Positive impacts of green finance on environmental protection investment: Evidence from green finance reform and innovations pilot zone

Abstract

After long-term development, the global economic level has improved significantly, but environmental issues generated by early extensive development seriously threaten the survival of human beings. China, in particular, urgently needs to promote sustainable development through green finance policies. For this reason, this paper regards the 2017 eight pilot zones in five provinces for green finance reform and innovations (GFRIs) as a quasi-natural experiment, and explores whether it can encourage investment in environmental protection in heavily polluting enterprises by using difference-in-differences-in-differences (DDD) model. The paper finds that: First, GFRIs can bolster investment in environmental protection in heavy polluting enterprises. The results remain consistent after several robustness checks, covering the placebo test, PSM-DID test and so on. Second, mechanism tests find that the policy promotes environmental protection investment by alleviating financing constraints and cutting financing costs. Third, heterogeneity analysis shows that the promotion effect of GFRIs on environmental protection investment is more pronounced for provinces with higher percentages of secondary industry GDP, large-scale enterprises, and enterprises with better ESG management. This paper demonstrates the beneficial influence of GFRIs on promoting the transformation of heavy polluting enterprises and provides suggestions for the improvement of such policies.

1. Introduction

After long-term development, the global economic level has significantly improved. However, the early extensive development strategy resulted in significant environmental and ecological issues. China, in particular, has urgent environmental concerns. In 2016, China's emissions of sulfur dioxide and nitrogen oxides reached 21.21 and 21.84 million tons, respectively. To address these challenges, China has committed to sustainable development, increased its investment in environmental preservation, and regards green finance as a crucial tool.

In fact, in order to deal with environmental problems, all countries in the world have carried out innovation in green finance, and achieved certain results. For example, based on panel data from India, Nenavath and Mishra found that green finance promotes environmental protection and high-quality growth of the economy [1]. Based on 53 transnational panel data, Hou et al. found that green finance promoted development of renewable energy, and its influencing mechanism was to promote investment in renewable energy fixed assets and technological industry innovation [2].

Early on, China used mandatory environmental legislative measures but gradually adjusted its strategy as its socio-economic development and environmental philosophy matured. In particular, green finance practices and academic research are widely carried out in China. Scholars have mainly used the Green Credit Guidelines released in 2012 as a quasi-natural experiment, or the construction of a regional green finance development index, to study the development of green finance in China. For example, Lu et al. constructed the Difference-in-differences (DID) model based on the Green Credit Guidelines, and concluded that China's green finance policies increased the financing constraints and debt financing costs of heavily polluting enterprises, thus encouraging their green transformation [3]. Li et al. constructed the Green Finance Index and concluded that green finance can promote the quality and quantity of green innovation of Chinese enterprises and play a more positive role in heavily polluting industries [4]. Yu et al. constructed China's regional green finance development index, and found that green finance promoted the R&D and innovation of green enterprises, thus improving the financial performance [5].

However, there are some limitations in the research methods of these studies. First, in many research about the Green Credit Guidelines, most scholars distinguish the experimental group from the control group according to whether the enterprise belongs to the heavy polluting industry or not, which makes the empirical results unable to exclude the interference of other environmental protection policies in the same period. Second, some scholars build green finance development index based on a series of indicators such as regional green credit, green insurance and green bonds, and make regression analysis, which may have endogenous problems.

China has introduced many policies to boost green finance since 2016, such as the Guidance on Promoting the Development of Green Finance, the Guidance on Effectively Strengthening the Management of Green Credit, and the Measures for the Management of Green Bonds, which boost green finance growth. In 2017, China set up eight pilot zones in five provinces for green finance reform and innovations (GFRIs), this provides a real and good quasi-natural experiment for us to study the development of green finance in China, and empirical research based on this can effectively alleviate the endogenous problem.

At present, scholars have made some achievements in the study of GFRIs. In the macro research, the existing literature generally believes that GFRIs has a positive impact on regional green development. Li et al. took GFRIs as a quasi-natural experiment based on the panel data of prefecture-level cities in China, and found that it could significantly improve the development level of regional green finance and environmental quality [6]. Hou et al. also believe that GFRIs has improved regional environmental quality, and the influencing mechanism is digital finance and green innovation [7]. Further, Zhang et al.’ research found that GFRIs can promote regional high-quality green innovation and thus improve energy utilization efficiency [8]. Lin and Zhong found that GFRIs can improve the efficiency of resource allocation, promote green innovation and industrial structure upgrading, and further enhance regional green total factor productivity [9]. Hu et al.’ study found that GFRIs also promoted the low-carbon transformation of regional economy, but only in eastern and central China [10]. In terms of methods, in addition to DID model, Hu and Wang studied the impact of GFRIs on regional GDP through PSM-DID model. Tang et al. studied the influence of GFRIs on regional green innovation based on the synthetic control method [11].

In the micro research, the existing literature mainly studies the impact of GFRIs on enterprise investment and financing decisions, green innovation and so on. By constructing a DID model, Liu et al. found that GFRIs reduces the debt financing cost of enterprises, increases R&D investment and foreign investment, and thus promotes enterprise innovation [12]. By constructing a DID model, Yan et al. found that GFRIs improves the investment efficiency of enterprises [13]. In addition, GFRIs has a positive impact on enterprises' risk taking, reducing the risk of stock price crash, and total factor productivity [[14], [15], [16]]. Although the literature on the positive impact of GFRIs on businesses is extensive, the conclusions of many studies are inconsistent. In terms of financing constraints, Shi et al.’ research believe that GFRIs reduces the financing cost of heavily polluting enterprises [17], while Zhang believes that GFRIs has no impact on the financing cost of enterprises and reduces the financing scale of heavily polluting enterprises [18]. In terms of green innovation, Wang and Zhang found that GFRIs had a positive impact on the R&D and innovation of enterprises’ green invention patents and green utility model patents [19]. Jia et al. and Zeng et al. believe that the effect of GFRIs on enterprise green innovation is more obvious in non-heavy polluting enterprises [20,21]. Ran and Zhang argued that GFRIs significantly inhibited the green innovation of heavy polluting enterprises [22].

