The Impact of Green Finance Policy on Environmental Performance: Evidence from China

Abstract

We investigate whether and how the policy of establishing green finance pilot zones affects corporate environmental performance in China, by employing the DID model and taking 2324 Chinese A-share listed companies as the empirical sample. The main findings show that the green finance policy can significantly improve corporate environmental performance in the green finance pilot zones. The policy effect varies according to enterprise ownership, sector, and degree of environmental supervision. In particular, compared with private enterprises and enterprises subject to key pollution monitoring, the environmental performance of state-owned firms and non-key pollution-monitored firms is more positively affected by the green finance policy. Through a mechanism analysis, we find that corporate innovation and financial constraints can play partially mediating roles in the linkage of green finance policy and corporate environmental performance. Among them, the mediating effects of green innovation and financial constraints are more prominent in private enterprises and key pollution-monitored enterprises. However, although the green finance policy can positively influence bank loans obtained by enterprises, there is no evidence to suggest that bank credit plays a significant mediating role between the green finance policy and corporate environmental performance. Our findings are helpful for understanding the effect of green finance policy on environmental sustainability and could provide some references for policymakers. In particular, we suggest that private and key pollution-monitored enterprises should actively respond to the green finance policy, broaden financing channels, and enhance capability of green innovation, thereby improving their environmental performance.

1. Introduction

Sustainable development is a hot topic around the world, and many countries have introduced different policies for driving people to pay attention to environmental, economic, and social sustainability. The increased focus on sustainable development has led to a shift in the financial landscape, incorporating sustainable practices within nations’ financial systems [1]. In the past decade, green finance, which refers to investing financial resources in environment-friendly products, policies, and strategies, has been regarded as an innovative solution to low-carbon economic growth [2]. For instance, China established the green finance innovation and reform pilot zones in the five selected provinces, namely, Zhejiang, Jiangxi, Guangdong, Guizhou, and Xinjiang in 2017, to encourage financial institutions to set up green finance divisions, and promote the issuance of green credit, green insurance, green bonds, etc., hoping that these measures can have a significant impact on sustainable development.

In recent years, the relationship between green finance and environmental sustainability has become increasingly emphasized in the economic research field. Some previous studies have tried to figure out the impact of green finance on environment sustainability from different perspectives. From the perspective of the research area, they focus on studying whether and how green finance affects sustainable development in different regions, such as developing countries [3], the European Union [4], the OECD economies [5], the BRICS economies [6], Africa [7], Indonesia [8], or Malaysia [9]. In terms of specific research content, these studies can be summarized as follows: On the one hand, some scholars are interested in the economic and environmental effects generated by green finance products, such as financial support for sustainable industrialization [10], green credit instruments [11], green bond instruments [12], green investments [13], and some compositive green finance indexes constructed from different green finance products [14]. The main findings of these studies show that green finance products can have a positive impact on sustainable development through different channels, such as decreasing CO₂ emissions [15], improving wastewater treatment [16], promoting corporate sustainability [17], improving green total factor productivity development, and increasing green GDP [18]. On the other hand, several previous studies focus on investigating the impact of green finance policy on environmental sustainability. As for China, there are mainly three kinds of green finance policy highlighted by the previous literature, that is, the Green Credit Guideline implemented in 2012 [19], the Green Financial System Guidelines implemented in 2016 [20], and the Green Finance Innovation and Reform Pilot Zones established in 2017 [21]. Although there has been some evidence that green finance has a significantly positive impact on sustainable development [22], the nexus between nature and finance is very complex and it is not yet totally clear how it works [23]. The impact of green finance policy on environmental sustainability remains a research topic that is highly worthy of attention.

The purpose of this study is to investigate the environmental impact of green finance policy, taking the policy of establishing green finance innovation and reform pilot zones (hereafter referred to as GFPZ) in 2017 as the research object in China. Our empirical sample covers 2324 Chinese A-share listed companies from 2012 to 2022. The difference-in-difference (DID) method is used in this study. The main empirical results are as follows: First, the baseline regression results provide evidence that the effect of the GFPZ policy on corporate environmental performance is positive, and this result is supported by multiple robustness tests (i.e., parallel trend test, placebo test, PSM-DID regression, and replacing dependent variables). Second, there is heterogeneity in the effect of the GFPZ policy on corporate environmental performance. Compared with private firms and key pollution-monitored firms, the impact of the GFPZ policy on environmental performance is stronger in state-owned firms and non-key pollution-monitored firms. Third, corporate innovation and financial constraints play partially positive mediating roles between the GFPZ policy and corporate environmental performance. In particular, the mediating roles of green innovation and financial constraints are more pronounced for private enterprises and enterprises subject to key pollution monitoring.

The contribution of this study can be described in three aspects. Firstly, different from the previous literature focusing on CO₂ emission [24] or ESG comprehensive performance [25], we focus on examining whether and how the green finance policy (GFPZ) can improve corporate environmental performance. Secondly, expanding upon the previous studies that focus on green finance and corporate innovation [26,27], we split the channels through which green finance policy affects corporate environmental performance into two parts, namely financial mechanisms and non-financial mechanisms. Among them, we take corporate innovation and green innovation as the non-financial mediating variables, while taking financial constraints and bank credit as the proxy variables for financial mediating channels. Thirdly, we explore the heterogeneity of the effect of green finance policy from multiple perspectives, including enterprise ownership, industry, and the degree of environmental regulation. In addition, we analyze the heterogeneity of the mechanism through which green finance policy affects corporate environmental performance, comparing state-owned enterprises with private enterprises, and key pollution-monitored enterprises with non-key pollution-monitored enterprises, which is another contribution of this study.

The rest of this paper is structured as follows: Section 2 reviews relevant studies and proposes the research hypotheses, Section 3 introduces the research methodology, Section 4 describes the sample data, Section 5 shows and discusses the empirical results, and Section 6 makes a conclusion.

2. Theoretical Analysis and Research Hypotheses

2.1. Green Finance Policy and Corporate Environmental Performance

The core of green finance theory lies in guiding social funds to green industries such as environmental protection and clean energy, promoting sustainable development through the allocation of financial resources [28]. As an innovative policy tool, the original intention of China in establishing the green finance innovation and reform pilot zones is to direct capital towards green and environmentally friendly industries, thereby promoting the optimization of economic structure and ecological improvement [29]. According to the resource-based theory, financial support from green finance may reduce the cost disadvantage of environmental protection projects faced by those enterprises with scarce resources [30]. Strengthening environmental impact assessment is urgent and necessary for promoting the collaboration between sustainability science and finance [31]. Although there is not yet a uniform result on the complex relationship between the policy of establishing green finance pilot zones (GFPZs) and environmental sustainability, some previous studies suggest that green finance has a positive impact on decreasing CO₂ emission [6], improving wastewater treatment [16], promoting corporate green investment [32], supporting sustainable renewable energy [33], and thereby promoting sustainable development [22]. Thus, based on the theoretical analysis above, we formulate the first research hypothesis of this study.

H1. 

The policy of establishing green finance pilot zones (GFPZs) has a positive impact on corporate environmental performance.

2.2. The Channels Between Green Finance Policy and Corporate Environmental Performance

Institutional theory holds that the external institutional environment would have an impact on corporate decision-making [34]. Regarding the pathway by which green finance policy influences the decision and behavior of enterprises in environmental protection, some previous studies have investigated it from different aspects. For example, some scholars argue that green finance policy may lead to corporate greenwashing behavior due to executive short-term orientation and easing financial constraints [35]. Some scholars suggest that green finance policy can reduce carbon emissions from polluting enterprises, by way of reallocating financial resources and promoting corporate environmental governance [21]. Some studies argue that corporate technological innovation, carbon emission efficiency, and economic agglomeration have a mediating effect on the relationship between green finance policy and green economic transformation [24]. Some other previous studies suggest that green innovation capabilities and financial constraints may be the mechanism by which the green finance policy affects business ESG participation [36]. In addition, some previous studies argue that green credit [11], green public finance [26], or green subsidies [27] can support corporate innovation [37]. Meanwhile, some scholars argue that corporate innovation is important for promoting environmental sustainability [7]. Summarily, the pathway between green finance policy and corporate environmental performance can be divided into financial and non-financial mechanisms. Regarding the non-financial channel through which the green finance policy affects corporate environmental performance, we propose the second hypothesis of this study based on the above theoretical analysis as follows:

H2a. 

Corporate innovation can have a partially mediating effect on the relationship between the green finance policy (GFPZ) and corporate environmental performance.

In addition, based on the above theoretical analysis, as for the financial channel through which the green finance policy affects corporate environmental performance, we propose two hypotheses as follows:

H2b. 

Financial constraints can have a partially mediating effect on the relationship between the green finance policy (GFPZ) and corporate environmental performance.

H2c. 

Bank credit can have a partially mediating effect on the relationship between the green finance policy (GFPZ) and corporate environmental performance.

