Does Size Matter?
An event study exposing the relative size of a green bond issue and its impact on value
creation for corporations.
Bachelor’s Thesis 15 hp
Department of Business Studies Uppsala University
Fall Semester of 2020
Date of Submission: 2021-01-15
Sophia Bragd Lovisa Lindgren
Supervisor: Jiří Novák
Abstract
Capital markets have changed profoundly since green bonds were first introduced in 2013 as a way of financing programs benefiting social and environmental sustainability. The purpose of this study is to investigate whether green bond issues are related to value creation as compared with conventional bond issues, and how the relative size of a bond issue may impact this relationship. Using bonds issued on Nasdaq Stockholm from 2014 to 2020, our study finds no significant abnormal returns for green bonds, the benchmark of conventional bonds nor the comparison of the means. Further, we could not find an interaction effect between the relative issue size and the green bond. Hence, this study finds no indication that green bonds create value. However, we show that relative size has a positive and significant regression coefficient in all models, meaning that the larger the relative size of a bond issue, the more the stock price is expected to increase.
Keywords: Green Bonds, Bonds, Bond Size, Relative Bond Size, Greenwashing, Abnormal Returns, Event Study.
Table of Contents
1. Introduction 1
1.1 Background 1
1.2 Problematization 2
1.3 Purpose 5
1.4 Disposition 5
2. Theory 6
2.1 Theoretical Framework 6
2.1.1 Capital Structure Irrelevance 6
2.1.2 Efficient Market Hypothesis 7
2.1.3 Signaling Theories 7
2.1.4 Greenwashing 8
2.2 Hypotheses 9
3. Methodology 10
3.1 Sample and Data Collection 10
3.1.1 Data Collection 10
3.1.2 Sample and Research Lapse 10
3.1.3 Sample Limitations 12
3.2 Variables 13
3.2.1 Dependent Variable 13
3.2.2 Independent Variables 13
3.2.3 Control Variables 16
3.3 Event Study 17
3.3.1 Event Window 17
3.3.2 Calculations 18
3.3.3 Multiple Linear Regression Analysis 20
3.4 Criticism of Chosen Methodology 20
3.4.1 Criticism of Event Study 20
3.4.2 Criticism of Regression Analysis 22
3.5 Descriptive Statistics 23
4. Results & Analysis 25
4.1 Hypothesis 1 25
4.2 Hypothesis 2 28
4.3 Hypotheses Results 30
5. Conclusion 31
5.1 Limitations and Further Research 31
References 33
Publications 33
Other Sources 35
Appendices 38
Appendix 1 - Issued Bonds 38
1. Introduction
1.1 Background
Few have missed the sustainability hype that has characterized the past decade (Clark, Feiner
& Viehs, 2015). It is commonly known that our current way of living cannot be sustained without causing severe damage to our planet and this has led to several climate agreements, perhaps most famously the Paris Agreement (Maizland, 2020). An important piece in the sustainability puzzle is the role of capital markets and how funds are channeled into future projects.
The capital raised from green bonds is dedicated to programs benefiting social and environmental sustainability, in contrast to conventional bonds where the capital raised is used for general corporate purposes (ICMA, 2018). The first green bond was issued in November 2008 by the World Bank and SEB (The World Bank, 2018). Five years later, Vasakronan followed suit as the first company to do so in November 2013 (SEB, 2013), which later was followed by SCA who became the first listed company to issue a green bond in 2014 (SCA, 2014). The green bond market in 2019 was worth USD 257.7 billion (Fatin, 2020), as opposed to USD 11 billion in 2013 (OECD & Bloomberg, 2015), culminating in an average annual increase of 69.03 %. Even though the conventional bond market is substantially larger, with a total worth of USD 128.3 trillion in August 2020 (ICMA, 2020), green bonds are a growing phenomenon (Gunter, Kraemer, & Vazza, 2019).
The Global Sustainable Investment Alliance (GSIA) is the monitoring organization of sustainable investing and releases a report on developments in sustainable investing biannually: Global Sustainable Investments Review (GSIA, 2018; Jain, 2019). The most recent was published in 2018 and shows that sustainable investing grew 34% globally from 2016 to 2018. According to the 2017 Cone Communications CSR Study (2017) conducted in the United States, 63% of the general public and 71% of millennials hope companies will lead the way in social and environmental change. According to Morningstar (2020), total assets held in sustainable mutual funds by the end of September 2019 were approximately USD 900 billion. By the end of September 2020, the same number was USD 1,258 billion,
nearly a 40% increase in one year (Ibid). A 2018 study by Bank of America found that i) 9 out of 10 wealthy millennial investors consider ESG work to be an important factor when choosing companies to invest in, and ii) 78% of millennials have reviewed their portfolios with regards to ESG history and ratings (Kull, 2020). In addition, a transfer of funds from boomers to more environmentally conscious millennials is currently occurring, and likely to continue in the coming years (Ibid). The trend with sustainable investing seems to be here to stay.
1.2 Problematization
A study by Arvidsson (2014) concludes that stock market actors view CSR information with sparse interest and that corporations' CSR work does not bring actual value to the stock market. On the contrary, Cellier and Chollet (2011) find that high CSR ratings lead to a positive reaction to the stock market. Ender and Brinckmann (2019) further concluded that positive CSR relevant news generates abnormal positive returns, in other words, create value.
Pérez, López-Gutiérrez, and Salmones (2019) find similar results.
Flammer (2020) provides two reasons why corporations would be interested in issuing green bonds as opposed to conventional bonds. First, since green bonds are restrained in what the companies can fund, it is a credible signal that they will undertake certain investments to improve environmental and social sustainability. Second, issuing green bonds could be a way for companies to portray themselves as environmentally and socially conscious even if they do not intend on operating more sustainably (Flammer, 2020).
Issuing green bonds may incur improved brand image and overall sustainability assessment scores, but it is also associated with additional costs. The International Capital Market Association (ICMA) has developed the Green Bond Principles (GBP) which are recommendations for corporations issuing green bonds (ICMA, 2018). These recommendations include that the issuing companies report back to the investors post-issue to guarantee the appropriate investment of funds, track their project by a third-party actor, and more (Ibid). The tracking by a third-party actor could be certifications, verifications, or green bond ratings (ICMA, 2018). Third-party tracking, as well as reporting back to investors, are two out of numerous cost centers associated with issuing green bonds. However, the ICMA
principles are merely recommendations and the ICMA can thus not enforce legal actions if corporations choose to not follow them (Ibid).