Based on the theory of environmental legitimacy, the behavior of enterprises should meet environmental protection provisions of policies and laws, and also meet expectations of the public, media, investors and other stakeholders on environmental protection. Not only does GFRIs raise society's bar for the environmental legitimacy of enterprises, but it can provide sustained economic incentives for the green transition of heavy polluters through green finance. Therefore, this study believes that GFRIs is conducive to heavy polluting enterprises to increase environmental protection investment, and the influencing mechanism is to provide financing incentives to them and reduce financing costs [17], rather than further restrict their sustainable development.

However, the research on the impact of GFRIs on the green development of enterprises is not perfect, especially the conclusions of some research are contradictory. This study suggests that two problems contribute to this problem. First, most literature study on the impact of GFRIs on enterprises based on the DID model, because it is widely used in the evaluation of the effect of pilot policies. However, China's listed enterprises belong to a wide variety of industries. Some research samples only include heavily polluting industries, while others include all industries except the financial industry. The results of DID model are not robust. Second, when most literature study the impact of GFRIs on enterprises' green innovation, the measurement index is green patent. However, invention patents and utility model patents are very different, and patents in different industries are also very different. Therefore, green patents can not always well reflect green transformation effect of enterprises.

With this in mind, we constructed the Difference-in-Difference-in-Differences (DDD) model to assess the impact of GFRIs on environmental investments in heavily polluting enterprises. Our motivation is that compared with DID model, DDD model can add a difference term to control the difference between heavy and non-heavy polluting industries and achieve a cleaner effect [16,23,24]. At the same time, we focus on enterprises' environmental investment rather than enterprises' green patents, because enterprises’ environmental investment can be measured in monetary terms, its horizontal and vertical comparability is higher. Although some scholars believe that compared with green innovation, environmental protection investment is a means of terminal governance rather than source governance [25], it is a means suitable for the green transformation of various types of enterprises with lower research and development difficulty, less investment and faster results. In 2022, 17,468 environmental protection investment projects have been authorized in China overall, with an investment of more than 3.36 trillion RMB in nearly 30 fields, highlighting the priority of environmental protection investment in the development of ecological civilization, industrial structure upgrading and stable economic growth.

The conclusion of this study has following three contributions. First, we built a DDD model to evaluate the impact of GFRIs on the environmental protection investment of heavy polluting enterprises, and further provided evidence for its positive impact on the green transformation of heavy polluting enterprises. As mentioned above, compared with DID model, DDD model has more advantages in evaluating the effect of pilot policies.

Second, based on the legitimacy theory, we provide evidence for GFRIs to ease financing constraints and reduce debt financing costs for heavy polluting enterprises. To a certain extent, this also enriches the research results on the relationship between green finance and enterprise financing, and supports its incentive effect rather than constraint effect on heavily polluting enterprises.

Finally, according to the aforementioned theoretical analysis and empirical research, this paper puts forward a series of specific policy suggestions, aiming to promote further optimization and popularization of GFRIs, and also provides valuable references and lessons for green finance practice in other developing countries.

2. Institutional background and theoretical mechanisms

2.1. Institutional background

Environmental issues include diminishing forest coverage rate, disappearing wetlands, species extinction, ecological imbalance, growing greenhouse impact, and pollution emissions have gained prominence as a result of China's fast urbanization. As these problems pose a serious threat to human survival and health, China has implemented administrative environmental control measures. Specifically, China promulgated the Law of Environmental Protection (for Trial Implementation) in 1979, the Regulations of the Prevention and Control of Environmental Noise Pollution in 1982, the Law on Marine Environmental Protection in 1989, and the Law on Environmental Impact Assessment in 1994, these regulatory documents mainly cover the fields of air, water, sea, solid waste, environmental protection and noise, and are implemented as mandatory command instruments. Environmental restrictions have been somewhat successful, but they have also had some negative effects. For example, they have made businesses less enthusiastic and increased their financial burden, which has hampered their growth and ability to compete in the market.

In 2012, China issued the Green Credit Guidelines to prioritize green credit in important sectors and encourage loan support for environmental and energy conservation. In 2015, China launched the Green Bond Standard to clarify the conditions and criteria for issuing green bonds to stimulate the growth of the environmental protection business. In 2016, the Guidance on Building a Green Financial System elaborated on the concept of green finance, which is a financial service that facilitates ecological improvement, climate issues and resource efficiency. However, these policies are flawed in five main ways: (1) Lack of standardization. The complexity of the green finance sector lacks consistent definitions, norms and standards leading to a lack of clarity and specificity in policy design. (2) Difficulty in assessing feasibility. The neglect of long-term feasibility assessment of green projects leads to difficulty in measuring risk and hinders China and investors from making informed decisions. (3) High-risk and low-return. Early green financial projects are challenged by the immaturity of technology and market, and enterprises prefer traditional commercial projects. (4) Lack of incentives. Policy subsidies to support the development of green enterprises, but there are side effects such as the inequitable distribution of funds. (5) Regulatory shortcomings. The green finance market regulatory system still needs to be improved to make up for regulatory loopholes and other problems.

On June 14, 2017, China adopted the policy of eight pilot zones in five provinces (GFRIs), including Zhejiang, Guangdong, Xinjiang, Guizhou and Jiangxi Provinces. GFRIs is committed to fostering policy innovation and the prosperity of green economy, and achieved a series of remarkable results. Gansu's approval to participate in the pilot program in 2019 represents a further development and enhancement of China's green finance reform. In the same year, China issued the Notice on Increasing Green Financial Support to the Fight Against Pollution, which proposes to support GFRIs. Financial institutions continued to take specific steps to boost green financial assistance and encourage green financial businesses growth.

The pilot policy was officially implemented in China in 2017. Specific features are as follows: (1) The pilot area represents the whole country, mainly concentrating on the east coast and the central and western regions, ensuring that many geographical areas benefit from the fruits of green financial innovation. (2) The pilot aims to expand the application area of green finance and breed rich financial products by involving numerous industries, including clean energy, smart manufacturing, bio-information technology, and other emerging industries. (3) The pilots adopt various models, including regional integrated, key areas and niche pilot areas, in order to reflect local characteristics and more effectively promote the practice of green financial innovation strategies.

