The real impact of corporate green bonds
A study on the environmental and financial impact of corporate green bonds. by: Wiebe Groen1
University of Groningen Faculty of Economics and Business
MSc Finance
Supervisor: J.V. Tinang Nzesseu Date: June 28, 2019
Abstract
This study examines the real impact of corporate green bonds. Due to the fact that there are no binding regulations on green bonds, there is a fear that green bonds do not deliver on their promise of being green. This study therefore examines whether corporate green bonds do really have an environmental impact, despite the lack of these regulations. Furthermore, this study examines if there is a yield premium on corporate green bonds and if the environmental performance of bonds affects this premium. This study finds that issuing a green bond results in a significant increase of 5.56 ESG score points after two years and that green bonds have a significant negative yield premium of 8.2 basis points. Most importantly, this study shows that the negative premium is more prevalent for better environmentally performing bonds. These results emphasize that corporate green bonds do deliver on their promise, and that investors can and do differentiate on the environmental performance between green bonds.
Keywords: Green bonds, Impact investing, sustainable finance, ESG portfolios Word count: 12404
2 1. Introduction
One of the biggest challenges for humankind in the 21st century is climate change. The goal, agreed to at the Paris climate conference in 2015, to limit global warming to below 2 degrees Celsius is one step in overcoming this big challenge (United Nations, 2015). In order to meet this requirement, several initiatives have been started to reduce pollution and create a more sustainable environment. One of these initiatives are the Sustainable Development Goals, set by the United Nations. To achieve the Sustainable Development Goals by 2030, investments of approximately 5-7 trillion US dollars per annum are needed (UNDP, 2015). This huge investment requirement needs innovative and specific green instruments to be successful. One instrument which can help in this transition towards a more sustainable investment environment are so-called green bonds.
Bonds are issued by companies to finance projects or investments. Green bonds are similar to ‘conventional’ bonds except that the firm uses the proceeds for specific green projects. Examples of such projects are renewable energy projects, green buildings, and technology aimed at reducing emissions. These green bonds show promising prospects.
However, whether or not green bond issuing firms deliver on their promise to finance green projects is questionable. So-called ‘greenwashing’ with green bonds is becoming more and more of a problem (WWF, 2016). Greenwashing is the practice of making a false or misleading claim about a projects’ environmental promise (Kirchhoff, 2000). For firms, greenwashing is tempting since nowadays it is critical for firms to have a positive social image (Liao, 2018). Especially green bonds are vulnerable for this greenwashing since due to lack of public governance and legal enforcement, firms are able to issue green bonds without actually financing green projects with the proceeds.
The current market is dependent on non-binding regulations from private institutions, such as the Green Bond Principles and the Climate Bonds Initiative, to classify the greenness of a green bond. Due to possibility of greenwashing and the lack of universal binding regulations, the real impact of green bonds remains unclear.
To solve this ambiguity on green bonds, both issuers and investors demand making clear universal regulations about what a green bond should deliver before it can be labelled as a green bond (WWF, 2016; Tang and Zhang, 2018; Flammer, 2018). In the meantime, before these regulations are made, there is a need for both investors and issuers to grasp the effectiveness of green bonds. This paper will, therefore, try to give a better view on the effectiveness of green bonds.
3 whether green bonds have a yield premium by matching and comparing green bonds to similar conventional bonds from the same issuer. At last, this paper studies the relationship between the environmental impact and the premium on corporate green bonds by using environmental performance portfolios. Green bonds are assigned to in either a ‘High’ or ‘Low’ portfolio based on their environmental impact.
The three analyses in this paper do depend on each other. This paper starts with a dataset of all corporate green bonds issued between 1-1-2013 and 31-12-2018. Of these bonds, this paper uses corresponding ESG data of the green bond firms to estimate the environmental impact. Due to the limited availability of ESG data, multiple bonds are omitted from the dataset. Thereafter, this paper uses the green bonds that are still in the dataset to estimate the green bond premium. Again, due to the limited availability of bond-specific data, multiple bonds are omitted. Finally, the bonds that are still in the dataset are used to estimate the relationship between the environmental impact and the premium on corporate green bonds.
This paper finds that firms’ environmental scores significantly increase after green bond issuance, especially in the long run. Furthermore, this paper finds that corporate green bonds have a significant negative yield premium compared to their conventional matches. Moreover, this paper finds that the bonds in ‘High’ portfolios have a significantly lower yield premium compared to bonds in ‘Low’ portfolios.
The contribution of this study is that it provides larger-scale evidence on the environmental impact of green bonds, following Flammer (2018). Furthermore, it enhances the findings of Zerbib (2019) that corporate green bonds have a negative yield premium, by using a more recent dataset. Most importantly, this paper introduces a new empirical framework to differentiate between the actual impact of green bonds. The results of this paper show that investors do differentiate between green bonds on their use of proceeds. Green bonds with a real impact have a lower premium in the secondary market compared to green bonds that have no or not a noticeable impact. This indicates that investors prefer green bonds that actually have an impact on the environment over green bonds that do not have a noticeable effect on the environment.
4 2. Literature review.
This literature review is organized in-line with the three-fold analysis of this paper. This section starts by introducing green bonds and describing the current state of the green bond market. Thereafter, section 2.2 provides the literature and formulates the hypothesis for the environmental impact analysis in section 3. Next, section 2.3 provides the empirical background and formulates the hypothesis for the estimation of the yield premium in section 4. Finally, section 2.4 concludes by relating corporate social responsibility (CSR) with financial performance and formulates the hypothesis for the analysis in section 5.
2.1 Green bonds
Issuing green bonds is a method for firms to express their commitment to corporate social responsibility (Shishlov, Morel, and Cochran, 2016). Green bonds differentiate from conventional bonds by the type of funding project they are connected to. Green bonds are any type of bond instrument where the proceeds will be exclusively applied to finance or re-finance eligible green projects (UNDP, 2019). Multiple papers argue that green bonds still lack a universal binding definition (OECD, 2016; Tripathy, 2017; Ehlers and Packer, 2017). However, there are two commonly used and widely accepted standards. The Green Bond Principles (GBP) and the Climate Bonds Initiative (CBI). Green bonds are defined by the Green Bonds Principle as bonds that fund projects which contribute to environmental objectives such as, but not limited to, climate change mitigation, climate change adaptation, natural resources conservation, biodiversity conservation, and pollution prevention (ICMA, 2018). The CBI runs a certification scheme to verify and keep track of green bonds. The GBP are incorporated into the CBI scheme (Climate Bonds initiative, 2017).
An important difference between green bonds is that some are labelled and some are unlabelled (UNDP, 2019). A labelled green bond is certified by an independent verifier which uses the CBI scheme. However, the certification process can be costly, and sometimes withhold a company from letting its bond getting certified (Jun, Kaminker, Kidney, and Pfaff, 2016). Unlabelled bonds do not have such certification but are called green because the issuing companies define them as green bond (Barclays, 2015).
5 are considering green bonds as serious investment vehicles. Figure 1 displays the growth of the corporate green bond market.
2.2 Environmental impact green bond
This section outlines the benefits and concerns related to green bonds to indicate why empirical evidence on the environmental impact of corporate green bonds is needed.