3. Methodology

3.1. Difference-in-Difference (DID) Model

Referring to previous studies [19], we employ the difference-in-difference (DID) model to examine the effect of the green finance policy on corporate environmental performance, for testing the first research hypothesis mentioned in the subsection above. The traditional DID model and the classical DID model are, respectively, specified as follows:

$$E{P}_{it}=\alpha +\varphi Trea{t}_i+\phi Pos{t}_t+\beta Trea{t}_i\times Pos{t}_t+\gamma Control{s}_{it}+{\epsilon }_{it}$$

(1)

$$E{P}_{it}=\alpha +\beta Trea{t}_i\times Pos{t}_t+\gamma Control{s}_{it}+{\kappa }_i+{\lambda }_t+{\epsilon }_{it}$$

(2)

In the model equations above, $E{P}_{it}$ denotes the environment performance of firm $i$ at time $t$, and $Trea{t}_i\times Pos{t}_t$ denotes the policy of establishing green finance innovation and reform pilot zones (GFPZs) of China, which was proposed in 2017. $Trea{t}_i$ is a dummy variable, which is set to 1 if the firm $i$ is registered within the green finance pilot zones; otherwise, it is set to 0. $Pos{t}_t$ is a dummy variable, which is set to 1 for the year of 2017 and the years after 2017; otherwise, it is set to 0. ${\kappa }_i$ and ${\lambda }_t$ are individual-fixed effect and time-fixed effect, respectively. $Control{s}_{it}$ represents a set of control variables for firm $i$ at time $t$. We select five control variables based on previous research [37,38], including the growth rate of total assets ($AG$), the cash ratio of total assets ($Cash$), the total asset–liability ratio ($Lev$), the return on total assets ($ROA$), and the logarithm of total assets ($Size$). They are used to control the growth capacity, debt-paying ability, operational ability, profitability, and firm size of enterprises, respectively. $\alpha$ and ${\epsilon }_{it}$, respectively, represent constant and residual terms.

3.2. Mediating Effect Model

The mediating roles of corporate innovation, financial constraints, and bank credit in the relationship between the GFPZ policy and corporate environmental performance are tested by three methods, namely the stepwise regression method, the Sobel method, and the Bootstrap method. Firstly, the mediating effect is tested by the stepwise regression method as follows:

$$E{P}_{it}={\alpha }_0+{\beta }_{0}Trea{t}_i\times Pos{t}_t+{\gamma }_{0}Control{s}_{it}+{\epsilon }_{it}$$

(3)

$$M{R}_{it}={\alpha }_1+{\beta }_{1}Trea{t}_i\times Pos{t}_t+{\gamma }_{1}Control{s}_{it}+{\epsilon }_{it}$$

(4)

$$E{P}_{it}={\alpha }_2+{\beta }_{2}Trea{t}_i\times Pos{t}_t+\kappa M{R}_{it}+{\gamma }_{2}Control{s}_{it}+{\epsilon }_{it}$$

(5)

where $M{R}_{it}$ denotes the mediating variables, including corporate innovation ($Ino{v}_{it}$), green innovation ($GIno{v}_{it}$), financial constraints ($F{C}_{\mathrm{it}}$), and bank credit ($B{C}_{\mathrm{it}}$). Corporate innovation ($Ino{v}_{it}$) is measured by the logarithms of the total number of technology patents for firm $i$ at time $t$. Green innovation ($GIno{v}_{it}$) is measured by the logarithms of the number of green technology patents for firm $i$ at time $t$. We use the absolute value of the SA index proposed by Ref. [39] to measure financial constraints. Bank credit ($B{C}_{\mathrm{it}}$) is measured by the logarithm of the sum of long-term and short-term borrowings in the balance sheet of firm $i$ at time $t$. If the estimated coefficients ${\beta }_0$, ${\beta }_1$, and $\kappa$ are statistically significant while ${\beta }_2$ is not statistically significant, it demonstrates that $M{R}_{it}$ has a complete mediating effect. If the estimated coefficients ${\beta }_0$, ${\beta }_1$, $\kappa$, and ${\beta }_2$ are all significant, it suggests that $M{R}_{it}$ has only a partially mediating effect. Subsequently, the mediating roles are also re-tested by the Sobel method and the Bootstrap method.

4. Data and Descriptive Statistics

We collect the environment performance rating data of Chinese listed companies provided by the China Securities Index Company from the Wind database, spanning the period from 2012 to 2022. Owing to the fact that the original data is ranked across 10 grades from D to AAA, we manually quantify these ratings using numbers from 1 to 10. Financial data and patent data on Chinese A-share listed firms are all obtained from the CSMAR database. According to the Green Technology Patent Classification System (https://www.gov.cn/zhengce/zhengceku/202308/content_6901253.htm, accessed on 20 May 2024) issued by the State Intellectual Property Office of China and the International Patent Classification Table (https://www.cnipa.gov.cn/art/2023/5/26/art_3161_185369.html, accessed on 20 May 2024), we calculate the number of green technology patents. In addition, we exclude firms which are listed after the year of 2017 and those whose patent data is missing. Consequently, there are 25,564 observations in total generated by 2324 Chinese A-share listed firms in our final sample.

As reported in Panel B of

Table 1. Sample distribution.

	Obs.	Percent
Panel A: Total sample
	25,564	100
Panel B: Sample distribution categorized by sector
Consumer Discretionary (CD)	3806	14.89
Consumer Staples (CSs)	1529	5.98
Energy (EN)	660	2.58
Financials (FI)	715	2.80
Health Care (HC)	2178	8.52
Industrials (IN)	6358	24.87
Information Technology (IT)	3850	15.06
Materials (MA)	4301	16.82
Real Estate (RE)	1067	4.17
Telecommunication Services (TSs)	33	0.13
Utilities (UT)	1067	4.17
Panel C: Sample distribution categorized by province
Shanghai	1980	7.75
Yunnan	286	1.12
Neimeng	231	0.9
Beijing	2277	8.91
Jilin	374	1.46
Sichuan	891	3.49
Tianjin	418	1.64
Ningxia	121	0.47
Anhui	825	3.23
Shandong	1617	6.33
Shanxi	374	1.46
Guangdong	3652	14.29
Guangxi	308	1.2
Xinjiang	418	1.64
Jiangsu	2310	9.04
Jiangxi	451	1.76
Hebei	473	1.85
Henan	693	2.71
Zhejiang	2651	10.37
Hainan	242	0.95
Hubei	858	3.36
Hunan	770	3.01
Gansu	275	1.08
Fujian	913	3.57
Xizang	110	0.43
Guizhou	220	0.86
Liaoning	605	2.37
Chongqing	396	1.55
Shan-xi	407	1.59
Qinghai	99	0.39
Heilongjiang	319	1.25

Notes: This table reports the sample distribution of this study. The sample period is from 2012 to 2022, including 2324 Chinese A-share listed companies. The data in the table is collected, analyzed, and organized by the authors of this paper. The same note applies to the subsequent tables.

, according to the number of observations from high to low, the ten sectors are, respectively, Industrials (6358), Materials (4301), Information Technology (3850), Consumer Discretionary (3806), Health Care (2178), Consumer Staples (1529), Real Estate (1067), Utilities (1067), Financials (715), Energy (660), and Telecommunication Services (33). As seen from Panel C of Table 1, which shows the sample distribution grouped by province, there are 7392 observations in the green finance pilot zones, namely Guangdong (3652), Zhejiang (2651), Jiangxi (451), Guizhou (220), and Xinjiang (418).

5. Empirical Results

5.1. Baseline Regression Results

Table 2. Baseline regression results.

	(1)	(2)	(3)	(4)
Dependent Variables	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$
$Treat\times Post$	0.0795 ***	0.0714 **	0.0795 ***	0.0748 ***
	(2.628)	(2.445)	(3.451)	(3.231)
$Post$	0.2429 ***	0.1186 ***
	(15.034)	(7.492)
$Treat$	0.0219	0.0628 ***
	(1.017)	(2.974)
$AG$		−0.0067		0.0057
		(−0.915)		(1.040)
$Cash$		−0.0071 ***		−0.0008
		(−3.390)		(−0.332)
$Lev$		−0.0020		−0.0005
		(−0.706)		(−0.297)
$ROA$		0.0022		−0.0062 **
		(1.182)		(−2.101)
$Size$		0.1980 ***		0.1095 ***
		(36.875)		(9.344)
Constant	1.7936 ***	−2.6043 ***	1.9324 ***	−0.5425 **
	(156.073)	(−21.915)	(301.744)	(−2.064)
Obs.	25,564	25,121	25,564	25,121
Year FE	No	No	Yes	Yes
Firm FE	No	No	Yes	Yes
Adj.R2	0.0153	0.0786	0.4163	0.4033
F	134.0	230.7	11.91	19.61

Notes: This table reports the baseline regression result of the DID models. Column (1) shows the estimation results of the traditional DID model without control variables. Column (2) shows the estimation results of the traditional DID model with control variables. Column (3) shows the estimation results of the classical DID model without control variables. Column (4) shows the estimation result of the classical DID model with control variables. There are robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05.

reports the estimation results of the DID models for examining the policy effect of establishing green finance innovation and reform pilot zones (GFPZs) in China. We firstly estimate the traditional DID model without control variables, and the estimated parameters are shown in column (1) of Table 2.

The estimated parameter for $Treat\times Post$ is 0.0795, which is statistically significant at the 1% level, indicating the GFPZ policy has a positively significant impact on corporate environmental performance in the pilot zones. Secondly, we add five control variables into the traditional DID model, and the estimation results are reported in column (2) of Table 2. When the control variables are added, the estimated parameter for $Treat\times Post$ is 0.0714, which is statistically significant at the 5% level. Finally, we also employ the classical DID model in which the fixed effects of both time and firm are controlled to examine the policy effect. As shown in columns (3) and (4) of Table 2, the estimated parameters for $Treat\times Post$ are, respectively, 0.0795 and 0.0748, which are also statistically significant at the 1% level. It highlights that establishing green finance pilot zones in China has a significantly positive policy effect on corporate environmental performance. In other words, after the establishment of green finance pilot zones in 2017, Chinese enterprises in the pilot zones significantly improved their environmental performance, and this improvement is partially attributable to the GFPZ policy. The first research hypothesis (H1) proposed in this study can hold true.

