Green Finance Policy and Green Development Resilience: Evidence From a Quasi-Natural Experiment

Abstract

Amid frequent geopolitical and trade conflicts, countries are striving to strike a new balance between green development and resilience building. Although many nations have adopted green finance policies (GFP) to promote green development, it remains uncertain whether such development possesses the resilience needed to withstand external shocks. Based on the establishment of China’s Green Finance Reform and Innovation Pilot Zones, this study employs a difference-in-differences approach to examine the impact of GFP on green development resilience (GDR). The results show that: (a) GFP significantly enhances GDR. (b) This effect of GFP on GDR operates primarily through two channels: easing firms’ financing constraints and stimulating green technological innovation. (c) The positive impact is stronger in regions with stricter environmental regulation, greater trade openness, and more segmented markets. These findings provide valuable policy implications for optimizing green finance frameworks, strengthening GDR, and advancing long-term sustainable development.

Plain Language Summary

Studying Whether National Policies That Direct Financial Resources Toward Green Industries Can Improve a Region’s Green Development Ability to Cope With External Shocks Remains a Crucial Area of Research.

While the immediate goal of green development is important, building resilience against external shocks is even more crucial. However, there is limited information on how national green finance policies impact resilience, especially when it comes to balancing economic growth, social progress, and environmental protection. Understanding this resilience is vital for achieving sustainable development goals. We analyze China’s Green Finance Reform and Innovation Pilot Zones policy as a case study, using data from 271 cities across China. We also measure green development using several indicators and assess resilience by comparing current green development to its initial stage. The results show that green finance policies increased resilience by 14.2% on average, with areas that have stricter environmental regulations or higher openness seeing resilience improvements of 14.9%. In regions with limited movement of capital, labor, or technology, the improvement was even greater, reaching 18.2%. This increase in resilience is largely driven by green finance policies that promote innovation and ease financing challenges. The findings highlight pathways to boost green development resilience. This provides strong support for policies that direct financial resources to green industries.

Introduction

Amid growing climate challenges and expanding human activities, green development has become a globally recognized objective. However, recent global shocks—most notably the Russia–Ukraine conflict and the COVID-19 pandemic—have revealed vulnerabilities accumulated over the past three decades (Bandyopadhyay, 2022). These crises revealed the limited resilience of many economies to unexpected disruptions. In response, international agendas such as SDGs 9 and 11 of the United Nations emphasize the urgent need to strengthen resilience against environmental and social shocks (Caprotti et al., 2017). Against this backdrop, enhancing green development resilience (GDR) has become a central concern for achieving global sustainability.

GDR refers to an economy’s ability to absorb external shocks, restructure internally to maintain functionality, and adapt to environmental changes when faced with uncertainties such as climate change, energy crises, or economic volatility (Holling, 1973). In theory, green finance policy (GFP) directs financial resources toward green technologies and renewable energy (Jiang et al., 2025). By fostering green technological innovation and environmental protection (Saeed Meo & Karim, 2022), GFP has the potential to strengthen GDR. In practice, several countries have launched targeted green finance initiatives. In 2004, Japan’s Ministry of Finance introduced an environmental rating-based loan program that provided subsidies based on environmental performance (Steffen, 2021). In 2005, the U.S. Department of the Treasury supported the Property Assessed Clean Energy program, offering subsidies for green lending products (United Nations Environment Programme, 2019). In 2018, the European Union introduced the Action Plan on Financing Sustainable Growth, marking a shift toward a coordinated public green finance strategy (United Nations Environment Programme Finance Initiative, 2020). China has also taken significant steps to integrate green finance into its sustainability agenda. In 2017, five provinces became the first green finance reform and innovation pilot zones (GFRIPZs), with comprehensive policies introduced covering green credit, green bonds, tax incentives, financial regulation, and green technology. These initiatives place China at the forefront of global green finance experimentation, offering one of the most comprehensive and coordinated policy frameworks in the developing world. Although many countries have adopted green finance strategies to support green development, it remains unclear whether such green development is resilient to external risks. Therefore, assessing the impact of GFP on GDR is essential for evaluating their effectiveness and informing future strategies for sustainable development.

Relevant literature falls into two strands: the economic effects of GFP and the concept, measurement, and determinants of GDR. Regarding the economic effects of GFP, the majority of studies highlight its positive role. Empirical evidence shows that green finance significantly reduces global energy security risks (Vo et al., 2025) and contributes to improvements in energy efficiency (Ullah et al., 2025). Alam and Ullah (2025) find that green finance accelerates the development and adoption of green innovation across 57 developing countries. Similarly, Afolabi (2025) shows that green finance mitigates the negative impact of uncertainty on sustainable development in West African economies. Evidence from four ASEAN countries suggests that green finance reduces the adverse effects of climate risk on financial stability, thereby enhancing financial resilience (Nguyen, 2025). In the context of China, existing research indicates that GFP helps reduce carbon and pollutant emissions from manufacturing enterprises (Guo et al., 2025) and strengthens multiple dimensions of resilience, including energy resilience (Lv & Guo, 2025), ecological resilience (Y. Li & He, 2025), and supply chain resilience (Ruan et al., 2025). However, some studies argue that financial development may increase the ecological footprint (Baloch et al., 2019), degrade environmental quality (Ahmad et al., 2022), and exert a negative impact on ecological resilience (Lee et al., 2024).

