Can green finance effectively mitigate PM2.5 pollution? What role will green technological innovation play?

Abstract

This study employs panel data encompassing a time frame from 2012 to 2020, collected from 30 provinces in China. By employing a geographic Durbin model and introducing green technological innovation as an intermediary variable, the study explores the relationship between green funds and PM2.5 levels on a spatial scale. The research takes a spatial perspective to explore the links between green finance and PM2.5 emissions, with a specific focus on the intermediary role played by green technology innovation. The findings offer comprehensive insights into enhancing air quality in China, promoting the country's transition towards sustainability, improving the overall human living environment, and generating novel ideas for tackling air pollution challenges. The findings of this study are as follows: (1) The progress of green finance proves to be an effective means of reducing local PM2.5 emissions. Additionally, it generates spillover effects on neighboring regions, promoting the growth of green finance and consequently leading to a decrease in PM2.5 emissions in adjacent areas. (2) In the study exploring the relationship between green financing and PM2.5, green technological innovation plays a crucial mediating role. By efficiently allocating financial resources during China's pivotal green revolution phase, green finance offers funding support to enterprises for the advancement of green technology. This, in turn, contributes to the reduction of PM2.5 emissions. As a consequence, this leads to a decline in energy consumption, pollution emissions, and PM2.5 levels. Additionally, with the continuous improvement in green technological innovation, the reverse effect between green finance and PM2.5 is becoming stronger and stronger. (3) The relationship between the two has obvious regional heterogeneity between the north and south regions of China.

Keywords

Green finance green technological innovation PM2.5

Introduction

Since the initiation of economic reform and opening up, China has witnessed a remarkable and extraordinary growth rate, creating a true marvel. However, this rapid development has had negative consequences for the environment, resulting in its ongoing degradation.¹ Among the various environmental challenges, pollution poses a significant threat to human health, with particular concern surrounding the presence of PM2.5 particles. PM2.5 is a major contributor to haze pollution and serves as an indicator of air pollution levels. PM2.5 particles are characterized by their small size and high concentration of toxic and hazardous substances. These particles can remain suspended in the air for prolonged durations and have high transportability, thereby presenting substantial risks to human health and overall air quality. On 2 November 2021, the authorities released the Opinions on Fighting Pollution Prevention and Control, which outlined specific objectives for addressing PM2.5 pollution and reducing haze pollution. It is clear that substantial efforts are still needed to successfully reduce PM2.5 concentrations and address the issue of particulate pollution.

The growth of China's green industry provides an opportunity to replace these inefficient and polluting sectors. However, emerging green enterprises often face challenges in securing affordable financial support, which hampers their potential impact on pollution reduction. To address this, green finance has emerged as a solution by offering accessible financing for environmentally friendly projects. Green finance encompasses a range of financial services, including credit, bonds, investment funds, insurance, and other support mechanisms for projects related to environmental protection, energy conservation, clean energy, green transportation, and sustainable construction.² It also provides investment and financing assistance for initiatives focused on clean energy and green materials. Figure 1 illustrates the progressive growth of green finance in China, with particular growth observed in provinces such as Shanxi, Shaanxi, Ningxia, Sichuan, Chongqing, and Jiangxi. These regions have received substantial support for their inland economic development, which has stimulated the enhancement of green finance in their central cities. Of great significance is the rapid advancement of green finance in these regions, which has had a notable influence on PM2.5 emissions. This is reinforced by data that demonstrates the beneficial effects of green finance initiatives in mitigating pollution.

Figure 1.

Gfi and PM2.5 concentration distribution map.

The term “green technology innovation” refers to the utilization of technological advancements in environmental protection, resource exploitation, and energy transformation to finish sustainable development goals.³ The role of green technology innovation in mitigating PM2.5 pollution is of utmost importance, as it leverages the advantages offered by green finance to achieve significant reductions in harmful particulate matter. Through the allocation of financial resources towards environmental preservation and the promotion of sustainable development, green finance plays a crucial role in fostering the harmonious and integrated progress of the economy, society, and environment. With the continuous advancement of green sustainable technology, there is an increasing momentum in the advancement of clean and environmentally friendly technology.⁴ The adoption of clean technologies in place of polluting ones can effectively curb PM2.5 emissions and mitigate air pollution. Nevertheless, the journey of green technology innovation encompasses various stages and factors, and it inherently carries certain risks. Financial institutions often hesitate to provide funding to innovative businesses, resulting in a lack of low-cost and high-quality financing sources that can impede their growth. To address this challenge, green finance plays a pivotal role by providing financial support for technological innovation. Through mechanisms such as green securities and green credit, green finance channels substantial and high-quality funds into innovative businesses.⁵ This support enables them to enhance productivity, improve energy efficiency, and promote green economic development. In the end, this can result in a decrease in the emissions of pollutants, such as PM2.5, aligning with the objectives set forth in the “14th Five-Year Plan” for the transition towards a greener and more sustainable economic development.

While numerous studies have been conducted by scholars both domestically and internationally on the correlation between green finance and PM2.5, there are still several limitations that need to be addressed. These limitations can be primarily observed in the following aspects: (1) The analysis has been relatively one-dimensional, lacking a comprehensive and integrated examination of the relationship between green finance and PM2.5. (2) Previous research has mainly focused on the correlation between green finance and PM2.5 without considering the specific context of the green transformation period. As a result, the investigation of the impact of green finance on PM2.5 within this unique period has been neglected. (3) There has been a lack of comprehensive and in-depth research on green technological innovation when studying the influence of green finance on PM2.5. To address these research gaps, this study aims to make the following improvements and developments: (1) The study conducts empirical tests from multiple dimensions, including spatial effects, mediating effects, and threshold effects. By employing spatial Durbin model regression and threshold effect model regression, the previous limitation of one-dimensional analysis is rectified. This approach enables a comprehensive and in-depth exploration of the relationship between green finance and PM2.5, providing novel research perspectives and methods that offer robust support for research and practical applications in related fields. (2) The study narrows its focus to the critical period of China's green development transformation, specifically from 2012 to 2020. By conducting an in-depth investigation during this period, the study not only enriches and enhances relevant theories but also provides scientific decision-making guidance for policymakers to address the opportunities and challenges presented by this distinctive transformation period. (3) By considering green technological innovation as both a mediating variable and a threshold variable, the study aims to better understand the effectiveness of green finance in mitigating PM2.5 pollution and the specific role of green technological innovation within this process.

Literature review

The relationship between green finance development and PM2.5 pollution

In response to China's growing emphasis on environmental preservation and sustainable development, green finance has start as a new research discipline. Feng et al.^6,7 state that financial activities impact the environment and are essential to a nation's sustainable development. Given the increasing severity of environmental issues and the limited financing options available for environmental industries, academics have begun to consider how to defend the environment from the perspective of financial incentives.⁸ Firstly, according to cost theory, cost reduction can incentivize enterprises and individuals to adopt more environmentally friendly and sustainable practices, thereby reducing pollutant emissions. Green finance can provide more favorable financing terms, such as low-interest loans and preferential tax policies, to enterprises and individuals, thus reducing the cost of implementing environmental protection measures.⁹ This can encourage them to take more actions for environmental protection and reduce PM2.5 emissions. Secondly, from a political economy perspective, market economies often suffer from problems such as monopolies and oligopolies, which can lead enterprises to prioritize their own interests and ignore environmental costs.¹⁰ Green finance can offer low-cost financial assistance to firms, making it easier for them to embrace eco-friendly technologies and sustainable practices.¹¹ Additionally, green finance can regulate market behavior and direct capital flow towards the green industry and technological innovation, thereby achieving the integration of environmental protection and economic development. From a Marxist perspective, environmental issues are considered inherent contradictions of capitalist production methods, specifically the contradiction between the pursuit of profit maximization and the need for environmental protection. As an emerging financial tool, green finance has the potential to influence market behavior and steer capital towards the green industry and technological innovation, facilitating the seamless integration of environmental conservation and economic growth.¹² The growth of green finance is in line with the principles of the “externality” theory. The release of PM2.5 has adverse effects on both the environment and human health, but these impacts are often not considered by enterprises and individuals, resulting in market failure. The development of green finance can correct this externality mechanism and encourage enterprises and individuals to consider the environmental costs of their actions.^13,14 This can contribute to the establishment of a more sustainable economic development model. Additionally, the expansion of green finance can encourage existing sectors to transition towards sustainable development. Traditional businesses can invest in energy-saving and emission-reduction technologies and create green goods and services with the support of green finance.¹⁵ This not only contributes to environmental protection but also creates new market opportunities and enhances the competitiveness of enterprises. Furthermore, the expansion of green finance has the potential to attract more social capital towards environmental conservation. As more investors recognize the value of environmental protection, they become interested in investing in green finance products such as green bonds and green funds, which offer stable returns while also contributing to environmental protection.^16,17

Moreover, the expansion of green finance has a ripple effect on pollution management techniques. Not only can it promote local economic development, but it can also stimulate economic growth in neighboring areas. It is important for developing green finance industries, especially in industrialized countries. Thus, green funding reduces regional PM2.5 emissions and stimulates energy and industrial infrastructure modernization, which reduces neighboring PM2.5 emissions and haze-related air pollution.¹⁸

Hypothesis 1 (a):

Green finance holds the potential to reduce PM2.5 emissions in the province, leading to a reduction in local atmospheric pollution.