GFRIs has achieved remarkable results. Data show that by 2020, China's green bond issuance totaled nearly 3 trillion RMB, and the size of green financial loans in Guangdong Province has increased to nearly 200 billion RMB.

2.2. Theoretical mechanisms

Legitimacy theory emphasizes the social acceptability of an organization's behavior. It asserts that the legitimacy of enterprise behavior is not limited to compliance with regulations, but also lies in meeting a wide range of social expectations, including the needs of the public, public opinion, and investors [[26], [27], [28]]. Enterprises can only obtain social resources if they meet diversified needs. Conversely, if enterprise behavior is illegal, they may suffer negative impacts such as administrative fines, financing constraints and public opinion pressure.

The theory of environmental legitimacy posits that enterprise investments ought to conform to environmental regulations in order to be deemed legitimate. This means that for enterprise investment behavior to gain societal recognition, it must not only meet environmental regulations but also the public's environmental expectations. GFRIs has played a key role in raising environmental legality standards: first, the stringent application of the policy has barricaded environmental legality. Second, a widespread environmental consciousness has developed as a result of the public's growing environmental consciousness and the propagation of environmental propaganda, which further advances environmental legitimacy. Thus, this paper makes the case that the innovation pilot zone policy and green financial reform can raise the bar for environmental legitimacy, help enterprises transition to more environmentally friendly models, and encourage investment in environmental preservation.

By reducing financing costs and loosening financial limitations on large polluters, GFRIs can encourage investment in environmental protection. First, according to the environmental legitimacy hypothesis, businesses that adhere to environmental regulations and goals and make efforts to lessen their environmental impact in order to fulfill society expectations are more likely to acquire social resources. [29,30]. Therefore, the implementation of GFRIs provides greater financing space for enterprises by strengthening their environmental legitimacy, gradually establishing and developing market-based green funds, encouraging banks to increase lending to energy-saving and environmental protection projects, and establishing special credit lines. Second, GFRIs has greatly enhanced the level of environmental information disclosure by enterprises and further strengthened the transparency of information in the market [31,32]. The policy significantly reduces the asymmetry in information between enterprises and financial institutions by incentivizing enterprises to actively publicize their environmental protection actions and green business practices, promoting the transparent information management to minimize lending costs and relieve the financial strain on businesses. Finally, GFRIs plays an important role in facilitating green transformation of enterprises. The policy guides those enterprises to invest more resources in ecological preservation to build and enhance their green image, which not only highlights their enterprise commitment to environmental responsibility, but also releases clear signals to the market about the transformation, demonstrating their determination to step into green transformation and attracting continuous public attention and support [33,34].

In summary, GFRIs can alleviate the financing constraints of heavy polluters and reduce financing costs. Further, enterprise financing plays a crucial part in determining investment behavior, and financing strategies and their associated costs directly affect enterprise investment behavior and investment scale. In the context of green transformation, the promotion of green development is greatly aided by green financing. According to existing research, easing enterprise finance restrictions and lowering financing costs are key factors in promoting enterprises’ transition to a more environmentally friendly business model [35,36].

Based on what mentioned above, the author intends to propose two hypotheses.

Hypothesis 1

GFRIs can greatly encourage investment in environmental protection in heavy polluters.

Hypothesis 2

GFRIs can improve ecological protection investment through optimizing the financing environment of enterprises, i.e., alleviating financing constraints and reducing financing costs.

3. Research design

3.1. Model construction

3.1.1. Baseline regression model

DID model has been widely used in the impact assessment research of pilot policies. It is to make two differences, one is the time dimension which evaluates the difference before and after the policy impact, and one is the individual dimension which evaluates the difference between the experimental group and the control group in the policy. However, environmental protection policies do not have a significant impact on all enterprises, and DID model may not be able to get a clean policy effect.

Therefore, we adopt DDD model to estimate policy effects. Compared with DID model, DDD model added a difference term, Pollution, which we use to represent the difference between heavy polluters and non-heavy polluters, allow us to assess the effect of cleaner policies. The model is shown below:

$$EnvironmentalInves{t}_{i,t+\mathit{1}}=\alpha +\beta Tim{e}_t\times Trea{t}_i\times Pollutio{n}_{i,t}+\chi Tim{e}_t\times Trea{t}_i+\delta Tim{e}_t\times Pollutio{n}_{i,t}+\eta Trea{t}_i\times Pollutio{n}_{i,t}+X_{i,t}+{\mu }_t+{\lambda }_i+{\epsilon }_{i,t}$$ (1)

Where $i$ stands for individual enterprise, $t$ stands for year; ${EnvironmentalInvest}_{i,t+1}$ denotes enterprise's level of environmental investment; ${Time}_t\times {Treat}_i\times {Pollution}_{i,t}$ denotes policy variable; $X_{i,t}$ denotes a set of control variables; ${\mu }_t$ stands for year fixed effect, ${\lambda }_i$ denotes individual fixed effect; and ${\epsilon }_{i,t}$ is a random error term. Meanwhile, to alleviate endogenous problem, the policy variables are given a one-period lag to all control variables in this paper.

The coefficient $\beta$ should be significantly positive if GFRIs can facilitate environmental protection investment by heavy polluters.

3.1.2. Mechanism testing model

To test the mechanism in this study, a three-step technique was chosen and the model was built as followed:

$$FC/Co{d}_{i,t+\mathit{1}}=\alpha +\beta Tim{e}_t\times Trea{t}_i\times Pollutio{n}_{i,t}+\chi Tim{e}_t\times Trea{t}_i+\delta Tim{e}_t\times Pollutio{n}_{i,t}+\eta Trea{t}_i\times Pollutio{n}_{i,t}+X_{i,t}+{\mu }_t+{\lambda }_i+{\epsilon }_{i,t}$$ (2)

$$EnvironmentalInves{t}_{i,t+\mathit{1}}=\alpha +\beta Tim{e}_t\times Trea{t}_i\times Pollutio{n}_{i,t}+\vartheta FC/Co{d}_{i,t+\mathit{1}}+\chi Tim{e}_t\times Trea{t}_i+\delta Tim{e}_t\times Pollutio{n}_{i,t}+\eta Trea{t}_i\times Pollutio{n}_{i,t}+X_{i,t}+{\mu }_t+{\lambda }_i+{\epsilon }_{i,t}$$ (3)

Where ${FC}_{i,t+1}$ denotes the degree of enterprise financing constraint, ${Cod}_{i,t+1}$ denotes the degree of financing of the enterprise, and the others are the same as in model (1). If green finance may ease the funding constraints of heavy polluters, lower their financing costs and thus promote their environmental protection investment, $\beta$ in model (2) should be statistically significantly positive and $\beta$ in model (3) is reduced from $\beta$ in model (1).