Meanwhile, research on the concept, measurement, and determinants of GDR has grown rapidly. Scholars generally view green development as the coordination between economic performance and ecological health (A. Hu, 2017). Resilience is understood not merely as survival capacity but as a proactive strategy to transform through crises and sustain long-term prosperity (Folke et al., 2010; Pendall et al., 2010). Accordingly, GDR refers to the ability of a green development system to withstand shocks, absorb disturbances, and adapt through transformation (H. Hu et al., 2025; X. Xu & Yang, 2025). In measurement, most studies adopt the United Nations’ framework of economic, social, and environmental subsystems. They construct composite indices that integrate indicators such as per capita GDP and PM2.5 concentrations (Sueyoshi, 2017). Although the dynamic nature of resilience is widely acknowledged, its measurement still relies primarily on static composite indices (Rizzi et al., 2018). For instance, Shamsipour et al. (2024) evaluated ecological resilience using 25 indicators capturing environmental gains and losses. A few studies use income sensitivity metrics to measure tourism economic resilience (Tang, 2024), or apply changes in output relative to a baseline period to assess economic resilience (Chao & Xue, 2023). Regarding the determinants of GDR, green technological innovation (Johnstone et al., 2010), environmental regulation (Wu et al., 2023), and trade openness (Eichengreen et al., 2024) are widely recognized as key drivers. Recent studies increasingly focus on the role of green finance in promoting green development and resilience. Findings suggest that green finance channels capital toward low-carbon industries (Muganyi et al., 2021) and improve the flexibility of resource allocation (S. Liu & Wang, 2023). Green credit and bonds have been shown to effectively incentivize green innovation (Z. Li et al., 2018) and support pollution control efforts (Jin & Zhang, 2024). However, whether green finance can strengthen the risk resistance and adaptive capacity of green development remains uncertain.

In summary, although existing studies have extensively explored green development and resilience separately, several important gaps remain. First, most research focuses on single dimensions such as economic or ecological resilience, with limited attention to GDR, which simultaneously integrates economic, social, and environmental systems. Unlike green development, which emphasizes the balance between economic growth, social progress, and environmental cost, GDR reflects a system’s capability to recover and stabilize in response to external shocks (H. Hu et al., 2025). As such, it holds broader implications for achieving the SDGs. Second, although scholars widely acknowledge the complexity and dynamic nature of resilience, most measurements rely on static composite indices or single dynamic indicators. Few studies provide dynamic assessments that capture the multidimensional nature of resilience. Third, while GFP has been shown to facilitate green innovation and direct capital flows, and while technological and financial innovations are recognized as key drivers of resilience, systematic analysis linking GFP and GDR within a unified analytical framework remains limited. Moreover, empirical evidence on this relationship remains scarce.

In light of these gaps, this study develops a dynamic and multidimensional indicator of GDR and empirically assesses the effect of GFP through a quasi-natural experiment involving China’s GFRIPZs. It also examines the underlying mechanisms and regional heterogeneity of this effect. This study offers several key contributions. First, it conceptualizes GDR as a form of adaptive capacity within green development and develops a new method for its measurement. Specifically, it constructs an evaluation framework based on economic, social, and environmental subsystems, and quantifies GDR by measuring the deviation between current and baseline levels of green development. This approach addresses methodological limitations in existing studies that rely on static or single-dimensional resilience metrics. Second, the study offers a comprehensive analysis of how GFP influences GDR and outlines the theoretical pathways for strengthening GDR. The findings offer valuable insights for refining GFP and promoting higher resilience across cities. Third, the study identifies heterogeneous effects of green finance on GDR across regions, offering evidence-based insights for designing regionally tailored green finance policies.

Policy Background and Theoretical Mechanism

Policy Background

In 2017, the Chinese government designated GFRIPZs in eight regions across five provinces: Guizhou, Guangdong, Jiangxi, Xinjiang, and Zhejiang. The initiative aimed to advance GFP as a tool for supporting economic growth, promoting environmental protection, and enhancing ecological balance (S. Zhang & Cheung, 2025). The GFRIPZs encouraged financial institutions to develop green financial products that meet the financing needs of green industries and environmental projects (Shi et al., 2022; Tu et al., 2021). They also strengthened the regulatory framework for green finance by improving oversight of relevant instruments (Springer et al., 2019). Additional measures involved implementing tax incentives to promote the issuance and investment in green financial products (S. Zhang et al., 2021), improving information platforms to reduce information asymmetries in green finance markets, and supporting financial technology innovation to accelerate green finance development (S. Liu & Wang, 2023).

GFRIPZs initiative integrates a comprehensive set of policies, including innovation in green financial products, regulatory oversight, tax incentives, information services, and technological development. Spanning eastern, central, and western regions, GFRIPZs offer a representative setting for examining major green finance policies in China. Therefore, this study employs GFRIPZs as a policy proxy for GFP to examine their impact on GDR and explore the underlying mechanisms.

Theoretical Mechanism

The resource-based view posits that organizational resources play a decisive role in shaping sustainable performance and building competitive capabilities (Barney, 1991). The development of GDR similarly depends on access to critical resources, particularly financial capital. However, traditional finance tends to prioritize profits, exhibit a bias in credit allocation, and misallocate capital (Yang et al., 2024). These characteristics, when coupled with the externalities of pollution and pervasive information asymmetries, often lead to a reluctance to support green investments or resilience-building initiatives (Gong et al., 2023). In contrast, green finance, by design, internalizes environmental externalities through regulatory mechanisms (Lan et al., 2025). It restricts funding for highly polluting and resource-intensive industries or activities by increasing their financing costs (Shahbaz et al., 2021), while simultaneously offering low-interest capital and tax incentives to support practices that are environmentally friendly and resource-efficient (Azhgaliyeva & Liddle, 2020). This helps mitigate externalities by incentivizing firms to reduce carbon and pollutant emissions (Guo et al., 2025). These mechanisms stimulate green investment and enhance GDR.

Moreover, the establishment of GFRIPZs fosters technological innovation in green finance and supports green financial information platforms. These efforts reduce information asymmetries in green lending markets, lower the costs of project identification and risk management for financial institutions, and enhance the efficiency of green capital allocation. Efficient capital utilization, in turn, enhances the stability of green development systems and strengthens their capability to recover from and adapt to adverse shocks. Accordingly, the following hypothesis is formulated:

• Hypothesis 1: GFP enhances GDR.

Nonlinear stability theory conceptualizes resilience as a dynamic characteristic of a particular system or a set of interrelated systems (Walker et al., 2004). A resilient system exhibits the capacity to absorb shocks, self-repair, and re-establish functional stability, which is essential not only for economic recovery but also for maintaining long-run ecological and social equilibrium (Eichengreen et al., 2024; H. Hu et al., 2025). However, the timing and nature of future shocks remain highly uncertain. Addressing such uncertainty requires not merely reactive adjustments but a systematic strengthening of the system’s adaptive capacities, resource allocation efficiency, and innovation potential. From an economic perspective, the ability of firms and regions to respond effectively to uncertainty is strongly conditioned by the availability of financial resources and the technological capabilities embedded within the production structure. GFP plays a foundational role in shaping both dimensions.