Even though legal actions will not be pursued by ICMA, other consequences can be costly for
“green” bond issuers if they do not follow the guidelines of green bonds. In certain countries, it could be classified as false advertising, which may lead to lawsuits, high associated costs, and a damaged reputation (Heilpern, 2016). Also, not following the green bond frameworks can be damaging to the brand of the issuing company (Scott, 2014). Brand damage could potentially lead to loss of brand trust and loyalty, where PwC Consumer Insight Survey (2018) states that more than 33 % of consumers ranked trust as one of their principal three reasons when selecting a retailer, and that “trust in brand” was the second most cited in the purchase decision process. The great importance of maintaining a good brand reputation, and the fact that most green bonds today are certified (Ehlers & Packer, 2017), indicate that the risks associated with issuing a green bond without actual commitment are difficult to justify.
Many recent studies have researched the effect of green bond issues on the stock price, however, the outcomes are conflicting. Tang & Zhang (2018) found positive effects on the stock price after the issue of bonds classified as green by Bloomberg, as did Wang et al.
(2020) on green bonds issued in China. Flammer (2020) observes a positive stock price reaction to the issue of green bonds, especially for a corporation’s first green bond, or when the bonds are certified by third-party actors. On the contrary, a study from January 2020 suggests no statistical evidence that issuing green bonds in the Nordics affects the stock price (Aste & Åström, 2020). In Europe, statistical evidence was only found for bonds issued in 2018 or later (Larsson & Zakrisson, 2020). Aste and Åström (2020) propose several reasons why their results might not be statistically significant, two of them being due to their small sample and that Nordic countries may not value the issue of green bonds. However, the differences in these results might also be explained by other variables that are yet to be taken into consideration.
All the above have designed their studies merely to investigate whether a positive or negative stock market reaction exists or not, but not why. The only exception is Flammer (2020) who has accounted for certification. As a result, previous studies have not accounted for the factors of corporation size or bond size. These variables may be able to explain the
discrepancies in the outcomes of their studies - simply issuing a green bond may not be enough.
The importance of this variable is mainly based upon the theoretical concepts of
“greenwashing”. Greenwashing occurs when actual sustainability efforts made are deemed too small in comparison to how it is portrayed, and the whole project is deemed unconscientious (Jeevan 2014). An example is the Volkswagen emission-scandal which had them set aside billions of euros to cover costs, and led to their first quarterly loss in 15 years (Hotten, 2015). Being accused of greenwashing is costly, and the risk of being accused of this is, therefore, an incentive for corporations to make a substantial effort. For green bonds, it implies that the size of them, relative to the size of the corporation, matters. As an example, a green bond of SEK 10 000 000 should have a more positive impact than a green bond of SEK 5 000 000, for the same issuing corporation, as it has more substance and affects the company more.
We find this variable intensely noteworthy, and would on these grounds like to unfold the topic of research examining and comparing how the relative size of a green bond emission affects the stock price of the issuing corporations. Thus, our research fills the gap in current research on the green bond market by investigating the important variable of relative bond size, where corporation size and bond size are taken into consideration. Besides, these studies have not benchmarked their results with non-green bonds and their market reactions. This could help explain why previous research has found varying results, and we further fill the research gap by including conventional bonds in our sample to isolate the effect of a bond being “green”.
In sum, green bonds embody a way to incorporate sustainability into one’s business.
Preceding research indicates that issuing these bonds creates value in the form of abnormal returns since people value the signals of environmental commitment. However, not all papers have been able to prove significance. Previous studies have not accounted for variables such as firm size and bond size, which could affect the result. Relative bond size matters as these provide strong indications of how genuine a corporation's environmental commitment is.
Issuing a larger green bond could henceforth imply that a corporation is more committed to fighting the environmental challenges we see today. Whereas, issuing a relatively smaller
green bond could imply that corporations seek to utilize the upsides of a green bond issue as a marketing ploy, in other words greenwashing.
1.3 Purpose
The purpose of this study is to investigate whether green bond issues are related to value creation as compared with conventional bond issues, and how the relative size of a bond issue may impact this relationship. If green bonds are more related to value creation this poses a financial incentive for corporations to issue green bonds instead of conventional bonds, and if the relative size is positively correlated with value creation this could further incentivize full commitment and genuineness with their green investments.
1.4 Disposition
In the second section, the theoretical framework which lays ground for our work will be presented. The second section will eventually lead to the formulation of hypotheses relevant to answer the study’s research questions for our thesis. The third section includes data collection, limitations, relapse, and sample. Further, variables of the study and the chosen methodologies are introduced. The fourth section will present the results, respond to our hypotheses, and analyze the results with the theoretical framework presented earlier. Lastly, the fifth section will include the conclusions of our work, thesis limitations, and suggestions for further research.
2. Theory
2.1 Theoretical Framework
2.1.1 Capital Structure Irrelevance
Corporate bonds constitute a way for a company to obtain debt, and therefore, bonds change a company's financial leverage. The capital structure irrelevance theorem proposes a relationship between leverage and the cost of equity. Modigliani & Miller (1958) built their discussion on the assumption of perfectly efficient capital markets (no asymmetric information, taxes, bankruptcy costs, nor agency costs), stating that choice of funding is irrelevant for corporate value which is calculated as the present value of the future earnings of the corporation (Ibid). These assumptions rarely hold and are more commonly used to show why capital structure matters. Ross (1977) showed how corporations can issue bonds, to give investors a strong indication of future growth and thereby increase the stock price, also named the signaling theory of debt. On the contrary, increasing debt makes stock ownership riskier, leading to a higher weighted average capital cost which decreases the value of the corporation and subsequently predicts a decline in stock price (Berk & DeMarzo, 2017, p. 586). A factor not included in these models but central to this paper is that green bonds bear the additional information of commitment to a new, green project (Flammer, 2020).
Many researchers and theories agree that capital structure matters, but not on what it is affected by and how it is affected. However, preceding research does not account for a bond being green or not. A green bond brings new information on sustainability focus and using conventional bonds as a benchmark, this is expected to filter out the effects that are the same for green and conventional bonds. With this in mind, and that previous studies do not provide a unanimous prediction of the effect of debt issue on stock price, we assume capital structure does not affect the expected change in stock price after a bond issue.
