The Polarization Effect and Mechanism of China’s Green Finance Policy on Green Technology Innovation

Abstract

The advancement of green technology innovation (GTI) is crucial for facilitating green development. China, the largest carbon-emitting economy, should prioritize the acceleration of GTI to augment global green economic growth and reduce carbon emissions. Green finance policy (GFP) is a common instrument for encouraging enterprises to develop GTI. This study, therefore, takes the pilot policy of China’s Green Finance Reform and Innovation Pilot Zone as a “quasi-natural experiment” and uses the difference-in-differences method to explore the impact and mechanism of GFP on Chinese enterprises’ GTI. Based on the empirical analysis using microdata from Chinese industrial enterprises from 2015 to 2021, the following conclusions can be drawn. First, GFP has a green innovation polarization effect. It facilitates the development of GTI in green enterprises while hindering the progress of GTI in polluting enterprises. Second, GFP enhances the GTI of green enterprises by promoting innovative behaviors and factor allocation optimization behaviors. However, GFP reduces the GTI of pollution enterprises by promoting non-innovative investments and reducing the efficiency of factor allocation optimization. Third, the combination of policies utilizing GFP, environmental subsidy, and R&D subsidy can effectively increase the GTI of polluting enterprises without compromising the GTI of green enterprises. This study offers empirical evidence and policy recommendations for establishing a green finance system in developing countries.

1. Introduction

Human survival and development have been seriously threatened by the ongoing deterioration of the environment and climate due to the discharge of pollutants since the beginning of the 21st century. Consequently, an increasing number of countries have chosen to pursue a “green development” path, which seeks to strike a balance between economic growth and environmental protection. The advancement of green technology innovation (GTI) is crucial to facilitate green development. The outlook report for the 28th meeting of the United Nations Framework Convention on Climate Change (COP28) proposed “enhancing access to advanced technologies and large-scale deployment” [1]. This report highlights the significance of using GTI in addressing and mitigating the impact of climate change. In 2021, carbon emission in China, the world’s largest source of carbon emissions and the second-largest economy, was 11.47 billion tons, twice that of the United States and four times that of the European Union [2]. Meanwhile, its industries, the main drivers of economic growth, produced 7.9 billion tons of carbon, constituting 69% of the country’s total [3]. Accordingly, accelerating the GTI of Chinese enterprises, especially those operating in substantial economic sectors such as industry, is crucial to facilitate global green economic growth and reduce carbon emissions.

Green finance policy (GFP) is widely used to incentivize enterprises to develop GTI. GFP guides diversified financial capital from high-polluting enterprises and energy-consuming projects to eco-friendly enterprises and cleaner production projects by factoring in the costs of environmental risk into investment decision-making within the financial sector [4,5,6]. This approach encourages enterprises to adopt efficient, energy-saving, and eco-friendly production methods through the development of GTI. For instance, IFC’s Environmental and Social Sustainability Performance Standards serve as a tool for clients to identify environmental and social risks associated with their investment activities, thereby incentivizing them to pursue green production and technological improvement [7]. ASSURPOL, a co-reinsurance group based in France, encourages the development of the GTI by reinsuring environmental risks transferred to pools by insurers on behalf of its members. This approach incentivizes participants to adopt environmentally friendly practices while protecting the interests of the ultimate insured [8].

Nevertheless, the effectiveness of GFP in developing countries such as China remains uncertain due to their relative lack of experience in designing and implementing policies compared with international organizations and developed nations. Especially in China’s limited history of environmental policy application, GFP is often designed to support the green industry rather than the eco-friendly behavior of enterprises [9], and there is an obvious tendency to support the development of the green industry by restraining the polluting industry [9,10]. If GFP is interpreted or implemented as an industrial subsidy policy, it may not provide financial support to traditional industrial enterprises seeking to adopt green production technology, while easing the financial constraints of non-innovative behavior by green industrial enterprises. This approach is not conducive to promoting the rational allocation of financial resources and the development of GTI among Chinese enterprises [11,12]. So, does China’s GFP provide appropriate financial resources for green enterprises and force polluting enterprises to change their production modes, so as to realize the synchronous development of GTI in different industries? Or does it support green enterprises at the cost of excessively squeezing the financing channels and technological improvement willingness of polluting enterprises, resulting in a green innovation polarization effect that is detrimental to the overall green development of the economy? This is a practical problem worth discussing.

The relationship between GFP and GTI has attracted interest among scholars in recent years. While the literature on this relationship is still relatively scarce, the number of studies on the topic is rapidly expanding [13]. However, the discussion of the relationship between GFP and GTI in relevant literature has yielded different conclusions in developed countries and developing countries represented by China. For developed countries, the majority of literature supports the positive role that GFP plays in promoting the improvement of GTI. Specifically, the implementation of GFP has been shown to enhance GTI through the improvement of regional financing environments [14], the willingness of industries to develop green technologies [15], the preference of investors for environmentally friendly investments [16], and the effectiveness of complementary policies [17]. For developing countries represented by China, the relationship between GFP and GTI remains a topic of controversy in the academic community. Some studies suggest that GFP has improved the efficiency of regional technological innovation, particularly in areas with insufficient financial development where GFP has recently been implemented [13,18]. Conversely, other studies have found that the lack of efficiency in GFP may hinder the enthusiasm for green production and technological innovation in industries [19].

In fact, the GFP in developing countries may have differential impacts on GTI for different types of enterprises. Therefore, the key issue lies in whether green enterprises or polluting enterprises should be the research sample, which directly affects the empirical research results on the relationship between GFP and GTI. This emphasizes the importance of exploring the heterogeneous impacts of GFP on GTI for different types of enterprises, especially for developing countries like China [20]. Regrettably, existing research has mostly treated enterprises’ GTI as a whole, with little investigation into whether GFP has differential impacts on GTI for green enterprises and polluting enterprises [15,19]. Consequently, there is a scarcity of research on the formation mechanisms and measures to address such differential impacts.

China’s GFP has witnessed significant growth and improvement over the last decade, with a focus on high-quality development strategy leading to its refinement, systematization, and increased operability. The Green Credit Guidelines released in 2012 is considered a positive attempt to leverage green credit instruments and encourage the green industry to take initiative [21]. Furthermore, Guidelines for Establishing the Green Financial System released in 2016 marked the beginning of a systematic policy framework for green finance, which included green credit, green securities, and green insurance [22]. This framework provides further support for the green transformation of significant sectors of the economy [23]. Next, the Green Finance Reform and Innovation Pilot Zone, which was launched in 2017, represents the latest exploration and practice in China of implementing green finance at the grassroots level by mobilizing local governments. This marks a new stage in China’s GFP that combines top-level design with regional exploration [24]. With the gradual expansion of the pilot scope of this policy, GFP is expected to positively facilitate the green transformation process of China’s economy in the future [25,26]. Accordingly, exploring the heterogeneous impacts and mechanisms of GFP, represented by Green Finance Reform and Innovation Pilot Zone, on GTI within Chinese enterprises is essential.

Within this context, this study primarily discusses three questions:

(1)

What is the impact of China’s GFP on GTI developed by enterprises? In particular, does it asymmetrically affect the GTI of green and polluting enterprises?

(2)

What are the micro mechanisms through which China’s GFP affects GTI developed by enterprises?

(3)

How can policymakers enhance the GFP system to promote the GTI of diverse enterprises simultaneously, given a clear understanding of the impact and mechanism of GFP on GTI?

To this end, microdata from Chinese industrial enterprises from 2015 to 2021 are considered as samples and the pilot policy of China’s Green Finance Reform and Innovation Pilot Zone is considered as a “quasi-natural experiment”. In addition, the difference-in-differences method is used to explore the impact and mechanism of China’s GFP on GTI developed by enterprises. Furthermore, the pathway for optimizing China’s green finance system is explored based on the given analysis.

