Green Bonds Drive Environmental Performance: Evidences from China

Abstract

Faced with the urgent challenge of global warming, green bonds play an important role in promoting economic transformation and improving environmental quality by financing environmentally friendly projects. However, the actual effects of green bonds, especially their impact on corporate environmental performance, and the mechanisms behind it, still need to be studied and validated. Based on the time-varying difference-in-differences (DID) model, this study uses 85 Chinese A-share listed companies that have issued green bonds from 2013 to 2022, to study the impact of green bond issuance on corporate environmental performance and the potential mechanisms. The results show that green bonds issuance effectively promotes the improvement of corporate environmental performance; this promotion is more significant for labor-intensive enterprises, larger enterprises, and enterprises with more government subsidies. In terms of the influencing mechanism, R&D investment and green innovation play partial mediating roles, media attention and analyst attention play positive moderating roles. This study further validates and complements the signal theory of green bonds and makes relevant suggestions for the development of green bonds in China.

1. Introduction

Faced with the urgent challenge of global warming, green finance has become a key strategy to promote environmental protection and achieve sustainable development. Green bonds, as a core tool of green finance, play an important role in promoting economic transformation and improving environmental quality by financing environmentally friendly projects. In 2007, the European Investment Bank issued the first green bond, called “Climate Awareness Bond”, which kicked off the development of the global green bond market [1,2]. The scale of green bonds is growing globally, and many regions or countries have launched policies to promote the development of green bonds. For example, the European Union launched EU green bond standards, regulating the EU green bond market. And the United States provided tax incentives for Qualified Energy Conservation Bonds and Clean Renewable Energy Bonds, stimulating the growth of these types of green bonds [2].

China joined the Paris Agreement in 2016 and had actively set goals for carbon peaking and carbon neutrality. Green bonds have received unprecedented development and attention in China [2,3,4]. In 2016, the Shanghai Stock Exchange of China issued the “Notice on launching Pilot projects for Green Corporate Bonds”, encouraging enterprises to support the development of green industries by issuing green bonds. At the same time, the Shanghai Stock Exchange of China issued the “Green Bond Support Project Catalog (2015 edition)”, which played an important role in defining the scope of projects supported by green bonds and regulating the development of the green bond market [5,6,7]. The catalog clearly pointed out that green industry includes six aspects: energy conservation, pollution prevention and control, resource conservation and recycling, clean transportation, clean energy, ecological protection and climate change adaptation. In particular, in 2021 and 2022, China further expanded the scope of projects supported by green bonds and issued the China Green Bond Principles [8,9,10] to further unify standards with international standards and promote the high-quality development of the green bond market.

It is particularly noteworthy that since the implementation of the green bond policy, the issuance scale of green bonds in China has continued to grow, reflecting the strong market demand and recognition of this financial instrument. By the end of 2022, China has ranked among the top in the global green bond market [11]. According to the data of China Bond Information Network (CBIN) Green Bond Environmental Benefit Information Database [10,12,13], China issued more than 600 billion RMB green bonds in 2022, reaching a record high. The total scale of green bond issuance from 2016 to 2022 was about 2 trillion RMB, and the cumulative issuance scale from 2020 to 2022 accounted for more than 60%, indicating that the period from 2020 to 2022 is the stage of explosive growth of China’s green bond issuance.

The rapid development of China’s green bond market reflects not only the strategic decision of the national level to green transition, but also the positive response of enterprises to achieve environmental sustainability goals. The original intention of green bonds is to improve environment, however, the impact of green bond issuance on environmental performance has not been determined [3,4,14,15], in particular the mechanisms and pathways of impact are not clear. So, it is significant to study and validate the actual effects of green bonds on corporate environmental performance, and identify the influence mechanisms from the perspectives of the internal and external effects. In addition, with the rapid development of the market, the problems that green bonds may face, such as “green washing”, also need to be discussed and solved through empirical research.

Against this backdrop, this paper aims to explore in depth the impact of green bond issuance by Chinese listed companies on their environmental performance and its potential mechanism. By using a time-varying difference-in-differences (DID) model to analyze the data of listed companies in China’s A-share market issuing green bonds from 2013 to 2022, we find that the issuance of green bonds significantly promotes the improvement of corporate environmental performance. Further analysis reveals the mediating effect of R&D investment and green innovation, as well as the positive moderating effect of media attention and analyst attention. This finding provides a new perspective for understanding the role of green bonds in promoting corporate sustainable development, and puts forwards valuable suggestions for the further development of China’s green finance market.

Our paper makes the following contributions to the existing literature. First, under the circumstance of the impact of issuing green bonds on corporate environmental performance unclear [3,4,14,15], this study verifies that green bond issuance is an effective signal for enterprises to convey their commitment to environmental protection and sustainable development to the outside world, which is in line with the expectations of signal theory [3]. This finding not only further confirms the applicability of signal theory in the field of green finance, but also emphasizes how green bonds can effectively improve the environmental performance in the context of a rapidly developing green bond market in China. Second, this study contributes to the literature about the relationship of environmental, social, and governance (ESG) or corporate social responsibility (CSR) to green bond issuance [11,16,17,18]. This study proceeds from the perspective of environmental performance, different from previous papers from the ESG or CSR aspect. It belongs to a more specific and subdivided research direction [19], examining the direct effect of green bond issuance, and enriches the research perspective in this field.

The following are the main novelties of this paper: First, few researches have looked into the mechanism of green bonds on environmental performance [3,4,6]. Our paper systematically explores the mechanism of green bond issuance on improving corporate environmental performance from both the perspectives of the internal effects of R&D investment and green innovation and the external effects of media attention and analyst attention. This approach, which comprehensively considers internal and external factors, provides a new perspective for understanding how green bonds affect corporate environmental performance. In addition, our paper conducts research specific to China’s unique policy environment and market characteristics, considering the Chinese government’s efforts in promoting green finance and the market’s uniqueness. This makes our study highly relevant and practically valuable both theoretically and in practice.

The remainder of the essay is structured as follows: Section 2 presents literature review and proposes hypotheses. Section 3 then introduces research design, including sample selection, variable definition, and models. Section 4 presents the empirical results, robustness analysis, mechanisms analysis and heterogeneity effect. Section 5 finally concludes and gives policy implications.

