How Can Government Regulation Reinforce the Low-Carbon Effects of Green Finance in China? Heterogeneity of Resource-Based Cities and High-Energy-Consuming Cities

Abstract

Based on panel data of 273 prefecture-level cities in China from 2006 to 2022, this study empirically examines the impact and mechanism of green finance on urban low-carbon development from the perspective of government regulation. The study yields the following key results: First, green finance significantly promotes urban low-carbon development, and this finding remains valid after addressing endogeneity issues and conducting a series of robustness tests. Second, government regulation strengthens the low-carbon effect of green finance through two pathways: market-oriented reform and urban land planning. Third, the low-carbon effect of green finance is more pronounced in resource-based cities and low-energy-consuming cities, which corresponds to urban disparities in development stages and resource constraints. Given these results, this study proposes two targeted recommendations: institutionalizing the coordination mechanism between land use and carbon markets and implementing context-specific green finance strategies via urban differentiation approaches.

1. Introduction

In recent years, escalating global climate change, progressive depletion of natural resources, and intensifying environmental pollution have propelled sustainable development strategies to the forefront of socioeconomic agendas worldwide. Green finance serves dual functions: combating climate change through capital reallocation to low-carbon sectors and propelling high-quality economic development (Xu & Dong, 2023; Liu & Li, 2024). According to the 2023 Statistical Report on Financial Institution Loan Allocations released by the People’s Bank of China (PBOC), China’s outstanding green loans in both domestic and foreign currencies amounted to RMB 30.08 trillion by the end of 2023, reflecting a 36.5% year-on-year growth. Notably, loans allocated to projects with direct and indirect carbon reduction benefits totaled RMB 10.43 trillion and RMB 9.81 trillion, respectively, jointly constituting 67.3% of the aggregate green loan portfolio. The 2024 PBOC Guiding Opinions further outline a five-year blueprint for establishing a world-leading green finance system.

Green finance refers to financing investments that provide environmental benefits within the broader context of environmental sustainability (G20 Green Finance Study Group, 2016). It refers to a systematic arrangement that uses financial instruments such as green credit, green investment, green insurance, green bonds, green support, green funds, and green equity to guide capital flows toward environmental protection, energy conservation, clean energy, and other areas. It has been widely recognized as a key driver in promoting the development of a low-carbon economy (Chen & Chen, 2021; Meo & Karim, 2022; Umar & Safi, 2023; Alharbi et al., 2023). However, its specific mechanisms and economic effects require further exploration. Existing research has primarily focused on the micro-level impacts of green finance, such as corporate green technology innovation (Feng et al., 2022; H. Huang et al., 2022) and improvements in energy efficiency (L. Zhang et al., 2022; Li & Umair, 2023). Studies have also examined the role of green financial instruments—including green bonds and green funds—in directing capital toward low-carbon industries (Monk & Perkins, 2020; Appiah & Essuman, 2024). Moreover, China’s green finance policies are transitioning from direct intervention to market-oriented approaches, with the government continuing to play a crucial role in this process (S. Zhang et al., 2021; Shao & Huang, 2023). Regarding economic effects, studies suggest that green finance contributes to high-quality economic development and facilitates a green and inclusive transition in China (Xu & Dong, 2023; Van Niekerk, 2024).

Government regulation serves as a critical policy tool in facilitating the transition to a low-carbon economy. Scholars widely advocate for the use of composite policy instruments—such as green credit and carbon trading mechanisms—to advance low-carbon objectives (Chen & Lin, 2021; Debrah et al., 2023). Research indicates that government regulation can alleviate financing constraints, promote green innovation, and drive green technological transformation through the improvement of environmental legislation (Yu et al., 2021; Zhou & Du, 2021). Furthermore, by maintaining market order and enhancing marketization levels, regulation improves the efficiency of financial resource allocation, thereby supporting low-carbon development (Pereira et al., 2023). In terms of spatial governance, regional collaborative mechanisms and resource integration have also been shown to contribute to achieving emission reduction targets (Ke et al., 2022). Overall, through market incentives and industrial restructuring—particularly in areas such as green industrial reorganization and carbon emissions trading—government regulation plays an essential role in promoting structural low-carbon transitions (Gu et al., 2021; Bayer & Aklin, 2020; Rahma et al., 2019).

In summary, while existing studies have largely examined the separate impacts of government regulation or green finance on the low-carbon economic transition, few have integrated both to analyze their synergistic effects. From the perspective of government regulation, this study employs quantitative data from 273 Chinese cities to explore how to enhance the low-carbon effects of green finance, and investigates the interplay and mechanisms between the two, thereby proposing practical pathways for urban low-carbon transition. The contributions of this paper are threefold: First, it examines how government interventions—such as market-oriented reforms and land planning—strengthen the carbon reduction effect of green finance, revealing heterogeneous effects under different regulatory and financial instrument interactions. Second, using prefecture-level city data offers finer granularity to verify the internal logic through which green finance, under government regulation, accelerates urban low-carbon transition, enhancing the reliability and policy relevance of the findings. Third, it proposes policy options to deliver practical value for achieving the dual carbon goals, including promoting the integration of land management systems with carbon market mechanisms and advancing context-adapted green finance development that aligns with local conditions.

Examining the mechanisms and effects of green finance in urban low-carbon transitions from the perspective of government regulation not only enriches the theoretical framework of existing research but also provides innovative practical pathways for unleashing the full potential of green finance, holding significant academic and practical implications.

The subsequent sections of the paper are structured as follows: Section 2 introduces the theoretical framework and research hypotheses regarding how policy regulations enhance the low-carbon effects of green finance; Section 3 constructs a two-way fixed-effects model using panel data from 273 prefecture-level cities in China between 2006 and 2022; Section 4 presents the baseline empirical results and robustness checks; Section 5 examines the underlying mechanisms; Section 6 explores heterogeneity effects; and Section 7 concludes with policy implications.