5.2. Robustness Check

5.2.1. Parallel Trend Test

The sample conforms to the parallel trend, which is the prerequisite for using the DID model.

Table 3. Results of parallel trend test.

Dependent Variables	(1)	(2)
	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$
$Treat\times pr{e}_2$	−0.0017	0.0134
	(−0.051)	(0.385)
$Treat\times pr{e}_1$	0.0143	0.0289
	(0.449)	(0.901)
$Treat\times Current$	−0.0068	−0.0122
	(−0.193)	(−0.345)
$Treat\times Pos{t}_1$	0.0360	0.0397
	(0.805)	(0.881)
$Treat\times Pos{t}_2$	0.1207 **	0.1268 **
	(1.980)	(2.064)
$Treat\times Pos{t}_3$	0.1126 *	0.1256 **
	(1.790)	(1.988)
$AG$		0.0057
		(0.986)
$Cash$		−0.0009
		(−0.244)
$Lev$		−0.0007
		(−0.374)
$ROA$		−0.0063 ***
		(−3.829)
$Size$		0.1099 ***
		(5.308)
Constant	1.7789 ***	−0.6425
	(94.132)	(−1.416)
Obs.	25,564	25,121
Number of firms	2324	2286
Adj.R2	0.0346	0.0375
F	37.17	32.29

Notes: This table reports the results of the parallel trend test. $Treat\times pr{e}_2$ and $Treat\times pr{e}_1$ are the cross-multiplicated variables before the implementation of the GFPZ policy, in which $pr{e}_2$ and $pr{e}_1$ are dummy variables representing two years and one year before the green policy, respectively. $Treat\times Current$ is the cross-multiplicated variable in the current year of the GFPZ policy. Similarly, $Treat\times Pos{t}_1$, $Treat\times Pos{t}_2$, and $Treat\times Pos{t}_3$ are cross-multiplicated variables one year, two, and three years after the implementation of the GFPZ policy, respectively. There are robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

reports the results of the parallel trend test. As seen in this table, the estimated parameters for the cross-multiplicated variables before the implementation of the GFPZ policy ($Treat\times pr{e}_2$ and $Treat\times pr{e}_1$) are all not statistically significant. The estimated parameters for the cross-multiplicated variables in the current year for the implementation of the GFPZ policy ($Treat\times Current$) are negative but not statistically significant, while the corresponding estimated parameters after the implementation of the GFPZ policy ($Treat\times Pos{t}_1$, $Treat\times Pos{t}_2$, and $Treat\times Pos{t}_3$) are all positive. In addition, the estimated parameters for $Treat\times Pos{t}_2$ and $Treat\times Pos{t}_3$ are, respectively, 0.1268 and 0.1256, which are statistically significant at the 5% level. It indicates that the empirical sample conforms to the parallel trend assumption, suggesting that Chinese corporate environmental performance began to be significantly improved in the second year after the implementation of the GFPZ policy.

5.2.2. Placebo Test

Next, we conduct a placebo test to examine whether the baseline regression results change due to the impact of unobservable variables such as other policies. Specifically, we randomly select the interaction term ($Treat\times Post$) for 500 times and re-estimate the baseline models for examining whether the coefficients differ significantly from the baseline estimation results. Figure 1 depicts the distribution of the estimated coefficients and that of the corresponding t-values for the placebo test. As seen from the left subfigure of Figure 1, the estimated coefficients are mostly around zero and significantly different from the baseline result marked by the vertically red dashed line. The corresponding t-values are mostly distributed around 0 and also significantly different from the baseline result, which is seen from the right subfigure of Figure 1. It indicates that in the case of random sampling, the baseline estimation coefficient of 0.0748 is a low-probability event. This suggests that our placebo test holds true; that is, there is no bias in the positive impact of the GFPZ policy on corporate environmental performance.

5.2.3. PSM-DID Model

For alleviating the endogeneity problems caused by selection bias, the PSM-DID model is used to re-estimate the effect of the GFPZ policy on corporate environmental performance. The propensity score matching method can make the treatment group and control group firms as similar as possible in all selected aspects, which are measured by the five control variables, namely, the growth rate of total assets, the cash ratio of total assets, the total asset–liability ratio, the return on total assets, and firm size. We employ seven matching methods to re-draw the empirical samples, including the nearest neighbor 1:1 matching method, the spline matching method, the local linear regression matching method, the kernel matching method, the radius matching method, the caliper matching method, and the martingale matching method, of which the estimation results are, respectively, shown in the columns from (1) to (7) of

Table 4. Results of the PSM-DID model.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Variables	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$
$Treat\times Post$	0.1665 ***	0.1480 ***	0.1480 ***	0.1389 ***	0.1389 ***	0.1665 ***	0.1408 ***
	(4.449)	(4.345)	(4.349)	(4.144)	(4.139)	(4.449)	(4.105)
$AG$	−0.0353	−0.0512 **	−0.1149 ***	−0.1291 ***	−0.0590 **	−0.0353	−0.0039
	(−1.495)	(−2.153)	(−3.345)	(−4.027)	(−2.354)	(−1.495)	(−0.539)
$Cash$	−0.0099	0.0102	0.0090	0.0087	0.0099	−0.0099	0.0086
	(−0.749)	(0.786)	(0.725)	(0.722)	(0.790)	(−0.749)	(0.667)
$Lev$	−0.1539 *	−0.0127	−0.0407	−0.0513	−0.0247	−0.1539 *	−0.0305
	(−1.775)	(−0.154)	(−0.531)	(−0.697)	(−0.315)	(−1.775)	(−0.370)
$ROA$	−0.0363	−0.0420	−0.0225	−0.0319	−0.0508	−0.0363	−0.0504
	(−0.572)	(−0.756)	(−0.422)	(−0.598)	(−0.897)	(−0.572)	(−1.381)
$Size$	0.2395 ***	0.2347 ***	0.2383 ***	0.2484 ***	0.2450 ***	0.2395 ***	0.2376 ***
	(14.011)	(16.231)	(16.516)	(18.999)	(18.688)	(14.011)	(16.243)
Constant	−3.4893 ***	−3.4135 ***	−3.4713 ***	−3.6887 ***	−3.6338 ***	−3.4893 ***	−2.8546 ***
	(−9.227)	(−10.803)	(−10.974)	(−12.863)	(−12.681)	(−9.227)	(−3.887)
Obs.	6090	11,379	11,370	13,670	13,679	6090	11,260
R2	0.0952	0.0889	0.0894	0.0993	0.0987	0.0952	0.0919
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes

Notes: This table reports the regression result of the PSM-DID models, specified by seven matching methods, namely, the nearest neighbor 1:1 matching, the spline matching, the local linear regression matching, the kernel matching, the radius matching, the caliper matching, and the martingale matching. The estimation results of these seven matching methods are reported in columns (1), (2), (3), (4), (5), (6), and (7). There are robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

. As shown in Table 4, all the estimated coefficients of $Treat\times Post$ are positive and significant at the 1% level. It implies that the environmental performance of Chinese A-share listed companies in the green finance innovation and reform pilot zones is positively improved by the GFPZ policy. The regression results obtained in the baseline model are reliable.

5.2.4. Changing the Dependent Variable

In order to alleviate the endogeneity problem caused by measurement errors, we change the method of environmental scoring, by transforming the original environmental performance scores into logarithms.

Table 5. Regression results for the alternative dependent variable.

	(1)	(2)	(3)	(4)
Dependent Variables	$\mathit{ln}\mathit{E}\mathit{P}$	$\mathit{ln}\mathit{E}\mathit{P}$	$\mathit{ln}\mathit{E}\mathit{P}$	$\mathit{ln}\mathit{E}\mathit{P}$
$Treat\times Post$	0.0269 ***	0.0234 ***	0.0269 ***	0.0246 ***
	(2.888)	(2.590)	(3.765)	(3.411)
$Post$	0.0817 ***	0.0453 ***
	(16.383)	(9.159)
$Treat$	0.0068	0.0197 ***
	(1.004)	(2.944)
$AG$		−0.0024		0.0019
		(−0.990)		(1.114)
$Cash$		−0.0027 ***		−0.0004
		(−3.762)		(−0.445)
$Lev$		−0.0010		−0.0005
		(−1.056)		(−0.756)
$ROA$		0.0004		−0.0023 **
		(0.692)		(−2.339)
$Size$		0.0601 ***		0.0315 ***
		(38.793)		(8.645)
Constant	0.9684 ***	−0.3669 ***	1.0149 ***	0.3026 ***
	(266.356)	(−10.672)	(511.760)	(3.698)
Obs.	25,564	25,121	25,564	25,121
Adj.R2	0.0183	0.0801	0.4110	0.4003
Year FE	No	No	Yes	Yes
Firm FE	No	No	Yes	Yes
F	160.2	272.2	14.18	18.03

Notes: This table reports the regression results of the DID models with the alternative dependent variable. Robust t-statistics are in parentheses. *** p < 0.01, ** p < 0.05.

reports the re-regression results for the alternative dependent variable. As seen from this table, whether employing the traditional DID model or the classical DID model, the estimated coefficients of $Treat\times Post$ are positive and statistically significant at the 1% level. And the main finding does not change regardless of whether we include control variables in the model or not. It once again demonstrates that the baseline regression results are robust. The first research hypothesis of this study can still hold; that is, the green finance policy has a positive effect on corporate environmental performance.

5.3. Heterogeneity Analysis

We split the total sample into two groups based on enterprise ownership to re-run the regressions, for examining whether there is a difference between private firms and state-owned firms. As seen from

Table 6. Comparing results of state-owned firms and private firms.