2. Literature Review and Hypothesis

2.1. Literature Review

Enterprises issuing green bonds will face stricter information disclosure and third-party information review, which will increase the disclosure cost and compliance cost [3]. To issue green bonds, the raised funds need to be used for green projects, which have greater risk, longer cycle and greater profit uncertainty than non-green projects. Therefore, it seems that issuing and operating costs are lower for enterprises to raise funds by issuing ordinary bonds and then invest in green projects [3]. So why are companies issuing green bonds? There are usually three theories to explain the issuance of green bonds by enterprises: First, signal theory [20]. By issuing green bonds, enterprises can convey their determination to support environmental protection to the outside world, gain attention from the securities market and investors, and enhance social reputation and corporate value [3,11,21]. In addition, the issuance of green bonds by enterprises conveys their commitment to environmental protection to the outside world. In order to ensure the authenticity of the transmitted information, enterprises will take practical measures to reduce emissions and pollution, protect the environment, and improve their environmental performance. Second, greenwashing theory [22,23]. Some people believe that companies issuing green bonds only enhance their social reputation, and the investment and operation of green projects do not improve environmental governance [4]. The greenwashing of green bonds can lead to reputational risk, ostensibly concerned with environmental protection, but may fund projects that do not meet environmental performance expectations and fail to deliver the promised environmental goals. Third, the theory of capital cost [24,25,26]. Issuing green bonds has a lower cost of capital than ordinary bonds, which is one of the motivations for issuing green bonds. Because the issuance of green bonds discloses more information about enterprises and reduces investor information asymmetry, it can require lower financing costs. At the same time, stakeholders attach more importance to the responsibility of enterprises to society, and are willing to provide relatively cheap funds. However, the signal theory and the greenwashing theory are contradictory. The core of the signal theory is that the issuance of green bonds does play its due role in improving environmental performance, while the greenwashing theory holds that the issuance of green bonds does not play its due role. Therefore, if the results of a study support the signal theory, it reflects that the issuance of green bonds may discourage companies from “greenwashing” behavior, it means that the greenwashing theory is not supported to some extent [3,14,27,28]. Our study mainly focuses on the verification of signal theory and greenwashing theory, and does not involve the capital cost theory. Therefore, the following part mainly introduces the literature from these two aspects.

Flammer (2021) used DID model to study the green bond issuance from 2013 to 2018 globally, and found that the companies get higher environmental ratings and produce lower CO₂ emissions post-issuance, which support the signal theory and do not support the greenwashing theory. Although he found listed companies that issued green bonds improve their environmental performance, the mechanism of the effect was not tested [3]. Fatica and Panzica (2021) also used DID model to study the green bond issuance up to 2019 globally, and found that firms issuing non-refinanced green bonds have significantly improved carbon intensity, and green bonds issued after the Paris Agreement have greater emission reductions [14], which support the signal theory. At the same time, the authors argued that although it cannot be said that there is a causal relationship between green bond issuance and reductions in firms’ carbon intensity, it does not support the “greenwashing” theory to a certain extent. Wu Shinong et al. (2022) took Chinese A-share listed companies issuing green bonds from 2010 to 2019 as samples and concluded DID model to study the impact the issuance of green bonds on corporate environmental performance and corporate value. The results showed that the issuance of green bonds effectively promotes the corporate environmental performance and corporate value, which in accordance with the signal theory, but they did not propose the impact mechanism [6]. Using a sample of Chinese A-share listed companies issuing green bonds from 2011 to 2022, Chen Fenggong and Zhang Yihao (2023) also built DID model to examine the relationship between green bond issuance and environmental performance. The results implied that green bond issuance improves firms’ environmental performance through the mediating effects of social and media attention, environmental institutions, and legal constraints [27]. From a macro perspective, Zhang Ke et al. (2023) took the panel data of 279 prefecture level cities from 2011 to 2020 as samples, used DID model to study the impact of green bond issuance on carbon intensity, and found that green bonds can significantly reduce urban carbon emission intensity by easing the financing constraints and increasing the social attention of issuers [7]. The above studies are all consistent with the signal theory of green bonds, but do not support the greenwashing theory.

Compared with the above scholars, there are also researches supporting the greenwashing theory. Compare to Flammer (2021)’s study, García et al. (2023) expanded the research sample of global green bond issuances, and the sample time window was extended from 2013 to 2021. Using the DID model, the empirical study found that there was no significant relationship between the issuance of green bonds and the improvement of corporate environmental performance, the greenwashing theory was supported [4]. From a macro perspective, Sinha et al. (2021) studied the relationship between S&P 500 Global Green Bond Index and S&P 500 Environmental and Social Responsibility Index from 2010 to 2020, using the model of combination of Quantile-on-Quantile Regression and Wavelet Multiscale Decomposition, and found that green bond financing may have a negative impact on environmental and social responsibility [15]. Wang Yong and Lu Xueyao (2023) selected the panel data of 30 provinces in China from 2010 to 2020, used DID model to test the green washing risk of green bonds and found that compared with non-labeled green bonds, labeled green bonds have higher risk of greenwashing [10].

As green bond is a new financing tool, the limitation of the current empirical research on green bond is that the observation sample is too small [3,4,14]. With the expansion of the observation sample for empirical research, long-term effects can be observed and different conclusions can be drawn. Therefore, different scholars chose different samples, research scopes and research methods to study the relationship between green bond issuance and corporate environmental performance, and reached different conclusions. Whether it is in an international context, or focusing on the specific situation in China, the impact of green bond issuing on environmental performance and sustainability has not been confirmed. More importantly, the specific mechanisms by which green bond issuance affects environmental performance have not been clearly defined, nor have been systematically studied and validated. It is still worth studying whether green bonds have become a tool for issuers to “wash green”, and whether issuers have raised their social image by issuing green bonds without significantly improving their environmental performance [29]. Especially under the unique policy background of China, it is necessary to expand the observation sample and conduct empirical research to observe the long-term impact.

2.2. Hypothesis

Signaling theory states that by issuing green bonds, companies are able to send positive signals to the market about their environmental responsibility and sustainability [3]. This signal is valuable to investors because it helps them identify companies that are committed to environmental improvement and have the potential for long-term sustainability. By issuing green bonds, companies can demonstrate their commitment to environmental responsibility, thus attracting more sustainability-conscious investors, thereby obtaining more funds to support environmentally friendly projects [16]. Not only to send signals to investors, such a clear signal of a company’s commitment to environmental responsibility and sustainability can also meet the expectations of the stakeholders. Stakeholder theory emphasizes that enterprises need to balance and satisfy the needs of their various stakeholders [30], By issuing green bonds and using the funds raised for environmentally friendly projects, companies can enhance their relationships of trust with stakeholders, including consumers, employees and communities, as it demonstrates their commitment to social responsibility.

In order to maintain the authenticity of the signal sent by green bonds, companies will strive to improve their environmental performance to meet expectations. Funds from green bond issuance are usually limited to environmental or sustainable development projects, such as clean energy, energy conservation and emission reduction, green transportation and water management [6,7]. This specialization ensures that the use of funds is directly linked to improved environmental performance. Through the implementation of these projects, companies can directly improve their environmental impact, such as reducing greenhouse gas emissions, improving energy efficiency or adopting sustainable resources [7,27]. In addition, green bonds are typically accompanied by greater transparency and reporting requirements, including environmental impact assessments of projects and the use of funds. Increased transparency helps build trust between companies and investors, while forcing companies to regularly evaluate the performance of their environmental projects, thereby continuously improving their environmental performance. Therefore, hypothesis 1 is proposed:

Hypothesis 1 (H1). 

