Can Regional New Digital Infrastructure Promote the Level of Green Finance? Empirical Evidence from Chinese Cities

Abstract

Using panel data for 135 Chinese prefecture-level cities from 2007 to 2023, this study investigates the impact of new digital infrastructure on green finance development. The new digital infrastructure indicator is constructed based on the proportion of relevant keywords appearing in government work reports, while the green finance index is reconstructed using the entropy-weighting method across seven dimensions. The estimation results indicate that new digital infrastructure exerts a significant positive effect on green finance development. This conclusion remains robust after a series of robustness checks, including alternative variable measurements, winsorization treatment, and instrumental-variable estimation. Mechanism analysis reveals that industrial structure upgrading, particularly the advancement of industrial structure, serves as an important transmission channel. Further heterogeneity analysis shows that the promoting effect is more pronounced in cities with larger economic scale, those located outside major urban agglomerations, and cities with higher levels of financial resource aggregation. These findings provide empirical evidence for the role of digital infrastructure in fostering green finance and facilitating sustainable regional development.

1. Introduction

As digital technologies continue to evolve, innovations such as 5G, big data, and artificial intelligence have become increasingly embedded in a wide range of industries, playing a pivotal role in advancing industrial modernization and supporting sustainable economic growth (Y. Zhao et al., 2025; Liu et al., 2024). In addition to improving resource allocation efficiency and transforming production organization, this process has stimulated growing demand for new-type digital infrastructure. Growing concerns over environmental sustainability, coupled with China’s shift toward a higher-quality growth model, have placed green and low-carbon development at the forefront of policy agendas. Green finance plays an increasingly important role in directing capital toward clean industries and promoting economic structural upgrading, serving as a crucial policy instrument for achieving the dual-carbon goals and enabling green transition (X. Wang & Wang, 2021).

New digital infrastructure exhibits significant advantages in improving information environments, strengthening environmental governance capacity, and promoting green innovation and the development of green industries, thereby creating favorable conditions for the development of green finance (Guo et al., 2024). However, existing research mainly relies on broad indicators such as digital economy development, smart cities, or informatization levels, and lacks focused examination of “new-type infrastructure construction” as a specific policy variable in shaping local green finance. The existing literature has devoted considerable attention to the economic consequences of new-type infrastructure, particularly its influence on regional development and productivity enhancement. (Hu et al., 2025; J. Feng & Qi, 2024), while the mechanisms through which such infrastructure influences green finance remain insufficiently identified. Given that green finance development is deeply shaped by regional industrial foundations, the degree of industrial upgrading, and city size, current studies remain limited in offering a structural perspective that systematically investigates the relationship between infrastructure and green finance. This study extends the existing literature by assessing the role of new digital infrastructure in shaping green finance development and by uncovering the potential pathways through which such influence is realized. Furthermore, industrial structure upgrading is incorporated into the theoretical framework and empirical model as a mediating channel to explore its structural transmission role.

This paper provides empirical evidence on how new digital infrastructure enhances cities’ capacity for green finance development, partly through industrial structure upgrading. This study contributes to the existing literature in three main respects. First, from the perspective of research focus, this study links new digital infrastructure construction with green finance development, thereby enriching the literature on the determinants of green finance in the context of digital transformation. Second, in terms of mechanism identification, the study incorporates industrial structure upgrading into the analytical framework and provides empirical evidence that the advancement of industrial structure serves as an important channel through which new digital infrastructure promotes green finance. Third, from a methodological perspective, this paper uses prefecture-level city data rather than provincial-level aggregates, allowing for a more fine-grained examination of regional differences and heterogeneous effects across cities with different economic scales, urban agglomeration characteristics, and levels of financial development.

2. Literature Review

2.1. Digital Infrastructure as a Driver of Green Finance

The rapid development of digital infrastructure, including 5G networks, data centers, cloud computing, and industrial internet platforms, has reshaped how financial systems collect information, evaluate risks, and allocate capital in the context of China’s new infrastructure construction. Existing studies generally argue that digitalization effectively reduces information asymmetry, expands the efficiency frontier of financial markets, and thereby creates institutional and technological conditions favorable to green finance (He et al., 2024; Shen et al., 2025). Through the integration of digital financial technologies such as AI-driven risk control, big data, and digital carbon-accounting platforms, digital infrastructure strengthens the ability of financial institutions to identify green projects and monitor environmental performance (Luo et al., 2023). In such an environment, sustainability-oriented financial instruments can develop more rapidly and at lower transaction costs.

Research on the digital economy further reinforces this logic. Digitalization has been widely shown to promote green development by improving resource allocation efficiency, encouraging green innovation, and facilitating cleaner production models (S. Zhao et al., 2023; Xiao et al., 2023). Recent studies have increasingly examined green finance as an integral component of green development (S. Zhang et al., 2021; Xu et al., 2024). The improvements brought about by digital infrastructure in productivity, information transparency, and environmental governance indirectly translate into higher demand for and supply of green finance. For example, smart-city governance and digital environmental-regulation mechanisms strengthen the capacity of local governments and firms to generate green and innovation-intensive projects, which rely heavily on scalable green-finance tools (Y. Zhang et al., 2023).