2. Theoretical Analysis and Research Hypotheses

2.1. Green Finance and Urban Low-Carbon Development

Green finance indirectly amplifies the effect of government regulation on carbon emission efficiency by facilitating the efficient flow of capital, technology, information, and resources within the market. From a capital perspective, it lowers barriers to market entry, alleviates corporate financing constraints, and provides sufficient funding for energy-saving technological innovation. Technologically, green finance stimulates enterprises to increase investment in the research, development, and application of energy-saving technologies, thereby promoting green technological upgrading. From an informational angle, it establishes market-based incentive mechanisms that optimize the allocation of energy resources. Through signaling theory and market-driven price signals, it enables more accurate assessment and guidance of carbon usage efficiency, reduces waste, and further enhances overall carbon emission performance. Accordingly, the following hypothesis is proposed:

H1. 

Green finance facilitates urban low-carbon transition.

2.2. Green Finance, Government Regulation, and Urban Low-Carbon Development

The core of green finance lies in promoting the application and adoption of low-carbon technologies through optimized capital allocation, risk pricing, and enhanced information transparency. However, certain limitations persist in its operational mechanisms. First, there is a conflict between market profit-seeking motives and public environmental goals, leading to frequent instances of greenwashing. Some enterprises utilize green financial instruments to meet public objectives without implementing substantial emission reductions, resulting in superficial compliance. Second, the coverage of green finance remains limited. Small and medium-sized enterprises (SMEs) often struggle to obtain financing due to the high costs associated with green certification. In China, only 12% of green credit is directed toward SMEs, and significant regional disparities exist—80% of global green investments are concentrated in Europe, North America, and East Asia, while most developing countries face substantial financing gaps in clean energy projects. Third, the short-term profit orientation of financial markets conflicts with the long-term nature of low-carbon transition, making it difficult for renewable energy projects with extended development cycles to secure funding.

To address these market failures in green finance, government regulation serves as an essential complementary mechanism. Marketization level reflects the maturity of regional market mechanisms and embodies the dynamic balance between government regulation and market forces. In regions with higher marketization levels, well-developed institutions, transparent market mechanisms, and efficient resource allocation provide a solid foundation for green finance. In contrast, in regions with lower marketization levels, ineffective institutions, market fragmentation, and information asymmetries often hinder the effectiveness of green finance and limit its positive environmental impact. Thus, the following hypothesis is proposed:

H2. 

Government regulation strengthens the low-carbon effect of green finance by enhancing the level of marketization.

Land use efficiency reflects the intensity of land resource allocation and is a direct outcome of governmental spatial planning and policy constraints. In regions with high urban land use efficiency, the intensive utilization of land resources, rational spatial layout, and well-developed infrastructure provide strong support for the development of green finance. In contrast, in areas where land use efficiency is low, issues such as land resource wastage, inefficient spatial planning, and inadequate infrastructure often hinder the effectiveness of green finance, limiting its potential to contribute meaningfully to environmental improvement. By optimizing the allocation of land resources and promoting intensive development, urban land use efficiency shapes the operational environment of green finance and indirectly enhances the impact of government regulation on environmental outcomes. For instance, the Chengdu–Chongqing Economic Circle has mitigated local protectionism through unified governance standards and incentive mechanisms, facilitating cross-regional flow of production factors and industrial coordination. Meanwhile, Chongqing’s multi-plan integration reform has consolidated spatial resource data—including urban–rural planning, land use, and environmental controls—into a unified One Blueprint platform, enabling highly efficient allocation of regional spatial resources. Thus, the following hypothesis is proposed:

H3. 

Government regulation strengthens the low-carbon effect of green finance by improving land use efficiency.

The heterogeneous impact of green finance on low-carbon transition in resource-based and high-energy-consuming cities stems from systematic differences in institutional embeddedness and factor restructuring costs. Resource-based cities, constrained by the resource curse and state-capital dominance, face diminishing marginal returns on emission reduction and high transition costs—exemplified by Datong’s 37% increase in financing costs for coal capacity replacement funds over five years, with labor resettlement accounting for a significant portion of total transition expenses. These cities require publicly oriented instruments such as government credit enhancement, ecological compensation, and restoration REITs to offset sunk costs and facilitate transition. In contrast, high-energy-consuming cities are often hindered by technological lock-in. For instance, Suzhou employed patent pledges to advance hydrogen-based steelmaking, demonstrating how green finance can break technological barriers and spur innovation through risk-pricing market tools like carbon futures. Ignoring such heterogeneity may lead to policy mismatch, inefficient green investment, and stalled transitions. Hence, tailored green financial instruments are essential to avoid one-size-fits-all policy pitfalls. Accordingly, the following hypothesis is proposed:

H4. 

Green finance exerts heterogeneous effects on the low-carbon transition in resource-based and high-energy-consuming cities.

In summary, green finance leverages diverse instruments—including green credit, green bonds, green funds, green subsidies, green insurance, green equity, and green investment—to establish a funding and resource supply network that underpins urban low-carbon development, thereby providing a continuous source of sustained funding for cities’ low-carbon transition. In this process, government regulation exerts moderating effects through two primary pathways: marketization level and land use efficiency. On one hand, by enhancing marketization, it improves the efficiency of green financial resource allocation and directs capital toward low-carbon sectors. On the other hand, by elevating land use efficiency—such as increasing carbon productivity per unit of land—it strengthens the capacity of green finance to enable low-carbon spatial development. Together, these two regulatory mechanisms shape the heterogeneous characteristics of urban carbon emission pathways and green transition processes, thereby ensuring that green financial instruments effectively and precisely contribute to urban low-carbon development goals. The mechanism is illustrated in Figure 1.