Variables	(1)	(2)
	SOE	POE
$Treat\times Post$	0.1119 ***	0.0631 **
	(3.216)	(2.027)
$AG$	0.0027	0.0071
	(0.268)	(1.148)
$Cash$	0.0036	−0.0032
	(0.500)	(−1.229)
$Lev$	−0.0587	0.0011
	(−1.018)	(0.913)
$ROA$	−0.0423	−0.0055 *
	(−0.876)	(−1.865)
$Size$	0.0942 ***	0.1289 ***
	(5.491)	(7.861)
Constant	−0.1599	−0.9809 ***
	(−0.411)	(−2.713)
Obs.	12,889	12,232
Number of firms	1173	1113
Year FE	Yes	Yes
Firm FE	Yes	Yes
Adj.R2	0.4322	0.3738
F	6.955	14.68

Notes: This table reports the result of the impact of the GFPZ policy on the environmental performance of state-owned firms (SOE) compared with private firms (POE). There are robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

, the estimated parameter for $Treat\times Post$ is 0.1119 and statistically significant at the 1% level in the subsample involving state-owned firms, while it is 0.0631 and statistically significant at the 5% level in the subsample involving private firms. Obviously, this suggests that the GFPZ policy has a stronger effect on improving the environmental performance of state-owned enterprises, compared with private firms. This might be because state-owned enterprises are more proactive in responding to national policies and have more advantages in obtaining green financial supports such as the green credit.

We continue to compare the effects of the GFPZ policy on environmental performance in different sectors. As reported in

Table 7. Comparison of results for different sectors.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	CD	CSs	EN	FI	HC	IN	IT	MA	RE
$Treat\times Post$	−0.1251 **	−0.1022	−0.0988	0.1737	0.1091	0.0145	0.2807 ***	0.0243	0.3126 ***
	(−2.249)	(−1.373)	(−0.981)	(0.718)	(1.449)	(0.307)	(4.645)	(0.429)	(3.991)
$AG$	−0.0190 *	−0.0382	0.0167	0.0301 ***	0.0137	−0.0328 *	0.0111	0.0104 ***	−0.1175 **
	(−1.761)	(−1.255)	(0.114)	(4.395)	(0.287)	(−1.850)	(1.098)	(3.513)	(−2.258)
$Cash$	0.0028	0.0088	−0.0076	−0.0164	−0.0210 **	0.0196	0.0011	0.0108	−0.0371
	(0.332)	(1.058)	(−1.073)	(−1.282)	(−2.496)	(1.449)	(0.329)	(1.092)	(−0.753)
$Lev$	−0.1136 **	−0.3046 **	0.8366 ***	0.4036	−0.2083	0.0029 **	0.0637	0.0273	0.0127
	(−2.079)	(−2.091)	(3.875)	(1.233)	(−1.605)	(2.166)	(0.947)	(0.486)	(0.207)
$ROA$	−0.0393	−0.5293 **	0.0487	0.6167	−0.0999	−0.0040	0.3758 **	0.0241	−0.2200 *
	(−1.340)	(−2.455)	(1.204)	(1.400)	(−0.574)	(−1.242)	(2.393)	(0.448)	(−1.909)
$Size$	0.1235 ***	0.2077 ***	0.0537	−0.0723	0.2623 ***	0.1470 ***	0.0623 *	0.1269 ***	0.1765 ***
	(4.914)	(4.729)	(1.045)	(−1.284)	(5.247)	(5.773)	(1.891)	(4.597)	(4.914)
Constant	−0.7674	−2.6078 ***	0.2653	3.4725 ***	−4.1796 ***	−1.4009 **	0.5020	−0.7994	−2.0102 **
	(−1.376)	(−2.727)	(0.226)	(2.740)	(−3.814)	(−2.448)	(0.690)	(−1.286)	(−2.337)
Obs.	3806	1529	660	277	2178	6358	3845	4301	1067
Number of firms	346	139	60	27	198	578	350	391	97
Adj.R2	0.4885	0.3967	0.4593	0.3146	0.3874	0.4172	0.3711	0.4365	0.6783
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
F	5.531	4.406	4.246	4.407	5.913	8.525	5.670	10.01	7.472

Notes: This table reports the results of the impact of the GFPZ policy on corporate environmental performance in different sectors, including Consumer Discretionary (CD), Consumer Staples (CSs), Energy (EN), Financials (FI), Health Care (HC), Industrials (IN), Information Technology (IT), Materials (MA), and Real Estate (RE). What should be noted is that the sample size of listed companies in the two sectors (i.e., telecommunications services and utilities) is too small, so we have disregarded them here. There are robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

, the GFPZ policy positively affects the environmental performance in six industries, including Financials (FI), Health Care (HC), Industrials (IN), Information Technology (IT), Materials (MA), and Real Estate (RE). Among them, the estimated coefficients of $Treat\times Post$ in information technology and real estate are statistically significant at the 1% level. Interestingly, we do not find the positive effect of the GFPZ policy in three sectors, namely Consumer Discretionary (CD), Consumer Staples (CSs), and Energy (EN).

As far as we know, in order to strengthen the supervision and management of environment, China designates a portion of enterprises as key pollution-monitored units based on the pollution situation of a company’s business operations every year. Thus, we further examine whether the effect of the GFPZ policy is different between key pollution-monitored firms and non-key pollution-monitored firms. The results of the comparison are shown in

Table 8. Comparison of results of key pollution-monitored firms and non-key pollution-monitored firms.

Variables	(1)	(2)
	KPMs	NON-KPMs
$Treat\times Post$	−0.0376	0.1549 ***
	(−1.078)	(5.042)
$AG$	0.0115 *	0.0057
	(1.809)	(0.745)
$Cash$	0.0012	−0.0010
	(0.165)	(−0.404)
$Lev$	0.0038 ***	−0.0909 **
	(3.601)	(−2.420)
$ROA$	−0.0033	−0.0325
	(−1.055)	(−1.230)
$Size$	0.1711 ***	0.0710 ***
	(9.264)	(4.609)
Constant	−1.8753 ***	0.2862
	(−4.462)	(0.842)
Obs.	11,792	13,329
Adj.R2	0.4322	0.3738
Year FE	Yes	Yes
Firm FE	Yes	Yes
F	17.60	8.552

Notes: This table reports the results of the impact of the GFPZ policy on the corporate environmental performance of the key pollution-monitored firms (KPMs) compared with the non-key pollution-monitored firms (NON-KPMs). There are robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

. There is evidence that the estimated coefficient of $Treat\times Post$ in the group of non-key pollution-monitored firms is 0.1549 and has a 1% significance level. But the estimated coefficient of $Treat\times Post$ in the group of key pollution-monitored firms is negative and not statistically significant. It implies that the positive impact of the green finance policy on corporate environmental performance is mainly reflected in non-key pollution-monitored enterprises, while its positive effect on key pollution-monitored enterprises is very limited or even negative.

5.4. Mechanism Analysis

5.4.1. Corporate Innovation

We continue to explore the role of corporate innovation in the relationship between the GFPZ policy and environmental performance. Firstly, we test whether corporate innovation can serve as a non-financial channel through which the GFPZ policy affects corporate environmental performance. The estimation results of the mediating effect model are reported in

Table 9. The mediating effect of corporate innovation tested by the stepwise regression method.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)
	Total Sample			Enterprise Ownership						Degree of Environmental Supervision
				SOE			POE			KPMs			NON-KPMs
	$\mathit{E}\mathit{P}$	Inov	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	Inov	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	Inov	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	Inov	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	Inov	$\mathit{E}\mathit{P}$
$Treat\times Post$	0.1342 ***	0.2728 ***	0.1142 ***	0.1114 ***	0.0948 *	0.1054 ***	0.1548 ***	0.3922 ***	0.1224 ***	0.1445 ***	0.3485 ***	0.1156 ***	0.1257 ***	0.2692 ***	0.1087 ***
	(6.668)	(8.367)	(5.699)	(3.619)	(1.843)	(3.433)	(5.768)	(9.358)	(4.581)	(4.700)	(8.611)	(3.803)	(4.747)	(5.748)	(4.094)
Inov			0.0730 ***			0.0633 ***			0.0828 ***			0.0831 ***			0.0631 ***
			(18.859)			(11.952)			(14.570)			(13.126)			(12.257)
$AG$	−0.0038	−0.0077	−0.0033	−0.0129	−0.0178	−0.0118	0.0019	−0.0024	0.0021	0.0070	−0.0047	0.0074	−0.0059	0.0114	−0.0067
	(−0.542)	(−0.736)	(−0.447)	(−1.007)	(−1.420)	(−0.920)	(0.253)	(−0.165)	(0.253)	(0.936)	(−0.238)	(0.844)	(−0.601)	(0.792)	(−0.671)
$Cash$	−0.0081 ***	−0.0087 *	−0.0075 ***	−0.0090	−0.1099 ***	−0.0020	−0.0122 ***	−0.0042	−0.0118 ***	−0.0164 ***	−0.0304 ***	−0.0139 **	−0.0075 ***	−0.0299 ***	−0.0056 ***
	(−3.880)	(−1.722)	(−3.629)	(−1.507)	(−6.380)	(−0.337)	(−4.724)	(−0.905)	(−4.752)	(−2.825)	(−3.317)	(−2.380)	(−3.425)	(−2.879)	(−2.661)
$Lev$	−0.0026	−0.0346	−0.0001	−0.1412 ***	−0.9589 ***	−0.0805 *	−0.0005	−0.0221	0.0013	0.0006	−0.0043	0.0010	−0.0887 ***	−1.0346 ***	−0.0234
	(−0.779)	(−1.099)	(−0.084)	(−2.933)	(−5.941)	(−1.823)	(−0.407)	(−1.161)	(1.249)	(0.407)	(−0.588)	(0.914)	(−2.712)	(−5.652)	(−0.786)
$ROA$	0.0013	0.0088	0.0006	0.0018	−0.0276	0.0036	0.0007	0.0098	−0.0002	0.0018	0.0128 ***	0.0007	−0.0106	−0.0504	−0.0074
	(0.659)	(1.587)	(0.318)	(0.030)	(−0.279)	(0.058)	(0.415)	(1.616)	(−0.085)	(1.240)	(4.225)	(0.478)	(−0.269)	(−0.525)	(−0.188)
$Size$	0.1981 ***	0.5160 ***	0.1605 ***	0.2171 ***	0.5940 ***	0.1795 ***	0.1883 ***	0.4944 ***	0.1474 ***	0.2250 ***	0.6744 ***	0.1690 ***	0.1715 ***	0.3446 ***	0.1498 ***
	(36.945)	(54.741)	(29.245)	(29.049)	(40.175)	(23.200)	(21.120)	(32.425)	(16.734)	(26.674)	(57.864)	(18.703)	(23.041)	(19.550)	(20.401)
Constant	−2.5481 ***	−9.0836 ***	−1.8845 ***	−2.9334 ***	−10.4098 ***	−2.2741 ***	−2.2918 ***	−8.4691 ***	−1.5909 ***	−3.0865 ***	−12.3481 ***	−2.0603 ***	−1.9847 ***	−5.1478 ***	−1.6597 ***
	(−21.454)	(−44.252)	(−15.725)	(−18.082)	(−35.454)	(−13.724)	(−11.873)	(−25.904)	(−8.394)	(−16.325)	(−47.367)	(−10.467)	(−12.429)	(−15.176)	(−10.541)
Obs.	25,121	25,121	25,121	12,889	12,889	12,889	12,232	12,232	12,232	11,792	11,792	11,792	13,329	13,329	13,329
Adj.R2	0.0818	0.1845	0.0955	0.0971	0.2127	0.1078	0.0625	0.1632	0.0786	0.0846	0.3015	0.0977	0.0755	0.1048	0.0870
F	119.9	301.9	131.9	73.65	184.5	75.64	48.67	130.5	56.21	58.28	308.9	64.79	62.59	70.60	66.14

