Industrial Structure, Green Finance, and Energy Resilience Enhancement in China

Abstract

Against the backdrop of global energy transition and multiple uncertainties, enhancing energy resilience has become a core priority for China’s pursuit of secure and sustainable development. Using Chinese provincial panel data from 2011 to 2019, this study applies a two-way fixed effects model, mediation effect tests, and interaction term analysis to empirically investigate the relationship between industrial structure, green finance, and energy resilience. The main findings are as follows. First, the increases in gross regional product (GRP) and the added value of the secondary and tertiary sectors significantly enhance energy resilience. Second, heterogeneity analysis indicates that in regions with a high level of green finance, both GRP and the secondary sector’s added value exhibit stronger positive effects on energy resilience, whereas in regions with lower levels of green finance, the tertiary sector’s added value contributes more significantly to energy resilience improvement. In areas with high coal dependency, the secondary sector’s added value shows a significantly positive effect on energy resilience. Increases in industrial and construction industry added value significantly enhance energy resilience, suggesting that the expansion of the secondary industry contributes positively to the stability and resilience of the energy system. Third, the mechanism analysis shows that green finance contributes to energy resilience partly through the optimization of the energy consumption structure. Specifically, by effectively curbing coal consumption and, to a lesser extent, fuel oil production, green finance reduces the structural dependence of the economy on high-carbon energy. By contrast, channels such as electricity generation yield weaker and less robust evidence. These findings suggest that energy resilience is fundamentally shaped by the interplay of industrial structure, financial intermediation, and energy structure adjustment. Therefore, policy should shift from single instruments to integrated governance, synergizing industrial policy, green finance, and energy optimization to bolster energy resilience.

1. Introduction

Against the backdrop of global energy market volatility, geopolitical uncertainty, and the pressures of low-carbon transition, enhancing energy resilience (ER) has become a critical policy challenge for countries, especially with large economies [1,2]. Energy resilience not only refers to the ability to maintain the basic stability of energy supply in the face of shocks but also encompasses the system’s capacity to absorb disturbances, recover efficiently, and adapt to long-term structural transformations [1,3]. For China, a country characterized by uneven regional development and a complex energy structure, understanding and strengthening energy resilience is essential in achieving the synergistic advancement of its dual carbon goals and energy security.

The existing literature has explored the factors influencing energy resilience from various perspectives. A body of research emphasized the foundational role of industrial structure (IS) optimization. It has been argued that upgrading the industrial structure from high-energy-consuming traditional sectors to technology-intensive and service-led industries enhances robustness and adaptability to various shocks by reducing overall energy intensity and increasing the flexibility of the economic system [4,5,6]. Another strand of literature has delved into the important role of green finance (GF). Studies indicate that green finance not only provides critical funding for green technology innovation and clean energy projects [7,8], but also promotes the transformation and optimization of high-carbon industries through capital reallocation, thereby positively impacting the long-term resilience and transition capacity of the energy system [9,10].

Despite the substantial progress made in the aforementioned studies, research that systematically examines industrial structure, green finance, and energy resilience within an integrated framework remains relatively scarce, leaving the following three notable research gaps.

First, the role of green finance in the relationship between industrial structure and energy resilience remains insufficiently clarified. In practice, green finance may influence energy resilience both by supporting cleaner industrial transformation and by reshaping energy production and energy consumption patterns through capital allocation [11,12]. However, although existing studies have shown that green finance can promote industrial upgrading, optimize capital allocation, and support low-carbon transformation [11,12,13,14], they have not sufficiently examined how the green finance dimension conditions the industrial structure–energy resilience relationship, nor have they clearly distinguished whether the evidence points to moderation, selected mechanism channels, or a complete mediation chain. Second, most existing studies have failed to adequately distinguish between the different impacts of industrial added value (scale effect) and industrial structure share (proportional effect) on energy resilience when analyzing industrial structure. These two dimensions differ fundamentally in their economic implications and operational mechanisms, and the failure to distinguish between them may lead to misinterpretations of policy priorities. Finally, regional heterogeneity has not been sufficiently examined. Significant differences in green finance foundations and coal dependency across regions may lead to differentiated resilience effects, which cannot be fully captured by national-level average estimates.

To address the aforementioned research gaps, this paper develops an integrated theoretical framework that links industrial structure upgrading, green finance, and energy resilience and conducts an empirical investigation using Chinese provincial panel data from 2011 to 2019. This study aims to answer the following three core research questions: (1) Does industrial structure upgrading, including increases in the added value of various industries and changes in structural proportions, significantly enhance energy resilience at the provincial level in China? (2) If so, through what channels and under what conditions does green finance shape the relationship between industrial structure and energy resilience? (3) Does this effect exhibit significant industrial heterogeneity, as well as differences based on the level of green finance development and variations in energy structure?

The contributions of this paper can be summarized in three aspects. First, this paper introduces an industrial structure perspective and examines how sectoral value-added upgrading and industrial share changes are associated with energy resilience. Second, the paper distinguishes between industrial-scale effects and structural-share effects and further examines whether green finance moderates the relationship between industrial development and energy resilience. Third, the paper explores selected energy structure channels through which green finance may be related to energy resilience, with particular attention to coal consumption and fuel oil production. These contributions extend the application of existing ER/GF indicators to the industrial structure context and provide policy evidence for differentiated industrial–financial coordination in China.

The remainder of the paper is organized as follows: Section 2 presents the literature review and research hypotheses. Section 3 introduces the data, variables, and model specification. Section 4 reports the baseline regression results, robustness checks, and heterogeneous effects. Section 5 presents the mechanism analysis. Section 6 concludes the paper.

2. Literature Review and Research Hypotheses

2.1. Literature Review

2.1.1. Industrial Structure and Energy Resilience

The literature establishes a robust link between industrial structure and energy resilience, positing that structural upgrading—shifting from energy-intensive, low-value-added sectors toward technology-intensive and service-oriented industries—serves as a critical pathway to enhance systemic resilience. This transition strengthens resilience by reducing overall energy intensity and dependence on single energy sources, thereby improving the system’s robustness to withstand shocks [4,5]. Empirical studies, such as those on smart city policies, confirm that industrial advancement, often driven by technological innovation, directly boosts energy resilience by optimizing resource allocation and fostering cleaner production [6,15]. Furthermore, the effect exhibits heterogeneity, being more pronounced in resource-based cities where restructuring is crucial for long-term stability [16]. The synergy between industrial upgrading and green finance is also evident, as green credit can incentivize structural shifts toward less energy-vulnerable sectors, creating a virtuous cycle that enhances adaptive capacity [9]. Ultimately, industrial structure is not a static background but a dynamic lever, where its optimization, particularly through green finance mechanisms, fundamentally shapes the energy system’s ability to anticipate, absorb, and recover from disruptions.

2.1.2. Green Finance and Energy Resilience

Green finance enhances energy resilience through macro-, micro-, and policy-coordination mechanisms. At the macro level, its implementation positively correlates with energy resilience indices, with the core pathway being systematically strengthening the energy system’s shock resistance through green technological innovation [7] and industrial structure upgrading [8,10]. Micro-level policy evaluations confirm that instruments such as green credit [9] and green finance pilot schemes [17] provide evidence of significant effects by alleviating resource misallocation, driving innovation, and optimizing industries. Synergies with fintech further amplify these outcomes [18]. The impact of green finance on China’s energy resilience exhibits regional heterogeneity, being more pronounced in eastern and western regions [8,17], and it is constrained by factors such as the financial environment and marketization levels [9]. Cutting-edge research indicates that the convergence of green finance with artificial intelligence [19,20] and digital infrastructure [21] jointly propels the evolution of energy systems towards digitalization, intelligence, and inclusivity by providing financial support and shaping new business models.

2.1.3. Green Finance and Industrial Structure

Green finance drives industrial structure upgrading through capital reallocation—via green credit, bonds, and insurance—channeling funds from high-pollution to low-carbon, tech-intensive sectors, thereby shifting production from labor- to capital/knowledge-intensive models [11,12,22]. Green technological innovation (GTI) serves as a core intermediary—green finance eases financing constraints, lowers R&D costs, and fosters both clean technologies in emerging industries and efficiency gains in traditional sectors [23,24,25]. A synergistic effect exists whereby green finance and GTI mutually reinforce industrial structure upgrading [26]. Institutional factors, particularly voluntary environmental regulation, enhance this synergy, while coherent policy frameworks combining stringent targets with green finance mechanisms accelerate industrial greening [27,28]. Green finance effects exhibit marked regional heterogeneity and spatial spillovers, its upgrading impact is stronger in central and western China than in the east [13,14]. Green finance policies may generate negative spillovers on neighboring non-pilot areas, necessitating regional coordination [29]. Functionally, green finance drives industrial upgrading through value chain ascent in the east, and industrial rationalization via intersectoral coordination in the central and western regions [12]. Government intervention exhibits a dual role, with moderate support enhancing green finance impacts and excessive intervention causing market distortions [30]. Emerging research links green finance with the digital economy and carbon finance, suggesting that digitalization enhances green finance’s role in upgrading by improving information flow and technology diffusion [31]. carbon finance provides price signals, with green finance enabling compliance and transition [32]. Green finance also mediates the link between industrial restructuring and renewable energy by overcoming project financing barriers [10]. Evidence from Vietnam confirms green finance’s contribution to green recovery, but its effectiveness depends on phased industrial policy beyond green finance alone [33]. Finally, green finance pathways are mechanism-specific. While green finance improves energy efficiency through green technology innovation, industrial structure upgrading is not a significant channel in this context [34]. Conversely, green finance facilitates low-carbon urban transformation via industrial restructuring, particularly when technology channels alone are insufficient [35].