3. Empirical Strategies

3.1. Sample Selection and Data Sources

As of 2022, there were 333 prefecture-level administrative divisions in China. After excluding regions with severe data shortages and special administrative regions, the final sample included 273 prefecture-level cities—covering 82% of all prefecture-level cities nationwide and accounting for 91% of China’s total economic output. This study takes these 273 prefecture-level cities as research objects, with the research period spanning from 2006 to 2022.

The data in this study were mainly sourced from the CSMAR database, Wind database, China Environmental Yearbook, Science and Technology Statistical Yearbook, China Statistical Yearbook, China Science and Technology Statistical Yearbook, and public information released by environmental protection departments. Regarding missing data, the volume was relatively small: linear interpolation was only adopted to supplement data for a few individual indicators, and this approach had little impact on the overall results. After integrating the above data and addressing missing values, we obtained a balanced panel dataset, which contained 4641 observations.

3.2. Modeling

In order to test how green finance impacts low-carbon development, this paper constructs a two-way fixed-effects model of city and time:

$${ci}_{i,t}={\beta }_0+{\beta }_1{gf}_{i,t}+\eta {Control}_{i,t}+{\mu }_t+{\lambda }_i+{\epsilon }_{i,t}$$

(1)

where i, t represents city and year, respectively; ${ci}_{i,t}$ is the explanatory variable, representing regional low-carbon development; ${gf}_{i,t}$ is the core explanatory variable, representing the regional development of green finance; and ${Control}_{i,t}$ is the set of control variables. This paper controls a series of variables that may affect the level of regional energy consumption in the baseline regression. ${\lambda }_i$ is the individual fixed effect, ${\epsilon }_i$ is the random perturbation term, and ${\beta }_0$ is the constant term. To reduce the estimation error of the samples, this paper controls both the regional individual fixed effect and the year fixed effect, and sets the clustering robust standard error to the prefecture-level city level, to ensure that the inference is as strict and error-free as possible.

3.3. Descriptive Statistics

The statistical description of the relevant variables in this paper is shown in

Table 1. Descriptive statistics.

Variables	Obs.	Mean	Median	Std.Dev.	Min.	Max.
ci	4641	4.116	2.896	4.172	0.064	50.22
gf	4641	0.411	0.323	0.824	0.057	11.37
ued	4641	0.304	0.121	0.760	0.001	17.02
edl	4641	10.54	10.58	0.730	4.595	13.06
isl	4641	0.409	0.400	0.102	0.112	0.839
human	4641	0.019	0.010	0.024	0	0.151
fa	4641	0.798	0.745	0.389	0.001	3.907
employment	4641	0.557	0.552	0.150	0.0850	1.661
fd	4641	0.471	0.437	0.227	0.0540	1.541

. The mean value of carbon emission intensity (ci) is 4.116, the median is 2.896, and the standard deviation is 4.172, which also indicates that the difference in carbon emission intensity between the study areas is very significant, with a maximum value of 50.22 and a minimum value of 0.064. The green finance composite index (gf) has a mean value of 0.411 and a standard deviation of 0.824, with maximum and minimum values of 11.37 and 0.057, respectively, which means that the gap between the green finance composite indexes of the study areas is also significant.

3.4. Definition of Variables

(1) Explained Variables. This study adopts carbon intensity (ci) as the measure of low-carbon development, following the approach of F. Zhang et al. (2020). Carbon intensity refers to the amount of carbon dioxide emissions per unit of economic activity, energy use, or production process. It is a key indicator of the efficiency of greenhouse gas emissions generated by a country, region, industry, or enterprise during economic activities or energy consumption, typically expressed in tons of carbon dioxide per unit of production value.

(2) Main Explanatory Variables. Drawing on the research framework of Zhao et al. (2021), this study constructs a comprehensive evaluation index system to measure the development level of green finance based on its core dimensions. The system encompasses seven key aspects: green credit, green investment, green insurance, green bonds, green support, green funds, and green equity. Among these, green credit and green investment serve as the core sources of capital input, with green insurance forming a solid risk prevention and control barrier for both. Green bonds and green credit constitute a complementary structure of direct and indirect financing. Government green financial support effectively stimulates the vitality of market participation through fiscal and tax policy guidance and cost reduction. Green funds pool social capital to invest in green sectors and foster high-quality targets for green equity; green equity, in turn, facilitates the financing and expansion of green enterprises, feeds back into green credit and green investment, and collectively establishes a multi-level green finance system. The specific composition of indicators is detailed in

Table 2. Composition and measurement of the green finance composite index.

Variables	Performance Indicators	Variable Measure
Green Credit	Proportion of nvironmental protection project loans	Total environmental protection project loans in the province/total loans in the province
Green Investment	Proportion of environmental pollution control investment in GDP	Environmental pollution control investment/GDP
Green Insurance	Promotion of environmental pollution liability insurance	Environmental pollution liability insurance premium income/total insurance premium income
Green Bonds	Development of green bonds	Total green bond issuance/total bond issuance
Green Support	Proportion of fiscal environmental protection expenditure	Fiscal environmental protection expenditure/total fiscal general budget expenditure
Green Funds	Proportion of green funds	Total market value of green funds/total market value of all funds
Green Equity	Depth of green equity development	Carbon trading, energy rights trading, and emissions trading/total transaction value of the rights market

. Using the entropy weight method to comprehensively assess this system, a variable representing the level of green finance development (denoted as gf) is derived.

