with significant data gaps. The methodological framework draws inspiration from the practices of Liu, Hu et al.^56,57 and handles the samples as follows: (1) some of the missing data in the individual samples are filled by the linear interpolation method; (2) to mitigate the impact of outliers, the continuous variables in this study were adjusted via the winsor2 command in Stata, which truncates the top and bottom 1% of values. The data in this study came from the China Energy Statistical Yearbook, China Statistical Yearbook, Wind database, and statistical yearbooks of provinces and cities. All empirical analyses were performed using Stata 17.0. Table 3 presents the descriptive statistics of the variables. Among the 3240 samples, the mean value of carbon was 2.715, with a median of 2.689, which was a small difference, and the overall distribution was normal. The extreme deviation is 6.268, and the standard deviation is 1.457, which is greater than 1, indicating that the carbon emission intensity varies greatly between samples, which provides a possibility to study regional carbon emission reductions.

Table 3.

Variables’ descriptive statistics.

Variables	N	Mean	Median	SD	Max.	Min
Carbon	3,240	2.715	2.689	1.457	6.434	0.166
DID	3,240	0.009	0	0.096	1	0
Greeinn_qua	3,240	4.219	4.043	1.731	8.725	0.693
Greeinn_num	3,240	4.558	4.511	1.607	8.548	1.099
Greeinn_tot	3,240	5.130	5.011	1.642	9.285	1.609
Str	3,240	1.023	0.891	0.537	3.386	0.293
Ind	3,240	0.197	0.177	0.109	0.641	0.017
Fin	3,240	2.485	2.167	1.147	6.755	1.008
Eco	3,240	0.381	0.373	0.106	0.690	0.129
DEco	3,240	0.306	0.139	0.495	3.179	0.006
Gov	3,240	0.197	0.172	0.094	0.573	0.074
SGov	3,240	0.464	0.434	0.220	0.999	0.098
Pop	3,240	429.0	359.5	309.7	1,555	11.86
Ed	3,240	0.176	0.175	0.038	0.267	0.087
Cap	3,240	0.020	0.010	0.026	0.121	0.001

Empirical analysis

Benchmark regression

Table 4 presents the estimation results of the effects of green finance policy on regional carbon emission reduction. Specifically, column (1) examines the net effect of this estimation result, that is, by incorporating only city and year fixed effects, the regression coefficient of DID is -0.2654 and statistically significant at the 1% level. Column (2) further demonstrates that, after including a series of control variables, the regression coefficient of green finance policy is − 1.1195, which remains statistically significant at the 1% level. Consequently, the results in Table 4 indicate that green finance policy can significantly contribute to regional carbon emission reduction, substantiating H1. The conclusion remains robust after a series of robustness tests, the results of which are presented in the Appendix.

Table 4.

Benchmark regression results.

Variables	(1)	(2)
	Carbon	Carbon
DID	− 0.2654***	− 1.1195***
	(0.0839)	(0.1175)
_cons	0.0438***	− 0.0521**
	(0.0102)	(0.0227)
Controls	Yes	Yes
Urban fixed effects	Yes	Yes
Year fixed effects	Yes	Yes
N	3237	3237
R2	0.084	0.117

***p < 0.01, **p < 0.05, *p < 0.1; Standard errors in parentheses.The same below.

Mechanistic analysis

Drawing on the aforementioned theoretical analysis, this study explores the underlying mechanisms through which green finance policy mitigates carbon emission intensity, focusing on two pathways: green technology innovation and industrial structure. Referring to Jiang⁵⁸, we constructed a two-step model of the mediating effect. Step 1: Test whether the green finance policy has an inhibitory effect on regional carbon emission levels. If the corresponding regression coefficient is significantly negative, then proceed to Step 2: Test whether the green finance policy has a promotional effect on the mediating variables. If the corresponding regression coefficient is significantly positive, then argue for the inhibitory effect of the mediating variables on carbon intensity through the existing references to determine whether there is a mediating effect of the mediating variables. The specific test model is as follows:

6

where is the set of mechanism variables measuring green technological innovation and industrial structure upgrading, and the rest of the variables are the same as those in the baseline regression model (1).