Issuing green bonds can improve companies’ environmental performance.

The issuance of green bonds provides companies with funds earmarked for environmentally friendly projects. Environmentally friendly projects include energy conservation and carbon reduction, resource recycling, clean energy, ecological protection, restoration and utilization, and green upgrading of infrastructure, etc. These projects usually involve major technological updates and modifications, require a large amount of research and development funds. As a means of financing, green bonds can raise funds to alleviate the financing constraints of enterprises [6,7]. Companies have more incentive to invest more in research and development, especially in the development of green technologies and products. Therefore, green bonds have become a financial strategy for enterprises to increase investment in research and development.

R&D investment is the key factor to promote green innovation [31]. Through research and development, companies can develop new technologies or improve existing technologies, thereby enhancing their green innovation capabilities. The issuance of green bonds is often accompanied by the production of green patents [32]. Green innovation has a longer cycle, slower returns and higher risks than traditional technological innovation [33,34]. The profits brought by green innovation for enterprises are not as good as those brought by traditional technological innovation in the short term, and green innovation mainly bears the social responsibility of protecting ecological environment [35]. It is difficult for ordinary financing methods to raise funds for green innovation, while green bonds inject funds into enterprises’ green projects, with the upgrading of production technology and the transformation of production structure, enterprises form certain technical barriers through the application of green patents, protect enterprises’ unique economic returns, deal with environmental pollution, and enhance enterprises’ environmental performance [21,36,37]. Therefore, hypothesis 2 is proposed:

Hypothesis 2 (H2). 

In terms of internal effect, issuing green bonds can improve the environmental performance of enterprises by increasing R&D investment and enhancing the green innovation capability.

According to signal theory, companies send positive signals to the outside world that they are committed to sustainable development and environmental protection by issuing green bonds [3]. When the media gives high attention to such behavior, the propagation effect of this signal will be further amplified, and more investors and stakeholders who pay attention to environmental protection will be attracted, thus improving the social reputation. This enhanced social reputation prompts companies to work harder to fulfill their environmental commitments in order to maintain and enhance their image in the eyes of the public [16,38].

As the guide of public opinion and the main source of public information, media can influence the views and reactions of other stakeholders, including consumers, government agencies and investors, on corporate behavior [27]. The high level of media attention not only increases the public exposure of companies, but also raises the public’s expectations of corporate environmental behavior. To meet these expectations, companies are likely to further enhance their environmental management practices and invest more resources in environmental projects and research and development of green technologies to achieve substantial improvements in environmental performance [39,40,41]. In addition, media attention can improve the transparency and quality of corporate environmental information [42]. In the face of constant media and public attention, companies are encouraged to disclose more detailed and accurate environmental information, including the progress of their green projects and the effectiveness of environmental management measures. This increased transparency helps to reduce information asymmetries, enabling investors and consumers to make better decisions, and thereby promoting positive feedback from capital markets on companies’ environmentally friendly behavior. Therefore, hypothesis 3 is proposed:

Hypothesis 3 (H3). 

In terms of external effects, media attention positively moderates the effect of green bond issuance on environmental performance.

Issuing green bonds is more likely to receive analyst attention [16]. First, analyst attention increases the visibility and acceptance of green bonds issued by companies. Through in-depth analysis and reporting on companies’ environmental projects and their potential contribution to their long-term value, analysts help market participants better understand the environmental commitment behind these bonds, thus attracting more investors with a focus on sustainable investing. This increased market recognition resulting from analyst attention not only increases the attractiveness of green bonds, but also gives companies more resources to implement and scale up environmentally friendly projects, directly improving their environmental performance.

In addition, the constant attention and evaluation of analysts amounts to a kind of external supervision of companies, forcing companies to take their environmental protection commitments seriously [43]. In the face of continued analyst and market attention, companies are more likely to take practical actions, such as increasing investment in green research and development and promoting green innovation, to ensure that the funds raised by green bonds are used effectively and truly improve their environmental performance [16]. This external pressure encourages enterprises not only to be satisfied with the surface environmental image, but to improve their environmental management level and performance through substantive improvements. In summary, analyst attention positively moderates the relationship between green bond issuance and corporate environmental performance by increasing the market recognition of corporate green bonds, implementing external supervision pressure. This not only strengthens the direct effect of green bonds on improving environmental performance, but also provides external motivation and support for enterprises to continuously improve environmental performance. Therefore, hypothesis 4 is proposed:

Hypothesis 4 (H4). 

In terms of external effects, analyst attention positively moderates the effect of green bond issuance on environmental performance.

3. Samples, Variables, and Models

3.1. Samples

As China’s green bond policy was implemented in 2016, there were no labeled green bonds issued before 2016, and the financial data of listed companies in 2023 are not all publicly available when the paper is written. So green bond issuance from 2016 to 2022 is obtained from China Bond Information Network (CBIN) Green Bond Environmental Benefit Information Database [12]. After obtaining the bond information, manually identify whether it is listed company, and cross-check with the green bond issuance data obtained from WIND database and CSMAR database [44,45]. After excluding the financial industry, 85 listed companies that have issued green bonds are finally obtained as samples. According to the industry classification of the 2012 edition of the China Securities Regulatory Commission, the industries of 85 listed companies issuing green bonds include mining, electricity, heat, gas and water production and supply industry, real estate industry, construction industry, transportation, warehousing and postal industry, wholesale and retail industry, water conservancy, environment and public facilities management industry, manufacturing industry, a total of 8 first-level industries and 24 second-level industries. Refer to previous studies [11,21,27], the control group select listed companies in the same secondary industry as the 85 listed companies issuing green bonds, which contains 24 second-level industries, such as extraction of petroleum and natural gas, nonferrous metal mining and beneficiation industry, electricity and heat production and supply industry, water production and supply industry, electrical machinery and equipment manufacturing, motor industry ang so on. To further ensure the comparability of the control group, the robustness test used PSM to match the samples. In order to obtain sufficient pre-policy data for the treatment group, the research period of variables is advanced to 2013, 3 years before the implementation of China’s green bond policy [3]. So, the research period is from 31 December 2013 to 31 December 2022. Financial companies and enterprises with ST, *ST are excluded. In order to avoid the influence of extreme outliers on the results, all continuous variables are winsorized by 1% and 99%, and samples with missing variables in the sample period are not included. The final treatment group includes 85 listed companies that has issued green bonds, and the control group includes 3185 listed companies that has not issued green bonds. The observation period is from 2013 to 2022, with a total of 18,194 samples. In order to balance the difference between the sample size of the treatment and control group, the sample size is further reduced after PSM matching in the robustness test. The Hua Zheng environmental performance scores are obtained from WIND database, the CNRDS and Bloomberg environmental performance scores are gained from CNRDS database and Bloomberg database. The green patent data are obtained from State Intellectual Property Office patent search website, which are captured by Python, and identify through the IPC code of the green patent of the World Intellectual Property Organization. The data of media attention come from China Research Data Service Platform (CNRDS); Other data were taken from CSMAR database.