2.1.4. Research Gaps and Theoretical Positioning

Although the existing literature has explored the relationship between green finance, industrial upgrading, and energy security from different perspectives, there remains a significant gap in systematic research on the intrinsic connections among industrial structure, green finance, and energy resilience. The specific research gaps are reflected in the following three aspects.

First, the structural dimensions are inadequately distinguished. The existing literature fails to clearly differentiate between the scale effect (value added) and the structural effect (share proportion) of industries when analyzing the impact of industrial structure. The three major industries differ significantly in energy intensity, carbon dependency, and technological adaptability. The mechanisms and intensity of their value-added expansion and share changes on energy resilience may vary substantially. Conflating these two dimensions may lead to misjudgments in policy focus.

Second, regional heterogeneity is insufficiently considered. China’s regions vary significantly in developmental stages, resource endowments (particularly in coal dependency), and the foundation of green finance. Nevertheless, existing studies lack systematic examination of heterogeneity based on green finance development levels and energy structure (e.g., high/low coal dependency). Such oversight may obscure regionally differentiated policy needs under national-level average conclusions.

Third, the transmission mechanism linking industrial structure and energy resilience remains underexplored. Existing studies have largely focused on the direct environmental effects of green finance or its macro-level relationship with energy resilience while paying limited attention to the role of industrial structure in enhancing the capacity of the energy system to withstand, recover from, and adapt to shocks. More importantly, green finance may not affect energy resilience only through a direct channel but also by moderating the impact of industrial structure optimization on energy resilience. However, this moderating mechanism has not been sufficiently tested in the existing literature.

In response to these gaps, this study extends the existing analytical framework on green finance and energy resilience by incorporating industrial structure variables. By simultaneously investigating the value added and structural shares of the three major industries, the moderating and mechanism-related roles of green finance, and regional heterogeneity based on green finance levels and coal dependency, this paper aims to shift the research focus from a narrow emphasis on environmental outcomes to a broader discussion of structural transformation and systemic resilience. It seeks to provide supplementary empirical evidence on how China can enhance systemic resilience through industry–finance linkages in the context of energy transition.

2.2. Theoretical Analysis and Research Hypotheses

Based on the literature review and in the context of China’s energy transition and industrial restructuring, this study constructs a theoretical analytical framework integrating industrial structure–green finance–energy resilience. The core logic of this framework is that industrial structure optimization serves as the fundamental driver in shaping the resilience of the energy system, while green finance plays a dual role as an intermediary channel and an enabling amplifier in this process. The specific pathways of influence involve direct effects, indirect mechanisms, moderating effects, and heterogeneous manifestations, which ultimately lead to enhanced resilience through the core transmission channel of energy structure optimization. Corresponding research hypotheses are proposed around this framework as follows.

Industrial upgrading implies the reallocation of production factors from traditional sectors characterized by high energy consumption and high carbon dependency to technology-intensive, more efficient modern manufacturing and service industries. This transformation can directly enhance the energy system’s ability to withstand external shocks and adapt to long-term transitions by improving overall economic efficiency, promoting technological innovation, and fostering cleaner production [4,15]. Specifically, the green transformation and scale expansion of the secondary industry (particularly industry and construction) and the quality improvement and efficiency enhancement of the tertiary industry (services) are expected to be the main sources of resilience enhancement [6]. Meanwhile, coordinated optimization of the industrial share structure (i.e., the proportions of the three major industries) may help reduce excessive dependence on single high-carbon sectors, thereby enhancing systemic diversity to absorb disturbances.

H1. 

Industrial structure upgrading can significantly enhance energy resilience, but the roles of different industrial sectors and structural dimensions vary.

At the same time, green finance may play a critical supporting role in transforming the resilience effects of industrial upgrading into more durable and system-wide outcomes. By reallocating capital toward cleaner production, green technology, and low-carbon infrastructure, green finance can ease financing constraints for industrial transformation while raising the cost of high-carbon activities. This implies that green finance may affect energy resilience not only by directly promoting energy structure optimization, but also by strengthening the capacity of industrial upgrading to generate resilience gains.

H2. 

Green finance is expected to shape the relationship between industrial upgrading and energy resilience through two empirically distinguishable pathways: an energy structure channel and a moderating channel.

For analytical clarity, this dual role of green finance can be separated into an energy structure channel and a moderating channel, which are articulated as H2a and H2b, respectively.

H2a. 

Green finance may improve energy resilience partly through energy structure adjustment. By easing financing constraints for cleaner production and low-carbon transition while increasing the relative cost of carbon-intensive activities, green finance is expected to reduce dependence on high-carbon energy, especially coal, and thereby strengthen systemic resilience [8,10].

H2b. 

The level of green finance development positively moderates the marginal effect of industrial upgrading on energy resilience. In regions with well-developed green finance systems, industrial upgrading activities can more readily access low-cost financial support for technological upgrades and equipment retrofitting, thereby amplifying the resilience-enhancing effect of industrial upgrading.

However, the resilience effects of industrial restructuring and green finance are unlikely to be uniform across regions. China’s provinces differ markedly in development stage, financial depth, industrial composition, and energy endowment, which means that the same upgrading strategy may generate very different resilience outcomes under different regional conditions. In particular, differences in green finance foundations affect the availability of low-cost capital for green transformation, while differences in coal dependence shape the urgency and marginal benefits of structural adjustment.

H3. 

The effects of industrial structure and green finance on energy resilience exhibit significant heterogeneity, which is particularly evident in the differences in green finance foundations and energy structure.

Accordingly, the heterogeneity hypothesis can be further specified along the two contextual dimensions most relevant to this study: the level of green finance development and the structure of coal dependence.

H3a. 

In regions with a high level of green finance development, the expansion of GRP and the value added of the secondary industry can be more effectively translated into energy resilience, as robust financial support accelerates their greening processes. In contrast, in regions with a lower level of green finance, the contribution of the tertiary industry’s added value to resilience may be more pronounced, as its development relies relatively less on traditional heavy-asset financial support.

H3b. 

In regions with high coal dependency, the green transformation of the secondary industry, especially the industrial sector, is crucial for the marginal contribution to energy resilience, as it directly addresses the core structural contradiction of coal reduction. In regions with low coal dependency, high-quality economic development and upgrading of the tertiary industry play a more significant role in enhancing resilience.

Beyond its enabling and moderating roles, green finance may also influence energy resilience through a more concrete transmission channel: the optimization of the energy structure. By altering the relative financing costs of high-carbon and low-carbon activities, green finance can constrain coal-intensive production and consumption while facilitating cleaner and more diversified energy use. If this channel holds, improvements in energy resilience should be closely associated with reductions in dependence on high-carbon energy, especially coal. Based on the above mechanism-based reasoning, the following hypothesis H4 is proposed.

H4. 

The core mechanism by which green finance enhances energy resilience lies in optimizing the energy structure, particularly by curbing the consumption of high-carbon energy, especially coal.

Green finance influences resilience not only indirectly by supporting industrial upgrading but also directly by constraining and guiding investment and financing in the energy sector, thereby driving the transformation toward cleaner energy on both the supply and demand sides. Its core mechanism of action involves increasing the financing costs and environmental risks associated with high-carbon projects while easing financing constraints for clean energy initiatives. This may effectively curb the production and consumption of traditional high-carbon energy sources, such as coal, promote the transition of the energy structure toward low-carbon and diversified development, and ultimately enhance the system’s adaptability and risk resistance at a fundamental level [9].

In summary, the theoretical hypotheses of this study link industrial structure, green finance, and energy resilience through direct associations, moderation, and selected energy structure mechanisms.

3. Data, Variables, and Model Specification

3.1. Data

Due to data availability, this study uses a panel dataset covering 29 Chinese provinces from 2011 to 2019, excluding Xinjiang and Tibet. Data were primarily obtained from official statistical publications. Specifically, data for GRP, PRI_VA, SEC_VA, TER_VA, PRI_SH, SEC_SH, TER_SH, POP, WAGE, and TRA were sourced from or calculated based on data from the National Bureau of Statistics (NBS). Additionally, the ER and GF indices were directly adopted from Nepal et al. [8,19].

Table 1. Descriptive statistics.

Variables	Definition	N	Mean	SD	Min	Max
lnER	The logarithm of energy resilience	261	−1.271	0.279	−1.853	−0.558
lnGRP	The logarithm of regional gross domestic product	261	9.786	0.886	7.237	11.612
lnPRI_VA	The logarithm of the value added of the primary sector	261	7.176	1.068	4.662	8.483
lnSEC_VA	The logarithm of the value added of the secondary sector	261	8.863	0.953	6.317	10.575
lnTER_VA	The logarithm of the value added of the tertiary sector	261	9.069	0.906	6.527	10.762
PRI_SH	The proportion of the primary sector’s added value within the gross domestic product	261	0.096	0.052	0.003	0.252
SEC_SH	The proportion of the secondary sector’s added value within the gross domestic product	261	0.408	0.080	0.17	0.588
TER_SH	The proportion of the tertiary sector’s added value within the gross domestic product	261	0.496	0.089	0.333	0.825
lnPOP	The logarithm of population density	261	7.874	0.411	6.64	8.669
lnWAGE	The logarithm of wage levels	261	7.875	0.865	5.663	9.649
lnTRA	The logarithm of foreign trade levels	261	11.225	1.562	7.003	14.311
lnGF	The logarithm of green finance	261	−1.774	0.430	−2.779	−0.232

presents the descriptive statistics of the main variables.