2.1.2 Efficient Market Hypothesis
If markets are efficient, we expect them to respond to bond announcements to adjust the stock price to the new situation (Fama et al., 1969). The efficient market hypothesis states three forms of efficiency: (i) the weak form where only historical prices are incorporated into prices, (ii) the semi-strong form where the price reflects all public information and new information is quickly incorporated into the price, and (iii) the strong form where both public and private information lay ground for the price and no information can be used to generate abnormal returns (Fama, 1970).
As we assume the semi-strong form of the EMH, we expect the value of the new information (the issue) to incorporate rather quickly into the stock price after the information has reached the public.
2.1.3 Signaling Theories
Signaling theories are valuable when two parties do not have access to the same information, usually organizations and individuals (Connelly et. al, 2011), and are employed to shrink the information asymmetry between those parties (Spence, 2002). It is considered a well-functioning corporate strategy when the signaling generates a better outcome than the status quo (Connelly et. al, 2011). Corporations know more than their investors about the details of their operations, and due to this, corporations are senders who choose when and how to signal this information. The receivers, the investors, choose how to deduce and use the information. Stiglitz (2000) defines two types of signals where the information sharing is of essence and these are i) information about quality and ii) information about intent. A quality and credible signal are easy to observe and costly to mimic (Connelly et. al, 2011).
Signaling theory can be applied to bonds, and even more so on green bonds. Issuing green bonds is a corporate strategy, where the corporations have more information than investors on their environmental and social sustainability, and investors lack information to assess the corporation's pledges. Issuing a green bond could signal the intent to investors that the corporation is devoted to working towards a more sustainable future, which would mean that the corporate strategy would be considered adequate. These signals are deemed well-functioning and hard to mimic among competitors due to the associated costs of
following the GBP guidelines of certifications, and the risks associated with issuing green bonds while not being committed such as brand damage and legal actions.
However, signaling theory is not solely focused on the sending of a signal, but also investigates the response to the signal (Connelly et. al, 2011). Examples of these are presented in the marketing studies where the customers are the receivers (Ibid). The response from the receiver is feedback of the signal sent (Ibid). The feedback could, in the case of green bonds, be an instant positive or negative market reaction to the green bond issue. The feedback could also be the increased demand for sustainable alternatives that force corporations to act more sustainably.
2.1.4 Greenwashing
Greenwashing is a marketing ploy where corporations perform environmental efforts mainly to improve their reputation (Poh & Chasan, 2019), and is according to Jeevan (2014, p. 3) an
“act of misleading consumers regarding the environmental practices of a company or the environmental benefits of a product or services”. Examples of these are car companies claiming their cars are green or putting pictures of green meadows on a flask filled with destructive substances (Jeevan, 2014). Green bonds could potentially be considered greenwashing by stakeholders if the corporations do not employ the funds raised through the bond in projects considered green enough by stakeholders.
To combat greenwashing, there are different certifications for green bonds. As mentioned earlier, the ICMA has presented a set of voluntary guidelines for issuing green bonds which today are known as the Green Bond Principles (GBP). These include four principles: that the investments made are to be reported in legal documents, there are processes for the evaluation of projects which use capital from green bonds, the green bond capital should be kept separated from the remaining assets of the capital, and that the issuer needs to report back to the investor at least once a year (Brundin, 2015; ICMA, 2018). Furthermore, it is recommended to allow external review in the form of second party opinion, verification, certification, or green bond rating (ICMA, 2018). These principles outline the criteria of most certifications. For example, the largest external reviewer in 2018 and 2019 according to Climate Bonds, Sustainalytics, follows the GBP principles when certifying green bonds (Climate Bonds Initiative, 2020; Sustainalytics, 2020).
2.2 Hypotheses
Previous research has found rapid market reactions for when new information is available, indicating that the semi-strong form of the efficient market hypothesis holds. Therefore, the market should react to the issue of a green bond and thus affect the stock price as there will be new information to lay ground for the asset valuation. The issue will also send out a signal to the investor, reducing the information asymmetry, that the company henceforth pledges to be more sustainable (Connelly et. al, 2011; Spence, 2002). It will be considered a well-functioning signal since it is deemed difficult to be imitated by competitors due to the high costs associated with issuing green bonds. The pledge will also be attractive to investors due to the changes in investment trends that seem to value sustainability more today than before.
Our first hypothesis is therefore phrased as follows:
H1: Issuing green bonds creates more shareholder value relative to conventional bonds.
Investors today are more aware of companies using greenwashing, making themselves look more sustainable than they are (Poh & Chasan, 2019; Jeevan, 2014). The signaling value of the bond size relative to corporation size can therefore inform the investor of how committed the issuing corporation is - signaling its intentions. Issuing a large green bond might therefore have a stronger impact on the stock value due to the signaling value of the issue size, and vice versa, due to the underlying intentions of the corporations.
We have therefore phrased our second hypotheses as follows:
H2: Relative size is more strongly associated with shareholder value creation for green bonds than for conventional bonds.
3. Methodology
3.1 Sample and Data Collection
3.1.1 Data Collection
This study uses secondary data through a sample of companies listed on Nasdaq Stockholm and Nasdaq First North, that have issued green and/or conventional bonds between the years 2014-2020. Both our bond data and our stock data is collected from Refinitiv’s Eikon. All issues are labeled green or conventional by Eikon, which defines the green bonds as those who follow the GBP principles. The Nasdaq Stockholm PI Index will be used as the base for calculating expected returns and formulas derived from there.
Refinitiv’s Eikon has all the above-mentioned financial data in their database, which increases the validity of the research since it can easily be replicated.
3.1.2 Sample and Research Lapse
Table 1. Crosstable with bond type issued, as well as the industry which is divided into real estate or others.
Our research paper consists of 366 observations, with 55 being green bonds (see table 1), constituting 15 % of our sample. These are divided into industries where circa half of our sample is in the real estate industry. However, a large majority of the green bonds are issued by real estate firms, and as residential development is a highly capital-intensive industry we will add control variables controlling for this.
Figure 1. Bonds per year of issue, divided into green and conventional bonds.