(3) Control Variables. To account for other factors that may influence carbon emission intensity, this study selects a series of control variables: ① Economic Factors: Urban economic density (ued) and economic development level (edl) are chosen as control variables (J. Huang et al., 2024; Xing & Yang, 2022). Urban economic density reflects resource allocation efficiency, whereas higher economic density may enhance carbon emission efficiency and promote low-carbon urban development. The economic development level is selected because technological advancements and improved carbon emission efficiency associated with economic growth typically reduce carbon intensity. ② Industrial Structure Level (isl): The transition from secondary to tertiary industries often improves energy consumption efficiency, thereby reducing energy intensity (Ni et al., 2024). ③ Human Capital Level (human): A well-educated and skilled workforce can utilize energy more efficiently and encourage industries to adopt advanced production technologies and management practices (Umar et al., 2022). ④ Fixed Asset Level (fa): The adoption of modern, energy-saving equipment and technologies can significantly reduce energy consumption and lower carbon emission intensity (Dan et al., 2023). ⑤ Employment Structure (employment): The proportion of employment in the tertiary sector to total employment is used as an indicator. The tertiary sector, typically associated with low-energy-consumption sectors, may reduce energy intensity as its employment share increases (Du et al., 2024). ⑥ Fiscal Decentralization (fd): Higher fiscal decentralization grants local governments greater financial autonomy, enabling them to implement low-carbon policies, promote green infrastructure, and support environmental protection projects, thereby reducing carbon emissions (Kırıkkaleli et al., 2021).

(4) Regulatory variables. Market-oriented reform and land planning, as two parallel paths of government regulation, promote the improvement of land resource allocation efficiency through institutional optimization and spatial governance, respectively. Marketization reform takes the marketization level as the core indicator to reflect the allocation efficiency of green finance under the market mechanism; land planning takes urban land use efficiency as the key indicator to assess the output efficiency of land resources in spatial layout. Therefore, this paper selects marketization level (market) and urban land use efficiency (land) as the regulating variables.

In summary, the variable selection and measurement of this paper’s empirical research are shown in the following

Table 3. Variables and their measurements.

Variables	Variable Name	Variable Symbol	Variable Measure
Green finance	Green Finance Composite Index	gf	Entropy method of measurement
Low-carbon development	Carbon emission level	ci	Tons of carbon dioxide (CO2) per unit of production value
control variable	Urban economic density	ued	GDP/land area of administrative area
	Economic development level	edl	ln (regional GDP per capita)
	Industrial structure level	isl	Tertiary value added/GDP
	Human capital level	human	Number of students in higher education institutions/registered population
	Level of fixed assets	fa	Fixed asset investment/GDP
	Employment structure	employment	Number of persons employed in the tertiary sector/total employment
	Fiscal decentralization	fd	General government revenue/general government expenditure
moderator variable	Marketization level	market	Urban private and self-employed workers/urban employment
	Urban land use efficiency	land	Area of district roads/total district population

.

4. Benchmarking Results and Robustness Analysis

4.1. Benchmark Regression Results

This study employs a two-way fixed-effects model. The regression results, obtained by sequentially adding control variables to the explanatory variables, are presented in

Table 4. Benchmark regression results.

	ci	ci	ci	ci	ci	ci	ci	ci
gf	−0.636	−1.136 ***	−1.161 ***	−1.205 ***	−1.195 ***	−1.244 ***	−1.227 ***	−1.226 ***
	(0.522)	(0.435)	(0.352)	(0.361)	(0.366)	(0.352)	(0.352)	(0.356)
ued		0.691 **	0.542 *	0.576 *	0.571 *	0.518 *	0.509 *	0.510 *
		(0.325)	(0.294)	(0.305)	(0.294)	(0.276)	(0.273)	(0.269)
edl			−2.698 ***	−2.462 ***	−2.455 ***	−2.534 ***	−2.606 ***	−2.531 ***
			(0.459)	(0.457)	(0.456)	(0.477)	(0.491)	(0.500)
is				4.389 **	4.265 **	4.352 **	4.298 **	4.324 **
				(1.808)	(1.800)	(1.772)	(1.768)	(1.768)
human					7.898	7.626	7.452	7.277
					(12.027)	(11.846)	(11.807)	(11.831)
fa						−0.441 *	−0.451 *	−0.436 *
						(0.237)	(0.236)	(0.236)
employment							−0.892	−0.969
							(0.628)	(0.633)
fd								−0.819
								(0.497)
_cons	4.385 ***	4.381 ***	32.905 ***	28.614 ***	28.414 ***	29.599 ***	30.888 ***	30.508 ***
	(0.214)	(0.147)	(4.847)	(5.096)	(5.101)	(5.436)	(5.647)	(5.653)
N	4641	4641	4641	4641	4641	4641	4641	4641
r2	0.881	0.884	0.897	0.899	0.899	0.900	0.900	0.900
r2_a	0.873	0.876	0.891	0.892	0.892	0.893	0.893	0.893

Note: Standard errors in parentheses. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

. The panel regression results reveal that, with carbon emission intensity as the dependent variable, the coefficient of green finance is −1.226, significant at the 1% level. This indicates that, after controlling for relevant variables, green finance exhibits a significant negative correlation with carbon emission intensity. These findings suggest that green finance significantly promotes urban low-carbon development, supporting the validity of Hypothesis H1 proposed in this study.

4.2. Endogeneity Problem

To address potential endogeneity bias in the core regression, this study employs an instrumental variable (IV) approach based on the method proposed by Bartik (1991). To ensure the validity and relevance of this instrument within our specific research context, we draw support from two recent studies that have applied this methodology in related fields. First, Borusyak et al. (2022) constructed a Bartik-type instrument that effectively controls for unobserved local confounders. Second, Goldsmith-Pinkham et al. (2020) demonstrated that the Bartik instrument captures exogenous variation stemming from aggregate shocks, making it suitable for addressing endogeneity concerns.