Since the United Nations Climate Change Conference in 2015, an extensive amount of firms agreed to fight the current state of the environment. Multiple papers argue that green bonds are a perfect financial tool to do so (Shishlov et al., 2016 and Tripathy, 2017). Tripathy (2017), for example, mentions that the issuance of a green bond enhances a company’s reputation and signals its sustainability commitments. Furthermore, Shishlov et al. (2016), argues that issuance of a green bond encourages a company to measure the contribution they are making to tackle climate change. The process of issuing green bonds improves the sustainability awareness in a company since the evaluation of an environmental investment project helps to build the ties between sustainability and financial departments.
However, as mentioned in section 2.1, the green bond market still lacks universal standards and regulations. This results in that firms can self-label green bonds without having to meet any requirements. This lack of regulations leads to a variety of issues and concerns in the green bond market. The AFM (2019), for example, argues that due to the lack of regulations firms issue green bonds without even knowing how the issuing process and the use of proceeds works. Furthermore, there is not enough transparency on how the proceeds of green bonds are used. This leads to greenwashing, companies misusing the funds they get from green bonds. The existence of greenwashing and the lack of transparency lowers the
6 integrity and thus the credibility of green bonds. This lower credibility is also underlined by the WWF (2016). WWF (2016) states that growth is restrained due to such uncertainties as greenwashing. All these uncertainties create a strong demand for standards. Various papers have, therefore, called upon creating a uniform standard and a viable way to assess the environmental impact on green bond (WWF, 2016; Tang and Zhang, 2018; Flammer, 2018). Nonetheless, these papers only raise the concern that greenwashing could exist in the green bond market. To my knowledge, there is no existing literature which proves that this greenwashing prevails on a large scale and Flammer (2018) is currently the only paper that examines the environmental effectiveness of green bonds. These above-presented benefits and concerns in the current green bond market lead to the first hypothesis of this paper: Hypothesis 1: Green bonds have a significant effect on the environment and therefore succeed in their purpose.
2.3 Financial impact green bonds
This section provides an overview of the literature related to the pricing of green bonds, to form the empirical background for the premium analysis in section 4.
Various papers underline the financial benefits of issuing a green bond instead of a normal bond. First of all, Gianfrate and Peri (2019) show that the issuance of green bonds attracts a more diverse investor base, compared to the issuance of conventional bonds. In their paper, they find that there is an increasing amount of institutional investors that are decarbonizing their portfolios and directing their funds towards green bonds, due to the growing threat of climate change. Furthermore, Tang and Zhang (2019) show comparable results and they argue that this redirection of funds is enlarged both by increased media exposure and because of national regulations, such as tax incentives. On top of that, the issuance of green bonds tends to be heavily oversubscribed in the primary market (Cowan, 2017; DuPont, Levitt, and Bilmes, 2015). Harrison and Boulle (2017) show, for example, that European green bonds have an average oversubscription of 2.3 times and US green bonds of 2.8 times.
7 bond. For the secondary market, Barclays (2015), Bloomberg (2017), Hachenberg and Schiereck (2018), and Zerbib (2019) studied the difference in yield between a normal and a green bond in the secondary market.
For the secondary market, all these papers found a significant negative yield difference between green and conventional bonds. This yield difference is defined as the green bond premium, which can either be positive or negative.
Significant determinants of this green premium are the credit rating, sector, type of issuing entity and ESG (Zerbib, 2019; Hachenberg and Schiereck, 2018). Furthermore, Zerbib (2019) and Hachenberg and Schiereck (2018) also specify the type of issuing entity: either government or corporate bonds.
However, only Hachenberg and Schiereck (2018) study whether the environmental impact of a green bond is a determinant of yield premium. Hachenberg and Schiereck (2018) test this relationship by creating a dummy variable which takes the value one if the issuer has at least one ESG rating from Sustainalystics or RobecoSam and otherwise zero.
I consider this approach as too narrow since their method does not measure the actual environmental impact of a green bond. As I discussed in section 2.2, the environmental impact is an important factor for issuers and investors to issue or invest in green bonds.
Based on the above-presented gap in the literature, this paper studies whether the actual environmental impact of a green bond affects the yield premium of a green bond. In order to do this, this paper first estimates the green bond premium, based on the methodology introduced by Zerbib (2019). The title of this section, financial impact, represents thus the impact of the label ‘green’ on the yield of a bond. The second hypothesis of this paper is:
Hypothesis 2: Pro-environmental preferences of investors translate into a yield premium of corporate green bonds compared to conventional bonds.
2.4 Relationship between the environmental and financial impact of a corporate green bond. Since this paper evaluates whether the environmental performance of green bonds will have an effect on the yield difference of green bonds, I first provide the literature on the relationship between CSR and the financial performance of a firm. Various papers have covered this relationship, however, whether this relationship is neutral, positive or negative is still a debate.
8 not significantly reduce the risk. They imply that CSR only leads to higher capital costs, and thus to a decrease in financial performance.
Kempf and Osthoff (2007) find, however, that buying stocks with a high socially responsible rating and selling stocks with slow responsible rating results in abnormal returns of 8.7% per year. They perform their analysis by constructing a portfolio which ranks firms based on KLD data. Furthermore, Oikonomou, Brooks, and Pavelin (2014) investigate the impact of CSR on the pricing of corporate debt and credit ratings. They find that high corporate social performance results in lower risk spreads and higher rating on corporate bonds. The most significant differences are found for the highest and lowest rated bonds. Both the results of Kempf and Osthoff (2007) and Oikonomou et al. (2014) show that being a socially responsible firm can have a positive impact on stock returns and corporate bonds.
Emerson (2003) argues that CSR does not imply that there has to be a trade-off between social and financial returns, based on the blended value propositions. This is empirically shown by Menz (2010), which studies the relation between the value of corporate green bonds and the CSR performance of the issuing companies. Menz (2010) hypothesizes that the social responsibility of firms affect the risk premium of bonds these firms issue. The results of his study show, however, that the risk premium on bonds from socially responsible firms is not significantly different from non-socially responsible firms.
Summing up, based on the above-mentioned papers, there is evidence that CSR affects financial performance. Especially the study of Oikonomou et al. (2014) shows that CSR affect corporate bonds. This suggests that if a green bond is enhancing the environmental performance of the company, it affects the pricing of this bond.
Based on the concerns of the actual impact of corporate green bonds, which I present in section 2.2, and the above-discussed relationship between CSR and the financial performance of a firm, I argue that both for firms and investors, it is useful to know whether it is worthwhile to effectively use the proceeds of green bonds or that greenwashing pays off. Therefore, the last hypothesis of this paper is:
9 3. Environmental impact
This section covers the estimation of the environmental impact of corporate green bonds. Section 3.1 gives an overview of the data and describes the matching method that I use to study the environmental impact. Section 3.2 describes the methodology and section 3.3 presents and discusses the results of the environmental impact of corporate green bonds. 3.1 Data
This section presents the data that is used to measure the environmental impact of corporate green bonds. The corporate green bonds and the variables that are used in the analysis are described in section 3.1.1. Thereafter, section 3.1.2 describes the firm-level matching method is described, which is necessary to isolate the environmental impact of corporate green bonds. 3.1.1. Green bonds and environmental measures
This paper uses the entire corporate green bond universe, of which the data is provided by the CBI. This data consists of 590 unique listed corporate green bonds, in a time span of 1-1-2013 till 31-03-2019. The first restriction of our data is that only green bonds issued by corporates can be used. This paper analyses the impact of green bonds on environmental scores, which are not available for governmental organizations. Next, since this paper analyses the effect of green bonds on firm-level outcomes, only unique firm entries in a given year can be used in the dataset. Some firms issued multiple green bonds in a given year, so these excess entries were eliminated. This resulted in a data set of 244 corporate bonds. The summary statistics are presented in table 1.