Studies on digital finance also show that as financial services become more inclusive through digital channels, green finance penetrates more deeply into firms and regions that previously lacked access to green credit. Financial inclusiveness strengthens the connection between green capital and green projects, improving the overall efficiency of the green finance system (H. Li & Xu, 2023). Overall, digital infrastructure is increasingly viewed as a key factor that improves the enabling conditions for the growth of green finance.

2.2. Infrastructure Construction and Industrial Structure Upgrading

Infrastructure has long been regarded as a key determinant of industrial transformation. Classical development theories and extensive empirical evidence from China’s reform period suggest that improvements in transportation, communication, and energy systems reshape cost structures, enhance market accessibility, and strengthen production capabilities, thereby promoting regional shifts toward higher value-added activities (Zou et al., 2022). Infrastructure investment supports the transition of regional economies from resource-intensive industries to producer services, advanced manufacturing, and knowledge-intensive sectors, though the magnitude of this impact varies with regional development foundations and stages (Xia et al., 2024).

With the rise of new infrastructure, scholars have increasingly focused on the transformative potential of digital and innovation-oriented infrastructure. Compared with traditional infrastructure, digital infrastructure significantly enhances information flows, human-capital formation, innovation spillovers, and cross-regional factor mobility. Provincial- and city-level studies show that digital infrastructure accelerates industrial upgrading by improving technological innovation capacity, attracting high-quality talent, and guiding capital flows toward high-tech and service-oriented industries (Gong et al., 2023). These changes are driven by deeper structural factors: digital platforms lower the cost of knowledge diffusion and coordination, while smart infrastructure improves urban governance and service-sector productivity.

Evidence from quasi-natural experiments such as the “Broadband China” policy further confirms that digital infrastructure helps shift regional economies toward more advanced and technology-intensive sectors (Sun & Masron, 2025). Furthermore, new infrastructure and traditional infrastructure function as complements, jointly supporting the formation of industrial clusters and innovation ecosystems.

2.3. Industrial Structure, Green Development, and Green Finance

Industrial structure is widely regarded as a core factor shaping regional environmental performance and the development of green finance. Prior studies generally regard industrial upgrading as an important driver of green total factor productivity improvement and carbon emission reduction, as it reflects a transition toward technology-intensive, cleaner, and service-oriented industries (Cheng et al., 2018; T. Feng et al., 2024). The underlying logic is that advanced industries rely more on knowledge, innovation, and efficiency rather than resource-intensive and pollution-intensive inputs. Studies show that industrial-structure rationalization effectively reduces emissions, while structural upgrading exhibits even nonlinear effects capable of generating significant long-term environmental improvements (Yang & Shen, 2023; Zhou et al., 2024).

These structural improvements provide fertile ground for green-finance development. A more advanced industrial base contains a greater number of green, low-carbon, and innovation-driven projects, including the types of projects that green bonds, green credit, and other sustainable-finance instruments target (C. Zhao et al., 2025). In other words, industrial upgrading expands the “investment universe” for green finance. As industries move up the value chain, their financing needs become more aligned with green investment principles and more consistent with environmental-risk management and regulatory requirements (Y. Zhang & Dilanchiev, 2022).

However, limited attention has been paid to the effect of new digital infrastructure on regional green finance, particularly the role of industrial structure upgrading as a potential transmission channel. Moreover, few empirical studies use prefecture-level data to capture intra-provincial heterogeneity. To address these gaps, this study investigates how new digital infrastructure influences green-finance development across Chinese prefecture-level cities and incorporates industrial-structure upgrading as a transmission channel. The analysis contributes to the existing literature by providing new evidence on how digital infrastructure and green finance can develop in a mutually reinforcing manner.

3. Theoretical Analysis and Hypothesis Development

3.1. The Effect of New Digital Infrastructure on Green Finance

From the perspective of information asymmetry, financial institutions often face difficulties in accurately capturing firms’ environmental performance and green-technology information, resulting in adverse selection and moral hazard in green-project financing (Akerlof, 2002). New digital infrastructure serves as an essential productive foundation in the digital economy by improving information transmission, technological support, and governance efficiency, thereby reducing information opacity. This enables financial institutions to obtain more accurate information on firms’ environmental performance, carbon emissions, and technological innovation, effectively reducing information asymmetry in the process of green-finance development.

From the perspective of transaction-cost economics, digital infrastructure enhances the digitalization of the financial system and reduces the transaction costs associated with the screening, auditing, and supervision of green projects. The environmental characteristics, carbon-emission performance, and innovation capacity of green projects can thus be identified more accurately (Goldfarb & Tucker, 2019), allowing green credit, green bonds, and other green-finance instruments to operate with lower risks and lower transaction costs (Coase, 2012; Williamson, 2008). Moreover, new digital infrastructure improves resource allocation efficiency, alleviates firms’ financing constraints, and strengthens the ability of the financial system to identify and screen green projects, thereby supporting the expansion of green-finance scale (Pan & Yang, 2024). At the same time, smart-governance systems and digital regulatory tools enhance the capacity of governments and firms to fulfill environmental responsibilities, thus expanding the demand base for green finance (Al-Aiban, 2024). By strengthening technological capacity and improving institutional conditions, new infrastructure construction promotes the regional development of green finance.