3.2. Variable definitions

3.2.1. Explanatory variable: GFRIs

The dummy variable Time is constructed to indicate the pilot time of the policy, and Time is assigned to 1 if the year is after 2017, otherwise it is assigned to 0. The dummy variable Treat is constructed to indicate the policy pilot area, and Treat is assigned to 1 if the province where the enterprise is registered as pilot province, otherwise, it is 0. The dummy variable Pollution is constructed to introduce another pair of experimental and control groups, and Pollution is assigned a value of 1 if the type of industry where the enterprise is situated is a heavy polluter and 0 otherwise. Therefore, $Time\times Treat\times Pollution$ is 1 only when the sample is a heavy polluter in a pilot province after 2017, which can accurately assess the effect of GFRIs.

As like many previous research, regarding Pollution of Heavy Pollution Industry, this paper is based on the Management List of Environmental Protection Verification Industries for Listed Enterprises issued by China in 2008, which is determined by comparing 2012 edition of China's industry classification[16]. If a listed enterprise's sector produces a lot of pollution, it is considered to be a heavily polluting enterprise.

3.2.2. Explained variable: environmental investment

Based on the findings of Zhong et al. [37], this work in enterprises capitalized expenditures directly associated with environmental preservation in the line items of “construction in progress” in the annual reports of listed enterprises, including but not limited to research and development of environmental protection technology, investment and renovation of environmental protection facilities, investment in environmental protection technology improvement projects, et. We sum them up to obtain data on the year's growth in enterprise environmental protection spending.

3.2.3. Control variables

Using references from the literature on business technology innovation [38,39], Table 1 shows a collection of control variables used in this investigation. Among them, enterprise financial indicators include listed years (Age), enterprise size (Size), asset-liability ratio (Leverage), and return on total assets (ROA) to control for the effects of the enterprise life cycle, scale effect, financial leverage, and profitability on environmental investment, respectively.

Table 1.

Definition table of control variables.

Variable Name	Variable Definition	Variable Measurements
Age	The number of years since the enterprise was listed	Ln(Years of listing)
Size	Enterprise size, the natural logarithm of total assets	Ln(Total assets)
Leverage	Leverage ratio, the ratio of total debt to total assets	Total debt/Total assets
ROA	Return on assets	Net Profit/Total Assets
Board	Board size	Ln(Number of board members)
Independent	Proportion of independent directors	Number of Independent Directors/Number of Board of Directors
Holding	The shareholding ratio of managers	Number of shares held by managers/Total number of shares
Top10	The shareholding ratio of the top ten shareholders i	Shareholding of top ten shareholders
SOE	State-owned property	1 for state-owned enterprises, 0 for non-state-owned enterprises
GDP	Level of economic development	Ln(Regional GDP)
Industry	Economic structures	Gross secondary sector/Gross regional product
Fiscal	Environmental fiscal expenditure	Environmental fiscal expenditure/Regional GDP

Enterprise governance characteristic includes board size (Board), board independence (Independent), management shareholding (Holding), equity concentration (Top10), and nature of property rights (SOE).

Regional characteristics include economic level (GDP), economic structure (Industry) and environmental protection fiscal expenditure (Fiscal) to control the impact of investment on environmental preservation due to the economic growth and ecological protection.

3.3. Sample and data

The listed enterprises in Shanghai and Shenzhen A-shares from 2012 to 2021 make up the study's initial sample. Further, the following steps are used in this investigation to filter the sample: (1) eliminating ST and *ST enterprises; (2) eliminating enterprises in the banking and financial sectors; (3) eliminating enterprises listed for one year or less; (4) eliminating enterprises with asset-liability ratio larger than 1; and (5) eliminating enterprises with missing values of core variables. Finally, this study obtained 3210 listed enterprises with a total of 21621 observations.

Based on the sample identified above, this study collects enterprise financial data from the China Stock Market and Accounting Research (CSMAR) database, environmental investment data manually from financial statements, and relevant data at the provincial level from the China Statistical Yearbook.

4. Empirical analyses

4.1. Descriptive statistics

Descriptive statistics for our key variables are presented in Table 2. The mean value of the explained variable Environmental Investment in this study is 0.188, while the standard deviation is 1.508, with a minimum value of 0 and a highest value of 64.689, indicating a large dispersion of variables and considerable disparities in the degree of environmental investment among businesses, providing a suitable research context for this paper.

Table 2.

Descriptive statistics.

Variable	Obs	Mean	Std. Dev.	Min	Max
Environmental Invest	21817	0.188	1.508	0.000	64.689
Age	21817	2.193	0.768	0.693	3.434
Size	21817	22.280	1.330	15.577	28.636
Leverage	21817	0.422	0.203	0.008	1.280
ROA	21817	0.042	0.068	−0.805	0.880
Board	21817	2.241	0.178	0.000	2.944
Independent	21817	0.376	0.056	0.000	0.800
Holding	21817	0.132	0.194	0.000	0.900
Top10	21817	0.585	0.150	0.013	1.012
SOE	21817	0.363	0.481	0.000	1.000
GDP	21817	10.377	1.341	0.489	11.615
Industry	21817	0.402	0.098	0.074	0.577
Fiscal	21817	0.249	3.161	0.002	77.785

4.2. Baseline regression

4.2.1. Baseline regression results

As can be seen in column (1) of Table 3, the coefficient of the interaction term is positive at the 5 % significance level when no control variables are included. As shown in column (2), when control variables are included, the coefficient is still positive at the 5 % significance level, implying that the pilot zone policy can encourage heavy polluters in the pilot area to boost their investment in ecological preservation.