3.2. Variables

3.2.1. Key Variables

Energy resilience ($\mathrm{E}\mathrm{R}$) is used as the dependent variable capturing the stability, recovery capacity, and adaptive capability of the regional energy system under shocks and transition pressures. Following Nepal et al. [8], we directly adopt the established entropy-weighted ER index and weights (In measurement terms, this study directly adopts the energy resilience indicator system, dimensional classification, and published indicator weights reported in Nepal et al. [8,19] and uses that established index framework as the measurement basis for the present empirical analysis. Equations (1)–(7) are presented solely to illustrate the measurement logic of the adopted indices, not to represent a newly constructed index).

For positive indicators:

$$z_{ij}={\displaystyle \frac{x_{ij-min\left(x_j\right)}}{max\left(x_j\right)-min\left(x_j\right)}}$$

(1)

For negative indicators:

$$z_{ij}={\displaystyle \frac{max\left(x_j\right)-x_{ij}}{max\left(x_j\right)-min\left(x_j\right)}}$$

(2)

Here, $\mathrm{i}$ denotes the province, $\mathrm{j}$ denotes the year, ${\mathrm{x}}_{\mathrm{i}\mathrm{j}}$ represents the initial value, $\mathrm{m}\mathrm{a}\mathrm{x}\left({\mathrm{x}}_{\mathrm{j}}\right)$ represents the maximum value, and $\mathrm{m}\mathrm{i}\mathrm{n}\left({\mathrm{x}}_{\mathrm{j}}\right)$ represents the minimum value.

Step 1. Calculate the weight ${\mathrm{q}}_{\mathrm{i}\mathrm{j}}$:

$$q_{ij}={\displaystyle \frac{z_{ij}}{{\sum }_{i=1}^m{z}_{ij}}},0\le q_{ij}\le 1$$

(3)

Step 2. Calculate the entropy value ${\mathrm{H}}_{\mathrm{j}}$:

$$H_j=-{\displaystyle \frac{1}{lnm}}{\displaystyle {\sum }_{i=1}^m}{q}_{ij}^{*}ln{q}_{ij}$$

(4)

Step 3. Calculate the coefficient of variation ${\mathrm{V}}_{\mathrm{j}}$: The coefficient of variation measures the degree of influence of an indicator on the research subject. A larger coefficient of variation indicates that the indicator has a greater impact on the research subject.

$$V_j=1-H_j$$

(5)

Step 4. Calculate the weight $G_j$:

$$G_j={\displaystyle \frac{V_j}{{\sum }_{i=1}^m{V}_j}}$$

(6)

Step 5. Calculate the energy resilience index ${ER}_i$:

$${ER}_i={\displaystyle {\sum }_{i=1}^n}{y}_{ij}^{*}{G}_{j }$$

(7)

Here, $\mathrm{n}$ represents the number of sub-indicators. A larger ${\mathrm{E}\mathrm{R}}_{\mathrm{i}}$ value indicates a higher level of energy resilience for province $\mathrm{i}$.

3.2.2. Other Variables

Regional gross domestic product (GRP) reflects the final output of economic activities in a region over a specific period. Value added of the primary industry (PRI_VA) refers to the final output, measured at market prices, generated by all resident units of a country (or region) engaged in production activities of the primary industry over a specific period. The primary industry includes agriculture, forestry, animal husbandry, and fishery (excluding supporting services in these sectors). Value added of the secondary industry (SEC_VA) refers to the final output, measured at market prices, generated by all resident units of a country (or region) engaged in production activities of the secondary industry over a specific period. The secondary industry includes mining (excluding mining support activities), manufacturing (excluding repair services of metal products, machinery, and equipment), production and supply of electricity, heat, gas, and water, and construction. Value added of the tertiary industry (TER_VA), refers to the final output, measured at market prices, generated by all resident units of a country (or region) engaged in production activities of the tertiary industry over a specific period. The tertiary industry includes all other sectors not classified under the primary or secondary industries. GRP, PRI_VA, SEC_VA and TER_VA are denominated in 100 million RMB.

The composition of the three sectors refers to the proportion of their value added in GDP, calculated at current prices. Primary sector share (PRI_SH) denotes the proportion of the primary sector’s added value within the gross domestic product (GDP). Secondary sector share (SEC_SH) denotes the proportion of the secondary sector’s added value within the GDP. Tertiary sector share (TER_SH) denotes the proportion of the tertiary sector’s added value within the GDP.

The impact of population density (POP) on energy resilience remains unclear. On the one hand, densely populated areas are more conducive to deploying smart grids and distributed energy systems; moreover, concentrated energy provision in high-density regions can reduce per-unit transmission costs and improve energy-use efficiency. On the other hand, high-density cities typically exhibit elevated cooling loads, which can intensify pressures on the energy system. We measure urban population density as the year-end registered population divided by the built-up area, expressed in persons per square kilometer.

The impact of wage levels (WAGE) on energy resilience is twofold. On the one hand, higher wages generally indicate stronger household purchasing power and higher living standards, which may drive demand for energy-related products such as smart homes, electric vehicles, and energy-efficient appliances, thereby promoting the upgrading of energy consumption structures and improving system efficiency. On the other hand, higher wages may also be associated with increased total energy consumption and peak loads, posing potential pressure on the stability and supply security of the energy system. We measure wage levels using the total wage of urban employed persons, expressed in ten thousand yuan.

The impact of foreign trade levels (TRA) on energy resilience is complex. On the one hand, deep participation in global trade helps achieve diversification of energy import sources, reducing dependence on a single region or country, thereby enhancing the security and resilience of energy supply. On the other hand, foreign trade, especially an export-oriented economy acting as the world’s factory, significantly increases domestic energy consumption and carbon emissions, making the national energy system more vulnerable to international market price fluctuations and geopolitical risks. We measure the level of foreign trade using the total import and export volume of each province, expressed in thousands of yuan.

Green finance is proxied by a composite green finance index. This study directly adopts the green finance indicator system, published weights, and corresponding index measures from Nepal et al. [8], which uses the entropy-weighting method, encompassing five dimensions (see Appendix A,

Table A1. The framework for green finance indicators in China.

Indicator	Definition or Calculation
Green credit	Measured by the interest-expenditure share of energy-intensive industries, defined as the ratio of interest expenditures in six major energy-consuming industrial sectors to total industrial interest expenditures.
Green bonds	Capture the extent of green bond development and measured as the ratio of total green bond issuance to total bond issuance.
Green insurance	Proxied by agricultural insurance depth, measured as agricultural insurance premium income divided by total agricultural output value.
Green investment	Measured by the share of investment in environmental pollution control in GDP, computed as environmental pollution control investment divided by GDP.
Government green support	Measured by the share of fiscal expenditures on environmental protection, calculated as fiscal environmental protection expenditure divided by general-public budget expenditure.

Note: The construction of indicators directly adapted from Nepal et al. [8], and the data come from Nepal et al. [19].

). Following Nepal et al. [8], the calculation of weight $G_j$ is the same as in Formulas (1)–(6). The measurement of green finance is shown in Formula (8):

$${GF}_i={\displaystyle {\sum }_{i=1}^n}{y}_{ij}^{*}{G}_j$$

(8)

Here, $\mathrm{n}$ represents the number of sub-indicators. A higher $\mathrm{G}\mathrm{F}$ value indicates a higher level of green finance in province $\mathrm{i}$.

3.3. Model Specification

The baseline model in this paper employs a two-way fixed-effects specification to estimate the impact of the value-added variables and industrial share variables on energy resilience. The two-way fixed-effects specification can be expressed as follows:

$${lnER}_{it}={\beta }_0+{\beta }_1{X}_{it}+{\beta }_2{Controls}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(9)

Here, ${\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{R}}_{\mathrm{i}\mathrm{t}}$ is the explained variable, energy resilience. The core explanatory variable ${\mathrm{X}}_{\mathrm{i}\mathrm{t}}$ includes $\mathrm{l}\mathrm{n}{\mathrm{G}\mathrm{R}\mathrm{P}}_{\mathrm{i}\mathrm{t}}$, ${\mathrm{l}\mathrm{n}\mathrm{P}\mathrm{R}\mathrm{I}\text{\_}\mathrm{V}\mathrm{A}}_{\mathrm{i}\mathrm{t}}$, $\mathrm{l}\mathrm{n}{\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}}_{\mathrm{i}\mathrm{t}}$, $\mathrm{l}\mathrm{n}{\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}}_{\mathrm{i}\mathrm{t}}$, ${\mathrm{P}\mathrm{R}\mathrm{I}\text{\_}\mathrm{S}\mathrm{H}}_{\mathrm{i}\mathrm{t}}$, ${\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{S}\mathrm{H}}_{\mathrm{i}\mathrm{t}}$, and ${\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{S}\mathrm{H}}_{\mathrm{i}\mathrm{t}}$, respectively. The control variables include $\mathrm{l}\mathrm{n}{\mathrm{P}\mathrm{O}\mathrm{P}}_{\mathrm{it}}$, $\mathrm{l}\mathrm{n}{\mathrm{W}\mathrm{A}\mathrm{G}\mathrm{E}}_{\mathrm{it}},$ and $\mathrm{l}\mathrm{n}{\mathrm{T}\mathrm{R}\mathrm{A}}_{\mathrm{it}}$. The value-added variables capture the scale effects of industrial upgrading, whereas ${\mathrm{P}\mathrm{R}\mathrm{I}\text{\_}\mathrm{S}\mathrm{H}}_{\mathrm{i}\mathrm{t}},{\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{S}\mathrm{H}}_{\mathrm{i}\mathrm{t}}$ and ${\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{S}\mathrm{H}}_{\mathrm{i}\mathrm{t}}$ capture the structural composition effects.