The theoretical mechanism linking GFP to GDR can be understood through the lens of corporate finance and investment under uncertainty. Limited external financing—arising from information asymmetry, high collateral requirements, or risk mispricing—remains a major barrier to firms’ green investment decisions. Classic models of capital market imperfections indicate that credit rationing disproportionately suppresses long-term and high-risk projects (Stiglitz & Weiss, 1981), including green transformation and resilience-enhancing investments. GFP directly alleviates these frictions through multiple channels. First, policy-oriented green financial instruments reduce adverse selection by improving the visibility and verifiability of firms’ environmental performance, thereby lowering monitoring costs for financial intermediaries. Second, GFP encourages banks to reallocate capital toward environmentally beneficial sectors, reducing financing costs for firms engaged in green upgrading. Third, by crowding in private capital through public–private risk-sharing arrangements, GFP broadens the financial base available for green transformation projects. As firms’ financing constraints ease, they become capable of undertaking strategic investments that enhance adaptive capacity—such as upgrading to cleaner production technologies, adopting resource-efficient processes, or strengthening environmental management systems. These investments improve both economic efficiency and environmental performance, increasing GDR. Hence:

• Hypothesis 2: GFP enhances GDR by easing firms’ financing constraints.

Resilience fundamentally reflects a system’s capacity, which is closely linked to technological capabilities (Lan et al., 2025). Endogenous growth theory posits that technological progress expands the production possibility frontier and supports long-term sustainable development (Crafts, 1996). In the green development context, green innovation generates cleaner inputs, more efficient production methods, and products with reduced environmental externalities—each of which contributes to a system’s ability to adapt to environmental and economic shocks. The establishment of GFRIPZs obtains this mechanism through targeted financial incentives and regulatory pressures. First, green credit and investment screening standards compel financial institutions to incorporate environmental risk into credit assessments (Necib & Gmati, 2024), effectively raising the marginal benefit of technological innovation for firms seeking external financing. Second, risk-sharing instruments—such as environmental liability insurance—reduce the downside risks of engaging in green R&D, particularly for smaller firms that face substantial uncertainty and long payback periods (Guo et al., 2025). Third, green bonds and other innovative financing vehicles provide long-term funding crucial for scaling green technologies. These financial and regulatory changes reshape firms’ innovation calculus. By lowering the expected cost and raising the expected return of green innovation, GFP encourages firms to develop more efficient, adaptive, and environmentally friendly technologies. Green innovations expand the range of feasible, sustainable production combinations, enhance resource-use efficiency, and allow economies to pivot more flexibly in response to shocks (Aydin et al., 2025). Ultimately, enhanced technological capability translates into stronger GDR.

• Hypothesis 3: GFP enhances GDR by stimulating green technological innovation.

Empirical Design

Model Setting

To assess the impact of GFP on GDR, this study treats the establishment of GFRIPZs as a quasi-natural experiment and empirically tests its effect with a difference-in-differences (DID) model. The baseline specification is as follows:

$$\begin{array}{c}\multicolumn{1}{c}{{\mathrm{GDR}}_{\mathrm{it}}={\alpha }_0+{\beta }_0{\mathrm{treated}}_i\times {\mathrm{time}}_t+{\phi }_0{\mathrm{Control}}_{\mathrm{it}}}\\ \multicolumn{1}{c}{+{\mu }_i+{\lambda }_t+{\epsilon }_{\mathrm{it}}}\end{array}$$

(1)

In the equation, ${\mathrm{GDR}}_{\mathrm{it}}$ is the green development resilience of city i in year t. ${\mathrm{treated}}_i$ is a treatment indicator that equals 1 if city i is included in the GFRIPZs and 0 otherwise. ${\mathrm{time}}_{\mathrm{t}}$ is a time dummy that equals 1 for years after 2017 and 0 for earlier years. ${\mathrm{Control}}_{\mathrm{it}}$ denotes a set of control variables. ${\mu }_i$ and ${\lambda }_t$ capture city and year fixed effects, respectively. ${\epsilon }_{\mathrm{it}}$ is the error term.

Variable Selection and Descriptions

Dependent Variable

According to Martin (2011), GDR refers to the capacity of a green development system to withstand external shocks, self-adjust and recover after disruption. Thus, GDR reflects both the decline and rebound of the system in response to disturbances. The 2008 global financial crisis is widely regarded as one of the most severe and prolonged economic crises in recent history (Bordo & Haubrich, 2017). The crisis led to a sharp decline in economic growth, severely constrained technological innovation (Brem et al., 2020), and caused a dramatic rebound in global carbon intensity (Wang et al., 2021), thereby having profound and lasting effects on green development. As a result, 2008 is taken as the baseline year, with the annual deviation of each city’s green development level from this baseline representing its GDR. The specific calculation is as follows (Chao & Xue, 2023):

$${\mathrm{GDR}}_{\mathrm{it}}=(E_{\mathrm{it}}-E_{i2008})/E_{i2008}$$

(2)

Here, $E_{\mathrm{it}}$ denotes the green development level, and $E_{i2008}$ is the baseline level in 2008. ${\mathrm{GDR}}_{\mathrm{it}}$ indicates green development resilience. To compute the green development level, an evaluation system is constructed based on three dimensions: economic development, social progress, and environmental protection (Table 1). The entropy weight method determines the weights of indicators, which are then aggregated into a composite green development index.

Table 1.

Indicator Framework for Measuring Green Development.