Notes: This table reports the estimation results of the mediating effect model specified in Equations (3)–(5), for testing the mediating effect of corporate innovation (Inov) between the GFPZ policy and corporate environmental performance. Columns (1)–(3) display the empirical results of the mediating effect model for the total sample. Columns (4)–(9) display the corresponding results for the subsamples grouped by enterprise ownership, namely state-owned firms (SOE) and private firms (POE). Columns (10)–(15) display the empirical results for the subsamples grouped by the degree of environmental supervision, namely key pollution-monitored firms (KPMs) and non-key pollution-monitored firms (NON-KPMs). There are robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

. In the total sample, when taking the corporate innovation ($Inov$) as the dependent variable, the estimated coefficient of $Treat\times Post$ is 0.2728, which is statistically significant at the 1% level. When both the GFPZ policy and corporate innovation are involved in the model in which corporate environmental performance is taken as the dependent variable, the estimated coefficients of $Treat\times Post$ and $Inov$ are, respectively, 0.1142 and 0.0730, which are both statistically significant at the 1% level. It indicates that the GFPZ policy can improve corporate innovation in the pilot zones, and corporate innovation ($Inov$) plays a partially mediating role between the GFPZ policy and environmental performance. Thus, the second research hypothesis (H2a) proposed in this study can hold true.

We also divide the total sample into different groups according to enterprise ownership and the degree of environmental supervision, for examining whether there is heterogeneity in the mediating role of corporate innovation. As reported in the columns from (4) to (9) of Table 9, the estimated positive impact of the green finance policy on corporate innovation for private firms (POE) is 0.3922, which is larger than that for state-owned firms (SOE). Whether it is for state-owned enterprises or private enterprises, corporate innovation has played a certain mediating role between the GFPZ policy and corporate environmental performance. As seen in the columns from (10) to (15) of Table 9, using similar analysis methods, we can find that the green finance policy has a more significant impact on enhancing the innovation capabilities of key pollution-monitored firms, compared with non-key pollution-monitored firms. The partially mediating role of corporate innovation between the green finance policy and corporate environmental performance exists in both key and non-key pollution-monitored firms.

In particular, we further explore whether only green innovation, which is included in corporate innovation, plays a mediating role between the green finance policy and corporate environmental performance. As shown in

Table 10. The mediating effect of green innovation tested by the stepwise regression method.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)
	Total Sample			Enterprise Ownership						Degree of Environmental Supervision
				SOE			POE			KPMs			NON-KPMs
	$\mathit{E}\mathit{P}$	GInov	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	GInov	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	GInov	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	GInov	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	GInov	$\mathit{E}\mathit{P}$
$Treat\times Post$	0.1342 ***	−0.0162	0.1343 ***	0.1114 ***	−0.1395 ***	0.1069 ***	0.1548 ***	0.1151 ***	0.1478 ***	0.1445 ***	0.0771 **	0.1421 ***	0.1257 ***	−0.0517	0.1216 ***
	(6.668)	(−0.594)	(6.118)	(3.619)	(−3.232)	(3.194)	(5.768)	(3.252)	(5.022)	(4.700)	(2.114)	(4.476)	(4.747)	(−1.368)	(3.991)
GInov			0.0858 ***			0.0658 ***			0.1103 ***			0.0868 ***			0.0871 ***
			(13.762)			(7.697)			(12.080)			(9.835)			(9.400)
$AG$	−0.0038	−0.0264 **	0.0029	−0.0129	−0.0358 ***	−0.0005	0.0019	−0.0183	0.0032	0.0070	−0.0191	0.0079	−0.0059	−0.0132	−0.0011
	(−0.542)	(−2.095)	(0.397)	(−1.007)	(−3.205)	(−0.039)	(0.253)	(−1.174)	(0.373)	(0.936)	(−0.852)	(0.932)	(−0.601)	(−1.181)	(−0.090)
$Cash$	−0.0081 ***	−0.0215 ***	−0.0131 ***	−0.0090	−0.0757 ***	−0.0117	−0.0122 ***	−0.0166 ***	−0.0133 ***	−0.0164 ***	−0.0344 ***	−0.0249 ***	−0.0075 ***	−0.0185 ***	−0.0104 ***
	(−3.880)	(−4.323)	(−4.511)	(−1.507)	(−5.775)	(−1.529)	(−4.724)	(−3.987)	(−4.209)	(−2.825)	(−3.878)	(−3.818)	(−3.425)	(−3.752)	(−3.774)
$Lev$	−0.0026	−0.0980 *	−0.1333 ***	−0.1412 ***	−0.2051 **	−0.1664 ***	−0.0005	−0.0582	−0.0388	0.0006	−0.2161 ***	−0.1287 **	−0.0887 ***	0.0735	−0.1649 ***
	(−0.779)	(−1.846)	(−3.184)	(−2.933)	(−2.481)	(−2.628)	(−0.407)	(−0.788)	(−0.661)	(0.407)	(−2.886)	(−2.144)	(−2.712)	(1.062)	(−2.750)
$ROA$	0.0013	0.5266 ***	−0.0751	0.0018	0.4730 ***	0.0002	0.0007	0.6022 ***	−0.1140	0.0018	0.3522 ***	−0.0594	−0.0106	0.7299 ***	−0.0859
	(0.659)	(5.976)	(−1.049)	(0.030)	(2.823)	(0.002)	(0.415)	(4.431)	(−1.337)	(1.240)	(3.456)	(−0.807)	(−0.269)	(6.250)	(−0.659)
$Size$	0.1981 ***	0.4755 ***	0.1815 ***	0.2171 ***	0.5194 ***	0.1958 ***	0.1883 ***	0.3777 ***	0.1768 ***	0.2250 ***	0.6134 ***	0.1818 ***	0.1715 ***	0.2562 ***	0.1831 ***
	(36.945)	(49.101)	(24.178)	(29.049)	(41.192)	(19.865)	(21.120)	(25.047)	(14.660)	(26.674)	(50.878)	(16.610)	(23.041)	(17.833)	(16.939)
Constant	−2.5481 ***	−9.1577 ***	−2.1333 ***	−2.9334 ***	−10.0575 ***	−2.4582 ***	−2.2918 ***	−7.0677 ***	−2.0437 ***	−3.0865 ***	−12.1067 ***	−2.1300 ***	−1.9847 ***	−4.5741 ***	−2.1666 ***
	(−21.454)	(−45.020)	(−13.579)	(−18.082)	(−37.510)	(−11.832)	(−11.873)	(−22.505)	(−8.096)	(−16.325)	(−47.652)	(−9.217)	(−12.429)	(−15.255)	(−9.666)
Obs.	25,121	20,283	20,283	12,889	10,241	10,241	12,232	10,042	10,042	11,792	10,557	10,557	13,329	9726	9726
Adj.R2	0.0818	0.2216	0.0890	0.0971	0.2633	0.0977	0.0625	0.1393	0.0770	0.0846	0.3186	0.0902	0.0755	0.0961	0.0792
F	119.9	250.3	89.58	73.65	168.5	52.30	48.67	82.80	38.62	58.28	255.2	47.80	62.59	49.09	39.08

Notes: This table reports the estimation results of the mediating effect model specified in Equations (3)–(5), for testing the mediating action of green innovation (GInov) between the GFPZ policy and environmental performance. Columns (1)–(3) display the empirical results of the mediating effect model for the total sample. Columns (4)–(9) display the corresponding results for the subsamples grouped by enterprise ownership, namely state-owned firms (SOE) and private firms (POE). Columns (10)–(15) display the empirical results for the subsamples grouped by the degree of environmental supervision, namely key pollution-monitored firms (KPMs) and non-key pollution-monitored firms (NON-KPMs). There are robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

, the estimated coefficient of $Treat\times Post$ for $GInov$ is −0.0162 but not statistically significant in the total sample. When both $Treat\times Post$ and $GInov$ are considered in one model, the estimated coefficient of $Treat\times Post$ and $GInov$ for $EP$ are, respectively, 0.1343 and 0.0858, both of which are statistically significant at the 1% level. It seems that the connection between the GFPZ policy and green innovation can enhance corporate environmental performance, but the mediating role of green innovation is not obvious. When the total sample is divided into two groups by enterprise ownership, the empirical results show that the green finance policy significantly and positively affects the green innovation of private firms, but the corresponding impact on the green innovation capability of state-owned firms is significantly negative. Similarly, when the total sample is divided into two groups by the degree of environmental supervision, there is evidence that the green finance policy has a significantly positive impact on the green innovation of key pollution-monitored enterprises, but the impact on the green innovation of non-key pollution-monitored enterprises is negative and not significant.

