The effectiveness of corporate green bonds
Bastiaan Bonnema, University of Groningen, June 2019Supervisor: Dr. S. Homroy
Abstract
This thesis examines the implications of corporate green bonds for financial and environmental outcomes as well as the stock market reaction to the announcement of this relatively new sustainable finance tool. I document that corporate green bonds are effective in increasing firms’ overall environmental performances and decreasing their emissions relative for their sizes. I further document that corporate green bonds yield 1) negative announcement returns for single bond issuers; 2) higher announcement returns for highly leveraged firms; 3) decreases in long-term value; 4) improvements in firms’ sizes; and 5) decreases in firms’ levels of debt. Overall, these findings suggest that corporate green bonds are indeed effective in reducing firms’ environmental footprints, but are out of line with the “doing good while doing well” hypothesis or the perception that investing in greener technologies contributes to long-term value creation.
2
1. Introduction
Climate change is a serious global threat and it demands an urgent global response (Stern, 2006). The energy sector is responsible for two-thirds of all greenhouse gas emissions as well as 80 percent of carbon emissions (International Energy Agency, 2016). This highlights the urgency of a transformation towards sustainable energy to tackle climate change. Countries have started to implement strategies in hope of reducing their impact on climate change and strive to develop in more environmental sustainable ways (Mathews and Kidney, 2012). In 2015, the United Nations held the 21st yearly Climate Change Conference in Paris. The Paris Agreement was negotiated, which stated the commitment of 197 countries of reducing their greenhouse gas emissions to keep the global warming below 2°C. The Paris Agreement is generally considered as a turning point. The International Energy Agency (2014) estimates the needed climate-aligned investments to be 53 trillion US Dollars by 2035, while the New Climate Economy estimates that it is necessary to invest 93 trillion US Dollars in the whole economy by 2030 to keep global warming below 2°C (Climate Bonds Initiative, 2016). This leads to the question how such investments should be financed.
3 which have considerable cash reserves that need to be invested (Dhaliwal et al., 2011; Dimson et al., 2015; Flammer 2018).
A good example of a recent corporate green bond issuance is that of Apple in June of 2017. The bond amounts to 1 billion US Dollars to finance renewable energy and energy efficiency at its facilities and in its supply chain (Forbes, 2017). A good example of a very recent sovereign green bond issuance is that of the Dutch State on May 21 2019, which was the first sovereign green bond with a triple A credit rating. The bond amounts to 6 billion Euros, while investors signed up to buy over 21 billion Euros before the issuance. The Dutch State wants to make sure that the use of proceeds of the bonds are fully used for green projects, including wind farms, large bike parks, strengthening dykes and improving home insulation. Therefore, the Dutch State is happy with the current amount of 6 billion Euros, but will increase the amount in the future (NOS, 2019). The development of the green bond market is believed to have a vital role in facilitating the allocation of capital to projects that tackle climate change (Schroders, 2015). With the Paris Green Bond Statement (2015), asset owners, investment managers and individual funds, managing a total of 11.2 trillion US Dollars of assets, agreed to cooperate to foster the development of the emerging green bond market. The green bond market shows rapid growth over the last couple of years, in which investors see an opportunity for more responsible green funding (Mandel, 2015). Corporate green bonds started to really come into play in 2013 and are still increasing in popularity in the years after (Flammer, 2018). The business community plays an essential part in addressing the threat of climate change and protecting our planets environment (Forbes, 2017). As of September 2018, green bonds amounted to 389 billion US Dollars, which accounts for 26.8 percent of the wider climate-aligned bond universe of 1.45 trillion US Dollars (CBI, 2018). The issuance of green bonds drastically increased from 3 billion US Dollars in 2012 to 167 billion US Dollars in 2018 (CBI, 2019). A critical role in this increase has been the introduction of broad industry guidelines in 2014, called Green Bond Principles (GBP). Before the introduction of these guidelines, the green-labelling process was largely unregulated and mainly subject to self-labelling, leading to little transparency and integrity in the green bond market (NEPC Impact Investing Committee, 2016). The development of the GBP provides confidence to investors on the integrity of the bond and the use of its proceeds. The guidelines define specific criteria that a bond needs to meet in order to be labelled as ‘’green’’. Only a small number of bonds can be labelled as ‘’green’’, so as expected, the green bond market is still relatively small (Schroders, 2015). Green bonds represent less than 1% of the total fixed-income market (NEPC Impact Investing Committee, 2016).
4 environmental sustainability of firms. Especially for corporate green bonds, there is substantial scepticism about the effectiveness of the bonds. A lot of practitioners argue that green bonds may merely be a form of ‘’greenwashing’’, which is the practice of a company declaring a commitment to environmental responsibility greater than in reality (Bachelet et al., 2019). Despite the introduction of the GBP in 2014, the main reason for this scepticism is the lack of public governance. A legal enforcement mechanism to ensure that the proceeds of the green bonds are used to invest in greener technologies doesn’t exist. The green bond market instead relies on private governance systems such as voluntary certification standards (Park, 2018). While the use of corporate green bonds has become increasingly more popular and possibly has an important role in the fight against climate change, very little is known about this new form of investment opportunity, its effectiveness in terms of financial and environmental performance and the implications for shareholder wealth. In practice, what are the implications of corporate green bonds for the financial and, more interestingly, the environmental performance? Do they deliver on their promise of decreasing their environmental footprints? And how does the stock market react to the announcement of corporate green bond issuances?
5
2. Corporate green bonds
This thesis examines the implications of corporate green bonds for firms’ financial and environmental outcomes as well as the implications for shareholder wealth. For this, I first explain how the bond market operates in general and give some stylized facts about my corporate green bond dataset.
2.1 The bond market
6 Figure 1. Bond credit ratings of Moody’s, Standard & Poor’s and Fitch
This figure depicts the several credit ratings for bonds as provided by the credit rating agencies Moody’s, Standard & Poor’s and Fitch. Source: InvestorJunkie, 2018.
2.2 Corporate green bonds
My dataset contains all corporate green bonds that are issued between 2014 and 2018. It yields 1391 corporate green bond issuances between January 1, 2014 and December 31, 2018. For each bond, I collected a wealth of information including the amount, coupon, several credit ratings and many others. In order to compare the green bonds that are issued in different currencies, the amounts issued are all converted into US Dollars.
2.2.1 Corporate green bonds over time
7 Figure 2. Corporate green bonds over time
This figure plots the total annual issuance amounts (in Bln $) of corporate green bonds issued between 2014 and 2018.