Based on the above analysis, this study proposes the following hypothesis:

H1. 

New infrastructure construction significantly promotes the development of regional green finance.

3.2. The Mediating Role of Industrial Structure Upgrading

The transformation of industrial structure from traditional industries to advanced manufacturing, modern services, and knowledge-intensive sectors represents a long-term trend in economic development (Kuznets, 1973; Chenery, 1986). As a system centered on digital networks, smart platforms, and innovation resources, new infrastructure exhibits strong technological spillover effects. It accelerates firms’ adoption of advanced technologies and promotes the upgrading of industrial chains toward higher value-added segments (Baldwin, 2016). Such digital transformation not only enhances firms’ production efficiency and innovation capabilities but also fosters the rapid development of high-tech industries, green industries, and digital services, making it a key driver of industrial-structure upgrading. Furthermore, new infrastructure promotes the agglomeration of high-quality human capital and innovation resources in digitally advanced cities, further shifting industrial structures from low-end activities toward research and development, design, and professional services (Tang & Zhao, 2023). Existing evidence indicates that digital infrastructure development contributes to the upgrading, greening, and intelligent transformation of regional industrial structures (L. Wang & Shao, 2024; Xia et al., 2024).

Industrial-structure upgrading is regarded as a fundamental basis for green development and green finance. Such upgrading improves regional green-development performance and strengthens both the demand and supply foundations of green finance. On the demand side, a higher proportion of high-tech and service-oriented industries indicates a shift toward innovation, knowledge capital, and technological efficiency, while reducing dependence on resource consumption and pollution-intensive inputs. This transformation increases the need for green financing (Szirmai, 2012). The transition toward advanced industries expands the pool of green investment opportunities by promoting green technological innovation, energy-efficiency improvements, and new-energy applications. It also strengthens regional absorptive capacity, as higher human-capital and technological levels make regions better positioned to obtain and utilize green financial resources (Hall & Rosenberg, 2010). Moreover, industrial upgrading reduces dependence on high-emission sectors, making it easier for green capital to be allocated and utilized.

Therefore, industrial-structure upgrading serves not only as an important economic outcome of new infrastructure construction but also as a necessary condition for the development of green finance, functioning as a typical mediating channel through which new infrastructure promotes green finance.

Based on the above analysis, this study proposes the second hypothesis:

H2. 

Industrial-structure upgrading channels the positive effect of new infrastructure construction on regional green finance.

4. Research Design

4.1. Sample Selection

The analysis is based on panel data for Chinese prefecture-level cities over the period 2007–2023, excluding centrally administered municipalities. Due to missing government work report texts in some prefecture-level cities and incomplete statistics on green finance indicators in the early years, this study excludes cities with severe data gaps to ensure the continuity and comparability of the sample.

The data were collected from multiple sources. The original indicators used to construct the Green Finance Index, including green credit, green investment, green insurance, green bonds, green support, green funds, and green equity, were obtained from the Macrodatas database. Based on these indicators, this study reconstructed the Green Finance Index using the entropy-weighting method. Data on new digital infrastructure were also obtained from the Macrodatas database and further processed according to the sample selection criteria. Mediating variable and control variables were manually compiled from the China Urban Statistical Yearbook and other official statistical sources. The final dataset is a balanced panel consisting of 135 prefecture-level cities from 27 provinces.

4.2. Variable Definition

4.2.1. Dependent Variable: Regional Green Finance (Grefin)

Green Finance reflects the level of support provided by the regional financial sector to local clean industries. This paper breaks down the green finance indicator system into seven indicators: green credit, green investment, green insurance, green bonds, green support, green funds, and green equity. The original data for the seven green finance dimensions were obtained from the Macrodatas database. The calculation methods for each indicator are shown in the table. The entropy method is used for measurement, and the indicator construction and weight results are presented in

Table 1. Green finance indicator system and indicator weights.

Criterion Level	Indicator Level	Unit	Indicator Attributes	Weight
Green Credit	Total amount of credit for environmental protection projects/total credit	%	positive	0.1351
Green Investment	Investment in environmental pollution control/GDP			0.1388
Green Insurance	Revenue from environmental pollution liability insurance/total premium income			0.1371
Green Bonds	Total issuance of green bonds/total issuance of all bonds			0.1431
Green Support	Fiscal expenditure on environmental protection/general public budget expenditure			0.1622
Green Funds	Total market value of green funds/total market value of all funds			0.1309
Green Equity	Carbon trading, energy-use rights trading, and pollution-discharge rights trading/total equity market transaction volume			0.1528

Note: The Weight column reports the entropy-method weight of each sub-indicator, indicating its relative contribution to the overall green finance index. All indicators are positive indicators.

.