Hypothesis 2 (b):

Green finance can hinder the release of PM2.5 emissions in nearby provinces, thereby lessening haze pollution in adjacent areas.

Relationship between green finance development and the level of green technology innovation

Green finance has the capacity to drive innovation in green technology primarily by alleviating the financial limitations encountered by environmental protection companies. This, in turn, facilitates investment in research and development of environmental technologies.¹⁹ Firstly, green finance can provide increased funding support, making it easier for environmental protection companies to secure funds and enhance their investment in research on green technology and development. In traditional financial markets, environmental protection companies face high risks when seeking financing.²⁰ However, green finance offers options such as issuing green bonds and establishing green funds, which provide more funding support and reduce the cost of financing for environmental protection companies, thereby encouraging investment in green technology.²¹ Secondly, green finance can utilize incentive mechanisms to promote research, development, and innovation in green technology. By incentivizing companies to develop environmental protection technology, improve the environmental performance of products, and reduce emissions during production processes, the implementation of green finance initiatives stimulates companies to allocate more resources towards the research and development of environmental technology.²² Additionally, green finance can enhance the environmental image and reputation of companies through the rating and certification of environmentally friendly companies, thereby promoting investment in environmental technology research and development.^23,24 Xu et al.²⁵ think that green finance encourages research and development investments and technological innovation in environmental protection firms. However, as green finance develops, environmental regulatory requirements will also increase, and the regulatory costs for companies that do not meet the criteria for green finance funding will rise. This may result in reduced funding for technology innovation and companies facing obstacles in technological innovation due to a lack of funds. Consequently, energy-intensive, inefficient, and polluting businesses will gradually be phased out while “green” businesses experience rapid growth. By promoting the transformation and upgrading of the industrial structure, increasing output, reducing environmental regulatory costs, and improving the long-term market competitiveness of businesses, these efforts will contribute to the acceleration of these positive outcomes.²⁶ From a Marxist perspective, green finance can promote green technology innovation primarily by regulating market behavior, providing financial support, and offering policy guarantees to environmental protection enterprises, thereby stimulating them to engage in green technology application. From the perspective of Marxist political economy, there is an inherent contradiction between private ownership of the means of production and profit maximization in the capitalist economic system.²⁷ Environmental protection and economic growth are also inherently incompatible. In this context, green finance can resolve this contradiction by regulating market behavior and directing capital flow towards environmental protection industries and green technology innovation.

Relationship between PM2.5 pollution and the level of green technology innovation

Innovation in green technology is crucial for shaping a new global energy landscape and attaining sustainable growth through environmentally-friendly economic development.²⁸ At the initial stage, the innovation of green technologies can enhance energy efficiency, decrease energy consumption costs, and subsequently reduce emissions and pollutant generation, leading to a decline in PM2.5 emissions.²⁹ For example, the adoption of new clean energy technologies such as solar, wind, and hydropower can replace conventional fossil fuels, reducing reliance on coal, oil, and other conventional energy sources and, as a result, decreasing PM2.5. According to Jahanger et al.,³⁰ technological advancements between 1990 and 2016 improved energy efficiency in 73 developing countries, leading to reduced carbon footprints and environmental degradation. Additionally, green technology innovation can enhance pollution control effectiveness, reduce the cost of pollution control, and ultimately decrease pollutant emissions and PM2.5 levels.³¹ For example, the adoption of efficient filtering, denitrification, and desulfurization technologies can effectively minimize pollutant emissions, leading to a subsequent reduction in PM2.5. Additionally, the innovation of green technologies can bolster the capability of renewable energy provision, optimize the structure of energy consumption, and facilitate the transition of China's economic development from traditional energy sources to clean energy.³² From a Marxist standpoint, the innovation of green technology can tackle the inherent contradictions of capitalist production methods and diminish the emission of pollutants, which serve as the primary contributor to PM2.5 levels.³³ Through the reduction of energy consumption, utilization of clean energy sources, and efficient management of pollutants, the innovation of green technologies can effectively decrease the cost of pollutant emissions and lead to a reduction in the release of hazardous substances such as PM2.5. Furthermore, green technology innovation can generate new business prospects, stimulate industrial upgrading, and create a mutually beneficial scenario for both economic gains and environmental conservation.³⁴ Specifically, it is common for businesses to implement technological transformations of energy-consuming equipment or invest in higher quality equipment from external sources to achieve high efficiency, low consumption, and cost savings in production factors, ultimately reducing energy consumption costs. Efficient energy usage and consumption will undoubtedly minimize waste and haze pollution from enterprises, enabling polluting firms to quickly achieve green upgrades.⁸

Methodology and data

Variables and data interpretation

Dependent variable

PM2.5 concentration is used as the dependent variable. Since PM2.5 is the primary haze component, its concentration may be used as a proxy for the intensity of the haze problem. The analysis was conducted using grid data of global PM2.5 concentration, which was derived from satellite monitoring data.

Core explanatory variable

This article employs the entropy method to construct green finance indicators, with the following specific steps: First, select indicator $X_{i j}$ , where i is the region, $i = 1, 2, \dots \dots n$ , j is the selected index, $j = 1, 2, \dots \dots, m$ . Following the approach of Yiniu Cui et al.,³⁵ this study constructed the Provincial Green Finance Development Index ( $g f i$ ) through the construction of an indicator system encompassing five key aspects: green finance, green insurance, green securities, green investment, and carbon finance. The data for each province were sourced from the “Social Responsibility Report of the Banking Industry,” the “China Industrial Statistics Yearbook,” the “China Insurance Yearbook,” the “China Environmental Statistics Yearbook,” and each province's “Statistical Yearbook.” Data from 30 provinces (due to missing data, excluding Tibet, Taiwan, Hong Kong, and Macau) from 2012 to 2020 were selected to calculate the green finance indicators, as shown in Table 1.

Table 1.

Green finance indicator system.

Primary indicators	Secondary indicators	Third level indicators
Development level of green finance	Green credit	Level of green credit development
	Green credit	Percentage of high energy-consuming industrial interests
	Green securities	Percentage of market value attributed to environmental protection enterprises
	Green securities	Percentage of market value attributed to high energy-consuming industries
	Green Insurance	Ratio of agricultural insurance scale
	Green Insurance	Scale ratio of agricultural insurance
	Green investment	Ratio of investment in pollution control
	Carbon finance	Degree of carbon decoupling

Second, the dimensionless treatment of indicators is carried out to eliminate the impact of different measurement units. The method is to take the maximum value and minimum value as the endpoints and carry out linear processing. The result of dimensionless processing is between 0 and 1. Positive and negative indications exist. Positive indicators suggest greater scores. Negative indicators reduce scores. Formulas (7) and (8) explain:

Y_{i j} = \frac{X_{i j} - X_{\min}}{X_{\max} - X_{\min}}

(7)

Y_{i j} = \frac{X_{\max} - X_{i j}}{X_{\max} - X_{\min}}

(8)

Where,

X_{\max}

and

X_{\min}

represent the maximum and minimum values of an index,

Y_{i j}

and

X_{i j}

are dimensionless indices.

Third, determine the information entropy. The system's degree of disorder is gauged by the information entropy. While the two have the same absolute values, their signs are the exact opposite. Information is more disordered and has a lower utility value when the information entropy is higher. The opposite is also true. The precise formula is displayed in the following equation:

e_{i j} = - k \sum_{i = 1}^{n} Y_{i j} l n Y_{i j}, k > 0

(9)

Y_{i j} = \frac{r_{i j}}{\sum_{i = 1}^{n} r_{i j}}

represents the proportion of index value of the i region under the j index.

Fourth, calculate the difference coefficient $d_{j}$ . See Formula (10) for details.

d_{j} = 1 - e_{j}

(10)

Fifth, the weight of each indicator is obtained by normalizing the difference coefficient. The specific algorithm is shown in Formula (11).

W_{j} = \frac{d_{j}}{\sum_{i = 1}^{n} d_{j}}

(11)

Sixth, calculate the green finance index indicators of each region. The specific algorithm is shown in Formula (12):

Y_{i j} = \sum_{j = 1}^{n} W_{j} e_{i j}

(12)

Mediating variable

Patents may be regarded as a key indicator of technical innovation since they are the main form of technological innovation. Prior study has utilized patent data to measure technological innovation.^36,37 Despite the fact that patent production may not fully represent the real quality of invention activities, patent data has distinct advantages. Lanjouw et al.³⁸ suggest using patent indicators to measure innovation output, as patent creation and output are related to innovation variables. Since patents are linked to innovation, Kim³⁹ suggests using patents to measure technological innovation. Green patents ( $g r e e n z l$ ) in this paper are considered to be a more accurate reflection of green technological innovation.