Dimension	Indicator	Direction	Weight
Economic development	GDP per capita (100,000 yuan)	+	0.0528
	Local budget revenue per capita (yuan)	+	0.0750
	Disposable income per capita (yuan)	+	0.0167
	Share of tertiary industry in GDP (%)	+	0.0071
	Share of science and technology expenditures in general budget (%)	+	0.0423
	Total trade per capita (yuan)	+	0.1331
Social development	Per capita education expenditure (yuan)	+	0.0319
	Physicians per 1,000 population	+	0.0224
	Urban workers enrolled in basic medical care per 100 people	+	0.0450
	Public libraries holdings per 100 people (volumes/100 persons)	+	0.0836
	Public vehicles per 10,000 people (vehicles/10,000 persons)	+	0.0636
	Urban road area per capita (m2)	+	0.0542
Ecological protection	Annual average PM2.5 concentration (μg/m3)	−	0.0157
	Comprehensive utilization rate of solid waste (general industry) (%)	+	0.0070
	Per capita residential electricity consumption (kWh)	−	0.0775
	Greening coverage in built-up areas (%)	+	0.0108
	Industrial SO2 emissions per unit of output (tons/10,000 yuan)	−	0.0530
	Centralized sewage treatment rate (%)	+	0.0144
	Industrial soot emissions per 10,000 yuan of GDP (tons/ 10,000 yuan)	−	0.1389
	Per capita wastewater discharge (tons)	−	0.0550

Explanatory Variable

The dummy variable ${\mathrm{treated}}_i\times {\mathrm{time}}_t$ indicates whether a city is in the GFRIPZs group. It equals 1 if a city implemented GFRIPZs in or after 2017, and 0 otherwise.

Control Variables

Following relevant literature (Kim, 2022; X. Xu & Yang, 2025), several control variables are selected, including economic growth, as measured using the nighttime light index to minimize potential data collection and reporting biases; industrial structure, as determined by the percentage of GDP attributed to tertiary industry; environmental regulation, proxied by the utilization rate of solid waste from general industry; openness to international trade, measured as the ratio of total imports and exports to GDP; financial development, as determined by the ratio of year-end balances of financial institution deposits and loans to GDP; financial self-sufficiency, defined as the ratio of budgeted revenues to budgeted expenditures; and infrastructure level, represented by urban road area per capita.

Data Sources

Considering data availability and completeness, this study uses panel data from 271 Chinese cities over the period 2009 to 2021. Among them, eight cities included in the first batch of GFRIPZs established in 2017—Huzhou, Quzhou, Ganjiang New Area, Guangzhou, Gui’an New Area, and Karamay—are designated as the treatment group. The remaining cities serve as the control group. Data sources are as follows: information on GFRIPZs was manually compiled by the research team; nighttime light data were obtained from the Harvard Dataverse; green technology innovation data were collected from the National Intellectual Property Administration; firm-level data were sourced from the CEIC and EPS databases; and other variables were derived primarily from the China City Statistical Yearbook and local statistical yearbooks and bulletins issued by local statistical bureaus. Regarding data processing, (a) observations with missing values in key variables, such as those for cities in the Tibet Autonomous Region, were excluded; (b) missing values for certain indicators in specific years were interpolated; and (c) continuous variables were winsorized at the 1% level on both tails to minimize measurement errors and the influence of outliers. Table 2 reports summary statistics.

Table 2.

Descriptive Statistics.

Var	Definition	Obs	Mean	SD.	Min	Max
GDR	Green development resilience	3,523	0.4797	0.3752	−0.4020	1.7915
EG	Economic growth	3,523	8.0377	9.1878	0.3499	46.5593
Stru	Industrial structure	3,523	41.5032	10.0045	20.6000	71.5600
ER	Environmental regulation	3,523	80.5294	21.6037	12.8650	100
Open	Openness to international trade	3,523	0.1803	0.2775	0.0021	1.6062
Fisc	Fiscal autonomy	3,523	0.4661	0.2193	0.1151	1.0088
Infr	Infrastructure level	3,523	5.1403	5.3700	0.4797	33.8932
Fina	Financial development	3,523	2.4380	1.1370	0.9646	6.7265

Empirical Results

Benchmark Regression Results

The baseline regression is conducted using the DID model, with results presented in Table 3. Column (1) reports the regression results without including fixed effects or control variables. The coefficient of ${\mathrm{treated}}_i\times {\mathrm{time}}_t$ is positive and statistically significant, offering initial evidence that the establishment of GFRIPZs significantly improves GDR. Column (2) introduces year and city fixed effects. The coefficient remains positive, reinforcing the initial finding. Column (3) includes all control variables, along with fixed effects. The coefficient on ${\mathrm{treated}}_i\times {\mathrm{time}}_t$ remains significantly positive, suggesting that GFP increased average GDR in pilot cities by 14.2%.

Table 3.

Benchmark Regression Results.

Variable	GDR
	(1)	(2)	(3)
${\mathrm{treated}}_i\times {\mathrm{time}}_t$	0.2858***(0.0695)	0.1660***(0.0312)	0.1420***(0.0324)
_cons	0.4765***(0.0063)	0.4779***(0.0034)	0.2905***(0.0811)
Controls	No	No	Yes
City fixed effect	No	Yes	Yes
Year fixed effect	No	Yes	Yes
N	3,523	3,523	3,523
Adj. R2	.0062	.7125	.7310

*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively, with robust standard errors in parentheses. This notation is consistent across all subsequent tables.

Robustness Tests

Estimation of the Parallel Trend Assumption

Following Z. Liu et al. (2025), the year before the establishment of GFRIPZs is set as the baseline. This study window includes 4 years before and after the establishment of GFRIPZs. This study estimates the following event-study regression model:

$$\begin{array}{c}\multicolumn{1}{c}{{\mathrm{GDR}}_{\mathrm{it}}={\alpha }_1+\sum _{m=-4}^4{\beta }_m\times {\mathrm{treated}}_i\times {\mathrm{time}}_{\mathrm{mt}}+{\phi }_1{\mathrm{Control}}_{\mathrm{it}}}\\ \multicolumn{1}{c}{+{\mu }_i+{\lambda }_t+{\epsilon }_{\mathrm{it}}}\end{array}$$

(3)

Here, ${\mathrm{time}}_{\mathrm{mt}}$ represents a set of time dummies, where m is the number of years before or after the baseline period. If the interaction coefficients for the pre-policy periods are statistically insignificant, and those for the post-policy periods are significant, this supports the parallel trends assumption and suggests a significant dynamic effect of GFRIPZs on GDR. All other variables are set as in Equation 1.