4.2.2. Independent Variable: New-Type Digital Infrastructure Development (Newdiginfra)

New-type digital infrastructure denotes a new generation of information infrastructure built upon technologies such as 5G networks, artificial intelligence, the industrial internet, and the Internet of Things. This study measures the level of new-type digital infrastructure development using the proportion of keywords related to new-type digital infrastructure in local government work reports. The keyword-frequency data and total word counts of government work reports were obtained from the Macrodatas database. Based on these data, the ratio of new-type digital infrastructure-related keywords to the total number of words in each report was calculated to construct the NewDigInfra indicator.

4.2.3. Mediating Variable: Industrial Structure Upgrading (UpIndus)

Industrial structure upgrading (UpIndus) is measured by the ratio of the output value of the tertiary industry to total output value. This measurement follows the official three-sector industrial classification used in Chinese statistical reporting, rather than a reclassification of industries into newly defined categories. In this framework, the economy is divided into primary, secondary, and tertiary industries. A higher share of the tertiary industry indicates that a city’s economic structure is shifting toward service-oriented and higher-value-added activities, which is commonly regarded as a sign of industrial structure upgrading.

4.2.4. Control Variables

This paper controls for other city-level factors that may influence the level of green finance. The selection of control variables is based on the potential determinants of city-level green finance development.

Urbanization rate (Urban), measured by the ratio of non-agricultural population to registered population, is included to capture the stage of urban development. Urbanization changes the scale of infrastructure demand, environmental governance pressure, and the concentration of financial resources, which may further influence the demand for green credit, green investment, and other green financial instruments (Fodouop Kouam & Catche, 2025).

Higher Education Resource Intensity (EduRes) is measured by the number of students enrolled in regular higher education institutions divided by the total year-end population. This indicator is used to capture the intensity of higher education resources in a city (Jing et al., 2025). Cities with stronger higher education resources generally have a stronger knowledge base and greater potential for innovation, which may influence the generation of green projects and the demand for green financial services (Lee et al., 2025).

Fiscal Investment Intensity (FisInv), measured by fixed asset investment divided by general government fiscal expenditure, is included to capture the relative intensity of regional investment activities and the fiscal capacity supporting infrastructure construction. Fiscal investment and public expenditure may shape the supply of green projects, environmental infrastructure, and the institutional environment for green finance. Previous studies have shown that fiscal investment and government expenditure are closely related to green finance development and the effectiveness of green financial instruments. For example, L. Li et al. (2023) emphasize the role of public expenditure and fiscal execution efficiency in green finance and environmental governance.

Openness level (OpenOut) is controlled because foreign capital may introduce advanced technologies, environmental standards, management experience, and external financing channels. Cities with higher openness are more likely to be exposed to international green investment practices and environmental governance standards, which may influence the development of green finance (H. Wu, 2022).

Environmental protection level (EnvPro), measured by the harmless treatment rate of household waste, is included to reflect local environmental governance capacity. Local environmental governance capacity may affect the demand for green financial instruments. A higher harmless treatment rate of household waste reflects stronger urban environmental management capacity. Stronger environmental protection may increase the demand for pollution control, clean production, and low-carbon transformation, thereby affecting green credit, green investment, green bonds, and other green financial activities (Y. Wu, 2024).

Mobile phone penetration rate (MobPop), measured by the average number of mobile phones owned per person, is controlled for because it reflects the accessibility of digital communication infrastructure (Teng et al., 2024). Higher mobile phone penetration can improve information transmission, financial inclusion, and the accessibility of digital financial services, which are important conditions for green finance development in the digital economy era.

The definitions and calculation methods of all variables are presented in

Table 2. Variable definitions.

Variables	Symbol	Definition
Green Finance	GreFin	Measured using the entropy method based on seven indicators: green credit, green investment, green insurance, green bonds, green support, green funds, and green equity (see earlier section).
New Digital Infrastructure	NewDigInfra	Number of terms related to new digital infrastructure/total number of words in government reports.
Industrial Structure Upgrading	UpIndus	Output value of the tertiary industry/total output value.
Urbanization Rate	Urban	Non-agricultural population/registered population.
Higher Education Resource Intensity	EduRes	Number of students enrolled in regular higher education institutions/total year-end population.
Fiscal Investment Intensity	FisInv	Fixed asset investment/general government fiscal expenditure.
Level of Openness	OpenOut	Amount of actual utilized foreign capital/regional GDP.
Environmental Protection	EnvPro	Harmless treatment rate of household waste.
Mobile Phone Penetration Rate	MobPop	Average number of mobile phones owned per person.

.

4.2.5. Model

Based on the above theoretical analysis, in order to accurately identify the impact of new-type digital infrastructure on the level of green finance, this paper constructs the following basic econometric models:

Baseline Regression Model

$${\mathrm{GreFin}}_{\mathrm{i},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{NewDigInfra}}_{\mathrm{i},\mathrm{t}}+{\sum }_{\mathrm{k}=1}^{\mathrm{n}}{\mathsf{\alpha }}_{\mathrm{k}}{\mathrm{control}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(1)

Mechanism Test Models

$${\mathrm{UpIndus}}_{\mathrm{i},\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\mathrm{NewDigInfra}+{\sum }_{\mathrm{k}=1}^{\mathrm{n}}{\mathsf{\beta }}_{\mathrm{k}}{\mathrm{control}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(2)

$${\mathrm{GreFin}}_{\mathrm{i},\mathrm{t}}={\mathsf{\gamma }}_0+{\mathsf{\gamma }}_1{\mathrm{NewDigInfra}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\gamma }}_2{\mathrm{UpIndus}}_{\mathrm{i},\mathrm{t}}+{\sum }_{\mathrm{k}=1}^{\mathrm{n}}{\mathsf{\gamma }}_{\mathrm{k}}{\mathrm{control}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$$

(3)

4.3. Empirical Results

4.3.1. Descriptive Statistics

The descriptive statistics of the sample variables are presented in

Table 3. Descriptive statistics.

