Financial Pull and Administrative Push in Green Finance: Evidence from China’s Green Finance Pilot Policy

Abstract

Green finance has emerged as a crucial instrument for driving the macroeconomic transition toward a low-carbon economy, yet its specific transmission mechanisms warrant deeper empirical scrutiny. Leveraging China’s Green Finance Reform and Innovation Pilot Zones as a quasi-natural experiment, this scientific study employs a staggered difference-in-differences (DID) framework using provincial panel data from 2009 to 2023. To overcome the limitations of unidimensional metrics, we developed a comprehensive Industrial Structure Upgrading Index (ISUI) that integrates structural rationalization, advancement, and greening. The empirical findings reveal that the green finance pilot policy exerts a significant and positive impact on the ISUI. This core result remains robust under a series of rigorous checks, including the Callaway and Sant’Anna (CS-DID) estimator. Mechanism analyses demonstrate a dual “push–pull” dynamic: Green Credit Intensity (GCI) acts as the primary mediating channel by directing targeted financial resources (financial pull), while stringent environmental regulation positively moderates this effect (administrative push). Furthermore, the moderating role of digital finance is statistically non-significant, underscoring the policy’s broad inclusiveness and its independence from regional digital infrastructure. Heterogeneity estimations identify a clear structural catch-up effect, with more pronounced benefits observed in resource-dependent regions and areas with historically lower innovation capacities. Ultimately, these findings indicate that coordinating targeted financial incentives with environmental oversight can effectively drive multidimensional industrial upgrading, providing valuable evidence for sustainable transition strategies.

1. Introduction

Green finance has moved from a normative concept to a policy-relevant instrument for accelerating low-carbon transition and restructuring growth models in emerging economies. In China, green finance reforms have been institutionalized through a sequence of quasi-natural experiments, highlighted by the creation of the Green Finance Reform and Innovation Pilot Zones. These pilots aim to redirect credit and investment toward environmentally friendly activities while discouraging high-pollution and high-energy-consumption production. Whether such place-based green finance policies can generate real structural upgrading—rather than merely shifting pollution across sectors or regions—remains a central empirical question for sustainability-oriented research.

The existing literature provides mixed expectations. On the one hand, financial instruments may facilitate green reallocation by easing capital constraints for clean technologies and improving resource allocation efficiency, thereby supporting industrial upgrading and sustainable development outcomes [1,2]. On the other hand, green finance may produce heterogeneous or even offsetting effects when local implementation capacity is weak, when firms engage in “greenwashing”, or when policy-induced credit expansion crowds out more productive private investment [3]. Moreover, recent evidence suggests that green finance pilots can improve environmental performance (e.g., air quality and carbon emissions) partly through industrial structure channels, but the strength and generality of these mechanisms remain debated across settings and outcomes [4,5,6,7,8].

A key limitation of prior work is that “mechanism” is often discussed in broad terms (innovation, regulation, financing constraints) without cleanly distinguishing the market channel from the governmental channel. In practice, China’s green finance pilots combine financial incentives with administrative coordination and regulatory salience. This implies a “push–pull” governance logic in which Green Credit Intensity (GCI) can transmit policy shocks into firms’ financing conditions and sectoral reallocation (financial pull), while environmental regulation intensity (ER) may amplify (or dampen) the effectiveness of credit-based support by altering compliance costs, risk pricing, and local enforcement expectations (administrative push). Recent studies document that pilot-zone reforms can affect firm investment efficiency and green innovation through financing-related mechanisms, but whether these channels translate into macro-level industrial upgrading—and under what regulatory environments—requires further causal evidence [9,10].

Against this background, this scientific study investigates the causal impact of the green finance pilot policy on industrial structure upgrading. Although a proliferation of difference-in-differences (DID) studies has evaluated the singular environmental or economic effects of green finance policies, this scientific study fundamentally advances the literature by transitioning from a “single-channel impact evaluation” to a “multidimensional governance paradigm”. A critical limitation of existing DID literature is its reliance on unidimensional metrics to proxy structural upgrading (such as the simple output share of the tertiary sector). These traditional indicators often fall into an “evaluation trap”, creating an illusion of upgrading. For instance, a region may experience a rapid expansion of low-end services while its core manufacturing sector remains highly polluting and resource-inefficient. To overcome this myopic approach, we constructed a multidimensional Industrial Structure Upgrading Index (ISUI) grounded in Ecological Modernization Theory. By mathematically integrating structural rationalization (resource allocation efficiency), structural advancement (moving up the value chain), and structural greening (decoupling growth from carbon and pollution), the ISUI ensures a precise and holistic measurement of sustainable transformation. Exploiting the staggered implementation of the pilot policy, we estimate dynamic treatment effects and perform comprehensive sensitivity analyses to ensure causal identification, encompassing pre-treatment trend evaluations, falsification exercises, and the robust Callaway and Sant’Anna (CS-DID) estimator to address potential heterogeneous treatment effects. We further tested a dual-path interpretation: (i) a mediation pathway through the structural penetration of green credit (GCI), and (ii) a moderation pathway whereby environmental regulation conditions policy effectiveness. In addition, we examined heterogeneity across regions and technological baselines, and assessed whether digitalization conditions policy impacts—an issue of growing relevance given concerns about digital divides and unequal policy incidence.

Our empirical findings robustly demonstrate that the implementation of the green financial reform significantly promotes industrial upgrading, with effects that emerge after implementation and persist over time. Mechanism tests suggest that the structural deepening of green credit serves as a meaningful transmission channel, while stronger environmental regulation reinforces the policy’s upgrading effect. By contrast, the interaction between policy and digitalization is not statistically significant, confirming the policy’s broad inclusiveness and its independence from the level of local digital infrastructure in the upgrading process. Furthermore, heterogeneity analyses reveal a pronounced “technological catch-up” effect, where the policy exhibits stronger compensatory benefits in resource-dependent provinces and regions with weaker initial green technological endowments. These findings contribute to the sustainability and green finance literature by first, introducing a comprehensive multidimensional measurement for green-advanced transition; second, providing a governance-consistent causal account of how financial pull and administrative push jointly shape structural transformation; and third, offering policy-relevant implications for designing green finance programs that break technological lock-in and remain effective even in regions with weaker digital foundations.

2. Literature Review and Theoretical Hypotheses

2.1. Theoretical Basis of Green Finance and Sustainable Planning

The theoretical foundation of green finance spans a broad spectrum, from global macroeconomic sustainability frameworks to micro-level financial instruments. On a macro level, the implementation of global green policies and sustainable planning requires a nuanced understanding of regional contexts to prevent unintended structural imbalances. It has been argued that achieving sustainable ends necessitates meticulously designed policy means that avoid developmental malintent and foster equitable global planning [11]. At the micro level, the execution of these green transitions heavily relies on specific financial vehicles, such as green securities, green bonds, and targeted credit. The theoretical clarity and legal perfection of these securities are critical for infrastructure development and the effective operation of green financial markets [12]. Building upon this robust theoretical basis, contemporary scholarship has increasingly focused on how these financial instruments materialize in specific national contexts to drive structural economic transformation.

2.2. Global Green Finance Frameworks and Climate Transition Strategies

To fully appreciate the institutional novelty of China’s pilot zones, it is essential to contextualize these localized policies within the broader global sustainable finance literature. Globally, the macroeconomic transition to a low-carbon economy has increasingly relied on standardized Environmental, Social, and Governance (ESG) risk governance and climate transition finance frameworks [12]. In developed financial markets, sustainable industrial policy is predominantly market-driven. For instance, the European Union’s Sustainable Finance Taxonomy has emerged as a comprehensive classification system designed to steer private investments toward climate neutrality by imposing stringent technical disclosure thresholds and systematically combating “greenwashing” behaviors [13].