2.2.2 Corporate green bonds across industries and countries
Table 1 provides an overview of all the corporate green bonds issued between 2014 and 2018 by industry and country. Industries are partitioned according to Bloomberg’s general industry classification. The countries are the main operating countries of the firms that issued green bonds. As can be seen in Table 1, the highest issuance amount of corporate green bonds is issued by either industrial firms (51.3%) or banks (32.9%). Most of the remaining issuance amount is issued by non-bank financial firms (8.5%) and the rest is issued by either utility- (3.7%), real estate- (2.5%), insurance- (0.9%) or other firms (0.1%). The main operating country with the highest issuance amount of corporate green bonds between 2014 and 2018 is the United States (30.0%). While, China has the second highest (17.6%) and supranational the third highest issuance amount (10.1%). These supranational issuers include development banks and supranational entities such as the European Bank for Reconstruction and Development. These entities qualify as corporates due to their private status. However, these development banks and supranational entities are not corporations in the traditional sense of the word, but are still included in this overview. On the third, fourth and fifth spot, we find European countries: the Netherlands, Germany and France. This is remarkable, since these countries are much smaller in terms of population sizes compared to the United States or China and have a relatively high issuance amount of corporate green bonds between 2014 and 2018. This is in line with the view that Europe tends to be greener (Doh and Guay, 2006; Wall Street Journal, 2017). 0 20 40 60 80 100 120 2014 2015 2016 2017 2018 Gr ee n b o n d issu an ce a mo u n (in Bln $) Year of issuance
8 Table 1. Corporate green bonds across industries and countries
Industry Amount (in Bln $) Country Amount (in Bln $)
Bank 151.3 United States 137.4
Financial 39.2 China 80.7
Real Estate 11.5 Supranational 46.5
Insurance 4.2 Netherlands 40.0 Germany 31.7 Industrial 235.6 France 27.8 Utility 16.9 Mexico 11.6 Others 0.6 Sweden 10.4 United Kingdom 9.4 Total 459.3 Canada 6.5 Japan 6.3 Spain 6.2 Italy 5.6 Luxembourg 5.3 Australia 5.2 India 4.6 Norway 4.3 Austria 3.0 Denmark 2.0 Mauritius 2.0 South Korea 1.7 Others 11.3 Total 459.3
This Table depicts the issuance amounts (in Bln $) across industries and main operating countries of the corporate green bonds issued between 2014 and 2018.
2.2.3 Summary statistics of the corporate green bonds
9 fixed-rate bonds is 3.89% and public firm issuers pay a slightly higher coupon rate than private firm issuers. The median credit rating of the corporate green bonds is A based on S&P’s, Moody’s and Fitch’s rating scales. Therefore, the corporate green bonds have a relatively low risk of default. The bonds issued by private firms have a slightly better median credit rating than the bonds issued by private firms. This means that green bonds issued by public firms have a slightly higher default risk than the green bonds issued by private firms.
Table 2. Summary statistics of the corporate green bonds
All (1) Public (2) Private (3)
Number of green bonds 1391 580 811
Amount (in Mln $) 330.2 (539.35) 374.8 (648.73) 298.3 (442.625)
Maturity (in years) 7.70
(5.60) 7.46 (5.39) 7.88 (5.75) Fixed-rate bond (1/0) 0.796 (0.403) 0.800 (0.400) 0.793 (0.406) Coupon in % (for fixed-rate bonds) 3.89
(2.23) 4.11 (2.16) 3.73 (2.26) Credit ratings
S&P rating (median) A BBB A
Moody’s rating (median) A Baa A
Fitch rating (median) A BBB A
This Table depicts summary statistics for all the corporate green bonds issued between 2014-2018 in column (1) and separately for corporate green bonds issued by public [column (2)] and private [column (3)] firms. Amount is the issuance amount in millions US Dollars. Maturity is the time to maturity of the corporate green bonds in years. Fixed-rate bond is a dummy variable equal to 1 for bonds with a fixed-rate coupon and 0 otherwise. Coupon is the coupon fixed-rate in % for fixed-fixed-rate bonds. Credit ratings are the three different credit ratings of the bonds. For each characteristic, the table reports the sample mean and corresponding standard deviations in parentheses, except for the Credit Ratings, where the medians are reported.
3. Literature review
10 green bonds for the financial and the environmental outcomes are reviewed and hypotheses are formed.
3.1 Literature review on announcement returns
Krüger (2015) studies how stock markets react to positive and negative events concerned with a firm’s corporate social responsibility (CSR). He provides evidence that investors respond strongly negative to negative events and weakly negative to positive events. However, he shows that this negative reaction to positive events is correlated with the likelihood of the positive event being caused by agency problems. Agency problems in the context of CSR, argues that CSR primarily benefits managers who earn a good reputation among key stakeholders at the expense of shareholders. Krüger shows that investors do value “offsetting CSR”, that is positive CSR events of firms that have a poor history of stakeholder relations. This is in line with the alternative perspective that firms engage with stakeholders for value-enhancing purposes. This view is commonly referred to as the “doing good while doing well” hypothesis.
Klassen and McLaughlin (1996) examine the effect of environmental management on future financial performance. Their model links strong environmental management to improved perceived future financial performance by measuring the stock market performance. They empirically test the relationship, using financial event methodology and historical data of the firm’s environmental and financial performance. They find significant positive returns for firms with environmental performance awards and significant negative returns for firms with environmental crises. Further, they find differences in results between first-time award winners and industries. The first-time award announcements were associated with greater positive returns and smaller increases were observed in browner industries, indicating market scepticism.
11 returns, meaning that firms that already behave environmentally responsible experience smaller positive stock returns around the announcement of an environmental responsible event.
Flammer (2018) examined corporate green bonds and their effectiveness. My thesis examines the same research questions as Flammer does in this paper. However, the approach and data she used is slightly different. Like me, she first examines how the stock market responds to the issuance of corporate green bonds. Flammer uses an event study methodology and finds positive announcement returns around the announcements of corporate green bonds. The event windows that she used are different than the ones used in this paper. She finds a Cumulative Abnormal Return (CAR) in the two-day event window [-1, 0] of 0.67%. Her findings suggest that the stock market expects green bonds to contribute to value creation.
Eckbo et al. (2007) examine the stock market reactions to the issuance of securities. They find that the stock market reacts negatively to equity issues, but show no significant reaction to the debt issues (bonds). Their findings are consistent with the pecking order theory of Myers and Majluf (1984).
The existing literature on the announcement returns of green bonds is mixed, but favours a positive effect. Flammer (2013, 2018) as well as Klassen and MacLaughlin (1996) find that stock markets react positively to the announcements of positive environmental events, whereas Krüger (2015) finds a (weakly) negative stock market reaction to the announcements of positive environmental events. Eckbo et al. (2007) find no significant stock market reactions to the announcement of bond issuances.
3.1.1 Hypotheses on announcement returns
To examine the stock market reaction around the announcement of a green bond issuance, hypotheses are made. Empirical research will be conducted based on the question: ‘how does the stock market react to the announcement of corporate green bond issuances?’ This question will be answered by testing the following null hypothesis and alternative hypothesis:
Ho: There is no stock market reaction to the announcement of corporate green bond issuances.
12 A two-sided test is used, since the announcement returns could be positive, negative or non-existent according to the existing literature. If there are significant announcement returns, the null hypothesis will be rejected and the alternative hypothesis will be accepted. If these significant returns are positive, the result will be in line with the “doing good while doing well” hypothesis.