Control variables

$p o p$ : The birth rate of the natural population might indicate a country's population increase. An increase in population growth rate can lead to accelerated urbanization and expansion of economic activities, which may increase PM2.5 emissions. Therefore, controlling for population growth rate can eliminate this potential confounding factor. Population growth rate is an important economic and social indicator, and it has a close relationship with environmental quality, resource utilization, and other aspects. Therefore, controlling for population growth rate when study the affection of $g f i$ on PM2.5 can help to better understand the interactions among the economy, society, and environment. Controlling variables can improve the accuracy and reliability of the research. If population growth rate is not controlled, the influence of green finance may be wrongly attributed to population growth rate, leading to biased research conclusions.

$f i n a n c e i m$ : The financial development level of a province, which quantifies the level of financial development within the province, is computed by dividing the number of deposits and loans from banks in each province by the GDP of each province. The amount of financial market growth is often tied to economic activity and urbanization, both of which may have an impact on PM2.5 emissions. Controlling for the amount of financial market development, for example, may prevent this possible influence. The degree of development of financial markets is a significant economic indicator that is directly linked to economic growth and social advancement. Including the degree of financial market development as a control variable can help to provide a more comprehensive understanding of the connections when examining the affection of $g f i$ on PM2.5 across the economy, society, and the environment. Controlling factors may increase the research's accuracy and dependability.

$g a p$ : The disparity in disposable income between urban and rural populations is referred to as the urban–rural income gap, and it is one of the metrics that represent economic inequality in a nation. The income gap between urban and rural areas is often connected to economic activity and urbanization level, which may have an impact on PM2.5 emissions. For example, since urban populations have a better quality of life and more capacity for production and consumption, they may produce more PM2.5. Controlling for the urban–rural income disparity may remove this potentially confusing issue. It is a significant social indicator that is closely linked to issues of social equity and sustainable development. When investigating the influence of green finance on PM2.5 emissions, adjusting for the urban–rural income disparity may assist in better understanding the connection between the economy, society, and the environment. Controlling factors may increase study accuracy and dependability. If the urban–rural income divide is not addressed, the influence of green finance may be incorrectly attributed to it, leading to biased study outcomes.

$g t f p$ : It is calculated based on unexpected output, which measures and calculates industrial development and economic growth using the undirected SBM model and the GML index. Controlling for green total factor production may thus remove this potentially confusing effect. Green total factor productivity is a key statistic for assessing the efficacy of green finance. If the influence of green finance is not properly controlled, the influence of green finance may be incorrectly attributed to green total factor productivity, leading to biased study results. So, we can better find the regional affection of $g f i$ on PM2.5 by adjusting for green total factor production. Simultaneously, accounting for green total factor production may assist academics in better understanding the link between economic growth, environmental protection, and sustainable development.

$e r i$ : The environmental regulation index serves as a tool for regulatory authorities to guide businesses in reducing pollutant emissions, achieving environmental protection, and promoting robust economic growth. In this study, entropy technology was employed to compute the composite index of environmental indicators in each region, including modern wastewater, sulfur dioxide, and residue emissions. By controlling for environmental regulation, the potential impact of a company's production and business behavior on pollution emissions can be accounted for, eliminating confounding factors. Environmental regulation is a crucial factor in assessing the effectiveness of green finance, which aims to foster sustainable economic and environmental growth. Therefore, incorporating controls for environmental regulation allows for a more comprehensive evaluation. Controlling for variables enhances research accuracy and reliability, as failing to account for environmental regulation may lead to biased conclusions and wrongly attribute the impact of green finance on environmental regulation.

$h u m a n$ : The mean years of education for individuals aged six and above in the population is an indicator of a country's overall education level and quality. The improvement of education level may enhance environmental awareness and reduce PM2.5 emissions. Furthermore, the improvement of education level may also improve economic development and reduce pollution to the environment. Therefore, education level is an important social indicator that is closely related to social equity and sustainable development. Therefore, controlling education level when studying the influence of green finance on PM2.5 emissions can enhance our understanding about the interaction between the economy, society, and environment. Controlling for variables can improve research accuracy and reliability. Failure to control for education level may cause researchers to attribute the influence of green finance to education level, leading to biased research conclusions.

$e n e r g y c$ : PM2.5 emissions are primarily attributed to energy consumption. In areas with higher energy consumption, there may be more pollution emissions, resulting in higher levels of PM2.5 emissions. Therefore, controlling for energy consumption can eliminate this potential confounding factor. Energy consumption is a significant metric for assessing the efficacy of green finance. The primary objective of green finance is to foster sustainable economic and environmental development, and a reduction in energy consumption is a crucial indicator of its success. Hence, incorporating controls for energy consumption enables a more comprehensive evaluation of its effectiveness. Controlling for variables can improve research accuracy and reliability. Failure to control for energy consumption may cause researchers to attribute the influence of green finance to energy consumption, leading to biased research conclusions.

Methodology

OLS panel regression model construction

This research initially creates a simple OLS model to analyze the affection of $g f i$ on PM2.5 in order to test the aforementioned assumptions. The following shows the model:

\begin{aligned} p m {2.5}_{i t} = & a_{0} + β_{1} g f i_{i t} + β_{2} p o p_{i t} + β_{3} f i n a n c e i m_{i t} + β_{4} g a p_{i t} + β_{5} g t f p_{i t} \\ + β_{6} e r i_{i t} + β_{7} e n e r g y c_{i t} + β_{8} h u m a n_{i t} + ε_{i t} \end{aligned}

(13)

In equation (13), the variables

p m {2.5}_{i t}

and

g f i

represent respectively dependent variable and core explanatory variable.

a_{0}

represents a constant. The coefficients β₁ to β₈ represent the regression variables. Control variables include

p o p

f i n a n c e i m

g a p

g t f p

e r i

h u m a n

, and

ε

represents the stochastic error.

Spatial Durbin model

Prior to spatial model application, geographic autocorrelation model regression must be performed to assess the applicability of the model. This study uses a geographic distance matrix to calculate global Moran's I. The solution is depicted by equation (14):

Mora n^{'} s I = \frac{\sum_{i = 1}^{n} \sum_{j = 1}^{n} w_{i j} (Y_{i} - \bar{Y}) (Y_{j} - \bar{Y})}{s^{2} \sum_{i = 1}^{n} \sum_{j = 1}^{n} w_{i j}}

(14)

s^{2} = \frac{1}{n} \sum_{i = 1}^{n} {(Y_{i} - \bar{Y})}^{2}, \bar{Y} = \frac{1}{n} \sum_{i = 1}^{n} Y_{i}

Mora n^{'} s I

represents the spatial autocorrelation, and

- 1 \leq I \leq 1

Y_{i}

denotes the region

i

；

w_{i j}

is the spatial weight matrix.

The spatial Durbin model (SDM) can be written as:

p m {2.5}_{i t} = ρ \sum_{j = 1}^{n} W_{i j} p m {2.5}_{j t} + β_{1} g f i_{i t} + θ_{1} \sum_{j = 1}^{n} W_{i j} g f i_{j t} + λ X_{i t} + μ_{i} + λ_{t} + ε_{i t}

(15)

In equation (15), W,

θ

, and

ρ

represent respectively the geographical distance spatial weight matrix, the coefficient of explanatory variables in spatial regression, and the spillover effect of explanatory variables in spatial analysis.

μ_{i}

is the space-fixed effect,

μ_{i}

denotes spatial specific effect,

λ_{t}

is the time specific effect, and

ε_{i t}

is a random error term.

Mediating effect model

In this model, M represents the intermediate variable between X and Y, indicating that the main variable X influences the explanatory variable Y through another variable. The mediation effect model is useful for establishing connections between original research on a particular phenomenon and its underlying causes. In the context of investigating the impact mechanism of $g f i$ and PM2.5, the intermediate effect model is particularly suitable for examining the role of the intermediary variable green technological innovation $(g r e e n z l$ ). Figure 2 illustrates the precise mechanism of this impact.

Figure 2.

Path analysis diagram.