Figure 1 shows the estimation results. Before the implementation of GFRIPZs, the coefficients are close to zero and statistically insignificant, indicating that the null hypothesis of parallel pre-trends cannot be rejected. After policy implementation, the 95% confidence interval of the interaction terms no longer includes zero, and the coefficients for all post-policy years are statistically significant and positive. This suggests that GFP exerts persistent and positive effects on GDR.

Figure 1.

Parallel trend test.

Placebo Test

Following G. Zhang and Guo (2025), cities and implementation years are randomly reassigned to construct a pseudo-treatment variable for GFRIPZs. The model is then re-estimated using Equation 2, and the corresponding placebo coefficients and p-values are recorded. After repeating the procedure 1,000 times, the distribution of placebo coefficients and their p-values is depicted in the kernel density and scatter plots in Figure 2. As shown, the placebo coefficients are approximately normally distributed around zero, in sharp contrast to the actual treatment effect in the baseline regression. In addition, most placebo p-values exceed the .1 threshold. These results further confirm that the observed positive effects of GFP on GDR are unlikely to have occurred by chance.

Figure 2.

Placebo test.

Endogeneity Test

Drawing on the watershed governance objectives outlined in the master plan for GFRIPZs and prior research (Z. Dong & Wang, 2019), this study employs the 1999 average annual precipitation as the first instrumental variable (IV1). As a key natural factor affecting watershed ecological quality, precipitation influences water conditions by transporting suspended solids, organic matter, and heavy metals (Y. Zhang et al., 2020). This environmental impact increases regulatory pressure and financial demand for water management, thereby strengthening local incentives to establish green finance pilot zones. Thus, historical precipitation is theoretically correlated with the establishment of GFRIPZs. At the same time, precipitation—being a natural variable—has no direct effect on GDR and predates the implementation of the policy, satisfying the condition of exogeneity. Following Borusyak et al. (2021), this study also constructs a panel-based instrumental variable by interacting the 1999 precipitation with the lagged national financial development (one period). In addition, drawing on Han et al. (2025), a second instrumental variable (IV2) is constructed by interacting the two-period lagged GFRIPZs implementation dummy with the one-period lagged national financial development. This helps address potential reverse causality between GFRIPZs and GDR.

Table 4 reports the instrumental variable (IV) estimation results. Columns (1) and (3) display the first-stage regressions, where both instruments show a statistically significant and positive relationship with GFRIPZs implementation. The corresponding first-stage F-statistics exceed the commonly used threshold of 16.38, indicating that the instruments are both strong and relevant. The second-stage regressions are reported in Columns (2) and (4), with the estimated coefficients for GFRIPZs aligning with those from the benchmark regression. These findings provide further evidence that GFP has a significant promoting effect on GDR.

Table 4.

Instrumental Variable Estimation Results.

Variables	(1)	(2)	(3)	(4)
	${\mathrm{treated}}_{\mathrm{i}}\times {\mathrm{time}}_{\mathrm{t}}$	GDR	${\mathrm{treated}}_{\mathrm{i}}\times {\mathrm{time}}_{\mathrm{t}}$	GDR
${\mathrm{treated}}_i\times {\mathrm{time}}_t$		2.7775*** (0.0000)		0.1798***(0.0000)
IV1	0.1992***(0.0000)
IV2			0.21833***(0.0000)
Controls	Yes	Yes	Yes	Yes
City fixed effect	Yes	Yes	Yes	Yes
Year fixed effect	Yes	Yes	Yes	Yes
Under identification test	32.4200*** (0.0000)		68.4640*** (0.0000)
Weak identification test	30.7100[16.38]		207.6420[16.38]
N	3,523	3,523	3,523	3,523

Note. Bracketed values represent the 10% critical values from the Stock–Yogo weak instrument test.

*** indicates statistical significance at the 1% level.

SCM-DID

Following Clarke et al. (2023), the synthetic control DID (SCM-DID) approach is applied to address potential deviations from the parallel trends assumption during the pre-treatment phase. Specifically, for each city implementing GFRIPZs, 150 bootstrapped samples are drawn from the control group to generate weighted synthetic control cities. This approach enables a more accurate comparison of GDR between cities and their synthetic controls after the policy intervention. As shown in Figure 3, prior to policy implementation, the actual treated cities and their synthetic counterparts exhibited highly similar levels of GDR. After the GFRIPZs policy was introduced, the treated cities showed a noticeable and sustained increase in GDR relative to the synthetic controls. Table 5 confirms this, showing an average treatment effect on the treated (ATT) of 0.1230, indicating a robust and positive impact of GFP on GDR.

Figure 3.

SCM-DID results.

Table 5.

SCM-DID Results.

GDR	ATT	Std. err.	t	p>|t|	[95% Conf. Interval]
${\mathrm{treated}}_i\times {\mathrm{time}}_t$	0.1230	0.0354	3.4700	.0010	0.0536	0.1924

PSM-DID

In practice, the selection of pilot cities may be non-random. Propensity score matching (PSM) identifies one or more non-pilot cities from the control group with similar characteristics to each pilot city, thereby constructing a comparable “counterfactual” city that approximates the conditions of a randomized experiment (Shi et al., 2022). This study applies the PSM-DID method with nearest-neighbor 1:1 matching. As shown in the PSM balance test plot (Figure 4), covariate imbalances are substantially reduced after matching, indicating satisfactory matching quality. Column (1) of Table 6 presents a statistically significant and positive coefficient for GFRIPZs, suggesting that the establishment of GFP significantly promotes GDR.

Figure 4.

PSM balance test plot.

Table 6.