Next, I use firm environmental scores to measure the environmental impact of green bonds. Unfortunately, there are no quantitative environmental variables specific for green bonds available yet. Therefore, it is important to note that this paper tries to measure the impact of a green bond by relating it to the environmental scores of the issuing firm.
10
Table 1. Summary statistics corporate green bonds
This table provides summary statistics for all corporate green bonds issued between 1-1-2013 and 31-12-2018.
Total Median Mean
No. of green bonds 580 - -
Issuing firms 244 - -
Amount issued in $ millions 167,911.16 92.72 289.50
By country China 35,695.90 337.10 524.94 United States 21,772.61 37.50 329.89 France 19,996.60 14.24 161.26 Spain 12,664.97 876.15 745.00 Sweden 10,141.22 56.33 92.19 Germany 10,088.98 568.65 504.45 The Netherlands 7,855.26 564.15 714.11 Japan 6,826.56 89.00 195.04 UK 6,300.31 384.04 484.64 Other 40,429.46 192.93 296.90
Maturity in years of green bonds - 3 4.03
3.1.2. Firm-level matching
This study measures how green bonds affect the environmental performance of firms. That is, the ESG variables are analysed following green bond issuance. An issue that arises by performing this analysis is endogeneity. Unobserved variables might affect the possible relationship between the issuance of green bonds and the ESG scores. For example, I could possibly find a positive relationship between the issuance of a green bond and an increase of the environmental pillar score. A logical conclusion would be that, due to good use of the proceeds of green bonds, the environmental rating has increased. However, this increase could also have occurred due to other unobserved factors that were not accounted for.
In a perfect world, this endogeneity issue would be addressed by using a randomized controlled trial (RCT). This would eliminate all other determinants of ESG score and makes sure that both the green bond issuing firms (hereafter, treated firms) and the non-green bond issuing firms (hereafter, control firms) are identical. However, in the setting of green bonds a RCT is not possible.
A second best alternative to estimate the impact of green bonds, is introduced by Flammer (2018), who uses a matching method. Flammer (2018) matches green bonds issuing firms with similar firms that did not issue a green bond, and estimates the environmental impact of green bonds by using a difference-in-difference estimation between the two groups. This paper uses a similar approach.
11 The matching restrictions are as follows. First of all, for the control group, only firms that issued conventional bonds are considered. Second, the control firm must operate in the same country as the treated firm. Third, the control firm must operate in the same industry as the treated firm, based on the SIC code. Matching based on these three restrictions ensures that the treated and the control firms face the same market conditions. Then, from all the remaining control firm candidates, the nearest neighbour, which is the firm with the lowest Mahalanobis distance, is selected. The Mahalanobis distance is calculated based on the variables: ESG, return on assets, total assets, leverage and the change of each of these four variables from two years to one year before the year of green bond issuance. The treated and control firms are matched in the year before issuance of the green bond.
The Mahalanobis distance is calculated by:
𝑀𝐷 = ((𝑋𝑖− 𝑋𝑗)𝑡𝑆−1(𝑋𝑖− 𝑋𝑗))
1
2 (1)
Where MD is the Mahalanobis distance between a treated and control firm, Xi is a (8*1) vector containing the 8 matching variables of a green bond firm, Xj is a (8*1) vector containing
the 8 matching variables of a treated firm, superscript t indicates that the vector is transposed and S-1 is the inverse covariance matrix of the 8 matching variables. This method is based on
methodology introduced by Mahalanobis (1936) and is also used by Flammer(2018) to match green bond firms to non-green bond firms.
Matching on these firms characteristics ensures that firms have similar environmental performance, similar growth opportunities, similar profitability, and the same level of access to capital markets. Note that I want to find a unique matched control firm for each treated firm. This ensures that each treated firm has a comparable control firm, instead of that overall the treated firms are equal to control firms. Furthermore, I need a unique matched control firm for the analysis in section 5.
12 Table 2. Summary statistics firm data
This table provides summary statistics of the treated and control firms for all variables that were used in the matching process. The variables are measured in the year before issuance, and the delta variables are measured as the difference between t-2 and t-1, where t is the year of issuance. ROA represents return on assets, leverage is the ratio of debt to total assets. The P-value of the difference in means is presented to show that the variables of the green bond firm and their corresponding matched firm do not significantly differ from each other.
N Mean Std. Dev P-value
ESG score Green bond firm 80 63.14 17.77 0.876
Matched firm 80 62.72 16.47
ROA Green bond firm 80 2.711 0.305 0.611
Matched firm 80 2.925 0.287
Log(assets) Green bond firm 80 26.852 0.302 0.195
Matched firm 80 26.291 0.307
Leverage Green bond firm 80 0.267 0.019 0.994
Matched firm 80 0.267 0.022
∆ESG Green bond firm 80 0.657 4.902 0.110
Matched firm 80 1.910 4.517
∆ROA Green bond firm 80 0.255 5.215 0.647
Matched firm 80 -0.159 0.886
∆Log(assets) Green bond firm 80 0.073 0.114 0.664
Matched firm 80 0.065 0.124
∆Leverage Green bond firm 80 0.001 0.044 0.868
Matched firm 80 0.002 0.036
Note: *p<0.1, **p<0.05, ***p<0.01
3.2 Methodology
This methodology assesses how corporate green bonds affect the environmental performance of the issuing company, for this I use a difference-in-difference (DiD) regression. This approach compares the difference of the outcome variable before versus after the issuance of a green bond of the treatment group with the difference on the outcome variable before versus after the issuance of a green bond of the control group. I use firm-year observations from 2013-2019. The regression is specified as:
13 where 𝑦𝑖,𝑡 is the outcome variable (e.g., ESG score, environmental score, environmental innovation score), 𝛼𝑖 stands for firm fixed effects, 𝛼𝑐 ∗ 𝛼𝑡 for country by year fixed effects, 𝛼𝑠∗ 𝛼𝑡 for industry by year fixed effects, 𝐺𝐵𝑖,𝑡 is a dummy variable which is one if the firm has issued a green bond in year t and zero otherwise, and 𝜀𝑖,𝑡 is the error term. Subscripts i,t,c and s indexes respectively firms, years, countries and SIC industry codes. 𝛽 is the coefficient of interest and measures the additional effect on the outcome variable due to the issuance of a green bond.
To get a better insight in the time-effect of green bond issuance, I extend this specification by using different treatment dummies for which I replace the 𝐺𝐵𝑖,𝑡 dummy in equation (2). I run the regression while using a dummies of which one tests whether there are pre-trends in the data. 𝐺𝐵𝑖,𝑡−1 is equal to one in the year before the green bond issuance and zero otherwise. If there are pre-trends in the data this indicates that the matching is biased and that the treated group already has an increase in the outcome variable despite the issuance of the green bond. Furthermore, I use a dummy which tests for the short term effect of green bond issuance. 𝐺𝐵𝑖,𝑡+1 is equal to one in the year after the green bond issuance and zero otherwise. At last, I use a dummy which tests for the long term effect of green bond issuance. 𝐺𝐵𝑖,𝑡+2 is equal to one in the second and subsequent years after green bond issuance and zero otherwise.