The specific mediation model is indicated by Formulas (16) to (18):

\begin{aligned} p m {2.5}_{i t} = & a_{0} \cdot + β_{1} g f i_{i t} + β_{2} p o p_{i t} + β_{3} f i n a n c e i m_{i t} + β_{4} g a p_{i t} + β_{5} g t f p_{i t} \\ + β_{6} e r i_{i t} + β_{7} e n e r g y c_{i t} + β_{8} h u m a n_{i t} + ε_{i t} \end{aligned}

(16)

\begin{aligned} g r e e n z l_{i t} = & a_{0} + β_{1} g f i_{i t} + β_{2} p o p_{i t} + β_{3} f i n a n c e i m_{i t} + β_{4} g a p_{i t} + β_{5} g t f p_{i t} \\ + β_{6} e r i_{i t} + β_{7} e n e r g y c_{i t} + β_{8} h u m a n_{i t} + ε_{i t} \end{aligned}

(17)

\begin{aligned} p m {2.5}_{i t} = & a_{0} + β_{1} g r e e n z l_{i t} + β_{2} g f i_{i t} + β_{3} p o p_{i t} + β_{4} f i n a n c e i m_{i t} + β_{5} g a p_{i t} \\ + β_{6} g t f p_{i t} + β_{7} e r i_{i t} + β_{8} e n e r g y c_{i t} + β_{9} h u m a n_{i t} + ε_{i t} \end{aligned}

(18)

g r e e n z l

is an mediating variable;

a_{0}

is a constant,

β_{1} \sim β_{9}

are the coefficients of the relevant influencing factors; control variables include

p o p

f i n a n c e i m

g a p

g t f p

e r i

h u m a n

, and

ε

represents the random error term.

Construction of the threshold model

To investigate the influence of carbon-neutral development on high-quality economic growth and the role of total factor productivity in this process, this study employs a panel threshold regression model inspired by methodology. The formulation of the threshold regression model is as follows:

Y_{i t} = α_{i} + 1 (q_{i t} \leq γ_{1}) X_{i t}^{^{'}} β_{1} + 1 (γ_{1} < q_{i t} \leq γ_{2}) X_{i t}^{^{'}} β_{2} + 1 (q_{i t} < γ_{2}) X_{i t}^{^{'}} β_{3} + β_{4} X_{i t} + ε_{i t}

(19)

In Formula (7), the threshold value

γ_{1} < γ_{2}

, the indicative function (·). 1

(q_{i t} \leq γ_{1}) = {\begin{matrix} 1 if q_{i t} \leq γ_{1} \\ 0 if q_{i t} > γ_{1} \end{matrix}

1 (γ_{1} < q_{i t} \leq γ_{2}) = {\begin{matrix} 1 if γ_{1} < q_{i t} \leq γ_{2} \\ 0 if q_{i t} \leq γ_{1} or q_{i t} > γ_{2} \end{matrix}

；

1 (q_{i t} > γ_{2}) = {\begin{matrix} 0 if q_{i t} \leq γ_{2} \\ 1 if q_{i t} > γ_{2} \end{matrix}

X_{i t}^{^{'}}

represents the threshold variable,

X_{i t}

represents the control variable, and

γ

represents the threshold value.

Descriptive statistics

To address the issue of heteroscedasticity in the regression analysis, this study employs a logarithmic transformation on key variables such as population growth, PM2.5 emissions, and the number of green patents. Table 1 presents the descriptive statistics for each variable. It is evident that PM2.5 levels vary significantly across regions, with a maximum value of 4.4507, a minimum value of 2.3079, a mean of 3.6347, and a median of 3.6615. Some regions experience lower levels of PM2.5 pollution, while others exhibit higher levels. Additionally, the maximum value of the gfi development level is 0.7930, the minimum value is 0.0620, the average value is 0.1854, and the median value is 0.153, indicating substantial regional disparities and a notable growth in gfi in recent years (Table 2).

Table 2.

Descriptive statistics.

	Observation	Median	Mean	SD	Min	Max
pm2.5	270	3.6615	3.6247	0.3844	2.3079	4.4507
gfi	270	0.153	0.1854	0.1082	0.0620	0.7930
pop	270	5.655871	5.4830	1.2727	2.0719	8.2560
financeim	270	2.97863	3.1754	1.1344	1.5296	7.9006
gap	270	1.695	1.5259	0.9489	0.0000	4.4900
gtfp	270	1.606105	1.7248	0.5884	0.8489	4.9789
eri	270	0.34	0.5205	0.5335	0.0000	2.5900
human	270	2.206625	2.2118	0.0923	2.0109	2.5479
energyc	270	9.380758	9.4277	0.6497	7.3781	10.6308
greenzl	270	7.588824	7.5546	1.3532	3.2581	10.4360

Empirical design and results analysis

Data stability test

This paper conducts data stationarity test on the data, including unit root test and panel cointegration test. Table 3 shows that: (1) Each variable has passed at least three-unit root tests, namely FISHER, LLC, IPS, and HADRI, so the selected variable is reasonable; (2) Panel data passed the Kao cointegration tests, which means that there is a long-term stable balance between variables, so panel regression can be performed.

Table 3.

Data stability test results.

Unit root test		pm2.5	gfi		pop	financeim	gap
FISHER	Inverse chi-squared	201.660***	129.663***		149.908***	51.359	2162.619***
	Inverse normal	−4.096***	−2.107**		−1.189	3.855	−44.507***
	Inverse logit t	−7.511***	−3.207***		−3.817***	4.090	−109.200***
	Modified inv. chi-squared	12.932***	6.359***		8.206***	−0.789	191.942***
LLC	Adjusted t	−17.373***	−42.704***		−31.370***	−29.996***	1.944
IPS	W-t-bar	0.197	−17.578***		−5.181***	−7.182***	−0.037
HADRI	z	19.592***	22.107***		16.886***	18.323***	17.961***
Unit root test		gtfp	eri		energyc	greenzl	human
	Inverse chi-squared	98.643***	110.185***		189.558***	177.542***	309.866***
FISHER	Inverse normal	2.683	0.593		−0.402	−3.023***	−6.427***
	Inverse logit t	1.334	−0.924		4.053***	−5.462***	−13.204***
	Modified inv. chi-squared	3.528***	4.581***		−3.695***	10.730***	22.810***
LLC	Adjusted t	−1.929**	−13.024***		−8.375***	4.816	−54.248***
IPS	z	5.550	−4.913***		2.480	−9.367***	−9.355***
HADRI	W-t-bar	18.114***	22.246***		17.300 ***	18.056***	21.882***
panel cointegration tests				statistic			p-value
Kao	Modified Dickey-Fuller t			−2.0482	0.0203
	Dickey-Fuller t			−4.6431	0.0000
	Augmented Dickey-Fuller t			−2.9516	0.0016
	Unadjusted modified Dickey-Fuller t			−3.6772	0.0001
	Unadjusted Dickey-Fuller t			−5.4404	0.0000

*p < 0.1, **p < 0.05, ***p < 0.01.

OLS panel regression and analysis

With a value of −0.836, Table 4 demonstrates a substantial inverse relationship between $g f i$ and PM2.5 at the 1% level. This shows that green capital makes a big difference in lowering PM2.5 pollution. $g f i$ can help new green businesses get funds through green credit, green loans, etc. It can also help traditional businesses go green by transforming and improving,⁴⁰ which cuts down on pollution like PM2.5. So, 1a is true. At the same time, we found a significant negative association between the income gap between urban and rural areas and PM2.5, with a coefficient of −0.103 at the 1% level. The narrowing of urban–rural income gap means the economic development and improvement of residents’ living standards in rural areas, which may lead to changes in the energy consumption structure in rural areas. Relatively low-income rural residents may be more inclined to use traditional, environmentally friendly energy sources such as firewood and loose coal, which generate more PM2.5 emissions due to their combustion. As rural residents’ income increases, they may be more able to purchase cleaner and environmentally friendly energy, thereby reducing PM2.5 emissions. And as income levels increase, people's attention to environmental issues may increase and they are more willing to take environmental measures to reduce pollutant emissions, such as using cleaner energy and improving rural living environments. The outcome from Table 4 indicates a significant impact of both the birth rate and energy consumption on PM2.5 emissions. The coefficients for these variables are 0.148 and 0.151, respectively, and they are statistically significant at the 1% level. Hence, we can deduce that an elevation in both the birth rate and energy consumption results in elevated PM2.5 emissions. The burning of fossil fuels, including coal, oil, and natural gas, in various energy-consuming activities releases significant amounts of particulate matter and harmful gases, including PM2.5. As the population increases, resource consumption also rises, leading to a subsequent increase in PM2.5 emissions.

Table 4.

OLS, PCSE, FGLS regression results.