Variable	N	Mean	Min	P50	Max	Sd
GreFin	2238	0.330	0.0127	0.343	0.719	0.118
NewDigInfra	2238	0.00164	0	0.00140	0.0138	0.00141
UpIndus	2238	1.178	0.299	1.014	6.383	0.662
Urban	2238	0.439	0.100	0.397	1	0.210
EduRes	2238	0.0266	0.000226	0.0143	0.185	0.0304
FisInv	2238	5.601	0.0164	4.793	36.84	3.793
OpenOut	2238	0.00252	0	0.00182	0.0194	0.00258
EnvPro	2238	90.76	0	100	100	17.88
MobPop	2238	1.070	0.128	0.928	10.17	0.744

. For the proxy variable of green finance (GreFin), the green finance index ranges from 0.0127 to 0.719, with a mean value of 0.330 and a standard deviation of 0.118. These statistics suggest that green finance has developed to a certain extent across Chinese cities, but substantial regional differences remain. The proxy variable for new digital infrastructure (NewDigInfra) ranges from 0 to 0.0138, with a mean of 0.00164 and a standard deviation of 0.0014, suggesting that not all cities have carried out new-infrastructure construction and that significant differences remain in the extent of such development across cities. The proxy variable for industrial-structure upgrading (UpIndus) has a maximum value of 6.383 and a minimum of 0.299, with a mean of 1.178 and a standard deviation of 0.662, reflecting substantial heterogeneity in industrial-structure levels across regions. The descriptive characteristics of the remaining variables are generally consistent with existing studies.

4.3.2. Main Test

Table 4. Results of the main regression test.

	(1)	(2)
	GreFin	GreFin
NewDigInfra	9.1296 ***	8.7256 ***
	(3.31)	(3.02)
Urban		0.1196 ***
		(8.87)
EduRes		−0.6325 ***
		(−6.50)
FisInv		0.0017 **
		(2.28)
OpenOut		1.7900 **
		(2.03)
EnvPro		0.0004 ***
		(3.07)
MobPop		0.0032
		(0.85)
_cons	0.2538 ***	0.1771 ***
	(35.83)	(14.96)
N	2238	2238
r2_a	0.1369	0.1806
F	21.9457	26.1259

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

reports the benchmark regression results examining the impact of new digital infrastructure construction on the level of green-finance development across prefecture-level cities. As shown in Column (1), NewDigInfra remains positive and significant at the 1% level in the baseline model without controls, implying a strong link between digital infrastructure and green finance. Column (2) incorporates a full set of control variables. The coefficient of NewDigInfra remains significantly positive at the 1% level, with an estimated magnitude of 8.7256. This finding indicates that, holding other factors constant, a 0.001 increase in NewDigInfra is associated with an increase of approximately 0.0087 in the green finance index. The consistent findings across specifications indicate that the positive relationship between digital infrastructure and green finance is robust, thereby supporting Hypothesis H1.

Regarding the control variables, Urban and OpenOut exhibit significantly positive coefficients, implying that urbanization and openness contribute to the expansion of green finance. FisInv and EnvPro also show significantly positive effects, suggesting that fiscal investment intensity and environmental protection efforts facilitate green-finance growth. The significantly negative coefficient of EduRes may reflect structural adjustment costs or transitional frictions associated with higher human-capital accumulation. Taken together, the estimation results provide robust evidence that new digital infrastructure contributes positively to regional green finance development.

4.3.3. Mechanism Test

The theoretical analysis suggests that new infrastructure construction can promote green finance both directly and indirectly by facilitating industrial-structure upgrading, which affects the demand and supply conditions of green finance. New infrastructure facilitates the shift in industrial structures toward higher value-added and service-oriented sectors by promoting technological upgrading, data-factor flows, and the concentration of innovation resources. This expansion increases the number of green investment projects and enhances the alignment between green projects and financial services. Meanwhile, an upgraded industrial structure is typically associated with stronger innovation capacity and improved environmental governance, which in turn substantially raises a region’s receptiveness to and reliance on green finance. Industrial-structure upgrading therefore constitutes a key channel through which new infrastructure affects green finance development.

Table 5. Mechanism test—Regression results.