To further clarify the mediating roles of corporate innovation between the GFPZ policy and corporate environmental performance, the Sobel method and the Bootstrap method are adopted to measure the degree of the mediating effect. As shown in Panel A of

Table 11. The mediating effect of corporate innovation and green innovation tested by the Sobel method and the Bootstrap method.

	GFPZ → Inov → EP					GFPZ→ GInov → EP
	Total Sample	Enterprise Ownership		Degree of Environmental Supervision		Total Sample	Enterprise Ownership		Degree of Environmental Supervision
		SOE	POE	KPMs	NON-KPMs		SOE	POE	KPMs	NON-KPMs
Panel A: The mediating role tested by the Sobel method
Total effect	0.187 ***	0.183 ***	0.181 ***	0.151 ***	0.221 ***	0.142 ***	0.132 ***	0.141 ***	0.140 ***	0.147 ***
	(10.329)	(6.541)	(7.595)	(5.445)	(9.256)	(7.080)	(4.252)	(5.299)	(4.836)	(5.257)
Indirect effect	0.033 ***	0.020 ***	0.040 ***	0.040 ***	0.032 ***	0.006 ***	−0.002	0.019 ***	0.014 ***	0.004
	(11.863)	(5.753)	(9.713)	(8.840)	(8.714)	(2.796)	(−0.731)	(5.268)	(4.623)	(1.265)
Direct effect	0.154 ***	0.163 ***	0.141 ***	0.111 ***	0.189 ***	0.137 ***	0.134 ***	0.122 ***	0.126 ***	0.144 ***
	(8.536)	(5.845)	(5.904)	(4.012)	(7.945)	(6.829)	(4.324)	(4.615)	(4.357)	(5.153)
Ratio of indirect effect	17.65%	10.93%	22.10%	26.49%	14.48%	4.23%	−1.52%	13.48%	10.00%	2.72%
Panel B: The mediating role tested by the Bootstrap method
Indirect effect	0.0329 ***	0.0202 ***	0.0404 ***	0.0396 ***	0.0316 ***	0.0057 ***	−0.0018	0.0189 ***	0.0143 ***	0.0037
	(11.590)	(5.460)	(9.280)	(9.130)	(7.990)	(2.800)	(−0.690)	(5.020)	(4.570)	(1.250)
Direct effect	0.1539 ***	0.1626 ***	0.1407 ***	0.1112 ***	0.1890 ***	0.1366 ***	0.1342 ***	0.1220 ***	0.1261 ***	0.1435 ***
	(8.260)	(5.700)	(5.780)	(4.000)	(7.620)	(6.820)	(4.290)	(4.530)	(4.100)	(5.130)
Ratio of indirect effect	17.60%	11.06%	22.32%	26.28%	14.33%	3.99%	−1.37%	13.41%	10.16%	2.48%

Notes: This table reports the results of the mediating effects of corporate innovation and green innovation tested by the Sobel method and the Bootstrap method, which are shown in Panels A and B, respectively. SOE denotes the subsample of state-owned firms, while POE denotes the subsample of private firms. KPMs and NON-KPMs represent the subsample of key pollution-monitored firms and non-key pollution-monitored firms, respectively. There are robust t-statistics in parentheses. *** p < 0.01.

, both corporate innovation and green innovation have significant mediating effects on the relationship between the green finance policy and corporate environmental performance on the whole. The proportion of the total effect that is mediated by corporate innovation is 17.65%, while the proportion mediated by green innovation is 4.23%.

In terms of corporate innovation, its mediating effect is 10.93% for state-owned firms and 22.10% for private firms, while the ratio of the indirect effect of corporate innovation in the groups of key and non-key pollution-monitored firms is 26.49% and 14.48%, respectively. It indicates that the mediating effect of corporate innovation is greater in state-owned enterprises and key pollution-monitored enterprises, compared with private and non-key pollution monitoring enterprises. However, as for green innovation, it only plays a partially mediating role for private firms and key pollution-monitored enterprises, of which the ratios of indirect effects are, respectively, 13.48% and 10%. As reported in Panel B of Table 11, these findings above still can hold when the Bootstrap method is used to re-measure the mediating roles of corporate innovation and green innovation.

5.4.2. Financial Constraints

Considering that corporate innovation only plays a partially mediating role, we further analyze the mediating role of financial constraints between the green finance policy and corporate environmental performance. As reported in

Table 12. The mediating effect of financial constraints tested by the stepwise regression method.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)
	Total Sample			Enterprise Ownership						Degree of Environmental Supervision
				SOE			POE			KPMs			NON-KPMs
	$\mathit{E}\mathit{P}$	FC	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	FC	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	FC	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	FC	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	FC	$\mathit{E}\mathit{P}$
$Treat\times Post$	0.1342 ***	0.0007	0.1343 ***	0.1114 ***	0.0343 ***	0.1208 ***	0.1548 ***	−0.0240 ***	0.1463 ***	0.1445 ***	0.0046	0.1457 ***	0.1257 ***	−0.0037	0.1242 ***
	(6.668)	(0.157)	(6.699)	(3.619)	(4.892)	(3.928)	(5.768)	(−4.148)	(5.470)	(4.700)	(0.742)	(4.749)	(4.747)	(−0.576)	(4.708)
FC			−0.3128 ***			−0.2742 ***			−0.3439 ***			−0.2490 ***			−0.3437 ***
			(−11.010)			(−6.989)			(−8.035)			(−5.505)			(−9.361)
$AG$	−0.0038	0.0027 *	−0.0030	−0.0129	0.0041 *	−0.0118	0.0019	0.0012	0.0023	0.0070	−0.0009	0.0068	−0.0059	0.0042 *	−0.0045
	(−0.542)	(1.679)	(−0.430)	(−1.007)	(1.738)	(−0.948)	(0.253)	(0.477)	(0.310)	(0.936)	(−0.604)	(0.902)	(−0.601)	(1.701)	(−0.464)
$Cash$	−0.0081 ***	−0.0114 ***	−0.0117 ***	−0.0090	−0.0122 ***	−0.0124 **	−0.0122 ***	−0.0074 ***	−0.0147 ***	−0.0164 ***	−0.0147 ***	−0.0201 ***	−0.0075 ***	−0.0081 ***	−0.0103 ***
	(−3.880)	(−6.890)	(−5.238)	(−1.507)	(−4.412)	(−2.156)	(−4.724)	(−5.658)	(−5.225)	(−2.825)	(−5.712)	(−3.548)	(−3.425)	(−5.265)	(−4.434)
$Lev$	−0.0026	−0.0081 ***	−0.0052	−0.1412 ***	0.0507	−0.1273 ***	−0.0005	−0.0070 ***	−0.0029 **	0.0006	−0.0113 ***	−0.0022	−0.0887 ***	0.0436 **	−0.0740 **
	(−0.779)	(−5.825)	(−1.551)	(−2.933)	(1.383)	(−2.923)	(−0.407)	(−6.655)	(−1.977)	(0.407)	(−6.261)	(−1.154)	(−2.712)	(2.200)	(−2.443)
$ROA$	0.0013	−0.0057	−0.0005	0.0018	0.0544	0.0168	0.0007	−0.0052	−0.0011	0.0018	−0.0086	−0.0004	−0.0106	0.0371 *	0.0023
	(0.659)	(−0.951)	(−0.277)	(0.030)	(0.978)	(0.309)	(0.415)	(−1.375)	(−0.635)	(1.240)	(−1.482)	(−0.347)	(−0.269)	(1.730)	(0.065)
$Size$	0.1981 ***	−0.0316 ***	0.1883 ***	0.2171 ***	−0.0726 ***	0.1972 ***	0.1883 ***	0.0179 ***	0.1946 ***	0.2250 ***	−0.0618 ***	0.2096 ***	0.1715 ***	−0.0128 ***	0.1672 ***
	(36.945)	(−16.652)	(35.679)	(29.049)	(−22.270)	(26.755)	(21.120)	(7.160)	(21.784)	(26.674)	(−23.994)	(24.992)	(23.041)	(−3.992)	(23.026)
Constant	−2.5481 ***	4.3735 ***	−1.1805 ***	−2.9334 ***	5.2910 ***	−1.4827 ***	−2.2918 ***	3.2653 ***	−1.1713 ***	−3.0865 ***	5.0681 ***	−1.8245 ***	−1.9847 ***	3.9210 ***	−0.6386 ***
	(−21.454)	(103.364)	(−7.546)	(−18.082)	(86.601)	(−6.270)	(−11.873)	(59.806)	(−5.177)	(−16.325)	(86.872)	(−6.813)	(−12.429)	(59.436)	(−3.233)
Obs.	25,121	25,120	25,120	12,889	12,889	12,889	12,232	12,231	12,231	11,792	11,792	11,792	13,329	13,328	13,328
Adj.R2	0.0818	0.2318	0.0865	0.0971	0.2794	0.1006	0.0625	0.2892	0.0672	0.0846	0.2862	0.0870	0.0755	0.2224	0.0817
F	119.9	443.2	115.4	73.65	250.7	70.79	48.67	294.9	47.22	58.28	250.1	55.57	62.59	234.5	62.24