3.2 Literature review on financial and environmental outcomes
Guenster et al. (2011) study whether corporate environmental management align with the economic objectives of a firm. They perform an extensive analysis on the relationship between eco-efficiency and several financial performance measures. They use a large database that consists of monthly data for the period December 1996 to December 2002. Guenster et al. find evidence that strong corporate eco-efficiency policy can yield significant improvements in the financial performance of the firm. They look at a measure for a firm’s value, namely Tobin’s Q and a measure for the firm’s operating performance, namely Return on Assets. The evidence suggests that the firms that are deemed more eco-efficient do not yield higher values in both variables than the control group, but the firms that are deemed least eco-efficient yield significantly lower values in both measures. Thus, they argue that the benefits of adopting a strong environmental policy are likely to outweigh the costs.
Arafat et al. (2012) examine the effect of environmental disclosure together with environmental performance on firm performance for 33 Indonesian manufacturing firms. The study extends the literature that has been done mostly in western countries or regions, but is rarely investigated in developing countries or regions. Their empirical results show that the environmental performances of the Indonesian firms have significantly positive effects on the firms’ financial performances. However, they find that environmental disclosures have no significant relation with the financial performances. This result could describe the behaviour of the Indonesian firms in disclosing its environmental management systems, which is on voluntary basis in Indonesia. They argue that firms with a good environmental performance will have a better financial performance due to a decrease in environmental-related costs.
13 Flammer (2018) examined corporate green bonds and their effectiveness. Like me, she uses a difference-in-differences specification estimation to examine the financial and environmental performance of firms after the issuance of corporate green bonds. Her evidence suggests that corporate green bonds are indeed effective: they yield improvements in firms’ environmental footprint. On top of that, Flammer found that corporate green bonds contribute to the financial performance in terms of Tobin’s Q and Return on Assets. Moreover, corporate green bonds help attract responsible long-term investors. Her findings suggest that corporate green bonds are an effective finance tool to create long-term value and improve the environmental performance. Therefore, she argues that corporate green bonds could serve as a powerful tool to fight climate change.
Literature on the effects of corporate green bonds on the environmental performance of the firms is scarce. However, there is some literature that examines the effect of the environmental performance on a firm’s financial performance. This existing literature generally sketches the same picture of the environmental performance having a positive effect on a firm’s financial performance. However, the increased environmental performance, as a consequence of corporate green bond issuances, is not assumed in this thesis. Therefore, we examine whether the corporate green bonds actually improve the firms’ environmental footprints. Flammer (2018), who did just this, found that corporate green bonds indeed improve the firms’ environmental performances and also contribute to the firms’ financial performances.
3.2.1 Hypothesis on financial and environmental outcomes
To examine how the corporate green bond issues affect financial and environmental outcomes, hypotheses are made. Empirical research will be conducted based on the question: ‘what are the implications of corporate green bonds on the financial and, more interestingly, the environmental performance?’ The implications for the financial outcomes will be tested by the following null hypothesis and alternative hypothesis:
Ho: There are no implications of corporate green bond issuances for the financial performance.
14 A two-sided test is used, since multiple financial performance measures with different implications are used in my analysis. This means that corporate green bonds could yield a positive effect on one measure, while yielding a negative effect on the other. If there is a significant effect on any of the financial outcomes, the null hypothesis will be rejected and the alternative hypothesis will be accepted.
The implications for the environmental outcomes will be tested by the following null hypothesis and alternative hypothesis:
Ho: There are no implications of corporate green bond issuances for the environmental performance.
Ha: Corporate green bond issuances yield a positive effect for the environmental performance.
An one-sided test is used, since we expect the corporate green bonds to yield a positive effect on the environmental performance. If there is a significant positive effect on any of the environmental outcomes, the null hypothesis will be rejected and the alternative hypothesis will be accepted. If corporate green bonds yield a significant positive effect on both a firm’s financial and environmental performance, it will be in line with the “doing good while doing well” hypothesis.
4. Data Analysis
15 collected from Bloomberg is also used for the overviews and summary statistics of the corporate green bonds in subsection 2.2. To be able to compare the bonds, all issuance amounts are converted into U.S. Dollars. In the empirical analysis of this thesis, the sample is restricted to only the corporate green bonds issued by public firms, because firm-level data is much more available for public firms than for private firms. Also, because the implications of corporate green bonds for shareholder wealth can only be examined for public firms and not for private firms. A few green bond issuers in the remaining sample are subsidiaries of larger public firms. For these few firms, data of the parent companies are collected and examined in the empirical analysis.
Data on the firms’ historic daily stock prices and corresponding daily market returns are required to estimate the abnormal returns, which are used to examine the announcement returns. The historic daily stock prices of the firms are gathered from DataStream. In this examination, three different approaches with different daily market returns are used. For the first two approaches, only the historic daily market returns are required to estimate the normal returns. Daily market price indexes are used to compute these daily market returns and are collected from DataStream. For the third approach, historic data on two additional daily risk factors is required as well as the historic daily excess market returns and the daily risk free rate. These are all gathered from the Kenneth R. French Data Library.
Financial and environmental data are required for the environmental and financial outcomes analysis. In this analysis, a control group of public firms that did not issue any green bonds is constructed. The public firms that did issue one or multiple green bonds between 2014 and 2018 are considered the treatment group. For all firms, the financial data is collected from DataStream. To measure the financial performance of the firms, annual data are collected on the Book Value of Total Assets denoted in US Dollars as well as in local currencies, the Leverage ratios, the Return on Assets ratios and the Total Market Values of the firms denoted in local currencies for the years 2011-2018. The main financial performance variables examined in this thesis are either constructed from two of these variables or directly provided by DataStream.
16 reflects how well a firm uses best management practices to avoid environmental risks and capitalizes on environmental opportunities in order to generate long-term shareholder value.
Data on the main operating countries of the firms are collected from Bloomberg’s fixed income database for the green bond issuers and from DataStream for the control group. For all firms’ industries, data on the general industry classification as provided by ASSET4 are collected from DataStream. This variable classifies all firms into 6 general industry sectors.
5. Methodology
In this section, the methods and models that are used to perform empirical research will be described. First, the methods and the models that are used to examine the stock market reaction around the announcement of a green bond issuance are demonstrated. Then, the method and the model that are used to examine how the announcement returns differ among levels of multiple financial performance measures of firms are explained. Lastly, the method and the model that are used to examine the effects of corporate green bonds on firms’ financial and environmental performances are described.
5.1 Announcement returns methodology
17 1 trading day prior to the event. To measure the announcement returns, the abnormal returns of the trading days within the event windows are summed, giving the cumulative abnormal returns (CARs) for every event window. To be able to do this, we first have to estimate the firms’ alphas and betas, which are used to compute the normal returns in the event windows. This is done by looking at the returns of the 120 trading days prior to the first trading day of the event window (t=-5), as depicted in Figure 3. For only 501 out of the 580 green bonds issued by public firms, the data on these 120 trading days are sufficient.
Figure 3. Estimation and event window
This figure depicts the estimation- and event window used in the announcement returns analysis. The
estimation window consists of the 120 trading days that range from t1 = -125 to t2 = -5. The event
window is the 11 day event window [-5, 5] around the announcement of the corporate green bonds, which also contains the event windows [-3, 3] and [-1, 1]. The estimation window is used to estimate the firms’ alphas and betas, which are then used to compute the normal returns in the event windows.