	(1)	(2)
	UpIndus	GreFin
NewDigInfra	87.7083 ***	9.9939 ***
	(7.69)	(3.54)
UpIndus		0.0333 ***
		(8.84)
Urban	0.4601 ***	0.1098 ***
	(5.05)	(8.19)
EduRes	5.0683 ***	−0.8360 ***
	(9.74)	(−8.63)
FisInv	−0.0282 ***	0.0032 ***
	(−9.28)	(4.10)
OpenOut	−31.5175 ***	1.8263 **
	(−6.54)	(2.11)
EnvPro	0.0024 ***	0.0004 ***
	(3.71)	(3.58)
MobPop	0.0158	0.0029
	(0.98)	(0.77)
_cons	0.6965 ***	0.1437 ***
	(11.39)	(11.71)
N	2238	2238
r2_a	0.2032	0.2037
F	62.1853	32.1036

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

reports the results of the mechanism analysis. As shown in Column (1), NewDigInfra has a positive and statistically significant effect on industrial-structure upgrading (UpIndus) at the 1% level. The result implies that cities with more developed new infrastructure tend to achieve higher levels of industrial upgrading. When NewDigInfra and UpIndus are jointly included in Column (2), both coefficients remain significantly positive at the 1% level, providing empirical support for the hypothesized transmission channel.

To provide further evidence on the reliability of the transmission channel,

Table 6. Mechanism test—Bootstrap results.

Effect	Observed Coefficient	[95% Conf. Interval]
Indirect Effect	3.5187	[2.5535, 4.6408]
Total Effect	13.3788	[9.0973, 17.6432]

reports the Bootstrap-based mediation test results. The confidence intervals for both sets of Bootstrap indirect effects do not include zero, indicating that the mediating effect is statistically significant and robust.

Taken together, Table 5 and Table 6 confirm that new infrastructure construction promotes green finance development partly through industrial-structure upgrading, supporting Hypothesis H2. This result clarifies the role of industrial transformation in connecting digital infrastructure with green capital allocation.

4.3.4. Robustness Test

This study further assesses the robustness of the baseline estimates by adopting alternative dependent-variable measures, winsorizing key variables, and applying an instrumental-variable strategy. The results are reported in

Table 7. Robustness test—Variable replacement and winsorization.

	(1)	(2)
	GreLoan	GreFin_w
NewDigInfra	100.5769 **
	(2.03)
NewDigInfra_w		11.7821 ***
		(3.78)
_cons	2.7663 ***	0.1754 ***
	(12.90)	(14.84)
Controls	YES	YES
N	2223	2238
r2_a	0.1238	0.1822
F	15.8770	27.6576

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

and

Table 8. Robustness test—Instrumental variable method.

	(1)	(2)
	NewDigInfra	GreFin
L.NewDigInfra	0.5507 ***
	(14.28)
NewDigInfra		19.2628 ***
		(3.53)
_cons	−0.0000	0.2672 ***
	(−0.14)	(12.48)
Controls	YES	YES
N	2075	2075
r2_a	0.6818	0.1556
Kleibergen–Paap rk LM statistic	78.21 [0.0000]
Cragg–Donald Wald F statistic	832.91 {16.3800}
Kleibergen–Paap Wald rk F statistic	202.96 {16.3800}
Anderson–Rubin Wald test	11.90 [0.0006]

Note: t statistics in parentheses. *** p < 0.01. Values in brackets [ ] denote p-values, and values in braces { } represent the 10% critical values from the Stock–Yogo weak identification test.

.

First, green credit (GreLoan) is used as an alternative dependent variable to test whether the findings are sensitive to the definition of green-finance development. As shown in Column (1) of Table 7, the coefficient of new infrastructure (NewDigInfra) remains significantly positive at the 5% level, indicating that the main conclusion continues to hold even when green finance is replaced with green credit.

Second, to address potential bias arising from extreme values, all continuous variables are winsorized at the top and bottom 1% and re-estimated. Column (2) of Table 7 shows that, after winsorization, the coefficient of NewDigInfra_w remains positive and significant at the 1% level. This indicates that the positive effect of new infrastructure on green finance is not sensitive to extreme values.

Finally, considering that the construction of new infrastructure may suffer from potential endogeneity issues such as reverse causality or omitted-variable bias, this study further employs an instrumental variable (IV) approach. To construct the instrument, this study uses the one-period lagged value of NewDigInfra. This instrument is correlated with current new infrastructure development but is unlikely to directly affect contemporaneous green finance, thus helping to address endogeneity concerns. As shown in Column (1) of Table 8, L.NewDigInfra is significantly and positively associated with current new infrastructure, indicating strong explanatory power of the instrument. The two-stage least squares estimation reported in Column (2) further shows that the coefficient of new infrastructure remains significantly positive at the 1% level, consistent with the direction of the baseline results.

The instrumental-variable diagnostics support the relevance and strength of the instrument. The significant Kleibergen–Paap LM statistic rejects under-identification, while the Cragg–Donald Wald F statistic and Kleibergen–Paap Wald F statistic both exceed the Stock–Yogo 10% critical value of 16.38, ruling out weak-instrument concerns. The significant Anderson–Rubin Wald test further supports the robustness of the IV estimates.

4.3.5. Heterogeneity Analysis

To further examine whether the impact of new infrastructure construction on green-finance development varies across cities with different characteristics and development conditions, this study conducts heterogeneity analyses from three dimensions: city size, whether a city belongs to an urban agglomeration and financial resource aggregation. The regression results are presented in

Table 9. Heterogeneity analysis—city scale and urban agglomeration.