The explanatory variables in columns (1) to (4) of Table 5 are Greeinn_qua, Greeinn_num, Greeinn_tot, and Str, respectively, and the regression coefficients of the green finance policy are significantly positive at the 1% confidence level. According to the theoretical analysis, green technological innovation and upgrading the industrial structure can effectively inhibit the intensity of carbon emissions. Therefore, it can be concluded that a green finance policy can reduce regional carbon emissions by promoting green technological innovation and industrial structure upgrading, and research hypotheses H2 and H3 are established.

Table 5.

Mechanistic test results.

Variables	(1)	(2)	(3)	(4)
	Greeinn_qua	Greeinn_num	Greeinn_tot	Str
DID	1.8903***	1.8980***	1.8995***	0.2574**
	(0.1256)	(0.0778)	(0.0901)	(0.1217)
_cons	0.1864***	0.1527***	0.1821***	0.1366***
	(0.0267)	(0.0232)	(0.0240)	(0.0147)
Controls	Yes	Yes	Yes	Yes
Urban fixed effects	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes
N	3240	3240	3240	3240
R2	0.126	0.078	0.090	0.132

Extensibility analysis

Heterogeneity analysis

Based on the above theoretical analysis, this study further explores the carbon emission reduction effect of a green finance policy based on the levels of regional heterogeneity and resource endowment heterogeneity.

1. Regional heterogeneity.

Shen Xiaobo et al.⁵⁹ divided 270 prefecture-level cities into two regions, eastern and central-western, and substituted them into model (1) for comparative analysis. The results of the comparison are presented in Columns (1) and (2) of Table 6, where the regression coefficients of DID are all significantly negative at the 1% confidence level. However, the current double machine learning model is unable to test the regression coefficient differences between groups using the Fischer combination. Thus, the triple-difference model can be combined effectively with the DDML model. Specifically, refer to Qian et al.⁶⁰; cross-multiply the city dummy variable (zone = 1 if it is an eastern city; zone = 0 if it is a central-western city) with DID to obtain its core explanatory variable DDD (DID*zone) and re-substitute this variable into Model (1) for estimation. According to Column (3) of Table 6, DDD is significantly negative, indicating that the green finance policy has a greater inhibitory effect on carbon emission intensity in the eastern regions than in the central-western regions. Hypothesis 4 states that there is regional heterogeneity in the impact of green finance policy on regional carbon emission reductions.

Table 6.

Heterogeneity test results.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	DDML		DDD-DDML	DDML		DDD-DDML
	Eastern	Central-western	Carbon	Resource	Non-Resource	Carbon
DID	− 0.7633***	− 1.6140***		− 1.4378***	− 0.7479***
	(0.1004)	(0.0777)		(0.1240)	(0.1472)
DDD			− 0.8487***			− 1.2783***
			(0.1316)			(0.1504)
_cons	0.0514*	− 0.1826***	− 0.0306	− 0.0839**	− 0.0116	− 0.0727***
	(0.0284)	(0.0272)	(0.0235)	(0.0337)	(0.0282)	(0.0246)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Urban fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
N	1199	2038	3240	1354	1883	3240
R2	0.100	0.078	0.132	0.124	0.147	0.150

• (2)Resource endowment heterogeneity.

According to Zhang et al.¹², 270 prefecture-level cities are categorized into resource cities and non-resource cities and regressed in groups based on the relevant provisions of the Circular on the National Sustainable Development Plan for Resource Cities (2013–2020). Columns (4) and (5) of Table 6 present the results of the comparison, where the regression coefficients of DID are all significantly negative at the 1% confidence level. Therefore, similar to the above, the dummy variable of resource endowment (resource = 1 in the case of resource cities and resource = 0 in the case of non-resource cities) is cross-multiplied by DID to obtain the core explanatory variable, DDD (DID*resource), and this variable is substituted into model (1) for estimation. According to Column (6) of Table 6, the coefficient of the triple interaction term has a notably negative value, suggesting that the carbon emission intensity of resource-based cities is more profoundly impacted by green finance policy than that of non-resource-based cities. Hypothesis 4 states that there is resource endowment heterogeneity in the impacts of a green finance policy on regional carbon emission reductions.

Policy synergies

Although China has formed a multi-level green financial system that includes green bonds, green loans, green trusts, and green funds, from the perspective of product supply, green financial products are mainly based on green credit and bonds. The market supply of green funds, green trusts, green insurance, and other products is relatively small, and the innovativeness of new green financial products is insufficient⁶¹.