Table 3.

Baseline regression results.

	(1)	(2)
Time × Treat × Pollution	0.173**	0.181**
	(0.083)	(0.083)
Time × Treat	−0.023	−0.026
	(0.045)	(0.045)
Time × Pollution	−0.031	−0.027
	(0.040)	(0.040)
Treat × Pollution	0.086	0.097
	(0.233)	(0.233)
Listing Time		0.005
		(0.060)
Size		0.068***
		(0.026)
Leverage		−0.081
		(0.100)
ROA		−0.100
		(0.121)
Board		−0.021
		(0.089)
Independent		−0.035
		(0.229)
Holding		−0.152
		(0.113)
Top10		0.320**
		(0.143)
SOE		0.158
		(0.099)
GDP		−0.016
		(0.017)
Economic Structure		0.696
		(0.450)
Fiscal Expenditure		−0.001
		(0.005)
_cons	2.860***	−8.713***
	(0.011)	(0.538)
Year FE	Yes	Yes
Id FE	Yes	Yes
N	21621	21621
R2	0.459	0.460

Note: Robust standard errors in brackets, ***, ** and * indicate significant at the 1 %,5 % and 10 % levels respectively.

Baseline regression results show that green finance can promote the green transformation of heavy polluting enterprises. Just like the research results of Li et al. [4], green finance will not restrict the development of heavy polluting enterprises, but also enhance the sustainability and green competitiveness of enterprises.

4.2.2. Dynamic effects test

The consistency is founded on the assumption that the experimental and control groups follow the parallel trend hypothesis, i.e., there is no discernible change in environmental investment between enterprises in the pilot and non-pilot provinces. In this paper, we refer Jacobson et al. [40] and apply the Event Study Approach (ESA) to test the dynamic effects of the pilot policy and construct the model shown below:

$$EnvironmentalInves{t}_{i,t+1}=\alpha +{\sum }_{t=2012}^{2020}{\beta }_t{\mu }_t\times Trea{t}_i\times Pollutio{n}_{i,t}+\chi Tim{e}_t\times Trea{t}_i+\delta Tim{e}_t\times Pollutio{n}_{i,t}+\eta Trea{t}_i\times Pollutio{n}_{i,t}+X_{i,t}+{\mu }_t+{\lambda }_i+{\epsilon }_{i,t}$$ (4)

The dynamic effect test result is shown in Fig. 1.

Fig. 1.

Dynamic effects test.

As shown in Fig. 1, the top panel shows the dynamic effects plot with the first year of the sample period (2012) as the base period. The hollow points represent the interaction terms' coefficients, while the dashed segments represent the 95 % confidence intervals. First, prior to 2017, the coefficients are not significantly different from 0, presenting that no discernible change exists between the two groups and therefore the parallel trend hypothesis is satisfied. Second, after 2017, the coefficients of the interaction terms are substantially different from 0, indicating that the pilot zone policy has the result of encouraging environmental investment. Finally, the coefficient and significance of the interaction term declines in 2020, and this paper contends that the shock of the COVID-19 pandemic is to blame for this downward tendency. The epidemic in the early 2020s hit the global economy, which leads to the cut or postponement of heavy polluting enterprises’ investment in environmental protection.

4.3. Robustness tests

4.3.1. Placebo test

In order to eliminate the possible influence of potential unobservable factors on the empirical test of this paper, the placebo test was conducted by randomly sampling pilot provinces. Specifically, this paper selects the same number of enterprises as the experimental group (Treat×Pollution) from the sample through 500 random sampling, and regards them as the “pseudo-experimental group”, and multiplies with Time variable to generate a new “pseudo-policy variable”, and conducts regression test.

As shown in Fig. 2 above, the solid line represents the kernel density of the sampling results, the blue hollow point represents the regression coefficient of the “pseudo policy variable”, the horizontal dotted line on the vertical axis represents the significance level of 0.1, and the vertical dotted line on the horizontal axis represents the actual regression results in this paper.

Fig. 2.

Placebo test.

First of all, it can be found through the kernel density curve that the sampling results roughly present a typical normal distribution shape, indicating that the sample has enough randomness to meet the basic conditions of random sampling. Secondly, the kernel density curve reaches the maximum value at 0, indicating that most of the coefficients of the sampling results are concentrated at 0, and the “pseudo-policy variable” has not achieved significant effect. Finally, the vertical dotted line represents the real regression results of this paper (see column (2) in Table 3), which is a kind of outlier. Therefore, these results show that there is basically no potential impact of unobservable factors in the estimation results of this paper.

4.3.2. PSM-DID test

In the policy formulation process, the selection of pilot provinces is usually not completely random, which may lead to the experimental group being influenced by other factors in the selection process, thus making the difference between these two groups not entirely determined by the policy itself. To eliminate this possible endogenous concern, we use the PSM matching method to test. Specifically, first, all control variables that may affect the outcome are selected as covariates to calculate the propensity score. Second, with the help of the “nearest neighbor” matching approach, the author matches the experimental and control groups 1:2, so that the distribution of the covariates in the two groups tend to be similar.

In column (1) of Table 4, the coefficient Time×Treat×Pollution remains significantly positive, implying that there is no substantial distictions between the two groups. Therefore, our main findings are robust.

Table 4.

Robustness tests.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Time × Treat × Pollution	0.279**	0.181**	0.178**	0.186**	0.227**		0.080**
	(0.131)	(0.083)	(0.083)	(0.083)	(0.089)		(0.033)
Time × Treat	0.022	−0.032	−0.022	−0.025	−0.061	0.147**	−0.019
	(0.097)	(0.047)	(0.044)	(0.045)	(0.050)	(0.071)	(0.017)
Time × Pollution	−0.138*	−0.027	−0.031	−0.031	−0.064		0.0002
	(0.078)	(0.040)	(0.040)	(0.040)	(0.046)		(0.017)
Treat × Pollution	0.270	0.098	0.103	0.114	0.226		−0.060
	(0.486)	(0.233)	(0.234)	(0.234)	(0.324)		(0.075)
Environmental Tax		−0.013
		(0.041)
Big Data			−0.050
			(0.037)
_cons	0.723	−1.585***	−1.581***	−1.536***	−2.622***	−0.596	−0.844***
	(1.443)	(0.554)	(0.554)	(0.555)	(0.658)	(0.861)	(0.237)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Id FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	5090	21621	21621	21621	18623	8071	21621
R2	0.498	0.460	0.460	0.461	0.531	0.407	0.534

Note: Robust standard errors in parentheses, ***, **, * indicate significant at the 1 %, 5 %, and 10 % levels, respectively.