18 returns are used instead of the ordinary market returns. The excess market returns are simply the market returns minus the risk-free rate of return. The three factor model also includes the risk-free rate of return as well as the two additional risk factors: the Small Minus Big (SMB) and the High Minus Low. The SMB factor accounts the returns for the size effect: firms with smaller market capitalization tend to outperform firms with larger market capitalization. The HML factor accounts the returns for the fact that firms with high book-to-market ratios outperform firms with low book-to-market ratios. All returns and market returns are computed by the following equations:
𝑅𝑖,𝑡 = 𝑃𝑖,𝑡 − 𝑃𝑖,𝑡−1 𝑃𝑖,𝑡−1 (1) 𝑅𝑀,𝑡= 𝑃𝑀,𝑡 − 𝑃𝑀,𝑡−1 𝑃𝑀,𝑡−1 (2)
where Pi,t is the closing price of stock i at date t, Pi,t−1 is the closing price of stock i at date t − 1, PM,t is the closing price of benchmark at date t and PM,t−1 is the closing price of the benchmark at date t − 1.
The alphas and betas of the firms are estimated by Ordinary Least Squares regressions based on the estimation window of 120 trading days prior to the first day of the event window (t=-125 to t=-5). For the first two approaches, I estimate the market model:
𝑅𝑖,𝑡 = 𝛼𝑖 + 𝛽𝑖 ∗ 𝑅𝑀,𝑡+ 𝜀𝑖,𝑡 (3)
where Ri,t is the return on the stock of firm i on day t, αi is the constant term, RM,t is the daily market return and εi,t the residual. The daily market returns are either firm-specific (approach 1) or global (approach 2).
The normal return of firm 𝑖 on day 𝑡 for the market model approaches are then given by:
𝑅̂𝑖,𝑡 = 𝛼̂𝑖 + 𝛽̂𝑖 ∗ 𝑅𝑀,𝑡 (4)
The global three factor model of approach 3 requires the estimation of two additional beta coefficients for the SMB and HML factors. The alphas and betas are again estimated by Ordinary Least Squares regressions based on the estimation window of 120 trading days prior to the first day of the event window (t=-125 to t=-5). For the third approach, I estimate:
19 where Ri,t is the return on the stock of firm i on day t, αi is the constant term, Rf,t is the daily global risk-free rate of return, (RM,t− Rf,t) is the daily global excess market return, SMBt is the daily global SMB factor, HMLt is the daily global HML factor and εi,t the residual.
The normal return of firm 𝑖 on day 𝑡 for the three factor model approach is then given by:
𝑅̂𝑖,𝑡 = 𝛼̂𝑖+ 𝑅𝑓,𝑡 + 𝛽̂𝑀,𝑖∗ (𝑅𝑀,𝑡− 𝑅𝑓,𝑡) + 𝛽̂𝑆𝑀𝐵,𝑖∗ 𝑆𝑀𝐵𝑡+ 𝛽̂𝐻𝑀𝐿,𝑖∗ 𝐻𝑀𝐿𝑡 (6)
The abnormal returns can then be computed in a similar fashion for both the market model approaches and the three factor model approach. It is simply the actual returns minus the estimated normal returns. The abnormal daily return (AR) of firm 𝑖 on day 𝑡 is calculated as follows:
𝐴𝑅𝑖,𝑡 = 𝑅𝑖,𝑡− 𝑅̂𝑖,𝑡 (7)
From the abnormal daily returns, the cumulative abnormal returns (CAR) for each considered event window can then be computed by summing up the daily abnormal returns within the event windows [-1, 1], [-3, 3] and [-5,5] around the announcements of the 501 corporate green bonds issued between 2014 and 2018 with sufficient stock market data.
5.2 Cross-sectional characteristics of announcement returns methodology
20 𝑇𝑜𝑏𝑖𝑛′𝑠 𝑄 = 𝑇𝑜𝑡𝑎𝑙 𝑀𝑎𝑟𝑘𝑒𝑡 𝑉𝑎𝑙𝑢𝑒
𝐵𝑜𝑜𝑘 𝑉𝑎𝑙𝑢𝑒 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠 (8) Cross-sectional regressions with the CARs of the three approaches and the three event windows as dependent variable are run. In these regressions, I account for country- industry- and year of issuance-fixed effects. The CARs are regressed against the firm’s four lagged financial performance measures, country dummies, industry dummies and year of issuance dummies. The financial performance measures consists of four lagged variables to measure the different aspects of the firm’s financial performance in the year preceding the green bond issuance. The lagged variables of these measures are used, because the annual financial data collected from DataStream is composed at the end of every year. Green bonds could be issued in the beginning of a year and therefore the lagged variables are certain to be composed before any of the corporate green bond issuances occurred. The country dummies are dummy variables with a value of 1 if the firm is mainly operating in that country and 0 otherwise. To prevent outliers from being included in the dataset, countries with less than 5 firm observations are dropped. Additionally, firm observations that have insufficient financial data are also dropped in this analysis. This decreases the amount of main operating countries from 26 to 11; the number of firm observations from 501 to 469 and the number of unique firms that issued one or multiple green bonds from 191 to 169. The industry dummies are dummy variables with a value of 1 if the firm is classified in that general industry classification and 0 otherwise. The firms are divided into 6 general industry classifications: 1. Industrial, 2. Utility, 3. Transportation, 4. Bank/Savings & Loans, 5. Insurance and 6. Other Financial. The year of issuance dummies take the value 1 if the green bond is issued in that year and 0 otherwise. The cross-sectional regression then looks as follows:
𝐶𝐴𝑅[−𝑡, +𝑡]𝑖= 𝛼𝑖+ 𝑇𝑄𝑖,𝑇−1+ 𝑅𝑜𝐴𝑖,𝑇−1+ 𝐿𝑜𝑔(𝐴𝑠𝑠)𝑖,𝑇−1+ 𝐿𝑒𝑣𝑖,𝑇−1+ 𝐶𝑡𝑟𝑦𝑖+ 𝐼𝑛𝑑𝑖+ 𝑌𝑖+ 𝜀𝑖,𝑡 (9)
21
5.3 Financial and environmental outcomes methodology
22
5.3.1 Control group
To construct a control group, a group of firms that did not issue any green bonds has to be selected. My selection process is based on the main operating countries of the firms that issued green bonds between 2014 and 2018. To be able to construct the control group, all firm observations of green bond issues, where the main operating country has less than 30 observations, are dropped from the remaining 469 firm observations that have sufficient financial data and no main operating countries with less than 5 observations. This decreases the amount of main operating countries from 11 to 5; the number of firm observations from 469 to 415 and the number of unique firms that issued one or multiple green bond from 169 to 136. Table 3 depicts this process and it also shows that a clear cut has been made at 30 observations, since there are no main operating countries with a number of firm observations between 13 (India) and 30 (Japan).
Table 3. Country frequency of remaining firm observations of green bond issues
This table reports the number of firm observations of main operating countries of the remaining firm observations of corporate green bond issues between 2014 and 2018. The first column depicts the number of firm observations of main operating countries after the removal of countries with less than 5 observations and firms with insufficient financial data. The second column depicts the number of observations of main operating countries after removal of countries with less than 30 observations.