From the perspective of service efficiency, China’s green financial services are characterized by cumbersome approval procedures, complex issuance processes, restricted radiation coverage, insufficient employment, and high service costs⁶². Against the backdrop of accelerating global digitization, digital economic policies are key to addressing the existing constraints of green finance⁶³. Digital policy, with digital infrastructure construction, digital technology innovation, and information consumption development as its main connotations, encourages and guides financial capital to continuously support the low-carbon and green sectors by realizing the complementary roles of policy-driven and financial support, thus exerting a synergistic effect of carbon reduction^52,64,65.

1. Digital infrastructure: forging a solid skeleton for the digital economy, taking the Broadband China” policy as an example. In terms of information sharing, the development and improvement of digital infrastructure can continue to expand the information coverage of the overall financial market, deepen the degree of digital network applications, optimize the information-sharing mechanism among financial institutions, and smooth communication channels⁶³. In terms of transaction risk reduction, the combination of “green finance + blockchain” ensures that the financing smart contract is open and transparent and that the relevant transactions are irreplaceable, making the flow of information more efficient and timely, significantly reducing the risk of financial transactions, and effectively solving the problem of “financing difficulties” for SMEs with poor qualifications or imperfect information in green transformation⁶⁶. In terms of enhancing the allocation efficiency of green finance, the mechanism is more efficient and timely. To enhance the allocation efficiency of green finance, first, the digital financial platform provides trading venues and cognitive scenarios for green credit, bonds, green insurance, and other forms of green financial products; avoids information asymmetry in the process of advancing green finance; and improves the matching rate of financing information⁶⁷. Second, the “ToC + ToB” port is fully developed with the support of digital infrastructure, and the two-way investment and financing enterprise further enriches the realization of green finance, multiplies the total amount of funds, improves the speed of aggregation, and improves the efficiency of service so that financial institutions can reasonably adjust the ratio of funds while effectively reducing transaction costs⁶⁶. To analyze the synergistic carbon reduction and mitigation effects of digital infrastructure policies and green finance policy, the “Broadband China” policy (broadband) and green finance policy (DID) are cross-multiplied to obtain the core explanatory variable DID*Broadband, which is substituted into Model (1) for estimation. According to column (1) of Table 7, DID*Broadband is significantly negative, showing that the Broadband China policy and green finance policy synergize to suppress carbon emission intensity.

2. Digital technology: An efficient engine for enhancing economic operations, taking the big data pilot zone as an example. In terms of increasing green financial product innovation, advanced digital technologies such as artificial intelligence and cloud computing can, on the one hand, accurately perceive the needs of different users in different life cycle stages and consumption scenarios by increasing the rate of massive information collection, processing and transmission and providing arithmetic and data support for the innovation and supply of green financial products⁶⁸. On the other hand, it can accurately predict the yield of each financial product and stimulate individual users to increase their demand for purchasing green financial products while enhancing the degree of innovation in green financial products by financial institutions. For example. In terms of enhancing the level and efficiency of green financial services, consequently, this technology reduces the duration required by financial institutions to identify, verify, and review green enterprises, thereby accelerating the process. Digital technology can first interface the business systems of financial institutions with the foundational data platforms of information systems. It facilitates the establishment of a social network encompassing financiers of green projects and enables the generation of precise environmental benefit assessment reports, thus shortening the time for financial institutions to identify and certify the review of green enterprises⁶⁹. Second, it has the potential to enhance decentralized and streamlined service procedures within the traditional financial system, reduce the approval time for each link, effectively simplify the green financial service process, and provide a good experience for customers⁷⁰. Third, it can effectively apply massive customer resources and form strong network externalities, which makes the marginal cost for financial institutions to serve a single new customer decrease continuously⁷¹. To analyze the synergistic carbon reduction and mitigation effects of digital technology policies and green finance policy, big data pilot zones (big data) and green finance policy (DID) are cross-multiplied to obtain the core explanatory variable DID*big data, which is substituted into Model (1) for estimation. According to column (1) of Table 7, DID*Big Data is significantly negative, indicating that the Big Data pilot zone and green finance policy synergize to suppress carbon emission intensity.