Notes: This table reports the estimation results of the mediating effect model specified in Equations (3)–(5), for testing the mediating effect of financial constraints (FC) between the GFPZ policy and corporate environmental performance. Columns (1)–(3) display the empirical results of the mediating effect model for the total sample. Columns (4)–(9) display corresponding results for the subsamples grouped by enterprise ownership, namely state-owned firms (SOE) and private firms (POE). Columns (10)–(15) display the empirical results for the subsamples grouped by the degree of environmental supervision, namely key pollution-monitored firms (KPMs) and non-key pollution-monitored firms (NON-KPMs). There are robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

, the impact of the green finance policy on corporate financial constraints is not statistically significant in the total sample. When the total sample is split according to enterprise ownership, the estimated coefficient of $Treat\times Post$ for $FC$ is 0.0343 in the subsample of state-owned firms and −0.0240 in the subsample of private firms. Both of them are statistically significant at the 1% level.

When we split the total sample by the degree of environmental supervision, it shows that the estimated coefficients of $Treat\times Post$ for $FC$ in the subsamples of key and non-key pollution-monitored firms are 0.0046 and −0.0037, respectively. This further implies that the green finance policy (GFPZ) has tightened the financial constraints of state-owned enterprises and key pollution-monitored enterprises, while easing the financial constraints of private enterprises and non-key pollution-monitored enterprises.

Combining the estimation results from Table 12 and

Table 13. The mediating effect of financial constraints tested by the Sobel method and the Bootstrap method.

	GFPZ → FC → EP
	Total Sample	Enterprise Ownership		Degree of Environmental Supervision
		SOE	POE	KPMs	NON-KPMs
Panel A: The mediating role tested by the Sobel method
Total effect	0.187 ***	0.183 ***	0.181 ***	0.151 ***	0.220 ***
	(10.326)	(6.541)	(7.588)	(5.445)	(9.251)
Indirect effect	−0.021 ***	−0.017 ***	−0.021 ***	−0.024 ***	−0.016 ***
	(−6.930)	(−3.215)	(−5.767)	(−4.876)	(−4.339)
Direct effect	0.208 ***	0.199 ***	0.202 ***	0.175 ***	0.237 ***
	(11.354)	(7.021)	(8.391)	(6.234)	(9.828)
Ratio of indirect effect	−11.23%	−9.29%	−11.60%	−15.89%	−7.27%
Panel B: The mediating role tested by the Bootstrap method
Indirect effect	−0.0210 ***	−0.0166 ***	−0.0208 ***	−0.0242 ***	−0.0163 ***
	(−6.940)	(−3.160)	(−6.250)	(−4.620)	(−4.320)
Direct effect	0.2077 ***	0.1995 ***	0.2017 ***	0.1750 ***	0.2367 ***
	(10.810)	(7.150)	(8.190)	(5.860)	(9.520)
Ratio of indirect effect	−11.27%	−9.08%	−11.51%	−16.08%	−7.39%

Notes: This table reports the results of the mediating effect of financial constraints tested by the Sobel method and the Bootstrap method, which are shown in Panels A and B, respectively. SOE denotes the subsample of state-owned firms, while POE denotes the subsample of private firms. KPMs and NON-KPMs represent the subsample of key pollution-monitored firms and non-key pollution-monitored firms, respectively. There are robust t-statistics in parentheses. *** p < 0.01.

, we next analyze the mediating effect of financial constraints between the green finance policy and corporate environmental performance. As a whole, financial constraints can play a partially mediating role, and the ratio of indirect effects is around −11.23%. It suggests that, to a certain extent, the positive impact of the green finance policy on corporate environmental performance is achieved by reducing their financial constraints. As shown in Panel A of Table 13, upon further classification of the total sample based on enterprise ownership and the degree of environmental supervision, it is found that the mediating effects of financial constraints in private enterprises (−11.60%) and key pollution-monitored enterprises (−15.89%) are higher than that in state-owned enterprises (−9.29%) and non-key pollution-monitored enterprises (−7.27%), respectively. Similar findings are supported by the empirical results estimated by the Bootstrap method (Panel B, Table 13). The results provide evidence that the third research hypothesis (H2b) proposed in this study can hold true.

5.4.3. Bank Credit

Since we have already identified the mitigating effect of the green finance policy on the financial constraints faced by enterprises, we further analyze whether the green finance policy can influence corporate environmental performance by affecting bank credit. Specifically, we construct a new mediating variable, namely taking the logarithm of the amount of bank loans obtained by enterprises (including short-term and long-term loans in the balance sheet) as a proxy variable for bank credit, to analyze its mediating role in the positive impact of the GFPZ policy on corporate environmental performance. The results on the mediating role of bank credit tested by the stepwise regression method, the Sobel method, and the Bootstrap method are, respectively, reported in

Table 14. The mediating effect of bank credit tested by the stepwise regression method.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)
	Total Sample			Enterprise Ownership						Degree of Environmental Supervision
				SOE			POE			KPMs			NON-KPMs
	$\mathit{E}\mathit{P}$	BC	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	BC	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	BC	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	BC	$\mathit{E}\mathit{P}$	$\mathit{E}\mathit{P}$	BC	$\mathit{E}\mathit{P}$
$Treat\times Post$	0.1342 ***	0.2773 **	0.1347 ***	0.1114 ***	0.0932	0.1117 ***	0.1548 ***	0.3883 **	0.1547 ***	0.1445 ***	0.4759 ***	0.1433 ***	0.1257 ***	0.1100	0.1262 ***
	(6.668)	(2.468)	(6.695)	(3.619)	(0.577)	(3.629)	(5.768)	(2.513)	(5.757)	(4.700)	(3.265)	(4.655)	(4.747)	(0.690)	(4.770)
BC			−0.0020 **			−0.0032 **			0.0004			0.0027			−0.0045 ***
			(−2.007)			(−2.036)			(0.311)			(1.643)			(−3.438)
$AG$	−0.0038	0.0401	−0.0037	−0.0129	0.1631	−0.0124	0.0019	−0.0065	0.0019	0.0070	0.0115	0.0070	−0.0059	0.0869	−0.0056
	(−0.542)	(1.038)	(−0.533)	(−1.007)	(1.619)	(−0.982)	(0.253)	(−0.140)	(0.253)	(0.936)	(0.264)	(0.928)	(−0.601)	(1.626)	(−0.570)
$Cash$	−0.0081 ***	−1.0142 ***	−0.0101 ***	−0.0090	−1.3790 ***	−0.0134 **	−0.0122 ***	−0.7917 ***	−0.0118 ***	−0.0164 ***	−1.6808 ***	−0.0119 *	−0.0075 ***	−0.6510 ***	−0.0104 ***
	(−3.880)	(−6.097)	(−4.420)	(−1.507)	(−11.757)	(−2.213)	(−4.724)	(−4.977)	(−4.433)	(−2.825)	(−8.374)	(−1.861)	(−3.425)	(−4.653)	(−4.206)
$Lev$	−0.0026	0.3311	−0.0020	−0.1412 ***	6.3467 ***	−0.1211 **	−0.0005	0.2431	−0.0006	0.0006	0.1300	0.0003	−0.0887 ***	6.7339 ***	−0.0583 *
	(−0.779)	(1.303)	(−0.680)	(−2.933)	(5.849)	(−2.536)	(−0.407)	(1.548)	(−0.467)	(0.407)	(1.458)	(0.162)	(−2.712)	(5.399)	(−1.843)
$ROA$	0.0013	−0.1670 **	0.0010	0.0018	−1.7692 ***	−0.0038	0.0007	−0.1338 *	0.0007	0.0018	−0.1786 ***	0.0023	−0.0106	−0.9787	−0.0150
	(0.659)	(−2.220)	(0.473)	(0.030)	(−2.744)	(−0.062)	(0.415)	(−1.764)	(0.450)	(1.240)	(−6.334)	(1.590)	(−0.269)	(−1.294)	(−0.389)
$Size$	0.1981 ***	2.1872 ***	0.2025 ***	0.2171 ***	1.5780 ***	0.2221 ***	0.1883 ***	2.6505 ***	0.1872 ***	0.2250 ***	1.8058 ***	0.2202 ***	0.1715 ***	2.0131 ***	0.1806 ***
	(36.945)	(45.712)	(34.305)	(29.049)	(22.541)	(28.051)	(21.120)	(37.545)	(19.170)	(26.674)	(34.577)	(24.206)	(23.041)	(22.280)	(22.576)
Constant	−2.5481 ***	−30.5421 ***	−2.6090 ***	−2.9334 ***	−19.4435 ***	−2.9951 ***	−2.2918 ***	−41.1421 ***	−2.2746 ***	−3.0865 ***	−20.7152 ***	−3.0308 ***	−1.9847 ***	−30.6961 ***	−2.1234 ***
	(−21.454)	(−24.580)	(−20.976)	(−18.082)	(−16.821)	(−18.037)	(−11.873)	(−23.139)	(−11.178)	(−16.325)	(−15.909)	(−15.613)	(−12.429)	(−18.299)	(−12.760)
Obs.	25,121	25,121	25,121	12,889	12,889	12,889	12,232	12,232	12,232	11,792	11,792	11,792	13,329	13,329	13,329
Adj.R2	0.0818	0.2934	0.0819	0.0971	0.3502	0.0973	0.0625	0.2793	0.0624	0.0846	0.3031	0.0847	0.0755	0.3212	0.0762
F	119.9	513.6	112.7	73.65	314.7	69.51	48.67	242.3	45.85	58.28	214.1	55.08	62.59	305.3	59.11

Notes: This table reports the estimation results of the mediating effect model specified in Equations (3)–(5), for testing the mediating effect of bank credit (BC) between the GFPZ policy and corporate environmental performance. Columns (1)–(3) display the empirical results of the mediating effect model for the total sample. Columns (4)–(9) display the corresponding results for the subsamples grouped by enterprise ownership, namely state-owned firms (SOE) and private firms (POE). Columns (10)–(15) display the empirical results for the subsamples grouped by the degree of environmental supervision, namely key pollution-monitored firms (KPMs) and non-key pollution-monitored firms (NON-KPMs). There are robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

and

Table 15. The mediating effect of bank credit tested by the Sobel method and the Bootstrap method.