Following their approach, we construct a Bartik instrumental variable using a shift-share design. The general idea is to interact local initial conditions with subsequent national-level changes in green finance exposure. Specifically, the instrument is constructed by calculating the interaction between the initial share of green finance activities at the city level and the subsequent national growth rate of green finance, thereby generating exogenous variation that reflects regional differential exposure to broad green finance policy or market shocks. The model is specified as follows:

$${\mathit{aver\_gf}}_{i,t}={\displaystyle \frac{1}{N_t}}\sum _{i=1}^{N_t}{gf}_{i,t}$$

(2)

$${{dgf}_{i,t}=\mathit{aver\_gf}}_{i,t}-{\mathit{aver\_gf}}_{i,t-1}$$

(3)

$${{\mathit{gf\_bartik}}_{i,t}=L.gf}_{i,t}-{dgf}_{i,t}$$

(4)

Equation (2) calculates the average value of green finance share, where ${\mathit{aver\_gf}}_{i,t}$ is the average value of green finance share of region i in year t, ${gf}_{i,t}$ is the green finance share of region i in year t, and $N_t$ is the total number of districts in year t. Equation (3) calculates the annual change in the green finance share, where ${dgf}_{i,t}$ is the change in the green finance share of region i in year t. Equation (4) calculates the Bartik instrumental variable, where ${\mathit{gf\_bartik}}_{i,t}$ is the Bartik instrumental variable of green finance.

After constructing the instrumental variable, the two-stage least squares (2SLS) method is employed to address endogeneity issues. The regression results are presented in

Table 5. Endogeneity test regression results.

Variables	(1)	(2)
	First Stage Regression	Second Stage Regression
	gf	ci
gf_bartik	30.46 ***
	(1.624)
gf		−1.086 ***
		(0.355)
controls	YES	YES
city	YES	YES
year	YES	YES
Kleibergen–Paap Wald rk F statistic	351.61
N	4368	4368
r2	0.989	0.187
r2_a	0.988	0.183

Note: Standard errors in parentheses. *** indicate significance at the 1% levels, respectively.

. In the first stage, the instrumental variable (${\mathit{gf\_bartik}}_{i,t}$) shows a significant positive correlation with the endogenous variable (${gf}_{i,t}$) at the 1% level, satisfying the relevance assumption. In the second stage, green finance exhibits a significant negative impact on carbon intensity, consistent with the main regression results and supporting Hypothesis H1. Additionally, the Stock–Yogo weak instrument test yields an F-statistic of 351.61, which exceeds the 10% critical value (16.38), indicating that the instrumental variable is sufficiently strong to effectively correct endogeneity bias.

4.3. Robustness Test

• Replacement of explanatory variables

Compared to carbon emission data, green patents better reflect a city’s activity in green technology development and application, serving as an important supplementary measure of low-carbon development potential and long-term emission reduction capacity. Therefore, we replace the original dependent variable (carbon emission intensity) with a green innovation indicator, namely the number of green patents granted. As shown in Column (1) of

Table 6. Robustness test.

	(1)	(2)	(3)	(4)
	green_palent	ci	ci	ci
gf	1588.422 ***	−2.927 ***	−2.317 ***	−1.135 ***
	(570.387)	(0.689)	(0.354)	(0.351)
controls	YES	YES	YES	YES
city	YES	YES	YES	YES
year	YES	YES	YES	YES
_cons	4477.009 **	31.346 ***		31.227 ***
	(2269.388)	(2.631)		(6.045)
N	4641	4641	4641	4216
r2	0.755	0.936	0.068	0.897
r2_a	0.738	0.931		0.890
rho			0.274 *** (0.017)
Direct Effect			−2.497 *** (0.365)
Indirect Effect			−3.517 *** (0.798)

Note: Standard errors in parentheses. ***, and ** indicate significance at the 1%, and 5% levels, respectively.

, the coefficient of green finance remains significantly positive, indicating that, ceteris paribus, green finance contributes to technological advancement in the urban low-carbon transition. This confirms the robustness of the baseline results.

2.

Sample adjustment

To mitigate the potential influence of outliers on the estimates, we winsorize the urban carbon emission variable at the 1% level on both tails. After this adjustment, as presented in Column (2) of Table 6, the regression coefficient of green finance remains statistically significant, and its negative effect is consistent with the baseline findings. This suggests that the results are not driven by extreme observations.

3.

Spatial econometric approach

Given the notable spatial interdependence characteristic of urban low-carbon development, a spatial econometric method is employed for robustness testing. A Spatial Durbin Model (SDM) based on a geographic adjacency matrix is constructed to capture potential spatial spillover effects of green finance policies. Spatial autocorrelation tests indicate that both carbon emission intensity and green finance exhibit significant spatial spillover effects at the 10% level. The Hausman test supports the use of fixed effects, and a likelihood-ratio (LR) test (chi²(8) = 189.85, p = 0.0000) confirms that the SDM specification is preferred over both the Spatial Autoregressive (SAR) and Spatial Error (SEM) models. As shown in Column (3) of Table 6, green finance continues to exert a significant promoting effect on urban low-carbon development after accounting for spatial dependence, confirming the robustness of the baseline results.

4.

Excluding provincial capital cities

Provincial capitals, as administrative centers, often receive more concentrated green finance policy resources—such as provincial-level green credit discounts and carbon trading pilots—leading to systematic differences in policy intensity compared to non-capital cities. Such institutional advantages may overestimate the carbon reduction effects of green finance. To mitigate this potential bias, we exclude data from 25 provincial capitals (e.g., Shijiazhuang, Nanjing, Guangzhou). Regression results presented in Column (4) of Table 6 show that the emission reduction effect of green finance remains statistically significant even after excluding provincial capitals, further supporting the robustness of the baseline findings.