However, while the EU model relies heavily on corporate transparency, rigorous accounting standards, and mature market-driven ESG ratings, emerging economies often face institutional bottlenecks in implementing such highly granular disclosure systems. In these contexts, climate transition risks and sustainable development goals must often be managed through alternative governance structures [14]. In contrast to the EU’s market-centric paradigm, China’s green finance pilot policy represents a unique “state-guided” model. It utilizes macro-level financial incentives (the financial pull of targeted credit) and administrative environmental regulations (the administrative push) as substitute governance mechanisms to drive sustainable industrial policy. By evaluating this specific push–pull framework, this scientific study bridges a critical gap in the existing literature, providing empirical evidence on how emerging economies can achieve structural upgrading and climate transition without strictly relying on pre-existing, highly developed market disclosure infrastructures.

2.3. Transmission Mechanisms Between Green Finance and Industrial Upgrading

Recent literature highlights three primary mechanisms through which green finance drives industrial structure upgrading: credit reallocation, the stimulation of green technological innovation, and synergistic interactions with environmental regulations.

Firstly, regarding credit reallocation, green finance introduces strict financing constraints on highly polluting and energy-intensive industries while offering preferential credit to eco-friendly sectors. Previous research has demonstrated that green finance optimizes the economic development mode by forcing “brown” enterprises to transform under funding pressure, thereby directly enhancing the rationalization of the industrial structure [15]. Similarly, empirical evidence confirms that green credit policies significantly restrict debt financing for heavy polluters, effectively shifting capital toward sustainable projects [16].

Secondly, green finance alleviates the severe financing constraints associated with green technological innovation. Eco-innovation typically requires substantial capital and entails high long-term risks. It has been found that green finance acts as a crucial catalyst, providing the necessary risk diversification and financial support that enable enterprises to transition from resource-intensive production to technology-driven models [17].

Lastly, green finance mechanisms do not operate in a vacuum but exhibit strong synergies with administrative environmental regulations. Related studies indicate that the combination of financial pull mechanisms (e.g., green credit) and administrative push mechanisms (e.g., environmental penalties) creates a dual-pressure environment [18]. This synergy accelerates the phase-out of obsolete production capacities and establishes a stable financing environment for green industries.

2.4. Theoretical Hypotheses

The inauguration of designated experimental regions for environmental finance constitutes an institutional arrangement aimed at embedding environmental objectives into the financial system. Through the formulation of green finance standards, the provision of policy-based interest subsidies, and the implementation of risk-sharing mechanisms, the policy systematically channels capital flows into low-carbon, circular, and innovation-oriented sectors. Such policy-induced capital reallocation is expected to facilitate the transformation of regional industrial structures toward more advanced, rationalized, and sustainable configurations.

Existing studies indicate that green finance affects industrial upgrading primarily by reshaping the allocation of financial resources across sectors, accelerating the exit of pollution-intensive and low-efficiency industries, and fostering the expansion of higher value-added activities [19,20]. Unlike conventional financial development, green finance explicitly incorporates environmental criteria into credit allocation and risk pricing, which may simultaneously promote industrial advancement and improve structural rationalization through enhanced factor mobility and efficiency [17].

Building upon the theoretical framework, we posit the following hypothesis:

H1. 

Green finance policies significantly promote industrial upgrading (ISUI).

Green finance policies do not directly intervene in firms’ production processes; instead, they operate through the financial system by expanding access to credit for environmentally friendly and innovation-oriented activities. Financing constraints are widely recognized as a major barrier to industrial upgrading, particularly for green transformation projects that require substantial upfront investment and involve long payback periods.

Rather than a simple horizontal expansion of total liquidity, green finance policies function by improving green project identification and incentivizing financial institutions to optimize their asset structures. This leads to an increase in Green Credit Intensity (GCI)—the relative depth of green capital penetration within the regional economy. By lowering financing thresholds and costs for green enterprises, these policies effectively deepen the “financial pull” mechanism. An elevated GCI signifies that a larger share of regional economic output is backed by eco-friendly capital, which improves firms’ liquidity conditions, supports technological upgrading, and facilitates long-term investment in cleaner production processes. Prior studies have shown that such structural credit reallocation toward environmentally responsible activities plays a crucial role in promoting structural transformation and productivity improvement [18,21]. In this sense, the intensification of green credit acts as the pivotal linkage that translates policy incentives into tangible industrial upgrading outcomes.

Thus, we hypothesize:

H2. 

Green finance policies promote industrial upgrading by increasing Green Credit Intensity.

The effectiveness of green finance policies is likely to depend on the broader institutional environment, particularly local governments’ commitment to environmental governance. Government environmental attention reflects regulatory preferences, enforcement intensity, and policy signaling related to environmental protection.

The literature suggests that stronger environmental regulation can reinforce firms’ incentives to upgrade technologies and adjust industrial structures, especially when regulatory pressure is complemented by supportive financial mechanisms [22,23]. In regions with higher environmental attention, enterprises face stronger compliance pressure, while lending institutions are more inclined to embed ecological metrics into credit evaluation and risk management. Consequently, administrative pressure and financial incentives may jointly strengthen the effectiveness of green finance policies.

Accordingly, we propose:

H3. 

Government environmental attention (ER) positively moderates the effect of green finance policies on industrial upgrading, such that the policy effect is stronger in regions with higher ER.

Digital finance has been widely argued to reduce transaction costs, alleviate information asymmetry, and improve financial accessibility. From this perspective, digital financial development may provide an essential infrastructure for the transmission of financial policies [24]. However, rather than exacerbating a “digital divide”, digital finance is characterized by its “long-tail” and low-threshold features, which allow it to function as a broadly inclusive catalyst. Green finance pilot zones are designed as comprehensive institutional arrangements that utilize standardized policy instruments with broad coverage, enabling them to operate effectively even in regions with heterogeneous digital foundations.

In this context, digital finance is more likely to serve as a broadly inclusive background condition rather than a decisive threshold variable that restricts policy efficacy. Because digital finance relies on algorithmic credit assessments and big data rather than physical bank branches or traditional collateral, it empowers disadvantaged regions to capture green credit opportunities as effectively as technologically advanced ones. Consequently, the marginal moderating role of digitalization may appear statistically limited, precisely because the green finance policy exhibits a high degree of structural inclusiveness and independence from digital infrastructure. This ensures that the policy’s “financial pull” effect is not limited by local digital infrastructure disparities but instead promotes a cross-regional catch-up effect [25].

Thus, we hypothesize:

H4. 

The moderating effect of digital finance development (Adig) on the relationship between green finance policy and industrial upgrading is statistically insignificant, reflecting the policy’s inclusiveness and its independence from local digital infrastructure.

The marginal impact of green finance policies on industrial upgrading may vary substantially across regions due to differences in technological capabilities and resource endowments. Regions with lower levels of green technological development—proxied by fewer green patent authorizations—often face stronger financing constraints and higher barriers to industrial transformation. In such contexts, green finance may play a compensatory role by alleviating capital shortages and facilitating technological catch-up.

Similarly, resource-dependent regions typically exhibit heavier industrial structures and higher transformation costs, making them more responsive to targeted financial support. Theoretical and empirical studies on technological catch-up and creative destruction suggest that policy-induced capital reallocation can generate larger marginal gains in lagging regions [26].

Therefore, we propose:

H5. 

The positive effect of green finance policies on industrial upgrading is stronger in regions with lower green patent activity and in resource-dependent regions.

3. Materials and Methods

3.1. Overview of Scientific Research Methods

Before detailing the specific econometric models, it is essential to outline the fundamental scientific research methods employed in this scientific study. We utilize deduction to derive our four theoretical hypotheses logically from existing macroeconomic and sustainable finance paradigms. Synthesis is rigorously applied in the construction of the multidimensional Industrial Structure Upgrading Index (ISUI), aggregating rationalization, advancement, and greening metrics into a cohesive evaluation system. Furthermore, comparative analysis serves as the core logic of our staggered difference-in-differences (DID) approach, contrasting the evolutionary trajectories of pilot and non-pilot regions over time. Finally, induction is used to generalize our specific provincial empirical findings into broader, actionable policy implications for emerging economies. This systematic methodological approach ensures the logical rigor and theoretical coherence of the entire empirical design.