3.3 Results
14 Table 3. Environmental performance green bond.
This table presents the 𝛽 coefficients of the difference-in-difference(DiD) equation 𝑦𝑖,𝑡= 𝛼𝑖+ 𝛼𝑐∗ 𝛼𝑡+𝛼𝑠∗ 𝛼𝑡++ 𝛽 ∗ 𝐺𝐵𝑖,𝑡+ 𝜀𝑖,𝑡. I
regress the DiD estimation multiple times with the different dependent variables and different dummies. Each column (1)-(6) thus shows the 𝛽 coefficient(s) of a regression. ESG, Environmental pillar and innovation score present the dependent variables of the regression. Green bond is a dummy variable, which is one if a firm issued a green bond. Pre-issue is a dummy variable, which is one if a firm issued a green bond in the year before issuance. Short-term is a dummy variable, which is one if a firm issued a green bond in the year after issuance. Long-term is a dummy variable, which is equal to one if a firm issued a green bond in two and subsequent years after issuance. The full sample of treated and control firms is used, which ranges from 2013-2019. N reports the number of observations and SE are reported in parentheses and are clustered at the industry level.
ESG score Environmental pillar score Environmental innovation score (1) (2) (3) (4) (5) (6) Green bond 2.409*** (0.497) 3.089*** (0.762) 3.737*** (1.163)
Green bond (pre-issue) 0.426
(0.350)
0.805 (0.793)
1.923* (0.969)
Green bond (short-term) -0.738**
(0.305)
-0.864* (0.481)
-0.077 (1.022)
Green bond (long-term) 5.560***
(0.740)
7.910*** (0.133)
7.082*** (2.352
Firm fixed effects Yes Yes Yes Yes Yes Yes
Country-year fixed effects Yes Yes Yes Yes Yes Yes
Industry-year fixed effects Yes Yes Yes Yes Yes Yes
N 1,234 1,234 1,234 1,234 1,234 1,234
R2 0.65 0.66 0.66 0.67 0.45 0.47
Note: *p<0.1, **p<0.05, ***p<0.01
In columns (1)-(6), I find overall that green bonds have a positive significant effect on the environmental behaviour of firms. Below I will describe for each dependent variable in detail the results and their implications.
ESG score. Green bonds have a positive significant effect in the year of issuance. Column (1) shows that the ESG score increases with 2.4 points. Given the ESG mean of 63.14 of green bond firms, presented in table 2, this is an increase of 3.80%. More interesting, however, are the results of column (2). This column shows the effect of green bond issuance in different time periods. The pre-issuance year is the same year as the year in which the control firms were matched. The coefficient is not significantly different from zero, and thus indicates that the treated firms are similar to the control firms. In the short run there is a slightly negative significant effect on the ESG rating. In the long run, however, the green bond has a positive significant effect on ESG score. The long run is after two and subsequent years following green bond issuance. The ESG score increases with 5.56 points which is an increase of 8.80%, given the ESG mean of 63.14 of green bond firms, presented in table 2.
15 the most interesting, since it is likely that a green investment would take some time to have a substantial effect. From this positive long-term effect I conclude that the issuance of a green bond increases the ESG score of a company considerably.
Environmental pillar score. Next to the ESG score, I also test whether green bonds have an effect on the environmental pillar score of a company. This environmental score is part of the overall ESG score. With the total ESG score the environmental implications of the effect on green bonds are somewhat diluted due to the fact that ESG score not only consists of environmental, but also social and governance scores. From table 3, I find that the environmental pillar score increases significantly with 7.91 points in the long run. This is an increase of 11.66%, given the environmental pillar score mean of 67.81 for green bond firms, presented in table A.4 in the appendix. Note that in the short run there is a minor decrease, similar to the results of the ESG score . These results indicate that the issuance of a green bond has a positive effect on the environmental pillar score, especially in the long run.
Environmental innovation score. The innovation score indicates the capacity of a company to create new environmental technologies and to which degree it reduces its environmental costs through innovations. The results in column (5) and (6) in table 3 indicate that green bonds substantially improve the environmental innovation score. In the long run, the innovation score of a company significantly increases with 7.082 points. This is an increase of 9.29%, given the environmental innovation score mean of 76.26 for green bond firms, presented in table A.4 in the appendix. These results indicate that the issuance of a green bond has a positive effect on the environmental innovation score, especially in the long run.
Overall, the results indicate that green bonds have a significant positive effect on the environmental scores of a firm. This indicates that a firm, after issuance of a green bond, improves its environmental behaviour. The first hypothesis of this paper, introduced in the literature review stated:
Hypothesis 1: Green bonds have a significant effect on the environment and therefore succeed in their purpose.
The results are in line with this hypothesis. I conclude that green bonds do have a positive effect on the environment. Therefore, greenwashing does not prevail on a larger scale in the green bond market.
16 4. Financial impact
This section covers the financial impact of corporate green bonds. In this paper the financial impact is regarded as the impact on the yield due to the green label of a bond. Section 4.1 gives an overview of the data and describes the matching method which is used to study the financial impact. Section 4.2 describes the methodology and section 4.3 presents and discusses the results.
4.1. Data
This section uses the same 80 corporate green bonds that were used in the estimation of section 3. Important to note is that section 3 uses corresponding firm data of the green bonds and this section uses data of the green bonds itself.
As mentioned earlier, this paper analyses the yield spread to study whether corporate green bonds are priced differently compared to exactly similar, but not green, bonds. Normally, an estimation on the yield value is done by determining all independent variables that could possibly explain this yield value of a bond. However, such an estimation has various drawbacks. For example, it is complicated to determine which variables exactly affect a bond yield. Furthermore, including too many variables could result in lack of robustness or collinearity in the date. Most importantly, however, is the fact that this paper analyses a small upcoming market, and thus data is not yet widely available.
Since this paper studies whether only the label green of a bond affects the yield, a more superior approach can be used. I match each green bond with two similar conventional bonds from the same issuer to analyse the difference in yield. The next section explains the matching method.
4.1.1. Bond-level matching
Matching each green bond with two similar bonds from the same issuer results in the fact that most factors that explain a bond yield are similar across each triplet. The differences left in each matched bond triplet are that the maturities and liquidities are not equal and that one bond is a green and the other two are conventional bonds. The matching method this paper uses follows the method of Zerbib (2019), which also studies the green bond premium. Helwege (2014) uses a similar method to analyse the cost of liquidity. Furthermore, Kreander (2005), Renneboog (2008) and Bauer (2005) use a similar matching method to analyse excess returns of ethical funds over conventional funds. These studies show that this approach is robust and appropriate to analyse differences across bonds.
17 the closest maturity to the green bond are selected. To limit the difference in maturity, only bonds which have a maturity that is within the range of two years compared to the green bonds are considered.
To limit the difference in liquidity, only bonds with an issue amount which is within the range of four times the green bond’s issue amount are considered. On top of that, for a bond to be eligible to match, the issue date must be within the time span of six years of the green bond’s issue date.
To estimate the yield spread, the triplets should have identical characteristics except for their liquidity. The liquidity is the independent variable in the estimation of the yield spread. Due to the above described matching criteria, each matched triplet now has the same characteristics, except for their maturity and liquidity. Table A.5 in the appendix presents an overview of each green bond and its two matched conventional bonds.