As shown in Table 3 above, the remaining 5 countries are: 1. China, 2. France, 3. Japan, 4. Sweden and 5. the United States. For the remaining 415 firm observations, I construct a control group based on the constituents of the remaining 5 main operating countries’ leading
23 benchmark stock market index(es). Of course, I remove the firms that issued a green bond between 2014 and 2018 from these lists. Since the green bond market was more or less non-existent prior to 2013 (Flammer, 2018), we can assume that firms that did not issue green bonds during this period are likely to not have issued green bonds at all. In general, only large firms issue corporate green bonds and the constituents of the leading benchmark stock market indexes are generally large firms as well. An overview of this construction process is given in Table 4 below.
Table 4. Control group construction process
This table shows the number of firms of each benchmark stock market index that are used to construct a control group for the difference-in-differences specification estimations that measure the effect of corporate green bond issuances on financial and environmental outcomes. The number of control firms results in 483 after the removal of the firms that issued corporate green bonds between 2014 and 2018.
As shown in Table 4 above, a control group of 483 firms is constructed. I tried to match the number of firms per country of the treatment and control group. For this reason, I took the top 50/300 constituents of certain benchmark stock market indexes as provided by the DataStream list, when the list constituted of too many firms. For China, two benchmark stock market indexes are used, since firms’ equities that issued green bonds are also mainly being traded on both the Hong Kong and Shanghai stock exchanges.
Country Benchmark stock market index
Constituents Green bond issuers
Remaining control group
China Hang Seng Enterprises (Hong Kong)
50 1 49
China Top 50 Shanghai SE A Share 50 6 44
France CAC40 40 6 34
Japan Top 50 TOPIX 50 3 47
Sweden OMX Stockholm 30 30 5 25
United States Top 300 S&P 500 300 16 284
24
5.3.2 Model to examine financial and environmental outcomes
25 years after a green bond issuance. The difference-in-differences specification estimation regressions then look as follows:
𝐷𝑒𝑝𝑉𝑎𝑟𝑖,𝑡 = 𝛼𝑡+ 𝐹𝑖𝑛𝑖,𝑡 + 𝐶𝑡𝑟𝑦𝑖 + 𝐼𝑛𝑑𝑖+ 𝐺𝐵𝑖,𝑡 + 𝜀𝑖,𝑡 (10)
where i indexes firms and t indexes the year; DepVari,t is the dependent variable of interest; Fini,t are the financial control variables; Ctryi are the country dummies; Indi are the industry dummies; GBi,t is the green bond current-dummy variable and εi,t is the error term.
The coefficient of the current-dummy GB is the coefficient of interest and captures the difference-in-differences in outcomes of the dependent variables for firms that issued a green bond in that year. To test for pre-trends in the data and for the longer-term effect of the green bond issues, the current-dummy GB is replaced in two separate regressions with the pre- and post-dummy variables: GBPre and GBPost. The coefficients of these dummy variables capture the difference-in-differences in outcomes of the dependent variables for firms in the year preceding- and the subsequent years after a green bond issuance.
6. Results
In this section, the outcomes of the empirical analyses will be discussed. First, the results of the announcement returns analysis will be discussed. Then, the outcomes of the cross-sectional characteristics analysis of these announcement returns will be examined. Lastly, the results of the financial and environmental outcomes analysis will be discussed.
6.1 Announcement returns
26 get the abnormal returns for the three event windows. The abnormal returns are then summed, giving us the CARs, which are summarized in Table 12 in the appendix.
In addition to looking at the average CARs for the three event windows, I tested whether these averages are significant and thus significantly different from zero. To determine the significance of the results, the CARs are tested on significance by using t-tests with null hypotheses of the mean of the single sample being equal to 0 and alternative hypotheses of the mean being different from 0. The average CARs and the results of the t-tests are summarized in Table 5 below and a full overview of the t-tests can be found in Table 13 in the appendix.
Table 5. Stock market reactions to the announcements of corporate green bond issuances Approach Event window Mean CAR (in %) t-test
Ho: mean = 0 t-statistic 1 [-1, 1] -0.290 (0.235) -1.23 1 [-3, 3] 0.157 (0.228) 0.69 1 [-5, 5] 0.386 (0.266) 1.45 2 [-1, 1] -0.256 (0.240) -1.06 2 [-3, 3] 0.191 (0.233) 0.82 2 [-5, 5] 0.418 (0.275) 1.52 3 [-1, 1] -0.298 (0.238) -1.25 3 [-3, 3] 0.112 (0.227) 0.49 3 [-5, 5] 0.290 (0.265) 1.10
27 When looking at Table 5, it shows us that the results are very similar across approaches. Our main focus, the 3 day event windows [-1, 1], have negative means for every approach, but are not significantly different from zero with t-statistics between -1.06 and -1.25. The 7 day event windows [-3, 3] and the 11 day event windows [-5, 5] have positive means for every approach, but are also not significantly different from zero with t-statistics between 0.49 and 1.52. Thus, the cumulative abnormal returns are not significantly different from zero for any approach or any event window. This means that there is no significant stock market reaction in the 3, 7 and 11 trading days around the announcement date of corporate green bond issuances. Therefore, the null hypothesis of there being no stock market reaction can’t be rejected and the alternative hypothesis of there being a stock market reaction is rejected, when considering the 501 corporate green bonds issued between 2014 and 2018. This finding is in line with the paper of Eckbo et al. (2007), who found that the stock market has no significant reaction to debt issues (bonds).
6.1.1 Announcement returns for single bond issuers
As already noted, some firms issued multiple green bonds between 2014 and 2018. The number of unique issuers in the sample of the 501 green bond issues is 191 and the number of firms that issued only one green bond in the sample period is 124. This means that 67 firms issued multiple green bonds between 2014 and 2018, accounting for 377 out of the 501 green bonds issued. To check whether the stock market reacts in a different manner to firms that issue a single green bond compared to firms that issue multiple green bonds, the CAR event study is redone with the firms that only issued a single green bond between 2014 and 2018. Since the green bond market was more or less non-existent prior to 2013 (Flammer, 2018), these issuances are likely to be their first green bond issuances. The average CARs of the single green bond issuers are computed and also tested for their significance. The t-tests of the average CARs of single green bond issuers yield slightly different results and are summarized in Table 6 below. A full overview of the t-tests can be found in Table 12 in the appendix.
Table 6. Stock market reaction to the announcements of corporate green bond issuances of single bond issuers
Approach Event window Mean CAR (in %) t-test Ho: mean = 0
t-statistic
28 (0.855) 1 [-3, 3] -0.113 (0.645) -0.17 1 [-5, 5] 0.339 (0.708) 0.48 2 [-1, 1] -1.473* (0. 867) -1.70 2 [-3, 3] 0.005 (0.634) 0.01 2 [-5, 5] 0.362 (0.689) 0.53 3 [-1, 1] -1.504* (0.861) -1.75 3 [-3, 3] -0.017 (0.646) -0.03 3 [-5, 5] 0.368 (0.714) 0.52
This table reports a summary of the single sample t-tests of the cumulative abnormal returns (CARs) as well as the mean and corresponding standard error in parentheses for the different time windows around the announcements of corporate green bond issues of firms that issued a single bond between 2014 and 2018. The CARs are tested for Ho: mean = 0 and Ha: mean ≠ 0. Three different approaches are used to estimate the CARs: 1. The market model with firm-specific market returns; 2. The market model with global market returns and 3. The global three factor model of Fama and French. The sample consists of N = 124 green bond issues. *, ** and *** denote significance at the 10%, 5% and 1% level, respectively.