In figure 1 we have divided our sample into years of issue, where most issues of green bonds occurred in 2018-2020 and conventional bonds in 2017-2019. The first Nasdaq Stockholm listed company in Sweden to issue a green bond was SCA, which did so in March 2014 (SCA, 2014). The data will be retrieved during November 2020 and we have therefore chosen October 31st 2020 as the cut-off date and will gather data starting from January 1 st 2014.
Further, we will exclude corporations that have declared dividends, had a change in a key executive, announced a merger, or announced new products within the event window, since these could intervene with the effects of the bond issues.
Table 2. Research lapse generated by our limitations.
Our research lapse is explained in table 2. The extreme outliers are defined by computing box plots (in the statistics program SPSS) of our dependent and independent variables. The values outside the whiskers of the box plot are marked with an asterisk and thereafter defined as extreme outliers. We found five of these extreme values in our dependent and independent variables and removed these observations.
We will also exclude financial institutions in our sample. Tang and Zhang (2018) state that financial institutions issuing bonds differ from corporations issuing bonds since financial institutions “issue green bonds to make green loans to and invest in other firms to finance other firms’ projects” (Tang & Zhang, 2018, p. 6). This might make the market reactions less trustworthy since the signals will differ from corporate issues that are fully responsible for following the guidelines of their green bonds. We furthermore noticed that our results would be skewed by the vast amount of bond issues constituted by financial institutions.
3.1.3 Sample Limitations
This study will be conducted on all green bonds issued in Sweden by companies listed on Nasdaq Stockholm and Nasdaq First North at the time of issue to be able to evaluate the stock price. Sweden has been chosen due to its high rank, 8 thout of 180 countries, in the 2020 EPI - Environmental Performance Index (Yale University, 2020), which is a sign of an environmentally conscious population. It has also been chosen due to the relatively many green bonds issued compared to other countries. To generate a larger sample, the study could use green bond issues in the entire Nordics, however, this idea was dismissed as Sweden has issued about 50% of the green bonds in this region and this might skew the result. If adjusted
for, the sample size for each of the other countries might be too small to generate reliable or significant results.
3.2 Variables
This study is ordained to research whether the issue of green bonds is related to value creation, and how the relative size of a green bond issue may impact this relationship. False correlations occur regularly where the regression shows that a relationship exists, but in fact, it depends on one or several other variables (Laerd, 2018). To avoid this, we will use several control variables: industry, year of issue, and interest-coverage ratio.
3.2.1 Dependent Variable
Our dependent variable is the market reaction to an event: whether the issue of a green bond generates positive or negative reactions. The market reaction will be measured by calculating the average abnormal returns (AAR). By choosing AAR as our dependent variable, we can i) measure if an event is significant to investors and ii) determine the effects of relative bond size. This will be measured through the market model where we will subtract the expected returns from the actual stock prices of our sample and divide this number by the number of observations. This is elaborated on, with formulas, in 3.3.2.
3.2.2 Independent Variables
Relative Size of the Bond Issue
Our first independent variable is the relative size of the bond issue (bond size divided by corporation value). With this variable, we can examine whether the market reaction (the dependent variable) varies with the relative size of the issue, and thereafter determine if the size of a green bond issue matters. The reasoning behind this stems from the theoretical frameworks of greenwashing and signaling theory – where issuing a relatively larger bond might indicate that the company is more committed to green projects, and issuing a smaller relative bond could suggest that a company is issuing the bond to make themselves look better.
We decided to use the relative size as our independent variable since measuring the bond size in absolute numbers is incompatible with the greenwashing theory. Also, if solely using the size of the bond, we would need to ensure that all issuing corporations are the same size at
the issue date to be able to draw conclusions from the data. We deemed that statement to be practically impossible to follow and still have a significant sample size.
To determine the relative size of the bond issue we have measured the value of corporations.
There is no consistent measurement for determining the value of corporations. Forbes Global 2000 utilizes four different methods of measuring the largest corporations in the world today – total assets, total sales, total profits, and market capitalization (Coyne, et. al, 2020). Dang, Li and Yang (2018) furthermore states that if all companies in the sample are listed, market capitalization is a good measurement for the corporation’s value. Market capitalization is a universal way of measuring companies and is easy to obtain from the stock exchanges. For these stated reasons, we believe using market capitalization is a good measure of the corporation’s value.
However, Dang, Li and Yang (2018) does mention that when using market capitalization as a measurement of corporation value, it is of great importance to understand that it can be affected by the capital structure and corporate performance. Since this can change swiftly, our study will investigate a change in the capital structure, we hope to evade this problem by measuring the market capitalization on the last day of the estimation window and not during the event window. This is to escape the problem that market capitalization can increase as an effect of our event (issuing green bonds), and make sure the right relative size is obtained.
Nonetheless, the relative size variable was not found suitable due to the many outliers and the sample not being normally distributed. Therefore, the independent variable was logarithmized with the base of ten, lg(relative size). Using logarithms is an effective way to compare values since they all will have the same base and therefore be “directly interpretable as approximate proportional differences” (Gelman and Hill, 2006, p. 60-61).
Figure 2. Box plot of the relative sizes per bond type.
The independent variable of lg(relative size) is shown in figure 2. As seen in the figure, the sizes are comparable.
Green Dummy
As our proposition makes a distinction between bonds classified as green or conventional, this has to be clearly expressed in the regression analysis. We will therefore create a variable for the classification of bonds, GreenDummy. It is a dummy variable since we are using Eikon as a database where the bonds are already classified as either green or conventional, so there is no scale or similar to how green a bond is. The dummy variable equals 1 if a bond is green, and 0 if a bond is conventional.
Interaction Effect
Given the theoretical framework and in particular greenwashing, our hypotheses suggest size will affect green and conventional bonds differently. This cannot be accounted for using only a dummy variable for green bonds and a variable for size, as the size variable will combine both kinds of bonds. As a result, to find the joint effect of relative bond size and the bond being classified as green, we create an interaction effect by multiplying the green dummy with the size variable. Since the GreenDummy variable equals 0 if the bond is conventional the interaction effect will also equal 0 for conventional bonds and is therefore only applicable to green bonds.
3.2.3 Control Variables
Industry
As about half of our sample are corporations in the real estate industry, this could skew the result in case there are industrial differences in how green bond issues affect corporations.