To fully eliminate the maturity difference, I create a synthetic bond from the two matched conventional bonds. From the Thomson Reuters Eikon database, the daily ask yields for all bonds are extracted from the issue date until December 31, 2018. I use the ask yield in this analysis since Zerbib (2019) argues that it gives the best representation of the issuers supply and investors’ demand of corporate green bonds. If at a given day, at least one ask yield of a triplet is missing, I also remove the ask yields of the other two matched bonds on that day. The synthetic bond yield is then created by either linear interpolation or extrapolation2 of the two conventional bonds at the maturity date of the green bond. Figure A.1 in the appendix describes in detail the method of interpolation. This synthetic conventional bond now has the same characteristics as the green bond. The only difference now left between the synthetic bond and the green bond is the liquidity.
Since the green bond and its matched synthetic bond are now exactly the same, aside from the liquidity. The difference between their yields can be explained by the effect of their liquidity difference plus a green bond premium. The yield difference between the synthetic and the green bond is defined as:
∆𝑦𝑖,𝑡 = 𝑦𝑖,𝑡𝐺𝐵− 𝑦
𝑖,𝑡𝑆𝐵 (3)
Where 𝑦𝑖,𝑡𝐺𝐵 is the ask yield of green bond i at time t, and 𝑦𝑖,𝑡𝑆𝐵 is the ask yield of the synthetic bond i at time t. The yield difference is the dependent variable in the analysis. The data ranges from January 1st 2013 until December 31st, 2018. Due to the matching restrictions and the requirement that for each day all three yields of a triplet must be available,
2 If both conventional matched bonds have a lower/ higher maturity I use extrapolation. However, in the majority
18 multiple green bonds dropped out. This leads from the 80 green bonds in section 3 to 45 green bonds. The descriptive statistics of the remaining sample are presented in table 4.
Table 4. Descriptive statistics bond matching
Descriptive statistics of the bond matching sample. This table gives the distribution of several variables of interest, to compare the accuracy of the matching process. It presents the ask yields of the green bonds and the matched conventional bonds 1 (CB1) and 2 (CB2). Furthermore, it shows the distribution of the ask yield of the synthetic bond, calculated based on CB1 and CB2. The yield difference is the difference between the ask yield of the green bond and the synthetic bond. The maturities are noted in years left from 31 December 2018.
No. of Green bonds 45 No. of match bonds 1 45 No. of match bonds 2 45
Min 1st Quart Median Mean 3rd Quart Max.
No. of days per bond 15 218 305 402 541 833
Ask yield GB -0.141 0.419 1.277 1.776 3.006 9.319
Ask yield Syn. bond -0.089 0.743 1.290 1.905 2.987 8.684
Ask yield CB1 -0.083 0.415 1.013 1.856 3.415 8.719 Ask yield CB2 -0.084 0.740 1.490 1.952 3.007 8.629 Yield difference ∆𝑦𝑖,𝑡 -0.089 -0.176 -0.061 -0.127 0.003 0.785 GB Maturity. 0.893 2.611 3.490 4.329 5.797 9.244 CB1 Maturity 1.501 2.071 2.901 3.879 4.562 9.616 CB2 Maturity. 1.578 3.052 4.340 5.194 8.288 9.940
I also need a variable which is an approximation of the liquidity of each bond. This variable will be constructed in line with the methodology of Zerbib (2019). Various papers indicate that approximations based on trading activity, such as the bid-ask (BA) spread, have the highest quality (see, e.g., Fong, Holden, and Trzcinka, 2017; Beber, Brandt, and Kavajecz 2009; and Chen, Lesmond and Wei, 2007). This paper will therefore use the bid-ask spread as liquidity approximation. For each triplet of bonds, the bid and ask price of each day are extracted from the Thomson Reuters Eikon database. Again, if for one bond a value at a given day is missing, the value for the two other bonds are deleted. The synthetic BA spread is constructed by the BA spreads of the two corresponding conventional bonds, in line with the green bonds maturity. The BA spread of the synthetic bond is the distance-weighted average of the BA spread of the conventional bonds 1 and 2 to the green bonds maturity. This is specified as:
𝑑1 = 𝑀𝐺𝐵 − 𝑀𝐶𝐵1 (4)
19 𝐵𝐴𝑖,𝑡𝑆𝐵= 𝑑2 𝑑1 + 𝑑2𝐵𝐴𝑖,𝑡𝐶𝐵1+ 𝑑1 𝑑1 + 𝑑2𝐵𝐴𝐶𝐵2𝑖,𝑡 (6) ∆𝐵𝐴𝑖,𝑡 = 𝐵𝐴𝑖,𝑡𝐺𝐵− 𝐵𝐴 𝑖,𝑡 𝑆𝐵 (7)
Where 𝑀𝐺𝐵 is the green bond maturity, 𝑀𝐶𝐵1 the conventional bond 1 maturity, and 𝑀𝐶𝐵2 the conventional bond 2 maturity. 𝐵𝐴
𝑖,𝑡
𝑆𝐵 is the bid-ask spread of the synthetic bond, 𝐵𝐴𝑖,𝑡𝐺𝐵 is the bid-ask spread of the green bond, and ∆𝐵𝐴
𝑖,𝑡 is the difference between these two. The descriptive statistics of the liquidity approximation can be found in table 5.
Table 5. descriptive statistics liquidity.
This table presents the distribution of the bid-ask spread for the green bond and conventional bond 1 (CB1) and conventional bond 2 (CB2). Furthermore, it shows the bid-ask spread distribution of the synthetic bond, calculated based on CB1 and CB2. The variable of interest is Δ BA, which stands for the difference between the bid-ask spread of the green bond and the synthetic bond at a specific point in time. The Δ BA is the liquidity approximation of the bonds and is used as an independent variable in equation ∆𝑦𝑖,𝑡= 𝑝𝑖+ 𝛽∆𝐵𝐴𝑖,𝑡+ 𝜀𝑖,𝑡(8)
Min 1st Quart Median Mean 3rd Quart. Max.
BA spread GB -0.143 0.124 0.148 0.226 0.296 1.680
BA spread Syn. bond -0.012 0.147 0.213 0.236 0.291 1.464
BA spread CB1 -0.049 0.149 0.207 0.228 0.260 1.725
BA spread CB2 -0.175 0.142 0.214 0.280 0.387 1.662
Δ BA -0.395 -0.084 -0.020 -0.010 0.020 1.440
These statistics show that Δ BA is concentrated around zero, and thus indicates that the matching criteria on liquidity were sufficient for a good approximation. The final dataset that is used in the estimation has a total of 10,867 data points, consisting of the BA spread difference and the yield difference between the green and synthetic bonds. Noteworthy is that table 5 indicates that the minimum values of the BA spreads in the sample are negative. This is counterintuitive since ask prices are normally higher than bid prices, but can be explained by a crossed market3 situation. Only 7 out of the 10,867 data points have a negative BA spread, so it is unlikely that it contaminates the estimation.
4.2. Methodology
This section describes how the green bond premium is estimated. The data section described the bond matching method, which ensured that both the maturity difference and all unobservable factors common to the matched bonds were removed. The only difference left
3 Temporary situation where bid prices are higher than ask prices. Source:
20 between the green and synthetic bond is the liquidity, which is specified as the BA spread, ∆𝐵𝐴𝑖,𝑡.