29 that issue a green bond for the first time and therefore reacts more negatively 2. the stock market is less surprised by green bond issuances of firms that already issued one or multiple green bonds and therefore reacts less negatively 3. a combination of both 1 and 2. Furthermore, this finding is out of line with the “doing well” aspect of the “doing good while doing well” hypothesis and in line with the paper of Krüger (2015), who finds that the stock market reacts weakly negative to positive CSR events.
6.2 Cross-sectional characteristics of announcement returns
I examine the cross-sectional characteristics of the cumulative abnormal returns. Specifically, I look at how the announcement returns differ among levels of multiple financial performance measures of firms. Following equation (9), the cross-sectional regressions are run with the 9 CARs as dependent variables. The amount of regressions equals 9, because we have 3 different event windows with 3 different approaches of estimating the CARs. A full overview of these regression outputs can be found in Tables 15, 16 and 17 in the appendix. The results of the cross-sectional regressions with the [-1, 1] event window CARs, the main focus, as dependent variables are shown in Table 7 below.
Table 7. Cross-sectional regression output of the 3 day event windows [-1, 1] CAR[-1, 1] Approach 1 (1) CAR[-1, 1] Approach 2 (2) CAR[-1, 1] Approach 3 (3) TQ i,T-1 0.0078 (0.0053) 0.0066 (0.0053) 0.0068 (0.0053) RoA i,T-1 0.0318 (0.0960) 0.0021 (0.0944) 0.0091 (0.0956) Log(Ass)i,T-1 -0.0006 (0.0010) -0.0010 (0.0010) -0.0011 (0.0010) Lev i,T-1 0.0516*** (0.0196) 0.0472** (0.0194) 0.0392** (0.0190) Constant 0.0080 (0.0164) 0.0127 (0.0164) 0.0095 (0.0171)
Country fixed effects Yes Yes Yes
Industry fixed effects Yes Yes Yes
Year of iss. fixed effects Yes Yes Yes
30
This table reports the regression outputs of the cross-sectional characteristics analysis of the cumulative abnormal returns (CARs) of the 3 day event windows [-1, 1] around the announcements of corporate green bond issues between 2014 and 2018. Three different approaches are used to estimate the CARs: 1. The market model with firm-specific market returns; 2. The market model with global market returns and 3. The global three factor model of Fama and French. The table depicts the coefficients of the four financial performance measures of the firms in the year preceding a green bond issuance and a constant term. The robust standard errors are reported in the parentheses. The sample consists of N = 469 green bond issues. *, ** and *** denote significance at the 10%, 5% and 1% level, respectively.
The results of the cross-sectional regressions across the three different approaches are very similar. As shown in Table 7, the only financial performance measure that has a significant relationship with the CARs of the 3 day event windows [-1, 1] is the Leverage ratio. The relationship is positive and significant at the 1% level for the CAR[-1, 1] computed via approach 1 (column 1) and significant at the 5% level for the CAR[-1, 1] computed via approach 2 and 3 (columns 2 and 3). Approach 1, the market model with firm-specific market returns, is considered the main approach and results in a coefficient for the lagged Leverage ratio of 0.0516 for the event window [-1, 1]. The other two approaches yield similar results and therefore strengthen this outcome. This means that a one unit increase in the lagged Leverage ratio, increases the CAR[-1, 1] with 0.0516. In other words: an 100% increase of the level of debt in the year preceding a green bond issuance leads to a 5.16% higher Cumulative Abnormal Return in the 3 days around the announcement of a corporate green bond issuance. Firms that have a high Leverage ratio in the year preceding a green bond issuance are therefore more likely to experience positive announcement returns and less likely to experience negative announcement returns in the 3 days around the announcement of a corporate green bond issuance.
When looking at Table 15, 16 and 17 in the appendix, the relations between the CARs of the event windows [-3, 3] and [-5, 5] and the lagged financial performance measures can be seen. The CAR[-3, 3] and the CAR[-5, 5] of approach 1 also show a positive significant relationship with the lagged Leverage ratio, but only at the 10% significance level. For approaches 2 and 3, the relationship is still positive, but not significantly different from zero at the 10% level. This means that the positive relationship between the lagged Leverage ratio and the announcement returns around the announcement of a green bond issuance becomes less strong when more days are included in the event window. When looking at the [-3, 3] and [-5, 5] event windows, we find that the lagged natural logarithm of the book value of total assets denoted in US Dollars has a significant negative relationship with 5 of the 6 CARs of these two event windows. For approach 1, the CAR[-5, 5] is negatively related with Log(Ass)
31
1 at the 10% significance level. For approach 2 and 3 both the CAR[-3, 3] and the CAR[-5, 5] are negatively related with Log(Ass)i,T-1 at the 10% significance level. Therefore, the firms that have a high book value of total assets denoted in US Dollars are more likely to experience negative announcement returns in the 7 and 11 days around the announcement of a corporate green bond issuance.
6.3 Results financial and environmental outcomes analysis
In this section, the results of the financial and environmental outcomes analysis are discussed. There is a big difference in the data availability between the financial performance measures and the environmental performance measures. The remaining 415 treatment group observations (green bond issues) are already filtered on the four variables used to measure the financial performance and the 483 control group observations also have sufficient data on the financial performance measures. Data on the environmental performance measures is more scarce, reducing the number of firm observations of the environmental outcomes analysis. An overview of this reduction is given in Table 18 in the appendix. A particular reason for this reduction is that the environmental performance measures have almost no annual data for the year 2018. Therefore, the environmental outcomes analysis can hardly be performed for green bonds issued during 2018. First, the financial outcomes and after that the environmental outcomes of the difference-in-differences specification regressions are discussed.
6.3.1 Results financial outcomes
32 Table 8. Implications of green bond issuances for financial outcomes
This table reports the relevant estimated coefficients of the difference-in-differences specification
regressions with the four financial performance measures as dependent variables. GBPre, GB and
GBPost are the green bond dummy variables. The sample includes all firm-year observations of the treated and control firms from 2011 to 2018. The robust standard errors are reported in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level, respectively.