Ideally, we would create a category for each industry, however as there is only a limited number of green bonds issued by listed companies, we risk having too small sample sizes for the other industries and hence, not receive significant or reliable results. To control for skewness caused by an overrepresentation of real estate corporations, a dummy variable for real estate corporations is included in the regression analysis. The dummy variable is constructed by assigning real estate companies the value of 1, and the rest the value of 0.
Interest Coverage Ratio
The financial state of a corporation may influence the market reaction to a debt issue, both due to the expansion of the balance sheet, and that the purchasers of the bond need to be able to assess the corporation's ability to pay back its debts (Corporate Finance Institute, 2020).
For the latter reason, many studies have used credit rating as a proxy for a corporation’s financial state. However, different rating institutions use different criteria when rating corporations, and credit ratings, therefore, become subjective. Instead, we will follow the example of Godlewski et. al. (2013) and use the Interest Coverage Ratio (ICR) to account for the financial state. The ratio is defined as EBIT divided by interest expenses and is commonly used by financial institutions, investors, and creditors, to define the risks of lending capital to a company (Ibid).
The risk of a bond is highly likely to impact its demand and the ease of obtaining new capital may influence the stock price. The ICR is more directly relevant for debtholders, but it is also an indicator of risk on equity. Because debtholders have a prior claim in the event of liquidation, increasing debt increases the risk for shareholders. A high ICR signals that the company can afford its debt, indicating there are enough assets left for the shareholders, while a low ICR is indicative of higher risk for shareholders. The ICR is a continuous variable and will be added to our regression to account for the differences in companies’
ratios. This variable will also help account for the large portion of green bonds issued by real estate firms.
Year of Bond Issue
As time changes, so does external effects on the stock market. Different years have different challenges and macroeconomic factors play into the stock market; the global pandemic of 2020 being one of them. To account for these external effects, we will include six dummy variables for the year each bond was issued (2015-2020), where the dummy equals 1 if the bond was issued in that year. 2014 is the reference category, meaning that when the six dummies all equal 0, the bond was issued in 2014.
3.3 Event Study
To decide if the issue of green bonds is related to value creation, and how the relative size of a green bond issue may impact this relationship, this study will conduct quantitative research with a deductive approach. A deductive approach is to produce theories and thereafter apply them to hypotheses (Bryman, 2016). To test our hypothesis, we will perform an event study, which is a way of investigating the “behavior of firms’ stock prices around corporate events”
(Kothari & Warner, 2006, p. 4). The use of event studies has grown rapidly lately and has come to play an essential role in financial economics (Ibid).
3.3.1 Event Window
We are assuming a semi-strong form of market efficiency and will therefore have a short event window of 7 business (trading) days [-3,0,3], meaning we expect that the new information should affect the stock price quickly. Previous event studies have had rather long event windows, which can imply that these authors did not believe that the event information is quickly incorporated in the stock price (McWilliams & Siegel, 1997). The event window is the period relevant to examine whether the event of issuing a bond affects abnormal returns or not (MacKinlay, 1997). MacKinlay (1997) further states that the event window is stretched with multiple days depending on the event study, but at least includes the day of the announcement and the following day. We will examine the 3 trading days after the event as the market does not always react immediately (Ibid). Due to the easy access and spread of information today we believe this timeframe should be sufficient. Since we are assuming a semi-strong form, insider information may occur and therefore it is of essence to examine the period prior to the event (Ibid). We will hence measure 3 business days before the event.
The estimation window is needed to calculate the expected, normal, return. Through the data extracted from the estimation window, we can calculate the abnormal returns of the event window (MacKinlay, 1997). MacKinlay (1997) further explains that it is most common to use the time before an event, and the time frame for this is suggested to be 120 days when using daily data and assuming the efficient market hypothesis. It is to be measured before the event window to “prevent the event from influencing the nominal performance model parameter estimates” (Ibid, p. 15). Thus, our estimation window will begin 124 trading days prior to the event and end on the fourth trading day before the event [-124, -4].
Figure 3 . Illustration of the study’s estimation and event window. The timeline of the event study will be 124 trading days prior to the event and end 3 trading days after the event has occurred.
3.3.2 Calculations
To measure how bond issues affect the stock price of the issuing company, we need to calculate the abnormal returns during the event window. Abnormal returns are the difference between the actual returns and the expected returns. Before calculating the abnormal returns, we need a model for normal returns. MacKinlay (1997) presents the following market model for normal returns, where the expected value of the error term is 0 and the variance of the error term is the squared standard deviation of the error term:
The abnormal returns are defined as the residuals after estimated alpha and estimated beta multiplicated with the return of the market portfolio, have been subtracted from the observed return of the stock. The market portfolio (R mt) in this study is the Nasdaq Stockholm PI Index. The formula for abnormal returns is as follows (MacKinlay, 1997):
The average abnormal returns per day during the event window are thereafter calculated. The abnormal returns for each company are summed up and divided by the number of observations (MacKinlay, 1997):
Then, the cumulative average abnormal returns are calculated by summing up the average abnormal returns in all event windows. T1 represents the first day of the event window and T2 represents the seventh and final day. (MacKinlay, 1997).
To decide whether our results are statistically significant, we will conduct a one-sample t-test.
Performing this test requires the variance of CAAR for which we need the variance of AAR first (MacKinlay, 1997):
the variance of CAAR can now be calculated. T1 represents the first day, and T2 represents the final day in the respective event window (MacKinlay, 1997).
After the variance of CAAR is calculated, all values needed to perform the t-test are available (MacKinlay, 1997):
Further, to test whether issuing green bonds creates more shareholder value relative to conventional bonds, we analyze the difference in means between green bonds and conventional bonds. This is done through a two-sample t-test (Kent State University, 2021):
3.3.3 Multiple Linear Regression Analysis
After calculating the Average Abnormal Returns ( AR ), we can test the second hypothesis and determine how the relative size impacts the relationship. To do so, we conduct a multiple linear regression analysis of the independent variables of relative bond size, green dummy, and interaction effect; and the dependent variable of AR . It is important to ensure that there is a causal correlation – that the independent variables affect the dependent variable, AR – and that there is no reverse causality (Bryman & Bell, 2013). In our case, this could be that companies with higher abnormal returns issue relatively larger bonds. To evade the risks of reverse causality it is necessary to add control variables to the regression (Leviton, 2001).