	(1)	(2)	(3)	(4)
	Cityscale- High	Cityscale- Low	Cluster	Non- Cluster
NewDigInfra	4.8696 **	−0.4387	−2.7766	12.7467 ***
	(2.45)	(−0.10)	(−0.89)	(3.76)
Urban	0.1382 ***	0.1656 ***	0.2384 ***	0.1598 ***
	(11.62)	(8.08)	(13.94)	(10.58)
EduRes	−0.7117 ***	1.0054 ***	−0.5044 ***	0.6537 ***
	(−8.69)	(4.99)	(−4.82)	(4.39)
FisInv	0.0004	0.0012	0.0010	0.0024 ***
	(0.56)	(1.08)	(0.82)	(2.87)
OpenOut	−3.2345 ***	0.8769	−7.0135 ***	3.9900 ***
	(−4.04)	(0.53)	(−5.54)	(3.62)
EnvPro	0.0004 ***	0.0001	0.0002	0.0006 ***
	(3.81)	(0.89)	(1.04)	(4.33)
MobPop	0.0195 ***	−0.2492 ***	0.0151 ***	−0.1571 ***
	(6.59)	(−15.93)	(4.68)	(−11.86)
_cons	0.2065 ***	0.2484 ***	0.2024 ***	0.1804 ***
	(16.80)	(15.79)	(12.47)	(13.23)
N	1111	1112	688	1534
r2_a	0.4910	0.3163	0.5396	0.2389
F	53.2837	18.4064	36.8482	25.3844

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

,

Table 10. Heterogeneity analysis—banking-based financial depth.

	(1)
	GreFin
NewDigInfra	−3.7805
	(−0.80)
FinanDep	−0.0102 ***
	(−3.24)
c.NewDigInfra#c.FinanDep	3.2525 ***
	(3.08)
Urban	0.1187 ***
	(8.76)
EduRes	−0.5381 ***
	(−4.48)
FisInv	0.0013 *
	(1.67)
OpenOut	2.0186 **
	(2.28)
EnvPro	0.0004 ***
	(3.32)
MobPop	0.0046
	(1.22)
_cons	0.1957 ***
	(14.84)
N	2238
r2_a	0.1855
F	24.4240

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

and

Table 11. Marginal effects of new digital infrastructure at different levels of banking-based financial depth.

Quantile of Traditional Financial Depth	Financial Depth Value	Marginal Effect of NewDigInfra	Std. Err.	t-Value	p-Value	95% Confidence Interval
25th percentile	1.701892	1.754975	3.475472	0.50	0.614	[−5.060555, 8.570505]
50th percentile	2.417752	4.083329	3.120567	1.31	0.191	[−2.036219, 10.202880]
75th percentile	3.509634	7.634704 ***	2.892301	2.64	0.008	[1.962794, 13.306610]

Note: t statistics in parentheses. *** p < 0.01.

.

For city-size heterogeneity, cities are classified into high- and low-density groups based on the median population density. As shown in Columns (1) and (2) of Table 9, NewDigInfra significantly promotes green finance in high-density cities, with a coefficient of 4.8696 significant at the 5% level, but has no significant effect in low-density cities, where the coefficient is −0.4387. This suggests that in large cities, where technological infrastructure, digital resources, and financial demand are more robust, new infrastructure can more effectively stimulate green-finance development. In contrast, small cities may be constrained by less developed industrial structures, fewer green projects, or limited financial resources, reducing the ability of new infrastructure to translate into green-finance growth.

Second, heterogeneity is examined from the perspective of urban agglomerations. Based on the China Urban Agglomeration Integration Report, cities are classified into urban-agglomeration and non–urban-agglomeration groups. Columns (3) and (4) show that new infrastructure construction has a significantly positive effect on green finance in non–urban-agglomeration cities, while the effect is insignificant in cities within urban agglomerations. This finding is informative: within urban agglomerations, market mechanisms are more mature, information transparency is higher, and green-finance resources are already concentrated, reducing the marginal effect of new infrastructure. In contrast, in non-agglomeration areas, where infrastructure and financial resources are relatively weaker, new infrastructure substantially improves the information environment and industrial foundation, thereby generating a stronger positive effect on green-finance development.

Finally, regarding heterogeneity in regional financial development, the ratio of outstanding loans and deposits of financial institutions to local GDP is used to measure traditional financial depth. This indicator mainly reflects the degree of financial resource aggregation within the formal banking system. As shown in Table 10, the coefficient of the interaction term is significantly positive at the 1% level, indicating that traditional financial depth strengthens the effect of new digital infrastructure on green finance.

Furthermore, Table 11 shows the marginal-effect results. The marginal effect of new digital infrastructure on green finance is positive at all three quantiles, but its statistical significance differs across financial-depth levels. At the 25th and 50th percentiles, the marginal effects are positive but statistically insignificant. At the 75th percentile, the marginal effect becomes significantly positive at the 1% level. This indicates that the positive effect of new digital infrastructure on green finance becomes more pronounced when traditional financial depth is relatively high. In other words, cities with deeper banking-based financial systems are more capable of transforming the technological and informational advantages of new digital infrastructure into green finance development.