	Model (1a) (OLS)	Model (1b) (PCSE)	Model (1b) (FGLS)	vif
gfi	−0.836***	−0.836***	−0.836***	3.87
	(0.276)	(0.114)	(0.271)
pop	0.148***	0.148***	0.148***	1.82
	(0.016)	(0.008)	(0.016)
financeim	−0.015	−0.015	−0.015	2.78
	(0.022)	(0.012)	(0.022)
gap	−0.103***	−0.103***	−0.103***	1.72
	(0.021)	(0.035)	(0.021)
gtfp	−0.052*	−0.052	−0.052*	1.37
	(0.030)	(0.043)	(0.030)
eri	0.050	0.050*	0.050	2.80
	(0.048)	(0.028)	(0.048)
energyc	0.151***	0.151***	0.151***	3.04
	(0.041)	(0.024)	(0.040)
human	1.734***	1.734***	1.734***	2.42
	(0.256)	(0.146)	(0.252)
_cons	−2.022***	−2.022***	−2.022***
	(0.587)	(0.270)	(0.577)
Hausman(p-value)	0.000
R²	0.593	0.593
Mean vif				2.48
ID	yes	yes	yes
YEAR	yes	yes	yes
Obs	270
Province	30

*p < 0.1, **p < 0.05, ***p < 0.01.

At the same time, this study used PCSE and FGLS regression to rule out the effects of heteroscedasticity, autocorrelation, and cross-sectional correlation on the regression results. Table 4 displays the regression findings. A variance inflation factor test was also performed to rule out the impact of variable multicollinearity. The results indicated that the VIF value of each variable did not exceed 5, indicating that multicollinearity between variables had no effect on the regression outcomes (Table 5).

Table 5.

Value of Moran's index.

Year	pm2.5	gfi
2012	0.2380***	0.0270***
	(0.0920)	(0.0810)
2013	0.2120***	0.2230***
	(0.0900)	(0.0810)
2014	0.2350***	0.2190***
	(0.0930)	(0.0800)
2015	0.2880***	0.2230***
	(0.0920)	(0.0780)
2016	0.2760***	0.2040***
	(0.0930)	(0.0780)
2017	0.2660***	0.1750***
	(0.0930)	(0.0750)
2018	0.2540***	0.1460***
	(0.0910)	(0.0730)
2019	0.2170***	0.2140**
	(0.0940)	(0.0770)
2020	0.2320***	0.2210***
	(0.0920)	(0.0770)

*p < 0.1, **p < 0.05, ***p < 0.01.

Spatial regression and analysis

Moran index test

We calculated the Moran index for global PM2.5 and $g f i$ from 2012 to 2012 using equation (2) prior to constructing the spatial model. The results of the geographical distance matrix are displayed in Table 4. According to Table 4, from 2012 to 2020, PM2.5 and the $g f i$ Moran index have substantial spatial autocorrelation, with the exception of 2019, when the $g f i$ Moran index is significant at the 5% significance level. In all other years, PM2.5 and the $g f i$ Moran index are significant at the 1% significance level. Therefore, the spatial model can be used for related study.

In Figure 3, the Moran Scatter plot illustrates the spatial relationship between PM2.5 and $g f i$ in 2020.

Figure 3.

The scatter plot of Moran's I in 2020.

Selection and testing of spatial regression models

As previously indicated, PM2.5 exhibits a significant spatial correlation. Consequently, spatial modeling regression is necessary. As shown in Table 6, we then conducted selection tests for spatial models and regression types. Fixed effects were chosen since the Hausman test was significant at 1%. The likelihood ratio test showed strong geographical and fixed effects over time. The Wald and LR tests of the SLM and SEM rejected the null hypothesis that the spatial Durbin model can be reduced to either SLM or SEM at the 1% level. This work used the spatial Durbin model for regression analysis.

Table 6.

Testing of spatial regression models.

Hausman	80.26***
LR_ind	18.28**
LR_time	826.44***
R²	0.400
Wald_error	22.440***
Wald_lag	15.270**
Log-likelihood	7 427.9110
LR_error	20.310***
LR_lag	15.020**

*p < 0.1, **p < 0.05, ***p < 0.01.

Spatial Durbin model regression and analysis

Based on the findings presented in Table 7, the effect of $g f i$ on PM2.5 is strongly negative. This suggests that the advancement of $g f i$ at the provincial level not only reduces PM2.5 emissions locally but also has a spill-over effect on adjacent provinces, as evidenced by the significant negative correlation coefficient of −1.042 at the 5% level. This is due to the possibility of $g f i$ providing financial help to both internal and foreign provinces, resulting in greener industrial structures, more effective resource utilization, and fewer PM2.5 emissions. Furthermore, $g f i$ encourages the progressive strengthening of regional economic integration and the establishment of industrial chains.^36,37,41 Therefore, the implementation of $g f i$ has accelerated regional green development and established a reciprocal influence and promotion model, resulting in a decline in spatial PM2.5 emissions. This supports hypothesis 1(b). The research findings also indicate a significantly positive direct influence of $f i n a n c e i m$ on PM2.5 emissions, with a coefficient of 0.025 at the 10% significance level. Additionally, the geographical spillover effect is strongly positive, with a value of 0.042. These findings confirm that financial activities have a substantial favorable influence on PM2.5 emissions. The level of economic development within a province has a highly positive impact on its own PM2.5 emission levels. Meanwhile, the spatial promotion effect is even more pronounced. This suggests that the level of economic development in a province has a strongly promotional affection on PM2.5 of neighboring regions. The fact that this province's financial development level has a considerable favorable influence on its own PM2.5 emissions demonstrates this. The degree to which financial institutions have developed is a key indicator of the financial development of an area, and well-developed financial institutions are more effective at absorbing financial resources from nearby areas. This causes production factors in the nearby areas to move to the financially developed areas, lowering the economic output of the nearby areas. Therefore, compared to neighboring provinces, this province's PM2.5 emissions are less affected by its degree of financial development.

Table 7.

Regression results of spatial Durbin model.

	Main	WX	Direct	Indirect	Total
gfi	−1.084***	−1.042**	−1.312***	−3.874***	−5.187***
	(0.220)	(0.470)	(0.256)	(1.444)	(1.586)
pop	0.382	1.623**	0.639**	4.418***	5.057***
	(0.266)	(0.652)	(0.280)	(1.688)	(1.816)
financeim	0.025*	0.042*	0.034**	0.132**	0.166**
	(0.013)	(0.024)	(0.014)	(0.063)	(0.072)
gap	0.062***	0.081*	0.078***	0.273**	0.350**
	(0.022)	(0.048)	(0.024)	(0.131)	(0.143)
gtfp	−0.003	0.018	−0.001	0.036	0.035
	(0.012)	(0.023)	(0.012)	(0.056)	(0.060)
eri	0.048**	−0.060	0.045**	−0.065	−0.020
	(0.020)	(0.062)	(0.022)	(0.150)	(0.160)
energyc	0.142***	−0.433***	0.092	−0.834***	−0.742**
	(0.054)	(0.134)	(0.058)	(0.316)	(0.337)
human	0.067	−0.039	0.058	−0.062	−0.004
	(0.209)	(0.638)	(0.251)	(1.595)	(1.761)
RHO	0.631***
	(0.056)
sigma2_e	0.002***
	(0.0002)
OBSERVATION	270
PROVINCE	30

*p < 0.1, **p < 0.05, ***p < 0.01.

By decomposing spatial effects into direct and indirect effects using partial differentiation, spatial spillover effects can be analyzed more precisely and with fewer errors. $g f i$ has significant direct affection, indirect affection, and total affection on PM2.5 with coefficients of −1.312, −3.874, and −5.187, respectively. And the regression results are significant at the 1% significance level. $g f i$ has a considerable negative influence on PM2.5 in this province, as well as a negative geographical spillover effect that grows with distance. $g f i$ development in this province may spur growth in neighboring provinces, particularly in economically wealthy places, which may have a “radiation effect” on the development of underdeveloped areas. The development of $g f i$ exhibits a stronger indirect inhibitory impact on PM2.5 emissions in neighboring provinces compared to its effect within the province itself.

Path analysis

Since the traditional three-step method omits endogenous variables in the regression, this paper conducts Sobel test with reference to Sobel's method. The outcomes are shown in Table 8, with a coefficient of −0.500, which is significant at the level of 1%, indicating that green technology innovation has a negative mediating affection in the mechanism of $g f i$ on PM2.5 emissions. The results of Goodman-1 (Aroian) and Goodman-2 are consistent, which proves that the mediation mechanism testing conducted in this study is scientific and robust.

Table 8.

Mediating effect mechanism test results.

Soble	−0.5000 ***(0.1880)
Goodman-1	−0.5000 ***(0.1890)
Goodman-2	−0.5000 ***(0.1870)

*p < 0.1, **p < 0.05, ***p < 0.01.