Our study will include the control variables of industry segment, interest-coverage ratio, and year of issue.
The following regression model will be used:
AbnormalReturn (Y) = α + β GreenDummy + β IndustryDummy + β InterestCoverage + β YearofIssue + β Size + β Size*GreenDummy + e
Table 3. Summary and explanation of variables in regression models.
3.4 Criticism of Chosen Methodology
3.4.1 Criticism of Event Study
Since we will conduct quantitative research with a deductive approach in this thesis, Bryman (2016) states that a certain amount of logic is needed to produce theories and thereafter apply them to hypotheses. After gathering data, a researcher’s perception of a theory can change
and thus modify the original theoretical framework which hollows out the deductive approach (Bryman, 2016). Since our research topic is not completely new, we hope to evade this problem by analyzing the theoretical framework already used on this topic. We will furthermore conclude our theoretical framework and hypotheses prior to gathering and analyzing the data.
McWilliams and Siegel (1997) conducted a study regarding the event study implications and have concluded three main points to consider ensuring a valid research conclusion. First, is the consideration that other factors can affect the value of the stock price and it is hard to isolate the effect of the measured event (Ibid). McWilliams and Siegel (1997) call these
“confounding events”, and it has to be assumed that these do not exist to ensure a valid research conclusion. Confounding events could be “declaration of dividends, an announcement of an impending merger, signing of a major government contract, an announcement of a new product, filing of a large damage suit, announcement of unexpected earnings, and change in a key executive” (Ibid, p. 634). The second implication is to conduct an event study, we must assume that the market is efficient as this lays the ground for an event study (McWilliams & Siegel, 1997). The effect of the stock price should encompass all relevant and available information traders, and the newly published information should rapidly be incorporated into the stock price (Ibid). The last point to consider according to McWilliams and Siegel (1997) is that the event could be foreseen or have leaked before the official announcement. The implications of these in terms of event studies is that it is difficult to establish when traders received the information and thereafter (assuming an efficient market) affect the stock price. Thus, it would be impossible to draw any conclusions regarding the data set as people would receive the information at different times.
We have studied these three considerations presented by McWilliams and Siegel (1997) and have made decisions to evade the implications of receiving an invalid result. To escape the effects of “confounding events” we have disqualified corporations who have had any of the following events during the event window: declaration of dividends, change in a key executive, announcements of a merger, and announcements of new products. We will not take into consideration the other confounding events mentioned by McWilliams and Siegel as it is hard to define a “large” damage suit, an announcement of “unexpected” earning, signing of a “major” government contract. These subjective phrasings leave room for interpretation that can affect our sample depending on how we choose to decode the words. We further
deem that our large sample will eliminate the effects of not taking these confounding events into account. The two latter implications were evaded by assuming the semi-strong form of EMH and measuring the days prior to the event to be able to identify leaked information.
An event study always has limitations in the form of the joint-test problem, meaning that while testing the hypothesis it also measures market capital efficiency (Ibid). It functions as a mechanism to measure capital market efficiency, as abnormal returns based on a corporate event proves that the market is not efficient (Ibid). The research on market capital efficiency made by Ball and Brown (2014) and Fama et al. (1969), culminated the event study principles where the theory argued that the stock price should be affected immediately after a corporate event. To further evade the implications presented by McWilliams and Siegel above, such as the second point that we need to assume that the market is efficient, we are assuming that the market is semi-strong by having an event window of seven business days.
Also, we are measuring the three days prior to the event to evade the third, and last, implication mentioned of foreseen or leaked information affecting the event study.
3.4.2 Criticism of Regression Analysis
Certain assumptions must be met for the statistics software SPSS to perform the regression:
(i) the dependent variable (in this study, abnormal returns) is measured on a continuous scale, (ii) there are at least two independent variables which are continuous or categorical, (iii) independence of residuals, (iv) a linear relationship which will be checked for using a scatter plot, (v) the data shows homoscedasticity, (vi) does not show multicollinearity, (vii) there are no significant outliers, high leverage points or highly influential points - these will be trimmed from the data, and (viii) the residual errors are approximately normally distributed.
No. i) and ii) are fulfilled as our dependent variable is shareholder value (abnormal returns) which is measured in percentage, and both our independent and control variables are continuous (issue size, interest-coverage ratio). No. iii-viii is controlled for with a correlation matrix, VIF values, and descriptive statistics of our variables (found below).
One of the prerequisites for conducting a trustworthy regression is that multicollinearity, when explanatory variables affect each other, cannot exist. If the sample exhibits multicollinearity, the coefficients of the regression may be less precise which in turn affects the statistical significance of the result. To ensure our sample does not have multicollinearity,
we compute correlation values and variance inflation factors (VIF) which are compiled in the matrix below.
Table 4. Correlation Matrix and VIF Values.
For the results to be reliable the correlation value between variables should be less than 0.8, which all of them are except for Size*GreenDummy which has a correlation of -0.907 with GreenDummy (see table 4). Nevertheless, one factor being a dummy variable can explain the correlation. The other values are reasonable and we deem the regression to be reliable. For VIF, 1 is considered to be the lowest possible value but the value should not be higher than 10, and all observed values are below 4.
3.5 Descriptive Statistics
Table 5. Descriptive statistics of our sample.
Table 5 presents the descriptive statistics of our sample from the regression models. The number of observations, minimum value, maximum value, mean value, standard error,
median, skewness, and kurtosis, is presented to give the reader a greater understanding of the sample before we present our results.
The means of our dummy variables represent the portion of the sample that receives a value in that variable (i.e., the mean of GreenDummy is 0.153 meaning that 15% of our sample is green bond issues). To accurately perform a regression, the sample needs to fulfill certain assumptions (explained in section 3.4.2). Some of these assumptions are that the sample can not have too many outliers and the residual errors need to follow a normal distribution. Two measurements that control for this are skewness and kurtosis. Most of our variables have reasonable values of skewness, which should be around 0 to determine whether a sample is following a normal distribution for skewness or not. The variables of InterestCoverage, Size*GreenDummy, and year of issue ( 2015, 2016), have high values of skewness and can therefore not be said to follow a normal distribution for skewness. The values of kurtosis should be around 3 to be considered to follow a normal distribution for kurtosis. The variables of Industry, InterestCoverage, and year of issue (2015, 2016, 2018) receive high values for kurtosis, and can thus not be considered to follow a normal distribution.