Robustness Tests.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	PSM-DID	Change measure of GDR	Excluding related policy	Two-period DID model	Changing clustering level	Adding city-time trend	Adding controls
${\mathrm{treated}}_i\times {\mathrm{time}}_t$	0.1548*** (0.0333)	0.1594*** (0.0227)	0.1344*** (0.0331)	0.0875** (0.0338)	0.1420** (0.0576)	0.1197*** (0.0329)	0.1375*** (0.0329)
_cons	0.2524*** (0.0968)	−0.1411*** (0.0531)	0.3289*** (0.0862)	−0.5325 (0.6678)	0.2905** (0.1276)	−32.0700*** (5.2170)	−0.0027 (0.1662)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
City fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year fixed effect	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	3,008	3,523	3,252	542	3,523	3,523	3,519
Adj. R2	.7210	.4810	.7388	.8830	.7310	.7339	.7310

Note. **, and *** indicate statistical significance at the 5%, and 1% levels, respectively.

Change the Measure of GDR

To further test the robustness, this study adopts the Sensitivity Index (SI), a widely used alternative measure of GDR in prior literature (Holl, 2018). The SI is defined as follows:

$${\mathrm{SI}}_{\mathrm{it}}=({\mathrm{lnE}}_{\mathrm{it}}-{\mathrm{lnE}}_{i,t-k})-({\mathrm{lnE}}_t-{\mathrm{lnE}}_{t-k})$$

(4)

where $E_{\mathrm{it}}$ denotes the green development level, $E_t$ refers to the national average in year t, and $t-k$ indicates k years prior. A positive ${\mathrm{SI}}_{\mathrm{it}}$ suggests relatively strong GDR in city i at time t, while a negative value indicates weaker resilience.

Following A. Chen (2022), the period from 2009 to 2012 is defined as the “shock-resistance phase” to assess each city’s ability to withstand economic downturns, while the period from 2012 to 2021 is defined as the “recovery phase” to evaluate resilience during economic stabilization. Column (2) of Table 6 reports the results using the SI-defined GDR measure. The interaction term remains significantly positive, confirming the positive effect of GFP on GDR.

Other Robustness Tests

The 2012 release of the Green Credit Guidelines represented a major milestone in advancement green finance. To mitigate potential confounding effects from this national policy, the analysis follows Jin and Zhang (2024) by excluding observations from the year 2012 and re-estimating the model. Column (3) of Table 6 shows a significantly positive interaction term, suggesting that the influence of GFP on GDR is not driven by the Green Credit Guidelines policy.

To address potential estimation bias caused by serial correlation in DID models, a two-period DID model is employed as a robustness check (Bertrand et al., 2004). Specifically, regressions were conducted using data from 1 year before and 1 year after the policy implementation. Column (4) of Table 6 reports a significantly positive coefficient for GFP, confirming the robustness of its effect on GDR under this alternative specification.

Given that cities within the same province may share similar performance in terms of GDR, standard errors are clustered at the provincial level as a robustness check. Column (5) of Table 6 provides additional confirmation of the baseline result.

To control for city-specific time trends, city-by-year interaction terms are added to Model (1). Column (6) of Table 6 shows a positive coefficient, indicating that the effect of GFP on GDR is not driven by time-varying trends across cities.

To control for potential confounding factors, three additional variables were included: government digital focus (measured by the proportion of digital-related terms in government reports), government environmental protection effort (measured by the share of environmental protection expenditure in fiscal spending), and market competition (measured by the natural logarithm of newly established firms at the prefecture level). The regression results, reported in the last column of Table 6, indicate that both the magnitude and significance of the GFP coefficient remain stable, further supporting its robustness.

Extended Analysis

Mechanism Exploration

Drawing on Alesina and Zhuravskaya (2011), mediation models are constructed to examine the mechanisms through which GFP influences GDR.

$$\begin{array}{c}\multicolumn{1}{c}{{\mathrm{GDR}}_{\mathrm{it}}={\alpha }_2+{\beta }_2\times {\mathrm{treated}}_i\times {\mathrm{time}}_t+{\phi }_2{\mathrm{Control}}_{\mathrm{it}}}\\ \multicolumn{1}{c}{+{\mu }_i+{\lambda }_t+{\epsilon }_{\mathrm{it}}}\end{array}$$

(5)

$$\begin{array}{c}\multicolumn{1}{c}{{\mathrm{Mediator}}_{\mathrm{it}}={\alpha }_3+{\beta }_3\times {\mathrm{treated}}_i\times {\mathrm{time}}_t+{\phi }_3{\mathrm{Control}}_{\mathrm{it}}}\\ \multicolumn{1}{c}{+{\mu }_i+{\lambda }_t+{\epsilon }_{\mathrm{it}}}\end{array}$$

(6)

$$\begin{array}{c}\multicolumn{1}{c}{{\mathrm{GDR}}_{\mathrm{it}}={\alpha }_4+{\beta }_4\times {\mathrm{treated}}_i\times {\mathrm{time}}_t+\delta {\mathrm{Mediator}}_{\mathrm{it}}}\\ \multicolumn{1}{c}{+{\phi }_4{\mathrm{Control}}_{\mathrm{it}}+{\mu }_i+{\lambda }_t+{\epsilon }_{\mathrm{it}}}\end{array}$$

(7)

${\mathrm{Mediator}}_{\mathrm{it}}$ includes corporate financing constraints (CFC) and green technological innovation (GTI). If both models (6) and (7) yield statistically significant coefficients for the mediator, and the coefficient of the policy treatment variable decreases after including the mediator in the baseline model, a partial mediation effect is inferred. If the coefficient of treatment variable becomes statistically insignificant upon the inclusion of the mediator, this suggests a full mediation effect. Other variables follow the same specification as in model (1).

Corporate Financing Constraints

The FC index, following Hadlock and Pierce (2010), is used as a proxy for corporate financing constraints. Column (1) of Table 7 presents the estimated result of GFP on these constraints. The significantly negative coefficient of GFP indicates that GFP alleviates corporate financing constraints. Column (2) indicates a negative association between financing constraints and GDR. Including financing constraints in the baseline model reduces the coefficient of GFP to 0.085, though it remains statistically significant. This suggests that GFP enhances GDR partly by easing corporate financing constraints, providing empirical support for Hypothesis 2.

Table 7.