Column (1) of Table 8 shows us the coefficients of the green bond dummies with Tobin’s Q as dependent variable, which measures a firm’s value. We find a positive coefficient of 0.191 for the pre-dummy that is significant at the 1% level. This means that a firm’s value increases significantly in the year preceding a green bond issuance with 0.191 or 19.1% in terms of Tobin’s Q. Thus, the results show that there is a pre-trend in the data with respect to Tobin’s Q. The estimated coefficient of the current-dummy is positive and significant at the 1% level and has a value of 0.307. This means that a firm’s value increases even more, namely with 30.7% in terms of Tobin’s Q, in the year of the green bond issuance. This is inconsistent with the CAR event study results, which found a significant decrease in shareholder value for single green bond issuers and an insignificant decrease in shareholder value in the analysis of all the green bond issuers. However, the coefficient of the post-dummy (-0.238) is negative and significant at the 1% level, meaning that the firm’s value decreases significantly in the subsequent years after a green bond issuance. This decrease of 23.8% in terms of Tobin’s Q is more in line with the results of the announcement returns analysis. Overall, the outcomes with respect to Tobin’s Q are quite contrasting, since we found a very large positive effect
Tobin’s Q (1) RoA (2) Log(Assets) (3) Leverage (4) GBPre 0.191*** (0.023) 0.024*** (0.003) -0.051*** (0.016) -0.0039* (0.0022) GB 0.307*** (0.029) -0.005* (0.003) 0.207*** (0.019) -0.0076*** (0.0018) GBPost -0.238*** (0.031) 0.012*** (0.003) 0.635*** (0.042) -0.0104*** (0.0035)
Financial control variables Yes Yes Yes Yes
Country fixed effects Yes Yes Yes Yes
Industry fixed effects Yes Yes Yes Yes
Observations 6849 6849 6849 6849
33 (+30.7%) on firms’ values in the year of the green bond issuance and a very large negative effect in the subsequent years after a green bond issuance (-23.8%). However, the large positive effect in the year of the issuance should be nuanced and is probably not solely caused by the green bond issuance, since green bond issuers are already increasing in value in the year preceding an issuance. Given that the green bond issuers are increasing in value in the year preceding- and in the year of the issuance, the large negative effect on a firm’s value in the subsequent years after an issuance is even more substantial. My findings with respect to Tobin’s Q are in contrast with those of Flammer (2018), who found that green bond issuances cause long-term value creation.
In column (2) of Table 8, the estimated coefficients of the green bond dummies with the Return on Assets (RoA), which measures a firm’s profitability, as dependent variable are depicted. For the pre-dummy, we find a positive coefficient of 0.024 that is significant at the 1% level. This means that there is a pre-trend in the data and a firm’s profitability increases with 2.4% in terms of RoA in the year preceding a green bond issuance. The estimated coefficient for the current-dummy (-0.005) is negative and weakly significant at the 10% level. This means that a firm’s profitability decreases slightly with 0.5% in terms of RoA in the year of the green bond issuance. This result is more substantial, since there is a positive pre-trend in the data. The firms’ profitability increases significantly in the year preceding the issuance and then has a weakly significant decrease in the year of the issuance. Looking at the estimated coefficient for the post-dummy (0.012), we find a positive effect that is significant at the 1% level. This means that the firm’s profitability increases significantly in the subsequent years after a green bond issuance with 1.2% in terms of RoA. Given the positive pre-trend, the increase in profitability in the subsequent years after an issuance might not be solely caused by the green bond issuance. Unobserved variables, which also caused the pre-trend in the data, might have influenced the post-treatment period outcome. Moreover, the coefficient of the post-dummy is smaller and less significant than the coefficient of the pre-dummy, increasing the seriousness of this concern. Therefore, the most substantial result with respect to the RoA is the decrease of 0.5% in the year of the green bond issuance. My findings with respect to RoA are in contrast with those of Flammer (2018), who found no pre-trend and a positive effect on a firm’s profitability in the year of a green bond issuance. However, both results show that the profitability increases in the long run, but this finding is weakened by the pre-trend in my analysis.
34 in the year preceding a green bond issuance. The coefficients for the current-dummy (0.207) and post-dummy (0.635) are both positive and significant at the 1% level. This means that a firm’s size increases significantly in the year of- and subsequent years after a green bond issuance. Given the negative pre-trend, the positive effects in the year of- and subsequent years after a green bond issuance are even more substantial. The firms that issue green bonds are decreasing in the year preceding an issuance, increasing the likelihood of the positive effects being caused by the green bond issuances. In column (4) of Table 8, the outcomes with respect to the Leverage ratio are depicted. The estimated coefficient of the pre-dummy (-0.0039) is negative and weakly significant at the 10% level. This means that there is a weakly significant pre-trend in the data and a firm’s debt level decreases with 0.39% in terms of the Leverage ratio in the year preceding a green bond issuance. The estimated coefficients for the current-dummy (-0.0076) and post-dummy (-0.0104) are also negative, but significant at the 1% level. This means that a firm’s level of debt decreases with 0.76% in terms of the Leverage ratio in the year of the green bond issuance and with 1.04% in the subsequent years after. These findings are weakened by the negative pre-trend in the data. The decreases in debt levels are less likely to be solely caused by the green bond issuances, because the green bond issuers were already decreasing their debt levels in the year preceding the issuance. However, the decreases are larger and more significant in the year of- and especially in the subsequent years after a green bond issuance.
35
6.3.2 Results environmental outcomes
Following equation (10), The difference-in-differences specification regressions are run with the two environmental performance measures as dependent variables. The coefficients of the green bond dummies measure the difference-in-differences in outcomes of the dependent variable between treated and control firms. In other words, the coefficients of the green bond dummies measure the effect of the green bond issue on the dependent variable. The results of these difference-in-differences specification estimations are depicted in Table 9 below.
Table 9. Implications of green bond issuances for environmental outcomes
This table reports the relevant estimated coefficients of the difference-in-differences specification
regressions with the two environmental performance measures as dependent variables. GBPre, GB and
GBPost are the green bond dummy variables. The sample includes all firm-year observations of the treated and control firms from 2011 to 2018. The robust standard errors are reported in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% level, respectively.
Since the corporate green bond issuances are expected to yield a positive effect on the environmental outcomes (increase in Env. Score and decrease in CO2 / Assets) in the year of- and subsequent years after an issuance, one-sided z tests are used to determine the significance of the current- and post-dummy coefficients. An overview of these one-sided z tests can be found in Table 19 in the appendix. The significance of the coefficients of the pre-dummies are still determined based on two-sided z tests, since we have no expectations for
Env. Score (1) CO2 / Assets (2) GBPre 0.334 (0.436) 0.083** (0.035) GB 0.560 (0.508) -0.122** (0.057) GBPost 2.331*** (0.715) -0.430*** (0.086)
Financial control variables Yes Yes
Country fixed effects Yes Yes
Industry fixed effects Yes Yes
Observations 4874 5445
36 this period with respect to the environmental outcomes. The regression outputs, depicted in Table 9, denote the significance levels of two-sided z tests.
In column (1) of Table 9, the estimated coefficients for the green bond dummies are depicted with the overall environmental score as dependent variable, which measures the overall environmental performance of a firm on a scale from 0 to 100. For the GBPre dummy, we find a positive and insignificant coefficient of 0.334. This means that there is no significant pre-trend in the data with respect to the environmental score as dependent variable. The estimated coefficient for the current-dummy (0.560) is positive and also insignificant, looking at Table 19 in the appendix. This means that there is no significant effect on a firm’s environmental performance in the year of a green bond issuance. Looking at the estimated coefficient for the GBPost dummy (2.331), we find a positive effect that is significant at the 1% level (Table 19). This means that the firm’s overall environmental performance increases by 2.331 points in terms of the environmental score in the subsequent years after a green bond issuance. Given that the effects in the year before- and the year of the green bond issuance are insignificant, we can argue that the investments in greener technologies, caused by the green bond issuances, take at least one or a couple of years to actually affect firms’ overall environmental performances.