3.2. Variable

3.2.1. Explained Variable

The explained variable is environmental performance (E). Generally, there are three methods to measure environmental performance. One is to use third-party environmental performance scores [3,6]. The second is the use of CO₂ emissions and CO₂ emission intensity [14,46]. Third, self-established environmental performance evaluation system [4,27]. As no authoritative database has been found to fully publish the carbon emission data of listed companies in China, the carbon emission data disclosed by listed companies are not comprehensive, and the carbon emission data calculated lack authority and cannot be verified for authenticity. Therefore, carbon emission data are not used as a measure of environmental performance. The self-built environmental performance evaluation system has some subjectivity, while the third-party evaluation system is widely used, and the method is relatively mature and objective. Therefore, this study uses the third-party scores to measure environmental performance. There are many third-party scoring methods used, such as the environmental performance score of Hua Zheng, Bloomberg, and CNRDS. The Hua Zheng environmental performance scores are widely used, which provide the widest coverage and most comprehensive data for environmental performance in China [11,13,44,47,48]. The environmental assessment system includes five first-level indicators of climate change, resource utilization, environmental pollution, environmental friendliness and environmental management, including 17 second-level indicators such as greenhouse gas emission, sustainable certification and so on. Considering the Chinese characteristics and the specific practical experience of enterprises, the evaluation is relatively objective and comprehensive. Therefore, the baseline regression uses the environmental performance scores of Hua Zheng as explained variable, and standardizes it. In order to avoid the special effects of the Hua Zheng environmental performance scores, environmental performance scores of CNRDS and environmental disclosure scores of Bloomberg ESG Disclosure Scores are used as robustness tests.

3.2.2. Explanatory Variable

Different listed companies issue green bonds at different times and are subject to policy shocks at different times. Therefore, the time-varying DID model is used in this study [3,11,16,17]. In the model, the explanatory variable is the interaction term of Treat and Post (Treat × Post). Treat is an individual dummy variable that takes 1 if the firm has issued green bonds in 2016 and after, since China’s green bond policy was implemented in 2016, and 0 otherwise. Post is a time dummy variable that takes 1 for the year in which the firm issues a green bond and the year after, 0 for the year prior to the green bond issuance, and 0 for firms that do not issue a green bond. If a firm issues green bonds multiple times, the first issuance time is used [4]. In order to capture the impact of green bond issuance over time, Formula (6) in the robustness test is used to examine the dynamic effect of green bonds on environmental performance [11,49].

3.2.3. Mediating Variable

The mediating variables include R&D investment and green innovation. The time from patent application to authorization is as long as 1 to 2 years in China, and the number of patent applications can better represent the innovation ability of enterprises [38]. The larger the number of green patent applications, the stronger the green innovation ability of enterprises. Green innovation (GP) is represented by the number of green patent applications plus 1 and then logarithm [6,41,42], to deal with the zero value of the green patent.

3.2.4. Moderator Variable

Moderator variables include media attention and analyst attention. Media Attention (ME) is measured by the total number of online news headlines about the company in a year. The greater the number of news, the greater the media attention. The logarithm of the number of media attention plus 1 is taken as the moderating variable. Analyst Attention (AN) represents the number of analysts or teams that have covered the company in a year. The number of a team is 1, and its members are not counted separately. The number of analysts or teams is added by 1 and the logarithm is taken as the moderating variable. The analyst attention data are obtained from CSMAR database.

3.2.5. Control Variable

Based on the studies of other researchers [3,4,6,11,16], the following control variables reflecting the characteristics of corporate operation and corporate governance are selected. They include: (1) Enterprise Size (Size): the natural logarithm of total annual assets; (2) Proportion of fixed assets (FIXED): the net fixed assets divided by the total assets; (3) Asset-liability ratio (Lev): total liabilities divided by total assets at the end of the year; (4) Rate of return on total assets (ROA): net profit divided by the average of total assets; (5) Total assets growth rate (Asset Growth): the total assets of the current year divided by the total assets of the previous year minus 1; (6) Tobin Q: (market value of tradable shares + number of non-tradable shares × net asset value per share + book value of liabilities)/total assets; (7) Company establishment years (Firm Age): ln(current year − year of company establishment + 1); (8) The number of directors (Board): the natural logarithm of the number of directors.

3.3. Model

3.3.1. Baseline Regression

According to other scholars’ research on green bonds [3,6,11,16], the benchmark regression uses the time-varying DID model, as shown in Equation (1). DID is a quasi-natural experimental method often used to explore the effect of policies. In this study, DID model is used to study the impact of green bonds issued by enterprises on environmental performance. Since the research data contain both time and individual dimension, which belong to panel data, in order to eliminate the influence of potential unobserved variables, fixed effect model is used in DID to solve the endogeneity problem caused by unobserved variables. E_it is used to measure the environmental performance of each company in year t and is standardized for regression. Treat_i is the variable that distinguishes the treatment group from the control group. Since China’s green bond policy was implemented in 2016, if the company has issued green bonds in 2016 and later, the variable is 1, otherwise 0. Post_it is the time dummy variable, 1 is taken for the year in which the enterprise issues green bonds and the years after issuance, 0 is taken for the year before the issuance of green bonds, and 0 is also taken for companies that have not issued green bonds. If the enterprise issues green bonds multiple times, the first issue time shall prevail [4]. Controls_it is the control variable for the model, µ_i is the firm fixed effect, δ_t is the year fixed effect, σ_i is the industry fixed effect, ε_it is the residual term, α₀ is the constant term, and the model uses robust standard error. α₁ represents the impact of issuing green bonds on corporate environmental performance. If the coefficient is significantly positive, it indicates that issuing green bonds can significantly improve the environmental performance. The main assumptions of DID model is the parallel trend hypothesis, that is, the environmental performance of the treatment and control group have the same trend before the policy occurs. Therefore, the parallel trend test is carried out in the subsequent robustness test to verify the rationality of the model.