5. Analysis of Mechanisms

5.1. Regulatory Effects of Green Finance Under Market-Oriented Reforms

The market-oriented reform of the green financial system carries the important responsibility for promoting the marketization of interest rates and the competition of financial institutions, whose main method is to make the price of funds reflect market supply and demand more realistically. Such reforms enhance the resilience and transparency of the green financial system, as well as the openness and diversity of the financial market. They also encourage competition between traditional and emerging financial institutions, reduce the cost of financial services, and improve service quality. Moreover, through signaling effects, enterprises and individuals can rationally allocate funds according to market signals, thereby reducing the risk of resource mismatch. Therefore, this paper argues that the low-carbon effect of green finance may be strengthened by the level of marketization, and establishes the moderating effect model as follows by adding the cross-multiplier term between the level of marketization and green finance in the benchmark model:

$${ci}_{i,t}={\beta }_0+{\beta }_1{gf}_{i,t}+{\beta }_2{gf}_{i,t}\times {market}_{i,t}+{\beta }_3{market}_{i,t}+\eta {Control}_{i,t}+{\mu }_i+{\lambda }_i+{\epsilon }_{i,t}$$

(5)

The test results are shown in Column (1) of

Table 7. Mechanism test results.

	(1)	(2)
	ci	ci
gf	−1.209 ***	−0.694 *
	(0.352)	(0.359)
market	0.002 ***
	(0.001)
gf × market	−0.005 ***
	(0.002)
land		0.007
		(0.010)
gf × land		−0.042 **
		(0.016)
controls	YES	YES
city	YES	YES
year	YES	YES
_cons	30.512 ***	29.539 ***
	(5.654)	(5.477)
N	4641	4607
r2	0.900	0.920
r2_a	0.893	0.915

Note: Standard errors in parentheses. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

. It can be seen that the coefficient of the interaction term (gf × market) is −0.005 and is significant at the 1% level, indicating that the increase in the level of marketization significantly strengthens the positive impact of green finance on urban low-carbon development. With the increase in the level of marketization, the impact of green finance on urban low-carbon development is gradually enhanced. Therefore, the level of marketization proposed in this paper plays a moderating role in the impact of green finance on urban low-carbon development, and Hypothesis H3 holds.

5.2. Moderating Effect of Urban Land Planning on Green Finance

Urban land use efficiency is one of the important driving forces promoting China’s high-quality development. By enhancing land use efficiency, cities can divide functional areas more scientifically, optimize the distribution of resources among different regions and industries, promote the flow of land resources to high-value-added and high-technology industries, and help upgrade the industrial structure and transform the economic growth model. This raises the question of whether its development has an obvious spillover effect on the development of green finance so as to promote urban low-carbon development. Therefore, this paper adds the cross-multiplier term of urban land use efficiency and green finance in the benchmark model and establishes the regulation effect model as follows:

$${ci}_{i,t}={\beta }_0+{\beta }_1{gf}_{i,t}+{\beta }_2{gf}_{i,t}\times {land}_{i,t}+{\beta }_3{land}_{i,t}+\eta {Control}_{i,t}+{\mu }_i+{\lambda }_i+{\epsilon }_{i,t}$$

(6)

In Equation (6), ${land}_{i,t}$ is the urban land use efficiency, and ${gf}_{i,t}\times {land}_{i,t}$ is the interaction term between green finance and urban land use efficiency. The test results are shown in Column (2) of Table 7. It can be seen that urban land use efficiency plays an important moderating role in green finance and low-carbon development. The coefficient of the interaction term (gf × land) is −0.042 and is significant at the 5% level, indicating that higher land use efficiency promotes the energy-saving and emission-reduction effect of green finance. The efficient use of urban land helps reduce the ecological damage caused by urban expansion, reduces the intensity of carbon emissions, and promotes the transformation of the land use pattern toward a green and sustainable direction. Therefore, the urban land use efficiency proposed in this paper should play a moderating role in the impact of green finance on urban low-carbon development, and Hypothesis H3 holds.

6. Heterogeneity Tests

6.1. Heterogeneity Analysis Based on Green Financial Instruments

Different green financial instruments vary in their operational mechanisms and resource allocation pathways, leading to divergent impacts on urban low-carbon transition. Green credit incentivizes corporate transformation through financing constraints, green investment accelerates the adoption of low-carbon technologies via capital injection, green insurance reduces project uncertainty with risk guarantees, while green bonds and funds aggregate social capital to broaden financing channels. Green support and equity, on the other hand, contribute to building a low-carbon ecosystem through policy coordination and incentive mechanisms.

Empirical results, as shown in

Table 8. Heterogeneity analysis of green financial instruments.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	ci	ci	ci	ci	ci	ci	ci
gf_loan	−6.510 ***
	(1.769)
gf_invest		−21.090 ***
		(7.099)
gf_insurance			−13.869 ***
			(3.906)
gf_bond				−11.411
				(10.999)
gf_support					−20.618 ***
					(6.522)
gf_fund						−7.039 ***
						(1.472)
gf_equity							−5.686 **
							(2.446)
_cons	30.281 ***	30.191 ***	30.282 ***	29.782 ***	29.911 ***	30.214 ***	29.909 ***
	(2.603)	(2.596)	(2.605)	(2.569)	(2.578)	(2.582)	(2.578)
controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
city	Yes	Yes	Yes	Yes	Yes	Yes	Yes
year	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	4641	4641	4641	4641	4641	4641	4641
r2	0.935	0.935	0.935	0.935	0.935	0.935	0.935
r2_a	0.931	0.931	0.931	0.930	0.931	0.931	0.930

Note: Standard errors in parentheses. ***, and ** indicate significance at the 1%, and 5% levels, respectively.