${\mathsf{\mu }}_{\mathrm{i}}$ represents the province fixed effect, which controls for time-invariant province-specific characteristics. ${\mathsf{\lambda }}_{\mathrm{t}}$ denotes the year fixed effect, controlling for common macroeconomic shocks. ${\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$ is the random disturbance term, which captures unexplained variation. ${\mathsf{\beta }}_1$ is the core coefficient of interest, representing the marginal effect of ${\mathrm{X}}_{\mathrm{i}\mathrm{t}}$ on energy resilience.

4. Empirical Results

4.1. Baseline Estimation Results

Table 2. The impact of industrial structure on ER.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Var.	lnER
lnGRP	0.338 **
	(0.128)
lnPRI_VA		−0.002			−0.076
		(0.070)			(0.061)
lnSEC_VA			0.157 *		0.166 *
			(0.079)		(0.082)
lnTER_VA				0.173 **	0.171
				(0.072)	(0.103)
PRI_SH						0.577
						(0.877)
SEC_SH							0.359		−0.448
							(0.351)		(0.803)
TER_SH								−0.512	−0.955
								(0.395)	(0.894)
lnPop	0.052 *	0.048	0.043	0.052	0.046	0.054	0.045	0.050	0.054
	(0.026)	(0.036)	(0.032)	(0.031)	(0.029)	(0.034)	(0.037)	(0.034)	(0.035)
lnWAGE	0.073	0.221 ***	0.132 *	0.168 ***	0.096	0.229 ***	0.197 ***	0.194 ***	0.200***
	(0.082)	(0.056)	(0.074)	(0.058)	(0.073)	(0.060)	(0.064)	(0.063)	(0.063)
lnTRA	0.039	0.069 **	0.046	0.050	0.027	0.076 *	0.065 **	0.070 **	0.076**
	(0.029)	(0.033)	(0.030)	(0.035)	(0.031)	(0.038)	(0.031)	(0.031)	(0.037)
Obs.	261	261	261	261	261	261	261	261	261
R2	0.875	0.861	0.870	0.865	0.874	0.862	0.864	0.865	0.866

Notes: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Robust standard errors in parentheses. Province fixed effects and year fixed effects are controlled.

presents the regression results obtained using the two-way fixed effects model, showing the impact of the added value of the three major industries and their respective proportions on energy resilience.

Column (1) uses $\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}$ as the core explanatory variable, with a coefficient of 0.338 that is significant at the 5% level, indicating that regional economic scale expansion significantly enhances energy resilience. Columns (2)–(4) examine the individual effects of the added value of the three major industries: the coefficient for $\mathrm{l}\mathrm{n}\mathrm{P}\mathrm{R}\mathrm{I}\text{\_}\mathrm{V}\mathrm{A}$ is −0.002 and is not statistically significant; $\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}$ has a coefficient of 0.157, which is significantly positive at the 10% level;$\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}$ has a coefficient of 0.173, which is significantly positive at the 5% level. These results indicate that, compared to the primary industry, the development of the secondary and tertiary industries plays a more prominent role in promoting energy resilience.

In column (5), when $\mathrm{l}\mathrm{n}\mathrm{P}\mathrm{R}\mathrm{I}\text{\_}\mathrm{V}\mathrm{A}$, $\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A},$ and $\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}$ are all included in the model, the coefficient for $\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}$ is 0.166, significant at the 10% level, while the coefficient for $\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}$ is 0.171, not significant, and $\mathrm{l}\mathrm{n}\mathrm{P}\mathrm{R}\mathrm{I}\text{\_}\mathrm{V}\mathrm{A}$ remains insignificant. This suggests that, after checking for inter-industry correlations, the primary contribution to energy resilience still comes from the secondary industry. Columns (6)–(8) separately test the individual effects of the industry share variables $\mathrm{P}\mathrm{R}\mathrm{I}\text{\_}\mathrm{S}\mathrm{H}$, $\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{S}\mathrm{H}$, and $\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{S}\mathrm{H}$, none of which are significant. Column (9) uses the share of the primary industry as the baseline group, including only the shares of the secondary and tertiary industries to avoid perfect multicollinearity resulting from the sum of the three industrial shares equaling 1. The regression results show that the coefficients for $\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{S}\mathrm{H}$ and $\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{S}\mathrm{H}$ are −0.448 and −0.955, respectively, but neither reaches statistical significance. This indicates that compared to the primary industry, an increase in the proportion of the secondary or tertiary industry does not exert a robust effect on energy resilience. The economic implication of this finding is that relying solely on structural adjustments in industrial shares may not significantly enhance energy resilience.

Overall, the baseline regression supports the following conclusions: an increase in the level of economic development significantly enhances energy resilience, and the expansion of the secondary industry is a key driver within the industrial structure. The explanatory power of individual industry share variables is limited. Therefore, the evidence is broadly consistent with Hypothesis H1, but the industrial structure channel appears to operate more through sectoral value-added upgrading than through simple share shifts.

4.2. Robustness Checks

4.2.1. Instrumental Variable Estimation

Table 2, the baseline regression, indicates that $\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}, \mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}$ and $\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}$ have a significant positive impact on energy resilience. In the robustness tests, this study employs an instrumental variable (IV) approach with province and year fixed effects for further examination. To mitigate potential endogeneity issues in the core explanatory variables, the first-order and second-order lagged terms of the endogenous explanatory variables are used as instruments while retaining province and year fixed effects.

The model is specified as shown in Equations (10)–(12).

First Stage:

$$X_{it}={\pi }_0+{\pi }_1{L}_1·X_{it}+{\pi }_2{L}_2{·X}_{it}+{\pi }_3{Controls}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(10)

Specification of instrumental variables:

$$Z_{it}=\left\{L_1·X_{it},L_2·X_{it}\right\}$$

(11)

Second Stage:

$$ln{ER}_{it}={\beta }_0+{\beta }_1{\widehat{X}}_{it}+{\beta }_2{Controls}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(12)

In the IV specification with fixed effects, i denotes provinces and t denotes years. $X_{it}$ represents $\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}$, $\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}$, or $\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}$ in separate estimations. Province and year fixed effects are retained, and standard errors are clustered at the provincial level to control for within-group correlation and heteroskedasticity. The control variables include $\mathrm{l}\mathrm{n}{\mathrm{P}\mathrm{O}\mathrm{P}}_{\mathrm{it}}$, $\mathrm{l}\mathrm{n}{\mathrm{W}\mathrm{A}\mathrm{G}\mathrm{E}}_{\mathrm{it}},$ and $\mathrm{l}\mathrm{n}{\mathrm{T}\mathrm{R}\mathrm{A}}_{\mathrm{it}}$.

Table 3. Results of instrumental variable estimation.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Var.	lnER
lnGRP	0.533 ***
	(0.130)
lnPRI_VA		0.055
		(0.053)
lnSEC_VA			0.213 ***
			(0.082)
lnTER_VA				0.329 ***
				(0.089)
PRI_SH					1.003
					(0.962)
SEC_SH						0.229
						(0.390)
TER_SH							−0.449
							(0.419)
First-stage F (Excluded instruments)	108.500	125.800	95.380	178.900	247.500	105.700	64.070
First-stage p-value	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Weak-ID F (KP/CD)	109.100	126.500	95.880	179.800	248.800	106.200	64.400
Under-ID LM	16.170	6.585	8.295	14.280	11.500	12.540	14.470
Under-ID p-value	0.000	0.037	0.016	0.001	0.003	0.002	0.000
Hansen J	0.222	1.040	0.813	1.129	1.460	0.347	0.343
Hansen J p-value	0.638	0.308	0.367	0.288	0.227	0.556	0.558
Obs.	203	203	203	203	203	203	203

Notes: *** indicates statistical significance at the 1% level. Robust standard errors in parentheses. Province fixed effects and year fixed effects are controlled.

presents the IV estimation results for energy resilience. In column (1), $\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}$ is instrumented with its first and second lags, and the second-stage coefficient is 0.533, significant at the 1% level, indicating a positive association between energy resilience development and $\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}$. In column (2), $\mathrm{l}\mathrm{n}\mathrm{P}\mathrm{R}\mathrm{I}\text{\_}\mathrm{V}\mathrm{A}$ is instrumented with its first and second lags, and its coefficient is 0.055 and statistically insignificant. Column (3) instruments $\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}$, yielding a coefficient of 0.213, significant at the 1% level. Column (4) instruments $\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}$ similarly, and its coefficient is 0.329, also significant at the 1% level. In column (5), $\mathrm{P}\mathrm{R}\mathrm{I}\text{\_}\mathrm{S}\mathrm{H}$ is instrumented with its lags, and the estimated coefficient is 1.003 and not statistically significant. Column (6) instruments $\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{S}\mathrm{H}$ with its lags, giving a coefficient of 0.229, which is also statistically insignificant. Column (7) instruments $\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{S}\mathrm{H}$ with $\mathrm{i}\mathrm{t}\mathrm{s} \mathrm{l}\mathrm{a}\mathrm{g}\mathrm{s}$, and the coefficient is −0.449 and statistically insignificant.