The possible marginal contributions of this study are threefold. First, the study focuses on the innovation incentive effect of green finance by examining the heterogeneous role of GFP in stimulating the GTI of different types of enterprises. This is based on identifying the asymmetry of GFP content, which adds to the existing research on the topic. Second, this study analyzes the influence mechanism of GFP on GTI from the micro perspective of enterprise behavior, providing a more specific and detailed explanation for the heterogeneous impact of GFP on GTI. This expands and deepens the existing literature. Third, the study highlights a gap in research on the design of GFP for promoting innovative development in China and other developing nations. By exploring the synergy between GFP and other environmental policy instruments in stimulating GTI, the study provides insights and strategies for the development and implementation of a green finance policy system in underdeveloped regions and countries.

The remainder of the paper is organized as follows. Section 2 overviews the relevant literature. Section 3 develops hypotheses. Section 4 describes the research design. Section 5 summarizes and analyzes the empirical research results. Section 6 presents conclusions and policy implications.

2. Literature Review

While there is a scarcity of studies that directly investigate the connection between GFP and GTI, there has been a significant surge in the number of studies examining the impact of GFP and its accompanying policies on technological innovation. The literature on this topic can be divided into three broader groups.

2.1. The Impact of GFP on Technological Innovation

The first group of studies has focused on the impact of GFP on technological innovation. Relevant literature has examined this relationship from two perspectives: the impact at a national-regional level and the impact on industries and firms. The vast majority of studies have concluded that GFP increases the efficiency of regional technological innovation, especially in developing countries and underdeveloped regions [27,28,29,30]. Irfan et al. [19] utilized the dataset covering 30 provinces in China from 2010 to 2019 and concluded that financial institutions’ green funds were applied to R&D, which resulted in the promotion of green innovation at the regional level. Lin and Ma [31] proved that efficient financial activities can enhance the quantity and quality of GTI in both developed and underdeveloped cities, and cities with higher levels of human capital are conducive to green innovation activities. Other studies have explored the impact of GFP on innovation in enterprises [32,33,34], particularly those in heavily polluting and high energy-consuming industries [35,36,37]. Zhang et al. [38] stated that China’s Green Credit Policy has enhanced the investment and financing environment for large enterprises with high energy consumption and high pollution levels, referred to as “two-high” enterprises, thereby promoting innovation efficiency in such enterprises. Xu et al. [9] observed that China’s Green Finance Pilot Policy can significantly enhance GTI in industrial enterprises.

2.2. The Influence Mechanism of GFP on Technological Innovation

The second group of literature has focused on examining the influence mechanism of GFP on technological innovation, with the scope of discussion in the relevant literature divided into two categories: industry-level and firm-level mechanisms. To explore the mechanisms from an industry-level perspective, some of the existing studies have proposed that GFP influences enterprise innovation by affecting the financing environment of the industry [39,40,41]. Li et al. [42] discovered that the introduction of green credit policy led to a tightening of external financing for heavy polluters, which adversely affected the efficiency and quality of green innovation for such enterprises. Su et al. [43] contended that the overall financing environment of the industry will improve when the intensity of green credit exceeds a certain threshold, which can promote green technological innovation in firms. Another part of the research has suggested that GFP primarily influences firm innovation by optimizing the allocation of production and R&D factors within the industry [15,44]. Irfan et al. [19] proposed that green finance can optimize the allocation of private idle capital by aggregating the participation of public and institutional investors in green investments, thereby enhancing the efficiency of green projects and providing adequate financial support for corporate green innovation. For the exploration of the mechanism from the firm’s perspective, most of the relevant papers have contended that GFP affects the innovation performance of enterprises by modifying micro factors, such as the enterprise’s living environment, operating costs, and government–enterprise relations [45,46]. Liu and Xiang [47] documented that China’s Green Finance Reforms have incentivized corporate innovation by reducing the cost of debt, increasing innovation inputs, and attracting foreign investment. Wang et al. [48], however, argued that GFP creates an implicit network of linkages between firms and government. This network internalizes the costs of environmental management activities and ultimately induces firms to improve their environmental performance through GTI.

2.3. The Countermeasures to Stimulate the Innovation Incentive Effect of GFP

Based on the summary of the perspectives on the impact and influence mechanisms of GFP on technological innovation (including GTI), the third group of studies has further discussed the means and countermeasures to stimulate the innovation incentive effect of GFP. A part of this literature has focused on the conditions conducive to the positive effect of GFP on technological innovation, i.e., the moderators of the innovation incentives of GFP. These studies have concluded that the appropriate intensity of government regulation [48,49,50], industry concentration [51], and the level of corporate environmental information disclosure [43] can positively moderate the impact of GFP on technological innovation and GTI of enterprises. Another part of the literature has contended that the development and implementation of GFP, especially in developing countries, are not yet well-developed. This implies that GFP needs to be used in conjunction with other environmental policy instruments to maximize its policy effects [14,52]. Although this research path is gaining prominence in academic circles for examining the effects of GFP policies, the number of relevant studies is currently insufficient. Currently, studies on the synergistic effects or coupling effects of environmental policy instruments such as environmental regulation [34], environmental subsidies [53], R&D subsidies [54], carbon trading [55], and public monitoring [56] on GTI are available. Few studies have incorporated GFP into the development of an environmental policy system. Only a few studies have discussed the establishment of a GFP system from the perspective of innovation incentives [57,58]. However, these studies are limited to theoretical analysis and policy recommendations, lacking detailed and quantitative empirical tests.

2.4. Literature Critique

In summary, on the one hand, the existing literature has comprehensively explored the impact and mechanism of GFP on GTI, along with the optimization strategy of GFP, thereby obtaining diverse conclusions. On the other hand, some weaknesses exist in related studies. First, while studies have been conducted on the innovation incentive effects of GFP on regions, industries, and high-polluting enterprises, few studies have comprehensively examined the diverse impacts of GFP on the GTI of different types of enterprises in comparison, as well as the economic and social consequences of such impacts. Second, most of the existing studies have explained the influence mechanism of GFP on GTI in terms of industry-level factors such as financing environment and industry characteristics (including the survival environment and financial characteristics of companies). However, these factors affect the enterprises’ innovation efficiency by influencing their business decisions and behaviors. Few studies have examined the role of micro factor of enterprise behavior as a channel between GFP and GTI. Third, studies discussing the synergy between GFP and other environmental policy instruments from the perspective of stimulating GTI are insufficient. Furthermore, discussions on the design of GFP systems in developing countries such as China are relatively scarce.

In view of this, this study considers the latest policy practice of green finance in China, the pilot policy of Green Finance Reform and Innovation Pilot Zone, as a basis to determine the heterogeneous effects of GFP on different types of enterprises. Moreover, it reinterprets the influence channel of GFP on GTI from the perspective of enterprise behavior. It further discusses how to reasonably construct the GFP system to stimulate Chinese industrial enterprises to develop GTI in a balanced manner and finally realize the “green development” model in which economic growth and environmental protection can be unified.

3. Research Hypothesis

3.1. The Impact of China’s GFP on GTI

According to Green Finance Development Theory, green finance promotes industrial transformation and technological innovation through the “structural effect” and “horizontal effect” [59]. The structural effect means that green finance directs diversified capital to green enterprises, thus alleviating their financing constraints. This ultimately promotes the greening of industrial structure and enterprise technology [18]. The horizontal effect refers to the fact that green finance has guided the transfer of corporate capital from polluting and high energy-consuming projects to environmentally friendly projects. This ultimately promotes the clean transformation of production technologies and processes in traditional industries and polluting enterprises [60]. In developed countries with a higher level of financial development, GFP is believed to trigger both the structural and horizontal effects, achieving the simultaneous promotion of GTI in both green and polluting enterprises [61]. For example, the EU Taxonomy for Sustainable Finance sets differentiated environmental thresholds for the financing needs of 67 economic activities. This allows green enterprises to easily obtain credit support for developing technological innovation, while polluting enterprises can raise funds for technological improvements through the issuance of green bonds and green insurance [62]. However, in developing countries such as China, the immaturity of the green finance system may lead to a greater “structural effect” than “horizontal effect” of its GFP [63]. In other words, green finance resources may be allocated to emerging green enterprises with high returns, while pollution enterprises with low expected returns and high risks may find it difficult to obtain financing support for clean production transformation projects. For example, based on a sample of Chinese enterprises, Wang [13] asserted that the introduction of green policies has transferred the responsibility of environmental management from society to the high-pollution industries, leading to additional costs associated with environmental regulation. Consequently, enterprises lack the incentive and capability to increase their investments in technological R&D for innovation due to added costs [64,65]. Based on this, we propose the following hypotheses:

Hypothesis H1. 