, indicate that green credit (gf_loan) has a significantly negative coefficient (−6.510), suggesting that credit constraints may pressure high-carbon enterprises to transition. Green investment (gf_invest, coefficient = −21.090) and green insurance (gf_insurance, coefficient = −13.869) both show statistically significant effects at the 1% level, substantially promoting urban low-carbon transition—the former through targeted capital allocation and the latter by enhancing project feasibility with risk coverage. Green support (gf_support, −20.618), green funds (gf_fund, −7.039), and green equity (gf_equity, −5.686) are also statistically significant, indicating that policy coordination, professional capital management, and equity incentives facilitate transition through resource integration and long-term motivation. Notably, green bonds (gf_bond) show a coefficient of –11.411 but are not statistically significant at conventional levels. This may be because some raised funds were not strictly directed toward low-carbon projects, reflecting “greenwashing” tendencies within green finance supply. This finding aligns with Sinha et al. (2021), who reported that green securities could have neutral to negative environmental and social impacts over the medium to long term.

6.2. Heterogeneity Analysis of Resource Endowment

Following Zheng and Niu’s (2023) methodology, this study classifies cities into resource-based and non-resource-based categories according to the National Sustainable Development Plan for Resource Cities (2013–2020). Group regression analyses reveal significantly negative coefficients for green finance’s impact on both city types at the 1% level (

Table 9. Heterogeneity analysis based on resource endowment and energy consumption.

	Resource-Based	Non-Resource-Based	Energy-Efficient c	Energy-Inefficient
	ci	ci	ci	ci
gf	−9.876 ***	−1.189 ***	−6.494 ***	−1.735 **
	(2.533)	(0.317)	(0.749)	(0.332)
controls	YES	YES	YES	YES
city	YES	YES	YES	YES
year	YES	YES	YES	YES
_cons	34.349 ***	33.047 ***	34.202 ***	16.166 **
	(9.831)	(5.851)	(3.064)	(7.564)
N	1870	2771	3876	765
r2	0.931	0.871	0.934	0.889
r2_a	0.926	0.862	0.929	0.878
Fisher test p-value	0.000		0.000

Note: Standard errors in parentheses. ***, and ** indicate significance at the 1%, and 5% levels, respectively.

). The Fisher test (p = 0.000) confirms significant inter-group differences. Resource-based cities demonstrate a stronger coefficient (−9.876) compared to non-resource-based cities (−1.189), indicating green finance’s greater impact on low-carbon development in resource-dependent areas. This disparity stems from resource-based cities’ reliance on natural resource extraction, characterized by industrial structures and high emissions from traditional industries. These cities face substantial challenges in low-carbon transition, requiring substantial financial and technical support. Green finance facilitates this transition through instruments like green loans and support, promoting green technology adoption and industrial transformation. Moreover, it addresses resource-based enterprises’ limited innovation capacity and motivation for green transformation. In contrast, non-resource-based cities’ diversified economic structures, with established service and high-tech industries, exhibit stronger endogenous innovation capacity and relatively lower dependence on green finance support.

6.3. Heterogeneity Analysis Based on Energy Consumption Levels

To further investigate the heterogeneous effects of green finance across cities with different energy consumption levels, this study employs a systematic clustering method to classify the sample into distinct groups. A heterogeneity test was subsequently conducted for each group. The results, presented in Table 9, show that green finance has a significantly negative impact on both groups. The empirical p-value of Fisher’s permutation test is 0.000, indicating significant differences in variable distributions between the two groups.

More specifically, the marginal effect of green finance is more pronounced in the low-energy-consumption city group, with a coefficient of −6.494, compared to −1.735 in the high-energy-consumption group. This suggests that green finance plays a stronger role in reinforcing and expanding existing low-carbon ecosystems in cities with already low energy usage. Through diversified instruments such as green investment and green equity, it helps accelerate a “multiplier effect” in low-carbon development. Conversely, in high-energy-consumption cities, industries often exhibit deep reliance on fossil fuels and face strong technological lock-in effects. As a result, the transition driven by green finance—such as the adoption of low-carbon technologies—encounters higher barriers and may yield less visible outcomes in the short term.

6.4. Heterogeneity Analysis Based on Geographical Regions

Following the classification standards of the National Bureau of Statistics, this study divides the sample into eastern, central, and western regions to examine the regional heterogeneity in the impact of green finance on low-carbon development. The results are presented in

Table 10. Heterogeneity analysis based on geographical regions and temporal dimensions.

	East	Central	West	2006–2011	2012–2022
	ci	ci	ci	ci	ci
gf	−0.593 ***	−3.701 **	−2.597 ***	−2.773 ***	−0.866 **
	(0.147)	(1.501)	(0.772)	(0.784)	(0.399)
controls	YES	YES	YES	YES	YES
city	YES	YES	YES	YES	YES
year	YES	YES	YES	YES	YES
_cons	25.680 ***	28.049 ***	29.043 **	40.64 ***	19.78 ***
	(5.587)	(4.458)	(12.833)	(5.414)	(1.701)
N	1683.000	1666.000	1292.000	1638	3003
r2	0.919	0.860	0.921	0.972	0.969
r2_a	0.913	0.849	0.914	0.966	0.966
Chow test p-value	0.000
Fisher Test p-value				0.030

Note: Standard errors in parentheses. ***, and ** indicate significance at the 1%, and 5% levels, respectively.

. A Chow test was conducted to assess the structural stability of the model across the regional subgroups. The empirical p-value of 0.000 indicates significant structural differences between the groups. The results show that green finance significantly reduces carbon emission intensity across all regions, though with notable variations in effect size. The estimated coefficient for the eastern region is −0.593, significant at the 1% level; for the central region, it is −3.701, significant at the 5% level; and for the western region, it is −2.597, also significant at the 1% level. These findings indicate that green finance effectively promotes low-carbon development in all three regions, with the strongest effect observed in the central region, followed by the western and eastern regions, reflecting distinct regional heterogeneity.