The variables that receive poor results on both normal distribution measurements are the InterestCoverage variable and certain years of issue ( 2015, 2016). Since interest-coverage ratio has a minimum value being 247.89 and the maximum value being 63 000, this can explain the high values of skewness and kurtosis. The high values of skewness and kurtosis in 2015 and 2016 can be explained by the fact that few bonds were issued during those years compared to other years (presented in figure 1). The remaining variables, however, can be deemed to follow some sort of normal distribution, which indicates that the data is somewhat congregated near the mean.
4. Results & Analysis
4.1 Hypothesis 1
H1: Issuing green bonds creates more shareholder value relative to conventional bonds.
Table 6. Results from the years 2014-2020. ***, ** and * represents significance levels 1%, 5% and 10%.
The study finds no significant results on whether the issue of any of the two bond types generates positive or negative abnormal returns. The green bond issue reaction was a positive abnormal return of 0.0464% over a period of seven days, whereas the benchmark of conventional bond issues had a negative abnormal return of -0.0747%. However, neither of these results are significant, and therefore no conclusions can be drawn from them.
Table 7. Results from the years 2014-2020. ***, ** and * represents significance levels 1%, 5% and 10%.
Our study indicates that the issue of green bonds has a 0.1209% higher abnormal return than a conventional bond issue, but as this is not significant we cannot conclude whether issuing green bonds result in higher abnormal returns relative to the issue of conventional bonds.
Table 8. Results from regression models 1-3 using the entire sample. The values are presented in percentage form. The values in parentheses are t-scores. ***, ** and * represents significance levels 1%, 5% and 10%.
The GreenDummy on its own is not a good predictor of the data, proved by the insignificance of theGreenDummy coefficient in model 1 and the low F-test value and R square value of the model. When control variables are added, improving the model (see F-test and R square values for models 2 and 3), some control variables have statistical significance but the GreenDummy variable remains insignificant in all models. This regression table further examines the results from tables 6 and 7 and likewise concludes that whether a bond is green or not does not have any significant effect on abnormal returns. Hence, H1 is rejected.
Several factors may explain why we reject the first hypothesis. First, the signaling value of green bond issues is not valued by the Swedish market. Preceding research on Chinese, European, and global markets have found positive market reactions to the issue of green bonds. This is explained as a consequence of the reduced risk of capital, as well as current investment trends which value sustainability. However, Sweden is one of the most sustainability-conscious countries in the world. The results could therefore be explained by the fact that the Swedish market does not react to green bond issues because the signals these send as it might not be considered to be extraordinary events. Hence, the news of a green bond issue might not generate as big of a market reaction as it might in other less sustainably conscious countries. This statement is aligned with the propositions by Aste and Åström (2020) on the Nordic market when they were unable to prove that the issue of green bonds in the Nordics affects the stock price.
Marketing factors should also be considered. In our sample, bonds are only classified as green if they are certified. As a result, bonds marketed as green but not certified will be part of our conventional bonds sample. Although greenwashing usually turns out detrimental, it will only have a negative impact if discovered. Green bonds constitute a new product and the effect of marketing as green versus certifying as green is yet to be clarified. Hence, there is a possibility greenwashing in these instances go undiscovered, thereby omitting the negative impact.
Another source of error when only using certified green bonds is that certifying a green bond is a lengthy process, giving the market time to predict the issue. The process includes gathering investors, preparing legal documents, presenting frameworks for evaluation and communication with investors, and more. This process is public, and since it might improve the reputation of the company there is no reason to conceal it from investors. Therefore, it might be difficult to decide exactly when the new information reached the market, meaning there is a risk we might have missed the market’s actual reaction. It might also be explained by the signaling theory, where investors have sent feedback to corporations with demands of sustainable alternatives, which made the issue predictable.
The length of the event window may also matter, as a shorter (or longer) event window could have yielded different results. We assumed the semi-strong form of the EMH since preceding research has found rapid market reaction when new information is published, indicating that
this form holds. Nonetheless, we do not know exactly how efficient the market is and this joint-test problem could be an explanation for our non-significant results.
4.2 Hypothesis 2
H2: Relative size is more strongly associated with shareholder value creation for green bonds than for conventional bonds.
Table 9. Results from regression models 3-6 from the entire sample. The values are presented in percentage form. The values in parentheses are t-scores.
The size variables improve the quality of the regressions given the higher values of the F-test and R squared. The interaction variable between size and green bonds is not significant in any regression and we can therefore not draw any conclusions regarding the size effect on green
bonds specifically. However, the size variable on its own is proven positive on a 5%
significance level in models 4 and 5, and a 1% significance level when the year of issue is being controlled for (model 6). This variable could have been skewed if the type of bonds had different distributions of size, however, as seen in figure 2, green and conventional bonds have approximately the same relative size distribution. This variable is therefore applicable on both bond issues, and it is further concluded since the variable specific for green bonds (Size*GreenDummy and GreenDummy) are insignificant in all models. Hence, the larger the relative size of a bond, the higher is the abnormal return on stock price expected to be - regardless of whether the bond is green or not. Consequently, we reject hypothesis 2.
The size variable becomes more significant and is proven on the 1% significance level when the year control variables are added in model 6. As abnormal returns are expected to increase as relative size increases, the market may interpret a larger issue as a signal that the issuing corporation has confidence in its future earnings, or will invest in new projects. Had the interaction variable for green dummy and size been significant, this would have indicated that the greenwashing theory might hold as that would have been specifically for green bonds.
Now, however, size has a positive effect on both kinds of bonds. So, these regressions indicate that relative size, and thereby substance, matters.
The regression model which best explains the data is model 6. The model has the most significant variables and the highest F-test value and adjusted R squared value. The year of issue, especially if a bond was issued in 2019 or 2020, seems to be the strongest determinator of abnormal returns. The relative size is significant as well but, due to the low β value, does not impact AAR much.
4.3 Hypotheses Results
The foundation of the paper is to accept hypotheses 1 and 2 if the paper can prove a significance level of less than, or equal to, 10 %.
Table 10. Results from the study’s hypotheses.