Mechanism Exploration Results.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	CFC	GDR	GTI 1	GDR	GTI 2	GDR
${\mathrm{treated}}_{\mathrm{i}}\times {\mathrm{time}}_{\mathrm{t}}$	−0.0248*** (0.0056)	0.0850*** (0.0054)	1.0417** (0.4238)	0.1341*** (0.0323)	0.5901*** (0.2209)	0.1338*** (0.0327)
GTI I				0.0076*** (0.0020)
GTI 2						0.0148*** (0.0037)
CFC		−0.0271*** (0.0047)
Control variable	Yes	Yes	Yes	Yes	Yes	Yes
City fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Year fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
N	58,432	58,432	3,519	3,519	3,511	3,511
Adj. R2	.7767	.8068	.7456	.7316	.7407	.7318

Note. **, and *** indicate statistical significance at the 5%, and 1% levels, respectively.

Green Technology Innovation

Following Song et al. (2021), green innovation scale (GTI 1) is represented by the number of green patent applications per 10,000 people, while green innovation quality (GTI 2) is proxied by the number of green invention patent applications per 10,000 people. Columns (3) and (5) of Table 7 report significantly positive coefficients for GFP on both green innovation scale and quality, indicating that GFP effectively enhances both dimensions. Columns (4) and (6) report that the coefficients for green innovation scale and quality remain significantly positive, while the coefficients of the policy treatment variable decrease compared to the baseline model. These results provide robust evidence that GFP promotes GDR through enhanced green technological innovation, thereby supporting Hypothesis 3.

Heterogeneity Exploration

To explore the heterogeneous effects of GFP on GDR, this study examines three moderating factors: environmental regulation intensity, openness level, and market segmentation degree.

Environmental Regulation Intensity

The Porter Hypothesis suggests that carefully crafted environmental regulations compel firms to adopt green technological innovations, thereby promoting regional green development (Porter & van der Linde, 1995). Consequently, the effect of GFP on GDR may differ based on the environmental regulation intensity, which is proxied by the comprehensive utilization rate of general industrial solid waste. The sample is grouped by the mean environmental regulation intensity, with results presented in Table 8. GFP significantly enhances GDR in cities with stricter environmental regulation, whereas no significant effect is found in cities with lower regulation intensity. This finding aligns with prior literature emphasizing the reinforcing effects of environmental regulation (Jing & Liu, 2024; Tariq & Hassan, 2023). A possible explanation is that the GFRIPZ policy provides a “green pull” for firms, while stringent environmental regulations exert a “green push,” creating both the capacity and the pressure for firms to invest in green initiatives and undertake green transformation. Specifically, in GFRIPZ pilot cities, financial institutions tend to support firms with stronger environmental responsibility (Tan & Zhu, 2022). To align with financial institutions’ preferences, firms integrate their financing activities with environmental responsibility to secure green funding. Meanwhile, regions with stringent environmental regulations impose stricter environmental standards, prompting local firms to accelerate green technological innovation (Jing & Liu, 2024), which increases demand for green financial instruments such as green loans and bonds. Therefore, the combination of GFRIPZ policy and strict environmental regulation equips firms with both the ability and the incentive to pursue green transformation, partially mitigating the negative externalities of environmental issues and enhancing the GDR of the cities.

Table 8.

Heterogeneity Exploration Results.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	High environmental regulation	Low environmental regulation	High openness	Low openness	High market segmentation	Low market segmentation
${\mathrm{treated}}_{\mathrm{i}}\times {\mathrm{time}}_{\mathrm{t}}$	0.1498*** (0.0421)	0.1085 (0.0757)	0.1493*** (0.0303)	0.0687 (0.0886)	0.1825*** (0.0382)	0.0902 (0.0594)
_cons	0.4393* (0.2378)	0.1176 (0.1307)	0.2762* (0.1476)	0.2016*** (0.0733)	0.1843 (0.1308)	0.4741*** (0.0959)
Control variable	Yes	Yes	Yes	Yes	Yes	Yes
City fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
Year fixed effect	Yes	Yes	Yes	Yes	Yes	Yes
N	1,739	1,738	1,757	1,749	1,993	1,529
Adj. R2	.7109	.7607	.7507	.7331	.7264	.7395

Note. *, and *** indicate statistical significance at the 10%, and 1% levels, respectively.

Openness

The pollution haven hypothesis posits that foreign firms may shift pollution-intensive industries to host countries with weaker environmental regulations, thereby worsening environmental degradation in the host countries (Levinson & Taylor, 2008). In contrast, the pollution halo hypothesis suggests that foreign investment enhances the host country’s environmental quality through knowledge spillovers and technology transfers (Aghasafari et al., 2021). Given these competing views, the level of openness may shape the impact of GFP on GDR by altering local pollution dynamics. This study measures openness using the ratio of total trade (imports and exports) to GDP and conducts subgroup regressions according to the median trade openness level. The results from the grouped regression show that GFP has no statistically significant effect on GDR in cities with low openness, but it has a significantly positive impact in cities with greater openness (Table 8). This indicates that, compared with highly open cities, low-openness cities have not fully realized the contribution of GFP to GDR. A possible explanation is that low-openness cities lack international communication and exchange, resulting in relative shortages of key resources and green technologies required for green development and resilience building. This finding aligns with previous studies on South Asian countries (J. Xu et al., 2023). Therefore, low-openness cities should prioritize attracting external development resources and green technologies to better leverage GFP in enhancing GDR.

Market Segmentation Degree

Regional differences in market segmentation can hinder the efficient allocation of production factors, thereby influencing the effect of GFP on GDR. Market segmentation is measured using inter-city relative consumer price indices, following Lyu et al. (2022). Cities are classified into high- and low-segmentation groups based on the median value of the indicator. Table 8 indicate that GFP’s influence on GDR is statistically insignificant in cities with low market segmentation but significantly positive in those with high segmentation. A possible explanation is that regions with high market segmentation experience insufficient capital flows, leading to underinvestment in green initiatives. In contrast, regions with low market segmentation benefit from more efficient capital flows (W. Chen & Li, 2025), enabling green investment to largely meet market demand. The green financial resources provided by GFRIPZ exhibit higher marginal returns in high market segmentation regions due to diminishing returns in green investment, resulting in a more pronounced effect on GDR in these regions compared to low-segmentation areas. This is consistent with literature emphasizing the critical role of market mechanisms (L. Dong et al., 2025). Therefore, support for the development of green finance in high market segmentation regions should be strengthened to better harness its role in enhancing GDR.