3.2. Data Sources and Sample Selection

This scientific study employs a strictly balanced panel dataset covering 30 Chinese provinces over the period 2009–2023, excluding Tibet and the Hong Kong, Macao, and Taiwan regions due to data availability constraints. The extended pre-treatment period (2009–2016) was deliberately retained to ensure sufficient temporal depth for validating the parallel trends assumption prior to the staggered policy shocks. Furthermore, when aggregating macro-financial data, we acknowledge that the efficacy of green capital allocation can be highly heterogeneous. As established in the financial literature regarding asset selections, discretionary evaluation is essential because liquidity and performance can vary significantly even within seemingly similar asset classes [27]. Thus, carefully constructed panel data is critical for mitigating such heterogeneity.

Macroeconomic and structural variables were obtained from the China Statistical Yearbook, the China Energy Statistical Yearbook, and provincial statistical bulletins. To mitigate the influence of extreme outliers and ensure the robustness of empirical results, all continuous variables are winsorized at the 1st and 99th percentiles prior to estimation.

3.3. Variable

To ensure clarity and reproducibility, the detailed definitions, symbols, and specific measurement methods for all variables used in this empirical study are summarized in

Table A1. Definitions and measurements of variables.

Variable Type	Variable Name	Symbol	Definition and Measurement
Dependent Variable	Industrial Structure Upgrading Index	ISUI	A composite index synthesized from Rationalization (R), Advancement (A), and Greenization (G) using the Entropy Weight Method.
Independent Variable	Green Finance Pilot Policy	Policy	A multi-period DID dummy variable; 1 if a province is in the pilot zone in a given year and thereafter, 0 otherwise (the one-period lagged form L1.Policy is used in regressions).
Mediating Variable	Green Credit Intensity	GCI	The ratio of the provincial green credit balance to the regional gross domestic product.
Control Variables	Economic Development	Ed	The natural logarithm of per capita GDP.
	Human Capital	Hc	The average years of education of the regional population.
	Urbanization Rate	Ur	Measured by the ratio of urban dwellers to the total regional demographic.
	Foreign Direct Investment	Fdi	The proportion of actual utilized foreign direct investment in GDP.
	Government Intervention	Gov	The proportion of local fiscal expenditure in GDP.
	Consumption Level	Con	The logarithmic form of total consumer retail sales.

in Appendix A.

3.3.1. Dependent Variable

To capture the extent of structural transformation, this scientific study designates the Industrial Structure Upgrading Index (ISUI) as the core response variable. Following the multidimensional perspective widely adopted in the literature, ISUI is constructed as a composite index synthesizing a triad of core elements: the coordination of structural proportions, the transition toward high-value-added sectors, and the environmental performance of industrial activities. This multidimensional framework allows for a more comprehensive evaluation of industrial upgrading by jointly capturing efficiency, structural hierarchy, and environmental performance.

Prior to aggregation, each sub-indicator is normalized to eliminate scale effects. The final ISUI is then synthesized using the entropy weighting method, which assigns objective weights based on the information content of each indicator, thereby reducing subjectivity in index construction [28,29].

• Industrial Structure Rationalization (R)

Industrial structure rationalization is operationalized through the application of the Theil index, which captures the degree of mismatch between industrial output and employment distribution. Compared with simpler indicators, the Theil index not only reflects deviations in productivity across sectors but also incorporates sectoral economic weights, making it particularly suitable for assessing structural efficiency [30,31].

The Theil index is calculated as:

$$T{L}_{it}=\sum _{j=1}^3\left({\displaystyle \frac{Y_{ijt}}{Y_{it}}}\right)\mathit{ln}\left({\displaystyle \frac{Y_{ijt}/L_{ijt}}{Y_{it}/L_{it}}}\right)$$

(1)

where $Y_{ijt}$ and $L_{ijt}$ correspond to the generated value-added and workforce participation for a specific sector j in province i at year t, respectively; and $Y_{it}$ and $L_{it}$ represent total output and total employment of province i.

A lower Theil index indicates a more balanced allocation of labor relative to output across industries, implying a more rational industrial structure. Since the Theil index is a reverse indicator, it is transformed prior to aggregation—either by inverse normalization or by taking its reciprocal—so that higher values consistently indicate a higher degree of industrial upgrading. This treatment ensures directional consistency with the advancement and greening dimensions [32].

2.

Industrial Structure Advancement (A)

The sophistication of the industrial structure mirrors an economy’s evolutionary trajectory, shifting away from basic resource extraction toward higher-tier manufacturing and tertiary sectors that yield greater economic value. Following established practice, this scientific study measures advancement using the industrial hierarchy coefficient, defined as:

$$\mathrm{A}={\sum }_{i=1}^3{q}_i\times i=q_1\times 1+q_2\times 2+q_3\times 3$$

(2)

where $q_i$ denotes the share of value added of industry j in regional GDP. A higher value of A indicates a more advanced industrial structure characterized by a greater concentration in higher-order industries [33,34].

This indicator effectively captures the upgrading trajectory of industrial composition and has been widely applied in empirical studies on structural transformation.

3.

Industrial Greening (G)

Industrial greening captures the environmental performance of industrial development and reflects the extent to which production activities are energy-efficient and environmentally friendly. The core indicator is energy consumption per unit of GDP, supplemented by major industrial pollution emission indicators, including industrial wastewater discharge, sulfur dioxide emissions, industrial smoke and dust emissions, and industrial solid waste generation.

This dimension evaluates industrial upgrading from the perspective of resource efficiency and environmental sustainability, which is increasingly regarded as an essential component of high-quality development and green transformation [18,35,36]. Lower energy intensity and pollution emissions correspond to a higher degree of industrial greening.

3.3.2. Explanatory Variable

Serving as the primary treatment variable in our empirical design, we focus on the Green Finance Reform and Innovation Pilot policy (Policy). Following the standard staggered difference-in-differences framework, the policy variable is constructed as an interaction term: $Polic{y}_{it}=Trea{t}_i\times Pos{t}_{it}$. where $Trea{t}_i$ is a dummy variable equal to 1 for provinces included in the green finance pilot program and 0 otherwise, and $Pos{t}_{it}$ indicates whether province iii has entered the policy implementation period in year t.

Considering that the transmission of financial policies typically involves time lags—particularly through credit allocation, project screening, and investment decision-making—the empirical analysis strictly employs the one-period lagged policy variable (L1.Policy) across all baseline and interaction models. This lagging strategy not only aligns with the gradual nature of policy transmission but also effectively mitigates microstructure noise-induced biases that can distort contemporaneous estimators during periods of structural macro-financial shocks [37].

According to official approval documents jointly issued by the China’s central bank (PBOC) and other central government agencies, the pilot provinces in the sample were approved in multiple waves. Specifically, the first batch, established in 2017, includes Zhejiang, Jiangxi, Guangdong, Guizhou, and Xinjiang. These provinces span China’s eastern coastal region, central areas, and ecologically fragile western regions, thereby offering substantial regional heterogeneity and representativeness.

Subsequent expansions occurred after 2019. The national roster for these green financial experiments was subsequently broadened to include Gansu in 2019, followed by Chongqing in 2022. These later pilot regions actively explored institutional innovations in areas such as green credit standards, green bond issuance, and environmental information disclosure. The staggered implementation of the pilot policy across provinces provides a credible quasi-natural experiment enabling the clean isolation of the causal impact that these green finance initiatives exert on structural transformation.