To examine the mediating role of $g r e e n z l$ between $g f i$ and PM2.5 emissions, this study employs a mediating effect model. The results of the indirect regression analysis for path (b) reveal a significantly positive relationship between $g f i$ and the development of $g r e e n z l$ , with a coefficient of 6.189 at the 1% significance level. This indicates that the advancement of $g r e e n z l$ is strongly influenced by the progress of $g f i$ , thus supporting Hypothesis 2. Furthermore, as per Path (c), $g r e e n z l$ exhibits a prominent negative impact on PM2.5 emissions at the 1% significance level, with a coefficient of −0.081, thereby confirming Hypothesis 3. Table 9 demonstrates that the cumulative effect of $g f i$ on the reduction of PM2.5 emissions in Path (a) surpasses its direct impact in Path (c), with values of −0.836 and −0.335 respectively. This is due to Path (a) incorporating the suppressive effect of $g r e e n z l$ on PM2.5 emissions, while Path (c) controls for the restraining impact of $g r e e n z l$ on PM2.5 emissions in $g f i$ , resulting in a reduction of the suppressive influence of $g f i$ on PM2.5 emissions.

Table 9.

Regression of mediating effect.

	Path (a)	Path (b)	Path (c)
	$p m 2.5$	$g r e e n z l$	$p m 2.5$
$g r e e n z l$	-	-	−0.081***(0.029)
$g f i$	−0.836***(0.276)	6.189***(0.573)	−0.335*(0.328)
$p o p$	0.148***(0.016)	0.421***(0.033)	0.182***(0.020)
$f i n a n c e i m$	0.015(0.022)	−0.168**(0.046)	−0.029(0.023)
$g a p$	−0.103***(0.021)	0.417***(0.044)	−0.069***(0.024)
$g t f p$	−0.052*(0.030)	0.025(0.063)	−0.054*(0.030)
$e r i$	0.050(0.048)	−0.660***(0.099)	−0.004(0.051)
$e n e r g y c$	0.151***(0.041)	1.246***(0.085)	0.252***(0.055)
$h u m a n$	1.734***(0.256)	−2.217***(0.532)	1.555***(0.261)
_cons	−2.022***(0.587)	−2.599**(1.218)	−2.232***(0.585)
YEAR	YES	YES	YES
ID	YES	YES	YES
R²	0.593	0.870	0.604
OBSERVATION	270	270	270
PROVINCE	30	30	30

*p < 0.1, **p < 0.05, ***p < 0.01.

According to empirical data, $g f i$ may support the development of $g r e e n z l$ , and green technology upgrade can diminish PM2.5 emissions.^42,43 By matching risks and returns in a reasonable manner, green finance tools can alleviate financing constraints for businesses, thereby increasing funding sources for high-risk technology research and development activities and ensuring the development of $g r e e n z l$ activities for research-based businesses. In the interim, $g r e e n z l$ is crucial to accomplishing coordinated economic growth and environmental protection, comprised primarily of new technologies and services that diminish environmental contamination and promote sustainable development. One prominent form of innovation in use today is $g r e e n z l$ , which focuses on upgrading industrial and energy structures. This type of innovation aims to enhance production efficiency, reduce reliance on traditional energy sources, and minimize pollutant emissions.^{36,37,44–46} As a consequence, the fast progress of environmentally friendly technological innovation contributes to the reduction of PM2.5 emissions.

Discussion on robustness test result

Robustness test and endogenous test of OLS regression

Using the method of replacing the explained variable and two-stage least squares, this study yields robust and trustworthy regression results. Since the emission of sulfides into the atmosphere is primarily responsible to produce PM2.5, this study tests the robustness by substituting sulfur dioxide emission concentration for PM2.5 emission concentration, and the results are shown in Table 10 (1). Moreover, employing the two-stage least squares technique, the results are shown in Table 10 (2). To address the issue of endogeneity in the regression analysis, this study incorporates first-order lag on all control variables, and the results are shown in Table 10 (3).

Table 10.

Robustness & endogeneity test.

	(1) so2	(2) 2sls	(3) Lagged first-phase regression
gfi	−1.156***	−1.743**	gfi	−1.279***
	(0.302)	(0.556)		(0.281)
pop	0.0001	0.502	pop1	0.508
	(0.017)	(0.317)		(0.390)
financeim	−0.079***	−0.003	financeim1	0.035**
	(0.019)	(0.015)		(0.017)
gap	−0.034	0.145	gap1	0.079***
	(0.052)	(0.104)		(0.028)
gtfp	−0.035	0.007	gtfp1	−0.063*
	(0.032)	(0.012)		(0.036)
eri	0.193***	0.040	eri1	0.005
	(0.038)	(0.022)		(0.026)
energyc	0.842***	0.101	-	0.143**
-	(0.038)	(0.091)	-	(0.071)
human	0.319	0.190	-	0.216
-	(0.226)	(0.247)	-	(0.265)
cons	5.139***	−0.579	cons	−0.78
	(0.544)	(2.254)		(2.541)
Obs	270	270	Obs	240
R²	0.915	0.982	R²	0.977

*p < 0.1, **p < 0.05, ***p < 0.01.

Robustness test of spatial Durbin regression

In this study, we conducted a spatial regression analysis by substituting the spatial geographic distance weight matrix with a binary adjacency spatial weight matrix. The results of this analysis are reported in Table 11, and the regression findings align consistently with those presented in Table 7, which utilized the geographical geographic distance weight matrix. Consequently, we can infer that the regression results obtained from the spatial Durbin model are dependable and trustworthy.

Table 11.

0-1 matrix space Durbin regression model regression results.

	Main	WX	Direct	Indirect	Total
gfi	−0.726***	−0.793*	−1.005***	−3.115**	−4.120**
	(0.203)	(0.399)	(0.250)	(1.096)	(1.258)
pop	0.201	1.129*	0.489	3.240*	3.730**
	(0.222)	(0.472)	(0.254)	(1.292)	(1.446)
financeim	0.013	0.025	0.022	0.083	0.105
	(0.011)	(0.019)	(0.013)	(0.048)	(0.056)
gap	0.045*	0.051	0.063**	0.2	0.263*
	(0.019)	(0.043)	(0.023)	(0.112)	(0.126)
gtfp	0.011	−0.022	0.007	−0.041	−0.034
	(0.011)	(0.022)	(0.012)	(0.051)	(0.056)
eri	0.051**	−0.051	0.048*	−0.043	0.004
	(0.018)	(0.031)	(0.021)	(0.084)	(0.096)
energyc	0.120*	−0.201	0.091	−0.338	−0.247
	(0.051)	(0.108)	(0.056)	(0.247)	(0.271)
human	0.008	−0.608	−0.142	−1.588	−1.73
	(0.182)	(0.404)	(0.228)	(1.129)	(1.289)
rho	0.579***
	(0.077)
OBSERVATION	270
PROVINCE	30

*p < 0.1, **p < 0.05, ***p < 0.01.

Extended analysis

Threshold regression analysis

To further investigate the role of green technology innovation in the relationship between $g f i$ and PM2.5, a threshold effect regression analysis was conducted in this study. The obtained threshold values were tested using the self-service sampling method, and the results are presented in Table 12. The F-value for the single threshold is 29.37, with a corresponding P-value of 0.000. For the double threshold, the F-value is 15.35, and the P-value is 0.090. However, the F-value for the three thresholds is 18.22, with a P-value of 0.610. As a result, the single-threshold test and double-threshold test were deemed significant, while the three-threshold test did not meet the significance criteria. Consequently, a double threshold regression analysis was conducted in this study.

Table 12.

Threshold inspection results.

Threshold	Rss	Mse	Fstat	P-value	crit10	crit5	crit1	Bs-time	Th-1
Single	1.253	0.005	29.37	0.000	15.429	16.707	21.431	1000	0.146
Double	1.183	0.005	15.35	0.090	14.706	18.425	28.992	1000	0.182
Triple	1.106	0.004	18.22	0.610	36.588	39.735	47.521	1000	-

By observing Figure 4, the red dotted line represents the critical value of 5%. It can be seen that the threshold values in this paper are all below it, indicating that the selection of threshold values is appropriate.

Figure 4.

Threshold regression result graph.

Based on the results depicted in Table 13, the influence of $g f i$ on PM2.5 is contingent upon the threshold degree of green technology innovation. As the level of green technology innovation increases, the inhibitory effect of $g f i$ on PM2.5 becomes stronger. Moreover, as the green technology innovation level continues to improve, the restraining effect of $g f i$ on PM2.5 becomes increasingly significant, with a significance level of 1%. Specifically, when the green technology innovation level is less than or equal to 0.146, the restraining effect of $g f i$ on PM2.5 is −0.123. When the green technology innovation level falls between 0.146 and 0.182, the restraining effect of $g f i$ on PM2.5 is −0.137. Lastly, when the green technology innovation level is equal to or greater than 0.182, the restraining effect of $g f i$ on PM2.5 is −0.146. The reasons for this nonlinear relationship may be as follows: firstly, there is a technological threshold effect in green technology innovation, which may require a certain technological threshold to achieve significant environmental benefits. In the early stages of technological innovation, there may be higher costs and lower technological maturity, so the inhibitory effect on PM2.5 may be limited. However, with the further development and maturity of technological innovation, the technological threshold gradually decreases, and the inhibitory effect of $g f i$ on PM2.5 may significantly increase. Secondly, the environmental system is a complex system, and the causal relationship may not be a simple linear relationship. The relationship between green technology innovation and PM2.5 inhibition may be influenced by various factors, such as environmental background, industrial structure, policy support, etc. The interaction of these factors may lead to the emergence of nonlinear effects. In conclusion, it is important to consider the possibility of threshold effects when examining the relationship between green technology innovation and green finance in relation to PM2.5. Beyond a certain threshold level of green technology innovation, breakthrough effects or positive feedback effects may occur. This means that when the level of green technology innovation reaches a certain point, there could be a significant increase in the inhibitory effect of $g f i$ on PM2.5, potentially leading to explosive technological progress. Therefore, it is crucial to take these factors into account and employ appropriate analytical methods to understand and explain the nonlinear relationship. By doing so, better-informed decisions can be made regarding the implementation of green technology innovation and $g f i$ to effectively address PM2.5 pollution.