E_it = α₀ + α₁ Treat_i × Post_it + α₂ Controls_it + µ_i + δ_t + σ_i + ε_it

(1)

3.3.2. Model for Mediating Effects

To perform mechanism analysis and verify hypothesis 2, Formulas (2)–(4) are used to examine the mediating effect of R&D and green innovation. The mechanism test adopts the three-step method to test the mediating effect [50]. The control variables and fixed effects are consistent with the baseline regression. In Formulas (2)–(4), Mediator _it represents the mediating variable of the year t of the i enterprise, Treat_i × Post_it is the interaction term, the meaning of which is consistent with the baseline model, ε_it is the residual term, and the model uses robust standard error. If β₁, 𝛶 ₁, ρ₁, ρ₂ are significant, there are partial mediating effects.

E_it = β₀ + β₁Treat_i × Post_it + β₂Controls_it + µ_i + δ_t + σ_i + ε_it

(2)

Mediator_it = 𝛶₀ + 𝛶₁ Treat_i × Post_it + 𝛶₂ Controls_it + µ_i + δ_t + σ_i + ε_it

(3)

E_it = ρ₀ + ρ₁ Treat_i × Post_it + ρ₂ Mediator_it + ρ₃ Controls_it + µ_i + δ_t + σ_i + ε_it

(4)

3.3.3. Model for Moderating Effects

To perform mechanism analysis and verify hypothesis 3 and 4, Formula (5) is used to examine the moderating effect of media attention and analyst attention [51]. In Formula (5), Moderator_it represents the moderator variables for the year t of the i firm, Treat_i × Post_it is the interaction term, the meaning of which is consistent with the baseline regression, control variables and fixed effects are the same as those of the baseline model. ε_it is the residual term, and the model uses robust standard error. If θ₃ is significantly positive, there is a positive moderating effect.

E_it = θ₀ + θ₁ Treat_i × Post_it + θ₂ Moderator_it + θ₃ Treat_i × Post_it × Moderator_it + θ₄ Controls_it + µ_i + δ_t + σ_i + ε_it

(5)

4. Results

4.1. Descriptive Statistics

Table 1. Descriptive statistics.

Variable	N	Mean	p50	SD	Min	Max
E	18,194	61.04	60.99	7.209	45.72	80.15
Treat × Post	18,194	0.0130	0	0.115	0	1
Size	18,194	22.42	22.21	1.371	19.16	26.44
FIXED	18,194	0.218	0.184	0.161	0.00200	0.698
Lev	18,194	0.451	0.447	0.203	0.0640	0.934
ROA	18,194	0.0370	0.0350	0.0620	−0.239	0.223
AssetGrowth	18,194	0.164	0.0900	0.315	−0.323	2.009
TobinQ	18,194	1.954	1.548	1.254	0.832	8.434
FirmAge	18,194	2.966	2.996	0.306	1.946	3.555
Board	18,194	2.115	2.197	0.195	1.609	2.639

is descriptive statistics of the data after winsorization, in which the mean value of environmental performance (E) is 61.04 points, indicating that most companies perform well in environmental performance. The standard deviation of E is 7.209, revealing that environmental performance of companies is uneven. The mean value of Treat × Post is 0.0130, suggesting that the number of companies issuing green bonds is small. The average rate of return on total assets (ROA) is about 3.7%, with a standard deviation of 0.062, showing that most companies are profitable, with little variation.

Table 2. VIF test.

Variable	VIF	1/VIF
Treat × Post	1.04	0.9582
Size	1.84	0.5424
FIXED	1.05	0.9569
Lev	1.75	0.5704
ROA	1.41	0.7071
AssetGrowth	1.15	0.8677
TobinQ	1.26	0.7913
FirmAge	1.06	0.9459
Board	1.10	0.9092
Mean VIF	1.30

is VIF test results for multicollinearity, all VIF values are less than 10 and near 1, the model has no multicollinearity problem.

4.2. Baseline Regression Result

As can be seen from

Table 3. Baseline regression results.

Variables	(1) E	(2) E	(3) E	(4) E	(5) E	(6) E
Treat × Post	0.863 ***	0.734 ***	0.276 ***	0.431 ***	0.288 ***	0.253 ***
	(0.0627)	(0.0591)	(0.0676)	(0.0601)	(0.0599)	(0.0679)
Size				0.220 ***	0.175 ***	0.165 ***
				(0.00711)	(0.00718)	(0.0178)
FIXED				0.0655	0.261 ***	0.0756
				(0.0551)	(0.0434)	(0.0829)
Lev				0.0149	0.00154	−0.170 ***
				(0.0447)	(0.0457)	(0.0618)
ROA				0.551 ***	0.687 ***	−0.284 **
				(0.133)	(0.139)	(0.123)
AssetGrowth				−0.245 ***	−0.208 ***	−0.123 ***
				(0.0224)	(0.0234)	(0.0186)
TobinQ				−0.0625 ***	−0.0565 ***	−0.0105
				(0.00587)	(0.00635)	(0.00689)
FirmAge				0.124 ***	−0.0836 ***	0.156
				(0.0222)	(0.0248)	(0.136)
Board				−0.00651	−0.0459	−0.0373
				(0.0364)	(0.0382)	(0.0529)
Constant	−0.0192 ***	−0.0175 **	−0.0146 ***	−5.172 ***	−3.528 ***	−3.999 ***
	(0.00706)	(0.00735)	(0.00486)	(0.171)	(0.168)	(0.539)
Industry	Yes	No	Yes	Yes	No	Yes
Year	No	Yes	Yes	No	Yes	Yes
Firm	No	No	Yes	No	No	Yes
Observations	18,194	18,194	17,860	18,194	18,194	17,860
R-squared	0.107	0.032	0.653	0.204	0.110	0.657

Robust standard errors in parentheses, *** p < 0.01, ** p < 0.05.

, the issuance of green bonds has a significant promoting effect on the environmental performance of enterprises. Columns (1), (2) and (3) are the regression results without adding control variables, while columns (4), (5) and (6) add control variables. Columns (1) and (4) only control the industry fixed effect, columns (2) and (5) only control the year fixed effect, and columns (3) and (6) control the year, industry and individual firm fixed effect at the same time. The regression coefficients of interaction term are all significantly positive at the 1% level, indicating that the issuance of green bonds has played its due role and effectively improved the environmental performance of enterprises, and hypothesis 1 has been verified. According to the interaction coefficient of column (6), the average treatment effect is 25.3%, indicate that the environmental performance of companies after issuing green bonds enhance by 25.3% [16,21]. The baseline regression results support the signal theory, that is, companies do signal subsequent improvements of environmental performance following the issuance of green bonds [3]. However, the results are inconsistent with the greenwashing theory [22], which suggests that companies issuing green bond have no intention and practical action to improve their environmental performance.

4.3. Robustness Test

4.3.1. Alternative Measurement of Environmental Performance

As environmental scores of CNRDS and Bloomberg are also widely used, environmental scores of CNRDS and Bloomberg are standardized and used as explained variables for baseline regression. The regression results are shown in columns (1), (2), (3) and (4) in

Table 4. Results with alternative measurement of environmental performance.