37 My findings suggest that the investments in greener technologies, funded by the corporate green bond issuances, increase the environmental performance of firms in the year of the issuance somewhat, but are better reflected by the two environmental performance measures after at least one or a couple of years. Therefore, the null hypothesis of the corporate green bond issuances having no implications for the environmental performance is rejected and we accept the alternative hypothesis of the corporate green bond issuances yielding a positive effect on the environmental performance. My findings with respect to the environmental outcomes analysis are in line with the “doing good” aspect of the “doing good while doing well” hypothesis and are essentially in line with the findings of Flammer (2018). However, Flammer found a significant positive effect on the overall environmental score in the year of corporate green bond issuances, while my results show an insignificant positive effect. Like with the financial outcomes analysis, this difference might be caused by the different range of issuance years of the examined corporate green bonds. Corporate green bonds issued in 2013 might yield a more positive immediate effect on the overall environmental score and/or corporate green bonds issued in 2018 might yield a less positive immediate effect on the overall environmental score.
7. Conclusion
39
References
Arafat, M., Warokka, A., Dewi, S., 2012. Does environmental performance really matter? A lesson from the debate of environmental disclosure and firm performance. Journal of Organizational Management Studies 2012, Article ID 213910.
Bachelet, M., Becchetti, L., Manfredonia, S., 2019. The green bonds premium puzzle: the role of issuer characteristics and third-party verification. Sustainability 11(4), Article ID 1098. Bodie, Z., Kane, A., Marcus, A., 2011. Investments and portfolio management. McGraw-Hill Education, New York.
Choudhry, M., 2004. Corporate bonds and structured financial products. Elsevier Butterworth Heinemann, Amsterdam.
Climate Bonds Initiative, 2015. The Paris green bonds statement. Online.
Climate Bonds Initiative, 2016. Bonds and climate change- the state of the market in 2016. Online.
Climate Bonds Initiative, 2018. Bonds and climate change- the state of the market 2018. Online.
Climate Bonds Initiative, 2019. 2018 Green bond market summary. Online.
CICERO, 2016. Framework for CICERO’s ‘Second Opinions’ on green bond investments. Online. Della Croce, R., Kaminker, C. and Stewart, F., 2011. The role of pension funds in financing green growth initiatives. OECD Publishing, Paris.
Dhaliwal, D., Li, O., Tsang, A., Yang, Y., 2011. Voluntary nonfinancial disclosure and the cost of equity capital: the initiation of corporate social responsibility reporting. Accounting Review 86, 59-100.
Dimson, E., Karakas, O., Li, X., 2015. Active ownership. Review of Financial Studies 28(12), 3325-3268.
Eckbo, B., Masulis R., Norli O., 2007. Security offerings. In: Eckbo, B., Handbook of Corporate Finance. North Holland, Amsterdam, pp. 233-373.
40
Fama, E., French, K., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33, 3-56.
Flammer, C., 2013. Corporate social responsibility and shareholder reaction: the environmental awareness of investors. Academy of Management Journal 56(3), 758-781.
Flammer, C., 2018. Corporate green bonds. SSRN, Online.
Forbes, 2017. Should you invest in ‘green bonds?’. Forbes, June 29, Online.
Guenster, N., Bauer, R., Derwall, J., Koedijk, K., 2011. The economic value of corporate eco-efficiency. European Financial Management 17(4), 679-704.
Hillier, D., Clacher, I., 2011. Fundamentals of corporate finance: European edition. McGraw-Hill Education, New York.
International Energy Agency, 2014. World energy investment outlook. Online.
International Energy Agency, 2016. Emissions data critical for targeted climate action. Online. Jo, H., Kim, H., Park, K., 2014. Corporate environmental responsibility and firm performance in the financial services sector. Journal Bus Ethics 131, 257-284.
Klassen, R., McLaughlin, C., 1996. The impact of environmental management on firm performance. Management Science 42(8), 1199-1214.
Krüger, P., 2015. Corporate goodness and shareholder wealth. Journal of Financial Economics, 115(2), 304-329.
MacKinlay, A., 1997. Event studies in economics and finance. Journal of Economic Literature 35, 13-39.
Mandel, K., 2015. The rise and rise of green bonds. ENDS Report 479, 44-48.
Mathews, J., Kidney, S., 2012. Financing climate-friendly energy development through bonds. Development Southern Africa 29(2), 337-349.
Morningstar Investing Classroom, 2015. The process of issuing bonds. Online.
Myers, S., Majluf, N., 1984. Corporate finance and investment decisions when firms have information that investors do not have. Journal of Financial Economics 13(2), 187-221.
41
NOS, 2019. ‘Donkergroene’ staatslening haalt met gemak 6 miljard euro op. NOS, 21 May, Online.
OECD, 2015. Green bonds: Mobilising the debt capital markets for a low carbon-transition. OECD Publishing, Paris.
Park, S., 2018. Investors as regulators: green bonds and the governance challenges of the sustainable finance revolution. Stanford Journal of International Law 54(1), 1-47.
Schroders, 2015. Green bonds – a primer. Online.
42
Appendix
Table 10. Summary statistics of the returns, market returns and risk factors
Obs. Mean Std. Dev. Min Max
Ri,t 60,120 .0001731 .0192426 -.4063836 .2929577 Firm-specific RM,t 60,120 .0001787 .0090822 -.0848752 .0575856 Global RM,t 60,120 .0002157 .006538 -.0490418 .0259849 RM,t - Rf,t 60,120 .000277 .0063836 -.0511 .0239 Rf,t 60,120 .0000207 .0000405 0 .0001 SMBt 60,120 1.26e-06 .0032734 -.0199 .0148 HMLt 60,120 -.0001743 .0032265 -.0154 .0158
This table depicts the summary statistics of the daily returns and daily market returns used in the market model estimations. The table also depicts the summary statistics of the daily global risk factors used in the global three factor model estimations: the daily global excess market returns, the daily risk-free rate, the daily SMB factor and the daily HML factor. The sample consists of N=501 firms that issued corporate green bonds between 2014 and 2018. The estimation window is the 120 days prior to the first day of the event window around the announcement of these bonds. Therefore, the number of observations per input variable equals 501*120 = 60120.
Table 11. Summary statistics of the estimated alphas and betas
Approach Variable Obs. Mean Std. Dev. Min Max
1 α̂i 501 -5.66e-06 .0014888 -.007788 .0065806 1 β̂i 501 .9580372 .4658225 -.4317795 3.98491 2 α̂i 501 -7.61e-06 .0015319 -.0074298 .0066216 2 β̂i 501 .9062591 .615541 -.765439 3.839513 3 α̂i 501 .000093 .0016515 -.0090342 .0066577 3 β̂M,i 501 .9449437 .5941841 -.6877655 4.485862 3 β̂SMB,i 501 -.0174311 .8993632 -2.534061 3.633254 3 β̂HML,i 501 .1438695 1.278273 -5.69584 5.926345