6.5. Heterogeneity Analysis Based on Time Periods

Given the distinct phase-specific characteristics in policy orientation and practical models during the development of green finance, this study conducts a temporal heterogeneity analysis. The year 2012 is selected as the cutoff point, as the China Banking Regulatory Commission (CBRC) issued the Green Credit Guidelines in that year, which explicitly required banks to support projects meeting environmental standards—particularly in clean energy, energy conservation, emission reduction, and pollution control. The guidelines also mandated environmental impact assessments of borrowing enterprises to prevent funding from flowing to projects that fail to comply with environmental requirements. As a result, 2012 is widely regarded as a milestone year marking the formal inception of green finance development in China. The results of the temporal heterogeneity analysis are presented in Table 10. Fisher’s permutation test yields an empirical p-value of 0.030, indicating a statistically significant difference in variable distributions between the pre- and post-2012 periods. The decrease in coefficient values reflects the phase-specific characteristic of diminishing marginal effects of green finance on carbon reduction, representing a transition from “easier carbon reduction” to “harder carbon reduction” stages.

7. Conclusions and Policy Recommendations

Based on the panel data of 273 prefecture-level cities in China from 2006 to 2022, this paper empirically examines the effect and mechanism of green finance on the development of urban low-carbon transition from the perspective of government regulation. It is found that, first, green finance significantly reduces urban carbon emission intensity by guiding capital allocation and technological innovation, and has a robust promotion effect on urban low-carbon transition; second, government regulation strengthens the low-carbon effect of green finance through dual paths—on the one hand, it enhances green finance efficiency by strengthening the city’s environmental pricing mechanism through perfecting market-oriented reforms, and on the other hand, it enhances green finance efficiency through urban land planning to optimize the land use pattern and enhance spatial carrying capacity; third, the heterogeneity analysis shows that resource-based cities and low-energy-consuming cities benefit more significantly from low-carbon transformation, reflecting differences in resource constraints.

Based on the above findings, in order to effectively exert the positive impact of green finance on urban low-carbon development, this paper puts forward the following three suggestions:

First, developing green finance in light of local conditions is a key priority. Resource-based cities should innovate tools such as mine ecological restoration loans and successor industry funds, link resource tax rebates to green credit quotas, and drive industrial restructuring. High-energy-consuming cities need to establish low-carbon technology whitelists and tiered financing incentives and incorporate the rate of reduction in energy intensity into performance evaluations. Additionally, they should develop a city carbon transition financial efficiency index, integrate satellite remote sensing and carbon account data to dynamically assess policy effects, and set a five-year flexible window period—with financial guarantees providing support in the early stage and a shift to market-oriented risk-sharing in the later stage. This approach avoids policy rigidity, promotes the R&D transformation, large-scale application, and iterative upgrading of low-carbon technologies, facilitates the green transformation of high-energy-consuming industries and the optimization of energy consumption structures, and ultimately achieves the goal of low-carbon transition.

From a regional perspective, green financial resources should be strategically directed toward central and western China: Central China’s resource-based cities should prioritize the deployment of green bonds and carbon quota pledge loans, with supporting financial interest subsidies and risk compensation, focusing on supporting low-carbon technological transformation of traditional industries. High-energy-consuming cities in Western China should be provided with targeted green refinancing and energy-consumption-linked interest rate tools to strengthen clean energy substitution and energy efficiency improvement. Eastern China should guide capital to feed back into central and western regions through cross-regional carbon account linkage mechanisms.

Second, establish a coordinated mechanism for market incentives and spatial governance. At the economic level, centered on expanding the carbon emission rights trading market, we should refine the environmental risk pricing mechanism, strengthen the fair competition review system, eliminate monopoly barriers, and improve market resource allocation efficiency. At the spatial level, deeply integrate green financial instruments with land use efficiency: design products such as special bonds for low-carbon land use to advance the green renewal of inefficient land, and concurrently establish a regulatory system to optimize mixed land use and the layout of renewable energy infrastructure. At the regional level, eastern China should prioritize integrated innovation of market mechanisms and spatial planning, while central and western China enhance the value conversion of ecological space.

Third, consolidate the coordinated support system for green finance, market-oriented reform, and land use efficiency. Centering on market-oriented reform, improve supporting mechanisms: establish a unified environmental information disclosure platform and incorporate land use efficiency into the scope of information disclosure; meanwhile, optimize the green finance pricing mechanism, integrate the benefits of low-carbon land use into the risk pricing model, and strengthen the guiding role of market-oriented reform in the allocation of green financial resources. Promote the alignment of domestic low-carbon land use standards with international green financial rules, attract international green capital to participate in local low-carbon land development and infrastructure construction, form a new green finance development ecosystem, and provide a replicable path for similar developing countries to balance resource constraints and low-carbon transition.

This study attempts to provide new empirical evidence for the role of green finance in enhancing the low-carbon effect of cities from the perspective of government regulation, but it still has certain limitations. Among these, the issue of endogeneity is the core challenge that this study has not fully resolved. Although this study has mitigated potential endogeneity problems by constructing Bartik instrumental variables, and a series of robustness test results show that the baseline results remain robust, achieving more rigorous causal inference still requires further control of key confounding variables that may lead to estimation biases—such as the implementation of new energy policies and technology spillover effects—in future research, so as to enhance the generalizability and reliability of the research conclusions. Therefore, a cautious attitude should be maintained when interpreting the heterogeneous causal relationship between green finance and the low-carbon effect of cities under government regulation. Nevertheless, the main conclusions of this study still hold significant reference value for green finance practice.