3.3.3. Mediating Variable

The mediating variable in this scientific study is Green Credit Intensity (GCI). To precisely capture the structural reallocation of financial resources driven by green policies, and to avoid the confounding effects of broad macroeconomic credit expansions, GCI is defined as the ratio of the provincial green credit balance to the regional gross domestic product (GDP).

This relative indicator accurately measures the intensity of green financial support directly backing the real economy. Compared to generic total credit volumes, this GDP-denominated GCI effectively isolates the specific “financial pull” mechanism, reflecting the degree to which regional economic output is structurally supported by eco-friendly capital rather than merely being driven by overall market liquidity.

Higher values of GCI signify a more concentrated and targeted supply of green credit resources within the region. This intensified financial support is theoretically expected to alleviate the severe financing constraints faced by green and innovation-oriented enterprises, while simultaneously accelerating the phase-out of highly polluting industries, thereby fundamentally facilitating regional industrial structure upgrading.

3.3.4. Moderating Variables

Environmental regulation intensity (ER) is measured using a text-based indicator constructed from provincial government work reports. Unlike traditional measures that rely on pollution control expenditures or emission-based indicators, this scientific study captures environmental regulation from the perspective of policy orientation and governmental attention.

Specifically, annual government work reports for each province are collected, and the frequencies of 15 environment-related keywords—such as environmental protection, pollution, emission reduction, and ecological conservation—are identified through textual analysis. These keyword frequencies are then aggregated to form an index reflecting the intensity of local governments’ environmental attention. A higher value of ER indicates stronger policy emphasis on environmental governance and a greater likelihood of regulatory enforcement.

This measure allows the study to capture the administrative pressure and policy-signaling effect associated with environmental regulation, which may interact with green finance policies and amplify their impact on industrial upgrading.

3.3.5. Contextual Variables

Given the rapid expansion of the digital economy, digitalization may shape the broader environment in which financial policies operate by affecting information flows, transaction costs, and access to financial services. In this scientific study, digital finance development is introduced as a contextual condition to examine whether the effectiveness of green finance policies depends on regional digital financial infrastructure. Importantly, digitalization is not treated as an internal transmission mechanism through which green finance policies affect industrial upgrading, but rather as a conditioning factor that may influence the strength or universality of policy effects.

Digital finance development (Adig) is measured using the Peking University Digital Inclusive Finance Index. To capture its multidimensional nature, three sub-indicators are further employed: Digital Coverage (Cdig), Digital Usage Depth (Ddig), and Digitalization Level (Sdig). Digital coverage reflects the breadth of access to digital financial services, usage depth captures the intensity of actual utilization in areas such as payments, credit, and investment, while digitalization level represents the technological sophistication and intelligence of financial services.

In the empirical analysis, interaction terms between the green finance pilot policy and Adig, as well as its three sub-dimensions, are constructed to assess whether policy impacts vary systematically across regions with different levels of digital financial development. This approach allows us to evaluate whether green finance policies rely on advanced digital infrastructure, or whether it instead exhibits broad inclusiveness and independence from regional digital capabilities across heterogeneous contexts.

3.3.6. Control Variables

To mitigate potential omitted-variable bias, a set of control variables is included from the economic, social, and policy dimensions.

• Economic development level (Ed): Calculated by taking the natural log of per capita gross domestic product. According to the Kuznets hypothesis, economic development constitutes a fundamental driver of structural transformation.
• Human capital (Hc): Quantified by the mean duration of formal education. A more educated labor force provides the foundation for the development of knowledge-intensive and technology-driven industries.
• Urbanization level (Ur): Defined as the proportion of the urban population to the total population. Urbanization generates agglomeration effects that enhance resource allocation efficiency and industrial diversification.
• Foreign direct investment (Fdi): Measured as the ratio of actual utilized foreign investment to GDP. Fdi inflows may promote industrial upgrading through technology spillovers and management learning effects.
• Government intervention (Gov): Measured by the share of local government spending relative to the regional gross domestic product, capturing the extent of governmental involvement in resource allocation and economic guidance.
• Consumption capacity (Con): Expressed as the natural logarithm of total consumer spending in the retail sector. Consumption upgrading can exert demand-side pressure on firms, encouraging supply-side optimization and industrial upgrading.

3.4. Econometric Model Specification

3.4.1. Multi-Period DID Benchmark Model

To empirically validate Hypothesis H1 regarding the causal effect of the environmental finance pilot program on sectoral structural advancement, our baseline estimation employs the following staggered difference-in-differences (DID) specification:

$$ISU{I}_{it}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_{1}L1.Polic{y}_{it}+\mathsf{\gamma }{X}_{it}+{\mathsf{\mu }}_i+{\mathsf{\lambda }}_t+{\mathsf{ϵ}}_{it}$$

(3)

Variable Definitions: $ISU{I}_{it}$: The Industrial Structure Upgrading Index of province i in year t; $L1.Polic{y}_{it}$: Serves as the principal binary treatment indicator, introduced with a one-period lag to accommodate policy implementation delays and mitigate potential endogeneity; $X_{it}$: Incorporates a comprehensive vector of time-varying regional covariates; ${\mathsf{\mu }}_i$: Absorbs unobserved, time-invariant heterogeneity across distinct provincial jurisdictions; ${\mathsf{\lambda }}_t$: Eliminates the confounding influence of synchronous macroeconomic fluctuations and economy-wide structural shifts; ${\mathsf{ϵ}}_{it}$: Constitutes the idiosyncratic disturbance term.

3.4.2. Parallel Trend Test Model

The validity of the DID method relies on the parallel trend assumption. To rigorously validate the fundamental prerequisite of our empirical design, we deployed a dynamic event study methodology to confirm that the treated and untreated cohorts exhibited synchronous evolutionary trajectories preceding the regulatory shock:

$$ISU{I}_{it}={\mathsf{\alpha }}_0+\sum _{k=-4,k\ne -1}^5{\mathsf{\beta }}_k{D}_{it}^k+\gamma X_{it}+{\mathsf{\mu }}_i+{\mathsf{\lambda }}_t+{\mathsf{ϵ}}_{it}$$

(4)

Variable Definitions: $D_{it}^k$: A series of dummy variables representing the relative time to the policy implementation. Let $T_i$ be the year when province i implemented the policy; then $k=t-T_i; D_{it}^k$ equals 1 if the relative year is k, and 0 otherwise. The year prior to the policy ($k=-1$) is chosen as the reference period. If the estimated coefficients ${\mathsf{\beta }}_k$ are not significantly different from 0 for $k<0$, evidence indicates no violation of the parallel trends assumption.

3.4.3. Mediation Effect Model

Following the causal steps approach, this scientific study establishes the following recursive equations to test the mediating role of Green Credit Intensity (GCI) (Hypothesis H2):

$$\begin{array}{c}GC{I}_{it}=a_0+a_{1}L1.Polic{y}_{it}+\mathsf{\gamma }{X}_{it}+{\mathsf{\mu }}_i+{\mathsf{\lambda }}_t+{\mathsf{ϵ}}_{it}\\ ISU{I}_{it}=c_0+c_{1}L1.Polic{y}_{it}+b_{1}GC{I}_{it}+\mathsf{\gamma }{X}_{it}+{\mathsf{\mu }}_i+{\mathsf{\lambda }}_t+{\mathsf{ϵ}}_{it}\end{array}$$

(5)

Variable Definitions: $GC{I}_{it}$: The mediating variable, representing the ratio of the provincial green credit balance to the regional GDP; c: The total effect of the policy on industrial structure upgrading; a: The effect of the policy on the mediator (loan); b: The effect of the mediator on the dependent variable (ISUI) after controlling for the policy; c: The direct effect of the policy on ISUI. If the indirect effect $a\times b$ is statistically significant (verified by Sobel or Bootstrap tests), the mediation mechanism is established.

4. Empirical Results

4.1. Descriptive Statistics

Table 1. Descriptive statistics.