5. Discussion

5.1. Discussion and Conclusions

Based on prefecture-level city panel data in China from 2007 to 2023, this study empirically examines the impact of new infrastructure development on green finance and the underlying mechanisms, and conducts heterogeneity analyses from the perspectives of city size, whether a city is part of an urban agglomeration, and financial development level. The empirical results show that new digital infrastructure significantly promotes the development of green finance. This finding is consistent with previous studies suggesting that digitalization and digital infrastructure can improve information transparency, reduce transaction costs, and enhance the efficiency of financial resource allocation. Different from studies using broader indicators such as the digital economy or informatization, this study focuses specifically on new digital infrastructure and provides city-level evidence on its role in promoting green finance.

The mechanism analysis confirms that industrial structure upgrading serves as a key transmission channel between new digital infrastructure and green finance. This finding suggests that digital infrastructure affects green finance not only through the financial system, but also by reshaping the real economy. By improving information flows, promoting digital transformation, and fostering service-oriented and higher-value-added industries, new digital infrastructure creates more viable green investment opportunities and strengthens demand for green financial instruments. Therefore, green finance development relies on both financial supply and the industrial foundation that supports green investment.

The heterogeneity analysis further reveals that the effect of new digital infrastructure on green finance is not uniform across cities, but varies with local development conditions. Specifically, the positive effect is more pronounced in large cities, cities outside major urban agglomerations, and cities with stronger traditional financial depth. Large cities usually possess more advanced digital infrastructure, a higher concentration of financial resources, and a larger pool of potential green investment projects, which may enable them to convert digital infrastructure advantages into green finance development more effectively. The stronger effect observed in non-urban-agglomeration cities may suggest that new digital infrastructure generates a larger marginal contribution in regions where digital and financial foundations were previously relatively weak. In addition, cities with deeper banking-based financial systems are better positioned to use digital infrastructure to improve green project identification, risk assessment, and financial resource allocation.

Overall, these findings contribute to the green finance literature by identifying new digital infrastructure as a key technological and institutional factor shaping green finance development. They further suggest that the role of digital infrastructure is neither automatic nor homogeneous across cities. Instead, its effectiveness depends on the industrial base, the depth of local financial systems, and broader regional development conditions. Accordingly, policies designed to promote green finance should not be limited to the expansion of green financial instruments. They should also emphasize the coordinated development of digital infrastructure, industrial upgrading, and local financial capacity, so that digital advantages can be more effectively transformed into green financial outcomes.

5.2. Policy Implications

This study provides several implications for government agencies and regional development strategies:

(1)

Strengthening new digital infrastructure as a foundation for green finance development.

Governments should continue to expand the deployment of 5G networks, artificial intelligence, industrial internet, data centers, and other forms of new infrastructure, while advancing reforms in the data factor market and improving digital governance. A more comprehensive digital infrastructure enhances information transparency in green projects, reduces the risk and cost of green finance, and promotes effective capital allocation.

(2)

Promoting regional industrial upgrading to enhance the demand and absorption capacity for green finance.

Policymakers should support the development of advanced manufacturing, green industries, and digital service sectors to expand the scope and quality of green investment projects. Encouraging green innovation and low-carbon technology adoption can strengthen enterprises’ eligibility for green financing, thereby forming a positive cycle between industrial upgrading and green finance.

(3)

Implementing differentiated new infrastructure and green finance policies based on regional characteristics.

Since the effects are stronger in large cities, financially developed regions, and non–urban-agglomeration areas, policies must be tailored accordingly. Large cities should focus on digital governance, data sharing, and innovation in green financial products. Regions with weaker financial systems should simultaneously enhance financial service capacity like expanding local green credit and improving guarantee mechanisms to avoid a mismatch between technological foundations and financial resources. Non-agglomeration cities should leverage new infrastructure to strengthen their information and industrial bases, thereby improving their ability to attract green capital.

5.3. Limitations and Future Directions

Despite providing useful city-level evidence, this study has several limitations that suggest directions for future research:

First, the construction of the new infrastructure index can be improved. The index derived from text analysis of government work reports captures policy intensity but may be influenced by reporting preferences. Future studies could incorporate more objective physical and investment-related indicators, such as the number of 5G base stations, data center computing capacity, broadband infrastructure, and digital economy investment, to develop a more comprehensive measure of new digital infrastructure development.

Second, the green finance indicator system can be further enriched. Although the entropy-based composite index includes green credit, investment, bonds, and other components, it does not cover emerging dimensions such as ESG performance or financing for green technological innovation due to data limitations. Future work may introduce more dynamic and comprehensive green finance indicators.

Third, this study does not fully investigate micro-level mechanisms related to green technological innovation or firm behaviour. While industrial upgrading is identified as a key mediator, microfoundations such as enterprise innovation capacity, environmental governance improvements, and financing behaviour remain to be explored. Future research could incorporate firm-level data to uncover how new digital infrastructure influences green finance through innovation and governance channels.