China’s GFP facilitates the progress of GTI in green enterprises while hindering the progress of GTI in polluting enterprises.

3.2. The Influence Mechanism of China’s GFP on GTI

Previous studies have demonstrated that the innovation performance of enterprises can be improved through two types of behavior: innovative behavior and factor allocation optimization behavior [66]. Innovative behavior refers to changes in a company’s R&D investment and efficiency, while factor allocation optimization behavior refers to activities such as capital renewal, asset restructuring, and strategic decision-making, all of which can create a favorable external environment for innovation [67]. In fact, both innovative behavior and factor allocation optimization behavior are essentially composed of enterprise investment and management activities, with the efficiency of these activities relying on financing [68]. If GFP provides financing convenience for green enterprises while simultaneously raising the financing cost of polluting enterprises, it means that GFP will be beneficial to the innovation behavior and factor allocation optimization behavior of green enterprises, while suppressing the aforementioned behaviors of polluting enterprises [69]. This difference will ultimately lead to GFP exerting a heterogeneous impact on the GTI of different types of enterprises. Based on this, we propose the following hypotheses:

Hypothesis H2. 

By guiding innovative behaviors and factor allocation optimization behaviors, China’s GFP can exert a heterogeneous impact on the GTI of different types of enterprises.

3.3. The Synergistic Effect of China’s GFP and Its Supplementary Policies

According to Tinbergen’s Rule, attaining policy goals requires that the number of policy instruments be at least equal to the number of target variables [70]. Therefore, to achieve the goal of weakening the green innovation polarization effect of China’s GFP, a combination of policy instruments needs to be constructed. Specifically, to achieve the dual policy goals of maintaining the improvement of the GTI of green enterprises while promoting the growth of the GTI of polluting enterprises, GFP needs to be used in conjunction with at least one other environmental policy instrument [14,52]. At present, strengthening the green financial system and enhancing the capacity of green finance have become crucial directions for financial policy reform in China [23]. In this context, creating a supportive external policy environment that leverages the synergies between GFP and other environmental policy instruments is crucial for establishing a green financial system that supports GTI. As Ma et al. [71] discovered, green credit policies have a suppressive effect on environmental innovation in Chinese enterprises, but government subsidy policies can effectively correct the negative impact of green credit policies. Using a combination of green credit and government subsidy policies can achieve a win–win situation for environmental protection and innovation performance. Based on this, we propose the following hypotheses:

Hypothesis H3. 

The judicious combination of China’s GFP and its supplementary policies can promote synchronous improvement of the GTI of both polluting and green enterprises.

4. Research Design

4.1. Methodology

4.1.1. Basic Model

This study considers the pilot policy of China’s Green Finance Reform and Innovation Pilot Zone (GFRIpilot) as a quasi-natural experiment to construct a difference-in-differences (DID) model. The reasons for employing the DID method to estimate the relationship between GFP and GTI are as follows: (1) The DID method can largely avoid the problem of endogeneity. This is because policies are generally exogenous to microeconomic entities and there is no reverse causality [72]. In other words, as long as the explanatory variables of the treatment group and the control group samples show consistent time trends when there is no policy shock, the potential endogeneity problem of the DID model is small [73]. (2) Traditional policy evaluation methods mainly use a dummy variable representing whether the policy has occurred and conduct regression analysis. In comparison, the DID model uses two dummy variables and difference processing, which largely eliminates the fluctuation of random disturbance term, thus more accurately estimates the policy effect [74]. (3) The DID model can better address the issue of confounding variables. Meeting the parallel trends assumption indicates that the policy shock is exogenous, which theoretically means that there are no variables related to the policy shock [74]. This also suggests that adding control variables to the DID model does not have a large effect on the coefficients of the independent variable, but only reduces its standard error.

Starting from June 2017, China has set up nine green finance reform and innovation pilot zones (Including Huzhou City, Quzhou City, Guangzhou City, Hami City, Changji Prefecture, Karamay City, Gui’an New Area, Ganjiang New Area, and Lanzhou New Area) across six provinces (Including Zhejiang Province, Jiangxi Province, Guangdong Province, Guizhou Province, Gansu Province, and Xinjiang Uygur Autonomous Region). The establishment of pilot zones is led by the central government, and regional enterprises are not informed before receiving the notice. Therefore, GFRI can be regarded as an exogenous factor independent of enterprise decision-making behavior. That is, it can be considered a quasi-natural experiment. We have chosen to use GFRI as an indicator of GFP for the following three reasons. First, this policy has designated pilot cities in different regions of China, including eastern cities represented by Huzhou, Quzhou, and Guangzhou, central cities represented by Ganjiang New Area, and western cities represented by Hami, Changji, Karamay, and Lanzhou New Areas. These cities fully embody the diverse economic, environmental, and resource characteristics of China. Therefore, the pilot policy’s effects can be considered highly representative [75]. Second, the pilot programs implemented in different zones have varying focuses; however, they have all stimulated local interest in developing green finance systems. This, in turn, has enabled enterprises in the regions to receive financial support through various channels such as credit, securities, funds, and insurance. Furthermore, these programs have enhanced government support for the establishment of a robust environmental rights market and green credit system in the regions. As of June 2022, the balance of green loans in the pilot zones in six provinces and nine cities reached RMB 1.1 trillion, constituting 11.7% of the total loan balance. The balance of green bonds was RMB 238.832 billion, indicating a year-on-year increase of 41.18%. The policy effects are significant and can be easily evaluated [76]. Third, the scheme has set a 5-year goal for each green finance reform and innovation pilot zone, requiring the pilot zone to reach a specific acceptable level of constructing a green finance system in about five years. This approach further ensures the effectiveness of the pilot policy and represents the latest exploration by the central government to mobilize local enthusiasm for green finance governance at the grassroot level.

As pilot cities are included in the GFRI in different batches, we have constructed a time-varying DID model for a sample of listed Chinese industrial companies, as presented in Equation (1). The primary function of Equation (1) is to test Hypothesis H1, which is expressed as follows:

GTI_it = β₀ + β₁GFRIpilot_it + Σ_jλ_jX_it + f_i + f_t + ε_it

(1)

where i and t denote firms and years, respectively. GFRIpilot_it = 1 implies that firm i has been included in the pilot scope of GFRI in year t; conversely, GFRIpilot_it = 0 implies that firm i has not been included in the pilot scope of GFRI in year t. X_it denotes the set of control variables. f_i and f_t denote firm-fixed and year-fixed effects, respectively. ε_it denotes a random disturbance term.

4.1.2. Mechanism Test Models

This study explores the micro mechanisms of GFP influencing GTI. Based on Hypothesis H2, two mechanism variables, innovation behavior (IB) and factor allocation optimization behavior (FB), are set in Equation (2) to test the mechanism of GFP influencing GTI.

GTI_it = α₀ + α₁GFRIpilot_it × EB_it + α₂GFRIpilot_it + α₃EB_it + Σ_jλ_jX_it + f_i + f_t + ε_it

(2)

In Equation (2), EB_it denotes enterprise behaviors encompassing IB and FB. In this equation, we focus on the coefficient ${\alpha }_1$ of the interaction term between EB_it and GFRIpilot_it. The coefficient reflects whether the green finance pilot policy can influence GTI by changing enterprise behavior. Accordingly, we have constructed a mediating effect model to further clarify the specific channels of GFP influencing GTI. The specific model is presented in Equation (3):

EB′_it = α′₀ + α′₁GFRIpilot_it + Σ_jλ_jX_it + f_i + f_t + ε_it

(3)

where $E{B}_{it}^{\prime }$ represents various types of specific enterprise behaviors. Specifically, we replace the dependent variable in the Equation (1) with indicators that capture and reflect four types of activities of firms: R&D activities, capital renewal activities, asset restructuring activities, and strategic decision-making activities. Among them, R&D activities belong to IB, and the latter three types of activities belong to FB. In this equation we focus on the coefficient ${\alpha }_1^{\prime }$ which reflects the impact of GFRIpilot_it on $E{B}_{it}^{\prime }$.