I use a bond fixed-effect (FE) panel regression with the yield difference as dependent variable and the liquidity difference as independent variable. This regression is specified as:
∆𝑦𝑖,𝑡 = 𝑝𝑖 + 𝛽∆𝐵𝐴𝑖,𝑡+ 𝜀𝑖,𝑡 (8)
Where ∆𝑦𝑖,𝑡 is the yield difference between the matched bonds, ∆𝐵𝐴𝑖,𝑡 is the liquidity difference between the matched bonds, and 𝜀𝑖,𝑡 the error term. Subscripts i and t stand for bond and time in days, respectively. The variable of interest is the fixed effect estimator of regression (8), which is defined as 𝑝𝑖.
I use a FE regression based on several arguments. The economic argument is that the goal of this paper is to estimate the yield premium on green bonds. Therefore, I need to estimate the bond-specific and time-invariant unobserved effect, which is not affected by information of the other bonds in the sample. Besides that, the FE regression is more efficient compared to pooled OLS and Random Effect based on a Hausman test. Results of the Hausman tests can be found in table A.6 and A.7 in the appendix.
4.3. Results
21 Table 6. Liquidity effect
This table reports the results of regression ∆𝑦𝑖,𝑡= 𝑝𝑖+ 𝛽∆𝐵𝐴𝑖,𝑡+ 𝜀𝑖,𝑡. ∆𝑦𝑖,𝑡 is the yield difference between the green
and synthetic bond. ∆𝐵𝐴𝑖,𝑡 is the bid-ask spread difference between the green and synthetic bond.
Furthermore, this table presents the standard fixed effects regression to estimate the effect of the bid-ask differential on the yield differential. Furthermore, to account for serial correlation, this table presents the results of a Newey-West and Beck-Katz robust standard errors regression. The standard errors are reported in parentheses. R2 and N are the same for
each regression. ∆𝑦𝑖,𝑡 Fixed effects Newey-West robust std. error Beck-Katz robust std. error. ∆𝑩𝑨𝒊,𝒕 -0.107*** (0.017) -0.107*** (0.015) -0.107*** (0.020) R2 0.009 N 10,687 Note: *p<0.1, **p<0.05, ***p<0.01
Table 6 shows that a 1 basis point increase in the BA spread difference result in a 10.7 basis points decrease in the yield difference. The Newey-west and Beck-Katz regressions attain the same results, with slightly different standard errors. Since this paper studies the green bond premium, I will not go into further detail on the results of table 6. The fixed effects estimator of regression (8) represent the green bond premium, and table 7 summarizes its distribution.
Table 7. Distribution green bond premium
This table reports the distribution of the green bond premiums across the sample. The green bond premium is the fixed effect estimator of the regression ∆𝑦𝑖,𝑡= 𝑝𝑖+ 𝛽∆𝐵𝐴𝑖,𝑡+ 𝜀𝑖,𝑡. ∆𝑦𝑖,𝑡 is the yield difference between the green and
synthetic bond. ∆𝐵𝐴𝑖,𝑡 is the bid-ask spread difference between the green and synthetic bond .
𝑝
Min. 1st Qtr. Median Mean 3rd Qtr. Max
-0.829 -0.153 -0.046 -0.082 0.016 0.663
The green bond premium in the sample ranges from -82.9 basis point to 66.3 basis point, where the median is -4.6 basis points and the mean is -8.2 basis points. Figure A.2 in the appendix presents the distribution of the green bond premium in a graph.
22 distributed4 and Zerbib (2019) suggests to use this kind test for this analysis. Table 8 presents the mean, median and significance of the green bond premium per characteristic.
As also earlier presented in table 7, the median of the entire sample is -8.2 basis point and is significant different from zero at a 99% confidence level. Furthermore, table 8 shows that green bonds from financial firms have a negative premium of -12 basis points, which is also significantly different from zero. Besides that, green bonds from real estate firms, AA- rated bonds and not rated bonds show negative premium which are significantly different from zero. For all other characteristics I do not find any significance. This is likely due to the low amount of bonds in each subsample.
The results provide evidence that there is a negative premium on corporate green bonds. Relating the results to the second hypothesis in this paper:
Hypothesis 2: Pro-environmental preferences of investors translate into a yield premium of corporate green bonds compared to conventional bonds.
Since this paper finds a significant negative premium on corporate green bonds, the results are in line with this hypothesis. A negative premium implies that investors in the secondary market are willing to accept a lower yield on corporate green bonds compared to corporate conventional bonds. These results suggest that investors have a strong preference for corporate bonds with a pro-environmental factor.
23 Table 8. Green bond premiums across several segments.
This table reports how the estimated green bond premium is distributed across either the industries, ratings or currencies. The green bond premium is estimated as is the fixed effect estimator of
∆𝑦𝑖,𝑡= 𝑝𝑖+ 𝛽∆𝐵𝐴𝑖,𝑡+ 𝜀𝑖,𝑡 (8) using the entire sample, and this table shows how it is subdivided across
different segments. For each segment the medians and the means of the premiums are reported. Furthermore, I use a Wilcoxon signed rank test to test whether the medians are significantly different from zero. The null hypothesis is thus p̂ =0.
24 5. Environmental performance portfolios
This section relates the environmental impact of corporate green bonds to the yield premium on corporate green bonds. If the environmental performance does affect the green bond premium, it implies that investors can differentiate between green bonds that do or do not deliver on their promise of being green. This section uses the data of section 3 and 4. Therefore, I do not cover the data in this section. Furthermore, this section uses the green bond premium estimated in section 4. First, I discuss the methodology I use to assess the environmental performance. After that, I present the results.
5.1. Methodology
This paper studies whether environmental performance affects the green bond premium. To determine this relationship, I assign green bonds to portfolios based on the environmental performance of their issuing firm. The average premium of the bonds in a portfolio can then be related to the nature of the firms in this portfolio. This section thus combines the studies outlined in section 3 and 4.
Kempf and Osthoff (2007), Statman and Glushkov (2009), and Halbritter and Dorfleitner (2015) use a similar empirical framework, which uses portfolios to determine the relationship between ESG and financial performance. These papers, however, compose the portfolios by solely looking at ESG scores and do not capture the increase of ESG and the differential effect between a treated and control firm. Therefore, this paper uses their method but I adjust it in such a way that it captures the effect of green bond issuance on the environmental performance.
The portfolios in this paper are constructed as follows, each year firms are ranked from best to worst based on their environmental performance. Based on this ranking, firms are assigned to either a ‘High’ or ‘Low’ portfolio. The environmental performance is measured as the difference between the change over time in ESG score of the green bond firm and the change over time in ESG score for their matched non-green bond issuing firm. This is comparable to the difference-in-difference methodology specified in section 3. However, in this section I calculate the difference-in-difference for each specific pair of firms. The environmental performance is thus measured by the following formula:
𝐸𝑃𝑖,𝑡 = (𝐸𝑆𝐺𝑖,𝑡𝐺𝐹− 𝐸𝑆𝐺𝑖,𝑡𝑜𝐺𝐹) − (𝐸𝑆𝐺𝑖,𝑡𝐶𝐹− 𝐸𝑆𝐺𝑖,𝑡0𝐶𝐹) (9)
25 t0 stands for the year of the green bond issuance. Besides portfolios based on ESG score, I create similar portfolios for the environmental pillar and the environmental innovation score. ESG score in formula (9) is then changed with either the environmental pillar or the environmental innovation score.