The first-stage relevance is strong across all columns: the excluded-instrument F-statistics range from 64.070 to 247.500 (all $p$ < 0.001), well above conventional weak-instrument thresholds. Weak-identification $\mathrm{F}-\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{s}(\mathrm{K}\mathrm{P}/\mathrm{C}\mathrm{D})$ are similarly high (64.400 to 248.800). Under-identification is rejected in every column (LM test p-values: 0.000–0.037). Hansen J tests are insignificant throughout (p-values: 0.288–0.638), supporting the exogeneity of the instruments.

Overall, the IV results provide robust evidence for the positive effects of $\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}$, $\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}$ and $\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}$ on energy resilience, while the industrial share variables remain statistically weak in the instrumental variable framework.

4.2.2. Excluding the Impact of Financial Crisis

To exclude the impact of the 2008 financial crisis, we removed data from before 2013 and retained only data from 2013 onward. The regression results are shown in

Table 4. Results of excluding the impact of the financial crisis.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Var.	lnER
lnGRP	0.468 ***
	(0.120)
lnPRI_VA		0.045
		(0.060)
lnSEC_VA			0.150 **
			(0.068)
lnTER_VA				0.257 **
				(0.100)
PRI_SH					1.106
					(0.862)
SEC_SH						−0.036
						(0.349)
TER_SH							−0.133
							(0.353)
Obs.	203	203	203	203	203	203	203
R2	0.862	0.837	0.844	0.847	0.839	0.836	0.836

Notes: *** and ** indicate statistical significance at the 1% and 5% levels, respectively. Robust standard errors in parentheses. Province fixed effects and year fixed effects are controlled.

. After excluding samples before 2013, the regression results remain largely consistent: both the GRP and the added value of the secondary and tertiary industries continue to have a significant positive impact on energy resilience, whereas the added value of the primary industry and the industrial structure share variables do not exhibit robust statistical significance. This result suggests that enhancing energy resilience is more likely driven by the scale expansion of industries, rather than by mere mechanical changes in their tertiary industry composition.

4.2.3. FE-Cluster and Driscoll–Kraay Evidence for Standardized Variables

Table 5. Robustness check with two-way fixed effects and Driscoll–Kraay standard errors for standardized variables.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Var.
z_lnGRP	1.074 **
	(0.408)
z_lnPRI_VA		−0.010
		(0.268)
z_lnSEC_VA			0.538 *
			(0.269)
z_lnTER_VA				0.563 **
				(0.234)
z_PRI_SH					0.108
					(0.164)
z_SEC_SH						0.103
						(0.101)
z_TER_SH							−0.164
							(0.127)
Obs.	261	261	261	261	261	261	261
R2	0.875	0.861	0.870	0.865	0.862	0.864	0.865

Notes: ** and * indicate statistical significance at the 5% and 10% levels, respectively. Cluster-robust standard errors at the province level are reported in parentheses. Province fixed effects and year fixed effects are controlled.

reports robustness estimations based on standardized variables under province and year fixed effects. The dependent variable and key regressors are z-score standardized; therefore, coefficients can be interpreted as the change in energy resilience in standard deviation units induced by a one-standard-deviation increase in each explanatory variable.

Column (1) shows that $\mathrm{z}\text{\_}\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}$ is positive and significant at the 5% level ($\mathsf{\beta }=1.074$, $p=0.009$), indicating that overall economic expansion exerts a strong and statistically robust promoting effect on energy resilience. In economic terms, a one-standard-deviation increase in $\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}$ is associated with approximately a 1.070-standard-deviation increase in energy resilience. Column (2) reports the effect of $\mathrm{z}\text{\_}\mathrm{l}\mathrm{n}\mathrm{P}\mathrm{R}\mathrm{I}\text{\_}\mathrm{V}\mathrm{A}$. Its coefficient is negative but statistically insignificant (coefficient = −0.010), suggesting that the value added of the primary industry does not exhibit a stable standardized contribution to energy resilience. Column (3) reports the effect of $\mathrm{z}\text{\_}\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}$. The coefficient remains significantly positive at the 10% level ($\mathsf{\beta }=0.538$, $p=0.045$), suggesting that growth in secondary-industry value added contributes to improving energy resilience, although the magnitude is smaller than that of aggregate economic scale. Column (4) indicates that $\mathrm{z}\text{\_}\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}$ is significantly positive at the 5% level ($\mathsf{\beta }$= 0.563, $p=0.016$). This result supports the view that the expansion of the tertiary-industry value added is conducive to strengthening regional energy resilience, likely via efficiency gains and structural upgrading. Columns (5) and (6) examine $\mathrm{z}\text{\_}\mathrm{P}\mathrm{R}\mathrm{I}\text{\_}\mathrm{S}\mathrm{H}$ and $\mathrm{z}\text{\_}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{S}\mathrm{H}$, respectively. Both coefficients are positive but statistically insignificant under FE-Cluster ($\mathrm{z}\text{\_}\mathrm{P}\mathrm{R}\mathrm{I}\text{\_}\mathrm{S}\mathrm{H}:$ $\mathsf{\beta }=0.108$, $p=0.510$; $\mathrm{z}\text{\_}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{S}\mathrm{H}$: $\mathsf{\beta }=0.103$, $p=0.307$), implying that changes in single structural shares alone do not provide stable explanatory power for resilience improvements. Column (7) reports $\mathrm{z}\text{\_}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{S}\mathrm{H}$, which is negative but insignificant under FE-Cluster ($\mathsf{\beta }=-0.164$, $p=$0.195). This suggests that, when considered in isolation, the tertiary share does not show a stable linear contribution under clustered inference.

Overall, the standardized robustness evidence confirms that GRP, PRI_VA and SEC_VA are the most stable drivers of energy resilience, while share-based structural indicators exhibit higher specification sensitivity and should be interpreted with greater caution.

4.3. Heterogeneous Effects

4.3.1. The Impact of Value Added of a Single Industry on Energy Resilience

Table 6. Results of the impact of value added of a single industry on energy resilience.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Var.	$\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{R}$
lnAFAHF_VA	0.011
	(0.072)
lnInd_VA		0.186 **
		(0.079)
lnCon_VA			0.190 **
			(0.076)
lnVAWRT				0.219 *
				(0.124)
lnVATSP					0.108
					(0.066)
lnVAACS						0.059
						(0.119)
lnVAFI							−0.073
							(0.079)
Obs.	261	261	261	261	261	261	261
R2	0.851	0.864	0.869	0.865	0.855	0.852	0.853

Notes: ** and * indicate statistical significance at the 5% and 10% levels, respectively. Robust standard errors in parentheses. Province fixed effects and year fixed effects are controlled. $\mathrm{A}\mathrm{F}\mathrm{A}\mathrm{H}\mathrm{F}\text{\_}\mathrm{V}\mathrm{A}$, $\mathrm{I}\mathrm{n}\mathrm{d}\text{\_}\mathrm{V}\mathrm{A}$, $\mathrm{C}\mathrm{o}\mathrm{n}\text{\_}\mathrm{V}\mathrm{A}$, $\mathrm{W}\mathrm{A}\mathrm{W}\mathrm{R}\mathrm{T}$, $\mathrm{V}\mathrm{A}\mathrm{T}\mathrm{S}\mathrm{P}$, $\mathrm{V}\mathrm{A}\mathrm{A}\mathrm{C}\mathrm{S}$, and $\mathrm{V}\mathrm{A}\mathrm{A}\mathrm{C}\mathrm{S}$ are expressed in logarithmic form.

presents the regression results of the added value of individual industries on energy resilience. In column (1), the coefficient on the added value of agriculture, forestry, animal husbandry, and fishery $\left(\mathrm{l}\mathrm{n}\mathrm{A}\mathrm{F}\mathrm{A}\mathrm{H}\mathrm{F}\text{\_}\mathrm{V}\mathrm{A}\right)$ is 0.011 and statistically insignificant, suggesting that the $\mathrm{l}\mathrm{n}\mathrm{A}\mathrm{F}\mathrm{A}\mathrm{H}\mathrm{F}\text{\_}\mathrm{V}\mathrm{A}$ does not exert a significant effect on energy resilience. In column (2), with industrial added value $\left(\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{n}\mathrm{d}\text{\_}\mathrm{V}\mathrm{A}\right)$ as the core explanatory variable, the coefficient is 0.186, significant at the 5% level. This suggests that a 1% increase in industrial added value leads to an approximately 0.186% increase in energy resilience, reflecting a positive contribution of the industrial sector to energy resilience. In column (3), with construction industry added value $\left(\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{o}\mathrm{n}\text{\_}\mathrm{V}\mathrm{A}\right)$ as the key explanatory variable, the coefficient is 0.190, significant at the 5% level. This shows that expansion in the construction industry is significantly correlated with improvements in energy resilience, and the economic significance is notable. In column (4), with the added value of wholesale and retail trade $\left(\mathrm{l}\mathrm{n}\mathrm{W}\mathrm{A}\mathrm{W}\mathrm{R}\mathrm{T}\right)$ as the core explanatory variable, the coefficient is 0.219, significant at the 10% level. In column (5), with the added value of transportation, storage, and postal services $\left(\mathrm{l}\mathrm{n}\mathrm{V}\mathrm{A}\mathrm{T}\mathrm{S}\mathrm{P}\right)$ as the main explanatory variable, the coefficient is 0.108, which is not significant. This indicates that the development of the transportation and logistics system does not significantly contribute to the stability and resilience of the energy system. In column (6), with the added value of accommodation and catering services $\left(\mathrm{l}\mathrm{n}\mathrm{V}\mathrm{A}\mathrm{A}\mathrm{C}\mathrm{S}\right)$ as the core explanatory variable, the coefficient is 0.059, which is not significant, suggesting that the direct role of consumer-oriented services in promoting ER is not yet evident. In column (7), with the added value of the financial industry $\left(\mathrm{l}\mathrm{n}\mathrm{V}\mathrm{A}\mathrm{F}\mathrm{I}\right)$ as the core explanatory variable, the coefficient is −0.073, which is not significant. This suggests that, under the current specification, the expansion of the financial industry itself has not yet directly translated into improvements in energy resilience.