4.1.3. Synergy Analysis Models

To test Hypothesis H3, we include interaction terms of GFP variable with other environmental policy instrument variables in the basic model to examine their synergistic effects on GTI. The specific model is shown in Equation (4):

GTI_it = γ₀ + γ₁GFRIpilot_it × EP_it + γ₂GFRIpilot_it + γ₃EP_it + Σ_jλ_jX_it + f_i + f_t + ε_it

(4)

where EP_it denotes the environmental policy instruments that support GFP, that is, the supplementary policies of GFP, which include subsidy and regulatory policies. In this equation, we have focused on the interaction term coefficient ${\gamma }_1$, which reflects whether GFRIpilot_it and EP_it have synergistic effects on GTI.

4.2. Variable Measurement and Description

This study focuses on a sample of Chinese industrial listed enterprises from 2015 to 2021. Prior to 2015, only a few companies had disclosed detailed information on R&D expenses, environmental management expenses, and related investment matters in their annual reports. This is evident from the revision and reform history of China’s corporate accounting standards. Therefore, pre-2015 statistics may not provide data for some indicators. Moreover, some indicators for 2022 are still awaiting collation. The relevant variables considered in this study are measured or defined as follows:

(1)

Dependent and independent variables. The dependent variable, GTI, is measured by the natural logarithm of the number of patent applications in the field of environmental technology [77]. We have selected the environmental technology field corresponding to the International Patent Classification (IPC) as the selection criterion. Based on the “Technology fields and IPC classification table” published by the OECD, we have established a correspondence between the environmental technology field and IPC classification. Based on this, we have screened and obtained the number of patent applications in the environmental technology field for enterprises. The independent variable, Green Finance Pilot (GFRIpilot), has been determined based on the approval time of pilot programs in various regions mentioned in relevant meetings or documents by the Chinese central or local governments [25].

(2)

Mechanism variables. We have conducted a mechanism analysis from the perspective of enterprise behavior (EB), which encompasses innovation behavior (IB) and factor allocation optimization behavior (FB). IB is measured by the natural logarithm of the ratio of the number of invention patents applied for by the enterprise in the current year to its operating revenue [78]. A higher IB indicates that the enterprise has a stronger motivation and higher efficiency in developing technological innovation. FB is measured by the enterprise investment efficiency according to the method proposed by Richardson [79]. A higher FB indicates that the enterprise has a higher efficiency in allocating its resources. Appendix A provides a detailed explanation of the specific construction method for variable FB.

(3)

Supplementary policy variables for GFP. We have identified four categories of environmental policy instrument variables to investigate the synergistic effects between GFP and supplementary policies. ① R&D subsidy (EP₁), which entails using a keyword search method to locate specific projects in government subsidy details that fall within the scope of research and development subsidies. These projects are then aggregated to obtain the total amount of subsidies. ② Environmental subsidy (EP₂), which involves identifying projects that fall within the scope of environmental subsidies by screening through management expense details, government subsidy details, and social responsibility reports in financial statements. These projects are then aggregated to obtain the value of this indicator. ③ Pollution charges (EP₃), which involves identifying projects that fall within the scope of pollution charges by screening through management expense details in the financial statements of companies. These projects are then aggregated to obtain the value of this indicator. ④ Pollution control (EP₄), which is measured by a binary variable that determines whether a company belongs to the list of key polluting units published by the Chinese Ministry of Environmental Protection. Appendix B provides a detailed explanation of the specific construction methods for variables EP₁, EP₂, and EP₃.

(4)

Control variables. Controlling for the interference caused by enterprise characteristics and certain industry characteristics in the regression results is crucial to obtain a clearer causal relationship between GFP and GTI. The present study considers the following variables: ① enterprise cash (Cash), measured by the natural logarithm of a company’s monetary funds; ② operating revenue (Revenue), measured by the natural logarithm of a company’s main business income; ③ debt-paying ability (Lev), measured by a company’s asset-liability ratio; ④ operational capability (ATO), measured by a company’s current asset turnover rate; ⑤ profitability (ROE), measured by a company’s return on equity; ⑥ supervisors’ annual salary (SAS), measured by the natural logarithm of the total annual salary of regulatory officials; ⑦ proportion of independent directors, measured by the proportion of independent directors to the total number of directors (Ind); ⑧ industry concentration (HHI), measured by the Herfindahl–Hirschman Index of the industry to which the company belongs. The summary of measurement and definition of all variables are shown in

Table 1. Summary of the measurements of the variables used in this study.

Variable Type	Name	Symbol	Definition or Measurement
Dependent variable	Green technology innovation	GTI	The natural logarithm of the number of patent applications in the field of environmental technology
Independent variable	China’s Green Finance Reform and Innovation Pilot Zone	GFRpilot	Based on the approval date of pilot projects in various regions mentioned in relevant meetings or documents of the Chinese central or local governments
Mechanism variables	Innovation behavior	IB	The natural logarithm of the ratio of the number of invention patents applied for by the enterprise in the current year to its operating revenue
	Factor allocation optimization behavior	FB	The enterprise investment efficiency
Supporting policy variables	R&D subsidy	EP1	Aggregate the amounts of relevant items in the enterprise’s financial reports
	Environmental subsidy	EP2	Aggregate the amounts of relevant items in the enterprise’s CSR reports
	Pollution charges	EP3	Aggregate the amounts of relevant items in the enterprise’s financial reports
	Pollution control	EP4	Construct a dummy variable by determining whether the enterprise is on the list of key polluting units.
Control variables	Enterprise cash	Cash	The natural logarithm of a company’s monetary funds
	Operating revenue	Revenue	The natural logarithm of a company’s main business income
	Debt-paying ability	Lev	The enterprise’s asset–liability ratio
	Operational capability	ATO	The enterprise’s current asset turnover rate
	Profitability	ROE	The enterprise’s return on equity
	Supervisors’ annual salary	SAS	The total annual salary of regulatory officials
	Proportion of independent directors	Ind	The proportion of independent directors to the total number of directors
	Industry concentration	HHI	The Herfindahl–Hirschman Index of the industry to which the enterprise belongs

4.3. Data Sources and Sample Processing

The study sample comprises industrial enterprises listed on the Shanghai and Shenzhen stock exchanges in China from 2015 to 2021. As defined by Classification of National Economic Industries of the People’s Republic of China, industrial enterprises are classified into three categories, namely “mining”, “manufacturing”, and “the production and supply of electricity, gas, and water”. To examine the heterogeneous impact of GFP, we define polluting enterprises as those in the high-energy-consuming industries identified in China’s Eleventh Five-Year Plan and the 13 heavily polluting manufacturing industries published by the Chinese Ministry of Environmental Protection. According to this regulation, the polluting enterprises are those operating in the following industries: ferrous metal mining, nonferrous metal mining, textiles, leather, fur, feathers, and their products and footwear, paper and paper products, petroleum processing, coking and nuclear fuel processing industry, chemical materials and chemical products manufacturing, chemical fiber manufacturing, rubber and plastic products industry, non-metallic mineral products industry, ferrous metal smelting and rolling processing industry, and non-ferrous metal smelting and rolling processing industry. Furthermore, we define industrial companies outside of this scope as green enterprises.

The data on enterprise characteristics have been sourced from the CSMAR, CNRDS, and Wind databases. Some policy data indicators have been manually collected and compiled from annual reports and corporate social responsibility reports of listed companies. The initial sample has been screened according to the following criteria: exclusion of samples with missing data for relevant variables; exclusion of samples with incomplete or unusually high indicators; exclusion of samples labeled as ST, ST*, and PT; exclusion of samples with total assets lower than net current and fixed assets; and exclusion of samples with negative equity and accumulated depreciation lower than current depreciation. Eventually, an imbalanced panel dataset of 5022 observations from 1751 enterprises was obtained. This includes 2889 observations of polluting enterprises and 2133 observations of green enterprises. Continuous variables were subjected to winsorization at the upper and lower 1% levels.

Table 2. Descriptive Statistics Results.