3. Digital consumption: A strong impetus to stimulate the economic cycle, using the information consumption policy as an example. From the perspective of consumption willingness, the continuous expansion of information consumption enterprises and the widespread popularization of mobile communication devices, smart terminals, and other information products are conducive to breaking through the time and space barriers to information dissemination, broadening access to knowledge, enhancing the rate of information dissemination, and providing the public with the necessary means to search for environmental information. Ultimately, information consumption can enhance consumers’ perceptions of environmental pressure and promote a shift in residents’ consumption attitudes toward green and low-carbon technologies⁷². From a power consumption perspective, lowering the threshold of access to green products can enhance residents’ green consumption behavior. Information consumption has continuously spawned e-commerce, digital inclusive finance, the sharing economy and other green and low-carbon emerging consumption modes, which reduces the difficulty of obtaining green and low-carbon products, broadens the channels for residents to practice green and low-carbon consumption, and improves the popularity and acceptance of such products in social groups, thus realizing the objective of “reaching out” with products⁷³. In addition, the “reachability” of products can increase the public’s awareness of environmental protection and reduce regional carbon emissions. To analyze the synergistic carbon reduction and mitigation effects of digital consumption policies and green finance policy, information consumption policy (Information) and green finance policy (DID) were cross-multiplied to obtain the core explanatory variable DID*Information, which was substituted into Model (1) for estimation. According to column (1) of Table 7, DID*Information is significantly negative, showing that the pilot information consumption policy and green finance policy synergize to suppress carbon emission intensity.

Table 7.

DDD-DDML test results.

Variables	(1)	(2)	(3)
	Carbon	Carbon	Carbon
DID*Broadband	− 1.5052***
	(0.0674)
DID*Big Data		− 1.2419***
		(0.0909)
DID*Information			− 1.5937***
			(0.0731)
_cons	− 0.0918***	− 0.0670***	− 0.1027***
	(0.0206)	(0.0211)	(0.0208)
Controls	Yes	Yes	Yes
Urban fixed effects	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes
N	3240	3240	3240
R2	0.067	0.091	0.073

Results and discussion

Taking the green financial reform pilot zone as the entry point, this study selected 270 prefecture-level cities in China from 2010 to 2021 as the research sample and analyzed the role of green finance policy in influencing regional carbon emission reduction via a dual machine learning model. The main findings of this study are as follows.

The conclusion that a green finance policy can significantly suppress the carbon emission intensity of pilot cities still holds after various robustness tests, similar to the findings of Zhang et al.¹² and Li et al.²². The former focuses on the positive role of green finance policy in promoting regional carbon emission reduction by influencing credit allocation through environmental constraints, whereas the latter discusses the role of green finance policy in carbon emission reduction in terms of the dual impact of government promotion and market regulation. This study integrates previous studies and argues that green finance policy achieves the synergistic effect of carbon emission reduction through the dual path of market regulation and government guidance. At the market level, green finance policy through the construction of a green credit system and special financing mechanism not only directly alleviates the burden of financing costs of low-carbon enterprises but also promotes the application of clean energy technology with the help of the capital allocation guidance effect to reduce the intensity of carbon emissions from the source of production and operation activities of enterprises; at the level of government regulation, green financial policy uses differentiated credit policy, environmental information disclosure and other regulatory means to form an innovative forcing mechanism to force traditional high-polluting enterprises to implement green technological transformation, which effectively reduces regional carbon emissions²³. At the government regulation level, green finance policy uses differentiated credit policies, environmental information disclosure, and other regulatory means to form a push mechanism for innovation, forcing traditional high-polluting enterprises to implement green technological transformation, which effectively reduces regional carbon emissions²⁵.

Mechanistic research shows that green finance policy mainly promotes green technology innovation and industrial structure upgrading to suppress carbon emission intensity in pilot cities, which is similar to the conclusions of studies conducted by Chen et al.⁵, Zhang et al.¹², and Tian et al.²⁹. These scholars select green technology progress, optimization of the energy structure, and industrial structure as explanatory variables or intermediary variables and explore the impact of green finance policy on provincial- or municipal-level carbon emission reduction. With respect to the impact of carbon emissions at the provincial or municipal level, this study selects two intermediary variables, green technology innovation and industrial structure upgrading, and applies the dual machine learning model to validate and analyze the role of green finance policy in promoting regional carbon emission reduction. The possibility of such a mechanism is as follows: Green finance policy promotes green technological innovation and the application of low-carbon technologies by guiding the flow of funds to green projects, which in turn reduces the intensity of carbon emissions. Simultaneously, the policy encouraged upgrading the industrial structure, supported the development of green and low-carbon industries, and eliminated traditional industries with high pollution and high energy consumption. Additionally, green finance combines policy support (e.g., tax incentives) and market mechanisms (e.g., carbon trading) to optimize resource allocation, promote green technological innovation, and promote industrial structure upgrading, thereby effectively curbing regional carbon emissions.