The regression results in Table 6 indicate that the impact of different industrial sectors on energy resilience varies significantly. Among them, increases in industrial and construction added value significantly enhance energy resilience, suggesting that the expansion of the secondary industry contributes positively to the stability and resilience of the energy system. In contrast, the added value of the primary industry and most service sectors (such as transportation, storage, and postal services; transportation, accommodation and catering, and finance industry) do not exhibit statistically significant effects, indicating a weak direct relationship with energy resilience or the need for other indirect pathways to exert influence. This suggests that, when formulating energy resilience policies, attention could be focused on the structural optimization of the secondary industry, while further exploring the indirect supporting mechanisms through which the service sector contributes to the energy system.

4.3.2. Green Finance Heterogeneity

Table 7. Regional heterogeneity regression results: green finance.

	(1)	(2)	(3)	(4)	(5)	(6)
Vars.	$\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{R}$
	High	Low	High	Low	High	Low
lnGRP	0.575 *	0.222
	(0.278)	(0.138)
lnSEC_VA			0.381 **	0.082
			(0.181)	(0.077)
lnTER_VA					0.160	0.278 ***
					(0.165)	(0.083)
Obs.	130	131	130	131	130	131
R-squared	0.878	0.871	0.877	0.864	0.862	0.874

Notes: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Robust standard errors in parentheses. Province fixed effects and year fixed effects are controlled.

reports the subgroup regression results by the level of green finance foundation. The results suggest that the relationship between industrial structure variables and energy resilience differs across regions with high and low levels of green finance foundation.

Overall, the subgroup results reveal heterogeneous patterns across regions with different green finance foundations. In particular, the coefficients on $\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}$ and $\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}$ are larger and statistically more robust in the high green finance group, whereas the coefficient on $\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}$ is statistically significant only in the low green finance group. These findings suggest that the role of industrial structure in shaping energy resilience may vary with the regional green finance foundation.

First, the coefficient on $\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}$ is 0.575 and statistically significant at the 10% level in the high green finance group, whereas the corresponding coefficient in the low green finance group is 0.222 and statistically insignificant. This indicates that the positive association between economic development and energy resilience is observed more clearly in regions with a higher level of green finance foundation.

Second, the coefficient on $\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}$ is 0.381 and statistically significant at the 5% level in the high green finance group, while the coefficient in the low green finance group is 0.082 and not statistically significant. This suggests that the positive association between secondary-industry development and energy resilience is mainly reflected in regions with a stronger green finance foundation.

Third, the coefficient on $\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}$ is positive in both groups, but it is statistically insignificant in the high green finance group and significantly positive in the low green finance group. This indicates that the positive association between tertiary-industry development and energy resilience is more evident in regions with a lower level of green finance foundation.

Overall, a stronger green finance foundation significantly enhances the economic growth effect and the industrial structure effect, but it does not show a similar amplifying effect on the tertiary industry. This suggests that policies should continue to strengthen the supply capacity of green finance, with a focus on enhancing its support for the green transformation of industry and the improvement of growth quality. At the same time, more targeted green finance instruments should be developed for the service sector to avoid structural mismatches caused by a one-size-fits-all policy approach. These findings are broadly consistent with Hypothesis H3a.

4.3.3. Coal Dependency Heterogeneity

Table 8. The heterogeneity of coal dependency.

	(1)	(2)	(3)	(4)	(5)	(6)
Vars.	lnER
	High	Low	High	Low	High	Low
lnGRP	0.277 ***	0.592 **
	(0.062)	(0.258)
lnSEC_VA			0.189 ***	0.247
			(0.047)	(0.162)
lnTER_VA					0.139 *	0.293 **
					(0.069)	(0.121)
Obs.	130	131	130	131	130	131
R2	0.917	0.847	0.918	0.830	0.909	0.824

Notes: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Robust standard errors in parentheses. Province fixed effects and year fixed effects are controlled.

presents the subgroup regression results by the level of coal dependency. The results show that the associations between industrial structure variables and energy resilience vary across high- and low-coal-dependent regions.

First, $\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}$ is significantly positive in both groups, with coefficients of 0.277 in the high-coal-dependent group and 0.592 in the low-coal-dependent group. This suggests that economic development is positively associated with energy resilience in both types of regions, and the estimated coefficient is larger in the low-coal-dependent group.

Second, the coefficient on $\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}$ is 0.189 and statistically significant at the 1% level in the high-coal-dependent group, whereas the corresponding coefficient in the low-coal-dependent group is 0.247 but statistically insignificant. This indicates that the positive association between secondary-industry development and energy resilience is more clearly observed in high-coal-dependent regions.

Third, $\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}$ is positive and statistically significant in both groups, with coefficients of 0.139 in the high-coal-dependent group and 0.293 in the low-coal-dependent group. This suggests that tertiary-industry development is positively associated with energy resilience in both groups, and the estimated coefficient is larger in the low-coal-dependent group.

Overall, the subgroup regressions suggest that coal dependency is associated with different empirical patterns in the relationship between industrial structure and energy resilience. The coefficient on $\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}$ is statistically significant only in the high-coal-dependent group, whereas the coefficients on $\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}$ and $\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}$ are larger in the low-coal-dependent group. These results indicate heterogeneity across groups, although formal tests of coefficient differences across subgroups are beyond the scope of the current analysis.

Synthesizing the two heterogeneity analyses, it can be inferred that green finance foundations and coal dependency jointly shape the conditions under which industrial structure affects energy resilience, although this combined pathway is not directly estimated in a single model. The results suggest that economic growth is more likely to generate sustained resilience-enhancing effects in regions with stronger green finance foundations and lower coal dependency. By contrast, in regions with high coal dependency, priority should be given to promoting green industrial transformation while simultaneously strengthening green finance supply capacity, thereby facilitating a gradual transition from high-coal lock-in to resilience enhancement. Accordingly, policies should adopt grouped strategies to avoid resource misallocation caused by uniform policy approaches. Taken together, these findings are broadly consistent with Hypothesis H3b.

5. Mechanism

5.1. The Impact of Green Finance on Energy Resilience

Here, we adopt a model with interaction terms to explore the role of green finance in economic scale and energy resilience:

$$\mathit{ln}{ER}_{it}=\alpha +{\mathsf{\beta }}_1{X}_{it}+{\mathsf{\beta }}_{2}ln{GF}_{it}+{\mathsf{\beta }}_3(X_{it}\times ln{GF}_{it})+{\gamma \mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(13)

In the model with interaction terms, ${\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{R}}_{\mathrm{i}\mathrm{t}}$ denotes energy resilience; the core explanatory variable ${\mathrm{X}}_{\mathrm{i}\mathrm{t}}$ includes ${\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}}_{\mathrm{i}\mathrm{t}}$, ${\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}}_{\mathrm{i}\mathrm{t}}$, and ${\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}}_{\mathrm{i}\mathrm{t}}$ in turn, depending on the model specification; $\mathrm{l}\mathrm{n}{\mathrm{G}\mathrm{F}}_{\mathrm{it}}$ denotes green finance. The control variables include $\mathrm{l}\mathrm{n}{\mathrm{P}\mathrm{O}\mathrm{P}}_{\mathrm{it}}$, $\mathrm{l}\mathrm{n}{\mathrm{W}\mathrm{A}\mathrm{G}\mathrm{E}}_{\mathrm{it}},$ and $\mathrm{l}\mathrm{n}{\mathrm{T}\mathrm{R}\mathrm{A}}_{\mathrm{it}}$. Province and year fixed effects are included. Standard errors are clustered at the provincial level.

Table 9. Mechanism: green finance on energy resilience.

	(1)	(2)	(3)
Var.	lnER
lnGRP	0.521 ***
	(0.108)
lnSEC_VA		0.402 ***
		(0.097)
lnTER_VA			0.383 ***
			(0.081)
lnGF	−0.689 **	−0.718 **	−0.520 **
	(0.267)	(0.274)	(0.224)
lnGRP × lnGF	0.088 ***
	(0.025)
lnSEC_VA × lnGF		0.105 ***
		(0.031)
lnTER_VA × lnGF			0.080 ***
			(0.020)
Obs.	261	261	261
R2	0.897	0.894	0.890

Notes: ***, ** indicate statistical significance at the 1% and 5% levels, respectively. Robust standard errors in parentheses. Province fixed effects and year fixed effects are controlled.

reports the interaction regression results testing green finance’s moderating effect on industrial structure and energy resilience. In all three models, the coefficients of $\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}$, $\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}$, and $\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}$ are significantly positive, indicating that economic expansion and the development of both secondary and tertiary industries generally contribute to enhancing energy resilience. The interaction coefficients are also significantly positive at the 1% level:$\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{R}\mathrm{P}\times \mathrm{l}\mathrm{n}\mathrm{G}\mathrm{F}$ is 0.088, $\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}\times \mathrm{l}\mathrm{n}\mathrm{G}\mathrm{F}$ is 0.105, and $\mathrm{l}\mathrm{n}\mathrm{T}\mathrm{E}\mathrm{R}\text{\_}\mathrm{V}\mathrm{A}\times \mathrm{l}\mathrm{n}\mathrm{G}\mathrm{F}$ is 0.080. These results suggest that green finance not only directly affects the energy system but also significantly enhances the marginal promoting effect of industrial development on energy resilience.