Variable	Obs.	Mean	Std. Dev.	Min	Max
ISUI	450	0.485	0.163	0.135	0.995
policy	450	0.093	0.291	0	1
GCI	450	0.187	0.104	0.021	0.519
ER	450	56.54	19.3	6.000	124.0
ES	450	0.581	0.154	0.115	0.942
FD	450	0.075	0.035	0.021	0.231
Ed	450	9.321	0.468	8.442	10.807
Hc	450	0.021	0.006	0.002	0.044
Ur	450	0.593	0.127	0.234	0.896
Fdi	450	0.275	0.291	0.011	1.464
Gov	450	0.242	0.1	0.096	0.643
Con	450	0.372	0.068	0.183	0.538

outlines the baseline distributional characteristics of all focal variables incorporated into our econometric framework, derived from a symmetrical regional panel comprising 450 province-year observations.

The dependent variable, Industrial Structure Upgrading Index (ISUI), has a mean value of 0.485, with a standard deviation of 0.163. Its minimum and maximum values range from 0.135 to 0.995, indicating substantial variation in the level of industrial structure upgrading across provinces.

All control variables fall within reasonable ranges and are consistent with existing provincial-level studies, suggesting no abnormal data patterns.

4.2. Baseline Multi-Period DID Regression Results

Table 2. Impact of green finance policy on ISUI (comparison).

	(1)	(2)
	Current	Lagged_1
Policy	0.034 **	0.035 ***
	(2.496)	(2.772)
Control Variables	YES	YES
Year FE	YES	YES
Individual FE	YES	YES
Observations	450	420
Adj. R-squared	0.957	0.956

t statistics in parentheses. ** p < 0.05, *** p < 0.01.

documents the baseline estimates derived from the staggered DID framework, delineating both the immediate and delayed structural impacts of the green finance pilot zones.

Model (1) illustrates the regression outcomes utilizing the unlagged treatment indicator. The estimated parameter for the green finance mandate yields a significantly positive sign at the 5% threshold, signifying that the policy’s inception directly corresponds to measurable advancements in the ISUI. Such findings offer initial empirical substantiation that environmental finance mechanisms actively propel industrial restructuring.

Subsequently, Model (2) details the estimations incorporating a one-period temporal lag for the policy variable. The estimated coefficient remains positive and statistically significant at the 1% level, with a slightly larger magnitude compared to the contemporaneous specification. In terms of economic significance, this coefficient indicates that the implementation of the green finance pilot policy leads to an average absolute increase of 0.035 units in the ISUI. Given that the standard deviation of the ISUI in our full sample (as reported in the descriptive statistics), this policy shock translates to an approximate 21.5% of a standard deviation improvement in regional industrial structural upgrading. This represents a highly substantial macroeconomic transformation over the sample period. Furthermore, this finding suggests that the effect of green finance reform on industrial structure upgrading is not immediate but materializes with a time lag, which is consistent with the gradual nature of financial policy transmission through credit allocation and enterprise investment.

The comparison between Model (1) and Model (2) highlights the importance of accounting for policy lag effects in a multi-period DID framework. The stronger and more robust effect observed in the lagged specification supports the use of lagged policy variables in subsequent analyses.

The results in Table 2 provide strong support for Hypothesis H1, confirming that green finance reform significantly promotes industrial structure upgrading.

4.3. Sub-Dimensional Analysis of Industrial Structure Upgrading

While the baseline regression confirms the overall promotional effect of the green finance pilot on the composite Industrial Structure Upgrading Index (ISUI), relying solely on a composite measure may mask the specific structural channels through which the policy operates. To unpack these heterogeneous effects, we decomposed the ISUI into its three constituent sub-dimensions: Industrial Rationalization (R), Industrial Advancement (A), and Industrial Greening (G). The variables were appropriately normalized and winsorized at the 1% and 99% levels to prevent outlier bias. We subsequently re-estimated the baseline staggered DID model for each sub-dimension.

As reported in

Table 3. Regression results of sub-dimensions of industrial structure upgrading.

Variables	(1) Rationalization (R)	(2) Advancement (A)	(3) Greening (G)
L.Policy	0.0427 **	0.0056	0.0083
	(0.0210)	(0.0125)	(0.0099)
Controls	YES	YES	YES
Province FE	YES	YES	YES
Year FE	YES	YES	YES
Observations	420	420	420
Adjusted R2	0.9263	0.9604	0.9248

Note: ** denotes significance at the 5% level.

, the lagged green finance policy exerts a robust and statistically significant positive effect on Industrial Rationalization ($\mathsf{\beta }=0.0427, p<0.05$). Conversely, although the estimated coefficients for Industrial Advancement and Greening are directionally positive, they do not achieve statistical significance within the current sample period.

These divergent sub-dimensional results reveal a profound economic mechanism: In the initial phase of the green finance pilot, structural upgrading is fundamentally driven by the “Rationalization” channel. By strictly enforcing green credit constraints, the financial system rapidly curtails funding for backward and highly polluting capacities, forcing a swift correction of resource misallocation. However, “Advancement” (which requires profound technological leapfrogging) and “Greening” (which entails a fundamental overhaul of the macro-energy consumption structure) inherently possess much longer technological and investment cycles. Consequently, their quantitative transformations are still incubating and have not yet reached statistical significance. This asymmetrical sequence suggests that rapid rationalization serves as the foundational prerequisite for the subsequent long-term advancement and greening of the industrial structure.

4.4. Robustness Test

4.4.1. Event Study Approach

To examine the validity of the identifying assumption underlying the multi-period difference-in-differences strategy, this scientific study employs an event study specification that traces the dynamic effects of the Green Finance Reform and Innovation Pilot policy on industrial structure upgrading (ISUI). The policy implementation year is normalized to zero.

Figure 1 reports the estimated coefficients for a sequence of leads and lags relative to the policy adoption. In the pre-treatment period (years −4 to −1), the estimated parameters evaluate the cumulative abnormal deviations of the treated cohort relative to the control group, conceptually analogous to the evaluation of cumulative abnormal returns in event studies under standard distribution assumptions [38]. The pre-treatment coefficients are statistically indistinguishable from zero, as confirmed by joint F-tests, and their 95% confidence intervals safely encompass the zero line. This finding suggests that, prior to policy implementation, treated and non-treated provinces followed similar trajectories in terms of industrial structure upgrading, and the parallel trend assumption cannot be rejected.

After the introduction of the policy, the estimated coefficients turn positive and exhibit a gradual increase over time. While the immediate post-treatment effect is modest, the magnitude of the coefficients becomes larger in subsequent years, indicating that the policy impact unfolds progressively rather than instantaneously. Such a dynamic pattern is consistent with the institutional characteristics of green finance policies, whose effects typically materialize through delayed adjustments in credit allocation and investment behavior.

4.4.2. Placebo Test

To further assess whether the estimated policy effect is driven by unobserved shocks or spurious correlations, a placebo test was conducted based on random policy assignment. Specifically, 500 placebo policies were generated by randomly assigning both the treatment status (which provinces are treated) and the treatment timing (the specific implementation years), while strictly preserving the original cohort sizes to accurately mimic the actual staggered adoption structure. The baseline multi-period DID model was then re-estimated for each simulation.

Figure 2 visualizes the density curve of the parameters derived from these randomized simulations. The resulting distribution strictly concentrates around the zero-mean, confirming that fictitious policy assignments fail to trigger any systematic structural shifts. Crucially, our authentic benchmark estimate (0.036) is positioned at the extreme right extremity of this simulated spectrum, drastically deviating from the artificial norm. This pronounced divergence definitively confirms that our primary findings are robust against stochastic noise or unobserved synchronous confounders.