Variable Type	Variable Symbols	N	Mean	Std. Dev	Min	Max
Dependent variable	GTI	5022	1.252	1.181	0	5.181
Independent variable	GFRpilot	5022	0.067	0.251	0	1
Mechanism variables	IB	4937	−18.811	1.356	−23.835	−14.297
	FB	5022	0.037	0.043	0.0004	0.394
Supporting policy variables	EP1	1418	8.087	2.206	2.154	16.956
	EP2	534	9.727	2.042	3.923	17.786
	EP3	708	7.338	1.705	3.036	16.684
	EP4	5022	0.441	0.496	0	1
Control variables	Cash	5022	20.631	1.340	17.524	24.644
	Revenue	5022	22.054	1.437	19.035	26.142
	Lev	5022	0.439	0.176	0.061	0.915
	ATO	5022	1.304	0.863	0.203	5.683
	ROE	5022	5.980	14.171	−118.16	36.76
	SAS	5022	15.562	0.699	13.670	17.875
	Ind	5022	0.377	0.057	0.230	0.8
	HHI	5022	0.131	0.105	0.032	0.631

presents the descriptive statistics of the variables. Table 2 reports the sample size (N), mean, standard deviation (Std. Dev), minimum (Min), and maximum (Max) values of the variables utilized in this empirical analysis. These data aid in comprehending the data structure and characteristics of every variable, thereby allowing for a preliminary evaluation of the rationality of the empirical research design [80]. As indicated by Table 2, the mean and standard deviation of the dependent variable GTI are 1.252 and 1.181, respectively. The large standard deviation at this mean level indicates significant variation in GTI levels among the sampled firms. As the independent variable GFRpilot is a dummy variable, its standard deviation is far greater than its mean, which can be inferred from its minimum and maximum values. However, both the dependent and independent variables exhibit strong discrete tendencies, indicating that the issue examined in this study is objectively present. This indirectly highlights the academic and practical value of this paper.

Similar methods can be used to analyze the data characteristics of other variables, which will not be further elaborated here. However, we believe that there are some details in Table 2 that deserve detailed explanation. (1) The number of samples with GTI value of 0 is 702, accounting for 13.97% of the total number of samples. Although the proportion is not large, it will not have a serious impact on the results of econometric analysis. (2) Since IB is constructed as the natural logarithm of a ratio, its mean, minimum, and maximum values are all negative. (3) The number of samples for IB is 4937, which is less than the total number of samples (5022). However, the missing samples only account for 1.71% of the total number of samples, so it will not have a serious impact on the results of econometric analysis. (4) The sample missingness of EP₁, EP₂, and EP₃ is relatively severe, which is related to the statistical method used for these variables. As they only appear as control variables in the model, the sample missingness will not have a serious impact on the results of econometric analysis. However, it is still a factor that interferes with empirical research, so we describe it in the section on future work prospects. (5) In the control variables, the minimum value of ROE is −118.16, which is due to the fact that the net profit of some companies is negative.

5. Empirical Results

5.1. Basic Regression

To test hypothesis H1, this section empirically investigates the impact of GFP, represented by China’s Green Finance Reform and Innovation Pilot Zone (GFRIpilot), on GTI, focusing on whether GFP has asymmetric effects on green and polluting enterprises.

5.1.1. Parallel Trend Test

The use of the DID method for policy effect evaluation requires that the changes in GFP of the treatment and control groups before policy implementation must follow a parallel trend. Specifically, the explanatory variables of the treatment group should exhibit a consistent time trend with the control group samples in the absence of policy shocks [73]. We have adopted an event analysis method and have followed the processing approach of Jacobson et al. [81] to construct Equation (5) for testing the parallel trend of GFP’s effect:

GTI_it = β′₀ + Σ_tβ′_tD_it + Σ_jλ_jX_it + f_i + f_t + ε_it

(5)

where D_it denotes a set of dummy variables. If the city in which enterprise i is located implements the green finance pilot policy in year t, then D_it takes the value 1; otherwise, it takes the value 0. The time span for year t ranges from three periods before policy implementation to three periods after policy implementation. The symbols of other variables have the same meanings as those in Equation (1). The parameter ${\beta }_t^{\prime }$ denotes the difference in GTI between the treatment and the control groups of enterprises in year t of implementing GFRIpilot, the focus of our study.

We have used the third period before the implementation of GFRIpilot as the base period to estimate the parameters in Equation (5). Figure 1 depicts the trend of the estimated values of the ${\beta }_t^{\prime }$ coefficient and the 95% confidence interval during the observation period. The figure indicates no significant difference in the coefficients between the treatment and control groups before the implementation of GFRIpilot. However, a significant difference can be observed in the coefficients after the implementation of the policy. The research sample has passed the parallel trend test. This result suggests that we can use DID to analyze the impact of GFRIpilot on GTI.

5.1.2. Impact of GFP on GTI

Table 3. The Impact of GFP on GTI.

Variables	GTI
	(1) TE	(2) TE	(3) PE	(4) GE
GFRIpilot	0.124 *** (0.041)	0.163 *** (0.049)	−0.135 * (0.077)	0.460 *** (0.083)
Cash		−0.033 (0.032)	−0.008 (0.045)	−0.100 * (0.053)
Revenue		0.226 *** (0.080)	0.176 * (0.094)	0.368 *** (0.132)
Lev		−0.291 * (0.181)	−0.367 (0.286)	−0.107 (0.303)
ATO		−0.142 ** (0.050)	−0.106 * (0.057)	−0.266 ** (0.121)
ROE		0.0001 (0.001)	−0.001 (0.001)	0.001 (0.001)
SAS		0.074 * (0.042)	0.083 (0.066)	0.055 (0.067)
Ind		0.270 (0.519)	0.282 (0.742)	−0.041 (0.651)
HHI		−1.640 ** (0.202)	−1.374 *** (0.244)	−2.583 *** (0.393)
Constant	1.099 *** (0.0504)	−4.078 ** (1.714)	−3.821 ** (1.750)	−4.864 * (2.915)
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
N	5022	5022	2889	2133
Adj-R2	0.105	0.158	0.159	0.171

Note: The numbers in parentheses are robust standard errors. *, **, and *** denote the 10%, 5%, and 1% significance levels, respectively.

presents the regression results based on Equation (1). Columns (1) and (2) of Table 3 indicate that the coefficient of GFRIpilot is significantly positive at the total enterprises level (TE), regardless of the inclusion of control variables. This suggests that GFP promotes enterprise GTI overall. Columns (3) and (4) of Table 3 indicate that the coefficient of GFRIpilot is significantly negative at the polluting enterprises level (PE) and significantly positive at the green enterprises level (GE). This result suggests that GFP has a polarization effect on green innovation, promoting the GTI of green enterprises while inhibiting the GTI of polluting enterprises. These empirical results support hypothesis H1.

As previously mentioned, in developing countries, such as China, that lack experience in designing and implementing green finance policies, GFP is essentially a form of industrial policy [69]. GFP is mainly used to transfer loan quotas and financial resources originally allocated to polluting enterprises to green enterprises [20]. It enhances the GTI initiative of the latter while suppressing that of the former. Thus, green enterprises can access clean production technologies and products, whereas polluting enterprises may find it challenging to transition toward an eco-friendly mode.

5.1.3. Robustness Test

(1)

Variable selection bias test

First, we have used the natural logarithm of the number of patents granted for inventions in the field of environmental technology (GTI′) as a surrogate indicator for the dependent variable GTI to conduct a regression analysis. Columns (1) and (2) of

Table 4. Robustness Test for Variable Selection Bias.