	GFPZ → BC → EP
	Total Sample	Enterprise Ownership		Degree of Environmental Supervision
		SOE	POE	KPMs	NON-KPMs
Panel A: The mediating role tested by the Sobel method
Total effect	0.187 ***	0.183 ***	0.181 ***	0.151 ***	0.221 ***
	(10.329)	(6.541)	(7.595)	(5.445)	(9.256)
Indirect effect	0.000	0.001	0.000	0.000	0.000
	(0.201)	(1.087)	(0.285)	(0.850)	(0.503)
Direct effect	0.187 ***	0.182 ***	0.181 ***	0.150 ***	0.220 ***
	(10.328)	(6.518)	(7.591)	(5.431)	(9.244)
Ratio of indirect effect	0.00%	0.55%	0.00%	0.00%	0.00%
Panel B: The mediating role tested by the Bootstrap method
Indirect effect	0.00005	0.0007	0.0001	0.0004	0.0004
	(0.200)	(1.040)	(0.240)	(0.730)	(0.500)
Direct effect	0.1867 ***	0.1822 ***	0.1810 ***	0.1504 ***	0.2202 ***
	(10.470)	(6.400)	(7.260)	(5.150)	(9.030)
Ratio of indirect effect	0.03%	0.36%	0.04%	0.27%	0.17%

Notes: This table reports the results of the mediating role of bank credit tested by the Sobel method and the Bootstrap method, which are shown in Panels A and B, respectively. SOE denotes the subsample of state-owned firms, while POE denotes the subsample of private firms. KPMs and NON-KPMs represent the subsample of key pollution-monitored firms and non-key pollution-monitored firms, respectively. There are robust t-statistics in parentheses. *** p < 0.01.

.

As seen from Table 14, the estimated coefficient of $Treat\times Post$ for $BC$ is 0.2773, which is statistically significant at the 1% level. This is indeed in line with our speculation that the GFPZ policy does have a positive impact on corporate borrowing. Further analysis reveals that this positive impact is more pronounced in private enterprises compared with state-owned ones. Moreover, compared with non-key pollution-monitored enterprises, the green finance policy has a more significant promoting effect on the bank loans obtained by key pollution-monitored enterprises.

However, when we analyze the estimation results in Table 14 and Table 15 together, it can be found that although the green finance policy can effectively increase the bank credit obtained by enterprises, bank credit does not play a significant mediating role between the green finance policy and corporate environmental performance. It implies that the fourth research hypothesis (H2c) proposed in this study cannot hold.

5.5. Further Discussion

In the previous subsections above, we have analyzed the impact of the green finance policy on corporate environmental performance, by using the policy for establishing green finance pilot zones (GFPZs) in five selected provinces of China as a sample. On the whole, the baseline regression results provide evidence that the effect of the GFPZ policy on corporate environmental performance is positive, and this conclusion is supported by multiple robustness tests such as the parallel trend test, the placebo test, the PSM-DID regression and changing dependent variable. This finding is in line with some previous studies. For example, Ref. [16] finds that green finance has a positive impact on environment sustainability, by using the S&P green bond index total return as a proxy of green finance. Some scholars also argue that green finance products have a positive impact on sustainable development such as decreasing CO₂ emission and improving wastewater treatment [3]. The original intention of the policy for establishing green finance pilot zones in China is precisely to encourage financial institutions to launch a wide variety of green finance products or strategies to guide enterprises to protect the environment. It is quite obvious that the empirical research results of our study have provided evidence for the expected positive policy effect to some extent.

However, we find that there is heterogeneity in the effect of the GFPZ policy on corporate environmental performance, when the total sample is categorized by enterprise ownership, industry, and the degree of environmental supervision. In particular, compared with private firms and key pollution-monitored firms, the impact of the GFPZ policy on environmental performance is stronger for state-owned firms and non-key pollution-monitored firms. We speculate that this might be related to the fact that Chinese state-owned enterprises have established party organizations, and they are more proactive in responding to governmental policies. Meanwhile, due to the strengthened environmental pollution supervision under the green finance policy, non-key pollution-monitored firms may be more likely to obtain green financial support.

We have tried to find out the channels through which the green finance policy affects corporate environmental performance. Corporate innovation, green innovation, financial constraints, and bank credit have all been taken into consideration by us one by one. We find that corporate innovation has a partially positive mediating action between the GFPZ policy and corporate environmental performance. This also supports some views of previous studies, which argue that green finance policy or green finance products can enhance innovation capabilities of firms [27]. However, when we confine corporate innovation to green innovation, we find that green innovation only plays a significant mediating role in private enterprises and enterprises subject to key pollution monitoring. However, its positive mediating effect is not observed in state-owned enterprises and non-key pollution-monitored enterprises. The empirical results further provide evidence that corporate financial constraints have been decreased by the green finance policy, especially for private enterprises and key pollution-monitored enterprises. Although the green finance policy has increased the bank credit available to enterprises, bank credit has not played a significant mediating role between the green finance policy and corporate environmental performance. Based on the above analysis, we argue that the green finance policy can reduce the financial constraints faced by enterprises, enabling them to obtain more financial support. This, in turn, promotes corporate innovation (especially green innovation), thereby improving environmental performance. This influence path is more pronounced for private enterprises and enterprises subject to key pollution monitoring.

6. Conclusions

In this study, we take the policy for establishing green finance pilot zones (GFPZs) of China as an example to investigate the impact of green finance policy on corporate environmental performance. The empirical sample includes 2324 Chinese A-share listed companies during the period from 2012 to 2022. We started to investigate the effect of the GFPZ policy on corporate environmental performance by the difference-in-difference (DID) model, using four methods (i.e., the parallel trend test, the placebo test, the PSM-DID model, and replacing dependent variables) to check the robustness of baseline regression results. Subsequently, we explored the heterogeneity of the effect of green finance policy from three dimensions, namely enterprise ownership, sector, and the degree of pollution regulation. We also analyzed the financial and non-financial channels through which the green finance policy affects corporate environmental performance, in which the mediating roles of corporate innovation, green innovation, financial constraints, and bank credit are examined.

The main findings can be concluded as follows: First of all, the empirical results show that there is a statistically significant and positive impact of the GFPZ policy on corporate environmental performance. Secondly, there are heterogeneities in the effect of the green finance policy. Compared with private firms and key pollution-monitored firms, the GFPZ policy has a relatively stronger effect on state-owned and non-key pollution-monitored firms. Thirdly, corporate innovation plays a partially mediating role between the GFPZ policy and corporate environmental performance in the total sample and the subsamples grouped by enterprise ownership and the degree of pollution supervision. Green innovation only plays a significantly partial mediating role in private enterprises and enterprises subject to key pollution monitoring. Fourthly, the GFPZ policy can decrease corporate financial constraints, and financial constraints can also play a partially mediating role between the GFPZ policy and corporate environmental performance, especially for private and key pollution-monitored enterprises. However, although the GFPZ policy can positively influence bank credit, there is no evidence that bank credit can have a significantly mediating effect in the relationship between green finance policy and environmental performance.

Our findings support the view of some previous studies, suggesting the positive effect of green finance policy on environmental sustainability. Unlike previous studies, we not only discuss both financial and non-financial mechanisms through which green finance policy affects corporate environmental performance but also explore the heterogeneity of the influence mechanism from different aspects (i.e., enterprise ownership and the degree of environmental supervision). Among them, we find that green finance policy can significantly alleviate the financial constraints of private enterprises and key pollution-monitored enterprises, thereby promoting their improvement in green innovation capabilities and environmental performance. This is of great significance for understanding the policy effects and the influence mechanism of establishing green finance pilot zones in China. Based on the empirical results of this study, we suggest that private enterprises and those enterprises subject to key pollution monitoring in China should broaden their financing channels and increase investment in green innovation, thereby promoting their environmental performance under the guidance of the GFPZ policy.

However, this study also has certain limitations and requires further research in the future. For instance, we only concentrate on whether the green finance policy can affect corporate environmental performance through corporate innovation, financial constraints, and bank credit in this paper. It could be necessary and interesting to further investigate the inter-relationship among bank credit (especially green credit), financial constraints, and corporate innovation in the future. In addition, instead of using the environment rating data provided by the China Securities Index Company, constructing more innovative indicators for environmental assessment would also be an interesting direction for further research in the future.