For each year, I rank all firms based on their environmental performance. Based on the ranking, the 20% best environmental performing firms are assigned to the ‘High’ portfolio and the 20% worst environmental performing firms are assigned to the ‘Low’ portfolio. For ESG, environmental performance and environmental innovation, I create portfolios for each year of which there is data available. The majority of the corporate green bonds were issued in 2016 and later. Therefore, I only construct portfolios for the years 2017 and later since ranking is not possible for earlier years.
This model ensures that the effect of the issuance of green bonds is taken into account, since it measures the ESG score difference between the treated and the control firm. If the green bond has a large impact on ESG score, it can be expected that the ESG score difference between the treated and the control firm increases after issuance. A portfolio therefore presents a green bond premium which can be related to environmental performance, either high or low. Similar to the estimation in section 4, I use a Wilcoxon signed-rank test to test for each portfolio whether the median is significantly different from zero.
To examine whether the results are dependent on the above explained method of the portfolio construction, I also consider variations in the cut-off percentage. Next to the initial cut-off of 20%, I examine a 10% and 50% cut-off limit.
Limitations of this approach are that the robustness of the environmental performance highly depends on the quality of the matching method. The environmental performance of a treated firm that is matched with control firm could be affected by factors other than the green bond issuance. Another limitation is the time effect. The effect of a green bond issuance most likely diminishes over time, thus this method can only be applied in a limited time period. Despite these limitations, this newly introduced empirical method could provide interesting insights for issuers and investors on how the actual environmental impact of green bonds affect their corresponding yields. Other approaches, such as only including a dummy variable if a firm has an ESG score in the estimation of a green bond premium, used by Hachenberg and Schiereck(2018), do not differentiate between how well green bonds use their proceeds.
5.2. Results
26 firm on their environmental score and based on this ranking, bonds are assigned to either a ‘High’ or ‘Low’ portfolio. A portfolio consists thus of a number of bonds, and table 9 presents the aggregate mean and median of these green bond premiums per portfolio.
Table 9. 20% cut-off portfolios
This table presents the medians and means of the green bond premiums of the green bonds that are placed in the High or Low portfolios. The green bond premiums are calculated as the fixed effect estimator of the regression ∆𝑦𝑖,𝑡= 𝑝𝑖+ 𝛽∆𝐵𝐴𝑖,𝑡+ 𝜀𝑖,𝑡 using all bonds in the sample. The High (Low) portfolios consists of the 20% best (worst)
performing green bond firms compared to their matched conventional peer, in terms of either the ESG, environmental pillar or environmental innovation score. The year indicates that the portfolio is constructed and that the premiums are calculated based on data from up and till that year. However, for 2019-portfolios ESG, environmental pillar or environmental innovation scores up and till 2019 are used but the premiums are calculated based on the entire sample which ranges till 31-12-2018. No. of bonds indicates how many bonds were placed in the portfolio. Furthermore, I use a Wilcoxon signed rank test to test whether the medians are significantly different from zero. The null hypothesis is thus p =0
27 Table 9 shows that bonds which are assigned to the ‘High’ ESG portfolios have a significant negative yield premium, this holds for each year. The ‘Low’ ESG portfolios show a smaller negative or even a positive premium, which is not significantly different from zero. Most interesting are the 2019 ESG portfolios. The results I present in section 3 indicate that green bonds have the highest positive effect on ESG in the long-run ( 2+ years). The 2019 portfolios consists thus of green bonds that had the highest environmental impact, since the ranking is based on the longest time period. Table 9 shows that the average green bond premium in the 2019 ESG ‘High’ portfolio is -16.3 basis points compared to 0.3 basis points for the ‘Low’ portfolio. Similar differences between the ‘High’ and ‘Low’ portfolios hold for 2017 and 2018. Turning to the portfolios that are created based on the environmental pillar score, only the ‘High’ portfolios of 2018 and 2019 show a negative green bond premiums with significance. The two 2019 environmental pillar score portfolios show that the negative yield premiums are more prevalent in the ‘High’ portfolios. However, the difference between ‘High’ and ‘Low’ portfolios is smaller compared to the difference between ‘High’ and ‘Low’ of 2019 ESG portfolios.
Finally, the environmental innovations score. None of the medians of the premiums in any portfolio is significantly different from zero. When comparing the premiums of the ‘High’ and ‘Low’ portfolios, 2017 and 2019 show that ‘High’ portfolios have a higher negative premium and 2018 ‘Low’ portfolios have a higher negative premium.
Tables A.11 and A.12 in the appendix present the premiums of portfolios with a 50% and 10% cut-off limit. With the 50% cut-off more portfolios are significantly different from zero. For ESG, the 50% cut-off has same relation as the 20% cut-off. ‘High’ portfolios have lower premiums compared to ‘Low’ portfolios. Differences between ‘High’ and ‘Low’ portfolios of the environmental pillar and innovation scores are, however, not present. More interesting is the 10% cut-off, which shows that all ‘High’ portfolios have a lower premium compared to Low portfolios. However, these results show less significance.
The results presented in tables 9, A.11 and A.12 suggest that ‘High’ portfolios correspond with lower green bond premiums compared to ‘Low’ portfolios. This is true for all portfolios constructed based on ESG score, of which the ‘High’ portfolios are also significantly different from zero. Although, for the environmental pillar and innovation score portfolios, the results are less prevalent .
Finally, recalling the last hypothesis of this paper:
Hypothesis 3: There is a relationship between the environmental impact of a corporate green bond and the premium on this bond.
28 score of their issuing company have a lower premium compared to green bonds which have no or negative results on the ESG score of their issuing company. This implies that investors, on the secondary bond market, are willing to accept a lower return on green bonds that had a real environmental impact. For issuers this implies that it is worthwhile to actually invest in environmental enhancing project when issuing green bonds.
29 6. Conclusion
This paper examined the real impact of corporate green bonds. The objective of this paper was to analyse the impact of corporate green bonds on the environment and to estimate if there is a yield premium on corporate green bonds. Furthermore, this paper tried to evaluate whether the environmental impact of corporate green bonds affects this yield premium on corporate green bonds.
The first conclusion of this study is that the corporate green bonds improve the firms’ environmental footprint. I find that after the issuance of a green bond, firms ESG, environment and environmental innovation scores significantly improve. These findings are similar to findings of Flammer(2018). However, I enhance her findings by using more recent data and using different indicators to estimate the environmental impact. Especially in the long run, I find that the issuance of green bond improve all considered environmental scores. This suggests that green bonds are a powerful tool to fight climate change, and that greenwashing is currently not present.
The second conclusion of this study is that there is a yield difference between a green bond and a similar conventional bond after controlling for the difference in liquidity. I find that there is a significant negative premium on corporate green bonds, indicating that investors do have a pro-environmental preference on the bond market. They are willing to pay a higher price, or in this case, get a lower yield on their investment, in return for an environmental enhancing investment. For issuers, this entails that issuing a green bond is a cheaper way of raising capital compared to issuing a conventional bond. Furthermore, this negative premium shows that investors prefer green bonds over conventional bond, thus for issuers it is an opportunity to broaden their bondholder base. These findings are similar to those of Zerbib (2019).