Variables	(1) CNRDS	(2) CNRDS	(3) Bloomberg	(4) Bloomberg
Treat × Post	0.191 ***	0.177 ***	0.301 ***	0.289 ***
	(0.0630)	(0.0625)	(0.0902)	(0.0870)
Controls	No	Yes	No	Yes
Industry	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
Firm	Yes	Yes	Yes	Yes
Observations	17,955	17,955	6079	6079
R-squared	0.725	0.726	0.719	0.727

Robust standard errors in parentheses, *** p < 0.01.

respectively, all of which are significantly positive at the 1% level.

4.3.2. Parallel Trend Test

The assumption of DID method is to meet the parallel trend hypothesis, that is, the environmental performance of the treatment group and control group has the same trend before the implementation of the green bond policy, and there is no significant difference. The year dummy variable is added to the baseline regression model to detect whether there is a significant difference in the years before the policy [11,49]. Where, Before is the dummy variable before the green bond issuance, and Current is the dummy variable in the year when issuing. After is the dummy variable after the issuance. In the first year before the issuance, Before1_it of treatment group equals 1, otherwise equals 0. In the first year after the issuance, After1_it of treatment group is equal to 1, otherwise equals 0, and so on. The year dummy variables of the control group are all 0, and other variables are the same as the baseline regression. The specific model is shown in Formula (6), and the results are shown in Figure 1.

E_it = 𝜆₀ + 𝜆₁ Before4_it + 𝜆₂ Before3_it + 𝜆₃ Before2_it + 𝜆₄ Before1_it + 𝜆₅ Current_it + 𝜆₆ After1_it + 𝜆₇ After2_it + 𝜆₈ After3_it + 𝜆₉ After4_it + 𝜆₁₀ Controls_it + µ_i + δ_t + σ_i + ε_it

(6)

As can be seen from Figure 1, the environmental performance trends of the treatment and control group are the same before the issuance of green bonds, and there is no significant difference, which meet the conditions of DID model. From the results of dynamic effect, the environmental performance of the treatment and control group begin to show significant differences in the year of green bond issuance, indicating that the policy began to have an impact in the year of issuance, and the impact lasts until the fourth year.

4.3.3. Placebo Test

Different enterprises issue green bonds at different times. In this study, enterprises are randomly selected to build treatment groups, randomly select virtual policy time, and interaction items are generated and put into baseline regression [11,16]. Control variables and fixed effects are the same as those of the baseline model, and 500 random samples are taken. The p-value distribution and estimated coefficient of 500 samples are shown in Figure 2, where the red dot represents the p-value distribution and the red line represents the kernel density of the estimated coefficient. Most of the p-values are greater than 0.1, indicating that the results of random sampling are not significant. The real estimated coefficient is 0.253, and all coefficients of sampling 500 are far from the real estimated coefficient, indicating that the results of baseline regression pass the placebo test and are robust.

4.3.4. PSM

The adoption of PSM-DID alleviates the endogenous problems of green bond issuance and corporate environmental performance. The difference between the treatment and control group is reduced by screening the samples of the control group through nearest neighbor matching, so as to alleviate the selection bias. The nearest neighbor matching of 1:20 is used to perform logit regression, and the control variables which are the same with baseline regression are used as covariates, including enterprise size, fixed assets ratio, asset-liability ratio, ROA, total assets growth rate, Tobin Q, company establishment years, and the number of directors. It can be seen from

Table 5. Differences before and after PSM matching.

	Unmatched	Mean		%Reduct	t-Test		V(T)/V(C)
Variable	Matched	Treated	Control	%Bias	|Bias|	t
Size	U	24.448	22.33	159.000		43.870	0.000
	M	24.442	24.405	2.800	98.300	0.520	0.603
FIXED	U	0.32413	0.21289	58.100		18.830	0.000
	M	0.32314	0.33156	−4.400	92.400	−0.740	0.457
Lev	U	0.5901	0.44468	82.300		19.510	0.000
	M	0.59001	0.58866	0.800	99.100	0.160	0.869
ROA	U	0.03379	0.037	−6.200		−1.400	0.162
	M	0.03373	0.03628	−4.900	20.500	−1.140	0.252
Asset Growth	U	0.17243	0.16333	3.100		0.780	0.435
	M	0.17098	0.19135	−6.900	−124.000	−1.190	0.234
Tobin Q	U	1.2663	1.9838	−74.300		−15.560	0.000
	M	1.2669	1.2591	0.800	98.900	0.300	0.765
Firm Age	U	3.014	2.9644	16.700		4.380	0.000
	M	3.0139	3.0237	−3.300	80.400	−0.610	0.544
Board	U	2.2056	2.1113	46.400		13.100	0.000
	M	2.2049	2.1918	6.400	86.100	1.230	0.218

that the deviation of standardized mean value of covariates after matching is within 10%, and the p-values of all covariates after matching are not significant, indicating that there is no significant difference between the covariates of the treatment and control group after matching, and the matching effect is good. After PSM matching, the treatment group maintain 85 listed companies, the control group is reduced to 1548 listed companies, with a total of 6250 samples. The matched samples are used for regression, and the results were shown in

Table 6. PSM-DID regression results.

Variables	(1) E	(2) E	(3) E	(4) E	(5) E	(6) E
Treat × Post	0.704 ***	0.487 ***	0.261 ***	0.431 ***	0.308 ***	0.242 ***
	(0.0631)	(0.0611)	(0.0722)	(0.0607)	(0.0607)	(0.0728)
Controls	No	No	No	Yes	Yes	Yes
Industry	Yes	No	Yes	Yes	No	Yes
Year	No	Yes	Yes	No	Yes	Yes
Firm	No	No	Yes	No	No	Yes
Observations	6247	6250	5739	6247	6250	5739
R-squared	0.114	0.042	0.668	0.213	0.103	0.671

Robust standard errors in parentheses, *** p < 0.01.

. The coefficient of interaction terms is still positive, indicating that the results are robust.

4.3.5. Research Model Transitioning

The baseline regression uses fixed effects model, which is now replaced by pooled regression. The control variables are the same as the baseline regression, and heteroscedasticity-robust standard errors are used. The regression results are shown in

Table 7. Pooled regression results.

Variables	(1) E	(2) E	(3) E	(4) E	(5) E	(6) E
Treat × Post	0.863 ***	0.734 ***	0.276 ***	0.431 ***	0.288 ***	0.259 ***
	(0.102)	(0.0967)	(0.0991)	(0.0972)	(0.102)	(0.0981)
Controls	No	No	No	Yes	Yes	Yes
Firm	No	No	Yes	No	No	Yes
Industry	Yes	No	Yes	Yes	No	Yes
Year	No	Yes	Yes	No	Yes	Yes
Observations	18,194	18,194	18,194	18,194	18,194	18,194
R-squared	0.107	0.032	0.656	0.204	0.110	0.659

Robust standard errors in parentheses, *** p < 0.01.