5. Conclusion
This study examines whether green bond issues are related to value creation as compared with conventional bonds, and how the relative size of these impact value creation. Our results show no significant abnormal returns for green bonds, the benchmark of conventional bonds, nor the comparison of means. Further, we could not find an interaction effect between the relative issue size and the green bond. Hence, this study finds no financial incentives for issuing green rather than conventional bonds. In other words, no evidence that issuing green bonds creates more value than issuing conventional bonds has been found.
However, we show that relative size has a positive and significant regression coefficient in all models, meaning that the larger the relative size of a bond issue, the more the stock price is expected to increase.
Our research contributes to the existing research on both green and conventional bonds by illustrating the impact of relative size, a factor which has yet not been thoroughly investigated. Our conclusions are useful for both corporations considering issuing a bond, regardless of type, as well as for analysts and other investors. We can conclude that size does matter, yet not in the way economic theory predicts.
5.1 Limitations and Further Research
A limitation of this study has been access to databases. We had access to Refinitivs Eikon which only includes GBP certified green bonds, while Bloomberg’s database includes these as well as bonds marketed as green. Access to Bloomberg would have made our research more comparable to previous research due to a more similar sample and enabled us to analyze whether bond certifications matter.
To complement this study and further develop its results, a qualitative study could be conducted to investigate whether marketing a bond as green (even if not certified) matters, and how well aware investors are of marketing campaigns related to bond issues. It could also address the difficulty of finding an accurate date for when information on bond issues actually reaches the market. A qualitative approach could also examine the underlying reasons for the bond issue, comparing offensive bonds (associated with a new project) and
defensive bonds (issued to cover solvency problems) market reactions might vary depending on whether the issue is related to a certain project or not.
The time aspect poses another research opportunity. With the growing importance of sustainability/trend of sustainable investing this may indicate that i) green bonds will make up a larger portion of all bonds in the future, and ii) green bonds may have just recently started to generate higher abnormal returns. Millennials are just recently entering the capital market which could explain the growing importance that the control variables of recent years play in the green bonds market reaction.
References
Publications
Arvidsson, S. 2014. Corporate social responsibility and stock market actors: A comprehensive study. Social Responsibility Journal, vol. 10 (2), 210-225.
Aste, E. & Åström, H. 2020. Emittering av gröna obligationer för börsnoterade nordiska bolag - en eventstudie om aktieavkastning. Bachelor thesis, Uppsala University. Retrieved:
http://www.diva-portal.se/smash/get/diva2:1392212/FULLTEXT01.pdf [2020-11-06]
Ball, R. & Brown, P.R. 2014. Ball and Brown (1968): A Retrospective. The Accounting Review, vol. 89 (1), 1–26.
Berk, J. & DeMarzo, P. 2017. Corporate Finance . 4th Global Edition. Essex: Pearson Education Limited.
Bryman, A. 2016. Samhällsvetenskapliga metoder. 2nd edition. Malmö: Liber.
Bryman, A., & Bell, E. 2013. Företagsekonomiska forskningsmetoder. 2nd edition. Stockholm: Liber.
Connelly, B., Certo T., Ireland, D., & Reutze, C. 2011. Signaling Theory: A Review and Assessment. Journal of Management, vol 37 (1), 39-67.
Dang, C.D., Li, F., Yang, C. 2018. Measuring Firm Size in Empirical Corporate Finance. Journal of Banking & Finance, vol 86 (C), 159-176.
Ender, M. & Brinckmann, M. 2019. Impact of CSR-Relevant News on Stock Prices of Companies Listed in the Austrian Traded Index (ATX). International Journal of Financial Studies, vol. 7 (3), 2-18.
Fama, E. F. 1970. Efficient Capital Markets: A Review of Theory and Empirical Work. The Journal of Finance., vol 25 (2), 383–417.
Fama, E. F., Fisher L., Jensen M. C., & Roll R. 1969. The adjustment of stock prices to new information.
International Economic Review, vol. 10 (1), 1–21.
Flammer, C. 2020. Corporate Green Bonds. Journal of Financial Economics, forthcoming publication.
Gelman, A. & Hill, J. 2006. Data Analysis using Regression and Multilevel/Hierarchical Models. 1st edition.
Cambridge: Cambridge University Press.
Godlewski, C.J., Turk-Ariss, R. & Weill, L. 2013. Sukuk vs. conventional bonds: A stock market perspective.
Journal of Comparative Economics, vol. 41 (3), 745-761.
Jeevan, P. 2014. Green washing - A conceptual framework. Development Prospects of Indian Economy, Bharati Publications, 270-276.
Kothari, S.P. & Warner, J.B. 2007. Econometrics of Event Studies. Handbook of Corporate Finance: Empirical Corporate Finance, vol. 1 (1), 4-36.
Larsson, E. & Zakrisson, J. 2020. Värdeskapande med gröna obligationer - En studie om marknadsreaktioner vid emittering av gröna obligationer i Europa. Bachelor thesis, Uppsala University. Retrieved:
https://www.diva-portal.org/smash/get/diva2:1448515/FULLTEXT01.pdf [2020-11-06]
MacKinlay, A.C. 1997. Event Studies in Economics and Finance. Journal of Economic Literature, vol. 35(1), 13-39.
McWilliams, A. & Siegel, D. 1997. Event Studies in Management Research: Theoretical and Empirical Issues.
The Academy of Management Journal, vol 40 (3), 626-57.
Modigliani, F. & Miller, M.H. 1958. The Cost Of Capital, Corporate Finance, And The Theory Of Investment.
American Economic Review, vol. 48(4), 261-97.
Pérez, A., López-Gutiérrez, C. & Salmones, M. 2019. Do all CSR news affect market value equally? Social Responsibility Journal, vol. 16 (8), 1107-1123.
Ross, S. 1977. The Determination of Financial Structure: The Incentive-Signalling Approach. Bell Journal of Economics, vol. 8 (1), 23-40.
Spence, M. 2002. Signaling in retrospect and the informational structure of markets. American Economic Review, vol 92 (3), 434-459.
Stiglitz, J. 2000. The Contributions of the Economics of Information to Twentieth Century Economics. The Quarterly Journal of Economics, vol. 115 (4), 1441-1478.
Tang, D.Y. & Zhang, Y. 2018. Do Shareholders Benefit from Green Bonds? Journal of Corporate Finance, vol.
61.