The study of regional heterogeneity shows that, compared with the central and western regions, the green finance policy has a greater inhibiting effect on carbon emission intensity in the eastern region, contrary to the results of Ma et al.⁴³ and Chen et al.⁵ and similar to the results of Li et al.⁷⁴. This is probably because of the following reasons. First, advanced economic development in the eastern region provides a robust foundation for a green finance policy. In economically developed areas, governments and firms possess greater fiscal and technological capacity to invest in clean energy and environmental technologies (e.g., renewable infrastructure and carbon capture), thereby reducing emission intensity more effectively⁷⁴ Second, the eastern region’s industrial structure is more diversified and service-oriented, aligning with the later stages of the Environmental Kuznets Curve⁷⁵ where economic growth correlates with declining pollution levels. Green finance policy accelerates this transition by channeling capital toward renewable energy and high-tech sectors (e.g., via green bonds or preferential loans), thereby phasing out carbon-intensive industries⁷⁶. This structural shift is less feasible in the central and western regions, where industrialization remains resource dependent. Finally, the eastern region’s mature financial markets, characterized by higher liquidity and innovation capacity, adhere to green finance theory⁷⁷ which emphasizes the role of financial instruments in mitigating environmental risk. With well-developed green credit systems, carbon trading platforms, and ESG (environmental, social, and governance) investment products, the region leverages market mechanisms to incentivize low-carbon transitions¹. In contrast, weaker financial intermediation in the central and western regions limits policy penetration.

Resource endowment heterogeneity suggests that green finance policy has a greater inhibitory effect on carbon emission intensity in resource cities than in nonresource cities, contrary to the results of Zhang et al.¹² and Ma et al.⁴³. The probable reasons for this are as follows: the economic development of resource-based cities is highly dependent on energy-intensive industries, with high-energy-consuming industries, such as coal, oil, iron, and steel, which dominate and emit high levels of carbon emissions. By implementing green financial policies and providing financial support for the research, development, and application of clean technology, energy saving, and emission reduction technologies, these cities can more effectively address the problem of high-emission sources and achieve pollution and carbon reduction targets. By contrast, nonresource cities with limited resource endowments have underdeveloped industrial structures, lower absorption capacities for green financial policies, and relatively limited carbon emission reduction effects.

Research on policy synergies shows that there are synergistic effects between the green finance policy and the “Broadband China” policy, the big data pilot zone, and the information consumption policy, which together improve regional carbon emission reduction. This shows that the digital economy has become a key way to overcome the current bottleneck in green finance in the context of accelerating the global digitization process. Through the complementary mechanism of policy guidance and financial support, digital policy with digital infrastructure construction, digital technology innovation, and information consumption development continue to incentivize and guide the flow of financial capital to green and low-carbon areas, thus realizing a synergistic carbon reduction effect while promoting the development of green finance^63,64.

In view of the above empirical findings, the following conclusions are drawn:

First, to further expand the green financial reform pilot zones as the primary focus, the impact of green finance policy on carbon emission reduction should be maximized. At present, China’s green financial reform remains experimental in primary terms, and from the perspective of the policy pilot effect, the green finance policy has suppressed carbon emissions in the pilot zones. Hence, there is an urgent need to strengthen the momentum of green finance to fully utilize its key role as an enabler. On the one hand, based on the successful experience of the pilot zones, the number of pilot reform zones should continue to increase moderately, with the aim of accumulating practical experience for the comprehensive implementation of green finance policy and ensuring that the pilot zones will play a leading role in promoting synergies between carbon emission reduction and pollution mitigation. However, in the subsequent process of expanding the scope of the pilot area, full consideration should be given to regional and industrial characteristics, and clearer constraints and incentives should be designed for resource-oriented cities to improve the precision and orientation of green financial policies. With respect to heavily polluting enterprises, it is still necessary for financial institutions to increase financial support for their transformation and upgrading to ensure that more funds are used to promote energy conservation, emission reduction and technological transformation; to facilitate the transformation of heavily polluting enterprises; and to achieve “double optimization” of environmental protection and economic benefits.