. After adding control variables, industry, year and firm fixed effects, the coefficient of the interaction term is still significantly positive, which further verifies the robustness of our results.

4.4. Mechanism Analysis

4.4.1. Mediating Effect Analysis

The results of the mediating effect are shown in

Table 8. Mediating effect analysis results.

Variables	(1) E	(2) RD	(3) E	(4) E	(5) GP	(6) E
RD			0.0368 ***
			(0.0111)
GP						0.0400 ***
						(0.00814)
Treat × Post	0.227 ***	0.397 ***	0.212 ***	0.252 ***	0.346 ***	0.238 ***
	(0.0748)	(0.0866)	(0.0747)	(0.0679)	(0.0712)	(0.0682)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Industry	Yes	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes	Yes
Firm	Yes	Yes	Yes	Yes	Yes	Yes
Observations	15,709	15,709	15,709	17,858	17,858	17,858
R-squared	0.647	0.896	0.648	0.657	0.781	0.658

Robust standard errors in parentheses, *** p < 0.01.

. Columns (1), (2) and (3) report the results of mediating effect of R&D. Column (2) shows the impact of green bond issuance on R&D expenditure of enterprises, the interaction terms of columns (2) implies that the R&D expenditure of enterprises increase by 39.7% after issuing green bonds (𝛶₁ = 0.397, p < 0.01). Column 3 shows the influence of R&D expenditure on environmental performance, the coefficient of R&D conveys that the advancement of R&D can enhance environmental performance (ρ₂ = 0.0368, p < 0.01). Since β₁, 𝛶 ₁, ρ₁, ρ₂ are all significantly positive at the 1% level, R&D expenditure plays partial mediating effects. The results indicate that the issuance of green bonds improves environmental performance through increasing the R&D expenditure. Hence, hypothesis 2 is partially verified. The issuance of green bonds has raised sufficient funds for environmentally friendly projects, which will stimulate R&D investment, enterprises can improve production technology, reduce environmental pollution and improve environmental performance through R&D activities.

Columns (4), (5) and (6) in Table 8 report the results of mediating effect of green innovation. In column 5, the coefficient of interaction term is 0.346 and significantly positive at 1% level, implies that the number of green patents increase by 34.6% after issuing green bonds. In column (6), the coefficient of GP is 0.0400 and significantly positive at 1% level, conveys that the advancement of green innovation can enhance environmental performance. Since β₁, 𝛶 ₁, ρ₁, ρ₂ are all significantly positive at the 1% level, green innovation plays partial mediating effects. The results indicate that the issuance of green bonds improves environmental performance through strengthen green innovation ability of enterprises. Hence, hypothesis 2 is supported. To sum up, after issuing green bonds, enterprises will increase R&D investment, convert R&D investment into green patents, form competition barriers, enhance green innovation ability, improve core competitiveness, and strengthen environmental protection.

4.4.2. Moderating Effect Analysis

Columns (1), (2), and (3) in

Table 9. Moderating effect results.

Variables	(1) E	(2) E	(3) E	(4) E	(5) E	(6) E
Treat × Post	0.0521	−0.148	−0.150	0.238	0.0479	−0.0718
	(0.236)	(0.226)	(0.222)	(0.159)	(0.164)	(0.187)
ME	0.0847 ***	0.0175 **	−0.0169 **
	(0.00783)	(0.00760)	(0.00777)
Treat × Post × ME	0.146 ***	0.110 **	0.0910 *
	(0.0496)	(0.0478)	(0.0497)
AN				0.114 ***	0.0328 ***	0.00453
				(0.0103)	(0.0120)	(0.0122)
Treat × Post × AN				0.203 ***	0.153 **	0.155 **
				(0.0648)	(0.0655)	(0.0757)
Controls	No	Yes	Yes	No	Yes	Yes
Industry	Yes	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes	Yes
Firm	No	No	Yes	No	No	Yes
Observations	18,064	18,064	17,803	11,760	11,760	11,367
R-squared	0.138	0.220	0.657	0.153	0.228	0.694

Robust standard errors in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1.

are the results of media attention as a moderating variable. Column (1) has no control variables, and columns (2) and (3) have control variables. Columns (1) and (2) control industry and year fixed effects, while column (3) controls industry, year and individual fixed effects. The coefficient of Treat × Post × ME in column (1) is significantly positive at the level of 1%, which in column (2) is significantly positive at the level of 5%, and in column (3) is significantly positive at the level of 10%, indicating that media attention plays a positive moderating role. The more media attention enterprises receive, the greater the improvement in environmental performance will be from issuing green bonds. and hypothesis 3 is verified. After issuing green bonds, enterprises gain more media attention, which plays the role of external supervision. In the face of increasing media attention, in order to maintain their good social image and reputation, enterprises are more motivated to take practical actions to fulfill their commitment to environmental protection and improve environmental performance.

Columns (4), (5), and (6) in Table 9 are the results that analyst attention as moderating variables. Column (4) has no control variables, and columns (5) and (6) have control variables. Columns (4) and (5) control industry and year fixed effects, while column (6) controls industry, year and individual fixed effects. The coefficient of Treat × Post × AN in columns (4) is significantly positive at the level of 1%, which in columns (5) and (6) are significantly positive at the level of 5%, indicating that analyst attention plays a positive moderating role. Analyst attention positively reinforce the impact of green bond issuance on corporate environmental performance, and hypothesis 4 is verified. After companies issue green bond, analyst attention and continuous tracking act as external monitors, forcing companies to fulfill commitment to improve environmental performance to meet stakeholder expectations.

4.5. Heterogeneity Analysis

The above studies have confirmed that issuing green bonds can significantly improve the environmental performance of enterprises, but enterprises with different characteristics of issuing green bonds have different effects on the environmental performance, therefore, this part analyzes the heterogeneous effects of issuing green bonds on the environmental performance of enterprises from the characteristics of enterprises.

4.5.1. Type of Enterprises

According to Yin Meiqun et al. (2018)’s classification method of industries [52], Labor-intensive enterprises include coal mining, metal ore mining, most manufacturing industries, construction, transportation, real estate, scientific research and technical services, and public facilities management, which are belong to the industries supported by China’s green bond policy. Therefore, for labor-intensive industries, with the support of policies, enterprises are more motivated to issue green bonds and realize their commitment to the environment, which will have more significant effects on improving environmental performance.