Variables	GTI′		GTI
	(1) PE	(2) GE	(3) PE	(4) GE	(5) PE	(6) GE	(7) PE	(8) GE
GFRIpilot	−0.134 ** (0.066)	1.101 *** (0.074)	−0.228 ** (0.100)	0.301 * (0.163)	−0.135 * (0.077)	0.477 *** (0.083)	−0.125 (0.091)	0.424 *** (0.078)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Industry FE	No	No	No	No	Yes	Yes	Yes	Yes
Industry × Year FE	No	No	No	No	No	No	Yes	Yes
N	2859	2105	2889	2133	2889	2133	2889	2133
Adj-R2	0.274	0.290	0.159	0.171	0.170	0.174	0.175	0.179

Note: The numbers in parentheses are robust standard errors. *, **, and *** denote the 10%, 5%, and 1% significance levels, respectively.

present the results. Second, we have redefined the establishment year of the green finance reform and innovation pilot zones based on their unveiling time in each pilot region. Specifically, we have considered the pilot zones unveiled in the first half of the year as established in the same year. Those unveiled in the second half of the year have been considered as established in the following year. For those without publicized unveiling times, the establishment year has been determined by the State Council’s approval time. This alternative indicator has been used for the core explanatory variable GFRIpilot and has been included in the regression analysis, as indicated in columns (3) and (4) of Table 4. Third, we have controlled for other forms of fixed effects. As indicated in columns (5) and (6) of Table 4, we have added industry fixed effects to the model while controlling for firm and year effects. As presented in columns (7) and (8), we have further added interaction fixed effects between industry and year. A comprehensive analysis of Table 4 reveals that the coefficient of the core explanatory variable GFRIpilot remains significant, and the sign remains unchanged at the level of both polluting and green enterprises. This indicates that the model does not suffer from serious variable selection bias.

(2)

Sample selection bias test.

One potential concern is that the selection of areas for the green finance pilot project may not have been random, leading to differences in characteristics between the enterprises included in the pilot project and those that were not. To address this issue, we have used a multi-period PSM-DID approach. We have selected companies included in the green finance pilot project during the sample period as the treatment group. Furthermore, we have conducted step-by-step matching during the construction of the PSM: first, we have set GTI, monetary funds, revenue, operating capacity, and profitability of the enterprises as matching variables. Thereafter, we have matched the enterprises in the sample year by year, combined the matched data from each year vertically, and generated the panel data required for regression. Finally, we have used DID to re-estimate the matched sample. The results are presented in columns (1) and (2) of

Table 5. Robustness Test for Sample Selection Bias and Placebo Test.

Variables	GTI
	(1) PE	(2) GE	(3) PE	(4) GE	(5) PE	(6) GE	(7) PE	(8) GE
GFRIpilot	−0.148 * (0.088)	0.465 *** (0.083)	−0.135 * (0.077)	0.460 * (0.083)	0.005 (0.118)	−0.121 (0.080)	0.037 (0.098)	−0.074 (0.105)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	2628	2394	2889	2133	2889	2133	2889	2133
Adj-R2	0.163	0.169	0.159	0.171	0.159	0.171	0.159	0.171

Note: The numbers in parentheses are robust standard errors. * and *** denote the 10% and 1% significance levels, respectively.

. Notably, the method of classifying the entire sample into “polluting” and “green” categories may overestimate the scope of polluting or green enterprises, as it may include industries that are only slightly polluting or not green at all, thereby affecting the regression results. Therefore, we have redivided the industry sample: we have only defined enterprises in the “high-energy-consumption and high-pollution” industries designated as polluting enterprises in the 14^th Five-Year Plan for Industrial Green Development. We have defined 35 concept blocks in the Wind database, such as new energy vehicles, combustible ice, hydrogen energy, and sewage treatment, as green enterprises. The results are presented in columns (3) and (4) of Table 5. As indicated in Table 5, the coefficient of the core explanatory variable GFRIpilot remains significant, and the sign remains unchanged at the levels of polluting and green enterprises. This suggests that the model is not affected by serious sample selection bias.

(3)

Placebo test.

We have conducted a time placebo test to ensure that the differences between the treatment and control groups are not caused by the implementation of other policies or changes in time during the sample period. The Guiding Opinions on Building a Green Financial System issued by seven departments including the People’s Bank of China in 2016 may have prompted some local governments to prioritize the development of local green finance systems and attract supporting policy resources for subsequent green financial reforms to tilt toward their regions. To account for the potential influence of the Guiding Opinions on Building a Green Financial System issued in 2016, the policy pilot time was advanced to 2016, and the GFRIpilot variable was reconstructed to participate in the regression. Columns (5) and (6) of Table 5 indicate that the GFRIpilot coefficient is not significant, implying no significant difference between the treatment and control groups before the implementation of the green financial pilot. Next, we have conducted an individual placebo test to ensure that the policy affects only the treatment group and not the control group. National pilot policies may have policy spillover effects on surrounding areas in addition to directly affecting pilot areas. To address the potential spillover effects of the green finance pilot project on neighboring regions, we have expanded the treatment group to include adjacent prefecture-level units to the pilot area and have reconstructed the GFRIpilot variable for inclusion in the regression analysis. Columns (7) and (8) of Table 5 indicate that the coefficient of GFRIpilot is not significant. This suggests that the green financial pilot policy did not significantly affect surrounding areas of the pilot and that it is reasonable to include only the pilot area as the treatment group in this study.

5.2. Mechanism Analysis

5.2.1. Overall Mechanisms

To test hypothesis H2, the regression results based on Equation (2) are shown in

Table 6. Mechanisms of GFP Impacting GTI.

Variables	GTI
	(1) PE	(2) GE	(3) PE	(4) GE
GFRpilot × IB	−0.329 * (0.178)	0.307 ** (0.141)
GFRpilot × FB			0.019 (0.044)	0.180 * (0.095)
GFRpilot	4.362 (3.528)	6.659 ** (2.841)	−0.450 (0.808)	2.751 (2.002)
Cash	−0.022 (0.047)	−0.081 (0.055)	−0.101 * (0.056)	−0.019 (0.053)
Revenue	0.195 * (0.100)	0.351 ** (0.140)	0.266 (0.171)	0.181 * (0.105)
Lev	−0.452 (0.285)	−0.126 (0.297)	−0.371 (0.376)	−0.454 (0.302)
ATO	−0.105 * (0.059)	−0.230 * (0.123)	−0.321 ** (0.138)	−0.132 ** (0.063)
ROE	−0.001 (0.001)	0.001 (0.001)	0.019 *** (0.007)	−0.008 (0.006)
SAS	0.086 (0.069)	0.046 (0.072)	0.002 (0.083)	0.119 (0.074)
Ind	0.289 (0.770)	0.218 (0.619)	0.298 (0.680)	0.640 (0.730)
HHI	−1.406 *** (0.248)	−2.361 *** (0.405)	−2.526 *** (0.414)	−1.338 *** (0.259)
Constant	−3.942 *** (1.872)	−4.892 * (3.014)	−0.837 (3.486)	−5.023 *** (1.888)
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
N	2889	2133	2889	2133
Adj-R2	0.163	0.172	0.166	0.181

Note: The numbers in parentheses are robust standard errors. *, **, and *** denote the 10%, 5%, and 1% significance levels, respectively.

. As shown in Table 6, for polluting enterprises, the coefficient of the interaction term between GFRIpilot and IB is significantly negative, while the coefficient of the interaction term between GFRIpilot and FB is not significant. This suggests that GFP reduces the GTI of polluting enterprises by encouraging non-innovative investment to crowd out R&D investment. On the other hand, for green enterprises, the coefficients of both types of interaction terms are significantly positive, indicating that GFP enhances the GTI of green enterprises by promoting R&D investment and factor allocation optimization investment. These empirical results support Hypothesis H2.

As previously mentioned, both innovative behavior and factor allocation optimization behavior are essentially composed of enterprise investment and management activities, with the efficiency of these activities relying on financing [68]. When GFP is implemented, green enterprises exhibit an increased willingness to engage in two types of behavior as their financial resources grow [82]. In contrast, the willingness of polluting enterprises to engage in these two types of behavior decreases [69]. This constitutes the internal mechanism by which GFP exerts an asymmetric impact on enterprises’ GTI.

5.2.2. Specific Channels

After identifying the overall mechanism, we further investigated the specific channels through which GFP impacts GTI. Specifically, we sought to identify the concrete enterprise activities through which green finance impacts GTI. We replaced the dependent variable in the basic model with indicators that capture and reflect four types of enterprises’ activities: R&D, capital renewal, asset restructuring, and strategic decision-making. Among them, R&D activities belong to enterprise innovation behavior, and the latter three types of activities belong to enterprise factor allocation optimization behavior.

(1)

R&D activities.