Table 13.

Threshold effect regression.

Variable
pop	−0.039
	(0.335)
financeim	−0.036**
	(0.150)
gap	0.033***
	(0.010)
gtfp	−0.085***
	(0.013)
eri	0.045
	(0.030)
energyc	0.039
	(0.070)
human	−0.818***
	(0.291)
gfi( $g r e e n z l \leq$ 0.146)	−0.123***
	(0.149)
gfi(0.146 $< g r e e n z l <$ 0.182)	−0.137***
	(0.148)
gfi( $g r e e n z l \geq$ 0.182)	−0.146***
	(0.151)
_cons	6.429***
	(1.703)
OBSERVATION	270
PROVINCE	30

*p < 0.1, **p < 0.05, ***p < 0.01.

Regional heterogeneity analysis

Heterogeneity research involves considering the characteristics and differences among individuals or groups to accurately assess the impact of influencing factors. When studying the effect of $g f i$ on PM2.5, it is important to recognize the significant distinctions between the northern and southern regions of China in terms of economic structure, environmental governance, and climate. Failing to account for these differences may lead to inaccurate conclusions, which can hinder the formulation and implementation of relevant policies. To provide a more precise evaluation of the influence of $g f i$ on PM2.5 and inform policy-making, this study conducted heterogeneity research. The 30 provinces of China were divided into two distinct groups, namely the north and south, based on geographical division. Separate regression analyses were then conducted for each group, and the results are presented in Table 14. The results demonstrate a significant coefficient of −1.234 at a 1% significance level, indicating a robust inhibitive effect of $g f i$ on PM2.5 in the northern regions. However, in the southern parts of China, the influence of $g f i$ on PM2.5 is negatively correlated but relatively moderate. This leads to the question of why the influence of $g f i$ on PM2.5 is prominent in the north but less pronounced in the south. Several factors may contribute to this disparity, including the following possibilities:

Differences in regional economic structures：the economy of southern provinces is mainly composed of the service industry and light industry, while the economy of northern provinces is dominated by heavy industry and energy. Light industry and the service industry have a relatively small impact on environmental pollution. Therefore, in southern provinces, due to the difference in economic structure, the role of $g f i$ in reducing PM2.5 is not as significant as in northern provinces. In northern provinces, however, due to the higher proportion of heavy industry and energy, PM2.5 emissions are correspondingly higher, making the role of $g f i$ more significant.

Historical differences in environmental governance: environmental governance in northern provinces is relatively lagging. Due to historical reasons, the environmental governance in northern provinces started relatively late, and air pollution problems are relatively more serious. Consequently, $g f i$ is more prominent in northern provinces. In the southern provinces, both air pollution and environmental governance are comparatively low, so $g f i$ plays a smaller role than in the northern provinces.

Table 14.

Regional heterogeneity regression results.

	North	South
gfi	−1.234***	−0.741
	(0.395)	(0.468)
pop	0.279	−0.238
	(0.258)	(0.571)
financeim	−0.024	−0.003
	(0.018)	(0.036)
gap	0.070*	0.106***
	(0.037)	(0.028)
gtfp	−0.005	0.007
	(0.015)	(0.015)
eri	0.071**	0.083*
	(0.028)	(0.038)
energyc	0.034	0.389***
	(0.078)	(0.107)
human	−0.159	0.067
	(0.420)	(0.262)
_cons	3.009	2.226
	(2.081)	(4.835)
R²	0.979	0.978
AIC	−332.7	−424.6
BIC	−242.7	−334.6
ID	yes	yes
YEAR	yes	yes
Obs	135	135
Province	15	15

*p < 0.1, **p < 0.05, ***p < 0.01.

Therefore, due to the above differences, $g f i$ plays a less significant role in reducing PM2.5 in southern provinces than in northern provinces.

Research conclusions and policy recommendations

In this paper, the influence of green finance and green technology innovation on PM2.5 emissions is examined using the spatial Durbin model and the mediating effect model. Considering the substantial regional disparities and spatial correlation in the progression of green finance in China, as well as the spatial contagiousness of PM2.5, it is more suitable to investigate the effect of green finance on PM2.5 from a spatial perspective, considering the specific circumstances of China. Moreover, green technology innovation, as a significant endogenous variable impacting economic growth and environmental contamination, is often overlooked in existing research. Innovations in green technology have significant implications for promoting green finance and reducing PM2.5 emissions. To summarize the findings, this study reveals several key points. Firstly, green finance exerts a modest restraining effect on PM2.5 emissions at the provincial level but significantly suppresses the geographical spillover effect of PM2.5 emissions in neighboring provinces. Secondly, green finance plays a role in promoting green technology innovation and can provide financial support for businesses involved in such innovation. Thirdly, the presence of green finance can have a spatial spillover effect, effectively curbing PM2.5 emissions in adjacent provinces and mitigating haze pollution in neighboring regions, optimizing, and modernizing industrial structures, and facilitating the transition to clean energy. Lastly, there is notable heterogeneity in the impact of green finance on PM2.5 emissions between the northern and southern regions, with a significant effect observed in the north but not in the south. Based on these findings, the following recommendations are proposed in this article:

Promoting green finance can effectively reduce environmental pollution, particularly PM2.5 emissions. Green finance, led by financial institutions and involving businesses and the public, channels funds towards enterprises that meet green finance support standards. Firstly, it is crucial to accurately identify companies that qualify for green finance support and ensure the efficient and timely allocation of funds to these companies so they can fulfill their responsibilities. By eliminating information asymmetry, demand for green funds and supply from providers can be effectively matched, preventing non-compliant companies from misusing funds and resources. Secondly, given limited resources, proactive strategies and the development of green products should be encouraged to assist fund providers in identifying and promoting high-quality green products and enterprises.

Green finance can support technology innovation companies in pursuing sustainable innovation and prioritizing environmentally related innovations to reduce PM2.5 emissions. From an “incentive” perspective, the government can expand green finance channels for companies and increase subsidies for approved green projects, enabling enterprises to actively engage in ecological innovation. From a “control” perspective, the government can establish an efficient evaluation process for green finance policies to incentivize green companies. Strengthening the qualification audit process for companies can facilitate easier access to green financial assistance, encouraging the creation and innovation of green technologies.

It is essential to increase investments in green finance, with a particular emphasis on regions with severe air pollution in the north. This can be achieved through various means, such as providing low-interest loans, offering incentives to encourage enterprises and individuals to invest in green industries. In regions with relatively lighter air pollution in the south, promoting green technology should be prioritized. By promoting the adoption of green technology, energy consumption can be reduced, and pollutant emissions can be decreased, ultimately leading to a reduction in PM2.5 levels. In the northern regions, efforts should be focused on strengthening air pollution control and promoting green industries. In the southern regions, the emphasis should be on promoting green technology and enhancing environmental protection. By implementing tailored policy measures, the role of green finance can be maximized, facilitating sustainable economic and environmental development.

One limitation of this article is the availability of data, specifically the limited availability of data for certain sub-indicators of green finance, which is only accessible up until 2020. As a result, it is not possible to extend the research period to include 2021 and 2022, potentially causing a lag in the research results.

Footnotes

Credit author statement

Yiniu Cui: Responsible for the overall planning and design of the paper, and drafting the initial version. Cheng Zhong: Responsible for data collection and preprocessing in the paper. She is responsible for gathering and organizing the required datasets, as well as performing data cleaning and preprocessing. Jianhong Cao: Responsible for conducting literature research and data analysis in the paper. Her work mainly focuses on model design and empirical regression. Mengyao Guo: Responsible for the model training and optimization work in the paper. She utilizes the collected data and research findings to conduct iterative testing and optimization.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by the Ministry of Science and Technology of China under Grant 2020AAA0108402 and the National Natural Science Foundation of China (NSFC) under Grant 71825007, in part by the National Natural Science Foundation of China (NSFC) under Grant 72210107001, in part by the Fundamental Research Funds for the Central Universities.