Discussion

This study provides clear evidence that green finance, as implemented through the GFRIPZ policy, significantly enhances GDR across Chinese cities. This finding is consistent with Y. Li and He (2025), who show that green finance improves ecological resilience; Zhao et al. (2025), who demonstrate that green finance promotes regional green growth; and A. Xu et al. (2025), who find that GFP fosters inclusive green growth. However, the results differ from studies suggesting that financial development can harm environmental quality or weaken resilience (Ahmad et al., 2022; Lee et al., 2024). This divergence likely arises from the fundamental distinction between general financial development and green finance. Unlike conventional financial expansion, green finance explicitly targets environmental goals by directing capital toward low-carbon sectors and sustainable projects. Consequently, its overall impact on ecological outcomes and resilience is more constructive.

Two mechanisms explain how GFP enhances GDR. First, green finance eases firms’ financing constraints by redirecting financial resources toward environmentally responsible industries. Access to such funding allows firms to invest in cleaner technologies, improve production efficiency, and strengthen their capacity to absorb and recover from external shocks. This mechanism echoes G. Zhang and Guo (2025), who emphasize the role of green finance in mitigating financial barriers to sustainable transformation. Second, green finance fosters GDR by stimulating green technological innovation. It encourages financial institutions to incorporate environmental risks into credit assessments (Necib & Gmati, 2024), while green insurance and similar instruments provide risk coverage for research and development activities. These financial supports lower uncertainty around green innovation and foster continuous innovation. In turn, innovation expands an economy’s adaptive and restorative capacities, enabling it to respond more effectively to disruptions. This pathway is consistent with Zhao et al. (2025) and reinforces the findings of Y. Li and He (2025), who show that green innovation contributes to ecological resilience.

The heterogeneity analysis further reveals that the impact of GFP on GDR is stronger in regions with stricter environmental regulation, higher trade openness, and greater market segmentation. These findings align with prior research (Aghasafari et al., 2021; Tariq & Hassan, 2023). In regions where environmental standards are more rigorous, firms face stronger incentives to innovate and higher demand for green finance, amplifying the positive effect of GFP—this pattern is consistent with the Porter Hypothesis (Porter & van der Linde, 1995). Contrary to the Pollution Haven Hypothesis, greater trade openness strengthens rather than weakens the impact of GFP on GDR. This finding diverges from the argument that trade liberalization leads to the relocation of polluting industries to less regulated countries (Levinson & Taylor, 2008). Finally, in regions with high market segmentation, limited factor mobility restricts access to green capital. In such contexts, GFP plays a greater role in bridging financing gaps and correcting resource mismatches, thereby enhancing GDR.

Conclusion and Implications

Conclusion

This study treats China’s 2017 GFRIPZs, implemented by the State Council, as a quasi-natural experiment to evaluate GFP’s effect on GDR. A DID approach is employed to evaluate this relationship. The results show that: (a) GFP significantly enhances GDR, with this conclusion remaining robust across various sensitivity checks. (b) The positive effects of GFP operate primarily through two channels: easing corporate financing constraints and fostering green technological innovation. (c) The effect of GFP on GDR is stronger in regions with stricter environmental regulations, greater openness, and more segmented markets.

Policy Implications

First, improve green finance reform efforts and guide the allocation of green capital. The government should gradually expand the coverage of GFP and encourage participation from local financial institutions and private capital in supporting green finance. The government should also explore the development of environmental rights trading markets, including emissions and energy use permits. Also, pilot programs for green patent valuation and pledge financing should be introduced. Comprehensive support measures, including tax incentives and low-interest loans, should be adopted to further strengthen GDR. In addition, regulatory authorities should expedite the development of a green finance risk prevention system. Full supervision of green capital usage must be enforced to prevent “greenwashing” and ensure that recipients are accountable for proper fund utilization.

Second, strengthen green technological innovation and alleviate corporate financing constraints. By leveraging green capital provided through the pilot green finance reforms, greater collaboration should be fostered among technology suppliers, market demand entities, and regulators. This coordination can support breakthroughs in green technologies, such as carbon capture and storage, intelligent energy management systems, and pyrolysis technologies, thus advancing the green innovation system. Meanwhile, the financial market should be further developed to guide capital toward high-growth green enterprises. Improving the market-based allocation of green capital can help relieve financing constraints—especially those related to green investments—and contribute to enhanced GDR.

Third, optimize the development environment and unlock the potential of GFP in enhancing GDR. Environmental regulatory authorities should strengthen environmental regulations and encourage enterprises to utilize GFP to support green technology research and application, green investment, and green consumption. These efforts will improve local systems’ capacity to absorb and recover from external shocks. Governments should enhance institutional openness by aligning with international trade and economic agreements, such as the CPTPP. Pilot programs involving cross-border negative list management and “whitelist” mechanisms should be explored to improve the business environment for foreign investment. Such measures can help attract external green capital, technologies, and advanced management practices to strengthen GDR. Additionally, the central government should accelerate the creation of a unified national market. Reducing regional market segmentation will promote the efficient cross-regional flow of green finance, enhancing financial support for strengthening GDR.

Limitations and Future Research Directions

Several limitations remain. First, due to data limitations, GDR is measured at the city level. If county-level or rural data become accessible in the future, the analysis could be refined to explore intra-city or urban–rural variations in GDR. Second, the sample is limited to mainland China. While Japan, the European Union, and the United States have also implemented comparable GFP, this study does not analyze their effects because of data constraints. Cross-country comparisons of GFP and its heterogeneous effects could be a valuable avenue for future research. Finally, although multiple robustness checks were conducted to control for potential confounding variables, some unobserved factors may still exist. As green finance policies and economic structures evolve, future studies could further address endogeneity concerns and test the generalizability of the conclusions under different policy and institutional environments.