4.3.3. Controlling the impact of other policies

In the process of public policy implementation, there are often interfering factors from some other policies. The Environmental Protection Tax Law promulgated in 2016 proposed that China starts to levy environmental protection tax in 2018, and the target enterprises of this policy highly overlap with the enterprises studied in this paper. Thus, it may interfere with our study. This paper selects the subject of “Tax payable” by checking the enterprise. If the enterprise has paid “Environmental Tax” after 2018, the value of environmental tax is 1, otherwise it is 0, as shown in Table 4, columns (2) and (3). The results show that the coefficient Time×Treat×Pollution is still quite positive, denoting the implementation of the Environmental Protection Tax Law does not interfere with the study. Therefore, the findings are still solid.

In addition, many literature show that big data and digital economy will affect the green transformation of enterprises. This study refers Hu et al.[41] and Wang et al.[42], which took the National Big Data Comprehensive Pilot Zone established in China in 2016 as a quasi-natural experiment. The variable Big Data is constructed based on DID model and added to the regression model as a control variable. As shown in Table 4, columns (3), the results show that the coefficient Time×Treat×Pollution is still quite positive, denoting the implementation of the National Big Data Comprehensive Pilot Zone does not interfere with the study. Therefore, the findings are still solid.

4.3.4. Alternative measures

In reality, an enterprise's commitment to environmental preservation can take several forms, ranging from direct funding of environmental protection initiatives to a variety of costs borne in the course of conducting business as usual in order to meet environmental protection objectives. Despite the fact that these costs may be shown in the financial statements as “administrative expenses”, they actually contribute to the work and expense required to meet environmental goals.

In order to further ensure the robustness of the measurement, on the basis of the previous data, the expenses related to environmental protection fees, sewage fees and green fees in the item of “management expenses” of enterprises are also considered as the scope of environmental protection investment. According to the new measure, the discoveries are shown in column (4) of Table 4, and the coefficient is still larger than 0. It shows that even under the more comprehensive and accurate measurement of environmental protection investment, the conclusions of this paper are still valid. Therefore, the research conclusion of this paper is robust.

4.3.5. Supplementary data of Gansu Province

To ensure the accuracy of the study results, some factors that may interfere with the results need to be excluded when selecting the experimental sample. Gansu Province was selected as a new pilot zone in 2019, which may have an impact on the sample, so Gansu Province was directly excluded from the previous analysis to avoid its influence on the results.

In order to more accurately test the robustness of the research conclusions, Gansu Province was included in the sample in this part, and the study time was shortened to 2019, that is, the year when Gansu Province was selected as GFRIs, so as to further ensure the authenticity of the control group. Through further analysis of the data, as shown in column (5) of Table 4, the coefficient is rather positive. Therefore, the conclusion of this paper is robust.

4.3.6. Change the research model and sample

Although DDD model is more advantageous, this study replaced it with DID model and reduced the sample to heavy polluting industries to ensure the robustness of the results. As shown in column (6) of Table 4, GFRIs is represented by Time×Treat in heavily polluting enterprises, and its coefficient is significantly positive. Therefore, the conclusion of this paper is robust.

4.3.7. Replace singular values

In order to reduce the impact of singular values on the research results, all continuous variables in the regression model winsorized by 1 % and 99 %. As shown in column (7) of Table 4, the coefficient is rather positive. Therefore, the conclusion of this paper is robust.

5. Further analysis

5.1. Mechanism testing

5.1.1. Financing constraints

Regarding the measurement of financing constraints, referring Hadlock [43] and Pierce, Fee et al. [44], this paper constructs an FC index, and the specification is as follows:

$$\mathrm{P}\left(QUFC=1|Z_{i,t}\right)=\frac{{\mathrm{e}}^{Z_{i,t}}}{\left(1+e^{Z_{i,t}}\right)}$$ (5)

$$Z_{i,t}={\propto }_0+{\propto }_{1}siz{e}_{i,t}+{\propto }_{2}le{v}_{i,t}+{\propto }_3{\left(\frac{CASHDIV}{ta}\right)}_{i,t}+{\propto }_{4}M{B}_{i,t}+{\propto }_5{\left(\frac{NWC}{ta}\right)}_{i,t}+{\propto }_6{\left(\frac{EBIT}{ta}\right)}_{i,t}$$ (6)

The following steps are used to calculate the financing constraint variable FC. To begin with, three variables of business size, age, and cash dividend payout ratio are standardized, and financing constraint dummy variable QUFC is determined according to the average value of those standardized variables, whose mean value is higher than the third quartile of the enterprise's financing constraint degree is lighter, and the corresponding QUFC is 0, and the financing constraint degree of the enterprise below the third quartile is heavier, and the corresponding QUFC is 1. Second, the financial constraint index FC (taking values between 0 and 1) is used in logit model to fit the likelihood of a financing constraint occurring in each year, and it is defined as the financial constraint index FC (taking values between 0 and 1). The larger the FC, the more significant the enterprise's funding constraint situation. CASHDIV in model (2) denotes cash dividends declared in the year, ta stands for total assets, NWC stands for fortotal assets, and EBIT denotes earnings before interest and taxes.

Shown in column (1), in Table 5, the coefficient of Time×Treat×Pollution stays significantly negative, indicating that GFRIs will alleviate the financial constraints of heavy polluters in the pilot area.

Table 5.

Mechanism tests.