4.4.3. PSM-DID with Alternative Matching Methods

Addressing the endogenous assignment of the green finance experimental areas, we introduced a Propensity Score Matching–Difference-in-Differences (PSM-DID) framework to systematically purge potential selection bias. Specifically, we first employed a Logit model, utilizing all baseline regional control variables (such as economic development, human capital, and urbanization) as covariates to estimate the propensity scores for each province. After conducting balance tests to confirm that the common support assumption was strictly satisfied and that no systematic differences remained between the covariates of the treated and control groups, we deployed a triangulation of pairing protocols for the untreated provincial sample: optimal one-to-one neighbor calibration (Column 1), caliper-based radius delimitation (Column 2), and nonparametric kernel weighting (Column 3).

Table 4. PSM-DID with alternative matching methods.

	(1) 1:1 Nearest Neighbor	(2) Radius Matching	(3) Kernel Matching
	ISUI	ISUI	ISUI
L.Policy	0.041 **	0.038 **	0.039 **
	(2.939)	(2.781)	(2.855)
Control Variables	YES	YES	YES
Year FE	YES	YES	YES
Individual FE	YES	YES	YES
N	140	148	150

t statistics in parentheses. ** p < 0.05.

reports the PSM-DID estimation results. Across all three matching specifications, the estimated coefficients of the policy variable ($L.policy\_\mathrm{m}\mathrm{u}\mathrm{l}\mathrm{t}\mathrm{i}$) remain significantly positive at the 5% level. These consistent findings provide compelling evidence that, even after eliminating the observable initial systematic differences between the pilot and non-pilot provinces, the green finance policy still exerts a substantial and stable promoting effect on industrial structure upgrading. This robustly confirms the reliability of our baseline conclusions.

4.4.4. Heterogeneous Treatment Effects (CS-DID)

In a staggered difference-in-differences (DID) framework, the traditional two-way fixed effects (TWFE) estimator may suffer from “negative weighting” biases when treatment effects are heterogeneous across cohorts and time. To strictly address this potential endogeneity, we re-estimated the dynamic effects and the aggregated average treatment effect on the treated (aggregated ATT) using the Callaway and Sant’Anna robust estimator (CS-DID). This method computes group-time average treatment effects (ATT(g,t)) by comparing newly treated cohorts strictly against never-treated units, effectively circumventing the bias of using already-treated units as controls.

As reported in

Table 5. Robustness check using Callaway and Sant’Anna [39] CS-DID estimator.

Variables	(1) Aggregated ATT (CS-DID)
Treatment Effect (Aggregated ATT)	0.0229 **
	(0.0097)
Controls (Pre-Treatment Attributes)	YES
Observations	450

Note: Standard errors clustered at the province level are reported in parentheses. ** p < 0.05.

, the aggregated ATT estimated via the CS-DID methodology remains robustly positive and statistically significant (β = 0.0229, p < 0.01). This rigorous estimation conclusively confirms that, even after purging the potential bias introduced by heterogeneous treatment timing, the green finance pilot significantly promotes industrial structure upgrading. The consistency between the CS-DID aggregated ATT and the baseline TWFE estimates firmly establishes the reliability of our causal inference and dispels concerns regarding negative weighting biases.

4.4.5. Adding Additional Control Variables

To rigorously address potential omitted-variable bias and validate the stability of our baseline estimates, we conducted an additional robustness check by expanding our core control variable set. Following recent sustainable macro-finance literature, we introduced two crucial regional structural factors: Energy Structure (ES) and Financial Development (FD). Energy Structure (ES) is measured by the proportion of regional coal consumption in total energy consumption, capturing the region’s inherent industrial dependence on fossil fuels. Financial Development (FD) is proxied by the ratio of the added value of the financial industry to regional GDP, which controls for the baseline capacity of the local financial system.

The regression results with the augmented control variables are presented in

Table 6. Augmented model with additional control variables.

Variables	(1) Baseline	(2) With Additional Controls
L.Policy	0.0354 ** (0.0142)	0.0315 ** (0.0138)
ES		−0.0421 ** (0.0185)
FD		0.1250 *** (0.0412)
Core Controls	YES	YES
Province FE	YES	YES
Year FE	YES	YES
Observations	420	420
Adjusted R2	0.9552	0.9573

Note: Standard errors clustered at the province level are reported in parentheses. *** p < 0.01, ** p < 0.05.

. The results reveal that even after strictly controlling for variations in coal dependence and regional financial depth, the estimated coefficient of the green finance pilot policy (L.Policy) remains positive and highly significant (β = 0.0315, p < 0.05). Furthermore, the coefficient of Energy Structure (ES) is significantly negative, as expected, indicating that high coal dependence hinders structural upgrading, whereas Financial Development (FD) exerts a significant positive impact. The stability of our core policy coefficient firmly demonstrates that our main empirical findings are not driven by these structural omitted variables, thereby powerfully attesting to the robustness of the causal identification.

4.4.6. Alternative Measure of Environmental Regulation

To ensure that our findings regarding the moderating effect of environmental regulation are robust and not sensitive to the choice of measurement, we replaced the text-mining-based index with an outcome-based metric. Following the suggestions from the literature on regulatory enforcement, we employed Pollution Control Investment Intensity (ER_inv), measured by the ratio of regional industrial pollution control investment to regional GDP. While government report keyword frequency captures ex-ante policy signaling, the actual investment intensity rigorously reflects the ex-post stringency and execution of regulatory enforcement, effectively eliminating the interference of regional economic scale.

The re-estimated moderation model is reported in

Table 7. Robustness check: Alternative measure of environmental regulation (ER_inv).

Variables	(1) ISUI
L.Policy	0.0265 ** (0.0121)
ER_inv	0.0184 (0.0135)
L.Policy × ER_inv	0.0412 ** (0.0194)
Controls	YES
Province FE	YES
Year FE	YES
Observations	420
Adjusted R2	0.9561

Note: Standard errors in parentheses. ** p < 0.05.

. The results reveal that the coefficient of the interaction term (Policy × ER_inv) is 0.0412 and is statistically significant at the 5% level. This complementary finding strongly solidifies our theoretical framework: the efficacy of the green finance pilot policy is structurally amplified in regions with stringent actual environmental enforcement. This rules out the concern that our previous results might merely capture political rhetoric, fully corroborating the synergistic “administrative push” hypothesis.

4.5. Heterogeneity Analysis

4.5.1. Regional and Resource Endowment

Acknowledging the profound spatial heterogeneity characterizing China’s macro-economy—spanning diverse institutional capacities, financial depths, and structural endowments—we posit that the green finance mandate exerts non-uniform, asymmetric effects. Notably, the coastal eastern tier exhibits stark contrasts with the inland central and western territories regarding capital market sophistication and industrial baselines. To rigorously capture this spatial divergence, our empirical strategy partitions the provincial panel into eastern and inland (mid-western) cohorts, executing independent staggered DID estimations to unmask region-specific structural upgrading dynamics.

The heterogeneity analysis by region is summarized in Columns (1) and (2) of

Table 8. Regional and resource endowment heterogeneity analysis results.

	(1)	(2)	(3)	(4)
	East	Mid-West	Resource	Non-Resource
L.Policy	0.022 *	0.043 ***	0.052 **	0.019
	(1.906)	(3.195)	(2.601)	(1.604)
Control Variables	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Individual FE	YES	YES	YES	YES
N	154	266	112	308
Adjusted R2	0.990	0.882	0.833	0.971

t statistics in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

. The coefficient on the lagged green finance policy variable is positive in the eastern region but only marginally significant at the 10% level (β = 0.022, p < 0.10). In contrast, the estimated coefficient is substantially larger and highly significant in the central and western regions (β = 0.043, p < 0.01). These findings suggest that the green finance policy exerts a stronger promoting effect on industrial upgrading in less-developed regions.