Labor-intensive enterprises are those that use a large amount of labor in the production process, and labor consumption accounts for a relatively large proportion of the operating costs of such enterprises. According to Yin Meiqun et al. (2018)’s classification method of industries, enterprises are classified into labor-intensive industries, capital-intensive industries and technology-intensive industries [52]. According to “2012 China Securities Regulatory Commission Sectoral Guidelines for Listed companies”, which uses sum of squares method of deviation in cluster analysis to classify industries, those with a higher proportion of fixed assets (net fixed assets/average total assets) are capital-intensive industries, those with a higher proportion of R&D expenditures (R&D expenditures/payable employees’ remuneration) are technology-intensive industries, and the rest are labor-intensive industries. According to the industries supported by the 2015 and 2021 editions of “Green Bond Support Project Catalog” in China [5,9], most labor-intensive industries are included. The sample firms are divided into labor-intensive firms and non-labor-intensive firms according to the labor-intensive industry code, and column (1) in

Table 10. Heterogeneity analysis.

Variables	(1) E	(2) E	(3) E	(4) E	(5) E	(6) E
Treat × Post	0.547 ***	0.100	0.244 ***	0.185	0.233 ***	0.257
	(0.121)	(0.0850)	(0.0719)	(0.388)	(0.0800)	(0.219)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Industry	Yes	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes	Yes
Firm	Yes	Yes	Yes	Yes	Yes	Yes
Observations	4527	13,321	8932	8674	8841	8444
R-squared	0.669	0.664	0.650	0.690	0.670	0.695

Robust standard errors in parentheses, *** p < 0.01.

shows the regression results of labor-intensive firms, and column (2) shows the results of non-labor-intensive firms. The interaction term in column (1) is significantly positive at the 1% level and the interaction term in column (2) turns out to be insignificant. The regression results indicate that in labor-intensive firms, green bond issuance has a significant effect on environmental performance, while in non-labor-intensive firms, green bond issuance has a non-significant effect on environmental performance.

4.5.2. Enterprise Scale

Large scale enterprises receive more social attention and bear greater public pressure, so they pay more attention to social reputation [53]. It is found that the larger the enterprise scale, the larger the enterprise ESG investment and the better the ESG performance [54]. Therefore, large enterprises issuing green bonds pay more attention to the environmental protection commitment to society and will have more significant effects on the promotion of environmental performance. The samples are divided into two groups according to the median size of enterprises. Larger than the median size, the group of large-scale enterprises; otherwise, the group of small-scale enterprises. Column (3) in Table 10 shows the regression results of large-scale enterprises, and column (4) shows the results of small-scale enterprises. The interaction terms of column (3) are significantly positive at the 1% level, and the interaction terms of column (4) are not significant. The regression results show that the issuance of green bonds in large-scale enterprises has a significant impact on environmental performance, while that in small-scale enterprises has no significant impact on environmental performance.

4.5.3. Government Subsidy

Companies issuing green bonds are more likely to receive government subsidies, for these companies’ sustainable development strategies respond to the government’s call to take social responsibility to protect the environment. And then, companies with government subsidies have eased financing constraints and are more motivated to invest funds in environmental improvement projects, respond to national environmental protection policies, and improve corporate environmental performance [11].

Many countries offer subsidies, tax breaks and other policies to companies that issue green bonds [2]. At present, China’s green bond subsidies are mainly related policies issued by provincial governments. For example, the Implementation Rules of Jiangsu Province’s Green Bond Discount Policy (Trial) points out that 30% of the actual interest paid by non-financial enterprises that successfully issue green bonds will be discounted for 2 years to reduce the financing cost of enterprises. Hence, the samples are divided into two groups according to the median of government subsidies received by enterprises. Greater than the median is the group with high government subsidies, and vice versa is the group with low government subsidies. Column (5) in Table 10 is the regression results of the group with high government subsidies, and column (6) is the low government subsidies group. The interaction term of column (5) is significantly positive at the 1% level, which of column (6) is not significant. The results imply that the green bond issuance of enterprises with high government subsidies has a significant effect on environmental performance, while the low government subsidies group has no significant effect.

5. Conclusions

This paper studies the impact of green bond issuance on environmental performance. The results show that the environmental performance of Chinese listed companies is significantly improved after issuing green bonds, and the results remain significant after a series of robustness tests. From the internal influence mechanism, R&D investment and green innovation play partially mediating roles. In terms of the external influence mechanism, media attention and analyst attention play positive moderating roles. The heterogeneity analysis demonstrates that the improvement of green bond on environmental performance only significant in labor-intensive enterprises, larger enterprises, and enterprises receiving more government subsidies. Overall, the research results are consistent with the conclusions of Flammer (2021), Fatica and Panzica (2021), Wu Shinong et al. (2022) etc., further supporting the signal theory of green bonds rather than the greenwashing theory.

This finding not only further confirms the applicability of signal theory in the field of green finance, but also emphasizes how green bonds can effectively improve environmental performance in the context of China’s rapidly developing green bond market, which promote existing theoretical frameworks in the field of green finance. Specifically, green bonds as a means of financing, the enterprises will invest the raised funds in the R&D activities of green friendly projects, especially the development of green technologies and products, and take practical and effective measures to improve the environmental performance of the enterprise by improving its green innovation capability. Media attention and analyst attention play roles in external monitoring and regulation, forcing companies to take their environmental commitments seriously and positively promote environmental performance. Especially for labor-intensive enterprises, large-scale enterprises, and enterprises enjoying more government subsidies, issuing green bonds can not only effectively improve environmental performance, but also obtain more market and policy support.

This study puts forward the following suggestions for the development of China’s green bond market: First, the government can consider further expanding the industries and fields supported by green bonds, and provide different policy tools such as tax incentives and subsidies according to the development of financial markets in different regions, so as to reduce the cost of issuing green bonds and stimulate enterprises to issue green bonds for environmentally friendly projects. Second, promote the international convergence of green bond policies, establish and improve the third-party certification of green bonds, improve the information disclosure mechanism of green bonds, strengthen the supervision of the quality of information disclosure of the issuers, and put forward higher requirements for the disclosure content and frequency of disclosure. Third, strengthen the supervision of the whole process of green funds raised by green bonds, and through third-party supervision and audit intervention, ensure that green funds are invested in research and development activities conducive to ecological environmental protection, green innovation and other behaviors, so as to achieve the purpose of improving the environment.

The limitations of this study are as follows: First, the scope and depth of this study are limited. Due to the difficulty and complexity in obtaining and quantifying some variables, it is impossible to consider all factors affecting corporate environmental performance and green bond issuance, such as government policy changes or international environmental agreements. Future studies can consider how to quantify policy factors and incorporate them into the research model. Secondly, we propose the mechanism of green bond issuance on environmental performance from both internal and external aspects. However, the mechanism research is not rich enough at present, and the role of other potential mechanisms can be studied in the future.