	(1)	(2)	(3)	(4)
	FC	Env_invest	Cod	Env_invest
Time × Treat × Pollution	−0.024***	0.176**	−0.315***	0.174**
	(0.007)	(0.083)	(0.070)	(0.083)
Time × Treat	−0.001	−0.026	0.096***	−0.024
	(0.005)	(0.045)	(0.035)	(0.045)
Time × Pollution	−0.012***	−0.030	0.609***	−0.015
	(0.004)	(0.040)	(0.050)	(0.040)
Treat × Pollution	0.038**	0.105	−0.107	0.095
	(0.016)	(0.234)	(0.096)	(0.233)
FC		−0.195***
		(0.058)
Cod				−0.021**
				(0.009)
_cons	3.372***	−0.928	−2.270**	−1.633***
	(0.109)	(0.564)	(1.075)	(0.555)
Controls	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Id FE	Yes	Yes	Yes	Yes
N	21621	21621	21621	21621
R2	0.877	0.460	0.596	0.460

Note: Robust standard errors in parentheses, ***, **, * indicate significant at the 1 %, 5 %, and 10 % levels, respectively.

Further, we consider it as a control variable, as shown in column (2), the coefficient of FC is negative, illustrating that alleviating financing constraints of heavy polluters in the pilot area will promote enterprise investment in ecological protection. Meanwhile, the coefficient of the interaction term decreases compared to model (1) (see column (2) in Table 3), proving that GFRIs can encourage enterprise investment in environmental preservation by relieving financial limitations of heavy polluters in the pilot area.

5.1.2. Financing costs cod

The ratio of “interest expense” to “total liabilities” is used to calculate financing costs. Compared with “finance costs”, “interest expense” can reflect the cost of financing more accurately by deducting the interest income of the enterprise.

Table 5 summarizes the findings of the mechanism test in this paper. As shown in column (3), the coefficient of the interaction term Time×Treat×Pollution proves negative, which shows that GFRIs reduces the financing cost of heavy polluters in the pilot area.

Furthermore, using it as a control variable, the coefficient of Cod is quite negative in column (4), revealing that reducing the financing cost of heavy polluters will promote enterprise investment in environmental preservation. Meanwhile, the coefficient of Time×Treat×Pollution decreases compared to model (1) (see column (2) of Table 3), showing that GFRIs can boost enterprises’ investment to protect environment by reducing the financial cost of heavy polluters in the pilot zone.

The results of mechanism test are consistent with Shi et al. [17]. GFRIs alleviates the financing constraints of heavy polluting enterprises and reduces the debt financing costs. The conclusion of this study further proves that GFRIs has a financing incentive effect on heavy polluting enterprises, rather than a constraint effect.

5.2. Heterogeneity analysis

5.2.1. Economic structure

In this paper, we use Industry as a proxy to measure the regional economic structure, and divide the provinces into non-industrial and industrial groups with the median of each year to group regressions on the benchmark regressions. The regression findings of economic structure heterogeneity are presented in Table 6. As displayed in column (1), the coefficient is not quite negative, indicating that GFRIs did not work in the non-industrial group. However, as seen in column (2), the coefficient remains significantly positive, indicating that the policy is effective in the industrial group.

Table 6.

Economic structural heterogeneity.

	(1)	(2)
	Non-industrial	Industrial
Time × Treat × Pollution	−0.012	0.296**
	(0.122)	(0.122)
Time × Treat	0.041	−0.089
	(0.059)	(0.078)
Time × Pollution	0.008	−0.029
	(0.047)	(0.071)
Treat × Pollution	−0.001	0.213
	(0.180)	(0.428)
_cons	−0.376	−6.088***
	(0.697)	(2.252)
Controls	Yes	Yes
Year FE	Yes	Yes
Id FE	Yes	Yes
N	9692	11089
R2	0.670	0.424

Note: Robust standard errors in parentheses, ***, **, * indicate significant at the 1 %, 5 %, and 10 % levels, respectively.

It is suggested that the explanation for this situation is as follows. First off, as industrial enterprises make up the bulk of heavy polluters, there may be a comparatively larger distribution density of strongly polluting enterprises within the industrial category. Second, green financing plays a crucial role in supporting the green transformation of these organizations, since the secondary industry in the industrial group is essential to economic growth. This group of enterprises also places a greater emphasis on this function.

5.2.2. Enterprise size

In this paper, total assets are utilized as a proxy to measure enterprise size, and enterprises are classified into small-scale and large-scale groups using the median of each year to group regressions on the benchmark regression. The regression results of enterprise size heterogeneity are displayed in Table 7. The coefficient in column (1) is not substantially positive, indicating that GFRIs did not work in the small-scale group. However, as can be seen in column (2), the coefficient proves positive, indicating that the innovation pilot zone is effective in the large-scale group.

Table 7.

Enterprise size heterogeneity.

	(1)	(2)
	Small scale	Large scale
Time × Treat × Pollution	0.005	0.208**
	(0.148)	(0.101)
Time × Treat	0.093	−0.093
	(0.066)	(0.072)
Time × Pollution	0.119*	−0.116**
	(0.064)	(0.054)
Treat × Pollution	0.020	0.297
	(0.118)	(0.543)
_cons	−1.399	−2.023*
	(0.988)	(1.127)
Controls	Yes	Yes
Year FE	Yes	Yes
Id FE	Yes	Yes
N	10598	10738
R2	0.367	0.526

Note: Robust standard errors in parentheses, ***, **, * indicate significant at the 1 %, 5 %, and 10 % levels, respectively.

The reasons are twofold. First, since environmental expenditures may not always result in better commercial performance, small-scale enterprise may find it challenging to justify them when faced with resource and budgetary limitations. Larger enterprises, on the other hand, are more likely to be able and willing to assume responsibility for environmental preservation and to cover the expenses associated with the capital and technology needed for such investments. Second, large-scale enterprises naturally benefit from easier financing terms offered by banks and financial institutions. This lowers their capital costs during the environmental investment process and enables large-scale businesses to more effectively utilize outside resources to meet their environmental objectives.