Second, we focus on green technological innovation to promote the refinement and upgrading of industrial structures. The advancement of eco-friendly technologies and refinement of industrial structures are pivotal in advancing regional carbon emission reductions through green finance policies. Governments can implement measures, including tax relief and policy incentives, to incentivize enterprises to devote themselves to engaging in the research, development, and deployment of green technologies, effectively steer green financial resources toward green innovation and clean energy utilization, and simultaneously narrow the scope for polluting industries, compelling them to undergo upgrades and transformations, thereby reducing carbon emissions in their operational processes. However, in the process of policy design and implementation, it is necessary to fully consider the actual conditions and clarify that the phased contraction of polluting industries is not the ultimate objective. The government should take the initiative to dock special funds and technical support, strengthen the construction and improvement of regional infrastructure, increase capital investment in green finance, build a green financial science and technology service platform, and realize intelligent identification, evaluation, and tracking and monitoring of emerging green projects to ensure that the green finance policy can be implemented accurately and produce effective results. In addition, the bridging role of technological innovation and energy structure adjustment in promoting green finance can help regional carbon reduction, deepen the practice of green finance, and help polluting enterprises to successfully transform into green enterprises, thus promoting the fundamental upgrading of the industrial structure.

Third, to achieve significant reductions in carbon emissions, it is imperative to implement tailored emission mitigation strategies that take into account the unique circumstances of each region. In the actual implementation of the policy, the differences in the geographic environment and resource conditions of the pilot cities must be fully considered for integrated planning. Specifically, in economically developed regions, green credit service programs should be continuously explored and innovated, the intensity of green financial innovation should be increased, and the types of green financial products should be broadened. In regions rich in natural resources, the efficient use of energy should be accelerated to achieve a harmonious unity of ecological protection and economic development. In contrast, for cities in central and western China or those with scarce resources, on the one hand, local governments, enterprises and the public can be provided with professional training in green finance and sustainable development to increase their awareness of and attach importance to green development. However, investments in transportation and other infrastructure should be increased further to improve the overall level of infrastructure development.

Fourth, the establishment of a synergistic mechanism for carbon emission reduction policies should be accelerated to fully unleash the maximum effectiveness of each policy and realize “1 + 1 > 2”. First, focusing on the deep integration of digital economy policies and green finance policy to drive the achievement of urban carbon emission reduction targets, actively exploring and promoting synergistic carbon emission reduction paths with universal applicability, and promoting the sharing of carbon emission reduction experience and resources to achieve carbon emission reduction targets; second, ensuring the maximization of the synergistic effect of the policies, it is necessary to establish a cross-sectoral coordinating mechanism and effectively integrate green finance, “Broadband China,” the big data pilot zone and information consumption policy, to build a unified and coordinated policy framework, and conduct regular policy evaluation and adjustment to ensure that all policies are closely linked in terms of target setting, implementation measures, and incentive mechanisms to form strong policy synergy. Finally, the development of various digital economic policies should be improved to reduce carbon emissions in the region. Deepen the “Broadband China” pilot policy, utilize advanced broadband technology to increase green technological advancements, increase energy efficiency, mitigate carbon emissions, increase the construction of comprehensive pilot zones for big data, utilize big data technology to accurately identify green projects and carry out risk assessment, introduce big data risk control models to improve the security and efficiency of green financial enterprises, expand information consumption policy to encourage the R&D and innovation of intelligent terminal products, expand emerging service industries, and promote the demonstration and popularization of e-commerce, the Internet of Things and other applications to promote the in-depth fusion and development of information consumption and green and low-carbon industries.

The research presented in this paper has certain limitations that warrant further investigation. First, owing to the inability of the current dual machine learning methods to address the parallel trend hypothesis, only traditional parallel trend test methods can be employed. Second, in recent years, China has implemented new location-oriented policies, such as “artificial intelligence innovation pilot zones” and “data elements,” and the synergistic effect of these policies with green finance merits additional exploration in future studies. Finally, while this study’s research sample included 270 prefecture-level cities in China, the impact of green finance on carbon emissions in other countries requires further examination in subsequent research.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

H.L., Z.Y. and G.C. designed the study. H.L. carried out the experimental research work. Z.Y. and H.L. analyzed the data. G.C. and Y.N. supervised the study. All authors participated in writing and reviewing the manuscript. All authors read and approved the final manuscript.

Funding

The project was supported by Post funding project of China National Social Science Fund (21FTYB007).

Data availability

The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.