Further comparison of the three columns reveals that the interaction coefficient of $\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{E}\mathrm{C}\text{\_}\mathrm{V}\mathrm{A}\times \mathrm{l}\mathrm{n}\mathrm{G}\mathrm{F}$ is the largest, indicating a stronger enabling effect of green finance on the secondary industry. This may be attributed to the concentrated financing needs and higher sensitivity to funding constraints in the industrial sector for energy-saving technological upgrades, equipment renewal, and clean production substitution. In contrast, although the interaction term for the tertiary industry is also significantly positive, its marginal effect is slightly lower, suggesting that the transmission mechanism of green finance in the service sector is relatively longer, and its incremental contribution to energy resilience in the short term is weaker than that of the industrial sector. Meanwhile, the main effect of $\mathrm{l}\mathrm{n}\mathrm{G}\mathrm{F}$ is negative and significant at the 5% level in all three columns, implying that, when not interacting with industrial variables, green finance may entail adjustment costs or resource reallocation frictions. However, the significantly positive interaction terms indicate that such short-term frictions can be offset by the industrial development foundation and transformed into a net positive resilience effect. These findings are broadly consistent with Hypothesis H2b.

5.2. Green Finance and Energy Structure

This section provides a more in-depth and detailed investigation into the pathways through which green finance influences the enhancement of energy resilience.

In Panel A of

Table 10. The impact of green finance on ER and energy structure variables.

Panel A	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Vars.	lnER	lnCoke	lnCrude	lnGP	lnCTP	lnDIP	lnFOP	lnNGP
lnGF	0.299 **	−0.368	−0.461	−0.335	−0.602	−0.432	−3.156 **	1.010
	(0.121)	(0.413)	(0.942)	(0.996)	(1.059)	(0.452)	(1.475)	(1.320)
Observations	261	243	166	237	210	232	228	208
R2	0.871	0.101	0.138	0.227	0.502	0.0674	0.117	0.141
Panel B	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Vars.	lnEGP	lnHPG	lnTPG	lnCCM	lnCC	lnCOC	lnGAC	lnKWH
lnGF	−0.331 *	−0.358	−0.202	−1.624 **	−1.456	−0.996	0.048	−0.208
	(0.194)	(0.397)	(0.246)	(0.707)	(0.961)	(1.341)	(0.362)	(0.148)
Obs.	261	250	261	261	258	236	261	261
R2	0.764	0.457	0.342	0.252	0.164	0.123	0.555	0.883

Notes: ** and * indicate statistical significance at the 5% and 10% levels, respectively. Robust standard errors in parentheses. Province fixed effects and year fixed effects are controlled.

, column (1) uses energy resilience as the dependent variable, and the coefficient for green finance is 0.299, which is significant at the 1% level, indicating that green finance significantly enhances energy resilience. Column (2) uses coke production (Coke) as the dependent variable, and the coefficient for green finance is −0.368, which is not significant, suggesting a potential negative direction but lacking statistical robustness. Column (3) uses crude oil production (Crude) as the dependent variable, and the coefficient for green finance is −0.461, which is insignificant, implying that the inhibitory effect on crude oil output is not yet stable. Column (4) uses gasoline production (GP) as the dependent variable; the coefficient for green finance is −0.335, insignificant, indicating no consistent impact. Column (5) uses coal tar production (CTP) as the dependent variable; the coefficient for green finance is −0.602, insignificant, providing no clear evidence for either a promoting or inhibiting effect. Column (6) uses diesel production (DIP) as the dependent variable; the coefficient for green finance is −0.432, insignificant, showing a negative trend but with insufficient evidence. Column (7) uses fuel oil production (FOP) as the dependent variable; the coefficient for green finance is −3.156, which is significant at the 1% level, demonstrating that green finance significantly suppresses high-carbon fuel oil production, which stands as one of the strongest pieces of pathway evidence. Column (8) uses natural gas production (NGP) as the dependent variable; the coefficient for green finance is 1.010, insignificant, and suggesting a positive substitution trend that is not yet robust.

In Panel B of Table 10, column (1) uses electricity generation (EGP) as the dependent variable, the coefficient for green finance is −0.331, which is not significant. Column (2) uses hydropower generation (HPG) as the dependent variable, the coefficient for green finance is −0.358 and insignificant. Column (3) uses thermal power generation (TPG) as the dependent variable, the coefficient for green finance is −0.202 and insignificant. Column (4) uses coal consumption (CCM) as the dependent variable: the coefficient for green finance is −1.624, which is significant at the 5% level, indicating that green finance significantly reduces coal consumption. This finding provides key evidence supporting the mechanism by which green finance contributes to energy structure optimization.

Column (5) uses coke consumption (CC) as the dependent variable, the coefficient for green finance is −1.456 and insignificant. Column (6) uses crude oil consumption (COC) as the dependent variable, the coefficient for green finance is −0.996 and insignificant. Column (7) uses gasoline consumption (GAC) as the dependent variable, the coefficient for green finance is 0.048 and insignificant. Column (8) uses electricity consumption (kWh) as the dependent variable, the coefficient for green finance is −0.208, insignificant, and suggesting that green finance does not yet have a clear direct inhibitory effect on end-use electricity demand.

The extended path regressions indicate that green finance is positively associated with energy resilience and is also linked to selected changes in the high-carbon energy chain. Specifically, the statistically robust evidence is concentrated in lower fuel oil production and lower coal consumption. Therefore, the mechanism evidence should be interpreted as supporting an energy structure optimization channel, especially through coal consumption reduction, rather than as showing that all production and consumption channels are significant. Overall, the empirical findings provide cautious support for Hypothesis H4.

5.3. Mechanism Tests: Mediating Channels of Green Finance

To examine the mediating role of the pathways through which green finance influences energy resilience, a mediation effect model is constructed:

$$ln{ER}_{it}={\alpha }_0+{\alpha }_{1}ln{GF}_{it}+{\alpha }_2{Controls}_{it}+{\mu }_i+{\lambda }_t+{ϵ}_{it}$$

(14)

$$M_{it}={\beta }_0+{\beta }_1{lnGF}_{it}+{\beta }_2{Controls}_{it}+{\mu }_i+{\lambda }_t+v_{it}$$

(15)

$$ln{ER}_{it}={\gamma }_0+{\gamma }_{1}ln{GF}_{it}+{{\gamma }_2{M}_{it}+\gamma }_3{Controls}_{it}+{\mu }_i+{\lambda }_t+{\eta }_{it}$$

(16)

In the mediation tests, the dependent variable is energy resilience ($\mathrm{l}\mathrm{n}{\mathrm{E}\mathrm{R}}_{\mathrm{i}\mathrm{t}}$), the mediators ${\mathrm{M}}_{\mathrm{i}\mathrm{t}}$ include $\mathrm{l}\mathrm{n}{\mathrm{E}\mathrm{G}\mathrm{P}}_{\mathrm{i}\mathrm{t}}$, $\mathrm{l}\mathrm{n}{\mathrm{F}\mathrm{O}\mathrm{P}}_{\mathrm{i}\mathrm{t}}$, and $\mathrm{l}\mathrm{n}{\mathrm{C}\mathrm{C}\mathrm{M}}_{\mathrm{i}\mathrm{t}}$, green finance is proxied by $\mathrm{l}\mathrm{n}{\mathrm{G}\mathrm{F}}_{\mathrm{i}\mathrm{t}}$, and the standard control variables and province/year fixed effects are retained. The control variables include $\mathrm{l}\mathrm{n}{\mathrm{P}\mathrm{O}\mathrm{P}}_{\mathrm{it}}$, $\mathrm{l}\mathrm{n}{\mathrm{W}\mathrm{A}\mathrm{G}\mathrm{E}}_{\mathrm{it}},$ and $\mathrm{l}\mathrm{n}{\mathrm{T}\mathrm{R}\mathrm{A}}_{\mathrm{it}}$. To identify the internal mechanisms through which green finance influences energy resilience, this paper selects electricity generation ($\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{G}\mathrm{P}$), fuel oil production ($\mathrm{l}\mathrm{n}\mathrm{F}\mathrm{O}\mathrm{P}$), and coal consumption ($\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{C}\mathrm{M}$) as mechanism variables and employs a mediation-effect model for testing. The results show that the effect of green finance on energy resilience is not realized through a single channel, and the transmission strength varies significantly across different energy variables.