5.2.3. ESG management

The benchmark regressions are grouped into non-excellent and excellent groups by using the ESG rating of Sino-Securities Index Information Service (Shanghai) Co.Ltd as a proxy to measure enterprise ESG performance, and dividing the enterprises into non-excellent and excellent groups by whether they are excellent (i.e., grade A or above). Table 8 presents the regression findings for economic structural heterogeneity. In column (1), the coefficient is not significant, indicating the policy did not work in the non-excellent group. However, as shown in column (2), the coefficient is significantly positive, implying that in the excellent group, GFRIs is effective.

Table 8.

ESG management heterogeneity.

	(1)	(2)
	Non-excellent	Excellent
Time × Treat × Pollution	0.060	0.339*
	(0.110)	(0.177)
Time × Treat	0.009	−0.034
	(0.056)	(0.087)
Time × Pollution	−0.009	−0.033
	(0.057)	(0.067)
Treat × Pollution	0.171	0.094
	(0.350)	(0.277)
_cons	−1.549**	−2.683**
	(0.641)	(1.312)
Controls	Yes	Yes
Year FE	Yes	Yes
Id FE	Yes	Yes
N	11777	9168
R2	0.575	0.463

Note: Robust standard errors in parentheses, ***, **, * indicate significant at the 1 %, 5 %, and 10 % levels, respectively.

The reasons are also twofold. First of all, enterprises that demonstrate excellent ESG management tend to show higher sensitivity and enthusiasm in environmental protection. These enterprises are well aware of the importance of environmental protection for the long-term development of enterprises and therefore develop rigorous policies and measures to mitigate the potential impact of production activities on the environment. Second, excellent ESG management will send a green signal to enterprises and make financial institutions and banks more relaxed about their financing restrictions, which is a phenomenon of the increasing emphasis on green finance in the financial sector and the common pursuit of sustainable development goals. In this context, enterprises with excellent ESG management will have easier access to financial support, thus providing a strong financial guarantee for their environmental projects and green transformation.

6. Conclusions and recommendations

The global economic level has increased dramatically in sustainable development, yet many environmental problems caused by earlier extensive development model pose a threat to human survival. China, in particular, urgently needs to promote economic transformation through the guided policy of green financing. Using the eight pilot zones in five provinces for in 2017 as a quasi-natural experiment, this paper employs a DDD model to provide insights into whether GFRIs promote environmental investments by heavy polluters. The results of robustness tests, such as the PSM-DID test, the placebo test, controlling for the impact of other policies, and so on, show that the pilot zone has a positive effect on increasing environmental protection investment by heavy polluters. The results of mechanism testing show that by easing financial limitations and lowering financing costs, GFRIs encourages investments in the environment. Finally, heterogeneity studies for provinces with a greater percentage of secondary GDP, large-scale enterprises, and enterprises with outstanding ESG management show that the promotion effect of GFRIs on environmental investments of heavy polluters is particularly noticeable.

In light of the previously described findings, this paper makes the following recommendations. As for governments, first, in order to maximize the policy atmosphere and boost publicity, China should support the pilot zone policy even more. This can be achieved through a variety of channels, including promotion of GFRIs, which will strengthen China's position in the policy atmosphere. Furthermore, China should work to raise public awareness of environmental issues in the broader community. This will help to reinforce the benefits of green finance and motivate more enterprises to make direct investments in environmental protection, creating a positive feedback loop that starts with policy and ends with enterprise practices. Secondly, China should increase support for green finance business of financial institutions, strictly regulate environmental information disclosure of enterprises, and guide financial institutions to ease funding restrictions on efficient and environmentally friendly enterprises. Finally, when promoting GFRIs in China, it is important to avoid a “one-size-fits-all” policy orientation and to set more precise and reasonable development strategies for distinctive areas and enterprises to achieve optimal policy effects in all types of subjects and further promote the development of environmental investment.

For enterprises and managers, first of all, they should have a deep understanding of the relevant regulations of green finance policies, so as to adjust their strategies to obtain green incentives. Especially for heavy polluting enterprises, they need to pay high costs and investments to complete the green transformation. GFRIs provides them with financial support from the government, and they should seize the opportunity to drive sustainable development through environmentally friendly investments. Second, managers should improve their decision-making and management capabilities, and deeply learn the concepts and knowledge of green finance and environmental protection investment. Although the practice of green finance and environmental investment has been carried out in China for many years, many business managers do not really accept and understand it. GFRIs have promoted the green transformation of many heavy polluting enterprises, and if managers do not keep up with time, they will eventually be eliminated from the market. Third, they should develop a green management system and culture. The government, investors and other stakeholders also always pay attention to enterprise culture, green management system and culture are conducive to the construction of green image of enterprises, and further obtain the support of GFRIs.

For the public and investors, first, they should deepen the concept of green consumption and green investment, influence enterprise decisions through green actions, and further promote the implementation of GFRIs. The public and investors can provide support for enterprises’ environmental protection investment through consumption behavior and investment behavior. Market competition and resource allocation will accelerate the speed at which heavy polluting enterprises are eliminated, and enterprises with green transformation will be further supported by GFRIs. Second, they should exercise external oversight. The public and investors can ask enterprises to disclose more information about environmental investment to urge enterprises to make a green transition. For banks and other financial institutions, they should also supervise the use of funds from green credit and green bonds by enterprises.

The study in this paper still has certain limitations. First, a thorough investigation of the reducing effect of any particular kind of pollution is lacking. Furthermore, the execution and upholding of green finance policies could entail intricate viewpoints and procedures, including policy regulation, which are crucial avenues for further research. In the future, scholars can further study the impact of GFRIs on enterprises, such as emissions of waste gas, wastewater, solid waste, and equity financing costs. In addition, scholars can try to explore the synergies between GFRIs and other policies, such as government environmental protection inspections.

Data availability statement

Most of the data used in this study can be obtained from the China Stock Market and Accounting Research (CSMAR) database at https://data.csmar.com/.

If you need to repeat or further research on this topic, you can contact the corresponding author for full data access. The corresponding author's e-mail address is zcc19960508@foxmail.com.

CRediT authorship contribution statement

Zhao Cheng: Writing – original draft, Software, Data curation, Conceptualization. Chengcheng Zhu: Writing – review & editing, Supervision, Software, Methodology, Conceptualization.

Declaration of competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.