A plausible explanation lies in the diminishing marginal returns of financial intervention. In eastern regions, where financial systems are relatively mature and industrial structures are already well-developed, the additional impact of green finance policies is inherently limited. By contrast, central and western regions typically face more severe financing constraints and greater structural rigidities, making them more responsive to policy-induced credit expansion and resource reallocation. From this perspective, green finance policies function as a compensatory mechanism that helps mitigate regional development imbalances.

Beyond regional disparities, differences in resource endowments may also condition the effectiveness of green finance policies. Resource-based regions tend to rely heavily on extractive industries, face higher environmental pressures, and encounter greater obstacles to industrial transformation. To explore this dimension, the sample is further divided into resource-based and non-resource-based regions.

Columns (3) and (4) of Table 8 show that the lagged policy coefficient is positive and statistically significant in resource-based regions (β = 0.052, p < 0.05), whereas it is smaller and insignificant in non-resource-based regions (β = 0.019). This indicates that green finance policies have a more pronounced effect on industrial upgrading in resource-dependent areas.

These results suggest that green finance plays a critical corrective role in resource-based regions by redirecting financial resources away from pollution-intensive and resource-dependent sectors toward cleaner and more technology-intensive industries, thereby accelerating structural transformation. In contrast, non-resource-based regions typically possess more diversified industrial bases and more market-driven upgrading dynamics, which reduces the marginal impact of external financial policy interventions.

4.5.2. Regional Technological Infrastructure

To further examine whether the effects of green finance policies vary with regional technological foundations, this scientific study conducts a heterogeneity analysis based on green technological innovation capacity. Specifically, provinces are divided into high green patent (High GP) and low green patent (Low GP) groups according to the median number of green patent authorizations, and staggered DID model is re-estimated for each subsample.

Table 9. Regional technological infrastructure heterogeneity analysis results.

	(1)	(2)
	High_GP	Low_GP
L.Policy	0.015	0.067 ***
	(1.198)	(5.233)
Control Variables	YES	YES
Year FE	YES	YES
Individual FE	YES	YES
N	221	197
Adjusted R2	0.963	0.916

Note: GP stands for green patents. t statistics in parentheses. *** p < 0.01.

(Specifications 1 and 2) exposes stark asymmetries contingent on regional innovation endowments. For provincial cohorts boasting advanced green patenting volumes, the treatment parameter of the delayed financial mandate remains statistically indistinguishable from zero ($\mathsf{\beta }=0.015$). Conversely, within territories exhibiting constrained eco-innovation capacities, the policy proxy demonstrates a robust, positive magnitude, passing the 1% significance threshold ($\mathsf{\beta }=0.067, p<0.01$).

These findings indicate that green finance policies exert a stronger promoting effect on industrial upgrading in regions with weaker technological foundations. A plausible explanation lies in the technology catch-up effect. Regions with lower green innovation capacity typically face more severe financing constraints and higher barriers to industrial transformation. Policy-induced expansion of green credit can effectively alleviate capital constraints, enabling firms to adopt cleaner technologies and advanced production processes, thereby accelerating industrial upgrading. By contrast, in regions with stronger green innovation bases, industrial upgrading relies more heavily on endogenous technological progress and market-driven mechanisms, which limits the marginal impact of external financial policy interventions.

4.5.3. Heterogeneity by Digital Finance Development

Amid the accelerated proliferation of the digital economy, technological integration has emerged as a pivotal structural determinant dictating the efficacy of macroeconomic financial instruments. Contemporary scholarship posits that digital financial frameworks are instrumental in alleviating information frictions, compressing transaction overheads, and optimizing the flow of credit. Consequently, these digital conduits magnify the transmission of regulatory shocks onto corporate decision-making and sectoral reorganization. Viewed through this lens, provincial territories endowed with superior digital ecosystems and mature digital financial networks are theoretically positioned to capture amplified dividends from green finance mandates.

However, the empirical results reported in

Table 10. The coordinated effect of green finance and digitalization.

	(1)	(2)	(3)	(4)
	Total_Index	Breadth	Depth	Sophistication
L.Policy	0.039 **	0.025	0.052 ***	0.065 **
	(2.135)	(1.460)	(3.326)	(2.230)
L.Policy × Adig	−0.015
	(−0.341)
L.Policy × Cdig		0.011
		(0.303)
L.Policy × Ddig			−0.042
			(−1.086)
L.Policy × Sdig				−0.085
				(−0.988)
Control Variables	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Individual FE	YES	YES	YES	YES
Observations	390	390	390	390
Adjusted R2	0.957	0.957	0.956	0.957

t statistics in parentheses. ** p < 0.05, *** p < 0.01.

do not support this hypothesis. Neither the overall digitalization index (Adig) nor its sub-dimensions—digital coverage (Cdig), depth of use (Ddig), and degree of digitalization (Sdig)—exhibit statistically significant interaction effects with the green finance policy variable. Rather than implying a failure of the policy, this non-significant moderating effect highlights the policy’s broad inclusiveness and its capacity to bridge the regional “digital divide”.

Specifically, the effectiveness of the green finance policy is not bounded by pre-existing digital financial infrastructure. In digitally lagging regions—where severe information asymmetries typically hinder upgrading—the green finance pilot acts as a crucial compensatory mechanism. By leveraging state-backed green credit channels, the policy effectively substitutes for the lack of digital financial penetration. Interpreted alongside the earlier findings of stronger effects in low-patent regions, this robust response in digitally disadvantaged areas points to a clear catch-up effect. Overall, these findings suggest that China’s green finance policy successfully overrides digital divide constraints, ensuring a more equitable process of industrial transformation.

4.6. Mechanism Analysis

Drawing upon existing theoretical and empirical studies, this scientific study selects Green Credit Intensity (GCI) as the primary mediating channel through which green finance policies affect industrial structure upgrading, and treats environmental regulation intensity as a critical moderating factor. Green finance policies mainly operate through the financial system by structurally guiding the allocation of credit resources, while environmental regulation influences the policy effect by reshaping investment returns and the institutional environment, thereby amplifying or constraining the transmission of financial incentives.

Accordingly, this scientific study incorporates Green Credit Intensity (GCI) into the mediation analysis to examine whether green finance policies promote industrial upgrading not merely by expanding total liquidity, but precisely by deepening the penetration of eco-friendly capital within the regional economy. At the same time, environmental regulation (ER) is introduced as a moderating variable to assess whether regulatory pressure strengthens the effectiveness of green finance policies through an administrative enforcement and policy-signaling mechanism.

Figure 3 outlines the “push–pull” synergistic governance and inclusive catch-up framework of this scientific study. The core pathway demonstrates how the lagged green finance pilot policy (L.Policy) drives industrial upgrading (ISUI) through a dual mechanism: the financial pull of Green Credit Intensity (H2) and the administrative push of environmental regulation (H3). Furthermore, the framework integrates two boundary conditions: a distinct catch-up effect in resource-dependent and low-patent regions (top right), and the broadly inclusive background of digital finance (bottom), highlighting the policy’s independence from regional digital infrastructure.

Guided by this framework, the subsequent sections empirically test the mediating effect of Green Credit Intensity and the moderating effect of environmental regulation in a sequential manner.

4.6.1. Mediating Effect of Green Credit Intensity

To uncover the transmission mechanism through which the Green Finance Reform and Innovation Pilot policy affects industrial structure upgrading, this scientific study examines the mediating role of Green Credit Intensity (GCI). The corresponding empirical results are reported in Columns (1)–(3) of

Table 11. Mediating effect of GCI and moderating effect of ER.

	(1)	(2)	(3)	(4)
	ISUI (Total)	GCI (Path a)	ISUI (Path b)	Moderation: ER
L.Policy	0.035 ***	0.042 ***	0.028 **	0.038 ***
	(2.772)	(3.344)	(2.151)	(2.924)
GCI			0.156 **
			(2.190)
ln_ER				0.005
				(0.627)
Policy × LnER				0.015 **
				(2.142)
Control Variables	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
Individual FE	YES	YES	YES	YES
N	420	420	420	420
Adjusted R2	0.956	0.893	0.957	0.955

t statistics in parentheses. Standard errors are clustered at the province level. ** p < 0.05, *** p < 0.01.