Columns (1) to (6) of

Table 11. Channels and transmission mechanisms of green finance on energy resilience.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Var.	lnEGP	lnER	lnFOP	lnER	lnCCM	lnER	lnER	lnER	lnER
lnGF	−0.331 *	0.348 ***	−3.156 **	0.334 **	−1.624 **	0.205 *	−0.263	0.339 **	−0.235
	(0.194)	(0.126)	(1.475)	(0.125)	(0.707)	(0.108)	(0.529)	(0.132)	(0.251)
lnEGP		0.149 **					0.298 **
		(0.054)					(0.133)
lnFOP				0.004				−0.003
				(0.007)				(0.036)
lnCCM						−0.058 ***			−0.045 ***
						(0.019)			(0.015)
lnEGP × lnGF							0.082
							(0.066)
lnFOP × lnGF								−0.004
								(0.019)
lnCCM × lnGF									0.043 *
									(0.023)
Obs.	261	261	228	228	261	261	261	228	261
R-squared	0.764	0.879	0.170	0.885	0.234	0.866	0.884	0.889	0.881

Notes: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Robust standard errors in parentheses. Province fixed effects and year fixed effects are controlled.

report the results of the mediation regression analysis. Columns (1)–(2) show that with $\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{G}\mathrm{P}$ as the mediator, green finance negatively affects $\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{G}\mathrm{P}$ at the 10% significance level; after its inclusion, the direct effect of green finance on energy resilience is 0.348, significant at the 1% level. $\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{G}\mathrm{P}$ itself also has a significantly positive effect on energy resilience, with a coefficient of 0.149, significant at the 5% level. This suggests that the electricity generation channel plays a mediating role, but the direction of its indirect effect is inconsistent with the direct effect of green finance, reflecting a suppression effect.

Columns (3) and (4) show that with $\mathrm{l}\mathrm{n}\mathrm{F}\mathrm{O}\mathrm{P}$ as the mediator, green finance significantly reduces fuel oil production, with a coefficient of −3.156, which is significant at the 5% level. However, the effect of $\mathrm{l}\mathrm{n}\mathrm{F}\mathrm{O}\mathrm{P}$ on energy resilience is not significant, indicating that the mediation chain of this channel is incomplete and lacks statistical support.

Columns (5) and (6) show that with $\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{C}\mathrm{M}$ as the mediator, green finance significantly suppresses coal consumption, with a coefficient of −1.624, which is significant at the 5% level. Moreover, after including $\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{C}\mathrm{M}$, coal consumption exerts a negative effect on energy resilience, with a coefficient of −0.058, significant at the 1% level, while the coefficient for green finance remains 0.205, significant at the 10% level. These results indicate that coal consumption is a key transmission channel through which green finance enhances energy resilience. In other words, green finance significantly improves the resilience of the energy system by reducing reliance on coal and optimizing the structure of energy consumption.

Meanwhile, we adopt an interaction term model to explore the role of green finance in economic scale and energy resilience:

$${lnER}_{it}={\delta }_0+{\delta }_1{M}_{it}+{\delta }_2{lnGF}_{it}+{\delta }_3(M_{it}\times {lnGF}_{it})+{\delta }_4{Controls}_{it}+{\mu }_i+{\lambda }_t+{\xi }_{it}$$

(17)

Here, energy resilience is measured by ${\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{R}}_{\mathrm{i}\mathrm{t}}$; mediating variables ${\mathrm{M}}_{\mathrm{i}\mathrm{t}}$ comprise ${\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{G}\mathrm{P}}_{\mathrm{i}\mathrm{t}}$, ${\mathrm{l}\mathrm{n}\mathrm{F}\mathrm{O}\mathrm{P}}_{\mathrm{i}\mathrm{t}}$, and ${\mathrm{l}\mathrm{n}\mathrm{C}\mathrm{C}\mathrm{M}}_{\mathrm{i}\mathrm{t}}$; green finance is proxied by $\mathrm{l}{\mathrm{n}\mathrm{G}\mathrm{F}}_{\mathrm{i}\mathrm{t}}$; and control variables include $\mathrm{l}\mathrm{n}{\mathrm{P}\mathrm{O}\mathrm{P}}_{\mathrm{it}}$, $\mathrm{l}\mathrm{n}{\mathrm{W}\mathrm{A}\mathrm{G}\mathrm{E}}_{\mathrm{it}},$ and $\mathrm{l}\mathrm{n}{\mathrm{T}\mathrm{R}\mathrm{A}}_{\mathrm{it}}$. Provincial and year fixed effects are denoted by ${\mathsf{\mu }}_{\mathrm{i}}$ and ${\mathsf{\lambda }}_{\mathrm{t}}$, respectively. Standard errors are clustered at the provincial level.

Furthermore, in the reported interaction term results, the main effect coefficient of $\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{G}\mathrm{P}$ is 0.298, which is significant at the 5% level. However, the coefficient of the interaction term $\mathrm{l}\mathrm{n}\mathrm{E}\mathrm{G}\mathrm{P}\times \mathrm{l}\mathrm{n}\mathrm{G}\mathrm{F}$ is positive but not significant, indicating that the moderating effect of green finance on the relationship between electricity generation and energy resilience lacks sufficient statistical support. Combining the mediation and interaction tests, it can be argued that the mechanism of green finance is more prominently reflected in suppressing coal consumption rather than in strengthening the marginal effect of electricity generation.

Overall, the evidence suggests that green finance enhances energy resilience partly through a low-carbon shift in energy consumption, with the clearest evidence from coal consumption reduction. By contrast, traditional fossil energy production channels exhibit limited explanatory power for energy resilience. Policymakers should strengthen the constraints and incentives of green credit and investment for high-energy-consuming and high-emission sectors. Priority should be given to directing capital flows toward clean energy substitution, energy-saving technological upgrades, and improvements in end-use energy efficiency, thereby achieving synergistic gains between energy resilience and low-carbon transition. This evidence is consistent with Hypothesis H2a and provides more direct support for Hypothesis H4.

6. Conclusions

Based on provincial panel data in China, this paper systematically examines the theoretical linkages and empirical relationships among industrial structure upgrading, green finance, and energy resilience. By constructing a two-way fixed effects model, conducting mediation mechanism tests, interaction effect tests, and performing heterogeneity analysis, the following main conclusions are drawn.

First, industrial structure upgrading is a key driver in enhancing energy resilience. Empirical results indicate that the expansion of value added in the secondary and tertiary industries significantly strengthens the robustness, recoverability, and adaptability of the energy system. This effect is particularly pronounced in the industrial and construction sectors, where the expansion of value added plays a key role. In contrast, mere changes in industrial share proportions or the development of the primary industry do not significantly contribute directly to resilience. This suggests that resilience improvement relies more on intrinsic quality and efficiency upgrades within industries rather than on quantitative or proportional adjustments.

Second, green finance serves as an important link between industrial upgrading and energy resilience. This study finds that the level of green finance development significantly and positively moderates the promoting effect of industrial upgrading on energy resilience. More importantly, the mechanism analysis shows that green finance contributes to resilience mainly through the optimization of the energy consumption structure. Specifically, by effectively curbing coal consumption—and, to a lesser extent, fuel oil production—green finance reduces the structural dependence of the economy on high-carbon energy. This fundamentally improves the energy system’s capacity to cope with price fluctuations, supply disruptions, and transition-policy shocks. By contrast, channels such as electricity generation provide weaker and less stable evidence.

Finally, the impact of industrial upgrading and green finance on energy resilience exhibits significant regional heterogeneity. In regions with a stronger green finance foundation, the resilience effects of economic growth and secondary-industry upgrading are more pronounced, whereas the contribution of tertiary-industry upgrading is stronger in regions with weaker green finance foundations. In addition, in regions with higher coal dependency, the secondary industry plays a more important role in resilience enhancement, while in regions with lower coal dependency, the effects of economic growth and tertiary-industry upgrading are stronger. These results suggest differentiated policy priorities across regional contexts.

This study also has several limitations that should be acknowledged. First, the empirical analysis is based on provincial panel data for 2011–2019; therefore, it does not capture more recent changes in energy markets, green finance development, and the post-pandemic economic and policy context. In particular, the period after 2020 witnessed substantial fluctuations in global energy prices, accelerated low-carbon policy adjustments, and new developments in green financial instruments, which may influence the relationship among industrial upgrading, green finance, and energy resilience. Future research could extend the sample period and incorporate more recent data to further verify the robustness and dynamic evolution of the findings.

Based on the findings above, this paper proposes the following policy implications.

The empirical findings and policy insights drawn from China’s experience can serve as a valuable reference for other developing economies that face similar challenges in pursuing industrial transformation and energy resilience. The integrated industrial upgrading plus green finance pathway, if appropriately adapted to local institutional and developmental contexts, may offer a replicable framework for strengthening energy system resilience alongside sustainable growth.

First, it is essential to promote synergy between industrial and financial policies. Energy resilience policies should transcend sectoral boundaries and strive for deep alignment between industrial upgrading objectives and green finance supply. Specifically, policymaking must shift from a focus on tool provision to mechanism building, thereby ensuring that financial resources such as green credit and green bonds are precisely directed toward key areas that can substantially optimize the energy structure, such as industrial energy-saving retrofits and clean energy substitution.

Second, differentiated regional strategies should be implemented. Uniform policies should be abandoned in favor of region-specific measures tailored to local green finance foundations and energy structure characteristics. In practice, regions with high-coal dependency should strengthen green finance support for industrial transformation and alternative energy projects, whereas regions that are pioneers in low-coal dependency should focus on innovating green finance products for the service sector to consolidate their resilience advantages.

Moreover, there is a need to strengthen resilience-centric policy evaluation. It is recommended to incorporate energy resilience indicators into the performance evaluation system for green finance, which would help establish a closed-loop management framework that progresses from capital allocation to structural optimization, and ultimately to resilience enhancement. Consequently, this will guide financial resources to effectively serve the long-term stability and low-carbon transition of the energy system.

In summary, fostering energy resilience requires systematic and coordinated efforts. By aligning industrial policies with green finance mechanisms and adopting place-based approaches, policymakers can better support the transition toward a more resilient and sustainable energy system.