R&D activities are classified into two categories: the first category measures R&D input through R&D personnel input (RDP) and R&D funding input (RDF), and the second category measures R&D efficiency through the number of independent patent applications (IPA) and the number of joint patent applications (JPA). Columns (1) to (4) of

Table 7. Specific Channels of GFP Impacting GTI: R&D Activities.

Variables	RDP		RDF		IPA		JPA
	(1) PE	(2) GE	(3) PE	(4) GE	(5) PE	(6) GE	(7) PE	(8) GE
GFRIpilot	−0.050 (0.042)	0.111 *** (0.023)	1.055 *** (0.062)	0.208 *** (0.022)	−0.045 (0.088)	0.355 *** (0.066)	−0.799 *** (0.150)	−0.629 *** (0.219)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	2889	2133	2889	2133	2889	2133	2889	2133
Adj-R2	0.075	0.071	0.116	0.216	0.225	0.262	0.139	0.088

Note: The numbers in parentheses are robust standard errors. *** denotes the 1% significance level.

indicate that GFP increases the RDF of polluting enterprises but does not increase their RDP. Furthermore, GFP synchronously increases the RDP and RDF of green enterprises. Columns (5) to (8) indicate that GFP significantly reduces the JPA of polluting enterprises, with no significant impact on their IPA. Furthermore, GFP significantly reduces the JPA of green enterprises but significantly increases their IPA.

(2)

Capital renewal activities.

Capital renewal activities are divided into two categories: the first group reflects noncurrent asset transactions, which are measured by the value of fixed asset additions (FAA) and the value of fixed asset disposals (FAD). The second group reflects asset consumption, which is measured by the depreciation rate of fixed assets (DR) and the comprehensive amortization rate of fixed assets, intangible assets, and productive biological assets (CAR). Columns (1) to (4) of

Table 8. Specific Channels of GFP Impacting GTI: Capital Updates Activities.

Variables	FAA		FAD		DR		CAR
	(1) PE	(2) GE	(3) PE	(4) GE	(5) PE	(6) GE	(7) PE	(8) GE
GFRIpilot	0.411 *** (0.009)	−0.063 *** (0.008)	2.220 *** (0.224)	−1.966 *** (0.135)	−0.157 *** (0.045)	0.083 * (0.049)	−0.246 *** (0.033)	0.068 * (0.034)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	2889	2133	2889	2133	2889	2133	2889	2133
Adj-R2	0.972	0.952	0.071	0.066	0.060	0.075	0.088	0.168

Note: The numbers in parentheses are robust standard errors. * and *** denote the 10% and 1% significance levels, respectively.

indicate that GFP increases the FAA and FAD for polluting enterprises and decreases these for green enterprises. Columns (5) to (8) reveal that GFP reduces the DR and CAR for polluting enterprises but increases these for green enterprises. When the noncurrent asset turnover increases but the depreciation rate decreases, it typically indicates that the enterprise is pursuing a strategy of scaling up its operations. With the increase in financing costs due to GFP, polluting enterprises are more likely to focus on improving their profitability through scaling up their operations rather than investing in GTI. For green enterprises, GFP reduces their inclination toward excessive expansion and enables them to allocate more resources toward improving their GTI.

(3)

Asset restructuring activities.

Asset restructuring activities can be divided into two categories: the first group reflects corporate acquisitions (CA), measured by the amount spent on acquiring subsidiaries. The second group reflects changes in equity ownership (EO), measured by the concentration of equity ownership. Columns (1) to (4) of

Table 9. Specific Channels of GFP Impacting GTI: Asset Restructuring Activities and Strategic Decision-making Activities.

Variables	CA		EO		FL		CF
	(1) PE	(2) GE	(3) PE	(4) GE	(5) PE	(6) GE	(7) PE	(8) GE
GFRIpilot	4.369 *** (0.844)	−0.508 * (0.316)	−1.570 ** (0.707)	1.704 *** (0.485)	7.849 *** (4.583)	−4.063 *** (0.870)	−0.255 ** (0.118)	0.140 (0.091)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	1163	858	2889	2133	2889	2133	2889	2133
Adj-R2	0.052	0.058	0.102	0.220	0.009	0.015	0.038	0.033

Note: The numbers in parentheses are robust standard errors. *, **, and *** denote the 10%, 5%, and 1% significance levels, respectively.

indicate that GFP stimulates acquisition activities of polluting enterprises and reduces their equity concentration while suppressing acquisition activities of green enterprises and increasing their equity concentration. Acquisition activities may benefit polluting enterprises in achieving green technology transformation through non-innovative means by acquiring the technology of the acquired firm. However, the reduction in equity concentration resulting from acquisitions will further undermine the enterprises’ determination to make long-term decisions such as investing in R&D, which is detrimental to their GTI. By contrast, GFP has the opposite effect on green enterprises.

(4)

Strategic decision-making activities.

We consider the financial leverage (FL) and cash flow per share from investment activities (CF) to reflect the corporate operating strategy. Columns (5) to (8) of Table 9 demonstrate that GFP increases the FL of polluting enterprises while reducing their CF. This suggests that GFP leads to increased debt financing but decreased investment efficiency for polluting enterprises. This inevitably results in a decline in the efficiency of factor allocation, thereby reducing GTI. GFP lowers the FL of green enterprises but has no significant impact on their CF. This indicates that GFP reduces the willingness of green enterprises to engage in debt financing. However, it is difficult to determine whether this is the reason for the improvement in the efficiency of factor allocation.

5.3. Synergy Analysis

Based on the empirical findings, it can be concluded that the implementation of GFP in China has resulted in a polarization effect on green innovation. This policy has promoted green enterprises’ GTI while inhibiting the GTI of polluting enterprises through different impacts on their behavior. As a result of this policy, a potential outcome could be that traditional industrial firms striving to enhance their production techniques in an environmentally sustainable manner might not have access to financing support [69,82]. Meanwhile, the financial restrictions for non-innovative activities of green industrial enterprises could be alleviated. This is clearly detrimental to the rational allocation of financial resources and GTI [11,12]. The challenge of differentiating between GFP and industrial policy in developing countries poses a significant obstacle to implementing short-term changes. Therefore, it is becoming increasingly important and urgent to establish a supportive external policy environment that can mitigate the polarization effect of GFP on GTI. Therefore, in this section, we focus on exploring the synergistic effects of GFP and other environmental policy instruments in stimulating GTI, with the aim of establishing effective incentive and constraint mechanisms for GFP. Our objective is to enhance the positive effects of GFP while offsetting its negative impact, thereby minimizing the polarization effect on green innovation. This approach can promote a steady and balanced increase in the willingness and efficiency of various types of enterprises toward green innovation.

To test Hypothesis H3, we have employed four types of environmental policy instruments to investigate the synergistic effects between GFP and supplementary policies: R&D subsidy (EP₁), environmental subsidy (EP₂), pollution charge (EP₃), and pollution control (EP₄). As presented in

Table 10. Synergy of GFP and Supporting Policies (I).

Variables	GTI
	(1) PE	(2) GE	(3) PE	(4) GE	(5) PE	(6) GE	(7) PE	(8) GE
GFRpilot × EP1	−0.052 * (0.030)	0.280 *** (0.063)
GFRpilot × EP2			0.499 * (0.258)	−1.482 *** (0.522)
GFRpilot × EP3					−0.187 ** (0.076)	0.450 * (0.273)
GFRpilot × EP4							0.083 (0.162)	−0.413 (0.384)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	814	604	392	142	556	152	2889	2133
Adj-R2	0.232	0.234	0.295	0.920	0.203	0.283	0.159	0.173

Note: The numbers in parentheses are robust standard errors. *, **, and *** denote the 10%, 5%, and 1% significance levels, respectively.