31 References
AFM, The Netherlands Authority for the Financial Markets., 2019. Position Paper on Green /
Social/Sustainable Bonds. Retrieved from
https://www.afm.nl/~/profmedia/files/onderwerpen/duurzaamheid/position-paper-green-bonds-amf-afm.pdf
Baker, M., Bergstresser, D., Serafeim, G., & Wurgler, J., 2018. Financing the response to climate change: The pricing and ownership of US green bonds (No. w25194). National Bureau of Economic Research.
Barclays., 2015. “The cost of being green.” Credit Res . Retrieved from https://www. environmental-finance.com/assets/files/US _ Credit _ Focus _ The _ Cost _ of _ Being _ Green.pdf
Bauer, R., Koedijk, K. and Otten, R., 2005. International evidence on ethical mutual fund performance and investment style. Journal of Banking & Finance, 29(7), pp.1751-1767. Beber, A., Brandt, M. W., & Kavajecz, K. A., 2008. Flight-to-quality or flight-to-liquidity? Evidence from the euro-area bond market. The Review of Financial Studies, 22(3), 925-957.
Bloomberg., 2017. Investors are willing to pay a ”green”premium. Bloomberg New Energy Finance report .
Chen, L., Lesmond, D. A., & Wei, J., 2007. Corporate yield spreads and bond liquidity. The Journal of Finance, 62(1), 119-149.
Climate Bonds Initiative., 2015. Green Bond Market Roundup. Retrieved from https://www.climatebonds.net/files/files/2015%20GB%20Market%20Roundup%200 3A.pdf
Climate Bonds Initiative., 2017. Post issuance reporting in the green bond market. Retrieved from https://www.climatebonds.net/resources/reports/post-issuance-reporting-green-bond-market-trends-best-practice-june-2017
Climate Bonds Initiative., 2019. 2018 Green Bond Market Summary. Retrieved from https://www.climatebonds.net/files/files/2018%20green%20bond%20market%20hig hlights.pdf
Cochu, A., Glenting, C., Hogg, D., Georgiev, I., Skolina, J., Eisinger, F., ... & Chowdhury, T., 2016. Study on the potential of green bond finance for resource-efficient investments. Report, European Commission.
Cowan, G., 2017. Investors warm to ‘Green bonds’. The wall street Journal.
DuPont, C., Levitt, J., & Bilmes, L., 2015. Green bonds and land conservation: The evolution of a new financing tool.
32 Emerson, J., 2003. The Blended Value Proposition: Integrating Social and Financial Returns.
California Management Review, 45(4), pp. 35-51
Flammer, C., 2018. Corporate Green Bonds. Available at
SSRN:https://ssrn.com/abstract=3125518
Fong, K. Y., Holden, C. W., & Trzcinka, C. A., 2017. What are the best liquidity proxies for global research?. Review of Finance, 21(4), 1355-1401.
Friedman, M., 1970. The social responsibility of business is to increase its profits. New York Times Magazine September 13, 32-33.
Frooman, J., Zietsma, C., & McKnight, B., 2008. There is no good reason not to be good. In ASAC (Vol. 29, No. 29).
Gianfrate, G., & Peri, M., 2019. The Green Advantage: Exploring the Convenience of Issuing Green Bonds. Journal of Cleaner Production.
Hachenberg, B., & Schiereck, D., 2018. Are green bonds priced differently from conventional bonds?. Journal of Asset Management, 19(6), 371-383.
Halbritter, G., & Dorfleitner, G., 2015. The wages of social responsibility—where are they? A critical review of ESG investing. Review of Financial Economics, 26, 25-35.
Harrison, C. & Boulle, B., 2017. Green bond pricing in the primary market
Helwege, J., Huang, J. Z., & Wang, Y., 2014. Liquidity effects in corporate bond spreads. Journal of Banking & Finance, 45, 105-116.
ICMA., 2018. Green Bond principles. Retrieved from https://www.icmagroup.org/green-social-and-sustainability-bonds/green-bond-principles-gbp/
Jun, M., Kaminker, C., Kidney, S., & Pfaff, N., 2016. Green bonds: Country experiences, barriers and options.
Kempf, A., & Osthoff, P., 2007. The effect of socially responsible investing on portfolio performance. European Financial Management, 13(5), 908-922.
Kirchhoff, S., 2000. Green business and blue angels. Environmental and Resource Economics, 15(4), 403-420.
Kreander, N., Gray, R. H., Power, D. M., & Sinclair, C. D., 2005. Evaluating the performance of ethical and non‐ethical funds: a matched pair analysis. Journal of Business Finance & Accounting, 32(7‐8), 1465-1493
Liao, Z., 2018. Environmental policy instruments, environmental innovation and the reputation of enterprises. Journal of cleaner production, 171, 1111-1117.
Mahalanobis, P. C., 1936. On the generalized distance in statistics. National Institute of Science of India.
33 OECD., 2016 Green bonds. Retrieved from
https://www.oecd.org/environment/cc/Green%20bonds%20PP%20[f3]%20[lr].pdf Oikonomou, I., Brooks, C., & Pavelin, S., 2014. The effects of corporate social performance on
the cost of corporate debt and credit ratings. Financial Review, 49(1), 49-75.
Renneboog, L., Ter Horst, J. and Zhang, C., 2008. The price of ethics and stakeholder governance: The performance of socially responsible mutual funds. Journal of corporate finance, 14(3), pp.302-322.
Shishlov, I., Morel, R., & Cochran, I., 2016. Beyond transparency: unlocking the full potential of green bonds. Institute for Climate Economics Report, https://www. i4ce. org/download/unlocking-the-potential-of-green-bonds/(Erişim Tarihi: 20.08. 2017). Statman, M., & Glushkov, D., 2009. The wages of social responsibility. Financial Analysts
Journal, 65(4), 33-46.
Tang, D. Y., & Zhang, Y., 2018. Do shareholders benefit from green bonds?. Journal of Corporate Finance.
Tripathy, A., 2017. “Translating to risk: The legibility of climate change and nature in the green bond market”. Economic Anthropology, 4(2), 239-250.
UNDP., 2015. Business and the SDGs. Retrieved from http://www.sdg-accelerator.org/content/sdg-accelerator/en/home/sdg-presa/SDGbiz.html
UNDP., 2019. Green bonds. Retrieved from
http://www.sdfinance.undp.org/content/sdfinance/en/home/solutions/green-bonds.html#mst-708893653-0
United Nations., 2015. Paris Agreement. Retrieved from https://unfccc.int/process-and-meetings/the-paris-agreement/what-is-the-paris-agreement
Wilcoxon, F., 1945. Individual comparisons by ranking methods. Biometrics bulletin, 1(6), 80-83
WWF., 2016. Green bonds must keep the green promise. Retrieved from http://d2ouvy59p0dg6k.cloudfront.net/downloads/20160609_green_bonds_hd_rep ort.pdf
34 Appendix.
Table A.1 Literature overview
This table summarizes the research and findings on the premium of green bonds.
Author Year Primary/ Secondary market Sample size Yield premium
Barclays 2015 Secondary N.A. -17 bps
Bloomberg 2017 Secondary 12 Not tested
Climate bonds initiative 2017 Primary N.A. Not tested
Ethlers and Packer 2017 Primary 21 -18 bps
Baker et al. 2018 Primary 2083 -7 bps
Hachenberg and Schiereck 2018 Secondary 63 -1 bp