ORCID iD

Cheng Zhong

Yiniu Cui is a doctoral student of World Economics at Yunnan University. His research interests include international political economy, green economy, and environment. He has published some papers in some SCI and SSCI journals as the first author and corresponding author.

Cheng Zhong is currently a lecturer at the Business School of Pingxiang University, and she has published some related papers in major journals in the environmental field.

Jianhong Cao is currently a Post Ph.D in the Yuquan Institute of the University of Chinese Academy of Sciences. Currently, her main research interests include green finance, energy economics, environmental policy. So far, She has published over 10 high-quality peer-reviewed papers in SCI/SSCI-indexed.

Mengyao Guo is a first-year postgraduate student in the history of economic thoughts in the School of Economics of Yunnan University. She has participated in a number of research projects, such as the Yunnan Provincial Department of Education project and the Yunnan University School of Economics project.

References

Zhang

Cao

, et al. Forty years of reform and opening up: China’s progress toward a sustainable path. Sci Adv 2019; 5: eaau9413.

Zhou

Tang

Zhang

. Influence of green finance on economic development and environmental quality: a study based on provincial panel data from China. Environ Sci Pollut Res 2020; 27: 19915–19932.

Miao

Fang

Sun

, et al. Natural resources utilization efficiency under the influence of green technological innovation. Resour Conserv Recycl 2017; 126: 153–161.

Ahmed

Cary

Ali

, et al. Moving toward a green revolution in Japan: symmetric and asymmetric relationships among clean energy technology development investments, economic growth, and CO2 emissions. Energy Environ 2022; 33: 1417–1440.

Huang

Mbanyele

Wang

, et al.

Climbing the quality ladder of green innovation: does green finance matter?

Technol Forecast Soc Change 2022; 184: 122007.

Feng

Chong

, et al. Digital finance and innovation inequality: evidence from green technological innovation in China. Environ Sci Pollut Res 2022; 29: 87884–87900.

Feng

Zhang

. Environmental decentralization, digital finance and green technology innovation. Struct Change Econ Dyn 2022; 61: 70–83.

Zhou

, et al. How environmental regulations affect the development of green finance: recent evidence from polluting firms in China. Renew Energy 2022; 189: 917–926.

Zhu

Jin

, et al. Green financial behavior and green development strategy of Chinese power companies in the context of carbon tax. J Cleaner Prod 2020; 245: 118908.

10.

Mori

. Socio-technical and political economy perspectives in the Chinese energy transition. Energy Res Soc Sci 2018; 35: 28–36.

11.

Dorfleitner

Braun

. Fintech, digitalization and blockchain: possible applications for green finance. The rise of green finance in Europe: opportunities and challenges for issuers, investors and marketplaces, 207–237. 2019.

12.

Wang

Zhi

. The role of green finance in environmental protection: two aspects of market mechanism and policies. Energy Procedia 2016; 104: 311–316.

13.

, et al. The impact of financial development and green finance on regional energy intensity: new evidence from 30 Chinese provinces. Sustainability 2022; 14: 9207.

14.

Ziolo

Filipiak

Bąk

, et al. Finance, sustainability and negative externalities. An overview of the European context. Sustainability 2019; 11: 4249.

15.

Liu

Song

, et al. Impact of financial technology on regional green finance. Comput Syst Sci Eng 2021; 39: 391–401.

16.

Wang

Lee

. The impact of clean energy consumption on economic growth in China: is environmental regulation a curse or a blessing? Int Rev Econ Finance 2022; 77: 39–58.

17.

Sachs

Woo

Yoshino

, et al. Importance of green finance for achieving sustainable development goals and energy security. In: Handbook of green finance. Singapore: Springer, 2019, pp.3–12.

18.

Zeng

Wang

. The influence of green finance on urban haze pollution in China: a technological innovation perspective. Energies 2022; 15: 801.

19.

Hou

Zhu

Wang

, et al. Will green financial policy help improve China’s environmental quality? The role of digital finance and green technology innovation. Environ Sci Pollut Res 2023; 30: 10527–10539.

20.

Kong

Sun

, et al. Effects of digital finance on green innovation considering information asymmetry: an empirical study based on Chinese listed firms. Emerging Markets Finance Trade 2022; 58: 4399–4411.

21.

Banga

. The green bond market: a potential source of climate finance for developing countries. J Sustain Finance Invest 2019; 9: 17–32.

22.

Cao

Law

Samad

ARA

, et al. Internal mechanism analysis of the financial vanishing effect on green growth: evidence from China. Energy Econ 2023; 120: 106579.

23.

Safitri

Sari

Gamayuni

. Research and Development (R&D), environmental investments, to eco-efficiency, and firm value. Indonesian J Account Res 2020; 22.

24.

Bhutta

Tariq

Farrukh

, et al. Green bonds for sustainable development: review of literature on development and impact of green bonds. Technol Forecast Soc Change 2022; 175: 121378.

25.

Zhu

Wang

. Micro green technology innovation effects of green finance pilot policy—from the perspectives of action points and green value. J Bus Res 2023; 159: 113724.

26.

Zhang

Duan

Yang

, et al. Can green finance improve China's haze pollution reduction? The role of energy efficiency. Environ Develop 2023; 45: 100833.

27.

Harvey

. The political economy of health: revisiting its Marxian origins to address 21st-century health inequalities. Am J Public Health 2021; 111: 293–300.

28.

Jiang

Deng

. Multi-dimensional threshold effects of the digital economy on green economic growth?—new evidence from China. Sustainability 2022; 14: 12888.

29.

Chien

Sadiq

Nawaz

, et al. A step toward reducing air pollution in top Asian economies: the role of green energy, eco-innovation, and environmental taxes. J Environ Manag 2021; 297: 113420.

30.

Jahanger

Usman

Murshed

, et al. The linkages between natural resources, human capital, globalization, economic growth, financial development, and ecological footprint: the moderating role of technological innovations. Resources Policy 2022; 76: 102569.

31.

Wang

Qiu

. Can “energy saving and emission reduction” demonstration city selection actually contribute to pollution abatement in China? Sustain Product Consump 2021; 27: 1882–1902.

32.

Fan

. Renewable energy technology innovation, industrial structure upgrading and green development from the perspective of China's provinces. Technol Forecast Soc Change 2022; 180: 121727.

33.

Ikävalko

. The Contradictions of Environmental Statehood: Air Pollution and Environmental Policy in Delhi and the National Capital Region, India. 2021.

34.

Zhuo

Xie

Mao

, et al.

Can cross-regional environmental protection promote urban green development: zero-sum game or win-win choice?

Energy Econ 2022; 106: 105803.

35.

Cui

Wang

Irfan

, et al. The effect of green finance and unemployment rate on carbon emissions in China. Front Environ Sci 2022; 10: 887341.

36.

Cao

Law

Samad

ARBA

, et al. Effect of financial development and technological innovation on green growth—analysis based on spatial Durbin model. J Cleaner Prod 2022; 365: 132865.

37.

Cao

Law

, et al. Impact of local government competition and land finance on haze pollution: empirical evidence from China. Emerging Markets Finance and Trade, 1–23. 2022.

38.

Lanjouw

Schankerman

. Patent quality and research productivity: measuring innovation with multiple indicators. Econ J 2004; 114: 441–465.

39.

Kim

Lee

. Patent databases for innovation studies: a comparative analysis of USPTO, EPO, JPO and KIPO. Technol Forecast Soc Change 2015; 92: 332–345.

40.

Shen

Ridwan

Raimi

, et al. Recent developments in green hydrogen–environmental sustainability nexus amidst energy efficiency, green finance, eco-innovation, and digitalization in top hydrogen-consuming economies. Energy Environ 2023: 0958305X231153936.

41.

Omokanmi

Ibrahim

Ajide

, et al.

Exploring the dynamic impacts of natural resources and environmental pollution on longevity in resource-dependent African countries: does income level matter?

Resources Policy 2022; 79: 102959.

42.

Khan

SAR

Ridwan

, et al. Energy efficiency, carbon neutrality and technological innovation: a strategic move towards green economy. Econ Res-Ekonomska Istraživanja 2023; 36: 2140306.

43.

Wang

Cui

, et al. How does imported pork regulate the supply and demand of China's pig market during the epidemic?—based on the analysis of African swine fever and COVID-19. Front Vet Sci 2022; 9: 1028460.

44.

Ren

Tong

Sun

, et al. Dynamic impacts of energy consumption on economic growth in China: evidence from a non-parametric panel data model. Energy Econ 2022; 107: 105855.

45.

Cao

Law

Samad

ARBA

, et al. Impact of financial development and technological innovation on the volatility of green growth—evidence from China. Environ Sci Pollut Res 2021; 28: 48053–48069.

46.

. The contributions of socioeconomic indicators to global PM2. 5 based on the hybrid method of spatial econometric model and geographical and temporal weighted regression. Sci Total Environ 2020; 703: 135481.