.

Column (1) demonstrates that the one-period lagged green finance policy variable has a significantly positive effect on industrial structure upgrading, which is consistent with the baseline regression results. Column (2) indicates that the policy variable significantly enhances Green Credit Intensity (GCI), suggesting that the green finance reform effectively redirects financial resources and deepens the penetration of eco-friendly capital in the real economy of pilot regions. In Column (3), when simultaneously including both the policy variable and GCI, the mediating variable (GCI) exhibits a robust and significant positive association with industrial structure upgrading. Concurrently, the coefficient of the policy variable decreases in magnitude. This pattern provides compelling preliminary evidence that the structural optimization of green credit serves as a vital transmission channel through which green finance policies facilitate industrial upgrading.

To rigorously verify the stability of this estimated mediation effect, both the Sobel test and a nonparametric percentile bootstrap approach were employed. The Sobel test yields a Z-statistic of 2.1842, which is significant at the 5% level (p < 0.05), confirming that the “policy–GCI–industrial upgrading” pathway is statistically valid. Furthermore, the Bootstrap test, based on 1000 resampling iterations, reveals that the 95% confidence interval of the indirect effect ranges from [0.0018, 0.0135]. Because this confidence interval strictly excludes zero, it provides robust additional evidence that Green Credit Intensity plays a critical and significant mediating role in the policy’s impact on structural economic transformation.

4.6.2. The Moderating Role of Environmental Regulatory Pressure

Expanding upon our fundamental estimations, this scientific study investigates the contingent role of environmental regulatory stringency in shaping the efficacy of green financial mandates. Model (4) in Table 11 documents these interactive dynamics, complemented by the marginal effect trajectories visualized in Figure 4. The empirical outputs reveal that regulatory strictness lacks a statistically validated standalone impact on structural advancement. Crucially, however, the multiplicative term between environmental oversight and the policy proxy yields a robust positive coefficient. This dynamic implies that stringent environmental regulation acts not as a solitary catalyst, but rather as a critical complementary force that amplifies the transmission efficacy of green credit interventions.

The marginal effect plot presented in Figure 4 provides further insights into this moderating relationship. As environmental regulation intensity increases, the marginal impact of green finance policies on industrial structure upgrading rises steadily. When environmental regulation is relatively weak, the estimated policy effect is limited and statistically insignificant. In contrast, in regions with stricter environmental regulation, the promoting effect of green finance policies on industrial upgrading becomes substantially stronger.

These results imply that environmental regulation serves as an important institutional complement to green finance policies. By increasing regulatory pressure and strengthening policy signals, environmental regulation creates a governance environment in which green finance instruments can more effectively guide financial resources toward cleaner and higher-value-added industries, thereby amplifying their impact on industrial structure upgrading.

5. Discussion

Employing a staggered difference-in-differences (DID) estimation strategy, this scientific study rigorously quantifies the causal impact of green finance policies on industrial upgrading. Furthermore, we examine the underlying transmission mechanisms and the regional heterogeneity of this structural transition. Overall, the empirical results provide robust evidence for the proposition that “green finance can effectively promote industrial structure upgrading”. Drawing upon the research hypotheses and existing literature, the findings are discussed in depth below.

5.1. The “Push–Pull” Mechanism: The Synergistic Logic of a Multidimensional Governance System

The analysis demonstrates that environmental regulation (ER) enhances the performance of green finance policies by reinforcing their intended effects, verifying the synergistic effect of administrative and financial instruments in industrial transformation. Mechanistically, this phenomenon can be summarized as a “push–pull” logic. On the one hand, environmental regulation, measured by the frequency of related keywords in government work reports, releases strong administrative signals and creates “Exit Pressure” for highly polluting industries. This resonates with the classic “Porter Hypothesis” and its derivative studies in the financial context [40,41]. On the other hand, the green finance policy significantly lowers the “Entry Barrier” for green industries through the expansion of Green Credit Intensity (GCI), providing endogenous motivation. As pointed out by Chen et al. [1], the precise guidance of financial resources is crucial for achieving structural transformation. This scientific study further demonstrates that the efficiency of resource allocation can only be maximized when the “push” of administrative regulation and the “pull” of financial credit form a cohesive synergy.

5.2. The Inclusiveness of Digital Finance: A Path to Overcoming Technological Barriers

Our estimations reveal that the moderating effect of digital finance is statistically insignificant. While previous studies often emphasize FinTech’s catalytic role in ecological governance [42,43], our findings highlight the structural inclusiveness of green finance policies. Specifically, policy effectiveness does not rigidly depend on regional digital infrastructure. For developing regions with nascent digital infrastructure, this offers an encouraging signal: green macroeconomic transition is not strictly limited by technological preconditions. Instead, targeted green credit allocation and well-calibrated regulations can effectively bypass digital bottlenecks.

5.3. Technological Catch-Up: Financial Solutions for “Technological Lock-In”

The policy impact is found to be more evident in resource-dependent provinces and in regions with lower levels of green patent authorizations, highlighting notable regional heterogeneity. This indicates that green finance plays a compensatory role, providing “timely assistance” in these areas. For a long time, resource-dependent regions have faced severe “path dependency” and the “technological lock-in effect” [44,45,46,47]. This scientific study proves that green finance, through targeted support, helps these technologically disadvantaged regions bridge the gap of initial capital accumulation. This not only supports the perspective that finance can compensate for technological shortcomings but also highlights the unique social value of green finance policies in narrowing regional transformation gaps and promoting coordinated regional development.

5.4. Limitations and Future Research Directions

Finally, this scientific study acknowledges certain limitations that provide valuable avenues for future research. As our empirical design relies on provincial macro-level data, it may be susceptible to ecological fallacies regarding individual corporate decision-making. Future studies should utilize firm-level micro-datasets to unpack the granular behavioral responses of enterprises to these macro-financial policies. Furthermore, while this study provides robust evidence for the efficacy of China’s state-guided “push–pull” model, generalizations to fully market-driven environments (such as the EU framework) should be made with caution, accounting for distinct institutional and regulatory contexts. Lastly, modeling the long-term effectiveness of green finance under varying macroeconomic conditions remains an open challenge. Future research should consider integrating regime-switching analysis and inflation-driven beta dynamics to better capture how green financial “pulls” operate across different economic cycles [48].

6. Conclusions and Policy Implications

Drawing on a provincial panel dataset from 2009 to 2023, this scientific study investigates the impact of China’s green finance pilot zones on industrial structure upgrading. Employing a staggered difference-in-differences (DID) framework, we systematically evaluate both the direct structural effects and the underlying transmission mechanisms. The empirical analysis yields four salient conclusions:

First, the green finance pilot policy provides robust empirical support for industrial structure upgrading (confirming Hypothesis 1). This positive structural effect remains highly consistent across multiple robustness specifications, including the CS-DID estimator, indicating a reliable causal relationship.

Second, Green Credit Intensity (GCI) acts as a primary mediating channel (confirming Hypothesis 2). Rather than relying on general credit expansion, the policy effectively directs targeted financial resources toward sustainable sectors, thereby alleviating financing constraints for green transformation.

Third, environmental regulation exerts a significant positive moderating effect (confirming Hypothesis 3). Stringent local environmental oversight complements the financial incentives, creating a synergistic “push–pull” dynamic that accelerates regional industrial upgrading.

Fourth, the moderating role of digital finance is not statistically significant (accepting Hypothesis 4). This finding underscores the broad inclusiveness of the green finance framework, as its effectiveness does not strictly depend on pre-existing digital infrastructure. Consequently, the policy facilitates a structural catch-up effect, conferring particular benefits to resource-dependent regions and areas with historically lower innovation capacities.