, first, the coefficient of the interaction term between GFRpilot and EP₁ is significantly negative for polluting enterprises and significantly positive for green enterprises. This finding suggests that R&D subsidy policy reinforces the polarization effect of GFP on GTI. The use of these two policies will further widen the GTI gap between polluting and green enterprises. Second, the coefficient of the interaction term between GFRpilot and EP₃ is significantly negative for polluting enterprises and significantly positive for green enterprises. This suggests that pollution charge exhibits similar attributes to R&D subsidy. Third, the coefficient of the interaction term between GFRpilot and EP₂ is significantly positive for polluting enterprises and significantly negative for green enterprises. This suggests that the environmental subsidy policy weakens the polarization effect of GFP on GTI. The use of these two policies will narrow the GTI gap between polluting and green enterprises at the cost of a decrease in the GTI of green enterprises. Finally, the coefficient of the interaction term between GFRpilot and EP₄ is not significant for polluting and green enterprises. This suggests that pollution control policy does not influence the polarization effect of GFP on GTI, and no synergistic effect exists between these two policies.

In the previous part, the synergistic effects of GFP and different supplementary policies are discussed. The results showed that only the environmental subsidy policy has the potential to weaken the polarization effect of GFP on green innovation. None of the other supplementary policies were found to have a simultaneous positive impact of GTI on GFP for both polluting and green enterprises. Therefore, we have further explored the synergistic effects of GFP and two types of supplementary policies to identify a policy combination that can promote the simultaneous improvement of GTI for green and polluting enterprises. Columns (1) to (3) of

Table 11. Synergy of GFP and Supporting Policies (II).

Variables	GTI
	(1) TE	(2) PE	(3) GE	(4) TE	(5) PE	(6) GE
GFRpilot × EP1 × EP2	0.334 ** (0.154)	0.520 ** (0.204)	0.246 *** (0.071)
GFRpilot × EP2 × EP3				0.045 (0.071)	−0.020 (0.133)	0.056 (0.073)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
N	404	275	129	118	105	101
Adj-R2	0.343	0.472	0.975	0.772	0.779	0.810

Note: The numbers in parentheses are robust standard errors. ** and *** denote the 5% and 1% significance levels, respectively.

indicate that the coefficients of the three-way interaction terms (GFRpilot, EP₁, and EP₂) are significantly positive for the total sample, polluting enterprises, and green enterprises. Moreover, the coefficient for the polluting enterprises is larger than that for the green enterprises. This suggests that the policy combination of GFP, environmental subsidy, and R&D subsidy can weaken GFP’s polarization effect on green innovation without reducing GTI for green enterprises, thereby promoting the simultaneous improvement of GTI for polluting and green enterprises. However, columns (4) to (6) indicate that the coefficients of the three-way interaction terms (GFRpilot, EP₂, and EP₃) are not significant at any level. This finding suggests that the policy combination of GFP, environmental subsidy, and pollution charge does not significantly impact GTI. In fact, the implementation of environmental subsidies and pollution charges is relatively contradictory, and their effects offset each other, as can be anticipated [71].

6. Conclusions and Policy Implications

6.1. Conclusions

The advancement of GTI is fundamental for green development. Accelerating GTI in China, the world’s largest carbon-emitting economy, is crucial for promoting global green economic growth and reducing carbon emissions. Exploring the impact of China’s GFP on GTI is crucial as GFP is a widely used approach for promoting the development of GTI among enterprises. The findings from this exploration can provide valuable insights and guidance for designing carbon reduction policies, especially in developing countries around the world. Therefore, we utilized microdata from Chinese industrial enterprises between 2015 and 2021 as a sample, with the pilot policy of China’s Green Finance Reform and Innovation Pilot Zone as a “quasi-natural experiment”. Furthermore, we employed the DID method to explore the impact and mechanism of China’s GFP on enterprise GTI. In addition, we discussed the pathway for the optimization of China’s green financial system. The main conclusions are as follows:

(1)

GFP has a green innovation polarization effect. It promotes the GTI of green enterprises but inhibits the GTI of polluting enterprises. Thus, green enterprises can easily access cleaner production technology and products; however, polluting enterprises may find it challenging to transform to an eco-friendly model. This conclusion holds true even after replacing the variable measurement method, altering the sample selection method, transforming the econometric model, and conducting a counterfactual placebo test.

(2)

The mechanism test reveals that GFP enhances the GTI of green enterprises by promoting innovative and factor allocation optimization behaviors. Conversely, it reduces the GTI of polluting enterprises by encouraging non-innovative investments that crowd out R&D investments and by lowering the efficiency of factor allocation optimization. The role of GFP in guiding enterprise behavior is specifically manifested in R&D, capital renewal, asset restructuring, and strategic decision-making activities.

(3)

The analysis results of the policy instrument synergy indicate that combining policies utilizing GFP, environmental subsidy, and R&D subsidy can effectively increase the GTI of polluting enterprises without compromising the GTI of green enterprises. This approach will help mitigate GFP’s polarization effect on green innovation and promote the simultaneous improvement of GTI for polluting and green enterprises.

6.2. Policy Implications

The findings have the following two policy implications:

(1)

Acknowledging the polarization effect of GFP on green innovation and expediting the refinement of its policy objectives are crucial. In China, the existing support of GFP for technological innovation in green enterprises has been at the cost of hindering polluting enterprises’ motivation to enhance their production technologies. This situation can lead to a scenario where both traditional industries and polluting enterprises may lose their ability to independently conduct R&D activities. They may rely on measures such as asset restructuring or end-of-pipe treatment to evade policy regulations. In the long run, it may result in an influx of inefficient investments in eco-friendly industries and green enterprises, which could ultimately weaken the overall competitiveness of national enterprises and slow down the momentum of sustainable development. Therefore, at the policymaking level, developing countries, such as China, should accelerate the improvement and innovation of GFP. First, to promote a balanced development of GTI across various industries, a standardized and measurable green finance classification system should be established. This would guide policy direction towards enterprise “green project orientation” and “environmental behavior orientation” instead of “green industry orientation” in terms of financial asset allocation. Second, the development of GFP instruments should be accelerated, and the current monopoly of green credit should be addressed. Diversification of green financial products, such as green bonds, insurance, funds, and others, would strengthen the overall support for clean production projects and potentially transformative environmental enterprises. This would lead to a more balanced promotion of GTI across different industries.

(2)

To achieve synergistic effects on GTI in enterprises, it is crucial to construct a green financial system rationally and fully utilize various environmental policy instruments. Prioritizing the simultaneous implementation of appropriate intensity environmental and R&D subsidies is vital while actively reforming GFP. Specifically, there should be an increase in direct R&D subsidies and tax reductions for enterprises to encourage green enterprises to promote innovation investment, thereby maintaining and enhancing their technological innovation capabilities. Furthermore, subsidies for green product production, green technology innovation, and high-carbon industry transformation should be increased to encourage polluting enterprises to increase innovation investment, thereby weakening the polarization effect of GFP on green innovation. Furthermore, it is important to accelerate the enhancement of pollution charges, such as environmental taxes, green fees, and pollution control fees, and concentrate on establishing pollution rights and carbon emissions trading markets. This measure will gradually promote the transformation of environmental regulation from a command-oriented approach to a market-oriented approach and enable environmental regulation policies to play a pivotal role in driving the incentive effects of GFP on GTI.

6.3. Future Work

For future study on the efficiency evaluation of China’s GFP, two potential directions are provided as follows:

(1)

Focus on analyzing the economic and social effects of GFP. While this paper explores the relationship between China’s GFP and enterprise GTI, in reality, enterprise GTI further affects the production and operational efficiency of enterprises, the scale and structure of industries, the quality of regional economic growth, and the prospects for China’s green development. Compared to technological innovation, it is also worthwhile to explore the specific impact of GFP on the aforementioned factors. For instance, to investigate the impact of GFP on enterprise production transformation or the quality of regional economic growth, we believe that green total factor productivity (GTFP) is a suitable indicator. Analyzing the relationship between GFP and indicators such as GTFP and green investment efficiency is a research direction worth considering.

(2)

Further discuss the synergistic effect of GFP and its supplementary policies. This paper analyzes the synergistic effect of different types of environmental policy instruments, including GFP, on GTI, but the analysis is still relatively crude. The biggest problem is the severe lack of data representing the indicators of environmental policy instruments. However, due to the limitations of data sources, we have not yet found a more comprehensive and reliable data source. Therefore, we suggest attempting to construct reasonable proxy variables and including more types of environmental policy instruments in the research scope to comprehensively analyze the synergistic effect of GFP and its supplementary policies.